UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Ecological site quality and productivity of western hemlock ecosystems in the coastal western hemlock… Kayahara, Gordon John 1992

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

Item Metadata

Download

Media
831-ubc_1992_spring_kayahara_gordon.pdf [ 10.28MB ]
Metadata
JSON: 831-1.0075221.json
JSON-LD: 831-1.0075221-ld.json
RDF/XML (Pretty): 831-1.0075221-rdf.xml
RDF/JSON: 831-1.0075221-rdf.json
Turtle: 831-1.0075221-turtle.txt
N-Triples: 831-1.0075221-rdf-ntriples.txt
Original Record: 831-1.0075221-source.json
Full Text
831-1.0075221-fulltext.txt
Citation
831-1.0075221.ris

Full Text

ECOLOGICAL SITE QUALITY AND PRODUCTIVITY OF WESTERNHEMLOCK ECOSYSTEMS IN THE COASTAL WESTERN HEMLOCK ZONEOF BRITISH COLUMBIAbyGORDON JOHN KAYAHARAB.Sc.F., University of Toronto, 1978A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIES(DEPARTMENT OF FOREST SCIENCE)We accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAFebruary 1992© Gordon John Kayahara, 1992In 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.(Signature)Department of ^Forest Science  The University of British ColumbiaVancouver, CanadaDate February ^21  1992DE-6 (2/88)iiAbstractTo assess relationships within the biogeoclimatic ecosystemclassification(BEC) system, 102 sample plots were established in immaturewestern hemlock [Tsuga heterophylla (Raf.) Sarg.] stands distributed withinthe submontane very wet maritime variant of the Coastal Western HemlockZone in coastal British Columbia. Using the methods described in the BECsystem, plant associations, and field derived soil nutrient regimes (SNR) andsoil moisture regimes (SMR) were identified. Direct measures of SNRs, i.e.,soil chemical measures of the forest floor and mineral soil expressed on aconcentration basis, and site index (height at a reference age) weredetermined for each stand.Despite a lack of understory species, both plant associations and 6diagnostic species were linked to an underlying nutrient gradient. For theformer, the use of canonical discriminant analysis on the principalcomponents analysis (PCA) scores of the soil chemical measures showed adefinite but overlapping trend. This trend was correlated most positivelywith nitrogen, and negatively with the C:N ratio and potassium. For thelatter, canonical correlation analysis of 6 diagnostic species with 4 forestfloor chemical measures resulted in the 4 chemical canonical variatesexplaining 37% of the variance in the species domain. Generally,oxylophytic species varied negatively with total and mineralizable nitrogen,and positively with available potassium and magnesium. The relationshipswith nitrophytic species were reversed.A PCA ordination showed the soil nitrogen measures to be the mostimportant factors in accounting for the variation in the soil chemical data.Discriminant analysis, used to see how well the nitrogen properties couldiiidistinguish SNRs, correctly classified 91% of the plots into their sourcegroup. However, this high success rate was not repeatable; on a validationset, the discriminant function correctly classified only 54% of the plots.All regression models reported, relating site index (m/50 yr) toindirect and direct variables, showed significant (p<0.05) results, and hadadjusted R2 values ranging from 0.35 to 0.81. The standard errors ofestimate (SEE) were relatively high, ranging from 4.5 m to 8.8 m. The bestfit regression equation, having the highest adjusted R2 value and lowestSEE, was the categorical model which related site index as a function ofSMRs and SNRs. The best fit analytical model related site index as afunction of nitrogen positively and potassium negatively (adjusted R 2 = 0.67,SEE = 5.61m). However, this model failed to adequately predict, within 3meters, the site index of the data set from which it was derived. Applied toa test data set, the prediction results improved but the equation tended tounderestimate site index. Multicollinearities among the soil nutrientproperties were noted and therefore, PCA regression used to supplement theinterpretation. Examination of the loadings of the axes of the finalregression equation (adjusted R2 = 0.63, SEE = 5.89) indicated that thestrongest relationships with site index were positively related with sulphurand logarithmic transformations of nitrogen, and negatively with the C:Nratio and potassium. Supplementary relationships were also noted.It was concluded that there exists a relationship between nitrogenmeasures and field derived SNRs. Further, the SNRs can be combined withSMRs to predict western hemlock site index. As well, soil chemicalmeasures can also predict site index. However, there was a relatively largevariation associated with predictions made using the regression equations.TABLE OF CONTENTSABSTRACT^TABLE OF CONTENTS^ ivLIST OF TABLES viLIST OF FIGURES^ viiiACKNOWLEDGMENTSPROLOGUE^ xiCHAPTER 1. INTRODUCTION1.1 Background^  11.2 Purpose and Objectives^ 4CHAPTER 2.^LITERATURE REVIEW2.1 Introduction^  62.2 Site Index.  62.3 Soil-Site Studies^ 102.4 Biogeoclimatic Classification^162.5 Western Hemlock Soil-Site Studies .. 202.6 Western Hemlock Nutrition^ 252.7 Summary^ .32CHAPTER 3.^MATERIALS AND METHODS3.1 Study Area^  .333.2 Study Sites  .353.3 Sampling and Site Description. ^363.4 Soil Chemical Analysis^413.5 Vegetation Data Analysis 443.6 Soil Nutrient/Moisture Regime Analysis. ^483.7 Analysis of Productivity Relationships . . ^51ivVCHAPTER 4.^RESULTS AND DISCUSSION4.1 Vegetation Classification^554.2 Soil Nutrient Regimes 794.3 Vegetation and SNRs.^894.4 Soil Moisture Regimes  .914.5 Indirect Site Index Relationships^934.6 Direct Site Index Relationships^ 1094.7 Site Index Relationships Discussion. . ^ .123CHAPTER 5^SUMMARY AND CONCLUSIONSSummary and Conclusions^129REFERENCES^  141LIST OF TABLES2.1 Reported site index relationships for Douglas-fir using the BECsystem^ 202.2 Relation between site index and of a hemlock-spruce type andindicator plants in southeastern Alaska^  212.3 Western hemlock site index in different plant communities atVancouver and Haney, B.C.(Eis 1962)^  224.1 Diagnostic table identified from western hemlock communitieson Vancouver Island, B.C.^ 574.2 Correlation values of species and associated PCA scores onthe first 2 PCA axes^ 614.3 Frequencies of nitrogen indicator species for the four majorplant associations. ^ 664.4 Correlation of the PCA chemical scores with the axes derivedfrom CDA.^ 674.5 Correlation of PCA scores used in CDA with soil chemicalproperties^ 704.6 Canonical redundancy analysis of the species PCA scores andforest floor chemical PCA scores^ 724.7 Canonical correlation analysis between six diagnostic species andeight forest floor chemical properties.^ 734.8 Summary of positive and negative correlations of six main specieswith forest floor chemical properties^ 784.9 Means and standard deviations of selected soil chemical propertiesand their relationship to SNRs . ^ 804.10 Correlation of soil chemical measures with the first 2 PCA axes ^83vivii4.11 Correlations of the nitrogen measures with the three CDA axes ^864.12 Comparison of soil nitrogen measures between the original andtest data sets.^ 874.13 Confusion matrix based on the source SNRs^874.14 Confusion matrix based on the validation set using the sourcediscriminant function.^ 874.15 Definition of moderately dry, slightly dry, and fresh moistureregimes^ 914.16 Thirty year precipitation normals from Tahsis, B.C.  ^924.17 Tukey-Kramer multiple range test for 4 plant associations. . . . 964.18 Correlation of species that are significantly correlated with thevegetation PCA axes.^ 994.19 Tukey-Kramer multiple range test for 6 site units.^1034.20 Correlation matrix of soil chemical properties. 1104.21 Tolerance values for soil chemical properties^1114.22 Eigenvalues and cumulative variation explained by the soilchemical properties^ 1114.23 Results of all combinations multiple regression^ 1134.24 Comparison of the original and test data sets. 1164.25 Regression model used on the original and test data sets^1164.26 PCA regression correlation for the original data^1194.27 PCA regression correlation for the validation data 1204.28 Literature review summary of past western hemlock soil-sitestudies. ^ 1235.1 Summary of models developed^ 133LIST OF FIGURES2.1 Western hemlock with the stilt-root characteristic^ 273.1 Western hemlock height growth curves of Wiley (1978) showingextrapolated curves^  .384.1 Immature western hemlock stand characterized by a closedcanopy and lack of understory vegetation.^ 564.2 Ordination of plots along the first two new axes of PCA on allspecies showing 80 % elliptical outlines.^ 594.3 Spectra histograms of indicator species of soil nitrogen for thefour major plant associations^ 634.4 Spectra histograms of indicator species of soil moisture for thefour major plant associations^ 644.5 CDA plot of the first two canonical variates showing 80% ellipticaloutlines.^ 684.6 Relationship between the 6 diagnostic species scores and theforest floor properties^ 764.7 Box plots of selected soil chemical properties showing theirrelationship to SNRs^ 814.8 Plot of the first 2 PCA axes of the soil chemical measures with80% elliptical outlines.^ 824.9 Plot of the first two canonical variates derived from soil nitrogenmeasures. 854.10 Plots showing the relationship of plant associations with SNRs .904.11 Box plots of site index and all plant associations derived usingthe Braun-Blanquet method.^ 95viii4.12 A 3-dimensional plot of the site indices on SMRs and SNRs withtheir orthogonal projections^ 1014.13 Edaphic grid showing site index for each site unit^1044.14 Edaphic grid with a site index isoline superimposed and derivedby extrapolation.^ 1054.15 Site series grid with site index values.^ 1084.16 Three dimensional plot and contour plot of the relationship ofnitrogen and potassium with site index.^ 1154.17 Actual site index vs. estimated site index for chosen regressionmodels^ 1254.18 Soil profile illustrating the thick mor humus forms and associateddecaying wood.^ 1274.19 Plotted relationship of site index and soil chemical PCA axis 1.137ixACKNOWLEDGMENTSIn hopes of not forgetting to mention anyone, I thank both mycommittee members and my supervisor for their guidance on the writing ofthis thesis and, more importantly, for their positive influence on mypersonal growth. I thank Dr. Tim Ballard for his enthusiasm for learningand showing me the excitement of soils, Dr. Peter Marshall for rekindling anold interest in computer programming and biometrics, and Reid Carter, notonly for his input, but also for the inspirational discussions on science ingeneral. A very special thank you goes to my supervisor, Dr. Karel Klinka,for his guidance, patience, understanding, and for sharing his enthusiasmand insights. Of course, I thank those funding bodies that providedresearch money and support -- Canadian Pacific Forest Products, FletcherChallenge Canada, the South Moresby Replacement Fund, and the B.C.Science Council.I thank all my fellow graduate students, Audrey Pearson, DonaldMcClennan, John Neumann, and Gaofeng Wang, and my field assistantHarry Williams for their moral support. Especially, I thank Qingli Wang forhis help in directing my way through the sometimes "chaotic" world ofmultivariate statistics. Last, but definitely not least, a thank you for mymother, for her monetary and spiritual support, and who in her typical quietnisei fashion continues to always be there for me.This has been an exciting, enjoyable and intellectually stimulatingperiod, not only an academic education but a personal life experience wellworth the energy. All only made possible with the support of thosementioned and those I have forgotten to mention.PROLOGUE...without classation there would be no science ofecosystems, and no ecology. And indeed, noscience.V.J. Krajina 1Human interest in the relationship between plants and the land whichsupports them is undoubtedly as old as human existence. The recognitionof land capable of supporting food plants and the subsequent mentalcategorizing of this information must have been a natural reaction forsurvival. Thus, the seeds of the need for classification start; it is a naturaland inherent process to create some semblance of order from an apparentdisorderly assembly, and an endeavour which, as far as we know, is auniquely human attribute.During the spring of 1985, while I was employed as a forester onVancouver Island, British Columbia, a treeplanting crew was working on asite which we foresters would designate as "moist/very rich". One of thetreeplanters, having many years of planting experience and having grown upin the rural areas of the Maritimes, decided that this area was "rich" enoughto plant some potatoes among the rows of Sitka spruce. Watching thissmall attempt at agroforestry, the question arose: how could thistreeplanter with entirely different experiences than myself come to the sameconclusion that this site was "very rich"?The treeplanter's assessment was based on planting in severaldifferent areas in Canada and on a Maritime experience of recognizing1 Krajina, V.J. 1960. Ecosystem classification of forests. Silva Fennica 105: 107-110.xixii"farmland". As a forester, my assessment was based on a universityeducation supplemented by several years of field application. Therefore,was our similar but separately derived decisions, based on two very differentlearning experiences, merely subjective? And if it was subjective, why thendid this treeplanter not plant potatoes on a dry, poor site and why would Inot choose to plant Sitka spruce on that same site? Subjective? I thinknot.Rather, both of us based our decisions on combined, althoughdifferent, experiences and a knowledge base using what is referred to as anheuristic procedure. This allowed both of us to identify this small, seeminglyinconsequential, tract of land as having similar vegetation potential. Forconfirmation of my heuristically derived decision, I had the benefit of beingable to analyze the soil with various chemical tests and also measuring theperformance of the planted trees years later. Both confirmed that, indeed,this site was "very rich", and by recognizing other similar areas I am able toinfer that they are also "very rich". Accompanying this recognition are avariety of accessory characteristics associated with this "class", such as soilchemical properties and productivity. Using these basic principles andprocesses it is also possible to extend this recognition of land areas todifferent ones and eventually formalize them into a generalizedclassification. Thus the basic, albeit simplistic, principle of ecosystemrecognition, classification, confirmation, and inference of associatedrelationships is demonstrated.As far as the treeplanter's decision, the confirmation lay in the harvestthat fall of a modest but substantial potato crop. I received some of thatcrop and now I am able to confirm that this treeplanter's heuristicallyarrived conclusion was also correct -- the potatoes were delicious!1. INTRODUCTION1.1 BackgroundProductivity, recognition, and classification of land are three interestswhich are as old as agriculture and whose recorded intellectual interest inthe Western World dates to the centuries B.C. (Kimmins 1988). Productivityof sites was a driving force behind recognition and classification. For trees,the first recorded observations of growth on sites of various moisture statuswas made by Theophrastus (370-285 B.C.), a student of Aristotle(Makkonen 1968; Tesch 1981). Recorded attempts at recognition andclassification of sites soon followed. The Roman Cato (234-139 B.C.)associated certain plants with soil conditions indicating good wheat land(Kelly 1922), and then developed a subjectively based land classificationwhich included land best suited for wine production, pasture-forest land,and commercial forest land (Makkonen 1969; Tesch 1981).The subjective nature of site classification, being based on broadquality classes or loosely based on physical characteristics such as soilexposure, remained as the primary method until the eighteen hundreds(Cajander 1926). Then, throughout Europe, plant geographers developedformalized vegetation classifications, and throughout the 1800's divergentapproaches developed into regional traditions adapting to different kinds oflandscapes and research interests (Shimwell 1971; Whittaker 1962, 1980).It was also during this period that there was the realization of theimportance of soil physical and chemical properties in controlling growth,which ultimately resulted in such familiar ecological rules such as Liebeg's"Law of the Minimum" (reported in Tesch 1981). The application of this12knowledge of soil chemistry resulted in some success in forestryapplications where, during the 1870's in northern Germany, Schiitzefollowed by v. Falckenstein (reported in Cajander 1926) demonstrated thecorrelation between forest yield capacity and some soil chemical measuresthat were available at that time. However, it was not until the late 1920'sand early 1930's that significant contributions to soil testing in evaluation ofsoil nutrients were obtained (Melsted and Peck 1973).The earliest comprehensive study that synthesized the relationshipbetween a vegetation classification system, soil chemical properties andproductivity was that of Cajander (1926) in Finland. Using the NorthernTradition of vegetation classification (Frey 1980), the boreal forest of Finlandwas recognized to be of five types. Ilvessalo (1927) related productivity toeach of these types. Valmari (reported in Cajander 1926) found differencesin amounts of loss of weight on ignition, of electrolytes (e.g sodium chloride),and nitrogen in the upper eight inches of soil, and Aaltonen (reported inCajander 1926) found differences in nitrogen and pH of the humus betweenthe types.At present, interest in the relationship between productivity,recognition, and classification of land is becoming more important aspressure for resources in a growing population continues (Kimmins 1988).Additionally, there is the realization of the need to develop an understandingof the underlying processes of what is holding a community intact andfunctioning as an organic entity (OrlOci 1988). To this end there has beenrecognition of the value of a multifactor, integrative, hierarchical forestclassification system (Krajina 1972; Kimmins 1977; Spun and Barnes1980; Barnes et al. 1982; Barnes 1984, 1986; Monserud et a/. 1990).This concept and approach are inherent to the biogeoclimatic3ecosystem classification (BEC) system (Pojar et al. 1987), now widely used inBritish Columbia, which provides a framework for organization of knowledgeof forest ecosystems. The classification is based, with some modifications,on the work of Dr. V.J. Krajina and his students in the 1960's and 1970's atthe University of British Columbia (see Krajina 1959, 1965, 1969, 1972;Wall 1988). Sites are classified according to their potential to producesimilar vegetation. Plant communities at a relatively stable stage ofdevelopment (mature, climax or near-climax) are considered a reflection ofthis potential and form the basis for classification. Classes derived throughthis process, termed "site units", are subsequently characterized by theirclimate, soil moisture and soil nutrients. A result of this classification ispresented as a series of edatopic grids of soil moisture regimes (S1\411) andsoil nutrient regimes (SNR) (Pogrebniak 1930 cited in Krajina 1972). Thuseach grid is developed within a regional climate, and distinguishes site unitswith environmental properties inferred to be similar or equivalent and,hence, having similar vegetation potential.Having classified, the next step is that of testing, validating anddemonstrating the importance of the derived units. A relationship betweenforest productivity and site units derived from the BEC system wasproposed by Krajina (1969). Subsequently, the relationship of productivity,using site index'', to vegetation units, site associations, site series, and theirformative elements [biogeoclimatic unit (subzone or variant), actual soilmoisture regime, soil nutrient regime] have been demonstrated (Kojima1983; Courtin et al. 1988; Green et al. 1989; Klinka et al. 1989; Carter andKlinka 1990; Klinka and Carter 1990).1 Site index is taken to be the height of a specified character and number of trees at a specified age.This is used as an empirical gauge of a site's capacity to support forest growth for the specificspecies measured.4As well, characterization of soil nutrient regimes, to relate the siteunits to objective and meaningful criteria, has been investigated (Kabzemsand Klinka 1987a) and an approach for an objective means of defining SNRsproposed (Courtin et al. 1988). To objectively define soil moisture regimes,Klinka et at. (1984) used the occurrence and duration of phases of wateruse, complemented by the ratio between actual and potentialevapotranspiration, and the occurrence and depth of the water table.1.2 Purpose and ObjectivesIn an effort to further assess the classification, and in order todescribe western hemlock [Tsuga heterophylla (Raf.) Sarg.] ecosystemsthrough the use of statistical patterns, this thesis is an investigation intothe relationships between western hemlock ecosystems, as derived from thebiogeoclimatic ecosystem classification system (Pojar et a/. 1987), associatedsoil chemical and physical properties, and productivity as measured by siteindex (height at 50 years breast height age). It seeks to provide a practicaltool for silviculturists, to describe relationships as related to westernhemlock and site units, to increase understanding of western hemlocknutrition, to increase understanding of the many interacting factors thatmake up an ecosystem, and to provide direction for further research.The specific objectives are:(i)^to explore the relationship of individual understory species, andthe collective expression of individual species through plantassociations derived using the Braun-Blanquet method, to soilnutrient measures.5(ii) to explore whether the inferred soil nutrient regimes and soilmoisture regimes, as derived from the field proceduresprescribed by the BEC system, are related to direct measures ofsoil nutrients for the former, and to the ratio between actualand potential evapotranspiration for the latter.(iii) to assess relationships of western hemlock site index withindirect and direct measures of ecological site quality. Thisincludes vegetation units and site units, as derived from theBEC system, individual species, and individual soil nutrientmeasures.2. LITERATURE REVIEW2.1 IntroductionProductivity, recognition, and classification of sites all developed in aparallel and concurrent manner, resulting in many different approacheswith much overlap. Productivity, however, is the common link and, in NorthAmerica, was the focus in the development of classification systems. Sincethe measure of productivity used in this thesis will be site index, thefollowing literature review will start with the concept of site index toestablish the validity of this method in assessing productivity. A review ofthe many studies involving correlation between a measure of potentialproductivity with individual factors of the site, the so called soil-site studies,is unnecessary since many excellent reviews have been done in the past.Similarly, a review of the many varied classification systems throughout theworld is beyond the scope of this study. However, commentary on certainsoil-site studies will be made with emphasis on the principles used as aframework for this thesis, and reference will be made to some of the morepopular and familiar (without a connotation of "better") classificationsystems. The three concepts of site recognition, productivity, andclassification will then be brought together in a review of some neededconcepts of the BEC system. Finally, a review of studies related directly towestern hemlock productivity and nutrition will be made.2.2 Site IndexModern methods and principles for evaluating forest productivitydeveloped in Europe starting at the end of the 18th century. From the671790's to the 1870's individual tree volume curves were developed fromearly anamorphic and stem analysis procedures, and from permanentsample plots (Cajander 1926; Tesch 1981). In 1824, Huber introduced stemanalysis on dominant trees to determine the normal development of standheight, named the index method (reported in Cajander 1926). This was thefirst method founded on the assumption that the dominant trees of a standhad always been dominant. The claim that height is a sensitive measure ofdifferences in site was based on comparisons of volume and height of 100year-old stands reported by Professor Schwappach in Germany in 1908(cited in Roth 1916). Schwappach regarded "... height, leaving out someabnormal cases, as the best criterion of the site in the stands of middle ageand older, while the volume of the main stand is suited for this purpose ifthe stand has been properly cared for and is in normal condition for a longperiod". Roth (1916) supplemented this claim with comparisons of volumeand height on different sites from five studies carried out in the UnitedStates.The use of height of the dominant tree as an indicator of volumeproduction was imported into the United States during the early 1900's(Mader 1963; Tesch 1981). The reasons for adopting height as a method ofsite classification were: (1) height is a sensitive measure of differences insite; (2) height is independent of stocking and species mixture within broadlimits; and (3) the height/age relationship is easy to determine and thus, isconvenient, simple, and practical (Roth 1916, 1918; Watson 1917; Society ofAmerican Foresters 1923).For several decades beginning in the 1920's, the site index conceptwas used in reporting most yield tables in the United States (Tesch 1981).Bruce (1923, 1926) established the anamorphic approach for height curve8construction, where a central guide curve or average curve was fit throughthe data and curves for each site index class were then derivedproportionately, as the primary method of determining site index.Since then, site index has become the most widely accepted measureof productivity, based on the observation that the height of dominant trees(that have always been dominant) of a given species and age is more relatedto the capacity of a given site to produce wood than any other singlemeasure, besides volume itself (Spurr and Barnes 1980). However,problems with using site index as a measure of productivity have beenreported (Monserud 1984a). Mader (1963) felt that height and volumegrowth may not react in exactly the same way to site differences, thusheight differences alone may not fully reflect site differences. Monserud(1984a) pointed out that sites having the same height growth potential neednot necessarily have the same basal area growth potential. Others (Sammi1965; Hall 1983; Monserud 1988) have stated that since volume is themeasure of interest, volume itself should be used directly.Also, specific problems with the anamorphic approach have beenreported (Carmean 1970; Beck and Trousdell 1973; Monserud 1984a).These include (i) the possibility of sampling bias due to a disproportionalsample of site and age during curve construction, and (ii) the assumptionthat the shape of the height growth curve is the same for all sites. This wasrecognized by Bruce (1926) and was soon after demonstrated for red pinewith the construction of polymorphic height curves on sites groupedaccording to height at 15 years (Bull 1931). Since then, a polymorphictrend between different sites has been demonstrated for mixed hardwoodplant associations using the Braun-Blanquet approach (Lemieux 1964), jackpine (Pines banksiana Lamb.) using soil pore pattern, soil moisture and9vegetation as distinguishing factors of different sites (Jameson 1963), blackspruce [Picea mariana (Mill.) B.S.P.] using the regolith system(VanGroenwoud and Ruitenberg 1982), interior Douglas-fir [Pseudotsugamenziesii (Mirb.) Franco var. glauca (Beissn.) Franco] using Daubenmire'shabitat types (Monserud 1984b) and others (see Jones 1969; Monserud1984a).The height growth of chosen site index trees is assumed to beindependent of stand density over a wide range of stocking (Hd.gglund 1981).Empirical evidence from thinning experiments indicates that within widelimits of stand density, height growth of dominants and codominants seemsto be unaffected by thinning (Lynch 1958; Clutter et a/. 1983). Someexceptions occur where height increment is severely reduced in very densestands, and is most apparent on poor sites (Lynch 1958). Studies involvingnatural stands, planting trials or spacing trials found that this was the casefor lodgepole pine (Pinus contorta Dougl.) (Parker 1942; Smithers 1956;Alexander et al. 1967), ponderosa pine (Pinus ponderosa Dougl.) (Lynch1958; Barrett 1965, 1970, 1973), and slash pine (Pinus elliottii Engelm.)(Collins 1967; Bennett 1975). Reukema (1979) found that the height at age51 years increased with increasing spacing for Douglas-fir [Pseudotsugamenziesii (Mirb.) Franco] on a poor site. For western hemlock specifically,evidence suggests that there is no difference in top height in stands ofdensities from 300-1000 stems/hectare when growing on medium sites(Reukema and Smith 1987). Wiley (1978) found that densities at age 50from less than 200 to 600 trees per acre did not have an effect on site index.In breeding programs, trees superior in production (so-called "plustrees"), can be identified on the basis of superior height growth (Wright1976). Yet, in site index, this factor influencing height growth is ignored.10Presumably, the idea is that with a measure such as "top height", where100 of the largest diameter trees are chosen as site index trees, this geneticfactor will somehow "even out". However, a study by Monserud andRehfeldt (1990) gives evidence that a "genetic index" of standardized 3-yearseedling heights (Rehfeldt 1989) explains more of the variation in heightgrowth for Douglas-fir than site factors, although King (1966) showed thatthis effect falls off as trees get older.Physiological ecological theory suggests that characteristics such asheight are controlled by many genes of small effect. Thus plant growth isextremely plastic, allowing it to respond to an unpredictable environment,and a plant's size directly reflects the conditions under which it is growing(Waller 1986). It is generally accepted that these variations are related tothe acquisition of limiting resources, at least in part. Changes inenvironmental conditions are met by allocation of growth resources to theorgan that is capable of alleviating the limitation (Fitter 1986). Keyes andGrier (1981) found differences in partitioning of resources between Douglasfir on a low and high productivity site. They proposed that on harsh siteswhich may impose water or nutrient deficiencies, this shift in productionallocation from above to belowground may be an essential mechanism toavoid or alleviate stress. This subsequently has been supported forlodgepole pine (Comeau 1986) and for Douglas-fir (Vogt et a/. 1983; Kurz1989).2.3 Soil-Site studiesIn the United States during the early 1900's, there was considerabledebate over what method should be adopted to unify the approaches to siteevaluation. The standards suggested and debated were height growth,11volume growth, and site types. Although in the end height growth becamethe accepted standard for the classification of sites (Society of AmericanForesters 1923), a general recognition seemed to exist that the use of sitetype and either volume or height measures were not in conflict but ratherserved complementary purposes. Even proponents of using height growthas a measure of site quality stated that a classification plan based onphysical factors was fundamental. However, since the knowledge at thattime was inadequate for such a classification, an indicatory measure suchas height growth was suggested for use in the interim (Mader 1963).Emphasis, from that time, was placed on the classification of sitesbased on productivity, and the development of site index curves using ananamorphic approach. Subsequently, many investigations of therelationship between a productivity measure, usually site index, andchemical and physical components of the site have been carried out. Thesehave come to be known as "soil-site" studies and have had an objective offinding an alternative method to predict productivity for areas where siteindex could not be determined (for reasons such as lack of the species ofinterest, age, top damage, excessive stand density, etc.).It was not until the 1960's that other systems involving climate,physiography (geomorphology, landform, terrain), soil, and vegetationreceived a major focus (Barnes 1986). The components of an ecosystem (theclimate, landform, soils, and vegetation) can be used together in variouscombinations to form a multi-component ecosystem classification system.The varieties of classification involving forests in North America are brieflyreviewed by Barnes (1986), by Burger and Pierpoint (1990) for Canada, andby Pojar and Meidinger (1991) specifically for British Columbia.Soil-site studies have been reviewed thoroughly (e.g. Coile 1952;12Rennie 1963; Ralston 1964; Jones 1969; Tamm 1971; Carmean 1970,1975, 1986; Shrivastava and Ulrich 1976; Daniel et al. 1979; Spurr andBarnes 1980; Tesch 1981; Hägglund 1981; Gessel and Oliver 1981; Pluthand Corns 1983; Clutter et al. 1983; Grigal 1984; and Packee 1988);therefore, only general description and comments will be given.The methods reported can generally be organized into two majorgroupings: (1) methods combining individual components of the site todirectly estimate a measure of productivity, what has been termed afactorial approach; and (2) methods utilizing classes or taxa of ecological orland classification systems to directly estimate a measure of productivity, anintegrative ecological approach.Relationships between a productivity measure and various soil,environmental and chemical factors, have been studied extensively, withalgorithms developed for a wide range of species, using a wide range offactors, over most of the world. The environmental variables thought tohave the greatest influence on tree growth are selected, combined and thenrelated to a measure of productivity. This approach has been referred to asa "factorial approach" to measuring productivity, for it tries to approximateforest productivity by relating it to one or more limiting factors of thephysical environment (Coile, 1952; Jones, 1969).The purpose of the derived equations has been either (1) to predictforest productivity from site factors when the species of interest is absent, or(2) to try to establish factors that are correlated with the productivity of aparticular species. The most common method is the use of multiple linearregression where a measure of productivity, usually site index, is used asthe dependent variable and environmental components of the site thoughtto have an influence on productivity are used as the independent variables.13Both field and laboratory physical and chemical measures have been usedas independent variables. However, equations using chemical propertiesmeasured in the laboratory usually have a different purpose than thosederived from field measurable properties. The practical usefulness ofregression analysis for predicting site index from soil chemical variables isseverely restricted by the cost and the difficulty of assessing nutrientsupply. As noted by Hdgglund (1981), properties measured throughlaboratory analysis usually will not be included in functions intended forpractical field work. However, these techniques may have value when tryingto evaluate some of the nutrient properties correlated with tree growth.Considering the multitude of varying results and successes, it isdebatable whether the factorial method, as currently used, is really leadinganywhere. The number of different techniques and results causedHagglund (1981) to pose the question of whether we are on our way towardsdevelopment of aids for practical productivity estimation or whether siteevaluation research is trapped in a "maze". At best it seems that successfulequations can be obtained on small, relatively homogeneous areas withuniform climates and distinctive soil conditions (Hodgkins, 1956; Monserudet al. 1990). The one generalization from the many factorial studies wasthat the growth potential of trees is chiefly affected by the amount of soiloccupied by tree roots and by the availability of soil moisture and nutrientsin this limited space (Spurr and Barnes 1980).There are inherent problems in studies that choose arbitrarily tosample across a diversity of sites. Multicollinearity of both physical andchemical factors can cause instability of multiple regression equations andconflicting or even misleading results if not specifically addressed. If theequations are used for interpretation, interactions with soil moisture,14nutrients, aeration and temperature can make this difficult. Only byholding moisture, aeration and temperature constant and varying nutrientsonly, may a nutrient effect become clear. However, for example, if soiltexture varies from sand to clay, then with decreasing soil particle size theremay be an increase in nitrogen, but a concomitant decrease in aeration.Productivity may then seem to vary inversely with nitrogen content. Thiswas the case in a soil-site study of western catalpa (Catalpa speciosa,Ward), where Walker and Reed (1960) found significant negative correlationsbetween height and total nitrogen (%). However, total nitrogen was highlycorrelated with the silt-plus-clay content of the soil, which in turn wasnegatively correlated with height. They concluded that the negativecorrelation between height and nitrogen was a "trailer effect" of the negativecorrelation between height and the silt-plus-clay content.Broadfoot (1969) felt that there were insurmountable problems withusing a factorial method to objectively select measurable soil properties topredict site index for southern hardwoods over wide geographic areas. Acombined subjective and objective approach was developed to predict siteindex, based on an understanding of the site requirements of the species.Using this experience, site factors were "subjectively" chosen and combinedto derive a site index value. For southern hardwoods this technique wassatisfactory (Baker and Broadfoot 1977, 1979). Harrington (1986) usedstepwise discriminant analysis to derive the factors in an objective mannerfor red alder (Alms rubra Bong.).The second group of soil-site studies involves adding forestproductivity measures to a taxonomic system which then can allow derivedtaxa or their groupings to be used directly to estimate a productivitymeasure. Taxonomic classifications are not restricted to productivity since15the units can be used for other management or scientific purposes. Sincethe taxonomic units usually are thought of as an expression of manydifferent environmental factors, multicollinearities are thought simply toexpress themselves into a few emergent properties. This identification ofunits also has potential for polymorphic site index curve development.Soil classification systems (Soil Survey Staff 1975; Agriculture CanadaExpert Committee on Soil Survey 1987) are the obvious systems with whichto relate productivity measures, soil being the rooting medium of vegetation.However, in many studies based on earlier versions of soil classificationsystems, the variation in productivity measures, usually site index, withinand between soil series or soil mapping units has been too high for soil taxaalone to be related to productivity (Jones 1969; Carmean 1970, 1975; andothers). The range of physical and chemical properties of the soil taxa of thepast seemed to be too wide to accurately reflect the major edaphic andtopographic factors that influence tree growth (Coile 1952; Ralston 1964;Jones 1969; Carmean 1975). Generally, the position was that highcorrelations between a site productivity estimate and soil classification unitoccurs only in restricted areas of limited variation (Doolittle 1957; Trimbleand Weitzman 1956). However, further study is required to assess thecorrelation with productivity using present-day versions of soil classification(for example Agriculture Canada Expert Committee on Soil Survey 1987),especially at the soil series and phase level.Another classification system, which has popularity in the UnitedStates, is Daubenmire's habitat system (Daubenmire 1968). Briefly, all landareas potentially capable of producing similar plant communities at climaxmay be represented by one habitat type in this system. The climax plantcommunity, because it is the end result of plant succession, is thought to16reflect the most meaningful integration of the environmental factorsaffecting vegetation. Thus, each habitat type represents a relatively narrowsegment of environmental variation and delineates a certain potential forvegetative development. Habitat types can be identified during mostintermediate stages of succession by comparing the relative reproductivesuccess of the tree species present with known successional trends and byobserving the existing undergrowth vegetation.Studies relating habitat type to productivity have had promising butinconclusive results. Pfister et al. (1977) reported the estimated yieldcapabilities of habitat types in Montana showing a large variation withinhabitat types, but a gradient between habitat types. In other studies,habitat types were grouped into productivity classes with significantdifferences in productivity (Roe 1967; Johnson et ca.. 1987). Using stemanalysis Monserud (1984b) not only found different site indices betweenhabitat types but also found different shaped height curves for each type.Derived units from soil, vegetation and ecosystem classificationsystems have potential for use in estimating site index. Correlation ofproductivity to interpretive units would in fact be an integral part of thetesting of classification systems to discover relationships. The derived taxaalso allow for the construction of site-specific polymorphic height curves,but perhaps more importantly, the classification system itself provides aneeded framework of which productivity is one integral, but not the only,function.2.4 The System of Biogeoclimatic Ecosystem ClassificationThe biogeoclimatic ecosystem classification system organizesecosystems according to the principal of ecological equivalence, which17results in groups that have similar ecological site quality and potentialvegetation (Polar et al. 1987). This basic unit of site classification is the siteassociation. Site association represents a group of ecologically equivalentsites that have a similar vegetation potential. They are subsequentlycharacterized by a certain range of climates, soil moisture and soil nutrientregimes, plus additional features if required. On these sites, similar plantcommunities will develop at late successional stages. The concept ofecological equivalence and the use of indicator species acknowledgesinteraction as an inherent property of ecosystems.A site association can contain ecosystems from several differentclimates thus varying in actual site conditions. Dividing the siteassociations into site series using subzones and variants produces site unitsthat are climatically, and therefore usually edaphically, more uniform. Siteunits are presented through edatopic grids, a form of an environmentalmatrix, composed of two major gradients, hygrotope (soil moisture regimes)and trophotope (soil nutrient regimes) (Pogrebniak 1930 cited in Krajina1972).Soil moisture regime (SMR) represents the long-term balance betweenthe amount of available water and the demand for that water by vascularplants. Krajina (1969) adopted nine classes (0 to 8) of relative SMRs andapplied them consistently in each climate (variant). These, in effect, reflectmoisture holding capacity, and the actual water available for plants isdependent on the circumscribing climate. Actual SMRs for coastal BritishColumbia, based on the ratio between actual and potentialevapotranspiration, the annual water balance, and the depth of the growingseason groundwater table, were proposed by Klinka et a/. (1984).Soil nutrient regime (SNR) is the average amount of essential soil18nutrients that are available to vascular plants over several years. Krajina(1969) adopted six classes (A to F) of SNRs and applied them in differentclimates and for soils with different SMRs.The identification of site series is based on a combination of climate,soil moisture regime and soil nutrient regime. Climate is identified throughthe variant, while SMR and SNR are assessed according to a heuristicsynthesis of individual physiographic and soil properties, expressed in anode and link key (Klinka et al. 1984; Banner et a/. 1990), and modified byvegetation. The soil properties chosen use identifiers which are themselvesemergent properties integrating causal factors into a physically identifiablecharacteristic. Thus humus order (Klinka et a/. 1981), which has beenshown to reflect decomposition rate (Klinka et a/. 1990), soil colour, which isrelated to soil organic matter content (Soil Survey Staff 1975), and thecharacter of the A horizon, which indicates pedogenic processes, forexample, are used to help identify a SNR. The presence of mottling whichindicates reduction-oxidation reactions caused by a fluctuating water tableis a property to help identify a SMR. Heuristic procedures, although notcompletely objective, take advantage of the current knowledge of "experts"and are especially useful when no algorithm exists. (Pearl 1984; see alsoGroner et a/. 1983).The concept of nutrient availability can be viewed from twoperspectives, (i) as a nutrient availability, the rate at which nutrients can besupplied for plant uptake, which is site dependent, and (ii) as nutrientlimitation, which is the extent to which productivity is reduced by aninadequate rate of nutrient supply, and is not only site dependent, butspecies dependent as well (Chapin et al. 1986). Nutrient characterization ofSNRs derived from the BEC system is a measure of the former and the19relationship to productivity a measure of the latter.On an initial quantification of the field-estimated SNRs recognized inthe BEC system (poor, medium, rich and very rich), Kabzems and Minim.(1987a) found that the sums of mineral soil plus forest floor mineralizablenitrogen, total nitrogen, and exchangeable calcium and magnesium,expressed in kg/ha, were the properties that best differentiated the SNRs.Individually, ln(x+1) transformations of mineral soil, and the sum of mineralsoil and forest floor mineralizable nitrogen and total nitrogen showedsignificant differences (Student-Neumann-Keulls test, p<0.05). Usingcluster and discriminant analysis, Courtin et a/. (1988) were successful indifferentiating soil nutrient regimes in coastal British Columbia on the basisof pH and carbon-nitrogen ratio of the forest humus form, and total soilnitrogen and the sum of available calcium, magnesium, and potassium(kg/ha) within the soil rooting zone.The relationship of site units, derived from the BEC system, withproductivity has been reported for Douglas-fir (Pseudotsuga menziesii(Mirb.) Franco) located within the very-dry-maritime and the dry-maritimesubzones of the Coastal Western Hemlock biogeoclimatic zone of BritishColumbia. Kabzems and KLinka (1987b) found significant (p<0.05)differences in site index (height at 50 years breast height age) for Douglas-firbetween each SMR (very dry, dry, and fresh) and each SNR (poor, medium,rich, and very rich) with the ordering of increasing site index correspondingto the ordering of SMR and SNR. Mineral soil (alone) and forest floor plusmineral soil mineralizable nitrogen, total nitrogen, exchangeable calcium,exchangeable magnesium, and mineral soil extractable phosphorus were thesoil chemical properties most highly correlated with site index. Otherresults (Green et al. 1989; Minia et al. 1989; Minim. and Carter 1990; and20Carter and Klinka 1990) show that Douglas-fir site index can be reliablypredicted using site associations, certain indicator species groups, SMRsand SNRs, and mineralizable nitrogen combined with growing-season water-deficit (Table 2.1).Table 2.1 Reported Douglas-fir site index relationships with variables derived from the BEC systemwithin the very-dry-maritime and the dry-maritime subzones of the CWH zone.R2 S.E. (m) n Author0.86 3.18 88 Green et al. 19890.77 3.95 88 Green et al. 19890.86 1.99 56 Klinka et al. 19890.85 2.00 99 Klinka and Carter 19900.67 3.10 53 Carter and Klinka 19900.71 2.90 53 Klinka and Carter 1990Variablessite associationsplant indicator species groupsfield estimated SMRs and SNRsgrowing-season water deficit andnatural log of mineral soil mineralizablenitrogen (anaerobic incubation)actual evapotranspiration duringMay and June, natural log of forestfloor and mineral soil mineralizablenitrogen (anaerobic incubation)2.5 Western Hemlock Soil-Site StudiesThere have been several soil site studies on the relationship ofwestern hemlock to its environment and productivity. The first reportedstudy was by Taylor (1929) on what was referred to as a hemlock-sprucetype in southeastern Alaska. One hundred and sixty-six tenth-acre plotswere sampled and a simple compilation of per cent occurrence of dominant21vegetation and the associated site index was made. As Table 2.2 shows,there was a correlation between the occurrence of these five species and siteindex (total height at 100 years for combined hemlock and spruce). Byapplying current knowledge as to the indicator values of these species, therelationship between site index and soil nutrients can be inferred. 0.horridus and R. spectabilis are considered nitrophytic species (species thatinhabit substrates that contain easily available nitrogen as a result of strongnitrification) while V. ovalifolium, C. canadensis, and R. pedatus areTable 2.2: Relation between site index of a hemlock-spruce type and indicator plants in southeasternAlaska (Taylor 1929). Site index is based on both hemlock and spruce combined.Species Site index [height(m) 100yrs breast height age]12.1^15.1^18.2 21.2 24.2 27.3 30.3 33.3% of species in total vegetationSalmonbeny (Rubus spectabilis) 0.2 3.2 8.0 14.6 23.6 32.2Devil's club (Oplopanax horridus) 0.4^3.0 6.5 10.5 16.0 23.0 32.0Blueberry (Vaccinium ovalffolium) 32.0^28.6 25.4 22.0 18.8 15.4Bunchberry (Corpus canadensis) 24.5^20.6^16.5 1.4 8.3 4.2Trailing raspberry (Rubus pedatus) 20.8^15.6^11.5 8.0 5.2considered oxylophytic species (species that inhabit acid (approximately pH< 4.5) substrates with low nitrification) (Klinka et al. 1989). Hence, siteindex of western hemlock also seemed to be correlated indirectly with thenitrogen status of the soils. However, the number of spruce and hemlockon which site index is based is not known.Eis (1962) sampled 139 plots on specific plant communities aroundVancouver and Haney, B.C. The results (Table 2.3) indicated that site indexfor western hemlock varies according to plant community, although22statistical tests were not carried out.Stephens et al. (1969, reported in Heilman 1976) found a correlationbetween site index and total nitrogen in the surface organic matter (r = 0.83)based on twenty-five samples in Alaska. In Oregon, several studies ofwestern hemlock site relationships with site index have been carried out.Wooldridge (1961) investigated the relationship between site index andchosen physical and chemical properties. Relationships worth noting weretotal soil depth (r = 0.80), elevation (50-1500ft.) (r = -0.70), pH (1:1 soil-water ratio with pH meter) (r = 0.55), clay% (r = -0.77), cation exchangecapacity (ammonium acetate method) (r=-0.69), and extractable potassium(ammonium acetate method) (r = -0.76). The association of low site indexeswith high per cent clay was interpreted as being an effect of poor aeration.Although Wooldridge interpreted that the per acre available levels of sodium,potassium and phosphate were related positively to growth, one has to becareful since these may simply be a reflection of soil depth (which wascorrelated to site index with r = 0.80) from which they were calculated.Table 2.3: Western hemlock site index in different plant communities at Vancouver and Haney B.C.(Eis 1962).Plant community No. of site indexl SD SEplots (m@100 yrs) (m) (m)Vaccinium-Gaultheria 9 17.5 3.5 1.2Gaultheria 16 27.7 5.9 1.5Vaccinium-Lysichitum 9 29.0 6.4 2.1Mahonia 7 31.2 6.6 2.4Vaccinium-Moss 26 32.2 6.5 1.6Ribes-Oplopanax 9 33.7 7.1 2.4Moss 26 36.7 5.3 1.1Blechnum 24 36.7 7.1 1.5Polystichum 24 39.3 5.3 1.21 Site index was calculated from the average height of dominant and codominant tree using curvesfrom Barnes (1949).23In 38 second growth and 18 old growth hemlock in the WashingtonCascades, Heilman (1976) found no relationship between site index (heightat reference age of 100 years breast height age) and the soil depth or depthof the humus layer, although it was noted that site index appears to bereduced when rooting depth is less than about 24 inches. Again elevation(0-2000ft) showed a negative correlation with site index (r = -0.55). Verylittle correlation with site index was found between forest floor totalnitrogen(%) (r = 0.03), pounds of nitrogen per acre in the top 10 inches ofthe soil (r = 0.03) and pounds of nitrogen per acre in the rooting depth (r =0.09). A better correlation was obtained with the carbon-nitrogen ratio inthe Al horizon (r = -0.39), with a reduction in site index when the ratio wasabove 15. The high available phosphorus (Bray method no. 1) of 84 ppm inthe forest floor suggested that the variation in site index was not related tophosphorus. There was no evidence that pH or concentration ofexchangeable potassium, calcium, or magnesium was related to site index.In the Oregon Coast Range, Meurisse (1972, 1976) investigated therelationship between site index (height at 100 years breast height age) andselected soil properties from fourteen sites. Once again, elevation (90-200feet) and effective soil depth showed a significant (p < 0.05) relationship withsite index with r =-0.91 and r = 0.71 respectively. There was littlerelationship between pH, as measured in water and in 1M KC1, and siteindex, with all soils having a low pH value (3.9-5.0). Site index hadsignificant but weak correlations (r = 0.50) with total nitrogen (kg/ha) 1 , andthe sum of extractable calcium, magnesium, and potassium (kg/ha) (r =1 Except for pH, all other chemical analysis in Meurisse (1972, 1975) was referred to the OregonState University Soil Testing Laboratory (Roberts, S.R., R.V. Vodraska, M.D. Kauffman and E.H.Gardner. 1971. Methods of soil analysis used in the soil testing laboratory at Oregon StateUniversity. Corvallis, Oregon State University, Department of Soils. Agricultural ExperimentStation Special Report 321).240.30). Increasingly higher correlations with site index were found withcation exchange capacity (meq/100g) having an r = -0.66, organicmatter(kg/ha) (r = 0.67), and available phosphorus (kg/ha) (r=0.78). Usingmultiple regression, the best fit equation was:SI = 176.6 - 0.101(elevation in m) + 0.742(P in kg/ha) - 0.026(Na in kg/ha) +0.008(K in kg/ha)SEE = 9.5 ft^R2=0.93Variables representing: P = available phosphorus; Na = extractable sodium;K = extractable potassium.The model not including elevation was:SI = 101.0 + 2.8(P in kg/ha) - 0.038(total N X 10- 3 in kg/ha)2 + 0.95 x 10-6 (Sumof bases in kg/ha)2SEE = 15.6 ft.^R2 = 0.78Variables representing: P = available phosphorus; total N = total nitrogen;sum of bases = sum of extractable calcium, magnesium, andpotassium.However, caution must be used in interpreting these equations sincemulticollinearities exhibited by the factors were not thoroughly investigatedand high bivariate correlations were reported.Using physical field measurable or estimable properties, Eis (1962)developed the following regression equation, among several others, for workin the field if site index can not be measured directly. The model, based onsoil and moisture factors and including a variable for plant community,was:SI = 100.2 + 2.102(depth of soil) + 4.197(soil moisture) - 4.633(soilpermeability) - 2.494(thickness of organic horizons) +4.668(plant community)R2 =0.63Variables such as depth of soil were expressed in actual measurement andvariables such as soil permeability and plant community were coded.25Steinbrenner (1976) developed a regression equation for westernhemlock site index for predictive and comparative studies, based on a totalof 103 sample plots. The area was stratified into glaciated and non-glaciated soils and the following two equations were reported:Non-glaciated soils:SI = 99.5 + 2.29(depth of "A") - 0.028(depth of "A")2 - 8.48 (Log [depth of "A"])+ 340/(silt+clay in "A") - 0.035(Elevation)R2=0.82^ SE=5.4 feetGlaciated soils:SI = -184.9 + 2.75*(total depth) + 11314/(total depth) - 12296/(total depth) 2 +39.2/(clay in "A") + 5.77*(elevation) 2 + 151.6/(elevation)2 -0.025*(elevation X precipitation) - 1526513/(elevation X precipitation) 3- 5.6Log(elevation X slope position)R2 = 0.77^ SE = 5.8 feetAlthough these equations are unwieldy with virtually no capacity forinterpretation, they were intended for prediction only in this restricted area.2.6 Western Hemlock NutritionThe nutritional requirements of western hemlock are low with the bestgrowth occurring where there is a well balanced supply of nutrients in smallquantities (Krajina 1969). It has been found that western hemlock survivesany deficiency treatment better than other conifers growing in BritishColumbia (with the exception of mountain hemlock which reacts similarly),and will grow even in such nutrient-poor soils where nutrients are availablein the smallest possible quantities (Krajina et al. 1982). However, acomplicated system of feedback, symbiosis and specialization seems to beinvolved with western hemlock nutrition.26Western hemlock seedlings seem to have a preference for ammoniumsources of nitrogen as compared to nitrate (Taylor 1935; Swan 1960; Vanden Driessche 1971; Krajina et a/. 1973), although there is a significantinteraction between nitrogen source and pH (Van den Driessche 1976).Turner and Franz (1985) suggested that the absence of nitrification is one ofthe characteristics of the hemlock nutrient cycling regime. This wasbelieved to involve the inhibition of nitrifying bacteria caused by the low pHof the soil and forest floor, higher phenolic contents of the foliage andconsequently the litter, and the higher forest floor fungal biomass whichwould exude organic acids.In a trial using a hydroponic solution, western hemlock seedlingswere distinctly more tolerant than Douglas-fir or western redcedar to pH 3.0(Ryan et al 1986). Rygiewicz et al. (1984) reported that the ratio of H+ ionsextruded to ammonium ions taken up is higher than Douglas-fir, althoughmycorrhizae appear to act as a rhizosphere buffer where mycorrhizal plantstake up ammonium at faster rates than nonmycorrhizal plants, but do notrelease H+ at a faster rate. This pH relationship suggests that hemlock mayhave exploited a niche on acid soils, the higher acidity possibly reducingcompetition for available nutrients from bacteria, other fungi and otherplant roots (Rygiewicz et al. 1984).This may be a reason why western hemlock is so closely associatedwith rotting wood. Regeneration of western hemlock has been reported tooccur frequently on stumps and prostrate logs (Fowells 1965; Minore 1972),with nurse-logs and stilt-rooted trees being conspicuous in the hemlockforests of the Pacific Northwest (Franklin and Dyrness 1973; Christy andMack 1984) (Figure 2.1). Christy and Mack (1984) and Harmon andFranklin (1989) concluded that the role of decaying logs is to provide27Figure 2.1. Western hemlock growing on the side of a blown-over Sitkaspruce, exhibiting the stilt-root characteristic. The location is theCarmanah Valley on Vancouver Island, British Columbia. (Photoby A. Inselberg)28elevated safe sites in a forest understory where seedling establishment onthe forest floor is thwarted by litter burial and competition with herbs andmosses. Organic matter also seems to stimulate root branching, resultingin dense mats of fine roots near the soil surface. Eis (1974, 1987) foundthat the greatest concentration of fine roots was in the organic horizon andthe top 10 cm of mineral soil, and roots also tended to follow decaying rootsof the previous forest or buried rotten wood.Although rotting wood is nutritionally a substantially poorer substratewhen compared to mineral soil (Harmon et al. 1986), there is evidence thatwood with a high moisture content and in an advanced stage of decay is apotential environment for nitrogen fixation by bacterial asymbiotes (Larsenet al. 1978; Spano et al. 1982; Jurgensen et al. 1984, 1987; Harvey et a/.1989). Larsen et al. (1978) suggested a strong nutritional relationshipbetween mycorrhizae and nitrogen fixation. Only a small amount fixedseems likely (a few kilograms per hectare per year), although this amountrepresents a net input to the ecosystem. However, the accuracy formeasuring such low-level nitrogen fixation, as expressed by the technique ofacetylene reduction to ethylene, has been questioned (Silvester et al. 1982;Harmon et al. 1986).Ectomycorrhizal association is a widespread phenomenon, infectionbeing a normal and regular event in nature (Richards 1987). The benefits tothe plant are through enhanced water, nutrient (particularly phosphorus)uptake, the production of enzymes, and possibly with pathogen resistance(Marks and Kozlowski 1973; Laursen 1985). It has been observed thatphosphorus and nitrogen levels are enhanced in mycorrhizal plants, andsince mor humus forms have large reserves of organic phosphorus andnitrogen, mycorrhizal infection could be of importance if it provides access29to these reserves (Read 1983).Specifically for western hemlock, fifty fungi have been demonstratedto form ectomycorrhizae in pure culture synthesis (Molina 1980; Kropp1982a; Kropp and Trappe 1982; Molina and Trappe 1982). An additional102 were considered to be probable mycorrhizal formers based on fieldobservation (Molina 1980b; Kropp 1982b, 1982c; Kropp and Trappe 1982).This probably does not represent the total potential mycorrhizal fungiassociated with western hemlock as most were found as the opportunity forfield work arose (Kropp and Trappe 1982).Most of the mycorrhizal fungi were non-host-specific, unlike thoseassociated with earlier successional species such as alder or Douglas-fir,which seem to be host specific. Kropp and Trappe (1982) hypothesized thatlate successional species, such as western hemlock, enter and adapt to themycorrhizal system already established with the overstory hosts. Hence,the selection pressure would be against hemlock specificity. However,rotten wood seemed the most likely substrate for evolution of hemlock-specific fungi, if they do exist, since this is a rather specialized microhabitatin which fungi might encounter relatively little competition from previouslyestablished mycorrhizal fungi. Harvey et al. (1986) reported a trend ofreduced active ectomycorrhizal short root types in the deep mineral fractionand high numbers in the organic fractions, particularly humus and decayedwood. They suggested that the apparent ability of mycorrhizal fungi todetoxify soil phenolics may contribute to the ability of conifer roots to thrivein decayed wood on and in forest soils.Krajina (1969) suggested that western hemlock is also adapted tonitrogen supply in the form of amino acids. This may be mycorrhizamediated since some mycorrhizal fungi have been reported capable of30utilizing simple organic nitrogen in laboratory culture trials (Lundberg1970). Stribley and Read (1980) have demonstrated, in sand culture, thatyoung mycorrhizal plants of Vaccinium macrocarpon, an oxylophytic speciescharacteristic of mor-humus soils of low nitrogen availability, could utilizeamino acids as a nitrogen source. This capacity was a specific feature ofmycorrhizal infection. Evidence also confirms this uptake mechanism forthe simplest of the organic phosphorus sources (Mitchell and Read 1981).However, the significance of these for total nitrogen and phosphorus uptakeunder field conditions is yet unknown (Raisin et a/. 1987), and notinvestigated specifically for western hemlock.Western hemlock has been noted to exert an influence on soildevelopment, especially as it pertains to podzolization. Crampton (1982,1984) reported differences in the thickness of the Ae horizon in variouslocations under the canopy of individual western hemlock trees. Lowe andKlinka (1981) found that productive growth of western hemlock wasassociated with chemical indicators of podzol development and of lowbiological activity. However, they cautioned that the correlation did notestablish either a cause or an effect. On three study sites, Alban (1969),compared soil properties between western hemlock and western redcedargrowing on the same site, and reported larger values of pH, calcium andcation exchange capacity under western redcedar. Turner and Franz (1985)found there were significantly lower (p<0.05) total microbial counts, numberof ammonium oxidizing bacteria, and significantly higher fungal sporecounts in the litter and Al horizon associated with western hemlockcompared to western redcedar.The most common criterion for defining nutritional status is growthresponse to fertilization (Binkley 1986). For western hemlock, the response31(generally basal area or radial increment increase) to nitrogen fertilizationwith urea has been inconsistent, typically with wide ranges of responsesand frequent negative responses being recorded (Webster et a/. 1976; Olsenet al. 1979). Growth responses range from increases of 50 per cent or moreto apparent reductions of about 20 percent. Webster et al. (1976) concludedthat Inland Washington responses are generally more positive than CoastalWashington, although the range of responses is still large. However, Olsenet al. (1979) noted that strong geographical trends were not apparent. Againthere are conflicting reports on the effect of stand condition on the responseto nitrogen fertilization. Webster et al. (1976) stated that generally, spacedstands respond positively to fertilization while Olsen et al. (1979) notedcomparable responses between unspaced and spaced stands. In agreenhouse study, Radwan and Debell (1980b) suggested that the source ofnitrogen in the fertilizer does not appear to be responsible for the reportedvariability in response of natural stands. Radwan and DeBell (1989) foundsignificant basal area and volume growth with the use of sulphur-coatedurea, but attributed the effect to the slow release of nitrogen from thefertilizer. Greenhouse fertilization studies (Heilman and Ekuan 1973;Anderson et al. 1982) and plantation fertilization trials (Gill and Lavender1983b; Radwan and Shumway 1983) suggest that low supplies of otherimportant nutrient elements, such as phosphorus, may be important inexplaining the lack of success with nitrogen fertilization of western hemlock.However, no clear trend has emerged in the relationship between responseto fertilizers and soils, site class, or site index. Gill and Lavender (1983a)found that urea fertilization of hemlock stands initially increased mortalityof mycorrhizae and then changed relative populations of mycorrhizal types.They concluded that significant changes in total mycorrhizae and in relative32populations of mycorrhizal types after fertilization could substantially affectthe nutrient status and growth of western hemlock.2.7 Literature Review SummarySite index is the most widely accepted measure of productivity, and ismore related to the capacity of a given site to produce wood than any othersingle measure, besides volume itself. Studies relating site index toindividual physical and chemical properties factorially has demonstrated theproblem of multicothnearity. However, the BEC system provides aframework within which interacting site factors are integrated heuristicallyinto site units, or integrated through plants acting as "phytometers".Soil-site studies with western hemlock, have demonstrated somesuccess, but there are no consistent nutrient variables between studies.Research into western hemlock nutrition indicate that western hemlockproductivity will likely be positively correlated with ammonium production.Productivity should not show a relationship, or may even show a negativerelationship, to areas with high nitrification. Western hemlock relationshipswith decaying wood, mycorrhizae, amino acid uptake, and nonsymbioticnitrogen fixers are yet to be clearly understood. All of this adds to theinconsistent results of fertilization studies and the elusive nature of westernhemlock nutrition.3. METHODS3.1 Study AreaThe study sites were located in three areas of southwestern BritishColumbia in the Vancouver Forest Region. Two areas were on VancouverIsland, the first near the municipalities of Gold River, Tahsis and Zeballos,the second near the municipality of Port MacNeil, and the third area waslocated in the Seymour Valley, near Vancouver (49° North latitude and 123°-126° West longitude).Most plots were within the Submontane Very Wet Maritime CoastalWestern Hemlock (CWHvm1) variant but some plots were in the Outer VeryWet Hypermaritime (CWHvh 1) and Western Very Dry Maritime (CWHxm2)variants (Klinka et at 1984; Klinka et at 1991). The use of mainly onevariant, with extreme atonal sites only being sampled from adjacentvariants, concentrated analysis on the within-variant effects.The climate is marine, with a relative lack of sunshine, cool summersand mild winters, with heavy precipitation concentrated in the winter.Mean annual precipitation ranges from 1500 to 4400 mm with less than15% as snowfall (Valentine et al. 1981; Pojar and Klinka 1983). The CWHzone has been characterized by Krajina (1969) as part of the Pacific CoastalMesothermal Forest region, within the Mesothermal formation. UsingKOppen's classification of climates (Trewartha and Horn 1980) this would beclassified as Cfb, mild temperate rainy climates, with mild winters, nodistinct dry season, and cool summers.The area is located within the Vancouver Island Ranges and EstevanCoastal Plain of the Vancouver Island Mountains and the Nahwitti Lowland3334in the Western System of the physiographic subdivision, the CanadianCordilleran Region (Holland 1976). This area is characterized mainly byvolcanic rock with numerous granitic batholiths, and some sedimentaryrock. Surficial materials that form the parent material of the soils in thisarea were formed during and since the time British Columbia was coveredby the Cordilleran Glacier Complex which disappeared at the end of thePleistocene some 10,000 years ago (Armstrong et al.. 1965; Fulton 1971).The most extensive of the surficial material encountered in this study wasPleistocene till. This till is a non-sorted and non-stratified sediment whichcontains a heterogeneous mixture of particle sizes which includes largeboulders. Fluvioglacial materials, deposited by glacial meltwater, fluvialmaterials transported and deposited by rivers, and colluvial materials,products of mass wasting, were other surficial materials encountered.The main features distinguishing the soils of the coastal area are theprevalence of deep reddish to yellowish brown B horizons enriched withsesquioxides and organic matter, the absence of a continuous Ae horizon,and thick mor forest floors (Lewis 1976; Valentine et al. 1981). Rootchannels (Martin and Lowe 1989) were common in the Bf horizons. Thesoils in the study plots were mainly Ferro-Humic Podzols, and Humo-FerricPodzols, with some Dystric Brunisols and Humic Folisols (AgricultureCanada Expert Committee on Soil Survey 1987). The forest floors of thestudy plots were classified mainly as Hemihumimors and Humimors, withsome Mormoders, Leptomoders and Mullmoders (Minka et al. 1981). Themajority of forest floors were thick (up to 60 cm) and included woodymaterial in various stages of decomposition.3.2 Study SitesSites were selected to represent a wide range of environmentssupporting the growth of western hemlock. The majority of the ecosystemsselected for study supported naturally established, even-aged, immature(the range of ages of the stands was approximately 30-100 years old),uniformly stocked, western hemlock stands. Fourteen plots were located onextremely low productive sites in non-forested ecosystems to establish verylow productivity sites for comparison. Twenty plots from a fertilizer trialdata set were located in stands that had been juvenile-spaced. All treeswere free of damaging agents, and the plots represented the nearly completeinferred soil moisture and soil nutrient gradients for the growth of westernhemlock within the CWHvm1 variant. The sites sampled outside this variantwere sites needed for poorly represented soil moisture and nutrientgradients. These sites were azonal, thus are mainly controlled by localconditions (Mueller-Dombois and Ellenberg 1974). Consequently, the plotswill be combined and used as if sampled from the CWHvm 1 variant. Notfound, and therefore not included, in this study were rich sites which wereslightly dry through extremely dry and very rich, fresh to moist sites.Within each ecosystem, a representative sample plot of size 0.04 ha.was used based on the recommendations for relevê sizes by Mueller-Dombois and Ellenberg (1974) for temperate forests. Plots were locatedsubjectively without preconceived bias (Mueller-Dombois and Ellenberg1974) for typification and pattern analysis of a segment of the landscapejudged to be relatively uniform in climate, soil, and plants (Knapp 1984;OrlOci 1988; Kenkel et at 1989). For the non-forested extreme ecosystems,selected open-grown hemlock were chosen for study with sampling and35description based on a subjective assessment of an area thought to beoccupied by the root system.3.3 Sampling and Site DescriptionDescription of vegetation followed the methods outlined in Pojar et a/.(1987). Vegetation description included identification of all vascular plants,bryophytes, and lichens and estimates of cover for each species using theDomin-Krajina cover abundance scale (Krajina 1933 cited in Mueller-Dombois and Ellenberg 1974) modified for local conditions. The use ofsignificance class values allows both quantitative (density) and qualitative(presence-absence) information to be expressed without either dominatingthe other (Gauch 1986) (i.e., not complete equality of presence as in theformer measure nor complete equality of species as in the latter measure).Species growing as epiphytes and on decaying coarse woody debris orcoarse fragments and rock were not included in the analysis.Complete site description (elevation, aspect, slope position, shape,gradient, etc.), followed the procedure of Walmsley et al. (1980) [revised byLuttmerding et al. 1990]. Relative soil moisture regimes (SMRs) and soilnutrient regimes (SNRs) of each plot were identified according to keys whichconsider selected topographic and soil morphological attributes (Minim et a/.1984; Banner et al. 1990). The relative SMRs were adjusted to actual SMRsusing the respective biogeoclimatic subzone to circumscribe the relativeSMR according to Banner et al (1990).Site index was used as an indicator of productivity and tree selectionwas based on top height (100 largest trees per hectare). Thus it wasdetermined from the heights of the four largest trees per 0.04 hectare plotusing the height growth curves of Wiley (1978). It is recognized that the3637Wiley curves were derived based on a sample consisting of the 10 trees oflargest dbh from a group of 50 adjacent individual hemlocks, and that thebest estimate of site index would thus be obtained by following theprocedure used in developing the curves. However, it was the opinion ofMitchell and Polsson (1987) that the difference in basing site index on topheight instead is small.The actual heights and ages measured on several sites in this studywere beyond the limit used in developing the Wiley curves, makingextrapolation beyond the intended range necessary. However, as figure 3.1indicates, the general form of the Wiley curves is maintained for the lowersite index values; therefore, such extrapolation was judged as acceptable.On the non-forested sites, where four individual trees were chosen as abasis for site index, the trees in many cases were very short. For trees ofsuch short height, age was taken at 0.3 meters above point of germinationand adjusted linearly for a breast height age to be used in the Wileyequation. Although this was using the equation well beyond the range ofthe data from which it was derived, the margin of error in estimating siteindex was very small because the height growth was so slow. For all otherforested sites, age was taken at breast height of 1.3 meters above the pointof germination.The soil description in the survey phase was qualitative, and includedthickness of forest floor, humus form, soil particle size, coarse fragmentcontent, root distribution, and both potential and an estimate of actualrooting depth. A soil pit was excavated approximately 1 m 2 in plane areaand, where possible, to the depth of the control section of each site. Soilswere identified using the Canadian System of Soil Classification (AgricultureCanada Expert Committee on Soil Survey 1987) and humus forms were38SI4035302520151052250807060-5- 504- 40tko30201020050^100^150breast height ageFigure 3.1. Western hemlock height growth curves of Wiley(1978) showing extrapolated curves (SI 2,5,10 and 15)beyond the range used in developing the curve.39identified according to Minim et al. (1981).To assess the available nutrient status of the soil, chemical analysesof both the forest floor and mineral soil were carried out. Forest floor andmineral soil (visually estimated to rooting depth) physical and chemicalproperties were examined and sampled separately at four sampling pointsselected at random. The methods prescribed by Ballard and Carter (1986)and used by Klinka et al. (1989), Carter and Klinka (1990), and Klinka andCarter (1990) for the sampling, and the physical and chemical analysis offorest floor and mineral soil were used in this study. Composite samples offorest floor and composite bulk samples of the mineral soils were taken ateach point. The bulk density of both forest floor and mineral soil wasmeasured near the point from which each soil sample was collected bycutting out a core, measuring its volume (by measuring the volume of glassbeads required to fill the resulting hole lined with a thin plastic sheet), andmeasuring its mass after oven-drying at 105° C to constant weight.There were 102 total plots available for this study. Sixty-onecomprised the original data set, and an additional 41 plots were added forpossible use as a test data set. For the original data set, samples weretaken from 3 sides of an approximate 1 m X 1 m soil pit from a depth basedon a subjective weighted ocular estimate of rooting depth. For the test dataset, soil samples were taken from three points of an equilateral triangle (2 mon a side), and from a depth of 0-30cm.Although most studies of nutrient availability examine only the uppermineral soil (Binkley and Hart 1989), the decision to sample to rootingdepth was based on some conflicting results of nitrogen mineralization withdepth and extractable phosphorus with depth. For the former, typicallymineralization in forests generally decreases with depth, although major40exceptions occur (Binkley and Hart 1989). Binkley (1983) found that in aDouglas-fir plantation, the nitrogen mineralized during anaerobicincubations actually increased with depth. The same is noted for samplingphosphorus where Ballard (1980) noted that in published reports thesurface layer of soil is generally better correlated to yield or foliarphosphorus. Contrary to this, Webber (1974) found that sampling from theupper or the entire profile did not effect results appreciably.Because of these conflicting results, it was decided to allow the fineroot location determine where nutrients were being taken up, thusdetermining the sampling depth. Fine roots (i.e., those below 1-2 mmdiameter) are the main agents in ion uptake together with mycorrhizae(Bowen 1984). Roots in higher plants have been noted to be opportunisticin discovering and growing towards locally rich patches of water or nutrients(Cook 1983; Santantonio 1985). Experiments in solution culture havedemonstrated a localized stimulation of root growth to local concentrationsof nutrients with barley (Drew et al.. 1973; Drew 1975; Drew and Saker1975, 1978) and with Sitka spruce seedlings (Coutts and Philipson 1976;Philipson and Coutts 1977). Field experiments also concur with thislocalized stimulation of root growth, termed "tropism for nutrients" by St.John (1983). Powers (1984) noted that the distribution of fine roots seem toparallel the mineralizable nitrogen profile, Kimmins and Hawkes (1978)concluded that the distribution of fine roots is strongly correlated withvertical variations in soil fertility and organic matter, and Coopersmith(1986), using in-growth bags, found that nutrient rich growth materialsstimulated fine root production. For western hemlock specifically, studies ofroot excavations (Eis 1974, 1987) have shown that organic material seemsto stimulate root branching, and explains why the greatest concentration of41fine roots in western hemlock is associated with organic matter. Therefore,the portion of the soil profile that was to be sampled was subjectivelychosen in proportion to the amount of fine roots visually estimated in thesoil profile.Horizontal variability in morphological characteristics was recognizedand noted, especially in western hemlock sites. Also recognized was theaccompanying variability in forest floor and mineral soil chemical properties(Quesnel and Lavkulich 1980; Courtin et al. 1983). However, Carter andLowe (1986) reported that analyses of composite forest floor samples werehighly correlated with the mean values for individual samples. Theyconcluded that for most purposes, composite samples appeared to providean adequate estimate of the mean value of samples analyzed individually.Thus, composite sampling of both forest floor and mineral soil was carriedout in this study. Temporal variability was minimized by confiningsampling to late spring and summer, although for total nutrients variabilitywas expected to be small because the annual fluctuations in total nutrientsare very small relative to the total pool (Binkley and Hart 1983).Successional temporal differences of an ecosystem's nutrient retentionproperties (Vitousek and Reiners 1975) were minimized by confining theages to mainly mid-successional stages. It was recognized that theanaerobic technique is also temperature-dependent, with field soiltemperature having a strong bearing on the interpretation of the laboratorytest (Powers 1980). However, by confining the study to one variant, soiltemperature differences due to climate or elevation were minimized.3.4 Soil Chemical AnalysisAll soil samples were air-dried to a constant weight in the laboratory42and then subsamples for each plot were composited. For chemical analysis,forest floor samples were ground in a Wiley mill to pass through a 2-mmsieve, while mineral soil samples were passed through a 2-mm sieve toseparate coarse fragments. Soil pH was measured with a pH meter using a1:1 suspension in water for mineral soil and a 1:5 suspension for forest floormaterial. Total carbon was determined using a Leco Induction Furnace(Bremner and Tabatabai 1971). Total nitrogen was determined bysemimicro-Kjeldahl digestion followed by colorimetric estimation ofammonium (NH4) (Bremner and Mulvaney 1982) using a TechniconAutoanalyzer (Anonymous 1976). Mineralizable nitrogen was determined byan anaerobic incubation procedure modified from Waring and Bremner(1964). Released NH4 was determined colorimetrically by use of a TechniconAutoanalyzer. Total phosphorus in the forest floor was measured by adigestion procedure (Bray and Kurtz 1945) followed by a colorimetricdetermination of the phosphorus in the digest. Mineral soil availablephosphorus was measured by the extraction procedure of Mehlich (1978).Total sulphur analyses in the forest floor material were conducted using aFisher Model 475 Sulphur Analyzer (Lowe and Guthrie 1981). AvailableSO4-S was determined by ammonium-acetate extraction (Bardsley andLancaster 1960, 1965) whereupon the extracted sulphate was reduced tosulphide by HI (Johnson and Nishita 1952) and the sulphide thus liberatedwas determined by the bismuth sulfide colorimetric procedure (Kowalenkoand Lowe 1972). Available potassium, magnesium, and calcium weredetermined by extraction with Morgan's solution of sodium acetate at pH4.8 (Greweling and Peech 1960). The extracted cations were then measuredby atomic absorption spectrophotometry. Pyrophosphate-extractable ironand aluminium in the B horizon were extracted overnight at 25°C using43sodium pyrophosphate solution as described by Bascombe (1968).Extracted iron and aluminium were then determined by atomic absorptionspectroscopy (Lavkulich 1978). Soil laboratory analysis was carried out byPacific Soil Analysis Incorporated which, based on past performance (R.E.Carter pers. comm. 1), had met the criteria for laboratory selection asoutlined in Ballard and Carter (1986).Anaerobic incubation for mineralizable nitrogen was chosen assuggested by Powers (1980), mineral soil available phosphorus wasdetermined using the new Mehlich method as suggested by Curran (1984),and Morgan's solution for available cations was used as suggested by Minimet al. (1980).Soil nutrient variables were expressed as concentrations on a drymass basis as opposed to the traditional use of mass per area basis. Thelatter calculation uses bulk density corrected for coarse fragment contentand represents mass of nutrient per hectare in the forest floor and themineral soil. Concentrations were used in this study because(i) in many cases the accuracy of field estimates of bulk density andsoil depth in soils derived from tills was hindered due to thepreponderance of coarse fragments and boulders. Mass per areacalculations were attempted but the depth and bulk densityintroduced considerable extra "noise" into the data;(ii)with western hemlock where deep mor humus forms are common,an expression in kg/ha for total nitrogen, for example, will giveundue emphasis to the storage of nutrients in the more slowlydecomposing thick mor humus forms, versus the relatively high1 R.E. Carter, Resource Analysis/Evaluation Forester Timberlands and Forests, Fletcher ChallengeCanada, Vancouver, B.C.44availability of rapidly decomposing, thus very thin, mull humusforms containing little nitrogen on a kg/ha basis;(iii)appreciating the view of Mehlich (1972) in proposing a uniformsystem for calculating and reporting soil analytical results foragricultural soils, the vastly differing humus depth mentionedpreviously and vastly differing soil depth, from organic matter overbedrock to deep alluvial soils, does not lend itself to the agriculturalconcept of a "cultivated plow layer" with a uniform 20 cm or 30 cmdepth;(iv)one objective of this study is to compare the soil chemicalmeasures to the field derived SNRs. The heuristic field procedureused mineral soil depth greater than 30 cm and coarse fragmentcontent in deriving SNRs. Mass per area calculations also use bothdepth and coarse fragment content. Soil nutrient measuresexpressed as concentrations on a dry mass basis are completelyindependent of the heuristic field key to which they are beingcompared.3.5 Vegetation Data AnalysisFor investigating vegetation relationships, the data collected by theauthor (61 plots) in 1988 were combined with data collected in 1987 (21plots) by Kabzems(1988). Study plots were classified into vegetationassociations using the methods described by Pojar et al.. (1987) and Minim.et W. (1989). This vegetation classification is based on the Zurich-Montpellier Tradition, using the tabular analysis method of Braun-Blanquet(Braun-Blanquet 1932; Becking 1957; Poore 1955; Shimwell 1971; Mueller-Dombois and Ellenberg 1974; Westhoff and van der Maarel 1980) modified45for use within the BEC system (Krajina 1969; Pojar et a/. 1987). Since theBraun-Blanquet method is, in a sense, conceptually between the organismviewpoint of Clements (1916) and the individualistic viewpoint of Gleason(1926), what Mueller-Dombois and Ellenberg (1974) have termed the"systematic viewpoint", both classification into communities and the use ofordination techniques can be used.The releves were compiled and sorted into floristically similar groupsand classified into a hierarchy of vegetation units using the computerizedtabling program VTAB (Emanuel 1989). Cluster analysis, using Ward'serror sum of squares method (Ward 1963) with the ordering procedure ofGruvaeus and Wainer (1972), and the reciprocal averaging technique (Hill1973) using a strategy of progressive data set fragmentation (Peet 1980) wasused to aid in identifying groups and hierarchies. To independentlyexamine the strength of the derived hierarchy, and as an hypothesissupporting technique, species were analyzed with centred, non-standardizedprincipal components analysis (PCA) ordination (Pearson 1901; Hotelling1933; Nichols 1977; Dillon and Goldstein 1984; Plelou 1977, 1984). Thecoordinates of the first and second axes were graphed, elliptical outlinescentred on the group means, oriented in the direction of maximum variationand drawn to encompass 80% of the plotted points of the respective groups,were overlain and then the entire pattern visually inspected. Ellipticaloutlines were used as polygons and constructed using a variation ofconfidence ellipse construction (Jolicoeur and Mosiman 1960; Owen andChimielewski 1985) where bivariate normality was not assumed; therefore,without statistical inference (Crovello 1970). This technique gave objectivityto the method of using polygons of various sizes and shapes.To see if the derived hierarchy represented some sort of gradient of46underlying environmental factors, spectral histograms of indicator specieswere constructed for each group in the derived hierarchy. Percentfrequency was calculated for an indicator species group (ISG) for a siteattribute k (e.g. soil nutrients), and is calculated according to (Minia et al.1989):F^.1( ni riC ..)NO()) k.^( )ji1.1 1=1 1=1F(k)j = percent frequency for site attribute k and ISG jCkii =is the midpoint percent cover for indicator species i for ISG j for a site attributek= indicator speciesj^= Indicator Species Groupk^=site attributen = number of indicator species in ISG jm =number of ISGs for attribute kThe derived spectra for each group were then compared to see the trends ofunderlying environmental factors.In order to describe soil chemical relationships with the vegetationand vegetation units, canonical correlation analysis (CCA) (Gittins 1979,1985; Dillon and Goldstein 1984; Tabachnick and Elden 1989) was used forthe former and canonical discriminant analysis (Gittins 1979, 1980; SAS1985) for the latter. The objective of CCA is to find a linear combination ofindependent variables that maximally correlates with a linear combinationof dependent variables. A check for multivariate normality was done byplotting the normalized Mahalanobilis distance D2 to the power 1/3 againsttheir expected order statistics for a chi-squared distribution with v d.f.,where v is the number of variables (Healy 1968; Campbell 1980; Seber1984). It should be recognized that this analysis is for descriptive purposes47only, so no distributional assumptions are required (Dillon and Goldstein1984); however, the analysis is enhanced (i.e., the description is capturingas close as possible the "true" relationship) if the assumptions are met(Tabachnick and Fidell 1989). No attempt was made to apply a multivariatetransformation to normalize the data since the analysis is used primarily fordescriptive purposes. If a transformation were to be made, it was felt thatthe parameters associated with the transformed data would not be asmeaningful as those with the original data (Seber 1984). However, it isinformative to plot the set scores of the CCA to demonstrate graphically howthe sets differ from one another (Pielou 1977).Before the relationship between the vegetation and soil chemicalproperties could be examined, the large number of species and chemicalproperty variables had to be reduced. This was done by performing centredPCA on the vegetation covariance matrix and the forest floor chemicalcorrelation matrix to express the overall vegetation composition of the sitesand the overall nutrient status of the forest floor. Canonical R 2 andredundancy were inspected to determine the correlation between thevegetation and forest floor sets.To describe the relationship between the vegetation units derived fromthe hierarchy and the soil chemical measures, canonical discriminantanalysis was used. This procedure is a canonical analysis where thevegetation units are in the form of binary valued "dummy" variables. Toreduce the number of soil chemical property variables, PCA was carried outon the correlation matrix. Soil chemical properties correlated with the mostimportant canonical variates were noted through correlations with therespective PCA axes. Canonical variates of the important axes were plottedand elliptical outlines drawn to encompass 80% of the plots in therespective vegetation units and visually inspected.3.6 Soil Nutrient Regime/Soil Moisture Regime AnalysisSNRs and SMRs of the study plots were identified based on aheuristic evaluation of physical properties of the plot and augmented by theindicator plants using the method described by Polar et al. (1987) andElinka et al. (1989). Plots were then assigned to a site series using themethod of Banner et al. (1990). Of the 61 plots originally collected, 6 plotsconsisting of humus forms over bedrock were eliminated since mineral soilchemical properties were non-existent. The remaining 55 plots formed oneset to derive relationships. A second data set was available for modelvalidation consisting of 21 plots collected by Kabzems (1988) and 20 plotsfrom a fertilizer study collected in 1989. These test plots were similar instand and site characteristics to the original plots, although collected from adifferent area.In order to explore whether SNRs as derived from the field heuristicprocedure prescribed by the BEC system were related to direct measures ofsoil nutrients, PCA was first used for exploratory purposes followed bycanonical discriminant analysis and discriminant analysis as a confirmatoryprocedure. To view the structure of the original data set, PCA was carriedout on the correlation matrix. The PCA scores of the important axes wereplotted and elliptical outlines were drawn to encompass 80% of the plots,and then visually inspected. Loadings of the soil chemical measures withthe PCA scores were inspected and the ones having the strongest correlationwere chosen for further analysis.Canonical discriminant analysis was then carried out on this reducedset and the canonical variates of the important axes plotted and elliptical4849outlines drawn to encompass 80% of the plots in the respective SNRs andvisually inspected. This display reveals the shape and extent of the scatterof each group, as well as the extent of the overlap or separation between thegroups (Gittins 1985). The soil chemical measures correlated with thecanonical variates were then chosen to be used in a discriminant analysis tosee how well these specific soil chemical properties can be used to predictfield derived SNRs. The data were checked for outliers by a visualexamination of plotted PCA scores, the multivariate normality checked asdescribed previously, and the equality of group dispersions checked using amultivariate generalization of Bartlett's test of homogeneity of K populationvariances (Tatsuoka 1988). Since discriminant analysis is a confirmatoryprocedure (Tukey 1980; Williams 1983), transformations to normality weremade when needed. Additionally, if the data exhibit multivariate normality,but lacks homogeneity of within covariance matrices, then Smith's (1947)quadratic function was used as the optimum rule to predict groupmembership, otherwise Fisher's (1936) linear discriminant function wasused. Since prior probabilities influence the forms of discriminant functions(Williams 1983), and the priors were not known, the discriminant analysiswas run first assuming equal prior probabilities and then with priorsassigned based on ancillary experience and knowledge. The validation setwas used with the two discriminant functions to test the portability of theresults.The rationale behind the above procedure was that if the field derivedSNRs are meaningful in terms of actual measured soil chemicals, PCA, asan exploratory technique, will indicate which chemical measures show thelargest variation. Canonical discriminant analysis will describe therelationship of the most meaningful chemical measures with the SNRs.50Finally, discriminant analysis and the subsequent testing of thediscriminant function on a validation data set will provide evidence, as aconfirmatory procedure, that the field derived SNRs are related to the actualsoil chemical measures chosen.Since comprehensive information gathered from an extensive soilwater measurement program was not available, the comparison of SMRs,derived by the heuristic method of Banner et a/. (1990), was made againstthe results of a simulation model. The Energy-Soil Limited (ESL) waterbalance model (Spittlehouse 1981; Spittlehouse and Black 1981) was usedto calculate the actual evapotranspiration and growing season water-deficitin order to estimate actual SMR of water-deficient (excessively dry throughslightly dry) and fresh sites. The depth to the soil gleyed horizon or thewater table was used to estimate the actual SMR of moist through wet sites.The ESL water balance model is driven by solar radiation,temperature and precipitation and uses soil depth and texture data tocalculate available water storage capacity. The coefficients used in themodel were derived by Giles et al. (1985) for use in immature stands ofcoastal Douglas-fir in the Eastern Very Dry Maritime variant of the CWHzone. This model is energy-limited until approximately 60% of the availablesoil water has been utilized, when it becomes increasingly soil-water-limited.Actual evapotranspiration was calculated as monthly totals during thegrowing-season (April-September) and as a growing-season total.The ESL model has several limitations as outlined in Carter andKlinka (1990). Further, calibration of the model has not been carried out inthe biogeoclimatic variant used in this study for immature stands of westernhemlock. However, the model results compared to field derived SMRs isuseful as a preliminary estimate.513.7 Analysis of Productivity RelationshipsThe investigation of site index relationships was divided into twogroups, the first using vegetation measures and the second usingenvironment measures. The relationship of site index with the vegetationmeasures was subdivided into two indirect measures: (1) categorical:vegetation units; and (2) analytical: PCA scores on vegetation, and frequencyof indicator species groups. The environment measures were subdividedinto an indirect measure and a direct measure, which were respectively: (1)categorical: SMR, SNR, SMR & SNR, and site series; and (2) analytical: soilnutrient measures.The relationship of site index to the plant associations was examinedusing "dummy" variable regression, where the independent variables wereplant associations and binary Pi coefficients equaled 1 if the site indexbelonged to plant association j, and 0 otherwise (Suits 1957; Chatterjee andPrice 1977; Ott 1988). The 67 plots used in the diagnostic vegetation tablederivation were used. The residuals were then investigated for normality bya visual inspection of a normal probability plot of the residuals (Chamberset a/. 1983), and homogeneity of variances were visually examined by a plotof the residuals.To investigate how the individual species were related to site index,regression analysis was carried out. To reduce the number of independentvariables from 120 species, PCA scores, which are a linear combination ofall species, were used instead of individual species. Only those axes whoseeigenvalues were greater than one were used in a backward stepwiseregression. The rationale behind this rule of thumb is that any componentshould account for more variance than any singe variable (Dillon and52Goldstein 1984). The respective loadings of the resulting regressionequation were analyzed to ascertain the individual species relationships.This type of analysis has been referred to as principal componentsregression. Although a statistically biased procedure (Jolliffe 1986;Chaterjee and Price 1977), it was appropriate for this type of exploratoryand descriptive use.To investigate the relationship between site index and the frequency ofindicator species groups (ISGs) (Klinka et al. 1989), species having anindicator value for moisture or nitrogen were grouped into ISGs and thefrequency values were used as independent variables in a regressionanalysis.SNRs and SMRs of study plots were identified based on a heuristicevaluation of physical properties of the plot, augmented by the indicatorplants using the method described by Pojar et at (1987) and Klinka et al.(1989). Plots were then assigned to a site series using Banner et al. (1990).Site index was used as a dependent variable in "dummy" variable regressionto examine its relationship to site units derived from soil moisture regimes(SMRs) and soil nutrient regimes (SNRs). SMRs and SNRs were used incombination and separately, with the magnitude of coefficients examined, todetermine the amount of influence each had upon site index. The residualswere then investigated for normality by a visual inspection of a probabilityplot of the residuals (Chambers et al. 1984), and homogeneity of varianceswas visually examined by a plot of the residuals. For this analysis, all 102plots were combined into one data set to fill in some noted gaps in the siteunits within the edatopic grid.To investigate the relationship between site index and soil chemicalmeasures, multiple regression was used on the primary data set of 55 plots53(the original 61 plots minus 6 plots of forest floor over bedrock which didnot have mineral soil chemical properties). The remaining 41 plots werereserved as a validation test data set. Because of suspectedmulticollinearity, three tests were carried out as suggested by Neter et al.(1990): bivariate correlation coefficients were examined, variance inflationfactors translated into tolerance indices were examined, and a PCA wasperformed followed by inspection of eigenvalues. An all combinationsmultiple regression approach was then carried out with equations havingrelatively high R2 and relatively low SEE noted. A "best fit" equation waschosen from inspection of all combinations based on higher adjusted R 2 andlower SEE, and factors that made "ecological sense". A 5% level ofsignificance was required in all model development. As an additional test ofmodel performance, the resulting equation was used to classify the originaldata into classes. Predictions within 3 m of the recorded site index wereconsidered correctly classified. Predictions between 3 m and 6 m of therecorded site index were considered one class off, and so on. The equationwas validated in the same manner on the test data set.Further, to supplement the all combinations multiple regression, PCAregression (Chatterjee and Price 1977; Jollife 1986; Morzuch and Ruark1991) was used. Loadings of the actual soil chemical measures withsignificant PCA axes scores derived from a backward stepwise procedurewere investigated to determine the soil chemical measures most highlycorrelated with site index.Additionally, an all combinations regression was done to explore therelationship between site index and forest floor chemical measures only.This was performed since the roots of western hemlock have been noted tooccur mainly in the forest floor. Finally, a regression was done to relate siteindex to mineral soil pyrophosphate extractable iron and aluminium, andtotal carbon. Although these three chemical measures are not nutrients,their amount in the B horizon has been reported to reflect the degree ofpodzolization (Lowe and Klinka 1981).For all multivariate statistical analysis and the all combinationsregression technique, the SAS statistical package (SAS 1985) was used.SYSTAT (Wilkinson 1990) was used for univariate statistical testing andgraphics.544. RESULTS AND DISCUSSION4.1 Vegetation ClassificationPreliminary use of cluster analysis and reciprocal averagingordination failed to produce obvious groupings of relev6s. This indicatedthat the vegetation derived from these second growth, mostly closed canopy,forests were relatively homogeneous (besides the small number of outliers).Therefore, the vegetation demonstrated low between plot species diversity,termed low 5 diversity by 'Whittaker (1975) (Figure 4.1). Tabular analysiswithout the use of further ordination was then employed following thetraditional Braun-Blanqet classification procedure based primarily on theuser's judgment and experience (Shimwell 1971).Four alliances and fourteen plant associations were distinguished andcharacterized in a diagnostic table (Table 4.1). A major portion of therelevès were concentrated in four major associations, the Gaultheria,Gaultheria-Polystichum, Polystichum, and Rubus plant associations having14, 11, 30 and 10 releves respectively. Only 1 to 3 releves were located ineach of the remaining plant associations.To test the strength of this derived hierarchy, and as an hypothesissupporting technique, all species were analyzed with centred non-standardized PCA ordination. The first and second axes of PCA scores weregraphed, elliptical outlines were overlain and then the entire pattern visuallyinspected (Figure 4.2). Points in associations having 1 or 2 plots only wereplotted without elliptical outlines.The first PCA axis explained 21% of the total variation and the secondPCA axis a further 12%. The elliptical outlines showed an ordered5556Figure 4.1. Immature western hemlock stand characterized by a closedcanopy and lack of understory vegetation.57Table 4.1. Diagnostic table identified from western hemlock plant communities on Vancouver Island,B.C.Vegetation unitNumber of plotsVegetation units and speciesTH-CaultherlaGaultherla shallonTH-MahoniaAchlys trlphylleUrines* borealisMahonla nervosaPfnus monticolaTM-Thule mall.Thuia plicateTH-Rhacomi trium a.Cledlna sp.Ofcranum sp.Pious monticolaPleurozfum schreberfRhecomitrlum canexcensTH-Vacclnlum(ovatum) a.Vaccfnlum ovatumTH-Gaultheria e.TH-Gaultheria-Polystichum a.Polystichum munitumTH-Chameacyparis a.Chamaecypario nootkatensfsTsuga mertenslanaVaccInfum aleshaenseTH-MalUs a.Afnus ruby.Malus fuscaissue brevilollaTH-Polystichum all.Blechnum Wean(Polystichum munitumTH-Polystichum sall.,a.TH-Rubus all.Rubus spectabilisTfarella trifoliateTH-Achlys a.Ables amabillsAchlys triphyllaAlnus rubraAruncus dfolcuslfnnaea borealisMahonia nervosaTrientalls latlfollaTH-Rubus a.Diagnosticvalue'2223^•^53^4^53^14^11Presence class' and mean species6^76^71^2significance.5a30991lo1010 2 111 2(dd) • 5 * II 15 7^5 7 5 6^V 7 V 5^5 9^ I"I(d) 5 64 3(d) 4 2 I • I • 5 3(d.c) 5 I I • I • I • 5 4(d.c) 5 3 5 3 5 2(d.cd) 3 J^5 2 II 3 5 53 2 5 5^IV 4 III 3^5 7^5 6 II(d) I +3 2(d.c) 5 4 I • II •(0.0 5 3 5 3 5 2-(d.c) 5 4 I •(d.c) 5 4 I •(dd) 5 I • I • I •5 7^I^I(d.c) I V 4 3 5 V 6 5 6 5 2II • V 2(d.c) 3 2 II 1 5 4(d.c) I • 5 2fcf,cd, II 3 II 2 5 7 III 2 III 3 5 4 5 5(d.c) 2 4 I • I 1 I f 5 4 II 35 4(d,c) 5 2 I • I •(d) I • 3 2(d.c) 2 2 II 1 IV 4^5 2 V 3 5 2 V 4 3 3 5 5(d.cd) II • V 2 V 4 5 5 v 6 5 6 5 2(d.cd)! I • II • 5 3 V 4 5 8 5 4(d) II • 5 6 III 1 3 1 5 5(d.c) I 1 II 3 I I II 2 35 2(d.cd) 4 3 5 6(d,c) 2 4 I • I I 5 4 I I 5 4 II 3(d.c) 5 3(d.c) 4 2 I • I • 5 3(d,c) 5 1 I * I • 5 4(d,c) 5 2TH-Saumbucus a.Hypnum cIrcfnale (d) I • I • I • 3 2Rubus pervfflorus (d) I • 3 2Sambucus raccoons (d.cd) 5 6TH-Veretrum e.Athyrfum fillx-femlne (d.c) • II 2Gymnocarpium dryopterfs (d.c) I • 3Ranunculus sp. (d,c) 4Trautvetterla carollnional• (d.cd) I • 3 5 6Verst rum vivid. (d,c) 4Viola sp. (d,cd) 6TP-SphagnumDrosera rotundifolla (d,c)Souris crIsta-pall) (d.cd)Lyslchltum mmerfcanum (d,cd)Men:testa Ferrugines (d.c) 2 • 4 2 II 3 5 2 I +Sphagnum glrgensohnll (d.cd) I • 5 1 • 5 3Sphagnum squorrosum (d.cd) 5 3TP-ledum ell.,a.Aulacomnfum Palustre (dd)EmPetrum ',forumledum groenlandlcum(d,cd)(d.cd)Pleurozfum schreberf (d.cd) 5 4 I •Vacclnfum oxycoccos (d,cd)Table 4.1 (continued)Vegetation unit^ 13^191313^14INumber of plots^Diagnost^ 1Vegetation units and species^value,^Presence class , and mean species significance'5804-Caultherie all.Gaultherfa SheltonTH - mahonta sall..a.Achlys triphyllaLinos,. borealisMahonfa nervosaPlnus monticoleITT - Thu(8 sell.Thuia offsetsTH-Rhacomitrium a.Clacline sp.Ofcranum SP.Pfnus moot/colaPleurozium schreberlRhacomitrfum canescensTH-Vaccinfum(ovatum) a.Vaccfnium ovstumTH-Geultheria a.TH-Gaultheria-Polystichum a.PolystIchum munftumTH-Chamaecyparis a.Chamsecyparis noothatensisbugs mertenslanaVaccfnium alasifsenceTH-Malus a.Alnus rubesMalus fuscataxus breyffolisTH-Polyetichum all.Olechnum splcantPolystfchum munftumTH-Polystichum sall.,a.TH-Rubus all.Rubus SpectabiliSTfarella trffolfataTH-Achlys a.Ables amabIllsAchlys triphyllaAlnus rubesAruncus dlolcusLim-lass borealisMahonia nervosaIrlentalls latlfollaTH-Rubus a.TH-Saumbucus a.Hypnum cfrcinaleRubus parviflorusSambucus racemosa• .. 4.1 •Athyrfum Mix-famineGymnocarpfum OryopterlsRanunculus sp.Trautvetterla carolinfensfsVerstrum vlrldeViola sp.TP-Sphagnum all....Orosera rotundffollaFaurfa CrIsta-galllLysichitum amerieanumMend testa ferrucilnesSphagnum girgensohnllSphagnum squarrosumTP-Ledum all.,a.Aulacomnfum palustreEmpetrum nigrumleclum groenlanOlcumPleurozfum schreberlVaccinfum oxycoccos(dd)(d)(d)(d.C)(d.c)(d,cd(d)(d.c)(d.c)(d.C)(d.c)(dd)(d, c)(d.c)(d.C)5553635^35^45^6(d.cd) 5 5(d.c) 5 I(d,c)(d)5^•(d.c)1(d.cd)(d,cd)(d)(d)(d)(d,cd)(d.c)(d.c)(d.c)(d,cd)(d.c((d.cd)(d.c) 5 •(d,cd) 5 5(d.cd) 5 6(d,c) 5 4(d.cd) 5 7(d.cd) 5 7(dd) 5 • 5^5(d,cd) 5^8(d.cd) 5^8(d.cd) 5^6(d.cd) 5^5(d.c)(d,cd)(d.c)(d.c)(d.C)(d,C)(d.c)559642'-'-cX)CV-,--1•—.....CVm0M—2—4—6—4^—2^0^2^4^6Axis 1 (21%)Vegetation Associations:R=Rubus; P=Polystichum; T=Gaultheria—Polystichum; G=Gaultheria;S=Sambucus; M=Mahonia; V=Vaccinium(ovatum); C=Chamaecyparis;X=Rhacomitrium; V=Veratrum; L=Ledum; D=Sphagnum; F=Malus;A=AchlysFigure 4.2: Ordination of plots along the first two axes of PCA on all speciesshowing 80% elliptical outlines for major plant associations Rubus (1),Polystichum (2), Gaultheria-Polystichum (3), Gaultheria (4), Mahonia (5), andVaccinium ovatum (6).60transition of the four major associations along the first component from theRubus association, grading into the Polystichum, Polystichum-Gaultheriaassociations, and into the Gaultheria association. Along the second axis,the Rubus association shows a separation from the Polystichum association.Only those species which had scores that correlated significantly (p<0.01)and equalled or exceeded a "rule of thumb" correlation coefficient of 0.3 (i.e.approximately 10% of the variation accounted for) were taken into accountand presented in Table 4.2. In total, 35 species met this criteria, with 10correlated to the first PCA axis and 18 to the second PCA axis. Five specieswere correlated to both axes 1 and 2. Gaultheria shalion and Polystichummunitum, the two species defining the two major alliances, were noted tohave the greatest positive and negative loadings on the first axisrespectively. On this axis, generally, those species in the Gaultheria alliancehad positive correlations, and those in the Polystichum alliance had negativecorrelations. On the second axis this trend is repeated, except for theaddition of the group of species having a correlation of -0.35 which occurprimarily in the Ledum plot. This was the lowest point on the vegetationPCA graph (Figure 4.2), located in the lower right quadrant. Of the 35species correlated with the first 2 axes, 24 were included in a pre-editeddiagnostic table and 15 were included in the final diagnostic table.Although the first two PCA axes combined only account for 33% of thevariation, this did provide for an independent and objective technique thatsupports the subjectively derived hierarchy. In general, the PCA of allvegetation supported the results of the vegetation table derived fromtraditional Blaun-Blanquet techniques. This is especially noteworthyconsidering the depauperate vegetation in second growth western hemlockstands.61Table 4.2: Correlation values (rul = 0.29) of species and associated PCA scores on the first 2 PCAaxis showing the respective nitrogen and moisture indicator values (Klinka et aL 1989).Symbols mean: P=poor, M=medium, R=rich, FVM=fresh to very moist, VMW=very moist towet, MDF=moderately dry to fresh, WVW=wet to very wet.n=82SPECIESLOADINGSAXIS1^AXIS2NITROGENINDICATORVALUEMOISTUREINDICATORVALUEGaultheria shallon 0.86 -0.11 PTsuga heterophylla -0.64 0.25Thuja plicata 0.57 -0.11Chamaecyparis nootkatensis 0.44 -0.02Blechnum spicant -0.40 0.28 P FVMPlagiothecium undulatum -0.39 -0.18 P FVMMenziesia ferruginea 0.39 0.03 P FVMComus canadensis 0.38 -0.03 PTsuga mertensiana 0.37 -0.12Dryopteris expansa -0.31 0.07 M FVMHylocomium splendens 0.69 0.34 PPolystichum munitum -0.65 0.47 RPinus contorta 0.42 -0.43Rubus spectabilis -0.40 0.50 R VMWKindbergia oregana -0.31 0.57 MDFPleurozium schreberi 0.30 -0.45 PRhytidiadelphus loreus 0.15 0.72 P FVMMaianthemum dilatatum -0.21 0.42 R VMWVaccinium parvifolium 0.26 0.38 PTiarella laciniata -0.09 0.37 R FVMRosa nutkana -0.11 0.37 R FVMRubus parviflorus -0.16 0.37 RTiarella trifoliata -0.13 0.36 R FVMAulacomium palustre 0.16 -0.35 M WVWEmpetrum nigrum 0.16 -0.35 PRhytidiadeiphus triquetrus 0.16 -0.35 MVaccinium oxycoccus 0.16 -0.35 P WVWLedum groenlandicum 0.16 -0.35 WVWGalium triflorum -0.16 0.33 R FVMVaccinium ovalifolium -0.05 0.33 P FVMPolytrichum alpinum -0.15 0.32Trautvefferia caroliniensis -0.13 0.32 R FVMRubus ursinus -0.07 0.32Plagiochila asplenioides -0.08 0.30Viola orbiculata -0.08 0.30 M MDF62Of interest also was whether the axes represented some sort ofgradient of underlying environmental factors as indicated either by thevegetation associations and/or by the individual plants. It was mentionedpreviously that there was a definite trend along the first PCA axis (Figure4.2) with four major associations. Spectra showing the distribution andfrequency of nitrogen ISGs for each of these four associations are given inFigure 4.3. Proceeding along the first PCA axis from the Rubus associationto the Gaultheria association, the spectral histograms changed also. Thefrequency of indicators of nitrogen-rich soils (Table 4.3) decreased from theRubus association to the Gavitheria association. The location of theassociations with increasing scores along the first PCA axis showed thatthere was a concurrent decrease in the occurrence of nitrogen-rich indicatorspecies. Comparing these frequencies to the proposed standard spectra forsoil nutrient regimes for Coastal British Columbia (Klinka et a/.. 1989), theRubus association was rich to very rich, the Po/ystichum associationmedium, the Gaultheria-Polystichurn association poor and the Gaultheriaassociation very poor. Therefore, the first PCA axis captured an inferred soilnutrient gradient, based on the frequencies of nitrogen-rich indicatorspecies for the four major plant associations, from rich to very poor. Theactual relationship between species and soil chemical measures arereported later.Spectra of soil moisture regimes (Figure 4.4) showed a predominanceof moderately dry to fresh and fresh to very moist ISGs. This suggested thatan inferred soil moisture gradient, based on frequencies of moistureindicator species for the four major plant associations, did not exist.The trend of the association between plants and an inferred nutrientgradient was also shown, but only in a general way, by looking atTH-GAULTHERIA-POLYSTICHUMTH-GAULTHERIATH-RUBUS11212%94%349%1 211 12%49%74% 24%CANC1Cs111LtiAtTH-POLYSTICHUM%COVER IPS1%CCM TENN.NUMBER OF SPECIESTOTAL^IN)^MIFF^UNCLAS142%26%39%52%4657384734412532913121433VEGETATION LNIT,Indfcator species grcups, Frequency (%)2_ SOIL AUTR1. REGIME:2 1N -medium1 N-poor3 N-rich6399%Figure 4.3. Spectra histograms, based on percentage of total cover ofindicator species of soil nitrogen and expressed as % frequency, forfour of the plant associations.TH-RtEL/53^4^539%n4 -poLYSTIcuum3^4^ I I546% 53%TH-GAULTHERIA-POLTSTICHUM337%^63%TN-GAULTHERIA3^454% 46%47% 14%SOIL MOIST. REGIME:- I Excess. to very dryVery to mod. dry2Mod. &y to freshFresh to very moist4Very moist to wetWet to very wet64 VEGETATION UNIT,Indicatcr species groups, Frequency (%)%COVER IPSf^NUMBER OF SPECIESXCOVER TOTAL. TOTAL IND INDIFF UNCLA5533%^46^29^14^331%^57^39^19^325%^38^22^1922%^47^26^20Figure 4.4. Spectra based on percentage of total cover of indicator speciesof soil moisture and expressed as % frequency. Spectra presented forfour plant associations.65correlations of individual species with each of the PCA axis scores and theassociated species diagnostic nutrient value (Klinka et a/. 1989) (Table 4.2).Of the 16 species which were correlated with PCA axis 1, two wereindicators of nitrogen rich soils, one of nitrogen medium soils and seven ofnitrogen poor soils. Negatively correlated with PCA axis 1 were two nitrogenrich indicators, the one nitrogen medium indicator and two nitrogen poorindicators. Positively correlated with PCA axis 1 were five nitrogen poorindicators. The two nitrogen poor indicators negatively correlated with axis1 confound the negative correlation of the two nitrogen rich indicators.However, five of the seven nitrogen poor indicators had positivecorrelations with axis 1. Based on these species, PCA axis 1 conformed onlyin a general way with that of the vegetation associations, indicating that thefirst PCA axis scores increased along a species gradient of decreasingnitrogen indicator values. Similarly, trends in environmental elements asindicated by plant species were noted for PCA axis 2. Of the 25 speciescorrelated with this axis, nine were nitrogen-rich indicators and were allpositively correlated with PCA axis 2. Of the three nitrogen-mediumindicators that were correlated, one was positive and two were negative. Ofthe seven poor-nitrogen indicators, five were positively correlated and twowere negatively correlated. This trend was due to the concentration of theextreme non-forested plant communities in the bottom right quadrant(Table 4.3).Using the soil moisture indicator value, the wet to very wet indicatorswere negatively correlated with the second axis. This too was attributed tofour very poor, wet to very wet non-forested plant communities in thebottom right quadrant. Also in this quadrant were two non-forestedecosystems whose soils were classified as folisols over rock.66Table 4.3. Frequencies of nitrogen indicator species for the four major plant associations.Plant Frequency (%) of ISGsAssociation N-rich N-medium N-poorRubus 49 2 49Polystichum 24 2 74Gaultheria-Polystichum 6 94Gaultheria 1 99In general, PCA axis 1 seemed to be negatively correlated withnitrogen-rich indicators and axis 2 seemed to be positively correlated withnitrogen-rich indicators. Both axes were not related to soil moistureindicators in a strong fashion.Further to investigating the relationship between the plantassociations and the measured soil chemicals, canonical discriminantanalysis was used. Only the five major plant associations with enough plotswere used for this analysis (Gaultheria, Gaultheria Polystichum, Polystichum,Rubus and Sambucus). To reduce the number of chemical variables and toaddress the multicollinearities in the data, centred PCA of the soil chemicalcorrelation matrix was used. Nine components were found to account for94.6% of the total variance. These components were considered toadequately characterize the soil chemical domain and accordingly were usedas one set of variables in the canonical analysis. The five plant associationsin the form of a constant and four binary valued "dummy" variablescomposed the second set of variables. This gave a ratio of plots to variablesof sixty-seven to twelve which was above the minimum ratio of five to one assuggested by Tabachnick and Fidell (1989).The ecological objectives were to describe and account for thestructure of the soil chemicals, expressed as PCA scores of linearcombinations of soil chemical measures, and in doing so, investigate their67relationships to the plant associations. The results of the analysis aresummarized in Table 4.4. All four canonical axes had squared canonicalcorrelations of greater than 0.1, thus they explained greater than 10% oftheir respective proportion of the variance in the soil chemical domain.To examine the relationships among the soil chemicals, plots ofdifferent combinations of canonical axes were examined. Only the first twoshowed a reasonable distinct visual separation on a scattergram. Figure 4.5displays the samples in the space of canonical variates 1 and 2 of the soilchemical PCA scores domain with elliptical outlines, drawn to encompass80% of the plotted points in the respective group, overlain.There appeared to be no clear separation of the four vegetation units,with canonical axis 1 showing a gradation from the Gaultheria, through theTable 4.4: Correlation of the soil chemical PCA scores with the canonical variatescores showing relationships between soil chemical PCA axes with five plantassociations.n=67Canonical variateLoadingsU 1 U2 U3 U4 h2wPCA axis of chemical variablesAxis1 0.867 -0.035 0.059 0.155 0.781Axis2 -0.334 -0.254 0.627 0.142 0.589Axis3 0.194 0.410 0.541 -0.276 0.575Axis4 0.154 0.068 0.274 0.478 0.331Axis5 -0.091 0.699 -0.204 0.340 0.655Axis6 0.030 -0.385 -0.075 0.621 0.540Axis? 0.059 -0.222 0.177 -0.129 0.101Axis8 0.247 -0.215 -0.035 -0.346 0.228Axis9 -0.019 0.168 0.396 0.120 0.200Squared CanonicalCorrelation 0.533 0.264 0.146 0.101 56831—1—3—5 —4 —3 —2 —1^0^1^2^3Canonical Variate 1Figure 4.5. Canonical discriminant analysis plot of the first two canonicalvariates showing 80% elliptical outlines for plant associationsGaultheria(1), Gaultheria-Polystichum(2), Polystichum(3), and Rubus(4).Plot labels are G=Gaultheria, T=Gaultheria-Polystichum,P=Polystichum, R=Rubus and S=Sambucus.69Gaultheria-Polystichum, the Polystichum and to the Rubus plant association.The Gaultheria plant association was clearly separated from the Rubus plantassociation. Canonical axis 2 further separated the Polystichum from theRubus plant association. The two plots of the Sambucus plant associationdid not separate from the Rubus plant association; therefore, they appearedto be chemically indistinguishable.Correlations between the canonical axes and the soil chemical PCAscores (Table 4.4) indicated that there were strong positive correlationsbetween canonical axis 1 with PCA axis 1, and canonical axis 2 with PCAaxis 5. The first PCA axis was dominated by a high positive association withforest floor and mineral soil nitrogen (both total and mineralizable), andstrong negative correlations with forest floor C:N ratio and availablepotassium (Table 4.5). PCA axis 5 showed a strong positive correlation withmineral soil carbon-nitrogen ratio and moderate positive correlations withforest floor pH and available potassium (Table 4.5). Hence, the underlyingnutrient gradient from the plant associations Gaultheria, Gaultheria-Polystichum to Polystichum and Rubus appeared to be associated positivelywith nitrogen and negatively with forest floor C:N ratio and potassium. Thegradation along canonical axis 2 appeared U-shaped, with both Gaultheriaand Rubus associations correlated positively with forest floor availablepotassium, and mineral soil C:N ratio.To investigate the relationship of soil nutrients with individualspecies, canonical correlation analysis (CCA) was performed. Only theoriginal data set of 61 plots was used for this analysis because differentlaboratory methods were used for sulphur and phosphorus in the two datasets. The aim of the following application of CCA was to clarify relationshipsbetween species' abundance and associated soil chemical characteristics.70Before embarking on the CCA, it was necessary to re-express the datain a form better suited to analyses of this kind. The obvious need was toreduce the large number of species and chemical variables, and to addressthe problem of multicollinearity in the variables (Gittins 1979). To reducethe number of species from 120, only the 43 diagnostic species were used.Additionally, only the 10 forest floor chemicals were used, since during fieldsampling the roots of the vegetation were noted to occur predominantly inthe forest floor. The roots located in the top portion of the mineral soil weremainly those of the overstory tree species. Klinka et al. (1990) also foundthat the structures of the humus form and vegetation data sets werestrongly related.Centred PCA performed on the vegetation covariance matrix, andTable 4.5: Correlation of PCA scores used in canonical discriminant analysis withsoil nutrient properties.Chemicaln=67Forest FloorpHtotal carbon (%)total nitrogen (%)mineralizable N (ppm)carbon-nitrogen ratioavailable calcium (ppm)available magnesium (ppm)available potassium (ppm)Mineral Soil pHtotal carbon (%)total nitrogen (%)mineralizable nitrogen (ppm)carbon-nitrogen ratioavailable calcium (ppm)available magnesium (ppm)available potassium (ppm)Axis 1LoadingsAxis 2 Axis 3 Axis 5 Ax is6-0.209 -0.158 0.741 0.414 -0.098-0.078 -0.097 -0.635 0.037 0.4590.846 -0.109 -0.236 0.284 0.1220.675 -0.434 0.070 0.230 0.395-0.798 0.090 -0.196 -0.206 0.165-0.018 -0.445 0.609 0.120 0.349-0.014 0.278 -0.167 0.066 -0.348-0.764 0.210 0.017 0.438 -0.0160.350 -0.678 0.267 -0.151 -0.1060.454 0.756 -0.056 0.083 0.1210.731 0.532 -0.023 0.146 -0.0970.637 0.590 0.086 0.118 -0.122-0.466 0.081 -0.115 0.594 -0.030-0.073 0.556 0.575 -0.344 0.2290.069 0.791 0.296 -0.124 0.078-0.354 0.772 0.041 0.110 0.29671performed on the forest floor chemical correlation matrix was used toexpress the overall vegetational composition of the sites and the overallnutrient status of the forest floor in relation to the ten forest floor nutrientproperties. In order to maintain a minimum ratio of plots to variables of fiveto one, eight vegetation PCA components and four chemical PCAcomponents were used. The eight vegetation components were found toaccount for 80% of the total variance, and the four forest floor chemicalcomponents accounted for 87% of the variance. These components wereconsidered to adequately characterize the vegetation and forest floorchemicals and accordingly, were used as the two sets of variables in thecanonical analysis. The ecological objective was to achieve a preliminaryunderstanding of the relationship between the diagnostic species with theforest floor nutrient properties measured.Table 4.6 indicates that the variance of the vegetation PCA scoresdomain explained by the four canonical axes is 0.53, with axis 1 onlyexplaining 0.13 of the variation. Therefore, canonical axis 1 was a ratherweak component of the vegetation scores domain. Although the varianceexplained between the first two variates was 0.62 (Canonical R 2), theredundancy or raw variance of the vegetation PCA scores explained by thechemical PCA scores was only 0.08. This indicated a weak relationshipbetween the diagnostic vegetation and the forest floor nutrient properties.However, the vegetation relationships were dominated by six majordiagnostic species at the alliance level. These individual species comprisedone set of variables in a further canonical analysis. Preliminary analysisshowed total sulphur was highly correlated with total nitrogen and that totalphosphorus was unimportant in the canonical correlations. Consequently,both soil chemicals were omitted, which allowed the two data sets to be72combined for a total of 82 plots. Thus, eight forest floor chemicalscomprised the second set with the ratio of variables to plots of 5.9:1.The ecological objective of this analysis was to identify theconnections between the six diagnostic species and the eight forest floorchemicals, and, in doing so, draw attention to the difference in magnitudebetween the effects of soil chemicals on the species. Since the ecologicalinterest in this study was confined to the possible effects of the forest floorsoil chemical variables on species' abundance, the analysis is directed inTable 4.6: Canonical redundancy analysis of the species PCA scores andforest floor chemical PCA scores.Their own^ The oppositeCanonical Variables^ Canonical VariablesVariate Proportion Cumulative Canonical Proportion CumulativeProportion R2^Proportion1 0.1313 0.1313 0.6219 0.0816 0.08162 0.0913 0.2226 0.3514 0.0321 0.11373 0.1231 0.3457 0.1647 0.0203 0.13404 0.1803 0.5260 0.0929 0.0168 0.1508nature. The analysis was confined to the species and directed to looking atthe variance of the species explained by the chemicals. The results of theCCA are summarized in Table 4.7.Only the squared canonical correlation coefficients associated withthe first four canonical axes exceeded 0.10 in magnitude. It thereforeappeared that four dimensions was sufficient to fully account for therelationship between the variables and consequently, discussion will involvethese four pairs of variates only.Table 4.7: Canonical Correlation Analysis showing relationships between six diagnostic species and eight forest floorchemical properties: loadings, cross loadings and redundancy.N = 82Canonical variatesLoadingsU1 U2 U3 U4 h2,NCross loadingsV1^V2 V3 V4 h2bSpeciesBlechnum spicant -0.370 -0.013 0.339 0.244 0.311 -0.281 -0.007 0.180 0.098 0.121Polystichum munitum -0.515 0.624 0.240 0.533 0.996 -0.391 0.374 0.127 0.214 0.355Gaultheria shallon 0.971 -0.172 0.094 0.137 1.000 0.737 -0.103 0.050 0.055 0.559Thuja plicata 0.551 0.339 0.324 -0.497 0.771 0.418 0.203 0.172 -0.199 0.285Rubus spectabilis -0.359 0.103 0.810 0.068 0.800 -0.273 0.062 0.429 0.027 0.263Tiarella trifoliata -0.341 -0.369 0.507 0.299 0.599 -0.259 -0.221 0.269 0.120 0.203Variance extracted 0.487 0.111 0.121 0.117 0.837 0.281 0.040 0.034 0.019 0.374LoadingsCanonical variate V1 V2 V3 V4 hewForest floor chemical0.215 0.211 0.814 -0.189 0.789pHtotal carbon (%) 0.187 0.371 -0.389 -0.318 0.425total nitrogen (%) -0.529 0.500 -0.170 0.219 0.603mineralizable nitrogen -0.647 0.422 -0.206 -0.268 0.711carbon-nitrogen ratio 0.618 -0.143 -0.151 -0.193 0.462available calcium -0.060 0.510 0.280 -0.682 0.808available magnesium 0.462 0.235 0.293 -0.042 0.356available potassium 0.885 -0.135 -0.019 -0.118 0.816Variance extracted 0.270 0.121 0.134 0.096 0.621Squared CanonicalCorrelation 0.576 0.360 0.281 0.161Redundancy R2y i x 0.281 0.040 0.034 0.019 0.37474The four species canonical variates (U k) accounted for 84% of thevariance in the species domain Of the three, U 1 was by far the strongest,accounting for 49% of the total variance; U2, U3 and U4 accounted for some11, 12, and 12 % respectively. The loadings on the first componentindicated that all six species had moderate (>0.3) to strong (>0.7)correlations with this axis. The greatest correlation was with Gaultheriashallon, followed by moderate associations with Polystichurn munitum, Thtgaplicata, Blechnum spicant, Rubus spectabilis, and Tiarella trifoliata. G.shallon and T. plicata had a positive association with U 1 while the remainingfour had a negative correlation. The second component (U2) is characterizedby a strong positive correlation with P. munitum, and a moderate positiveand negative correlation with T. plicata and T. trifoliata, respectively. Thethird component (U3) was mainly an expression of a strong positivecorrelation with R. spectabilis and moderate positive correlations with T.plicata, T. trifoliata and B. spicant. The fourth component (U4) wascharacterized by a moderate positive correlation with P. munitum and amoderate negative correlation with T. plicata.Considering the correlations between the original forest floor chemicalvariables and the canonical variates, Vk, defined on them, most chemicalvariables contributed positively to V 1 with the exception of total nitrogen,mineralizable nitrogen and available calcium,. Available potassium had ahigh correlation with this variate (0.89), and total and mineralizablenitrogen, C:N ratio, and available magnesium had moderate correlationswith this variate. V 1 therefore, seemed to be a strong expression ofpotassium and moderate one of magnesium and C:N ratio in the positivedirection, and a moderate expression of total and mineralizable nitrogen inthe negative direction. The correlations with V2 were all moderate and75positive with total carbon, total and mineralizable nitrogen and availablecalcium. Variate V3 was predominantly a strong positive expression of pHalong with a weaker negative expression of total carbon. V4 had a moderatenegative correlation with available calcium and total carbon. V 1 extracted27% of the total variance of soil variables, V2 12%, V3 13% and V4 10%.Together the four Vk canonical variates accounted for 62% of the variance ofthe forest floor chemical domain.The intraset communalities (h2w) of the species variables for the rank4 model indicated that Uk accounted for substantial proportions of thevariances of each of the species. The equivalent forest floor chemicalcommunalities showed that Vk also accounted for sizeable proportions ofthe chemical variances. Figure 4.6 shows U 1 and V1 plotted against eachother.From the redundancies, it is clear that canonical variate V 1 accountedfor most of the explained variance of the species examined. The redundancyin the species domain generated by the canonical variate V 1 was 0.28 --thus the actual forest floor chemicals explained roughly 28 % of thevariance in the species domain. The chemical variates V2, V3, and V4explained 4%, 3% and 2% additional variance respectively. A rank 4 modelseems to be satisfactory to provide the necessary combination of fit, insight,and parsimony. The four forest floor chemical variates explained 37% of thevariance of the species domain. As pointed out by Gittins (1979), in generalwhen the ratio of predictor to criterion variables is in the order of 1:1, it isexpected that the percentage of predictable variance will fall compared tomultiple regression. Gittins demonstrated that canonical analysis was ableto efficiently recover relationships of ecological interest between two sets ofvariables if about 40% of the variance of the variable-set of interest was—2 —1 30^1^23—2First Species Canonical VariateFigure 4.6. Relationships between the six diagnostic species and the eightforest floor chemical properties using the first canonical variate pair.7677explained.The preceding canonical correlation analysis suggested which of theforest floor nutrient properties were correlated with the six species(summarized in Table 4.8). The main correlations occurred on the first pairof variates where G. shallon and T. plicata varied negatively with totalnitrogen, mineralizable nitrogen and positively with C:N ratio, potassiumand magnesium. B. spicant, P. muniturn, R. spectabilis, and T. trifoliatavaried oppositely. Supplementary relationships also existed. B. spicant, T.plicata, R. spectabilis, and T. trifoliata had a positive relationship with forestfloor pH, B. spicant R. spectabilis, and T. trifoliata had a negative correlationwith total carbon, T. plicata had a positive correlation with calcium, and T.trifoliata showed an inverse relationship with calcium.The six predominant species distinguishing the major associationsshowed a trend in correlation, using CCA, with some of the forest floorchemical properties measured. In agreement with Klinka et al. (1989)Gaultheria shallon, considered an oxylophytic species, was negativelycorrelated with forest floor total and mineralizable nitrogen. Polystichummunitum, Rubus spectabilis, and Tiarella trifoliata, considered nitrophyticspecies, were positively correlated with mineralizable nitrogen. Blechnumspicant showed a positive correlation with nitrogen; this is opposite to whatwould be expected since it is considered an oxylophytic species. However,the sites were noted to have a large variation in the amount of decayingwood which, when combined with the low number of species, could result insuch a disagreement.These six species were dominant in defining four major associationsin the diagnostic table; therefore, the trend exhibited by the soil chemicalproperties with the four plant associations reflected the relationship found78Table 4.8: Summary of positive and negative correlation of six diagnostic species with eight forestfloor chemical properties. Large font size are from the first Canonical variate pair, small fontsize are from supplementary Canonical variate pairs.Forest floorChemicalpH^totalCtotalNmineral C:NNavailCaavailMgavailSpeciesBlechnum spicant + + +Polystichum munitum + +Gaultheria shallon + + +Thuja plicata + + + + +Rubus spectabilis + + +Tiarella trifoliata + + +for individual species. There was an overlapping gradation of soil chemicalproperties from the Gaultheria , through the Gaultheria-Polystichwn,Polystichurn , to the Rubus Association. Considering that the plantassociations themselves showed the same overlapping gradation, this sametrend with the six major diagnostic species was expected. As with the totalindividual species, this gradation was correlated positively with both forestfloor and mineral soil nitrogen (both total and mineralizable), and negativelycorrelated with forest floor C:N ratio and available potassium. Thus, thegradation of the four plant associations was associated with an increase ofnitrogen availability which is correlated with an decrease in the C:N ratioand available potassium of the forest floor.When examining the relationship between the six species and theplant associations with potassium, it should be kept in mind that thenutrients are based on a concentration basis. The negative correlationbetween the nitrogen measures and potassium may simply be due to adilution effect (i.e., the increase in supply of other nutrient elements on thesites with more nitrogen is proportionately greater than the presumedincrease in supply of potassium).4.2 Soil Nutrient RegimesThe primary data set of 55 plots was used to explore the relationshipbetween the field derived SNRs (hereafter referred to as simply "SNR") andthe soil chemical measures. The most promising individual soil chemicalmeasures showing a relationship with SNRs are listed in Table 4.9. Figure4.7 shows this relationship using a series of box plots. Although there aredefinite trends, the variation within each SNR is very wide with considerableoverlap between SNRs.To further explore the relationship between the field derived SNRs andthe actual soil chemical measures, a PCA ordination was carried out onboth the forest floor and mineral soil chemical measures. The first twoaxes, accounting for 32% and 18% of the variation respectively, weregraphed and elliptical outlines were overlain (Figure 4.8). The ellipticaloutlines showed a definite gradation from the very poor SNR to the richSNR. There was only one very rich plot and it was not considered further inthe analysis. The very poor and rich SNRs showed a distinct separation butthe poor and medium overlapped substantially with each other, and withthe very poor and rich SNRs.The correlations between PCA axes 1 and 2, and the soil chemicalmeasures appear in Table 4.10. PCA axis 1 showed a highly positivecorrelation with forest floor and mineral soil total and mineralizablenitrogen, and total sulphur. Highly negative correlations were found withforest floor and mineral soil carbon:nitrogen ratios, and forest floor availablepotassium. PCA axis 2 showed moderately positive correlations with forest7980Table 4.9: Means and standard deviations of selected soil chemical properties and their relationshipto SNRs.Chemical Very Poor^Poor^MediumMean Concentration I (Standard Deviation)Rich Very RichForest FloorNumber of plots 19 11 21 9 1mineralizable N (ppm) 168 205 258 285 250(59) (74) (75) (104)C:N 48 43 39 30 31(6) (7) (10) (6)available K (ppm) 801 577 492 367 270(245) (226) (182) (128)Mineral SoilNumber of plots 13 11 21 9 1pH 3.7 4.2 4.5 4.6 4.2(0.4) (0.3) (0.3) (0.3)mineralizable N (ppm) 23 38 37 55 115(12) (29) (22) (27)C:N 44 33 29 22 25(9) (7) (6) (2)available K (ppm) 63 33 22 23 57(35) (15) (7) (9)floor available magnesium and mineral soil available calcium. Also, highlypositive correlations were found with mineral soil magnesium andpotassium, and a highly negative correlation with mineral soil pH.The very poor SNR was most associated with low forest floor andmineral soil nitrogen, low forest floor sulphur, high forest floor and mineralsoil carbon-nitrogen ratio, high forest floor available potassium, low mineralsoil pH and high mineral soil available cations. The opposite relationshipswere found with the rich SNR. The poor and medium SNRs fell between ■•■VP P^R YR11NRYRVP^1.^11^RSNRR YR1T T VP PSNR1501100so0VP P 31SNRYR60605 40ISO2010160A 100a1•e$8120VP^P^11^R^YRSNRVP^P^11^YRSNE6060a40130600400i 9001100016006 1000e"0VP^P^11^R^YRSNRFigure 4.7. Box plots of selected soil chemical properties showing theirrelationship to SNRs. The symbols are very poor (VP), poor (P),medium (M)), rich (R) and very rich (VR)—6 2^4—2^0—4 66—6Principal Component Axis 1 (32%)total N, mineralizable N, C:NFigure 4.8. Plot of the first two PCA axes of the soil chemical measures with80% elliptical outlines separating SNRs very poor (1), poor (2),medium (3), and rich (4).8283Table 4.10: Correlation values (r001 = 0.34) of actual soil chemical properties withPCA axes 1 and 2.Chemicaln=55Forest FloorpHtotal nitrogen (%)mineralizable nitrogen (ppm)C:Ntotal phosphorus (%)total sulphur (%)available calcium (ppm)available magnesium (ppm)available potassium (ppm)Mineral Soil pHtotal nitrogen (%)mineralizable nitrogen (ppm)C:Navailable phosphorus (ppm)available sulphur (ppm)available calcium (ppm)available magnesium (ppm)available potassium (ppm)Axis 1 (32%)LoadingsAxis 2 (18%)-0.257 -0.2210.871 0.0180.779 -0.106-0.855 0.2890.246 -0.3610.818 0.075-0.210 -0.2800.048 0.561-0.724 0.1100.319 -0.7570.811 0.3430.783 0.374-0.784 0.258-0.515 0.008-0.508 -0.190-0.094 0.5280.150 0.733-0.323 0.748these two extremes and did not show definite associations. However, thestrongest trend existed along the first PCA axis which was most correlatedwith the nitrogen forest floor and mineral soil measures. This trend alsoagreed with knowledge that nitrogen is considered to be the most limitingnutrient in many forest ecosystems (Jenny 1941; Viets 1965; Heilman 1979;Ballard and Carter 1986). Previous studies involving characterization ofSNRs of the BEC system also supported the use of nitrogen as a criterion fordefining SNRs (Kabzems and Klinka 1987; Courtin et al. 1989).Consequently, the six forest floor and mineral soil nitrogen measureswere used in a CDA to explore the relationship of nitrogen with SNRs. Sincethere was only one plot in the very rich SNR, this plot and the very rich SNR84were dropped from further analysis. Thus, the six nitrogen measurescomprised one set of variables, and four SNRs, in the form of a constant andthree binary valued "dummy" variables, composed the second set ofvariables. With fifty-four plots, there was a ratio of plots:variables of 5.4:1.The ecological objectives were to describe and account for the structure ofthe soil chemical measures, and in doing so, investigate their relationshipsto the field-derived SNR's.Of the three CDA axes, the first accounted for 81%, the second 11%,and the third 8% of the variation. Figure 4.9 displays the samples in thespace of canonical variates 1 and 2. The first two axes were plotted withoutlines of ellipses which encompass 80% of the plotted points of therespective groups. Although the outlines showed somewhat more distinctgroupings than with PCA, there still was substantial overlap between thepoor and medium SNRs. The correlations between CDA axes 1 and 2, andthe soil chemical measures appear in Table 4.11. CDA axis 1 showed ahighly positive correlation with forest floor mineralizable nitrogen, andmoderately positive correlations with forest floor total nitrogen, and mineralsoil total and mineralizable nitrogen. This axis was highly negativelycorrelated with forest floor and mineral soil carbon:nitrogen ratio.As a validation procedure to investigate how well the nitrogenmeasures are related to the field derived SNRs, discriminant analysis wasconducted to test the ability of the six nitrogen chemical measures todifferentiate among the four SNRs. Prior to analysis, a check of multivariatenormality showed that the distribution was not normal. To achieve amultivariate normal distribution, five of the six variables were transformedusing the common logarithm. Only the forest floor carbon:nitrogen ratiowas left in the original units. A test of equality of the dispersion matrices43210—1—2—3—485—3 —2 —1^0^1^2^3^4^5Canonical Variate 1Figure 4.9. Plot of the first 2 canonical variates, derived from the soilnitrogen measures, with 80% elliptical outlines separating SNRs verypoor (1), poor (2), medium (3), and rich (4).86Table 4.11: Canonical Discriminant Analysis showing correlations between the threeCDA axes and six soil nitrogen measures. n=54Canonical variateForest floor chemical measuresloadingsU 1 U2 U3total nitrogen 0.670 -0.028 0.923mineralizable nitrogen 0.706 -0.008 0.504C:N -0.882 -0.327 -0.127Mineral soil chemical measures0.639 -0.046 0.317total nitrogenmineralizable nitrogen 0.584 0.187 0.008C:N -0.957 0.200 0.185Squared CanonicalCorrelation 0.581 0.159 0.127resulted in a test chi-square value significant at p<0.1; therefore, the withincovariance matrices were used in the discriminant function. For thisanalysis, equal prior probabilities were used. A validation data set of 41plots was available, and was collected from similar areas. Comparison ofthe nitrogen measures between the two sets (Table 4.12) showed that thetwo sets were comparable, but did have some differences in means andranges. The validation set also lacked very poor SNRs.Discriminant analysis correctly classified 91% of the plots into theirsource groups (Table 4.13). Of the five plots that were misclassified, allwere only wrong by one class. However, this high success rate was notrepeatable; on the validation data set, the discriminant function correctlyclassified only 54% of the plots (Table 4.14). None of the poor plots and onlyhalf of the rich plots were correctly classified, although misclassification wasonly one soil nutrient class off.Discriminant analysis was again carried out on the original data setbut with prior probabilities thought to reflect the actual level based on87Table 4.12: Comparison of soil nitrogen measures between the original data set of 54 plots and avalidation data set of 41 plots. original datamean^sd rangevalidation datamean^sd range1.09 0.28 0.65-2.06 1.09 0.22 0.59-1.53230 86 79-479 303 113 89-58641 10 21-58 42 9 31-780.19 0.12 0.03-0.47 0.21 0.08 0.08-0.4237 25 3-93 39 21 15-11132 10 20-56 26 4 19-40Forest floor measuretotal N (%)mineralizable N (ppm)C:NMineral soil measuretotal N (%)mineralizable NC:NTable 4.13 Confusion matrix showing the percentage of plots identified bydiscriminant analysis in the source SNRs on the basis of forest floor andmineral soil log(total N), log(mineralizable N), forest floor C:N ratio andmineral soil log(C:N ratio). This matrix was derived from the originaldata set.SNR PercentcorrectNumber of plots identified in SNRVery Poor^Poor^Medium RichVery Poor 93 13 0 1 0Poor 90 1 9 0 0Medium 84 0 0 16 3Rich 100 0 0 0 11Total 89 14 9 17 14Table 4.14 Confusion matrix showing the percentage of plots identified usingthe discriminant function derived from the original data set on avalidation data set in the source SNRs on the basis of forest floor andmineral soil log(total N) and log(mineralizable N), forest floor C:N ratioand mineral soil log(C:N ratio).SNR^Percent^Number of plots identified in SNRcorrect^Very Poor Poor^Medium^RichPoor^0^1^0^6^0Medium^71 1 5 17 1Rich^50^0^0^5^5Total^54^2^5^28^688knowledge of the area. The priors for the SNRs were changed to .05 for verypoor, 0.35 for poor, 0.45 for medium and 0.15 for rich. This reflected theproportion of the respective SNRs actually existing in the area. The resultswere only slightly different from using equal prior probabilities of 0.25 for allfour SNRs. The function this time correctly classified 89% of the plots inthe original data set, and once again, 54% for the validation data set.The discriminant function then, captured the within site variability of theoriginal data set but failed to capture between site variability with anindependent test data set. This result may partly be attributed to themulticollinearities between the variables chosen. Multicollinearity can leadto unstable discriminant functions and is known to be a problem (Gittins1985). Also, a substantial portion of the test data set was sampled fromdifferent geographical areas than the original data set. This included thefollowing differences: (i) the original data set included non-forested, extremevery poor sites, (ii) the soil parent material of the original data set waspredominantly of volcanic origin while the test data set included volcanic,granitic and limestone parent materials 1 , and (iii) the test data set includedareas on northern Vancouver Island, which had forest floors up toapproximately one meter deep. Thus, the discriminant function derivedfrom the original data set may only have use within the limited area and/orconditions that it was derived from. Further study on a wider geographicarea is needed to assess whether a more general function can be derived, orwhether a different function is required for smaller regions.The use of multivariate analysis has its strength in pattern analysis,but the traditional univariate statistical analysis has its strength in actual1 Although, Heilman and Gass (1974) suggested that the influence of parent materials on chemicalproperties in the upper soil horizons (organic layers to B22) is small.89hypothesis testing (Gauch 1986). Therefore, if it is accepted that nitrogen isthe major nutrient representing a SNR gradient, then the next step would beto statistically test for differences. However, the laboratory chemicalanalysis methods used in this study were chosen with pattern analysis inmind. Thus for instance, the anaerobic incubation procedure formineralizable nitrogen, originally developed for "routine analysis of soils"(Waring and Bremner 1964), was used. Because laboratory estimates ofnitrogen availability are insensitive to site environmental factors that areknown to influence the nitrogen supplying power of soil, severalinvestigators have chosen in situ assays (Raisin et al. 1987; Hart andFirestone 1989; Adams et al. 1989). Aber and Melillo (1984) also suggestedthat increased accuracy may come from the use of an in situ incubationprocedure. A measure of the on-site nitrogen mineralization potential,which includes measurements of nitrate, can be viewed as reflecting anoverall decomposition rate.4.3 Relationship Between Plant Associations and SNRsBoth the vegetation classification and the site classification ofedatopes were related, since understory species were used in supplementingthe heuristic procedure to derive a SNR. However, the relationship was notexact, as shown in Figure 4.10 where several plant associations overlap withthe SNRs. In this study, sites chosen were second growth hemlock standswhich were characterized by a closed canopy resulting in depauperatevegetation. However, the use of even-aged second-growth stands ensuredthat the top height trees were not suppressed beyond breast height, whichis necessary for site index estimates. It would be expected that therelationship between field derived SNRs and plant associations would be VERA -SAMB -ACHL -RUBU -POLY -GAPO -MAHO -GAULVOVT -PINU -RHAC -CHAM -SPHA -LEDU -1[1]1--IC.J 1—1^I^I^I0^10 20 30 40^50VP^P^M^R. VRNumber of Plots — Separated by SNRsLEDU=Ledum; SPHA=Sphagnum; CHAM=Chamcaepyris;RHAC=Rhacomitrium; PINU=Pinus; VOVT=Vaccinium(ovatum)GAUL=Gaultheria; MAHO=Mahonia; GAPO=Gaultheria—Polystichum;POLY=Polystichum; RUBU=Rubus; ACHL=Achlys; SAMB=Sambucus;VERA=VertatrumVP=Very Poor; P=Poor; M=Medium; R=Rich; VR=Very RichFigure 4.10. Box plots of the number of plots showing the relationship ofplant associations to SNRs. The numbers on the x-axis represent thenumber of plots within each SNR.90stronger using ecosystems showing a better expression of vegetation.4.4 Soil Moisture RegimesThe Energy-Soil Limited (ESL) water balance model (Spittlehouse1981; Spittlehouse and Black 1981) was used to calculate the growingseason water-deficit in order to estimate actual SMRs of water deficient(moderately dry through slightly dry) and fresh sites. The definitions ofmoderately dry, slightly dry and fresh sites as defined in Minia et al. (1990)are given in Table 4.15.However, even the driest non-forested ecosystem, with only 1 cm ofsoil, showed no water deficit upon using the model. Despite the highamount of rainfall associated with this variant (Table 4.16), this result doesnot seem reasonable. The occurrence of Cladina sp. and Rhacomitriumcanescens, which are reported to occur on moisture deficient sites (Klinka etal. 1989), suggests that these extreme sites, even in this wet variant, doexperience moisture deficits. Consideration must be given to the fact thatthe coefficients used in running the model were those of Giles et al. (1985)which were derived from a different biogeoclimatic subzone (CWHxm1) forDouglas-fir using daily data. It seems probable that the model needscalibration in the CWHvm1 variant, and for western hemlock. SpittlehouseTable 4.15 Key to tentative soil moisture regimes for coastal British Columbia for moderately dry,slightly dry and fresh sites (from Klinka et al. 1989). 91Soil moisture characteristicWater deficit > 1.5 but 3.5 months or AET/PET1 <=90 but > 60%Water deficit > 0 but 5_ 1.5 months or AET/PET > 90%Utilization (and recharge) occurs (current need for water exceedssupply and soil-store water is used)Soil moisture regimemoderately dryslightly dryfreshAET - actual evapotranspiration, PET - potential evapotranspiration92Table 4.16. Thirty year precipitation normals (1951-1980) from the climate station located at Tahsis,B.C. (Anonymous 1982).April^May^June^July^August Sept^Oct^Yearly totalPrecipitation (mm) 254.7^142.1^119.6 91.0^114.5 269.9 482.4^3828.8(1981) stated that measurements at more forest sites and with differentspecies are required to test the generality of this model. Also, Wang (1991)pointed out that coefficient calibration and model validation are requiredwhen applying the model to other ecosystems with different characteristics(under different regional climate or with different tree species).Despite the need for model calibration and validation, it was stillsurprising that the model results showed no water deficit modelled on thisextreme shallow soil site. There are three possible reasons why waterdeficits may not be detected, assuming that the ESL water balance modeland the coefficients used are at least somewhat accurate. The first reason,stated by Carter and Klinka (1990), was that the monthly time-step used inthe calculations likely resulted in an underestimation of soil water-deficits.Secondly, water deficits may occur in periods of weeks as opposed tomonths. The thirty year averages suggest that there is considerableprecipitation in all months throughout the growing season. However, amodel run with daily averages may detect shorter term deficits which mayplay an important role in western hemlock ecosystems. Daily averages alsoeliminate the effect of uniform distribution of precipitation over the wholemonth. The third possibility is that years of abnormally high rainfall wereforcing the thirty year average monthly rainfall up, giving the impressionthat rainfall occurs throughout the growing season every year. To overcome93this, rather than using thirty year averages in the model, yearly valuescould be used and water deficits calculated yearly for the same thirty yearperiod. The water deficits could then be averaged, eliminating the largeinfluence of heavy rainfall years.It is recommended for further use of this model in the very wetCoastal Western Hemlock Subzone that (i) model calibration and validationbe first carried out, (ii) that daily averages separated by years be used forcalculations, and the deficit periods averaged over the thirty-year datacollection period', and (iii) for western hemlock ecosystems, an approach tomodelling the effect of the thick humus layers is needed.4.5 Relationships Between Site Index and Indirect Measures ofEcological Site QualityIn order to visually summarize the distribution of plots and therelationship between site index with plant associations, box plots of siteindex values (Tukey 1977; McGill et al. 1978; Chambers et al. 1983) areshown in Figure 4.11. Despite the wide variation, there was a trend ofincreasing site index from the Ledum to the Polystichum plant association,where it then leveled off.To mathematically relate site index to the plant associations,"dummy" variable regression was used with the resulting full equation:SI = 2.5 + 0.45(PA2) + 1.1(PA3) + 5.5(PA4) +6.4(PA5) + 8.1(PA6) +9.2(PA7) + 13.7(PA8) + 17.6(PA9) + 24.7(PA10) + 27.5(PA11) +30.7(PAl2) + 33.1(PA13) + 33.2(PA14)Adjusted R2 = 0.78^F-ratio = 22.86^SEE = 5.3 m N = 821 Thus, despite the statement in the instructions to "SWBED1: Program to calculate simple waterbalances" (programmed by D.L. Spittlehouse 1987), that daily or monthly averages may be usedit is suggested that daily averages only be used.94Dummy variables representing various vegetation associations (assn):Constant=Ledum Assn; PA2=Rhaccomitrium Assn; PA3=SphagnumAssn; PA4=Maius Assn; PA5= Vaccinium(ovatum) Assn; PA6= VeratrumAssn; PA7=Chamaecyparis Assn; PA8= Gaultheria Assn; PA9=MahoniaAssn; PA10=Gaultheria-Polystichum Assn; PA11=Achlys Assn;PAl2=Polystichum Assn; PA13=Rubus Assn; PA14=Sambucus Assn.A check of the residuals showed a lack of homogeneity of variance, mainlydue to the few plots in several vegetation units. This weakens the analysis,if confidence or prediction intervals are required. However,heteroscedasticity does not invalidate the analysis since the linearrelationship between variables is still captured (Tabachnick and Fidell1989). A probability plot of residuals indicated that the error distributionwas near normal. One outlier in the Gaultheria Association was detectedbut was difficult to explain, except perhaps by the depauperate vegetation.This plot was classified into the Gaultheria Association mainly on thepresence of 2.3 to 5.0 % presence cover of Gaultheria shallon and lack ofother diagnostic species.From the box plots, it is apparent that there may not be a significantdifference in site index among the Gaultheria, Gaultheria-Polystichum,Polystichum, Rubus, and Sambucus plant associations. The regressioncoefficients indicated only approximately a 3 meter difference between thePolystichum, Rubus and Sambucus associations.To test whether these five major plant associations, which consist of67 of the plots sampled, had significantly (p < 0.05) different mean siteindexes, a Tukey-Kramer multiple range test (Ott 1988) was carried out onthe Gaultheria, Gaultheria-Polystichum, Polystichum, and Rubus combinedwith Sambucus plant associations. The Sambucus unit and Rubus unitwere combined since the former only had two plots, had Rubus spectabilisas a major component, represents very similar site conditions to the latter,1^I^1^I^I^I^I^I^1^I^I^I^I^11 1T T I^I^I^1^I^1^I^I^I^I^I^I^I^I9550of) 40,14,o1-430OEt 20U100vt.1 vN3 v.'. 40 vt.6 vio YAK IN9TOTOvN11.??.v3v,t4Plant AssociationFigure 4.11. Boxplots of site index and plant associations Ledum (PA1),Sphagnum (PA2), Rhacomitrium (PA3), Malus (PA4), Vaccinium(ovatum)(PA5), Veratrum (PA6), Chamaecyparis (PA7), Gaultheria (PA8),Mahonia (PA9), Gaultheria-Polystichum (PA10), Achlys (PAl 1),Polystichum (PAl2), Rubus (PA13), and Sambucus (PA14).96and had a mean site index almost identical to the latter. An initial Bartletttest for homogeneity of the four group variances indicated significant (p <0.05) differences between them. However, after eliminating the outlierdiscussed previously, differences were not significant. Bartlett's test isnoted to be unduly sensitive to departures from normality in the data (Sokaland Rohlf 1981), an effect that can be caused by outliers. Based on anormal probability plot of site index, it was visually judged that theassumption of normality had been met. As Table 4.17 indicates, there wasTable 4.17. Results of a Tukey-Kramer multiple comparison for the plant associations Gaultheria(PA8), Gaultheria-Polystichum(PA10), Polystichum(PAl2), and Rubus-Sambucus(PA13+PA14)Matrix of Pairwise Comparison ProbabilitiesPA8PA10PAl2PA13+PA14PA81.0000.2200.0250.032PA101.0000.6220.465PAl21.0001.000PA13+PA141.000no significant (p < 0.05) difference between the Gaultheria and Gaultheria-Polystichum units and none between the Gaultheria-Polystichum,Polystichum, and Rubus-Sambucus units. The only significant differencewas between the Gaultheria and both the Polystichurn and Rubus-Sambucusunits.The outlier in the Gaultheria group was noted previously. However,none of the relationships changed at the p < 0.05 significance level when itwas removed, although the Bartlett's test for equal variances did becomenon significant.A similar relationship between plant associations and site index ofwestern hemlock was described by Eis (1962). Three plant communitiesreported were similar to the ones in this study. The "Salal" plantcommunity had the lowest site index value, with the "Blechnum" plantcommunity somewhat larger, and the largest was the "Polystichum" plantcommunity.To investigate how the individual species were related to site index,PCA was first performed on the data set of 120 species. Eigenvaluesindicated that 68 PCA axes contributed to the explained variance in somemanner with the 68th axis accounting for only 0.0001% of the variance.Only the first eighteen axes accounted for greater than 1% of the varianceand cumulatively explained 92 % of the variation. These axes were thenused in a backward stepwise multiple regression with the resultingequation:SI = 25.5 - 4.3(AXIS1) + 0.8(AXIS2) - 1.2(AXIS7) - 1.4(AXIS9) +1.0(AXIS13)Adjusted R2 = 0.78 F-ratio = 58.05^SEE = 5.1 m N = 82Examination of the plot of residuals and the normal probability plotindicated a close to normal distribution and relative homogeneity ofvariances. Both the adjusted R2 and SEE values were close to thoseobtained for the plant associations. This is expected when considering theclose association shown previously between the first two PCA axes and thederived plant associations.Of a total of 120 species, 62 were significantly (p<0.05) correlated and hadcorrelation coefficients greater than 0.3, with at least one of the five axes9798(Table 4.18). Axis 7 mainly represented two plots on the extreme very wetsites - one is predominantly Ledurn groenlandicum, and the otherpredominantly of Sphagnum species. Axis 13 mainly represented the formersite only. Axis 9 most strongly represents one plot in a very wet/very richsite. Axes 1 and 2 captured, only in a general way, correlations withnitrophytic and oxylophytic species. The relationships were not very clearsince higher site indexes generally occurred on sites with closed canopies,while the lower site indexes occurred on open-grown, and in extreme cases,non-forested, sites. In particular, Tsuga heterophylla presence wascorrelated negatively with the first axis, thus positively with site index. Thedecrease in site index associated with hemlock presence was simply areflection of this canopy stand attribute. Conversely, the speciesHylocomium splendens and Rhytidiadelphus loreus increased withincreasing scores, with the higher scores coming from sites with decreasedcover. This was again mainly attributable to the light conditions and notthe site itself.A multiple linear regression showed a weak relationship between siteindex and the frequency of indicator species groups, with the onlysignificant variables being MOIST3 (moderately dry to fresh), MOIST4 (freshto very moist) and NITR3 (nitrogen rich).SI = -12.1 - 545.4(MOIST2) + 33.2(MOIST3) + 37.1(MOIST4) +16.6(MOIST5) + 13.0(MOIST6) + 19.5(NITR2) + 19.2(NITR3)Adjusted R2 = 0.35^F-ratio = 7.19^SEE = 8.8 m^N = 82constant = excessively dry to very dry and N-poor; MOIST2 = verydry to moderately dry; MOIST3 = moderately dry to fresh;MOIST4 = fresh to very moist; MOIST5 = very moist to wet;MOIST6 = wet to very wet; NITR2 = N-medium; NITR3 = N-rich.99Table 4.18. Correlation values (r001 = 0.280) of species that are significantly (p> 0.01) correlated andhave r> 0.3 with PCA axes 1,2,7,9 and 13.n=82SpeciesLoadingsAxis1^Axis2 Axis7 Axis9 Axis13Gaultheria shallon 0.86Chamaecyparis nootkatensis 0.44Menziesia ferruginea 0.39Dryopteris expansa -0.31Hylocomium splendens 0.69 0.34Polystichum munitum -0.65 0.47Rubus spectabilis -0.40 0.50Kindbergia oregana -0.31 0.57Pinus contorta 0.42 -0.43 0.35Pleurozium schreberi 0.30 -0.45 0.41 0.45 0.48Tsuga heterophylla -0.64 -0.39Thuja plicata 0.57 0.38Blechnum spicant -0.40 -0.41Corpus canadensis 0.38 0.43Tsuga mertensiana 0.37 0.39Plagiothecium undulatum -0.39 -0.33Rhytidiadelphus loreus 0.72Vaccinium parvifolium 0.38Tiarella laciniata 0.37Vaccinium ovalifolium 0.33Polytrichum alpinum 0.32Rubus ursinus 0.32Plagiochila asplenioides 0.30Viola orbiculata 0.30Aulacomium palustre -0.35 0.46 0.72Empetrum nigrum -0.35 0.45 0.72Rhytidiadelphus triquetrus -0.35 0.45 0.72Vaccinium oxycoccus -0.35 0.45 0.72Ledum groenlandicum -0.35 0.45 0.72Maianthemum dilatatum 0.42 0.47Rosa nutkana 0.37 0.41Rubus parviflorus 0.37 0.56Galium triflorum 0.33 0.56Trautvettetia caroliniensis 0.32 0.65Tiarella trifoliata 0.36 0.34 -0.32Coptis asplenifolia 0.53Sphagnum girgensohnii 0.52Sphagnum squarrosum 0.51Drosera rotundifolia 0.42Fauria crista-galli 0.42Lysichitum americanum 0.42Rhododendron albiflorum 0.42Sphagnum palustre 0.42Sphagnum rubra 0.42Chimaphila umbellata -0.33Linnaea borealis -0.31Mylia taylorii 0.31 0.37100Nardia scalaris 0.31 0.37Ranunculus sp. 0.31 0.37Veratrum wide 0.31 0.37Viola sp. 0.31 0.37Gymnocarpium dryopteris 0.31 0.37Diplophyllum albicans 0.31 0.37Sambucus racemosa 0.55Mycelis muralis 0.54Pellia neesiana 0.38Dicranu sp. 0.37Bazzania denudata 0.37Rhacomitrium canescens 0.36Pinus monticola 0.32Cladina sp. 0.31Alnus rubra -0.31Because of the lack of understory plants combined with thecorrelation of site index and canopy closure, the regression of site indexwith individual species, expressed through PCA scores or combined intoindicator species groups, seems to be of limited value in this case. Secondgrowth western hemlock stands do not have enough of an expression ofunderstory species for use with individual species. However, the collectiveexpression through plant associations seems to work to a certain extent.A total of 102 plots was used in the investigation of the relationshipbetween site index and SMRs and SNRs. A visual summary of therelationship of site index and field estimates of SMRs and SNRs is given inthe 3 dimensional plot with orthogonal projections onto a SMR facet andSNR facet in Figure 4.12. Site index increased as the SNRs proceeded fromvery poor to rich, but showed a wide spread with changing SMRs. The lowsite indices on the SMR facet are due to the very poor SNR which extendsacross the SMR gradient. However, rich/dry and rich/wet sites do notsupport stands of western hemlock, thus maintaining an increase of siteindex on the SNR facet.5040302010101Figure 4.12. A 3 dimension plot of site indices on SNRs and SMRs withtheir orthogonal projections onto a SMR and SNR facet. SNR axisrepresents very poor (1), poor (2), medium (3), rich (4) and very rich(5). SMR axis represents moderately dry (1), slightly dry (2), fresh (3),moist (4), very moist (5), wet (6), and very wet (7).102Mathematically in a dummy variable regression, this relationship wasexpressed as:SI = 5.7 + 4.2(MD) + 8.9(SD) + 14.3(F) + 16.2(MST) + 15.7(VM) + 4.9(W) -2.6(VP) + 7.2(P) + 11.7(MED) + 13.2(R)Adjusted R2 = 0.81^F-ratio = 43.7^SEE = 4.5 m^n = 102Dummy variables representing combinations of SMR and simultaneously SNR:Constant=very wet or very rich; MD=moderately dry; SD= slightly dry; F=fresh;MST=moist; VM=very moist; W=wet; VP=very poor; P=poor; Med=medium;R=rich.For the soil moisture gradient, the coefficients indicated that siteindex increased from moderately dry to slightly dry, leveled off at fresh andmoist, then decreased slightly at very moist, and rapidly at wet and very wetsites. For the soil nutrient gradient, site index increased from very poor,and poor to medium where it leveled off between medium and rich anddecreased at very rich SNRs.The similar coefficients indicated that there may not be a significantdifference in site index between fresh and moist SMRs and between poor,medium, and rich SNRs. Within the range of fresh to moist, poor to richsites, there were 75 sample plots. Between these six units, Bartlett's test forhomogeneity of group variances indicated that the variances could besignificantly different (p=0.06). An analysis of variance (ANOVA) indicatedthat at least one of the means was significantly different (p<0.05). A Tukey-Kramer HSD test between the edaphic units within this range indicated thatthere was a significant difference (p<0.05) between the site indexes ofmoist/rich sites and either fresh/poor or moist/poor sites (Table 4.19), andno significant differences among the others.A listing of mean site indexes on the edatopic grid is given in Figure103Table 4.19: Tukey-Kramer multiple comparison test matrix of pairwise comparison probabilities forcombinations of SMR and SNR. Testing for differences in site index.SMR/SNR^F/P^M/P^F/M F/R^M/M^MRFresh/Poor^1.00Moist/Poor^1.00^1.00Fresh/Medium^0.62^0.39^1.00Fresh/Rich^0.73^0.61^1.00^1.00Moist/Medium^0.37^0.21^0.94^0.99^1.00Moist/Rich^0.04^0.01^0.13^0.48^0.87^1.004.13. Although site index increased until a maximum at Moist/Rich sites,the wide variation in site indexes for the site units is apparent. Althoughnot enough samples were taken over the entire range, the dummy variableregression and the multiple comparison test indicated that the isolines ofsite index for a generalized edatopic grid of the submontane Very WetMaritime Coastal Western Hemlock variant is as in Figure 4.14. This isolinegraph was produced using a distance weighted least squares procedure witha tension of 0.2 (Wilkinson 1990).Of interest also is how well Just SMR or SNR alone would be inpredicting site index. Dummy variable regression was used with thefollowing results for SMRs:SI = 5.1 + 5.2(MD) + 14.6(SD) + 27.7(F) + 29.0(MST) + 30.0(VM) + 5.8(W)Adjusted R2 = 0.69^F-ratio = 38.0^SEE = 5.8 m^n = 102Dummy variables representing combinations of SMR:Constant=very wet or very rich; MD=moderately dry; SD= slightly dry; F=fresh;MST=moist; VM=very moist; W=wet.MDSDbea)1-4t-44)OEOniVM104Submontane Very Wet MaritimeCoastal Western Hemlock variant (CWHvm1)Soil nutrient regimeVP^P^M^R^VRSI=7.2sd=2.8n=9SI=17.7n=1.SI=13.6sd=2.1n=3SI=22.3sd=4.9n=2S1=20.3n=1SI=16.3sd=3.1n=2SI=28.2sd=6.5n=5SI=31.8sd=4.3n=31SI=32.0sd=3.1n=8SI=18.1sd=6.4n=2SI=28.3sd=6.7n=9SI=33.5sd=2.6n=9SI=35.8sd=4.1n=13SI=32.6sd=3.2n=2SI=8.0sd=0.3n=2SI=10.6n=1Figure 4.13. Edatopic grid showing site index (m © 50 years), standarddeviations (m) and the number of plots for site units sampled.105105 --^-----....-^- - - -_1520%:-.'--------35 l\ \ .-- ,— —MDWVP^P^M^R^VRSoil Nutrient RegimeFigure 4.14. Edatopic grid with a site index isoline superimposed. This wascalculated from actual mean site index values and extrapolatedsubjectively to areas lacking data (dashed lines).The following were the results for SNRs:SI = 10.6 - 0.8(VP) + 16.3(P) + 21.3(MED) + 23.7(R)Adjusted R2 = 0.75^F-ratio = 77.3^SEE = 5.2 m^n = 102Dummy variables representing combinations of SMR and simultaneously SNR:Constant=very wet; VP=very poor; P=poor; Med=medium; R=rich.Both equations had lower adjusted R2 values and higher SEE valuesthan the regression using both SNR and SMR. However, the equation usingSNR alone had values that were close. There was a lack of slightly andmoderately dry, and wet sites, with the majority of sites on fresh andmoist moisture regimes, and this had the effect of holding SMR constantwith a variable SNR gradient.The relationship between site index and certain CWHvm1 site series isgiven in Figure 4.15 with accompanying standard deviations and samplesizes. Dummy variable regression gave the following result:SI = 3.1 + 5.2(SS1) + 14.6(SS2) + 27.5(SS4) + 28.9(SS5) + 26.9(SS6) + 32.7(SS7) +4.9(SS8) + 7.5(SS9)Adjusted R2 = 0.71^F-ratio = 31.92^SEE = 5.6 m n = 102Dummy variables representing Site Series in the CWHvm variant:Constant= very wet very poor site series; SS1 = HwPI-Cladina; SS2 = HwCw-Salal;SS4 = HwBa-Blueberry; SS5 = BaCw-Foamflower; SS6 = HwBa-Deer Fern; SS7 =BaCw-Salmonberry; SS8=PI-Spagnum; SS9 = CwSs-Skunk cabbage.Five plots had large leverage; these were plots with the lowest siteindex and found on wet and very wet sites. In general, the plots had beencategorized into units similar to the site units; however, the units based onsite series combined poor and very poor SNRs. These two SNRs appeared tohave large differences in site index for western hemlock. Also, sampling was106VMwMDVery Wet Maritime submontane variantCoastal Western Hemlock (CWHvm1)Soil nutrient regimeVP^P^M^R^VR0 HwP1-CladinaSI=8.3ad=4.1^n=100 HwCw-SalalSI=17.1sc1=5.5^n=6CwHw-Sword Fern0HwBa-BlueberrySI=30.5sd=5.8^n=380 BaCw-FoamflowerSI=32.0sd=3.1^n=80HwBa-Deer FernSI=29.9sd=6.8^n=221BaCw-SalmonberrySI=35.8sd=4.1^n=13® Pl-SphagnumSI=8.0sd=0.3^n=20CwSs-Skunk CabbageSI=10.6n=1107Figure 4.15. Site Series grid with site index values, standard deviations andnumber of plots indicated.108based on site units, thus there is over representation on the medium/freshsite units, which raises the mean site index for the HwBa-Blueberry siteseries relative to the others.Krajina (1969) proposed that the most productive site for westernhemlock would be a site index of 35 (meters @ 50 years) on a moist/poorsite unit. This is different than what was found in this study where meansite index was at a maximum on moist/rich site units. However, there wereno significant (p<0.05) differences between poor-rich/fresh and poor-medium/moist site units, and between medium-rich/fresh and medium-rich/moist site units (Table 4.19). The only significant differences werebetween the moist/rich and fresh-moist/poor site units. Nevertheless, it isclear at least, that maximum site index does not occur on moist/poor sites.Krajina based his proposal on the fact that western hemlock takes upnitrogen in the form of ammonium, and that nursery experiments indicatedgrowth actually decreased in the presence of nitrate (Krajina et a/. 1973). Inthis study, site index, in fact, followed an ammonium gradient; therefore,the site index relationship is not in conflict with Krajina (1969) based onammonium. Maximum site index was found to be in the rich SNR, and theSNR gradient followed, among other nitrogen measures, ammonium, asmeasured by the anaerobic incubation procedure. Research is still neededto determine the nitrification potential across the SNRs, to investigatedifferences in western hemlock productivity with different levels of nitrate.Confounding this relationship was the association of western hemlock rootswith decaying wood. The roots were noted, even on the rich sites, to begrowing in rotten logs or old root channels. The complexity of therelationship between western hemlock and decaying wood, stated in theliterature review, makes it difficult to assess western hemlock productivityrelationships with SNRs.4.6 Relationship Between Site Index and Direct Measures of SoilNutrient RegimesTo investigate the relationship of site index with the soil chemicalproperties, the primary data set of 55 plots was used. The relationships ofmineral soil total carbon, pyrophosphate extractable iron and aluminiumwith site index was analyzed separately.Strong pairwise correlations were identified between several of thevariables (Table 4.20). As suggested by Neter et al. (1990), two moreindicators of predictor variable multicollinearities were calculated. Varianceinflation factors were transformed into tolerance indices (Table 4.21). Allvariables exhibit low tolerance values with forest floor total nitrogen, totalsulphur, and mineral soil total nitrogen and potassium having values closeto 0, indicating the presence of multicollinearity. As a third indicator, PCAwas used and the eigenvalues examined. From table 4.22, it can be seenthat the eigenvalues are far from equal and exhibit a strong trend of the firstaxes accounting for much of the variance in the data. These results supportthe fact that multicollinearity does exist. From knowledge of nutrientchemical behaviour, it is suggested that at least some of the collinearitiesare population-inherent. For example, there is an indication that nitrogenand sulphur in soil organic matter are mineralized at proportional rates(Bardsley and Lancaster 1960), and pH is known to effect several soilchemicals (Binkley and Richter 1987; Runge and Rode 1991).Because of multicollinearity, independent variables alone in aregression equation may not improve the fit significantly, but they maywhen in combination with one or more additional variables. Since forward109Table 4.20. Pearson Correlation matrix of soil chemical properties showing the presence of several large bivariate correlations.n=55SIForest FloorPH^LOGTN LOGMN CN TP TS CA MG KSI 1.000FFPH -0.224 1.000LOGFFTN 0.445 -0.270 1.000LOGFFMN 0.457 -0.056 0.735 1.000FFCN -0.509 0.042 -0.791 -0.651 1.000FFTP 0.058 0.370 0.310 0.262 -0.426 1.000FFTS 0.403 -0.416 0.937 0.685 -0.649 0.161 1.000FFCA 0.008 0.552 -0.275 0.033 0.006 0.127 -0.346 1.000FFMG -0.047 -0.081 0.104 0.020 0.141 -0.298 0.138 0.020 1.000FFK -0.627 0.379 -0.481 -0.381 0.592 0.146 -0.498 0.090 0.058 1.000MSPH 0.561 0.111 0.190 0.266 -0.423 0.281 0.152 0.240 -0.306 -0.432LOGMSTC 0.231 -0.182 0.507 0.368 -0.274 0.096 0.493 -0.159 0.368 -0.318LOGMSTN 0.458 -0.163 0.665 0.555 -0.542 0.155 0.600 -0.134 0.277 -0.524LOGMSMN 0.426 -0.181 0.599 0.556 -0.451 0.006 0.552 -0.131 0.283 -0.541MSCN -0.648 0.037 -0.595 -0.614 0.736 -0.166 -0.469 0.019 0.066 0.619MSP -0.263 0.136 -0.415 -0.416 0.390 -0.014 -0.353 0.049 -0.246 0.332MSS -0.288 0.308 -0.285 -0.218 0.339 0.128 -0.236 0.275 -0.129 0.473MSCA -0.044 0.128 -0.180 -0.123 0.102 -0.332 -0.227 0.186 0.109 -0.079MSMG -0.046 0.033 0.093 0.075 -0.013 -0.170 0.039 -0.053 0.422 -0.054LOGMSK -0.511 0.086 -0.141 -0.203 0.298 0.029 -0.167 -0.072 0.313 0.434Mineral SoilMSPH LOGTN LOGMN MSCN MSP MSS MSCA MSMG LOGKMSPH 1.000LOGMSTN 0.014 1.000LOGMSMN -0.102 0.878 1.000MSCN -0.475 -0.592 -0.534 1.000MSP -0.078 -0.359 -0.472 0.407 1.000MSS -0.001 -0.395 -0.417 0.378 0.494 1.000MSCA -0.308 0.122 0.226 0.059 0.177 -0.117 1.000MSMG -0.423 0.390 0.430 -0.067 -0.081 -0.291 0.685 1.000LOGMSK -0.764 0.246 0.220 0.458 0.061 0.086 0.294 0.464 1.000Table 4.21. Tolerance values for soil chemical properties, the low values indicatingthe presence of multicollinearity.Forest Floor Mineral SoilpH 0.353 pH 0.183LOG(Total N) 0.040 LOG(Total N) 0.064LOG(Min N) 0.293 LOG(Min N) 0.132C:N 0.108 C:N 0.111Total P 0.299 Available P 0.411Total S 0.055 Available S^0.415Available Ca 0.414 Available Ca 0.253Available Mg 0.432 Available Mg 0.263Available K 0.154 LOG(Available K) 0.074Table 4.22. Eigenvalues and cumulative variation explained by soil chemical PCAaxesPCA Axis Eigenvalue Proportion CumulativeAxis1 6.11802 0.339890 0.33989Axis2 3.34015 0.185564 0.52545Axis3 1.98247 0.110137 0.62658Axis4 1.67675 0.093153 0.72874Axis5 1.09820 0.061011 0.78976Axis6 0.87501 0.048612 0.83837Axis? 0.54991 0.030551 0.86892Axis8 0.53800 0.029889 0.89881Axis9 0.43665 0.024259 0.92306Axis10 0.32939 0.018300 0.94136Axis11 0.27665 0.015369 0.95673Axis12 0.23536 0.013075 0.96981Axis13 0.15669 0.008705 0.97851Axis14 0.14242 0.007912 0.98643Axis15 0.12653 0.007029 0.99346Axis16 0.06799 0.003777 0.99723Axis17 0.02547 0.001415 0.99865Axis18 0.02434 0.001352 1.00000111112and backward stepwise procedures only consider the addition or deletion ofvariables by their individual effect, all combinations or setwise regressionwas used as suggested by Tabachnick and Fidell (1987) for multicollineardata. The top ten equations were inspected, but residual analysis indicatedsome nonlinearity and an outlier having large leverage. For the formerproblem, forest floor and mineral soil total nitrogen and mineralizablenitrogen, and mineral soil potassium were transformed into commonlogarithms. For the latter problem, the outlier was noted as being the oneplot from a very rich and wet site. Thus, the nutrient levels were very high,but the site index relatively low. If there had been many plots in the wetSMR, dummy variables might have been used to separate the effects of awet site. However, since this was the only plot in this SNR and SMR, it wasdeleted from further analysis. All combinations multiple regression wasrepeated after the transformations and deletion of the outlier, on a total of54 plots.The ten equations having the highest adjusted R2 are listed in Table4.23. Adjusted R2 ranged from 0.56 to 0.69 and the SEE ranged from 5.4mto 6.5m. Although the maximum number of independent variables wasthree, there were eight different variables involved in different combinationsforming the ten possible relationships. The effects of multicollinearity wereevident, but generally pH, nitrogen and potassium reoccurred consistentlyin the equations. Site index leveled off with increasing nitrogen anddecreasing potassium, as indicated by the log transformations.The "best" fit equation, based on adjusted R 2 , and SEE was equation1. However, this equation was rejected as exhibiting the most "ecologicalsense" since the effects of multicollinearity appeared to dominate thesulphur variable. The tolerance value of the sulphur variable was 0.52,Table 4.23: Results of all combinations multiple regression on the primary data set.Regressionequation R2Ad'R4 SEE(m)Regression Coefficients of:Mineral soil chemical measureslog^logpH^N minN^C:NlogKForest floor chemical measurespH^K^S(1) .70 .69 5.4 26.87 -26.22 -63.59(2) .68 .67 5.6 22.03 -23.67(3) .66 .64 5.8 7.82 16.03 -11.25(4) .65 .63 5.9 8.58 12.38 -0.01(5) .63 .61 6.0 12.41 15.09(6) .62 .60 6.2 4.90 -0.34 -0.01(7) .60 .59 6.2 16.54 -22.63(8) .60 .59 6.3 11.10 17.19(9) .59 .56 6.5 7.0 -0.49 -7.51(10) .57 .56 6.5 -0.42 -0.02114indicating collinearity with the nitrogen variable which had a similartolerance value. The standard error associated with the sulphur regressioncoefficient was 31.3, almost 50% of the value of its coefficient. The negativecoefficient was opposite to that of simple correlation with site index,indicating that it is dependent on the other variables included or excludedfrom the equation (Neter et a/. 1990). Thus, the "best" fit equation, based onadjusted R2, SEE, and independent variables that made ecological sense,was:SI = 76.7 + 22.0 [Log(MSTN)] - 23.7 [Log(MSK)]Adjusted R2 = 0.67^F-ratio = 54.21^SEE = 5.6 m N = 54Variables representing:Log(MSTN) = common logarithm of mineral soil total nitrogen; Log(MSK) = commonlogarithm of mineral soil available potassium.The tolerance value of the two variables was 0.95, indicating that the twovariables were nearly completely independent of each other. A threedimensional plot and contour plot of the logarithmic relations are given inFigure 4.16.To test the portability of this equation, it was applied to the validationdata set of 41 plots, which were sampled from the same biogeoclimaticvariant. In both sets the same laboratory was used for soil chemicalanalyses. Comparison of the original and test data sets is given in Table4.24. The site index range of the test data set indicated that this set onlytested the upper portion of the equation. The results of the model test aresummarized in Table 4.25.The performance on the original data was not very good, with only46% of the cases classified within 3 meters of the measured site index.3020io115-160^-134^-1.08^-0.82^-0.56^-030LOGMSTN1.81.71.6131.4131.21.11.0Figure 4.16. Three dimensional plot and contour plot of the logarithmicrelations of mineral soil total nitrogen and available potassium withsite index.116Table 4.24. Comparison of the mineral soil chemical measures used in the regression equationbetween the original data and test data sets.original data set (54 plots)^test data set (41 plots)mean SD range mean SD rangetotal nitrogen(%)^0.19 0.12 0.03-0.47 0.21 0.08 0.08-0.42available potassium (ppm) 34 26 6-145 18 5 10-28Sl(ht 0 50 yrs bh age)^24.6 9.6 1.7-41.5 33.4 3.4 24.7-39.8Table 4.25. Classification test of site index on the original and test data based on the soil chemicalmeasures logarithm of mineral soil total nitrogen and available potassium.Number of cases classified (cumulative proportion in parentheses).original data test dataCorrect .' 25 (46) 27 (66)1 class off 15 (74) 9 (88)2 classes off 11 (94) 4 (98)3 classes off 2 (98) 1 (100)4 classes off 1 (100)TOTAL 54 41% above correct 15 (28) 4 (10)below correct 14 (26) 10 (24)1 Correct within 3 m of measured site index; 1 class off within 3-6 m; 2 classes off within 6-9 m; 3classes off within 9-12 m; 4 classes off within 12-15 m.Twenty-six per cent of the cases were greater than 6 meters off. Theperformance of the equation on the test data, based on correctclassification, was better, with 66% of the cases classified within 3 meters ofthe measured site index and 12% greater than 6 meters off. However, theequation underestimated site index on the test data by more than doublethe amount it did with the original data. The equation exhibits only a fairperformance in predicting site index on the data from which it was derived.It appears to have some portability within the upper ranges of site index,but tends to underestimate site index.The negative relationship of site index with mineral soil available117potassium was somewhat surprising since the parent material of theoriginal plots was dominated by rocks of volcanic origin. Basalts generallycontain plagioclase feldspars, whose chemical composition includes mainlysodium and calcium, as opposed to potassium, which is associated with thealkali feldspars of granite (Dietrich and Skinner 1979; Mengel 1985).Additionally, organic matter appears to exhibit a preference for divalentcations, such as Ca2+ and Mg2+, over K+ (Naylor and Overstreet 1969;Jardine and Sparks 1984).There are two possible reasons for this relationship. Mycorrhizaeappear to enhance K+ uptake and storage in roots by increasing thevacuolar pool sizes, with this effect particularly evident for western hemlockseedlings (Rygiewicz and Bledscoe 1984). Potassium in the root vacuolesare not subject to leaching. When sieving forest floor samples, fine rootsinevitably form part of the sample, and this may increase the concentrationof K+ in the sample being analyzed. Additionally, the inverse correlation ofpotassium concentration with site index indicates that this element is notlimiting tree growth. This inverse relationship may be explained by adilution effect; the increase in supply of other nutrient elements on thebetter site (higher site index) is proportionately greater than the presumedincrease in supply of potassium. The absolute quantities of potassiumsupplied annually on the sites with higher site index may well be higherthan on the sites having lower site index, but these values can bedetermined only by the measurement on a mass per area basis.In order to supplement the interpretations, PCA regression was usedto overcome the problem of multicollinearity in the regression analysis. ThePCA was carried out on the variables of the original data set used in theprevious regression analysis. Transformed variables were maintained and118the outlier noted eliminated. Every axis that accounted for at least 1% ofthe variation was used in a backward stepwise regression procedure. Thus,the first 13 axes accounting for 96.3% of the variation were used with thefollowing result:SI = 24.5 + 2.65(AXIS1) - 1.31(AXIS2) - 2.33(AXIS4) + 3.07(AXIS8)Adjusted R2 = 0.63^F-ratio = 23.92^SEE = 5.89^n = 54Examination of the residuals indicated that there was relativelyhomogeneous variance, and that a linear model was appropriate. A normalprobability plot indicated that the residual distribution was close to normal.Axis 1 was highly positively correlated (r>0.7) with forest floor totalnitrogen, mineralizable nitrogen, total sulphur and mineral soil totalnitrogen, mineralizable nitrogen, and highly negatively correlated with forestfloor C:N ratio and potassium (Table 4.26). Axis 1 was also moderatelynegatively correlated (r>0.5,r<0.7) with mineral soil available sulphur,available phosphorus and carbon to nitrogen ratio. Axis 2 was highlypositively correlated with mineral soil magnesium and potassium, andhighly negatively correlated with mineral soil pH. It was also moderatelypositively correlated with forest floor magnesium, and mineral soil availablecalcium. Axis 4 was moderately positively correlated with forest floor totalphosphorus, and moderately negatively correlated with mineral soilavailable calcium. Finally, axis 8 showed a weak (<0.5) negative correlationwith the logarithm of forest floor mineralizable nitrogen and a weak positivecorrelation with the logarithm of forest floor total nitrogen. Axis 1accounted for the majority of the variance (47%), with axes 2, 4 and 8accounting for considerably less (9%, 4%, and 4% respectively). Thisindicates that the variables most associated with axis 1 have the strongest119Table 4.26. Correlation of PCA Scores with actual chemical values for the originaldata set.LoadingsAxis 1 Axis 2 Axis 4 Axis 8-0.338 -0.190 -0.047 -0.1050.889 -0.030 0.334 -0.1490.774 -0.115 0.153 -0.330-0.823 0.276 -0.022 0.0720.174 -0.415 0.518 0.2130.831 -0.008 0.346 -0.165-0.204 -0.210 -0.336 0.0390.126 0.602 0.057 0.076-0.727 0.093 0.484 -0.1960.317 -0.773 -0.317 0.1770.792 0.382 0.084 0.3390.758 0.471 -0.021 0.183-0.795 0.228 0.266 -0.033-0.569 -0.100 0.023 0.126-0.514 -0.266 0.333 -0.011-0.084 0.588 -0.531 -0.1630.166 0.783 -0.202 -0.140-0.267 0.753 0.433 0.170ChemicalForest FloorpHLog (total nitrogen)Log (mineralizable nitrogen)C:Ntotal phosphorus (%)total sulphur (%)available calcium (ppm)available magnesium (ppm)available potassium (ppm)Mineral SoilpHLog (total nitrogen)Log (mineralizable nitrogen)C:Navailable phosphorus (ppm)available sulphur (ppm)available calcium (ppm)available magnesium (ppm)Log (Avail potassium)relationship with site index, while variables correlated to the remainingthree axes have supplementary relationships. Thus, nitrogen, forest floortotal sulphur and potassium were most strongly correlated with site index.The PCA regression showed the multicollinear nature of this data set,with several soil chemical measures correlated with the axes in the resultingregression equation. This result provides evidence that the productivity ofwestern hemlock is associated with a well balanced supply of nutrients.Further to the idea of a well balanced supply of nutrients, inspection of theforest floor carbon-nitrogen-sulphur-phosphorus ratios for different siteindex classes (Table 4.27) showed that nitrogen, sulphur and phosphorusare generally of the same ratio across site index classes. This ratio is also120Table 4.27. The total carbon-total nitrogen-total sulphur-totalphosphorus ratio of the forest floor for different siteindex classes.Site index (m @50 yrs) classC:N:S:P0-10 476 : 10 : 1.2 : 1.610-20 452 : 10 : 1.1^: 2.020-25 422 : 10 : 1.2 : 1.625-30 327 : 10 : 1.2 : 1.330-35 394: 10 : 1.2 : 1.735-41 318 : 10 : 1.1^: 1.5International average 140: 10 : 1.3 : 1.3similar to that reported for soil organic matter for an international average(Stevenson 1986). However, it should be emphasized that all direct nutrientrelationships developed in this study are correlations, and may either be acause or an effect, or even both.Since the roots of western hemlock were noted to occur mainly in theforest floor, setwise regression was performed on the forest floor chemicalmeasures only. The best fit regression was:SI = 42.8 + 3.4 [Log(FFMN)] - 0.21(FFK)Adjusted R2 = 0.50^F-ratio = 27.14^SEE = 6.9 m^n = 54Variables representing:Log(FFMN) = common logarithm of forest floor min-N; FFK = forest floor availablepotassium.Once again nitrogen and potassium were involved in the finalregression equation. In terms of adjusted R 2 and SEE, this regression wasnot as good as the best fit regression derived from both the mineral soil andforest floor chemical properties. However, in terms of knowledge of westernhemlock rooting patterns, this equation seemed to have met biologicalexpectation.121Pyrophosphate extractable iron and aluminium and total carbon ofthe mineral soil were also used in a backward stepwise regression with thefollowing result:SI = 28.4 + 18.6 [Log(MSAI))Adjusted R2 = 0.45^F-ratio = 44.6^SEE = 7.2 m^n = 54Variables representing:Log(MSAI) = common logarithm of pyrophosphate extractable aluminium.Examination of the residuals indicated that there was relativelyhomogeneous variance, and that a linear model was appropriate. A normalprobability plot indicated that the residual distribution was close to normal.The correlation of site index with mineral soil pyrophosphateextractable aluminum is opposite in sign to that of Lowe and Minim (1981).However, their correlation was with the soil chemical measure in the Bfhorizon only. In this study, the sampling of the leached soil layer in theextreme organic matter over rock sites makes a direct comparison invalid.The sites sampled in this study largely represented western hemlockecosystems without a water deficit or water excess. Thus, the conceptualmodel of Kimmins et al. (1990), which is a presentation of the variation inthe importance of moisture, nutrients, light, and soil aeration indetermining net primary production under various combinations of sitenutrient and moisture status, suggests that nutrients are the most limitingfactor within most of the range sampled. This range effectively eliminateslow site indexes that may occur in dry or wet nutrient rich sites. Dry andwet sites occurred mainly in the very poor and poor SNRs where the siteindex was low already. The one wet, very rich site was eliminated as an122outlier but this condition should be kept in mind -- i.e. that as the sitesbecome wet, site index falls regardless of an increase in SNR, due to pooraeration.The lack of sites with water deficits is due to the fact that westernhemlock does not exhibit two mechanisms that contribute to droughttolerance -- stress avoidance mechanisms and stress tolerance mechanisms.Ballard and Dosskey (1984) found that water uptake by western hemlockfrom moderately dry soils is limited by higher needle water potential andhigher uptake resistance compared to Douglas-fir. Western hemlockseedlings planted on a south-facing, high elevation clear-cut exhibited afailure to cope with drought because of lack of stress tolerance throughosmotic adjustment to enable the seedlings to maintain turgor during dryperiods (Livingston and Black 1987a,b).Similarly, western hemlock is more adaptable to wet sites where watertables are more than fifteen centimeters below the surface (Minore andSmith 1971). Flooding during the growing season has a marked negativeeffect on survival and growth (Brink 1954; Minore 1968)In the literature, site index regression relationships developed forwestern hemlock showed several possible soil chemical properties beingcorrelated with site index (Table 4.28). The multicollinear nature of the soilchemical measures creating unstable regression equations is one possibilityfor explaining the different results. However, it is probable that over thelarge area in which western hemlock is found, which includes Oregon,Washington, Alaska and British Columbia, there are differences in growthlimiting factors for western hemlock.123Table 4.28: Summary of site index regression relationships for western hemlockderived from the literature.[1] SI = f (- exchangeable K)R2 = 0.58^N = 82Wooldridge (1961)[2] SI = f [ +P (kg/ha), - total N (kg/ha), +sum of bases (kg/ha)]R2 = 0.78^N = 14^SEE = 15.6 ftMeurisse (1972, 1976).[3] SI = f (total-N in the organic matter)R2 = 0.69^N = 25Stephens et a/ (1969) - (cited in Heilman 1976)[4] Growth Class = f (-pyrophosphate extractable Fe + Al,)R2 = 0.58^N = 26Lowe and Klinka (1981)4.7 Site Index Relationships DiscussionAll regressions reported in this study were significant (p<0.05) and, inmost cases, had acceptable proportions of the variation in site indexexplained by the respective independent variables. The amount of varianceexplained was generally within the average range (65-85%) explained insuccessful soil-site models (Carmean 1975). For western hemlockspecifically, Radwan and DeBell (1980a) felt that foliar chemical variableswith correlation coefficients in the order of 0.85 (r 2 of 0.72) may be useful as124indicators to assess site index. Therefore, all regression models seemed tohave successfully captured the pattern of mean site index response over thevariables used.However, the relatively large standard error of estimates (SEE)indicated that there was still high variation in western hemlock site indexfor all regression models developed. Plots of the actual site index versus theestimated site index for some of the regression models (Figure 4.17) showedthis variation. The effect of the large SEE was demonstrated on theequation that was tested against an independent data set. The equationfailed to successfully predict the site index within three meters on 54% ofthe plots from the data set it was derived from.This relatively large variation within similar sites (see Figures 4.12and 4.13) and within similar nutrient levels (see Figure 4.16) is consistentwith what is known about the nutrition of western hemlock (covered in theliterature review). Soil-site studies involving western hemlock, reported siteindex to be correlated to several different soil physical and/or chemicalmeasures. Moreover, fertilization studies, the most common criteria fordefining nutritional status (Binkley 1986), are notorious for theinconsistency of the results, responses being positive, zero and negative.The causes for this variation about the regression model surfaces canbe divided into two general categories -- (1) the nature of the data; and (2)the nature of western hemlock ecosystems. For the first category, it wasnoted during field collection that there was high variation in microsites,caused mainly by decaying wood. The presence of large amounts ofdecaying wood, even on sites designated as rich, is a confounding factor.For example, on a rich site with decaying wood present, the question ariseswhether the tree is growing on the rich site, or simply on rotten wood5040V 3°161 20100 0^10^29^30^40EstimateIll Plant Associations5040U 3°7153 201000 10 20 30 40Brtimate121 SMRs/SNRs125131 log(total N) + log(available K)^141 Soil Chemical PCA ScoresFigure 4.17. Actual site indexes plotted against estimated site indexes forchosen regression models developed for western hemlock.126microsites?The nature of western hemlock ecosystems was noted in the literaturereview, as being a complicated system of feedback, symbiosis andspecialization. The rooting habit of western hemlock roots was reported tobe primarily associated with decaying wood and mor humus forms (Figure4.18). Even within the mineral soil profile, it was noticed that there was aproliferation of roots along root channels and buried wood. If roots areexploiting the area of maximum nutritional return, then mor humus formsand rotten wood seem to be most important for the nutrition of westernhemlock. Associated with this organic matter are strong mycorrhizalrelationships which, among other benefits, may enhance uptake of bothnitrogen, especially in the form of ammonium, and phosphorus. There alsoexists evidence that mycorrhizae of some oxylophytic species have theability to take up amino acids directly. Western hemlock is able towithstand relatively acid conditions which seems to be a condition favouringits mycorrhizal associates. Then, there is the suggestion that there areasymbiotic nitrogen fixers associated with decaying wood and fungi.Finally, Major (1951) stated that it is the "plexus" of environmental factorswhich determine both vegetation and soil in a concomitant manner.However, with western hemlock this may include the suggestion that thevegetation and soil also seem to co-determine each other. That is, althoughacid parent materials in the humid climate of the west coast seem beneficialto western hemlock, western hemlock also is reported as influencing the siteto cause more acid conditions.Thus the nature of the western hemlock ecosystem itself may preventplants or plant associations, an inferred nutrient/moisture gradient, or soilchemical measures from more precisely predicting site index. ConsideringFigure 4.18. Soil profile illustrating the thick mor humus forms andassociated decaying wood. The roots were noted to be predominantlyin the organic layer.127128the nature of western hemlock ecosystems, capturing the pattern of meansite index response, even though not as precise as what has been achievedfor species such as Douglas-fir (see literature review), was considered asuccessful description of western hemlock productivity.SUMMARY AND CONCLUSIONS(1) Despite the lack of understory plants in second growth westernhemlock ecosystems, the following relationships were found.Based on the frequency of nitrogen indicator species groups, the fourmain plant associations, derived from the traditional Braun-Blanquetmethod, captured an inferred nitrogen gradient, from the nitrogen poor(based on the dominance of oxylophytic species) Gaultheria plantassociation to the nitrogen rich (based on the dominance of nitrophyticspecies) Rubus plant association.An underlying nutrient gradient was supported through the use ofcanonical discriminant analysis on the PCA scores of the soil chemicalmeasures on a concentration basis. A definite, but overlapping, trend wasnoted. There was generally an increase in both mineral soil and forest floortotal and mineralizable nitrogen, a decrease in forest floor availablepotassium and a decrease in the carbon-nitrogen ratio in proceeding fromthe Gaultheria (oxylophytic) association to the Rubus (nitrophytic)association. Other supplementary soil chemical measure correlations withthis nutrient gradient were also noted.Analysis of the relationship between six predominant species, whichdistinguished the four major plant associations, with the forest floorchemical measures (on a concentration basis) was done using canonicalcorrelation analysis. Four forest floor chemical canonical variates explained37% of the variance of the species domain. Generally, species consideredoxylophytic (Klinka et al.. 1990) varied negatively with total andmineralizable nitrogen, and positively with available potassium andmagnesium. The relationships with nitrophytic species were reversed129130Both the plant associations and the six diagnostic species were linkedto an underlying nutrient gradient, despite the lack of understory plants.The nutrient gradient was generally correlated positively with nitrogen andnegatively with potassium. Potassium was thought to be simply a result ofa dilution effect on the nitrogen poorer sites rather than of a toxic nutrientstatus. Plant associations in the major variant sampled (CWHvm1), derivedthrough the Braun-Blanc:let method, may have use for managementpurposes to supplement the delineation and designation of an area to itssoil nutrient status. However, the variation and overlapping of forest floorchemical CDA scores between plant associations, emphasizes the need forfurther study, especially using western hemlock ecosystems having a betterexpression of understory species. Five of the six individual speciesexamined showed relationships with forest floor chemical measures whichsupport their nutrient indicator values reported in Klinka et al. (1990).(2a) The heuristic method of the BEC system, in conjunction withunderstory plants, was used to identify field estimated SNRs. A PCAordination was carried out on both the forest floor and mineral soil chemicalmeasures, expressed on a concentration basis. A plot of PCA axes 1 and 2,accounting for 32% and 18% of the variance respectively, captured adefinite gradation, primarily with the first axis, of increasing scores as SNRsproceeded from very poor to rich. The very poor and rich SNRs showed adistinct separation but the poor and medium overlapped substantially. PCAaxis 1 was highly positively correlated with forest floor and mineral soil totaland mineralizable nitrogen, and total sulphur. Highly negative correlationswere found with forest floor and mineral soil carbon-nitrogen ratios, andforest floor available potassium. Since this axis was most strongly131correlated with the nitrogen chemical measures (the four nitrogen measuresand the two carbon-nitrogen ratios), these six soil chemical measures wereused in a canonical discriminant analysis to further explore the relationshipof nitrogen with SNRs. Plots of the first two canonical variates showedsomewhat more distinct groupings than with PCA, but there still wassubstantial overlap between the poor and medium SNRs.As a validation procedure to investigate how well the nitrogenmeasures are related to SNRs, discriminant analysis was conducted to testthe ability of the six soil nitrogen chemical measures to differentiate amongthe four SNRs. To achieve a multivariate normal distribution, five of the sixvariables were transformed using common logarithms. Discriminantanalysis correctly classified 91% of the plots into their source groups.However, this high success rate was not repeatable; on a validation data set,the discriminant function correctly classified only 54% of the plots. Thediscriminant function then, captured the within site variability of theoriginal data set, but failed to capture between site variability with anindependent test data set.The results of this study provide evidence that supports the existenceof a relationship between nitrogen measures, expressed on a concentrationbasis, and field derived SNRs. However, unlike the study by Kabzems andKlinka (1987), significant differences between SNRs, using eithermineralizable or total nitrogen alone, could not be demonstrated. Forfurther actual statistical testing, as opposed to pattern analysis, it issuggested that a technique, such as in situ incubation to determinemineralizable nitrogen, be chosen to distinguish or revise site units, or torevise the heuristic field identification procedure.The success of the discriminant function in classifying the original132data into their source groups, and the lack of success in classifying anindependent test data set, suggests that the relationship developed on theoriginal data set was relative only. Consequently, there is a need for furtherstudy on a wider geographic area to assess whether a more general functioncan be derived, or whether a different function is required for smallerregions.(2b) The Energy-Soil Limited water balance model was used to correlatethe ratio of AET/PET with field derived SMRs. However, even the driest site,consisting of 1 cm of sand over bedrock, failed to show a moisture deficit.Even taking into account the fact that model validation and calibration forthis variant and for western hemlock has yet to be done, this result was stillconsidered unreasonable. It was suggested that in future runs of the ESLwater balance model for the CWHvm1 variant, the following time periodsshould be used: (i) a shorter time period of weeks or days, to see if waterdeficits occur only through part of the month; and (ii) yearly measurementsover the 30-year period from which climatic normals were derived.AET/PET could then be averaged, thus eliminating the over influence ofyears of very high growing season rainfall.Model calibration and validation for this variant and for westernhemlock needs to be carried out.(3) A summary of the regression equations developed in this study isgiven in Table 5.1. All models developed demonstrated an increase in siteindex to a certain point and then levelling off relative to the respectiveindependent variables. The SEEs of over 5 meters in all but model [4],showed that there was a relatively large variation around the regressionresponse surface.133Table 5.1: Summary of relationships between site index and indirect and directmeasures of ecological site quality. All variables are significant at p<0.05.Indirect Vegetation Measures of Ecological Site QualityCategorical[1] SI = 2.5 + 0.45(PA2) + 1.1(PA3) + 5.5(PA4) +6.4(PA5) + 8.1(PA6) + 9.2(PA7) + 13.7(PA8) +17.6(PA9) + 24.7(PA10) + 27.5(PA11) + 30.7(PAl2) + 33.1(PA13) + 33.2(PA14)Adjusted R2 = 0.78^F-ratio = 22.86^SEE = 5.3 m^n = 82Dummy variables representing various plant associations:Constant=Ledum; PA2=Rhaccomitrium; PA3=Sphagnum; PA4=Pinus(contorta);PAS= Vaccinium(ovatum); PA6= Veratrunr, PA7= Chamaecyparis; PA8=Gaultheria;PA9=Mahonia; PA10=Gaultheria-Polystichum; PA11=Achlys; PAl2=Polystichum;PA13=Rubus; PA14=Sambucus.Analytical[2] SI = 25.5 -4.3(PCA1) + 0.8(PCA2) - 1.2(PCA7) - 1.4(PCA9) + 1.0(PCA13)Adjusted R 2 = 0.78^F-ratio = 58.05^SEE = 5.1 m^n = 82Variables representing vegetation PCA axis scores.[3] SI = -12.1 - 545.4(MOIST2) + 33.2(MOIST3) + 37.1(MOIST4) + 16.6(MOIST5) +13.0(MOIST6) + 19.5(NITR2) + 19.2(NITR3)Adjusted R2 = 0.35^F-ratio = 7.19^SEE = 8.8 m^n = 82constant = excessively dry to very dry and N-poor; MOIST2 = very dry to moderately dry;MOIST3 = moderately dry to fresh; MOIST4 = fresh to very moist; MOIST5 = verymoist to wet; MOIST6 = wet to very wet; NITR2 = N-medium; NITR3 = N-rich.134Table 5.1: (continued)Indirect Environmental Measures of Ecological Site QualityCategorical[4] SI = 5.7 + 4.2(MD) + 8.9(SD) + 14.3(F) + 16.2(MST) + 15.7(VM) + 4.9(W) - 2.6(VP) + 7.2(P)+ 11.7(MED) + 13.2(R)Adjusted R 2 = 0.81^F-ratio = 43.7^SEE = 4.5 m^n = 102Dummy variables representing SMRs and SNRs:Constant=very wet or very rich; MD=moderately dry; SD= slightly dry; F=fresh; MST=moist;VM=very moist; W=wet; VP=very poor; P=poor; Med=medium; R=rich.[5] SI = 3.0 + 5.2(MD) + 14.6(SD) + 27.7(F) + 28.0(MST) + 30.0(VM) + 5.8(W)Adjusted R 2 = 0.69^F-ratio = 38.0.4^SEE = 5.8 m^N = 102Dummy variables representing SMRs:Constant=very wet; MD=moderately dry; SD= slightly dry; F=fresh; MST=moist; VM=verymoist; W=wet.[6] SI = 10.6 - 0.77(VP) + 16.3(P) + 21.3(MED) + 23.7(R)Adjusted R2 = 0.75^F-ratio = 77.3^SEE = 5.2 m^n = 102Dummy variables representing and SNRs:Constant=very rich; VP = very poor; p = poor; MED = medium; R = rich.[7] SI = 3.0 + 5.2(SS1) + 14.6(SS2) + 27.5(SS4) + 28.9(SS5) + 26.9(SS6) + 32.7(SS7) +4.9(SS8) + 7.5(SS9)Adjusted R 2 = 0.71^F-ratio = 31.92^SEE = 5.6 m^n = 102Dummy variables representing Site Series in the CWHvm1 variant:Constant= very wet/very poor; SS1 = HwPI-Cladina; SS2 = HwCw-Salal; SS4 = HwBa-Blueberry; SS5 = BaCw-Foamflower; SS6 = HwBa-Deer Fern; SS7 = BaCw-Salmonberry; SS8 = PI-Sphagnum; SS9 = CwSs-Skunk cabbage.135Table 5.1: (continued)Direct Measures of Ecological Site QualityAnalytical[8] SI = 76.7 + 22.0[Iog(total N)] - 23.7[log(available K]Adjusted R2 = 0.67^F-ratio = 54.21^SEE = 5.6 m^n = 54Variables representing soil chemicals:log(total N) = common log of mineral soil total nitrogen; log(available K) = common log ofmineral soil available potassium.[9] SI = 24.5 + 2.65(AXIS1) - 1.31(AXIS2) - 2.33(AXIS4) + 3.07(AXIS8)Adjusted R2 = 0.63^F-ratio = 23.92^SEE = 5.9 m^n=54Variables representing PCA axis scores of forest floor and mineral soil chemicals.[10]SI = 42.8 + 3.4[log(mineralizable N)] - 0.21 [available K]Adjusted R2 = 0.50^F-ratio = 27.14^SEE = 6.9 m^n = 54Variables representing soil chemicals:log(mineralizable N) = common log of forest floor mineralizable nitrogen; available K =forest floor available potassium.[11]SI = 28.4 + 18.6[Iog(MSAI)]Adjusted R 2 = 0.45^F-ratio = 44.6^SEE = 7.2 m^n = 54Variables representing soil chemicals:log(MSAI) = common log of mineral soil pyrophosphate extractable aluminium.136In terms of adjusted R2 and SEE, model [4] had the best fit,suggesting that the use of this model, which combines SNR and SMR, mayproduce the best prediction equation. This model suggests that the greatestproductivity of western hemlock is associated with moist/rich sites. Thedifference in mean site index between fresh and moist rich sites comparedto fresh and moist medium sites is not significant. The order of thedifference is less than 3 meters, which is probably considered in the samesite index class.Significant relationships with site index were developed for plantassociations and for individual species expressed in linear combinationsthrough PCA. However, due to the lack of understory plants, theserelationships may have questionable practical value. The use of plantassociations with strong expressions of understory plants has a potential ofbeing related to site index, considering these results with a poor expressionof understory plants.Of the direct analytical models, model [8] has the best fit. However,this model failed to predict successfully the site index of the data set fromwhich it was derived. Only 46% of the plots were correctly predicted within3 meters. The negative relationship with mineral soil potassium wasthought to be due to weathering of the parent material. The PCA regressionof mineral soil measures demonstrated the highly collinear structure of thedata. Based on the loadings, a high proportion of the soil chemicalmeasures are all related to site index.Graphically, a depiction of the two major results is given by plottingsite index against the first PCA axis of the soil chemical measures withSNRs indicated by elliptical outlines (Figure 4.18). The site indexrelationships show a logarithmic or asymptotic pattern, levelling of50-7-1 -) 40anctS30@) 20100137—6^—4^—2^0^2^4^6Soil Chemical PCA Axis 1Figure 4.19. Plotted relationship of site index and soil chemical PCA axes 1with 80% elliptical outlines indicating SNRs: 1 and V = very poor, 2and P = poor, 3 and M = medium, 4 and R = rich.138somewhere around poor-medium sites or the soil chemical measureequivalent. The high variation associated with site index for all soilchemical levels or SNRs is shown by the relatively large width of theelliptical outlines. This relatively large variation within similar sites andwithin similar nutrient levels is consistent with what is known about thenutrition of western hemlock.The relatively large variation in site index about the response surfacesmay be due to a combination of the nature of the data and the nature ofwestern hemlock ecosystems itself. The extreme heterogeneity of the site,especially with rotten logs, confounds the site factors used as independentvariables. Finally, the interactions of western hemlock and its associatesfurther complicate the system. Western hemlock seems to have a rootingpreference for wood, which seems to be a preferred location for theassociated mycorrhizae and some asymbiotic nitrogen fixers, since there isless competition. There is evidence that the mycorrhizae may provide theplant with increased ability to gain access to ammonium, phosphorus, andeven amino acids containing nitrogen and phosphorus. Not only doeswestern hemlock tolerate high acid conditions, but there is evidence that italso helps to maintain, or even, create it.Thus, the nature of the western hemlock ecosystem itself may notallow plants, an inferred nutrient/moisture gradient, or soil chemicalmeasures from more precisely predicting site index. Considering thisnature, capturing the pattern of mean site index response, even though notas precisely as what has been achieved for species such as Douglas-fir, wasconsidered to have successfully captured the character of western hemlockproductivity.139(4) A final consideration to explain the large variation in site indexassociated with soil nutrients, is to question whether there is an adaptiveadvantage for western hemlock to increase height growth in response toincreasing nutrient levels. For shade intolerant, early successional speciesa response in height growth would be an advantage. Since light becomesmore limiting relative to an increase in nutrient levels, there is a selectionpressure towards increased height growth in response to increased levels ofnutrients (Tilman 1986). However, for a shade tolerant, climax species,such as western hemlock, this may not necessarily still hold. Instead, itmay be expected that the selection advantage would be towardsphotosynthetic efficiency (as described by Kozlowski et a/. 1991). Radwanand DeBell (1980) reported significant correlations (p<0.01) between the siteindex of western hemlock with chlorophyll 'a' and total chlorophyll(extracted with 80% acetone) on a weight basis. The richer content ofchlorophyll in chloroplasts of shade-adapted leaves may allow for moreefficient light utilization (Boardman 1977). Thus increased nutrition mayonly be an advantage in its ability to enhance photosynthetic efficiencyunder shaded conditions. For shade tolerant species, further research isnecessary to determine whether increasing nutrient levels effects bothheight growth and volume increment concomitantly, or predominantly justone of the two.Additionally, since western hemlock is a climax species, there mayalso be an advantage to adapting to the nutritional soil characteristics of amature ecosystem. Odum (1969) suggested that mature ecosystems have agreater capacity for nutrient retention, and that the selection pressure forspecies adapted to mature ecosystems would be towards this feedbackcontrol, as opposed to growth. Humffication, a dominant process associated140with western hemlock (thus the development of thick mor humus forms),results in the removal of relatively simple organic compounds from theeasily metabolized soil organic matter pool and their addition to therelatively stable soil organic matter fractions (Tate 1987). The accumulationof humus for nutrient retention may be important in the high rainfallclimates associated with the distribution of western hemlock where there ispressure on the leaching of nutrient elements (podzolization being adominant pedogenic force). Although resistant, humus remains subject tomicrobial decomposition but at a slow rate. Thus western hemlock may beadapted to this slow release of nutrients and, as suggested by Krajina(1969), may grow best with a well balanced supply of nutrients in smallquantities.Perhaps then, the key to western hemlock nutrition is the ability ofthe site to conserve nutrients, yet release them slowly. The soil chemicalmeasure required to capture increasing western hemlock productivity maybe one that not only tests the amount of nutrients, but also the capacity toconserve nutrients, and the ability to slowly release nutrients. Methods ofevaluating organic matter, not only its chemical nature but also its chemicalstructure, need to be investigated as a possible influence on westernhemlock nutrition.141LITERATURE CITEDAber, J.D., and J.M Melillo. 1984. Nutrient cycling models and landclassification. In: Proceedings of the symposium on forest landclassification: experiences, problems, perspectives, ed. J.G. Bockheim,pp.205-217. University of Wisconsin.Adams, M.A., P.J. Polglase, P.M. Attiwil, and C.J. Weston. 1989. In situ studiesof nitrogen mineralization and uptake in forest soils; some comments onmethodology. Soil Biology and Biochemistry 21(3): 423-429.Agriculture Canada Expert Committee on Soil Survey. 1987. The Canadiansystem of soil classification. 2nd ed. Agriculture Canada Publication1646. 164 pp.Alban, D.H. 1969. The influence of western hemlock and western redcedar onsoil properties. Soil Science Society of America Proceedings 33: 453-457.Alexander, R.R., D. Tackle, and W.G. Dahms. 1967. Site indexes for lodgepolepine, with correction for stand density: methodology. USDA ForestService, Rocky Mountain Forest and Range Experiment Station, ResearchPaper RM-29.Anderson, S., R.J. Zasoski, and S.P. Gessel. 1982. Phosphorus and limeresponse of Sitka spruce, western hemlock seedlings, and romainelettuce on two coastal Washington soils. Canadian Journal of ForestResearch 12: 985-991.Anonymous. 1976. Technicon Autoanalyzer II. Methodology: Individualsimultaneous determination of nitrogen and phosphorus. IndustrialMethod No. 334-74A.Anonymous. 1982. Canadian climate normals, 1951-1980, temperature andprecipitation. Environment Canada, Atmospheric Environment Service.Ottawa, Ontario. 268pp.Armstrong, J.E., D.R. Crandell, D.J. Easterbrook, and J.B. Noble. 1965. LatePleistocene stratigraphy and chronology in southwestern BritishColumbia and northwestern Washington. Geological Society of AmericaBulletin 76: 321-330.Baker, J.B., and W.M Broadfoot. 1977. A practical field method of siteevaluation for eight important southern hardwoods. USDA Forest Service,Southern Forest Experiment Station, General Technical Report SO-14. 31pp.Baker, J.B., and W.M. Broadfoot. 1979. A practical field method of siteevaluation for commercially important southern hardwoods. USDA ForestService, Southern Forest Experiment Station, General Technical ReportSO-26. 51 pp.142Bakuzis, E.V. 1969. Forestry viewed in an ecosystem perspective. In: Theecosystem concept in natural resource management, ed. G.M. Van Dyne,pp 189-258. Academic Press, New York.Ballard, R. 1980. Phosphorus nutrition and fertilization of forest trees. In: Therole of phosphorus in agriculture, ed. F.E. Khasawneh, E.C. Sample andE.J. Kamprath, pp 763-804. National Fertilizer Development Center,Tennessee Valley Authority, Muscle Shoals, Alabama. Soil Science Societyof America, Madison, Wisc.Ballard, T.M., and R.E. Carter. 1986. Evaluating forest stand nutrient status.Land Management Report No. 20, B.C. Ministry of Forests and Lands,Victoria, B.C. 60 pp.Ballard, T.M., and M.G. Dosskey. 1985. Needle water potential and soil-to-foliage flow resistance during soil drying: a comparison of Douglas-fir,western hemlock and mountain hemlock. Canadian Journal of ForestResearch 15: 185-188.Banner, A., R.N. Green, K. Klinka, D.S. McLennan, D.V. Meidinger, F.C.Nuszdorfer, and J. Pojar. 1990. Site classification for coastal BritishColumbia: a first approximation. B.C. Ministry of Forests, Victoria, B.C. 2pp. (a coloured pamphlet).Bardsley, C.E., and J.D. Lancaster. 1960. Determination of reserve sulfur andsoluble sulfates in soils. Soil Science Society of America Proceedings24:265-268.Bardsley, C.E., and J.D. Lancaster. 1965. Sulfur. In: Methods of soil analysis,ed. C.A. Black, D.D. Evans, J.L White, L.E. Ensminger, and F.E. Clark, pp1102-1116. Agronomy No. 9, American Society of Agronomy.Barnes, B.V. 1984. Forest ecosystem classification and mapping in Baden-Warttenberg, West Germany. In: Forest land classification: experience,problems, perspectives, ed. J.G. Bockheim, pp 49-65. University ofWisconsin.Barnes, B.V. 1986. Varieties of experience in classifying and mappingforestland ecosystems. In: Site classification in relation to forestmanagement, ed. G.M. Wickware and W.C. Stevens, pp. 5-23. CanadianForestry Service, Great Lakes Forestry Centre, COJFRC Proceedings O-P-14.Barnes, B.V., K.S. Pregitzer, T.A. Spies, and V.H. Spooner. 1982. Ecologicalforest site classification. Journal of Forestry 80(8): 493-498.Barrett, J.W. 1965. Spacing and understory vegetation affect growth ofponderosa pine saplings. USDA Forest Service, Pacific Northwest Forestand Range Experiment Station, Research Note PNW 27.Barrett, J.W. 1970. Ponderosa pine saplings respond to control of spacing andunderstory vegetation. Pacific Northwest Forest and Range ExperimentStation Research Paper PNW-106.143Barrett, J.W. 1973. Latest results from the Pringle Falls ponderosa pinespacing study. USDA Pacific Northwest Forest and Range ExperimentStation Research Note PNW 209.Bascombe, C.L. 1968. Distribution of pyrophosphate-extractable iron andorganic carbon in soils of various groups. Journal of Soil Science 19:251-268.Beck, D.E., and K.B. Trousdell. 1973. Site index: accuracy of prediction.USDA Forest Service, Southeastern Forest Experiment Station ResearchPaper SE-108.Becking, R.W. 1957. The Ziirich-Montpellier school of phytosociology. TheBotanical Review 23(7): 411-488.Bennett, F.A. 1975. Slash pine: some intriguing growth and yieldrelationships. USDA Forest Service, Southeastern AgriculturalExperiment Station, Research Paper SE-133.Binkley, D. 1983. Ecosystem production in Douglas-fir plantations: interactionof red alder and site fertility. Forest Ecology and Management 5:215-227.Binkley, D. 1986. Forest nutrition management. John Wiley and Sons, NewYork, U.S.A. 290 pp.Binkley, D., and S.C. Hart. 1989. The components of nitrogen availabilityassessments in forest soils. Advances in Soil Science 10: 57-112.Binkley, D., and D. Richter. 1987. Nutrient cycles and H + budgets of forestecosystems. Advances in Ecological Research 16: 1-51.Boardman, K. 1977. Comparative photosynthesis of sun and shade plants.Annual Review of Plant Physiology 28: 355-377.Bowen, G.D. 1984. Tree roots and the use of soil nutrients. In: Nutrition ofPlantation Forests, ed. G.D. Bowen and E.K.S. Nambiar, pp 147-179.Academic Press, Orlando, U.S.A.Braun-Blanquet, J. 1932. Plant sociology: the study of plant communities.Translated by G.D. Fuller and H.S. Conrad. New York, New York. 439pp.Bray, R.H., and L.T. Kurtz. 1945. Determination of total, organic, andavailable forms of phosphorus in soils. Soil Science 59: 39-45.Bremner, J.M., and M.A. Tabatabai. 1971. Use of automated combustiontechniques for total carbon, total nitrogen and total sulphur analysis ofsoils. In: Instrumental Methods for Analysis of Soils and Plant Tissue,ed. L.M. Walsh, pp. 1-16. Soil Science Society of America, Madison,Wisconsin.144Bremner, J.M., and C.S. Mulvaney. 1982. Nitrogen -- total. In: Methods of soilanalysis, Part 2, ed. A.L Page, R.H. Miller, and D.R. Keeney, pp. 595-624.Agronomy Monograph Number 9 (2nd Edition). 1159 pp.Brink, V.C. 1954. Survival of plants under flood in the lower Fraser RiverValley, British Columbia. Ecology 35: 94-95.Broadfoot, W.M. 1969. Problems in relating soil to site index for southernhardwoods. Forest Science 15(4): 354-364.Bruce, D. 1923. Anamorphosis and its use in forest graphics. Journal ofForestry 21(8): 773-783.Bruce, D. 1926. A method of preparing timber-yield tables. Journal ofAgriculture Research 32: 543-557.Bull, H. 1931. The use of polymorphic curves in determining site quality inyoung red pine plantations. Journal of Agricultural Research 43(1): 1-28.Burger, D. and G. Pierpoint. 1990. Trends in forest site and land classificationin Canada. The Forestry Chronicle 66(2): 91-96.Cajander, A.K. 1926. The theory of forest types. Acta Forestalia Fennica 2(3):11-108.Campbell, N.A. 1980. Robust procedures in multivariate analysis. AppliedStatistics 29: 231-237.Carmean, W.H. 1970. Tree height growth patterns in relation to soil and site.In: Tree growth and forest soils, ed. C.T. Youngberg and C.B. Davey, pp499-512. Third North American Forest Soils Conference Proceedings.Carmean, W.H. 1975. Forest site quality evaluation in the United States.Advanced Agronomy 27:209-269.Carmean, W.H. 1986. Forest site quality research in the north central region.In: Site classification in relation to forest management, ed. G.M. Wickwareand W.C. Stevens, pp. 62-67. Canada-Ontario Joint Forestry ResearchCommittee Symposium Proceedings 0-P-14, Sault Saint Marie, Ontario.Carter, R.E., and K. Klinka. 1987. Relationships between Douglas-fir siteindex, actual evapotranspiration, and soil nutrients. A paper presented atthe IUFRO Seminar: Management of water and nutrient relations toincrease forest growth. Canberra, Australia., 19-22 October, 1987.Carter, R.E., and K. Klinka. 1990. Relationships between growing-season soilwater-deficit, mineralizable soil nitrogen and site index of coastal Douglas-fir. Forest Ecology and Management 30: 301-311.Carter, R.E., and L.E. Lowe. 1986. Lateral variability of forest floor propertiesunder second-growth Douglas-fir and the usefulness of compositesampling techniques. Canadian Journal of Forest Research 16:1128-1132.145Causton, D.R 1988. Introduction to vegetation analysis. Unwin Hyman,London, England. 342 pp.Chambers, J.M., W.S. Cleveland, B. Kleiner, and P.A. Tukey. 1983. Graphicalmethods for data analysis. Wadsworth & Brooks/Cole Publishing Co.Advanced Books & Software, Pacific Grove, California. 395 pp.Chapin, F.S.H. III, K. Van Cleve, and P. Vitousek. 1986. The nature of nutrientlimitation in plant communities. American Naturalist 127: 148-158.ChatterJee, S., and B. Price. 1991. Regression analysis by example. JohnWiley and Sons, Inc., New York. 278 pp.Chikishev, A.G. 1965. Plant indicators of soils, rocks, and subsurface waters.(Translated from Russian). Consultants Bureau, New York. 210 pp. .Christy, E.J., and R.N. Mack. 1984. Variation in demography of juvenile Tsugaheterophylla across the substratum mosaic. Journal of Ecology 72:75-91.Clements, F.E. 1916. Plant succession. An analysis of the development ofvegetation. Carnegie Institute, Washington, Publication Number 242,Washington, D.C.Clements, F.E. 1920. Plant indicators: the relation of plant communities toprocess and practice. Carnegie Institute of Washington, PublicationNumber 290. 388 pp.Clutter, J.L., J.C. Forston, L.V. Pienaar, G.H. Brister, and R.L. Bailey. 1983.Timber management: a quantitative approach. John Wiley and Sons, NewYork, New York. 333 pp.Coile, T.S. 1952. Soil and the growth of forests. In: Advances in agronomy, ed.A.G. Norman, pp. 329-398. Academic Press Inc., New York.Collins, A.B. 1967. Density and height growth in natural slash pine. USDAForest Service, Southeastern Forest and Range Experiment Station,Research Paper SE-27.Comeau, P.G. 1986. Net primary production of lodgepole pine (Pinus contortavar. latifolia Engelm.) in some ecosystems in southeastern BritishColumbia. Ph.D. dissertation, University of British Columbia, Vancouver,B.C. 232 pp.Cook, R.E. 1983. Clonal plant populations. American Scientist 71: 244-253.Coopersmith, D.J. 1986. Root and fungal biomass production on low, medium,and high productivity second-growth Douglas-fir stands on VancouverIsland. M.Sc. Thesis, University of British Columbia, Vancouver, B.C. 244pp.Courtin, P., M.C. Feller, and K. Klinka. 1983. Lateral variability in someproperties of disturbed forest soils in southwestern British Columbia.Canadian Journal of Soil Science 63:529-539.146Courtin, P.J., K. Klinka, M.C. Feller, and J.P. Demaerchalk. 1988. Anapproach to quantitative classification of nutrient regimes of forest soils.Canadian Journal of Botany 66:2640-2653.Coutts, M.P., and J.J Philipson. 1976. The influence of mineral nutrition onthe root development of trees. I. The growth of Sitka spruce with dividedroot systems. Journal of Experimental Botany 27(100): 1102-1111.Crampton, C. B. 1982. Podzolization of soils under individual tree canopies insouthwestern British Columbia, Canada. Geoderma 28: 57-61.Crampton, C.B. 1984. Concentric zonation of gleyed soils under individualcanopies in southwestern British Columbia, Canada. Geoderma 32: 329-334.Crawley, M.J. 1986. The structure of plant communities. In: Plant ecology, ed.M.J. Crawley, pp 1-50. Blackwell Scientific Publications, London, England.396 pp.Crovello, T.J. 1970. Analysis of character variation in ecology and systematics.Annual Review of Ecology and Systematics 1:55-98.Curran, P.M 1984. Soil testing for phosphorus availability to some conifers inBritish Columbia. M.Sc. Thesis, University of British Columbia,Vancouver, B.C. 130 pp.Daniel, T.W., J.A. Helms, and F.S. Baker. 1979. Principles of silviculture.McGraw-Hill Book Company, New York. 500pp.Daubenmire, R. 1961. Vegetative indicators of rate of height growth inponderosa pine. Forest Science 7(1): 24-34.Daubenmire, R. 1968. Plant communities: a textbook of plant synecology.Harper and Row, Publishers, New York. 300 pp.Dietrich, R.V., and B.J. Skinner. 1979. Rocks and rock minerals. John Wiley& Sons, New York. 319pp.Dillon, W.R., and M. Goldstein. 1984. Multivariate analysis: methods andapplications. John Wiley and Sons, Inc., New York. 579 pp.Doolittle, W.T. 1957. Site index of scarlet and black oak in relation to southernAppalachian soil and topography. Forest Science 3(2): 114-124.Drew, M.C. 1975. Comparison of the effects of a localized supply of phosphate,nitrate, ammonium, and potassium on the growth of the seminal rootsystem and the shoot in barley. New Phytologist 75:479-490.Drew, M.C., and L.R. Saker. 1975. Nutrient supply and the growth of theseminal root system in barley. II. Localized compensatory increases inlateral root growth and rates of nitrate uptake when nitrate supply isrestricted to only part of the root system. Journal of Experimental Botany26: 79-90.147Drew, M.C., and L.R. Saker. 1978. Nutrient supply and the growth of theseminal root system in barley. III. Compensatory increase in growth oflateral roots and in rates of phosphate uptake, in response to a localizedsupply of phosphate. Journal of Experimental Botany 29: 435-451.Drew, M.C., L.R. Saker, and T.W. Ashley. 1973. Nutrient supply and thegrowth of the seminal root system in barley. I. The effect of nitrateconcentration on the growth of axes and laterals. Journal of ExperimentalBotany 24: 1189-1202.Eis, S. 1962. Statistical analysis of several methods for estimation of foresthabitats and tree growth near Vancouver, B.C. Forestry Bulletin, Facultyof Forestry, University of British Columbia No.4. 76 pp.Eis, S. 1974. Root system morphology of western hemlock, western redcedar,and Douglas-fir. Canadian Journal of Forest Research 4:28-38.Eis, S. 1987. Root systems of older immature, cedar and Douglas-fir.Canadian Journal of Forest Research 17:1348-1354.Emanuel, J. 1989. VTAB: Ecosystem classification and analysis. Unpublishedmanuscript, Faculty of Forestry, University of British Columbia, Vancouver,B.C. 84 pp.Fisher, R.A. 1946. The use of multiple measurements in taxonomic problems.Annals of Eugenics 7:179-188.Fitter, A.H. 1986. Acquisition and utilization of resources. In: Plant Ecology,ed. M.J. Crawley. Blackwell Scientific Publications, London, England. 496pp.Fowells, H.A. 1965. Silvics and forest trees of the United States. USDA ForestService, Agriculture Handbook No. 271. 762pp.Franklin, J.F., and C.T. Dyrness. 1973. Natural vegetation of Oregon andWashington. USDA Forest Service General Technical Report PNW 8. 417pp.Frey, T.E.A. 1980. The Finish school and forest site types. In: Classification ofPlant Communities, ed. R.H. Whittaker ed. pp 81-110. Dr. W. Junk byPublishers, The Hague, Netherlands.Fulton, R.J. 1971. Radiocarbon geochronology of southern British Columbia.Geological Survey of Canada, Department of Energy, Mines and Resources,Paper 71-37.Gauch, H.G., Jr. 1986. Multivariate Analysis in Community Ecology.Cambridge University Press, New York. 298 pp.Gessel, S.P., and C.D. Oliver. 1981. Soil-site relationships and productivity oftrue firs. In: Proceedings of the biology and management of true fir in thePacific Northwest symposium, eds. C.D. Oliver and R.M Kenady, pp. 174-184, Seattle, Washington.148Giles, D.G., T.A. Black, and D.L. Spittlehouse. 1985. Determination of growingseason soil water deficits on a forested slope using water balance analysis.Canadian Journal of Forest Research 15: 107-114.Gill, R.S., and D.P. Lavender. 1983a. Urea fertilization; effects on primary rootmortality and mycorrhizal development of young-growth western hemlock.Forest Science 29: 751-760.Gill, R.S., and D.P. Lavender. 1983b. Urea fertilization and foliar nutrientcomposition of western hemlock (Tsuga heterophylla (Raf.) Sarg.). ForestEcology and Management 6: 333-341.Gittins, R. 1979. Ecological applications of canonical analysis. In: Multivariatemethods in ecological work, ed. L. OrlOci, C.R. Rao, and W.M Stiteler, pp.309-535. International Co-operative Publishing House, Fairland, Maryland.Gittins, R. 1985. Canonical analysis: a review with applications. Springer-Verlag, New York, New York. 351 pp.Gleason, H.A. 1926. The individualistic concept of plant succession. Bulletinof the Torrey Botanical Club 53:7-26.Green, R.N., P.L. Marshall, and K. Klinka. 1989. Estimating site index ofDouglas-fir (Pseudotsuga menziesii [Mirb.] Franco) from ecologicalvariables in southwestern British Columbia. Forest Science 35:50-63.Grewelling, T., and M. Peech. 1960. Chemical soil tests. Cornell AgricultureExperiment Station Bulletin 960.Grieg-Smith, P. 1983. Quantitative plant ecology. University of CaliforniaPress, Los Angeles, California. 359 pp.Grigal, D.F. 1984. Shortcomings of soil surveys for forest management. In:Proceedings of the symposium on forest land classification: experiences,problems, perspectives, ed. J.G. Bockheim, pp. 148-166. University ofWisconsin.Groner, R., M. Groner, and W.F. Bischof. 1983. Methods of Heuristics.Lawrence Erlbaum Associates, Hillsdale, New Jersey. 416 pp.Gruvaeus, G., and H. Wainer. 1972. Two additions to hierarchical clusteranalysis. The British Journal of Mathematical and Statistical Psychology25:200-206.Hägglund, B. 1981. Evaluation of forest site productivity. Forestry AbstractsReview Article 42 (11):515 -527.Hall, F.C. 1983. Growth basal area: a field method for appraising forest sitepotential for stockability. Canadian Journal of Forest Research 13: 70-77.Harmon, M.E., and J.F. Franklin. 1989. Tree seedlings on logs in Picea-Tsugaforests of Oregon and Washington. Ecology 70(1): 48-59.149Harmon, M.E., J.F. Franklin, F.J. Swanson, P. SoBins, S.V. Gregory, J.D.Lattin, N.H. Anderson, S.P. Cline, N.G. Aumen, J.R. Sedell, G.W.Lienkaemper, K. Cromack, Jr., and K.W. Cummins. 1986. Ecology ofcoarse woody debris in temperate ecosystems. Advances in EcologicalResearch 15: 133-302.Harrington, C.A. 1986. A method of site quality evaluation for red alder.USDA Forest Service, Pacific Northwest Research Station, GeneralTechnical Report PNW-192. 22 pp.Hart, S.C., and M.K. Firestone. 1989. Evaluation of three in situ soil nitrogenavailability assays. Canadian Journal of Forest Research 19: 185-191.Harvey, A.E., M.F. Jurgensen, M.J. Larsen, and J.A. Schlieter. 1986.Distribution of active ectomycorrhizal short roots in forest soils of theInland Northwest: effects of site and disturbance. Department ofAgriculture, Forest Service, Intermountain Research Station ResearchPaper INT 374. 8 pp.Harvey, A.E., M.J. Larsen, M.F. Jurgensen, and E.A. Jones. 1989. Nitrogenaseactivity associated with decayed wood of living northern Idaho conifers.Mycologia 81(5): 765-771Healy, M.J.R. 1968. Multivariate normal plotting. Applied Statistics 17: 157-161.Heilman, P. 1976. Soils and site index in coastal western hemlock forests ofWashington and Alaska. In: Western hemlock management conference,ed. W.A. Atkinson and R.J. Zasoski, pp. 39-48. University of Washington,Seattle, Washington. 317pp.Heilman, P.E., and G. Ekuan. 1973. Response of Douglas-fir and westernhemlock seedlings to lime. Forest Science 19: 220-224.Heilman, P.E., and C.R. Gass. 1974. Parent materials and chemical propertiesof mineral soil in southeast Alaska. Soil Science 117(1): 21-27.Hill, M.O. 1973. Reciprocal averaging: an eigenvector method of ordination.Journal of Ecology 61:237-251.Hill, M.O. 1979. DECORANA -- A FORTRAN program for detrendedcorrespondence analysis and reciprocal averaging. Ecology andSystematics, Cornell University, Ithaca, New York. 52pp.Hodgkins, E.J. 1956. Testing soil-site index tables in southwestern Alabama.Journal of Forestry 54(4): 261-266.Holland, S.S. 1976. Landforms of British Columbia. B.C. Department ofMines and Petroleum Research Bulletin 48.Hotelling, H. 1933. Analysis of a complex of statistical variates into principalcomponents. Journal of Educational Psychology 24:417-441, 498-520.Ilvessalo, Y. 1927. Methods for preparing yield tables. Silva Fennica 5: 1-32.150Jameson, J.S. 1963. Relation of jack pine height-growth to site in themixedwood forest section of Saskatchewan. In: Forest-soil relationships inNorth America, ed. C.T. Youngberg, pp. 299-316. Oregon State UniversityPress, Corvallis, Oregon.Jardine, P.M., and D.L. Sparks. 1984. Potassium-calcium exchange in a multi-reactive soil system: II. Thermodynamics. Soil Science Society of AmericaJournal 47:45-50.Jeglum, J.K. 1974. Relative influence of moisture-aeration and nutrients onvegetation and black spruce growth in northern Ontario. CanadianJournal of Forest Research 4: 114-126.Jenny, H. 1941. Factors of soil formation. McGraw-Hill, New York. 281 pp.Johnson, C.M., and H. Nishita. 1952. Microestimation of sulfur in plantmaterials, soils, and irrigation waters. Analytical Chemistry 24:736-742.Johnson, R.E., C.L. Haag, J.G. Bockheim, and G.G. Erdmann. 1987. Soil-siterelationships and soil characteristics associated with even-aged red maple(Acer rubrum) stands in Wisconsin and Michigan. Forest Ecology andManagement 21: 75-89.Jolicoeur, P., and J.E. Mosimann. 1960. Size and shape variation in thepainted turtle. A principal component analysis. Growth 24:339-354.Jolliffe, I.T. 1986. Principle component analysis. Springer-Verlag, New York.271 pp.Jones, J.R. 1969. Review and comparison of site evaluation methods. USDAForest Service, Rocky Mountain Forest and Range Experiment StationResearch Paper RM-51. 27pp.Jurgensen, M.F., M.J. Larsen, R.T. Graham, and A.E. Harvey. 1987. Nitrogenfixation in woody residues of northern Rocky Mountain conifer forests.Canadian Journal of Forest Research 17: 1283-1288.Jurgensen, M.F., M.J. Larsen, S.D. Spano, A.E. Harvey, and M.R. Gale. 1984.Nitrogen fixation associated with increased wood decay in Douglas-firresidue. Forest Science 30: 1038-1044.Kabzems, R.D. 1988. Some relationships between soil nutrients, vegetationand site index for western hemlock ecosystems in the Very Wet MaritimeCWH Subzone. Section 88 Research Report, CIP Inc., Gold River, B.C.Kabzems, R.D., and K. Klinka. 1987a. Initial quantitative characterization ofsoil nutrient re es. I. Soil properties. Canadian Journal of ForestResearch 17:1 57-1564.Kabzems, R.D., and K. Klinka. 1987b. Initial quantitative characterization ofsoil nutrient regimes. II. Relationships among soils, vegetation, and siteindex. Canadian Journal of Forest Research 17:1565-1571.151Kelly, A.P. 1922. Plant indicators of soil types. Soil Science 13(6): 411-423.Kenkel, N.C., P. Juhász-Nagy, and J. Podani. 1989. On sampling proceduresin population and community ecology. Vegetatio 82: 195-207.Keyes, M.R., and C.C. Grier. 1981. Above- and below-ground net production in40-year-old Douglas-fir stands on low and high productivity sites.Canadian Journal of Forest Research 11: 599-605.Kimmins , J.P. 1977. The need for ecological classification in B.C. B.C.Forestry volume 2, number 1. 2pp.Kimmins, J.P. 1988. Community organization: methods of study andprediction of productivity and yield of forest ecosystems. CanadianJournal of Botany 66:2654-2672.Kimmins, J.P., and B.C. Hawkes. 1978. Distribution and chemistry of fineroots in a white spruce-subalpine fir stand in British Columbia:implications for management. Canadian Journal of Forest Research 8(3):265-279.Kimmins, J.P., P.G. Comeau, and W. Kurz. 1990. Modelling the interactionsbetween moisture and nutrients in the control of forest growth. ForestEcology and Management 30: 361-379.King, J.E. 1966. Site index curves for Douglas-fir in the Pacific Northwest.Weyerhauser Forestry Paper 8, 49 p.Klinka, K., and R.E. Carter. 1990. Relationships between site index andsynoptic environmental variables in immature coastal Douglas-fir stands.Forest Science 36:815-830.Minim, K., and V.J. Krajina. 1987. Ecosystems of the University of BritishColumbia Research Forest, Haney, B.C. Faculty of Forestry, University ofBritish Columbia, Vancouver, B.C. 123 pp.Klinka, K., A.M Scagel, and P.J. Courtin. 1985. Vegetation relationshipsamong some seral ecosystems in southwestern British Columbia.Canadian Journal of Forest Research 15(3): 561-569.Minim., K., J. Pojar, and D.V. Meidinger. 1991. Revision of Biogeoclimatic unitsof coastal British Columbia. Northwest Science 65(1): 32-47.Klinka, K., M.C. Feller, L.M. Lavkulich, and A. Kozak. 1980. Evaluation ofmethods of extracting soil cations for forest productivity studies insouthwestern British Columbia. Canadian Journal of Soil Science 60:697-705.Klinka, K., R.E. Carter, M.C. Feller, and Q. Wang. 1989. Relations betweensite index, salal, plant communities, and sites in coastal Douglas-firstands. Forest Science 36:815-830.152Klinka, K., R.N. Green, P.J. Courtin, and F.C. Nuszdorfer. 1984. Sitediagnosis, tree species selection, and slashburning guidelines for theVancouver Region. Land Management Report No. 25, Ministry of Forests,Province of British Columbia, Victoria, B.C. 180 pp.Klinka, K., R.N. Green, R.L. Trowbridge, and L.E. Lowe. 1981. Taxonomicclassification of humus forms in ecosystems of British Columbia. LandManagement Report No. 8, Ministry of Forests, Province of BritishColumbia, Victoria, B.C. 54 pp.Klinka, K., V.J. Krajina, A. Ceska, and A.M Scagel. 1989. Indicator plants ofcoastal British Columbia. University of B.C. Press, Vancouver, B.C. 288pp.Knapp, R. 1984. Sample (relevd) areas (distribution, homogeneity, size, shape)and plot-less sampling. In: Sampling methods and taxon analysis invegetation science, ed. R. Knapp, pp. 101-119. Handbook of vegetationscience part 4, Dr. W. Junk Publishers, Boston. 370 pp.Kojima, S. 1981. Biogeoclimatic ecosystem classification and its practical usein forestry. Journal of the College of Liberal Arts, Toyama University14(1): 41-75.Kojima, S. 1983. Forested plant assemblies of the lower northern subalpineregions of Alberta and their productivity relationships. In: IUFROSymposium on forest site and continuous productivity, ed. R. Ballard andS.P. Gessel, pp 13-30. USDA Forest Service, Pacific Northwest Forest andRange Experiment Station, General Technical Report PNW-163.Kowalenko, C.G., and L.E. Lowe. 1972. Observations on the bismuth sulfidecolorimetric procedure for sulfate analysis in soil. Communications inSoil Science and Plant Analysis 3(1): 79-86.Kozlowski, T.T., P.J. Kramer, and S.G. Pallardy. 1991. The physiologicalecology of woody plants. Academic Press, Inc., San Diego, California. 657pp.Krajina, V.J. 1959. Bioclimatic zones in British Columbia. The University ofBritish Columbia Botanical Series No. 1.Krajina, V.J. 1960. Ecosystem classification of forests. Silva Fennica 105:107-110.Krajina, V.J. 1965. Biogeoclimatic zones and biogeocoenoses of BritishColumbia. Ecology of Western North America 1:1-17.Krajina, V.J. 1969. Ecology of forest trees in British Columbia. EcologyWestern North America 2(1): 1-146.Krajina, V.J. 1972. Ecosystem perspectives in forestry. H.R. MacMillanForestry Lecture Series. Faculty of Forestry, University of B.C.,Vancouver, B.C.153Krajina, V.J., K. Klinka, and J. Worrall. 1982. Distribution and ecologicalcharacteristics of trees and shrubs of British Columbia. The University ofBritish Columbia, Faculty of Forestry, Vancouver, B.C. 131pp.Krajina, V.J., Madoc-Jones, S., and G. Mellor. 1973. Ammonium and nitratein the nitrogen economy of some conifers growing in Douglas-fircommunities of the Pacific North-west of America. Soil Biology andBiochemistry 5: 143-147.Kropp, B.R. 1982a. Fungi from decayed wood as ectomycorrhizal symbionts ofwestern hemlock. Canadian Journal of Forest Research 12: 36-39.Kropp, B.R. 1982b. Rotten wood as mycorrhizal inoculum for containerizedwestern hemlock. Canadian Journal of Forest Research 12: 428-431.Kropp, B.R. 1982c. Formation of mycorrhizae on nonmycorrhizal westernhemlock outplanted on rotten wood and mineral soil. Forest Science28(4): 706-710.Kropp, B.R., and J.M. Trappe. 1982. Ectomycorrhizal fungi of Tsugaheterophylla. Mycologia 74(3):479-488.Kurz, W.A. 1989. Net primary production, production allocation, and foliageefficiency in second growth Douglas-fir stands with differing site quality.Ph.D. dissertation, University of British Columbia, Vancouver, B.C. 224pp.Larsen, M.L., M.F. Jurgensen, and A.E. Harvey. 1978. N2 fixation associatedwith wood decayed by some common fungi in western Montana.Canadian Journal of Forest Research 8: 341-345.Laursen, G.F. 1985. Mycorrhizae: a review of the importance of fungi fromhigh latitude forests of Alaska. Agroborealis 17(2): 58-66.Lavkulich, L.M. 1978. Methods manual. Pedology Laboratory. Department ofSoil Science, University of B.C., Vancouver, B.C. 224 pp.Lemieux, G.J. 1964. Ecology and productivity of the northern hardwoodforests of Quebec. Ph.D. Dissertation, University of Michigan. 144pp.Lewis, T. 1976. The till-derived podzols of Vancouver Island. Ph.D.dissertation, University of British Columbia, Vancouver, B.C.Livingston, N.J., and T.A. Black. 1987a. Water stress and survival of threespecies of conifer seedlings planted on a high elevation south-facing clear-cut. Canadian Journal of Forest Research 17: 1115-1123.Livingston, N.J., and T.A. Black. 1987b. Stomatal characteristics andtranspiration of three species of conifer seedlings planted on a highelevation south-facing clear-cut. Canadian Journal of Forest Research 17:1273-1282.154Lowe, L.E., and T.F. Guthrie. 1981. The development of an analytical methodto determine sulphur content of foliage. Contract Research Report,Research Branch, B.C. Ministry of Forests, Victoria, B.C.Lowe, L.E., and K. Klinka. 1981. Forest humus in the Coastal WesternHemlock biogeoclimatic zone of British Columbia in relation to forestproductivity and pedogenesis. Ministry of Forests, Province of BritishColumbia Research Note 89.Lundberg, G. 1970. Utilization of various nitrogen sources, in particularbound soil nitrogen, by mycorrhizal fungi. Studia Forestalia Suecica 79:1-95.Luttmerding, HA, D.A. Demarchi, E.C. Lea, D.V. Meidinger, and T. Vold (eds.).1990. Describing Ecosystems in the Field. 2nd ed., B.C. Ministry ofForests, Victoria, B.C. 213 pp.Lynch, D.W. 1958. Effects of stocking on site measurement and yield ofsecond growth ponderosa pine in the Inland Empire. USDA ForestService, Intermountain Forest and Range Experiment Station, ResearchPaper 56. 36 pp.Mader, D.L. 1963. Volume growth measurement -- an analysis of function andcharacteristics in site evaluation. Journal of Forestry 61(3): 193-198.Major, J. 1951. A functional, factorial approach to plant ecology. Ecology 32:392-412.Major, J. 1969. Historical development of the ecosystem concept. In: TheEcosystem Concept in Natural Resource Management, ed. G.M. Van Dyne,pp. 9-32. Academic Press, New York.Makkonen, 0. 1968. Ancient forestry: an historical study. Part I: facts andinformation on trees. Acta Forestalia Fennica 82:1-84.Makkonen, 0. 1969. Ancient forestry: an historical study. Part II: theprocurement and trade of forest products. Acta Forestalia Fennica 95:1-46.Marks, G.C., and T.T. Kozlowski (eds.). 1973. Ectomycorrhizae: their Ecologyand Physiology. Academic Press, New York. 444pp.Martin, D.E., and L.E. Lowe. 1989. Characterization and classification of rootmat horizons in some coastal British Columbia Podzols. CanadianJournal of Soil Science 69: 17-23.McGill, R., J.W. Tukey, and W.A. Larsen. 1978. Variations of box plots. TheAmerican Statistician 32(1):12-16.Mehlich, A. 1972. Uniformity of expressing soil test results: a case forcalculating results on a volume basis. Communications in Soil Scienceand Plant Analysis 3(5): 417-424.155Mehlich, A. 1978. New extractant for soil test evaluation of phosphorus,magnesium, calcium, sodium, manganese, and zinc. Communications inSoil Science and Plant Analysis. 9:477-492.Melsted, S.W., and T.R. Peck. 1973. The principles of soil testing. In: SoilTesting and Plant Analysis, eds. L.M. Walsh and J.B. Beaton, p 13-22.Soil Science Society of America, Madison, Wisconsin.Mengel, K. 1985. Dynamics and availability of major nutrients in soils.Advances in Soil Science 2: 65-131.Meurisse, R.T. 1972. Site quality of western hemlock and chemicalcharacteristics of some western Oregon andic soils. Ph.D. dissertation,Oregon State University, Oregon, Washington, USA. 164pp.Meurrise, R.T. 1976. Some chemical and other properties of western hemlocksoils in Oregon -- their relationship to productivity. In: Western hemlockmanagement conference, eds. W.A. Atkinson and R.J. Zasoski, pp. 49-55.University of Washington, Washington, U.S.A. 317 pp.Minore, D. 1968. Effects of artificial flooding on seedling survival and growthof six northwestern tree species. USDA Forest Service, Pacific NorthwestForest Range and Experiment Station Research Note PNW 92. 12pp.Minore, D. 1972. Germination and early growth of coastal tree species onorganic seed beds. USDA Forest Service, Pacific Northwest Forest andRange Experiment Station Research Note PNW 160.Minore, D., and C.E. Smith. 1971. Occurrence and growth of fournorthwestern tree species over shallow water tables. USDA Forest Service,Pacific Northwest Forest and Range Experiment Station Research NotePNW 160.Mitchell, D.T., and D.J. Read. 1981. Utilization of inorganic and organicphosphates by the mycorrhizal endophytes of Vaccinium macrocarpon andRhododendron pontictun. Transaction of the British Mycological Society76: 255-260.Mitchell, K.J., and K.R. Polsson. 1988. Site index curves and tables for BritishColumbia: coastal species. Canada/B.C. Economic & RegionalDevelopment Agreement FRDA Report 037. 29 pp.Molina, R.J. 1980. Patterns of ectomycorrhizal host fungus specificity in thePacific Northwest. Ph.D. dissertation, Oregon State University, Corvallis,Ore.Molina, R., and J.M. Trappe. 1982. Patterns of ectomycorrhizal host specificityandpotential among Pacific Northwest conifers and fungi. Forest Science28(3): 423-458.Monserud, R.A. 1984a. Problems with site index: an opinionated review. In:Forest Land Classification: Experiences, Problems, Perspectives, ed. J.G.Bockheim, pp.167-180. University of Wisconsin, Madison, Wisconsin.276 pp.156Monserud, R.A. 1984b. Height growth and site index curves for InlandDouglas-fir based on stem analysis data and forest habitat type. ForestScience 30(4): 943-965.Monserud, R.A. 1988. Variations on a theme of site index. In: Forest growthmodelling and prediction, ed. A.R. Ek, S.R. Shifley, and T.E. Burk, pp.419-427. USDA Forest Service, North Central Experiment Station,General Technical Report NC-120. 579 pp.Monserud, RA., and G.E. Rehfeldt. 1990. Genetic and environmentalcomponents of variation of site index in inland Douglas-fir. Forest Science36(1): 1-9.Monserud, R.A., U. Moody, and D.W. Breuer. 1990. A soil-site study for inlandDouglas-fir. Canadian Journal of Forest Research 20:686-695.Morzuch, B.J., and G.A. Ruark. 1991. Principal components regression tomitigate the effects of multicollinearity. Forest Science 37(1): 191-199.Mueller-Dombois, D., and H. Ellenberg. 1974. Aims and Methods of VegetationEcology. John Wiley and Sons, New York. 546 pp.Naylor, D.V., and R. Overstreet. 1969. Sodium-calcium exchange behaviour inorganic soils. Soil Science Society of America Proceedings 33: 848-851.Neiman, K.E. Jr. 1988. Soil characteristics as an aid to identifying foresthabitat types in northern Idaho. USDA Forest Service, IntermountainResearch Station, Research Paper INT-390.Neter, J., W. Wasserman, and M.H. Kutner. 1990. Applied linear statisticalmodels: regression, analysis of variance and experimental designs.Richard D. Irwin Inc., Homewood, IL. 1181 pp.Nicholls, S. 1977. On the interpretation of principal components analysis inecological contexts. Vegetatio 34(3):191-197.Odum, E.P. 1969. The strategy of ecosystem development. Science 164: 262-270.Odum, E.P. 1971. Fundamentals of ecology. W.B. Saunders Co., Philadelphia,Pa. 574 pp.Olsen, J., W. Atkinson, and M. Rinehart. 1979. Response of western hemlockto nitrogen fertilization and thinning in the Pacific Northwest. In:Proceedings of the forest fertilization conference, ed. S.P. Geessel, R.M.Kenady and W.A. Atkinson, pp 69-77. College of Forest Resources,University of Washington, Seattle, Washington. 275 pp.Orkici, L. 1988. Community organization: recent advances in numericalmethods. Canadian Journal of Botany 66:2626-2633.Ott, L. 1988. An introduction to statistical methods and data analysis. PWS-Kent Publishing Co., Boston. 945 pp.157Owen, J.G., and M.A. Chimielewski. 1985. On canonical variate analysis andthe construction of confidence ellipses in systematic studies. SystematicZoology 34(3): 366-374.Packee, E.C. 1988. Correlation of forest growth and yield with soilcharacteristics. In: Proceedings of the Alaska forest productivityworkshop, ed. C.W. Slaughter and T. Gasbarro, pp. 37-42. USDA ForestService, Pacific Northwest Research Station, General Technical ReportPNW-GTR-219.Parker, H.A. 1942. Dominant height and average diameter as a measure of sitein untreated even-aged lodgepole pine stands. Canada Department ofMines and Resources, Lands, Parks and Forests Branch, Dominion ForestService, Silvicultural Research Note No. 72.Pearl, J. 1984. Heuristics: intelligent search strategies for computer problemsolving. Addison-Wesley Publishing Company, Menlo Park, California.382 pp.Pearson, K. 1901. On lines and planes of closes fit to systems of points inspace. Philosophical Magazine Series 6. 2(11):559-572.Peet, R.K. 1980. Ordination as a tool for analyzing complex data sets.Vegetatio 42:171-174.Pfister, R.D., B.L. Kovalchik, S.F. Arno, and R.C. Presby. 1977. Forest habitattypes of Montana. USDA Forest Service, Intermountain Forest Range andExperiment Station, General Technical Report INT-34.Phillipson, J.J., and M.P. Coutts. 1977. The influence of mineral nutrition onthe root development of trees. II. The effect of specific nutrient elements onthe growth of individual roots of Sitka spruce. Journal of ExperimentalBotany 26(105): 864-871.Pielou, E.C. 1977. Mathematical ecology. John Wiley and Sons, New York,New York. 386 pp.Pielou, E.C. 1984. The interpretation of ecological data: a primer onclassification and ordination. John Wiley and Sons. New York, New York.263 pp.Pluth, D.J., and I.G.W. Corns. 1983. Productivity of conifers in WesternCanada boreal forests in relation to selected environmental factors. In:IUFRO symposium on forest site and continuous productivity, ed. R.Ballard and S.P. Gessel, pp. 13-30. USDA Forest Service, PacificNorthwest Forest and Range Experiment Station, General TechnicalReport PNW-163.Polar, J., and K. fflinka. Coastal Western Hemlock Zone. In: Forestryhandbook for British Columbia, ed. S.B. Watts, pp. 268-272. Faculty ofForestry, University of British Columbia, Vancouver, B.C. 611 pp.158Pojar, J., and D. Meidinger, eds. 1991. Ecosystems of British Columbia. B.C.Ministry of Forests, Special Report Series No. 6. 330 pp.Pojar, J., K. Klinka, and D.V. Meidinger. 1987. Biogeoclimatic ecosystemclassification in British Columbia. Forest Ecology and Management 22:119-154.Poore, M.E.D. 1955. The use of phytosociological methods in ecologicalinvestigations. I. The Braun-Blanquet system. Journal of Ecology 43:226-244.Powers, R.F. 1980. 1Vfineralizable soil nitrogen as an index of nitrogenavailability to forest trees. Soil Science Society of America Journal 44:1314-1320.Powers, R.F. 1984. Sources of variation in mineralizable soil nitrogen. In:Nitrogen assessment workshop May 19-20, 1982, RFNRP Report No. 2,College of Forest Resources, University of Washington, Seattle, pp. 25-31.Quesnel, H.J., and L.M Lavkulich. 1980. Comparison of the chemicalproperties of forest floors, decaying wood, and fine roots in threeecosystems on Vancouver Island. Canadian Journal of Forest Research11: 215-217.Radwan, M.A., and D.S. DeBell. 1980a. Site index, growth, and foliar chemicalcomposition relationships in western hemlock. Forest Science 26(2):283-290.Radwan, MA., and D.S. DeBell. 1980b. Effects of different sources of fertilizernitrogen on growth and nutrition of western hemlock seedlings. USDAForest Service, Pacific Northwest Forest and Range Experiment StationResearch Paper PNW 267.Radwan, MA., and D.S. DeBell. 1989. Effects of different urea fertilizers onsoil and trees in a young thinned stand of western hemlock. Soil ScienceSociety of America Journal 53: 941-946.Radwan, M.A., and J.S. Shumway. 1983. Soil nitrogen, sulfur andphosphorusin relation to growth response of western hemlock to nitrogen fertilization.Forest Science 29(3): 469-477.Raison, R.J., M.J. Connel, and P.K Khanna. 1987. Methodology for studyingfluxes of soil mineral-N in situ. Soil Biology and Biochemistry 19(5): 521-530.Ralston, C.W. 1964. Evaluation of forest site productivity. InternationalReview of Forestry Research 1: 171-201.Read, D.J. 1983. The biology of mycorrhiza in the Ericales. Canadian Journalof Botany 61: 985-1004.Rehfeldt, G.E. 1989. Ecological adaptations in Douglas-fir: a synthesis. ForestEcology and Management 28: 203-215.159Rennie, P.J. 1963. Methods of assessing forest site capacity. In: Transactionsof Joint meetings of commissions IV & V, ed. G.J. Neale, pp. 770-785.International Society of Soil Science, New Zealand.Reukema, D.L. 1979. Fifty-year development of Douglas-fir stands planted atvarious spacings. USDA Forest Service, Pacific Northwest Forest andRange Experiment Station, Research Paper PNW-253. 21pp.Reukema, D.L., and J.H.G. Smith. 1987. Development over 25 years ofDouglas-fir, western hemlock, and western redcedar planted at variousspacings on a very good site in British Columbia. USDA Forest Service,Pacific Northwest Research Station, Research Paper PNW-RP-381. 46 pp.Richards, B.N. 1987. The microbiology of terrestrial ecosystems. LongmanScientific and Technical, Essex, England. 399pp.Roe, A.L. 1967. Productivity indicators in western larch forests. USDA ForestService, Intermountain Forest and Range Experiment Station, ResearchNote INT-59.Roth, F. 1916. Concerning site. Forestry Quarterly 15: 3-13.Roth F. 1918. Another word on site. Journal of Forestry 16: 749-753.Rowe, J.S. 1956. Uses of undergrowth plant species in forestry. Ecology 37(3):461-473.Runge, M., and M.W. Rode. 1991. Effects of soil acidity on plant associations.In: Soil acidity, ed. B. Ulrich and M.E. Sumner, pp. 183-202. Springer-Verlag, New York. 224 pp.Ryan, P.J., S.P. Gessel, and R.J. Zasoski. 1986. Acid tolerance of PacificNorthwest conifers in solution culture I. Effect of high aluminiumconcentration at constant pH. Plant and Soil 96:259.Rygiewicz, P.T., and C.S. Bledscoe. 1984. Mycorrhizal effects on potassiumfluxes by northwest coniferous seedlings. Plant Physiologist 76: 918-923.Rygiewicz, P.T., C.S. Bledscoe, and R.J. Zasoski. 1984. Effects ofectomycorrhizae and solution pH on [ 151‘11 ammonium uptake byconiferous seedlings. Canadian Journal of Forest Research 14(6): 885-892.Rygiewicz, P.T., C.S. Bledscoe, and R.J. Zasoski. 1984. Effects ofectomycorrhizae and solution pH on ( 15N1 nitrate uptake by coniferousseedlings. Canadian Journal of Forest Research 14(6): 893-899.Sammi, J.C. 1965. An appeal for a better index of site. Journal of Forestry 63:174-176.Santantonio, D., and R.K. Hermann. 1985. Standing crop, production, andturnover of fine roots on dry, moderate, and wet sites of mature Douglas-fir in western Oregon. Annales des Sciences forestieres 42(2): 113-142.160SAS Institute Inc. 1985. SAS User's Guide: Statistics, Version 5 Edition. Cary,NC. 956 pp.Scanlin, D.C., and H. Loewenstein. 1979. Response of inland Douglas-fir andgrand fir to thinning and nitrogen fertilization in norther Idaho. In: Forestfertilization conference, ed. S.P. Gessel, R.M. Kenady, and W.A. Atkinson,pp. 82-88. University of Washington, College of Forest ResourcesContribution No. 40. 275 pp.Seber, G.A.F. 1984. Multivariate observations. John Wiley and Sons, Inc.New York, New York. 686 pp.Shimwell, D.W. 1971. The description and classification of vegetation.University of Washington Press, Seattle, Washington. 322 pp.Shrivastava, M.B., and B. Ulrich. 1976. Heredity deals the cards butenvironment plays the hands -- a method of assessing forest siteproductivity. The Pakistan Journal of Forestry 26:147-162.Silvester, W.B., P. Sollins, T. Verhoeven, and S.P. Cline. 1982. Nitrogenfixation and acetylene reduction in decaying conifer boles: effects ofincubation time, aeration and moisture content. Canadian Journal ofForest Research 12: 646-652.Smith, C.A.B. 1947. Some examples of discrimination. Annals of Eugenics13:272-282.Smithers, L.A. 1956. Assessment of site productivity in dense lodgepole pinestands. Canada Department of Northern Affairs and Natural Resources,Forest Research Division, Technical Note No. 30.Society of American Foresters. 1923. Committee for classification of forestsites report. Journal of Forestry 21:139-147.Soil Survey Staff. 1975. Soil taxonomy: a basic system of soil classification formaking and interpreting soil surveys. USDA Soil Conservation Service,Agriculture Handbook No. 436.Sokal, R.R., and F.J. Rohlf. 1981. Biometry: the principles and practices ofstatistics in biological research. W.H. Freeman and Company, New York.859 pp.Spano, S.D., M.F. Jurgensen, M.J. Larsen, and A.E. Harvey. 1982. Nitrogen-fixing bacteria in Douglas-fir residue decayed by Fornitopsis pinicola.Plant and Soil 68: 117-123.Sparks, D.L 1985. Potassium dynamics in soils. Advances in Soil Science 6:1-63.Spittlehouse, D.L 1981. Measuring and modelling evapotranspiration fromDouglas-fir stands. Ph.D. dissertation, The University of BritishColumbia, Vancouver, B.C.161Spittlehouse, D.L., and T.A. Black. 1981. A growing season water balancemodel applied to two Douglas-fir stands. Water Resources Research 17:1651-1656.Spun, S.H., and B.V. Barnes. 1980. Forest ecology. John Wiley and Sons,New York, New York. 687pp.St. John, T.V. 1983. Response of tree roots to decomposing organic matter intwo lowland Amazonian rain forests. Canadian Journal of ForestResearch 13: 346-349.Steinbrermer, E.C. 1976. Soil-site relationships and comparative yields ofwestern hemlock and Douglas fir. In: Western hemlock managementconference, ed. W.A. Atkinson and R.J. Zasoski, pp. 228-235. Universityof Washington, Washington, U.S.A. 317 pp.Stevenson, F.J. 1986. Cycles of soil: carbon, nitrogen, phosphorus, sulfur,micronutrients. John Wiley & Sons, New York. 380pp.Stribley, D.P., and D.J. Read. 1980. The biology of mycorrhiza in the EricaceaeVII. The relationship between mycorrhizal infection and the capacity toutilize simple and complex nitrogen sources. New Phytologist 86: 365-371.Sukachev, V., and N. Dylis. 1964. Fundamentals of forest biogeocoenology.(translated from Russian by J.M MacLennon). Oliver and Boyd, London.673 pp.Suits, D.B. 1957. Use of dummy variables in regression equations. Journal ofthe American Statistical Association 52:548-551.Swan, H.S.D. 1960. The mineral nutrition of Canadian pulpwood species: I.The influence of nitrogen, phosphorus, potassium and magnesiumdeficiencies on the growth and development of white spruce, black spruce,Jack pine and western hemlock seedlings grown in a controlledenvironment. Pulp and Paper Research Institute of Canada, WoodlandsResearch Index no. 116. 66pp.Tabachnick, B.G., and L.S. Fidell. 1989. Using multivariate statistics. Harperand Row, Publishers, Inc., New York, NY. 746 pp.Tatum, C.O. 1971. Nature's law or pure empiricism? Site studies as a guide tobetter use of forest land. In: XV IUFRO Congress, pp. 128-135, Universityof Florida, Florida, USA.Tansley, A.G. 1935. The use and abuse of vegetational concepts and terms.Ecology 16: 284-307.Tatsuoka, M.M. 1988. Multivariate analysis: techniques for educational andpsychological research. Macmillan Publishing Company, New York, NewYork. 479pp.Taylor, R.F. 1929. Plant indicators in Southeastern Alaska. Journal ofForestry 30:746.162Taylor, R.F. 1935. Available nitrogen as a factor influencing the occurrence ofsitka spruce and western hemlock seedlings in the forests of southernAlaska. Ecology 16(4): 580-602.Tesch, S.D. 1981. The evolution of forest yield determination and siteclassification. Forest Ecology and Management 3:169-182.Tilman, D. 1986. Resources, competition and the dynamics of plantcommunities. In: Plant Ecology, ed. M.J. Crawley, pp. 51-75. BlackwellScientific Publications, London, England. 596 pp.Trewartha, G.T., and L.H. Horn. 1980. An introduction to climate. 5th ed.,McGraw-Hill Inc., New York. 416 pp.Trimble, G.R., Jr., and S. Weitzman. 1956. Site index studies of upland oaksin the northern Appalachians. Forest Science 2(3): 162-173.Tukey, J.W. 1977. Exploratory data analysis. Addison-Wesley Publishing Co.,Reading Massachusetts. 688 pp.Tukey, J.W. 1980. We need both exploratory and confirmatory. The AmericanStatistician 34(1): 23-25.Turner, D.P., and E.H. Franz. 1985. The influence of western hemlock andwestern redcedar on microbial numbers, nitrogen mineralization, andnitrification. Plant and Soil 88:259-267.Valentine, K.W.G., P.N. Sprout, T.E. Baker, and L.M Lavkulich, eds. 1981. Thesoil landscapes of British Columbia. Queen's Printer, Victoria, B.C. 197pp.Van den Driessche, R. 1971. Response of conifer seedlings to nitrate andammonium sources of nitrogen. Plant and Soil 34:421-439.Van den Driessche, R. 1976. Mineral nutrition of western hemlock. In:Western hemlock management conference, ed. W.A. Atkinson and R.J.Zasoski, pp 56-70. College of Forest Resources, University of Washington,Seattle, Washington. 317 pp.Van Groenewoud, H., and A.A. Ruitenberg. 1982. A productivity orientedforest site classification system for New Brunswick. Canadian ForestryService, Maritimes Forest Research Centre, Information Report M-X-136.9pp.Vitousek, P.M., and W.A. Refiners. 1975. Ecosystem succession and nutrientretention: a hypothesis. Bioscience 25(6):376-381.Vogt, K.A., E.E. Moore, D.J. Vogt, M.J. Redlin, and R.L. Edmonds. 1983.Conifer fine root and mycorrhizal root biomass within the forest floors ofDouglas-fir stands of different ages and site productivities. CanadianJournal of Forest Research 13:429-437.163Wall, M.K. 1988. Reflections on the life, work, and times of Vladimir JosephKrajina. Canadian Journal of Botany 66:2605-2619.Walker, N., and R.M. Reed. 1960. Site evaluation for western catalpa in northcentral Oklahoma. Oklahoma State University, Oklahoma AgriculturalExperiment Station, Bulletin B-544. 15pp.Waller, D.M. 1986. The dynamics of growth and form. In: Plant ecology, ed.M.J. Crawley, pp. 291-320. Blackwell Scientific Publications, London,England. 496 pp.Walmsley, M., G. Utzig, T. Vold, D. Moon, and J. van der Barneveld, eds. 1980.Describing ecosystems in the field. B.C. Ministry of the Environment,RAB Technical Paper 2, B.C. Ministry of Forests, Land ManagementReport No. 7, Victoria, B.C. 225 pp.Wang, G. 1991. Height growth of white spruce in relation to ecological sitequality. Ph.D. dissertation, University of British Columbia, Vancouver,B.C. (in progress).Ward, J.H. 1963. Hierarchical grouping to optimize an objective function.Journal of the American Statistical Association 58:236-244.Waring, S.A., and J.M Bremner. 1964. Ammonium production in soil underwaterlogged conditions as an index of nitrogen availability. Nature 201:951-952.Watson, R. 1917. Site determination, classification, and application. Journalof Forestry 15:552-563.Webber, B.D. 1974. Soil and foliar nutrient relationships for Douglas-fir onthree different sites. Canadian Forest Service Information Report PFRCNo. BC-X-100. 19 pp.Webster, S.R., D.S. DeBell, K.N. Wiley, and W.A. Atkinson. 1976. Fertilizationof western hemlock. In: Western hemlock management conference, ed.W.A. Atkinson and R.J. Zasoski, pp 247-252. College of Forest Resources,University of Washington, Seattle, Washington. 317 pp.Westhoff, V., and E. van der Maarel. 1980. The Braun-Blanquet approach. In:Classification of plant communities, ed. R.H. Whittaker, pp. 287-399. DrW. Junk by Publishers, The Hague. 408 pp.Whittaker, R.H. 1962. Classification of natural communities. BotanicalReview 28: 1-239.Whittaker, R.H. 1982. Approaches to classifying vegetation. In: Classificationof plant communities, ed. R.H. Whittaker, pp. 1-32. Dr. W. Junk byPublishers, The Hague, Netherlands. 408 pp.Wiley, K.N. 1978. Site index tables for western hemlock in the PacificNorthwest. Weyerhauser Co., Forest Resource Centre, Forestry Paper 17.164Wilkinson, L. 1990. SYSTAT: The system for statistics. SYSTAT, Inc.,Evanston, Illinois.Williams, B.K. 1983. Some observations on the use of discriminant analysis inecology. Ecology 64(5): 1283-1291.Wooldridge, D.D. 1961. Environmental factors related to growth andmanagement of western hemlock (Tsuga heterophylla (Raf.) Sarg.). Ph.D.dissertation, University of Washington, Seattle, Washington. 211pp.Wright, J.W. 1976. Introduction to forest genetics. Academic Press, New York.463 pp.

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0075221/manifest

Comment

Related Items