@prefix vivo: .
@prefix edm: .
@prefix ns0: .
@prefix dcterms: .
@prefix skos: .
vivo:departmentOrSchool "Forestry, Faculty of"@en ;
edm:dataProvider "DSpace"@en ;
ns0:degreeCampus "UBCV"@en ;
dcterms:creator "Kurz, Werner Alexander"@en ;
dcterms:issued "2010-10-13T16:40:11Z"@en, "1989"@en ;
vivo:relatedDegree "Doctor of Philosophy - PhD"@en ;
ns0:degreeGrantor "University of British Columbia"@en ;
dcterms:description """Several of the current generation of computer models which simulate biomass production in forest ecosystems require a quantitative understanding of the effects of site quality on foliage efficiency (the amount of biomass produced per unit of foliage) and on carbon partitioning between above- and belowground stand components. This study investigated changes in foliage efficiency and carbon allocation in 12 Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) stands for which the site indices ranged from 19.5 to 41.3 m at 50 years. These stands are located on Vancouver Island, British Columbia, Canada.
Regression models for aboveground biomass components were developed from 39 destructively sampled Douglas-fir trees. Foliage biomass was predicted with a model which uses diameter at breast height (dbh) and a competition index as independent variables. This model predicts that stand foliage biomass stabilizes after canopy closure. Diameter and tree mortality data of the 12 Douglas-fir stands were available for 15 to 16 years, and were used to calculate aboveground and coarse root biomass and annual production estimates. In 1985, aboveground biomass in the 12 stands, with ages from 32 to 70 years, ranged from 135 to 573 Mg ha⁻¹. Coarse root biomass was estimated to be equal to 20 - 23% of aboveground biomass. In the period between 1985 and 1987, annual aboveground production (ANPP) in these 12 stands ranged from 4.7 to 16.0 Mg ha⁻¹ year⁻¹. Coarse root production was estimated to be equal to 13 -16% of aboveground production.
Fine (0-2 mm) and small (2-5 mm) root biomass and production estimates were derived by analyzing soil cores collected in five of the stands on 5 to 6 sampling dates over a 12 month period. All five stands showed similar seasonal dynamics in live fine root biomass, with the highest values occurring in May and the lowest values in October. In May 1985, biomass of living fine and small roots ranged from 1.82 to 7.91 Mg ha⁻¹ and from 0.59 to 4.10 Mg ha⁻¹, respectively. Three different methods of computing production and mortality were assessed. Different estimates were obtained for annual production and annual mortality in both fine and small roots, because fine root mortality exceeded production during the year which had a very dry summer. Estimates derived using one of the computational methods (decision-matrix) ranged from 1.12 to 5.14 Mg ha⁻¹ year⁻¹ for fine root production and from 2.15 to 4.89 Mg ha⁻¹ year⁻¹ for fine root mortality. Small root production and mortality estimates based on this computational method ranged from 0.51 to 2.22 and from 0.88 to 2.13 Mg ha⁻¹ year⁻¹, respectively.
With increasing site index, a decreasing proportion of total production was allocated to belowground stand components. The site with the lowest site index allocated about 31 to 51% of total net production to belowground components while the site with the highest site index allocated about 23 to 30% belowground. About 56% of the variation in 72 estimates (12 stands and 6 measurement periods) of foliage efficiency based on aboveground production was accounted for by a regression model with foliage biomass and site index as independent variables. This model suggests that there is an optimum foliage biomass at which total aboveground production is maximized and that this optimum foliage biomass increases with increasing site index.
The results of this study emphasize the importance of understanding variation in canopy function and shifts in carbon allocation from above to belowground stand components. Forest ecosystem production simulation models should include an explicit representation of changes in foliage efficiency and carbon allocation patterns to be able to accurately predict the responses of forest ecosystems to changes in environmental conditions and to silvicultural treatments."""@en ;
edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/29134?expand=metadata"@en ;
skos:note "NET PRIMARY PRODUCTION, PRODUCTION ALLOCATION, AND FOLIAGE EFFICIENCY IN SECOND GROWTH DOUGLAS-FIR STANDS WITH DIFFERING SITE QUALITY By WERNER ALEXANDER KURZ Diplom Holzwirt, University of Hamburg, West-Germany, 1976 THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Faculty of Forestry) We accept this thesis as conforming to the required standard The University of British Columbia September 1989 © Werner Alexander Ktirz, 1989 In presenting this thesis in partial fulfillment 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 representative. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Forestry University of British Columbia 2075 Wesbrook Place Vancouver, B.C., Canada V6T 1W5 Date: September, 1989 ii ABSTRACT Several of the current generation of computer models which simulate biomass production in forest ecosystems require a quantitative understanding of the effects of site quality on foliage efficiency (the amount of biomass produced per unit of foliage) and on carbon partitioning between above- and belowground stand components. This study investigated changes in foliage efficiency and carbon allocation in 12 Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) stands for which the site indices ranged from 19.5 to 41.3 m at 50 years. These stands are located on Vancouver Island, British Columbia, Canada. Regression models for aboveground biomass components were developed from 39 destructively sampled Douglas-fir trees. Foliage biomass was predicted with a model which uses diameter at breast height (dbh) and a competition index as independent variables. This model predicts that stand foliage biomass stabilizes after canopy closure. Diameter and tree mortality data of the 12 Douglas-fir stands were available for 15 to 16 years, and were used to calculate aboveground and coarse root biomass and annual production estimates. In 1985, aboveground biomass in the 12 stands, with ages from 32 to 70 years, ranged from 135 to 573 Mg ha\"1. Coarse root biomass was estimated to be equal to 20 - 23% of aboveground biomass. In the period between 1985 and 1987, annual aboveground production (ANPP) in these 12 stands ranged from 4.7 to 16.0 Mg ha\"1 year1. Coarse root production was estimated to be equal to 13 -16% of aboveground production. Fine (0-2 mm) and small (2-5 mm) root biomass and production estimates were derived by analyzing soil cores collected in five of the stands on 5 to 6 sampling dates over a 12 month period. All five stands showed similar seasonal dynamics in live fine root biomass, with the highest values occurring in May and iii the lowest values in October. In May 1985, biomass of living fine and small roots ranged from 1.82 to 7.91 Mg ha\"1 and from 0.59 to 4.10 Mg ha\"1, respectively. Three different methods of computing production and mortality were assessed. Different estimates were obtained for annual production and annual mortality in both fine and small roots, because fine root mortality exceeded production during the year which had a very dry summer. Estimates derived using one of the computational methods (decision-matrix) ranged from 1.12 to 5.14 Mg ha^ year\"1 for fine root production and from 2.15 to 4.89 Mg ha_1year_1 for fine root mortality. Small root production and mortality estimates based on this computational method ranged from 0.51 to 2.22 and from 0.88 to 2.13 Mg ha^year1, respectively. With increasing site index, a decreasing proportion of total production was allocated to belowground stand components. The site with the lowest site index allocated about 31 to 51% of total net production to belowground components while the site with the highest site index allocated about 23 to 30% belowground. About 56% of the variation in 72 estimates (12 stands and 6 measurement periods) of foliage efficiency based on aboveground production was accounted for by a regression model with foliage biomass and site index as independent variables. This model suggests that there is an optimum foliage biomass at which total aboveground production is maximized and that this optimum foliage biomass increases with increasing site index. The results of this study emphasize the importance of understanding variation in canopy function and shifts in carbon allocation from above to belowground stand components. Forest ecosystem production simulation models should include an explicit representation of changes in foliage efficiency and carbon allocation patterns to be able to accurately predict the responses of forest ecosystems to changes in environmental conditions and to silvicultural treatments. iv TABLE OF CONTENTS ABSTRACT ii LIST OF TABLES vii LIST OF FIGURES . . x ACKNOWLEDGEMENT xiii 1. GENERAL INTRODUCTION . 1 2. DESCRIPTION OF STUDY SITES . 5 3. BIOMASS REGRESSION EQUATIONS 14 3.1 INTRODUCTION 14 3.2 REVIEW OF EXISTING MODELS 17 3.2.1 Foliage biomass regression models for Douglas-fir 17 3.2.2 Branchwood biomass regression models for Douglas-fir 23 3.2.3 Stem component biomass regression models for Douglas-fir 23 3.3 MATERIALS AND METHODS 29 3.3.1 Site Description 29 3.3.2 Biomass Sampling . 29 3.3.3 Competition Indices 32 3.3.4 Statistical analysis 35 3.3.5 Data from the Shawnigan Lake study 35 3.4 RESULTS 36 3.4.1 Foliage biomass for individual branches 36 3.4.2 Branchwood biomass for individual branches 39 3.4.3 Foliage biomass regression models 39 3.4.4 A test of the models on an independent data set 50 3.4.5 Branchwood biomass regression models 55 3.4.6 Regression models for stem biomass components 60 3.5 DISCUSSION 66 3.6 CONCLUSIONS 70 V 4. ABOVEGROUND AND COARSE ROOT BIOMASS AND PRODUCTION 71 4.1 INTRODUCTION 71 4.2 LITERATURE REVIEW 71 4.2.1 Annual stemwood and stembark production 72 4.2.2 Annual branchwood production 72 4.2.3 Annual foliage production 74 4.3 MATERIALS AND METHODS 74 4.3.1 Site description 74 4.3.2 Field measurements and data processing 74 4.3.3 Calculation of biomass and net production 79 4.3.3.1 Stemwood 80 4.3.3.2 Stembark 80 4.3.3.3 Branchwood . 81 4.3.3.4 Foliage 87 4.3.3.5 Coarse roots 88 4.4 RESULTS 89 4.4.1 Branch biomass turnover 89 4.4.2 Foliage biomass turnover 91 4.4.3 Aboveground and coarse root biomass and production . 94 4.5 DISCUSSION 106 5. FINE AND SMALL ROOT BIOMASS AND PRODUCTION 112 5.1 INTRODUCTION 112 5.2 LITERATURE REVIEW 113 5.2.1 Estimating fine root biomass 113 5.2.2 Horizontal and vertical distribution of fine roots . . 114 5.2.3 Nutrient availability and fine root biomass 115 5.2.4 Estimating fine root production and turnover . 118 5.3 METHODS 122 5.3.1 Location and description of the study sites 122 5.3.2 Sample Collection . . . 122 5.3.3 Sample Preparation 126 5.3.4 Decision criteria for the sorting of roots 127 5.3.5 Ash content 129 5.3.6 Data processing and analyses 130 5.3.7 Calculation of production and mortality estimates 131 5.4 RESULTS AND DISCUSSION 136 5.4.1 Ash content of root samples 136 v i 5.4.2 Fine root biomass in May 1985 . 136 5.4.3 Seasonal dynamics of fine roots 142 5.4.4 Seasonal dynamics of small root biomass 154 5.4.5 Non-coniferous root biomass 157 5.4.6 Fine and Small Root Production 157 5.4.6.1 Calculating production and mortality estimates 157 5.4.6.2 Fine root production and mortality estimates 162 5.4.6.3 Small root production and mortality estimates 166 5.4.6.4 Turnover rates of fine and small roots 170 5.4.6.5 Site quality and fine and small root production 174 5.4.7 General Discussion 177 6. THE RELATIONSHIPS BETWEEN SITE INDEX AND FOLIAGE BIOMASS, FOLIAGE EFFICIENCY, PRODUCTION, AND PRODUCTION ALLOCATION 180 6.1 INTRODUCTION 180 6.2 MATERIALS AND METHODS 183 6.3 RESULTS 184 6.3.1 Foliage biomass versus site index 184 6.3.2 Foliage efficiency (ANPP) versus site index and foliage biomass . . 184 6.3.3 Production allocation versus site index 191 6.4 DISCUSSION .201 6.4.1 Foliage biomass versus site index 201 6.4.2 Foliage efficiency (ANPP) versus site index and foliage biomass . . 202 6.4.3 Optimum foliage biomass in Douglas-fir stands 204 6.4.4 Production allocation versus site index 206 7. SUMMARY AND CONCLUSIONS 210 8. REFERENCES 213 LIST OF TABLES 2.1. Installation and plot numbers as assigned by the Forest Productivity Committee, nearest town, latitude and longitude, and elevation of the study sites . 9 2.2. Mean annual precipitation and temperature at climate stations near the study sites 10 2.3. Regional climate (biogeoclimatic variants), soil moisture regime and soil nutrient regime of the six installations 11 2.4. Selected environmental characteristics of the five study sites 12 2.5. Stand age, stand origin, site index (SI), stand density, and basal area distribution in 1985 13 3.1. Regression models for calculating foliage biomass of Douglas-fir 18 3.2. Ten regression models for the prediction of foliage biomass (FOL) of Douglas-fir with dbh or basal area (BA) as independent variable 22 3.3. Six regression models for the prediction of foliage biomass (FOL) of Douglas-fir with sapwood area (SA) as independent variable 22 3.4. Six regression models for the prediction of branchwood biomass (BRA) of Douglas-fir with dbh as independent variable 24 3.5. Six regression models for the prediction of stemwood biomass (STW) of Douglas-fir with dbh as independent variable .26 3.6. Six regression models for the prediction of stembark biomass (STB) of Douglas-fir with dbh as independent variable 26 3.7. Stand characteristics of the six auxiliary plots (0.04 ha) in which trees were destructively sampled 30 3.8. Regression models to predict foliage biomass per branch (g) from branch fresh weight (g) and whorl number (Model 3.6). 38 3.9. Regression models to predict foliage biomass per branch (g) from branch diameter (cm) and whorl number (Model 3.7) 38 3.10. Regression models to predict branchwood biomass per branch (g) from branch freshweight (g) and whorl number (Model 3.8) 40 3.11. Regression models to predict branchwood biomass per branch (g) from branch diameter (cm) and whorl number (Model 3.9) 40 3.12. Mean, standard deviation (S.D.), and range of tree measurements, biomass data, and competition indices for the 39 sample trees. ........ 42 3.13. Statistics of ten models predicting Douglas-fir foliage biomass 45 3.14. Eight models predicting Douglas-fir foliage biomass. The significant models from Table 3.13 predicted foliage biomass and linear regression of the form ACTUAL = b0+b1*PREDICTED are calculated 45 3.15. Percent bias for 3 foliage biomass models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by diameter class. . 46 3.16. Percent bias for 3 foliage biomass models from this study and one viii regional model (RM, Gholz et al. 1979) for the combined data set and stratified by Installation 46 3.17. Five models for the prediction of foliage biomass for Douglas-fir applied to the Shawnigan Lake data set 52 3.18. Percent bias for 5 models for the prediction of foliage biomass for the combined data set and stratified by treatment 52 3.19. Contribution of four competition indices to the prediction of foliage biomass of Douglas-fir trees (n=76) from four treatments of the Shawnigan Lake experiment 53 3.20. Statistics of five models to predict branchwood biomass (grams) for Douglas-fir 56 3.21. Four models to predict branchwood biomass for Douglas-fir. The significant models from Table 3.20 predicted branchwood biomass and linear regression of the form ACTUAL = bo+b]*PREDICTED are calculated. ... 56 3.22. Percent bias for 2 branchwood biomass models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by diameter class 57 3.23. Percent bias for 2 branchwood biomass models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by Installation 57 . 3.24. Statistics of regression models to predict stem biomass components (kg) for Douglas-fir. 62 3.25. Seven models to predict stemwood biomass for Douglas-fir. The significant models from Table 3.24 and a regional model predicted stemwood biomass and linear regression of the form ACTUAL = b0+b^PREDICTED are calculated 63 3.26. Percent bias for stemwood and stembark biomass as calculated with 2 models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by diameter class. 64 3.27. Percent bias for stemwood and stembark biomass as calculated from 2 models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by Installation 65 4.1. Maximum percentage of total branchwood biomass encountered in a single whorl (whorl number) for each of the six destructively sampled plots 90 4.2. Coefficients, sample size (n), R2, and standard error of estimate (SEE) of six equations to calculate, for individual branches, the percentage of foliage in the first age class 92 4.3. Stand age, site index (SI), basal area (BA), and stand density in 1985 for the 12 study plots 96 4.4. Total stand biomass in 1985 of foliage, branches, stemwood, stembark, and coarse roots 99 4.5. The distribution of the biomass components listed in Table 4.4, expressed as a percentage of aboveground biomass 100 4.6. Annual production of foliage, branches, stemwood, stembark, and coarse root biomass 103 4.7. The distribution of the production listed in Table 4.6, expressed as a ix percentage of aboveground biomass production .104 5.1. Stand age, site index (SI), basal area (BA), and stand density in 1985 for the five study plots. Douglas-fir (Df) BA and stand density are listed in absolute amounts and as a percentage of the total 123 5.2. Sampling dates, sampling interval, and number of cores processed per stand 124 5.3. The populations of fine roots are subdivided into one (0-2 mm) or 2 diameter classes (0-1 mm, 1-2 mm) and/or one (FF-50 cm) or 3 soil horizons (FF, 0-30 cm, 30-50 cm) 135 5.4. Mean and standard error of the mean (...) of ash content expressed as percent of sample dry weight for three horizons and ten root classes. . . . 138 5.5. Mean and standard error of the mean (...) of live fine (0-2 mm) root biomass in May 1985 at the five study sites 140 5.6. Live fine root biomass in May 1985 in three horizons, expressed as a percentage of the total fine root biomass, and on a volume basis (g nr3) for each of three horizons and the total profile to a depth of 50 cm 141 5.7. Mean and standard error of the mean (...) of live small (2-5 mm) root biomass in May 1985 at the five study sites . 155 5.8. The effects of separating fine roots into populations according to diameter and/or soil layer (Groups I-VI) on estimates of production and mortality. . 161 5.9. Annual fine root production and mortality (g nr2 yr1) for the five stands calculated with three computational methods 164 5.10. Annual fine root production and mortality for the five stands expressed as a percentage of the estimates obtained from the SC computational method 165 5.11. Annual small root production and mortality (g m-2 yr\"1) for the five stands calculated using three computational methods 167 5.12. Turnover rates (year1) of fine roots calculated as the ratios of annual fine root production / fine root biomass in May 1985 and annual fine root mortality / fine root biomass in May 1985 171 5.13. Turnover rates (year1) of small roots calculated as the ratios of annual small root production / small root biomass in May 1985 and annual small root mortality / small root biomass in May 1985 172 5.14. The relationships between site index and several measures of fine and small root production and mortality 176 6.1. The relationship between site index and the proportion of total biomass and total production allocated to foliage, branches, stemwood, stembark, and coarse roots 198 6.2. The proportion of total production allocated belowground for five Douglas-fir stands 199 6.3. The relationship between site index and the proportion of total production allocated belowground for four different estimates of belowground production 200 X LIST OF FIGURES 1.1. A general overview of the major components of this dissertation. 4 2.1. The location of the six research installations on eastern Vancouver Island, British Columbia, Canada 8 3.1. Foliage biomass as predicted by 10 regression models which use dbh or basal area as the independent variable 19 3.2. Foliage biomass as predicted by 6 regression models which use sapwood area as the independent variable 21 3.3. Branchwood biomass as predicted by 6 regression models which use dbh as the independent variable 25 3.4. Stemwood biomass as predicted by 6 regression models which use dbh as the independent variable 27 3.5. Stembark biomass as predicted by 6 regression models which use dbh as the independent variable. 28 3.6. Foliage biomass as predicted by Model 3.19 47 3.7. Foliage biomass as predicted by Model 3.20 48 3.8. Foliage biomass as predicted by Model 3.21 49 3.9. Comparison of actual and predicted foliage biomass of 76 Douglas-fir trees in 4 treatments of the Shawnigan Lake experiment 54 3.10. Branchwood biomass as predicted by Model 3.26 58 3.11. Branchwood biomass as predicted by Model 3.27. 59 4.1. Contour maps of Growing Space Index (GSI) in Installation 2, Plot 6, without (A) and with (B) hypothetical buffer strip. 77 4.2. The scheme used to establish a hypothetical stand around each plot 78 4.3. Theoretical distribution of branch biomass by whorl number 83 4.4. Branch biomass distribution of 39 destructively sampled trees from 6 plots 84 4.5. Percentage of foliage in the first year age class for 267 sample branches from 6 Installations. 93 4.6. Total stand density plotted against time for the 12 sample plots 97 4.7. Basal area plotted against time 98 4.8. Aboveground biomass plotted against time 101 4.9. Foliage biomass plotted against time 102 4.10. Aboveground production plotted against time 105 4.11. Total aboveground biomass versus basal area from 42 different Douglas-fir stands. 110 4.12. Total aboveground production versus foliage biomass from 38 different Douglas-fir stands. HI 5.1. Schematic diagram of the soil corer and the sampling depths collected with it 1 2 5 xi 5.2. The decision matrix,(Fairley and Alexander 1985), modified. 134 5.3. Mean and standard error of the mean of ash content expressed as percent of dry-weight for live and dead roots from three soil horizons and four diameter classes 139 5.4. Seasonal dynamics of live (top) and dead (bottom) fine root biomass in the five stands 146 5.5. Seasonal dynamics of live and dead fine root biomass at each of the five stands. 147 5.6. Mean monthly precipitation (30 year average) at Nanaimo Airport, and actual precipitation for 1985 and the first six months of 1986 148 5.7. Daily precipitation at the Cowichan Lake Research Station for 1985 and the first six months of 1986, and the live fine root biomass at Stand E approximately 1 km from the climate station 149 5.8. Live fine root biomass in three soil layers at each of the five study sites. . 150 5.9. Dead fine root biomass in three soil layers at each of the five study sites. 151 5.10. Total (live plus dead) fine root biomass at each of the five study sites. . . . 152 5.11. Live and dead fine root biomass in May 1986 expressed as a percentage of the May 1985 values 153 5.12. Seasonal dynamics of live and dead small root biomass for each of the five stands 156 5.13. Seasonal dynamics of non-coniferous root biomass in the five study sites. 158 5.14. Estimates of fine root production and mortality for five plots based on the Decision Matrix and Significant Changes methods. 168 5.15. Estimates of small root production and mortality for five plots based on the Decision Matrix and Significant Changes methods 169 5.16. Fine root mean life span for 5 plots based on produciton and mortality estimates obtained with two computaional methods. 173 5.17. The relationships between fine and small root biomass and production and site index based on data from this and several published studies. . . . 179 6.1. The relationship between foliage biomass (Mg ha-1) and site index (m at 50 years) based on (A) the foliage biomass regression model developed in this study and (B) the model of Gholz et al (1979) 185 6.2. The relationships between foliage efficiency (Mg yr 1 Mg-1), calculated from total aboveground production (ANPP) and (A) foliage biomass (Mg ha\"1) and (B) site index (m at 50 years). . 186 6.3. Foliage efficiency (ANPP) (Mg yr\"1 Mg\"1) plotted against foliage biomass (Mg ha\"1) for each of the 12 plots 188 6.4. Graphical presentation of Equation 6.1, which predicts foliage efficiency (ANPP) (Mg yr-1 Mg'1) as a function of site index (m at 50 years) and foliage biomass (Mg ha'1) 189 6.5. Aboveground production (ANPP) as a function of foliage biomass and site index (Equation 6.3, n=72). 190 6.6. The difference between actual foliage biomass (FB) and site-specific optimum foliage biomass (FBopt) for each of the 12 plots 192 xii 6.7. Proportions of aboveground biomass allocated to foliage, branches, stemwood, stembark, and coarse roots plotted against site index 193 6.8. Proportions of aboveground production allocated to foliage, branches, stemwood, stembark, and coarse roots plotted against site index 194 6.9. Aboveground (ANPP) and total (TNPP) annual production increase with site index 196 6.10. The partitioning of total production to above and belowground components for the five stands of this study and the two stands of Keyes and Grier (1981) 209 xiii ACKNOWLEDGEMENT I thank my supervisor, Dr. Hamish Kimmins, for his guidance, his editorial contributions, and his support through all stages of this research. The comments on an earlier draft by the other members of my supervisory committee, Dr. Holger Brix, Dr. Karel Klinka, Dr. Tony Kozak, and Dr. Gordon Weetman, were much appreciated. This project would not have been possible without the dedication of several people who spent many months processing root samples. I thank Claire Trethewey, Ray Morello, Patricia Riley, Debra Stowe, Nora Galdert, Shamsah Mohamed, Gordon Weber, and Thomas Smith for their patient and reliable assistance in processing the root samples. Aboveground biomass sampling was assisted by Trevor Charles, John Karakatsoulis, Wesley Mussio, and Sabrina Rampersad. The cooperation of Min Tzse in providing logistic support was much appreciated. The British Columbia Ministry of Forests, Research Branch, provided growth and yield data for the study plots. The cooperation of Paul Barker and Stephen Omule was much appreciated. Dr. Holger Brix (Forestry Canada) kindly provided biomass information from the Shawnigan Lake Research site. Aboveground biomass sampling was supported by financial contributions of the federal Environment 2000 program. Financial support for the belowground biomass and production study (Chapter 5) was provided by Forestry Canada (FRDA, Direct Delivery Program, Contribution Number 68576-52-6-2). I thank the World University Service of Canada for a personal scholarship. Finally, I would like to thank Dr. Ann McGee for her editorial comments, her patience, and for the support and encouragement she provided throughout the evolution of this dissertation. Her assistance is much appreciated. 1 1. GENERAL INTRODUCTION Until recently, forest production ecology has focussed on aboveground biomass and production because stems are the primary harvestable component of forest ecosystems and because of the difficulties involved in obtaining data on belowground biomass and production. Faced with a changing and uncertain future, it is becoming increasingly important to be able to make accurate predictions about the responses of forest ecosystems to anticipated changes in environmental conditions and management regimes. Such predictions require a sound under-standing and quantification of both above- and belowground stand components and of the factors that determine the partitioning of net production between them. Computer simulation models have become a very important tool for the prediction of forest growth (cf. Ek et al. 1988a, 1988b), because they can integrate existing knowledge about complex systems and make projections over time scales which are of interest to foresters. In order to develop, calibrate, and use such models, forest science must provide a quantitative understanding of the major growth determining ecosystem processes. The central concept underlying many simulation models of forest growth is that foliage biomass or area is multiplied by some measure of foliage production efficiency to obtain an estimate of total photosynthate production. An allocation scheme or hierarchy is then used to partition total production to stand components (McMurtrie and Wolf 1983, Grier et al. 1986, Barclay and Hall 1986, rCimmins et al. 1986, Makela and Hari 1986, Ford and Bassow 1988). Such an approach to modelling requires a quantitative understanding of the factors which determine foliage biomass, foliage production efficiency, and carbon allocation to above and belowground stand components. Few production studies in forest ecosystems have 2 investigated belowground production and therefore few have been able to distinguish between the relative contributions of changes in photosynthate production and photosynthate allocation to the observed differences in aboveground production. In earlier studies, belowground production was often assumed to represent some fixed proportion of aboveground production (Newbould 1967). In 1981, Keyes and Grier reported that the proportion of total production allocated to belowground stand components is affected by site conditions. They found that the difference in total production between a high and low productivity Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) stand in western Washington was much smaller than the difference in aboveground production, because trees growing on the better site allocated much less photosynthate to fine and small root production (Keyes and Grier 1981). This observed plasticity of resource partitioning in response to environmental conditions has raised questions about the interpretation of the responses of forest ecosystems to silvicultural treatments (Kurz 1989). The increase in aboveground production following nitrogen fertilization, for example, can be due to an increase in photosynthate production, or a shift in carbon allocation away from fine roots, or both (Brix 1983, Brix in press, Friend 1988, Axelsson and Axelsson 1986). With information about aboveground production only, no unequivocal interpretations of observed aboveground responses are possible. The overall objectives of the research summarized in this dissertation were to answer two questions for coastal second growth Douglas-fir stands growing on Vancouver Island, British Columbia: 1) Does foliage efficiency change with site quality? 3 2) Does carbon allocation to belowground stand components decrease with increasing site quality? Specifically, for each of a series of Douglas-fir stands growing on sites covering a range in site quality, the objectives were: i) to develop regression models, which use diameter and a competition index as independent variables, for the prediction of aboveground biomass components, ii) to quantify aboveground and coarse root biomass using these and other regression models, iii) to quantify fine and small root biomass, production, and mortality, and iv) to quantify foliage biomass and foliage efficiency. Figure 1.1 gives an overview of the major components of the research reported in this dissertation. Chapter 2 contains a description of the study sites. Each Chapter that reports research results (Chapters 3 to 6) includes introduction, methods, results, and discussion sections. Biomass regression models for foliage and other aboveground biomass components developed from destructively sampled trees are reported in Chapter 3. In Chapter 4, long-term diameter growth and tree mortality data are combined with remeasurement data from this study and used to calculate competition indices and their change over time for all trees on the study plots. Also in Chapter 4, these data and the regression models developed in Chapter 3 are used to compute aboveground and coarse root biomass and production estimates. Chapter 5 reports estimates of fine and small root production and mortality. The results of all previous chapters are combined in Chapter 6 to investigate the effects of site quality on foliage biomass, foliage efficiency, and production allocation. Conclusions are presented in Chapter 7. • D a t e • PfOQrams C C • Results Figure 1.1 A general overview of the major components of this dissertation. 5 2. DESCRIPTION OF STUDY SITES This study was conducted in six growth and yield research installations which are part of a network of such installations established by the Forest Productivity Committee, British Columbia Ministry of Forests (Darling and Omule 1989). The six installations of this study are located on the east side of Vancouver Island, British Columbia, Canada (Figure 2.1). Table 2.1 gives the location, elevation, and installation and plot numbers of each of the six study sites. Table 2.2 presents long-term averages of annual precipitation, annual temperature, and temperatures for the coldest and warmest months as measured at nearby climatic stations (Environment Canada 1982). Five of the six installations are located in the Eastern variant of the Very Dry Maritime Coastal Western Hemlock subzone (CWHxml) of the British Columbia biogeoclimatic ecosystem classification (Pojar et al. 1987, Green et al. 1984, Krajina 1969), while the sixth (Installation 72) is in the Western variant of that subzone (CWHxm2) (Table 2.3). Actual soil moisture regimes (Pojar et al. 1987, Green et al. 1984) range from moderately dry to fresh and actual soil nutrient regimes range from poor to rich (Table 2.3). The parent material in five of the installations is a morainal blanket while in the sixth it is fluvial material (Table 2.4). Soils are either Eluviated Dystric Brunisols or Duric Humoferric Podzols (Agriculture Canada Expert Committee on Soil Survey 1987) that are moderately well to well drained (Table 2.4). Forest floors in all installations are less than 2 cm thick. Mor humus forms were found in five Installations and Installation 72 had a Mull humus form. Each of the six installations contains up to 18 experimental plots, including the two untreated control plots on which this study is based. Details of site 6 selection and plot establishment are provided in Darling and Omule (1989). Plots are square (22.36 m x 22.36 m), have an area of 0.05 ha, and are surrounded by a 0.05 ha buffer zone. A distance of at least 12 meters separates the outer edge of the buffer zone from the nearest adjacent buffer zone or road, stream, or other changes in stand structure. In 1985, Douglas-fir (Pseudotsuga menziesii Mirb. Franco) accounted for 90% or more of the overstory basal area in nine of the twelve stands (Table 2.5). Western hemlock (Tsuga heterophylla (Raf.) Sarg.) was present as a second overstory species. It generally represented less than 4% of total basal area, but in three of the plots western hemlock accounted for 17.7 to 32.4% of total basal area. In four of the five plots in which western redcedar (Thuja plicata Donn) was present, it contributed less than 2% of the basal area; in one plot it accounted for 9.1%. In Installation 72, Plot 14, one bigleaf maple (Acer macrophyllum Pursh.) contributed 3.8% of the basal area (Table 2.5). In 1985, Installations ranged in age from 32 to 70 years and were either naturally regenerated, planted, or were planted and additional natural regeneration occurred (Table 2.5). All sites appear to have been burned after logging. Stand densities in 1985 ranged from 3400 to 440 stems per hectare of which 40 to 98 % were Douglas-fir trees. The relative density distribution of stems of other species was fairly similar to the basal area distribution of these species (Table 2.4). The understory vegetation of the lower site index plots was generally dominated by Gaultheria shallon Pursh with some Mahonia nervosa (Pursh) Nutt. As site quality improved, the dominance of Gaultheria shallon declined and Mahonia nervosa became a larger component of the understory. Mosses encountered on the low to medium quality plots were mainly Kindbergia oregana 7 (Sull.) Ochyra and Hylocomium splendens (Hedw.) B.S.G., while in the high quality sites Rhytidiadelphus loreus (Hedw.) Warnst, Rhytidiadelphus triquetrus (Hedw.) Warnst., and Rhytidiopsis robusta (Hook.) Broth were dominant. The understory vegetation of Installation 72 was dominated by Polystichum munitum (Kaufl.) Presl and Achlys triphylla (Sm.) DC. Frequently encountered herbs were Tiarella laciniata Hook., Tiarella trifoliata L., and Trientalis latifolia Hook. Mosses (Kindbergia oregana, Hylocomium splendens) were mostly confinded to decaying logs. 8 0 50 100 > i i Washington Figure 2.1. The location of the six research installations on eastern Vancouver Island, British Columbia, Canada. Numbers in circles refer to installation numbers (cf. Table 2.1). 9 Table 2.1. Installation and plot numbers as assigned by the Forest Productivity Committee, nearest town, latitude and longitude, and elevation of the study sites. Instal- Elevation lation Plots Location Latitude Longitude (m) 2 6 11 Chemainus 48°50'N 123°50'W 419 4 1 17 Cassidy 49°05'N 124°01*W 333 5 8 10 Bowser 49°25'N 124°48'W 360 16 2 6 Campbell River 49°58'N 125°30'W 335 71 11 14 Campbell River 50°02'N 125°29'W 244 72 2 14 Cowichan Lake 48°49'N 124°07'W 229 10 Table 2.2. Mean annual precipitation and temperature at climate stations near the study sites. Data from Canadian Climatic Normals 1951-1980 (Environment Canada 1982). Mean annual Mean annual Mean Jan. Mean July Climate precipitation temperature temperature temperature Inst Station (mm) (°C) (°C) (°C) 2 Nanaimo Airport 1104 9.3 1.8 17.4 4 Nanaimo Airport 1104 9.3 1.8 17.4 5 Mud Bay Fisheries 1714 9.3 1.8 17.4 16 Campbell River Airport 1406 8.2 0.4 16.6 71 Campbell River Airport 1406 8.2 0.4 16.6 72 Cowichan Lake Res. Stn. 2123 9.1 1.4 17.4 11 Table 2.3. Regional climate (biogeoclimatic variants), soil moisture regime and soil nutrient regime of the six installations according to Smith (1979) but following the more recent terminology of Green et al. (1984) and Klinka (pers. comm.). CWHxm = Very Dry Maritime Coastal Western Hemlock subzone, 1 = Eastern variant, 2 = Western variant. Instal- Biogeoclimatic Soil Moisture Soil Nutrient lation Variant Regime Regime 2 CWHxml moderately dry poor 4 CWHxml slightly dry medium 5 CWHxml slightly dry medium 16 CWHxml slightly dry medium 71 CWHxml slightly dry medium 72 CWHxm2 fresh rich 12 Table 2.4. Selected environmental characteristics of the five study sites. Data from Smith (1979). Inst. Parent material Soil classification Drainage Texture Coarse frag. (%) Humus forma 2 morainal blanket Eluviated Dystric Brunisol moderately well drained loam 30 Mor 4 morainal blanket Eluviated Dystric Brunisol well drained sandy loam 56 Mor 5 fluvial Duric Humo-Ferric Podzol well drained sandy loam 65 Mor 16 morainal blanket Duric Humo-Ferric Podzol moderately well drained sandy loam 45 Mor 71 morainal blanket Eluviated Dystric Brunisol well drained sandy loam 33 Mor 72 morainal blanket Duric Humo-Ferric Podzol moderately well drained sandy loam 48 Mull a Forest floor thickness in all plots is less than 2 cm. 13 Table 2.5. Stand age, stand origin, site index (SI) (Bruce 1981), stand density, and basal area (BA) distribution in 1985 for the 12 study plots. Stand Density Basal Areab Age SI Inst Plot Origin* (yr) (m@50) (st. ha'1) %DF (m^a\"1) %DF %WH %RC %WP %BM 2 6 N 42 27.7 3000 40.0 53.4 66.5 32.4 1.0 2 11 N 41 19.5 2840 71.8 33.6 80.3 17.7 1.9 4 1 C 44 29.1 1840 97.8 45.3 99.7 0.3 -4 17 C 48 25.7 1640 86.6 37.4 94.8 3.2 -5 8 P 40 26.8 3400 53.5 58.7 76.0 23.9 0.1 5 10 P 39 29.5 1420 87.3 47.4 96.9 3.1 -16 2 P 32 29.4 2000 98.0 41.3 98.6 0.2 1.2 16 6 P 32 32.4 2520 80.2 45.1 90.4 0.5 9.1 71 11 P 41 24.6 1880 97.9 45.3 99.5 0.5 -71 14 P 41 23.3 2460 97.6 46.0 99.1 0.9 -72 2 c 70 41.3 440 90.9 69.1 97.0 3.0 -72 14 c 70 41.0 480 95.8 75.3 96.2 - -a Stand origin: N= natural, P=plantation, C=combination. (from Darling and Omule 1989). b DF = Douglas-fir, WH = western hemlock, RC = western redcedar, WP = western white pine, BM = bigleaf maple. 14 3 . BIOMASS REGRESSION EQUATIONS 3.1 INTRODUCTION Biomass regression equations are widely used in both forest science and management (Satoo and Madgwick 1982, Cannell 1982). Such equations typically relate a variable which is difficult to measure, such as foliage biomass, to one or more other variables which are easier to measure, such as diameter at breast height (dbh). Regression models that describe relationships between stem biomass components and stem diameter typically have very high coefficients of determination (R2), which indicates that the independent variable accounts for most of the observed variation in the dependent variable. In contrast, models which relate crown biomass components to stem characteristics (e.g. foliage biomass to dbh) have much lower R2 values, indicating that factors other than those accounted for by the independent variable influence the relationship. Regression equations are often developed for predictive purposes. A sample of a population is measured in detail and the observed relationships are employed to calculate the variables of interest for the entire population. This works well if the population from which the regression models were developed and the population to which they are applied for prediction are similar. There are, however, many examples in forestry where models derived in one study have been used in other studies without ensuring that the allometric relationships described by those models are indeed similar in the two populations. Part of the problem is that it is often unclear what factors influence the allometric relationships. Thus, the criteria by which to judge whether two stands can be described by the same regression models are often not available. 15 One approach to dealing with between-stand variation has been to develop regional regression models that are based on sample trees from a large number of stands (Gholz et al. 1979, Standish et al. 1985). Such models average the variation between stands, but their predictions for individual stands may be substantially in error (Marshall and Waring 1986). A second approach is to incorporate into the regression equation additional independent variables which account for some of the remaining sources of variability. For example, Baskerville (1983) found that stand age accounts for some of the between-stand variation in foliage biomass regression models developed for balsam fir (Abies balsamea (L.) Mill.). The pipe model theory (Shinozaki et al. 1964a, b) suggests that a functional relationship exists between the cross sectional area of the water-conducting sapwood and the amount of foliage which can be supported by this sapwood. Several studies have since shown that sapwood basal area can be used to predict foliage biomass (Grier and Waring 1974, Snell and Brown 1978, Whitehead 1978). More recent studies have found that factors such as sapwood permeability (Whitehead et al. 1984) and mean annual sapwood ring width (Albrektson 1984) can account for additional variation in the foliage area-sapwood area relationship. Although the use of sapwood basal area as an independent variable in regression models may account for some of the between-stand variation, this variable is not always measurable, particularly in permanent sample plots where the need for repeated diameter measurements often precludes the use of increment corers. Other variables which have been shown to account for some of the residual variation in canopy biomass regression models, such as crown length or stem diameter at the base of the live crown, are difficult to measure on large numbers of trees. 16 Brown (1978) investigated the suitability of stand density, determined with a prism plot, as an independent variable in biomass regression models and found that it reduced the residual variation somewhat for some conifer species. A competition index might be superior to stand density in reducing the residual variation in biomass regression models, because it integrates the cumulative competitive influence of the trees surrounding a subject tree. Competition indices are based on various computational methods that typically use either the distance and relative size of competing trees or the crown area overlap. For detailed reviews see Noone and Bell (1980) and Daniels et al. (1986). Traditionally, these indices have been used in growth and yield studies to predict growth rates of trees following thinning (Smith and Bell 1983). None of the published regression models includes a competition index as independent variable, although stand maps are frequently available for research installations and competition indices can be computed from such data. In this study, four competition indices were tested for their contribution to regression models that predict the biomass of crown and stem components. The first objective of the research reported in this chapter was to provide regression equations for the prediction of foliage, branchwood, stemwood, and stembark biomass for my study areas. The second objective was to investigate the contribution of several competition indices to these regression models. It was hypothesized that competition indices, which measure the growing conditions experienced by individual trees, will significantly improve biomass regression equations, especially those for canopy biomass components. It was hoped that by adding these indices, the generality of the regression models, and their value to other studies, might be increased. One of the criteria for the development of these new models was that their independent variables should be easily measurable. This excluded the use of sapwood basal area because it can 17 only be measured by taking increment cores which may not be desirable in permanent sample plots where diameters will be measured repeatedly. 3.2 REVIEW OF EXISTING MODELS 3.2.1 Foliage biomass regression models for Douglas-fir Investigations into the relationships between foliage biomass and other tree variables in Douglas-fir date as far back as Burger (1935) and Kittredge (1944). While these relationships were merely statistical, Shinozaki et al. (1964 a,b) proposed a functional relationship between foliage biomass and the cross-sectional area of the water-conducting tissue supporting the foliage. This pipe model theory has become the basis for many investigations which relate foliage biomass to sapwood basal area. In Douglas-fir, as in many other tree species, sapwood basal area has often, though not always, been found to be superior to dbh as a predictor of foliage biomass (Grier and Waring 1974, Snell and Brown 1978, Brix and Mitchell 1983). Table 3.1 lists 32 published models for the prediction of Douglas-fir foliage biomass. The relationship between foliage biomass and the independent variable, dbh or basal area (BA), varies greatly (Figure 3.1, Table 3.2). For example, the predicted foliage biomass for a tree of 25 cm dbh ranges from 9.0 kg (Model 10) to 26.4 kg (Model 1), an almost threefold difference (294%). Regression coefficients of the 10 models presented in Figure 3.1 are listed in Table 3.2. Some of the regression lines in Figure 3.1 are extrapolated beyond the range of diameters from which the models were derived, but the above numeric example is within the range of diameters of Models 1 and 10. 18 Table 3.1. Regression models for calculating foliage biomass of Douglas-fir. n = number of sample trees, In = logarithm base e, log = logarithm base 10, d.b.l.c.= diameter at base of live crown, cw = crown width, BA = basal area, SA = sapwood area Independent Range of Variable n dbh (cm) Source log dbh 22 6-46 Kittredge 1944 dbh 29 Ahmed 1956 log dbh 5 Swank 1960 log dbh 35 6-46 Heilman 1961 dbh cw2 23 2-120 Kurucz 1969 log dbh 10 2-23 Dice 1970 log dbh 104 a x=11.6 Dice 1970 log dbh 8 9-111 Woodard 1974 SA 36 Grier and Waring 1974 log dbh Gholz et al. 1976 In dbh 29 2-163 Grier and Logan 1977 In dbh 5 34-53 Kay 1978 In d.b.l.c. 5 34-53 Kay 1978 In dbh 18 1-11 Snell and Brown 1978 In SA 18 1-11 Snell and Brown 1978 log dbh 123 a 2-162 Gholz et al. 1979 dbh2ht 171 a 2-162 Shaw 1979 SA 14 Granier 1981 SA b.l.c. 13 Granier 1981 BA 15 Granier 1981 SA 96 b 5-25 Brix and Mitchell 1983 BA u. bark 96 b 5-25 Brix and Mitchell 1983 In dbh 8 c 6-29 Feller et al. 1983 In dbh 10 5-56 Feller et al. 1983 In height 10 Feller et al. 1983 In height 16 Feller et al. 1983 log dbh 26 d 9-30 Grier et al. 1984 dbh2ht 49 4.5-66.0 Standish et al. 1985 dbh 12 10-25 Borghetti et al. 1986 SA 12 10-25 Borghetti et al. 1986 In dbh 40 Espinosa Bancalari and Perry 1987 lndbh 18 1.4-13.4 Helgerson et al. 1988 a This study compiles data previously published elsewhere. b 24 trees in each of 4 treatments. c Old (height > 2.5m) and young ( < 2.5m) stands on good and poor sites separated, d 13 trees in a control and fertilizer treatment. 19 23 4 56 80 (/> (0 < O m LU CD < o 60 40 20 8 9 10 15 25 35 45 55 65 DBH (cm) Figure 3.1. Foliage biomass as predicted by 10 regression models which use dbh or basal area as the independent variable. Details of the models are listed in Table 3.2. 20 The relationship between foliage biomass and sapwood area (SA) is equally variable (Figure 3.2) for six regression models (Table 3.3). For example, a tree of 250 cm2 SA is predicted by Model 6 to have 10.0 kg and by Model 1 to have 20.2 kg of foliage biomass, a twofold difference. All models in Table 3.1 include, as the independent variable, one or several measures of dimensions of individual trees, but do not account for any additional sources of variability in the relationship between that variable and foliage mass. Factors which may affect this allometric relationship include stand density, and nutrient and moisture availability (Grier et al. 1986, Brix and Mitchell 1983). Estimating foliage mass for an individual stand using regression models from a different stand, or with combined regional models (e.g. Gholz et al., 1979), may yield large errors (Grier et al., 1984, Marshall and Waring 1986). Madgwick (1983) identifies the need to determine additional variables which may affect the relationship between crown weight and individual stem dimensions. As pointed out by Madgwick (1983), a new approach is required which identifies sources of variability not accounted for in the models listed in Table 3.1. Competition indices may account for some of the residual variability as they describe the competitive status of a tree relative to its neighbours. 21 Figure 3.2. Foliage biomass as predicted by 6 regression models which use sapwood area as the independent variable. Details of the models are listed in Table 3.3. 22 Table 3.2. Ten regression models for the prediction of foliage biomass (FOL) of Douglas-fir with dbh or basal area (BA) as independent variable. In = logarithm base e, log = logarithm base 10. Standard error of estimate (SEE) is in logarithmic units. Model R2 SEE Source Number* InFOL (kg) FOL (kg) InFOL (kg) InFOL (g) logFOL (g) logFOL (g) InFOL (kg) InFOL (kg) FOL (kg) InFOL (kg) -4.791+2.502*lndbh 0.92 -2.688+0.054*BA 0.74 -6.093+2.723*lndbh 0.93 3.329+2 0.643+2 1.159+2 -2.846+1 -4.151+1 -8.296+0 -3.890+1 031*lndbh 0.94 ,396*logdbh 0.87 097*logdbh 0.82 ,701*lndbh 0.86 ,982*lndbh 0.96 ,979*dbh 0.92 ,890*lndbh 0.88 0.160 Grier etal. 1984 1 Granier 1981 2 0.240 Espinosa Bancalari and Perry 1987 3 0.348 Helgerson et al. 1988 4 0.194 Dicel970b 5 0.279 Dicel970c 6 0.695 Gholz et al. 1979 7 0.176 Grier & Logan 1977 8 Borghetti et al. 1986 9 Gholz et al. 1976 10 a refers to labels in Figure 3.1 b Cedar River data set c combined data set Table 3.3. Six regression models for the prediction of foliage biomass (FOL) of Douglas-fir with sapwood area (SA) as independent variable. In = logarithm base e, log = logarithm base 10. SEE is in logarithmic units. Model R2 SEE Source Numbera FOL (kg) = -2.020+0.089*SA 0.85 FOL (kg) = -1.365+0.074*SA 0.92 FOL (kg) = -1.340+0.072*SA 0.97 FOL (kg) = -1.004+0.071*SA 0.86 FOL (kg) = -0.030+0.052*SA 0.78 InFOL (g) = 3.996+0.938*lnSA 0.96 Brix & Mitchell 1983c 1 Granier 1981 2 Grier & Waring 1974 3 Borghetti et al. 1986 4 Brix & Mitchell 1983b 5 0.274 Snell & Brown 1978 6 a refers to labels in Figure 3.2 b control plot data c all data 23 3.2.2 Branchwood biomass regression models for Douglas-fir Six regression models for the prediction of branchwood biomass in Douglas-fir are summarized in Table 3.4 and plotted in Figure 3.3. Model 1 (Helgerson et al. 1988) was derived from a data set that included young trees with a maximum of 13.4 cm dbh. The plotted line for Model 1 in Figure 3.3 is clearly an extrapolation beyond the range of the original data, but even at 13 cm dbh this model differs greatly from the other five models. As with foliage biomass models, there are no criteria by which to judge which model will yield the best biomass prediction for a specific stand. Note that predictions from the regional model (Table 3.4, Model 4, Gholz et al. 1979) are approximately in the middle of the range (Figure 3.3). 3.2.3 Stem component biomass regression models for Douglas-fir Stemwood and stembark biomass are treated as two separate biomass components. Both components are highly correlated with dbh and the range of biomass predictions is much narrower than in the crown biomass components. Table 3.5 lists 6 biomass regression equations for stemwood biomass which are plotted in Figure 3.4. Table 3.6 and Figure 3.5 present six stembark biomass prediction models. Note that both the predicted stemwood and stembark biomass of the regional models (Tables 3.5 and 3.6, Model 1, Gholz et al. 1979) are the highest of the range of models presented in Figures 3.4 and 3.5. 24 Table 3.4. Six regression models for the prediction of branchwood biomass (BRA) of Douglas-fir with dbh as independent variable. In = logarithm base e, log = logarithm base 10. SEE is in logarithmic units. Model R2 SEE Source Numbera lnBRA(g) = 2.856+2.503*lndbh 0.94 0.440 lnBRA(g) = ln(1.64)+2.96*lndbh 0.81 0.073 lnBRA(kg) = -4.456+2.469*lndbh 0.86 0.200 lnBRA(kg) = -3.694+2.138*lndbh 0.92 0.631 logBRA (g) = 0.945+2.388*logdbh 0.90 0.230 logBRA (g) = 1.112+2.162*logdbh 0.90 0.156 Helgerson et al. 1988 1 Barclay et al. 1986b 2 Grier et al. 1984 3 Gholz et al. 1979 4 Dice 1970c 5 Dice 1970d 6 a refers to labels in Figure 3.3 b control plot data c combined data d Cedar River data 25 1 5 15 25 35 45 55 65 DBH (cm) Figure 3.3. Branchwood biomass as predicted by 6 regression models which use dbh as the independent variable. Details of the models are listed in Table 3.4. 26 Table 3.5. Six regression models for the prediction of stemwood biomass (STW) of Douglas-fir with dbh as independent variable. In = logarithm base e, log = logarithm base 10. SEE is in logarithmic units. Model R2 SEE Source Numbera InSTW (kg) logSTW(g) InSTW (g) logSTW (kg) InSTW (kg) InSTW (g) =-3.040+2.595*lndbh 0.99 0.310 = 1.636+2.609*logdbh 0.98 0.064 =-4.747+2.967*lndbh 0.98 0.097 = 1.857+2.444*logdbh 0.89 0.320 =-2.603+2.367*lndbh 0.97 0.080 = ln(99.61)+2.28*lndbh 0.98 0.025 Gholz et al. 1979 1 Dice 1970b 2 Dice 1970c 3 Espinosa Bancalari and Perry 1987 4 Grier et al. 1984 5 Barclay et al. 1986d 6 a refers to labels in Figure 3.4 b combined data c Cedar River data d control plot data Table 3.6. Six regression models for the prediction of stembark biomass (STB) of Douglas-fir with dbh as independent variable. In = logarithm base e, log = logarithm base 10. SEE is in logarithmic units. Model R2 SEE Source Numbera InSTB (kg) InSTB (kg) InSTB (kg) logSTB(g) InSTB (g) logSTB (g) = -4.310+2.430*lndbh = -5.610+2.701*lndbh = -4.906+2.530*lndbh = l,169+2.328*logdbh =ln(16.31)+2.30*lndbh = 1.347+2.165*logdbh 0.99 0.322 Gholz et al. 1979 1 0.85 0.340 Espinosa Bancalari and Perry 1987 2 0.94 0.130 Grier et al. 1984 3 0.98 0.027 Dicel970b 4 0.98 0.027 Barclay et al. 1986c 5 0.97 0.115 Dicel970d 6 a refers to labels in Figure 3.5 b combined data c control plot data d Cedar River data 27 5 15 25 35 45 55 65 DBH (cm) Figure 3.4. Stemwood biomass as predicted by 6 regression models which use dbh as the independent variable. Details of the models are listed in Table 3.5. 28 Figure 3.5. Stembark biomass as predicted by 6 regression models which use dbh as the independent variable. Details of the models are listed in Table 3.6. 29 3.3 MATERIALS AND METHODS 3.3.1 Site Description The six research installations in which non-destructive measurements were made are described in detail in Chapter 2. Stand characteristics of the 6 auxiliary plots used for destructive sampling, and statistics of the sample trees from those plots are summarized in Table 3.7. 3.3.2 Biomass Sampling Destructive sampling of trees inside the control plots of the Productivity Committee Installations was not permissible. In each Installation, one temporary auxiliary plot (20 x 20 m) was therefore established in a section which was representative of the stand conditions in the control plots. All trees were numbered and dbh (1.3 m) measurements taken. Based upon a stratified random sampling scheme (by dbh-class), 6 to 8 trees in each auxiliary plot were selected for biomass sampling. Each tree was felled and processed within one day to minimize potential needle weight loss due to respiration (Forrest 1966). In total, 39 trees were sampled. Prior to felling of the sample trees, the distance and direction (compass bearing) to, and the dbh of, each of the neighbouring trees were determined. After felling a sample tree, total height, length of the live crown, and distance from the top of the stem for each live whorl were determined. 30 Table 3.7. Stand characteristics of the six auxiliary plots (0.04 ha) in which trees were destructively sampled. %DF = percent of total represented by Douglas-fir. Statistics are based on 1984 data. Basal area Density • Agea Heighta Sample Inst, (n^ ha\"1) %DF (st. ha'1) %DF (years) (m) size 2 35.5 92 1850 87 34 17.4 7 4 54.3 96 1975 87 40 21.7 6 5 22.8 81 1425 63 30 16.1 8 16 39.8 99 2125 99 24 17.4 6 71 43.8 99 2100 98 34 18.2 6 72 80.1 96 625 72 67 44.2 6 a mean of sample trees. 31 For every whorl, each living branch was cut from the stem, weighed and its length was measured to the nearest cm. Branch diameter near the base was determined to the nearest mm as the mean of two orthogonal caliper measurements. The condition of each branch was recorded to indicate whether sections were lost during the felling or whether other abnormalities were observed. One branch selected at random from all branches of every third whorl was processed further. Within each whorl the number and combined weight of all dead branches, as well as all non-nodal branches, were also recorded. The selected branches from every third whorl, counting from the top down, were clipped into sections by age class. All sections for which an age determination was not possible were combined in an \"unidentified\" age-class. The samples of twig sections and foliage, as well as all remaining branchwood, were taken to the laboratory and weighed the same day. A subsample of approximately 25g (or the entire sample if its freshweight was less than 25g) was randomly selected from each needle age-class and dried in a forced air drying oven at 105° C for 24 hours. Later, these subsamples were manually separated into needles and twigs, redried at 105° C for 24 hours, and the component weights determined to the nearest 0.1 gram. The stems were cut into 6 to 13 sections (depending on tree size) and the length of each section was determined. At the bottom of each section a disk was cut. Each disk was measured along four orthogonal radii to obtain: radius, thickness of bark and sapwood, and increment over the last 5 years. The number of annual rings in the sapwood and the total age of each disk were also recorded. Subsamples of sapwood, heartwood, and bark were collected from approximately every second disk for the determination of specific density (Forest Products Laboratory 1952). 32 Biomass values for stemwood, stembark, and total stem (wood plus bark) were obtained by calculating the volume of each component in each stem section and multiplying it by the appropriate specific density. Volumes were calculated using Smalian's formula (Husch et al. 1982). Specific densities for sapwood, heartwood and bark were obtained from stem disks cut at each section, or by linear interpolation between adjacent disks. The biomass of each section was summed to obtain stem totals. 3.3.3 Competition Indices In this study, four competition indices were tested for their contribution to regression models that predict the biomass of crown and stem components. The four indices are the Competitive Influence Zone Overlap (CIO) (Bella 1971), the Competitive Stress Index (CSI) (Arney 1973), the Diameter-Distance Competition Index (DCI) (Hegyi 1974), and the Growing Space Index (GSI) (Lin 1974). Competitive Stress Index (CSI) (Arney 1973): Arney's (1973) Competitive Stress Index (CSI) is based on the observation that in open-grown trees of a particular species, there is a high correlation between stem diameter (dbh) and crown width. This relationship for open-grown trees is used to compute the hypothetical crown width for the subject tree and each of its competitors. The sum of the area of overlap of these hypothetical crowns is defined as CSI which is computed as: 33 n CSI = ((( Z Oy) + Aj)/Aj) * 100 i=l [3.1] where Oy = area of zone overlap (m2), Aj = influence zone of the subject tree (m2), and n = number of competitors. The crown width in this study was computed from Arney's (1973) model for the combined data sets of western Oregon and B.C. which include 290 open-grown trees. For this study the model has been converted to metric units. Competitive Influence Zone Overlap (CIO) (Bella 1971): Bella's (1971) Competitive Influence Zone Overlap (CIO) sums the hypothetical overlap between the influence zones of the subject tree and its competitors. The influence zone is defined as a zone with three times the diameter of the crown width of the open-grown tree. Crown width is computed from the same regression equation as described above. This competition index also accounts for the diameter ratio of the competing trees and the subject tree. The equation to compute the CIO is: where Ojj = area of zone overlap (m2), Aj = influence zone of the subject tree (m2), = diameter of the i t h competing tree (cm), Dj = diameter of the j t h subject tree (cm), and n = number of competitors. n [3.2] 34 Diameter-Distance Competition Index (DCI) (Hegyi 1974): Hegyi (1974) included all competitors within a set radius around the subject tree, without specifying the radius used. The formula to compute the DCI is: n Dj 1 DCI= X ( * ) [3.3] i=l Dj DISy where Dj = dbh of the i t h competing tree (cm), Dj = dbh of the j t h subject tree (cm), and DISjj = distance between i t h competitor and j t h subject tree (m). Growing Space Index (GSI) (Lin 1974): Lin's Growing Space Index (GSI) considers only the cumulative influence of one tree in each of four quadrants surrounding the subject tree. The decision whether a neighbouring tree exerts a competitive influence is based on the angle (0) between two lines which, at breast height, connect the centre of the subject tree with the outside of the stem of the competing tree. This angle is a function of both the distance between subject tree and competitor and the dbh of the competitor. In each quadrant, the tree with the largest 0 is the competing tree. Each quadrant is initially assigned a growing space of 25 which is reduced depending on 0. If 0 < 2.15°, no competition occurs (GSIj=25) or if 0 > 5.25°, maximum competition occurs (GSL=0). For 2.15°>0<5.25°: Dj + Di GSL= 25-(0-2.15)* 8.0645* — [3.4] 2*Dj where Dj = diameter of the i t h competing tree (cm), and Dj = diameter of the j t h subject tree (cm). 35 The sum of the GSIj for the four quadrants is: 4 GSI= X GSIi [3.5] i=l which by definition yields 0 < GSI < 100. 3.3.4 Statistical analysis Numerical analyses were performed with MIDAS (Fox and Guire 1976) and SYSTAT (Wilkinson 1988b). Logarithmic transformations were applied to variables when scatter diagrams indicated that such transformations would yield linear models. Such logarithmic transformations also reduce heteroscedasticity, i.e. the increase of the variance of Y at any X in proportion to the value of X is reduced or eliminated (Zar 1984:286). The systematic bias introduced by such transformations (Baskerville 1972) is reduced by applying a correction factor (exp(SEE2/2)) (Sprugel 1983) whenever equations are converted to their anti-log form. 3.3.5 Data from the Shawnigan Lake study The Biology of Forest Growth study (Crown and Brett, 1975) at Shawnigan Lake, Vancouver Island, is the source of an independent data set which was used to test the foliage biomass regression models developed for this study. The Shawnigan Lake study is investigating the effects of thinning and fertilization on growth of Douglas-fir. Data from four treatments were used: T0F0 - control plots T2F0 - 2/3 of basal area removed at time of thinning 36 T0F2 - 448 kg/ha N applied T2F2 - combined thinning and fertilizer treatment. Details of the experimental design and the site are described by Crown and Brett (1975) and by Barclay and Brix (1985). The destructive sampling of trees for biomass equations is described in Brix and Mitchell (1983) and in Barclay et al. (1986). In addition to the information published previously, distance and bearing to the nearest competitors and their diameters have been recorded for 76 of the 96 trees used by Brix and Mitchell (1983). This information was made available by Dr. Holger Brix (Forestry Canada, Victoria). Based on these data, competition indices for the 76 sample trees have been computed using the same methods as described above. 3.4 RESULTS 3.4.1 Foliage biomass for individual branches In total, 3149 branches were measured for diameter, length, branch freshweight, whorl number and height in the crown. Of these, 230 branches were further separated into needles and twigs by age-class for the determination of dry weights. Regression equations were developed1 which predict foliage and branchwood biomass for individual branches. The highest R2 values were obtained when the 1 Morello, R. (1986). Regression models for predicting foliage and branchwood biomass per branch in Douglas-fir. B.Sc. Thesis, University of British Columbia, Vancouver, BC. 91 pp. 37 models included branch freshweight and whorl number (counting from the top down) as independent variables. Not all branches were intact after felling the trees, and freshweight, as measured after felling, was therefore not always representative. Branches which had sections missing after felling had been coded appropriately. Foliage and branchwood biomass of such branches were calculated from a second regression model which used branch diameter and whorl number as independent variables. Complete measurements, including branch freshweight, were available for 2594 branches, 82.4% of the total. Table 3.8 lists regression models for each of the six Installations and for the combined data set. Foliage dryweight per branch is predicted from the model InFOL = b0 + b^lnBFWT + b2*WHORLN2 [3.6] where InFOL is the natural logarithm of foliage dryweight (g), InBFWT is the natural logarithm of branch freshweight (g), and WHORLN is the number of the whorl. In Table 3.9, similar statistics are shown for the second model type which predicts foliage biomass per branch from branch diameter and whorl number: InFOL = b0 + b^lnDIA + b2*WHORLN3, [3.7] where InDIA is the natural logarithm of branch diameter (cm) and other variables are as in model 3.6. F-tests showed that, in each model type, some of the installation-specific equations are significantly different (p=0.05) from others and that therefore a combined model is not applicable in either case. Installation-specific regression models were therefore applied to the data from each installation to calculate foliage biomass for all branches of the sample trees. 38 Table 3.8. Regression models to predict foliage biomass per branch (g) from branch freshweight (g) and whorl number (Model 3.6). SEE is in logarithmic units. Inst R2 SEE n V b i b 2 2 .947 .286 47 -2.3074 1.1756 -0.0019196 4 .973 .227 36 -1.5759 1.0528 -0.0015907 5 .975 .187 44 -1.7268 1.0780 -0.0018445 16 .965 .192 28 -1.6962 1.0882 -0.0046709 71 .966 .200 33 -1.8342 1.0979 -0.0023376 72 .977 .191 42 -1.1793 0.9616 -0.0015755 ALL .959 .248 230 -1.6081 1.0478 -0.0018554 Table 3.9. Regression models to predict foliage biomass per branch (g) from branch diameter (cm) and whorl number (Model 3.7). SEE is in logarithmic units. Inst R2 SEE n bo bi b 2 2 .820 .527 47 3.5214 2.7813 -0.0000970 4 .873 .489 36 3.5602 2.7313 -0.0000913 5 .905 .365 44 3.4714 2.6966 -0.0001275 16 .865 .375 28 3.4855 3.0913 -0.0004972 71 .857 .407 33 3.5167 2.7843 -0.0001641 72 .904 .390 42 3.4121 2.4674 -0.0000764 ALL .856 .390 230 3.4568 2.6099 -0.0000969 39 3.4.2 Branchwood biomass for individual branches Regression models for the prediction of branchwood biomass were developed from the data set described in the previous section. The independent variables branch freshweight and whorl number yielded the highest R2 values and the lowest SEE in the model: ln(BRA) = b0 + b^ lnCBFWT) + b2* WHORLN, [3.8] where BRA = branchwood dryweight (g), BFWT = branch freshweight (g), and WHORLN is the number of the whorl. A second model was developed to be used for the prediction of branchwood biomass of those branches for which branch freshweight did not represent the condition of the branch prior to felling. The model which best predicted branchwood biomass was ln(BRA) = b0 + b]*ln(DIA) + b2*WHORLN, [3.9] where DIA = branch diameter (cm) and other variables as in model 3.8. Tables 3.10 and 3.11 list the regression equations for each of the six Installations and for the combined data set, for models 3.8 and 3.9, respectively. Some of the installation-specific equations of model 3.8 differed from others, as determined by F-tests (p=0.05). Model 3.9 had no significant differences between the equations. For both models, installation-specific equations were used to predict branchwood biomass. 3.4.3 Foliage biomass regression models Summary statistics for dbh, height, biomass, and the four competition indices for the 39 sample trees are presented in Table 3.12. Models 3.6 and 3.8, with specific regression coefficients for each installation, were applied to 82.4% of 40 Table 3.10. Regression models to predict branchwood biomass per branch (g) from branch freshweight (g) and whorl number (Model 3.8). SEE is in logarithmic units. Inst R2 SEE n b0 bx b2 -2.8292 0.98751 0.62063 -2.3409 0.94293 0.54197 -2.3733 0.99359 0.44686 -1.6902 0.81482 0.68041 -2.2889 0.91410 0.60983 -2.1937 0.97060 0.46828 -2.3911 0.96853 0.52261 Table 3.11. Regression models to predict branchwood biomass per branch (g) from branch diameter (cm) and whorl number (Model 3.9). SEE is in logarithmic units. Inst R2 SEE n bo bi b2 2 .952 .308 47 2.0114 2.3645 0.65479 4 .945 .374 36 2.1687 2.3794 0.60136 5 .966 .250 44 2.2397 2.4279 0.53269 16 .975 .181 28 2.5696 2.5075 0.41756 71 .966 .242 33 2.6310 2.5585 0.35416 72 .980 .233 42 2.2517 2.4453 0.53480 ALL .966 .274 230 2.3024 2.4678 0.52487 2 .975 .225 47 4 .990 .156 36 5 .980 .193 44 16 .989 .117 28 71 .976 .202 33 72 .992 .149 42 ALL .981 .202 230 41 the 3149 branches to calculate foliage and branchwood biomass for each branch, respectively. For the 17.6% of the branches of which sections were missing, models 3.7 and 3.9 were applied. For each tree, foliage biomass was summed to obtain total foliage biomass and total branchwood biomass per tree. The contribution of each competition index (CI) was highly significant (p<0.001) (Table 3.13) when included in the simple model: InFOL = b0 + b^ lndbh + b2*CI, [3.10] where InFOL is the natural logarithm (base-e) of foliage biomass (g) and lndbh is the logarithm of the diameter at breast height (dbh, in cm). When dbh was added to the model so that: InFOL = b0 + bi*lndbh + b2*dbh + b3*CI [3.11] the contribution of the competition indices remained significant (p<0.05). Dbh, however, did not contribute significantly to models 3.22 and 3.23 (Table 3.13). When computing predicted values from regression models based on logarithmic transformations, a correction factor was applied as suggested by Baskerville (1972) and Sprugel (1983): Predicted = exp(model) * exp(SEE212), [3.12] where model is the regression equation in its logarithmic form and SEE is the standard error of estimate of this model. In regression models based on logarithmic transformation of the dependent variable, the coefficient of determination (R2) and the standard error of estimate (SEE) are affected by the transformation. When the models are transformed back into the original units, the coefficient of determination and the SEE can be calculated from a new linear model Actual = b0 + b^ Predicted. [3.13] 42 Table 3.12. Mean, standard deviation (S.D.), and range of tree measurements, biomass data, and competition indices for the 39 sample trees. Variable Mean S.D. Min. Max. Tree measurements dbh (cm) 22.6 13.4 6.8 59.2 height (m) 22.0 10.4 8.1 48.4 age (years at b.h) 37.6 13.5 24.0 67.0 Biomass (kg) Total stema 322.1 554.1 12.9 2140.4 Stemwood 277.7 481.3 11.2 1816.2 Stembark 44.4 73.7 1.7 324.2 Foliage 8.97 9.38 0.31 35.69 Branchwood 16.47 34.69 0.43 91.22 Competition Indices'3 GSI 24.8 19.25 0.0 65.2 CSI 437.1 111.93 253.1 784.8 CIO 9.2 7.41 2.1 33.0 DCI 4.0 2.82 0.8 13.5 a stemwood plus stembark b see Methods for explanation 43 Table 3.14 shows R2 and SEE for the significant models (3.14 to 3.22) in anti-log form. Models 3.20 and 3.21 rank first and second with respect to highest R2 and lowest SEE. Model 3.19 is the best model which does not include a competition index. The three regression equations are: InFOL = -0.2681+3.4051*lndbh-0.0587*dbh [3.19] InFOL = 0.3081+2.9902*lndbh-0.0411*dbh+0.0104*GSI [3.20] InFOL = 2.1452+2.8449*lndbh-0.0456*dbh-0.0024*CSI [3.21] Additional statistics are listed in Tables 3.13 and 3.14. The regression coefficients of the competition indices in models 3.20 and 3.21 have opposing signs because of the way the indices are computed: intense competition is expressed by a low GSI value and by a high CSI value. Figure 3.6 shows a plot of model 3.19. The surfaces described by models 3.20 and 3.21 are plotted in Figures 3.7 and 3.8. The combinations of dbh and competition index in the data set are displayed in the lower sections of Figures 3.7 and 3.8. Note that some regions of the regression surfaces are not defined by the sample trees, e.g. large open-grown trees are absent. Such regions should therefore be regarded as tentative extrapolations of the model. Large, open grown trees do not occur in the data sets to which the model will later be applied. Bias of regression models is another valuable criterion by which to judge their performance. Here, percent bias is expressed as the mean residual (actual -predicted) divided by the mean actual value multiplied by 100: n A n Percent Bias = 100 * (( X (Yi - Yi))/n) / (( X (yi))/n) [3.24] i=l i=l A where yj = predicted value, yj = actual value, and n = number of sample trees. 44 Table 3.15 shows percent bias for each of 6 diameter classes and for the combined data set for models 3.19, 3.20, and 3.21 and, for comparative purposes, for a regional model (Gholz et al. 1979) frequently used to predict foliage biomass. Bias for the combined data set is less than 1% for the 3 models that were derived from this data set. In contrast, the model from the literature has an average bias of -55.2%, i.e. the model greatly overestimated foliage biomass of the sample trees. Models 3.19, 3.20, and 3.21 consistently overestimated foliage biomass of trees in the 10-15 cm diameter class. Model 3.20 has little bias (<8%) for all diameter classes except the 10-15 cm dbh class where bias is -29.9%. The regional model always overestimated foliage biomass. It had a bias of -233.4% and -99.9% for the smallest and largest dbh classes, respectively. In models 3.19, 3.20, and 3.21 over- and under-estimation alternate in successive diameter classes, indicating that there is no lack-of-fit in these three models. The regional model's bias suggests that lack-of-fit is a problem. The models' bias in predicting foliage biomass of the sample trees from the 6 Installations is summarized in Table 3.16. The bias of model 3.20 ranges from -15.1 to 14.4%. The ranges of models 3.19 and 3.21 are somewhat larger. Such bias will introduce some error in the predictions of foliage biomass on a stand basis, but these biases are much lower than those that result from the application of the regional model from the literature (Table 3.16). The only way to reduce the biases further is by using 6 separate regression equations for the 6 Installations. It seemed preferable, however, to select one model that adequately describes the entire data set (n=39) rather than 6 plot specific models, each based on a much smaller and inadequate sample size (n=6 to 8). While site specific equations may reduce bias in predicting foliage biomass of the small number of sample trees from each site, they may result in increased error 45 Table 3.13. Statistics of ten models predicting Douglas-fir foliage biomass (grams, n=39), CI = Competition Index. R2 is based on log-transformed data. SEE is in logarithmic units. Significance (p) Model R2 SEE Const, lndbh dbh CI No. C+lndbh 0.843 0.4889 <.001 <.001 [3.14] C+lndbh+GSI 0.890 0.4150 <.001 <.001 <.001 [3.15] C+lndbh+CSI 0.894 0.4078 <.001 <.001 <.001 [3.16] C+lndbh+CIO 0.896 0.4035 <.001 <.001 <.001 [3.17] C+lndbh+DCI 0.894 0.4075 <.001 <.001 <.001 [3.18] C+lndbh+dbh 0.884 0.4262 .756 <.001 .001 - [3.19] C+lndbh+dbh+GSI 0.907 0.3861 .702 <.001 .015 .005 [3.20] C+lndbh+dbh+CSI 0.917 0.3655 .035 <.001 .003 <.001 [3.21] C+lndbh+dbh+CIO 0.889 0.4024 .075 .002 .282 .026 [3.22] C+lndbh+dbh+DCI 0.904 0.3924 .048 <.001 .059 .010 [3.23] Table 3.14. Eight models predicting Douglas-fir foliage biomass. The significant models from Table 3.13 predicted foliage biomass and linear regression of the form ACTUAL = bn+bj*PREDICTED are calculated. R2 is based on non-transformed data and SEE is in actual (non-logarithmic) units. Model R2 SEE bo No. C+lndbh 0.687 5320.3 2988.8 0.575 [3.14] C+lndbh+GSI 0.797 4283.7 2058.6 0.702 [3.15] C+lndbh+CSI 0.704 5171.2 2699.7 0.636 [3.16] C+lndbh+CIO 0.787 4385.8 1012.4 0.868 [3.17] C+lndbh+DCI 0.755 4706.8 1432.9 0.806 [3.18] C+lndbh+dbh 0.756 4691.4 301.0 0.962 [3.19] C+lndbh+dbh+GSI 0.867 3467.0 -185.0 1.013 [3.20] C+lndbh+dbh+CSI 0.814 4098.5 33.7 1.001 [3.21] 46 Table 3.15. Percent bias (calculated as mean residual divided by mean actual foliage biomass, times 100) for 3 foliage biomass models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by diameter class. %Bias of Model DBH n Actual 3.19 3.20 3.21 RM 5-65 39 8971.2 -0.8 -0.9 0.5 -55.7 5-10 6 672.1 -14.8 -1.7 -0.8 -233.4 10-15 6 2063.7 -20.0 -29.9 -24.1 -122.6 15-20 8 5470.7 4.6 6.7 3.9 -39.4 20-25 8 7946.9 -7.0 -5.3 -7.5 -39.4 25-35 6 17259.6 15.1 7.6 13.8 -5.2 35-60 5 24513.0 -10.5 -5.6 -5.4 -99.9 Table 3.16. Percent bias (calculated as mean residual divided by mean actual foliage biomass, times 100) for 3 foliage biomass models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by Installation. %Bias of Model Inst n Actual 3.19 3.20 3.21 RM ALL 39 8971.2 -0.8 -0.9 0.5 -55.7 2 7 10103.1 29.7 14.4 18.2 5.4 4 6 9436.5 13.7 6.4 14.6 -15.2 5 8 4376.3 -0.9 9.2 4.0 50.7 16 6 5099.8 -0.7 0.7 -0.8 -45.3 71 6 5172.2 -14.6 -15.1 -13.6 -59.8 72 6 20982.5 -21.0 -12.6 -13.1 -111.2 47 5 15 25 35 45 55 65 DBH(cm) Figure 3.6. Foliage biomass as predicted by Model 3.19. See text for more details. 48 Figure 3.7. Foliage biomass as predicted by Model 3.20. See text for more details. 49 Figure 3.8. Foliage biomass as predicted by Model 3.21. See text for more details. 50 when predicting foliage biomass of trees from different sites. The 6 plots from which the sample trees were collected are in the vicinity of, but not identical to, the plots for which the predictions will be made. The general models, some of which incorporate a competition index, will therefore be used for biomass prediction. 3.4.4 A test of the models on an independent data set Brix and Mitchell (1983) used 96 trees for the development of foliage biomass regression equations. The auxiliary data required to compute competition indices were available for 76 of these trees. Five regression models were used to predict foliage biomass of the 76 trees: models 3.19, 3.20 and 3.21 of this study, the regional model of Gholz et al. (1979), and the model of Brix and Mitchell (1983) that predicts foliage biomass from basal area under bark which was derived from the combined data set (n=96). Table 3.17 shows the coefficients of determination (R2) and the standard error of estimate (SEE) for each of the five models in non-logarithmic units. As before, these were derived from two computational steps. The predicted values were derived from the five regression models described above. Then the regression between actual and predicted values (Equation 3.13) was computed. The R2 value of the latter expresses the proportion of the variation in the actual data that is accounted for by the predicted data. The five models accounted for 82.4 to 85.2% of the variation in the Shawnigan Lake data set (Table 3.17). Model 3.21, which uses lndbh, dbh and CSI as independent variables, has the highest R2 value. Note, however, that the slopes 51 (bj) of the regression lines differ greatly from one, which indicates that systematic bias in the prediction of foliage biomass occurs. All five models tended to underestimate foliage biomass (Table 3.18). The amount of bias varied with treatment and with model. Figure 3.9 shows predicted versus actual foliage biomass for each of the four treatments as predicted by model 3.20. The discrepancy was largest for the T2F2 treatment (34.7% bias), in which trees differed most from the untreated trees, and the bias was smallest for the control plots (T0F0, bias = 12.1%). Finally, the question was addressed whether competition indices would also contribute significantly to regression models derived from the Shawnigan Lake data set. Models of the form lnfol = b0 + bi*lndbh + b3*CI, [3.10] where CI is one of the four competition indices, were calculated. Table 3.19 lists five models and the significance of each of the regression coefficients. Three of the four competition indices contributed significantly to the regression model, the exception being CIO (p=0.061). 52 Table 3.17. Five models for the prediction of foliage biomass for Douglas-fir applied to the Shawnigan Lake data set. The models have the form ACTUAL = bo+bi *PREDICTED. R2 is based on non-transformed data and SEE is in actual (non-logarithmic) units. No. Model R2 SEE [3.19] C+lndbh+dbh 0.839 2593.1 -759.6 2.158 [3.20] C+lndbh+dbh+GSI 0.824 2707.2 289.2 1.327 [3.21] C+lndbh+dbh+CSI 0.852 2485.7 -1198.2 1.436 [ - ] Gholz et al. 1979 0.834 2633.3 -4228.6 1.949 [ - ] Brix & Mitchell 1983 0.837 2606.4 -733.9 0.890 Table 3.18. Percent bias (calculated as mean residual divided by mean actual foliage biomass, times 100) for 5 models for the prediction of foliage biomass for the combined data set and stratified by treatment. Treatment n Actual 3.19 %Bias of model 3.20 3.21 Gholza B&M 83b T0F0 16 4312.4 34.5 12.1 5.3 -12.2 21.4 T2F0 22 9958.8 50.4 20.6 16.5 26.7 8.7 T0F2 14 7078.8 40.0 22.9 13.2 8.4 -15.1 T2F2 24 14892.1 55.6 34.7 30.8 39.0 11.6 ALL 76 9797.4 50.1 26.9 21.9 26.5 4.1 a Gholz et al. 1979 b Brix and Mitchell 1983 53 Table 3.19. Contribution of four competition indices to the prediction of foliage biomass of Douglas-fir trees (n=76) from four treatments of the Shawnigan Lake experiment. CI = Competition Index. R2 is based on log-transformed data and SEE is in logarithmic units. Model R2 SEE Significance (p) Const. lndbh CI C+lndbh C+lndbh+GSI C+lndbh+CSI C+lndbh+CIO C+lndbh+DCI 0.877 0.914 0.928 0.910 0.926 0.276 0.265 0.243 0.271 0.246 .005 .001 <.001 .877 <.001 <.001 <.001 <.001 <.001 <.001 .008 <.001 .061 <.001 54 u 5 Actual Actual o 1 Q . a1 0 , Actual Actual Figure 3.9. Comparison of actual and predicted foliage biomass of 76 Douglas-fir trees in 4 treatments of the Shawnigan Lake experiment. Predictions are based on model 3.20 from this study. TOFO = control, T0F2 = fertilized, T2F0 = thinned, and T2F2 = thinned and fertilized. 55 3.4.5 Branchwood biomass regression models Three of the four competition indices contributed significantly to the model InBRA = b0 + b^ lndbh + b2*CI [3.25] where InBRA is the logarithm (base-e) of branchwood biomass, lndbh is the logarithm of the diameter at breast height (dbh, in cm), and CI is the competition index (Table 3.20). CIO was the only competition index that was not significant. Adding dbh as additional independent variable did not contribute significantly (p=0.05) to the models. As with the foliage biomass models, the branchwood models were converted to the anti-log form with the appropriate correction factors. The regression equations between actual and predicted values (see Equation 3.10) are summarized in Table 3.21. Model 3.27 yielded the highest R2 and the lowest SEE. Percent bias, as defined above, was lowest for model 3.27, averaging -2.0%. Percent bias for diameter classes and for the six installations for each of the 4 models and for the regional model of Gholz et al. (1979) is listed in Tables 3.22 and 3.23. The two models selected for computation of branchwood biomass are model 3.27, which includes a competition index, and model 3.26, which is the best model without a competition index. The coefficients of these two models are: InBRA = 1.9036+2.3522+lndbh [3.26] InBRA = 1.7018+2.3471*lndbh+0.0087*GSI [3.27] Figures 3.10 and 3.11 present the two equations graphically. Note that the combination of dbh and competition index in the data set does not cover the entire range displayed in the graphs (Figure 3.11). Therefore, the predicted branchwood biomass for large trees with little competition must be regarded as a tentative extrapolation of the model. Such trees, however, do not occur in the data sets to which the models will be applied for prediction purposes. 56 Table 3.20. Statistics of five models to predict branchwood biomass (grams) for Douglas-fir (n=39), CI = Competition Index. R2 is based on log-transformed data and SEE is in logarithmic units. Model Significance (p) R2 SEE Const, lndbh CI No. C+lndbh C+lndbh+GSI C+lndbh+CSI C+lndbh+CIO C+lndbh+DCI 0.922 0.938 0.947 0.929 0.938 0.3804 0.3446 0.3193 0.3679 0.3433 <.001 <.001 <.001 <.001 <.001 <.001 <.001 <.001 <.001 <.001 .005 <.001 .067 .004 [3.26] [3.27] [3.28] [3.29] [3.30] Table 3.21. Four models to predict branchwood biomass for Douglas-fir. The significant models from Table 3.20 predicted branchwood biomass and linear regression of the form ACTUAL = b0+b1*PREDICTED are calculated. R2 is based on non-transformed data and SEE is in actual (non-logarithmic) units. Model R2 SEE b0 No. C+lndbh C+lndbh+GSI C+lndbh+CSI C+lndbh+DCI 0.888 0.922 0.885 0.900 7967.1 6640.4 8053.9 7526.1 1815.2 804.2 1511.3 208.6 0.833 0.932 0.891 0.997 [3.26] [3.27] [3.28] [3.29] 57 Table 3.22. Percent bias (calculated as mean residual divided by mean actual branchwood biomass, times 100) for 2 branchwood biomass models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by diameter class. %Bias of model DBH n Actual 3.26 3.27 RM ALL 39 16472.5 -6.8 -2.0 -69.9 5-10 6 1001.1 -13.5 4.3 -146.3 10-15 6 2611.1 -17.4 -26.3 -132.7 15-20 8 6548.6 5.5 3.7 -75.9 20-25 8 10542.9 2.4 -1.3 -73.7 25-35 6 25639.0 19.8 10.2 -33.6 35-60 5 66038.0 -22.8 -7.8 -80.5 Table 3.23. Percent bias (calculated as mean residual divided by mean actual branchwood biomass, times 100) for 2 branchwood biomass models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by Installation. %Bias of model INST n Actual 3.26 3.27 RM ALL 39 16472.5 -6.8 -2.0 -69.9 2 7 14307.4 34.7 19.5 -13.7 4 6 12611.2 12.3 1.7 -50.7 5 8 5537.5 2.8 7.8 -79.8 16 6 6997.8 11.3 8.7 -62.9 71 6 6692.6 -9.0 -14.0 -96.6 72 6 56694.6 -26.4 -10.4 -87.2 58 150 5 15 25 35 45 55 65 DBH(cm) Figure 3.10. Branchwood biomass as predicted by Model 3.26. See text for more details. Figure 3.11. details. Branchwood biomass as predicted by Model 3.27. See text for more 60 3.4.6 Regression models for stem biomass components Separate regression models have been developed for each of the three stem biomass components: stemwood, stembark, and total stem (wood plus bark). The contribution of the competition indices to the regression models varied with the biomass component. In models for the prediction of stemwood or total stem biomass which use lndbh as independent variable, three of the four CIs contributed significantly, the exception being CSI (Table 3.24). Adding dbh to the models rendered all CIs except GSI insignificant. In stembark models, only CIO contributed significantly (Table 3.24). Coefficients of determination (R2) ranged from 0.976 to 0.989 for logarithmically transformed data. The regressions between actual values and those predicted from the significant models of Table 3.24 express the performance of the predictive models in non-logarithmic units (Table 3.25). For the prediction of stemwood biomass, model 3.37 yielded the highest R2 value, a low SEE, an intercept (b0) near zero, and a slope (bi) near 1.0. Of the two models which do not use a CI as an independent variable, model 3.36 is preferable because b0 is closer to zero and bi is closer to 1.0. With similar arguments, models 3.46 and 3.47 are identified as the best predictive models for total stem biomass. For the prediction of stembark, models 3.38 and 3.39 are selected as the best models with and without CI, respectively. The relationships between predicted values obtained from the regional models of Gholz et al. (1979) and actual values are very similar to those obtained with regression models from this study (Table 3.25). Two models each for the prediction of stemwood (models 3.36 and 3.37) and stembark (models 3.39 and 3.40) biomass were selected as the best models: InSW = -1.2850+1.7980*lndbh-0.0267*dbh [3.36] InSW = -1.4995+1.9524*lndbh+0.0202*dbh-0.0039*GSI [3.37] 61 InSB = -4.8505+2.5563*lndbh+0.0136*CIO [3.39] InSB = -3.4709+1.9925*lndbh+0.0185*dbh [3.40] where InSW = logarithm stemwood biomass (kg) and InSB = logarithm stembark biomass (kg). Additional statistics are listed in Tables 3.24 and 3.25. These four models and the regional models of Gholz et al. (1979) were further analyzed for their bias (see Equation 3.24). Percent bias ranged from -1.7 to -0.6% and from -2.4 to 2.3% for the stemwood and stembark regression models, respectively (Table 3.26). Percent bias in separate diameter classes was somewhat larger for the models from this study but was considerably larger in the regional model. Stratifying the data set by Installation showed that the bias of the stemwood models was small for five Installations (-3.1 to 7.4%), but in Installation 2 it was -25.2 and -15.7% for models 3.36 and 3.37, respectively (Table 3.27). Percent bias of the stembark models for all data combined was small (-2.4 to 0.8%) (Table 3.27). For the six Installations, percent bias ranged from -11.6 to 7.0% and was largest in Installation 5. The regional model had a small bias when averaged over all data, but it had a much larger bias in predicting stemwood biomass (-38.6% in Installation 2) and stembark biomass (-18.3% in Installation 5) of individual installations. 62 Table 3.24. Statistics of regression models to predict stem biomass components (kg) for Douglas-fir (n=39), CI = Competition Index. R2 is based on log-transformed data and SEE is in logarithmic units. Model R2 Significance (p) SEE Const, lndbh dbh CI No. STEMWOOD BIOMASS C+lndbh 0.978 C+lndbh+GSI 0.984 C+lndbh+CSI 0.980 C+lndbh+CIO 0.987 C+lndbh+DCI 0.982 C+lndbh+dbh 0.985 C+lndbh+dbh+GSI 0.987 0.2028 <.001 <.001 0.1746 0.1951 0.1588 0.1864 0.1703 0.1568 .001 .001 ,001 ,001 ,001 ,001 <.001 <.001 <.001 <.001 <.001 <.001 <.001 .004 <.001 .055 <.001 .008 .010 [3.31] [3.32] [3.33] [3.34] [3.35] [3.36] [3.37] STEMBARK BIOMASS C+lndbh C+lndbh+CIO C+lndbh+dbh 0.976 0.979 0.979 0.2109 0.2014 0.1983 001 001 001 <.001 <.001 <.001 .021 .039 [3.38] [3.39] [3.40] TOTAL STEM BIOMASS C+lndbh 0.981 0.1887 <.001 <.001 C+lndbh+GSI 0.985 0.1661 <.001 <.001 C+lndbh+CSI 0.982 0.1849 <.001 <.001 C+lndbh+CIO 0.989 0.1464 <.001 <.001 C+lndbh+DCI 0.984 0.1763 <.001 <.001 C+lndbh+dbh 0.987 0.1557 <.001 <.001 C+lndbh+dbh+GSI 0.989 0.1464 <.001 <.001 <.001 .002 .002 .120 <.001 .016 .022 [3.41] [3.42] [3.43] [3.44] [3.45] [3.46] [3.47] 63 Table 3.25. Seven models to predict stemwood biomass for Douglas-fir. The significant models from Table 3.24 and a regional model predicted stemwood biomass and linear regression of the form ACTUAL = bp+bj*PREDICTED are calculated. R2 is based on non-transformed data and SEE is in actual (non-logarithmic) units. Model R2 SEE bft bi No. STEMWOOD BIOMASS C+lndbh 0.981 67.0 -30.6 1.216 [3.31] C+lndbh+GSI 0.982 66.1 -20.4 1.139 [3.32] C+lndbh+CSI 0.985 59.9 -27.8 1.190 [3.33] C+lndbh+CIO 0.984 62.7 -10.8 1.065 [3.34] C+lndbh+DCI 0.985 59.7 -18.4 1.129 [3.35] C+lndbh+dbh 0.979 70.9 8.2 0.961 [3.36] C+lndbh+dbh+GSI 0.985 59.9 3.8 0.980 [3.37] Gholz et al. 1979 0.983 63.7 -10.5 1.021 [ ] STEMBARK BIOMASS C+lndbh 0.983 9.9 -2.8 1.113 [3.38] C+lndbh+CIO 0.980 10.6 -1.0 1.031 [3.39] C+lndbh+dbh 0.983 9.8 1.4 0.946 [3.40] Gholz et al. 1979 0.983 9.9 -2.7 1.086 [ ] TOTAL STEM BIOMASS C+lndbh 0.984 70.9 -33.9 1.205 [3.41] C+lndbh+GSI 0.984 72.5 -23.3 1.138 [3.42] C+lndbh+CIO 0.986 67.9 -12.0 1.062 [3.43] C+lndbh+DCI 0.987 63.9 -21.6 1.130 [3.45] C+lndbh+dbh 0.983 74.8 9.6 0.960 [3.46] C+lndbh+dbh+GSI 0.987 65.6 5.5 0.975 [3.47] 64 Table 3.26. Percent bias (calculated as mean residual divided by mean actual biomass, times 100) for stemwood and stembark biomass as calculated with 2 models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by diameter class. Stemwood Stembark DBH n Act. 3.36 3.37 RM Act. 3.40 3.39 RM (cm) (kg) % Bias (kg) % Bias 5-65 39 278.9 -1.0 -0.6 -1.7 44.5 0.8 -2.4 2.3 5-10 6 18.0 6.4 2.9 29.2 2.6 -5.1 -3.6 3.9 10-15 6 37.9 -7.3 -3.4 -0.5 6.8 1.6 -0.5 -3.0 15-20 8 80.1 1.8 1.0 -3.1 13.1 -3.4 -2.8 -10.3 20-25 8 128.3 -1.4 -1.5 -12.7 22.0 -3.4 -1.7 -10.7 25-35 6 268.4 -1.4 0.9 -16.0 47.9 -1.9 2.9 -4.2 35-60 5 1453.1 -1.0 -1.0 2.7 222.2 2.6 -3.9 7.4 65 Table 3.27. Percent bias (calculated as mean residual divided by mean actual biomass, times 100) for stemwood and stembark biomass as calculated from 2 models from this study and one regional model (RM, Gholz et al. 1979) for the combined data set and stratified by Installation. Stemwood Stembark Inst n Act. 3.36 3.37 RM Act. 3.40 3.39 RM (kg) %Bias (kg) %Bias ALL 39 278.9 -1.0 -0.6 -1.7 44.5 0.8 -2.4 2.3 2 7 97.3 -25.2 -15.7 -38.6 21.4 1.0 2.6 -4.4 4 6 145.8 0.0 2.7 -11.3 26.1 1.9 4.7 -1.8 5 8 70.4 1.6 -1.8 -3.2 10.6 -11.6 -11.0 -18.3 16 6 76.9 -3.1 -3.1 -9.6 12.5 -8.5 -8.5 -16.4 71 6 100.0 6.6 7.4 -1.2 16.3 0.7 1.7 -5.8 72 6 1283.0 0.4 -0.1 3.2 195.3 2.1 -3.3 7.1 66 3.5 DISCUSSION In the 45 years since Kittredge (1944) published the first paper on foliage biomass regression equations, over 30 additional models for the prediction of foliage biomass of Douglas-fir have been published (cf. Table 3.1). Collectively, they have identified a great variability in the relationships between dbh and foliage biomass. Some researchers found that sapwood basal area is a superior predictor of foliage biomass, but the relationship described by this variable appears to be affected by additional factors (Figure 3.2). Sapwood basal area can only be determined from increment cores or through destructive sampling. These measurements are often not available, however, especially in permanent sampling plots where the swelling of the stem following increment core sampling may not be desirable. In some projects, sapwood basal area has been predicted from a second regression equation established from sampling trees. This second model, however, represents an additional source of error. Despite the large number of existing regression models, a researcher who wants to predict foliage biomass is facing a difficult task. Given the existing variability between regression models, what criteria should be used to select and judge a model? Existing models for the prediction of foliage biomass typically use one or more variables which describe stem characteristics, such as dbh, height, or sapwood basal area. The size of the crown and the amount of foliage of a tree are also influenced by the amount of competition a tree is experiencing. The \"social\" position within a stand can be quantified by a competition index (CI). Four competition indices have been tested in this study for their contribution to regression models for the prediction of foliage and branchwood biomass of Douglas-67 fir. Each of these competition indices contributed highly significantly to the regression equations (Table 3.13 and 3.20). The regression coefficients associated with the competition indices in foliage and branchwood regression models have values that are ecologically meaningful: the coefficients are positive in models which include GSI, a CI that increases from 0 to 100 with decreasing competition. The coefficients are negative in models that use the other three CIs, all of which decrease with less competition. In both cases, the models predict that trees with less competition will have more foliage and branchwood biomass. Competition indices will account for the between and within stand variability in crown biomass allometric relationships that is attributable to variations in stand density. Competition indices can be computed from stand maps which are often available for research plots. Furthermore, repeated measurements of dbh and recording of tree mortality are sufficient to compute the changes in CI over time. Other variables occasionally used in regression models, such as height, sapwood basal area, or length of the live crown, are more difficult and more expensive to measure. In addition, these measurements would have to be taken repeatedly, if a prediction of biomass trends over time is desired. Competition indices by themselves may not be able to account for the changing allometric relationships following fertilization (Grier et al. 1986, Brix and Mitchell 1983, Barclay et al. 1986). In stands that have recently been thinned, a CI will express competition based on present stand density but tree crowns and foliage biomass may not have had the time to fully occupy the newly available space. This must be considered when using these equations to predict foliage and branchwood biomass. 68 Foliage biomass regression models developed in this study were applied to predict foliage biomass of 76 sample trees in 4 treatments at the Shawnigan Lake research project. The 3 models (3.19, 3.20, and 3.21) accounted for 83.9%, 82.4%, and 85.2% of the variation in this independent data set. Model 3.21 accounted for a greater proportion of the variation in the data than either the regional model (Gholz et al. 1979) or the model developed from the combined data of the Shawnigan Lake study (Brix and Mitchell 1983). All three models developed in this study consistently underpredicted foliage biomass. The two models which include a CI underpredicted foliage biomass of the trees in the unthinned and unfertilized control plots with 12.1% (Model 3.20) and 5.3% bias (Model 3.21). The model without CI had 34.5% bias (Model 3.19). Bias increased with thinning and fertilization treatments. In the Shawnigan Lake study, trees were destructively sampled 5 and 7 years after a thinning in which 2/3 of the basal area were removed. The remaining trees experienced much less competition than any of the sample trees from which the regression models were developed. Predicted foliage biomass of trees from thinned plots is in some cases based on an extrapolation of the model, which may account for some of the observed bias. The contribution of competition indices to regression models which predict stemwood and stembark biomass was smaller than their contribution to crown biomass component models (Table 3.24). The signs of the regression coefficients associated with the CIs were reversed compared to those in crown biomass equations. The models predict that of two trees with the same dbh, the one that experiences less competition will have the smaller stemwood and stembark biomass. Wood specific density in Douglas-fir is primarily a function of cambium age and not of ring width (Jozsa et al. 1989), This suggests that the form factor 69 (Husch et al. 1982) decreases with decreasing competition, as can be observed in Douglas-fir (R.E. Carter, pers. comm.). Bias in predicting stemwood biomass for the entire data set was -1.0% and -0.6% for models 3.36 and 3.37, respectively (Table 3.27). Bias increased when the data set was stratified by Installation. The greatest bias, -25.2% (Model 3.36) and -15.7% (Model 3.37) was observed in Installation 2. Specific gravity of the stemwood of the trees in this Installation (SG=0.403, S.D.=0.024, n=7) was significantly (p=0.005) lower than that of the trees in the other five Installations (SG=0.440, S.D.=0.032, n=32). Therefore stemwood biomass in Installation 2 is overestimated somewhat by the regression models. This study showed that competition indices can contribute significantly to biomass regression models. Sample trees for this study originated from 6 sites that covered a range of site indices. Further testing of the regression models should be conducted with additional independent data sets. The 39 sample trees of this study cover a large range of the possible combinations of dbh and competition index, but open-grown trees (no competition) and large trees with little competition are not included in the data set. Future studies should attempt to include such sample trees. 70 3.6 CONCLUSIONS The review of existing regression models for the prediction of biomass of Douglas-fir has shown that the range of predicted biomass for the same diameter differs widely between models. The lack of criteria by which to judge the suitability of existing models for the prediction of biomass components in specific stand conditions emphasizes the need to find a new approach to biomass regression models. This new approach should attempt to include an independent variable which might account for the between-stand differences in the biomass regression models. Competition indices, which account for the competitive status of individual trees, contributed significantly to regression models for the prediction of all above-ground biomass components. Their contribution was highest in foliage and branchwood biomass regression models. Of the competition indices tested, GSI (Lin 1974) contributed most to regression models for the prediction of foliage, branch, and stemwood biomass. The new models for the prediction of foliage biomass were tested against an independent data set from the Shawnigan Lake research project and performed very satisfactorily, in particular in unthinned and unfertilized plots. Thinning and fertilization increased the bias of the predicted values. 71 4. ABOVEGROUND AND COARSE ROOT BIOMASS AND PRODUCTION 4.1 INTRODUCTION A quantitative understanding of the production of all major above- and belowground biomass components is a necessary foundation for many studies of forest ecosystems: carbon and nutrient budgets, the biology of tree response to silvicultural treatments, and the study of environmental influences on forest yield are some important examples. In this chapter, biomass and production of those tree components which can be predicted from allometric relationships will be reported. This includes the major aboveground components and coarse roots. Fine and small root biomass and production will be reported in the next chapter. The first objective of the research presented in this chapter was to quantify the biomass of major tree components in twelve Douglas-fir stands growing over a range of site indices. The second objective was to quantify the net biomass production of these components over several measurement periods. 4.2 LITERATURE REVIEW Net primary production (NPP) of a forest ecosystem is defined as the amount of organic matter produced by plants over a time period, usually a year (Waring and Schlesinger 1985, Satoo and Madgwick 1982). NPP includes all increments in the biomass of stemwood, stembark, branches, foliage, reproductive organs, and roots, plus the amounts of plant material that become detritus or are consumed by animals. 72 NPP can be estimated by determining, for each of the major biomass components, the net increment (A Biomass), the mortality of individual trees or their parts (detritus), and the amount of plant material consumed by animals (consumption) over the time period. NPP = A Biomass + Detritus + Consumption [4.1] For general reviews of forest ecosystem biomass, production, and litterfall data see Cannell (1982), Reichle (1981), Bray and Gorham (1964), and Vogt et al. (1986). 4.2.1 Annual stemwood and stembark production Stemwood and stembark are the most commonly measured components in studies of biomass production and growth and yield (cf. Cannell 1982), most of which consider only net increment and mortality of trees. The turnover component of stembark production, due to shedding of the outermost bark layers, is generally not considered in production studies. 4.2.2 Annual branchwood production Annual branchwood production comprises the net increment of total branchwood biomass and the replacement of mortality. Few studies have attempted to include annual branchwood production in developing estimates of production at the stand level. 73 Satoo and Madgwick (1982) suggest two methods for deriving an estimate of branchwood production. The first method calculates the change in total branchwood biomass, based on regression equations, and adds a value for branchwood mortality derived from litterfall estimates. This method has several problems. Firstly, branch litterfall is likely to vary greatly between years depending on the occurrence of storms, snowfall, drought, and other factors. Secondly, large traps are required to obtain accurate estimates because of the spatial heterogeneity of branch litterfall. Thirdly, the origin of branchwood litterfall cannot be identified: some of it originates from dead trees, the rest from dead whorls of living trees. Only the latter component is of interest in the estimation of branchwood turnover, because branchwood mass of dead trees is not included when estimating stand branchwood biomass from regression equations. The second method suggested by Satoo and Madgwick (1982) is based on the assumption that the relative growth rate of branches is equal to the relative growth rate of stemwood. The authors warn, however, that this method may lead to underestimates of actual production. The study of Douglas-fir biomass production by Dice (1970) omitted branchwood from the total aboveground production estimates. Mohren (1987) calculated annual branchwood mortality by assuming an average lifespan of 30 years for a Douglas-fir branch. However, stand density will affect live crown length (Carter et al. 1986, Ritchie and Hanh 1987) and such an approach may not be appropriate in stands with differing density. Comeau (1986) calculated branchwood production of lodgepole pine stands by dividing tree branchwood biomass by the age of the oldest branches on the tree. This may result in the underestimation of branchwood production because the average lifespan, not the maximum lifespan, determines turnover rates. 74 4.2.3 Annual foliage production There are several different approaches to the quantification of annual foliage production in evergreen coniferous stands (Satoo and Madgwick 1982, Newbould 1967). The amount of first year foliage, which is equal to foliage production, can be calculated from regression equations. Alternatively, annual foliage litterfall can be collected and, assuming that the foliage biomass of a stand has reached a steady state, annual foliage litterfall will be equal to annual foliage production (Fogel and Hunt 1979). Between-year variation in annual foliage litterfall and the litterfall component from dying trees can complicate calculations based on this approach. 4.3 MATERIALS AND METHODS 4.3.1 Site description The study sites have been described in detail in Chapter 2. 4.3.2 Field measurements and data processing Details of plot selection and establishment are described in Darling and Omule (1989). The main points relevant to this study will be summarized briefly. Each plot was 22.36 x 22.36 m (0.05 ha) and was surrounded by a 4.63 m buffer strip (0.05 ha). All trees within each plot were marked at breast height (1.3m), numbered, and their x-y coordinates determined. Diameter at breast-height (dbh) was measured to the nearest 0.1 cm for all trees greater than or equal to 5.0 cm dbh. Installations 71 and 72 were measured in the fall or winter of 1971, 1974, 75 1977, 1980, and 1983. The remaining four Installations were measured in 1972, 1975,1978, 1981, and 1984. These dbh measurements were taken by the B.C. Forest Service Research Branch. Additional dbh measurements for this study were taken for all plots in the fall or winter of 1985 and 1987. The two data sets were merged and checked for inconsistencies. Obvious errors, such as trees classified as dead in one period and living in the next, were corrected. \"Shrinking\" of trees was occasionally observed if a suppressed tree was approaching death and then died in the next measurement period. \"Shrinkage\" was presumably caused by moisture loss in the stemwood and should not affect stemwood biomass on a dry weight basis. In the few situations where shrinkage was observed and followed by tree death in the subsequent measurement period, the dbh of such trees was held constant at its preceding maximum value. Such corrections amounted to increasing the dbh at the last measurement prior to mortality by one to two mm and avoided the calculation of decreasing biomass estimates. The merged and corrected data file was the basis for all subsequent computations. At each measurement period, height measurements were taken on approximately 10 trees per plot. In 1985, the height of the five largest (by dbh) trees per plot was determined with a tripod-mounted relascope. Ages were determined by the B.C. Forest Service Research Branch from cores collected 30 cm above the germination point in at least 5 trees per plot. Site indices (Bruce 1981, Mitchell and Poison 1988) for 1985 were calculated from total stand age (converted to age at breast height) and the mean height of the 5 largest trees per plot. Site indices for years prior to 1985 were calculated from total stand age (converted to age at breast height) and the mean height of the 5 largest trees for which height measurements were available. Site index calculations were based on height/age equations (Mitchell and Poison 1988) which 76 were iteratively solved for site index using the secant method (Gerald and Wheatley 1984). Competition indices were calculated using the equations described in Chapter 2. Growing Space Index (GSI) (Lin 1974) and Competitive Stress Index (CSI) (Arney 1973) were computed for each tree in the plot at each measurement date using dbh data and x-y coordinates. These data were unavailable for trees in the buffer strip. Competition indices for trees near the edge of the plot were erroneous because of the lack of neighbours for these trees in the data set. The contour map of GSI values of Installation 2 Plot 6 (Figure 4.1A) shows the edge effect as increasing GSI values near the plot edge. This artefact was reduced by establishing a hypothetical stand around each plot. The plots were subdivided into 16 quadrats of equal size (5.59 x 5.59m), labelled A through P in Figure 4.2. The stand information of these quadrats was copied to similar quadrats surrounding the plot (labelled a through p in Figure 4.2) following the scheme outlined in Figure 4.2. Competition indices were calculated for trees inside the original plot, assuming the hypothetical stand structure in the area surrounding the plot. Figure 4. IB shows that this procedure effectively removed the edge effect observed in Figure 4.1A. Figure 4.1. Contour maps of Growing Space Index (GSI) in Installation 2, Plot 6, without (A) and with (B) hypothetical buffer strip. Note the increase in GSI (less competition) towards the plot border in A. Tree locations are indicated by •. 78 k d h m =*= ==^= n —1— 0 p rH J a e 1 P = f = =1= i m g a b c d f Figure 4.2. The scheme used to establish a hypothetical stand around each plot. Sixteen quadrats (A to P) of the actual plot (shaded area) were copied to locations surrounding the plot (a to p). 79 4.3.3 Calculation of biomass and net production The tree components for which biomass and net primary production (NPP) were determined individually are stemwood, stembark, branches (including bark), foliage, and coarse roots. Biomass and production of reproductive organs were not quantified. The biomass of each major aboveground tree component and of coarse roots was calculated from regression equations applied to the data of every living tree with at least 5 cm dbh. All living trees were summed to obtain plot biomass which was converted to Mg ha\"1. The regressions for aboveground Douglas-fir biomass components were described in Chapter 3. Biomass components of western hemlock and western redcedar were calculated from published regressions (Gholz et al. 1979). Other species (western white pine, (Pinus monticola Dougl.), bigleaf maple {Acer macrophyllum Pursh.), and Hooker's willow (Salix hookeriana Barr. in Hook.)) were treated as Douglas-fir. White pine and bigleaf maple never occurred more than once per plot. In one plot, three willows were present, but they never accounted for more than 0.6% of total basal area. Net primary production for any period was calculated as the sum of the net increment in biomass, plus mortality of trees, plus turnover: Production = A Biomass + Mortality + Turnover [4.2] Consumption of plant biomass by animals was not considered in this study as no data were available. For the period for which diameter measurements were taken, no outbreaks of defoliating insects had been recorded in the six installations. Net annual increment for each of the biomass components was calculated as the difference between stand biomass at the beginning and end of each measurement period divided by the number of years in that period. The calculation of the mortality and turnover components of NPP differed between biomass components as described below. The exact year of tree mortality within each 80 measurement period was not determined. The biomass of all trees which died within a period was summed and evenly distributed among all years of the measurement period. Ingrowth, (i.e. trees which were recorded for the first time because they grew bigger than the diameter-recording limit of 5 cm), was treated as if the biomass of such trees had been produced entirely during the measurement period. Biomass production was evenly distributed among the years of that period. 4.3.3.1 Stemwood The biomass regression equation with dbh and growing space index (GSI) as independent variables (Equation 3.37) was applied to compute stemwood biomass of each living Douglas-fir tree at each measurement date. Net increment was calculated from the difference in plot stemwood biomass at successive dates. Mortality was calculated as described above. 4.3.3.2 Stembark Stembark production was computed in the same way as stemwood production. Douglas-fir biomass was predicted from dbh using regression equation 3.40. Mortality of individual trees was treated the same as that of stemwood production. Stembark turnover was not considered because no data were available. 81 4.3.3.3 Branchwood Annual branchwood production (Mg ha-1 year-1) in a stand can be partitioned into A biomass, mortality, and turnover. The change in total branchwood biomass (A biomass) is calculated as the difference in total live branchwood biomass at the beginning and end of a measurement period. Mortality refers to the branchwood biomass of trees which died during the measurement period. Turnover refers to death of branches at the bottom of the canopy and addition of new whorls on the top. At each measurement date, biomass is calculated from Equation 3.27 (Chapter 3). Mortality of branchwood, due to the death of individual trees, is treated in the same way as stemwood mortality (described above). Turnover of branchwood biomass is difficult to determine due to the large temporal and spatial variability in the occurrence of branchwood litterfall. Data on branchwood litterfall were not available for the 12 study plots. The following computational approach was developed to calculate the turnover component of branchwood production. Under normal growing conditions, Douglas-fir produces one whorl of branches annually Gammas growth involving the terminal bud occurs only rarely in Douglas-fir, R.E. Carter, pers. comm.). Whorl number is therefore equivalent to the age of the whorl (i.e. growing seasons of production), if whorl number is counted from the top down during the dormant season. In a closed canopy stand in which the canopy has started to lift off the ground, one whorl will die approximately every year, thus maintaining, within a tree, a number of whorls which is relatively constant within a measurement period. Annual branchwood production in Douglas-fir can be approximated from the following considerations. Suppose that annual branchwood production (BWP) within a whorl j is a function f of whorl number j. BWP(j) = flj) [4.3] 82 Suppose, further, that annual branchwood litterfall (BWL) within a whorl a function g of whorl number j. BWL(j) = g(j) [4.4] Branchwood biomass (BWB) of a whorl j can then be calculated as j BWB(j) = . X ( BWP(i) - BWL(i)) [4.5] i=l because whorl number and age of the whorl are approximately equivalent in Douglas-fir. 83 Jm - 10 Production, Litterfall (kg year\"') Production, Litterfall (kg year\"1) Figure 4.3. Theoretical distribution of branch biomass ( • ) by whorl number. At each whorl, annual branch production (hatched bars), and branch litterfall (cross-hatched bars) are shown, but at a different scale than biomass. A: Production occurrs in whorls 1 -10 and litterfall in whorls 11-15. B: Production occurrs in whorls 1 - 13 and litterfall in whorls 8 - 15. See text for further explanation. 84 4) E o JZ Percent of total branch biomass Figure 4.4. Branch biomass distribution of 39 destructively sampled trees from 6 plots. For each whorl, branch biomass (open circles) is expressed in percent of total plot branch biomass. The distribution is approximated by a distance weighted least square algorithm (solid line). 85 Tree branchwood production (TBWP) is denned as the sum of branchwood production of the n wood-producing whorls of the tree: n TBWP = X BWP(i) [4.6] i=l The distribution of branchwood biomass among whorls can be described by a bell-shaped curve (Figure 4.3a). One whorl in the lower section of the tree's crown will carry the maximum biomass, e.g. whorl 10 in Figure 4.3a. The biomass of this whorl (jm) can be defined from equation 4.5 as: Jm Jm BWB(jm)= X BWP(i)- XBWL(i) [4.7] i=l i=l If we assume (discussed below) that in a single whorl and measurement period either production or litterfall occur, but not both, and that production occurs in whorls 1 to whorl j m and litterfall in all whorls from j m + 1 to n (Figure 4.3b) then we can demonstrate that the branchwood biomass of whorl j m is equal to the total branchwood biomass production of the tree. In equation 4.7, we can substitute the first summation with equation 4.6 and the second summation with zero and rewrite equation 4.7 as: BWB(jm) = TBWP [4.8] . This equation states that the maximum amount of branchwood in a single whorl can be used as an estimate of annual branchwood production. This relationship holds if the branchwood biomass of a single tree is at or near steady state, which is true if annual branchwood production and litterfall at each whorl number remain constant, i.e. Equations 4.3 and 4.4 do not change over time. This would be the case if competition from neighbouring trees prevents further increase of the individual tree's branchwood biomass. In Chapter 3 (Equation 3.27) it was shown that competition influences the branchwood biomass of a tree. Any increase 86 in total branchwood biomass of a tree during a measurement period is accounted for by the A biomass component of equation 4.2. Violating the assumption that in a single whorl and measurement period either branchwood production or litterfall, but not both, occur will result in an underestimate of tree branchwood production (TBWP) in equation 4.8. Figure 4.3A gives an example of the assumption that production occurs only in whorls 1 through j m and litterfall only in whorls j m + i to the maximum number of live whorls. Figure 4.3B shows the simultaneous occurrence of branchwood production and litterfall in whorls 8 through 13. Data which could be used to assess how much overlap occurs between branchwood production and litterfall in particular whorls are not available. The site specific biomass regressions described in section 3.4.1 were applied to calculate branchwood biomass for each live branch in each of the 39 sample trees. Total branchwood biomass per whorl and per tree was calculated from these data. Within each sample plot, the mean branchwood biomass at each whorl number was computed by summing the whorl biomass of the sample trees. The distribution of live branchwood biomass was expressed as the percentage of the total of the destructively sampled trees in each plot. Calculating plot averages rather than a mean derived from individual trees reduces the influence of intermediate and suppressed trees, which have smaller than average branchwood biomass and which may have unrepresentative crown shapes. For the sample trees from each installation, the maximum percentage of total branchwood biomass present in a single whorl was determined. The turnover component of total annual branchwood biomass production was calculated by multiplying total branchwood biomass per tree at the beginning of the measurement interval by the maximum percentage value described above. 87 Within each plot all species were treated similarly. As no data are available, it was assumed that the turnover component of other species' annual branchwood production can be calculated in the same way as that of Douglas-fir. The mortality and A biomass component of western hemlock and western redcedar branch production were calculated from species specific regression equations (Gholz et al. 1979), as described for stemwood biomass. 4.3.3.4 Foliage Annual foliage production was computed similar to branch production from the net change in total biomass plus the replacement of mortality due to tree death, plus the amount of foliage which is replaced annually. The biomass regression equation based on dbh and growing space index (GSI) (Equation 3.20) was applied to calculate total foliage biomass at each measurement date. The loss of foliage biomass due to tree mortality during a measurement period was evenly distributed among the years of that period. Annual turnover of foliage was assumed to be equal to the biomass of one-year-old foliage. To derive an estimate of the proportion of total foliage in the first year age class, the percentage of foliage in the first year was calculated for each of the 267 sample branches which had been collected and analyzed (cf. Section 3.3.2). For each of the 6 Installations, a regression equation was developed which predicts the proportion of each whorl's foliage which is in the first age class. Using these regression equations and the data for all live branches of the 39 sampling trees, the amount and proportion of foliage in the first year age class was computed for each branch and summed to obtain the totals for each tree. A 88 regression equation was developed from data from the 39 sample trees to predict the total amount of first year foliage as a function of dbh. 4.3.3.5 Coarse roots Coarse root biomass for all species was calculated from a regression equation for Douglas-fir (Gholz et al. 1979, Dice 1970) which uses dbh as independent variable. The lower diameter limit for coarse roots used by Dice (1970) was 10 mm. No biomass regressions are available for roots 5 to 10 mm diameter. Mortality of individual trees was treated as described for stemwood. No attempt was made to quantify the turnover component of coarse root production. McMinn (1963) encountered no dead structural roots greater than 1 cm in diameter, except those of dead suppressed trees, when excavating root systems of Douglas-fir. 89 4.4 RESULTS 4.4.1 Branch biomass turnover As described in Section 4.3.3.3, branch biomass production is derived from three estimates: the net change in branch biomass per hectare, the replacement of mortality due to death of trees, and the replacement of mortality due to death of whorls. The latter estimate, branch turnover, is assumed to be equal to the maximum percentage of total biomass encountered in a single whorl (see Section 4.3.3.3 for a derivation). The percentage of total branch biomass plotted against whorl number is presented in Figure 4.4. To show the approximate distribution of branch biomass, lines based on a distance weighted least square smoothing algorithm (Wilkinson 1988a) have been added. The whorl number at which the largest proportion of total branch biomass was encountered ranged from whorl 12 in Installation 16 to whorl 19 in Installation 72 (Table 4.1). The maximum percentage of total branch biomass in a whorl ranged from 6.0 to 12.9%, which is equivalent to a mean lifespan of 16.7 to 7.8 years, respectively (Table 4.1). Total branch biomass at the beginning of each measurement period is multiplied by the maximum percentage values (Table 4.1) to estimate the branch turnover component of branch production. 90 Table 4.1. Maximum percentage of total branchwood biomass encountered in a single whorl (whorl number) for each of the six destructively sampled plots. Installation whorl percent of mean lifespan ' number total (years) 2 13 6.0 16.7 4 16 6.6 15.2 5 12 9.9 10.1 16 12 12.9 7.8 71 13 8.9 11.2 72 19 6.4 15.6 91 4.4.2 Foliage biomass turnover The percentage of foliage on a branch which is in the first year age class declines with whorl number, if whorl number is counted from the top down. Figure 4.5 shows percent foliage biomass in the first year age class as a function of whorl number for 267 sample branches. The model which best described the data was PERFOL = b0 + bi * 1AVHORL + b2 * l/CWHORL)2 [4.9] where PERFOL is the percentage of a whorl's foliage which is in the first year age class, and whorl is the number of the whorl, counting from the top down. Regression lines (Table 4.2), fitted to the data from each Installation, are shown in Figure 4.5. After applying the regression models described in section 3.4.1 to calculate foliage biomass per branch, and the regression equations listed in Table 4.2 to calculate the proportion of first year foliage, the results were summed for each tree. First year foliage biomass per tree (FOLD was calculated from the model InFOLl = -1.598 + 3.125 * In dbh -0.0514* dbh [4.10] where InFOLl is the natural logarithm of first year foliage biomass per tree (gram) and dbh is diameter at breast height (cm). Based on a sample size of 39 trees, the regression model is highly significant (R2=0.868, p < 0.001, SEE=0.432 In gram). 92 Table 4.2. Coefficients, sample size (n), R2, and standard error of estimate (SEE) of six equations to calculate, for individual branches, the percentage of foliage in the first age class. The general model is described in equation 4.9. (p < 0.001 for all models and for all coefficients (except see footnote). Inst n R2 SEE(%) b0 b: b2 2 54 .977 5.16 -9.91 213.23 -103.20 4 42 .982 4.79 -13.44 267.87 -154.37 5 52 .978 5.24 -11.52 245.95 -134.40 16 34 .987 4.49 -23.19 344.11 -220.84 71 39 .934 8.86 -5.77a 241.14 -135.41 72 46 .965 6.00 -3.42b 257.59 -154.34 a p=0.078 b p=0.068 93 1 6 11 16 21 26 31 1 6 11 16 21 26 31 Whorl number Figure 4.5. Percentage of foliage in the first year age class for 267 sample branches from 6 Installations (open circles). Regression equations (Table 4.2) are plotted as solid lines. 94 4.4.3 Aboveground and coarse root biomass and production The results for the major ecosystem variables will be presented graphically for all 12 plots. In addition, results for either the year 1985 or the period 1985 to 1987 will be presented in tabular format. Stand age, basal area, stand density, and site index for 1985 are summarized in Table 4.3. Stand age ranged from 32 to 70 years. Site index ranged from 19.5 to 41.3 meters at 50 years. Stand density in 1985 varied between 440 and 3400 stems per hectare (Table 4.3) and generally decreased over time (Figure 4.6). The only exception is Installation 2 Plot 11 where the number of both Douglas-fir (+8.6%) and western hemlock (+85.7%) stems with more than 5.0 cm dbh increased from 1972 to 1987. In 1985, basal area of all species ranged from 33.6 to 75.3 m2 ha-1 (Table 4.3) of which Douglas-fir represented 66.5 to 99.8%. Basal area increased with time in 11 of the 12 plots (Figure 4.7). Snowbreak in Installation 4 Plot 17 in the winter of 84/85 resulted in some tree mortality and a reduction in basal area in this plot. Aboveground biomass changed in a pattern which closely followed the change in basal area (Figure 4.8). In 1985, aboveground biomass ranged from 135.0 to 573.6 Mg ha-1 (Table 4.4). Table 4.4 lists the biomass of each major aboveground component and of coarse roots. The distribution of biomass as a percentage of aboveground biomass is presented in Table 4.5. Coarse root biomass is expressed as a percentage of aboveground biomass, but is not included in its calculation. In most plots, foliage biomass increased with time, but the rate of increase declined with time (Figure 4.9). In the two plots of Installation 72, however, foliage biomass was approximately constant (10 and 11 Mg ha\"1), despite the continuing 95 increases in basal area and total aboveground biomass. Installation 72 is both the oldest stand and the stand with the highest basal area. Mean annual biomass production for the period 1985 to 1987 ranged from 5.6 to 16.0 Mg ha-iyr1 (Figure 4.10, Table 4.6). The low production (4.7 Mg ha-iyr1) in Installation 4 Plot 17 is due to tree mortality and reduced tree growth following snowbreak. Stemwood production was the single largest component of total aboveground biomass production, representing 42.4 to 68.5% of the total (Table 4.7). Foliage, which represented 1.8 to 6.4% of total aboveground biomass, accounted for 8.7 to 31.0% of aboveground production. Branches and stembark accounted for an additional 12.4 to 26.7% and 7.1 to 10.6% of total aboveground production, respectively. Coarse root production was estimated to be equal to 13.2 to 17.3% of total aboveground production. 96 Table 4.3. Stand age, site index (SI), basal area (BA), and stand density in 1985 for the 12 study plots. Douglas-fir (Df) BA and stand density are listed in absolute amounts and as a percentage of the total. Inst Plot Age SI Total Df %Df Total Df %Df BA BA BA Stems Stems Stems (yr) (m@50) (m^ a^ Km^ Wty* of tot) (st. ha^ Kst. ha\"1)(% of tot) 2 6 42 27.7 53.4 35.5 66.5 3000 1200 40.0 2 11 41 19.5 33.6 26.9 80.1 2840 2040 71.8 4 1 44 29.1 45.3 45.2 99.8 1840 1800 97.8 4 17 48 25.7 37.4 35.4 94.7 1640 1420 86.6 5 8 40 26.8 58.7 44.6 76.0 3400 1820 53.5 5 10 39 29.5 47.4 45.9 96.8 1420 1240 87.3 16 2 32 29.4 41.3 40.7 98.5 2000 1960 98.0 16 6 32 32.4 45.1 40.7 90.2 2520 2020 80.2 71 11 41 24.6 45.3 45.1 99.6 1880 1840 97.9 71 14 41 23.3 46.0 45.6 99.1 2460 2400 97.6 72 2 70 41.3 69.1 67.0 97.0 440 400 90.9 72 14 70 41.0 75.3 72.5 96.3 480 460 95.8 97 CO CO E *•* CO co c 1) •u c CO CO 5000 4000 3000 2000 1000 1070 197S 1980 1985 1990 5000 4000 3000 2000 1000 -1970 1975 1980 1985 1990 5000 4000 3000 -2000 1000 1970 1975 1980 1985 1990 5000 4000 -3000 -2000 1000 -1970 1975 1980 1985 1990 5000 40O0 3000 2000 1000 1970 1975 1980 1985 1990 2500 2000 1500 1000 500 -«-1970 1975 1980 1085 1990 Year Figure 4.6. Total stand density plotted against time for the 12 sample plots. Installation numbers are in the upper right corner of each graph. Solid circles represent Installation (I) and Plot (P): I2-P6,14-P1,I5-P8,I16-P2,171-P11, and 172-P2. Open circles represent: I2-P11,14-P17,15-P10,116-P6,171-P14, and I72-P14. 98 CO CM CO CO CO CO 100 eo eo 40 20 1670 197S 1980 1865 1090 100 80 60 40 20 1970 197S 1980 1985 1990 100 80 60 40 20 100 80 60 40 20 1970 1975 1980 1985 1990 100 80 -80 -40 20 1970 1975 1980 1985 1990 100 80 60 40 20 1970 1975 1980 1985 1990 1970 1975 1980 1985 1990 Year Figure 4.7. Basal area plotted against time. Legend as in Figure 4.6. 99 Table 4.4. Total stand biomass in 1985 of foliage, branches, stemwood, stembark, and cOarse roots. Inst Plot Age SI Foliage Branches Stem Stem £ Above Coarse wood bark ground roots (yr) (m@50) (Mgha'1) 2 6 42 27.7 13.08 20.83 176.8 27.77 238.4 49.66 2 11 41 19.5 8.60 11.64 98.1 16.68 135.0 26.31 4 1 44 29.1 10.48 14.39 158.9 28.15 211.9 47.18 4 17 48 25.7 9.05 11.75 122.5 21.60 164.9 35.48 5 8 40 26.8 13.54 19.72 186.9 29.49 249.7 49.74 5 10 39 29.5 10.13 13.68 169.4 28.35 221.6 . 48.18 16 2 32 29.4 9.77 12.04- 127.9 22.47 172.2 35.13 16 6 32 32.4 11.49 14.34 142.3 24.11 192.2 39.52 71 11 41 24.6 10.35 12.93 145.1 25.35 193.7 40.34 71 14 41 23.3 10.00 12.55 146.4 25.17 194.1 38.96 72 2 70 41.3 9.87 26.47 441.5 71.96 549.8 123.00 72 14 70 41.0 10.50 25.32 463.1 74.61 573.6 129.70 100 Table 4.5. The distribution of the biomass components listed in Table 4.4, expressed as a percentage of aboveground biomass. Inst Plot Age SI Foliage Branches Stem Stem S Above Coarse wood bark ground roots (yr)(m@50) (% of S Aboveground) 2 6 42 27.7 5.5 8.7 74.2 11.6 100.0 20.8 2 11 41 19.5 6.4 8.6 72.6 12.4 100.0 19.5 4 1 44 29.1 4.9 6.8 75.0 13.3 100.0 22.3 4 17 48 25.7 5.5 7.1 74.3 13.1 100.0 21.5 5 8 40 26.8 5.4 7.9 74.8 11.8 100.0 19.9 5 10 39 29.5 4.6 6.2 76.4 12.8 100.0 21.7 16 2 32 29.4 5.7 7.0 74.3 13.0 100.0 20.4 16 6 32 32.4 6.0 7.5 74.0 12.5 100.0 20.6 71 11 41 24.6 5.3 6.7 74.9 13.1 100.0 20.8 71 14 41 23.3 5.2 6.5 75.4 13.0 100.0 20.1 72 2 70 41.3 1.8 4.8 80.3 13.1 100.0 22.4 72 14 70 41.0 1.8 4.4 80.7 13.0 100.0 22.6 Year Figure 4.9. Foliage biomass plotted against time. Legend as in Figure 4.6. 103 Table 4.6. Annual production of foliage, branches, stemwood, stembark, and coarse root biomass. Data represent the sum of all species for the period 1985 to 1987. Inst Plot Age SI Foliage Branches Stem Stem Z Above Coarse wood bark ground roots (yr) (m@50) (Mgha-iyr1) — 2 6 42 27.65 2.58 1.61 3.53 0.60 8.32 1.13 2 11 41 19.50 1.38 1.01 2.77 0.40 5.56 0.78 4 1 44 29.05 1.39 1.25 4.06 0.69 7.39 1.28 4 17 48 25.66 1.21 0.95 2.18 0.36 4.71 0.69 5 8 40 26.75 2.62 2.41 4.67 0.69 10.39 1.37 5 10 39 29.45 1.55 1.61 4.46 0.67 8.29 1.28 16 2 32 29.38 1.33 1.80 4.06 0.64 7.83 1.18 16 6 32 32.39 1.42 2.12 3.78 0.61 7.94 1.17 71 11 41 24.62 1.39 1.29 2.67 0.41 5.76 0.76 71 14 41 23.27 1.47 1.36 3.36 0.54 6.74 1.00 72 2 70 41.34 1.39 2.16 9.55 1.55 14.66 2.30 72 14 70 41.03 1.40 1.99 10.97 1.65 16.01 2.55 104 Table 4.7. The distribution of the production listed in Table 4.6, expressed as a percentage of aboveground biomass production. Data represent the sum of all species for the period 1985 to 1987. Inst Plot Age SI Foliage Branches Stem Stem Z Above Coarse wood bark ground roots (yr)(m@50) (% of Z Aboveground) 2 6 42 27.7 31.0 19.4 42.4 7.2 100.0 13.6 2 11 41 19.5 24.8 18.2 49.8 7.2 100.0 13.9 4 1 44 29.1 18.8 16.9 54.9 9.3 100.0 17.3 4 17 48 25.7 25.7 20.2 46.3 7.7 100.0 14.5 5 8 40 26.8 25.2 23.2 44.9 6.6 100.0 13.2 5 10 39 29.5 18.7 19.4 53.8 8.1 100.0 15.4 16 2 32 29.4 17.0 23.0 51.9 8.2 100.0 15.1 16 6 32 32.4 17.9 26.7 47.6 7.6 100.0 14.7 71 11 41 24.6 24.1 22.4 46.4 7.1 100.0 13.2 71 14 41 23.3 21.8 20.2 49.9 8.1 100.0 14.8 72 2 70 41.3 9.5 14.7 65.1 10.6 100.0 15.7 72 14 70 41.0 8.7 12.4 68.5 10.3 100.0 15.9 105 0 L 1 1 1 1 0 1 1 1 1 1 1970 1975 1980 1986 1990 1970 1975 1980 1985 1990 Year Figure 4.10. Aboveground production plotted against time. Legend as in Figure 4.6. 106 4.5 DISCUSSION The use of a competition index as an independent variable in biomass regression models makes it necessary that the locations of trees in research plots are known. The X-Y coordinates and the dbh and mortality data for up to 16 years were available for trees inside the research plots, but not for the trees in the surrounding buffer strips. Competition indices for trees near the plot edge are biased because of the lack of neighbouring trees in the data set. This problem can be reduced by either establishing a buffer strip inside the existing plot or by creating, based on available plot information, a hypothetical stand surrounding the plot. A buffer strip of adequate width inside the plot would have reduced the inner plot to one third to one quarter of its original size. An external buffer strip could have been created by \"mirroring\" the trees along the plot border. This method, however, amplifies existing irregular tree distribution patterns. A small gap or a large tree near the plot edge will be duplicated on the opposite side of the plot border. The method chosen in this study, whereby plot sections from one side of the plot were copied adjacent to the opposite side, overcomes this problem. Furthermore, it maximizes the use of available information because dbh data from all trees can be utilized. Although this method does not use \"actual\" X-Y coordinates or dbh data for the trees in the buffer strip, the error introduced by using these \"hypothetical\" stand data will probably be small because the plots have been established to include buffer strips of similar stand structure as in the plots themselves (Darling and Omule 1989). In this study, a competition index was used, in addition to dbh, as an independent variable in regression models for the prediction of foliage biomass. In the two plots with the highest basal area (Installation 72), total foliage biomass 107 reached a maximum of about 10 and 11 Mg ha-1 and decreased slightly thereafter (Figure 4.9), despite the continuing increase in basal area (Figure 4.4). In comparison, the dbh-based regression model of Gholz et al. (1979), which has frequently been used in biomass studies, predicts an increase in foliage biomass from 13.1 to 16.1 and from 15.5 to 16.6 Mg ha\"1 for the period 1971 to 1987 for plots 2 and 14 of Installation 72, respectively. The stabilization of foliage biomass predicted by the regression model developed in this study is consistent with the observation that, during stand development, foliage biomass levels off at a site specific maximum value (Tadaki 1966, Albrektson 1980). A similar prediction could be obtained from regression models that use sapwood basal area as independent variable. As is often the case, sapwood basal area data were not available in this study because coring of the trees in the long-term sample plots was not permissable. Using a competition index in combination with dbh data is a suitable alternative approach to predicting stand foliage biomass. Biomass of western hemlock and western redcedar, where present, was estimated from published regression equations (Gholz et al. 1979). Some of the problems of using regional rather than site-specific regression models have been discussed in Chapter 3, but site-specific models for these species were not available. In 1985, Douglas-fir represented at least 90% of total basal area in 9 of the 12 plots, and never less than 66.5% (Table 4.3). The error introduced by using regional models for a small component of the trees in a stand is probably small. For other tree species, which occurred only rarely in some of the plots, biomass components were predicted from Douglas-fir regression equations. The single bigleaf maple in Installation 72, Plot 14, accounted for 4.0% of total basal area in 1974, the year in which it reached the maximum proportion of total basal area. Similarly in 1987, a single white pine in Installation 4, Plot 17, accounted for 108 2.1% of total basal area. Installation 5, Plot 8, initially contained three small willows which, in 1972, accounted for 0.6% of total basal area. On a plot basis, the errors associated with using Douglas-fir equations for these few cases are probably minor. A comparison of biomass and production values between different studies is always complicated by differences in the methodology employed to derive the estimates. Cannell (1982) summarized many biomass studies for Douglas-fir stands. The total aboveground biomass data for non-fertilized, non old-growth (<150 years) Douglas-fir stands from Cannell (1982) and data from Espinosa Bancalari and Perry (1987), and Binkley (1983) are summarized in Figure 4.11. The scatter plot shows that the relationship between total aboveground biomass and total stand basal area obtained in this study (solid circles) is consistent with the relationship observed in other published studies (open circles). The turnover components of branch and foliage production estimates have been derived using approaches specifically developed in this study. Many production studies do not account for branch turnover at all (Dice 1970) and foliage turnover is often simply assumed to represent 20% of total foliage biomass (Keyes and Grier 1981). The branchwood turnover estimates derived in this study represent an improvement over the alternative of omitting this production component. The method used in this study is only applicable to determinate tree species, such as Douglas-fir, which produce one whorl per year. A second assumption which must be met is that the number of live whorls remains approximately constant over a measurement period. The fact that only one production estimate was derived for both plots in each Installation could introduce some error, in particular in Installations where stand density differs between the two plots, e.g. Installation 5 109 has 3400 and 1420 stems ha-1 in Plots 8 and 10, respectively. Errors are probably also introduced from the calculation of branch turnover for species other than Douglas-fir, because the same mean lifespan of branches was used for all species. The contribution of this source of error will increase as the proportion of Douglas-fir in the stand decreases. All three components of branch production combined, (A biomass, mortality, and turnover, cf. equation 4.2) represent 12.4 to 26.7% of total aboveground production (Table 4.7). The combined effects of errors in the branch turnover estimate is not likely to greatly affect total production estimates. A comparison of Douglas-fir production data from 38 stands reported in Cannell (1982), Espinosa Bancalari and Perry (1987), Binkley (1983), and in this study is complicated by the differences in methods used, in stand ages and in stand densities. Figure 4.12 shows the relationship between stand foliage biomass and total aboveground production for these Douglas-fir stands less than 150 years of age and not fertilized. Total aboveground production increases with increasing foliage biomass (r2=0.143, p=0.019). The large variation in the data is due to the different regression equations used to calculate foliage biomass, and differences in stand densities, stand ages, and between-year variation in annual production. The 12 data points from this study (solid circles) fall well within the range of data reported in the literature (open circles). 110 (9 E o !5 •o c 3 o O) 0 > o < 1000 800 600 400 200 1 1 o o • • o O • o . o | p o 5mm= CL3 or CD3). The 0-1 mm diameter class of both live and dead roots was further subdivided into fine roots with \"clay\" particles (probably a mixture of clay- and silt-sized particles) adhering to the surface (CLO-C and CDO-C, respectively) and those without. Soil particles embedded in the mycorrhizal mantle and in mycorrhizal clusters could not be removed but were accounted for by applying ash content correction factors (see below). Mycorrhizal hyphae and other fungal materials were classified in a separate category (M). Non-coniferous roots (NC) were not separated into subclasses. The very small diameter of the roots of many herbaceous and shrub species did not permit 128 classification into live and dead roots. In retrospect, between-sample variation in this category could have been reduced by separating the larger diameter rhizomes of ferns {Polystichum munitum, Pteridium aquilinum), which were found in some samples, from other non-coniferous roots. The most difficult and time consuming task was the differentiation between live and dead coniferous fine roots. We employed criteria from the literature (Santantonio and Hermann 1985, Persson 1983), from personal communications with other researchers, and from our own observations. These criteria included colour, texture and tensile strength of the roots. When in doubt, iodine stain was used to test for the presence of starch granules. If starch was present roots were classified as living at the time of sampling. Details of the criteria are as follows. COLOUR: The periderm of live coniferous roots is almost black, sometimes with some red. The secondary xylem is off-white. In contrast, the surface colour of dead roots is grayish. The secondary xylem is brownish to yellowish. TEXTURE: Dead roots are brittle with a low tensile strength. Tensile strength was assessed subjectively and varied with root diameter. Pulling a live fine root apart using stainless steel tweezers causes a \"snapping\" sound which is much less audible in dead roots. The secondary phloem of dead roots separates readily from the secondary xylem and the root often shows a \"distinct girdled pattern\" of separated bark segments along the central cylinder. Dead roots often feel soggy or mushy and are not pliable. FLOTATION: A root segment which floats on the surface is most likely dead. Only some of the dead roots float, however. We placed great emphasis on a consistent application of the sorting criteria. Samples were often cross-checked by an experienced technician. New staff were 129 carefully trained and their work was closely monitored and re-examined. After some training, very consistent results were obtained by all technicians. Processing times varied greatly depending on the characteristics of the samples. Forest floor samples, although of much smaller volume, tended to require more time than mineral soil samples. An average of approximately 8 hours per sample or about 24 hours per core was required for sorting. Additional time was required for drying, weighing, and, for a subset of samples, ash content determination. When the processing of a sample was completed, roots were dried in a forced air oven at 70°C for 48 hours. Root samples were weighed to the nearest .001 gram. In total, 986 samples were processed: this involved the drying and weighing of about 8000 petri dishes with classified root material. 5.3.5 Ash content Soil particles adhering to the surface of roots after careful rinsing and washing cannot be removed without loss of organic substance from the fine roots. If soil particles are included in the dry weights of roots, an overestimate of fine root biomass can occur. This error, which can be serious in fine textured soils, can be corrected by obtaining ash-free dry weights. Ovendry samples (redried at 70°C for 24 hours) were weighed into porcelain crucibles and heated to 470°C for 4 hours. Ash content was expressed as a percentage of ovendry weight. 130 5.3.6 Data processing and analyses A computer program, developed for the conversion of root dry weights to ash-free dry weights, verified, for each of the approximately 8000 samples, whether or not the ash content had been determined. If this was the case, the sample's actual ash content was used for the conversion to an ash-free dry weight. Otherwise, the mean ash content for that sample's installation, horizon, and root class was used for the conversion. If this mean was based on less than 3 observations, the mean of the sample's root class and horizon based on all five installations was used. The ash-free weights of the CLO plus CLO-C and the CDO plus CDO-C root classes were then added to obtain the ash-free dry weights of the live and dead roots in the 0-1 mm diameter classes, respectively. Two additional classes represent live and dead fine roots in the 0-2 mm diameter class. Root samples originated from three different soil layers: forest floor (1), 0-30 cm mineral soil (2), and 30-50 cm mineral soil (3). Three additional horizons were introduced which represent forest floor plus the upper mineral soil layer (4=1+2), the sum of the two mineral soil layers (5=2+3), and the sum of all three horizons (6=1+2+3). The computer program calculated, for each of the 2700 sampling strata (5 Installations x 6 sampling dates x 6 horizons x 15 root classes), the mean, standard deviation, and standard error of the ash-free root dry weights. These results were printed to a tabulated output file and to a second output file which subsequently served as the input data file for the computation of fine root production and mortality. 131 5.3.7 Calculation of production and mortality estimates An interactive computer program was developed to calculate fine root production and mortality estimates by applying three different computational methods to the data describing the seasonal dynamics of live and dead fine root biomass. The three methods are based on (1) the decision matrix (Fairley and Alexander 1985, McClaugherty et al. 1982), (2) all changes in live fine root biomass, and (3) significant changes in live fine root biomass. DECISION MATRIX: The decision matrix (DM) presents a series of equations which calculate fine root production, mortality, and disappearance based on the observed changes in both live and dead fine root biomass. The decision as to which set of equations to apply for a given sampling interval is based on a matrix which describes all possible combinations of increases and decreases in live and dead fine root biomass (Figure 5.2). The computer program calculates changes in live and dead fine root biomass between sampling dates and applies the appropriate set of equations for the computation of production and mortality estimates. The annual totals are based on the sum of the estimates obtained for each sampling interval. ALL CHANGES: The second method, (AC), is based only on the observed changes in the live fine root biomass. All increases in live fine root biomass from one sampling date to the next represent production, and all decreases represent mortality. The sums of the estimates for all sampling periods constitute the annual totals. SIGNIFICANT CHANGES: The third method (SC) is again based only on the observed changes in live fine root biomass, but in this method only significant changes in fine root biomass are attributed to production or mortality. The student 132 t-test (Zar 1984) is applied at p<.05 to successive means of live fine root biomass to test for statistical differences. If the means at tj and tj+i are not significantly different, the direction of change between the two means is determined. If this trend is continued from the mean at tj + 1 to the mean at tj+2 then the means at tj and t|+2 are tested for significant differences. If the direction of change reverses, the means at tj + 1 and t\\+2 are tested for significant difference. When two sample means are compared, the program first tests whether or not the two variances associated with the means are equal. If yes, the student t-test is applied, otherwise Welch's approximate t-test is applied (Zar 1984:131). The sum of all significant increases represents annual production and the sum of all significant decreases represents annual mortality. The computer program was used to calculate fine root production and mortality for each of five root classes (0-2 mm, 0-1 mm, 1-2 mm, 2-5 mm, >5 mm) and for each of six horizons (see above) using the three computational methods. The program also calculates a number of additional statistics which facilitate the comparison of the different computational methods. Fine roots in the three sampled soil horizons are experiencing different environmental conditions. Soil temperature and soil moisture amplitudes, for example, are greater in the forest floor than at 30-50 cm depth. It is therefore possible that fine roots in the three soil layers display different seasonal dynamics. Fine roots in the forest floor might stop growing or die due to soil moisture stress while others continue to grow at greater soil depth (Teskey and Hinkley 1981). Similarly, roots in the 0-1 mm diameter class may show seasonal dynamics that differ from those of the 1-2 mm diameter class. For the computation of production and mortality rates, fine roots were divided into four different groups of populations (Table 5.3). Group I includes all 133 fine roots in the 0-2 mm diameter class in the FF-50 cm depth range. Group II separates the diameter classes 0-1 mm and 1-2 mm into two populations and treats the FF-50 cm soil profile as one layer. Group III separates the roots according to their soil layer of origin but does not separate the different diameter classes. Group IV treats the two root diameter classes separately in each of the three soil layers and therefore recognizes 6 individual populations. In each case, the annual estimates for production and mortality are the sum of the estimates for each population. The ability to identify differences of the seasonal patterns in both the diameter classes and the soil layers increases from Group I to Group IV. This can be advantageous if, for example, a decrease in the biomass of the 0-1 mm diameter class in the forest floor occurs during the same interval as an increase in the 1-2 mm root biomass in the 30-50 cm layer. These opposing trends would not be detected if only one population is recognized (Group I). On the other hand, as the number of independent estimates increases from Group I to Group IV, so does the number of error terms associated with the estimates of the annual totals. 134 LIVE ROOT BIOMASS i n c r e a s e d e c r e a s e D i A B d e a d > ^ l i v e ^ l i v e > A B d e a d E n A c D r e p = A B l i v e + A B d e a d p = A B l i v e + A B d e a d P = 0 R 0 a s M = A B d e a d M = ^ d e a d M = - A B l i v e 0 T 1 d B e I 0 c r p = A B l i v e P = 0 M A e a M = 0 M = - A B l i v e S s S e Figvire 5.2 The decision matrix (Fairley and Alexander 1985), modified. The equations for estimating fine root production (P) and mortality (M) are selected on the basis of changes in live and dead fine root biomass (AB) during the interval between two sampling dates. 135 Table 5.3 The populations of fine roots are subdivided into one (0-2 mm) or 2 diameter classes (0-1 mm, 1-2 mm) and/or one (FF-50 cm) or 3 soil horizons (FF, 0-30 cm, 30-50 cm). Production and mortality estimates of fine roots for Groups I through IV are based on the sums of 1 to 6 individual estimates. Group Diameter Soil Number of classes horizons classes I I 1 1 II 2 1 2 III 1 3 3 rv 2 3 6 136 5.4 RESULTS AND DISCUSSION 5.4.1 Ash content of root samples Ash contents of 2225 root samples in 195 strata (5 study sites x 3 horizons x 13 root classes) were determined. Table 5.4 presents means for the combined data from the five study sites for each of ten root classes and 3 horizons. Ash contents increased with depth in the soil and, with one exception, were higher for dead than for live roots in the same diameter class (Figure 5.3). Similar trends were reported by Vogt and Persson (in press) for Abies amabilis in Washington. The roots in the 0-1 mm diameter class were divided into two categories: with and without \"clay\" particles adhering to the mycorrhizal clusters (see Section 5.3.4). Although only a small portion of the fine roots of most samples was classified as \"with clay\" (CLO-C, CDO-C), the data in Table 5.4 confirm the need for this extra category. The \"with clay\" classes had two to four times greater ash contents than the \"without clay\" classes. Failing to separate these two additional classes would have introduced considerable bias and increased the variability in the estimates of ash-free fine root dry weights. 5.4.2 Fine root biomass in May 1985 Seasonal patterns of live and dead fine root biomass will be described in the next section. The static comparison of live fine root biomass at the five study sites presented in this section is based on the values obtained in May 1985. In some cases these values differ considerably from the values in May 1986; this will be discussed later. 137 Total mean live root biomass and biomass in each of the three soil layers in May 1985 are shown in Table 5.5. Forest floor thickness (FFT) based on 20 measurements per plot is shown in the same table. Fine root biomass in the forest floor ranged from 23.6 g nr2 in Stand E to 212.3 g nr2 in Stand B, a nine-fold difference. The mineral soil layer to a depth of 30 cm (measured from the upper surface of the forest floor, Figure 5.1) contained the highest amount of fine roots, ranging from 109.9 to 436.6 g nr2. The 30 to 50 cm mineral soil layer contained from 48.5 to 142.5 g nr2 fine roots. Total fine root biomass ranged from 182 g nr2 in Stand E to 791 g nr2 in Stand B, more than a four-fold difference. The distribution of live fine roots among the 3 soil layers was approximately 1:2:1 for the forest floor : 0-30 cm : 30-50 cm layers (Table 5.6). Thirteen to 28% of all live fine roots to a depth of 50 cm were found in the forest floor. About 51% to 60% were present in the 0-30 cm mineral soil layer, and 20% to 27% were in the 30-50 cm soil layer. Mean forest floor thickness was less than 2 cm in all five study sites and fine roots were predominantly found at the interface of forest floor and mineral soil. The importance of the upper soil layer and the forest floor becomes even clearer when fine root biomass is expressed on a per volume (g nr3) rather than on a per area (g nr2) basis (Table 5.6). Live fine root biomass (g nr3) decreased from 11057 g nr3 in the forest floor to 713 g nr3 in the 30-50 cm layer in Stand B. In Stand E, the live fine root biomass ranged from 1934 g nr3 in the forest floor to 243 g nr3 in the 30-50 cm mineral soil layer. 138 Table 5.4 Mean and standard error of the mean (...) of ash content expressed as percent of sample dry weight for three horizons and ten root classes. Data represent mean values for the five study sites, n = sample size. Horizon Root Class Forest floor FF-30 cm 30-50 cm CLO1 0-1 mm2 10.30 (0.58) n= 104 14.49 (0.50) n= 113 17.11 (0.79) n= 94 CLO-C 0-1 mm 44.39 (2.55) n= 35 44.74 (1.20) n= 96 43.56 (1.65) n= 63 CLl 1-2 mm 6.45 (0.72) n=47 9.83 (0.55) n= 91 12.66 (0.75) n=61 CL2 2-5 mm 4.28 (0.51) n= 15 10.51 (0.81) n= 59 12.68 (0.92) n=41 CDO 0-1 mm 15.92 (0.86) n= 82 21.16 (0.78) n= 101 22.07 (0.88) n= 92 CD0-C 0-1 mm 49.31 (3.07) n= 20 53.54 (1.44) n= 73 53.29 (1.70) n= 47 CD1 1-2 mm 5.61 (1.21) n=6 14.22 (1.23) n= 39 17.17 (1.37) n= 33 CD2 2-5 mm 9.61 (1.79) n=5 14.65 (1.23) n= 39 21.99 (2.44) n= 17 NC 14.97 (1.47) n= 55 20.92 (1.35) n= 87 17.33 (1.25) n= 46 M 57.15 (2.41) n= 58 64.51 > (1.92) n= 48 64.19 (4.69) n=9 1 C = coniferous, L = live, D = dead, NC = non-coniferous, M = fungal hyphae. 2 diameter range. 139 Live roots Dead roots 60 c e o t> c • c 40 20 1% 4 FF FF-30 30-60 FF 4 FF-30 30-50 Horizon Horizon Figure 5.3 Mean and standard error of the mean of ash content expressed as percent of dry-weight for live and dead roots from three soil horizons and four diameter classes. In each horizon, bars from left to right represent diameter classes: 2-5 mm, 1-2 mm, 0-1 mm, and 0-1 mm with clay. 140 Table 5.5 Mean and standard error of the mean (...) of live fine (0-2 mm) root biomass in May 1985 at the five study sites. Mean and standard error of forest floor thickness (FFT) are based on 20 measurements taken in each stand. SI refers to the Site Index in meters at breast-height age 50, n = sample size. Live fine root biomass (g nr2) Stand SI FFT (m@50) (mm) n FF 0-30 cm 30-50 cm FF-50 ci A 23.3 18.1 15 63.4 243.5 98.0 404.9 (1.0) (16.7) (49.1) (23.9) (80.7) B 27.7 19.2 17 212.3 436.6 142.5 791.5 (2.5) (24.8) (44.1) (23.2) (56.2) C 29.4 13.3 17 97.6 228.1 106.7 432.5 (1.2) (17.5) (31.2) (20.0) (42.5) D 29.5 13.9 16 87.6 160.3 63.2 311.2 (1.3) (24.8) (24.5) (12.8) (43.3) E 41.0 12.2 17 23.6 109.9 48.5 182.0 (1.1) (11.7) (12.8) (14.9) (17.6) 141 Table 5.6 Live fine root biomass in May 1985 in three horizons, expressed as a percentage of the total fine root biomass, and on a volume basis (g nr3) for each of three horizons and the total profile to a depth of 50 cm. Stand % of total gram nr3 FF 0-30cm 30-50cm FF 0-30cm 30-50cm FF-50 cn • A 15.7 60.1 24.2 3503 864 490 810 B 26.8 55.2 18.0 11057 1555 713 1583 C 22.6 52.7 24.7 7338 796 534 865 D 28.1 51.5 20.3 6302 560 316 622 E 13.0 60.4 26.6 1934 382 243 364 142 5.4.3 Seasonal dynamics of fine roots Seasonal dynamics of live and dead fine (0-2 mm) root biomass are reported here. The dynamics of small (2-5 mm) roots and of non-coniferous roots will be presented below. Roots greater than 5 mm in diameter were only occasionally present in the soil cores, and data on this root class will not be presented. Seasonal dynamics of live and dead fine root biomass of all five study sites are shown in Figure 5.4. Figure 5.5 displays live and dead fine root biomass with standard errors for each stand individually. Live fine root biomass shows a very similar trend in all 5 study sites. It peaked in both the spring of 1985 and 1986 and was lowest at the early October or mid August sampling dates. From May to June (1985), live fine root biomass declined slightly in four study sites and increased slightly in Stand B. This was followed by a sharp decline at the August sampling date in all five stands. From August to early October, live fine biomass continued to decline in four stands and showed a small increase in Stand E. From the low in the fall, biomass increased again to the May 1986 sampling date. We were able to process the samples for one additional sampling date (early February) for Stand E. The resulting data point suggests that most of the increase in live fine root biomass occurred from February to May, i.e. during the spring rather than during the winter months. Dead fine root biomass also showed similar trends in all five stands (Figures 5.4 and 5.5). Values increased in Stands A, D, and E but decreased in Stands B and C from mid May to the end of June. In all stands, this was followed by a sharp increase to the August sampling date. For Stands A, D, and E, dead root biomass showed little change from August to October, but continued to increase during this period in Stands B and C. From October to May 1986 dead root biomass decreased in all five stands. The additional data point (February 1986) for Stand E suggests 143 that disappearance was greater from February to May than during the winter months. Live and dead fine root biomass dynamics displayed a very symmetrical pattern (Figure 5.5). Most of the decreases in live fine root biomass were accompanied by increases in dead fine root biomass. When live fine root biomass increased, the dead biomass component decreased, presumably through decomposition. Fine root mortality can be induced by moisture stress (Deans 1979). All five study sites experience a soil moisture deficit in a typical year. Figure 5.6 shows the 30 year average of mean monthly precipitation at Nanaimo Airport (Environment Canada 1982) and the actual monthly precipitation for 1985 and the first six months of 1986 (Environment Canada pers. comm.). The summer of 1985 was unusually dry, with below-average rainfall from May through August and no precipitation at all in July. Precipitation in October 1985 was well above the 30 year average at Nanaimo Airport, but this rainfall occurred after the sampling date in the first week of October. Daily precipitation data were obtained from the Cowichan Lake Research Station, about 1 km from the location of Stand E. Figure 5.7 shows daily precipitation at the Research Station and live fine root biomass in Stand E. Only 16.8 mm of precipitation fell from June 24 to August 22, most of it on two consecutive days in early August. About 54 mm of precipitation occurred in early September, followed by another rain-free period of 21 days prior to the October sampling date. These periods of extreme drought may explain why so few live fine roots were found at Stand E at both the August and the October sampling dates. This 144 rich lower slope stand receives seepage for most of the year (pers. observation), but the soil was very dry to a depth of 50 cm at the August and October 1985 sampling dates. The seasonal dynamics of live and dead fine roots showed similar trends in the three soil layers which were sampled. The amplitude of the changes, however, differed between horizons. Figures 5.8 and 5.9 show fine root biomass for live and dead roots respectively, separated by horizon. The greatest change in fine root biomass occurred in the forest floor layer. The amplitude of the seasonal changes decreased with increasing depth in the soil for both live and dead fine root biomass. The sum of live plus dead fine root biomass yields total biomass as shown in Figure 5.10. There is little seasonal change in the total fine root biomass of the five study sites, but considerable variation in the ratios of dead to live fine roots. This result confirms the importance of determining live and dead fine root biomass separately if production or mortality rates are to be derived from the data. Ratios of live to dead fine root biomass varied greatly between stands and also between the May 1985 and May 1986 sampling dates. The ratios ranged from 2.0 to 3.6 in May 1985, decreased sharply (0.03 to 0.6) in the summer and increased again (0.8 to 2.0) in May 1986. The differences in live and dead fine root biomass in May 1985 and May 1986 show that there is considerable between-year variation. Figure 5.11 shows live and dead fine root biomass expressed as a percentage of May 1985 values (=100%). In Stand A, the stand with the lowest site index, total fine root biomass was 12% lower in May 1986 than in May 1985. Live fine root biomass represented only 56% of the May 1985 value, while dead roots represented 163%. The drought of the summer 1985 was associated with a high mortality of fine roots. In May 1986, live fine root 145 biomass had not yet recovered and dead roots had not yet decreased to the May 1985 levels. Stand E, the site with the highest site index, showed a very different pattern. In May 1986, live, dead and total biomass were almost the same quantities as in May 1985 (95%, 97%, and 96%, respectively). The recovery of live fine root biomass occurred within seven months, mostly during the spring (Figure 5.5). In Stand B, live fine root biomass in May 1986 represented 103% of the May 1985 value, while the comparable value for dead fine root biomass was 171%. Live root biomass in stands C and D recovered to 84% and 67% of the 1985 values, respectively. Dead root biomass in the two stands represented 150% and 216% of the quantities in May 1985. 146 E to in o m 1000-1 800 600-400-200-T 1 1 U | MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 1986 V) V) D CD 1000 800-600-400 200 ~\\ 1 i 1 1 1 1 1 1 1 1 r MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 1986 Q = A A = B 0 = C X = D • = E Figure 5.4 Seasonal dynamics of live (top) and dead (bottom) fine root biomass in the five stands. Data represent the mean of 10 samples per sampling date (15-17 for May 1985). FINE ROOT BIOMASS live and dead 0 -2mm all layers 147 Figure 5.5 Seasonal dynamics of live and dead fine root biomass at each of the 5 Stands. Vertical bars represent + 1 standard error. 148 NANAIMO AIRPORT Figure 5.6 Mean monthly precipitation (30 year average) at Nanaimo Airport, and actual precipitation for 1985 and the first six months of 1986. (Environment Canada 1982 and pers. comm.) 149 Figure 5.7 Daily precipitation at the Cowichan Lake Research Station for 1985 and the first six months of 1986, and the live fine root biomass at Stand E approximately 1 km from the climate station. (Precipitation data from B.C. Forest Service Cowichan Lake Research Station). 150 FINE ROOT BIOMASS live 0-2mm by layer 600 i 1 i 1200 1000-800-600-200 MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 ' 1986 MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 1986 600 500 400 300 200 1 30-50 E 2 FF-30 LZZ3 MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 1986 Figure 5.8 Live fine root biomass in three soil layers at each of the five study sites. 151 FINE ROOT BIOMASS dead 0-2mm by layer 600 1 1200 I O O O H 800 400 I I I I I -i 1 r MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 1986 MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 1986 600-500-E 400-•3 10 300-v> o E o 200-in 100- 1 600 500 400 300 H 200 MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 1986 m m 100-m • m % 1 MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 1986 600 MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 1986 30-50 K Z 2 FF-30 FF Figure 5.9 Dead fine root biomass in three soil layers at each of the five study sites. 152 MAY JUN JUL AUG SEP OCT NOV DEC JAN FEB MAR APR MAY 1985 - 1986 DEAD E2 LIVE Figure 5.10 Total (live plus dead) fine root biomass at each of the five study sites. m oo <7> M Live Fine Roots in May 1986 • Dead Fine Roots in May 1986 ^ 250 D 2 0) > • — o o 0.05) PRODUCTION I 100.0 100.0 100.0 II 131.8 (20.8) 106.0 (6.8) 87.4 (34.8) III 123.0 (23.7) 106.4 (7.5) 95.8 (5.8) rv 154.8 (23.3) 120.2 (17.9) 78.6 (32.9) ITALITY I 100.0 100.0 100.0 II 113.8 (11.0) 103,2 (3.1) 103.2 (3.1) III 127.2 (23.0) 104.8 (6.1) 100.0.6 (5.6) rv 140.0 (15.4) 111.8 (8.1) 101.4 (8.9) 1 Refer to Table 5.3 for an explanation of groups. 162 5.4.6.2 Fine root production and mortality estimates Fine root production and mortality estimates based on three different computational methods for each of the five stands are reported in Table 5.9. All estimates are based on ash-free dry weights and are reported in g nr2, which can be converted to Mg ha-1 (i.e. t ha\"1) by dividing the g nr2 values by 100. For a comparison of the effects of the three computational methods, the same estimates have also been expressed as a percentage of the estimate obtained from the SC computational method (Table 5.10). Production estimates, obtained from the AC and the SC methods, were identical with one exception: in Stand B the AC method calculated 6.4% more production than the SC method. Estimates based on the decision matrix were on average 23.6% higher than the estimates obtained from the SC method. Mortality estimates based on both the AC or the SC method were identical, while these based on the decision matrix were on average 13.1% higher. As discussed in the preceding section, separation of fine roots into different populations according to diameter class and soil layer (Table 5.3) has different effects on production and mortality estimates depending on the computational method used. Consequently, the statement that the AC and SC computational methods yielded similar estimates cannot be generalized. Differences between the AC and SC methods were small in this study because the sampling frequency was low. The sampling intervals were large enough to show a clear seasonal trend with significant differences between the peaks and troughs in live and dead fine root biomass. The estimates of annual production and mortality of fine roots differed between stands (Table 5.9, Figure 5.14). Production in Stand B was 481 g nr2 and 163 514 g nr2 based on the SC and DM method, respectively. This was over four times higher than in Stand D, which had a production of 112 g nr2 for all computational methods. Based on the DM computational method, the stands with the lowest (Stand A) and highest (Stand E) site index had 199 g nr2 and 208 g nr2 of fine root production. The SC computational method yielded lower estimates and showed larger differences between the two stands: 118 g nr2 and 168 g nr2 for Stands A and E, respectively. Production estimates for Stand C were the second highest and ranged from 287 g nr2 for the DM method to 242 g nr2 for the SC method. Estimates of annual fine root mortality also differed between the five stands, but they covered a narrower range than the production estimates. Based on the DM method, mortality estimates are lowest for stands D and E: 215 g nr2 and 218 g nr2, respectively. The SC method calculated the lowest mortality estimates for Stand E (178 g nr2) and yielded the second lowest estimate for Stand D (215 g nr2). The highest mortality estimates, obtained for Stand B, range from 490 g nr2 to 487 g nr2 for the DM and SC method, respectively. Figure 5.14 presents a comparison of fine root production and mortality estimates for each stand. Only in Stands B and E were the two estimates approximately equal. Mortality estimates in Stands A and D were approximately double the production estimates. In Stand C, production represented about 80% of mortality. These differences between production and mortality estimates are explained by the lower fine root biomass in May 1986 compared to that in May 1985 (Figure 5.11). This may nave been due to the heavy mortality which occurred during the unusually dry summer of 1985 and the inability of some stands to re-establish the fine root biomass by the following spring. 164 Table 5.9 Annual fine root production and mortality (g nr2 yr1) for the five stands calculated with three computational methods: decision matrix (DM), all changes (AC) and significant changes (SC). All fine roots (0-2mm diameter, FF-50 cm depth) are treated as one population. Production Mortality Stand SI DM AC SC DM AC SC (m@50) (g nr2 yr\"1) (g nr2 yr\"1) A 23.3 199.5 118.2 118.2 377.7 296.4 296.4 B 27.7 514.3 511.9 481.1 489.5 487.1 487.1 C 29.4 286.6 241.8 241.8 355.6 310.7 310.7 D 29.5 112.1 112.1 112.1 214.8 214.8 214.8 E 41.0 207.9 168.2 168.2 217.7 178.0 178.0 165 Table 5.10 Annual fine root production and mortality for the five stands expressed as a percentage of the estimates obtained from the SC computational method. Production Mortality Stand SI DM AC SC DM AC SC (m@50) (%ofSC) (%ofSC) A 23.3 168.6 100.0 100.0 127.7 100.0 100.0 B 27.7 106.9 106.4 100.0 100.6 100.0 100.0 C 29.4 118.6 100.0 100.0 114.5 100.0 100.0 D 29.5 100.0 100.0 100.0 100.0 100.0 100.0 E 41.0 123.8 100.0 100.0 122.5 100.0 100.0 Mean 123.6 101.3 100.0 113.1 100.0 100.0 S.D. (26.9) (2.9) (12.6) 166 5.4.6.3 Small root production and mortality estimates As was observed with fine roots, estimates of small root production and mortality differ between stands and are affected by the computational method used. Small root production calculated by the DM method ranged from 50.7 to 222.3 g nr2 (Table 5.11). The high production estimate for stand E is largely due to the increase in dead root biomass from August to October 1985 (Figure 5.12). The other two computational methods consider only changes in live small root biomass and yield identical production estimates of 125.9 g nr2. In Stands B and C, the SC method yields zero production estimates because the observed increases in small root biomass were not statistically significant (p=0.05). Mortality estimates ranged from 88.5 to 212.0 g nr2 for the DM computational method. The two other methods both resulted in a range of mortality estimates from 52.2 to 189.6 g nr2. With the exception of Stand E, mortality estimates were always higher than production estimates (Figure 5.15). The differences between the three computational methods were generally small with the exception of the two zero production estimates (Stands B and C) obtained from the SC method, and the high production estimate obtained from the DM method in Stand E (Figure 5.15). 167 Table 5.11 Annual small root production and mortality (g nr2 yr*1) for the five stands calculated using three computational methods. DM = decision matrix, AC = all changes, and SC = significant changes. Production Mortality Stand SI DM AC SC DM AC SC (m@50) (g nr2 yr\"1) (g nr2 yr-1) A 23.3 105.0 81.7 81.7 212.9 189.6 189.6 B 27.7 84.7 66.1 0.0 203.7 185.0 177.3 C 29.4 61.9 61.9 0.0 166.7 166.7 166.7 D 29.5 50.7 50.5 50.5 88.5 88.4 88.4 E 41.0 222.3 125.9 125.9 148.6 52.2 52.2 168 6 I 1 1 1 r A B C D E Plot g | Product ion (DM) • Mortal i ty (DM) • Product ion (SC) S Mortal i ty (SC) Figure 5.14 Estimates of fine root production and mortality for five plots based on the Decision Matrix and Significant Changes methods. Bars within each plot, from left to right, represent production (DM), production (SC), mortality (DM), and mortality (SC). 169 T B 0 Plot Product ion (DM) Product ion (AC) Product ion (SC) • Mortal i ty (DM) H Mortal i ty (AC) g | Mortal i ty (SC) Figure 5.15 Estimates of small root production and mortality for five plots based on the Decision Matrix and Significant Changes methods. Bars for each stand represent, from left to right, production DM-, AC-, and SC-methods and mortality DM-, AC-, and SC-methods. 170 5.4.6.4 Turnover rates of fine and small roots Both production and mortality rates of roots are a measure of turnover, and, therefore, two sets of turnover rates are calculated based on the ratios of production to root biomass and of mortality to root biomass in May 1985. As discussed above, the estimates of annual root production and annual mortality differ in some stands. These differences are also apparent in the calculated turnover rates of fine (Table 5.12) and small (Table 5.13) roots. Turnover rates of fine roots based on production range from 0.36 to 1.14 year1 for production estimates calculated using the DM method and from 0.29 to 0.92 year1 for production estimates based on the AC or SC method. Between-stand differences for turnover rates based on mortality estimates are smaller and range from 0.62 to 1.20 year1 for DM-derived estimates and from 0.62 to 0.98 year1 for AC- or SC-derived estimates. Mean life span is calculated as the inverse of turnover rates. Figure 5.16 shows mean life span of fine roots in the 5 stands based on both production and mortality estimates. Mean life span ranges from 0.83 to 3.43 years. Table 5.12 and Figure 5.16 show that stand E, the site with the highest site index, has the highest turnover rate and consequently the shortest fine root lifespan. While there is some debate in the literature as to whether a more nutrient (specifically nitrogen) rich site has higher or lower turnover rates than a poorer site, it is not clear that the results obtained in this study can support either argument. The unusually long summer drought of 1985 led to about 100% mortality of fine roots in Stand E, whereas in the other stands, a smaller proportion of the live fine roots died. The differences in turnover rates are therefore primarily drought induced and are probably less related to nutrient availability. 171 Table 5.12 Turnover rates (year-1) of fine roots calculated as the ratios of annual fine root production / fine root biomass in May 1985 and annual fine root mortality / fine root biomass in May 1985. Production and mortality estimates are based on three computational methods which are described in the text. Production/biomass Mortality/biomass Biomass ot May 85 (g nr2) DM AC (year1) SC DM AC (year1) SC A 404.9 0.49 0.29 0.29 0.93 0.73 0.73 B 791.5 0.65 0.65 0.61 0.62 0.62 0.62 C 432.5 0.66 0.56 0.56 0.82 0.72 0.72 D 311.2 0.36 0.36 0.36 0.69 0.69 0.69 E . 182.0 1.14 0.92 0.92 1.20 0.98 0.98 172 Table 5.13 Turnover rates (year1) of small roots calculated as the ratios of annual small root production / small root biomass in May 1985 and annual small root mortality / small root biomass in May 1985. Production and mortality estimates are based on three computational methods which are described in the text. Production/biomass Mortality/biomass Biomass ot May 85 (gm-2) DM AC (year1) SC DM AC (year1) SC A 239.2 0.44 0.34 0.34 0.89 0.79 0.79 B 409.9 0.21 0.16 0.00 0.50 0.45 0.43 C 226.1 0.27 0.27 0.00 0.74 0.74 0.74 D 108.3 0.47 0.47 0.47 0.82 0.82 0.82 E 59.7 3.72 2.11 2.11 2.49 0.87 0.87 173 Plot H Product ion (DM) • Mortal i ty (DM) • Product ion (AC) 0 Mortal i ty (AC) Figure 5.16 Fine root mean life span for 5 plots based on produciton and mortality estimates obtained with two computaional methods. Bars within each plot, from left to right, represent mean life span calculated from estimates of production (DM), production (SC), mortality (DM), and mortality (SC). 174 Small root turnover rates, derived from production estimates, cover a much wider range than those of fine roots (Table 5.13). Extreme values are 0 year1, for the two stands in which the SC method calculated zero production, and 2.1 to 3.7 year-1 for Stand E, for which the highest turnover rates are calculated by all three methods. Turnover rates for other stands or methods fall in the range 0.16 to 0.47 year1. Mortality-based turnover rates range from 0.43 to 0.89 year1, with one additional very high value of 2.5 year*1 (Stand E, DM-method). The much higher estimates of turnover rates of small roots in Stand E are probably a consequence of sample variation rather than real turnover rates, because they are two to three times higher than the turnover estimates obtained for fine roots in the same stand. Furthermore, such high turnover rates would imply a mean life span of only 4 to 6 months for roots that are 2 to 5 mm in diameter. Roots of such diameters had to undergo secondary thickening and often have growth rings indicating that the roots have lived through more than one growing season. 5.4.6.5 Site quality and fine and small root production Many factors determine the \"quality\" of a site and there has been much discussion in the literature about an appropriate measure of site quality (Hfigglund 1981). In this study, site quality is expressed as site index (meters at breast-height age 50) (Bruce 1981, Mitchell and Polsson 1988). Several approaches can be taken to obtain estimates of fine root production and to date there appears to be no agreed-upon method of determining fine root production, mortality and turnover, as discussed above. 175 Table 5.14 lists coefficients of determination (r2), significance levels (p), slopes, and intercepts of the relationships between site index and several measures of fine root production, mortality, and turnover for the five sites of this study. The correlations between fine root production or mortality and site index were always negative, but never significant (Table 5.14). Fine root turnover was significantly (p=.042 to .068) negatively correlated with site index when turnover was calculated from production estimates. When turnover was calculated from mortality estimates, however, it was positively correlated with site index and was either less significant or non-significant (p=.078 to .233). Small root production was positively but not significantly correlated with site index, whereas small root mortality was negatively correlated with site index (significance levels rang from .062 to .172, Table 5.14). Turnover rates of small roots were significantly and positively correlated with site index (p=.028 to .051) when based on production estimates. Similarly, turnover rates based on mortality estimates of small roots were positively correlated with site index, but these estimates were only significant for estimates calculated with the decision matrix (p=.048). The four significant correlations between small root turnover and site index can be attributed to the single, very high turnover estimate obtained for Stand E. Concern about the reliability of this result has already been expressed above. 176 Table 5,14 The relationships between site index and several measures of fine and small root production and mortality. Three different computational methods (DM, AC, and SC) are used. All relationships are based on the data from all five stands (n=5), r2 = coefficient of determination, p = significance. Variable P r2 Intercept Slope Fine Root Prod. (DM) 0.781 .030 3.859 -0.040 Fine Root Prod. (AC) 0.845 .015 3.234 -0.031 Fine Root Prod. (SC) 0.857 .013 3.037 -0.026 Fine Root Mort. (DM) 0.269 .379 6.622 -0.110 Fine Root Mort. (AC) 0.342 .297 5.976 -0.099 Fine Root Mort. (SC) 0.342 .297 5.976 -0.099 Turnover Fine Root Prod. (DM) 0.068 .724 -0.501 0.039 Turnover Fine Root Prod. (AC) 0.057 .752 -0.446 0.033 Turnover Fine Root Prod. (SC) 0.042 .797 -0.471 0.034 Turnover Fine Root Mort. (DM) 0.233 .425 0.169 0.023 Turnover Fine Root Mort. (AC) 0.078 .698 0.220 0.017 Turnover Fine Root Mort. (SC) 0.078 .698 0.220 0.017 Small Root Prod. (DM) 0.128 .593 -1.394 0.081 Small Root Prod. (AC) 0.172 .516 -0.201 0.032 Small Root Prod. (SC) 0.348 .291 -0.830 0.045 Small Root Mort. (DM) 0.477 .179 2.612 -0.032 Small Root Mort. (AC) 0.067 .724 3.805 -0.081 Small Root Mort. (SC) 0.062 .738 3.756 -0.080 Turnover Small Root Prod. (DM) 0.030 .834 -5.346 0.211 Turnover Small Root Prod. (AC) 0.028 .844 -2.764 0.114 Turnover Small Root Prod. (SC) 0.051 .768 -2.959 0.117 Turnover Small Root Mort. (DM) 0.048 .778 -2.157 0.107 Turnover Small Root Mort. (AC) 0.498 .165 0.427 0.010 Turnover Small Root Mort. (SC) 0.508 .158 0.413 0.011 177 5.4.7 General Discussion Comparison of the results from different root studies has always been complicated by the lack of common methodology and definitions. Diameter classes recognized by different studies are not uniform. Sampling depth varies among studies, as do the root sorting methods used. Some studies report root dry weights on an ash-free basis (e.g. Vogt et al 1987, Keyes and Grier 1981) while others do not correct their data (Espinosa Bancalari and Perry 1987). Many studies separate live from dead fine root biomass; others report total root biomass data. Estimates of root biomass and production obtained in this study are well within the range of values reported in the literature for coastal Douglas-fir stands in Oregon and Washington (Keyes and Grier 1981, Santantonio and Hermann 1985, Vogt et al 1987). The combined data sets of Keyes and Grier (1981), Santantonio and Hermann (1985), and Vogt et al (1987) have been used to investigate the relationships between fine and small root biomass and production and site index. Site indices for Santantonio and Hermann (1985) and for Vogt et al. (1987) were estimated from the published productivity classes by assuming that all plots within a productivity class were at the midpoint of the range this class. Productivity Classes II, III, and IV have mean site indices of 38.1, 32.0, and 25.9 m at 50 years, respectively (King 1966). The methods used by Keyes and Grier (1981) were very similar to those used in this study, the only obvious difference being that they used a sample depth of 45 cm instead of 50 cm. Vogt et al. (1987) sampled only to a depth of 15 cm. Santantonio and Hermann (1985) classified roots with 0-1 mm diameter as fine and those with 1-5 mm diameter as small roots. They did not correct to ash-free values and sampled to a depth of 100 cm. The other three studies used definitions of 0-178 2mm and 2-5mm diameter for fine and small roots, respectively, and corrected for ash-free dry weights. The regression between site index and fine root biomass is highly significant (p=0.001, r2=0.45, n=20) (Figure 5.17), but that between small root biomass and site index is not significant (p=0.384, n=10) (Figure 5.17). The estimates of small root biomass obtained by Santantonio and Hermann (1985) are somewhat higher than those obtained by Keyes and Grier (1981) and in this study, probably because the former study only included roots in the 1-5 mm diameter class, compared to 2-5 mm in the latter two studies. Neither of the regressions between fine and small root production and site index was significant (Figure 5.17). The high estimate of small root production in Stand E in this study is probably not realistic, as discussed above. These results show that, within one species, fine root biomass decreases with increasing site quality, as expressed by site index. This suggests that as nutrient and moisture availability increase, fewer fine roots are required to meet the forest stands' demands for soil-provided resources. No significant relationship between site index and fine root production exists. Many factors can account for the between-stand variation in fine root production. Methodological differences between studies, differences in stand age and stand density, and between-year variation could all contribute to the variation in fine root production estimates. Fine and small root production are only two components of stand productivity. In this chapter, only the belowground components of biomass and production were addressed. The final chapter will summarize the results of the aboveground (Chapter 4) and belowground (Chapter 5) biomass and production studies, and will investigate allocation to aboveground and belowground stand components. 179 Figure 5.17 The relationships between fine and small root biomass and production and site index based on data from this and several published studies. V=Keyes and Grier 1981, A=Vogt et al. 1987, •=Santantonio and Hermann 1985, €>=this study. 180 6. T H E RELATIONSHIPS BETWEEN SITE INDEX AND FOLIAGE BIOMASS, FOLIAGE EFFICIENCY, PRODUCTION, AND PRODUCTION ALLOCATION 6.1 INTRODUCTION The prediction of biomass production and future yields is central to many aspects of forest science and management. Some of the different approaches to yield forecasting involve computer simulation models of forest stands or ecosystems (Mitchell 1975, Wykoff et al. 1982). The conceptual framework that is common to many simulation models of forest growth is centered around the relationships between foliage biomass (or area) and stand production (Mitchell 1975, Barclay and Hall 1986, Kimmins et al. 1986, Ford and Bassow 1988). A common conceptual approach for many models is to simulate site and stand factors, such as moisture and/or nutrient availability, which control the amount of foliage biomass carried by a stand. Foliage biomass, or a derivative thereof, is multiplied by some measure of \"foliage production efficiency\" to obtain an estimate of total production. Total production is then partitioned between different above- and belowground components of the forest stand using fixed or variable allocation strategies or allocation hierarchies (Makela and Hari 1986, Bossel 1986, Barclay and Hall 1986). Some models invoke feedback between fine root biomass, which supplies nutrients and moisture at considerable carbon cost, and foliage biomass which provides the required carbon through photosynthesis (Makela' 1986, 1988, Thornley 1972). While such approaches to simulating forest ecosystem production do not require the large amount of data commonly needed for physiologically-based simulation models, they do require an understanding of some of the fundamental determinants of forest ecosystem production. Some of the 181 important questions which should be considered when developing simulation models of forest ecosystems are: 1) What is the relationship between site quality and stand foliage biomass? 2) Does foliage efficiency change with site quality? 3) What are the relative contributions of changes in total production and changes in production allocation to the observed between-stand differences in aboveground production? A strong correlation between total overstory leaf area and site water balance has been demonstrated for forest ecosystems along west to east transects through central Oregon (Grier and Running 1977, Gholz 1982). These studies cover a very wide range of site moisture conditions and atmospheric moisture demand, but the observed changes in total ecosystem leaf area are confounded with changes in species composition from coastal to interior sites. Forest ecosystems respond to short term fluctuations in soil moisture status by controlling transpiration through stomata (Helms 1965, Helms 1976, Running 1976). Prolonged dry conditions lead to increased leaf litterfall and increased mortality of suppressed trees, thus effectively reducing total stand leaf area. It would therefore seem reasonable that leaf area of stands of a particular species would be correlated with moisture availability along a local, topographically- or edaphically-induced site moisture gradient. However, the existing studies have not quantified how foliage biomass is affected by such local variations in soil moisture availability within a given climatic regime. The influence of site nutrient conditions on total foliage biomass has frequently been demonstrated in fertilization studies. Increased availability of nutrients, in particular nitrogen, resulted in increased foliage biomass in the 182 fertilized stands (Brix 1983). These increases are most pronounced in stands that have been thinned prior to fertilization, and they may be less or not at all obvious in fully stocked stands because tree mortality may accelerate following fertilization (Lassoie et al. 1985). There appear to be no published studies, however, that have investigated the relationships between stand foliage biomass and the local variation in site quality in Douglas-fir ecosystems. Foliage efficiency, defined as the amount of biomass (total, aboveground, or stemwood, etc.) produced per unit of foliage (biomass or area), has been used to calculate biomass production from foliage estimates (Barclay and Hall 1986), and as an indicator of tree vigor (Waring et al. 1980). An estimate of foliage efficiency integrates many physiological processes over time and is particularly useful in experiments in which total production can be measured easily, as with seedlings or agricultural crops. In forest ecosystems, however, foliage efficiency has often been expressed on an aboveground or stemwood production basis (Waring et al. 1980, Satoo 1967, Lavigne 1988), because of the difficulty in quantifying belowground production. The interpretation of between-stand differences in foliage efficiency, or the comparison of foliage efficiency estimates from different stands, is complicated by the fact that the observed differences in aboveground or stemwood production may be due to either changes in net production, changes in production allocation, or both. Few studies, however, have the data required to be able differentiate the relative contribution of these two major determinants of aboveground production. One such study, in 40-year-old Douglas-fir stands in western Washington, reported that aboveground production represented 53.3% and 86.5% of total production on a low and a high productivity site, respectively (Keyes and Grier 1981). The objectives of the research reported in this chapter were to investigate whether total foliage biomass, foliage efficiency, and production allocation change 183 with site index in second-growth Douglas-fir stands. Site index (meters at 50 years) integrates the past growth performance of a stand, and, although competing vegetation and other factors affect it, site index is generally considered to be a good index of site growth potential (Hagglund 1981). Site index was therefore chosen in this study as the integrating measure of site quality. The data and results of the previous chapters are combined and extended to address the three questions identified above. The chapter concludes with a general discussion of the results and implications of the research reported in this thesis. 6.2 MATERIALS AND METHODS This part of the study is based on the 12 Douglas-fir stands described in Chapter 2. Biomass regression equations for the calculation of component biomass have been described in detail in Chapter 3. The application of the biomass regression equations to stand information to compute component biomass, production, and mortality has been described in Chapter 4. Biomass, production, and mortality estimates of fine and small roots for five of the 12 plots have been described in Chapter 5. Production and mortality estimates for all aboveground components and for coarse roots are reported as mean annual rates for the last available measurement period (1985 - 1987). Biomass estimates for these components are reported for 1985. Fine and small root production and mortality estimates are based on a single year, May 1985 to May 1986. Fine and small root biomass estimates refer to May 1985. All relationships which are presented are based on data from these years or measurement periods only, unless specifically stated otherwise. 184 6.3 RESULTS 6.3.1 Foliage biomass versus site index Foliage biomass calculated for the 12 plots using the regression model based on dbh and Growing Space Index (the \"GSI model\") as independent variables (Chapter 3, Equation 3.20), ranged from 8.6 to 13.5 Mg ha'1 (Figure 6.1A and Table 4.4). However, there was no significant relationship between total foliage biomass in 1985 and site index (r2=0.011, p=0.746). It is interesting to note that a different conclusion would have been reached if the regional Douglas-fir foliage biomass regression model (Gholz et al. 1979; the \"regional-model\") which has been used in previous studies (Gholz 1982) had been used instead of the regression model developed for this study. The relationship for the 12 plots between foliage biomass predicted by the regional model and site index (Figure 6. IB) is significant (r2=0.374, p=0.035). The mean and standard deviation of predicted foliage biomass in 1985 for the GSI model are 10.5711.47 Mg ha\"1 compared to 14.39±2.17 Mg ha'1 for the regional model. These means are significantly different (t-test, p<0.001). 6.3.2 Foliage efficiency (ANPP) versus site index and foliage biomass Foliage efficiency (aboveground production per unit foliage biomass, FE^ -pp) during the period between 1985 and 1987, increased as site index increased over the 12 plots in this study (r2=0.39, p<0.001) (Figure 6.2B), but was not correlated with foliage biomass (r2=0.003, p=0.871) (Figure 6.2A). Foliage efficiency (ANPP) ranged from 0.52 to 0.82 (Mg yr-1 Mg*1) for 10 of the plots, but was much higher for the two plots of Installation 72 (1.48 and 1.53 Mg yr 1 Mg\"1). The relationship 185 a E o a o u. 25 20 -15 10 -5 -10 20 30 40 50 Site Index (m at 50 years) 0) m m m E o S o u. 10 20 30 40 50 Site Index (m at 50 years) Figure 6.1. The relationship between foliage biomass (Mg ha'1) and site index (m at 50 years) based on (A) the foliage biomass regression model developed in this study and (B) the model of Gholz et al. (1979). 186 Figure 6.2. The relationships between foliage efficiency (Mg yr\"1 Mg-1), calculated from total aboveground production (ANPP) and (A) foliage biomass (Mg ha-1) and (B) site index (m at 50 years). 188 5 CD a. a, a c 1 h 4 « 8 10 12 14 4 8 8 10 12 14 HI CO (0 o 8 10 12 14 8 10 12 14 Foliage Biomass (Mg ha ) Figure 6.3. Foliage efficiency (ANPP) (Mg yr-1 Mg\"1) plotted against foliage biomass (Mg ha\"1) for each of the 12 plots. Installation numbers are in upper right corner of each graph. Legend as in Figure 4.6. 189 Figure 6.4. Graphical presentation of Equation 6.1, which predicts foliage efficiency (ANPP) (Mg yr 1 Mg'1) as a function of site index (m at 50 years) and foliage biomass (Mg ha-1). 190 20 Q L 15 10 i 1 1 \\ — \\ -SI-40 / \\ \\ < : \\ \\ \\ / / s i - 3 0 \\ \\ _ / / / SI-20 \\ \\ _ \\ \\ \\ \\ \\ \\ \\ i 1 I N 10 15 20 Foliage Biomass (Mg ha ) Figure 6.5. Aboveground production (ANPP) as a function of foliage biomass and site index (Equation 6.3, n=72). Natural stands will probably not occur much beyond the site-specific optimum foliage biomass (see Figure 6.6 and Discussion). 191 figure, as discussed in Section 6.4.2). The function also suggests that this optimum foliage biomass increases with site index, e.g. 8.3 and 13.3 Mg ha\"1 for site index 20 and 40, respectively. From equation 6.3, the optimum foliage biomass FBopt can be calculated as a function of SI, such that FBopt = 3.398 + 0.247 * SI [6.4] This equation can be used to establish the difference between actual and optimum foliage biomass. Foliage biomass in eight of the 12 plots appears to be converging towards the site-specific optimum (Figure 6.6). Two plots are diverging from the optimum foliage biomass, and the two plots of Installation 72 have stable foliage biomass below the expected optimum. 6.3.3 Production allocation versus site index The proportion of total aboveground biomass, which is represented in each of the main aboveground biomass components, changes with site index for all components except stembark (Table 6.1, Figure 6.7). Foliage and branch biomass represent a larger proportion of total aboveground biomass on sites with lower site index than on sites with higher site index. The proportion of total biomass in stemwood and coarse root biomass increases with increasing site index. Similar trends are observed in the change of proportions of total aboveground production allocated to the main biomass components (Table 6.1, Figure 6.8). As site index increases, a smaller proportion of total aboveground production is allocated to foliage and an increasing proportion to stemwood and stembark. Coarse roots and branches showed no significant change in allocation. 192 Figure 6.6. The difference between actual foliage biomass (FB) and site-specific optimum foliage biomass (FBopt) for each of the 12 plots. Installation numbers are in upper right corner of each graph. Legend as in Figure 4.6. 193 (0 (0 (0 E XI c o & > O < C o o a o OJO 0.1 s :-o . i o 0.06 ;-0.00 0.0 0.3 0.20 0.16 h 0.10 0.06 0.00 0.3 0.1 0.0 16 Stembark 25 35 Site Index (m at 50 years) 45 Figure 6.7. Proportions of aboveground biomass allocated to foliage, branches, stemwood, stembark, and coarse roots plotted against site index. 194 0.4 QL GL z < km o a o 0.4 0.3 H OJJ 0.1 0.0 16 0.16 0.10 0.06 0.00 Branches 28 36 45 16 26 35 45 Site Index (m at 50 years) Figure 6.8. Proportions of aboveground production allocated to foliage, branches, stemwood, stembark, and coarse roots plotted against site index. 195 Total production is partitioned between aboveground and belowground stand components. Belowground production in this study is the sum of coarse, small (2-5mm), and fine (0-2mm) root production. Fine and small root production data are only available for 5 of the 12 plots (cf. Chapter 5). As has been discussed in Chapter 5, there are two measures of fine root turnover: production and mortality. In a steady state system, the two estimates of fine root turnover should be approximately equal, otherwise fine root biomass would either disappear or increase indefinitely. The estimates of root production and mortality obtained in this study differed in some plots, with mortality estimates exceeding those of production (Figures 5.14 and 5.15) for reasons explained in Chapter 5. Therefore, results of belowground allocation are reported for both root production and root mortality estimates. Annual aboveground production (ANPP) increases significantly with site index (r2=0.791, p<0.001, n=12) (Figure 6.9, open circles). Estimates of total production (TNPP) have been obtained by adding four different estimates of belowground production to ANPP (Figure 6.9 A-D, filled squares). The increase of TNPP with increasing site index is statistically significant for both production and mortality-based estimates of belowground allocation (p=.035 to .038 and p=.053 to .062, respectively). The proportion of total production allocated belowground ranged from about 0.25 to about 0.5, depending on site conditions and computational method used (Table 6.2). This proportion tended to decrease with increasing site index, regardless whether the estimates of belowground allocation were derived from fine and small root production or mortality estimates, and regardless of the computational method used (Table 6.3). This decrease of the proportion of total 196 Figure 6.9. Aboveground (O, ANPP) and total (•, TNPP) annual production increase with site index. The statistics (r2, p) refer to the relationship between TNPP and site index (n=5). The relationship between ANPP and site index is highly significant (r2=.791, p<.001, n=12). Fine and small root turnover estimates are based on A: Production DM-method, B: Production AC-method, C: Mortality DM-method, D: Mortality AC-method. See text for further explanations. 197 production allocated belowground with increasing site index was significant for mortality derived estimates of fine and small root turnover (AC method) (p=0.038). If fine and small root turnover was calculated from mortality estimates derived with the DM method, the decrease in the proportion of total production allocated belowground was less significant (p=0.054). If the proportion of total production allocated belowground was calculated from root turnover estimates based on fine and small root production data, the same trend was observed but neither computational method yielded a significant relationship. 198 Table 6.1 The relationship between site index and the proportion of total biomass and total production allocated to foliage, branches, stemwood, stembark, and coarse roots. For all models n=12. r2 = coefficient of determination, p = significance. Component Intercept Slope r 2 P BIOMASS Foliage 0.106 -0.0020 0.753 <0.001 Branches 0.113 -0.0015 0.587 0.004 Stemwood 0.660 0.0033 0.756 <0.001 Stembark 0.121 0.0002 0.071 0.401 Coarse Roots 0.175 0.0012 0.573 0.004 PRODUCTION Foliage 0.448 -0.0084 0.696 <0.001 Branches 0.280 -0.0028 0.226 0.118 Stemwood 0.236 0.0097 0.633 0.002 Stembark 0.035 0.0016 0.675 0.001 Coarse Roots 0.119 0.0010 0.278 0.078 199 Table 6.2 The proportion of total production allocated belowground for five Douglas-fir stands. Fine and small root production (P) and mortality (M) estimates have been derived with two computational methods: Decision Matrix (DM) and All Changes (AC). Production-based Mortality-based Inst. Plot SI DM AC DM AC 2 6 27.6 0.461 0.454 0.492 0.486 5 10 29.4 0.260 0.259 0.342 0.342 16 2 29.4 0.374 0.350 0.450 0.432 71 14 23.3 0.375 0.308 0.506 0.465 72 14 41.0 0.300 0.255 0.280 0.233 200 Table 6.3 The relationship between site index and the proportion of total production allocated belowground for four different estimates of belowground production. Fine and small root turnover estimates derived from production- or mortality-based estimates using the decision matrix (DM) and all changes (AC). For all models n=5. r2 = coefficient of determination, p = significance. Method Intercept Slope r 2 Production-based: DM 0.520 -0.005 0.216 0.430 AC 0.488 -0.005 0.190 0.464 Mortality-based: DM 0.810 -0.013 0.761 0.054 AC 0.821 -0.014 0.807 0.038 201 6.4 DISCUSSION 6.4.1 Foliage biomass versus site index During stand development, foliage biomass of forest stands reaches a site-specific maximum value after canopy closure, which in temperate coniferous forests averages approximately 10 Mg ha'1 (Turner and Long 1975). The stand age at which this plateau of foliage biomass is reached depends on factors such as site quality and stand density (Turner and Long 1975). The total amount of foliage biomass in a stand at the time of stabilization is thought to depend on site conditions such as soil moisture and soil nutrient availability (Grier et al. 1986). Foliage biomass regressions which use dbh as the independent variable are not able to predict a stabilization of stand foliage biomass, and will instead predict continuing foliage biomass accumulation as stand basal area increases. None of the models investigated in Chapter 3 which use dbh alone, or a derivative thereof, had an inflexion point at greater stem diameters, which means that foliage biomass per tree predicted by such models continues to increase with increasing diameter. In contrast, models which use sapwood basal area or dbh plus a competition index as independent variables can potentially predict the stabilization of foliage biomass after canopy closure. The relationship between foliage biomass and site index was not significant when foliage biomass was calculated using the regression model developed in this study, which uses a competition index as the independent variable in addition to dbh. In contrast, foliage biomass estimates obtained with the regional model of Gholz et al. (1979) increased significantly with increasing site index. A possible explanation for this difference is that the regional model predicts a continuing increase of foliage biomass with greater stand basal area. Because stand basal area in the twelve stands is highly correlated with site index (r2=.686, p=0.001), the 202 correlation between site index and foliage biomass also becomes significant. In contrast, the GSI-based model predicts that foliage biomass stabilizes in some plots, despite continuing increases in stand basal area. The two regression models also differ in that the regional model consistently predicted higher foliage biomass values for each of the plots, and it predicted a greater range of values (10.8 to 17.7 Mg ha*1) than the the GSI-model (8.6 to 13.5 Mg ha\"1). The GSI-based model has been derived from destructively sampled trees from plots in the immediate vicinity of the stands to which the regression model was applied. The regional model overpredicted foliage biomass of the sample trees by an average of 55.7% (Table 3.16) and is therefore not suitable for the prediction of foliage biomass of the stands used in this study. Warnings against using regression models which are based on dbh only for the prediction of stand foliage biomass have been presented earlier (Marshall and Waring 1986). Furthermore, in studies that use such models, all variables that are correlated with basal area will also be correlated with foliage biomass (Satoo and Madgwick 1982, p.64). 6.4.2 Foliage efficiency (ANPP) versus site index and foliage biomass Foliage efficiency (ANPP) of a plant or tree canopy decreases with increasing biomass of the entire canopy because the amount of light that reaches its lower portions diminishes due to self-shading (Cannell 1979). This inverse relationship between foliage biomass and efficiency, calculated from total aboveground production, was not detected in this study when only data from the 12 stands and one measurement period were analysed (Figure 6.2A). Foliage efficiency increases with site index (Figure 6.2B) and the relationship displayed in Figure 6.2A is therefore confounded with site index. 203 Analysing the data from each plot separately showed that in five of the 12 plots, foliage efficiency decreased significantly with increasing foliage biomass. In five additional plots, foliage efficiency was negatively correlated with foliage biomass, but the relationship was not statistically significant. In the two plots of Installation 72, foliage biomass did not change over time and the observed variation in foliage efficiency is probably related to other factors, such as climatic variability. The two plots of each Installation tend to show very similar patterns (Figure 6.3) of changes in foliage efficiency with changing foliage biomass. Foliage biomass is confounded with time, and the observed similarities in the pairs of plots of each Installation could be interpreted as an expression of the influences of site or regional climate. The regression model (Equation 6.3) which predicts foliage efficiency from foliage biomass and site index accounts for 56.5% of the variation in the 72 data points (Figure 6.4). Climatic variability, and other factors which affect foliage efficiency but are not represented in the model, account for the remaining variation. This model also explains the lack of correlation between foliage efficiency and foliage biomass observed in Figure 6.2. Equation 6.3 (Figure 6.5) should not be extrapolated beyond the range of the data. At low stand foliage biomass values, very little self-shading will occur and foliage efficiency will not increase further but will asymptotically approach some maximum value as most of the foliage approaches light saturation. At very high foliage biomass values, an increasing proportion of the foliage would have to be maintained at light levels which are inadequate to support positive net-photosynthesis. While it is often stated that foliage which is a net sink for carbon will be shed (Cannell 1979) or will adapt to the low light intensities by reducing respiration rates (Loomis and Gerakis 1975), the small growth improvement following pruning of the lower 25 to 30% of live branches has been attributed to the removal of a net carbon sink (Stein 1955). 204 6.4.3 Optimum foliage biomass in Douglas-fir stands Scientists concerned with crop production have shown that there is an optimum foliage biomass or leaf area at which production is maximized (Watson 1958, Zavitkovski et aZ.1974). Respiration increases with greater foliage biomass while the rate of increase in photosynthesis diminishes due to increasing self-shading. Waring et aZ.(1980) showed that in five Douglas-fir stands of different densities, basal area growth rate was highest at a leaf area index of approximately 7. In their study, leaf area was calculated from sapwood basal area assuming that this relationship is not affected by stand density, an assumption that has since been demonstrated to be incorrect (Brix and Mitchell 1983, Keane and Weetman 1987, Long and Smith 1988). The results of this study suggest that there is an optimum foliage biomass at which aboveground production is maximized (Figure 6.5). They further demonstrate that this optimum foliage biomass increases with site index, i.e. the optimum amount of foliage biomass is greater for a better site than for a poorer site. There are at least two explanations for this observation. In Douglas-fir, a determinant species, the distance between whorls increases with site index because of greater height growth rates at better sites. Increasing the vertical spacing of foliage biomass allows for greater intrusion of diffuse light into the lower parts of the canopy, which results in greater total photosynthesis (Kira 1975, Cannell 1979, Jarvis and Leverenz 1983). The light extinction coefficient, which determines reductions in light intensity as a function of foliage biomass, is affected by the spatial arrangement of the foliage (Monsi and Saeki 1953). A second argument is that in Douglas-fir, foliage nitrogen concentrations are positively correlated with site index (R.E. Carter and K. Klinka, pers. comm.), as has been observed with other conifers such as black spruce (Gagnon 1964). Foliage 205 with higher N-concentration has a greater photosynthetic efficiency and compensation points at lower light intensities (Kuroiwa 1960, Brix 1981). The optimum foliage biomass will therefore increase with greater soil nitrogen availability. Equation 6.3 has been used to predict the optimum foliage biomass as a function of site index for each of the 12 plots. Figure 6.6 shows that most of the 12 plots of this study asymptotically approach the site-specific optimum foliage biomass. However, two plots are clearly above the expected optimum foliage biomass. Installation 2 Plot 6 and Installation 5 Plot 8 have stand densities of 3000 and 3400 stems per hectare, respectively. These two plots also have the highest component of western hemlock and western redcedar: 33.5 and 24% of basal area, in Installation 2, Plot 6, and Installation 5, Plot 8, respectively. Because no regression models for the prediction of western hemlock or western redcedar foliage biomass have been developed in this study, the regional models of Gholz et al. (1979) were used to predict foliage biomass of these species. The higher than expected foliage biomass of these two plots may be attributable to overprediction by the regional models which are not sensitive to stand densities, as discussed above. A stable but lower than expected foliage biomass was observed in Installation 72, the site with the highest site index (41 m@50) and also the highest stand age (70 years). In the two plots of Installation 72, foliage biomass was about 25% below the expected value. There are several possible explanations for this. The trees in these stands are very tall (45 to 55 meters) and during storm events their crowns will swing considerable distances. The resulting crown contact between trees is known to cause large branches or branch parts to be broken off, which can be a potentially significant cause of foliage loss (Grier 1988). The stands may therefore not be able to maintain the optimum amount of foliage biomass. 206 6.4.4 Production allocation versus site index The change with site index in the proportion of total biomass represented in the major aboveground biomass components is to some extent due to the influence of the two plots of Installation 72. These two plots are older than the remaining ten plots and have a lower stand density (Table 4.3). When the data from these two plots are removed, the changes in distribution among aboveground biomass components with changing site index become non-significant. The slopes maintain the same direction in all five models, however. The variablity in the remaining data, the reduction of the sample size to ten and of the site index range to about half its original spread, all contribute to the non-significance of the reduced model. The proportion of total aboveground production allocated to the major biomass components also changed with site index, although only changes in foliage, stemwood, and stembark were statistically significant. The three significant relationships were strongly influenced by Installation 72, because removing the data of this Installation resulted in non-significant slopes, although again, their direction did not change. Both biomass and production distribution are affected by stand age and stand density, the influence of which is not accounted for in the simple models which use only site index as the independent variable. The quantification of belowground biomass and biomass production has always been very difficult, which is reflected in the paucity of information about this biomass component. Newbould (1967) suggested that belowground production should be calculated using the equation: Aboveground Production _ ^ x Belowground Production j-g ^ Aboveground Biomass Belowground Biomass For a lack of better data, he also suggested that k be assumed to have a value of 1. More recently, Keyes and Grier (1981) have demonstrated that the proportion of total production allocated belowground changes with site quality, and that 207 therefore a constant ratio between aboveground and belowground production cannot be assumed. The results of this study strongly support the conclusions reached by Keyes and Grier (1981). Figure 6.10 contains the data from the five stands of this study and the two stands investigated by Keyes and Grier. The differences between production and mortality estimates of fine and small roots obtained in this study have been discussed above. The decrease in the proportion of total production allocated belowground with increasing site index is apparent in both production- or mortality-based estimates of belowground production. The production-based estimate of fine and small root turnover calculated with the decision matrix is significant at the 10% level (p=0.087). The two mortality-based estimates are highly significant, p=0.003 and 0.002 for the DM- and AC-based methods, respectively. Considering the differences in geographic location and methods used, the similarity of the estimates of allocation obtained by Keyes and Grier (1981) and in this study is striking. In this study, soil moisture regime (Green et al. 1984) of the twelve plots ranged from moderately dry to fresh and site index increased with improved soil moisture regime. Although Douglas-fir does not normally grow on wet or very wet sites, exceptions can be found as a result of off-site planting or due to other unusual circumstances. On wet sites, poor growth of Douglas-fir may result in a lower site index than on fresh sites (K. Klinka, pers. comm.). Growth limitations on these wet sites are different than those on drier sites which may, however, have the same site index. Some of the conclusions reached about the relationships between site index and carbon allocation may not apply to Douglas-fir stands on wet or very wet sites. The results of this study emphasize that shifts in carbon allocation from above- to belowground stand components are an important and frequently ignored 208 adaptation of forest ecosystems to changes in site quality. Future predictions of stand responses to silvicultural treatments and environmental change require a better understanding of the role of shifts in carbon allocation in forest ecosystems. 209 Figure 6.10 The partitioning of total production to above and belowground components for the 5 stands of this study (•) and the two stands of Keyes and Grier (1981) (A). Fine and small root turnover estimates are based on A: Production DM-method, B: Production AC-method, C: Mortality DM-method, D: Mortality AC-method. See text for further explanations. 210 7. SUMMARY AND CONCLUSIONS 1) Regression models for the prediction of aboveground biomass components of Douglas-fir that include a competition index in addition to dbh, are significant improvements over models which use dbh only. Of the four competition indices tested, the Growing Space Index (GSI) (Lin 1974) was the best independent variable in addition to dbh for the prediction of foliage, branchwood, and stemwood biomass. 2) Foliage biomass predicted with the GSI model appears to reach a steady state in some Douglas-fir stands, despite the continuing increase in basal area in these stands. Foliage biomass regression models based on dbh only do not predict this stabilization of foliage biomass. 3) In 1985, aboveground biomass in 12 Douglas-fir stands, with site index 19.5 to 41.3 m at 50 years and age 32 to 70 years, ranged from 135 to 573 Mg ha-1. The aboveground biomass was distributed among the major components as follows: foliage 2 - 6%, branches 4 - 9%, stemwood 73 - 81%, and stembark 12 - 13%. Coarse root biomass was equal to 20 - 23% of the aboveground biomass. 4) In the period between 1985 and 1987, annual aboveground production (ANPP) in these stands ranged from 4.7 to 16.0 Mg ha-1 year1. Aboveground production was distributed among the major components in the following proportions: foliage 9 - 31%, branches 12 -23%, stemwood, 42 - 68%, stembark 7 - 11%. Coarse root production was equal to 13 - 16% of the aboveground production. 211 5) Live fine root (0-2mm) biomass showed a similar seasonal pattern of variation in all five stands that were sampled. A maximum value was reached in May or June, and a minimum value in August or October. 6) Biomass of living fine roots (0-2 mm) in May 1985 ranged from 1.82 to 7.91 Mg ha\"1, and that of living small roots (2-5 mm) ranged from 0.59 to 4.10 Mg ha\"1. 7) Different estimates were obtained for production and mortality in both fine and small roots. Estimates derived using the decision-matrix ranged from 1.12 to 5.14 Mg ha\"1year1 for fine root production, and from 2.15 to 4.89 Mg ha'fyear1 for fine root mortality. Small root production and mortality estimates ranged from 0.51 to 2.22 and from 0.88 to 2.13 Mg ha\" Vear1, respectively. The differences in fine root turnover estimates derived from fine root production and mortality data are probably attributable to the unusually dry conditions during the summer of 1985. In particular, the lower site index stands were not able to replace by the subsequent spring all living fine roots which had died during the summer. 8) A highly significant regression model (Equation 6.1), which uses stand foliage biomass and site index as independent variables, accounted for 56.5% of the variation in foliage efficiency (based on ANPP) observed in 72 data points which represented 12 plots and 6 measurement periods. The model suggests that foliage efficiency decreases with increasing stand foliage biomass and increases with increasing site index. 9) Based on the above model, it would appear that there is an optimum foliage biomass at which total aboveground production is maximized. This optimum foliage biomass appears to increase with increasing site index. 212 10) With increasing site index, a decreasing proportion of total production is allocated to belowground stand components. The site with the lowest site index allocated from 31 to 51% of total production to belowground components while the site with the highest site index allocated from 23 to 30%. The ranges in estimates are due to the differences between production and mortality estimates of fine and small roots, as discussed above. 11) The relationship between the proportion of total production allocated belowground and site index that was observed in the five stands of this study strongly supports a similar observation by Keyes and Grier (1981) in two Douglas-fir stands of high and low site quality. 12) Future predictions of the responses of forest ecosystems to environmental change and silvicultural treatments should be based on a quantitative understanding of shifts in carbon allocation patterns in forest stands. 213 8. REFERENCES Agren, G.I., Axelsson, B., Flower-Ellis, J.G.K., Linder, S., Persson, H., Staaf, H., and Troeng, E. 1980. Annual carbon budget for a young Scots pine. In Structure and Function of a Northern Coniferous Forest - An Ecosystem Study. Edited by T. Persson. Ecol. Bull. (Stockholm) 32: 307-313. Agriculture Canada Expert Committee on Soil Survey. 1987. The Canadian System of Soil Classification. Agric. Can. Publ. 1646, 164 pp. Ahmed, S.S. 1956. The effect of application of artificial fertilizer on height, radial, and branch growth of 35-year old Douglas-fir. M.F. Thesis. Univ. Washington, Seattle. Albrektson, A. 1980. Relations between tree biomass fractions and conventional silvicultural measurements. In Structure and Function of Northern Coniferous Forests - An Ecosystem Study. Edited by T. Perrsson. Ecol. Bull. (Stockholm) 32: 315-327. Albrektson, A. 1984. Sapwood basal area and needle mass of Scots pine (Pinus sylvestris L.) trees in central Sweden. Forestry 57:35-43. Alexander, I.F. and Fairley, R.I. 1983. Effects of N fertilisation on populations of fine roots and mycorrhizas in spruce humus. Plant and Soil 71: 49-53. Arney, J.D. 1973. Tables for quantifying competitive stress on individual trees. Can. For. Serv., Pacific Research Centre, Victoria B.C., Inf. Rep. BC-X-78, 10 pp. Axelsson, E. and Axelsson, B. 1986. Changes in carbon allocation patterns in spruce and pine trees following irrigation and fertilization. Tree Physiology 2:189-204. Barclay, H.J. and Brix, H. 1985. Fertilization and thinning effects on a Douglas-fir ecosystem at Shawnigan Lake: 12-year growth response. Canadian Forestry Service, Pacific Forestry Centre, Information Report BC-X-271, 34 pp. Barclay, H.J. and Hall, T.H. 1986. SHAWN: a model of Douglas-fir ecosystem response to nitrogen fertilization and thinning: a preliminary approach. Canadian Forestry Service, Pacific Forestry Centre, Information Report BC-X-280,30pp. Barclay, H.J., Pang, P.C., and Pollard, D.F.W. 1986. Aboveground biomass distribution within trees and stands in thinned and fertilized Douglas-fir. Can. J. For. Res. 16: 438-442. Baskerville, G.L. 1972. Use of logarithmic regression in the estimation of plant biomass. Can. J. For. Res. 2: 49-53. Baskerville, G.L. 1983. Balsam fir foliage regressions are age dependent. Can. J. For Res. 13: 1248-1251. Bella, I.E. 1971. A new competition model for individual trees. For. Sci. 17: 364-372. 214 Bohm, W. 1979. Methods of studying root systems. Ecological Studies: Analysis and Synthesis 33. Springer Verlag, Berlin, Heidelberg, 188 pp. Borghetti, M., Vendramin, G.G., and Gianni. R. 1987. Specific leaf area and leaf area index distribution in a young Douglas-fir plantation. Can. J. For. Res. 16:1283-1286. Bossel, H. 1986. Dynamics of forest dieback: systems analysis and simulation. Ecol. Model. 34: 259-288. Bray, J.R. and Gorham, E. 1964. Litter production in forests of the world. Adv. Ecol. Res. 2:101-157. Brix, H. 1981. Effects of nitrogen fertilizer source and application rates on foliar nitrogen concentration, photosynthesis, and growth of Douglas-fir. Can. J. For. Res. 11:775-780. Brix, H. 1983. Effects of thinning and nitrogen fertilization on growth rate of Douglas-fir: relative contribution of foliage quantity and efficiency. Can. J. For. Res. 13:167-175. Brix, H. and Mitchell, A.K. 1983. Thinning and nitrogen fertilization effects on sapwood development and relationships of foliage quantity to sapwood area and basal area in Douglas-fir. Can. J. For. Res. 13: 384-389. Brix, H. in press. Mechanisms of response to fertilization. II Utilization by trees and stands. In Improving Forest Fertilization Decision-making in British Columbia. Edited by D. Lousier. B.C. Ministry of Forests, Vancouver, B.C., FRDA Report, in press. Brown, J.K. 1978. Weight and density of Rocky Mountain conifers. USDA, Forest Service, Odgen, Utah, Res. Pap. INT-197. Bruce, D. 1981. Consistent height-growth and growth-rate estimates for remeasured plots. For. Sci. 27: 711-725. Burger, H. 1935. Holz, Blattmenge und Zuwachs. II Mitteilung. Die Douglasie. Mitt. Schweiz. Anst. Forstl. Versuchswesen 19: 21-72. Cannell, M.G.R. 1979. Improving per hectare forest productivity. In Fifth North American Forest Biology Workshop. Edited by CA. Hollis and A.E. Squillace. University of Florida, School of Forest Resources, Gainesville, USA, pp. 120-148. Cannell, M.G.R. 1982. World Forest Biomass and Primary Production Data. Academic Press, London, New York, 391 pp. Carter, R.E., Miller, I.M., and Klinka, K. 1986. Relationships between growth form and stand density in immature Douglas-fir. Forestry Chronicle Oct. 1986: 440-445. Chapin, F.S.,111, Bloom, A.J., Field, C.B., and Waring, R.H. 1987. Plant response to multiple environmental factors. Bioscience 37: 49-57. 215 Comeau, P.G. 1986. Net primary production of lodgepole pine (Pinus contorta var. latifolia Engelm.) in some ecosystems in southern British Columbia. Ph.D. thesis, Faculty of Forestry, University of British Columbia, Vancouver, BC. Coopersmith, D.J. 1986. Root and fungal biomass production in low, medium and high productivity second growth Douglas-fir stands on Vancouver Island. M. Sc. thesis, Faculty of Forestry, University of British Columbia, Vancouver, B.C., 229 pp. Coutts, M.P. and Philipson, J.J. 1976. The influence of mineral nutrition on the root development of trees. I. The growth of Sitka spruce with divided root systems. J Exp. Bot. 27: 1102-1111. Crown, M. and Brett, CP. 1975. Fertilization and thinning effects on a Douglas-fir ecosystem at Shawnigan Lake: an establishment report. Can. For. Serv., Pac. For. Res. Cent. Inf. Rep. BC-X-110,45 pp. Cuevas, E. and Medina, E. 1983. Root production and organic matter decomposition in a Tierra Firme forest of the upper Rio Negro basin. In Root Ecology and its Practical Application. A contribution to the Investigation of the whole plant. Edited by W. Bohm, L. Kutschera, and E. Lichtenegger. Bundesanst. alpenl. Landwirtschaft, Irdning, pp. 653-666. Daniels, R.F., Burkhart, H.E., and Clason, T.R. 1986. A comparison of competition measures for predicting growth of loblolly pine trees. Can. J. For. Res. 16: 1230-1237. Darling, L.M. and Omule, S.A.Y. 1989. Extensive studies of fertilization and thinning coastal Douglas-fir and western hemlock: an establishment report. B.C. Min. For. FRDA Report 054, 22 pp. Deans, J.D. 1979. Fluctuations of the soil environment and fine root growth in a young Sitka spruce plantation. Plant and Soil 52: 195-208. Dice, S.F. 1970. The biomass and nutrient flux in a second growth Douglas-fir ecosystem: a study in quantitative ecology. Ph.D. thesis. Univ. Washington, Seattle, Wa., 166 pp. Drew, M.C, Saker, L.R., and Ashley, T.W. 1973. Nutrient supply and the growth of the seminal root system in barley. J. Exp. Bot. 24:1189- 1202. Ek, A.R., Shifely, S.R., and Burk, T.E., (eds.) 1988a. Forest Growth Modelling and Prediction. USDA For. Serv., Gen. Tech. Rep. NC-120, Vol 1: 1-579. Ek, A.R., Shifely, S.R., and Burk, T.E., (eds.) 1988b. Forest Growth Modelling and Prediction. USDA For. Serv., Gen. Tech. Rep. NC-120, Vol 2: 580-1149. Environment Canada. [1982]. Canadian Climatic Normals 1951-1980. Temperature and Precipitation, British Columbia. Environment Canada. Atmospheric Environment Service, 268 pp. Ericsson, A. and Persson, H. 1980. Seasonal changes in starch reserves and growth of fine roots of 20 year old Scots pine. In Structure and Function of Northern Coniferous Forests - An Ecosystem Study. Edited by T. Persson., Ecol. Bull. (Stockholm) 32: 239-250 pp. 216 Espiriosa Bancalari, M.A. and Perry, D.A. 1987. Distribution and increment of biomass in adjacent young Douglas-fir stands with different early growth rates. Can. J. For. Res. 17: 722-730. Fairley, R.I. and Alexander, I.J. 1985. Methods of calculating fine root production in forests. In Ecological Interactions in Soil. Edited by A. Fitter. British Ecological Society, Special Publication Number 4, pp. 37-42. Feller, M.C., Kimmins, J.P., and Scoullar, K.A. 1983. FORCYTE-10: Calibration data and simulation of long term effects of intensive forest management on site productivity, energy performance and energy benefit/cost ratio. In IUFRO Symposium on Forest Site and Continous Productivity. Edited by R. Ballard and S.P. Gessel, USDA For. Serv. Gen. Tech. Rep. PNW-163, pp. 179-200. Fogel, R. and Hunt, G. 1979. Fungal and arboreal biomass in a western Oregon Douglas-fir ecosystem: distribution patterns and turnover. Can. J. For. Res. 9: 245-256. Ford, E.D. and Bassow, S. 1988. Modelling the influence of pollution on allocation. In Forest Growth Modelling and Prediction, Edited by A.R. Ek, S.R. Shifely, and T.E. Burk., USDA For. Ser., Gen. Tech. Rep. NC-120, Vol 1: 450-457. Ford, E.D. and Deaiis, J.D. 1977. Growth of a Sitka spruce plantation: spatial distribution of length, weight and carbohydrate concentrations of fine roots. Plant and Soil 47: 463-485. Forest Products Laboratory. 1952. Methods of determining the specific gravity of wood. US For. Prod. Lab. Tech. Note No. B-14, (Revised 1956)., 6 pp. Forrest, W.G. 1966. The estimation of oven-dry weight. Aust. For. Res. 3: 41-46. Fox, D.J. and Guire, K.E. 1976. Documentation for MIDAS. Stat. Research Lab. U. of Michigan, 203 pp. Friend, A.L. 1988. Nitrogen stress and fine root growth of Douglas-fir. Ph.D Thesis, University of Washington. College of Forest Resources, Seattle, 131 pp. Gagnon, J.D. 1964. Relationship between site index and foliage nitrogen at two crown levels of mature black spruce. For. Chron. 4:169-174. Gerald, CF. and Wheatley, P.O. 1984. Applied Numerical Analysis, 3rd edition. Addison-Welsley Publishing Co., Reading. Gholz, H.L. 1982. Environmental limits on aboveground net primary production, leaf area, and biomass in vegetation zones of the Pacific Northwest. Ecology 63: 464-481. Gholz, H.L., Fitz, F.K., and Waring, R.H. 1976. Leaf area difference associated with old-growth forest communities in the western Oregon Cascades. Can. J. Res. 6:49-57. 217 Gholz, H.L., Grier, C.C., Campbell, A.G., and Brown, A.T. 1979. Equations and their use for estimating biomass and leaf area of Pacific Northwest trees, shrubs and herbs. Oregon State Univ., For. Res. Lab., Corvallis, Research Paper 41, 37 pp. Gill, R. and Lavender, D.P. 1983. Urea fertilization: effects on primary root mortality and mycorrhizal development of young-growth western Hemlock. For. Sci. 29: 751-760. Gislason, T.V. 1984. Fine root biomass and distribution in 50-year-old Douglas-fir stands of high and low productivity. B.S.F. thesis, Faculty of Forestry. University of British Columbia, Vancouver, B.C., 87 pp. Granier, A. 1981. Etude des relations entre la section du bois d'aubier et la masse foliare chez le Douglas (Pseudotsuga menziesii Mirb. Franco). Ann. Sci. for. 38: 503-512. Green, R.N., Courtin, P.J., Klinka, K., Slaco, R.J., and Ray, CA. 1984. Site diagnosis, tree species selection, and slashburning guidelines for the Vancouver Forest Region. Land Management Handbook 8, B.C. Ministry of Forests, Victoria, 143 pp. Grier, C.C 1988. Foliage loss due to snow, wind and winter drying damage: its effects on leaf biomass of some western conifer forests. Can. J. For. Res. 18: 1097-1102. Grier, C.C. and Logan, R.S. 1977. Old-growth Pseudotsuga menziesii communities of a western Oregon Watershed: biomass distribution and production budgets. Ecol. Monograph 47: 373-400. Grier, C.C. and Running, S.W. 1977. Leaf area of mature north-western coniferous forests: relation to site water balance. Ecology 58: 893-899. Grier, C.C. and Waring, R.H. 1974. Conifer foliage mass related to sapwood area. For. Sci. 20: 205-206. Grier, C.C, Hinkley, T.M., Vogt, K.A., and Gower, S.T. 1986. Net primary production in Douglas-fir ecosystems: its relation to moisture and mineral nutrition. In Douglas-fir: Stand Management for the Future. Edited by C.D. Oliver, D.P. Hanley, and J.A. Johnson. University of Washington, Seattle, Wa., pp. 154-161. Grier, C.C, Lee, K.M., and Archibald, R.M. 1984. Effect of urea fertilization on allometric relations in young Douglas-fir. Can. J. For. Res. 14: 900-904. Hagglund, B. 1981. Evaluation of forest site productivity. Forestry Abstracts 42: 515-527. Harris, W.F., Kinerson, R.S., and Edwards, N.T. 1977. Comparison of belowground biomass of natural decideous forests and loblolly pine plantations. Pedobiologia 17: 369-381. Harris, W.F., Santantonio, D., and McGinty, D. 1980. The dynamic belowground ecosystem. In Forests: Fresh Perspectives From Ecosystem Analysis. Edited by R.H. Waring. Oregon State University Press, Corvallis, Or., pp. 119-129. 218 Hegyi, F. 1974. A simulation model for managing jack pine stands, in Growth Models for Tree and Stand Simulation, Edited by J. Fries., Royal College of Forestry, Stockholm, Research Note No. 30, pp. 74-90. Heilman, P.E. 1961. Effects of nitrogen fertilization on the growth and nitrogen distribution of low-site Douglas-fir stands. Ph. D. thesis. Univ. Washington, Seattle. Helgerson, O.T., Cromack, K., Stafford, S., Miller, R.E., and Slagle, R. 1988. Equations for estimating aboveground components of young Douglas-fir and red alder in a coastal Oregon plantation. Can. J. For. Res. 18:1082-1085. Helms, J.A. 1965. Diurnal and seasonal patterns of net assimilation in Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco), as influenced by environment. Ecology 46: 698-708. Helms, J.A. 1976. Factors influencing net photosynthesis in trees: An ecological viewpoint. In Tree Physiology and Yield Improvement. Edited by M.G.R. Cannel and F.T. Last. Academic Press, London, pp. 55-78. Hermann, R.K. 1977. Growth and production of tree roots: a review. In The Belowground Ecosystem: A Synthesis of Plant Associated Processes. Edited , by J.K. Marshall. Range Science Department, Colorado State University, Fort Collins, Col., Science Series No. 26, pp. 7-27. Husch, B., Miller, C.I., and Beers, T.W. 1982. Forest Mensuration, 3 rd edition. John Wiley and Sons, New York, N.Y., 402 pp. Ingestad, T. 1979. Nitrogen stress in birch seedlings. II. N, K, P, Ca and Mg nutrition. Physiol. Plant. 45:149-157. Ingestad, T. 1981. Nutrition and growth of birch and grey alder seedlings in low conductivity solutions and at a varied relative rate of nutrient addition. Physiol. Plant. 52: 454-466. Ingestad, T. and Kahr, M. 1985. Nutrition and growth of coniferous seedlings at varied relative nitrogen addition rate. Physiol. Plant 65:109-116. Ingestad, T. and Lund, A.B. 1979. Nitrogen stress in birch seedlings. I. Growth technique and growth. Physiol. Plant. 45:137-148. Jarvis, P.G. and Leverenz, J.W. 1983. Productivity of temperate, deciduous and evergreen forests. In Encyclopedia of Plant Ecology IV. Ecosystem Processes: Mineral Cycling, Productivity and Man's Influence. Edited by O.L. Lange, P.S. Nobel, C.B. Osmond, and H. Ziegler. Springer-Verlag, Berlin, pp. 234-280. Jozsa, L.A., Richards, J., and Johnson, S.G. 1989. Relative density. In Second Growth Douglas-fir: Its Management and Conversion for Value. Edited by R.M. Kellogg, Special Publication SP-32, Forintek Canada Corp., pp 5-22. Kay, M. 1978. Foliage biomass of Douglas-fir in a 53-year-old plantation. N. Z. J. For. Sci. 8: 315-326. 219 Keane, M.G. and Weetman, G.F. 1987. Leaf area sapwood cross-sectional area relationships in repressed stands of Lodgepole pine. Can. J. For. Res. 17: 205-209. Keyes, M.R. and Grier, C.C. 1981. Above- and below-ground net production in 40-year-old Douglas-fir stands on low and high productivity sites. Can. J. For. Res. 11:599-605. Kimmins, J.P. and Hawkes, B.C. 1978. Distribution and chemistry of fine roots in a white spruce-subalpine fir stand in British Columbia: implication for management. Can. J. For. Res. 8: 265-279. Kimmins, J.P., Scoullar, K.A., Bigley, R.E., Kurz, W.A., Comeau, P.G., and Chatarpaul, L. 1986. Yield prediction models: the need for a hybrid ecosystem-level approach incorporating canopy function and architecture. In: Crown and Canopy Structure in Relation to Productivity. Edited by T. Fujimori and D. Whitehead. Forestry and Forest Products Research Institute, Ibaraki, Japan, pp. 26-48. King, J.E. 1966. Site index curves for Douglas-fir in the Pacific Northwest. Weyerhaueser Co. For. Res. Cent., Forest Paper 8: 50. Kira, T. 1975. Primary production of forests. In Photosynthesis and Productivity in Different Environments, IBP Programme 3. Edited by J.P. Cooper. Cambridge Univ. Press, London, pp. 5-40. Kittredge, J. 1944. Estimation of the amount of foliage of trees and stands. J. For. 42: 905-912. Kramer, P.J. and Kozlowski, T.T. 1979. Physiology of Woody Plants. Academic Press, New York, 811 pp. Kuroiwa, S. 1960. Ecological and physiological studies on the vegetation of Mt. Shimagare IV: Some physiological functions concerning matter production in young Abies trees. Bot. Mag. Tokyo 73:133-141. Kurucz, J. 1969. Component weights of Douglas-fir, western hemlock and western red cedar biomass for simulation of amount and distribution of forest fuels, M. F. thesis. Univ. of B.C., Vancouver, B.C., 116 pp. Kurz, W.A. 1989. The significance of shifts in carbon allocation to long-term site productivity research. In Research Strategies for Long-term Site Productivity. Proc, International Energy Agency BE A3 Workshop, Seattle, WA. August 1988. Edited by W.J. Dyck and CA. Mees. IEA/BE A3 Report No. 8., in press. Kurz, W.A. and Kimmins, J.P. 1987. Analysis of some sources of error in methods used to determine fine root production in forest ecosystems: a simulation approach. Can. J. For. Res. 17: 909-912. Krajina, V.J. 1969. Ecology of forest trees in British Columbia. Ecol. Western N. America 2:1-146. 220 Lassoie, J.P., Hinkley, T.M., and C.C. Grier. 1985. Coniferous forests of the Pacific Northwest. In Physiological Ecology of North American Plant Communities, Edited by B.F. Chabot and H.A. Mooney, Chapman and Hall, New York, pp. 127-161. Lavigne, M.B. 1988. Growth and net assimilation rates in thinned and unthinned stands of balsam fir. Can. J. For. Res. 18:1205-1210. Lin, J.Y. 1974. Stand growth simulation models for Douglas-fir and Western Hemlock in the northwestern United States. In Growth Models for Tree and Stand Simulation. Edited by J. Fries. Royal College of Forestry, Stockholm, Research Note 30, pp. 102-118. Linder, S. and Axelsson, B. 1982. Changes in carbon uptake and allocations pattern as a result of irrigation and fertilization in a young Pinus sylvestris stand. In Carbon Uptake and Allocation in Subalpine Ecosystems as a Key to Management, Edited by R.H. Waring. For. Res. Lab., Oregon State Univ., Corvallis, Or. pp. 38-44. Long, J.N. and Smith, F.W. 1988. Leaf area-sapwood area relations of lodgepole pine as influenced by stand density and site index. Can. J. For. Res. 18: 247-250. Loomis, R.S. and Gerakis, P.A. 1975. Productivity of agricultural ecosystems. In Photosynthesis and Productivity in Different Environments. Edited by J.P. Cooper. Cambridge University Press, pp. 145-172. Madgwick, H.A.I. 1983. Above-ground weight of forest plots - comparison of seven methods of estimation. N.Z. J. of For. Sci. 13:100-107. Makela, A. 1986. Implications of the pipe model theory on dry matter partitioning and height growth in trees. J. Theor. Biol. 123:103-120. Makela, A. 1988. Parameter estimation and testing of a process-based stand growth model using Monte-Carlo techniques. In Forest Growth Modelling and Prediction. Edited by A.R. Ek, S.R. Shifley, and T.E. Burk., N. Central Forest Exp. Stn., Gen. Tech. Rep. NC-120: 315-322. Makela, A. and Hari, P. 1986. A stand model based on carbon uptake and allocation in individual trees. Ecol. Mod. 33: 205-230. Marshall, J.D. and Waring, R.H. 1986. Comparison of methods of estimating leaf-area index in old-growth Douglas-fir. Ecology 67: 975-979. McClaugherty, C.A., Aber, J.D., and Melillo, J.M. 1982. The role of fine roots in the organic matter and nitrogen budgets of two forested ecosystems. Ecology 63: 1481-1490. McMinn, R.G. 1963. Characterization of Douglas-fir root systems. Can. J. Bot. 41: 105-122. McMurtrie, R. and Wolf, L. 1983. Above- and below-ground growth of forest stands: a carbon budget model. Ann. Bot. 52:437-448. 221 Meyer, F.H. 1967. Feinwurzelverteilung von Waldbaumen in Abhangigkeit vom Substrat. Forstarchiv 38: 286-290. Mitchell, K.J. 1975. Dynamics and simulated yield for Douglas-fir. For. Sci. Monograph 17: 39 pp. Mitchell, K.J. and Polsson, K.R. 1988. Site index curves and tables for British Columbia: Coastal species. Cdn. For. Serv. and B.C. Min. of For. FRDA Rep. 037: 29 pp. Mohren, G.M.J. 1987. Simulation of forest growth, applied to Douglas-fir stands in the Netherlands, Dissertation. University of Wageningen, Netherlands, 184 pp. Moir, W.H. and Bachelard, E.P. 1969. Distribution of fine roots in three Pinus radiata plantations near Canberra, Australia. Ecology 50: 658-662. Monsi, M. and Saeki, T. 1953. Uber den Lichtfaktor in den Pflanzengesellschaften und seine Bedeutung fur die Stoffproduktion. Jap. J. Bot. 14: 22-52. Nadelhoffer, K.J., Aber, J.D., and Melillo, J.M. 1985. Fine roots, net primary production and soil nitrogen availability: a new hypothesis. Ecology 66: 1377-1390. Newbould, P.J. 1967. Methods for Estimating the Primary Production of Forests. IBP Handbook No.2, London, 62 pp. Noone, C.S. and Bell, J.F. 1980. An evaluation of eight intertree competition indices, Forest Research Lab., Oregon State University, Corvallis, Or., Research Note 66, 6 pp. Persson, H. 1978. Root dynamics in a young Scots pine stand in Central Sweden. Oikos 30:508-519. Persson, H. 1980. Spatial distribution of fine-root growth, mortality and decomposition in a young Scots pine stand in central Sweden. Oikos 34: 77-87. Persson, H. 1983. The distribution and productivity of fine roots in boreal forests. Plant and Soil 71: 87-101. Philipson, J.J. and Coutts, M.P. 1977. The influence of mineral nutrition on the root development of trees: II. The effect of specific nutrient elements on the growth of individual roots of Sitka spruce. J. Exp. Bot. 28: 864-871. Pojar, J., Klinka, K., and Meidinger, D.V. 1987. Biogeoclimatic ecosystem classification in British Columbia. Forest Ecol. Manage. 22:119-154. Reichle, D.E.,(ed.). 1981. Dynamic Properties of Forest Ecosystems, International Biological Program 23. Cambridge University Press, 683 pp. Ritchie, M.W. and Hann, D.W. 1987. Equations for predicting height to crown base for fourteen tree species in southwest Oregon. Oregon State University, Forest Research Lab, Research Paper 50,14 pp. 222 Rost, T.L. and Jones, T.J. 1988. Pea root regeneration after tip excision at different levels: polarity of new growth. Ann. Bot. 61: 513-523. Running, S.W. 1976. Environmental control of leaf water conductance in conifers. Can. J. For. Res. 6: 104-112. Santantonio, D. and Hermann, R.K. 1985. Standing crop, production, and turnover of fine roots on dry, moderate, and wet sites of mature Douglas-fir in western Oregon. Ann. Sci. for. 42:113-142. Santantonio, D., Hermann, R.K., and Overton, W.S. 1977. Root biomass studies in forest ecosystems. Pedobiologia 17:1-31. Satoo, T. 1967. Efficiency and quantity of leaves of closed stands of Cryptomeria japonica as influenced by site quality. In 14th IUFRO Congress, Munich, Vol 2, pp. 396-404. Satoo, T.T. and Madgwick, H.A.I. 1982. Forest Biomass. M. Nijhoff/Dr. W. Junk Publ, The Hague, 152 pp. Shaw, D.L.,Jr. 1979. Biomass equations for Douglas-fir, western hemlock and red cedar in Washington and Oregon. In Proceedings, Forest Resource Inventories Workshop. Edited by W.E. Frayer. Colorado State University, Fort Collins, Colorado, pp. 763-781. Shinozaki, K., Yoda, K, Hozumi, K, and Kira, T. 1964a. A quantitative analysis of plant form - the pipe model theory I. Basic analysis. Jap. J. Ecol. 14: 97-105. Shinozaki, K, Yoda, K., Hozumi, K, and Kira, T. 1964b. A quantitative analysis of plant form - the pipe model theory II. Further evidence of the theory and its application in forest ecology. Jap. J. Ecol. 14:133-139. Singh, J.S., Lauenroth, W.K., Hunt, H.W., and Swift, D.M. 1984. Bias and random errors in estimators of net root production: a simulation approach. Ecology 65: 1760-1764. Smith, R.T. 1979. An ecological classification of installations in the forest productivity program. Internal Report. B.C. Ministry of Forests. Research Branch, Victoria, B.C., 100 pp. Smith, S.H. and Bell, J.F. 1983. Using competitive stress index to estimate diameter growth for thinned Douglas-fir stands. For. Sci. 29:491-499. Snell, J.A. and Brown, J.K. 1978. Comparison of tree biomass estimators DBH and sapwood area. For. Sci. 24: 455-457. Sprugel, D.G. 1983. Correction for bias in log-transformed allometric equations. Ecology 64: 209-210. St. John, T.V. 1983. Response of tree roots to decomposing organic matter in two lowland Amazonian rain forests. Can. J. For. Res. 13:346-349. Standish, J.T., Manning, G.H., and Derriaerschalk, J.P. 1985. Development of biomass equations for British Columbia tree species. Information Report BC-X-264, 48 pp. 223 Stein, W.L 1955. Pruning to different heights in young Douglas-fir. J. For. 53: 352-355. Swank, W.T. 1960. A quantitative analysis of vegetation on the Fern Lake Watershed. M.F. thesis. Univ. of Washington, Seattle. Tadaki, Y. 1966. Some discussions of the leaf biomass of forest stands and trees. Jap. For. Exp. Stn. Bull. 184:135-161. Teskey, R.O. and Hinckley, T.M. 1981. Influence of temperature and water potential on root growth of white oak. Physiol. Plant. 52: 363-369. Thornley, J.H.M. 1972. A model to describe the partitioning of photosynthate during vegetation growth. Ann. Bot. 36: 419-430. Turner, J. and Long, J.N. 1975. Accumulation of organic matter in a series of Douglas-fir stands. Can. J. For. Res. 5: 681-690. Vogt, K.A., Edmonds, R.L., and Grier, C.C. 1981. Seasonal changes in biomass and vertical distribution of mycorrhizal and fibrous textured conifer roots in 23-and 180-year-old subalpine Abies amabilis stands. Can. J. For. Res. 11: 223-229. Vogt, K.A., Edmonds, R.L., Grier, C.C, and Piper, S.R. 1980. Seasonal changes in mycorrhizal and fibrous-textured root biomass in 23- and 180- year-old Pacific silver fir stands in western Washington. Can. J. For. Res. 10: 523-529. Vogt, K.A., Grier, C.C, and Vogt, D.J. 1986. Production, turnover, and nutrient dynamics of above- and belowground detritus of world forests. Adv. Ecol. Res. 15:303-377. Vogt, K.A., Grier, C.C, Gower, S.T., Sprugel, D.C, and Vogt, D.J. 1986. Overestimation of net root production: a real or imaginary problem? Ecology 67:577-579. Vogt, K.A., Moore, E.E., Vogt, D.J., Redlin, M.J., and Edmonds, R.L. 1983. Conifer fine root and mycorrhizal root biomass within the forest floors of Douglas-fir stands of different ages and site productivities. Can. J. For. Res. 13:429-437. Vogt, K.A. and H. Persson. in press. Root methods. In: Techniques and Approaches in Forest Tree Ecophysiology. Edited by J.P. Lassoie and T.M. Hinkley. CRC Press, Boca Raton, Florida. Vogt, K.A., Vogt, D.J., Moore, E.E., Fatuga, B.A., Redlin, M.R., and Edmonds, R.L. 1987. Conifer and angiosperm fine-root biomass in relation to stand-age and site productivity in Douglas-fir forests. J. Ecol. 75: 857-870. Vogt, K.A., Vogt, D.J., Moore, E.E., Littke, W., Grier, C.C, and Leney, L. 1985. Estimating Douglas-fir fine root biomass and production from living bark and starch. Can. J. For. Res. 15:177-179. Waring, R.H. and Schlesinger, W.H. 1985. Forest Ecosystems. Academic Press, Orlando, Florida, 340 pp. 224 Waring, R.H., Thies, W.G., and Muscato, D. 1980. Stem growth per unit of leaf area: A measure of tree vigor. For. Sci. 26:112-117. Watson, D.J. 1958. The dependence of net assimilation rate on leaf-area index. Ann. Bot. 22: 37-54. Whitehead, D. 1978. The estimation of foliage area from sapwood basal area in Scots pine. Forestry 51:137-149. Whitehead, D., Edwards, W.R.N., and Jarvis, P.G. 1984. Conducting sapwood area, foliage area, and permeability in mature trees of Picea sitchensis and Pinus contorta. Can. J. For. Res. 14:940-947. Wilkinson, L. 1988a. SYGRAPH. Systat Inc., Evanston, H., 923 pp. Wilkinson, L. 1988b. SYSTAT: The System for Statistics. Systat Inc., Evanston, II., 822 pp. Woodard, P.M. 1974. Predicting crown slash weights in Douglas-fir, M. Sc. thesis. Univ. Washington, Seattle, 94 pp. Wykoff, W.R., Crookston, N.L., and Stage, A.R. 1982. User's guide to the Stand Prognosis Model. USDA For. Serv., Gen. Tech. Rep. INT-133,112 pp. Zar, J.H. 1984. Biostatistical Analysis. Prentice Hall Inc., Englewood Cliffs, NJ, 718 PP. Zavitkovski, J., Isebrands, J.G., and Crow, T.R. 1974. Application of growth analysis in forest biomass studies. Proceedings of the third North American Forest Biology Workshop, Colorado State University, Fort Collins, Col., pp. 196-227. "@en ;
edm:hasType "Thesis/Dissertation"@en ;
edm:isShownAt "10.14288/1.0098305"@en ;
dcterms:language "eng"@en ;
ns0:degreeDiscipline "Forestry"@en ;
edm:provider "Vancouver : University of British Columbia Library"@en ;
dcterms:publisher "University of British Columbia"@en ;
dcterms:rights "For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use."@en ;
ns0:scholarLevel "Graduate"@en ;
dcterms:title "Net primary production, production allocation, and foliage efficiency in second growth douglas-fir stands with differing site quality"@en ;
dcterms:type "Text"@en ;
ns0:identifierURI "http://hdl.handle.net/2429/29134"@en .