Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Ecological site quality, site index, and height growth of white spruce stands in the sub-boreal spruce… Wang, Gaofeng G. 1993

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

Item Metadata

Download

Media
831-ubc_1993_fall_phd_wang_gaofengo.pdf [ 13.37MB ]
Metadata
JSON: 831-1.0075206.json
JSON-LD: 831-1.0075206-ld.json
RDF/XML (Pretty): 831-1.0075206-rdf.xml
RDF/JSON: 831-1.0075206-rdf.json
Turtle: 831-1.0075206-turtle.txt
N-Triples: 831-1.0075206-rdf-ntriples.txt
Original Record: 831-1.0075206-source.json
Full Text
831-1.0075206-fulltext.txt
Citation
831-1.0075206.ris

Full Text

ECOLOGICAL SITE QUALITY, SITE INDEX, AND HEIGHT GROWTH OF WHITE SPRUCE STANDS IN THE SUB-BOREAL SPRUCE ZONE OF BRITISH COLUMBIA by GAOFENG G. WANG B.Sc., Nanjing Forestry University, 1983 M.Sc., Nanjing Forestry University, 1986 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Forest Science)  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA June 1993 ©Gaofeng.W,193  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission.  (Signature)  Department of  FeRS,CT --9 c/FA/CE-  The University of British Columbia Vancouver, Canada  Date  DE-6 (2/88)  %Pity 77  /9f  i  Abstract In order to quantify the relationship between ecological site quality and white spruce [Picea glauca (Moench) Voss] height growth, vegetation, soil, foliar nutrient, and stem analysis data were obtained from 102 white spruce-dominated stands across a wide range of sites throughout the Sub-Boreal Spruce (SBS) biogeoclimatic zone of central British Columbia. The data were analyzed using the principles and methods of biogeoclimatic ecosystem classification, statistical analysis, and growth modelling. Using vegetation classification, the study stands were organized into eight seral plant associations on the basis of differences in the floristic composition of understory vegetation. The ratio of actual/potential evapotranspiration and the depth to a gleyed layer and groundwater table were used to stratify the study stands into seven actual soil moisture regimes. Soil water saturation (implied by the seven soil moisture regimes), soil drainage, soil texture, and slope were used for qualitative characterization of three soil aeration regimes. Soil moisture and aeration regimes were then combined, and nine soil moisture-aeration regimes were recognized. Total soil mineralizable-N (kg ha -1 ) and C/N were used to characterize a regional soil nitrogen gradient and to stratify the study stands into five soil nutrient regimes. When understory vegetation, foliar nutrients, or white spruce site index were related to categorical or continuous measures of soil moisture, aeration, and nitrogen, the strength of relationships obtained indicated that these measures were meaningful estimates of ecological conditions experienced by plants. Using the recognized soil moisture, aeration, and nutrient regimes and zonal units as differentiae, the study stands were organized into seven site groups, 13 site associations, and 30 site series.  iii  Regression analysis showed that both continuous and categorical measures of ecological site quality were useful in predicting white spruce site index (0.66 5 R 2 0.94). The semi-empirical model, based on soil moisture-aeration and nutrient regimes and limiting factor analysis, was nearly as useful for predicting site index as the best developed regression model. Independent tests indicated that (1) integration of easily measurable soil variables into synoptic variables and (2) stratification of the study stands according to soil moisture conditions improved the strength of site index-ecological site quality relationships. Depending on the availability of data, the stratified soil model (mean prediction error = 1.44 m), the soil moisture-aeration regime plus soil nutrient regime model (mean prediction error = 0.85 m), and the semi-empirical model (mean prediction error = 0.96 m) were recommended for operational application. Z ratio (the height at 60 years of age divided by the height at 30 years of age) was used to quantify white spruce height growth pattern and to stratify the study stands into seven Z-groups. Developing height growth models for each Z-group improved prediction precision compared to traditional height growth models. In spite of the changing height growth pattern among stands, neither site index nor any measure of ecological site quality was found to be the major source of the variation in the study stands. Three ecologically based, site-specific height growth models (site series group-specific model, edaphic unit-specific model, and site groupspecific model) were developed. The edaphic unit- and site group-specific models, with mean error of estimate <1 m, were recommended for predicting white spruce dominant height in the SBS zone, without using site index as a predictor. By developing synoptic measures of ecological quality for sub-boreal sites, this study gave further evidence of usefulness of these measures in predicting white spruce site index or stand dominant height.  ^ iv  TABLE OF CONTENTS page Abstract  ^  ii  List of Tables  ^  ix  List of Figures  ^  xii  Acknowledgements ^  xv  1. GENERAL INTRODUCTION ^  1  2. STUDY AREA  7  ^  2.1.^THE SPECIES ^  7  2.2.^THE STUDY AREA ^  11  3. MATERIALS AND METHODS ^  15  3.1.^STAND SELECTION CRITERIA ^  15  3.2.^VEGETATION DESCRIPTION AND ANALYSIS ^ 15 3.3.^SOIL DESCRIPTION, SAMPLING, AND CHEMICAL ANALYSIS ^  17  3.4.^FOLIAR SAMPLING AND ANALYSIS ^  20  3.5.^STEM ANALYSIS ^  21  3.6.^STATISTICAL ANALYSIS ^  22  4. ECOLOGICAL CLASSIFICATIONS ^  24  4.1.^INTRODUCTION ^  24  4.2.^LITERATURE REVIEW ^  25  4.3.^METHODS ^  33  ^ v  4.3.1. Vegetation Classification ^  .33  4.3.2 Classification of Soil Moisture Regimes ^  34  4.3.2.1.^Et/Emax Ration Calculations ^ 4.3.2.2.^Classification Procedures ^ 4.3.2.3^Comparisons and Testing ^  36 38 39  4.3.3. Soil Aeration Regime Classification and Integrated Soil Moisture and Aeration Regimes ^ .39 4.3.4. Classification of Soil Nutrient Regimes ^  41  4.3.3.1.^Selection of Differentiating Characteristics ^ 4.3.3.2.^Classification Procedures ^ 4.3.3.3.^Comparisons and Testing ^  41 42 43  4.3.5. Site Classification ^ 4.4.^RESULTS AND DISCUSSIONS ^ 4.4.1. Vegetation Classification ^ 4.4.1.1.^Classification and Ordination ^ 4.4.1.2.^Indicator Plant Analysis ^  44 45 45 45 53  4.4.2. Characterization of Soil Moisture Regimes and Aeration Regimes ^ 57 4.4.2.1.^Delineation of Soil Moisture Regimes ^ 57 4.4.2.2.^Delineation of Soil Aeration Regimes and Soil MoistureAeration Regimes ^ 60 4.4.2.3.^Soil Moisture Regimes in Relation to Understory Vegetation ^ 62 4.4.2.4. Soil Moisture Regimes, Soil Aeration Regimes, and Soil Moisture-Aeration Regimes in Relation to White Spruce Foliar Nutrients and Site Index 67 4.4.2.5.^Discussion ^ 71 4.4.3. Characterization of Soil Nutrient Regimes ^72 4.4.3.1.^Delineation of Quantitative Soil Nutrient Regimes .72 4.4.3.2.^Testing Soil Nutrient Regimes ^ 78 4.4.3.3.^Relationships Between Quantitative and Heuristic Classifications ^ 88 4.4.3.4.^Discussion ^ 90  vi  4.4.4. Site Classification ^  94  4.4.4.1.^Delineation of Site Association and Site series . . . . 94 4.4.4.2.^Delineation of Site Groups ^ 96 4.5.^CONCLUSIONS ^ 5. WHITE SPRUCE SITE INDEX IN RELATION TO MEASURES OF ECOLOGICAL SITE QUALITY ^  102  104  5.1.^INTRODUCTION ^  104  5.2.^LITERATURE REVIEW ^  105  5.3^METHODS ^  113  5.3.1. Soil Capacity, Soil Nutrient, and Foliar Nutrient Indices ^ 113 5.3.2. Regression Analysis ^  118  5.3.3. Limiting Factor Analysis ^  119  5.4.^RESULTS AND DISCUSSION ^  122  5.4.1. Site Index in Relation to Soil, Understory Vegetation, and Foliar Nutrients ^ 122 5.4.2. Site Index in Relation to Synoptic Measures of Ecological Site Quality ^  134  5.4.3. Selecting and Testing Prediction Models ^ 136 5.4.4. A Semi-empirical Model Based on Limiting Factor Analysis . . . .152 5.4.5. Discussion ^  159  5.5.^CONCLUSIONS ^  162  6. HEIGHT GROWTH MODELS ^  164  6.1.^INTRODUCTION ^  164  6.2.^LITERATURE REVIEW ^  167  ^ vii  ^6.3.^METHODS ^ 6.3.1. Preliminary Site-curve Preparation ^  172 172  6.3.2. Selecting and Fitting Traditional Height Growth Models ^ 174 6.3.3. Height Growth Pattern and Growth Modelling ^ ..176 6.3.4. The Link between Height Growth Model and Ecological Site Classification ^ 6.4.^RESULTS AND DISCUSSION ^ 6.4.1. Development and Evaluation of Anamorphic and Polymorphic Height Growth Models ^  178 178 179  6.4.2. Characterization of Height Growth patterns ^ 188 6.4.3. Site-specific Height Growth Models ^  196  6.4.3.1.^Site-specific model with site index as predictor . . . .195 6.4.3.2.^Site-specific model without site index as predictor . 201 6.4.4. Comparison and Evaluation of Height Growth Models ^ 210 6.4.5. Discussion ^  213  6.5.^CONCLUSIONS ^  215  7. SUMMARY AND GENERAL CONCLUSIONS ^  217  8 LITERATURE CITED ^  219  Appendix 1.  List of scientific name of plant species identified in the study stands ^ .239  Appendix 2.  Coordinations between site associations recognized in this study and the B.C. Min. For. ^ 244  Appendix 3.  The edaphic grid showing site series distinguished for the study stands in each subzone or variant ^ 246  viii  Appendix 4^Residual analysis of the five traditional anamorphic and polymorphic height growth models developed in the study ^ 252 Appendix 5.^Height growth curves showing the difference between and similarity within each defined Z ratio group . . .257 Appendix 6.^A BASIC program used to calculate the dominant height of white spruce stands in the SBS zone ^ 260  ix LIST OF TABLES Table 2.1.  Means of selected climatic characteristics for the six biogeoclimatic units within the study area. ^ 14  Table 3.1.  Number of study stands, location, and the range in age and site index according to biogeoclimatic units. ^ 16  Table 4.1.  Coefficient a and b of the simple forest water balance model calibrated by different studies. ^ 30  Table 4.2.  The criteria used for characterization and classification of actual soil moisture regimes of study stands. ^ 35  Table 4.3.  The Coefficients used in the simple forest soil water balance models to calculate water deficit for the study stands. ^37  Table 4.4.  Coefficients for soil texture classes used in the simple forest water balance model to calculate water deficit for the study stands. ^37  Table 4.5.  A tentative key for identification of relative soil aeration regimes in the study area. ^ 40  Table 4.6.  Diagnostic combination of species for the plant alliances and associations distinguished in the study stands. ^ 46  Table 4.7.  Frequencies of indicator species groups (ISGs) of soil moisture and soil nitrogen for eight distinguished plant associations. ^55  Table 4.8.  Means and standard deviations (in parenthesis) of site index and inferred soil moisture regimes and soil nutrient regimes for the vegetation units distinguished in the study stands. ^ 56  Table 4.9.  Means and standard deviations (in parentheses) of the differentiating characteristics and the frequencies of ISGs stratified according to soil moisture regimes. 59  Table 4.10. Diagnostic combinations of species for the seven soil moisture regimes delineated in the study stands. ^ 63 Table 4.11. Models for the regression of the frequency of soil moisture indicator plants on ratio of actual and potential evapotranspiration (Et/Emax), depth to the gleyed layer or prominent mottles (GLEY), and depth to the groundwater table (GW). 67 Table 4.12. Means and standard deviations (in the parentheses) of white spruce site index and foliar macronutrients stratified according to soil moisture regimes. ^ 68 Table 4.13. Coefficients of discriminant functions and means and standard deviations (in parenthesis) of soil min-N and C/N for five groups delineated by cluster and discriminant analyses. 73  Table 4.14. Means and standard deviations (in parentheses) of other measured soil chemical properties stratified according to five delineated groups.. 76 Table 4.15. Diagnostic combination of species for the five soil nutrient regimes delineated in the study stands. ^ 79 Table 4.16. Means and standard deviations (in parentheses) of measured white spruce foliar nutrients summarized according to the five soil nutrient regimes delineated in the study stands. ^ 84 Table 4.17. Simple and multiple correlation coefficients describing the relationships between white spruce site index and foliar N, P, and S. ^ 85 Table 4.18. Means and standard deviations (in parentheses) of vegetation, soil, and stand properties selected for characterization of the final soil nutrient regimes identified in the study. ^ 89 Table 4.19. Means of selected soil nutrients stratified according to soil nutrient regimes distinguished in different studies. ^ 93 Table 4.20. Means and standard deviations (in parentheses) of selected soil and stand properties stratified according to the site associations distinguished in the study stands. ^ 97 Table 4.21. Means and standard deviations (in parentheses) of site index, and number study stands stratified according to soil moisture-aeration regimes and soil nutrient regimes. 101 Table 5.1. Simple correlation coefficients between white spruce site index and selected soil chemical and physical properties. ^ 123 Table 5.2. Selected models for the regression of white spruce site index on selected measures of soil and topography. ^ 124 Table 5.3. Selected models for the regression of white spruce site index on selected understory vegetation variables. ^ 128 Table 5.4. Selected models for the regression of white spruce site index on foliar nutrients and foliar nutrient index. ^ 129 Table 5.5. Selected models for the regression of white spruce site index on various combinations of soil/topography, understory vegetation, and foliar nutrient variables. ^ 132 Table 5.6. Selected models for the regression of white spruce site index based on the stratified data: (a) sites with groundwater table within 60 cm (n = 18), (b) sites with gleyed layer within 50 cm (n = 25), and (c) other sites (n = 59). ^ 133 Table 5.7. Selected models for the regression of white spruce site index on categorical measures of ecological site quality - soil nutrient regimes, soil moisture regimes, soil aeration regimes, soil moisture-aeration regimes, and site associations. ^ 135  xi  Table 5.8. Some statistics of the four candidate models selected for the prediction of white spruce site index from measures of ecological site quality. . 138 Table 5.9. The four candidate models recalibrated from the 68 randomly selected stands and selected statistics for these models. ^ 150 Table 5.10. Independent and non-independent tests of the four candidate models using 34 randomly selected stands as test data. ^ 151 Table 5.11. Stratification of biogeoclimatic units according to temperature and precipitation gradients. ^ 155 Table 5.12. Soil moisture-aeration limiting (Lma) and soil nutrient limiting (Ln) coefficients determined by limiting factor analysis for soil moistureaeration regimes and soil nutrient regimes. 158 Table 6.1. Coefficients and statistics of one anamorphic and four polymorphic height growth models fitted to the study stands. ^ 181 Table 6.2. Mean absolute error and bias (m) (upper and lower values, respectively) of height estimates obtained from the anamorphic and polymorphic height growth models fitted to the study stands according to site index classes. ^ 183 Table 6.3. Mean absolute error and bias (m) (upper and lower values, respectively) of height estimates obtained from the anamorphic and polymorphic height growth models fitted to the study stands according to age classes. ^ 184 Table 6.4. Height growth pattern-specific models: coefficients, coefficients of determination, standard error of estimate for equation [6.2] fitted to each of the six Z-groups. ^ 190 Table 6.5. Correlations coefficients between indices (Z ratio of each study stand and b coefficients of model [6.1] fitted to each study stand) of white spruce height growth pattern and measures of ecological site quality. ^ 194 Table 6.6. Site series group-specific models: coefficients, coefficients of determination, and standard errors of estimate for model [6.3] fitted to each of the four site series groups. ^ 199 Table 6.7. Site group-specific models: coefficients, coefficients of determination, and standard errors of model [6.1] fitted to each of the seven site groups characterized by a unique range of soil moisture-aeration regimes and soil nutrient regimes. ^ 205 Table 6.8. Test of the site group-specific height growth models specified in Table 6.7 using the data of this study and Wang et al. (1992). ^209 Table 6.9. Comparisons of the baseline, the height growth pattern-specific, and the three site-specific height growth models developed in the study.211  xii  LIST OF FIGURES Figure 2.1.  Approximate locations of sampling areas in the six biogeoclimatic units. ^  12  Figure 4.1.  PCA ordinations of 4 plant alliances (a) and 8 plant associations (b) showing 95% confidence ellipses for the means of alliances and associations superimposed on PCA ordinations. ^ 49  Figure 4.2.  95% confident ellipses for the means of distinguished plant associations superimposed on PCA ordinations showing the separation of plant associations within Spiraea (a), Streptopus (b), and Petasites (c) all.s. ^ 51  Figure 4.3.  95% confidence ellipses for the means of seven associations superimposed on RA ordination (based on all species from 100 plots). ^ 52  Figure 4.4.  RA ordination (based on all species from 100 plots) showing the patterns of soil moisture regimes (a) and soil nutrient regimes (b). . 54  Figure 4.5.  Means and standard deviations of measured site index by interpreted soil moisture regimes (a) and soil nutrient regimes (b). ^ 58  Figure 4.6.  Matrix showing relationships of soil moisture-aeration regimes to soil moisture regimes and soil aeration regimes. ^ 61  Figure 4.7.  Ordination of study stands along the first two PCA axes based on diagnostic species in Table 4.10, showing 95% confidence ellipses for the means of soil moisture regimes. ^ 66  Figure 4.8. Means and standard deviations of white spruce site index in relation to (a) soil aeration regimes and (b) soil moisture-aeration regimes. . . .69 Figure 4.9. Ordination of study stands on natural logarithm of min-N and C/N and 95% confidence ellipses of the means for the five delineated groups.74 Figure 4.10. Ordination of study stands along the first two axes of PCA based on all measured soil nutrient properties (except min-N and C/N) and 95% confidence ellipses of the means for the five delineated groups. . . . 77 Figure 4.11. Ordination of study stands along the first two axes of PCA based on diagnostic species, and 95% confidence ellipses of the means for soil nutrient regimes. .81  Figure 4.12. Relative frequencies (%) of nitrogen-poor (a) and nitrogen-rich (b) indicator plants stratified according to soil nutrient regimes. ^ 82 Figure 4.13. Box plot showing white spruce site index (m @ 50 yr b.h. age) stratified according to soil nutrient regimes. ^ 87 Figure 4.14. An environmental matrix showing the site associations distinguished in the study in relation to climate (biogeoclimatic subzones), relative and actual soil moisture regimes, and soil nutrient regimes. 95 Figure 4.15. Comparisons of white spruce site index among site series within the circumscribing site associations, showing the lack of effect of climate implied by biogeoclimatic subzone or variant: (a) Sheperdia, (b) Equisetum, (c) Oplopanax, and (d) Aulacomnium associations. . . . .98 Figure 4.16. An edaphic grid, defined by soil moisture-aeration regimes and soil nutrient regimes, showing the site groups distinguished in the study stands and white spruce site index. ^ 100 Figure 5.1. Relationships between selected soil and foliar nutrients in the study stands using Mitscherlich equations. ^ 115 Figure 5.2. Relationships between white spruce site index and selected foliar nutrients in the study stands. ^  117  Figure 5.3. Relationships between white spruce site index (m @ 50 yr b.h. age) and (a) soil nutrient index and (b) soil capacity index. ^ 127 Figure 5.4. Residual analysis for the selected soil/topographic model (equation [91): (a) residual versus estimated site index and (b) measured versus estimated site index. ^ 139 Figure 5.5. Residual analysis for the soil/topography plus ISG model (equation 23): (a) residual versus estimated site index and (b) measured versus estimated site index. ^ 140 Figure 5.6. White spruce site index in relation to soil C/N ratio and depth to groundwater table (equation (30a1) showing the regression surface, isolines, and distribution of measured site index ^ 141 Figure 5.7. White spruce site index in relation to soil total nitrogen and depth to gleyed layer or prominent mottles (equation [30b]) showing the regression surface, isolines, and distribution of measured site index. ^ 142  xiv  Figure 5.8. White spruce site index in relation to Et/Emax ratio and soil nutrient index (equation [300) showing the regression surface, isolines, and distribution of measured site index. ^143 Figure 5.9. Residual analysis for the stratified model (equation [30a]):(a) residual versus estimated site index and (b) measured versus estimated site index. ^ 144 Figure 5.10. Residual analysis of the stratified model (equation [30b]: (a) residual versus estimated site index and (b) measured versus estimated site index. ^ 145 Figure 5.11. Residual analysis of for the stratified model (equation [30d): (a) residual versus estimated site index and (b) measured versus estimated site index. ^  147  Figure 5.12. White spruce measured site index in relation to soil moisture-aeration regimes and soil nutrient regimes (equation [38]) showing the regression surface, isolines, and distribution of measured site index. ^ 148 Figure 5.13. Residual analysis for the combined SMAR and SNR model (equation [38]): (a) residual versus estimated site index and (b) measured versus estimated site index. ^ 149 Figure 5.14. A conceptual model delineating the growth limiting factors within the edaphic matrix for white spruce growth in the SBS zone. ^153 Figure 5.15. Categorical plot of measured white spruce site index in relation to (a) climate (represented by biogeoclimatic units), (b) soil moistureaeration regimes, and (c) soil nutrient regimes. ^ 154 Figure 5.16. Residual analysis for the limiting factor model (equation [43]): (a) residual versus estimated site index and (b) measured versus estimated site index. ^ 160 Figure 6.1. White spruce height growth curves produced from one anamorphic and four polymorphic models at site index 12 (bottom), 18 (middle), and 24 m (top). 180 Figure 6.2. White spruce height growth curves produced by Goudie and Mitchell's (1986) model based on the data from this study (solid line) and the original study (dotted line) at site index 12 (bottom), 18 (middle), and 185 24 m (top). ^  XV  Figure 6.3. White spruce height growth curves produced by Alemdag's model based on the data from this study (solid line) and the original study (dotted line) at site index 12 (bottom), 18 (middle), and 24 m (top). . 186 Figure 6.4. White spruce height growth curves produced by Ek's (1971) model based on the data from this study (solid line) and the original study (dotted line) at site index 12 (bottom), 18 (middle), and 24 m (top). . 187 Figure 6.5. White spruce height growth curves for each of the six Z-groups at site index 18 m obtained from by height growth pattern-specific models specified in Table 6.4. ^ 191 Figure 6.6. White spruce annual height increment curves for each of the six Zgroups at site index 18 m obtained from the differential forms of height growth pattern-specific models specified in Table 6.4. ^ 192 Figure 6.7. Box plots showing Z ratio stratified according to (a) soil nutrient regimes, (b) soil moisture regimes, (c) soil aeration regimes, and (d) soil moisture-aeration regimes. ^ 195 Figure 6.8. Box plots showing the Z ratio stratified according to (a) plant associations, (b) site associations, (c) biogeoclimatic units, (d) site ^ groups.  197  Figure 6.9. Box plot showing Z ratio stratified according to the 4 site series groups. ^ 198 Figure 6.10. White spruce height growth curves for each of the four site series groups at site index 18 m obtained from site series group-specific height growth models specified in Table 6.6. 200 Figure 6.11. White spruce height growth curves (equation [6.8]) for each of the five soil nutrient regimes under the optimum soil moisture and aeration 203 conditions. ^ Figure 6.12. White spruce height growth curves (equation [6.8]) for each of the nine soil moisture-aeration regimes under the optimum soil nutrient conditions. ^ 204 Figure 6.13. White spruce height growth curves for each of the seven site groups obtained from site group-specific height growth models specified in Table 6.7. 206 Figure 6.14. White spruce annual height increment curves for each of the seven site groups obtained from the differential forms of the site group-specific height growth models specified in Table 6.7. ^ 208 Figure A3.1. Site series distinguished in the SBSdw1 variant. ^ 246 Figure A3.2. Site series distinguished in the SBSdw3 variant. ^ 247 Figure A3.3. Site series distinguished in the SBSdk subzone. ^248  xvi Figure A3.4. Site series distinguished in the SBSmw subzone. ^ 249 Figure A3.5. Site series distinguished in the SBSmk subzone. ^ 250 Figure A3.6. Site series distinguished in the SBSwk subzone. ^ 251 Figure A4.1. Equation [6.2] (modified Richards's model). (a) - residual versus predicted height, (b) - measured height versus predicted height, (c) residual versus site index, (d) residual versus b.h. age. ^ 252 Figure A4.2. Equation [6.3] (Goudie and Mitchell's model). (a) - residual versus predicted height, (b) - measured height versus predicted height, (c) residual versus site index, (d) residual versus b.h. age. ^ 253 Figure A4.3. Equation [6.4] (Alemdag's model). (a) - residual versus predicted height, (b) - measured height versus predicted height, (c) - residual versus site index, (d) residual versus b.h. age. ^ 254 Figure A4.4. Equation [6.5] (Ek-Payandeh's model). (a) - residual versus predicted height, (b) - measured height versus predicted height, (c) - residual versus site index, (d) residual versus b.h. age. ^ 255 Figure A4.5. Equation [6.6] (logistic model). (a) - residual versus predicted height, (b) - measured height versus predicted height, (c) - residual versus site index, (d) residual versus b.h. age. ^ 256 Figure A5.1. Height growth curves from original stem analysis data stratified according to Z ration groups: (1) Z ratio group 7 and 8. ^ 257 Figure A5.2. Height growth curves from original stem analysis data stratified according to Z ration groups: (1) Z ratio group 9 and 10. ^ 258 Figure A5.3. Height growth curves from original stem analysis data stratified according to Z ration groups: (1) Z ratio group 11 and 12 ^ 259  xvi i  ACKNOWLEDGMENTS This dissertation represents the final chapter in a story that began ten years ago in China. Without the contributions of many people, it would not exist. Thanks first to my wife, Catherine, for all the encouragement and support she gave to me in this endeavor. Zhe(1) Ye(3) You(3) Ta(1) De(1) Yi(1) Fen(4)! Thanks to my supervisor, Dr. Karel Klinka, for his guidance, patience, understanding, and support through all stages of this research. Thanks to my supervisory committee members, Dr. Tim Ballard, Dr. Hamish Kimmins, Dr. Peter Marshall, and Dr. Mike Novak for their useful comments on earlier drafts of the dissertation. Thanks also to Dr. Andy Black for his valuable comments on the use of the water balance model. The thorough editorial work of Dr. Marshall was very much appreciated. Thanks to David New, Robert Slavik, Kevin Keys, and Daniel Klinka for the assistance in the field work. Thanks also to my friends and fellow forest ecology graduate students and research associates, Qingli Wang, for his help in identifying plant species, Gordon Kayahara, Reid Carter, and Donald McLennan for their input and advice, in general. Finally, a very special thanks to my parents. Their strong belief in education initiated my pursuit in knowledge. From them I received encouragement and strength to move forward in the face of adversity. Financial support for the research was provided by the British Columbia Ministry of Forests (Inventory Branch), Natural Sciences and Engineering Research Council of Canada, and Northwood Pulp and Timber Ltd. This support is gratefully acknowledged.  1  1. GENERAL INTRODUCTION Information on the productivity of a forest is essential to sustained yield management of that forest. If a forest consists of many different sites, then consideration must given to evaluation and classification of forest sites and relations between forest productivity and sites. From the beginning of the century, numerous studies have focused on evaluation of sites and estimation of forest productivity (e.g., Cajander 1926, Coile 1952, Rennie 1962, Ralston 1964, Jones 1969, Carmean 1975, Hagglund 1981). With the increasing demands for forest products, foresters have struggled to sustain and/or increase productivity from a decreasing land base. Accurate estimation of forest productivity for various sites and growth responses to stand/site treatments have become increasingly important as forest management has intensified. In spite of the great importance of white spruce (Picea glauca [Moench] Voss) as a timber species in British Columbia, very little is known about its growth and yield (Larocque and Marshall 1988), particularly in relation to ecological site quality. Conventionally, estimating forest productivity or site quality means taking measurements of selected stand and/or site attributes and using them as independent variables in functions, tables, and figures to describe their relationship to some indices of productivity (Hagglund 1981). All approaches to estimating forest productivity basically fall into two categories: (1) direct estimation, using stand variables, and (2) indirect estimation, using site variables. Direct estimation of forest productivity using site index has been widely applied in North America due to the fact that the height of dominant trees of a given species at a given age (most commonly at 50 years of b.h. age) is more closely related to the capacity of a given site to produce wood of that species than  2  any other measures (Spurr and Barnes 1980). Although there are some problems associated with the use of site index (Monserud 1984b, 1987), the site index method has been proven to be the best measure of forest productivity (i.e., the site production potential for a given species) (Monserud 1984a) and the most useful predictor of tree growth and yield in even-aged stands (Wykoff and Monserud 1987). To estimate site index directly for a given tree species and site, the relationship between site index and any pair of height and age, expressed as SI = f(H,A), has to be predetermined and presented as a group of height vs age curves or site index tables. Depending on the type of data and the procedure used for curve development, height vs age curves can either be anamorphic (assuming the same curve shape for any site index) or polymorphic (assuming different curve shapes for different site indices). The assumption that height growth for any site index follows the same trend has been widely criticized (e.g., Bull 1931, Stage 1963, Beck and Tronsdell 1973, Clutter et al. 1983). However, allowing curve shape to vary with site index may or may not solve the problem associated with the anamorphic curve system. Recent studies indicated that site index is a poor descriptor of height growth pattern for black spruce (Picea mariana [Mill.] B.S.P.), Douglas-fir (Pseudotsuga menzensii [Mirb.] Franco),  lodgepole pine (Pinus contorts Dougi. ex Loud), karri (Eucalyptus diversicolor F. Muell), and ponderosa pine (Pinus ponderosa Laws.) (Smith and Watts 1987, Milner 1987, Q. Wang 1992, Rayner 1991). In consequence, there is a potential to improve the widely used traditional polymorphic model, if important influences on the height growth pattern can be identified and used as predictors in height growth models. In view of the increasing evidence that height growth patterns vary between site types for which site index is similar (e.g., Carmean 1956 and 1972,  3  Zahner 1962, Newberry and Pienaar 1979, Pfister et al. 1979, Monserud 1984a), several studies have attempted to improve height and/or site index prediction by characterizing the site-specific variation of height growth patterns (e.g., Monserud 1984a, Milner 1987, Q. Wang 1992). As there has not been any such attempt for white spruce, the variation in its height growth pattern remains unknown, and whether its height and/or site index prediction can be improved by characterizing site-specific growth patterns remains uncertain. Indirect estimation of forest productivity by site attributes is very useful when site trees are not available in a stand or when our interest is in assessing the effects of environmental changes, due to either natural or artificial disturbances, on forest productivity. It not only provides an alternative method for predicting forest productivity, but also helps us to understand why the productivity varies. This understanding should help us to prescribe suitable management practices for sustaining productivity. The indirect estimation methods may be further divided into two categories: soil-site studies and ecological site classification. The purpose of soil-site studies is to develop prediction models based on the relationship between site index and environmental and/or biotic variables, which may be either categorical or continuous. Categorical variables that represent the units of ecological site classification could be considered as a special case of soil-site studies. By integrating the environmental and biotic factors, these units delineate ecologically-equivalent classes of sites (i.e., groups of sites with the same vegetation and productivity potential). Thus, site units should be suitable variables, not only for indirect estimation of site index, but also for inquiring into causes of its variation. Attempts have been made to estimate productivity for white spruce plantations in the Lake States using soil-site relationships, but the results are  4  far from conclusive (Harding 1982, Rauscher 1984). As soil-site relationships of natural stands have rarely been investigated, very few soil-site factors that are significantly correlated with white spruce productivity have been identified in natural stands (Pluth and Corns 1983). Thus, the problem of predicting white spruce productivity on a given site in the absence of suitable site trees has not yet been resolved. Over the last fifteen years, the system of biogeoclimatic ecosystem classification (BEC) has demonstrated its usefulness in providing a management and research framework. The classification and interpretation of 'sub-boreal' forest ecosystems was carried out by Krajina (1965, 1969), Wali and Krajina (1973), Annas and Coupe (1979), Meidinger and Pojar (1983, 1991), Pojar et al. (1984), Hope (1984), Delong et al. (1984, 1985), and Lewis et al. (1986). The usefulness of the system in determining relationships between site index and ecological site quality for several tree species has been demonstrated in several studies (e.g., Green et al. 1989, Klinka and Carter 1990, Kayahara 1992, Pearson 1992, Q. Wang 1992). However, the link between the height growth and ecological site quality for white spruce has not been established. The absence of this link was the impetus for this study. The purpose of the research carried out for this dissertation was to develop models for estimating white spruce site index and predicting average height of dominant trees at any age for naturally established, unmanaged, evenaged white spruce stands in the Sub-Boreal Spruce biogeoclimatic zone of central B.C. The research also addressed the following inquiries into the height growth of white spruce stands in relation to ecological site quality: (1)  How and why do height growth patterns vary among sites?  (2)  What is the usefulness of ecological variables in characterizing ecological site quality and predicting forest productivity (measured by site index)?  5  (3)^Can dominant stand height be reliably predicted from site-specific height growth models that use ecological parameters instead of site index as a predictor? To answer these questions, the research was organized to address the following nine specific objectives: 1.  to obtain qualitative and quantitative data for characterization of climate, soils, vegetation, stand nutrient status, and height growth of the study stands;  2.  to characterize soil moisture, nutrient, and aeration regimes of study stands;  3.  to classify study stands into vegetation and site units using the methods of biogeoclimatic ecosystem classification;  4.  to examine variation in site index in relation to different measures of ecological site quality;  5.  to develop empirical models for predicting site index using various measures of ecological site quality;  6.  to evaluate suitability of anamorphic and polymorphic height growth curves fitted to the data from stem analysis for the development of a baseline height growth model;  7.  to characterize height growth patterns and their relationship to the measures of ecological site quality;  8.  to develop pattern-specific height growth models according to the height growth patterns determined for the study stands, and site-specific height growth models according to site classification; and  9.^to evaluate the height growth models developed in the study. The results of this research are reported in three, relatively independent chapters. Chapter 4 deals with characterization of ecological site quality and  6  classification of white spruce study stands. Chapter 5 examines relationships between site index and various measures of ecological site quality. Chapter 6 presents the developed height growth models, quantifies the pattern of height growth, and suggests possible causes for the variation in height growth pattern.  7  2. STUDY AREA  2.1. THE SPECIES White spruce and Engelmann spruce (Picea engelmannii Parry) hybridize freely in the study area wherever their ranges overlap (Coates et al. 1992). Because identification of white spruce versus its hybrid is very difficult, the study stands actually included both white spruce and its hybrid. The term "white spruce", as used hereafter in the study, refers to both white spruce and its hybrid. White spruce is adapted to a wide range of edaphic and climatic conditions in the Northern Coniferous Forest (Nienstaedt and Zasada 1990). It has a transcontinental range, and grows from latitude 44 0 to 69 0 N and from sea level to 1520 m. By itself or with black spruce and tamarack [Larix laricina (Du Roi) K. Koch], white spruce forms the northern boundary of tree-form growth (Sutton 1969). As a major commercial tree species in northern and central British Columbia, white spruce is distributed mainly in the Spruce-Willow-Birch (SWB), Boreal White and Black Spruce (BWBS), and Sub-Boreal Spruce (SBS) zones, and also occurs in Interior Douglas-Fir (IDF), Interior Cedar-Hemlock (ICH), Sub-Boreal Pine-Spruce (SBPS), and Engelmann Spruce - Subalpine Fir (ESSF) zones of B.C. (Krajina 1969, Krajina et al. 1982). White spruce grows under highly diverse environmental conditions. Its climatic range may be described as cool temperate to subarctic with extreme low temperature below -56 0 C, mean daily July temperatures between 13 0 C and 21 0 C, extreme high temperature above 43 0 C, mean annual precipitation between 250 and 1270 mm, and growing season between 20 and 180 days (Sutton 1969,  8  Nienstaedt and Zasada 1990). With increasing climatic severity northward, the range of sites supporting the species becomes more limited (Sutton 1969). White spruce grows on a variety of soils developed from a wide variety of parent materials (Halliday 1937). Although Podzolic soils predominate over its range, white spruce also grows on Brunisolic, Luvisolic, Gleysolic, Regosolic, Cryosolic, and, less commonly, Organic soils (Sutton 1969, Rowe 1972, Nienstaedt and Zasada 1990). Texture of soils ranges from clays to sandy skeletal (Wilde et al. 1949, Nienstaedt 1957, Rowe 1959). Soil acidity ranges from pH 4.7 to 7.5 (Sutton 1969, Zasada et al. 1977, Brand and Janas 1988, Nienstaedt and Zasada 1990). Light intensity, temperature, and precipitation are the three most important factors which affect variation in the amount and rate of height growth of white spruce, although soil fertility, age, and intra-specific and intra-progeny variation may also be very important. Soil moisture, soil nutrients, and soil aeration have been recognized as the three most important factors which control white spruce productivity under the same regional climate. The effects of these factors on the growth of white spruce are interrelated. Sutton (1969) pointed out that white spruce can tolerate dry sites if they are fertile, and no fertile site is too moist unless the soil water is stagnant; however, Stiell (1958) indicated that white spruce is not suited to dry sites. Although white spruce can tolerate a wide range of moisture conditions, good growth occurs only on slightly dry, fresh, moist, and well-aerated, very moist sites according to the observation of this study, with the most productive growth occurring on moist alluvial sites (Kenety 1917, Rowe 1959, Sutton 1969). Kenety (1917) and Wilde et al. (1965) concluded that soil moisture is the best single predictor of white spruce productivity. Nienstaedt and Zasada (1990) suggested that the key to good growth of white spruce is a season-long  9  dependable supply of well-aerated water. Poor aeration has been found to be a growth-limiting factor (Cheyney 1942, Sutton 1969, Wilkins 1984, Kozlowski 1986, Rivard et al. 1990). Ahlgren and Hansen (1957) observed that white spruce height growth was reduced by flooding. As soil moisture increases up to some optimal level, availability of soil nutrients may become increasingly growth-limiting. Many studies indicated that nutrient deficiencies depressed the growth of white spruce more than that of black spruce, red spruce (Picea rubens Sarg.), Norway spruce (Picea abies (L.) Karst), and the pines (Heiberg and White 1951, MacLeod 1956, MacArthur 1957, Paine 1960, Swan 1960). Stone et al. (1962) estimated that the nutrient status of white spruce is higher than that of pine and equal to that of Norway spruce. Wilde (1966) found that minimum soil-fertility standards for white spruce are higher than those for jack pine (Pinus banksiana Lamb.), red pine (Pinus resinosa Mt.), and white pine (Pinus monticola Dougl. ex D. Don). To maintain  good growth, white spruce requires at least intermediate fertility (Sutton 1969). The best growth of white spruce is on nutrient-rich to very rich sites, with a sufficient supply of available calcium, magnesium, potassium, and a combination of nitrate and ammonium compounds (Krajina et al. 1982). However, Russell (1963), Wilde et al. (1965), and Payandeh (1986) found that soil nutrients are not good predictors of white spruce productivity (measured by site index). The shade tolerance of white spruce is intermediate or low, varying with climate and site conditions (Krajina 1969). It is more shade-tolerant than aspen (Populus tremuloides Michx.), paper birch (Betula papyrifera Marsh.), and  lodgepole pine, but less shade-tolerant than subalpine fir (Abies lasiocarpa [Hook.] Nutt.). On mesic sites, white spruce decreases its shade tolerance with decreasing temperature, from moderate in the southern IDF and SBPS zones through somewhat lower in the SBS the zone to the lowest in the BWBS zone.  10  Under the same regional climate, such as within the SBS zone, the tolerance of white spruce decreases from dry to moist sites (Krajina 1969, Krajina et al. 1982). Generally, white spruce seedlings can maintain its full height growth potential with anywhere between 45 to 100% sunlight up to nine years old (Logan 1969). Russell (1963) found that spruce planted under a nurse canopy of aspen or paper birch providing about 30% shade for the first 10 to 12 years were usually taller than those grown in the open. Beyond this age, overhead cover has a negative effect on height growth (Russell 1963, Stiell 1976). Reducing light intensity to 50% of full light reduced height growth by 25%, shoot growth by 50%, and root growth by 40% in 10-year-old seedlings, and no seedlings survived at 15% of full light (Eis 1970). The significant influences of soil moisture, light intensity, and temperature on bud morphogenesis, hence height growth potential, have been reported, but the effect of soil nutrients has not been examined (Pollard and Logan 1977). Severe deficiency in rainfall during the period of seedling height growth may retard or entirely curtail the potential growth predetermined by bud morphogenesis. However, as the tree grows older and its root system exploits an increasingly larger volume of soil, its ability to achieve the growth potential increases (Sutton 1969). Time of flushing of white spruce is controlled by climate, hence it varies with location. However, it is also influenced by microsites (Sutton 1969). Different trees may vary by two weeks or longer within a local population on apparently homogeneous sites (Rauscher 1984). White spruce has acquired the reputation of being a slow starting species. White spruce plantations usually take 6-18 years to pass the 'check' period in the Lake States (Rauscher 1984). In the SBS zone, naturally regenerated white spruce take 8-15 years to reach breast height under open growing conditions according to our measurements. White spruce naturally regenerated under a canopy may survive  11  a suppression period of up to 200 years (Nienstaedt and Zasada 1990). Significant response to release can be expected, although growth may be significantly reduced due to the suppression (Nienstaedt and Zasada 1990). Lateral crown competition curtails diameter growth without negatively affecting height growth (Stiell 1976), but height growth may be affected by root competition (Stiell 1976, Rauscher 1984). 2.2. THE STUDY AREA One hundred and two stands were located in the central and southern portions of the SBS biogeoclimatic zone, approximately from 52°30' to 54 0 18' N and from 122 0 0' to 125°54' W (Figure 2.1). This area is influenced by a continental montane boreal climate which is characterized by seasonal extremes of temperature, long snowy winters, and warm and moist summers. The study stands were distributed in five subzones, with two sets of stands being located in two variants of one subzone, making a total of six biogeoclimatic units (Meidinger and Pojar 1991): (1)  Dry Warm SBS (SBSdw) subzone, specifically in the Horsefly (SBSdw1) and Stuart (SBSdw3) variants;  (2)  Dry Cool SBS (SBSdk) subzone;  (3)  Moist Warm SBS (SBSmw) subzone;  (4)  Moist Cool SBS (SBSmk) subzone, specifically in the Mossvale (SBSmk1) variant; and  (5) Wet Cool SBS (SBSwk) subzone, specifically in Willow (SBSwk1) variant. Each subzone represents a segment of a combined precipitation gradient (ranging from relatively dry to moist to wet climates) and temperature gradient (ranging from relatively cool to warm climates) (Table 2.1).  130°  135°^  125°  115°  120°  Figure 2.1.^Approximate locations of sampling areas in the six biogeoclimatic units: Horsefly Dry Warm SBS variant (SBSdw1), (2) Stuart Dry Warm SBS variant (SBSdw3), (3) Dry Cool SBS subzone (SBSdk), (4) Moist Warn SBS subzone (SBSmw), (5) Moist Cool SBS subzone (SBSmk), and (6) Wet Cool SBS subzone (SBSwk).  0  —glaciers and icefields  60°  ALPINE TUNDRA  (1)  SPRUCE—WILLOW—BIRCH ^ BOREAL WHITE AND BLACK SPRUCE I^ J SUB-BOREAL PINE - SPRUCE SUB-BOREAL SPRUCE  ,re  MOUNTAIN HEMLOCK  r  ENGELMANN SPRUCE - SUBALPINE FIR MONTANE SPRUCE BUNCHGRASS  55°  PONDEROSA PINE INTERIOR DOUGLAS-FIR  55°  1 COASTAL DOUGLAS-FIR  on Entranc e  INTERIOR CEDAR— HEMLOCK COASTAL WESTERN HEMLOCK Queen  C")  Charlotte Islands O  as  -  v 1 15°  500  50°  Vancouver  Biogeoclimatic Zones of British Columbia 135°  130'  Research Branch, Ministry of Forests, 31 Bastion Square, Victoria, B.C. V8W 3E7  Island  —  49°  1 :7500 000 1 km 0^50^100^150  1254  120°  Prepared by Canadian Cartographics Ltd, 1 989 for the Province of British Columbia Ministry of Forests  13  Physiographically, the study area occurs within the interior plateau and is mantled with glacial drift which chiefly occurs in the form of drumlins and fluted till. The plateau is bounded by lava cliffs or steep rocky slopes adjacent to the entrenched rivers. Eskers, kames and meltwater channels are numerous; glaciolacustrine silt occurs in the Fraser valley (Valentine et al. 1978). Soils are primarily in the Luvisolic, Podzolic and Brunisolic orders. Imperfectly to poorly drained sites typically have Gleysols or gleyed subgroups of Luvisols, Podzols, and Brunisols, and occasionally Organic soils. Hybrid white spruce, white spruce, and subalpine fir are the dominant climax species in upland coniferous forests. Lodgepole pine appears to be a fire-climax species in the dry southwestern portion of the SBS zone. Alluvial forests of black cottonwood [Populus trichocarpa Torr. & Gray ex Hook.], often with a minor component of  spruce, occur on active floodplains. Wetlands are common and dot the landscape in poorly drained, postglacial depressions. Major wetland communities include sedges (Carex spp.), scrub birch (Betula glartclulosa Michx.), willows (Salix spp.), and hybrid white spruce or black spruce. More detailed descriptions of the study area were given by Meidinger and Pojar (1983, 1991).  Table 2.1.^Means of selected climatic characteristics for the six biogeoclimatic units within the study areal. Characteristics  SBSdw1  SBSdw3  SBSdk  SBSmw  SBSmk1  SBSwk1  Mean annual precipitation (mm)  615.6  524.7  474.6  664.4  658.5  879.7  Mean precipitation May-September (mm)  299.0  260.5  215.2  291.6  270.8  349.5  Mean precipitation of driest month (mm)  23.1  20.2  18.6  36.6  29.3  38.4  Mean precipitation of wettest month (mm)  97.1  55.7  56.5  83.0  76.9  100.0  Mean annual temperature (°C)  3.7  2.6  2.1  3.7  1.7  2.7  Mean temperature May-September ( 0C)  13.0  11.5  10.6  11.6  11.2  11.3  Mean temperature coldest month (°C)  -10.6  -12.6  -13.1  -10.0  -14.0  -12.7  Mean temperature warmest month (°C)  14.9  14.6  13.5  15.5  13.4  14.8  Frost free period (days)  79  83  70  NA  73  93  Accumulated degree days >5°C  1225  1061  925  1140  960  1084  'Sources are Meidinger (pers. comm.) and Anonymous (1982).  15  3. MATERIALS AND METHODS  3.1. STAND SELECTION CRITERIA Study stands were allocated into biogeoclimatic units according to the maps obtained from the Ecological Program Staff in the Cariboo, Prince George, and Prince Rupert forest regions. In each unit, study stands were selected to represent the widest possible range of soil moisture and nutrient conditions for white spruce growth. Only naturally regenerated, fully stocked, unmanaged, and even-aged white spruce-dominated stands without a history of damage were chosen for the study. In each stand a 20 x 20 m (0.04 ha) sample plot was located to represent an individual ecosystem relatively uniform in topography, soil, and vegetation characteristics. Two geographically separated sets of sample plots were selected in each biogeoclimatic unit, with each set being located along a field-identified soil moisture gradient and including at least six plots. A brief description of the 102 study plots is given in Table 3.1. 3.2. VEGETATION DESCRIPTION AND ANALYSIS In each plot, all plant species, except these growing as epiphytes, on decaying wood, or on coarse fragments, were identified. Their covers were estimated using the species significance scale (Mueller-Dombois and Ellenberg 1974) and coded into a file to be analyzed by a vegetation tabling (VTAB) program (Emanuel 1987). The nomenclature used in the study followed Taylor and MacBryde (1977) for vascular plants, Ireland et al. (1980) for mosses, Stotler and Crandall-Stotler (1977) for liverworts, and Hale and Culberson (1970) for  Table 3.1.^Number of study stands, location, and the range in age and site index (height @ 50 year b.h age) according to biogeoclimatic units. Symbols for biogeoclimatic units are given in Figure 2.1. Biogeoclimatic units  Number of plots  Longitude  Latitude  Elevation (m)  Age @ b.h. (years)  Site index (m)  SBSdw1  18  121°42' - 122°24'  52°30' - 52°30'  488 - 800  43 - 128  4.7 - 23.0  SBSdw3  13  122054' - 123°12'  53°48' - 53°54'  746 - 1006  56 - 111  11.8 - 22.6  SBSdk  13  125°30' - 125°48'  54012'  644 - 844  47 - 88  11.9 - 20.9  SBSmw  15  52°48'  788 - 1008  42 - 97  12.9 - 24.2  SBSmk  15  122°30' - 122°42'  53°48' - 54 0 12'  696 - 861  48 - 115  15.5 - 21.6  SBSwk  28  122°06' - 122°18'  53°36' - 54°18'  766 - 1026  32 - 88  7.5 - 24.1  Total  102  12000' - 122°54'  52°30' - 54°18'  488 - 1026  32 -128-  4.7 - 24.2  122°0'  rn  17  lichens. The scientific names of plant species identified in the study are listed in Appendix 1. Indicator species were stratified into one of the six ISGs (indicator species groups) for soil moisture and one of the three ISGs of soil nitrogen according to their indicator values assigned by Klinka et al. (1989a). The VTAB program was used to calculate the frequency of indicator plants for each ISG of soil moisture or nitrogen: m^m n  F. = ( E C•• / E E C•• ) x 100 J^i=1 1J^i=1j=1 13 where F.J is the frequency of jth ISG (j = 1, 2, ^ n; n = 6 for soil moisture and n = 3 for soil nitrogen) for soil moisture or soil nitrogen; C ij is the midpoint percent cover (%) for ith indicator species (i = 1, 2, ^ m) in jth ISG for soil moisture or nitrogen. The frequencies of the six soil moisture ISGs (MOIST1 - excessively dry to very dry, MOIST2 - very dry to moderately dry, MOIST3 - moderately dry to fresh, MOIST4 - fresh to very moist, MOIST5 - very moist to wet, and MOIST6 wet to very wet) and the three soil nitrogen ISGs (NITR1 - poor, NITR2 medium, and NITR3 - rich) were used as vegetation variables in statistical analyses.  3.3. SOIL DESCRIPTION, SAMPLING, AND CHEMICAL ANALYSIS  The topography and soil of each plot were described using a simplified version of the standard procedures employed by the Ecological Program Staff of the B.C. Forest Service (Walmsley et al. 1980). Relative soil moisture and soil nutrient regimes were estimated in the field using a combination of topographic, vegetation, and soil morphological properties and the methods described by  18  Klinka et al. (1984, 1989a). On each plot, elevation (m), aspect (degree), and slope (%), were measured by altimeter, compass, and clinometer, respectively. Slope position of each plot was described and recorded. Two subjectively located soil pits down to the root restricting layer were used to describe forest floor and mineral soil characteristics according to Klinka  et al. (1981) and Canada Soil Survey Committee (1978), respectively, and to determine the depth of the major (the depth with over 90% accumulated fine root distribution) and potential rooting zone (the depth to the root restricting layer). Coarse fragments >4 cm in diameter were isolated by sieving and weighed at the depths of 0-30 cm and >30 cm for each pit, and the dimension of each pit was recorded to determine the coarse fragment content. Five bulk density measurements were taken from five points which were located in the center and the four corners of each plot. The bulk density of the forest floor (FF), mineral soil (MS), or organic soil (OS) was measured by cutting a core (to the surface of the mineral soil, to the depth of 30 cm or the bottom of major rooting zone in the mineral soil, or to the depth of 30 cm or the bottom of the major rooting zone in the organic soil) and measuring its before drying volume and the mass after oven-drying at 105 0 C to a constant mass. Around each of the five points used for bulk density sampling, samples of forest floor and 0-30 cm mineral or organic soil layer were taken at the three points of an equilateral triangle (2 m on a side), and composited. If the actual rooting zone was >30 cm, another composite sample was taken from the four sides of a pit from the 30 cm depth to the end of the rooting zone. All soil samples for physical and chemical analyses were air-dried to constant mass; forest floor samples were ground in a Wiley mill to pass through a 2-mm sieve, while mineral soil samples were sieved through a 2-mm sieve to separate cm coarse fragments. The percent of the volume contributed by the coarse fragments  19  was determined and added to the content of coarse fragment determined in the field. The following soil analyses were done by Griffin Laboratories Corporation, Kelowna, B.C. Soil particle size distribution was determined using a hydrometer method (McKeague 1978). Soil texture was then determined according to the percentages of sand, silt, and clay of each soil sample (Canada Soil Survey Committee 1978). The pH in 0.01 M CaC1 2 was determined potentiometrically using a pH meter on a 1:1 soil-solution suspension for the mineral soil and a 1:5 soil-solution suspension for the forest floor (McKeague 1978). Total C was determined by loss-on-ignition at 500 0 C (McKeague 1978) using a LECO Carbon Analyzer (induction furnace). Total N was determined by a micro-Kjeldahl digest followed by colorimetric determination of [NH 4 ]+ using a Technicon Autoanalyzer (Anonymous 1976). Mineralizable-N was determined using an anaerobic incubation procedure modified from Waring and Bremner (1964), without deducting pre-incubation values, as suggested by Powers (1980). Released {NH4 ]+ was determined colorimetrically using a Technicon Autoanalyzer. Available phosphorus was determined colorimetrically using the ascorbic acid reductant method (John 1970) on a Bray P-1 (dilute acid ammonium fluoride) extract (Bray and Kurtz 1945). Available SO4-S was determined using an ICP (inductively coupled plasma) Spectrophotometer (Price 1978) on a 0.01 M CaC12 extract (Bardsley and Lancaster 1965). Extractable K, Mg, and Ca were determined using an ICP Spectrophotometer on Morgan's extract, i.e., sodium acetate at pH 4.8 (Greweling and Peech 1960). Analytical results were expressed as concentrations on a dry-mass basis and, where applicable, on a mass per unit area (kg ha -1 ) basis (including both forest floor and mineral soils). The calculation of kilograms per hectare in the mineral soil used depth of major rooting zone, bulk density, and coarse fragment  20  content. The calculation of kilograms per hectare in the forest floor used the depth and bulk density of the forest floor. The sum of any soil nutrient in the mineral soil and the forest floor represents kilograms per hectare within the entire major rooting zone (major rooting zone of mineral soil plus the depth of forest floor for mineral soils or major rooting zone in organic layer for organic soils). 3.4. FOLIAR SAMPLING AND ANALYSIS Foliar sampling and analyses followed the guidelines and procedures described by Ballard and Carter (1986). Foliar sampling was carried out in late September. On each plot, about 60 g of composite sample of current year foliage were collected from the 15 selected dominant and codominant white spruce trees. On each tree, a minimum of two branches from the uppermost 1/2 - 1/4 of the live crown were used to collect the foliar sample. On each branch, about 2 g of the current year needles were collected from at least three branchlets. Foliar samples were oven-dried at 70 0 C for 8 to 12 hours until the needles snapped cleanly in two when bent. From each composite sample, 100 needles were randomly selected to determine their dry weight. Dry foliage was ground for chemical analyses using a Braun type KSM-2 coffee grinder. After grinding, the sample was re-dried at 70°C and then stored in air-tight, screw-cap plastic bottles. After digestion by the method of Parkinson and Allen (1975), N and P were determined by colorimetric analysis using an autoanalyzer, and K, Ca, Mg, Fe, Mn, Zn, Al were determined by atomic absorption spectrophotometry. S was determined with a Fisher Sulfur Analyzer using the method described by Guthrie and Lowe (1984). Nitric acid and hydrogen peroxide digestion followed  21  by atomic absorption spectrophotometry was used to determine Cu. Dry asking followed by colorimetric analysis with the azomethine H method of Gaines and Mitchell (1979) was used to determine B. Active Fe was extracted by the method of Oserkowsky (1933), using 1 M HC1, and analyzed by atomic absorption. Foliar sulphate was determined using the method of Kelly and Lambert (1972). The results of chemical analysis were expressed both on a concentration basis (% for macronutrients and ppm for micronutrients) and on a total content basis (milligrams per 100 needles). 3.5. STEM ANALYSIS Diameter at b.h. (breast height) of each tree taller than 10 m was measured and the number of such trees was counted on each plot. The total heights and breast height ages of five dominant white spruce trees, which were among the 15 trees selected for foliar sampling, were measured using a clinometer and increment core, respectively. Three of the five dominant trees were felled for stem analysis. The total height and length of live crown were measured in the field. Stem disks were cut at 30, 60, and 130 cm above the ground surface and then were taken at 100 cm intervals between 130 cm and the top of each tree. On each disk, rings were counted. Site index was originally determined from Goudie and Mitchell's (1986) table using the average height and breast height age of the five dominant trees. This was later replaced by the actual measurement (i.e., the average height at 50 years from the three felled trees). Height/age data recorded from stem analyses can be biased if the height of the cross-cut is taken as the tree height for the given age, because of the presence of a 'hidden tip' above the cross-cut (Carmean 1972). Dyer and Bailey  22  (1987) compared six published algorithms for estimating the true height within a section and concluded that Carmean's (1972) method was the best. Therefore, the raw stem analysis data were adjusted using Carmean's (1972) algorithm to calculate tree height corresponding to the age at each cross-cut. For most plots with b.h. age > 50 years, actual site index was directly determined by calculating the average true height of the three dominant trees at an age of 50 years. Site indices of the remaining 20 plots with b.h. age < 50 years were estimated from the three parameter Richards's (1959) model fitted for each plot.  3.6. STATISTICAL ANALYSIS  MIDAS (Fox and Guire 1976) and VTAB (Emanuel 1987) were used for the vegetation analyses, including classification, ordination, tabular comparison, and indicator plant analysis. SYSTAT version 5.0 (Wilkinson 1990) was used for other statistical analyses. SYGRAPH version 5.0 (Wilkinson 1990) and Harvard Graphics version 3.0 (Software Publishing Corporation 1991) were used to produce the figures. Data transformations were made if they were required to improve the normality of the distribution of the data or the homogeneity of variance in order to meet the assumptions of the statistical analyses. Whenever necessary, multivariate normality, outliers, and the equality of the population covariance matrix were tested according to the methods suggested by Seber (1984), Morrison (1990), and Tatsuoka and Lohnes (1988), respectively, using programs written in BASIC. Univariate normality and equality of group variances were evaluated using probability plots and Bartelett's test (Sokal and Rohlf 1981), respectively, available through SYSTAT (Wilkinson 1990).  23  Analysis of variance (ANOVA) and Tukey's multiple comparison (Ott 1988) were used to detect differences in variables among groups. Simple and multiple linear correlations were used to examine the strength of relationships among variables. Linear and nonlinear regressions were used to develop prediction models. Unless specified, significance was set at a = 0.05. Ordinary least squares regression was used to estimate the parameters of all linear models. Adjusted R 2 was reported, instead of R 2 , for all the linear regression models, calculated as: Adjusted R 2 = 1 - (1 - R 2 ) x (N -1) / (N - K) where N is the number of cases and K is the number of predictors (including the constant) used in the model. The adjusted R 2 is the R 2 adjusted for the number of degrees of freedom used up in estimating the parameters (Greene 1990). As R 2 denotes the proportion of variance in the dependent variable accounted for by the predictor variable(s), adjusted R 2 reduces this proportion to a level expected when using this model in a new sample from the same population (Wilkinson 1990). Derivative-free Quasi-Newton methods (Wilkinson 1990, Greene 1990) were used to compute the least squares estimation of the parameters for all the nonlinear regression models. The R 2 reported for the nonlinear model was the corrected R 2 (Wilkinson 1990), calculated as: R2 = 1 - I e i2  /1  (yi - y)2  where y is the mean of the dependent variable, e i and y i are the residual and the measure of the dependent variable for ith observation, respectively. Although the R 2 of a nonlinear regression model is no longer guaranteed to be in the range of zero to one, it does provide a useful descriptive measure of the fit of the regression (Greene 1990).  24  4. ECOLOGICAL CLASSIFICATION  4.1. INTRODUCTION For decades, foresters have recognized the importance of classifying forest lands according to their vegetation and/or environmental attributes. Numerous studies have been carried out worldwide, and different classification systems have been developed. In British Columbia, the Ministry of Forests adopted the system of biogeoclimatic ecosystem classification (BEC) about fifteen years ago. Ecosystem studies carried out by Dr. V.J. Krajina and his students in British Columbia from 1950 to 1975 resulted not only in the development of ecosystem classification (Krajina 1969, Pojar et al. 1987, Meidinger and Pojar 1991), but also in understanding vegetation-environment relationships (e.g. Brooke et al. 1970, Wali and Krajina 1973, Kojima and Krajina 1975, Klinka and Krajina 1986). To assist foresters in understanding causes of variation in forest productivity, recent studies have focused on determining potential productivity of different tree species on different sites and on examining the usefulness of the BEC system for the study of site-productivity relationships. These studies showed that site index of coastal Douglas-fir (Green et al. 1989, Klinka and Carter 1990), lodgepole pine (Q. Wang 1992), Sitka spruce (Pearson 1992), and western hemlock (Kayahara 1992) can reliably be predicted by various measures of ecological site quality using the methods of the BEC system. The theme of the research described in this chapter was to classify the study stands using the principles and methods of the BEC system (Pojar et al. 1987), and to demonstrate the most important ecological relationships. The main objective of this work was to lay a foundation for examining white spruce site index in relation to measures of ecological site quality (Chapter 5) and for  25  developing height growth models within the framework of ecosystem classification (Chapter 6). Secondary objectives were to contribute to quantitative characterization of soil moisture regimes (SMRs) and soil nutrient regimes (SNRs), and to develop qualitative characterization of soil aeration regimes (SARs). These objectives were accomplished by analyzing and integrating vegetation and environmental data obtained from 102 study stands using phytosociological and numerical techniques. 4.2. LITERATURE REVIEW The characterization and categorization of the ecological site quality of forest lands has been the concern of foresters for centuries. As a result, various classification systems have been developed to provide a framework for management and scientific research (e.g., Burger 1972, Daniels et al. 1979, Spurr and Barnes 1980, Bailey 1981, Barnes 1986, Wang 1986a, Kimmins 1987, Burger and Pierpoint 1990). In spite of differences in approaches, methods, and systems, either a single or a combination of ecosystem attributes (climate, physiography, soil, and/or vegetation) have been used in classification, and systems that have attempted to integrate major ecosystem attributes have been recommended (Krajina 1972, Kimmins 1977, Barnes 1984, Wang 1986b). Recently, several attributes of forest ecosystem classification have been emphasized. Johnson (1985) advocated defining ecological units within which species' responses to silvicultural practices are similar and predictable. Barnes (1986) suggested that any classification should indicate the productivity of each unit. Crow and Rauscher (1984) suggested that growth modelling should be an integral part of classification.  26  The BEC system is a hierarchical scheme with three levels of integration: regional, local, and chronological. At the regional level, the vegetation and soils of climatic climax ecosystems are used to infer regional climate and to develop climatic or zonal classification producing a hierarchy of biogeoclimatic units.  At  the local level, ecosystems are classified according to their vegetation or soils into vegetation or site units, respectively. Vegetation classification follows the Braun-Blanquet approach and uses the VTAB program (Emanuel 1987). Site classification is based on the concepts of biological (Cajander 1926) or ecological equivalence (Bakuzis 1969), SMRs and SNRs. At the chronological level, ecosystems are organized, according to vegetation, into site-specific chronosequences. More detailed descriptions of the BEC system were given by Pojar et al. (1987) and Meidinger and Pojar (1991). Vegetation classification is the most important component classification of the BEC system. When classified, plant communities are arranged into classes, with the members of each class having in common one or more characteristics that set them apart from the members of other classes (Greig-Smith 1983). Because of the individualistic distributions of species and the continuity of communities, there is no single, natural unit of classification. Different ways of defining community-types imply different classifications of the same vegetation, each with some advantages and disadvantages depending on the aim of the study and the vegetation itself (Whittaker 1978). There has been an increasing tendency in the past three decades to use numerical methods in favour of traditional subjective approaches in vegetation classification (e.g., Pielou 1969, Goodall 1970, Whittaker 1978, Gauch 1982). A subjective approach groups stands on the basis of subjective assessment of similarity, or divides the whole set of stands into two or more groups on the basis of the presence of one or a few selected species deemed to produce a useful  27  classification. A numerical system of vegetation classification, which necessarily operates on a finite number of stands, uses an objective measure of similarity and allows data themselves to indicate the most satisfactory division criteria (Greig-Smith, 1983). The methods of vegetation classification used in the BEC system represent a compromise between traditional and numerical procedures. In addition to the Braun-Blanquet tabular method, the BEC system employs some numerical procedures, such as cluster analysis (CA), principle component analysis (PCA), and reciprocal averaging analysis (RA), to form floristically uniform groupings of stands. Since phytosociologists have neither precisely specified, nor agreed upon, the required composition for diagnostic combinations of species for particular categories, arbitrary criteria were proposed for vegetation classification in B.C. (Pojar et. al. 1987). Ecological site quality has been defined as the sum of all the environmental factors affecting the biotic community on a site (Daniels et al. 1979, Spurr and Barnes 1980). Different combinations of environmental factors may have similar effects on plants due to compensating effects (Odum 1971); consequently determining the interaction of many individual factors has been difficult. As routinely measured environmental factors are seldom causative, there have been attempts to integrate these environmental factors into a few synoptic ones (Klinka and Carter 1990) that directly affect plants. The most commonly used synoptic factors are climate, soil moisture, and soil nutrients (e.g., Cajander 1926, Pogrebnyak 1930, Hills 1952, Major 1963, Bakuzis 1969, Krajina 1969, Damman 1979). The principle underlying the integration is that plant growth can be expressed as a function of these synoptic factors, each of them being determined by the interactions of a set of individual environmental factors. These synoptic factors or effective properties (Stone 1984) (expressed as  28  categorical or continuous variables) have been adopted for characterization of ecological site quality and differentiation of site units in the BEC system. Sites with the same values for the synoptic factors are considered to be ecologically equivalent or to have the same vegetation and productivity potential (Pojar et al. 1987). This approach provides the most promising basis for evaluating ecological site quality and predicting forest productivity (Green et al. 1989, Klinka and Carter 1990). SMR represents a population of soils which provide a similar amount of soil water annually available for evapotranspiration by vascular plants over several years (Pogrebnyak 1930, Krajina 1969, Klinka et al. 1984, 1989a; Pojar et al. 1987). In theory, this amount can be accurately measured through  continuously monitoring the changes of soil water balance. However, SMRs have been inferred, traditionally, from a combination of qualitatively expressed and subjectively integrated topographic, soil, and vegetation attributes (Pogrebnyak 1930, Hills 1952, Loucks 1962, Bakuzis 1969, Krajina 1969, Klinka et al. 1984, 1989; Pojar et al. 1987). Krajina (1969) adopted nine SMR classes to describe the relative SMRs in different climates, with very xeric (0) representing the relative driest soil and hydric (8) representing the wettest soil in a given regional climate. Several workers have attempted to develop objective means for classifying SMRs (e.g., Waring and Major 1964, Soil Survey Staff 1975, Canada Soil Survey Committee 1978, Klinka et al. 1984, 1989a). Most recently, Brais and Camire (1992) proposed a key for SMR evaluation in northwestern Quebec using soil properties such as parent material, soil depth, mottles, slope position and shape, and humus thickness as differentiating characteristics. Their results agreed well with the expert classification of SMRs. Klinka et al. (1984) suggested using a water balance approach to assess the soil water supply of a site quantitatively.  29  They proposed an objective SMR classification using occurrence and duration of phases of water use, the ratio between actual and potential evapotranspiration, and the occurrence and depth of the water table as differentiae (Klinka et al. 1989a). Potential and actual evapotranspiration and water deficit during growing season were calculated using a slightly modified version (Carter and Klinka 1990) of a simple forest water balance model (Spittlehouse and Black 1981). While the model was applied successfully to characterize SMRs for Coastal Douglas-fir sites (Carter and Klinka 1990) and interior lodgepole pine sites (Q. Wang 1992), it failed to identify SMRs for Sitka spruce (Pearson 1992) and western hemlock (Kayahara 1992) stands. Several limitations associated with the model itself and its application may be responsible for the inconsistency. Firstly, the model does not consider runoff, upward flow, laterally flow (seepage), and the contribution of forest floor to the rooting zone water supply. Secondly, coefficients used in the above applications were calibrated for coastal Douglas-fir stands (Spittlehouse and Black 1981, Giles et al. 1984). Because most of these coefficients are subject to change according to climate, soil, and stand type (an example is given in Table 4.1), coefficient calibration and model validation are necessary when applying the model to other ecosystems (Spittlehouse 1985). Thirdly, monthly (instead of daily) precipitation, temperature, and solar radiation of 30 year normals were used in the model, which tends to underestimate the growing season water deficit (Black 1982). Given the fact that the same coefficients and monthly data were applied, the success of Carter and Klinka's (1990) study may be attributed to the application of the model to the same type of stands (Douglas-fir stands) under a  Table 4.1. Coefficient a and b of the simple forest water balance model calibrated by different studies. Study^Tree^Number al^b2^area^species^of plots  Spittlehouse &^0.8^10.0^Coastal^Douglas-fir^2 Black (1981)^(±0.07)^(±1.0)^B.C. Giles et al.^0.73^4.1^Coastal^Douglas-fir^7 (1984)^ (+0.07)^(+0.5)^B.C. Yarie et al.^1.0^4.28^Interior^White spruce^1 1 (1990)^ 1.0^2.69^Alaska^  'Priestly-Taylor coefficient, ratio of Emax (energy limited evapotraspiration rate) to Eeq (equilibrium evapotraspiration rate). 2 Ratio of Es (maximum rate of supply of water) to Oe (the fraction of extractable water in the rooting zone).  31  similar regional climate for which the model was calibrated. The failure of Kayahara's (1992) and Pearson's (1992) studies may be attributed to the application of the model to different stands (western hemlock and Sitka spruce stands) under a different regional climate. The success of Q. Wang's (1992) study may be attributed to the contrasting climate (Wet Cool SBS subzone vs. Very Dry and Cold SBPS subzone) among the study sites. A dummy variable equation, using only subzones as predictors, explained the variation of site index equally as well as equations using water deficit or potential evapotranspiration, calculated from the model, as predictors. To ensure optimum plant growth, the soil must be well aerated. Soil aeration can be directly measured by oxygen concentration and diffusion rate in soil pores (Letey 1985). The supply of oxygen in the soil has to be sufficient to meet not only the respiratory requirements of roots, but also those of soil organisms (Canneell 1977). Waterlogging, compaction, fine texture, and poor drainage may create anoxic or hypoxic conditions that adversely affect tree growth. Although white spruce can tolerate a wide range of moisture conditions, poor aeration results in reduced growth. Therefore, it would be useful to have some means for characterizing the soil aeration of the study stands. Although some studies have indicated that certain soil morphological properties are closely related to soil aeration (e.g., Jeglum 1974, Rivard  et al. 1990), no scheme  of soil aeration regime (SAR) classification has been found in the literature. SNR represents a population of soils that provide a similar amount of essential nutrients to vascular plants over a period of several years (Pogrebnyak 1930, Krajina 1969, Klinka  et al. 1984, 1989; Pojar et al. 1987). As soil fertility,  SNR is an inferred soil property (Soil Survey Staff 1951). Numerous workers have recognized the presence of a regional soil nutrient gradient (i.e., the gradual change of nutrient status among soils in an area) and, for convenience,  32  divided this gradient into several classes (SNRs) to describe the relationship between plant communities and soils (e.g., Pogrebnyak 1930, Bakuzis 1969, Krajina 1969, Pojar et al. 1987). Thus, interpretive classifications have been developed to reduce the multidimensional soil nutrient space into a onedimensional soil nutrient gradient, in which the delineated classes are referred to as trophotopes (e.g., Pogrebnyak 1930, Krajina 1969) or SNRs (e.g., Klinka et a/. 1984, Kabzems and Klinka 1987, Pojar et al. 1987; Courtin et al. 1988). Traditionally, SNRs have been inferred from a combination of qualitatively expressed and subjectively integrated topographic, forest floor, mineral soil, soil parent material, and vegetation attributes (Pogrebnyak 1930, Hills 1952, Locks 1962, Bakuzis 1969, Krajina 1969, Klinka et al. 1984). Despite their usefulness in describing vegetation-site relationships, the inferred SNRs have been criticized for the subjectivity of the integration (Ballard 1982, Kimmins 1987). In response to the criticism, Kabzems (1985), Kayahara (1992), and Pearson (1992) related the inferred SNRs to actual measurements of soil nutrients. Kabzems (1985) found that the inferred SNRs were most closely related to min-N (kg ha- 1 ) and extractable Mg (kg ha- 1 ) in the rooting zone. Kayahara (1992) and Pearson (1992) identified that the inferred SNRs were most closely related to total N (%), min-N (ppm), and C/N. Courtin et al. (1988) and Q. Wang (1992) attempted to develop a quantitative means for SNR classification using nutrient-related soil properties as differentiating characteristics. The differentiating characteristics proposed by Courtin et al. (1988) were pH and C/N of the forest floor and total soil N (kg ha- 1 ) and the sum of exchangeable Ca, Mg, and K (kg ha- 1 ) within the rooting zone; those proposed by Q. Wang (1992) used min-N (kg ha- 1 ) and the sum of exchangeable Ca, Mg, and K (kg ha- 1 ) within the rooting zone. All these studies implied that nitrogen-  33  related measures (i.e., total N, min-N, and C/N) were closely related to the soil nutrient gradient. Another criticism of SNR classification is the over simplicity of using a one-dimensional gradient to represent a multidimensional nutrient space (Ballard 1982). In theory, the multidimensional soil nutrient space is composed of all the growth-requiring soil nutrients. It is, indeed, very difficult to use a onedimensional SNR to represent the entire soil nutrient space. However, it is possible to approximate it using one or a few of the most limiting soil nutrient(s), especially when only one or a few soil nutrient(s) is/are limiting. Since soil nitrogen has been recognized as the most important limiting factor in the Pacific Northwest (Heilman 1979, Peterson and Gessel 1983, Ballard and Carter 1986) and British Columbia (Ballard 1983) and soil nitrogen-related measures have been identified as the important differentiating characteristics in previous studies (e.g., Kabzems 1985, Courtin et al. 1988, Kayahara 1992, Pearson 1992, Q. Wang 1992), SNR characterized by available soil nitrogen may often be a useful surrogate for the multidimensional soil nutrient space. 4.3. METHODS 4.3.1. Vegetation Classification As a first approximation, individual stands were combined into groups according to the pattern of the dendrogram produced by CA. The analysis was based on Euclidean distance calculated from all 230 species, and used average linkage as a clustering strategy. In the next step, the groups obtained from the CA were adjusted on the basis of comparisons with the distribution of some important indicator species using the species by plot matrix produced by RA. RA is an ordination technique related conceptually to weighted averages. However,  34  computationally, it is an eigen-analysis problem similar to PCA (Gauch 1982). The groups formed by CA and RA were then compared for affinities using VTAB (Emanuel 1987), and a two-level category system was proposed. The diagnostic combination of species (DCS) for each vegetation unit in the hierarchy was identified and one dominant species in each DCS was selected to name the delineated vegetation units—alliances and associations. Two ordination techniques, PCA (Tabachnick and Fidell 1989) and RA, were used to summarize community patterns among the study stands and vegetation units and to show their affinities in a two dimensional space. The 95% confidence ellipses of group means were superimposed on the ordination graphs to assist in visualizing the pattern and the position of each group mean. For each stand and plant association, frequencies of ISGs for soil moisture and soil nitrogen were calculated using the VTAB. The criteria of Klinka et al. (1989a) were used to infer the SMR and SNR for each stand and plant association. ANOVA and Tukey's multiple comparisons were used to test whether there were significant differences in site index among the vegetation units and the inferred SMRs and SNRs.  4.3.2. Classification of Soil Moisture Regimes The depths to groundwater table, gleyed layer, or prominent mottling were used as differentiae to characterize "water-surplus" SMRs (Table 4.2). Soil depth, coarse fragment content, texture, monthly mean temperature ( 0 C), precipitation (mm), and solar radiation (MJ m- 2 d- 1 ) were used to calculate potential (Emax) and actual (Et) evapotranspiration using the simple forest water balance model. The Et/Emax ratio was calculated as the index of soil water deficit and used to characterize "water-deficient" SMRs (Table  35  Table 4.2. The criteria used for characterization and classification of actual soil moisture regimes (SMRs) of study stands (modified from Klinka et al. 1984). Differentia  ^  SMRs1  Rooting-zone groundwater table present during the growing season (water supply exceeds demand) Groundwater table > 30 cm deep 2^VM ^ Groundwater table > 0 but 30 cm deep W ^ Groundwater table at or above the ground surface VW Soil is very poorly to imperfectly drained, rooting-zone fluctuating groundwater table occurs as indicated by gleying or prominent mottling ^ Gleyed layer > 20 cm deep ^ Gleyed layer S 20 cm deep  M VM  Soil is moderately well to excessively drained, rooting-zone groundwater table and gleyed layer absent Water deficit occurs as determined from the Spittlehouse and Black model 3 ^ Growing season Et/Emax < 50 % ^ Growing season Et/Emax < 70 but > 50 % ^ Growing season EtJEmax < 90 but > 70 % ^ Growing season Et/Emax > 90 % No water deficit occurs as determined from ^ the Spittlehouse and Black model  ED VD MD SD/F F/M  lED, VD, MD, SD, F, M, VM, W, and VW stand for excessively dry, very dry, moderately dry, slightly dry, fresh, moist, very moist, wet, and very wet, resepectively. 2 Depth is measured from the top of the forest floor. 3Actual transpiration (Et) and potential transpiration (Emax) are calculated using the Spittlehouse and Black (1981) water balance model.  36  4.2). Other soil characteristics (e.g., fine textured soil, weak mottling, gleyed layer below 50 cm, groundwater table below 60 cm, etc.) were used for finetuning the above classifications. 4.3.2.1.^Et/Emax Ratio Calculations The development of the simple forest water balance model was described in detail by Spittlehouse (1981), Spittlehouse and Black (1981), and Giles et al. (1984). The means of applying it for soil moisture regime classification were described by Carter and Klinka (1990). The growing season Et and Emax calculation begins in early May and ends in late September. The coefficients used in this study (Tables 4.3 and 4.4) were obtained from studies carried out for coastal Douglas-fir stands (Spittlehouse and Black 1981, Giles et al. 1984). Soil texture was used to estimate soil water content at field capacity  (Elmax)  and at  wilting point (0), water potential at air entry (w e ), water content at saturation (O s ), and an empirical coefficient (m). The depth of the rooting zone, corrected by coarse fragment content, was used as soil depth (mm). Monthly mean temperature ( 0 C) and precipitation (mm) of 30 year normals (Anonymous 1982) from the nearest atmosphere environment station (AES) within each subzone or variant were used for those stands located in that subzone or variant. Monthly mean solar radiation (MJ m" 2 d" 1 ) data from Prince George AES were used for all stands since it is the only AES available in the study area. Because the actual soil water balance was not measured in this study, the model output was empirically calibrated on a zonal site (Pojar et al. 1987) from each subzone or variant. The SMR of each zonal site was predetermined; the water deficit that a zonal site would experience within each subzone or variant was estimated from its SMR according to the criteria of Klinka et al. (1989a). Using the zonal site, the a values were estimated for each subzone or variant by  37  Table 4.3. Coefficients used in the ESL models to calculate water deficit for study stands. Parameter^  Values  ^  Priestley-Taylor constant (a) ^ Soil limited rate (b) ^ Evaporation of interception (g)  0.70 - 0.95 10.0 mm d -1 0.4  Rainfall interception 1^ Rainfall interception 2^  0.55 0.81  Rainfall increase multiplier^  1.0  ^ Yearly mean solar radiation ^ Amplitude  17.2 MJ m -1. d -1. 13.8 MJ m -2 d -1  Albedo^ Emissivity of vegetation^ EA correction^  0.88 0.96 0.92  Soil depth' Starting value of soil water content^  Omax  'Soil depth is the rooting depth corrected by coarse fragment content within the rooting zone, expressed as mm.  Table 4.4. Coefficients related to soil texture classes used in the ESL model to calculate water deficit for study stands'. Texture  Omax  Owp  xve  m  Os  Sand, loamy sand  0.18  0.05  0.003  4.25  0.40  Sandy loam  0.21  0.08  0.006  5.90  0.50  Loam, silt loam  0.30  0.11  0.012  5.30  0.45  Sandy clay loam Silty clay loam  0.37  0.19  0.015  7.80  0.48  Clay  0.40  0.25  0.015  11.0  0.48  'From SWBED1 -a program made by D. L. Spittlehouse for calculating soil water balances using the ESL model.  38  keeping the other parameters constant and changing the a value until the expected water deficit occurred. As a result, different a were subsequently used for different biogeoclimatic units (i.e., 0.95 for SBSwk, 0.8 for SBSdw1 and SBSmw, 0.75 for SBSdw3 and SBSmk, and 0.70 for SBSdk) to calculate Et and Emax. By dividing monthly precipitation into 5 and 6-day intervals, the model operates as follows. To the root zone water storage at the end of a given time interval, it adds the precipitation and subtracts the evapotraspiration and drainage for the subsequent time interval to give the storage at the end of this time interval; then it advance one time interval and repeat the process (Black 1982). The soil storage of water available to plants was calculated by subtracting water content at wilting point (Owp) from that at field capacity (Amax).  4.3.2.2.^Classification Procedure Firstly, study stands were stratified into three groups: (1) stands with a groundwater table within 60 cm from the ground surface, (2) stands with a gleyed layer or prominent mottling within 50 cm from the ground surface, and (3) stands with none of the above. If any stand fitted both groups (1) and (2), the stand was assigned to group (1). Secondly, the depth to the groundwater table, the depth to the gleyed layer or prominent mottling, and the ratio of Et/Emax, respectively, were used to differentiate SMRs of the study stands within each of the three groups according to the adopted criteria (Table 4.2). Thirdly, the SMRs of the study stands in group (3) were adjusted if the study stands had any evidence indicating other possible water input rather than precipitation (e.g., weak mottling in the soil, a gleyed layer below 50 cm, seepage  39  from the upper slope, groundwater table below 60 cm, etc.). In this case, the frequencies of the ISGs for soil moisture were considered. Finally, the resulting SMRs were compared to field-estimated SMRs that were converted from relative SMRs according to their occurrence in the biogeoclimatic units. The percentages of plots assigned into the same, adjacent (off one class), and different (off two or more classes) SMRs were reported.  4.3.2.3.^Comparison and Testing To examine the relationship between soil moisture and vegetation, a diagnostic table for the distinguished SMRs was developed using the diagnostic criteria proposed by Pojar et al. (1987) and the VTAB program (Emanuel 1987). Using diagnostic species as variables, PCA ordination and 95% confidence ellipses for the means were used to show the pattern of SMRs in a two dimensional space, presumably defined by soil moisture indicator plants. Spectral analysis (Klinka et al. 1989a) was used to determine frequencies of ISGs for soil moisture. The three major differentiating characteristics, Et/Emax ratio, depth to gleyed layer or prominent mottling, and depth to groundwater table, were related to the frequencies of ISGs for soil moisture through regression analysis. ANOVA and Tukey's multiple comparison were used to examine the differences of selected soil characteristics, foliar nutrients, and site index among the distinguished SMRs.  4.3.3. Soil Aeration Regime Classification and Integrated Soil Moisture-Aeration Regimes SARs were qualitatively characterized using depth to a groundwater table, depth to a gleyed layer or prominent mottling, slope, and soil texture as differentiating characteristics (Table 4.5). Since SMRs and SARs both  40  Table 4.5. A tentative key for identification of relative soil aeration regimes in the study area. Differentia  ^  Relative soil aeration regime "  Rooting-zone groundwater table absent during the growing season Mottles, gleyed layer, or very-fine-clayey 2 texture absent within rooting zone ^a Mottles, gleyed layer, or very-fine-clayey texture present within rooting zone Mottles, gleyed layer, or very-fine-clayey texture occurs below 30 cm 3^a Mottles, gleyed layer, or very-fine-clayey texture occurs within 30 cm Slope > 5 % or sandy, loamy, or sandy loamy soil texture^a ^ r Slope < 5 % Rooting-zone groundwater table present during the growing season Groundwater table > 30 cm deep ^ Slope > 5 % ^ Slope < 5 %  a r  Groundwater table > 0 but < 30 cm deep Slope > 5 %^ Slope < 5 %^ Groundwater table at the ground surface ^  "a, r, and d stand for adequate, restricted, and deficient aeration, respectively.  2 Soi1 has clayey fraction > 60%. 3 Depth is measured from the surface of the forest floor.  r d d  41  depended on the dynamics of soil water and aeration in the soil pores, integrated soil moisture-aeration regimes (SMARs) were derived by segregating the edatope defined by SMRs and SARs. Site index differences among SARs and SMARs were examined using ANOVA.  4.3.4. Classification of Soil Nutrient Regimes 4.3.4.1.^Selection of Differentiating Characteristics Total soil min-N (kg ha -1 ) and C/N [total C (kg ha -1 ) : total N (kg ha -1 )] within the major rooting zone (approximately in the range from 10 to 60 cm) were chosen to characterize the soil nutrient gradient of the study stands. This choice was based on two assumptions: (1)  available soil N is a primary measure of overall nutrient availability in the soil; and  (2)  min-N and C/N are reasonable indices of available soil N. Numerous studies have identified N to be the most important, growth-  limiting nutrient in the soil (e.g., Zinke 1960, Tamm 1964, Armson 1977, Carlyle 1986, Kimmins 1987, Pritchett and Fisher 1987). Severe N deficiency is common throughout British Columbia (Ballard 1983). Previous studies suggested that soil nitrogen might be the most promising differentiating characteristic of SNRs (Kabzems 1985, Courtin et al. 1988, Kayahara 1992, Pearson 1992, Q. Wang 1992). Foliar nutrient data collected in this study also suggested that N is the most limiting nutrient for the study stands, thus supported the choice of soil N as the differentiating characteristic. The usefulness of min-N as an index of available soil N has been demonstrated in many studies (e.g., Bremner 1965, Keeney and Bremner 1966, Keeney 1980, Powers 1980, Smith et al. 1981). However, a high C/N ratio may result in high microbial immobilization of N (Binkley and Vitousek 1989), which  42  is not accounted by min-N values. Under anaerobic conditions, which were used to determine min-N in the study, more ammonium may accumulate since microbial growth is depressed and immobilization is reduced, without necessarily reducing mineralization (Ponnamperuma 1972). Therefore, a combination of min-N and C/N may provide a better estimation of available soil N for plant growth than using min-N alone. 4.3.4.2.^Classification Procedures Nine study stands growing on organic soils with a high groundwater table (wet to very wet SMRs) were excluded from the analysis. The remaining 93 stands were classified, using multivariate analysis, into five groups as follows. In the first step, using ln(min-N) (the natural logarithm of min-N was used to improve normality) and C/N as variables, the study stands were segregated into five groups by KMEANS CA (Wilkinson 1990). If any group included only a few stands with extreme values relative to the majority of the population, these stands were deleted and the CA was repeated. The purpose of deleting the extreme stands was to prevent the overall data structure from being driven by a few extreme stands. In the second step, the same two variables were used in discriminant analysis (DA) (Dillon and Goldstein 1984) to develop classification functions for each group. DA was repeated until no misclassification occurred. In the third step, the classification functions developed were used to assign the sample plots deleted during the CA to one of the five delineated groups. Finally, DA was used to obtain new classification functions for the whole population. By examining the differences in soil nutrients and nutrient-related properties among the groups, each group was assigned into one of the five SNRs: very poor (VP), poor (P), medium (M), rich (R), and very rich (VR).  43  The quantitatively determined SNRs were compared with the fieldestimated SNRs. The percentages of the study stands assigned into the same, adjacent (one class difference), and different (two or more class difference) SNRs by the quantitative and field procedures were reported. Those stands that were assigned into different SNRs were re-examined. Adjustments were made if the stands had substantially lower or higher frequency of soil N indicator plants compared to the criteria of Klinka et al. (1989a) and/or if field observations indicated other possible nutrient input (e.g., nutrient input through seepage) that could not be accounted for by the soil sampling and analysis used in the study.  4.3.4.3.^Comparisons and Testing Using all of the measured soil nutrients/properties (excluding min-N and C/N) as independent variables, the first two axes of the PCA were used to ordinate the study stands to show the association of the other soil nutrients with the quantitative SNRs determined by min-N and C/N. To examine the relationship between the SNRs and understory vegetation, a diagnostic table based on stratification of study stands according to SNRs was developed using diagnostic criteria proposed by Pojar et al. (1987) and the VTAB program (Emanuel 1987). The first two PCA axes from the diagnostic species were used to show the quantitative relationships between vegetation and the SNRs. Computer-assisted (Emanuel 1987) spectral analysis (Mueller-Dombois and Ellenberg 1974; Klinka et al. 1989) was used to determine frequencies of nitrogen indicator plants. Multiple linear regression analysis (Chatterjee and Price 1977) was used to describe the relationships between frequencies of nitrogen poor and rich indicator plants and soil nutrient properties. The relationships between foliar nutrients and soil nutrient properties and between  44  site index and foliar nutrients were examined by correlation analysis (Ott 1988). If homogeneity of variances was indicated by Bartlett's test (Sokal and Rohlf 1981), ANOVA followed by Tukey's multiple comparison was used to examine the differences in soil nutrient properties, frequency of nitrogen indicator plants, foliar nutrients, and site index among the delineated SNRs.  4.3.5. Site Classification Site classification was based on the climate (implied by biogeoclimatic subzone or variant), SMR, and SNR determined for each study stand. Stratification according to these attributes showed that the study stands occurred in many unique associations. To differentiate site associations, it was necessary to consider the vegetation potential of the study stands. This was done by examining the relations of each stratification to the recognized vegetation units in this study and to site associations recognized by the Ecological Program Staff of the B.C. Ministry of Forests (D. Meidinger, pers. comm.). A site association was only recognized when it could be segregated from all other associations by an exclusive range of climate, soil moisture, and soil nutrient regimes and a unique combination of understory vegetation. Site associations identified were named after the dominant understory species. If a site association occurred in different subzones or variants, the site association was sub-divided into different site series. The study stands were stratified into site groups according to SMARs and SNRs. Edaphic units, each with a unique combination of SMARs and SNRs, were related to white spruce site index. Those edaphic units that were not significantly different in site index and ecologically adjacent in the gradation represented by SMARs and SNRs were defined as a site group.  45  Site index and other selected characteristics of the study stands were stratified according to site associations and site groups. Edaphic grids were used to show the location of site associations and groups in the SBS zone and the location of site series in each subzone.  4.4. RESULTS AND DISCUSSION  4.4.1. Vegetation Classification 4.4.1.1.^Classification and Ordination The study stands were classified into a hierarchy of four plant alliances (all.^) and eight associations (a.^) (Table 4.6). These vegetation units represented the mid-seral stages of white spruce-dominated forest communities. For simplicity, 'Picea' was omitted from the name and each vegetation unit was identified by the generic name of one species selected from the diagnostic combination of species for that unit. The diagnostic table indicated that the extent of floristic differentiation was strong and implied that there were eight different ecological strata within the population of the study stands. Using only the diagnostic species (17 species for the plant alliances, and 61 species for the plant associations) as variables (Table 4.6), the study stands were ordinated along the first two PCA axes to show the patterns of plant alliances and associations (Figure 4.1). The 95% confidence ellipses for the means of three alliances and seven associations were superimposed on the ordination to assist in visualizing the patterns. The confidence ellipses for the Betula alliance and association were not included in the graph, as these units were only represented by two stands.  ^  Table 4.6. Diagnostic combination of species for the plant alliances (all.^) and associations (a.^) distinguished in the study stands. Colum^identification Number of plots Vegetation units and species  Diagnostic value'  SHEP 18  PETA^ATHY^EQUI ARAL^STRE^DISP 37^8^9 10^15^7 Presence class'^and mean species significance'  BETU 2  Spiraea p.^all. Amelanchier^alnifolia Festuca occidentalis Peltigera aphthosa Spiraea betulifolia  (d,c) (d) (d) (d,c)  V  2 1 1 4  II  2  IV  3 3 2 5  I  3 2  III V  IV  V  3 1 3 4  IV  III III V  II  II II  1 1  III  1  IV  3  II  2  IV  I  +  +^II  2  IV  I  II  2  III  1 3  II  +  I  +  +  1  III  1  I  +  I  +  V  5  II V I  1 4 +  IV  3 + +  Shephedia (SHEP) p.^a. Cladina mitis Dicranum scoparium Epilobium angustifolium Shepherdia canadensis  (d) (d) (d) (d,cd)  IV  III  V  I  1^II +  +  4 4 3 2 2  II III III III III  1 3 1 4 3  III 4 III 2 I+  V V  5 4 5 3 4 2  I I  I II  +  III II  1 1 +  II  1  IV  4 1 3 4 5 2  IV III I III V  IV  3 3 + 3 4 2  III I  I III I  + 1 +  Aralia^(ARAL)^p.^a. Alnus^sinuata Aralia^nudicaulis Arnica cordifolia Rhytidiadelphus^triquetrus Rubus pubescens Smilacina racemosa  (d)  (d,c) (d)  (d)  (d)  (d)  III II II II  1  1 1  II  +  IV  4 2  IV IV IV IV  1  IV  1  I 3 II + I + 1 III I+ V 2  IV  III V  Streptopus^p.^all. Abies^lasiocarpa Clintonia^uniflora Gymnocarpium dryopteris Lycopodium annotinum Streptopus roseus Tiarella^trifoliata  (d,c)  (d,c) (d,cd) (d) (d,c) (d)  II  I+ 1 II  V III  V IV  II  1  IV  3 2  III III  2 4  III  I+  V V  4 3 4  III III II  3 2 2  II II V  3 + 5  V III  5 2  II I  1 2  II II  1 +  +  IV  3 2 + 5 3  I + I + II 3 V 4 I+  IV II  +  I  1  II  1  II II  II  1  I+  Streptopus^(STRE)^P.^a. Arnica cordifolia Geocaulon^lividum Rubus pedatus Rubus pubescens Spiraea betulifolia  (d) (d)  II III  1 3  (d)  V  4  (d) (d)  II II  1 1  III  IV V  IV IV  1 1 4 3 3  I  II 2 I+ II 2  III I V  IV  4=.  I II  V  + 3  4  al  Table 4.6. Continued 1. Disporum (DISP) p. a. Actaea rubra Aruncus dioicus Asarum caudatum Brachythecium asperimum Calamagrostis canadensis Disporum trachycarpum Dryopteris expansa Galium trifidum Oplopanax horridus Pseudotsuga menziesli Ribes lacustre Rubus parviflorus Thalictrum occidentale Tiarella unifoliata Vaccinium ovatum Veratrum viride Viola glabella  (d) (d) (d) (d) (d) (d,cd) (d,c) (d,c) (d,c) (d,c) (c) (d,c) (d,c) (d,cd) (d,cd) (d) (d,c)  II I I 1 I+ I+ II 2 I+ III 3 I 1 I+ 1 I  I+ I+ III 4 II + IV 4 II 1 I+  I+  1 +  I+  I+  I+ II 1 I+ II 2 IV 3 III 3 1 I III 2 I+ II 1 I+  III  2  IV  2  IV V V V V V V V V V V IV V  1 3 3 3 2 5 2 2 3 4 3 4 3 5 5 3 2  II I  2 1  III II  4 2  I  3  III  I II I  + + +  I  +  IV IV IV  IV  I+ III 1 I+ II 2 III 2 III 3 III 2 I+ I+ I+  I+ IV 2 II 3 II 2 V 3 1 II II 1 1 II I  2  III IV  3 5  III  1  III  1  I  +  Petasitis p. all. Alnus sinuata Hylocomium splendens Mitella nuda Petasites palmatus Rubus pubescens  + 1  III II II I IV  4 1 + + 2  II II II II III  1 3 1 1 3  1 1 2 1 3 + 4  IV III II II III III V  3 2 3 1 2 1 4  III I I I III III IV  1 + + + 1 1 3  I+ II + II +  II I V IV II  + + 5 3 1  (d)  I II  3 2  (d,c) (d,c)  I II  II II II II III II V  (d) (d,c)  IV V V V  4 4 3 3 5  IV IV V IV V  3 3 3 2 4  V IV IV V V  5 4 3 3 4  + 2 +  II  1  1 2  3 3 2 2 2 2 3  I II I  III II  IV IV III IV III IV IV  I II I  + + +  II I II  + + +  II II V V III  1 I +^I 4^II 3^III 2^II  + 1 2 2 +  V V V V V  5 4 5 3 2  I III II III  + 3 + 1  3  2  Petasites (PETA) p. a. Arnica cordifolia Aster ciliolatus Aster conspicuus Fragaria virginiana Geocaulon lividum Osmorhiza chilensis Spiraea betulifolia  (d) (d) (d) (d) (d) (d) (d)  Athyrium (ATHY) p. a. Athyrium filix-femina Equisetum sylvaticum Gymnocarpium dryopteris Ribes I acus t re Streptopus amplexifolius  (d,cd) (d,c) (d,cd) (d,c) (d,c)  I  iA  Table 4.6. Continued 2. Equisetum^(EQUI)^p.^a. Aulacomnium palustre Calamagrostis canadensis Equisetum palustre Sphagnum capillaceum Betula^(BETU)^p.^all^and p. Betula glandulosa Carex aquatilis  (d)  (d) (d,cd) (d)  I 1 I + I+  IV III  2 1  I I I I  1 1 2 +  II II III I  IV IV V IV  4 5 7 5  3  5  3  5  5 5  7 8  a. (d,cd) (d,cd)  'Species diagnostic values: d - differential, dd - dominant differential, cd - constant dominant, c - constant, is - important companion (Pojar et al. 1987). ,  1 2 3 1  Presence classes as percent of frequency: I^1-20. II = 21-40, III = 41-60, IV = 61-80, V = 81-100. If 5 plots or less, presence class is arabic value (1-5).  'Species significance class midpoint percent cover and range: + = 0.2 (0.1 - 0.3), 1 = 0.7 (0.4 - 1.0), 2 = 1.6 (1.1 - 2.1), 3 = 3.6 (2.2 - 5.0), 4 = 7.5 (5.1 - 10.0), 5 = 15.0 (10.1 - 20.0), 6 = 26.5 (20.1 - 33.0), 7 = 41.5 (33.1 - 50.0), 8 = 60.0 (50.1 70.0), 9 85.0 (70.1 - 100).  49  (a)  15  10  A = Spiraea  5  B = Streptopus C = Petasites  0  D = Betula -5  -10 ^ ^ ^ ^ ^ -10 -5 0 5 10 15 PCA axis 1  (b  )  20 15  A = ARAL  10  B = ATHY C = BET'U  5  D = DISP E = EQUI  0  F = PE'TA -5  G = SHEP H = STRE  - 10 - 15 -15 -10 -5^0  5  10  15  20  PCA axis 1  Figure 4.1. PCA ordinations of four pant alliances (a) and eight plant associations (b) showing 95% confidence ellipses for the means of alliances and associations superimposed on PCA ordinations.  50  Regardless of apparent separation, there were overlaps among the stands of the four alliances (Figure 4.1a). This was not surprising considering that the first two principal components accounted for only a part (43.8%) of the variance contributed by all the diagnostic species and the four alliances are, to some degree, similar as they represent mid-seral stands of secondary succession. The Betula all. was mixed with other alliances because its two diagnostic species had a very low loading on the first two PCA axes. Similar to plant alliances, large overlaps occurred among stands of different plant associations, especially between the DISP and STRE a.s, and between the SHEP and ARAL a.s (Figure 4.1b). Only a small portion (29.4%) of the total variance was explained by the first two PCA axes. Some diagnostic species, which differentiated plant associations within the same alliance, were probably responsible for the overlap, as they were not well represented by the first two PCA axes. This inference was supported by the PCA ordination of plant associations within each of the three alliances. Using 10, 21, and 16 diagnostic species, which differentiated plant associations within the Spiraea, Streptopus, and Petasites all.s, respectively, PCA ordinations (the first two PCA components explained 44.6%, 41.0%, and 46.9% of the total variance, respectively) and 95% confidence ellipses of the means suggested strong separation of plant associations within each alliance (Figure 4.2). A further examination of the floristic pattern among the study stands and vegetation units was done by RA ordinations which used all species in each stand as variables. The ordination corroborated the classification given above. According to the pattern shown in Figure 4.3, the study stands (excluding the two stands of the BETU a.) were stratified into seven groups corresponding  (a)  51  15 10  A = ARAL  5  B = SHEP  0  —5 —10 ^ ^ ^ ^ ^ —10 —5 10 0 5 15 PCA axis 1  (b)  15 10 cv •  A = DISP  5  X  a  o  B = STRE  —5 —10 ^ ^ ^ ^ —5 0^5 —10 10 15 PCA axis 1  (c)  15 10 cv  a  A = ATHY  5  B = EQUI  0  C = PETA  —5 —10 ^ ^ ^ ^ —10 —5 0^5 10 15 PCA axis 1  Figure 4.2. 95% confident ellipses for the means of the distinguished plant associations superimposed on PCA ordinations showing the separation of plant associations within the Spiraea (a), Streptopus (b), and Petasites (c) all.s.  52  A = ARAL B = ATHY C = BETU D = DISP E = EQUI F = PETA G = SHEP H = STRE  RA axis 1  Figure 4.3. 95% confidence ellipses for the means of seven associations superimposed on RA ordination (based on all species from 100 plots).  53  to the seven distinguished plant associations. The ordination showed an interpretable pattern of stands, with axis 1 representing a soil moisture gradient and axis 2 a soil nutrient gradient. As scores increased, soil moisture increased from moderately dry to wet and very wet (Figure 4.4a). For either low or high scores on axis 1, soil nutrients decreased with increasing scores on axis 2 (Figure 4.4b). In the middle part of axis 1, where soils were mainly fresh and moist, soil nutrients did not show an obvious trend along axis 2.  4.4.1.2.^Indicator plant analysis The frequencies of ISGs for soil moisture and nitrogen calculated for each plant association were arranged in the order of increasing soil moisture and nitrogen (Table 4.7). The eight plant associations reflected a more obvious soil moisture gradation than a soil nitrogen gradation. SMRs and SNRs, inferred from the frequencies of ISGs for soil moisture and nitrogen, respectively, are given in Table 4.8. Corresponding to differences in inferred SNR and SMR, differences in white spruce site index among plant alliances and associations were observed, although statistical tests were not carried out because of heterogeneous group variances. Site index was lowest for the Betula all., highest for the Streptopus all., and similar for the other two alliances. In relation to plant associations, site index increased from very low for the BETU and EQUI a.s, representing water-surplus and poor-aeration sites, to very high for the DISP a., representing moist and nutrient-rich sites. Significant differences in white spruce site index were observed when the stands were stratified according to their inferred SMRs and SNRs. Multiple comparisons indicated that site index on wet sites was significantly lower than on the other sites. Site index on moderately dry sites was also significantly lower than on the other sites, except for very moist sites. No significant differences in  54  (a)  110 A  90  A A  SM R:  F  A = Moderately dry  F  E  AA a  70  AA  B = Slightly dry  F  DitAA B  E  EE  A  C = Fresh  B n."11^D JIB  R^)1)  %BD :  F  DE  C  D = Moist i-E  E = Very moist F = Wet and Very wet  D  30 10  C D D C  -10 -10^10^30^50^70  90  110  RA axis 1  (b)^110  SNR:  A C  90  A  C  A = Very poor  A  70  C B  B  C  133 Bft c A ci L c^DD D  B = Poor  LC D rij4 C ^D FPc R^f D DS^11) C^DE ^CE  50  D  C = Medium  E  30  D = Rich 10  C  E C E  E  -10 -10^10^30^50^70  E = Very rich 90  110  RA axis 1  Figure 4.4. RA ordination (based on all species from 100 plots) showing the patterns of soil moisture regimes (a) and soil nutrient regimes (b).  Table 4.7. Frequencies of indicator species groups (ISGs) of soil moisture and soil nitrogen for eight distinguished plant associations. Symbols for the plant associations are given in Table 4.6).  Plant association^SHEP  ARAL  STRE  DISP  PETA  ATHY  EQUI  BETU  2 21 41 34 2 0  0 3 38 49 9 1  0 2 30 53 15 0  0 8 31 41 19 1  0 0 15 38 43 4  0 5 5 21 56 13  8 0 2 2 17 71  54 18 28  55 13 32  28 14 58  47 20 33  20 25 55  45 33 22  56 32 12  ISGs of soil moisture l : MOIST1 (%) MOIST2 (%) MOIST3 (%) MOIST4 (%) MOIST5 (%) MOIST6 (%)  4 33 48 12 3 0  ISGs of soil nitrogen 2 : NITR1 (%) NITR2 (%) NITR3 (%)  64 21 15  1 MOIST1, MOIST2, MOIST3, MOIST4, MOIST5, and MOIST6 designate excessively dry to very dry, very dry to moderately dry,  moderately dry to fresh, fresh to very moist, very moist to wet, and wet to very wet ISGs, respectively.  2 NITR1, NITR2, and NITR3 designate nitrogen-poor, nitrogen-medium, and nitrogen-rich ISGs, respectively.  56  Table 4.8. Means and deviations (in parenthesis) of site index and inferred soil moisture regimes and soil nutrient regimes for the vegetation units distinguished in the study stands. Vegetation unit  Number of plots  Site index (m @ 50 yr b.h. age)  Soil moisture regime 1  Soil nutrient regime2  Plant alliances 3 Streptop us  22  20.82 (1.50)  F to M  M to R (VR)  Petasites  52  17.67 (3.84)  M to W  M to R (VR)  Spiraea  26  17.60 (2.73)  MD to SD  (VP) P to M  Betula  2  8.95 (4.17)  VW  M  BETU  2  8.95 (4.17)  VW  M  EQUI  9  12.17 (4.18)  W  M  ATHY  8  18.93 (2.58)  VM  R (VR)  PETA  35  17.67 (3.84)  M  M  DISP  7  21.87 (1.58)  M  R (VR)  STRE  15  20.33 (1.22)  F  M  ARAL  9  20.71 (1.18)  SD  M  SHEP  17  15.95 (1.62)  MD  P (VP)  Plant associations 3  1 Symbols for soil moisture regimes are given in Table 4.2. 2 VP - very poor, P - poor, M - medium, R - rich, VR - very rich. 3 Symbols for vegetation units are given in Table 4.6.  57  site index were found between slightly dry, fresh, and moist sites. The differences in site index among the SNRs were not tested since the group variances were heterogeneous. However, the data suggested a slight increase in site index from the very poor to the rich and very rich SNRs (Figure 4.5).  4.4.2. Characterization of Soil Moisture and Aeration Regimes 4.4.2.1.^Delineation of Soil Moisture Regimes The Et/Emax ratio, the depth to a gleyed layer or prominent mottling, and the depth to a groundwater table were used to separate the 102 study stands, according to the criteria proposed in Table 4.2, into the following regimes: moderately dry, slightly dry to fresh, fresh to moist, moist, very moist, wet, and very wet. As a result, 21 stands were assigned into moderately dry, 36 stands into slightly dry to fresh, 2 stands into fresh to moist, 21 stands into moist, 13 stands into very moist, 6 stands into wet, and 3 stands into very wet. In view of the imprecise calculation of evapotranspiration, separations between the slightly dry and fresh SMRs and between the fresh and moist SMRs were made qualitatively by considering other possible water inputs as reflected in the frequencies of ISGs for soil moisture. As a result, the two fresh to moist stands were differentiated into one fresh and one moist stand; the 36 slightly dry to fresh stands were differentiated into 28 slightly dry and eight fresh stands. Compared to the field-estimated SMRs, 72 (70.5%) stands were classified into the same class. Twenty eight (27.5%) stands were classified into adjacent classes. Only two plots were classified more than one class apart. These two stands were reassigned according to the frequencies of ISGs for soil moisture. The means and standard deviations of the three differentiating characteristics for the SMRs are given in Table 4.9. On water deficit sites, the Et/Emax ratio sufficed to separate moderately dry from slightly dry and fresh  58  10 -  MD SD^F^M VM W/VW Soil moisture regime  VP^P^M  ^  R/VR  Soil nutrient regime  Figure 4.5. Means and standard deviations of measured site index by inferred soil moisture regimes (a) and soil nutrient regimes (b). Symbols for SMRs and SNRs are given in Table 4.2 and Table 4.8, respectively.  Table 4.9. Means and standard deviations (in parentheses) of the differentiating characteristics and the frequencies of ISGs stratified according to soil moisture regimes (SMRs) identified in the study. Symbols for SMRs are given in Table 4.2. Dominant indicator species groups are printed in bold. Property Number of stands  MD 21  SD 28  Soil moisture regimes F M 21 9  VM 14  W 6  VW 3  Et/Emaxl  0.87 (0.04)  0.92 (0.04)  0.95 (0.04)  1.0 (0)  1.0 (0)  1.0 (0)  1.0 (0)  Depth to a gleyed layer or prominent mottles (cm)  na  na  na  31.9 (5.7)  20.4 (1.5)  na  na  Depth to a groundwater table (cm)  na  na  na  na  43.6 (8.4)  23.3 (8.2)  0 (0)  MOIST1 2 (%)  3.0 (5.6)  0.7 (1.4)  0 (0)  0 (0)  0 (0)  5.5 (13.5)  0 (0)  MOIST2 2 (%)  30.9 (14.9)  13.3 (11.4)  3.8 (6.7)  2.1 (3.3)  0.9 (1.4)  0.3 (0.5)  14.7 (25.4)  MOIST3 2 (%)  47.3 (15.2)  45.1 (15.6)  34.3 (15.4)  23.9 (12.9)  12.9 (11.5)  5.5 (3.4)  3.0 (2.0)  MOIST4 2 (%)  13.4 (11.0)  32.2 (15.0)  49.7 (11.6)  50.7 (13.8)  44.0 (18.2)  19.5 (11.2)  17.7 (17.2)  MOIST5 2 (%)  4.8 (3.8)  8.6 (9.7)  11.2 (9.2)  22.6 (12.1)  38.4 (21.0)  50.7 (25.9)  46.7 (28.3)  MOIST6 2 (%)  0.7 (2.4)  0.1 (0.3)  1.0 (2.6)  1.0 (1.4)  3.8 (8.6)  18.5 (23.0)  18.0 (12.5)  lEt/Emax - the ratio of actual evapotranspiration and potential evapotranspiration. 2 MOIST1 through MOIST6 represents frequencies of soil moisture indicator species groups: MOIST1 - excessively to very dry, MOIST2 - very to moderately dry, MOIST3 - moderately dry to fresh, MOIST4 - fresh to very moist, MOIST5 - very moist wet, and MOIST6 - wet to very wet.  60  SMRs. Very dry and excessively dry sites were rarely occupied by white spruce in the study area and were not sampled. On water surplus sites, the depth to the gleyed layer or prominent mottling and the depth to the groundwater table sufficed to differentiate moist, very moist, wet, and very wet SMRs. There were apparent differences in the Et/Emax ratio, depth to the gleyed layer or prominent mottling, and depth to the groundwater table between moderately dry and slightly dry or fresh, between moist and very moist, and among very moist, wet, and very wet SMRs, respectively. 4.4.2.2.^Delineation of Soil Aeration and Moisture-Aeration Regimes According to the status of soil water saturation (as implied by water deficit, gleyed horizon, mottling, or groundwater table), drainage, slope, and soil texture, the study stands were qualitatively classified into three SARs: adequate (a), restricted (r), and deficient (d) (Table 4.5). As a result, 78 stands were assigned into the adequate SAR; 21 stands into the restricted SAR; three stands into the deficient SAR. As soil aeration decreased from adequate through to deficient, so did white spruce site index. Since soil water and air (oxygen), both occupying soil pores, are closely related to each other, and most factors used to determine SMRs and SARs were the same, it was logical to develop an integrated concept of soil moistureaeration regimes (SMARs) (Figure 4.6). Where applicable, these special regimes should be used to characterize the ecological site quality of forest sites. On excessively dry and very dry sites, deficient and restricted aeration would not be possible in undisturbed forest soils because most of the soil pores are occupied by air. Normally, restricted aeration on moderately dry, slightly dry, and fresh sites would not occur, but when soils are fine-textured and soil  61  Soil aeration regime a^ r ED  EDa  VD  VDa  MD  MDa  SD  SDa  F  Fa  M  Ma  VM  W  VMa  ^  d  DFr  -Mr  Wr - - - •Wd• - - --  VW  Figure 4.6. Matrix showing the relationships of soil moisture-aeration regimes to soil moisture regimes and soil aeration regimes. Symbols for soil moisture and soil aeration regimes are given in Table 4.2 and Table 4.5, respectively. Symbols for soil moisture-aeration regimes are as following: MDa - moderately dry and adequately aerated, DFr - dry to fresh and restrictedly aerated, SDa - slightly dry and adequately aerated, Fa - fresh and adequately aerated, Ma - moist and adequately aerated, VMa - very moist and adequately aerated, Mr moist to very moist and restrictedly aerated, Wr - wet and restrictedly aerated, and Wd - wet to very wet and deficiently aerated.  62  permeability is low (typically in Luvisols), restricted aeration conditions do occur for short and often recurring periods during the growing season. Therefore, moderately dry, slightly dry, and fresh sites with restricted soil aeration were combined into one regime (DFr). As soil water surplus increases from moist to very wet SMRs, aeration becomes increasingly limiting to tree growth. Thus, moist and very moist sites with restricted aeration were combined into one regime (Mr). Wet sites with restricted aeration were combined into one regime (Wr). Wet and very wet sites with deficient aeration were combined into one regime (Wd) (Figure 4.6). The combined soil moisture and aeration classification helps to explain which factors may affect plant growth on different sites (cf. Stone 1978). On excessively dry, very dry, and moderately dry sites with adequate aeration, soil moisture is the major limiting factor. On slightly dry, fresh, moist, and very moist sites with adequate aeration, soil nutrients, temperature, and light may play an important role. On moderately dry, slightly dry, and fresh sites with restricted aeration, both soil moisture and/or aeration may become limiting factors. On moist, very moist, wet, and very wet sites with both restricted and deficient aeration, soil aeration becomes the major limiting factor. 4.4.2.3.^Soil Moisture Regimes in Relation to Understory Vegetation As one of the major factors that affect the occurrence and abundance of plant species, soil moisture status (implied by SMR) of a site should be reflected in the floristic composition of vegetation, especially in the distribution pattern of indicator plants for soil moisture. This was examined by tabular comparison (Table 4.10). The diagnostic combinations of species showed that the seven SMRs were first stratified into three groups: (1) moderately dry and slightly dry, (2) fresh, moist, and very moist, and (3) wet and very wet. Each of these groups  Table 4.10. Diagnostic combinations of species for the seven soil moisture regimes (SMRs) delineated in the study stands. Symbols for SMRs as in Table 4.2. Soil moisture regimes^ MD^SD^F^M^VM^W^VW Number of stands^Diagnostic^21^28^9^21^14^6^3 Plant species^value'^Presence class 2 and mean species significance 3 Moderately to slightly dry SMRs Amelanchier alnifolia (mdf)4 Arnica cordifolia (mdf) Geocaulon lividum (?) Goodyera oblongifolia (mdf) Spiraea betulifolia (vdmd)  (d,c)  (d) (d) (d) (d,c)  V III III IV V  3 3 3 1 4  IV IV IV III V  2 3 3 1 4  IV III II  (d) (d) (d) (d) (d) (d)  IV IV III III IV IV  3 2 3 1 3 4  II I I II  I  1  II  I + I+ I 1  III IV III  III  IV  1 2 +  1  2  III II  1 2  III II  1 +  II III  1 2  II  +  II  +  I  +  I  III  1  III  1  I  +  IV III II  3 3 1  +  Moderately dry SMR Dicranum scoparium (?) Epilobium angustifolium (?) Lycopodium complanatum (mdf) Dryzopsis asperifolia (?) Peltigera aphthosa (vdmd) Shepherdia canadensis (vdmd)  II  II  +  II  I  +  I I  + +  I  +  1 1 4  IV III II  3 1 3  V III II  3 2 1  1  III IV IV IV  2 2 5  4  III IV IV  I II II II II III II II  1 1 3 2 3 3 1 2  1  1^II 1^I 1^I 1 2  + + +  1 2 2  I  +  1  Slightly dry SMR Mitella nuda (fvm) Pyrola asarifolia (mdf) Streptopus roseus (fvm)  (d) (d) (d)  II II IV  4 4  1  2  Fresh, moist, and very moist SMRs Actaea rubra (fvm) Galium trifidum (fvm) Gymnocarpium dryopteris (fvm) Ribes lacustre (?)  (d) (d) (d) (d)  I+  I+ I+  II  +  II  1  II  IV  IV  3  III  + 1 3 2 2 1 + +  III  IV  1 4 2  1  I  1  IV  2 4 2  I II III  + 2 3  I I I I II II II  + 1 1 2 1 1 3  I II  + 1  IV II  3 1  I I  + +  2 2  + 1  Fresh SMR Dryopteris expansa (fvm) Lycopodium annotinum (mdf) Pseudotsuga menziesii (?) Rubus pedatus (fvm) Streptopus roseus (fvm) Thalictrum occidentale (fvm) Tiarella trifoliata (fvm) Veratrum viride (?)  (d) (d) (d) (d) (d) (d) (d) (d)  I  +  I II I I II  + 2 1 1 1  I I II I III II I I  III  IV III IV IV IV III  1  2  4 3 4 2 2 2  0") w  Moist and very moist SMRs Athyrium filix-femina (mw) Equisetum palustre (?) Equisetum sylvaticum (mw) Hylocomium splendens (?) Mitella nuda^(fvm) Ribes^triste^(?) Rubus pubescens (fvm) Streptopus amplexifolius^(fvm)  (d) (d) (d) (d) (d,c) (d) (d,c) (d)  I I II I  + + 3 +  III  2  III III II I  2 2 3 2  I III III  + 4 1  I I II  + 1 1  IV I  3 +  II II  4 1  II III III III IV II IV IV  III 3 IV 4 III 4 I+  III IV III II  3 3 3 2  III IV IV III  I  +  1 3 1 3 3 1 4 1  III III III IV V III V IV  1 2 2 2  I II II  4 4 3 4 3 1 5 1  II IV II IV IV II V I  3 6^4 3^2 4^2 3^4 1^4 4^4 +  6 1 1 1 1 4  Moist SMR Aralia nudicaulis^(fvm) Clintonia uniflora^(mdf) Rubus parviflorus^(?) Symphoricarpos albus (?)  (d) (d) (d) (d)  2 1^II 2  +  III  3  III  5^2  4  3^4 5^5  5  Very moist SMR Carex disperma  I  (d)  +  Wet and very wet SMRs Aulacomnium palustre (wvw) Sphagnum capillaceum (wvw)  (d) (d,cd)  I  I  1  1  I  1  II I  2 +  IV IV  I I II I III V V IV II  +  I I II II IV V V IV I  +  IV IV IV IV IV V IV III IV  2 3 5 3 4 4 3 3 3  2 2 2 2 2 2 2  1 4 5 + 1 2 +  I II  1 1  I  2  2^IV 1^II +^IV 4^II  3 1 5 4  4 4 5 4  5 1 7 5  7  Wet SMR Angelica genuflexa (wvw) Aster subspicatus (?) Calamagrostis canadensis^(mw) Geum macrophyllum (fvm) Hylocomium splendens (?) Pleurozium schreberi ^(?) Ptilium crista-castrensis^(?) Ribes^lacustre^(?) Rubus pedatus (fvm)  (d) (d) (d) (d) (d) (d,c) (d) (d) (d)  +  I  II 3 V 6 IV 5 I+ 1 I  III V V II I  I  4 7 5 1 2  II  +  I V V IV III  1 5 5 2 3  I  1  2 1 + 3 6 5 4 2  1 3 + 4 5 5 2 2  Very wet SMR Aulacomnium palustre (wvw) Ribes^triste^(?) Sphagnum capillaceum (wvw) Spiraea douglasii^(vmw)  (dd) (d) (cd) (d)  "As in Table 4.6. 4 Indicator values are from Klinka et al (1989a).  I  1 I  +  II III I II  65  was further stratified into individual SMRs, except group (2) in which fresh sites were separated first from moist and very moist and then moist and very moist were separated from each other. The indicator values of most diagnostic species (from Klinka et al. 1989) are also listed in Table 4.10. These were found to correspond well with the SMRs. Using all diagnostic species as independent variables, the pattern of SMRs in the two-dimensional space defined by the first two PCA axes is given in Figure 4.7. Despite two overlaps, which were confined to adjacent SMRs, the ordination separated the study stands reasonably well according to SMRs. Frequencies of ISGs for soil moisture, which were calculated for each stand and stratified according to SMRs (Table 4.9), showed that their pattern was related to soil moisture. As SMRs changed from moderately dry to very wet, the dominant frequency changed from the MOIST2 to the MOIST5 ISG. Further statistical testing of the differences was not done because the group variances were heterogeneous. Significant (p < 0.001) relationships were found between the three differentiating characteristics of SMRs and frequencies of ISGs for soil moisture (Table 4.11). With increasing frequencies of the MOIST2 and MOIST3 ISGs, Et/Emax decreased (i.e., soil water deficit increased). The depth to the gleyed layer or prominent mottling was negatively related to the frequency of the MOIST5 ISG. A logarithmic relationship was found between the depth to the groundwater table and the frequency of the MOIST4 ISG. The higher frequency of the MOIST4 ISG was associated with a lower groundwater table. This also implied that sites with a high groundwater table were dominated by plants indicating very moist and to very wet conditions (MOIST5 and MOIST6 ISGs).  66  PCA axis 1  Figure 4.7. Ordination of study stands along the first two PCA axes based on diagnostic species in Table 4.10 showing 95% confidence ellipses for the means of soil moisture regimes (SMRs). Symbols for SMRs are: A - moderately dry, B - slightly dry, C - fresh, D - moist, E - very moist, and F - wet and very wet.  67  Table 4.11. Models for the regression of the frequency of soil moisture indicator plants on the ratio of actual and potential evapotranspiration (Et/Emax), depth to the gleyed layer or prominent mottling (GLEY), and depth to the groundwater table (GW). [4.1] Et/Emax = 0.983 - 0.00194(MOIST2) - 0.00096(MOIST3) adjusted R 2 = 0.41^SEE = 0.04^n = 59 where MOIST2 and MOIST3 are relative frequencies (%) of the very dry to moderately dry and the moderately dry to fresh ISGs, respectively. [4.21 GLEY = 22.1 - 0.074(MOIST5) adjusted R 2 = 0.26^SEE = 2.1 cm n = 25 where MOIST5 is the relative frequency of the very moist to wet ISG. [4.3] GW = 2.1 + 4.004In(MOIST4) adjusted R2 = 0.61^SEE = 3.1 cm n = 18 where MOIST4 is the relative frequency of the fresh to very moist ISG.  4.4.2.4.^Soil Moisture, Aeration, and Moisture-Aeration Regimes in Relation to White Spruce Foliar Nutrients and Site Index If the differences in available soil water and oxygen are expressed by different SMRs and SARs, they should be manifested in the foliar nutrient status and the productivity of the study stands. This was found when foliar nutrients and site index were stratified according to SMRs, SARs, and SMARs (Table 4.12 and Figure 4.8). Significant differences in foliar nutrients were found among different SMRs. As expected, slightly dry, fresh, moist, and very moist sites, which supported reasonably good growth of white spruce, were significantly higher in foliar N, P, and S (Table 4.12). Site index varied significantly among the SMRs. Tukey's multiple comparisons indicated that there were significant differences  Table 4.12. Means and standard deviations (in the parentheses) of white spruce site index and foliar macronutrients stratified according to soil moisture regimes (SMRs). Symbols for SMRs are given in Table 4.2. Property Number of stands  MD 21  SD 28  Soil moisture regime M F 21 9  Site index (m @ 50 yr b.h. age)  15.6 (1.36)  19.3 (1.72)  21.2 (1.17)  N (mg kg-1 )  1.06 (0.12)  1.21 (0.07)  P (mg kg-1 )  0.23 (0.02)  K (mg kg-1 )  VM 14  W 6  VW 3  21.3 1.56  17.9 (1.75)  12.5 (1.69)  6.1 (1.40)  1.22 (0.08)  1.21 (0.09)  1.24 (0.11)  1.11 (0.13)  0.90 (0.09)  0.24 (0.02)  0.26 (0.02)  0.24 (0.02)  0.24 (0.03)  0.21 (0.03)  0.14 (0.02)  0.57 (0.06)  0.60 (0.07)  0.63 (0.05)  0.60 (0.06)  0.60 (0.07)  0.54 (0.05)  0.42 (0.06)  S (mg kg-1 )  0.085 (0.009)  0.093 (0.009)  0.099 (0.004)  0.098 (0.007)  0.096 (0.012)  0.089 (0.011)  0.075 (0.006)  Ca (mg kg-1 )  0.56 (0.12)  0.59 (0.10)  0.53 (0.10)  0.52 (0.07)  0.50 (0.08)  0.43 (0.07)  0.40 (0.04)  Mg (mg kg-1 )  0.13 (0.02)  0.14 (0.01)  0.14 (0.02)  0.14 (0.01)  0.13 (0.01)  0.12 (0.02)  0.11 (0.03)  69  (a) ela  30  Fr  O  ul  20  f  0 0)  10  cn  ti (I)  0  a  ^ ^  r  d  Soil aeration regime  (b)  30  et  cti  4 Fw  Lc) 20  aU  x  a.)  a.) 10  0  MDa DFr SDa Fa Ma VMa Mr Wr Wd  Soil moisture and aeration regime  Figure 4.8. Means and standard deviations of white spruce site index in relation to (a) soil aeration regimes (SARs) and (b) soil moistureaeration regimes (SMARs). Symbols for SARs and SMARs are given in Table 4.5 and Figure 4.6, respectively.  70  in site index between all possible pairs of SMRs, except between slightly dry, fresh, and moist sites and between slightly dry and very moist sites. Water deficit reduced white spruce site index on moderately dry sites. The low site index on very moist sites and the extremely low site index on wet and very wet sites were attributed to deficient aeration, as implied by soil water saturation. The importance of aeration for white spruce growth reported in the literature implies that consideration of soil aeration (implied by SARs and SMARs) would facilitate the study of site-productivity relationships. Significant differences in site index were found among the three SARs and nine SMARs (Figure 4.8). In view of the relationships between white spruce site index and the proposed SMARs, four SMAR groups, each with site index significantly different from the others, were recognized. Group 1, including slightly dry, fresh, moist, and very moist sites with adequate aeration, represent the best soil moisture and aeration conditions for white spruce growth. No significant differences in site index were found among these SMARs, although slightly dry and very moist sites showed a slight decrease in site index likely due to slight water deficits (slightly dry sites) and a decrease in soil temperature from increased water surplus (very moist sites). Group 2, which included moderately dry sites, dry to fresh and moist to very moist sites with restricted aeration, represented limited soil moisture and/or aeration conditions for white spruce growth. Group 3, which included wet sites with restricted aeration, and group 4, which included wet to very wet sites with deficient aeration, represented marginal soil moisture and aeration conditions for white spruce growth.  71  4.4.2.5.^Discussion In British Columbia, SMR has been used as a synoptic measure of soil available water, and hence an important predictor of forest productivity. In order to provide an objective means to characterize SMR, the simple forest water balance model was introduced to calculate the Et/Emax ratio and water deficit  (Emax - Et) (Carter and Klinka 1990), and subsequently tested in several other studies (Kayahara 1992, Pearson 1992, and Q. Wang 1992). The results from the different studies appear to contradict each other. Carter and Klinka (1990) and Wang (1992) reported that water deficit or the ratio of Et/Emax were useful to characterize SMRs and to explain the variation of forest productivity in their studies. Kayahara (1992) reported that the driest non-forested ecosystem, with only 1 cm of soil, in his study area showed no water deficit using this model. Pearson (1992) reported that results calculated from this model could not be confidently applied to differentiate SMRs for Sitka spruce ecosystems in her study area. Keeping this problem in mind, the purpose of applying the model in my study was confined to providing an integrated index for the relative soil moisture status of a site, rather than quantifying the actual water deficit of the site. Unlike previous applications, the model was first calibrated on a zonal site from each of the biogeoclimatic units, and then applied only to those study stands with freely draining soils. Using the Et/Emax ratio, depth to the gleyed layer or prominent mottling, and depth to the groundwater table as differentiae, this study provided an objective means to characterize SMRs. The distinguished SMRs were meaningful when related to indicator plants and white spruce foliar nutrients and site index. Although the water deficit calculated from the model was useful as an index to compare the relative soil water deficit among sites, it may not be a precise estimation of the water deficit experienced by the sites. Several obstacles  72  encountered in this study made the calculation of the exact water deficit difficult. Firstly, data collected in the study could not be used to validate the model output. Secondly, due to the lack of daily data, monthly data for solar radiation, precipitation, and temperature were used in the calculations, which tends to underestimate soil water deficit although a consistently similar pattern (i.e., the change of soil available water over a growing season) was found compared to using daily or weekly data (Black 1982, Q. Wang 1992). Thirdly, the use of one AES station for each subzone or variant (precipitation and temperature) and one AES station for all six biogeoclimatic units (solar radiation) may be questionable. Soil aeration is recognized as one of the important factors controlling white spruce growth (e.g., Ahlgren and Hansen 1957, Nienstaedt and Zasada 1990). As soil water supply could not be a problem for trees growing on water surplus sites, the characterization of moist, very moist, wet, and very wet SMRs mainly implies differences in soil aeration. To explicitly express the difference in soil aeration, SARs were qualitatively determined in this study. The significant differences in white spruce site index among the distinguished SARs suggest that the SARs were meaningful in relation to white spruce growth. To justify the qualitative approach proposed in the study, the differences in oxygen supply among different SARs should be quantified in future studies by directly measuring oxygen concentration and diffusion rate in soil pores. 4.4.3. Characterization of Soil Nutrient Regimes 4.4.3.1.^Delineation of Quantitative Soil Nutrient Regimes CA and DA stratified the population of study stands into five groups (Table 4.13). These groups were arranged from A through E in order of increasing min-N and decreasing C/N—two variables which were used to  73  discriminate among the groups. Ordination of study stands based on ln(min-N) and C/N showed that there was a major trend of decreasing C/N with increasing ln(min-N) (Figure 4.9). Tukey's multiple comparison indicated that C/N was significantly different between all possible pairs of groups except between B and C and between D and E. These two pairs appear to be well differentiated by minN [23.8 (group B) versus 40.4 kg ha -1 (group C), and 56.6 (group D) versus 109.5 kg ha -1 (group E) although their differences were not statistically tested due to heterogeneous group variances (Table 4.13).  Table 4.13. Coefficients of discriminant functions and means and stand deviations (in parenthesis) of min-N and C/N for five groups delineated by cluster and discriminant analyses. Group^A^B^C^D^E Number of plots^(16)^(10)^(39)^(17)^(11)  Coefficients of discriminant functions Constant^-65.293^-67.339^-91.040^-110.385^-157.330 ln(min-N)^23.181^35.205^47.507^59.267^75.223 (kg ha-1) C/N  ^  1.603^0.680^0.190^-0.664^-1.480  Means and standard deviations of min-N and C/N: ^ ^ min-N ^ 16.3^23.8^40.4^55.6 ^ 109.5 (41.1) (kg ha-1) (3.8)^(7.6)^(9.5)^(17.8) C/N^39.3^33.1^33.6^28.2^26.0 (3.8)^(3.1)^(4.0)^(3.9)^(4.3)  74  Natural logarithm of min—N  Figure 4.9. Ordination of study stands on natural logarithm of min-N and C/N and 95% confidence ellipses of the means for the five delineated groups.  75  In general, variation in the other nutrient properties followed the above trend (Table 4.14), thus supporting the choice of total soil min-N and C/N as differentiating characteristics for the SNR classification of the study stands. ANOVA showed that forest floor pH and C/N and total soil C were significantly different among the five groups. Since Bartlett's test indicated that the group variances of mineral soil pH and C/N, total soil nitrogen, total soil exchangeable Ca, Mg and K, and available S were heterogeneous, the differences in these properties were not tested. However, these properties did increase from group A through group E (Table 4.14). Compared to other soil nutrient properties, available soil P showed an opposite trend (Table 4.14). However, due to a large within-group variation, between-group differences were not significant. Phosphorus nutrition is probably the most difficult one to be related to soil chemical analysis since it depends on several factors outside of P availability (Ballard and Carter 1986). The first two axes of the PCA based on all nutrient properties (excluding min-N and C/N) expressed 59% of the total variance. The 95% confidence ellipses imposed on the PCA ordination showed that groups A through E occurred along axis 1 from left to right and along axis 2 somewhat from top to bottom (Figure 4.10). The analysis suggested that the groups delineated by ln(min-N) and C/N also account for differences in other soil nutrient properties. As a result, the delineated groups could be considered to represent five SNRs (Krajina 1969, Klinka et al. 1984, 1989a) for the soils in the study stands: A - very poor (VP), B - poor (P), C - medium (M), D - rich (R), and E - very rich (VR).  76  Table 4.14. Means and deviations (in parentheses) of other measured soil chemical properties summarized according to five delineated soil groups. Soil group property  A (11=16)  B (n=10)  C (n=39)  D (n=17)  E (n=11)  Forest floor pH  4.4 (0.45)  4.5 (0.34)  4.9 (0.40)  5.2 (0.58)  5.5 (0.57)  Mineral soil pH  4.5 (0.36)  4.7 (0.38)  5.0 (0.58)  5.3 (0.58)  5.5 (0.92)  Forest floor C/N  41.9 (5.8)  41.7 (5.4)  40.8 (7.6)  38.1 (7.1)  33.9 (5.6)  Mineral soil C/N  49.7 (37.6)  40.6 (28.3)  34.3 (10.9)  26.3 (2.8)  23.6 (1.9)  Soil total C (kg ha -1)  58427 (22538)  59133 (19183)  73403 (22024)  92314 (20093)  140551 (35719)  Soil total N  (kg ha I)  1479 (581)  1731 (735)  2169 (660)  3332 (857)  5544 (1603)  Soil exchangable Ca (kg ha -1 )  1352 (982)  1710 (1224)  2984 (1717)  4790 (2226)  7515 (3706)  Soil exchangable Mg (kg ha - I)  314 (527)  331 (441)  478 (387)  913 (768)  906 (765)  Soil exchangable K (kg ha -1 )  112 (99)  90 (55)  154 (133)  134 (76)  148 (96)  Soil available S (kg ha -1 )  8.7 (4.0)  11.5 (5.9)  17.3 (9.6)  23.4 (10.2)  33.2 (15.9)  Soil available P (kg ha -1)  50.1 (66.0)  37.6 (45.9)  32.8 (39.2)  16.9 (12.8)  12.3 (10.7)  -  77  3 2 1 0 —1 —2 —3 —4  —3^—2^—1^0^1  ^ ^ ^ 2 3 4  PCA axis 1  Figure 4.10. Ordination of the study stands along the first two axes of PCA based on all measured soil nutrient properties (except min-N and C/N) and 95% confidence ellipses of the means for the five delineated groups.  78  4.4.3.2.^Testing Soil Nutrient Regimes 1.^Relationships Between Vegetation and Soil Nutrient Regimes Vegetation, especially indicator plants, has been used as an indirect index of ecological site quality (e.g., Cajander 1926; Daubenmire 1976; Klinka et al. 1989a). Thus, differences in available soil nutrients or SNRs should be manifested in the floristic composition of the understory vegetation and in the occurrence of nitrogen indicator (nitrophytic) plants. Variation in understory vegetation among the five SNRs was summarized in a diagnostic table together with the nitrogen indicator value for each diagnostic species (Klinka et al. 1989a) (Table 4.15). The tabular comparison demonstrates that the plant species were not randomly distributed along the soil nutrient gradient, as they appeared in an orderly arrangement of both SNRs (columns) and diagnostic combination of species (rows) from left to right from the very poor through to the very rich SNR. Affinities between the understory vegetation and the SNRs increased with the number, presence class, and significance class of species within the diagnostic combination of species. Each SNR was differentiated by two or three diagnostic combinations of species depending on hierarchy. For example, the very poor SNR was segregated from all other SNRs by the Chimaphila umbellata and Oryzopsis asperifolia combinations; the separation of the very -  -  rich SNR required three diagnostic combinations of species: Petasites palmatus, Actaea rubra, and Smilacina stellata (Table 4.15). Perfect separation, in which  each SNR was associated with species unique to that regime, was not expected due to the relatively wide ecological amplitude of most plant species in B.C. (Pojar et al. 1987) and possible variability in soil moisture and light conditions in each SNR.  ^ ^  79 Table 4.15. Diagnostic combination of species for the five soil nutrient rgimes (SNRs) delineated in the study stands.  Soil moisture regimes^ VP^P^M^R^VR Number of stands^Diagnostic^21^28^9^21^14 species (indicator value) 4 value'^Presence class 2 and mean species significance 3 Very poor and poor SNRs Chimaphila umbellata (p) ^I  Geocaulon^lividum (p) Goodyera oblongifolia (p) Pinus contorta (?) Spiraea betulifolia (m) Vaccinium membranaceum (p)  (d) (d,c) (d) (d) (d,c) (d)  III IV IV V V III  2 3 1 5 4 5  (d) (d) (d)  III IV III  1 3 4  1 2 2 5 4 4  ^1 II 2^I+^I+ III^1^III^1^I+ III 3^III 3^II^1 IV 4^IV 2^II 2 III^3^II^3^II^2  I+ II 2 I 4  I+^I^1 II^1^I^+ II 2^I+  III V IV IV IV IV  Very poor SNR Oryzopsis asperifolia (?) Peltigera aphthosa (p) Shepherdia canadensis (m)  Poor SNR Cornus sericea (r) Osmorhiza chilensis ^(r) Pseudotsuga menziesii^(?) Ribes^lacustre (r) Rubus parviflorus (r) Sorbus sitchensis (p)  (d) (d) (d) (d) (d) (d)  I+ I 1 I 1 II 3 II 2  III III III III IV IV  2 1 4 1 4 2  III^1^III 2^III 2 III^1^IV 1^IV 2 II^3^II^2^II^4 III 1^IV 3^V 4 III 3^IV 3^IV 3 II^+^II^1^III^1  I I III II III II II  + + 1 1 1 1 2  III III III III III  3^III 3 3^V 3 1^IV 2 3^V 3 3^V 4 V 4^IV 5^V 5 III^1^III 3^III^1  III^2^II^1^I+ V 2^IV 1^III 1  Medium,^rich,^and very rich SNRs Aster ciliolatus^(m) Mitella nuda (m) Osmorhiza chilensis^(r) Petasites palmatus (r) Ribes^lacustre (r) Rubus pubescens (r) Thalictrum occidentale (r)  I I  1 1  II I III  2 1 2  (d) (d,c)  II V  2 1  III IV  2 1  (d) (d) (d,c) (d)  I I I I  + + + +  I I II  + + +  (d)  III  2  III  (d) (d,c) (d)  II III V  3 1 4  III IV IV  3 2 4 2 4  I  +  (d) (d) (d) (d) (d) (d) (d)  2^III 2^IV 1^IV 3^IV 1^IV  Medium SNRs Aster conspicuus (r) Orthilia secunda (p)  Rich and very rich SNRs Actaea rubra (r)  Brachythecium asperimum (?) Galium trifidum (r) Streptopus amplexifolius (r)  IV III IV III  1^IV 2^IV 2^V 1^IV  3 3 2 2  Rich SNR Arnica cordifolia (m) Goodyera oblongifolia (p) Rhytidiadeiphus triquetrus (m) Smilacina racemosa (r) Spiraea betulifolia (m)  (d)  IV  1  IV  III 3^III 2^I + III^1^III^1^I^+ III 3^IV 3^II^1 III 1^V 2^III 2 IV 4^IV 2^II 2  Very rich SNR Smilacina stellata (r) Viola glabella^(r)  1-3As in Table 4.6.  4 lndicator values are from Klinka  (d) (d)  et al. (1989a).  I^1^II^1 I +^I +  IV 2 III 3  80  Using all the diagnostic species as variables, the first two PCA axes explained 35.7% of the total variance. Ordination of the study stands and the 95% confidence ellipses of the means for the five SNRs showed that the study stands were stratified, although with considerable overlap, according to SNRs along the first axis from left (the very poor SNR) to right (the very rich SNR) (Figure 4.11). Frequencies of nitrogen-poor and nitrogen-rich ISGs were calculated for each stand and stratified according to SNRs. A distinct and consistent decrease of nitrogen-poor indicators and an increase of nitrogen-rich indicators from the very poor through to the very rich SNRs were found (Figures 4.12). Tukey's multiple comparison indicated that the frequencies, using either nitrogen-poor or nitrogen-rich indicator plants, for the very poor SNR were significantly different from those for the medium, rich, and very rich SNRs, and those for the poor and medium SNRs were significantly different from those of the very rich SNR. Since the characterization of SNRs in this study and the value given to indicator species of soil N (Klinka et al. 1989) were both based on available soil N, a reasonable relationship might be expected between min-N and the relative frequency of the nitrogen ISG. Two regression models were used to describe these relationships: [4.4] NITR1 = 18.04 - 8,231n(min-N) + 1.83C/N Adjusted R 2 = 0.40^SEE = 16.2 %^n = 93 [4.5] NITR3 = 50.64 +itk.631n(min-N) - 1.52C/N Adjusted R 4 = 0.39^SEE = 14.6 %^n = 93 where NITR1 and NITR3 represent frequencies of the nitrogen-poor and nitrogen-rich ISGs, respectively. Both regressions were significant (p < 0.000)  81  PCA axis 1  Figure 4.11. Ordination of study stands along the first two axes of PCA based on diagnostic species, and 95% confidence ellipses of the means for soil nutrient regimes (A - very poor, B - poor, C - medium, D - rich, E very rich).  • 82  (a)  100  1:7  80  •-•  •  60  V  cr 40 a)  • 7)  20  .  0  A^B^C^D^E Soil nutrient regime  (b)  100  ;N:" c• •-) 80  O  60  V  cr 40  a) 4— ta) co  20  0  A^B^C^D^E Soil nutrient regime  Figure 4.12. Relative frequencies (%) of nitrogen-poor (NITR1) (a) and nitrogen rich (NITR3) (b) indicator plants stratified according to soil nutrient regimes (A - very poor, B - poor, C - medium, D - rich, E very rich).  83  and showed a strong relationship in view of the obvious disparities between the location and size of the area used for the sampling of soils and understory vegetation. 2.^Relationships Between Foliar Nutrients and Soil Nutrient Regimes All measured foliar nutrients were summarized according to SNRs (Table 4.16). ANOVA indicated that N, S, P, Mn, and Cu were significantly different among the SNRs, but significant differences were not found for Ca, Mg, K, Zn, Fe, B, and SO4-S. Active-Fe were not tested since its variance was heterogeneous across SNRs. Tukey's multiple comparison indicated that N and S on very poor soils were significantly lower than on rich and very rich soils, P on very poor soils was significantly lower than on rich soils, Cu was significantly higher on very rich soils, and Mn was significantly different between all pairs of SNRs except between very poor and poor and between rich and very rich soils. Considering all stands and using the guidelines proposed by Swan (1971) and Ballard and Carter (1986), Ca, Zn, and Mn were found to be in the range of luxury consumption; P, Mg, and K were found to be in the range of sufficiency for good to very good growth; Cu, Fe, and B were found to be in the range of slightly possible deficiency to no deficiency. Foliar N in most stands was found to be in the range of acute to moderate deficiency, with only 13 plots in the range of slight deficiency to sufficiency. Foliar S of 24 stands were in the range of possible deficiency. As both foliar and soil N and S concentrations varied along a soil nutrient gradient (Table 4.14 and Table 4.16), they were found to be significantly correlated with the natural logarithms of soil min-N (r = 0.41) and available S (r = 0.44), respectively. In addition, foliar Mg was also found to be significantly correlated with the natural logarithm of soil exchangeable Mg (r = 0.42).  84  Table 4.16. Means and standard deviations (in parentheses) of measured white spruce foliar nutrients summarized according to the five soil nutrient regimes delineated in the study stands. Foliar nutrient  Very poor (n=16)  Soil nutrient regime Poor^Medium (n=10)^(n=39)  Rich (n=17)  Very rich (n=11)  N  1.10 (0.12)  1.14 (0.13)  1.19 (0.09)  1.25 (0.10)  1.23 (0.10)  P  (eg g-1)  0.23 (0.02)  0.24 (0.02)  0.24 (0.02)  0.25 (0.02)  0.24 (0.03)  Ca (cg  0.55 (0.10)  0.57 (0.12)  0.57 (0.11)  0.51 (0.08)  0.53 (0.09)  Mg (cg g-1)  0.133 (0.011)  0.130 (0.012)  0.137 (0.015)  0.134 (0.015)  0.133 (0.017)  K  0.56 (0.06)  0.57 (0.07)  0.61 (0.06)  0.61 (0.06)  0.60 (0.08)  S  (eg g-1)  0.087 (0.009)  0.089 (0.010)  0.094 (0.009)  0.097 (0.009)  0.099 (0.009)  SO4-S (mg kg-1 )  118 (43)  117 (41)  129 (47)  124 (60)  134 (56)  Cu (mg kg-1 )  3 (0.6)  3 (0.3)  3 (0.6)  3 (0.7)  4 (0.8)  Zn (mg kg-1 )  45 (9)  45 (7)  45 (7)  43 (7)  42 (4)  Mn (mg kg -1 )  549 (138)  522 (147)  399 (160)  253 (160)  191 (164)  B (mg kg-1 )  12 (3)  14 (3)  12 (4)  14 (5)  15 (3)  Fe (mg kg-1 )  32 (10)  35 (13)  35 (13)  35 (13)  32 (15)  (cg  (eg g-1)  85  Within each delineated SNR, relationships between foliar nutrients and white spruce site index were examined using correlation analysis (Table 4.17). Only N, P, and S were found to be significantly correlated with site index within one or more SNRs. Decreasing correlation coefficient values for foliar N from the very poor through the very rich SNRs corresponded to increasing by available soil N. Regardless of the amount of available soil P in the soil (Table 4.14), P uptake by white spruce increased with increasing uptake and supply of N in the order from the very poor through the very rich SNRs. Thus, strong correlations between foliar P and site index on the very poor, poor, and medium soils may suggest that P is strongly correlated with N. Correlation coefficients between site index and foliar S increased from the very poor through the very rich SNRs. This could be caused by the increasing supply and uptake of N needing to be balanced by the increasing supply and uptake of S. Table 4.17. Simple and multiple (in bold print) correlation coefficients describing the relationships between white spruce site index (m @ 50 yr b.h. age) and foliar N, P, and S (* significant at a = 0.05). Soil nutrient regime  Very poor (n=16)  Poor (n=10)  Medium (n=39)  Rich (n=17)  Very rich (n=11)  N  0.713*  0.700*  0.413*  0.005  0.043  P  0.715*  0.565*  0.647*  0.366  0.069  S  0.245  0.403  0.424*  0.536*  0.504*  N,P  0.791*  0.793*  0.669*  0.425  0.098  86  3.^Relationships between Site Index and Soil Nutrient Regimes The variation in white spruce site index across five SNRs is shown in Figure 4.13. ANOVA followed by Tukey's multiple comparison indicated that site index on very poor soils was significantly lower than on medium, rich, and very rich soils; site index on poor and medium soils was significantly lower than that on very rich soils; and no significant differences were found among other possible pairs of SNRs. Two linear regression models were developed to examine the relationship between white spruce site index and soil nutrients. The categorical model (equation 4.6), which used SNRs as dummy variables, explained 26% of the total variance in site index. [4.6] SI = 16.2 +VP + 1.6P + 2.2M + 3.8R + 4.8VR Adjusted R = 0.26^SEE = 2.3 m^n = 93 where SI is site index (m CO 50 yr b.h. age), and VP, P, M, R, and VR are as previously defined. A similar result (R 2 = 0.23, SEE = 3.5 m) was reported for lodgepole pine in the SBS and the SBPS zones by Q. Wang (1992). The continuous model (equation 4.7), which used the natural logarithm of min-N and C/N, explained 29% of total variance in site index. [4.7] SI = 18.8 +.4141n(min-N) - 0.163C/N ^ Adjusted R = 0.29^SEE = 2.3 m n = 93  87  25  I  15  10  A^B^C^D^E Soil nutrient regime  Figure 4.13. Box plot showing white spruce site index (m @ 50 yr b.h. age) stratified according to soil nutrient regimes (A - very poor, B - poor, C - medium, D - rich, E - very rich).  88  where SI, min-N, and C/N are as previously defined. Using only the natural logarithm of min-N, a similar result (R 2 = 0.28, SEE = 3.4 m) was reported for lodgepole pine in the SBS and SBPS zones by Q. Wang (1992). These two models accounted for a reasonable amount of the variation in site index, considering that soil nutrients are only one of several primary environmental factors (i.e., climate, soil moisture, soil nutrients, and soil aeration) that control tree growth. These results compare favorably to a number of similar studies that had poor success when using soil nutrients to explain the variation in white spruce site index, particularly over a large area (e.g., Russell 1963, Wilde et al. 1965, Payandeh 1986). 4.4.3.3^Relationships between Quantitative and Field Classifications Reasonable agreement was found when comparing the assignment of study stands into SNRs using the quantitative and field procedures. Fifty-six (60.2%) stands were assigned into the same SNR. Thirty (32.3%) stands had a one-class difference. Only seven (7.5%) stands had a two-class difference. Inquiry into those stands where differences occurred showed that 13 stands (seven stands with two class differences and six stands with one class difference) had possible soil nutrient input from seepage and/or substantially higher or lower frequency of the nitrogen-rich ISG than the mean frequency of the SNR. Therefore, the SNRs of these stands were reassigned according to the frequency of the nitrogen-rich ISG. The nine plots that were excluded from the quantitative analysis kept their SNRs assigned by the field procedures. Following the final stratification of the 102 study stands, a summary of soil, vegetation, and stand properties selected for characterization of SNRs was prepared (Table 4.18).  89  Table 4.18.^Means and standard deviations (in parentheses) of vegetation, soil, and stand properties selected for characterization of the final soil nutrient regimes identified in the study.  Property Number of stands  Very poor (10)  Poor (20)  Soil nutrient regime Medium (41)  Rich (22)  Very rich (9)  1NITR1 (%)  75.9 (14.0)  56.5 (16.5)  46.8 (17.7)  37.3 (17.3)  27.2 (18.5)  2NITR3 (%)  8.0 (6.0)  21.0 (12.4)  31.2 (14.2)  40.8 (14.7)  58.4 (17.8)  Soil min-N (kg ha -1)  15.6 (3.6)  29.1 (13.5)  43.5 (20.7)  59.7 (21.5)  110.7 (45.8)  Soil total N (kg ha-1)  1183 (351)  1813 (605)  2240 (851)  3220 (875)  6049 (1220)  Soil C/N  41.1 (4.9)  38.4 (9.1)  32.5 (4.8)  29.0 (4.3)  25.6 (4.7)  Foliar N (mg kg 1)  1.07 (0.09)  1.12 (0.13)  1.17 (0.12)  1.24 (0.10)  1.23 (0.10)  Site index (m @ 50 yr b.h. age)  15.5 (1.77)  17.4 (3.11)  17.2 (4.10)  20.3 (2.12)  22.0 (1.72)  1 Frequency of nitrogen-poor indicator plants. 2Frequency of nitrogen-rich indicator plants.  90  4.4.3.4.^Discussion 1.^Problems with SNR classification Compared to a traditional approach, the quantitative SNR classification provides an objective means to characterize a regional soil nutrient gradient and to construe interpretive classification. However, any SNR classification based on selected differentiating characteristics may not describe the gradient of each soil nutrient equally well. Usually, the differentiating characteristics and its accessary properties are described much better than other soil nutrient properties. In this study, as SNR changed from the very poor to the very rich, N, Ca, Mg, and S increased while K showed no increase and P showed a decreasing trend (Table 4.14). Thus, selecting soil nutrient(s) as differentiating characteristic(s) is the first and the most important step in developing a meaningful SNR classification. To select differentiating characteristics, it is critical to identify growthlimiting soil nutrients and to measure the amounts of these nutrients available to plants. This requires understanding the soil as a source of plant nutrients and the relationships among soil nutrients, foliar nutrients, understory vegetation (indicator plants), and plant growth performance within a given area. Selection based only on statistical considerations may easily result in a meaningless SNR classification that does not represent the regional nutrient gradient, and hence shows no relationships with plant communities. For example, a soil nutrient property that has the largest variance might have the most influence on the SNR classification and could be selected as a differentiating characteristic without any consideration for its ramifications to plant nutrition. As foliar N has been identified as the most limiting nutrient in this study, using min-N and C/N as differentiating characteristics for SNR classification would be appropriate. In this study, the differences of understory vegetation and forest productivity have  91  been well explained by SNRs based on the gradient of available nitrogen. This result indicates that the regional nutrient gradient can be described reasonably well by characterizing the soil N gradient. As the quantitative SNR classification was based on a soil chemical analysis, soil nutrient input by laterally moving groundwater (seepage), which cannot be accounted for by soil chemical analysis, may cause problems by assigning sites to SNRs incorrectly. This soil nutrient input may significantly improve the overall soil nutrient status and eventually be reflected in the floristic composition of understory vegetation, humus form, and productivity (e.g. Krajina 1969, Wali and Krajina 1973, Klinka 1976). In this study, several  stands were reassigned due to the influence of lateral nutrient input. 2.^Are SNRs Relative or Absolute? Since the main processes (i.e., wet deposition, dry deposition, N 2-fixation, and mineral weathering, litter decomposition) contributing to the input of soil nutrients vary with region, the same SNR (e.g., medium) may have different amount of soil nutrients in different regions. To characterize absolute SNRs in a given area, study stands have to cover the multiple soil nutrient gradients or at least the gradients of limiting soil nutrients and to be distributed more or less evenly along these gradients. If these conditions are not met, the resultant SNR classification could be based on limited data and would not have the proper range and amount of soil nutrients for each delineated SNR, in relation to the regional soil nutrient gradient. However, this study and several previous efforts towards quantitative classification (Kabzems and Klinka 1987, Courtin et al. 1988, Q. Wang 1992) were all based on ecosystems dominated by a single tree species. Because each tree species has an unique ecological amplitude, sample stands from different studies may not cover the whole regional nutrient  92  gradient. Thus, the amount of soil nutrients for the same SNRs may differ in different studies. Differences in some selected soil nutrients for the same SNR were observed in different studies in both the coastal and interior British Columbia (Table 4.19). Compared to Q. Wang's (1992) study, the amount of minN and sum of exchangeable Ca, Mg, K of this study were higher for all SNRs except for the very rich SNR. This could be explained by the differences in the study area and species. Q. Wang's study (1992) included the SBPS zone, which is much drier than the SBS zone. The actual ecological niche of lodgepole pine may cover mainly the very poor, poor, and medium sites of the upland landscape and various wetland sites, while the actual ecological niche of white spruce may occur mainly on the medium, rich, and very rich sites. Comparing this study to the studies carried out in coastal B.C., the amount of total nitrogen was consistently lower for all SNRs. The amount of min-N of this study was lower on medium, rich, and very rich SNRs than that of Kabzems and Klinka (1987). However, the sum of exchangeable Ca, Mg, and K of this study was higher than that of Courtin et al. (1988). These differences are probably due to the differences in regional climate and geological background, although incomplete representation of regional nutrient gradients may complicate the comparison. To determine the proper range and amount of soil nutrients for each delineated SNR, further studies should concentrate on characterizing soil nitrogen gradients based on more extensive data, including all major ecosystem types under the same regional climate. 3.^Criteria for a good SNR classification Difficulties in measuring the exact amount of soil nutrients available to tree growth, expressing interactions among soil nutrients, and representing multidimensional soil nutrient space suggest that the goodness of a SNR  93  Table 4.19. Means of selected soil nutrients (kg ha- 1 ) stratified according to soil nutrient regimes distinguished in different studies. Property  ^  Very poor^Poor^Medium^Rich^Very rich  This study:.  (based on data from moderately dry to very moist white spruce ecosystems in the SBS zone) min-N  17.2  27.5  37.5  59.2  90.1  total N  1183  1838  2105  3194  6050  SEB 1  1270  2559  3459  5693  8929  Q. Wang (1992)  (based on data from excessively dry to wet lodgepole pine ecosystems in the SBS and SBPS zones) min-N  2.7  9.7  29.7  38.3  130.1  SEB  1202  1040  1376  3960  8278  Kabzems and Klinka (1987):  (based on data from very dry to fresh Douglas-fir ecosystems in drier maritime CWH subzone) min-N  18  54  113  242  total N  2328  3193  4108  7121  Courtin et al. (1988);  (based on data from very dey to wet Douglas fir ecosystems in the very dry and dry Maritime CWH subzones) total N  1743  3297  12989  4067  8404  SEB  1386  873  1255  1743  5066  1Sum of exangeable Ca, Mg, and K  94  classification has to be tested not only by actual measurements from soil chemical analysis, but also by its relationships with understory vegetation, foliar nutrients, and forest productivity. In this study, SNRs were first tested by actual measurements of soil nutrients. Increasing soil nitrogen, and other nutrients such as Ca, Mg, and S, were observed from the very poor to the very rich SNRs, which suggested that different SNRs had different amounts of nutrients. SNRs were then tested in relation to understory vegetation, foliar nutrients, and site index. Some reasonably strong relationships were observed, which suggested that the SNR classification was meaningful. Another important test may be fertilization trials. As different SNRs may represent soil groups with different soil nutrient supplies, different responses would be expected if fertilizer was applied to sites with different SNRs. Carter and Klinka (1992) found that both relative and absolute responses of Douglas-fir basal area to N fertilizer decreased as SNRs changed from the very poor to the very rich. Similarly, a decreased correlation between site index and foliar N was also found in this study as SNRs changed from the very poor to the very rich. This may imply differences in nutrient limitations are associated with different SNRs. 4.4.4. Site Classification 4.4.4.1.^Delineation of Site Associations and Site Series Considering the previously distinguished plant associations, knowing the regional climate (implied biogeoclimatic subzone or variant), SMRs, and SNRs for each study stand, and using the existing B.C. Forest Service site classification as a guide (Meidinger, pers. comm.), all the study stands were classified into 13 site associations (Figure 4.14). Another two site associations, Cladina and Carex, which rarely support the growth of white spruce, were  recognized by the B.C. Forest Service and added here for completeness.  95  SMRs  SBSdw  SBSdk  SBSmw  SBSmk  SBSwk  Very poor to medium SNRs  *10 Cladina  VD  0-1/VP-M  0/VP-M 20 Shepherdia  MD  1-2/VP-M  2/VP-M 30 Pleurozium  SD to F  3-4/VP-P  3-5/VP-P 40 Hylocornium  M to VM  5-6/VP-P  6/VP-P 50 Sphagnum  W to VW  7-8/VP-M Medium to very rich SNRS  SD  F  31 Aster 3-4/M-VR  32 Spiraea 3-4/M-VR  33 Aralia  34 Viburnum  36 Petasites 5/M-VR  3-4/M-VR  3-4/M-VR  3-4/M-VR  42 Oplopanax  M  5/M-VR  41 Aulacomnium VM  43 Equisetum 6 6/M-VR  6/M-VR  W to VW  35 Gymnocarpium  *51 Carex 7-8/R-VR  Figure 4.14. An environmental matrix showing the site associations distinguished in the study in relation to climate (biogeoclimatic subzones), relative and actual soil moisture regimes, and soil nutrient regimes.  96  Correlations between the site associations distinguished in this study and those recognized by B.C. Forest Service are given in Appendix 2. The environment and productivity affinities of site associations are shown in Figure 4.14 and Table 4.20. A site association was only recognized when it could be distinguished from all other site associations by an exclusive range of climatic, soil moisture, and soil nutrient regimes, expressed by both categorical (Figure 4.14) or continuous variables (Table 4.20). Differences in white spruce site index among site associations were found. The Sphagnum (50) association has an obviously lower, but highly variable site index. This was expected since the very wet sites support open-canopy communities with extremely poor growth. Excluding associations 36 (representing only by 1 stand) and 50, ANOVA suggested a significant difference among the other ten associations. Tukey's multiple comparison indicated that site association 20 had a significant lower site index than the others and that association 42 had a higher site index than associations 30, 32, 40, 41, and 43. To form climatically uniform units, site associations were divided into site series according to subzones or variants. The resultant 30 site series and their mean site indices are given in Appendix 3. Using the site series that were represented at least by three stands, four comparisons of site index between series within a circumscribing association did not show any significant differences, suggesting a minor influence of climate (implied by different subzones or variants) on white spruce height growth (Figure 4.15). 4.4.4.2.^Delineation of Site Groups Within the montane boreal climate of the SBS zone, local soil nutrient, soil moisture, and soil aeration conditions are the three primary factors controlling vegetation potential and forest productivity. A new version of the  Table 4.20.^Means and standard deviations (in parentheses) of selected soil and stand properties stratified according to the site associations distinguished in the study stands. Characteristics Number of stands  20 (21)  30 (12)  31 (7)  32 (4)  33 (4)  Site association 35 34 36 (1) (3) (6)  40 (2)  41 (12)  42 (11)  43 (10)  Et/Emaxl  0.87 (0.05)  0.91 (0.04)  0.92 (0.02)  0.93 (0.04)  0.90 (0.03)  0.96 (0.04)  0.97 (0.03)  0.96 na  (0)  1  1 (0)  0.992 (0.02)  (0)  (0)  Depth to a gleyed layer na or prominent mottling (cm)  na  na  na  na  na  na  na  28.0  28.5 (7.1)  33.8 (6.7)  21.0 3 (5.7)  na (1.7)  Depth to the groundwater table (cm)  na  na  na  na  na  na  na  na  na  45.04 (8.5)  na  43.1 (9.0)  15.6 (13.3)  min-N (kg ha -1 )  27.4 (13.1)  25.1 (13.4)  46.7 (25.1)  35.2 (14.8)  53.8 (12.2)  34.5 (4.4)  37.7 (17.0)  41.2 na  22.6 (1.49)  81.7 (49.7)  65.3 (23.1)  56.8 (37.4)  65.8 (31.8)  C/N  37.3 (4.1)  40.2 (11.7)  32.6 (3.1)  37.8 (4.2)  33.1  (5.1)  28.9 (2.4)  29.1 (3.5)  33.1 na  36.0 (8.2)  31.0 (6.2)  27.0 (5.1)  28.7 (4.1)  30.9 (7.2)  Soil nutrient index  0.39 (0.13)  0.44 (0.12)  0.53 (0.10)  0.55 (0.09)  0.61 (0.11)  0.53 (0.04)  0.44 (0.11)  0.50 na  0.42 (0.06)  0.56 (0.13)  0.56 (0.11)  0.47 (0.13)  0.22 (0.05)  Foliar N ( cg g-1)  1.06 (0.12)  1.18 (0.07)  1.22 (0.05)  1.22 (0.07)  1.32 (0.05)  1.26 (0.03)  1.17 (0.04)  1.26 na  1.20 (0.04)  1.22 90.06)  1.22 (0.11)  1.24 (0.12)  1.04 (0.15)  Foliar nutrient index  0.87 (0.08)  0.92 (0.02)  0.94 (0.02)  0.92 (0.02)  0.96 (0.03)  0.92 (0.02)  0.93 (0.02)  0.96 na  0.92 (0.02)  0.91 (0.03)  0.94 (0.03)  0.92 (0.02)  0.80 (0.08)  Site index (m @ 50 yr b.h. age)  15.6 (1.4)  18.9 (1.8)  20.6 (1.5)  19.4 (1.4)  20.7  19.6  21.0 (1.3)  20.3 na  19.7 (0.4)  19.5  22.0 (1.5)  18.3 (1.7)  (3.5)  (2.3)  (2.0)  lActual/potential evapotranspiration; 2 only two stands; 3 only three stands; 4only two stands.  (2.5)  1  50 (9) 1  10.33  98  (a)  (b)  30  i 0  (c)  ^i  ^ SBSdw3 SBSdk SBSwk ^ Variant or subzone  30  i  +  SBSmk  ^  SBSmw  Subzone  30  0  SBSmw  ^  SBSwk  Subzone  SBSdw3^SBSdw1 Variant  Figure 4.15. Comparisons of white spruce site index among site series within the circumscribing site associations, showing the lack of effect of climate implied by biogeoclimatic subzone or variant: (a) Sheperdia, (b) Equisetum, (c) Oplopanax, and (d) Aulacomnium associations.  99  edaphic grid defined by SMARs and SNRs was constructed to show all possible combinations of ten SMARs and five SNRs-a total of 50 edaphic units (Figure 4.16). Stratification according to the SMARs and SNRs showed that the study stands occurred only in twenty-three units (Table 4.21). Despite a poor representation in certain combinations, the variation in white spruce site index showed a definite pattern. Considering differences in white spruce growth as well as affinities in ecological site quality among the edaphic units, the 50 individual edaphic units were combined into thirteen site groups (A through M), each representing a segment of a regional edaphic gradient (Figure 4.16). Edaphic combinations, represented by site groups A, B, D, H, and M, which might support white spruce growth, were not identified and sampled in this study. Thus, only eight of the possible 13 site groups were represented. Tukey's multiple test indicated that there were significant differences in site index between all groups except between C and K, F and I, and I and G. Site groups could be further combined into 4 productivity classes: good (F, G, and I), fair (C, D, E, K, and H), poor (A, B, and J), and marginal (L and M). Compared to site associations and site series, the site classification group introduced the qualitatively determined soil aeration regimes as a important factor affecting productivity, but it did not take into account the climatic variation within the SBS zone, as the effect of climate, implied by subzones or variants, on white spruce height growth was not significant. Site groups identified in this study were based solely on the present differences in white spruce site index (thus, they most likely, species-specific units), while the site associations consider the vegetation potential of ecological-equivalent sites at a climax stage.  ^  100  Soil nutrient regime VP^P VDa  ^ ^ ^ M R VR  A  B  C^n=25  D  I.)  E  tlo^MDa a.) ... DFr  SI=15.7^m (1.35)  E  F  G n=32 SI=20.0^m (1.10)  Fa  n=22 SI=21.8 m (1.37)  Ma VMa  H  Mr  J  o^Wr cn Wd  L  I  n=6 SI=12.5^m (1.69m) n=3 SI=6.07^m (1.40)  K  n=4 SI=20.2 m (0.85) n=11 S I=17.0 m (0.96)  M  Figure 4.16. An edaphic grid, defined by soil moisture-aeration regimes and soil nutrient regimes, showing the site groups distinguished in the study stands and white spruce site index (m @ 50 yr b.h. age).  101  Table 4.21. Means and standard deviations (in parentheses) of site index, and number study stands stratified according to soil moisture-aeration regimes (SMAR) and soil nutrient regimes (SNRs). Symbols for SMARs and SNRs are given in Figure 4.6 and Table 4.8, respectively. Soil moisture and aeration regime  Very poor  15.0  Poor  15.9  Soil nutrient regime Medium  Rich  Very rich  16.4  MDa  (1.72) n=7  DFr  (1.41) n=2  (0.71) n=2  (na) n=1  17.8 (na)  19.9 (1.01)  20.0  21.0  20.7  21.7  20.1  21.4  22.5  20.4  20.0  SDa  16.2  n=1  (1.12) n=6 16.0  n=8  17.5  (1.10) n = 12 (1.44) n=4  Fa  Ma  (0.44) n=7  19.7  (0.42) n=2  (0.93) n =5  16.5  Mr  Wd  (0.77) n=5 (1.30) n=7 (0.71) n=2  VMa  Wr  (2.76) n =3  (1.06) n=5 10.7  (0.97) n=2  13.8  (0.82) n=4 6.1  (1.40) n=3  17.4  (0.67) n=5  (1.41) n=7 (1.20) n=2  102  4.5. CONCLUSIONS 1.  Using the methodology of the BEC system, vegetation classification was a useful means for the first-step organization and interpretation of the 102 study stands in the SBS zone. The resulting vegetation units were unique in their floristic composition, ecological site quality, and, hence, white spruce productivity measured by site index.  2.  Et/Emax, depth to a gleyed layer or prominent mottling, and depth to a groundwater table were useful differentiating characteristics in classifying soil moisture regimes. Testing in relation to understory vegetation, foliar nutrients, and white spruce site index indicated significant differences among soil moisture regimes.  3.  For sites with freely draining soils, the simple forest water balance model provided an integrated index (Et/Emax) to compare the relative status of soil water supply during the growing season, which was useful for estimating the actual soil moisture conditions of the study stands. However, to determine water deficit precisely, the model has to be calibrated and verified.  4.  Water saturation status (implied by soil moisture regimes), drainage, slope, and soil texture were useful differentiating characteristics in classifying soil aeration regimes. Significant differences in site index were found among the soil aeration regimes.  5.^Soil min-N and C/N were useful differentiating characteristics in classifying soil nutrient regimes. Testing in relation to understory vegetation, foliar nutrients, and white spruce site index implied that minN and C/N provide a good index of available soil N as well as other soil nutrients.  103  6.^Within the montane boreal climate of the SBS zone, soil moisture, aeration, and nutrient regimes were important measures of ecological site quality. Either site units (groups, associations, and series) or these regimes in combination explained the variation in white spruce site index.  104  5. WHITE SPRUCE SITE INDEX IN RELATION TO MEASURES OF ECOLOGICAL SITE QUALITY 5.1 INTRODUCTION  Over the past several decades, numerous studies have focused on predicting site index from environmental variables because direct estimation of site index from crop trees is not always possible. As forest management has become increasingly intensified, predicting the growth responses to site treatments through environmental variables has been emphasized (Ford 1983, Stone 1984). The relationships between site index and environmental factors have been examined for a wide range of tree species around the world (e.g., Coile 1952, Carmean 1975, Haggland 1981). However, there have been only few comparable studies that have addressed white spruce. The development of reliable site index prediction equations for white spruce plantations has been limited by a lack of older stands and by the use of inadequate statistical methods (Harding 1982, Rauscher 1984). Very few soil-site studies have been reported for natural white spruce stands (Pluth and Corns 1983). In British Columbia, correlations between white spruce site index and various environmental factors have not been examined. The usefulness of ecological site classification in predicting white spruce site index has not been tested. The objectives of this chapter are (1) to examine relationships between white spruce site index and various environmental variables, including soil and topography, vegetation, and foliar nutrients and (2) to evaluate the usefulness of ecological site classification in addressing the relationship between white spruce site index and ecological site quality. Two hypotheses are tested: (1) different ecosystem attributes tend to reinforce each other in predicting site index and (2)  105  ecological stratification improves the relationships between site index and environmental variables. From the various regression models developed, several models are selected, tested, and compared. Their reliability for the prediction of site index in the case of lack of suitable tree growth data is discussed. In addition to the routine regression analysis, a semi-empirical model is developed within the framework of ecological site classification using limiting factor analysis. 5.2. LITERATURE REVIEW There are two reasons for evaluating forest productivity on the basis of ecological site quality: (1) the use of site index curves may be restricted because of the lack of suitable trees with which to establish height-age relationships and (2) proper assessment of the effects of environmental changes on forest productivity may be critically important to forest managers. As pointed out by Stone (1984), when site quality is viewed as possibly mutable rather than fixed, the question is not only "what is the Quality Class of this soil or location?", but "what could it become?". Indirect estimation of forest productivity can be made in two ways: (1) by predicting site index from environmental variables by regression models (i.e., the traditional soil-site study approach) and (2) by classifying forest sites into groups uniform in some measures of primary environmental factors (i.e., the ecological site classification approach). Since Coile's (1935) study on southern pines, extensive research has been conducted on predicting site index from environmental variables in North America as well as in other parts of the world (e.g., Coile 1952, Rennie 1962, Ralston 1964, Jones 1969, Carmean 1975, Spurr and Barns 1980, Hagglund  106  1981, Wang 1986b). All this work can be placed into a category called soil-site studies (Carmean 1975). A similar methodology has been followed (i.e., fitting a regression model using site index of certain tree species as dependent variable and one or more direct or indirect measures of the sites on which the tree species grows as independent variables). Sites or stands are located in a study area which covers a variety of conditions. Site index is either measured directly from stem analysis data or from a set of site index curves or tables. A variety of field and/or laboratory measurements are made of soil properties. Observations of topographic features (slope, aspect, elevation, etc.) are commonly recorded, as are designation of the site according to various vegetation and soil classification schemes. Understory vegetation and climate data are also collected in some studies. Although the general approach taken in most soil-site studies is similar, a considerable variation in methodology and results exists. While most soil-site studies have employed multiple regression techniques to predict forest productivity using topographic, soil physical, and soil chemical properties (e.g., Coile 1935, Eis 1962, Carmean 1972, Jokela et al. 1988, Schmidt and Carmean 1988), some studies have included understory vegetation (e.g., Corns and Pluth 1984, La Roi and Strong 1988, Klinka et al. 1989b, Green et al. 1989, Strong et al. 1991) and foliar nutrients in crop trees (e.g., Watt and Heinselman 1965,  Alban 1974, Radwan and DeBell 1980, Radwan and Harrington 1986, and Radwan et al. 1989) as independent variables. Frequently, soil properties which influence the quality and quantity of growing space of tree roots were found to be closely related to site index (Coile 1952, Carmean 1975). Topographic and climatic features were also found to be closely associated with site index, when topography and climate varied greatly within a study area.  107  Soil-site studies for white spruce plantations in the Lake States were reviewed in detail by Harding (1982) and Rauscher (1984). Kenety (1917) concluded that soil moisture is the key factor for good growth of white spruce plantations. Russell (1963) found that white spruce growth is retarded by a high groundwater table and poor drainage, thus implying that soil aeration is an important factor for white spruce growth. Wilde et al. (1965) concluded that available soil moisture is the major factor controlling the growth of white spruce plantations. Soil nutrients, on the other hand, have been reported as poor predictors of white spruce productivity (Russell 1963, Wilde et al. 1965), although soil phosphorus and nutrient synecological coordinates, together with slope, were identified to be the important discriminators of productivity groups (Harding 1982). Corns (1978) found that the mean annual height growth was negatively correlated with elevation. Pluth and Corns (1983) reported that white spruce site index increased linearly (towards poor drainage) along a soil drainage gradient. Vegetation has been often used to assess forest productivity in North America (e.g., Ilvessalo 1929, Heimburger 1934, Rowe 1956, Minore 1972, McLean and Bolsinger 1973, Pfister and Arno 1980, Corns and Pluth 1984, Klinka et al. 1989, Strong et al. 1991, Host and Pregitzer 1991) and Europe (e.g., Cajander 1926, Jahn 1982). Plants act as 'phytometers' of site quality by integrating many growth related factors which are difficult to measure directly (Major 1951, Daubenmire 1976). Therefore, understory species with a relatively narrow ecological amplitude can be good indicators of ecological site quality and, hence, forest productivity. Relationships between white spruce site index and understory vegetation have been studied in Canada and United States. Gagnon and MacArthur (1959) identified three understory vegetation types for white spruce plantations, and  108  showed productivity differences among the types. Bakuzis and Hansen (1962) noted differences in weighted site indices of white spruce according to differences in the synecological coordinates on the moisture-nutrient ecograph. Van Groenewood (1965) identified white spruce community types of different productivities using understory vegetation in association with soil texture. Corns and Pluth (1984) reported an increase in the R 2 value from 0.58 to 0.91 with the addition of individual understory species to a white spruce site index regression model. They indicated that vegetation attributes, used in conjunction with soil and site properties as independent variables in site index prediction, can account for a significant amount of the variability in white spruce site index. However, their model may be subject to potential prediction bias because of a large number of independent variables (9) relative to small sample size (30) (Verbyla 1986). La Roi and Strong (1988) found that spatially segregated understory community types usually have significantly different site indices. Strong et al. (1991) developed several white spruce regression models using individual understory species as independent variables, but the best model explained less than 50% of the variability in site index. The chance presence or absence of individual species on certain sites may be responsible for the unreliable prediction from these regression models. Although the use of understory plants in estimating ecological site quality has demonstrated some promise in managing northern forests (e.g., La Roi and Strong 1988, Sims et al. 1989, Klinka et al. 1989, Host and Pregitzer 1991), there are some limitations as individual species occurrence reflect understory light conditions, disturbance, and chance (Spurr and Barnes 1980, Spies and Barnes 1985). In order to improve this situation, indicator species groups (Klinka et al. 1989) or ecological species groups (Barnes et al. 1982) have been recommended instead of individual species.  109  Foliar nutrients have been widely used as a diagnostic standards to determine the nutritional status of forest stands. Since the nutritional status of forest stands reflects the availability of soil nutrients on a site, foliar nutrients have also been related to ecological site quality and forest productivity. Relationships between foliar nutrients and site index, that have been studied for black spruce (Watt and Heinselman 1965), western hemlock (Radwan and DeBell 1980), western redcedar (Thuja plicata Donn ex D. Don) (Radwan and Harrington 1986), and Pacific silver fir (Abies amabilis Dougl. ex Forbes) (Radwan et al. 1989), suggest that levels of some foliar nutrients are useful indicators of both ecological site quality and site index. However, the variability of foliar nutrients with age and the requirement of the presence of crop trees may prevent foliar nutrients from being consistent and primary predictors of site index. Although most soil-site studies achieved some success, many limitations have been recognized (e.g., McQuikin 1976, Stone 1978, Milner 1987, Green et al. 1989). The major limitations of soil-site studies can be categorized as either  technical or ecological. 1.^Technical Limitations Most soil-site models have relied on stepwise regression to select significant explanatory variables from a large set of candidate independent variables (e.g., Wall and Loewentein 1969, McGrath and Loewenstein 1975). Although this procedure has produced some models with a high R 2 (>0.70) and low standard errors of estimates, those models that have been tested against independent data have generally performed poorly (Broadfoot 1969, McQuilkin 1976). This is most likely caused by chance correlation, overfit, and multicolinearity. The more candidate independent variables used in developing a  11 0  model, the greater the possibility that the model will suffer from chance correlation. If the number of independent variables is large relative to the number of observations, then the danger of overfitting arises (McQuilkin 1976, Hagglund and Lundmark 1977, Verbyla 1986). The only way to detect chance correlation and overfitting is to test the validity of the model against an independent data set. However, most of these studies have failed to do so due to lack of independent test data (McQuilkin 1976). Multicolinearity, which causes unstable coefficients and affects the applicability of the model to new data sets, is very commonly associated with soil-site studies due to the presence of strong correlations among the site variables. Morzuch and Ruark (1991) suggested using principal component regression to mitigate the effect of multicolinearity. Most importantly, the system controlling tree growth is multifactorial and highly interactive as suggested in some recent studies (Stone 1984, Milner 1987, Monserud et al. 1990). Thus, it seems unlikely that a linear, additive model that includes only individual factors could possibly approximate the dynamics of a complex ecosystem (Stone 1978, Milner 1987). Some efforts towards developing models which are less dependent on empirical data from one geographical area and which are more "process-oriented", depending on a tree's morphological and physiological characteristics (Landsberg 1986), have been made recently. Gale (1987) and Henderson et al. (1990) integrated soil and site characteristics and their relationship to root growth and to the vertical root distribution of a tree to establish a soil productivity index (PI) model. Wickramasinghe (1988) derived potential growth indexes (PGIs) by modelling effective evapotranspiration, an indicator of interactions among atmospheric energy, potential evapotranspiration (PET), and soil moisture supply. Stand age was incorporated as an indicator of physiological efficiency in his model. Both PI and PGIs were highly correlated with tree growth, and were used to predict forest productivity.  111  2.^Ecological Limitations Most soil-site studies have had limited success in accounting for a significant portion of the variation in site index, particularly over a large area (Broadfoot 1969). As the size of a study area increases, it is likely that ecological complexity also increases (Stone 1984, Monserud et al 1990). Best results were generally obtained in studies limited to a small area where one or a few ecological factors were isolated while all others were relatively constant (Jones 1969). However, the results from a small area may not be applicable elsewhere. As an alternative, developing site index prediction models over a relatively large area based on ecological stratification has been frequently recommended (e.g., Stone 1978, Schmidt and Carmean 1988, Monserud et al. 1990, Strong et al. 1991). However, the hypothesis that ecologically stratifying a study area into smaller and more homogeneous areas (ecological strata) would lead to improved regression models seldom has been directly tested. Independent variables considered in most soil-site studies are not measures of the primary or causative factors that control tree growth (i.e., light, heat, soil moisture, soil nutrients, and soil aeration) (Rennie 1962, Jones 1969, Stone 1978, Schmidt and Carmean 1988); rather, they are measures of secondary factors (i.e., topographic, soil, and climatic factors) that are indirectly related to the causative factors. In consequence, the effects of these secondary factors will vary depending on the primary input of energy (light and heat), materials (nutrients, moisture, aeration, etc.), and compensating effects. Since the primary inputs vary across any sizable area, the link between site index and the effects of these secondary factors will likely be weak and changeable (Milner 1987, Monserud et al. 1990). Aber and Melillo (1984) stated that traditional soil-  112  site studies are limited by the nature of the database rather than the precision with which the data were gathered and analyzed. It is obvious that these problems cannot be overcome without improving the database or changing the modelling strategy. In terms of using empirical (statistical) models, increasing accuracy can only come from using either the direct measurement of primary (causative) factors or by integrating the measures of secondary factors into synoptic measures of primary factors (Stone 1984, Aber and Mellilo 1984, Klinka and Carter 1990). Since measuring primary factors is expensive and sometimes impractical, the database for most studies only includes various measures of the secondary factors. Searching for expressions of the primary factors through these secondary factors has raised some concerns in soil-site studies. An alternative approach (i.e., ecological site classification) has been suggested to predict forest productivity based on evaluation of ecological site quality in British Columbia (Klinka et al. 1989, Green et al. 1989, Klinka and Carter 1990). Similarly, habitat type classification has also been used to estimate forest productivity in western United States (Pfister et al. 1979). Individual observations or measurements of site properties are integrated into synoptic measures of climate, soil moisture, and soil nutrients (i.e., site units or biogeoclimatic units, SMRs, and SNRs). Therefore, forest productivity can be predicted from regression models using biogeoclimatic units, SMRs, and SNRs or site units as dummy variables. Within climatically uniform areas, regression models explained 84% of the total variance of Douglas-fir site index (Klinka and Carter 1990), 81% of the total variance of western hemlock site index (Kayahara 1992), 84% of the total variance of lodgepole pine site index (Wang 1992), 80% of the total variance of Sitka spruce site index (Pearson 1992), using SMRs and SNRs as independent variables,.  113  5.3. METHODS  5.3.1. Soil Capacity, Soil Nutrient, and Foliar Nutrient Indices  If individual variables, such as various measures of soil physical and chemical properties and foliar nutrients, were independent and their effects on tree growth were linear, a regression model would be ideal to describe the relationship between tree growth and environmental factors. However, various site variables are frequently multicolinear and their relationships with site index are often nonlinear in the nature. To mitigate the effect of multicolinearity and to express the nonlinearity, this study attempted to predict white spruce site index by integrating soil properties and foliar nutrients into simple indices. Soils with the same rooting depth may be very different in the quality and quantity of growing space for tree roots because organic matter content, coarse fragment content, soil texture, and depth of humus form vary with site. Coarse fragments, which occupy soil volume that could otherwise be occupied by fine particles capable of furnishing moisture and nutrients for tree growth (Schmidt and Carmean 1988), are not able to hold soil moisture and nutrients. Sand in soils is partially similar to coarse fragments, thus 30% of sand is assumed to be equivalent to coarse fragments (Wang 1989). In this study, a simple soil capacity index (SCI) was proposed to integrate those soil properties (i.e., major rooting zone depth in the forest floor and mineral soil, organic content, soil texture, and coarse fragment content) which are closely related to the capacity of the soil to hold water and nutrients. It was calculated as:  [5.1] SCI = Dms x (1 - OM / 100) x (1 - CF / 100) x (CLAY + SILT + 0.70 x SAND)/100 + Dms x (OM / 100) + Dff  114  where Dms (cm) is the depth of major rooting zone in the mineral soil; OM (%) is the organic matter content in the mineral soil; CF (%) is coarse fragment (>2 mm) content; CLAY, SILT, and SAND are clay (<0.002 mm), silt (0.002-0.05 mm), and sand (>0.05 but <2 mm) content (%), respectively; Dff is the depth of the major rooting zone in the forest floor or organic soil. Although nitrogen has been identified as the most limiting soil nutrient in this study, and SNRs characterized by min-N and C/N have been used to describe the multidimensional soil nutrient space (Chapter 4), other soil nutrients may also affect white spruce height growth on some study stands. To provide another alternative to approximate the multidimensional soil nutrient space, soil min-N, available P and S, and exchangeable K were integrated into a simple soil nutrient index (SNI). The selection of the four soil nutrients was based on consideration of the foliar nutrient status of the study stands, as well as their relationships with tree growth (measured by site index). The SNI was based on critical levels of soil nutrients, which were inferred from the response curves of foliar nutrients to soil nutrients, as well as from Wilde (1966). Relationships between foliar and soil nutrients were quantitatively described using Mitscherlich equations (Mitscherlich 1909) (Figure 5.1). Although the data resulted in relatively low, but statistically significant, R 2 's (0.10, 0.37, 0.17, and 0.19 for min-N, P, K, and S, respectively), the analysis suggested that the pattern of variation in soil and foliar nutrients in the study stands is related. The non-response levels were, approximately, 100 for min-N, 40 for S, 200 for K, and 20 for P (all in kg ha -1 ). Considering the much higher soil available P level (47 kg ha -1 ) suggested by Wilde (1966) for good to very good growth of white spruce, and that white spruce site index consistently increased  115  0.30  0.8  027  •  SS • • •  0.7  IS 01 •  ak' 0.24  •  •  •■■•• • •  •••••  •  •^•  5 0.21  •^SO^•  ••  0.6  co  O'  • • ••••  -  • •  ••^•  • • ••  0.  •^•  •  •  IS^• •  • •^• • • •  0.5  0.18  0.15 ^ 0  OS •  0.4 ^ 0  10^20^30^40^50^60^70 ^ Soil available P (kg ha')  •  100^200^300^400 Soil exchangeable K (kg ha')  1.50  0.12  1.43 0.11  1.36 1.29  0.10  122 1.4  •co —  0.09  0.08  1.01 0.94  0.07  0.87 0.80  0^50^100^150^200 250 ^ Soil min—N (kg ha -3 )  0.06  0  10^20^30^40^50 Soil available S (kg ha - ')  Figure 5.1. Relationships between selected soil and foliar nutrients in the study stands using Mitscherlich equations.  60  116  with foliar P in the study stands (Figure 5.2), the critical P level was set at 45 kg ha-1 . The SNI was calculated as: [5.2] SNI = [(min-N/100) x (P/45) x (S/40) x (K/200)]^(1/4)  SNI should be within the range from 0 to 1 (i.e., from no nutrients available to complete nutrient sufficiency). If min-N/100, P/45, S/40, and K/200 were greater than 1, they were set equal to 1. In a manner similar to the SNI, a foliar nutrient index (FNI) was developed by considering the foliar nutrient status of white spruce stands (Swan 1971, Ballard and Carter 1986) and relationships between foliar nutrients and site index (Figure 5.2). Six foliar macronutrients (N, P, K, S, Ca, and Mg) were used to calculate the FM. To facilitate the determination of their critical levels (i.e., the level beyond which increasing foliar nutrients, most likely, will result in no significant increase in site index), quadratic functions were selected to describe the relationships between site index and foliar nutrients. The critical levels for N, P, K, S, Ca, and Mg were determined to be 1.5, 0.28, 0.50, 0.10, 0.50, and 0.12 (all cg g- 1 %), respectively. The FNI was calculated as:  [5.3]^FNI = [(N/1.5) x (P/0.28) x (K/0.5) x (S/0.10) x (Ca/0.5) x (Mg/0.12)1 ^(1/6)  FNI should be in the range of 0 to 1, with 1 representing complete foliar macronutrient sufficiency. If N/1.5, P/0.28, K/0.5, S/0.10, Ca/0.5, and Mg/0.12 were greater than 1, they were set equal to 1.  117  s  MI 0.07 0.08^0.09^0.10^0.11 Foliar S (X)  0.12  0.4^0.5^0.8^0.7  0.9  Foliar Ca (X)  0.10^0.15 Foliar P (X)  0.8  020  Foliar Mg (%)  Foliar N (X)  Figure 5.2. Relationships between white spruce site index (m @ 50 yr b.h. age) and selected foliar nutrients (expressed as concentrations) in the study stands.  118  5.3.2. Regression Analysis Using all available soil, vegetation, and foliar data, regression analysis was applied to construct empirical models for predicting white spruce site index. Original measurements, scores of linear combinations of original measurements from PCA analysis, indices, and synoptic variables integrated from the original measurements were used as independent variables. In the first step, the variation of white spruce site index was examined by multiple regression analysis using either soil/topography (including SCI and SNI), vegetation, or foliar nutrients (including FNI) as independent variables. The best models from each set of independent variables were selected for presentation (first model set). In the second step, various combinations of soil/topography, vegetation, and foliar nutrient variables were used to develop a second set of models. These models were evaluated, and compared to the first set of models In the third step, the 102 study stands were stratified into three groups according to the following soil moisture conditions (see Table 4.2 in Chapter 4): (1)  sites with a groundwater table within 60 cm from the ground surface;  (2)  sites with a gleyed layer or prominent mottling within 50 cm from the ground surface; and  (3)^sites with none of the above. Using various combinations of soil/topography, vegetation, and foliar nutrient variables, regression models were developed for each of the three strata. These models were evaluated and compared with those based on unstratified data. In the fourth step, synoptic measures (i.e., SNRs, SMRs, or SARs; combinations of SNRs, SMRs, and SARs; and site associations) were used as dummy variables to develop a fourth set of regression models. These models were then evaluated and compared with those based on continuous variables.  119  Finally, several candidate models were selected from the four sets of models developed above to predict white spruce site index in the SBS zone. The selection was made based on error and residual analyses, and the model's applicability to the situation where crop stands are lacking. In order to test the selected candidate models, one third of the 102 study stands was selected as an independent data set using a stratified random procedure. All stands were first stratified according to SMR; then one third of the study stands in each SMR were randomly selected. In total, 34 stands were selected. The remaining 68 stands were used to recalibrate the selected candidate models. The coefficients of new models were compared to the models based on all 102 stands. The selected 34 stands were used to test both the original candidate models (n = 102) and the recalibrated candidate models (n = 68). The results from testing the original models were used to serve as a baseline to judge the results from testing the recalibrated models. 5.3.3. Limiting Factor Analysis It has been recognized that numerous individual environmental factors which affect tree growth and forest productivity may be summarized into five synoptic ones: light (L), heat (H), soil nutrient (SN), soil moisture (SM), and soil aeration (SA) (e.g., Pogrebynak 1930, Hills 1952, Major 1963, Bakuzis 1969, Krajina 1969). Consequently, the actual tree growth or productivity (AP) can be expressed as: [5.4] AP = f(L,H,SN,SM,SA) Among the five factors, light and heat, which come from solar radiation and thus may be modified by local topographic features, determine the climatic  120  limit to potential productivity (PP) of a given species. Soil nutrients, moisture, and aeration determine the actual productivity (AP) (i.e., the fraction of potential productivity that can be achieved on a given site). Therefore, equation [5.4] can be rewritten as: [5.5] AP = f 1 (L,H)*f2 (SN,SM,SA) or AP = PP*f2 (SN,SM,SA) If (1) solar energy inputs to all the study stands are approximately the same, (2) the canopy of dominant trees can absorb the same amount of light and have the same photosynthetic efficiency under nonlimiting conditions, and (3) soil nutrients, moisture, and aeration control tree growth, then equation (5.5) would give an ideal model for describing forest productivity of a site. Considering that (1) all study stands are situated within a climatically uniform area, (2) each stand is dominated by white spruce and is even-aged, and (3) site index is a measure of AP, then adopting the above three assumptions appears to be reasonable. The concept of limiting factors proposed by Odum (1971) was used to analyze the limitations imposed by soil nutrients, moisture, and aeration on white spruce productivity. To specify equation [5.5], the limitations imposed by soil nutrients, moisture, aeration, as well as other environmental factors were expressed in a multiplicative way: [5.6] AP =PP x Lm x Ln x La x Le where Lm, Ln, La are limitation coefficients for soil moisture, nutrient, and aeration, respectively; Le is the limitation due to other possible factors. Values of Lm, Ln, La, and Le are in the range of 0 to 1.  121  Since soil aeration and soil moisture are highly interdependent, the limitation imposed by soil aeration and moisture can be combined into one, Lma, [5.7] AP = PP x Lma x Ln x Le where AP, PP, Lma, Ln, and Le are as previously defined. If we know the potential productivity (PP) and the limitations imposed by other factors (Le), the actual productivity can be easily calculated from equation [5.7] using the estimates of Lma and Ln. However, it was difficult to get a reasonable estimate for PP and Le from the available data. Therefore equation [5.7] was rewritten as: [5.8] AP = C x Lma x Ln where C = PP x Le, which can be estimated from empirical or statistical procedures, represents the average maximum productivity in the study area without any limitation from soil moisture, aeration, and nutrients. Using site index (SI) as the measure of productivity, the above model becomes [5.9] SI = C x Lma x Ln Theoretically, C, Lma and Ln can be determined by process-based submodels; however, in this study they were empirically estimated by limiting factor analysis. Since C is the potential site index on the sites without limitations of soil moisture, aeration, and nutrients, the mean site index on these sites gives a reasonable estimate of C. Because six subzones or variants were included in this study, the effect of climate (implied by subzones or  122  variants) on C was also evaluated using study stands on zonal sites (i.e., sites with a 3 or 4 relative SMR and a medium SNR; Pojar et al. 1987). Holding SNR constant, the limitation coefficient of soil moisture and aeration (Lma) on the optimum SMAR was set to 1. Lma's on the other SMARs were estimated by comparing their mean site indices to that on the optimum SMAR. Using all stands where soil moisture and aeration were not limiting, the limitation coefficient of soil nutrients (Ln) on the optimum SNR was set to 1. Similarly, Ln's on the other SNRs were estimated by comparing their mean site indices to that on the optimum SNR. Using these calibrated coefficients, white spruce site index on each study stand was calculated using equation [5.91, and compared with the measured site index. The prediction precision of this model was then compared to that of the regression models developed in this study.  5.4. RESULTS AND DISCUSSIONS 5.4.1. Site Index in Relation to Soil, Understory Vegetation, and Foliar Nutrients Simple correlations between white spruce site index and individual soil chemical and physical properties were examined (Table 5.1). Both concentration (separately for forest floor and mineral soil) and quantity (forest floor plus mineral soil) of measured soil nutrients were used. Among the nine soil chemical properties, soil total N (kg ha -1 ) was the most strongly and positively correlated with site index. Among the seven soil physical properties, the depth of the major rooting zone (including both mineral and organic soils) was the most significant. Overall, all correlations were weak for predicting site index, although some of them were statistically significant.  123  Table 5.1. Simple correlation coefficients between white spruce site index and selected soil chemical and physical properties. Coefficients marked by asterisk (*) are significant (a = 0.05).  Property  Total  Forest floor  Mineral soil  pH  -0.03  0.17  C/N  -0.00  -0.16  -0.18  exchangable Ca  -0.02  0.19  0.34*  exchangable Mg  -0.05  0.12  0.24*  exchangable K  0.26*  -0.14  0.16  total N  -0.13  0.24*  0.42*  min-N  -0.06  0.10  0.13  available P  0.25*  -0.12  0.07  available S  0.19  0.11  0.31*  0.02  0.47*  0.58*  Chemical properties  Physical properties  Depth of major rooting zone (RD) Coarse fragment (CF)  0.20  Slope  0.07  Bulk density (BD)  0.25*  -0.07  Sand  -0.10  Clay  0.05  Color value  -0.11  124  Since simple correlation can only be used to detect linear relationships, scatter plots were made to examine possible nonlinear relationships between site index and soil chemical properties. Again, no strong nonlinear relationships were found except for the depth of major rooting zone. The logarithmic transformation of this variable improved its correlation with site index (r = 0.70 versus r = 0.58). Multiple regression analysis was used to examine the relationships between site index and selected combinations of soil properties. The three best regression models using chemical properties of forest floor, mineral soil, or both (expressed as concentration) explained only a very small portion of the total variance (18%, 10%, and 14%, respectively). The best regression model using total soil nutrients (kg ha -1 ) explained 26% of the total variance (equation [1], Table 5.2). The best regression model using soil physical properties explained 57% of the total variance (equation [4], Table 5.2). The best regression model based on measured soil physical and chemical properties was equation [7] (Table 5.2) which used a combination of soil nutrients and soil physical properties and explained 63% of total variance in white spruce site index. As an attempt to mitigate the possible multicollinearity among soil chemical and physical properties, PCA was used to generate new independent variables (i.e., PCA scores calculated from linear combinations of original soil chemical and physical properties). Regression analysis using PCA scores from the selected PCA axes produced very similar results, in terms of R 2 and SEE, to those using original soil chemical and physical properties. These models developed explained 20% (forest floor properties), 10% (mineral soil properties), 13% (both mineral soil and forest floor, expressed as concentration), 28% (both mineral soil and forest floor, expressed as kg ha -1 ), and 55% (soil physical properties) of the total variance, respectively.  125  Table 5.2.  Selected models for the regression of white spruce site index on selected measures of soil and topography. All models are significant (p < 0.001, n = 102). SNI and SCI designate soil nutrient index and soil capacity indices, respectively. Other symbols as in Table 5.1.  Soil chemical properties [1]  SI = 14.72 - 0.001(TN) - 0.044(min-N) + 0.019(P) + 0.082(S) Adjusted R 2 = 0.26^SEE = 3.13 m  [2]  SI = 24.12 - 2.389/(SNI) Adjusted R 2 = 0.53^SEE = 2.50 m  [3]  SI = 33.51 - 2.446/(SNI) - 1.062(PHfI) - 0.119(C/N) Adjusted R 2 = 0.58^SEE = 2.37 m  Soil physical properties [4]  SI = -5.92 + 6.6621n(RD) - 0.039(SLOPE) + 21.978(BDff) - 2.522(CF) Adjusted R 2 = 0.57^SEE = 2.39 m  [5]  SI = 19.51 - 70.733/(SCI) + 0.116(SCI) Adjusted R 2 = 0.56^SEE = 2.42 m  [6]  SI = 7.11 - 47.984/(SCI) + 3.5331n(RD) + 17.745(BDff) Adjusted R 2 = 0.62^SEE = 2.24 m  Soil physical and chemical properties [7]  SI = -2.43 + 0.000951(TN) - 0.034(min-N) + 0.052(S) + 4.9451n(RD) + 17.558(BDff) Adjusted R 2 = 0.63^SEE = 2.22 m  [8]  SI = 9.07 - 1.484/(SNI) + 3.8641n(RD) - 0.033(SLOPE) Adjusted R 2 = 0.65^SEE = 2.17 m  [9]^SI = 6.37 - 33.490/(SCI) + 8.141(SNI) + 2.9931n(RD) Adjusted R 2 = 0.67^SEE = 2.10 m  126  Relationships between white spruce site index and the two integrated soil indices (SNI and SCI) were also examined. Both indices showed a nonlinear relationship with site index (Figure 5.3). Using SNI as the independent variable, equation [2] explained 53% of total variance in site index (Table 5.2). Compared to the best model using only soil chemical properties (equation [1]), equation [2] greatly improved the power of soil nutrients to explain the variability of white spruce site index by integrating the individual soil nutrients into a single soil nutrient index. By adding two other soil chemical properties (forest floor pH and C/N), equation [2] was slightly improved (equation [3]). Equation [5], using SCI as the independent variable, explained 56% of the total variance in site index, which is similar to that explained by equation [4] using four individual soil physical properties. By adding two other soil physical properties, equation [5] was slightly improved, explaining 62% of the total variance in site index (equation [6]). Using SNI and SCI together with all the other measured soil/topographic variables (equations [8] and [9]) improved the fit slightly compared to equation [7]. The relationship between site index and understory vegetation was examined using both analytical and categorical measures as independent variables. Six models were selected for presentation; all were significant at p < 0.001 (n = 102) (Table 5.3). The models using the frequencies of ISGs for soil moisture, nitrogen, or both as predictors (equations [10], [11], and [12]) explained 27, 49, and 53% of the total variance in site index, respectively. The model using plant associations as dummy variables (equation [13]) explained 56% of the total variance in site index, which is slightly more than that explained by the best model using frequencies of ISGs. As most of the 230 understory plant species identified in this study were not present in all 102 stands, two models using PCA scores (i.e., linear combinations of individual  127  (a)  Soil nutrient index  (b  )  0^10^20^30  ^  40  ^  50  Soil capacity index  Figure 5.3. Relationships between white spruce site index and (a) soil nutrient index and (b) soil capacity index.  128  Table 5.3. Selected models for the regression of white spruce site index (m @ 50 yr b.h. age) on selected understory vegetation variables.  [10]  SI = 7.735 + 0.104NITR1 + 0.173NITR3 Adjusted R2 = 0.27^SEE = 3.11 m where NITR1 and NITR3 as in Table 4.7.  [11]  SI = 10.818 + 0.083MOST3 + 0.137MOIST4 Adjusted R2 = 0.49^SEE = 2.61 m where MOIST3 and MOIST4 as in Table 4.7.  [12]  SI = 6.738 + 0.080MOIST3 + 0.106MOIST4 + 0.053NITR1 + 0.086NITR3 Adjusted R 2 = 0.53^SEE = 2.52 m  [13]  SI = 8.95 + 10.95ARAL + 9.36ATHY + 12.92DISP + 3.22EQUI + 10.07PETA + 6.96SHEP + 11.38STRE + 0.000BETU Adjusted R 2 = 0.56^SEE = 2.42 m where ARAL, ATHY, DISP, EQUI, PETA, SHEP, STRE, and BETU as in Table 4.6.  [14]  SI = 15.441 + 0.202PCA1 + 0.130PCA2 + 0.599PCA3 + 0.159PCA5 - 0.298PCA6 + 0.352PC12 + 0.270PCA13 - 0.285PCA16 Adjusted R 2 = 0.61^SEE = 2.30 m where PCA analysis is based on diagnostic specis (Table 4.6).  [15]  SI = 14.12 -0.059PCA1 +0.106PCA2 - 0.522PCA3 + 0.080PCA4 + 0.322PCA6 + 0.189PCA9 + 0.122PCAl2 + 0.288PCA14 + 0.281PCA15 -0.359PCA17 +0.227PCA19 Adjusted R2 = 0.74^SEE = 1.85 m where PCA analysis is based on all specis (Appendix 1).  129  species covers) as independent variables were developed (Table 5.3). The model using PCA scores from the diagnostic species (Table 4.6) explained a slightly larger portion of the total variance in site index than equation [13]. The model using PCA scores based on all species (equation [151) explained a significantly larger portion of the total variance than any other vegetation model, although 11 PCA axes derived from 230 plant species were involved. This model, albeit impractical, implies a strong relationship between understory vegetation and ecological site quality. Regression analysis using the cover of individual understory species as independent variables was also explored. In contrast to the strong relationships reported by Corns and Pluth (1984) and Strong et al. (1991), no significant relationships were found in this study. Given that this study determined relatively strong relationships between white spruce site index and other measures of understory vegetation, the failure of single understory species as predictors is likely the result of the mid-seral stage in the development of understory vegetation imposing a strong chance presence or absence on individual understory species. Quadratic functions were used to describe the relationships between white spruce site index and foliar macronutrient concentrations (Figure 5.2). Foliar N, P, K, S, Ca, and Mg explained 36%, 56%, 30%, 30%, 10%, and 14%, respectively, of the total variance in site index, with all models being statistically significant (p < 0.01) (Figure 5.2, Table 5.4). No significant relationships were found between site index and foliar micronutrients.  130  Table 5.4. Selected models for the regression of white spruce site index (m 50 yr b.h. age) on foliar nutrients and foliar nutrient index (n = 102). FN, FP, FS, and FNI represent foliar nitrogen, phosphorus, sulphur, and foliar nutrient index, respectively. [16]  SI = -21.93 + 52.605(FN) - 15.509(FN) 2 Adjusted R2 = 0.36^SEE = 2.93 m  [17]  SI = -25.45 + 288.889(FP) - 434.228(FP) 2 Adjusted R2 = 0.56^SEE = 2.42 m  [18]  SI = 12.49 + 24.431(FN) - 0.596(FN/FS) - 3.060(FN/FP) Adjusted R2 = 0.62^SEE = 2.24 m  [19]  SI = -32.10 + 276.800(FP) - 444.32(FP) 2 + 107.700(FS) Adjusted R2 = 0.64^SEE = 2.19 m  [20]  SI = -30.50 + 53.780(FNI) Adjusted R 2 = 0.63^SEE = 2.21 m  [21]  SI = -23.58 + 52.182(FNI) - 1.096(FN/FP) Adjusted R 2 = 0.66^SEE = 2.14 m  Using combinations of foliar nutrients and their ratios as independent variables, equations [18] and [19] explained a larger portion of the total variance in site index than that explained by any model using only a single foliar nutrient as the predictor. The model using foliar nutrient index (equation [20]) also explained a larger portion of the total variance in site index than any of the models using individual foliar nutrients; its performance was comparable to the models using a combination of nutrients and ratios. The best model using foliar nutrients to predict site index was equation [21], which explained 66% of the total variance in site index (Table 5.4).  131  Using various combinations of soil/topography, vegetation, and foliar nutrient measures as independent variables considerably improved the regression models over those that used the same measures individually (Table 5.5). The combined models that used combinations of several soil, vegetation, and foliar nutrient variables (equations [28] and [29]) explained 83% and 84% of the total variance in site index, respectively. Other models, which used combinations of soil and vegetation variables (equations [22] and [23]), soil and foliar nutrient variables (equations [24] and [25]), or vegetation and foliar nutrient variables (equations [26] and [27]), also showed some improvement (0.74 > R 2 < 0.80) over models based on soil, vegetation, or foliar nutrient variables alone. It appears that information from the various ecosystem attributes tends to reinforce each other in explaining the variability in site index. By using information from different ecosystem attributes in developing regression models, the precision of site index prediction can be improved. Regression models, using various combinations of soil/topographic, vegetation, and foliar nutrient variables, were also developed for the three strata (groups) of study stands defined according to soil moisture conditions (Table 5.6). In general, the models developed for stratum (a) and (b), using variables from soil alone, soil and vegetation, or soil, vegetation, and foliar nutrients as predictors, showed a stronger relationship with site index (in terms R 2 and SEE) compared to the models developed using unstratified data. The SEE's for the stratified and unstratified models were < 1.73 versus 2.10 (Table 5.2) using soil variables alone as predictors; < 1.52 versus > 1.76 (Table 5.5)] using soil and vegetation variables as predictors; < 1.23 versus > 1.45 (Table 5.5) using soil, vegetation, and foliar nutrient variables as predictors. The models developed for stratum (c) had improved SEE's but had slightly lower R 2 's. Adding either  132  Table 5.5. Selected models for the regression of white spruce site index (in @ 50 yr b.h. age) on various combinations of soil, understory vegetation, and foliar nutrient variables. All models are significant at p < 0.001 (n = 102). Symbols for independent variables are given in Tables 5.1, 5.3, and 5.4. Soil/topography and vegetation variables [22]  SI = -1.74 + 3.4981n(RD) + 22.892(BDff) + 0.000612(TN) - 0.0271(min-N) + 0.0446(S) + 0.0456(MOIST3) + 0.0815(MOIST4) Adjusted R2 = 0.75^SEE = 1.82 m  [23]  SI = 4.46 - 22.205/(SCI) + 6.588(SNI) + 2 .3421n(RD) + 0.04 l(MOIST3) + 0.054(MOIST4) + 0.030(NITR3) Adjusted R2= 0.77^SEE = 1.76 m  Soil/topography and foliar nutrients variables [24]  SI = 25.66 + 2.6311n(RD) + 0.00914(TN) - 0.0195(min-N) + 36.972(FNI) Adjusted R 2 = 0.79^SEE = 1.67 m  [25]  SI = -3.83 + 3.1671n(RD) + 0.000922(TN) - 0.0228(min-N) + 16.882(FN) - 0.2858(FN/FS) 1.229(FN/FP) Adjusted R 2 = 0.80^SEE = 1.65 m  Vegetation and foliar nutrients variables [26]  SI = -14.91 + 0.0241(MOIST3) + 0.0591(MOIST4) + 0.0202(NITR3) + 38.601(FNI) 1.102(FN/FP) Adjusted R 2 = 0.74^SEE = 1.86 m  [27]  SI = 7.40 + 0.0403(MOIST3) + 0.0633(MOIST4) + 0.0282(NITR11) + 0.0471(NITR3) + 16.981(FN) - 2.203(FN/FP) - 0.342(FN/FS) Adjusted R 2 = 0.74^SEE = 1.87 m  Soil/topography, vegetation, and foliar nutrients: [28]  SI = -20.11 + 2.3691n(RD) + 15.086(BDff) + 0.000698(TN) - 0.0212(min-N) + 0.0257(MOIST3) + 0.0500(MOIST4) + 27.767(FNI) Adjusted R2 = 0.83^SEE = 1.52 m  [29]  SI = -1.31 + 2.2701n(RD) + 0.000714(TN) - 0.0163(min-N) + 3.582(SNI) + 0.0298(MOIST3) + 0.0444(MOIST4) + 12.171(FN) - 0.988(FN/FP) - 0.214(FN/F'S) Adjusted R2 = 0.84^SEE = 1.45 m  133  Table 5.6. Selected models for the regression of white spruce site index (m @ 50 yr b.h. age) on combinations of soil, vegetation, and foliar nutrient variables using the stratified data: (a) sites with groundwater table within 60 cm (n = 18), (b) sites with a gleyed layer or prominent mottling within 50 cm (n = 25), and (c) other sites (n = 59). Symbols for independent variables are given in Tables 4.11, 5.2, 5.3, and 5.4. Soil/topography variables [30a]  SI = 12.58 + 0.265(GW) - 0.197(C/N) Adjusted R2 = 0.88^SEE = 1.73 m  [30b] SI = -18.28 + 6.8501n(GLEY) + 1.9151n(TN) Adjusted R 2 = 0.83^SEE = 0.99 m [30c] SI = -9.10 + 8.767(SNI) + 25.872(EtlEmax) Adjusted R 2 = 0.43^SEE = 2.07 m Soil/topography and vegetation variables [31a]  SI = 7.50 + 0.242(GW) - 0.173(C/N) + 0.078(NITR1) + 0.061(NITR3) Adjusted R 2 = 0.91^SEE = 1.52 m  [31b] SI = -13.74 + 4.9011n(GLEY) + 2.3051n(TN) - 0.044(MOIST5) Adjusted R 2 = 0.90^SEE = 0.77 m [31c] SI = 7.04 + 0.000417(TN) + 0.065(RD) + 31.90(BDFF) + 0.0475(MOIST3) + 0.0956(MOIST4) Adjusted R 2 = 0.57^SEE = 1.78 m Soil/topography, vegetation, and foliar nutrient variables [32a]  SI = -5.35 + 0.166(GW) + 12.763(FN) Adjusted R 2 = 0.94^SEE = 1.23  [32b]  (model not improved by including foliar nutrient varaibles)  [32c] SI = 5.57 + 0.00055(TN) + 0.0525(RD) + 19.675(BDFF) + 0.0308(MOIST3) + 0.0567(MOIST4) + 13.328(FN) - 2.21(FN/FP) Adjusted R2 = 0.71^SEE = 1.46 m  134  vegetation, foliar nutrients, or both to the original edaphic variables significantly improved the models developed for each stratum, with SEE's decreasing from 1.73 (equation [30a]) to 1.23 m (equation [32a]) for stratum (a), from 0.99 (equation [30b]) to 0.77 m (equation [31b]) for stratum (b), and from 2.07 (equation [300) to 1.46 m (equation [320) for stratum (c). It is evident that a general stratification of stands provided important information that could help improve model fit. 5.4.2. Site Index in Relation to Synoptic Measures of Ecological Site Quality As shown in the previous Chapter, SMRs, SNRs, SARs, and site units were closely related to white spruce site index. Assuming that the sampling procedure was unbiased and an adequate sample size was used for each class, then the sample means of site index for each SMR, SNR, SAR, and site unit are unbiased estimations of the population means. These sample means can be directly used to predict the values for new samples from the same population. In order to compare the results with the other developed models, regression models were developed using SNRs, SMRs, SARs, SMARs, and site associations as dummy variables (Table 5.7). The SMR model (equation [34]) explained 81% of the total variance in site index and was the best model that used a single measure of ecological site quality. This agrees with the work of Kenety (1917), Russell (1963), and Wilde et al. (1965) who concluded that soil moisture is the single best predictor of white  spruce site index. As SARs also explained a large proportion (R 2 = 0.50) of the total variance in site index, soil aeration might be an important factor affecting white spruce height growth. In contrast to SMRs, SNRs only explained a minor proportion (R 2 = 0.23) of the total variance in site index. This was attributed to the variation in soil moisture, particularly within the poor and medium SNRs.  135  Table 5.7.  Selected models for the regression of white spruce site index (m @ 50 yr b.h. age) on categorical measures of ecological site quality soil nutrient regimes (SNRs), soil moisture regimes (SMRs), soil aeration regimes (SARs), soil moisture-aeration regimes (SMARs), and site associations. Symbols for SNRs, SMRs, SARs, SMARs and site associations are given in Tables 4.8, 4.2, and 4.5, and Figures 4.6 and 4.14, respectively.  [33]  SI = 17.46 - 1.97(VP) - 0.08(P) + 0.00(M) + 2.95(R) + 4.49(VR) Adjusted R2 = 0.23^SEE = 3.17 m  [34]  SI = 6.07 + 9.61(MD) + 13.41(SD) + 15.18(F) + 15.25(MOIST) + 11.82(VM) + 6.49(W) + 0.00(VW) Adjusted R2 = 0.81^SEE = 1.56 m  [35]  SI = 6.07 + 13.26(a) + 9.47(r) + 0.000(d) Adjusted R2 = 0.50^SEE = 2.58 m  [36]  SI = 6.07 + 9.64(MDa) + 13.89(SDa) + 10.31(DFr) + 15.18(Fa) + 15.25(Ma) + 14.11(VMa) + 10.90(Mr) + 6.49(Wr) + 0.00(Wd) Adjusted R2 = 0.87^SEE = 1.33 m  [37]  SI = 6.07 - 1.78(VP) - 0.81(P) + 0.00(M) + 1.32(R) + 2.58(VR) + 10.57(MD) + 13.74(SD) + 14.44(F) + 14.03(MOIST) + 10.79(VM) + 6.82(W) + 0.00(VW) Adjusted R2 = 0.87^SEE = 1.33 m  [38]  SI = 6.07 + 10.41(MDa) + 14.10(SDa) + 11.18(DFr) + 14.56(Fa) + 14.28(Ma) + 12.56(VMa) + 10.35(Mr) + 6.76(Wr) + 0.00(Wd) - 1.51(VP) - 0.67(P) + 0.00(MED) + 1.12(R) + 1.99(VR) Adjusted R2 = 0.90^SEE = 1.17 m  [39]^SI = 10.32 + 5.25(SA20) + 8.54(SA30) + 10.32(SA31) + 8.78(SA32) + 9.80(SA33) + 9.28(SA34) + 10.71(SA35) + 9.98(SA36) + 9.38(SA40) + 9.32(SA41) + 11.63(SA42) + 7.67(SA43) + 0.00(SA50) Adjusted R2 = 0.72^SEE = 1.93 m  136  This could be one of the reasons why soil nutrients determined in routine chemical analysis have been reported as poor predictors of white spruce site index (Russell 1963, Wilde et al. 1965, Payandeh 1986). Improvement was found when various combinations of SMRs, SARs, and SNRs were used for predicting white spruce site index. The SMAR plus SNR model (equation [38]) is the best categorical model (R 2 = 0.90 and SEE = 1.2 m) among those developed in this study, followed by the SMR plus SNR model (equation [37]) and the SMAR model (equation [36]). These models accounted for 87% of the total variance in site index and had SEE's < 1.3 m. Given that routine identification of SMRs, SNRs, and SARs could be easily carried out by forestry field personnel, these models have promising operational potential. The site association model (model [39]) explained only 72% of the total variance in site index. This reflected the fact that variation in site index within the site associations was relatively large compared to that within individual combinations of either SMRs and SNRs, SMRs and SARs, or SMARs and SNRs. 5.4.3. Selecting and Testing Prediction Models Regression analysis demonstrated that white spruce site index was closely related to (1) direct measures of various ecosystem attributes (i.e., soil/topography, vegetation, and foliar nutrients), (2) synoptic measures inferred from soil/topography and vegetation (i.e., SMRs, SNRs, SARs), and (3) plant and site associations. The majority of the models presented explained more than 70% of the total variance in site index. However, not all of these models are suitable for predicting site index because some of the independent variables used in the models may be very difficult or impractical to obtain or simply unavailable under certain circumstances. For example, foliar nutrient data cannot be obtained in the absence of suitable stands where site index prediction is needed.  137  Understory vegetation changes with disturbance and time, and is not likely to be an efficient and stable predictor, especially when using individual species or PCA scores derived from all species in the stands. This situation may be avoided by using the frequency of ISGs, either in improving the success of prediction models based on soil/topography or in assisting identification of SMRs, SNRs, and site associations. Therefore, the most suitable models for predicting white spruce site index are those that use either individual measures of soil/topography, ISGs, synoptic measures of ecological site quality (i.e., SMRs, SNRs, SARs), or site associations (Table 5.8). Of the three types of continuous models selected for comparison, the soil/topography model (equation [9]) was the least precise (Figure 5.4). The largest standard error of estimate was less than 5.5 m and most of the errors were within 4 m. Adding the frequency of ISGs (equation [23]) reduced the mean absolute error from 1.72 to 1.41 m (Figure 5.5). The largest error of estimate was less than 4 m and most of errors were within 3 m. The three stratified models (equations [30a], [30b], and [30c]) used two predictors, one giving a measure of soil moisture and the other of soil nutrients. For each model, a threedimensional graph was used to show relationships between site index and the two variables (Figures 5.6, 5.7 and 5.8). The overall performance of the stratified models was similar to the soil/topography plus vegetation model (equation [23]). Residual analysis indicated that equations 30a and 30b provided better estimates of site index for the two respective strata than equations [9] and [23]. The largest error was less than 3 m and most of the errors were within 2 m (Figures 5.9 and 5.10). However, the model for upland sites (equation [30c]) had much larger errors compared to the models for waterlogged (equation [30a]) and gleyed sites (equation [30b]). The largest error was about 4 m, and most of the  138  Table 5.8. Some statistics of the four candidate models selected for the prediction of white spruce site index (m @ 50 yr b.h. age) from measures of ecological site quality. Model  N  R2  SEE (m)  Error range (m)  Mean error (m)  Equation [9] from Table 5.2  102  0.67  2.10  -5.4 - 4.8  1.721 (1.13) 2  Equation [23] from Table 5.5  102  0.77  1.76  -3.9 - 3.9  1.41 (0.95)  Equation [30a]  18  0.88  1.73  Equation [30b] from Table 5.6  25  0.83  0.99  -4.0 - 3.6  1.44 (0.98)  Equation [30c]  59  0.43  2.07  Equation [38] from Table 5.7  102  0.90  1.17  -3.2 - 2.2  0.85 (0.67)  1 The mean of absolute errors (measured SI - predicted SI). 2 Standard deviation of absolute error.  139  (a)  6 •  4 •  •  • •  •  *.^• • •II.  ••  —4 --6  0  10  ^  •  20  ^  30  Estimated site index (m)  (b  )  30  0  0  10  20  30  Estimated site index (m)  Figure 5.4. Residual analysis for the selected soil/topographic model (equation [9]): (a) residual versus estimated site index and (b) measured versus estimated site index (m @ 50 yr b.h. age).  140  (a)  6  4  'So^.  2  • #^ •^• • •^••:"..  •  $ ••••t••^. • 05 ,• •• • ."1..  .  to ..^. . 4• • . • •• .^. •••• ,  • • •  .  '..  •  ^  ••  —4  —6  0  10  20  30  Estimated site index (m)  (b)  30  E ,1 20 1 -)c) z co ..-) Cl) ^c) co ;... F, 10 co co .  0  0  10  20  30  Estimated site index (m)  Figure 5.5. Residual analysis for the soil/topography plus ISG model (equation 23): (a) residual versus estimated site index and (b) measured versus estimated site index (m @ 50 yr b.h. age).  141  20  10  Figure 5.6. White spruce site index in relation to soil C/N ratio and depth to groundwater table (equation [30a]) showing the regression surface, isolines, and distribution of measured site index (m @ 50 yr b.h. age)  142  Figure 5.7. White spruce site index in relation to soil total nitrogen and depth to gleyed layer or prominent mottles (equation [30b]) showing the regression surface, isolines, and distribution of measured site index (m @ 50 yr b.h. age).  143  30  20  4% 1* *tl ol 4' .•••• rift **4444.41:111 0*‘^ %^ •* •••• • " 4 • ••••■• I sd . 4::::":" 1 C:::::$ W.: rt,i, t) % s•••••:•:‘ • 'A, . 4 to0• 1,. .• : %%% 4 4 4 •• ■•:••••••• ! '1:it *, *♦0 • ••• ..••. •♦• • ••• 0♦4 . ‘ 41/4. t:** soyit itt ..... . ..♦ ."so.* sti": ";44::• . ....... ■ ■•*ft..:.•.•• •♦:••■!•.1♦1". -  VStrif  ,  •• ftlyttl. 1 eff:::::4::::::t♦ • *S"' • •♦••••:: pall:It). .1 ♦ .  . tr t 1,44:4;* Ve.: . ,igi _ . f•' t:4! p4 :**, •• _ i 7'1 I ar•I a' awarwhytd. • • 4 lig "0"*°••• . oh^, . z.-.- 1^*10,' • • SA 09'sb "'N -;- -,. 0› .c ( . 3 .° .6.-. - s %-. c ~. -4c:.^ G'  Figure 5.8. White spruce site index in relation to Et/Emax ratio and soil nutrient index (equation [306) showing the regression surface, isolines, and distribution of measured site index (m @ 50 yr b.h. age).  EST  144  (a)  6 4 2  —4 —6  0  10  ^  20  ^  30  Estimated site index (m)  (b)  30  0  0  10  20  30  Estimated site index (m)  Figure 5.9. Residual analysis for the stratified model (equation [30a]): (a) residual versus estimated site index and (b) measured versus estimated site index (m © 50 yr b.h. age)  145  (a)  6  4  2 711  .  0  En  a4 cu  —2  —4  —6  16^17^18^19 20 21 22 23 24 Estimated site index (m)  (b)  30  25 a) .d 20 -0 s. CO  15  10  10^15^20^25  ^  30  Estimated site index (m)  Figure 5.10. Residual analysis for the stratified model (equation [30b]: (a) residual versus estimated site index and (b) measured versus estimated site index (m @ 50 yr b.h. age).  146  errors were within 3 m (Figure 5.11). This was better than equation [9] and similar to equation [23]. The SMAR plus SNR model (equation [38]), using SMARs and SNRs as dummy variables, was the most precise in predicting white spruce site index. A three-dimensional graph was used to describe the regression surface, the isoline of site index, and the distribution of site index along the gradients measured by SMARs and SNRs (Figure 5.12). Except for one plot, all the errors of estimates were within 2 m, as shown by the residual analysis (Figure 5.13). The mean absolute error of site index prediction was 0.85 m, the least among the four types of models selected for comparison. The candidate models were recalibrated for testing using the data from 68 randomly selected stands (Table 5.9). Compared to the original models (Table 5.8), the performance of the continuous models (Equations [9], [23], and [30]) was slightly improved (i.e., they had higher R 2 's, lower SEE's, and lower mean absolute errors). The recalibrated categorical model performed similarly to the original model. The results of testing the models using the data from the remaining 34 stands are given in Table 5.10. The range of the residuals and the mean error of estimates were close to those presented for the recalibrated models in Table 5.9. However, biases were found. The largest bias was associated with the first model (equation [T9]), which overestimated site index by 1.39 m on the average. Overestimating was also detected in the non-independent test, which used the data from the 34 stands to test the four original models based on the 102 stands (Table 5.8). This may suggest that the two datasets split from the 102 stands were slightly different in terms of the quantitative relationships between site index and the ecological variables. Therefore, underestimating would be expected if the 68 randomly selected stands were used to test the original model  • 147  (a)  6 4 2  •  • U)  • • •  0 •  a)  —2 —4 —6  • • • .. • • • •  •  •  ••  ••  •  •  14^16^18^20^22  ^  24  Estimated site index (m)  (b)  30  E 25 a)  -  • 20 71 a) (4)  15  10  10^15^20^25  ^  30  Estimated site index (m)  Figure 5.11. Residual analysis for the stratified model (equation [30c1): (a) residual versus estimated site index and (b) measured versus estimated site index (m @ 50 yr b.h. age).  148  Figure 5.12. White spruce measured site index in relation to soil moistureaeration regimes (SMARs) and soil nutrient regimes (SNRs) (equation (381) showing the regression surface, isolines, and distributions of measured site index (m CO 50 yr b.h. age). The ranges of 0-1, 1-2, 2-3, 3-4, 4-5, 5-6, 6-7, 7-8, and 8-9 represent MDa, DFr, SDa, Fa, Ma, VMa, Mr, Wr, and Wd SMARs (Figure 4.6), respectively. The ranges of 0-1, 1-2, 2-3, 3-4, and 4-5 represent very poor, poor, medium, rich and very rich SNRs, respectively.  149  (a)  6 4  I  2  • .. •  • —6  0  10  20  30  Estimated site index (m)  (b  )  30  0  0  10  20  30  Estimated site index (m)  Figure 5.13. Residual analysis for the combined SMAR and SNR model (equation [38]): (a) residual versus estimated site index and (b) measured versus predicted site index (m @ 50 yr b.h. age).  Table 5.9. The four candidate models recalibrated from the 68 randomly selected stands and selected statistics for these models. Symbols for independent variables are previously defined. N  R2  SEE (m)  Error range (m)  Mean error (m)  [T9]^SI = 7.42 - 36.018/(SCI) + 9.106(SNI) + 2.7351n(RD)  68  0.71  1.92  -4.4 - 4.3  1.481 (1.14) 2  [T23]^SI = 5.86 -20.992/(SCI) + 7.329(SNI) + 5.861n(RD) + 0.034(MOIST3) + 0.053(MOIST4) + 0.025(NITR3)  68  0.78  1.67  -3.9 - 3.7  1.29 (0.93)  [T30a] SI = 7.23 + 0.236(GW) - 0.025(C/N)  12  0.92  1.37  [T30b] SI = -16.54 + 6.3331n(GLEY) + 1.9251n(TN)  18  0.85  0.86  -4.1 - 3.6  1.32 (1.07)  [T30c] SI = -10.71 + 9.466(SNI) + 27.414(Et/Emax)  40  0.45  2.03  [T38]^SI = 6.75 + 8.97(MDa) + 13.02(SDa) + 10.06(DFr) + 13.74(Fa) + 13.76(Ma) + 11.96(VMa) + 9.95(Mr) + 6.58(Wr) + 0.00(Wd) - 0.78(VP) - 0.30(P) + 0.00(M) + 1.19(R) + 1.74(VR)  68  0.89  1.19  -3.9 - 2.0  0.80 (0.71)  Model  1The mean of absolute errors. 2The standard deviation of absolute error.  151  Table 5.10. Independent and non-independent tests of the four candidate models using 34 randomly selected stands as test data.  Model  ^Residual^Std. dev. range^of residual^Mean bias^Mean error (m)^(m)^(m)^(m)  Independent test (test data were not included in the model calibration) [T9]  -5.9 - 2.0  2.15  -1.39  2.06 1 (1.50) 2  [T23]  -4.5 - 3.1  1.86  -0.86  1.66 (1.17)  [T30]  -5.0 - 2.5  1.97  -0.50  1.64 (1.17)  [T38]  -3.0 - 2.6  1.48  -0.24  1.27 (0.76)  Non-independent test (test data included in the model calibration) [9]  -5.4 - 2.5  2.17  -0.85  1.96 (1.50) 2  [23]  -3.9 - 3.4  1.79  -0.51  1.57 (1.17)  [30]  -3.2 - 2.7  1.72  -0.16  1.49 (1.17)  [38]  -2.2 - 2.2  1.33  -0.24  1.11 (0.76)  1 The mean of absolute errors. 2 The standard deviation of absolute  errors.  152  based on the 102 stands, since the overall bias of the 102 stands should be very close to zero. The combined SMAR plus SNR model (equation (381) was the best prediction model with the smallest bias (-0.24 m) and mean absolute error (1.27 m) and the narrowest residual range (from an over estimate of 3 m to an under estimate of 2.6 m).  5.4.4. A Semi-Empirical Model Based on the Limiting Factor Analysis As demonstrated in Figure 5.14, the effect of growth limiting factors varies depending on the position of a stand on a regional edaphic gradient. On water-deficient sites, soil water becomes the limiting factor to white spruce growth. On slightly dry, fresh, and moist sites, climatic and soil nutrient factors become growth limiting. On water-surplus sites, soil aeration along with soil temperature become growth limiting. Due to interactions among soil moisture, aeration, and nutrients, as well as variation in the influence of each factor, it is difficult to examine the effect of changes in each factor on white spruce growth independently. However, an attempt was made to test the independent effect of climate, soil moisture and aeration, and soil nutrients on white spruce growth, using the principles of ecological site classification. Stratifying the zonal sites selected from the study stands according to biogeoclimatic units showed that there was little variation in site index along a climatic gradient (Figure 5.15). Tukey's test indicated that the only significant difference in site index was between the SBSdw3 (18 m) and SBSmw (21.5 m) units (p < 0.05). Using the climatic data (Table 2.1), the six biogeoclimatic units were stratified according to broad segments of temperature and precipitation  153  Sub-boreal Spruce Zone Soil nutrient regime VP^P^M^R^VR ED Soil moistt)re limited sites .  VD  E  MD  oo SD -^-^nutrient limited sites  0 E  0 Lel  Light and/or ^ h ea.t. limted sites  VM  w  Soil. aeration. limited . sites:  VW  Figure 5.14. A conceptual model delineating the growth limiting factors within the edaphic matrix for white spruce growth in the SBS zone. Symbols for soil moisture and nutrient regimes are given in Table 4.2 and Table 4.8, respectively.  154  (a)  30  1  25  .^i^f 10  DW3 DK MK1 Will DWI MW Subzone or variant  30  (b)  i^i^+  0  i  MD. SDa Fa Ma Mr Wr Wd Soil moisture and aeration regime  (c  30  )  1  25  11  i  9  II 2°  ^  i  4  10  VP^P^14^R  ^  VR  Soil nutrient regime  Figure 5.15. Categorical plot of measured white spruce site index in relation to (a) climate (represented by biogeoclimatic units), (b) soil moistureaeration regimes, and (c) soil nutrient regimes.  155  gradients (Table 5.11). This stratification implied that the climatic adjectives selected by the B.C. Ministry of Forests (Meidinger and Pojar 1991) are not entirely appropriate. White spruce site index was the lowest (18.1 m) in the dry and cool units (SBSdk and SBSdw3); intermediate (19.3 m) in the moist and wet cool units (SBSmk1 and SBSwk1); and highest (21.0 m) in the moist warm units (SBSdw1 and SBSmw). When white spruce site indices were compared separately along precipitation and temperature gradients (Table 5.11), significant differences were found only between the cool and warm segments (p < 0.05). Using the segments of a temperature gradient as dummy variables, its relationship to site index was described as: [40]^SI = 18.70 + 0(cool) + 2.16(warm) Adjusted R 2 = 0.37^SEE = 1.36 m n = 19  Table 5.11. Stratification of biogeoclimatic units according to temperature and precipitation gradients. Temperature^ Precipitation gradient gradient^ Dry^Moist^Wet Cool^ Warm^  SBSdk, SBSdw3^SBSmk1^SBSwk1 SBSmw, SBSdw1  To test the effect of soil moisture and aeration on white spruce site index, 40 sites with the same (medium) SNR were selected for comparison. Stratifying the selected study stands according to the SMARs showed that site index increased with increasing available soil water to a maximum and then it decreased with increasing water saturation and decreasing aeration (Figure 5.15). Tukey's test indicated that site index was significantly different between  156  all possible pairs of SMARs except slightly dry- and fresh-adequate SMARs, slightly dry- and moist-adequate SMARs, and fresh- and moist-adequate SMARs. It should be pointed out that restricted to deficient soil aeration, not soil water saturation, in the rooting zone was the direct cause of decreased growth on very moist to very wet sites. Very moist sites with medium SNRs were all poorly aerated, whereas those with rich to very rich SNRs were nearly all adequately aerated; therefore, they supported very good growth of white spruce. Using SMARs as dummy variables, their relationship to site index was described as: [41]^SI = 6.07 + 9.89(MDa) + 13.92(SDa) + 14.61(Fa) + 14.03(Ma) + 10.45(Mr) + 7.26(Wr) + 0(Wd) Adjusted R 2 = 0.92^SEE = 1.16 m n = 40  where the symbols for the SMARs are given in Figure 4.6. To examine the variation in white spruce site index across different soil nutrient regimes, 50 stands with slightly dry, fresh, and moist SMRs and adequate SARs were selected for comparison. The decision to select these SMRs was made based on two considerations: (1) the selected stands should include all five SNRs and (2) neither soil moisture nor soil aeration should be growthlimiting for the selected stands. The Tukey's test described above confirmed that site index was not significantly different among slightly dry-, fresh-, and moistadequate SMARs. Stratifying the selected study stands according to the SNRs showed that site index tended to increase with increasing available nitrogen from the very poor through the very rich SNRs; however, the differences in site index between the poor and the medium and between the rich and the very rich SNRs were not  157  significant (Figure 5.15). Using SNRs as dummy variables, their relationship with site index was described as: [42]^SI = 20.25 - 2.75(VP) - 0.37(P) + 0(M) + 1.61(R) + 2.66(VR) Adjusted R 2 = 0.61^SEE = 0.97 m n = 50  where the symbols for SNRs are given in Table 4.8. Limiting coefficients for soil moisture and aeration (Lma) and soil nutrients (Ln) were empirically determined based on the above limiting factor analyses. The values of these coefficients varied from 0 to 1, with 0 representing absolute limitation and 1 representing no limitation (Table 5.12). The stands with medium SNR were used to calibrate Lma. Since optimum white spruce growth occurred on sites with the fresh-adequate (Fa) SMAR, the Lma of these sites was set to 1, while Lma for the other SMARs was determined by calculating the ratio of mean site index between the fresh-adequate SMAR and other SMARs. Because the stands with medium SNR within the dry to freshrestricted (DFr) SMAR and the very moist-adequate (VMa) SMAR were either under-represented or absent, the Lma's for these two SMARs were calculated using the data from all possible SNRs (Table 4.21). The ratios (DFr/Fa) of mean site index on the poor and medium sites were calculated and their mean was taken as the Lma of SDpa. Similarly, the ratios (VMa/Fa) of mean site index on the rich and very rich sites were also calculated and their mean was taken as the Lma of VMa. The stands with the slightly dry-, fresh-, and moist-adequate SMARs were used to calibrate Ln. The Ln for the very rich SNR was set to 1, as  very rich sites were found to support optimum growth. Ln for the other SNRs was determined by calculating the ratio of mean site index between the very rich SNR and other SNRs. Climatic limitations, imposed by different biogeoclimatic units, were not estimated due to a small number of (n = 19) available zonal sites  158  and the inconsistency between the climatic designations given to biogeoclimatic units and the climatic data (Table 2.1).  Table 5.12. Soil moisture-aeration limiting (Lma) and soil nutrient limiting (Ln) coefficients determined by limiting factor analysis for soil moisture and aeration regimes (SMARs) and soil nutrient regimes (SNRs). Symbols for SMARs are as given in Figure 4.6. SMARs  Lma  SNRs  Ln  MDa  0.772  Very poor  0.764  DFr  0.846  Poor  0.868  SDa  0.967  Medium  0.883  Fa  1.000  Rich  0.954  Ma  0.972  Very rich  1.000  VMa  0.940  Mr  0.799  Wr  0.645  Wd  0.294  As there appear to be no moisture, aeration, and nutrient limitations on sites with a fresh-adequate SMAR and a very rich SNR, the mean site index on these sites was assumed to represent the productivity potential (C) of the studied population of white spruce. Unfortunately, sites with a fresh-adequate SMAR and a very rich SNR were not found in this study. Thus, the mean site index on the sites with a moist-adequate SMAR and a very rich SNR was used to estimate C (23.2 m). This was calculated using the mean site index on the sites  159  with a moist-adequate SMAR and a very rich SNR (22.5 m) divided by the limiting coefficient for these sites (0.972). As a result, site index on any site in the study area was estimated as:  [43] SI = 23.2 x Lma x Ln  When this model was tested against the 102 study stands, only a small bias was detected, an under estimate of site index by 0.11 m. The largest error was 4.2 m, and most of the errors are within 2 m. The mean absolute error was 0.96 m, with a standard deviation of 0.75 m. The residual versus estimated site index and measured versus estimated site index indicated that this model produced slight overestimates for low site indices and underestimates for high site indices (Figure 5.16). Compared to the four candidate models (equations [9], [23], [30], and [38]; Table 5.8), the limiting factor model was satisfactory in terms of predicting white spruce site index in the SBS zone. Its prediction precision was comparable with the best regression model (equation [38]).  5.4.5. Discussion Failure of soil nutrients alone (expressed as concentrations, or mass per unit area) to explain a large proportion of the total variance in site index does not mean that soil nutrients are not an important growth-limiting factor. Firstly, there are many nutrients affecting tree growth and the limiting soil nutrient(s) may vary with site. Secondly, complicated interactions among soil nutrients and the nonlinear response of height growth to soil nutrients may not be well described by linear models. Thirdly, the unavoidable multicollinearity of soil nutrient data may also cause problems in regression analysis.  160  (a)  6  4 • •  • • •^• • •  •• •  • • ;^•  • • • •^•^• •• ••  • •  ••  te • • • •s: • ••  •  •  •  ••  •  •  Sip  •  —4  —6  0  10  ^  20  ^  30  Estimated site index (m)  (b)  30  0  0  10^20  30  Estimated site index (m)  Figure 5.16. Residual analysis for the limiting factor model (equation [43]): (a) residual versus estimated site index and (b) measured versus estimated site index (m @ 50 yr b.h. age).  161  Despite the difficulties in synthesizing individual soil nutrient measurements into a single index, two soil nutrient indices (SNI and SNR) showed reasonably strong relationships with white spruce site index. Using SNI as the independent variable, equation [2] (Table 5.2) explained 53% of the total variance in site index; using SNRs as independent variables, equation [41] explained 61% of the total variance in site index, when noise from other major limiting factors had been eliminated. Similar strong relationships between productivity index (PI), integrated from some measured soil physical and chemical properties, and site index of several tree species were reported by Gale (1987) and Henderson et al. (1990). All models that include min-N as an independent variable in this study showed a negative effect of min-N on height growth. This negative relationship was most likely caused by high values of min-N associated with some wet and very wet sites with extremely low site indices. When these wet and very wet sites were excluded, a positive effect of min-N was found (equation [4.4]; Chapter 4). This result illustrates the importance of ecological stratification in describing site-productivity relationships. The wrong sign for regression coefficients are caused frequently by multicollinearity (i.e., the measured variables are too highly intercorrelated to allow precise analysis of their individual effects). However, if the overall R 2 in the regression is larger than any of the individual Ri2 's for the regressions of each independent variable on the other independent variables, and the purpose of using the regression is to describe the overall effect rather than to analyze the effect of each predictor, this may not be considered a problem (Greene 1990). From this viewpoint, all the regression models developed in this study are valid. Among the models selected for application, the model based on synoptic measures of ecological site quality was most successful in predicting white  162  spruce site index. This result suggests the presence of a strong link between these measures and white spruce site index (i.e., between ecological site classification and forest productivity). With the success in predicting white spruce site index from various measures of ecological site quality, changes in environmental factors resulting from either anthropological or natural disturbances can be related to changes in forest productivity.  5.5. CONCLUSIONS 1.  SoiUtopography, vegetation, or foliar nutrient variables alone were useful predictors of white spruce site index, with R 2 > 0.66 for the best regression models.  2.  Using any combinations of soil/topography, vegetation, and foliar nutrient variables as predictors, regression models explained z 74% of the total variance in site index. This improvement suggest that different ecosystem attributes tend to reinforce each other in explaining site-productivity relationships.  3.  Soil moisture and soil aeration were the most effective predictors of white spruce site index.  4.  Soil nitrogen, measured by min-N, TN, or C/N, was the most important soil nutrient affecting white spruce growth. The integrated soil nutrient indices, SNI and SNR, provided useful synoptic measures of soil nutrient conditions.  5.^Independent testing indicated that the prediction precision was improved significantly when the population of study stands was stratified into three subsets according to soil moisture conditions.  163  6.  The semi-empirical model, based on the concept of limiting factors and the model coefficients empirically calibrated using ecological site classification, reliably predicted white spruce site index over a wide range of sites.  7.  Depending on available data, the stratified model (model [30]), the combined SMAR plus SNR model (model [38]), and the limiting factor model [model [43]) are recommended to predict white spruce site index in the SBS zone. The mean absolute errors of these models were 1.44, 0.85, and 0.96 m, respectively, and most of the prediction errors were within 2 M.  164  6. HEIGHT GROWTH MODELS 6.1. INTRODUCTION Forest management planning for sustained timber production requires accurate assessment of ecological site quality as well as forest productivity. Because the height growth of free growing (dominant) trees in even-aged stands is sensitive to differences in ecological site quality, strongly correlated with volume growth, and weakly correlated with density and species composition (Davis and Johnson 1987), it has been used as a measure of forest productivity and a 'driving' variable in many forest growth and yield models. As actual measurements of site index cannot be made in stands with ages other than site index age, site index models/curves are used to estimate site index from any known height and age pair. Without site quality information, height at a future age cannot be reliably predicted since height at a certain age may differ a great deal among sites with different ecological site qualities. The accuracy of these estimates, in turn, depends on the quality of height growth models/curves (i.e., how well can they simulate the real trajectory of height over age development on different sites). Height growth models/curves may differ a great deal in terms of mathematical expression, independent variables, prediction precision, application, and other factors. However, the basic reason for developing height growth models/curves is always the same (i.e., to forecast stand height and, through the close relationship between stand height and volume, to forecast the volume of forest stands on a specific site (Curtis et al. 1974)). Various types of height growth models/curves have been developed for important timber species in North America. Anamorphic and polymorphic height  165  growth curves have been constructed for white spruce plantations (Love and Williams 1968, Stiell and Berry 1973, Berry 1978) and natural stands (Gevorkiantz 1957, Heger 1971, Ek 1971, Alemdag 1971, 1988, and 1991, Payandeh 1974, Hahn and Carmean 1982, Goudie and Mitchell 1986, Cieszewski and Bella 1991) in Canada and the United States, but no systematic comparisons have been made to date. Similarly, the height growth model for white spruce developed by Goudie and Mitchell (1986) has not been independently tested and evaluated. Variation in height growth patterns have been observed among different stands of the same site index (e.g., Carmean 1956, 1972, and 1975, Zahner 1962, Beck and Trousdell 1973, Carmean and Kok 1974, Losch and Schlesinger 1975, Newberry and Pinnaar 1978, Pfister et al. 1978, Hoyer and Chawes 1980, Monserud 1984a, Milner 1987). However, traditional approaches used to describe height growth patterns have failed to consider the variance of height growth pattern within the designated group and to provide a quantitative measure for the pattern. Recent studies have suggested that Zeide's two-point system (Zeide 1978) is an adequate method to quantitatively characterize variation in height growth patterns (Hoyer and Chawes 1980, Milner 1987). When height growth pattern, characterized by Ziede's method, was related to site variables, different results have been found for different tree species (Milner 1987). The quantitative characterization of height growth patterns and the sources of their variation in natural white spruce stands have not been studied. Predicted changes to the environment have raised concerns about the ability of empirical growth models, based on 'historical bioassays', to predict future growth (Kimmins 1985). Replacing site index in empirical models with site descriptors has been suggested (West 1990). However, the assumption that site effects are adequately represented without site index has seldom been  166  directly tested (Wykoff and Monserud 1987). Considering the usefulness of site classification in delineating ecologically-equivalent sites and the strong relationships between site index and measures of ecological site quality (see Chapters 4 and 5), it should be possible to develop site-specific height growth models using only age and measures of ecological site quality as independent variables. Advantages of these site-specific models would be: (1) height at any age could be predicted without using any stand information, (2) variation in height growth pattern, either due to site index and/or site factors, would be implicitly included in the model, and (3) impact of environmental changes could be accounted for by the assessment of ecological site quality. In this study, selected anamorphic and polymorphic height growth curves were developed and evaluated. The best model was selected as the base-line model of the study. The suitability of Zeide's two-point system (Z ratio) to characterize height growth patterns was tested. The relationships between height growth pattern defined by the Z ratio (height at 60 yr. b.h. age : height at 30 yr. b.h. age) and the measures of ecological site quality and productivity were studied in order to identify the determinants of the height growth patterns. Pattern-specific models were developed for each height growth pattern group defined by the Z ratio. Within the framework of site classification using the BEC system, site-specific models were developed, in which site index was implicitly and explicitly represented by the measures of ecological site quality. Comparisons between models and modelling strategies used in the study were made, followed by the selection of candidate models for predicting dominant height of white spruce stands in the study area.  167  6.2. LITERATURE REVIEW By definition, a model is a formal and precise statement, or set of statements, embodying our current knowledge or hypotheses about the workings of a particular system and its response to stimuli (Landsberg 1986). Models can be divided into two basic categories: empirical (descriptive or statistical) models (based on data obtained by measurements), and mechanistic (explanatory or theoretical) models (based on the physiological and physical processes that describe how the system responds to stimuli). The advantages and disadvantages of each category were discussed in detail by Landsberg (1986), Kimmins et al. (1990), and Bossel (1991). Forest growth and yield models are traditionally based on 'historical bioassays' and, therefore, are empirical models. A height growth model is a mathematical expression of a set of height growth curves which portray the development of height over age. Usually, height growth models, expressed as H=f(SI,A), simulate only the mean height of dominant and codominant trees in a forest stand. Frequently, these simulation are graphically expressed as a set of height over age curves for different site indices, and sometimes are also called (incorrectly) as site index curves or site curves. As pointed out by Curtis (1974), when the parameters of height growth curves [i.e., H=f(SI,A)] are estimated via regression, suboptimal estimates of SI are obtained if the equation is algebraically inverted to give SI=f(H,A). Two different dependent variables, each with its own variance, are involved, and thus two separate sum of squares must be minimized. Unless H and SI are perfectly correlated, the estimated parameters will be different. Therefore, it seems best to differentiate between height growth curves [H=f(SI,A)] and site index curves [SI=f(H,A)] when making reference to a site curve system. The curve system that should be used is strictly dependent on the purpose of the study. Because of  168  the above problem, several studies have attempted to build compatible models to predict both SI and H (e.g., Kirby 1975, Cieszewski and Bella 1989). Various types of height growth models/curves have been developed to predict height growth based on site index and age. Although the models may differ in terms of selected functions and coefficients, the principle of height predictions remains the same [i.e., given a pair of H and A (so site index can be estimated) or site index, predict the height at any other given age]. Therefore, the accuracy of height predictions depend on how well models portray the trajectory of height development on a specific site and on how accurately site index can be estimated with or without suitable site trees. The history of height growth models/curves reflects the pursuit to perfectly simulate the real trajectory of height over age. The form of height growth curves have evolved from anamorphic through polymorphic to sitespecific (e.g., Bruce 1923 and 1926, Bull 1931, Carmean 1956, Stage 1963, Monserud 1984a, Milner 1987), with an increase in complexity and number of models and a decrease in the size of area to which the model is applicable. Anamorphic height growth curves usually are constructed using temporary plot data and the guide curve method. The implied assumption that the same height growth pattern holds on all sites, which results in a constant ratio at all ages for any two levels of site index, has been subjected to major criticisms (e.g., Bull 1931, Stage 1963, Beck and Tronsdell 1973). Even if this assumption holds, data often fail to meet the requirement of an uniform distribution of ecological site quality over all ages, which results in a distorted guide curve (Beck and Trousdell 1973, Simth 1984, Monserud 1984a). The inaccuracy of prediction using anamorphic height growth curves has long been recognized (e.g., Bull 1931, Stage 1963, Carmean 1970).  169  Polymorphic height growth curves are derived from true time series data from stem analysis or repeated measurements on permanent sample plots. Since these curves allow height growth on different sites (in terms of site index) to take different shapes, polymorphic height growth curves have been widely used and proven to have many advantages over anamorphic curves (Bull 1931; Carmean 1972, Monserud 1984a). Published polymorphic height growth curves for white spruce include: Heger's (1971) curves for the mixedwood forest section in Alberta, Ek's (1971), Payandeh's (1974), and Hahn and Carmean's (1982) curves for natural white spruce stands in the Great Lake area, Alemdag's curves for natural stands in the Yukon Territory (1971), Northwest Territories (1988), and Canada (1991), Goudie and Mitchell's (1986) curves for natural stands in Alberta and interior B.C., and Cieszewski and Bella's (1991) for natural stands in Alberta. The assumption that there only exists a single curve shape for a given level of site index is questionable, and has been challenged. Many studies indicate differences in height development for stands growing on different site types (i.e., the relationship between height growth and site index changes in response to soil and climatic factors even if site index remains the same) (e.g., Carmean 1956, Beck and Tronsdell 1973, Carmean and Kok 1974, Losch and Schlesinger 1975, Hoyer and Chaws 1980, Monserud 1984a, Milner 1987). Recent studies have shown that the anamorphic model for karri (Eucalyptus diversicolor F. Muell.) (Rayner 1991) and the linear model for black spruce  (Smith and Watts 1987) were superior to polymorphic models. Other studies have indicated that site index is a poor descriptor of height growth pattern for some tree species (e.g., Milner 1987, Q. Wang 1992), although it may adequately reflect the height growth potential (asymptotic level).  170  Carmean (1956, 1972, 1975), Zahner (1962), and Newberry and Pienaar (1978) showed that the pattern of height growth through a specified site index varies among different soil textural and drainage classes of soils. Pfister et al. (1979) and Monserud (1984a) found different growth patterns among habitat types. Using Heger's (1968) data summarization method, Monserud (1984a) identified three habitat series groups according to the average site curve shape of each habitat type and used them as dummy variables in height growth and site index models. Hoyer and Chawes (1980) showed that curve shapes varied according to precipitation more than it did for site index levels. On the other hand, Linteau (1955) found that the site index curve shape for white spruce on different site types in the northeastern coniferous section of the boreal forest region in Quebec was essentially the same. Heger (1971) compared differences in the shape of site index curves for white spruce in Alberta mixedwoods between different parent materials and different drainage classes. He found that these differences were so small as to be of limited practical importance. In all these studies, the basic approach was first to stratify the data according to some classification scheme, then to fit curves to the data within each group, and finally compare the curves via graphical, numerical or statistical methods (Milner 1987). The soil and site characteristics that were used in the site classification scheme were implicitly assumed to be the factors which control the pattern of height growth. This assumption may not be necessarily correct since some factors such as climate, stand structure, history, and stocking, and genetic differences, which are generally ignored in most of site classification and productivity studies, could also be controlling factors for the height growth pattern (Heger 1971, Smith and Watts 1987). Furthermore, these approaches failed to examine the variation in curve shape within each stratum.  171.  If there is a great within-stratum variation in the curve shape, comparing the shape of average curves fitted for each stratum would be misleading. Milner (1987) attempted to identify the major sources of variation in height growth patterns. He used the Z ratio (Zeide 1978) as the quantitative measure of height growth patterns, and related this ratio to certain environmental factors (precipitation for March through June, sum of average monthly daylight temperature for March through June, cumulative spring afternoon solar radiation for March through June, and percent silt in top 100 cm of soil profile). For the four species examined (ponderosa pine, Douglas-fir, western larch and lodgepole pine), he found that the environmental factors accounted for the significant sources of variation in curve shape for all but western larch. Since most of the environmental variables examined in his study were climatic, the result indicated that climate conditions are very important sources of variation in height growth patterns. Over the past several decades, empirical height growth models have successfully been used in daily forestry practice. As long as the future growth conditions are similar to the past, the use of empirical height growth models/curves remains justified (Kimmins et al. 1990). However, the possibly rapid change in environmental conditions from atmospheric pollution, forest overuse and deforestation in many part of the world, and intensified management regimes has resulted in a situation in which environmental conditions for growth can no longer be treated as immutable. In consequence, attempts to build mechanistic models for growth and yield prediction have been made (e.g., Agren and Avelsson 1980, Shugart 1984, Bossel 1986, Kienast 1987, Running and Coughlan 1988). Mechanistic models are believed to be superior to empirical models under changing environmental conditions (Landsberg 1986, Bossel 1991), but many  172  workers agree that more time is needed for existing mechanistic models to match the precision of the empirical models calibrated from forest-wide inventory and growth plot databases (Leech 1984, Rayner and Turner 1990). Empirical models will therefore continue to be used in both operational forest planning and silviculture response simulations. A major constraint in generalizing mechanistic models is the availability of site description parameters to replace the 'site index' parameter used presently in empirical models (West 1990). The successful replacement of site index in height growth models by measurable environmental factors is extremely helpful not only when site trees are not available, but also when ecological site quality changes with time. Direct incorporation of environmental factors as driving variables in growth models is presently limited by the resolution (time and spatial scale) and the nature of available climatic and edaphic data (Nautiyal and Cuoto 1984, Rayner and Turner 1990). Consequently, alternative site describers, such as those derived from site classification, have received considerable attention (e.g., Green et al. 1989, Klinka and Carter 1990, Inions 1990, Inions et al. 1990). However, growth and yield research and site classification studies have rarely been coordinated (Crow and Rauscher 1984). 6.3. METHODS 6.3.1. Preliminary Site-Curve Preparation Graphs of height versus age were examined for each site tree. If suppression or damage was apparent, data from the site tree was deleted or truncated. In consequence, six trees were deleted, and the remaining 300 site trees were used in further analyses.  173  An average height growth curve was determined for each plot from the individual tree stem analysis data using Richards's (1959) three-parameter model  [6.11 H = 1.3 + b1*(1 - e-b2*A)b3  where H is height (m), A is age (years) at breast height, e is the base of the natural logarithm, and bl, b2 and b3 are parameters to be estimated for each stand. Within-plot standard errors of estimates for the above model in the 102 study stands averaged 0.79 m, with a standard deviation of 0.28 m. Actual site index in each stand was determined either from the model by setting A = 50 years (site index age at breast height) for those stands with breast height age < 50 years or directly using the average of the true heights at b.h. age 50 years by applying Carmean's (1972) interpolation. The model was evaluated for each stand at every decade from age 10 to the decadal age nearest the age of the oldest tree in that stand to provide the database used for constructing height growth curves. As a result, 690 decadal observations of height, age, and site index for 102 stands were produced. Of these, 596 observations between age 10 and 100 from 82 stands were used to develop height growth models which required site index as a predictor. The remaining 94 observations from 20 stands, which had average b.h. age < 50 years, were not used to fit any type of height growth model that included site index as a predictor. For those models without site index as a predictor, data from all 102 stands were used to calibrate the model coefficients. The three parameters (bl, b2, and b3) of each average height growth curve were later related to site index, selected measures of ecological site quality, and height growth pattern. The possibility of using measures of  174  ecological site quality and/or site index to predict these coefficients was examined. 6.3.2. Selecting and Fitting Traditional Height Growth Curves Anamorphic model  The original three-parameter Richards' model (Equation 6.1) adequately describes the development of individual trees growing in relatively stable environment and, thus, has been widely used in growth and yield modelling (Pienaar and Turnbull 1973). The parameter bl is interpreted as the asymptote of height growth, while b2 and b3 are interpreted as the rate and shape parameters, respectively (Richards 1959). In this study, a slightly modified version of the Richards' model, in which b1 was replaced by b1*(SI-1.3) to allow the asymptote to vary with site index, was used to fit the data. [6.2]^H = 1.3 + b1*(SI .1 . 3)*(1 _ e (-b2*A))133  where H, SI, A, bl, b2, and b3 are as previously defined. Since SI in equation [6.2] can only affect the asymptotic value, not b2 and b3, height vs age curves generated from the model are anamorphic.  Polymorphic models  Four polymorphic models described in the literature, Goudie and Mitchell's (1986) (equation [6.3]), Alemdag's (1988) (equation [6.4]), Ek-  175  Payandah's (Ek 1971, Payandeh 1974) (equation [6.5]), and the logistic (Monserud 1984a) (equation [6.6]), were selected to fit the data. 1 + e [b1 + b2*ln(50) + b3*ln(SI - 1.3)] [6.3]^H = 1.3 + (SI - 1.3)*  [6.4] H = 1.3 +  1 + e [bl + b2*ln(A) + b3*ln(SI - 1.3)]  (SI -1.3) + b1*(A - 50) + b2*(A - 50) 2 1.0 + b3*1n(A/50) + b4*[1n(A/50)1 2  [6.5]^H = 1.3 + bl*SIID 2*[i_ e (-b3*A)](b4* SIb5) [6.6]^H = 1.3 + b1*(SI- 1.3)b2 / (1 +e [b3 - b4*ln(A) - b5*1n(SI-1.3)])  where H, SI, A, e, bl, b2, and b3 are as previously defined, and b4 and b5 are model coefficients. Equations [6.3] and [6.4] have been used to construct height growth curves, based on stem analysis data, for natural white spruce stands in British Columbia (Goudie and Mitchell 1986) and the Northwest Territories (Alemdag 1988), respectively. Equation [6.5] has also been fitted to data from site index tables for natural white spruce stands in the Great Lake area of United States (Ek 1971, Hahn and Carmean 1982) and Canada (Payandeh 1974). Comparisons between these original models and the models fitted in this study were made graphically. The selected anamorphic and polymorphic models were evaluated with regard to their goodness of fit to the original data using residual analysis. Based on this evaluation, a baseline model was chosen as the standard for comparison with the pattern-specific and site-specific height growth models developed in the study.  176  6.3.3. Height Growth Pattern and Growth Modelling In nature, dominant height growth on different sites may often be the same at a certain age and markedly different at other ages. The insufficiency of the one-point system, on which anamorphism depends, to determine height growth curves has been recognized since the use of polymorphic curves was proposed and advocated (Bull 1931, Stage 1963). However, the polymorphic concept does not suggest a minimum number of points necessary to determine height growth curves. Zeide (1978) found that two points were necessary and sufficient to determine any stand growth curve. This two-point system has been adopted by Hoyer and Chawes (1980) and Milner (1987) to characterize height growth patterns. In this study, Zeide's two-point system was used to characterize white spruce height growth patterns. The two points selected were the average dominant heights at the b.h. age of 30 (H30) and 60 (H60) years, the same as the two points suggested and used by Hoyer and Chawes (1980) and Milner (1987). The ratio of H60/H30 (the Z ratio) was calculated for each stand as the index of height growth patterns. The 82 ratios were then arranged in order and divided into several groups (Z groups) according to the criteria proposed by Zeide (1978) (i.e., the difference between the upper and the lower limit of each group was set to seven percent of the lower limit). In order to confirm that Z ratio was an adequate index of height growth patterns, both visual and statistical methods were used. The visual method involved plotting original unsmoothed stem analysis data and connecting each individual tree's data as an unsmoothed line according to Z group. The variations in growth patterns within each Z group and among all Z groups were observed from the graphs. Since the b2 and b3 coefficients in equation [6.1] were  177  interpreted as coefficients describing curve shape, a close relationship between them and the Z ratio was expected. Correlation analysis was used to quantify this relationship. In order to demonstrate the usefulness of the Z ratio in improving the precision of height prediction, pattern-specific height growth models were developed. Since each Z group was relatively uniform in height growth pattern, an anamorphic model was considered to be sufficient for simulating height-age relationships. These anamorphic models were compared with each other to observe the differences in the height growth patterns among groups, and with the selected baseline model to observe the improvement of prediction precision. Since height growth pattern was quantitatively determined by the Z ratio, a considerable improvement in the height prediction precision could be expected either by incorporating Z ratio into the height growth models or by developing separate height growth models for each Z group (i.e., developing pattern-specific height growth models). However, it is impossible to obtain the two points selected to calculate Z ratio without stem analysis data or a permanent plot record for height growth from the stands older than 60 years. Therefore, the applicability of the two-point system in developing height growth models depends on the predictability of the Z ratio from routinely measurable site variables. In order to establish a model to predict the Z ratio, various measures of ecological site quality, including SMR, SNR, SMAR, site association, and site index were related to the Z ratio by correlation or regression analyses. The possibility and reliability of predicting the Z ratio using these measures were also examined.  178  6.3.4. The Link between Height Growth Models and Ecological Site Classification In order to establish a link between height growth models and site classification, site-specific height growth models, with and without site index as a predictor, were proposed and tested. Site series with similar Z ratios were combined into site-series groups. Goudie and Mitchell's model (equation [6.3]) was fitted separately to the data of each group. The resultant site series-specific height growth models were then compared with the baseline model. Traditional height growth models (both anamorphic and polymorphic) express height as a function of age and site index [i.e., H = f(SI, A)], which assume that height at a certain age depends on ecological site quality. The disadvantages of using site index as an measure of ecological site quality were discussed by Monserud (1984a). Even if site index could be assumed to be a perfect measure of ecological site quality, a precise estimate of site index would be problematic in some situations. In this study, an attempt was made to replace site index in height growth models by measures of ecological site quality using two methods. In the first method, site index in the height growth model was explicitly replaced by measures of ecological site quality. Specifically the site index prediction model (equation [43]) developed in Chapter 5 and Goudie and Mitchell's model (Equation [6.3]) were used to form an equation system. Substituting equation [43] into equation [6.3] yields  [6.7] H = 1.3 + (23.2*Lma*Ln - 1.3)*  1 + exp(b1 - b2*ln(50) - b31n(23.2*Lma*Ln - 1.3) 1 + exp(b1 - b2*ln(A) - b3*ln(23.2*Lma*Ln - 1.3)  179  Equation [6.7] was fitted to the data, and used to predict height from age and measures of Lma and Ln (Table 5.12; Chapter 5). In the second method, site index was implicitly represented by measures of ecological site quality. As demonstrated in Chapter 4, each site group was relatively uniform in ecological site quality, especially in terms of white spruce height growth. Richards' model (equation [6.1]) was fitted to the data for each site group. Thus, the effect of ecological site quality on height growth was incorporated into each height growth model. These models were independently tested against the data from Q. Wang et al. (1992), and then compared with the height growth models that used site index as a predictor.  6.4. RESULTS AND DISCUSSION 6.4.1. Development and Evaluation of Anamorphic and Polymorphic Height Growth Models One anamorphic and four polymorphic models were developed in the study (Table 6.1). All five models achieved a very good fit with equations [6.3], [6.5], and [6.6] having slightly higher R 2 's and lower SEE's than equations [6.2] and [6.4]. Residual analyses suggested there was no lack of fit in any of these models (Appendix 4). Height over age curves produced by these models at site indices 12, 18, and 24 m are depicted in Figure 6.1. Estimations from these models were very similar or identical around 50 years. With decreasing or increasing age, estimates diverged. Differences among height growth curves on the medium productivity sites (SI = 18) were less than that on either high (SI = 24 m) or low (SI = 12) productivity sites. For ages younger than 40 years, equation [6.2] gave  180  40  Equation:  30  6.2 6.3 6.4 6.5 ^ 6.6  20  10  0  0^20^40^60^80  ^  100  Age @ b.h. (year)  Figure 6.1. White spruce height growth curves produced from one anamorphic and four polymorphic models for site index 12 (bottom), 18 (middle), and 24 m (top).  Table 6.1. Coefficients and statistics for one anamorphic and four polymorphic height growth models fitted to the study stands.  Equation  bl  b2  b3  b4  b5  R2  SEE  [6.2]  2.140  0.01728  1.396  0.984  0.95  [6.3]  9.565  -1.451  -1.236  0.987  0.84  [6.4]  0.1272  -0.0006044  -0.5804  0.4930  0.983  0.98  [6.5]  4.121  0.7287  0.01953  5.934  -0.4762  0.987  0.85  [6.6]  16.36  0.3975  7.884  1.217  0.8643  0.987  0.86  182  higher estimates on low productivity sites and lower estimates on high productivity sites. This trend was reversed at ages greater than 60 years. Mean absolute errors and biases of height estimates obtained from the five models were summarized according to site index and age classes (Tables 6.2 and 6.3). The trend of mean absolute errors and biases across site index classes indicated that the mean absolute errors and biases from the model calibration were not associated with site index (Table 6.2). Although the mean absolute error tended to increase with age as the total height increased with age, no obvious trend for mean bias was found across age classes (Table 6.3). Therefore, the mean absolute errors and biases can be used to evaluate the performance of the models. In general, no serious bias was found in any of the models, although minor biases were detected in equations [6.4] and [6.6] (Table 6.2). In terms of mean absolute errors, equations [6.3] and [6.5] were the best, followed by equations [6.6], [6.2], and [6.4]. Equations [6.3], [6.4], and [6.5] developed in this study were compared to the original models developed by Goudie and Mitchell (1986), Alemdag (1988), and Ek (1971), respectively. Equation [6.3] gave nearly identical results to the original model (Figure 6.2). This was not surprising when considering that the original model was developed partially with data from the interior B.C. Equations [6.4] and [6.5] were different from the original models. As indicated by the different curve shapes, serious errors will occur if the original models are used to estimate white spruce dominant height in the study area or vice versa (Figure 6.3 and 6.4). Compared to Alemdag's original model for each of the three site index classes, equation [6.4] gave lower estimates before 50 years and after 80 years, and gave slightly higher estimates between 50 and 80 years (Figure 6.3). In contrast, equation [6.5] gave constantly higher estimates, especially at younger and older ages, compared to Ek's original model (Figure 6.4).  Table 6.2.^Error and bias (m) (upper and lower values, respectively) of height estimates obtained from the anamorphic and polymorphic height growth models fitted to the study stands according to site index classes. Site index^Number of classl^observation  Equation [6.41  [6.5]^[6.6]  -0.01  1.05 0.03  0.70^0.73 0.16^0.31  0.93 -0.05  0.77 -0.15  0.80 -0.14  0.75^0.99 -0.04^-0.14  0.60 -0.05  0.64 -0.06  0.65 -0.06  -0.00^-0.09  18^99  0.67 -0.20  0.63 0.14  0.73 -0.22  0.63^0.65 -0.13^-0.18  20^188  0.60 -0.03  0.57 0.04  0.66 -0.08  0.57^0.58 0.01^0.04  22^70  0.56 0.14  0.50 0.03  0.57 -0.12  0.50^0.55 0.01^0.17  24^12  0.41 0.38  0.29 0.06  0.23 -0.05  0.21^0.37 0.07^0.36  Total^596  0.66  0.61  0.00  -0.00  0.70 0.09  -0.00^0.013  [6.2]  [6.3]  0.85 0.31  14^56 16^112  12^59  0.62  1 Site index classes 12 - SI < 13, 14 - > SI < 15, 16 - 15 > SI < 17, 18 - 17  SI < 19, 20 - 19  0.62^0.64  0.60^0.63  SI < 21, 22 - 21  SI < 23, and 24 - SI > 23 m.  Table 6.3.  Error and bias (m) (upper and lower values, respectively) of height estimates obtained from the anamorphic and polymorphic height growth models fitted to the study stands according to age classes.  Age class (yrs)  Number of observation  Mean height (m)  10  82  20  [6.2]  Equation [6.3]^[6.4]  [6.5]  [6.6]  3.86  0.67 -0.18  0.61 -0.02  0.91 -0.76  0.58 -0.06  0.64 0.22  82  7.59  0.85 -0.12  0.77 -0.05  0.80 -0.05  0.77 0.02  0.79 -0.13  30  82  11.40  0.71 0.01  0.67 -0.04  0.73 0.35  0.67 0.05  0.68 -0.11  40  82  14.93  0.42 0.10  0.39 0.02  0.47 0.33  0.39 0.11  0.40 0.02  50  82  17.99  0.06 0.06  0 0  0 0  0.05 0.05  0.11 0.08  60  79  20.48  0.47 -0.08  0.48 -0.06  0.54 -0.36  0.48 -0.07  0.50 0.04  70  46  21.60  1.06 -0.19  1.02 -0.21  1.11 -0.63  1.02 -0.26  1.01 -0.13  80  25  23.04  1.29 0.20  1.11 0.06  1.18 -0.20  1.10 0.02  1.12 0.10  90  23  24.08  1.54 0.55  1.39 0.15  1.56 0.36  1.36 0.21  1.37 0.17  100  14  24.30  1.42 0.58  1.17 0.11  1.58 0.69  1.08 0.07  1.22 -0.09  185  40  30  .4 tu •,-, a) 20 1..)  7c54 E-. 10  0  0^20^40^60^80  ^  100  Age @ b.h. (year)  Figure 6.2. White spruce height growth curves produced by Goudie and Mitchell's (1986) model based on the data from this study (solid line) and the original study (dotted line) for site index 12 (bottom), 18 (middle), and 24 m (top).  186  40  30  2 a) 20 0  10  0  0^20^40^60^80  ^  100  Age @ b.h. (year)  Figure 6.3. White spruce height growth curves produced by Alemdag's (1988) model based on the data from this study (solid line) and the original study (dotted line) for site index 12 (bottom), 18 (middle), and 24 m  (top).  187  40  30  0  0^20^40^60^80  100  Age @ b.h (year)  Figure 6.4. White spruce height growth curves produced by Ek's (1971) model based on the data from this study (solid line) and the original study (dotted line) for site index 12 (bottom), 18 (middle), and 24 m (top).  188  These comparisons imply that climate is a key factor in determining height growth patterns since different curve shapes for the same site index were observed in the three different regions (i.e., the interior B.C., the Northwest Territories, and the Lake States). For a widely distributed tree species like white spruce, stands with the same site index in different climatic regions likely occur on edaphically different sites. For example, the site index of a species growing on a water deficient and nutrient-poor site in a favorable climate may be the same as the species growing on a non-water deficient and nutrient-rich site in a less favorable climate. Based on the evaluations, equations [6.3] and [6.5] were the most appropriate models to describe height growth in the study stands. Considering that Goudie and Mitchell's curve is currently used in the province, the identical fit for the study stands, and the fact that the curves pass through site index at age 50, equation [6.3] was selected as the baseline model for further comparisons. 6.4.2. Characterization of Height Growth Patterns The calculated Z ratios, ranging from 1.51 to 2.25, were stratified into six groups, corresponding to Z numbers of 7, 8, 9, 10, 11, and 12, according to Zeide's (1978) methods. One plot with a high Z ratio (2.73) was tested as an outlier. Although the plot may represent a unique pattern of white spruce height growth, it is not adequate to carry out further analysis based on only one plot. Using the original stem analysis data, height-age curves plotted for each of the six groups showed relatively uniform growth patterns within each group and different growth patterns among groups (Appendix 5). The wide spread of site index within each group that was relatively uniform in height growth  189  pattern suggested that site index may not be the major factor that controls height growth pattern. Related to shape and rate coefficients b2 and b3 from the three-parameter Richards' model (equation [6.1]) fitted for each plot, the Z ratio was found significantly correlated with b2 (r = 0.33), b3 (r = 0.35), b2 and b3 together (r 0.91), and b2, b3, and the product of b2 and b3 (r = 0.94). As b2 and b3 were interpreted as curve shape determinants, these strong relationships justified the usefulness of the Z ratio as a quantitative descriptor of the height growth pattern. If the Z ratio is a good indicator of height growth pattern, height growth models based on the Z groups should improve the precision of height prediction. In order to confirm this reasoning, equation [6.2] was selected to fit each Z group. It was not necessary to use polymorphic models because the height growth pattern within each Z group was relatively uniform. As expected, a very low SEE was obtained compared to the baseline model, or the same model fitted to unstratified data (Table 6.4). Each model specified in Table 6.4 can be used to generate a set of anamorphic height growth curves for different site indices to show white spruce dominant height changes with age once the Z ratio is determined and assigned into the appropriate Z groups. Using the same site index (SI = 18 m), the curves produced by the six height growth pattern-specific models were plotted in Figure 6.5. Although all curves passed through the same points at b.h. age 0 and 50 years, differences among them existed at other ages. The curve for Z-group 7 gave the highest estimates for ages less than 50 and the lowest estimates for older ages, while the estimates given by the curve for the Z-group 12 were the reverse. At b.h. age 100 years, the difference in the mean dominant height between Z-groups 7 and 12 was as large as 9 m.  Table 6.4. Height growth pattern-specific models: coefficients, coefficients of determination, standard errors of estimates for equation [6.2] fitted to each of the six Z-groups.  Z-group  range of Z ratio  bl  b2  b3  R2  SEE (m)  7  1.51 - 1.59  1.376  0.03354  1.542  0.998  0.28  8  1.62 - 1.71  1.536  0.02801  1.524  0.997  0.36  9  1.72 - 1.83  1.893  0.02016  1.407  0.997  0.38  10  1.85 - 1.95  1.928  0.02245  1.666  0.998  0.39  11  1.99 - 2.09  1.990  0.02456  1.981  0.997  0.47  12  2.11 - 2.25  2.507  0.01942  1.929  0.996  0.47  191  40  Z—group: ^Z=7  30  Z=8 Z = 9 _ _ _ Z = 10 Z = 11 _ _ Z = 12  20  10  0  0^20^40^60^80  ^  100  Age @ b.h. (year)  Figure 6.5. White spruce height growth curves for each of the six Z-groups at site index 18 m obtained from the height growth pattern-specific models specified in Table 6.4. The codes for the Z-groups are given in Table 6.4.  • 192  Z—group: ^Z=7 _Z=8 Z = 9 _ _ _ Z = 10 Z = 11 — — Z = 12  cI), 0.48  0.42 0 c) r. 0.36 0 a) ;.. c.) 0.30 g .4—) 4 0.24 tip "al 4 0.18  7t;  O  0.12  0 -..t  Age @ b.h. (year)  Figure 6.6. White spruce annual height increment curves for each of the six Zgroups at site index 18 m obtained from the differential forms of the height growth pattern-specific models specified in Table 6.4. The codes for the Z-groups are given in Table 6.4.  193  The differential forms of these height growth pattern-specific models were used to obtain curves of annual height increment over b.h. age (Figure 6.6). As the Z ratio increased, the b.h. age of the maximum height growth rate increased from about 15 (Z-group 7) to 40 (Z-group 12) years. Before approximately 18 years, height growth rate increased as the Z ratio decreased, with the highest growth rate in Z-group 7 and the lowest growth late in Z-group 12. After approximately 30 years, the above trend was reversed. The unique pattern of height growth and annual height increment revealed by the six height growth pattern-specific models further supports the usefulness of the Z ratio as a quantitative measure of the height growth pattern. Site index and various measures of ecological site quality were related to the Z ratio and coefficients bi, b2, and b3 of the Richards' model (equation [6.11) fitted to each stand. In general, no very strong relationships were found (Table 6.5). Although there were several significant correlations, these relationships were weak in terms of producing a reliable prediction of the Z ratio or the b coefficients for the height growth model. Using site series, which had the highest correlation with bl (r = 0.84) and b2 (r = 0.70), as dummy variables for estimating coefficients bl and b2, the adjusted R 2 's of the resulting multiple linear regression models were 0.56 and 0.25, respectively. Using soil moisture regimes as dummy variables for estimating coefficient b3, the adjusted R 2 of the resulting multiple linear regression model was 0.20. Thus, the two-step procedure (parameter prediction approach) for developing site-specific height growth models (i.e., fitting a height growth model for each stand and then predicting the model coefficients using measures of ecological site quality) was rejected. No significant differences in the Z ratio between SNRs, SMRs, SARs, and SMARs were detected using ANOVA (Figure 6.7). In relation to plant  Table 6.5. Correlations coefficients between indices (the Z ratio of each study stand and the b coefficients of equation [6.1] fitted to each study stand) of white spruce height growth pattern and measures of ecological site quality. Asterisks mark the significant correlations (p < 0.05, n = 81). Measure of site quality  Z ratio  bl  b2  b3  Site index  0.24*  0.63*  0.10  0.38*  Soil moisture regime  0.34  0.66*  0.46*  0.51*  Soil nutrient regime  0.15  0.39*  0.27  0.20  Soil aeration regime  0.20  0.51*  0.26  0.41*  Soil moisture-aeration regime  0.38  0.67*  0.46*  0.52*  Site association  0.41  0.75*  0.50  0.53*  Site series  0.70*  0.84*  0.70*  0.64  Biogeoclimatic unit  0.25  0.40  0.33  0.30  Site group  0.27  0.61*  0.36  0.49*  195  (a) 2.4  (b) 2.4  22  22  2.0  2.0  0 ::.1  4  N  4 N  1.8  1.6  1.4  1.8  1.6  VP^P^M^R  1.4  VR  Soil nurtrient regime  (d) 2.4  22  22  J  4 N  1.6  1.4  0  7.). 4  N  1.8  I  i  T  TT^T MD SD F M VM W VW Soil aeration regime  (c) 2.4  0 2.0  I  1  2.0  1.8  .1.  ^ a^r^d ^ Soil moisture regime  1.6  1.4  MDa DFr SDa Fa Ma VMa Mr Wr Wd Soil moisture and aeration regime  Figure 6.7. Box plots showing the Z ratio stratified according to (a) soil nutrient regimes (SNRs), (b) soil moisture regimes (SMRs), (c) soil aeration regimes (SARs), and (d) soil moisture-aeration regimes (SMARs). Symbols for SNRs, SMRs, SARs, and SMARs are given in Table 4.8, Table 4.2, Table 4.5, and Figure 4.6, respectively.  196  associations, site associations, biogeoclimatic units, and site groups, ANOVA followed by Tukey's multiple comparisons suggested that the only significant difference was between the ARAL and PETA plant associations (Figure 6.8). Using site index and site series, the two factors significantly related to the Z ratio, as predictors for estimating the Z ratio, the adjusted R 2 's of the resulting simple and multiple linear regression models were 0.06 and 0.25, respectively. Obviously, the variation in the Z ratio cannot be explained by any of the measures of ecological site quality. As most of the 27 delineated site series were only represented by a few (1, 2, or 3) stands, site series with similar Z ratios were combined into four (A, B, C, and D) site series groups, approximately corresponding to Z-groups 7 and 8, 9, 10, and 11 and 12, respectively. Despite large variations within site series groups B and C, the differences in the Z ratio among the four groups were obvious (Figure 6.9) 6.4.3. Site-specific Height Growth Models 6.4.3.1.^Site-specific Models with Site Index as Predictor In view of the correlation between height growth pattern and site index and site series, using these variables could improve the performance of height growth models. Based on the differences in the Z ratio among the four site series groups, Goudie and Mitchell's model was calibrated for each site series group (Table 6.6). The reason for selecting a polymorphic model rather than an anamorphic model was the significant correlation between site index and the Z ratio. A better fit was achieved compared to the baseline model (equation [6.3}, Table 6.1). Given the same height at b.h. age 50 years (e.g., SI = 18 m), the height growth curves for each site series group obtained from the four site series  197  (a) 2.4  (b) 2.4  22  22  0  2.0  1.8  1.6  T^T  T  1.6  1.4  1.4 t+W'11+1/111103 10 0318'SCNSOSV°  Site association  Plant association  (c) 2.4  (CO 2.4  22  22  2.0  2.0  0  0  1.8  1.8  1.6  1.6  1.4  vasOoosysirosS  iox 55silvs13 N.I.  20 30 91 32 33 94 95 36 40 41 42 43 50  1.4  11 CEFGHI  I  J^L  Site group Subzone or variant  Figure 6.8. Box plots showing the Z ratio stratified according to (a) plant associations, (b) site associations, (c) biogeoclimatic units, and (d) site groups. Symbols for plant associations, site associations, biogeoclimatic units, and site groups are given in Table 4.6, Figure 4.14, Figure 2.1, and Figure 4.16, respectively.  198  2.4  I^I^f^1  2.2 T  I  T  1.6  1.4  I (^I  1^(  ^ ^ ^ A B C D Site series group  Figure 6.9. Box plot showing the Z ratio stratified according to four site series groups. Symbols for the site series groups are given in Table 6.7.  Table 6.6. Site series group-specific models: coefficients, coefficients of determination, and standard errors of estimates for equation [6.3] fitted to each of the four site series groups. The codes for site series are given in Appendex 3. Site series group^site series  bl  b2  b3  R2  SEE (m)  206, 301, 306  6.175  -1.582  -0.1382  0.998  0.28  302, 303, 305, 311, 330, 340, 350, 361, 412, 424, 426, 434, 435, 506  7.465  -1.429  -0.5512  0.992  0.69  201, 202, 203, 205, 320, 405, 411  9.473  -1.536  -1.066  0.987  0.87  436, 501, 502  9.696  -1.846  -0.8361  0.999  0.26  200  40  30  Site series group: A ^ B C -D ^ D  0  0^20^40^60^80  100  Age @ b.h. (year)  Figure 6.10. White spruce height growth curves for each of the four site series groups at site index 18 m obtained from the site series groupspecific height growth models specified in Table 6.6. Symbols for site groups are given in Table 6.6.  201  group-specific models specified in Table 6.6 are shown in Figure 6.10. For site series group A, white spruce showed a fast early growth and a sluggish later growth, while the opposite trend was apparent for site series group C and D. The shape of the height growth curves produced by these models varied with site index and site series groups. 6.4.3.2.^Site-specific Models without Site Index as a Predictor To explicitly represent site index in a height growth model, a site index prediction model (equation [43], developed in Chapter 5) was used with the baseline height growth model (equation [6.3], Table 6.1) as a model system to predict height growth at any age. Two strategies can be used to calibrate the complex model. The first one is to fit the two simple models separately, and then to substitute the site index prediction model into the height growth model. The second one, thought to be superior (Borders 1989, Greene 1990), is to fit the two simple models as a system. Since the site index prediction model (equation [43]) was not derived statistically, the model system was replaced by a complex model (equation [6.7]) in which white spruce site index was replaced by 23.2*Lma*Ln. The following equation was obtained using the data from the 102 study stands:  1 + exp(8.909 - 1.477*Ln(50) -1.007*1n(23.2*Lma*Ln - 1.3) [6.8] H = 1.3 + (23.2*Lma*Ln - 1.3)* ^ 1 + exp(8.909 - 1.477*Ln(A) - 1.007*ln(23.2*Lma*Ln - 1.3)  R2 = 0.97, SEE = 1.25 m The minor differences in the b coefficients between the complex model and the baseline model were probably due to the difference between measured site index and predicted site index. The SEE of the complex model was slightly larger than  202  that of the baseline model. This could be expected, as the measured site index was replaced by the estimated site index from the semi-empirical model (equation [43]), which introduced site index prediction errors into the model. As Lma and Ln vary with combination of SMAR and SNR, the complex height growth model is, in fact, edaphic unit-specific. Given optimum conditions of soil moisture and aeration, a plot of the height growth curves according to SNRs showed that the curves were well separated among SNRs, except between the poor and the medium SNR, and ordered from the very rich (top) to the very poor (bottom) SNRs (Figure 6.11). A similar plot stratifying the study stands according to SMARs under optimum soil nutrient conditions also showed separation (Figure 6.12). Height growth curves were ordered from the moist and adequately aerated sites (top) through wet to very wet and deficiently aerated sites (bottom). The curves for slightly dry, fresh, moist, and very moist sites with adequate aeration (high productivity), were close to each other, followed by the curves for adequately aerated and moderately dry sites and slightly dry and very moist sites with restricted aeration (medium aeration). The curves for wet sites with restricted aeration and wet to very wet sites with deficient aeration (low productivity), were separated from each other and from the other curves. As the variation in site index within a site group was relatively small (Figure 5.16), it should be possible to develop site group-specific height growth models using age as the single predictor. This means implicit representation of site index in these models (i.e., a constant role of site index in each site group). The b coefficients, R 2 , and SEE for each site group are given in Table 6.7 for equation [6.1]. Coefficient b1, which was highly correlated with the mean site index of each site group (r = 0.92), represents the average asymptotic value for each site group. The highest values were found for site groups G and I, and the  203  40  SNR:  30  VP P  -_-  M R VR  0  0^20^40^60^80  ^  100  Age @ b.h. (year)  Figure 6.11. White spruce height growth curves (equation [6.8]) for each of the five soil nutrient regimes under the optimum soil moisture and aeration conditions.  204  40  SMAR: MDa ^ DFr SDa  30  Fa _ _ Ma VMa _ _ _ Mr Wr ^ Wd  0  0^20^40^60^80  ^  100  Age @ b.h. (year)  Figure 6.12. White spruce height growth curves (equation [6.8]) for each of the nine soil moisture-aeration regimes (SMARs) under optimum soil nutrient conditions. Symbols for SMARs are given in Figure 4.6.  Table 6.7. Site group-specific models: coefficients, coefficients of determination, and standard errors of estimates for equation [6.1] fitted to each of the seven site groups (Figure 4.16) characterized by a unique range of soil moisture-aeration regimes (SMARs) and soil nutrient regimes (SNRs). Symbols for SMARs and SNRs are given in Figure 4.16 and Table 4.8, respectively. Site group^Ecological site quality^bl^b2^b3^R2^SEE SMARs^SNRs C^MD, DFr^VP to M^28.65^0.02061^1.544^0.954^1.40 F^SDa to Ma^P to M^30.93^0.02560^1.559^0.975^1.09 G  SDa to Ma^R to VR^37.50^0.02192^1.498^0.967^1.40  I^VMa^R to VR^38.63^0.01868^1.434^0.988^0.90 K  Mr^M to VR^22.42^0.03795^2.206^0.965^1.18 Wr^R to VR  J  Mr^VP to P^24.66^0.02223^1.921^0.938^1.17 Wr^VP to M  L  Wd^VP to M^10.70^0.01984^1.680^0.886^0.93  206  40  Site group:  30  C ^  F G ^ I^- - J ^ K _ _ L  10  0  0^20^40^60^80  100  Age @ b.h. (year)  Figure 6.13. White spruce height growth curves for each of the seven site groups obtained from the site group-specific height growth models specified in Table 6.7. Symbols for the site groups are given in Figure 4.16.  207  lowest value for site group L. The shapes of the average curve for each site group were also different, as indicated by coefficients b2 and b3 (Figure 6.13), which actually represent the average trend of height over age development (i.e., the average height growth pattern in each site group). However, this average trend may or may not reflect the real height growth pattern of the individual stands included in the site group. Height growth curves for site groups F, G, and I were very close to each other before age 20, but spread afterward. The height growth curve for the site group G was consistently above any of the other curves up to 100 years. Height growth curves for site groups F and I were nearly identical up to 60 years. After 60 years, the height growth for site group I surpassed that of site group F, and approximated the height growth on site group G after 100 years. Height growth curves for site groups C and J intersected twice (approximately at 15 and 70 years). Before the first and after the second intersections, height growth of the stands in site group C was superior to those in site group J. Height growth of the stands in site group L was the lowest of all the site groups. Although it was consistently lower, the height growth curve for site group K paralleled that of site group C. Similar trends and separation were found when the differential forms of the site group-specific models were plotted (Figure 6.14). The maximum annual height increment decreased in the order of G>F>I>J>C>K>L among the site groups and maintained this order until approximately 25 years of age. After this age, several shifts occurred. For example, the increment of the stands in site group I increased and, surpassed that of other site groups after 60 years. Similarly, after about 50 and 70 years, the increment of the stands in site groups C and K surpassed those of site groups J and F, respectively. Site group L  208  0.60 0.54 Site group: C ^  F G ^  (1.)  0.36 74  I^_ _ J ^  `, \ •  ;.4 c.) 0.30  L  4 0.24 a) .4 0.18  0.12 0.06 5 ••••••■  0.00  0  20  40^60  80  100  Age @ b.h. (year)  Figure 6.14. White spruce annual height increment curves for each of the seven site groups obtained from the differential forms of the site groupspecific height growth models specified in Table 6.7. Symbols for the site groups are given in Figure 4.16.  Table 6.8. Test of the site group-specific height growth models specified in Table 6.7 using the data of this study and Wang et al. (1992). Site^Residual range^Mean bias^Mean error^Relative error^Number of group^(m)^ (m)^(m)^(%)^observations Non-independent test (using the data of this study) C  -4.11 - 3.53  -0.002  1.07  8.2  180  F  -3.04 - 3.07  0.005  0.84  5.6  208  G  -3.00 - 3.12  0.008  1.02  6.6  125  I  -1.99 - 1.81  -0.006  0.74  4.6  27  K  -3.08 - 3.00  0.031  0.87  6.6  70  J  -2.33 - 2.14  -0.003  0.93  10.8  35  L  -1.48 - 1.41  -0.002  0.71  12.2  27  Independent test (using the data of Q. Wang et al. 1992) C  -3.24 - 1.30  -0.98  1.61  8.1  8  F  -4.43 - 4.93  -0.39  1.58  6.8  15  G  -4.17 - 5.06  0.28  1.81  6.7  25  ALL  -4.42 - 5.06  -0.18  1.70  6.9  48  210  maintained its lowest height growth rate until about 80 years, but afterward the rate increased and surpassed that of site group J. Basic statistics for the seven site group-specific models and the results of tests on independent data are given in Table 6.8. Due to the limited independent data, only three (site groups C, F, and G) of the seven models were tested. Although some biases were found and the average errors were slightly higher than those obtained from the non-independent tests, the relative errors were very comparable. Considering that the assignments of stands into site groups were only based on field estimation of SMRs and SNRs and that SARs were not identified in the study of Q. Wang et al. (1992), better results from the independent test could not be expected. Despite the limitations of the independent test, the developed site group-specific height growth models appear to be reliable. To demonstrate this conclusively, an adequate independent data set must be obtained for all site groups. 6.4.4. Comparison and Evaluation of Height Growth Models Based on the performance of the selected anamorphic and polymorphic models, Goudie and Mitchell's model was selected as the baseline model. Can the baseline model be improved? Can site index in the model be replaced and how? This study attempted to develop one height growth pattern-specific (based on characterization of height growth pattern) and three site-specific height growth models (based on ecological site classification). The five models (i.e., the baseline model, the height growth patternspecific model, and three site-specific height growth models) are compared in Table 6.9. No significant biases were found in any of the five models. In terms of the precision of height prediction, the height growth pattern-specific model was the best, as it had the narrowest range of residuals, the lowest average error,  Table 6.9. Comparisons of the baseline, the height growth pattern-specific, and the three site-specific height growth models developed in the study. Model  ^  ^ ^ ^ Residual range Mean^ bias Mean error^Relative error Data required ^ ^ (m)^(%) for prediction (m) (m)  Goudie & Mitchell model  (-4.10, 3.25)  -0.004  0.61  4.14  Site index  Height growth pattern-specific model  (-2.22, 1.54)  -0.001  0.25  1.72  Site index and height at 30 and 60 years  Site series group-specific model  (-3.56, 3.14)  -0.011  0.47  3.23  Ecological site classification and site index  Site group-specific model  (-4.11, 3.12)  0.005  0.93  6.53  Ecological site classification  Edaphic unit-specific model  (-3.77, 5.77)  -0.051  0.96  6.81  Ecological site classification  212  and the lowest relative error. It was followed by the site series group-specific height growth model, the baseline model, the site group-specific height growth model, and the edaphic unit-specific height growth model. In terms of data requirements, the height growth pattern-specific model required not only site index, but also a pair of heights at the b.h. age of 30 and 60 years. Without stem analysis or permanent sample plot records on height growth, this model can not be applied to height estimation. Thus, this model is not applicable unless the Z ratio can be reliably predicted without using the pair of heights. An attempt to circumvent this problem was made in this study, but no solution was found. Site index is required by the baseline model and the site series group-specific model. Since the site series group-specific model is, in fact, the baseline model with the coefficients specified for each site series group, site classification is also required. However, this model results in improved prediction precision over the baseline model. The models (i.e., site group-specific model and the edaphic unit-specific model) that do not require any stand information, not even the estimation of site index, had slightly inferior prediction precision compared to the baseline model. However, their average errors of 0.93 and 0.96 m and relative errors of 6.53 and 6.81%, respectively, likely are sufficient for estimating dominant height for most forest management purposes. Although these two models were derived using different strategies, they gave very similar results. The edaphic unit-specific model could be further improved if Lma, Ln, and C were more precisely determined from processoriented submodels. Based on the above comparisons, it is clear that selecting a model for predicting white spruce dominant height depends on the kind of information that is available. If crop stands, which are suitable for determining site index, and site classification for the study area are both available, the site series group-  213  specific model would be the best prediction model. If site index as well as heights at b.h. ages of 30 and 60 years are available, the height growth pattern-specific model would be the best prediction model. If site index was the only information available, Goudie and Mitchell's model would be the best prediction model. If site classification was the only available information, the site group-specific model or edaphic unit-specific model would be the best models. A computer program was written for calculating the average dominant height of any white spruce stand at any age on any site within the study area (Appendix 6). Depending on the available information, the program selects the height growth pattern-specific model, the baseline model, the site series groupspecific model, or the site group-specific model as the best prediction model.  6.4.5. Discussion If high quality data were available, the 'right' model was selected, and the correct technique was used to fit the model, the most important factors for determining the quality of the height growth curves would be the variation of the height growth pattern. Although many studies recognized that height growth patterns vary with climate and soils, there have been very few conclusive reports on how to characterize height growth patterns and how to identify the causes of the variation in height growth pattern. One of the most important reasons for this is lack of systematic studies on height growth pattern within a framework of ecological site classification. Most of the previous approaches to identifying height growth patterns were based on comparing the shape of average height growth curves fitted to strata according to some classification scheme. These approaches ignored the possible significant differences in height growth patterns within the same  214  stratum, and failed to provide a quantitative measure for the height growth pattern. Ziede's method was used in this study as well as in several other studies (Hoyer and Chawes 1980, Milner 1987) to overcome this problem. Although Ziede's method was successful in stratifying the study stands into groups with similar height growth patterns, neither site index nor edaphic factors (SNRs, SMRs, and SARs) were found to be the major cause of the different height growth patterns. Regional and local climate, stand history, and genetic differences may all contribute to the variation. Differences in height growth patterns among regional climates were also implied by comparing height growth curves developed for this study area, the Lake States, and the Northwest Territories. To test the impact of regional climate, future studies should analyze samples across the widest possible climatic range. To test the effect of local climate on the variation, local climatic conditions should be measured and added to other environmental data. If climate was identified as a major cause of the variation, significant improvement in height growth models by including climatic variables as additional predictors would be possible. Difficulties in identifying the exact stand history and genetic differences would prevent these factors from being additional predictors of height growth, even if they proved to be major causes of the variation. Coordination of growth and yield models and ecological site classification requires adapting growth and yield models to an ecological classification system (Crow and Rauscher 1984). Since the measures of ecological site quality provided by the BEC system have shown to be highly correlated with white spruce height growth, site classification seems to have a logical role in height growth modelling for this species. The three methods used in this study to link height growth modelling to ecological site classification were successful. The first method used site series groups as prediction strata, which were uniform in  215  height growth pattern but different in site index. Within each stratum, age and site index were used to predict dominant height. The second method used a site index prediction model to estimate site index in the height growth model. The resultant edaphic unit-specific model used age and three site-related parameters (Lma, Ln, and C) to predict the dominant height. The third method used each site group as a homogeneous productivity unit. Richards' model, using only age as predictor, was fitted to each site group. In terms of overall prediction precision, only the model based on site series groups was superior to the best traditional polymorphic height growth model. However, the site group-specific model and the edaphic unit-specific model do not need site index as predictor; they can be used under any site conditions. Most importantly, site-specific models can also be applied to conditions of changing environments. As long as the changes in ecological site quality resulting from environmental changes are predictable, the effects of the changes on height growth can be predicted through the use of these models. This could be a desirable feature to add to empirically-based height growth models given the fact that growth conditions can no longer be considered as immutable. 6.5. CONCLUSIONS 1.  All the tested anamorphic and polymorphic height growth models adequately described the height growth of white spruce dominants in the study stands, with Goudie and Mitchell's model being somewhat superior.  2.  Comparisons between height growth curves for the study stands and those used by Goudie and Mitchell, Alemdag, and Ek suggested that the pattern of white spruce height growth varies with large-scale climatic variations.  216  3.  The Z ratio was useful for quantitative characterization of white spruce height growth patterns. Height growth pattern-specific models developed for each height growth pattern group improved prediction precision over traditional height growth models.  4.  Despite the fact that white spruce height growth patterns changed among stands, both site index and synoptic measures of ecological site quality were not the major source of this variation.  5.  The three methods proposed in this study (i.e., the site series groupspecific, edaphic unit-specific, and site group-specific models) could be considered for developing ecologically based, site-specific height growth models.  6.  If suitable site trees are present in a stand, then Goudie and Mitchell's model is most suitable for predicting white spruce dominant height in the SBS zone; in other situations, the site-specific models are more appropriate.  217  7. GENERAL SUMMARY AND CONCLUSIONS  This study advanced the methodology of biogeoclimatic ecosystem classification and the knowledge of growth behavior of white spruce in relation to ecological site quality in the Sub-Boreal Spruce zone of British Columbia. The major contribution to the methodology was in (1) advancing the concept of soil aeration regime, (2) developing the concept of integrated soil moisture plus aeration regime, and (3) quantitatively characterizing soil moisture and nutrient regimes. The major contribution to the knowledge of white spruce height growth behavior was in (1) linking height growth and site index to ecological site quality, (2) developing empirical and semi-empirical models for site index prediction, (3) quantifying height growth patterns and identifying the possible causes of its variation, and (4) developing height growth models for average dominant height prediction at any age with or without stand variables. The results of the study justified the following conclusions: 1.  The methods and scheme of biogeoclimatic ecosystem classification offers useful measures and means for characterizing ecological quality of forest sites and for predicting white spruce site index in the Sub-Boreal Spruce zone of British Columbia.  2.  Soil/topography, understory vegetation, and foliar nutrient variables, as well as the synoptic variables (e.g., soil moisture, soil aeration, and nutrient regimes), explained a large amount of the total variance in white spruce site index in the Sub-Boreal Spruce zone.  3.^The similarity in relationships between white spruce site index and both categorical and continuous measures of ecological site quality suggests that the categorical measures of soil moisture, aeration, and nutrients are good estimates of their direct measures.  218  4.  When quantitatively characterized by Z ratios, white spruce height growth patterns varied with site, but neither site index nor the selected measures of ecological site quality were the major source of the variation in pattern. Although it could not be proven conclusively, it is suggested that climatic variations are the major source of this variation.  5.  Site-specific models developed within the framework of site classification gave reliable estimates of the average dominant height in the studied white spruce stands without using site index as a predictor.  6.  The results of this study corroborated two contemporary propositions: (1) stratification of a study area into smaller or ecologically more homogeneous segments results in strengthening site-productivity relationships, and (2) the effect of site can be adequately represented in growth models without using site index.  7.  The results of this study also suggested: (1) integrating or synthesizing easily measurable environmental variables into synoptic measures can facilitate our understanding of site-productivity relationships, and (2) different ecosystem attributes can reinforce each other in explaining the variation in productivity.  219  8. LITERATURE CITED Aber, J.D. and J.M. Melillo. 1984. Nutrient cycling models and land classification. Pp. 205-217 in J.G. Bockheim (ed.), Forest land classification: experience, problems, perspectives. Univ. of Wisconsin, Madison, WI. Agren, G.I. and B. Axelsson. 1980. PT - a tree growth model. in T. Persson (ed.), Structure and function of northern coniferous forests - an ecosystem study. Ecol. Bull. 32: 525-536. Ahlgren, C.E. and H.L. Hansen. 1957. Some effects of temporary flooding on coniferous trees. J. For. 55: 647-650. Alemdag, I.S. 1971. Preliminary site index curves for white spruce and lodgepole pine in Upper Liard River area, Yukon Territory. Inf. Rep. FMR-X-33, Dept. Environ., Can. For. Serv. Ottawa. 11 pp. Alemdag, I.S. 1988. Site index equations for white spruce in the Northwest Territory, Canada. For. Ecol. Manage. 23: 61-71. Alemdag, I.S. 1991. National site index and height growth curves for white spruce in natural stands in Canada. Can. J. For. Res. 21: 1466-1474. Annas, R.M. and R. Coupe (eds.). 1979. Biogeoclimatic zones and subzones of the Cariboo Forest Region. B.C. Min. For., Victoria, B.C. Anonymous. 1976. Technicon Autoanalyzer. II. Methodology: individual/simultaneous determination of nitrogen and/or phosphorus in BD acid digests. Industrial Method No. 328/74W/A. Technon Corp., Tarrytown, New York. Anonymous. 1982. Canadian climate normals (1951-1980): temperature and precipitation (Vol. 6). Environment Canada, Atmospheric Environment Service, Ottawa, Ontario. 276 pp. Armson, A.K. 1977. Forest soils. Univ. of Toronto Press, Toronto, Ontario. 390 pp. Bailey, R.G. 1981. Integrated approaches to classifying land as ecosystems. Pp. 95-109 in P. Laban (ed.), Proc. Workshop on Land Evaluation for Forestry. Intern. Workshop of the IUFRO/ISSS, Institute of Land Reclamation and ImprovementllLRl, Wageningen, The Netherlands. Bakuzis, E.V. and H.L. Hansen. 1962. Ecographs of shrubs and other understory species of Minnesota forest communities. Minn. For. Notes 117, Univ. of Minnesota, St. Paul, MN. Bakuzis, E.V. 1969. Forestry viewed in an ecosystem perspective. Pp. 189-258 in G.M. Van Dyne (ed.), The ecosystem concept in natural resource management. Academic Press, New York.  220  Ballard, T.M., 1982. Soil interpretation for forestry. Pp. 156-167 in Land Manage. Rep. 10. B.C. Min. of For., Victoria, B.C. Ballard, T.M. 1983. Forest soil and tree nutrition. Pp. 207-219 in S.B. Watts (ed.), Forestry Handbook for British Columbia (4th ed.), Forestry Undergraduate Soc., Univ. of British Columbia, Vancouver, B.C. Ballard, T.M. and R.E. Carter. 1986. Evaluating forest stand nutrient status. Land Manage. Rep. No. 20, B.C. Min. For., Victoria, B.C. 60 pp. Bardsley, C.E. and J.D. Lancaster. 1965. Sulphur. Pp. 1102-1116 in C.A. Black, D.D. Evans, J.L. White, L.E. Ensminger, and F.E. Clark (eds.), Methods of soil analysis. Agronomy No. 9, Am. Soc. Agron., Madison, WI. Barnes, B.V., KS. Pregitzer, T.A. Spies, and T.H. Spooner. 1982. Ecological forest site classification. J. For. 80: 493-498. Barnes, B.V. 1984. Forest ecosystem classification and mapping in BadenWiirttenberg, West Germany. Pp. 49-65 in J.G. Boc eim (ed.), Forest land classification: experience, problems, perspectives. University of Wisconsin, Madison, WI. Barnes, B.V. 1986. Varieties of experience in classifying and mapping forestland ecosystems. Pp. 5-23 in G.M. Wickware and W.C. Stevens (eds.), Site classification in relation to forest management. Can. For. Serv., Great Lakes For. Centre, COJFRC Proceedings 0-P-14. Sault St. Marie, Ontario. Beck, D.E. and D.B. Trousdell. 1973. Site index: accuracy of prediction. USDA For. Serv. Res. Pap. SE-108, Southeastern For. Exp. Stn., Asheville, NC. 7 pp. Berry, A.B. 1978. Metric yield tables based on site class and spacing for white spruce plantationa at the Petawawa Forest Experiment Station, Can. For. Serv. Inf. Rep. PS-X-70. Petawawa, Ontario. 14 pp. Binkley, D. and P. Vitousek. 1989. Soil nutrient availability. Pp. 75-96 in R.W. Pearcy, J. Ehleringer, H.A. Mooney, and P.W. Rundel (eds.), Plant physiological ecology: field methods and intrumentation. Chapman and Hall Ltd., London. Black, T.A., 1982. Determining soil climate for soil surveys. Pp. 216-235 in Land Manage. Rep. 10. B.C. Min. of For., Victoria, B.C. Borders, B.E. 1989. Systems of equations in forest stand modelling. For. Sci. 35: 548-556. Bossel H. 1986. Dynamics of forest dieback: systems analysis and simulation. Ecol. Modelling. 34: 259-228. Bossel H. 1991. Modelling forest dynamics: moving from description to explanation. For. Ecol. Manag. 42: 129-142.  221  Brais, S. and C. Camire. 1992. Keys for soil moisture regime evaluation for northwestern Quebec. Can. J. For. Res. 22:718-724. Brand, D.G. and P.S. Janes. 1988. Growth acclimation of planted white pine and white spruce seedlings in response to environment conditions. Can. J. For. Res. 18: 320-329. Bray, R.H. and L.T. Kurtz. 1945. Determination of total, organic, and available phosphorus in soils. Soil Sci. 59: 39-45. Bremner, J. 1965. Nitrogen availability indices. Pp. 1324-1345 in C.A. Black (ed.), Agronomy Monograph 9. American Society of Agronomy, Madison, WI. Bremner, J. and M.A. Tabatabai. 1971. Use of automated combustion techniques for total carbon, total nitrogen, and total sulfur analysis of soils. Pp. 1-16 in L.M. Walsh (ed.), Intrumental methods for analysis of soils and plant tissue. Soil Sci. Soc. Am., Madison, WI. Broadfoot, W.M. 1969. Problems in relating soil to site index for southern hardwoods. For. Sci. 15: 354-364. Brooke, R.C., E.B. Peterson, and V.J. Krajina. 1970. The subalpine Mountain Hemlock zone. Ecol. West. N. Amer. 2: 148-349. Bruce, D. 1923. Anamorphosis and its use in forest graphics. J. for. 21: 773783. Bruce, D. 1926. A method of preparing timber yield tables. J. Agri. Res. 32: 543-557. Bull, H. 1931. The use of polymorphic curves in determining site quality in young red pine plantations. J Agri. Res. 43: 1-28. Burger, D. 1972. Forest site classification in Canada. Mitt. Vereins Forstl. Standorsk. Forstpfl. zucht. (Stuttgar) 21: 20-36. Burger, D. and G. Pierpoint. 1990. Trends in forest site and land classification in Canada. For. Chron. 66: 91-96. Cajander, A.K. 1926. The theory of forest types. Acta. For. Fenn. 2: 11-108. Canada Soil Survey Committee (CSSC). 1978. The Canadian system of soil classification. Can. Dept. Agric. Publ. No. 1646. Supply and Services Canada, Ottawa, Ontario. 164 pp. Canneell, R.Q. 1977. Soil aeration and compaction in relation to root growth and soil management. Pp. 1-86 in T.H. Coaker (ed.), Applied Biology (Vol II). Academic Press, New York. Carlyle, J.C. 1986. Nitrogen cycling in forested ecosystems. For. Abst. 47: 307-336.  222  Carmean, W.H. 1956. Suggested modifications of the standard Douglas-fir site curves for certain soils in southwestern Washington. For. Sci. 2: 242-250. Carmean, W.H. 1970. Tree height growth patterns in relation to soil and site. Pp. 499-512 in C.T. Youngberg and C.B. Davey (eds.), Tree growth and forest soils. Third North American Forest Soils Conference Proceedings, Raleigh, NC. Carmean, W.H. 1972. Site index curve for upland oaks in the Central States. For. Sci. 18: 109-120. Carmean, W.H. and C.T. Kok. 1974. Site quality for Caribbean pine in peninsular Malaysia. The Malaysian Forester. 37: 109-119. Carmean, W.H. 1975. Forest site quality evaluation in the United States. Adv. Agron. 27: 209-269. Carter, R.E. and K. Klinka. 1990. Relationships between growing-season soil water-deficit, mineralizable soil nitrogen and site index of Coastal Douglas-fir. For. Ecol. Manage. 30: 301-311. Chatterjee, S. and B. Price. 1977. Regression analysis by example. WileyInterscience, New York. 288 pp. Cheyney, E.G. 1942. American silvics and silviculture. University of Minnesota Press, Minneapolis, MN. 477 pp. Cieszewski, C.J. and I.E. Bella. 1989. Polymorphic height and site index curves for lodgepole pine in Alberta. Can. J. For. Res. 19: 1151-1160. Cieszewski, C.J. and I.E. Bella. 1991. Polymorphic height and site index curves for the major tree species in Alberta. Forestry Canada, Northwest Region, For. Manage. Note 51. 8 pp. Clutter, J.L., J.C. Fortson, L.V. Pienaar, G.H. Brister, R.L. Bailey. 1983. Timber management: a quantitative approach. John Wiley & Sons, Inc., New York. 333 pp. Coates, K.D., S. Haeussler, S. Lindeburgh, R. Pojar, and A.J. Stock. 1992. Ecology and silviculture of interior spruce in British Columbia (draft). B.C. Ministry of Forests, Prince Rupert Forest region, Smithers, B.C. 253 pp. Coile, T.C. 1935. Relation of site index for shortleaf pine to certain physical properties of the soil. J. For. 33: 726-730. Coile, T.S. 1952. Soil and the growth of forests. Pp. 329-398 in A.G. Norman (ed.), Advances in Agronomy. Academic Press Inc., New York. Corns, I.G.W. 1978. Tree growth prediction and plant community distribution in relation to environmental factors in lodgepole pine, white spruce, black spruce, and aspen forests of western Alberta foothills. Ph.D. dissertation, University of Alberta, Edmonton. 229 pp.  223  Corns, I.G.W. and D.J. Pluth. 1984. Vegetational indicators as independent variables in forest growth prediction in west-central Alberta, Canada. For. Ecol. Manage. 9: 13-25. Courtin, P.J., K. Klinka, M.C. Feller, and J.P. Demerschalk. 1988. An approach to quantitative classification of nutrient regimes of forest soils. Can. J. Bot. 66: 2640-2653. Crow, T.R. and H.M. Rauscher. 1984. Forest growth model and land classification. Pp. 109-204 in J.G. Bockheim (ed.), Forest land classification: experience, problems, perspectives. Univ. of Wisconsin, Madison, WI. Curtis, R.O., D.J. DeMars, and F.R. Herman. 1974. Which dependent variable in site index-height-age regression? For. Sci. 20: 74-87. Damman, A.W.H. 1979. The role of vegetation in land classification. For. Chron. 55: 175-182. Daniels, T.W., J.A. Helms, and F.S. Baker. 1979. The principles of silviculture (2nd ed.). McGraw-Hill, New York. 500 pp. Daubermire, T. 1976. The use of vegetation in assessing the productivity of forest lands. The Botanical Review. 42: 115-143. Davis, L.S. and K.N. Johnson. 1987. Forest Management (3rd ed.). McGrawHill Book Company, New York. 790 pp. Delong, C., G. Hope, and A. Mcleod. 1984. A field guide for the identification and interpretation of ecosystems of the SBSe2 in the Prince George Forest Region. First approximation. B.C. Min. For., Prince George, B.C. (Draft report). Delong, C. and A. Mcleod (compilers). 1985. A field guide for the identification and interpretation of ecosystems of the SBSk2 in the Prince George Forest Region. First approximation. B.C. Min. For., Prince George, B.C. (Draft report). Dillon, W.R. and M. goldstein. 1984. Multivariate analysis: methods and applications. John Wiley & Sons Inc., New York. 587 pp. Dyer, M.E. and R.L. Bailey. 1987. A test of six methods for estimating true heights from stem analysis data. For. Sci. 33: 3-13. Eis, S. 1962. Statistical analysis of several methods for estimation of forest habitats and tree growth near Vancouver, B.C. For. Bull. No. 4, Faculty of For., Univ. of British Columbia, Vancouver, B.C. 76 pp. Eis, S. 1970. Root-growth relationships of Juvenile white spruce, alpine fir and lodgepole pine on three soils in interior British Columbia. Can. For. Serv., Pub. 1276. Ottawa, Ontario. 10 pp.  224  Ek, A.R. 1971. A formula for white spruce site index curves. For. Res. Note 161. University of Wisconsin, Madison, WI. 2 pp. Emanuel, J. 1987. A vegetation classification program (VTAB). Faculty of For., Univ. of British Columbia, Vancouver, B.C. (Mimeographed). 26 pp. Ford, D.E. 1983. What do we need to know about forest productivity and how can we measure it? Pp. 45-51 in R. Ballard and S.P. Gessel (eds.), IUFRO Symposium on Forest Site and Continuous Productivity. USDA For. Serv. Gen. Tech. Rep. PNW-163. Pacific NW For. and Range Exp. Stn., Portland, OR. Fox, D.J. and K.E. Guire. 1976. Documentation of MIDAS. Statistical Research Laboratory, Univ. of Michigan, Ann Anbor, MI. 203 pp. Gagnon, J.D. and J.D. MacArthur. 1959. Ground vegetation as an index of site quality in white spruce plantations. Can. Dept. N. Affairs and Nat. Res., For. Branch, For. Res. Div. Tech. Note 70. Ottawa, Ontario. 12 pp. Gaines, T.P. and G.A. Mitchell. 1979. Boron determination in plant tissues by the azomethine H method. Comm. Soil Sci. Plant Anal. 10:1099-1108. Gale, M.R. 1987. A forest productivity index model based on soil- and rootdistribution characteristics. Ph. D. Thesis. Univ. of Minnesota, St. Paul, MN. 159 pp. Gauch, H.G. Jr. 1982. Multivariate analysis in community ecology. Cambridge University Press, London. 298 pp. Gevorkiantz, S.R. 1957. Site index curves for white spruce in the Lake States. Tech. Note 474. USDA For. Serv., St. Paul, MN. 2 pp. Giles, D.G., T.A. Black, and D.L. Spittlehouse. 1984. Determination of growing season soil water deficits on a forested slope using water balance analysis. Can. J. For. Res. 15: 107-114. Goodall, D.W. 1970. Statistical plant ecology. A Rev. Ecol. Syst. 1: 99-124. Goudie, J.W. and K.J. Mitchell. 1986. The first approximation managed stand yield tables for interior white spruce: initial density. (draft). B.C. Min. For., Victoria, B.C. 35 pp. Greene, W.H. 1990. Econometric analysis. Macmillan Publishing Company. New York. 783 pp. Green, R.N., P.L. Marshall, and K. Klinka. 1989. Estimating site index of Douglas-fir (Pseudotsuga menziesii [Mirb.] Franco) from ecological variables in southwestern British Columbia. For. Sci. 35: 50-63. Greig-Smith, P. 1983. Quantitative plant ecology (3rd ed.). Univ. of Califolia Press, Berkeley and Los Angeles. 359 pp.  225  Greweling, T. and M. Peach. 1960. Chemical soil tests. Cornell Agric Exp. Sta. Bull. 960, New York, NY. 304 pp. Guthrie, T.F. and L.E. Lowe. 1984. A comparison of methods for total sulpher analysis of tree foliage. Can. J. For. Res. 14:470-473. Hagglund, B. and J-E. Lundmark. 1977. Site index estimation by means of site properties: Scots pine and Norway spruce in Sweden. Studia Forestalia Suecica No. 138. Hagglund, B. 1981. Evaluation of forest site productivity. For. Abst. 42: 515527. Hahn, J.T. and W.H. Carmean 1982. Lake States site index curves formulated. USDA For. Serv. Gen. Tech. Rep. NC-88, St. Paul, MN. 5 PP. Hale, M.E., Jr. and W.L. Culberson. 1970. A fourth checklist of the lichens of the continental United States. The Bryologist. 73: 499-543. Halliday, W.E.D. 1937. A forest classification for Canada. Can. Dep. Mines Resour. Domin. Forest Serv., Bull. 89. Ottawa, Ontario. 50 pp. Harding, R.B. 1982. Site quality evaluation for white spruce plantations in northern Minnesota. Ph.D. thesis. University of Minnesota, St. Paul, MN. 135 pp. Heger, L. 1968. A method of constructing site index curves from stem analysis. For. Chron. 44: 11-15. Heger, L. 1971. Site-index/soil relationships for white spruce in Alberta mixedwoods. Dept. Environ., Can. For. Serv. Inf. Rep. FMR-X-32. Ottawa, Ontario. 15 pp. Heiberg, S.O. and D.P. White. 1951. Potassium deficiency of reforested pine and spruce stands in northern New York. Soil Sci. Amer. Proc. 15: 369376. Heilman, P.E. 1979. Minerals, chemical properties, and fertility of forest soils. Pp. 121-136 in P.E. Heilman, W.H. Anderson, and D.M. Baumgartner (eds.), Forest soils of Douglas-fir region. Washington State Univ., Pullman, WA. Heimburger, C.C. 1934. Forest-type studies in the Adirondack region. N.Y. (Cornell) Agric. Exp. Stn. Mem. 165. New York. 122 pp. Henderson, G.S., R.D. Hammer, and D.F. Grigal. 1990. Can measureable soil properties be integrated into a framework for characterizing forest productivity? Pp. 137-154 in S.P. Gessel, D.S. Lacate, G.F. Weetman, and R.F. Powers (eds.), Sustained Productivity of Forest Soils. Proceedings of the 7th North American Forest Soil Conference, University of British Columbia, Faculty of Forestry Publication, Vancouver, B.C.  226  Hills, G.A. 1952. The classification of evaluation of site for forestry. Ontario Dept. of Lands and Forests Res. Rep. 24. Toronto, Ontario. 41 pp. Hope, G.D. (compiler). 1984. A field guide for the identification and interpretation of ecosystems of the SBSh in the Prince George Forest Region. First approximation. B.C. Min. For., Prince George, B.C. (Draft report). Host, G.E. and K.S. Pregitzer. 1991. Ecological species groups for upland forest ecosystems of northwestern Lower Michigan. For. Ecol. Manage. 43: 87-102. Hoyer, G.E. and R. Chawes. 1980. Application of Ziede's standardized growth curves and the two-point curve form estimation system to Pacific Northwest species. State of Washington, Dept. of Natural Resources, DNR Report No. 40. Olympia, Washington. 19 pp. llvessalo, Y. 1929. Notes on some forest (site) types in North America. Acta For. Finn. 34. Inions, G. 1990. Classification and evaluation of site in karri (Eucalyptus diversicolor F. Muell.) regeneration. I. Edaphic and climatic attributes. For. Ecol. Manage. 32: 117-124. Inions, G., G. Wardell-Johnson, and A. Annels. 1990. Classification and evaluation of site in karri (Eucalyptus diversicolor F. Muell.) regeneration. II. Floristic attributes. For. Ecol. Manage. 32: 125-134. Ireland, R.R., C.D. Bird, G.R. Brassard, W.B. Schofield, and D.H. Vitt. 1980. Checklist of the mosses of Canada. National Museum of Natural Science Publication in Botany, No. 8. National Museums of Canada, Ottawa, Ontario. 75 pp. Jahn, G. (ed.). 1982. Application of vegetation science to forestry. Dr W. Junk Publishers, The Hague, Netherland. 405 pp. Jeglum, J.K. 1974. Relative influence of moisture-aeration and nutrients on vegetation and black spruce growth in northern Ontario. Can. J. For. Res. 4: 114-126. John, M.K. 1970. Colorimetric determination of phosphorus in soil and plant materials with ascorbic acid. Soil Sci. 109: 214-220. Johnson, P.S. 1985. Regenerating oaks in the Lake States. Pp. 98-109 in J.E.Johnson (ed.), Proc. of a Conference on Challenges in Oak Management and Utilization. University of Wisicosin, Madison, WI. Jokela, E.J., E.H. White, and J.V. Berglund. 1988. Prediction Norway Spruce growth from soil and topographic properties in New York. Soil Sci. Soc. Am. J. 52: 809-915. Jones, J.R. 1969. Review and comparison of site evaluation methods. USDA Forest Service, Rocky Mountain Forest and Range Experiment Station Research Paper RM-51. Fort Collins, CO. 27 pp.  227  Kabzems, R.D. 1985. Characterization of soil nutrient regimes in Douglas-fir ecosystems in Drier Maritime Coastal Western Hemlock Zone. M.Sc. thesis. Faculty of forestry, University of British Columbia, Vancouver, B.C. Kabzems, R.D. and K. Klinka. 1987. Initial quantitative characterization of soil nutrient regimes. Can. J. For. Res. 17:1557-1564. II. Relationships among soils, vegetation, and site index. Can. J. For. Res. 17: 15651571. Kayahara, G.J. 1992. Ecological site quality and productivity of western hemlock ecosystems in Coastal Western Hemlock Zone of British Columbia. M.Sc. thesis. The University of British Columbia, Vancouver, B.C. 164 pp. Keeney, D.R. and J. Bremner. 1966. Comparison and evaluation of laboratory methods of obtaining an index of soil nitrogen availability. Agron. J. 58: 498-503. Keeney, D.R. 1980. Prediction of soil nitrogen availability in forest ecosystems: a literature review. For. Sci. 26: 159-171. Kelly, J. and M.J. Lambert. 1972. The relationship between sulphur and nitrogen in the foliage of Pinus radiata. Plant and Soil 37: 395-407. Kenety, W.H. 1917. A preliminary study of white spruce in Minnesota. Agri. Exp. Sta. Bull. 168, University of Minnesota, St. Paul, MN. 30 pp. Kienast, F. 1987. FORECE - A forest succession model for southern central Europe. Oak Ridge Nat. Lab., ORNDTM-10575, Environ, Sci. Div. Publ. No. 2980. Oak Ridge, TN. 73 pp. Kimmins, J.P. 1985. Future shock in forest yield forecasting: the need for a new approach. For. Chron. 61: 503-512. Kimmins, J.P. 1987. Forest Ecology. Macmillan, New York. 531 pp. Kimmins, J.P., P.G. Comeau, and W. Kurz. 1990. Modelling the interactions between moisture and nutrients in the control of forest growth. For. Ecol. Manage. 30: 361-379. Kirby, C.L. 1975. Site index equations for lodgepole pine and white spruce in Alberta. Can. For. Serv. North. For. Res. Cent. Inf. Rep. NOR-X-142. Edmonton, Alberta. Klinka, K. 1976. Ecosystem units, their classification, interpretation, and mapping in the University of British Columbia Research Forest. Ph.D. thesis. Faculty of Forestry, University of British Columbia, Vancouver, B.C. 622 pp. Klinka, K., R.N. Green, R.L. Trowbridge, and L.E. Lowe. 1981. Taxonomic classification of humus forms in ecosystems of British Columbia, Land Manage. Rep. No. 8. B.C. Min. For., Victoria, B.C. 54 pp.  228  Klinka, K., R.E. Green, P.J. Courtin, and F.C. Nuszdorfer. 1984. Site diagnosis, tree species selection, and slashburning guidelines for the Vancouver Forest Region. Land Manage. Rep. No. 25. B.C. Min. of For., Victoria, B.C. 180 pp. Klinka, K, V.J. Krajina, A. Ceska, and A.M. Scagel. 1989a. Indicator plants of coastal British Columbia. Univ. British Columbia press, Vancouver, B.C. 228 pp. Klinka, K., R.E. Carter, M.C. Feller, and Q. Wang. 1989b. Relations between site index, salal, plant communities, and sites in coastal Douglas-fir ecosystems. 63: 19-28. Klinka, K. and R.E. Carter. 1990. Relationships between site index and synoptic environmental factors in immature Douglas-fir stands. For. Sci. 36: 815-830. Kojima, S. and V.J. Krajina. 1975. Vegetation and Environment of the Coastal Western Hemlock Zone in Strathcona Provincial Park, British Columbia. Syesis, 8 (suppl. 1): 1-23. Kozlowski, T.T. 1986. Soil aeration and growth of forest trees (Review article). Scand. J. For. Res. 1: 113-123. Krajina, V.J. 1965. Biogeoclimatic zones in British Columbia. Ecol. West. N. Amer. 1: 1-17. Krajina, V.J. 1969. Ecology of forest trees in British Columbia. Ecol. West. N. Amer. 2: 1-146. Krajina, V.J. 1972. Ecosystem perspectives in forestry. H.R. Macmillan Lecture Series. Faculty of Forestry, University of B.C., Vancouver, B.C. 31 pp. Krajina, V.J. K. Klinka, and J. Worrall. 1982. Distribution and Ecological characteristics of trees and shrubs of British Columbia. The University Of B.C, Vancouver, B.C. 131 pp. La Roi, G.H. and W.L. Strong. 1988. Understory plant community classifications as predictors of forest site quality for lodgepole pine and white spruce in west-central Alberta. Can. J. For. Res. 18: 875-887. Landsberg, J.J. 1986. Physiological ecology of forest production. Academic Press Inc., London. 296 pp. Larocque, G. and P.L. Mashall. 1988. Growth and yield of spruce in the inland mountain west: a literature review. Pp. 192-196 in Proceedings - Future Forest of the Mountain West: A Stand culture symposium USDA For. Serv. Gen. Tech. Rep. INT-243. Intermt. For. and Range Exp. Stn., Ogden, Utah.  229  Leech, J.W. 1984. Modelling for forest management. Pp. 229-233 in J.J. Landsberg and W. Parsons (eds.), Research for Forest Management. CSIRO, Melb. Letey, J. 1985. Relationship between soil physical properties and crop production. Adv. in Soil Sci. 1: 277-293. Lewis, T., J. Pojar, D. Holmes, R. Trowbridge, and K.D. Coates. 1986. A field guide for identification and interpretation of the Sub-Boreal Spruce zone in the Prince Rupert Forest region. Land manage. Handbook No. 10. B.C. Min. For., Victoria, B.C. Linteau, A. 1955. Forest site classification of the Northeastern Coniferous Section, Boreal Forest Region, Quebec. Can. Dep. Northern Aff. Nat. Resour., For. Br., bull. 118. Ottawa, Ontario. 85 pp. Loucks, O.L. 1962. Ordinating forest communities by means of environmental scalars and phyto-sociological indices. Ecol. Monogr. 32: 137-166. Logan, K.T. 1969. Growth of tree seedlings as affected by light intensity. IV. Black spruce, white spruce, balsam fir, eastern white cedar. Canadian Forestry Service, Publication 1256. Ottawa, Ontario. 12 pp. Losch, C.K. and R.C. Schlesinger. 1975. Predicting site index in young black walnut plantations. USDA For. Serv. Res. Note NC-187. North Central For. Exp. Stn., St. Paul, Minn. 4 pp. Love, D.V. and J.R.M. Williams. 1968. The economics of plantation forestry in southern Ontario. Rep. 5. Dept. Regional Econ. Expansion, Can. Land Inventory. Ottawa, Ontario. 19 pp. MacArthur, J.D. 1957. The effects of manure on a white and Norway spruce plantation at Grand'Mere, Quebec. Can. Dept. North. Affaris Nat. resources, Forest. Br., Forest Res. Div. Tech. Note 64. Ottawa, Ontario. 15 pp. MacLeod, J.W. 1956. Plantations of the Acadia Forest Experiment Station. Can. Dept. North. Affairs Nat. Resources, Forest. Br., Forest Res. Div., Tech. Note 31. Ottawa, Ontario. 25 pp. Major, J. 1951. A functional factorial approach to plant ecology. Ecology. 32: 392-412. Major, J. 1963. A climatic index to vascular plant activity. Ecology. 44: 485498. McGrath, C.L. and H. Loewenstein. 1975. Soil-site quality relationships on the University of Idaho experimant forest. University of Idaho Forestry, Wildlife, and Range Sciences, Exp. Sta. Note 22. Moscow, ID. 4 pp. Mckeague, J.A. (ed.). 1978. Manual of soil sampling and methods of analysis (2nd ed.). Canadian Society of Soil Science, Ottawa, Ontario.  230  McLean, C.D. and C.L. Bolsinger. 1973. Estimating Dunnings site index from plant indicators. USDA For. Serv. Res. Note PNW-152. Portland, OR. McQuilkin, R.A. 1976. The necessity of independent testing of soil-site equations. Soil Sci. Soc. Amer. J. 40: 783-785. Meidinger, D. and J. Pojar. 1983. Sub-boreal spruce zone. Pp. 306-311 in Watts,S.B. (ed.), Forestry handbook for British Columbia. The University of British Columbia, Vancouver, B.C. Meidinger, D. and J. Pojar (eds). 1991. Ecosystems of British Columbia. Special Report Series 6. B.C. Min. For. Victoria, B.C. 330 pp. Milner, K.S. 1987. The development of site specific height growth curves for four conifers in western Montana. Ph. D. dissertation. Univ. of Montana, Missoula, Montana. 169 pp. Minore, D. 1972. A classification of forest environments in the south Umpqua Basin. USDA For. Serv. Res. Pap. PNW-129. Portland, OR. Mitscherlich, E.A. 1909. Das Gesetz des Minimums and das Gesetz abnchmenden Bodenertrags. Landwirtsch. Jahrb. 38: 537-552. Monserud, R.A. 1984a. Height growth and site index curves for inland Douglas-fir based on stem analysis data and forest habitat type. For. Sci. 30: 943-965. Monserud, R.A. 1984b. Problems with site indexes: An opinionated review. Pp. 167-180 in J. Bockheim (ed.), Forest land Proc. Dep. Soil Sci., Univ. Wisconsin, Madison, WI. Monserud, R.A. 1987. Variation on a theme of site index. Pp. 184-191 in Forest Growth Modelling and Prediction, Volume 1. USDA For. Serv. Gen. Tech. Rep. NC-120. North Central For. Exp. Stn., St. Paul, MN. Monserud, R.A., U. Moody, and D.W. Breuer. 1990. A soil site study for inland Douglas-fir. Can. J. For. Res. 20: 686-695. Morrison, D.F. 1990. Multivariate statistical methods (3rd ed.). McGraw-Hill Publishing Company, New York. 495 pp. Morzuch, B.J. and G.A. Ruark. 1991. Principal components regression to mitigate the effects of multicollinearity. For. Sci. 37: 191-199. Mueller-Dombois, D. and H. Ellenberg. 1974. Aims and methods of vegetation ecology. John Wiley and Sons. Toronto. 547 pp. Nautiyal, J.C. and L. Cuoto. 1984. The natural and use of the timber production function: Eucalyptus grandis in Brazil. For. Sci. 30: 761773.  231  Newberry, J.D. and L.V. Piennar. 1978. Dominant height growth models and site index curves for site-prepared slash pine plantations in the lower coastal plain of Georgia and North Florida. Univ. of Georgia. Plantation Manage. Res. Coop. Res. Paper No. 4. Athens, GA. Nienstaedt, H. 1957. Silvical characteristics of white spruce (Picea glauca). U.S. Dept. Agri., Forest Service, Lake States Forest Exp. Sta., Pap. 55. St. Paul, MN. 23 pp. Nienstaedt, H. and J.C. Zasada. 1990. Picea gluaca (Moench) Voss or White spruce. Pp. 204-225 in Burns, R.M. and B.H. Honkala (tech. coords). Silvics of North America: 1. Conifers. Agriculture Handbook 654. USDA For. Serv., Washington D.C. Odum, E.P. 1971. Fundamentals of ecology (3rd ed.). W.B. Saunders Company, Philadelphia, PA. 574 pp. Oserkowsky, J. 1933. Quantitative relation between chlorophyll and iron in green and chlorotic pear leaves. Plant Physiol. 8:449-468. Ott, L. 1988. An introduction to statistical methods and data analysis (3rd ed.). PWS-Kent Publishing Company, Boston. 835 pp. Paine, L.A. 1960. Studies in forest pathology XXII. Nutrient deficiencies and climatic factors causing low volume production and active deterioration in white spruce. Can. Dept. Agri., Sci. Serv., Forest Biol. Div., Pub. 1067. Ottawa, Ontario. 29 pp. Parkinson, J.A. and S.E. Allen. 1975. A wet oxidation procedures for the determination of nitrogen and mineral nutrients in bological material. Comm. Soil Sci. Plant Anal. 6:1-11. Payandeh, B. 1974. Nonlinear site index equations for several major Canadian timber species. For. Chron. 50: 194-196. Payandeh, B. 1986. Predictability of site index from soil factors and lesser vegetation in northern Ontario forest types. Great Lakes Forestry Centre, Canadian Forestry Service, Information Report O-X-373. Ottawa, Ontario. Pearson, A.F. 1992. Relationships between site index of sitka spruce and measures of ecological site quality in the eastern Queen Charlotte islands. M.Sc. Thesis. The University of British Columbia. Vancouer, B.C. 92 pp. Peterson, C.E. and S.P. Gessel. 1983. Forest fertilization in the U.S. Pacific Northwest: Results of the Regional Forest Nutrition Research Project. Pp. 365-368 in R. Ballard and S.P. Gessel (eds.), IUFRO symposium on forest site and continuous productivity. USDA For. Serv. Gen. Tech. Rep. PNW-63. Pacific Northwest For. and Range Exp. Stn., Portland, OR.  232  Pfister, R.D., B.L. Kovalchik, S.F. Arno and R.C. Presby. 1979. Forest habitat types of Montana. USDA Forest Service, Intermountain Forest and Range Experiment Station, General Technical Report INT-34. Ogden, UT. Pfister, R.D. and S.F. Arno. 1980. Classifying forest habitat types based on potential climax vegetation. For. Sci. 26: 52-70. Pielou, E.C. 1969. Association tests versus homogeneity tests: their use in subdividing quadrats into groups. Vegetatio, 18: 4-18. Piennar, L.V. and K.J. Turnbull. 1973. The Chapman-Richards generalization of Von Bertalanffy's growth model for basal area growth and yield in even-aged stands. For. Sci. 19: 2-12. Pluth, D.L. and I.G.W. Corns. 1983. Productivity of conifers in western Canada boreal forests in relation to selected environmental factors. Pp. 101-111 in Ballard, R. and P.G. Gesseel (Tech. eds), IUFRO Symposium on Forest Site and Continuous Productivity. USDA For. Serv. Gen. Tech. Rep. PNW-163. Pacific Northwest For. and Range Exp. Sta., Portland, OR. Pogrebnyak, P.S. 1930. "Ober die Methodik der Standortsuchungen in Verbidung mit den Waldtypen. Pp. 455-471 in Proc. Int. Congr. For. Exp. Station, Stockholm, 1929. [Translation in Chinese]. Pojar, J., R. Trowbridge, and D. Coates. 1984. Ecosystem classification and interpretation of the Sub-boreal Spruce Zone, Prince Rupert Forest Region, British Columbia. Land Management Report No. 17. B.C. Min. For., Victoria, B.C. 73 pp. Pojar, J., K Klinka, and D.V. Meidinger. 1987. Biogeoclimatic ecosystem classification in British Columbia. For. Ecol. Manage. 22: 119-154. Pollard, D.F.W. and K.T. Logan. 1977. The effect of light intensity, photoperiod, soil moisture potential, and temperature on bud morphogenesis in Picea species. Can. J. For. Res. 7: 415-421. Ponnamperuma, F.N. 1972. The chemistry of submerged soils. Advances in Agronomy 24: 29-96. Powers, R.F. 1980. Mineralizable soil nitrogen as an index of nitrogen availability of forest trees. Soil Sci. Soc. Am. J. 44: 1314-1320. Price, W.J. 1978. Analytic atomic absorption spectrophotometry. Heydon and Son Ltd., London. Pritchett, W.L. and R.F. Fisher. 1987. Properties and management of forest soils (2nd ed.). John Wiley & Sons, New York. 494 pp. Radwan, M.A. and D.S. DeBell. 1980. Site index, growth, and foliar chemical composition relationships in western hemlock. For. Sci. 26: 283-290.  233  Radwan, M.A. and C.A. Harrington. 1986. Foliar chemical concentrations, growth, and site productivity relations in western red cedar. Can. J. For. Res. 16: 1069-1075. Radwan, M.A., M.D. Murray, and J.M. Kraft. 1989. Growth and foliar nutrient concentrations of Pacific silver fir. Can. J. For. Res. 19: 14291435. Rayner, M.E. and B.J. Turner. 1990. Growth and yield modelling of Australia eucalyptus forests I. Historical development. Aust. For. 53: 224-237. Rayner, M.E. and B.J. Turner. 1990. Growth and yield modelling of Australia eucalyptus forests II. Future trends. Aust. For. 53: 238-247. Rayner, M.E. 1991. Site index and dominant height growth curves for regrowth karri (Eucalyptus diversicolor F. Muell). For. Ecol. Manage. 44: 261-283. Ralston, C.W. 1964. Evaluation of forest site productivity. Int. Rev. For. Res. 1: 171-201. Rauscher, H.M. 1984. Growth and yield of white spruce plantation in the Lake States (a literature review). USDA For. Ser. Res. Pap. NC-253. Northern Central For. Exp. Stn., St. Paul, Minn. 46 pp. Rennie, P.J. 1962. Methods of assessing forest site capacity. Pp. 3-18 in Trans. 7th Internatl. Soc. Soil Sci., Comm. IV and V. Richards, F.J. 1959. A flexible growth function for empirical use. J. Exp. Bot. 10: 290-300. Rivard, P.G., P.M. Woodard, and R.L. Rothwell. 1990. The effect of water table depth on white spruce (Picea glauca) seedling growth in association with the marsh reed grass (Calamagrositis canadensis) on wet mineral soil. Can. J. For. Res. 20: 1553-1558. Rowe, J.S. 1956. Uses of understory plant species in forestry. Ecology. 37: 461-474. Rowe, J.S. 1972. Forest regions in Canada. Based on W.E. D. Halliday's "A Forest Classification for Canada." Canadian Forest Service, Publication 1300. Ottawa, Ontario. 172 pp. Running, S.W. and J.C. Coughlan. 1988. A general model of forest ecosystem processes for regional application. I. Hydrological balance, canopy gas exchange and primary production processes. Ecol. Modelling, 42: 125154. Russell, K.W. 1963. Plantation white spruce growth related to some forest soil-site characteristics. Plan B paper. College of Forestry, University of Minnesota, St. Paul, MN. 42 pp.  234  Schmidt, M.G. and W.H. Carmean. 1988. Jack pine site quality in relation to soil and topography in north central Ontario. Can. J. For. Res. 18: 297-305. Seber, G.A.F. 1984. Multivariate observations. John Wiley & Sons, New York. 686 pp. Shugart, H.H. 1984. Theory of forest dynamics. Springer. New York. 278 pp. Sims, R.A., W.D. Towill, K.A. Baldwin, and G.M. Wickware. 1989. Field guide to the forest ecosystem classification for northwestern Ontario. Forestry Canada and Ontario Min. of Nat. Resour., Thunder bay, Ontario. Smith, J.L., B.L. McNeal, E.J. Owens, and G.O. Klock. 1981. Comparison of nitrogen mineralized under anaerobic and areobic conditions for some agriculture and forest soil of Washington. Soil Sci. Plant Anal. 12: 9971009. Smith, V.G. 1984. Asymptotic site-index curves, fact or artifact? For. Chron. 60: 150-156. Smith, V.G. and M. Watts. 1987. The assessment of the structural method of deriving a black spruce site equation. Can. J. For. Res. 17: 1181-1189. Software Publishing Corporation. 1991. User's Manual (Harvard Graphics 3.0). Software Publishing Corporation, Mountain View, CA. Soil Survey Staff. 1951. Soil survey manual. U.S.D.A. Handbook No. 18, U.S. Dept. of Agriculture. Washington, D.C. 503 pp. Soil Survey Staff. 1975. Soil Taxonomy. Handbook 436, USDA Soil Conserv. Serv., Washington, D.C. 754 pp. Sokal, R.R. and F.J. Rohlf. 1981. Biometry: the principles and practice of statistics in biological research (2nd ed.). W. H. Freeman and Company. New York. 859 pp. Spies, T.A. and Barnes, B.V. 1985. Ecological species groups of upland northern hardwood and conifer ecosystems of Sylvania Recreation Area, Upper Peninsula, Michigan. Can. J. For. Res. 15: 961-972. Spittlehouse, D.L. 1981. Measuring and Modelling evapotranspiration from Douglas-fir stands. Ph.D dissertation, The University of British Columbia, Vancouver, B.C. Spittlehouse, D.L., and T.A. Black. 1981. A growing season water balance model applied to two Douglas-fir stands. Water Resources Research. 17: 1651-1656. Spittlehouse, D.L. 1985. Determination of year-to-year variation in growing season water use of Douglas-fir stand. Pp. 235-254 in B.A. Hutchison and B.B. Hicks (eds.), The forest-atmosphere interaction. D. Reidel Publishing Company, London.  235  Spurr, S.H. and B.V. Barnes. 1980. Forest ecology (3rd ed). John Wiely and Sons, New York. 678 pp. Stage, A.R. 1963. A mathematical approach to polymorphic site index curves for grand fir. For. Sci. 9: 167-180. Stiell, W.M. 1958. Pulpwood plantations in Ontario and Quebec. Index No. 1770 (F-2). Canadian Pulp and Paper Association, Woodlands Section. 42 pp. Stiell, W.H. and A.B. Berry. 1973. Development of unthinned white spruce plantations to age 50 at Petawawa Forest Experiment Station. Can. For. Serv., Pub. 1317. Ottawa, Ontario. 18 pp. Stiell, W.M. 1976. White spruce: artificial regeneration in Canada. Canadian Forestry Service, Information Report FMR-X-85. Forest Management institute, Ottawa, Ontario. 275 pp. Stone, E.L. Jr., R. Feuer and H.M. Wilson. 1962. Approximate adaption of plantation species to some site factors. Supplement to New York State Coll. Agri., Cornell Ext. Bull. 1075. New York, NY. 1 p. Stone, E.L. 1978. A critique of soil moisture-site productivity relationships. Pp. 377-387. in W. E. Balmer (ed.). Soil moisture-site productivity symposium proceedings. USDA For. Serv., Southeast and Private Forestry, Atlanta, GA. Stone, E.L. 1984. Site quality and site treatment. Pp. 41-52 in E.L. Stone (ed.). Forest soil and treatment impacts: Proceedings of the sixth North American Forest Soil Conference. Dept. of Forestry, wildlife and Fisheries, University of Tennessee, Knoxville, TN. Stotler, R. and B. Crandall-Stotler. 1977. A checklist of the liverworts and hornworts of North America. The Bryologist 80: 405-428. Strong, W.L., D.J. Pluth, G.H. La Roi, and I.G.W. Corns. 1991. Forest understory plants as predictors of lodgepole pine and white spruce site quality in west-central Alberta. Can. J. For. Res. 21: 1675-1683. Sutton, R.F. 1969. Silvics of white spruce. Can. Dep. Fish. For., For. Br. Publ. 1250. Ottawa, Ontario. 57 pp. Swan, H.S.D. 1960. The mineral nutrition of Canadian pulpwood species. I. The influence of nitrogen, phosphorus, potassium and magnesium deficiencies on the growth and development of white spruce, black spruce, jack pine and western hemlock seedlings grown in a controlled environment. Pulp Pap. Res. Inst. Can. Tech. Rep. 168. Montreal, Quebec. 66 pp. Swan, H.S.D. 1971. Relationships between nutrient supply, growth, and nutrient concentrations in the foliage of white and red spruce. Woodlands Pap. W.P. 29. Pulp and Paper Research Institute, Montreal, Quebec. 27 pp.  236  Tabachnick, B.G. and L.S. Fidel!. 1989. Using multivariate statistics (2nd ed.). Harper & Row Publishers Inc., New York. 746 pp. Tamm, C.O. 1964. Determination of nutrient requirement of forest stands. Int. Rev. For. Res. 1: 115-170. Tatsuoka, M.M. and P.L. Lohnes. 1988. Multivariate analysis. Macmillan Publishing Company, New York. 479 pp. Taylor, R.L. and B. MacBryde. 1977. Vascular plants of British Columbia. Tech. Bull. No. 4, Univ. British Columbia Press, Vancouver, B.C. Valentine, K.W.G., P.N. Sprout, T.E. Baker, and L.M. Lavkulich (eds.). 1978. The soil landscapes of British Columbia. The resource Analysis Branch, Min. of Env., Victoria, B.C. 197 pp. Van Groenewoud, H. 1965. An analysis and classification of white spruce. communities in relation to certain habitat features. Can. J. Bot. 43: 1025-1036. Verbyla, D. 1986. Potential prediction bias in regression and discriminant analysis. Can. J. For. Res. 16: 1255-1257. Wali, M.K. and V.J. Krajina. 1973. Vegetation-environment relationships of sub-boreal spruce zone ecosystem in British Columbia. Vegetatio. 26: 237-381. Wall, H.G. and H. Loewenstein. 1969. The relationship of certain soil and topographic properties to site quality of grand fir in northen Idaho. University of Idaho Forestry, Wildlife, and Range Sciences, Exp. Sta. Note. Moscow, ID. 4 pp. Walmsley, M., G Utzig, T. Vold, D. Moon, and J. van Barneveld (eds.), 1980. Describing ecosystem in the field. B.C. Min. Env., RAB Technical paper 2. B.C. Min. For., Land Management Rep. No. 7. Victoria, B.C. 225 pp. Wang, G. 1986a. A study of forest site classification in the southern Purple Mountain in Nanjing, China. M.Sc. Thesis. Nanjing Forestry University, Nanjing, China. 118 pp. [in Chinese with English abstract]. Wang, G. 1986b. A review of forest site classification. J. Nanjing For. Univ. 9: 108-115. [in Chinese with English abstract] Wang, G. 1989. A quantitative method of forest site classification. J. Nanjing For. Univ. 12: 24-31. [in Chinese with English abstract]. Wang, Q. 1992. Ecological and height growth analysis of some sub-Boreal immature lodgepole pine stands in central British Columbia. Ph.D. dissertation. The University of British Columbia, Vancouver, B.C. 207 pp.  237  Wang, Q, G.G. Wang, K.D. Coates, and K. Klinka. 1992. Use of site classification in the prediction of lodgepole pine and white spruce site index in the SBS zone. Submitted for publication as Land Management Report, B.C. Min. For., Victoria, B.C. Waring, S.A. and J. Major. 1964. Some vegetation of the California coastal redwood region in relation to gradients of moisture, nutrients, light, and temperature. Ecol. Monogr. 34: 167-215. Watt, R.F. and M.L. Heinselman. 1965. Foliar nitrogen and phosphorus level related to site quality in a northern Minnesota spruce bog. Ecology. 46: 357-361. West, P.W. 1990. Thinning response and growth modelling. in Kerruish, C.M and W.H.M. Rawlins (eds.). The Young Eucalyptus Report. CSIRO, Melb. Whittaker, R.H. (ed.). 1978. Classification of plant communities. Dr. W. Junk Publisher, The Hague. 408 pp. Wickramasinghe, A. 1988. Modeling tree growth potential based on effective evapotraspiration. For. Sci. 34: 864-881. Wilde, S.A., F.G. Wilson and D.P, White. 1949. Soil of Wisconsin in relation to silviculture. Wisconsin Conserv. Dep. Pub. 525-49. Madison, WI. 171 pp. Wilde, S.A., J.G. Iyer, C. Tanzer, W.L. Trautman, and K.G. Watterston. 1965. Growth of Wisconsin coniferous plantations in relation to soils. Res. Bull. 262. University of Wisconsin, Madison, WI. 81 pp. Wilde, S.A. 1966. Soil standards for planting Wisconsin conifers. J. For. 66: 389-391. Wilkins, M.B. 1984. Advanced plant physiology. Pitman publishing Ltd., London. 514 pp. Wilkinson, L. 1990. SYSTAT: The system for statistics. SYSTAT Inc., Evanston, IL. 677 pp. Wilkinson, L. 1990. SYGRAPH: The system for graphics. SYSTAT Inc., Evanston, IL. 547 pp. Wykoff, W.R. and R.A. Monserud. 1987. Presenting site quality in increment models: A comparison of Methods. Pp. 184-191 in Forest Growth Modelling and Prediction (Volume 1). USDA For. Ser. Gen. Tech. Rep. NC-120. North Central For. Exp. Stn., St. Paul, MN. Yarie, J., K. Van Cleve, and R. Schlentner. 1990. Interaction between moisture, nutrients and growth of white spruce in Interior Alaska. For. Ecol. Manage. 30: 73-89. Zahner, R. 1962. Loblolly pine site curves by soil groups. For. Sci. 8: 104-110.  238  Zasada, J.C., K.Van Cleve, R.A. Werner, J.A. McQueen, and E. Nyland. 1977. Forest biology and management in high latitude North American forests. Pp. 137-195 in Proceedings, Symposium on North American Lands at Latitudes North of 60 Degrees, September 19-22, 1977. University of Alaska, Fairbanks, AK. Zeide, B. 1978. Standardization of growth curves. J. For. 76: 289-292. Zinke, P.J. 1960. Forest site quality as related to soil nitrogen content. Pp 411-418 in Proceedings of the 7th International Congress on Soil Science, Madison, WI (Vol. 3). Elsevier Publ., Amsterdam.  239  Appendix 1.^List of scientific name of plant species identified in the study stands. Abies lasiocarpa Acer glabrum Achillea millefolium Actaea rubra Adenocaulon bicolor Alnus sinuata Amelanchier alnifolia Anaphalis margaritacea Anemone multifida Angelica arguta Angelica genuflexa Antennaria neglecta Apocynum androsaemifolium Aquilegia formosa Aralia nudicaulis Arctostaphylos uva-ursi Arnica cordifolia Aruncus dioicus Asarum caudatum Aster ciliolatus Aster conspicuus Aster modestus Aster subspicatus Athyrium filix-femina Aulacomnium androgynum Aulacomnium palustre Barbilophozia barbata Betula glandulosa Betula papyrifera Botrychium virginianum Brachythecium albicans Brachythecium asperrimum Bromus vulgaris Calamagrostis canadensis Calamagrostis rubescens Calypso bulbosa Carex aquatilis Carex bebbii Carex concinnoides Carex disperma Carex nigricans Carex rossii Castilleja miniata Cephalozia bicuspidata Ceratodon purpureus Chimaphila umbellata Cicuta douglasii Cinna latifolia Circaea alpina Cladonia chlorophaea  (Hook.) Nutt. Torr. L. (Ait.) Willd. Hook. (Regel) Rydb. (Nutt.) Nutt. (L.) B. & H. Poir. Nutt. in Torr. & Gray Nutt. in Torr. & Gray Greene L. Fisch. in DC. L. (L.) Spreng. Hook. (Walt.) Fern. Lindl. Lindl. in Hook. Lindl. in Hook. Lindl. in Hook. Nees (L.) Roth (Hedw.) Schwaegr. (Hedw.) Schwaegr. (Schmid) Loeske Michx. Marsh. (L.) Sw. (Hedw.) B.S.G (C. Muell.) Sull. (Hook.) Shear (Michx.) Beauv. Buckl. (L.) Oakes in Thomps. Wahlenb. (Bailey) Olney ex Fern. Mack Dew. C. A. Mey. Boott in Hook. Dougl. ex Hook. (L.) Dum. (Hedw.) Brid. (L.) Barton (DC.) Coult. & Rose (Trey. ex Goepp.) Griseb L. (Florke ex Somm.) Spreng  240 Cladonia gracilis Cladina mitis Cladonia multiformis Cladonia uncialis Clematis occidentalis Climacium dendroides Clintonia uniflora Corallorhiza maculata Corallorhiza striata Corallorhiza trifida Cornus canadensis Cornus sericea Cystopteris fragilis Dicranum fuscescens Dicranum polysetum Dicranum scoparium Disporum hookeri Disporum trachycarpum Drepanocladus uncinatus Dryopteris expansa Elymus glaucus Empetrum nigrum Epilobium angustifolium Epilobium latifolium Epilobium palustre Equisetum arvense Equisetum fluviatile Equisetum palustre Equisetum scirpoides Equisetum sylvaticum Eriophorum angustifolium Festuca occidentalis Fragaria vesca Fragaria virginiana Galium boreale Galium triflorum Gaultheria hispidula Geocaulon lividum Geum aleppicum Geum macrophyllum Goodyera oblongifolia Gymnocarpium dryopteris Heracleum lanatum Hieracium albiflorum Hylocomium splendens Hypnum revolutum Isopterygium elegans Juniperus sibirica Kindbergia praelonga Lathyrus nevadensis Lathyrus ochroleucus Ledum groenlandicum Lepidozia reptans Lilium columbianum  (L.) Wild. (Sandst.) Hale & W. Culb Merr. (L.) Wigg. (Hornem) DC. (Hedw.) Web. & Mohr. (Schult.) Kunth Raf. Lindl. Chat. L. L. (L.) Bernh. in Scrad. Turn. Sw. Hedw. (Ton..) Nicholson (Wats.) Benth. & Hook. F (Hedw.) Warnst. (Presl) Fraser-Jenkins Buckl. L. L. L. L. L. L. L. Michx. L. Honck. Hook. L. Duchesne L. Michx. (L.) Muhlenb. ex Bigel. (Richards.) Fern. Jacq. Wind. Raf. (L.) Newm. Michx. Hook. (Hedw.) B.S.G. (Mitt.) Lindb. (Brid.) Lindb. L. (Hedw.) Ochyra Wats. Hook. Oeder (L.) Dum. Hanson ex Baker  241 Linnaea borealis Listera borealis Listera cordata Lonicera involucrata Lycopodium annotinum Lycopodium complanatum Lycopodium obscurum Mahonia aquifolium Maianthemum dilatatum Marchantia polymorpha Melampyrum lineare Melica subulata Menyanthes trifoliata Mimulus guttatus Mitella nuda Mnium medium Mnium spinulosum Moehringia lateriflora Moneses unit fora Mycelis muralis Oplopanax horridus Orthilia secunda Oryzopsis asperifolia Osmorhiza chilensis Parnassia fimbriata Paxistima myrsinites Peltigera aphthosa Petasites frigidus Petasites palmatus Petasitis sagittatus Picea glauca Picea mariana Pinus contorta Plagiochila aspleniformis Plagiomnium insigne Platanthera dilatata Platanthera hyperborea Platanthera obtusata Platanthera orbiculata Platanthera unalascensis Pleurozium schreberi Pogonatum alpinum Polemonium pulcherrium Polygonum bistortoides Polytrichum commune Polytrichum juniperinum Populus tremuloides Potentilla palustris Prunus virginiana Prunella vulgaris Pseudotsuga menziesii Pteridium aquilinum Ptilium crista-castrensis Pyrola asarifolia  L. Morong (L.) R. Br. in Ait. (Richards.) Banks ex Spr L. L. L. (Pursh) Nutt. (How.) Nels. & Macbr. L. Desr. (Griseb.) Scribn. L. DC. L. B.S.G. B.S.G. (L.) Fenzl (L.) Gray (L.) Dumort. (Sm.) Miq. (L.) House Michx. Hook. & Am. Koenig (Pursh) Raf. (L.) Willd. (L.) E. Fries (Mt.) Gray (Banks) Gray (Moench) Voss (Mill.) B.S.P. Dougl. ex Loud. Schust. (Mitt.) Kop. (Pursh) Lindl. ex Beck (L.) Lindl. (Banks ex Pursh) Lindl. (Pursh) Lindl. (Spreng.) Wats. (Brid.) Mitt. (Hedw.) Rohl. Hook. Pursh Hedw. Hedw. Michx. (L.) Scop. L. L. (Mirb.) Franco (L.) Kuhn in Decken (Hedw.) De Not. Michx.  242  Pyrola chlorantha Ranunculus eschscholtzii Ranunculus occidentalis Ranunculus repens Ranunculus uncinatus Rhacomitrium heterostichum Rhizomnium glabrescens Rhizomnium nudum Rhizomnium perssonii Rhytidiadelphus loreus Rhytidiopsis robusta Rhytidiadelphus triquetrus Ribes lacustre Ribes laxiflorum Ribes triste Rosa acicularis Rubus parviflorus Rubus pedatus Rubus pubescens Salix bebbiana Salix drummondiana Salix lasiandra Salix monticola Salix myrtillifolia Salix pyrifolia Salix rigida Salix scouleriana Salix sitchensis Sambucus racemosa Senecio pauperculus Senecio streptanthifolius Shepherdia canadensis Smilacina racemosa Smilacina stellata Solidago canadensis Sorbus sitchensis Sphagnum capillacem Spiraea betulifolia Spiraea douglasii Stellaria crispa Stellaria longifolia Steocaulon tomentosum Streptopus amplexifolius Streptopus roseus Symphoricarpos albus Taraxacum officinale Thalictrum occidentale Thuidium abitinum Thuja plicata Tiarella laciniata Tiarella trifoliata Tiarella unifoliata Tomenthypnum nitens Trientalis latifolia  Sw. Schlecht. Nutt. in Torr. & Gray L. D. Don in G. Don (Hedw.) Brid. (Kindb.) Kop. (Britt. & Williams) Kop • Koponen (Hedw.) Warnst. (Hook.) Broth. (Hedw.) Warnst. (Pers.) Poir. Pursh Pall. Lindl. Nutt. Sm. Raf. Sarg. Barratt Benth. Bebb. ex Coult. Anderss. Anderss. Muhl. Barratt in Hook. Sanson in Bong. L. Michx. Greene (L.) Nutt. (1.) Desf. (1.) Desf. L. M. J. Roem. (Ehrh.) Hedw. Pall. Hook. Cham. & Schlecht. Muhl. Fr. (L.) DC. Michx. (L.) Blake Weber in Wiggers Gray (Hedw.) B.S.G. Donn ex D. Don in Lamb. Hook. L. Hook. (Hedw.) Loeske Hook.  243  Trifolium repens^L. Trisetum cernuum^Trin. Vaccinium caespitosum^Michx. Vaccinium membranaceum^Dougl. ex Hook. Vaccinium myrtilloides^Michx. Vaccinium ovatum^Pursh Vaccinium oxycoccus^L. Valeriana sitchensis^Bong. Veratrum viride^Ait. Veronica americana^(Raf.) Schwein. ex Benth Viburnum edule^(Michx.) Raf. Viola adunca^Sm. in Rees Viola canadensis^L. Viola glabella^Nutt. in Torr. & Gray Viola orbiculata^Geyer ex Hook. Viola palustris^L. Viola renifolia^Gray Viola sempervirens^Greene  ^  Appendix 2. Correlations between site associations recognized in this study and the B.C. Min. For. (D.V. Meidinger, pers. comm.). Symbols for soil moisture and soil nutrient regimes as in Table 4.2 and Table 4.8. Symbols for tree species are: Bl - Subalpine fir, Fd - Douglas-fir, Pl - lodepole pine, Sb - black spruce. Sxw - hybrid spruce.  SMRs  ^  SBSdw3  ^  SBSdwI  ^  SBSdk  ^  SBSmw  ^  SBSmk1  ^  SBSwkI  Very poor to medium SNRs ^10^Cladonia ^P1 FdPI^- Cladonia VD  ^-^Juniper -^Ricegrass  FdBI^-^Huckleberry  ^20^Sheperdia PI^- Feathermoss ^PI MD - Cladina  Fd^-^Saskatoon -^Pinegrass  PI^-^Feathermoss -^Cladina  PI  -^Huckleberry Velvet-leaved blueberry  30^Pleurozium ^sxy, SawFd -^Ricegrass Sew^ Spirea SD-F -^Feathermoss  PI^-^Cladina -^Step^moss  PI^-^Feathermoss -^Cladina SzwFd^-^Knight's plume  PISb  -^Feathermoss  PI^-^Pinegrass -^Feathermoss  ^40^Hylocomium ^Sb M-VM  ^-^Labrador^tea -^Sphagnum  ^50^Sphagnum ^Sb W-VW  ^-^Labrador^tea -^Sphagnum  PI^-^Huckleberry -^Velvet-leaved blueberry SzwFd^-^Knight's plume  ^.^Huckleberry  -^  ^-  PI^-^Huckleberry -Cladina  Sb^-^Huckleberry -Spirea  Highbush-  ^  craneberry  Appendix 2. Continued. SMRs  SBSdk^I^SBSmw  SBSdwl  SBSdw3  SBSmk  SBSwk  Medium to very rich SNRs 32 Spirea  31 Aster  33 Aralia  34 Viburnum  Sxw . Spirea -  SD  Purple Peavine SxwFd - Pinegrass  Sxw - Twinberry -^Co It sfoot  36 Petasitis F  Sxw - Oak fern  SxwFd - Hazelnut  Sxw - Horsetail  Sxw - Twinberry  -Glow moss  -^Col tsfoot  SxwFd - Knight's plume  wheat grass Sxw - twinberry  SxwFd - Falsebox  cr aneberry  Sxw - Oak fern  SxwFd - Toadflax  SxwFd - Knight's plume  Sxw - Horsetail  42 Oplopanax  M  Oak fern  Sxw - Horsetail  - Hig hbush-  -^Co I t sfoot  Sxw - Oak fern Sxw - Twinberry  Bogs  Sxw - Huckleberry  Bluegrass^-S lender  41 Aulacomium  Sxw - Horestail  Sxw - Devil's club  - Glow moss  Sxw - Oak fern  Sxw - Horsetail Sxw - Devil's club  51 Carex SbSxw - Scrub birch - Sedge  - Hig hbushcrane berry Sxw - Twinberry - Oak fern  Sxw - Devil's club  Sxw - Horsetail  VM  Sxw - Huckleberry  Sxw - Devil's club  43 Equisetum  Act - Floodplain  w - vw  35 Gymnocarpium  Sxw - Horsetail Sxw - Devil's club  - Lady fern  246  Appendix 3. The edaphic grid showing site series distinguished in the study in each subzone or variant.  Horsefly Dry and Warm SBS Variant (SBSdw1) Soil nutrient regimes VP^P^M^R^VR VD  MD  SD  0  1 Sheperdia  2  (SI^=^16.5,^n^=^1)  4D  3 4  Pleurozium (SI^=^18.9,^n^=^1)  F  5  M  6  (11)  Aster (SI^=^20.6,^n^=^7)  Petasitis (SI^=^20.3,^n^=^1)  ID  Aulacomnium (SI^=^19.1,^n^=^7)  VM  6  W  7  0 Sphagnum (SI^=^4.7,^n^=^1)  VW  Figure A3.1.  8  Site series distinguished in the study for SBSdw1. Symbols for soil moisture and soil nutrient regimes as in Table 4.2 and Table 4.8, respectively. White spruce site index and number of stands for each site series are given in parentheses.  247  Appendix 3. Continued 1.  Stuart Dry and Warm SBS variant (SBSdw3) Soil nutrient regimes VP P^M^R^VR VD  0  1 MD  SD  2 3 4  F  5  M  6  Sheperdia 0^ (SI =^  15.0,^n^= 5)  CD Pleurozium (SI =^17.5,^n^= 2)  I  G  A ulacomnium  (SI = 20.1,^n^= 5)  VM  6  W  7  CD Sphagnum  VW  8  (SI =^14.5,^n^=^1)  Figure A3.2.^Site series distinguished in the study for SBSdw3. Symbols for soil moisture and soil nutrient regimes as in Table 4.2 and Table 4.8, respectively. White spruce site index and number of stands for each site series are given in parentheses.  248  Appendix 3. Continued 2.  Dry and Cool SBS Subzone (SBSdk) Soil nutrient regimes VP^P^M^R^VR VD  0  1  MD 2 3 4  F^5  CD  Sheperdia (SI^=^15.8,^n = 5)  CD  Spiraea (SI^=^19.4,^n^= 4)  Pleurozium (SI^=^20.3,^n^=^2)  M 6 A ulacomnium (SI^=^20.9,^n^=^1)  VM 6  W 7  CD Sphagnum  VW 8  (SI^=^11.9,^n^= 1)  Figure A3.3.^Site series distinguished in the study for SBSdk. Symbols for soil moisture and soil nutrient regimes as in Table 4.2 and Table 4.8, respectively. White spruce site index and number of stands for each site series are given in parentheses.  249  Appendix 3. Continued 3.  Moist and Warm SBS Subzone (SBSmw) Soil nutrient regimes VP P^M^R^VR VD 0  MD  1 2  (111)  ^SD^3  Aralia (SI^=^20.1,^n^=^4)  ^F^4  ,-.  .....4  M 5  0^  VM 6  0^  (SI^=^22.0,^n^=^6)  Equisetum  (SI^=^19.0,^n^=^4)  CD VD W 7  Oplopanax  0  Sphagnum VW 8  (SI^=^12.9,^n^=^1)  Figure A3.4.^Site series distinguished in the study for SBSmw. Symbols for soil moisture and soil nutrient regimes as in Table 4.2 and Table 4.8, respectively. White spruce site index and number of stands for each site series are given in  parentheses.  250  Appendix 3. Continued 4.  Moist and Cool SBS Subzone (SBSmk) Soil nutrient regimes VP P^M^R^VR VD 0  MD  1 2 3  F^4  0^  Sheperdia (SI^=^16.6,^n^=^2)  1:1) Pleurozium (SI^=^18.1,^n^=^5)  0 Viburnum (SI^=^19.6,^n^=^3)  0  VM 6 C  Hylocomium (SI^=^19.7, n^=^2)  0^  Equisetum (SI^=^17.4,^n^=^2)  7 VW 8  Figure A3.5.^Site series distinguished in the study for SBSmk. Symbols for soil moisture and soil nutrient regimes as in Table 4.2 and Table 4.8, respectively. White spruce site index and number of stands for each site series are given in parentheses.  251  Appendix 3 Continued 5.  Wet and Cool SBS Subzone (SBSwk) Soil nutrient regimes VP^P^M^R^VR VD  MD  o 1 2  SD  3  0^Sheperdia  CD Pleurozium  F  4  M  5  VM  6  Gymnocarpium  Oplopanax  Equisetum  0  W  Sphagnum VW  8  Figure A3.6.^Site series distinguished in the study for SBSwk. Symbols for soil moisture and soil nutrient regimes as in Table 4.2 and Table 4.8, respectively. White spruce site index and number of stands for each site series are given in parentheses.  ▪ 252  Appendix 4.^Residual analysis for the five traditional anamorphic and polymorphic height growth models developed in the study.  (c) 6  4  2  a 0 oc  —2  —4  —6 ^ ^ ^ ^ 10^20 30 0 40 10^20^30 ^ Site index (m) Predicted height (m)  (d  (b) 40  )  •  •^•  • I  • • •  • 0  0^10^20^30 Predicted height (m)  40  6 ^ 0  20^40^60^BO^100^120 Age @ b.h (year)  Figure A4.1. Equation [6.2] (modified Richards' model). (a) - residual vs predicted height, (b) - measured height vs predicted height, (c) residual vs site index, and (d) - residual vs b.h. age.  253  Appendix 4. Continued 1.  (a)  6  4  2  —4  —6  O  10^20^30 40 ^ Predicted height (m)  30  10^20 Site index (m)  (b) 40  ..— 30 E  WI^111 1.i . i^.^  be w 4 20  T. ea,  2  I^•  ^  10 0  . :  O  ^ ^ 10^20^30 40 ^ Predicted height (m)  •^•  !  I^•  :^.  20^40^60^80  ^  Age @ b.h (year)  Figure A4.2.^Equation [6.3] (Goudie and Mitchell's model). (a) - residual vs predicted height, (b) - measured height vs predicted height, (c) - residual vs site index, and (d) - residual vs b.h. age.  100  ^  120  254 Appendix 4. Continued 2.  (c) 6 4 ,.....,^2  E 7P° 0  1 I:4  —2 —4  10^20^30  —6  40  0  30  10^20 Site index (m)  Predicted height (m)  (d)  (b) 40  I 30  $  .  S : .  8  0  0^10^20^30 Predicted height (m)  40  20^40^60^80^100^120 Age @ b.h (year)  Figure A4.3.^Equation [6.4] (Alemdag's model). (a) - residual vs predicted height, (b) - measured height vs predicted height, (c) residual vs site index, and (d) - residual vs b.h. age.  255  Appendix 4. Continued 3.  (c) 6 4 2  • . I  a • • •  r g ' .4 go •!: to .1:^ti  a : • ... :  : ;^•  _ s^  ^.  .  ;  •  ^%• ^• ... •1 il tI 9.  . .• • ‘..  ^.  ^•  —4 —6  •  0^10^20  ^  30  Site index (m)  Predicted height (m)  (d) 6  (b) 40  4 ...... 30 2 t iii  4 20  E0  E to co  14)  10  .  11  I .  I I  o  !  , !  —4  0  0^10^20^30 Predicted height (m)  —6 40^0^20^40^60^80  100  Age @ b.h (year)  Figure A4.5.^Equation [6.5] (EK-Payandeh's model). (a) - residual vs predicted height, (b) - measured height vs predicted height, (c) - residual vs site index, and (d) - residual vs b.h. age.  ^  120  256  Appendix 4. Continued 4.  (a)  (c) 6 4  •• • •^• . . : .  2  I  ^I:  : . % 2:* C • N  I• !  : •^1: : .^• .  •  1^1  Ft s .11.  •• III •  •^• SI . :• •• •• • .••  01  . •^.  —4 —6 ^ ^ ^ ^ 10^20^30 40 0 10^20 30 ^ Predicted height (m) Site index (m)  (b) 40  (d) 6  4 ...., 30 2 Po Ei 4 20 c,  •  0  ,  E  I  S  I  Ia)  —2  2 10  —4  0  0^10^20^30 Predicted height (m)  40  —6  0^20  40^60^80 Age @ b.h (year)  Figure A4.5.^Equation [6.6] (Logistic model). (a) - residual vs predicted height, (b) - measured height vs predicted height, (c) residual vs site index, and (d) - residual vs b.h. age.  100  120  257  Appendix 5. Height growth curves showing the difference between and similarity within each defined Z ratio groups. 40  30  4-1  be  20  r.  E2  10  0  0^20^40^60^BO  ^  100  ^  120  Age @ b.h. (year)  40  30  10  0  0^20^40^60^80^100  ^  120  Age @ b.h. (year)  Figure A5.1.  Height growth curves from original stem analysis data stratified according to Z ratio groups: (1) Z ratio group 7 and 8.  258  Appendix 5. Continued 1. 40  30  0  0^20^40^60^80  ^  100  ^  120  Age @ b.h. (year)  40  30  10  0  0^20^40^60^80  ^  100  ^  120  Age @ b.h. (year)  Figure A5.2.^Height growth curves from original stem analysis data stratified according to Z ratio groups: (1) Z ratio group 9 and 10.  259  Appendix 5. Continued 2. 40  30  5 be  20  10  0  0^20^40^60^80  ^  100  ^  120  Age @ b.h. (year)  40  30 1 so  20  E2 E-1  10  0  0^20^40^60^80  ^  100  ^  120  Age @ b.h. (year)  Figure A5.3.^Height growth curves from original stem analysis data stratified according to Z ratio groups: (1) Z ratio group 11 and 12.  260  Appendix 6^A BASIC program used to calculate the dominant height of white spruce stands in the SBS zone.  5 REM This is the program used to calculate dominant 10 REM height of white spruce stands growing in the SBS 15 REM biogeoclimatic zone 16 PRINT "*********************************************************" 17 PRINT "* A COMPUTER PROGRAMME TO PREDICT^*" 18 PRINT "* DOMINANT HEIGHT OF WHITE SPRUCE STAND *" 19 PRINT "*********************************************************" 20 PRINT "If ecological site classification information" 21 PRINT "available, please answer ESC=YES. Otherwise" 22 PRINT "answer ESC=NO" 25 INPUT "ESC=";ESC$ 30 IF ESC$="YES" THEN GOTO 200 35 PRINT "If heights at 30 and 60 years available," 36 PRINT "please answer ZG=YES. Otherwise answer ZG=NO" 40 INPUT "ZG=";ZG$ 45 IF ZG$="YES" THEN GOTO 100 46 PRINT "If site index available, please answer SITE=YES" 47 INPUT "SITE=";SITE$ 48 IF SITE$="YES" GOTO 51 49 PRINT "Sorry, height can not be reliably predicted!" 50 END 51 PRINT "Goudie & Mitchel model is the best model" 52 PRINT "available for height prediction!" 55 INPUT "Age=";age 60 INPUT "SI=";SI 65 LET A=1+exp(9.565-1.451*log(50)-1.236*log(SI-1.3)) 70 LET B=1+exp(9.565-1.451*log(age)-1.236*log(SI-1.3)) 75 LET H=1.3+(SI-1.3)*A/B 80 PRINT "H=";H 85 END 100 PRINT "Pattern-specific model is the best model" 101 PRINT "available for height prediction!" 105 INPUT "SI=";SI 106 INPUT "age=";age 110 INPUT "Z ratio=" ;Z 115 IF Z>1.60 THEN GOTO 125 120 LET H=1.3+1.376*(SI-1.3)*(1-exp(-0.03354*age))A1.542 121 GOTO 170 125 IF Z>1.72 THEN GOTO 135 130 LET H=1.3+1.536*(SI-1.3)*(1-exp(-0.02801*age))A1.524 131 GOTO 170 135 IF Z>1.85 THEN GOTO 145 140 LET H=1.3+1.893*(SI-1.3)*(1-exp(-0.02016*age))A1.407 141 GOTO 170 145 IF Z>1.97 THEN GOTO 155 150 LET H=1.3+1.928*(SI-1.3)*(1-exp(-0.02245*age))A1.666 151 GOTO 170 155 IF Z>2.10 THEN GOTO 165  261 160 LET H=1.3+1.990*(SI-1.3)*(1-exp(-0.02456*age))A1.981 161 GOTO 170 165 LET H=1.3+2.507*(S1-1.3)*(1-exp(-0.01942*age))A1.929 170 PRINT "H=";H 180 END 200 PRINT "If site index available, please answer SITE=YES" 205 INPUT "SITE=";SITE$ 206 IF SITE$="YES" THEN GOTO 300 215 PRINT "For height prediction, you must know site groups" 216 PRINT "For assigning sites into site groups, please" 217 PRINT "refer to G. Wang's (1992) Ph.D Thesis at page ???" 218 PRINT "If you are able to assign sites into site" 219 PRINT "groups, please answer SG equals one of A, B," 225 PRINT "C, D, E, F, G, H, I, J, K, L,or M. Otherwise" 226 PRINT "answer SG=NO" 227 INPUT "SG=";SG$ 228 IF SG$="NO" THEN GOTO 49 229 PRINT "Site-specific model (2) is the best model for" 230 PRINT "height prediction!" 233 INPUT "age=";age 235 IF SG$="A" THEN GOTO 250 236 IF SG$="B" THEN GOTO 250 237 IF SG$="C" THEN GOTO 250 238 IF SG$="D" THEN GOTO 250 239 If SA$="E" THEN GOTO 255 240 IF SA$="F' THEN GOTO 255 241 IF SG$="G" THEN GOTO 260 242 IF SG$="H" THEN GOTO 265 243 IF SG$="I" THEN GOTO 265 244 IF SG$="J" THEN GOTO 270 245 If SA$="K' THEN GOTO 275 246 IF SA$="L" THEN GOTO 280 247 IF SA$="M" THEN goto 280 250 LET H=1.3+28.65*(1-exp(-0.02061*age))A1.544 251 PRINT "H=";H 252 END 255 LET H=1.3+30.93*(1-exp(-0.02560*age))^1.559 256 PRINT "H=";H 257 END 260 LET H=1.3+37.50*(1-exp(-0.02192*age))^1.498 261 PRINT "H=";H 262 END 265 LET H=1.3+38.63*(1-exp(-0.01868*age))^1.434 266 PRINT "H=";H 267 END 270 LET H=1.3+22.42*(1-exp(-0.03795*age))^2.206 271 PRINT "H=";H 272 END 275 LET H=1.3+24.66*(1-exp(-0.02223*age))^1.921 276 PRINT "H=";H 277 END 280 LET H=1.3+10.70*(1-exp(-0.01984*age))^1.680 281 PRINT "H=";H 282 END  262  300 PRINT "For height prediction, you must know how to" 301 PRINT "assign sites into site series groups" 310 PRINT "For assigning sites into site series groups," 311 PRINT "please refer to G. Wang's (1993) Ph.D thesis" 312 PRINT "at page 98" 320 PRINT "If you are able to assign sites into sites" 321 PRINT "into site series groups, please answer SSG" 322 PRINT "equals one of A, B, C, or D. Otherwise answer" 323 PRINT "SSG=NO" 330 INPUT "SSG=";SSG$ 335 IF SSG$="NO" THEN GOTO 51 340 PRINT "Site-specific model (1) is the best model for" 342 PRINT "height prediction!" 350 INPUT "SI=";SI 351 INPUT "age=";age 355 IF SSG$="A" GOTO 360 356 IF SSG$="B" GOTO 365 357 IF SSG$="C" GOTO 370 358 IF SSG$="D" GOTO 375 360 LET X=1+exp(6.175-1.582*log(50)-0.1382*log(SI-1.3)) 361 LET Y=1+exp(6.175-1.582*log(age)-0.1382*log(SI-1.3)) 362 LET H=1.3+(SI-1.3)*X/Y 363 PRINT "H=";H 364 END 365 LET X=1+exp(7.465-1.429*log(50)-0.5512*log(SI-1.3)) 366 LET Y=1+exp(7.465-1.429*log(age)-0.5512*log(SI-1.3)) 367 LET H=1.3+(SI-1.3)*X/Y 368 PRINT "H=";H 369 END 370 LET X=1+exp(9.473-1.536*log(50)-1.066*log(SI-1.3)) 371 LET Y=1+exp(9.473-1.526*log(age)-1.066*log(SI-1.3)) 372 LET H=1.3+(SI-1.3)*X/Y 373 PRINT "H=";H 374 END 375 LET X=1+exp(9.696-1.846*log(50)-0.8361*log(SI-1.3)) 376 LET Y=1+exp(9.696-1.846*log(age)-0.8361*log(SI-1.3)) 377 LET H=1.3+(SI-1.3)*X/Y 378 PRINT "H=";H 379 END Note: The program can be run in any PC under GWBASIC environment. You simply give the answers to the asked questions. The four of the five models, except site-specific model (3), listed in Table 6.10 have been used in the program. Any of them could be selected to predict the dominant heights of white spruce stands depending on the available information. The computer will choose the best possible model for the prediction.  

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:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0075206/manifest

Comment

Related Items