Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Effects of fire on vegetation in the interior douglas-fir zone Hanel, Claudia 2000

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

Item Metadata

Download

Media
831-ubc_2000-0213.pdf [ 11.81MB ]
Metadata
JSON: 831-1.0089392.json
JSON-LD: 831-1.0089392-ld.json
RDF/XML (Pretty): 831-1.0089392-rdf.xml
RDF/JSON: 831-1.0089392-rdf.json
Turtle: 831-1.0089392-turtle.txt
N-Triples: 831-1.0089392-rdf-ntriples.txt
Original Record: 831-1.0089392-source.json
Full Text
831-1.0089392-fulltext.txt
Citation
831-1.0089392.ris

Full Text

EFFECTS OF FIRE ON VEGETATION IN THE INTERIOR DOUGLAS-FIR ZONE by CLAUDIA HANEL B.Sc. The University of Western Ontario, 1991 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF Master of Science in THE FACULTY OF GRADUATE STUDIES (Department of Forestry) THE UNIVERSITY OF BRITISH COLUMBIA We accept this thesis as corvforrning to the required standard THE UNIVERSITY OF BRITISH COLUMBIA • April 2000 © Claudia Hanel, 2000 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of f-Qre$~J-r t, The University of British Columbia Vancouver, Canada Date DE-6 (2/88) ABSTRACT To investigate the effects of fire on vegetation in dry Douglas-fir (Pseudotsuga menziesii var. glauca) forests, as well as year to year fluctuations in understory vegetation, three studies were conducted. In the first study, the percent cover of different understory plant growth forms was monitored in 1 m2 plots at two sites for three years. Significant (p < 0.05) fluctuations over time were observed for all growth forms, but these depended on site, and were not always similar in adjacent stands with different canopy closure. Only forbs showed a generally increasing trend over the three-year period at all sites. None of these fluctuations was related in a simple way to the temperature or precipitation recorded at nearby weather stations in the same growing season. The second study examined the response of understory vegetation in 1 m 2 plots to fire severity and reduction of competition by tliirming of the overstory in moss and herb dominated communities. High fire severity was associated with a lower cover and number of species persisting in the plots, and a higher cover and species richness of invading species, than low fire severity. Existing moss and lichen species were almost elinninated by fire of any severity, and most persisting species were vascular plants. The invaders with the highest cover were pioneer bryophytes colonizing patches of mineral soil. The changes in cover and number of species of all growth forms after removal of overstory competition were less than those after burning, suggesting that fire effects in forest thickets are not just due to reduction of competition. Logistic regression models predicting the probability of mortality of fire injured trees two years after burning in standing forest were developed. Percentage of the crown length scorched was the best predictor of mortality, but species, diameter at breast height and maximum height of bark char were also significant. Lodgepole pines (Pinus contortd) were more likely to die than ponderosa pines (Pinus ponderosa) and Douglas-firs when all other variables were held constant. Six tree mortality models developed in the United States were ii also tested. Some, but not all of them adequately predicted the tree mortality of the species they were developed for, but the models created in this study were generally superior for the sites that were .studied. iii TABLE OF CONTENTS ABSTRACT ii TABLE OF CONTENTS iv LIST OF TABLES vii LIST OF FIGURES , ix LIST OF APPENDICES xiii ACKNOWLEDGEMENTS Xv 1. INTRODUCTION AND LITERATURE REVIEW 1 1.1 INTRODUCTION 1 1.2 LITERATURE REVIEW 2 1.2.1 Fire Characteristics and Lethal Temperatures 2 1.2.2 Characteristics of the Interior Douglas-fir Zone 3 1.2.3 Fire History of the Interior Douglas-fir Zone 4 1.2.4 Fire Suppression and Management Concerns 5 1.3 OVERVIEW OF THE PRESENT STUDY 6 2. YEAR-TO-YEAR VARIATION OF UNDISTURBED VEGETATION IN THE IDF ZONE7 2.1 INTRODUCTION 7 2.2 OBJECTIVES 8 2.3 METHODS '. 8 2.3.1 Study Sites 8 2.3.2 Plot Lay out 10 2.3.3 Data Collection 10 2.3.4 Data Analysis 14 2.4 RESULTS 15 2.4.1 Vegetation Variation over Three Years 15 2.4.2 Weather : 21 2.5 DISCUSSION .' 23 2.5.1 Vascular Plants 23 2.5.2 Cryptogams 25 2.5.3 Disturbance and Estimation Error 26 2.6 CONCLUSIONS AND RECOMMENDATIONS 27 3. RESPONSE OF UNDERSTORY VEGETATION TO BURNING AND THINNING 29 3.1 INTRODUCTION 29 3.1.1 Measures of Fire Severity 29 3.1.2 Community Response to Disturbance 30 3.1.3 Regeneration by Sprouting 31 3.1.4 Regeneration from Seeds and Spores 31 3.1.5 Regeneration by Fragments and other Vegetative Structures 32 3.1.6 The Post-fire Environment 32 3.1.7 Variables Related to Post-burn Response 34 3.2 OBJECTIVES 35 iv 3.3 METHODS .' .- 35 3.3.1 Study Sites : 35 3.3.2 Experimental Design 37 3.3.3 Treatment Descriptions 39 3.3.4 Data Collection 40 3.3.4.1 Vegetation 40 3.3.4.2 Environmental Variables 40 3.3.5 Data Analysis 41 3.4 RESULTS 44 3.4.1 Environmental Variables 44 3.4.1.1 Overstory Attributes 44 3.4.1.2 Burn Severity 45 3.4.2 Variables Related to Understory Response 45 3.4.2.1 Persisting Species 45 3.4.2.1.1 Pre-existing Vegetation and Burn Severity 45 3.4.2.1.2 Other Variables 53 3.4.2.2 Invading Species 54 3.4.2.2.1 Fire Severity 54 3.4.2.2.2 Pre-existing Vegetation 57 3.4.2.2.3 Other Variables 57 3.4.3 Response of Individual Plant Growth Forms and Species to Burning and Reduced Overstory Competiton 58 3.4.3.1 Response to Fire 58 3.4.3.1.1 Cryptogams 58 3.4.3.1.2 Vascular Plants -63 3.4.3.2 Response to Overstory Reduction 70 3.5 DISCUSSION 72 3.5.1 Variable Combinations Best Related to Plant Response 72 3.5.1.1 Fire Severity 72 3.5.1.2 Pre-existing Vegetation 74 3.5.1.3 Other Variables 75 3.5.2 Response of Individual Plant Growth Forms and Species 77 3.5.2.1 Response to Fire 77 3.5.2.1.1 Cryptogams 77 3.5.2.1.2 Vascular Plants 78 3.5.2.2 Response to Overstory Reduction 79 3.5.2.3 Complications due to Herbivory 80 3.6 CONCLUSIONS AND RECOMMENDATIONS 81 3.6.1 Conclusions 81 3.6.2 Management Implications 82 3.6.3 Recommendations 83 4. M O R T A L I T Y O F O V E R S T O R Y C O N I F E R S 85 4.1 INTRODUCTION 85 4.1.1 Fire Damage and Tree Mortality 85 4.1.2 Effects of Fire on Surviving Trees 88 4.1.3 Modelling of Post-fire Tree Mortality 89 4.2 OBJECTIVES 90 v 4.3 METHODS 90 4.3.1 Study Sites 90 4.3.2 Plot Lay out 90 4.3.3 Data Collection 91 4.3.4 Data Analysis 91 4.3.4.1 Mortality Prediction 93 4.3.4.2 Variable Cutoffs 94 4.3.4.3 Validation Data Sets 94 4.3.4.4 Evaluation of Models from the Literature 95 4.3.4.5 Composite Models 100 4.4 RESULTS 101 4.4.1 Mortality Prediction 103 4.4.2 Variable Cutoffs 112 4.4.3 Validation Data Sets 112 4.4.4 Evaluation of Models from the Literature 116 4.4.5 Composite Models 120 4.5 DISCUSSION 121 4.5.1 Mortality Prediction 121 4.5.2 Variable Cutoffs 122 4.5.3 Validation Data Sets , 123 4.5.4 Evaluation of Models from the Literature 123 4.5.5 Application of Mortality Models '. 127 4.6 CONCLUSIONS AND RECOMMENDATIONS 129 5. G E N E R A L C O N C L U S I O N S A B O U T P R E S C R I B E D F I R E I N T H E I N T E R I O R D O U G L A S - F I R Z O N E 131 L I T E R A T U R E C I T E D 135 A P P E N D I C E S 144 vi LIST O F T A B L E S Table 2.1 Site Characteristics of the four stands in the study study 9 Table 2.2 Two-way univariate (split plot) repeated measures ANOVA comparing plant cover development of the different growth forms over a three-year period (from 1994 to 1996) at two sites (Tulip Creek and Twin Lakes) and in a stand of open and closed canopy at each site, s = Huynh-Feldt multiplier of numerator and denorrtinator degrees of freedom for within-plot tests. P-values in bold are approximate interpolations where s was fractional (linear interpolation if p > 0.1, quadratic interpolation if p < 0.1) 16 Table 2.3 Summary of weather data at Castlegar Airport, Twin Lakes and Kikomun Creek Park from May 1 to the day preceding vegetation sampling : 22 Table 3.1 Treatments applied to the plots in a) forest openings and b) forest thickets at the two Knife Creek sites. The codes are used instead of full treatment descriptions in subsequent figures and tables 38 Table 3.2 Independent variables available for inclusion in the regression equations predicting cover and number of species of persisters and invaders. The codes are used instead of full variable names in subsequent text, figures and tables 43 Table 3.3 Spearman rank correlations between a) percent cover and b) number of species of persisters and the continuous variables available for inclusion in the regression equations in June 1996, July 1996 and July 1997 for burned and unburned plots. The variable codes are explained in Table 3.2 46 Table 3.4 Spearman rank correlations between a) percent cover and b) number of species of invaders and the continuous variables available for inclusion fn the regression equations in June 1996, July 1996 and July 1997 for burned and unburned plots. The variable codes are explained in Table 3.2 47 Table 3.5 Coefficient of determination (R2), standard error of the estimate (SEE) and coefficients of the variables in the multiple linear regression equations predicting a) percent cover and b) number of persisting species, in June 1996, July 1996 and July 1997. The variable codes are explained in Table 3.2 48 Table 3.6 Coefficient of determination (R2), standard error of the estimate (SEE) and coefficients of the variables in the multiple linear regression equations predicting a) percent cover and b) number of invading species, in June 1996, July 1996 and July 1997. The variable codes are explained in Table 3.2 49 Table 4.1 Relative fire resistance of conifers found in the IDF (from Fisher and Bradley, 1987).86 Table 4.2 Weather and FWI System values at noon on the day of the burn 92 Table 4.3 Means and standard deviations of tree dimensions and fire damage characteristics of dead and surviving Douglas-fir (Fd), lodgepole pine (PI) and ponderosa pine (Py). Standard deviations are shown in brackets 102 vi i Table 4.4 Regression coefficients and diagnostic statistics for each equation predicting probability of mortality for the equation generated using all species and the species specific equations. Variable codes as defined in Table 4.3 106 Table 4.5 Regression coefficients and diagnostic statistics for each of three species-specific equations predicting probability of mortality. For each species, the data were split into three equal groups (A, B, and C) and two of these groups were used to fit the model the process was repeated three times. Variable codes as in Table 4.3 114 Table 4.6 Classification of trees and Chi-square statistic for each of the three test groups (A, B, and C) using the species specific equations for a) Douglas-fir, b) lodgepole pine and c) ponderosa pine generated from 67% of the data 115 Table 4.7 Classification of trees using a 0.5 cutoff value and chi-square statistic for each of the models tested with the reduced and full data set of each of the three tree species. Classification percentages should only be compared within numbered categories. The 2nd year post-burn mortality was used unless otherwise noted. Models in bold were developed in the present study 118 0 viii LIST OF FIGURES Figure 2.1 Layout of the vegetation plots in the open stands at Twin Lakes (TL-o) and Tulip Creek (TC-o). At TC-o the dimensions (including spacing of vegetation plots) of triangles located in smaller forest openings were proportionally reduced 11 Figure 2.2 Layout of the vegetation plots in the closed stand at Twin Lakes (TL-c) in five tree plots randomly chosen from those established by Braumandl et al. (1995). Plots were located on transects at right angles to the existing burn pin transects '. 12 Figure 2.3 Layout of the vegetation plots in the closed stand at Tulip Creek (TC-c) in five tree pfots randomly chosen from those established by Braumandl et al. (1995). Plots were centered on either the first four or last four of the burn pins of one of the three rows of pins of existing burn pin transects 13 Figure 2.4 Percent cover (left side) and change in percent cover (right side) of the five life forms from 1994 to 1995 (on left) and from 1995 to 1996 (on right) in the open and closed stands at Tulip Creek and Twin Lakes. The boxes contain half of the sample plots, the lines connect plots up to 1.5 box lengths from the box edges and points falling further then 1.5 box lengths from the mean are shown as discrete circles or stars 18 Figure 3.1 Influence of a) depth of burn, b) pre-treatment cover of all plants, and c) pre-treatment cover of vascular plants on the predicted normalized ranks of cover of persisters in June 1996 (left side) and July 1997 (right side) 51 Figure 3.2 Influence of a) depth of burn, b) pre-treatment number of all plant species, and c) pre-treatment number of vascular plant species on the predicted normalized ranks of the number of persisting species in June 1996 (left side) and July 1997 (right side) 52 Figure 3.3 Influence of a) depth of burn, b) plot area burned, and c) pre-treatment moss cover on the predicted normalized ranks of the cover of invading species in June 1996 (left side) and July 1997 (right side) 55 Figure 3.4 Influence of a) depth of burn, b) plot area burned, and c) pre-treatment moss cover on the predicted normalized ranks of the number of invading species in June 1996 (left side) and July 1997 (right side) 56 Figure 3.5 Percent cover of residual species of different growth forms grouped by treatment at KC-1 and KC-2 in a) herb plots and b) moss plots. The four adjacent bars are values for June 1995, June 1996, July 1996 and July 1997 (pre-treatment, and 8,9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets 59 Figure 3.6 Number of residual species of different growth forms grouped by treatment at KC-1 and KC-2 in a) herb plots and b) moss plots. The four adjacent bars are values for June 1995, June 1996, July 1996 and July 1997 (pre-treatment, and 8,9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 60 ix Figure 3.7 Number of invading species of different growth forms grouped by treatment at KC-1 and KC-2 in a) herb plots and b) moss plots. The three adjacent bars are values for June 1996, July 1996 and July 1997 (8,9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets 61 Figure 3.8 Percent cover of invading species of different growth forms grouped by treatment at KC-1 and KC-2 in a) herb plots and b) moss plots. The three adjacent bars are values for June 1996, July 1996 and July 1997 (8,9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets 62 Figure 3.9 Mean percent cover of Rosa acicularis and Spirea betulifolia in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets 64 Figure 3.10 Mean Percent cover of Vaccinium caespitosum in herb plots at KC-1 in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets 65 Figure 3.11 Mean percent cover of Fragaria virginiana and Lathyrus ochroleucus in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 67 Figure 3.12 Mean percent cover of Arnica cordifolia and Aster conspicuus in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets 68 Figure 3.13 Mean percent cover of Taraxacum officinale in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Due to very low cover, data for moss plots at KC-1 is not shown. Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets 69 Figure 3.14 Mean percent cover of Calamagrostis rubescens in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets 71 Figure 4.1 Logistic regression models predicting the probability of mortality using various predictors created by a) Ryan and Reinhardt (1988), b) Bevins, (1980), c) Saveland, (1982) and d) Regelbrugge and Conard(1993) 96 Figure 4.2 Logistic regression models created by Harrington (1993) predicting the probability of mortality for ponderosa pine burned during the a) dormant season and b) growing season. Discrete classes of crown scorch and diameter were used to create the models 98 Figure 4.3 Distribution of dead and live a) Douglas-fir, b) lodgepole pine and c) ponderosa pine with respect to percent scorch height class 104 Figure 4.4 Probability of tree mortality as predicted from the equation for all trees using percentage of the live crown height scorched as the only predictor 105 Figure 4.5 Probability of mortality as a function of diameter at breast height (DBH) and percentage of the live crown height scorched (PSH) for the species specific equations for Douglas fir at maximum height of bark char (HBC) values of a) 0 m, b) 4 m, and c) 16 m, and for d) ponderosa pine and e) lodgepole pine at all HBC values 107 Figure 4.6 Probability of mortality of 0.5 for ponderosa pine from equations using the square, tenth power and fifteenth power of the percentage of live crown height scorched (PSH) and diameter at breast height (DBH). Below the line the probability of mortality is > 0.5 and above the line it is < 0.5 108 Figure 4.7 Distribution of dead and alive Douglas-fir correctly and incorrectly predicted by the species specific equation and the all species, PSH only equation with respect to the percentage of crown height scorched', diameter at breast height and maximum height of bark char 109 Figure 4.8 Distribution of dead and alive lodgepole pine correctly and incorrectly predicted by the species specific equation and the all species, PSH only equation with respect to the percentage of crown height scorched, diameter at breast height and maximum height of bark char 110 Figure 4.9. Distribution of dead and alive ponderosa pine correctly and incorrectly predicted by the species specific equation and the all species, PSH only equation with respect to the percentage of crown height scorched, diameter at breast height and maximum height of bark char I l l Figure 4.10 Number of trees correctly and incorrectly predicted dead or alive when different probability of mortality cutoff values (above which a tree is considered dead) are used. Values are from the species specific equations for a) Douglas-fir, b) lodgepole pine, c) ponderosa pine 113 xi Figure 4.11 Distribution of dead and alive a) and b) Douglas-fir and c) lodgepole pine correctly and incorrectly predicted by the a) Bevins (1980) model, b) and c) Ryan and Reinhardt (1988) model with respect to the independent variables used to create each model 117 Figure 4.12 Distribution of dead and alive ponderosa pine correctly and incorrectly predicted by the a) Saveland (1982), b) Regelbrugge and Conard (1993), c) Harrington (1993) growing season, d) Harrington dormant season and e) Ryan and Reinhardt (1988) models with respect to the independent variables used to create each model 119 xii LIST O F APPENDICES Appendix 2.1 Mean percent cover values for species found at Tulip Creek and Twin Lakes for 1994,1995 and 1999 144 Appendix 2.2 Normal probability plots and P-values from the Kolmogorov-Smirnov normality tests for the residuals of the repeated measures ANOVAs comparing change in vegetation cover at Twin Lakes and Tulip Creek for a) bryophytes in 1994 and b) lichens in 1994 where the residuals were not distributed normally. 149 Appendix 2.3 Daily precipitation (in mm) at Castlegar Airport in a) 1994, b) 1995, and c) 1996. 150 Appendix 2.4 Daily precipitation (in mm) at a) Twin Lakes in 1994, and at Kikomun Creek Park b) in 1994, c) in 1995, and d) in 1996 151 Appendix 3.1 Density of the overstory in a) forest openings and b) forest thickets for the various treatments at Knife Creek 152 Appendix 3.2 Severity of overstory thinning in a) forest openings and b) forest thickets for the various treatments at Knife Creek 153 Appendix 3.3 Spearman rank correlations between the independent variables used in the correlation and regression analyses in a) unburned plots and b) burned plots. The variable codes are explained in Table 3.2 : 154 Appendix 3.4 Burn severity statistics for the various treatments at Knife Creek 155 Appendix 3.5 Residuals from the equations vs. predicted normalized ranks of cover and number of residual species, as well as normalized ranks of cover and number of invaders for June 1996, July 1996, and July. 1997 156 Appendix 3.6 Normal probability plots and results of the Kolmogorov-Smirnov normality test for the residuals of the equations predicting a) cover of residual species in June 1996, the number of residual species in b) June 1996 and c) July 1996, and e) the number of invading species in June 1996 160 Appendix 3.7 Influence of a) basal area cut, b) pre-treatment canopy cover, and c) site on the predicted normalized ranks of cover of persisters in June 1996 (left side) and July 1997 (right side)..... 162 Appendix 3.8 Influence of a) basal area cut, b) pre-treatment canopy cover, and c) plot area burned on the predicted normalized ranks of the number of persisting species in June 1996 (left side) and July 1997 (right side) 163 Appendix 3.9 Influence of a)pre-treatment canopy cover on the predicted normalized ranks of the cover of invaders and b) number of trees surrounding the plot on the predicted normalized ranks number of invading species in June 1996 (left side) and July 1997 (right side) 164 xm Appendix 3.10 Mean percent cover of all plant species found at Knife Creek in A) KC-1 herb plots, B) KC-1 moss plots C) KC-2 herb plots and KC-2 moss plots. Values are shown for the six treatments (OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in forest thickets, FTL-thinned + light burn in forest thickets) at four sampling occasions: (1) June 95 (pre-treatment), (2) June 96, (3) July 96, and (4) July 1997 165 Appendix 4.1 Percentages of surviving, dead, and all trees correctly classified by the single species equation and the equation with all species using varying probability of mortality cutoff values 180 xiv ACKNOWLEDGEMENTS I wish to warmly thank my supervisor, Dr. Michael C. Feller, for providing guidance and focus to this project. I also wish to express my appreciation to Dr. Valerie Lemay, who spent many hours advising me on statistical matters. Many thanks to the other members of my committee for reading a substantial thesis and providing valuable feedback. I very much appreciated the time spent by Trevor Goward, VVilf Schofield, Olivia Lee, Terry Taylor, Ray Coupe and Tom Braumandl to initiate me into the art of identifying lichens, mosses and vascular plants. Many thanks to the field assistants Christine Dekens, Rebecca Johnson, Penny Olanski, Corinne Guillot, Heidi Slaymaker, Kim Johnston, Hillary Rudd, Jason Wagenaar, Cari Olson, Lucy Duso, and especially Anna Avilov, who helped to make the collection of vegetation, tree cover and tree mortality data a more manageable affair. I also wish to acknowledge the kind assistance of Steve Taylor from the Canadian Forest Service, who provided me with tree measurement data for the tree mortality models, the staff of the Alex Fraser Research forest who helped with plot selection and burning, and the Fire Injuns (Williams Lake Unit Crew forest fire fighters), who performed the thinning in the Knife Creek forest. This work was supported in part by the Canadian Forest Service under the Green Plan. xv 1. INTRODUCTION AND LITERATURE REVIEW 1.1 INTRODUCTION Prescribed burning can be a tool for manipulating both overstory and understory vegetation. It has been used in dry Douglas-fir (Pseudotsuga menziesii var. glauca) forests to kill some of the regenerating Douglas-fir, permitting the re-estabUshment of pioneer tree species, and to rejuvenate shrubs and herbs, increasing food supplies for cattle and wildlife. In Canada, 100,000 to 180,000 ha have been prescribed burned annually in recent years. About 35% of this area was burned for wildlife habitat and, since 1988, a very small fraction for ecosystem restoration (Feller, 1996). Over 90% of Canada's wildlife burning was conducted by the Province of British Columbia (BC), while ecosystem restoration burning was practiced mainly by the Canadian Parks Service. Ecosystem maintenance burning in forests is expected to comprise a larger fraction of the total area burned in the future, as its benefits are becoming more widely appreciated and guidelines that enable safe and effective prescriptions are developed (Feller, 1996). Some jurisdictions, such as the Nelson Forest Region, are investigating the feasibility of reintroducing fire into dry Dougas-fir ecosystems in the interior of BC (Braumandl et al, 1995). While in most parts of BC plants have evolved in the presence of a fire regime (Parminter, 1992), and the restoration of something similar to this fire regime seems desirable, at least in natural areas, planning has to focus on the immediate consequences of each burn. Planners of prescribed burns have to determine appropriate objectives for the burn and then select weather and fuel moisture conditions that produce effective and acceptable levels of fire behaviour to achieve these goals. Often goals of prescribed burns are structural, (i.e. a certain vegetation composition is aimed for with a specific prescription) (Agee and Huff, 1986). This necessitates a thorough understanding of fire behaviour and the succession ecology of the ecosystem in question. Canadian agencies have had to rely mainly on the findings of American fire research (e.g. Kilgore, 1985; Wright and Bailey, 1982) to develop prescriptions designed to kill a selected proportion of trees of smaller diameter classes, expose mineral soil and increase production of understory vegetation. It is questionable that these results can be mdisCTirninately applied in Canada, given the local ecosystem variation due to the large geographic and altitudinal range of dry Douglas-fir forests. Also, more quantitative information on the local 1 response of individual understory plant species, as well as on changes in species composition, is needed for different regimes of prescribed fire (Wikeem and Strang, 1983). The objectives of the present study were to quantify the yearly variation of understory vegetation in undisturbed areas and the response of understory vegetation to burning, with or without overstory tlurming. Also, models predicting the probability of mortality of fire damaged trees were developed, and model performance was compared to that of several American tree mortality models. 1.2 LITERATURE REVIEW 1.2.1 FIRE CHARACTERISTICS AND LETHAL TEMPERATURES Fire has the ability to kill at least portions of plants. The lethal temperature for a plant cell is 61° C for instantaneous death (Wade, 1986), 60° C for death after exposure for two minutes to an hour (Wright and Bailey, 1982), and 50° C for death after several hours (Wade, 1986). The relationship between lethal temperature and exposure time is inversely exponential (Hungerford et at, 1990; Wright and Bailey, 1982). Lethal temperatures are lower at all exposure time levels if the plant moisture content is higher, and the concentrations of salt, sugar, lignin and pectin are also believed to play a role (Wright and Bailey, 1982). To kill a plant, heat of lethal temperature and duration must pass through air, duff, soil or protective plant structures to reach a sufficient proportion of heat sensitive tissues. Whether this occurs depends on the type of fire, the thermal diffusivity of the air, duff or soil, as well as on the type of plant, its vigour and phenological state (Flinn and Wein, 1988; Hungerford et at, 1990; Kauffman, 1990; Wade, 1986; Wright and Bailey, 1982). The amount of heat released at different locations above and below the ground surface during a fire varies greatly depending on the type and amount of fuel consumed and on the speed at which the fire travels (Dauberurure, 1968; Wade, 1986). Crown, surface, and ground fires have different heat release and spread rates (Hungerford et at, 1990) and more intense fires with longer flames concentrate heat release higher above ground level (Wade, 1986). Temperatures during flaming combustion range from 1000 °C to 1500 °C, while those found in smouldering duff range from 500 °C to 600 °C (Hungerford et at, 1990). Temperature profiles at the ground surface or on the bark of tree boles characteristically show a sharp peak of maximum temperatures lasting from a few minutes to a few hours (Hartford and Frandsen, 1992; Noste et at, 1988), while those below the ground surface exhibit lower peak temperatures 2 of much longer duration (Braumandl et al, 1995; Hartford and Frandsen, 1992). In ecological studies, fire temperatures are usually measured at the ground surface or within the forest floor and upper mineral soil. A wide range of surface temperatures has been reported, from below 100° C during grass fires (Daubenmire, 1968) to 750° C in experimental small plot burns in boreal fores vegetation (Schirrvmel and Granstrom, 1996). Braumandl et al (1995) recorded maximum surface temperatures ranging from 203° C to 597° C in prescribed burns in standing forest in the southern Rocky Mountain Trench in BC. Several terms are used to describe the characteristics of a fire without relying on measured temperatures. Frontal fire intensity is the rate of heat release per unit length of fire front, and can be determined from flame length or calculated from the rate of fire spread and the amount of fuel consumed (Wade, 1986). It can be predicted by the codes and indices of the Canadian Fire Weather Index (FWI) system and the fuel type (Forestry Canada Fire Danger Group, 1992). Fire intensity is a good measure of the vertical heat pulse up, but not of the heat pulse down (Brown and DeByle, 1989), and Wade (1986) recommended its use only as a predictor of fire damage above the flame zone. Fire severity has been described as either the heat pulse down (Brown and DeByle, 1987; Flinn and Wein, 1988), or as the combined effect of the heat pulse up and down, and therefore including fire intensity (Brown and DeByle, 1989; Ryan and Noste, 1985). Since fire intensity can be characterized separately, the term fire severity will only be used in the former sense, the heat pulse down. Fire severity is determined by fuel consumption, fire residence time, and transmission of heat below-ground (Brown and DeByle, 1989). It determines the consumption or mortality of plant parts in the forest floor, as well as the amount of exposed mineral soil; therefore it is a descriptor of the impact of fire on an ecosystem (Brown and DeByle, 1989). 1.2.2 CHARACTERISTICS OF THE INTERIOR DOUGLAS-FIR ZONE Forests where Douglas-fir (Pseudotsuga menziesii var. glauca) is one of the dorninant climax species occupy extensive areas in the western United States and Canada, ranging from northern California to southern Alberta and British Columbia (Wright and Bailey, 1982). The Interior Douglas-fir Zone (IDF) in British Columbia is part of this area found in the Interior Plateau region of British Columbia and in the southern Rocky Mountain Trench at altitudes ranging from 350 m to 1450 m (Meidinger and Pojar, 1991; Tisdale and McLean, 1957). The zone is characterized by a continental climate with warm, dry summers, a fairly long growing season, cool winters, and often a substantial moisture deficit. Mean annual precipitation 3 generally ranges from 300 mm to 750 mm, but it exceeds 1000 mm in the wettest areas (Meidinger and Pojar, 1991). Twenty to 50% of the precipitation falls as snow, which forms a permanent cover for several months (Meidinger and Pojar, 1991; Tisdale and McLean, 1957). Most of the area is covered by till deposited during the Pleistocene glaciations (Tisdale and McLean, 1957). Soils are typically Orthic to Dark Grey Luvisols, and Eutric or Dystric Brunisols with a medium to rich nutrient regime. Humus forms range from Mors to Mullmoders. Douglas-fir is the dominant climax tree species and important serai tree species include lodgepole pine (Pinus contorta var. latifolia), ponderosa pine (Pinus ponderosa), trembling aspen (Populus tremuloid.es), paper birch (Betula papyrifera) and western larch (Larix occidentalis). Douglas-fir is also often found in serai stands. Ponderosa pine is found at low elevations south of Clinton and Little Fort, and western larch is restricted to the southern part of the zone. On moister sites, hybrid white spruce (Picea glauca x engelmannii) and western redcedar (Thuja plicata) are also found. Stand structure ranges from open to closed, depending on the disturbance history of the site (Meidinger and Pojar, 1991). Common understory species include Calamagrostis rubescens, Arctostaphylos uva-ursi, Aster conspicuus, Spirea betulifolia, Fragaria virginiana, Pleurozium schreberi, Amelanchier alnifolia, Peltigera spp. and Cladonia spp. (Tisdale and McLean, 1957). The zone is subdivided into seven subzones, which vary slightly in climate and vegetation characteristics. Some of these are further divided into variants. Large bunchgrass communities have developed in parts of the IDF as a result of edaphic and topographic conditions, as well as past disturbances by fire (Meidinger and Pojar, 1991). These grasslands, as well as many of the forested communities of this zone, have been extensively utilized for grazing of domestic livestock, which has led to shifts in species composition as well as conifer invasions (Meidinger and Pojar, 1991; Tisdale, 1950). 1.2.3 FIRE HISTORY OF THE INTERIOR DOUGLAS-FIR ZONE Fire recurs on a site in a sequence termed a fire regime (Agee, 1993), which over time produces selection pressures favouring certain plant strategies over others (Kauffman, 1990). In dry Douglas-fir forests in the United States fire histories are better documented than in BC, and Arno (1976) was able to determine the fire record in western Montana back to the 1500s or 1600s. On warmer and drier sites, where ponderosa pine and western larch are serai dominants, fire free intervals ranged from 7 to 19 years, but in somewhat moister and cooler sites, where lodgepole pine and Douglas-fir are common, fire-free intervals were slightly 4 longer, ranging from 17 to 28 years (Arno, 1976). In central Idaho the natural mean fire return interval ranged from 15 to 26 years in areas where ponderosa pine was the serai dominant (Barrett, 1988). Longer mean fire-free intervals of 26 to 52 years were found in western Montana in higher elevation ponderosa pine/Douglas-fir stands, which were less likely to be burned by First Nations, who are believed to have strongly irtfluenced fire intervals in many valleys (Arno, 1976; Arno et al, 1995). The fires at any site were not of the same intensity or severity. In Douglas-fir forests in Glacier National Park in Montana, individual western larches and ponderosa pines often survived up to seven non-lethal surface fires between stand replacing fires (Barrett et al, 1991). Fire intensities also varied somewhat among different ecosystems within the Douglas-fir zone. For example, in drier Douglas-fir/bunchgrass ecosystems crown fires were probably rare, while both light surface fires and crown fires burned in moister Douglas-fir/pinegrass ecosystems (Wright and Bailey, 1982). Large areas are often burned in years of unusual drought, such as during the time period from 1910 to 1936 which was the most severe drought cycle in the Pacific Northwest in the last three centuries (Barrett et al, 1991). Parminter (1992) attempted to reconstruct the historic fire regime in the IDF in British Columbia, using American data, as well as mostly unpublished data from the BC Ministry of Forests. He described the average mean fire return interval as ranging from 10 to 20 years for surface fires and from 150 to 250 years for severe surface fires and crown fires. The average size for surface fires was estimated at 5 to 50 ha, and that for crown fires at 50 to 500 ha, with some being as large as 5000 ha. However, Arno (1976) stated that historically even low to medium intensity fires could cover large areas. In several areas near Kamloops and Lillooet, BC, fire free intervals ranged from 2 to 46 years, with averages of 10 years and 13 to 18 years for drier and moister Douglas-fir forests, respectively (Low, 1988). At several IDF sites in the Cariboo region fire intervals ranged from 3 to 32 years, with means from 8 to 18 years (Daniels et al, 1995). 1.2.4 FIRE SUPPRESSION AND MANAGEMENT CONCERNS Since the beginning of this century fire suppression and fuel removal caused by cattle grazing have lengthened the mean fire return interval. While the fire free intervals in the 1700's and 1800's did not exceed 32 years in the Cariboo IDF stands described by Daniels et al (1995), no fires have occurred there in the last 40 to 100 years. Braumandl et al (1995) noted the absence of data on fire suppression in BC, but considered trends observed in National Forests in Washington, Idaho and Montana, to be quite similar. There from 1910-1919 over 2,000,000 5 ha were burned by wildfires, but this value decreased during every ten-year period until from 1950-1959 less than 40,000 ha were burned. Increased stocking of Douglas-fir attributed to fire suppression has been observed in many formerly open stands both in the United States (Agee, 1993; Arno, 1976; Arno et al, 1995; Barrett, 1988; Habeck, 1990; Wright and Bailey, 1982) and in BC (Low, 1988; Tisdale, 1950). This is considered detrimental to a stand for several reasons: a) the likelihood of severe crown fires is greater due to an increase in ladder and surface fuels; b) the aesthetic values of the stand and timber volume growth are decreased; c) the production of shrubs, herbs and grasses which provide forage for wildlife and livestock is reduced; and d) the trees become more susceptible to attacks of western spruce budworm (Choristoneura occidentalis) (Agee, 1993; Arno et al., 1995; Braumandl et al, 1995; Fisher and Bradley, 1987; Tisdale,1950; Wright and Bailey, 1982). Fire suppression has also caused shifts in stand composition from shade intolerant species, such as ponderosa pine, western larch and lodgepole pine, towards the more shade tolerant Douglas-fir (Agee, 1993; Arno et al, 1995; Fisher and Bradley, 1987). For example, in a ponderosa pine dominated forest in Montana the tree density more than quadrupled between 1900 and 1984 (Habeck, 1990). While Douglas-fir was almost absent in the diameter classes above 36 cm, over two thirds of the regeneration below 7 cm in diameter was Douglas-fir. Low (1988) implicated the loss of plant species diversity caused by fire suppression in local extirpations of animal species that depend on specialized habitats, such as the yellow badger near Kamloops. 1.3 OVERVIEW OF THE PRESENT STUDY The present study consisted of three parts, in which different aspects of vegetation dynamics were evaluated in burned and in undisturbed areas in the Interior Douglas-fir zone. In Chapter 2, the fluctuations in cover of understory plants over a period of three years were described, and it was attempted to determine if these cover fluctuations were related in any way to overstory canopy closure or to growing season weather variables. Chapter 3 describes a study in which small plots in forest openings and thickets were burned, thinned, or both to determine the effects of fire and overstory reduction on understory vegetation. In Chapter 4, data from prescribed burns were used to develop equations predicting the probability of mortality of fire-damaged conifers, and the performance of several American tree mortality models was evaluated. 6 2. Y E A R - T O - Y E A R V A R I A T I O N O F U N D I S T U R B E D V E G E T A T I O N I N T H E I D F Z O N E 2.1 INTRODUCTION Variations in plant cover from year-to-year can be due either to succession after disturbances, or to smaller, reversible fluctuations. These fluctuations are caused by changes in meterological and hydrological conditions, herbivory by animals, pathogens, peculiarities in the life cycles of plants, and human influences (Rabotnov, 1974). Many studies of plant succession have assumed that these year-to-year fluctuations are negligible, and that plant cover in undisturbed areas remains essentially steady. Those studies did not include control plots, and compared changes in disturbed areas to a single pre-disturbance measurement (e.g., Geier-Hayes, 1989; Lyon, 1971; Stickney, 1981). Changes in plant cover are a reflection of the complex interactions of plants with each other, the abiotic environment, and the other biota on a site. Increases in cover are caused by growth of existing plants, or by the establishment of new individuals, as favourable weather conditions or removal of competitors increase the availability of resources. Plants that are able to grow new leaves quickly, such as many forbs and graminoids, would be most likely to increase in cover in response to short term improvement in growing conditions. Reductions in cover are due to senescence or removal of leaves at a higher rate than their replacement, which can be caused by herbivory by a variety of organisms, trampling, diseases, environmental stress, or competition by other plants. While successional plant development after major disturbances, such as fire or logging, has been extensively studied, little attention has been focused on the changes occurring in undisturbed areas over time, especially in Douglas-fir ecosystems. A limited amount of information is available from control plots of successional studies in vegetation types having some similarity to Douglas-fir forests. In some cases, year-to-year variation in cover of various growth form groups was observed to be <2% (Riegel et al, 1995), while in others it exceeded 20% (Brown and DeByle, 1989). However, in the drier ecosystem where the cover variations were smaller, the changes were larger relative to the initial cover. Brown and DeByle (1989) and Crouch (1986) reported year-to-year changes in biomass in addition to percent cover, but in both studies instances were found where cover and biomass did not increase or decrease in unison. However, Crouch (1986) did not mention whether cover was assessed on permanent plots, or on the biomass plots prior to clipping. 7 2.2 OBJECTIVES The present study was conducted to examine trends in variation in plant cover of several understory growth form types over several years. Specifically the following objectives were addressed: 1. To determine if any annual variation was present and if this variation was similar at different sites and in adjacent stands with differing levels of canopy closure. 2. To determine if any year-to-year changes in vegetation cover were linked in an obvious way to weather patterns recorded at the nearest weather stations. 2.3 METHODS 2.3.1 STUDY SITES The understory vegetation was studied at two sites in the interior of British Columbia that were selected by Braumandl et al. (1995) as candidates for the Ecosystem Maintenance Burning Research and Evaluation (EMBER) project because they were representative of sites where operational burning could be used to reduce tree density and surface fuels. Both sites contained both a stand with many forest openings about 10 m to 30 m in diameter and an adjacent stand with a denser overstory, and were located in mesic ecosystems with medium moisture and nutrient regimes, and remained unburned for the duration of the study. The Tulip Creek (TC) site near Castlegar was located approximately 1 km from the western shore of Lower Arrow Lake within the Arrow Lake Unit of the Undifferentiated IDF (Braumandl and Curran, 1992). The closed forest was called TC-c and an area of openings TC-o. The overstory of both stands was dominated by Douglas-fir, with smaller amounts of ponderosa pine, western redcedar and western larch. The soils were Orthic Dystric Brunisols on glacial till 0ungen, 1980). The Twin Lakes site was near Lake Koocanusa, approximately 40 km southeast of Cranbrook, and fell witWn the FdPl Pinegrass-Twinflower Site Series of the Kootenay Dry Mild IDF Variant (dm2) (Braumandl and Curran, 1992). Both the closed stand (TL-c) and the open stand (TL-o) supported a Douglas-fir/western larch forest. The soils were Eutric Brunisols on morainal parent material (Kelley and Sprout, 1956, Lacelle, 1990). A more detailed description of the study sites is found in Table 2.1. 8 Table 2.1 Site Characteristics of the four stands in the study study. Slope Canopy Basal Tree Elevation (%), cover area density Site Latitude Longitude (m) Aspect (%) (m2/ha) (Stems/ha) Tulip Creek 49° 24'N 117° 59'W 610 (Castlegar) TC-o 30 28 11 670 (openings) SW TC-c 24 73 45 1600 (closed forest) SW Twin Lakes 49° 13'N 115° 21' W 1040 (Cranbrook) TL-o 0 42 27 1100 (openings) TL-c 0 59 32 1900 (closed forest) 9 2.3.2 PLOT LAYOUT A total of 100 vegetation plots were used in this study. In the less dense stands at both sites, five equilateral triangles with a side length of 30 m were located subjectively to represent the forest openings. At TC-o, some smaller triangles with side lengths of 23 m and 20 m were used. Six 1 m 2 square vegetation plots were placed at 10 m intervals along the perimeter line and 1 m beside the line of each triangle, for a total of 30 per stand (Figure 2.1). At the three smaller triangles at TG-o, the distance between the vegetation plots was proportionately reduced. Where obstructions were encountered, the plots were moved to the nearest unobstructed location. The plots were marked at two corners with metal pins. In the denser stands at both sites, systematically arranged plot centres had been previously established by Braumandl et al. (1995). These plots were located on a 100 m square grid at Twin Lakes, and on a 150 m by 75 m rectangular grid at Tulip Creek. At each of these sites, five of the previously established plot centres, were chosen at random. Twenty vegetation plots were located in both the TC-c and the TL-c stand. At TC-c, one transect with three rows of metal pins in a random direction had been established at each main plot by Braumandl et al. (1995). For each of the five chosen plots, one row of pins was picked at random and vegetation plots were centered on either the first four or last four pins of the row (Figure 2.2). At the TL-c site, one transect with only one row of pins at a random direction had been placed at each plot centre by Braumandl et al. (1995). Four vegetation plots were established every 4 m along, and 1 m beside, a separate transect located at 90° from the pin transect (Figure 2.3). The location of these subplots was marked with metal spikes. 2.3.3 DATA COLLECTION All plants less than 2 m in height were considered understory plants. The percent cover of all understory species was visually estimated to the nearest 1 % at each of the 30 vegetation plots in openings and the 20 plots in the closed stand at both sites. Due to the difficulty of estimating very small percentages, any cover values below 1% were coded as 0.5%. Vascular plants had to be rooted in the quadrats to be counted. Nomenclature followed Qian and Klinka (1998). At each vegetation plot the overstory canopy cover was visually estimated. All stands were sampled in July 1994,1995, and 1996. A photograph was taken of each vegetation plot at Twin Lakes in August 1994, about four weeks after vegetation sampling. In 1995 and in 1996, plots at both sites were photographed at the time of vegetation sampling. 10 I I 1 m 2 vegetation plot • rebar corner stake O 60 • • \< 10 m >\< 10 m >\< 10 m >\ Figure 2.1 Layout of the vegetation plots in the open stands at T w i n Lakes (TL-o) and Tulip Creek (TC-o). A t TC-o the dimensions (including spacing of vegetation plots) of triangles located in smaller forest openings were proportionally reduced. 11 Figure 2.2 Layout of the vegetation plots in the closed stand at Twin Lakes (TL-c) in five tree plots randomly chosen from those established by Braumandl et al. (1995). Plots were located on transects at right angles to the existing burn pin transects. 12 I I 1m 2 vegetation plot -X— burn pin transect • rebar corner stake Figure 2.3 Layout of the vegetation plots in the closed stand at Tulip Creek (TC-c) in five tree plots randomly'chosen from those established by Braumandl et al (1995). Plots were centered on either the first four or last four of the burn pins of one of the three rows of pins of existing burn pin transects. 13 Weather records from fire weather stations near each site were examined for the growing seasons under consideration. Data for 1994,1995, and 1996 were collected at Castlegar Airport, which was located approximately 28 km southeast of Tulip Creek, and at Kikomun Creek Park, which was located about 8 km northwest of Twin Lakes in the Rocky Mountain Trench. The Castlegar Airport and Kikomun Park stations were approximately 150 m and 240 m lower in elevation than the Tulip Creek and Twin Lakes site, respectively, and both were situated on flat terrain. A fire weather station was operational for the whole 1994 season directly at Twin Lakes, and for that year data from this station were also reported. For each day of the growing season from May 1 to the day preceding vegetation sampling, the precipitation and temperature at noon were obtained. The total precipitation and the average temperature at noon were calculated for this time period. 2.3.4 DATA ANALYSIS Because few species occurred with high frequency at all sites, and in some cases plants were too immature to be accurately identified to the species level, the individual plant species were grouped into the following growth form groups: woody plants, forbs, graminoids, bryophytes, and lichens. Sub-shrubs, such as Arctostaphylos uva-ursi and Linnaea borealis, were included in the woody plant group, and the vine Clematis columbiana was included in the forb group. To produce vegetation variables that are approximately normally distributed, rank transformations were performed separately on the cover percentages of each growth form using SYSTAT Version 7.0 (SPSS Inc., 1997). The data were then normalized using Blom's (1958) method. The ranked and normalized percent cover values of each growth form were separated by year and the values for the different years were then used as the dependent variables in a factorial repeated measures Analysis of Variance (ANOVA). SITE (TC or TL) was considered a random factor and CANOPY (openings or closed forest) a fixed factor. The residuals for each growth form group for each year were tested for normality with a Kolmogorov-Smirnov test, using Minitab Release 10.51 (Minitab Inc., 1995). For the within-plot significance tests, both the numerator and denominator degrees of freedom were multiplied by a value e (0 < e > 1, as calculated by SYSTAT Version 7.0; s < 1 if the variances within the repeated measures were not homogeneous, Huynh-Feldt, 1976). For tests where an interaction term was used as the error, the P-values were calculated using EXCEL 97. Since EXCEL 97 cannot calculate P-values using the fractional degrees of freedom obtained by the Huynh and Feldt correction, approximate p-values were interpolated from the two values obtained after 14 rounding both numberator and denominator degrees of freedom to the next highest, and both to the next lowest, integer. Linear interpolation was used except when one of the p-values from rounded degrees of freedom was < 0.1; in this case quadratic interpolation was used for more accurate approximations. The significance of all effects was determined at the accepted standard of a = 0.05. 2.4 RESULTS A list of species present at each study site during the three years of sampling and the average percent cover of each species in each stand can be found in Appendix 2.1. 2.4.1 VEGETATION VARIATION OVER THREE YEARS The interactions and main effects tested are shown in Table 2.2. If a higher level interaction was significant, and not ordered, lower level interactions and main effects were not interpreted. Interpretation of the significant interactions was done from graphics (Fig. 2.4). Any change in the mean cover of a plant growth form group was the net result of the cover changes of all species in all of the individual plots. None of the growth forms increased or decreased in cover in all of the plots at any of the sites (Fig. 2.4). A uniform increase or decrease of all species of a growth form was not observed in any stand (Appendix 2.1). Cover of woody plants responded differently with time at the two sites (TIME X SITE interaction, Table 2.2). At both Twin Lakes stands, woody plant cover decreased by about half in 1995, then increased again in 1996, while remaining relatively unchanged at Tulip Creek (Fig. 2.4). At TL, the pattern of decline and subsequent increase in cover was observed for most woody species, especially for Mahonia spp. which varied from ~ 10% at both sites in 1994 to < 5% in 1995 and > 7% in 1996, and Spirea betulifolia, which showed changes of slightly lower magnitude (Appendix 2.1). At TC, fluctuations were smaller and less consistent among species. The pattern of decline and subsequent increase at TL was reflected by correstponding changes of >5% on most plots in both stands at TL. At TC-c, most changes in cover were smaller (< 5% in over 75% of the plots), while at TC-o larger changes in cover occurred in many plots, but the increases and decreases tended to cancel each other. Although average woody plant cover at TC-o decreased slightly during both yearly intervals, increases in cover exceeding 5% were found on more than a quarter of the plots during both time intervals (Fig. 2.4). Regardless of the time period sampled, at TL the cover of woody plants was similar in the open and closed 15 Table 2.2 Two-way univariate (split plot) repeated measures ANOVA comparing plant cover development of the different growth forms over a three-year period (from 1994 to 1996) at two sites (Tulip Creek and Twin Lakes) and in a stand of open and closed canopy at each site, e = Huynh-Feldt multiplier of numerator and denominator degrees of freedom for within-plot tests. P-values in bold are approximate interpolations where s was fractional (linear interpolation if p > 0.1, quadratic interpolation if p < 0.1). A) Woody plants e = 1.000 Sum of Degrees of Mean Source of Variation Squares Freedom Square F-test F P Between Plots SITEi 0.856 1 0.856 MS S / M S EI 0.40 0.530 CANOPY2 19.352 1 19.352 MS c /MS sxc 1.18 0.560 SITE X CANOPY 16.355 1 16.355 MS sxc /MS EI 7.58 0.007 Error 1 207.265 96 2.159 Within Plots TIME3 16.026 2 8.013 M S T / M S T X S 2.30 0.303 TIME X SITE 6.960 2 3.480 MS T X S / M S E2 26.54 0.000 TIME X CANOPY 0.536 2 0.268 MS T X C / M S TXSXC 0.79 0.559 TIME X SITE X CANOPY 0.682 2 0.341 M S T X S X C / M S E2 2.60 0.077 Error 2 25.172 192 0.131 B) Forbs s = 0.9563 Sum of Degrees of Mean Source of Variation Squares Freedom Square F-test . F P Between Plots SITE 0.261 1 0.261 MS s /MS EI 0.13 0.781 CANOPY 36.429 1 36.492 MS c /MS sxc 25.53 0.124 SITE X CANOPY 1.427 1 1.427 MS sxc /MS EI 0.72 0.399 Error 1 190.762 96 1.987 Within Plots TIME 12.513 2 6.257 M S T / M S T X S 21.07 0.050 TIME X SITE 0.594 2 0.297 M S T X S / M S E2 2.90 0.243 TIME X CANOPY 1.209 2 0.605 MS T X C / M S TXSXC 2.04 0.334 TIME X SITE X CANOPY 0.741 2 0.371 M S T X S . X C / M S E2 1.78 0.174 Error 2 40.033 192 0.209 1 SITE levels were TC and TL. 2 CANOPY levels were open forest and closed forest. 3 TIME levels were 1994,1995 and 1996. 16 Table 2.3 (continued) C) Graminoids 8 = 1.000 Sum of Degrees of Mean Source of Variation Squares Freedom Square F-test F P Between Plots SITE 2.230 1 2.230 MS s/MS EI 1.39 0.241 CANOPY 73.385 1 73.385 MS c /MS sxc 224.42 0.042 SITE X CANOPY 0.327 1 0.327 MS sxc /MS EI 0.20 0.652 Error 1 153.881 96 1.603 Within Plots TIME 1.346 2 0.673 M S T / M S T X S 0.86 0.539 TIME X SITE 1.575 2 0.787 M S T X S / M S E2 3.08 0.048 TIME X CANOPY 0.757 2 0.378 MS T X C / M S TXSXC 1.43 0.412 TIME X SITE X CANOPY 0.530 2 0.265 M S T X S X C / M S E2 1.04 0.357 Error 2 49.085 192 0.256 D) Bryophytes s = 0.8839 Sum of Degrees of Mean Source of Variation Squares Freedom Square F-test F P Between Plots SITE 1.577 1 1.577 MS s/MS EI 0.81 0.371 CANOPY 20.847 1 20.847 MS c /MS sxc • 1.43 0.412 SITE X CANOPY 13.151 1 13.151 MS sxc /MS EI 6.75 0.011 Error 1 186.978 96 1.948 Within Plots TIME 1.184 2 0.592 MS T/MS TXS 1.59 0.199 TIME X SITE 0.252 2 0.126 M S T X S / M S E2 0.71 0.475 TIME X CANOPY 3.363 2 1.681 MS T X C / M S TXSXC 1.45 0.416 TIME X SITE X CANOPY 3.231 2 1.161 M S T X S X C / M S E 2 9.17 0.000 Error 2 33.844 192 0.176 E) Lichens s = 0.7577 Sum of Degrees of Mean Source of Variation Squares Freedom Square F-test F p Between Plots SITE 17.986 1 17.986 MS s/MS EI 9.83 0.002 CANOPY 16.192 1 16.192 MS c /MS sxc 13.75 0.168 SITE X CANOPY 1.178 1 1.178 MS sxc /MS EI 0.64 0.424 Error 1 175.597 96 1.829 Within Plots TIME 1.728 2 0.864 M S T / M S T X S 1.01 0.498 TIME X SITE 1.713 2 0.856 MS TXS/MS E 2 4.33 0.024 TIME X CANOPY 0.324 2 0.162 M S T X C / M S T X S X C 18.00 0.091 TIME X SITE X CANOPY 0.018 2 0.009 M S T X S X C / M S E 2 0.05 0.916 Error 2 36.993 192 0.198 17 W o o d y Plants -TC-c 95 year of sampling 3—TC-o _ 0 _ T L - c -TL-o 95 year of sampling 96 80 60 40 20 0 -20 -40 -60 -80 ^ # ^ TC-c TC-o TL-c TL-o Site Gramino id s 95 year of sampling Figure 2.4 Percent cover (left side) and change in percent cover (right side) of the five life forms from 1994 to 1995 (on left) and from 1995 to 1996 (on right) in the open and closed stands at Tulip Creek and Twin Lakes. The boxes contain half of the sample plots, the lines connect plots up to 1.5 box lengths from the box edges and points falling further then 1.5 box lengths from the mean are shown as discrete circles or stars. 18 Bryophytes year of s ampl ing » - T C - c —o— TC-o m TL-c —O—TL-o Lichens 94 95 96 year of sampl ing Figure 2.4 (continued) 19 stand, while at TC it was much higher in the open stand (CANOPY X SITE interaction, Table 2.2, Fig. 2.4). Forbs were the only growth form that showed a trend with time that was relatively consistent in all four stands (i.e., the marginally significant TIME main effect). This trend was a relatively stable cover from 1994 to 1995 and an increase froml995 to 1996. In. the two open stands, a decline in forb cover was observed in 1995 on less than half of the plots. At TC-o, very large increases and decreases of forb cover were found on a few plots. At all sites, the number of forb species was larger than the number of woody plant species. A few of the forb species showed larger fluctuations, but most forb species showed smaller changes. At Twin Lakes, Arnica cordifolia increased the most (4% and 5.5% at TL-c and TL-o, respectively), and the decline in cover in the open stand observed in 1995 was mostly due to a decrease of Antennaria neglecta. At TC- c, Aralia nudicaulis showed an increase in cover from 6% to 10%, even though this species was found on only three plots. At TC-o, the patterns observed for the most abundant species were not consistent. Trifolium repens declined from ~ 4% in 1994 to almost zero by 1996, but Lupinus sericeous and Potentilla gracilis steadily increased. Cover of Centaurea maculosa peaked in 1995, while the combined cover of Medicago lupulina and Trifolium aureum (which could be misidentified in the juvenile stage) was lowest in the same year. Graminoids were the only growth form whose cover was significantly higher in the open than in the closed stand at both Twin Lakes and Tulip Creek in all three years (CANOPY main effect, Table 2.2, Fig. 2.4). However, the changes in cover over time depended on site (significant TIME X SITE interaction). At both Twin Lakes stands, graminoid cover showed a decrease in 1995 followed by an increase in 1996, with an increase followed by decrease in the TC-c stand and a decrease for two years in a row in the TC-o stand. In the closed stands at both sites, graminoid cover changed only slightly on the majority of plots from 1994 to 1996. While the average changes in cover were similarly small in the two open stands, both larger increases and decreases were more common. In both of the open stands, the cover of Calamagrostis rubescens declined by > 3% from 1994 to 1995, but at TL-o, this grass increased again in 1996 and an unidentified Festuca species increased from 5% to 60% in one plot. In the closed stands, no single graminoid species fluctuated by much more than 2%. The changes in bryophyte cover over time depended on both site and canopy cover (significant TIME X CANOPY X SITE interaction). At TC-o, where average bryophyte cover exceeded 25% in 1994, bryophytes decreased to less than a third of their former cover over the 20 two-year period. However, during both yearly intervals bryophyte cover increased in 25% or more of the plots in this stand. In contrast, at TC-c bryophyte cover remained nearly stable, with fluctuations less than 10% in all plots. At TL-o bryophytes decreased in 1995, but increased slightly again in 1996. At TL-c, they declined slightly in both years. The large decreases in bryophyte cover at TC-o, as well as the smaller decreases in both Twin Lakes stands, were largely due to decreases in Brachythecium, the most abundant moss in all four stands. At TL-o, Brachythecium declined to less than 25% of its former abundance over the two-year time period. However, its proportion of the bryophyte cover, which was 86 % in 1994, declined to a much smaller extent. Cover of lichens also fluctuated in a different pattern at the two sites (TIME X SITE interaction). Lichen cover decreased during both yearly intervals at Twin Lakes, especially in the open stand. The decrease in lichen cover at TL-o was due to declines of a large proportion of the Peltigera and Cladonia species at the site. Fluctuations in lichen cover were small at TL-c and in both Tulip Creek stands. At all sites, changes in lichen cover were rmnimal on most plots, especially at TC-c, where lichen cover was very low. On occasional plots in the open stands at Twin Lakes and Tulip Creek, large increases or decreases were observed (Fig. 2.4). For most growth forms, the frequency of large cover changes at a site was somewhat proportional to the initial mean cover, with large changes in cover often being more common at sites where initial mean cover was large. At TC-c the percent cover of all growth forms changed very little in most plots, but a few large increases in herb cover were observed. At TC-o large changes in cover were especially common, and often large increases in some plots were offset by large decreases in others. In many plots where cover of one growth form changed greatly in one of the yearly intervals, it changed less during the other interval and often in the opposite direction. A violation of the assumption of ANOVA that the residuals are normally distributed was observed for bryophytes and lichens in 1996 (Appendix 2.2); therefore the P-values of the main effects and interactions of these two growth forms are not completely accurate. 2.4.2 WEATHER Values for total precipitation and average temperature at 12 noon for the growing season from May 1 to the day preceding vegetation sampling are shown in Table 2.3. The patterns of rainfall distribution during the growing seasons are shown in Appendix 2.3. At Castlegar Airport, total precipitation was highest in 1996, followed by 1994, and lowest in 1995. 21 Table 2.3 Summary of weather data at Castlegar Airport, Twin Lakes and Kikomun Creek Park from May 1 to the day preceding vegetation sampling. Total Average temperature precipitation at 12 noon Weather station Year (mm) (°Q Castlegar Airport 1994 156.3 20.2 1995 95.5 21.3 1996 216.8 17.8 Twin Lakes 1994 135.5 18.1 Kikomun Creek 1994 67.1 20.3 Park 1995 186.8 19.2 1996 126.6 17.9 22 During the three-year time span, Kikomun Creek Park received the highest amount of precipitation in 1995 and the lowest amount in 1994. At both weather stations, the average temperature at noon was highest in the year with the lowest rainfall and lowest in the year with the highest rainfall. 2.5 DISCUSSION 2.5.1 VASCULAR PLANTS In this study, percent cover of grasses was significantly higher in open than in closed stands at both Twin Lakes and Tulip Creek. However, cover of woody plants was higher in the open stand only at Tulip Creek. Cover and biomass of all vascular understory plants combined have been observed to be higher in open forests than in more closed stands in a variety of conifer dominated ecosystems, including the IDF (Anderson et al., 1969; Clary, 1975; Dodd et al., 1971; Tisdale and McLean, 1957). The overstory affects levels of light, precipitation throughfall, and root competition (McLaughlin, 1978). At Tulip Creek, the differences in basal area, canopy cover and number of trees between the two stands were more pronounced, suggesting that shrubs are affected only by large differences in overstory attributes.. However, the CANOPY factor did not represent treatment levels in a randomized experiment (i.e. the plots were pseudoreplicates within the two canopy types), therefore it cannot be assumed that the differences in vegetation between the open and closed stands are caused only by differences in overstory density. It is possible that the differences in overstory stand structure in the two stands at each site were not entirely due to disturbance history, but due to differences in soil moisture or nutrient conditions of a magnitude not detected at the site series level. This was more likely at Tulip creek, where the closed stand appeared to be slightly moister and the slope gradient was slightly lower than the open stand. AATule it is known that temperature and available moisture affect the rates of physiological plant processes, such as photosynthesis and respiration, the optima of temperature and soil moisture have been determined only for a very limited number of plant species, generally under controlled conditions. Clary (1974) documented a nearly linear increase in yield of herbage with increasing average yearly precipitation from a preliminary study of several sites with different precipitation regimes. He hypothesized that this precipitation from year to year, and that responses to annual weather differences would be larger where the forest overstory is less dominant. 23 In longleaf pine (Pinus palustris) forests in Louisiana, the biomass of herbaceous vegetation was related to the rainfall in the May-September period (Grelen and Lohrey, 1978), but no such studies have been performed in interior Douglas-fir ecosystems. Over the three-year duration of the present study, no evidence supporting clear relationships between the percent cover values of woody plants, forbs or graminoids and the weather variables recorded in the same year was found. Since the IDF is characterized by a dry climate, it would be expected that moisture is the limiting factor, and that therefore higher rainfall would result in higher plant cover. At Kikomun Creek Park, rainfall was highest in 1995, but at Twin Lakes most growth forms had their lowest cover in this year. In 1994 rainfall was lowest, but cover of most growth forms was highest in this year. At Castlegar Airport, the 1996 growing season was twice as wet as the preceding year, which was the driest of the three seasons examined, but only forbs in the open stand showed the lowest cover values in 1995 and the highest in 1996. Cover of woody plants in the closed stand also decreased in 1995 and increased again in 1996. It is possible that winter precipitation, or the rainfall in the preceding growing season were important to the current year's growth by influencing water availability and stored plant reserves, respectively. Temperature also did not appear to be directly related to vegetation cover. Because the higher temperatures associated with low rainfall years result in increased potential evapotranspiration, they would expect to magnify the effect of low precipitation. Temperatures and precipitation in fall, winter and spring were not recorded, but it is possible that they affected plant growth. Also, the weather stations were located 8 km or more from the study sites and were situated at lower elevations. The data from Kikomun Creek Park and Twin Lakes suggest that the weather data from the nearest weather station inadequately described the weather at the research site. Adjacent stands experience very similar weather; therefore, if weather were the overriding factor affecting changes in vegetation cover, these changes would be similar in adjacent stands. This was observed at Twin Lakes, but at Tulip Creek greater differences were found between changes in cover in the open and closed stand. For example, forbs decreased in cover and then increased again at TC-o, but increased for two consecutive intervals at TC-c. At Twin Lakes the differences in tree density, basal area and canopy cover between the open and closed stand were smaller than at Tulip Creek, suggesting that the density of the overstory or other site factors may affect the influence of weather. Changes in vegetation cover at nearby sites were also observed to be different by Brown and DeByle (1989), who found that during a 24 two year period shrub cover increased from 42 to 65% in a Colorado aspen stand, while it decreased from 33 to 27% in a nearby aspen mixedwood stand. However, the differences in overstory density or other site factors were not reported. This indicates a need for stand-level replication of treatments, in order to avoid mistaking chance events affecting only one stand for treatment effects. The vascular growth forms never increased or decreased in unison in any of the stands, indicating that changes in conditions favoured some growth forms over others. Differences in the year-to-year changes in cover also occurred on the species scale, with different species within each growth form showing different responses. Growth form- or species-specific responses to drought have been documented in the northern US Rocky Mountain states, where generally increasing trends in vegetation cover after disturbance were temporarily reversed. For example, Physocarpus malvaceous, Spirea betulifolia, and herbs were reduced in cover in 1986 in areas that were either logged, or logged and slashburned, nine years previously, while other shrubs, such as Salix scouleriana and Symphoricarpos albus were not affected (Geier-Hayes, 1989). Stickney (1980), who studied vegetation recovery after logging and slashburning in twenty stands in two different study areas, found that in most stands herb coverage was reduced in the drought year of 1976, but shrub cover was usually not affected. Unfortunately, neither of these studies included control plots in undisturbed areas, nor were precipitation values reported for either the drought years or for normal years. 2.5.2 CRYPTOGAMS Bryophytes and lichens differ greatly in their biology from vascular plants, and their growth is governed by different factors. Many bryophytes can photosynthesize in temperatures near freezing, and lichens photosynthesize best at temperatures below 15 °C, which is lower than the photosynthetic optima of most vascular plants (Schofield, 1985; Vitt et al, 1988). Most bryophytes and all lichens lack water-conducting tissues. They are able to photosynthesize only when moist, and have to obtain all nutrients from atmospheric deposition. Lichens and most mosses become dormant when they dry out and are often very drought tolerant, while liverworts are often much less able to tolerate drought (Vitt et al, 1988). ft is likely that mosses and lichens could have started growing before May 1, the date of the beginning of weather data collection for this study. On the whole, bryophyte growth rates are slow compared to those of most vascular plants (Schofield, 1985), and according to Vitt et al (1988) lichen growth rates of more than 1 cm per year are restricted to foliose lichens in very moist habitats. Therefore, unless colonization by a large number of individuals occurs, 25 increases in moss and lichen cover would be expected to be very gradual even in a good year, and if damaged they would only slowly regain their former cover. The large increases in lichen cover observed in a few plots at TC-o and TL-o are difficult to explain; they may have been due to shifts in plot location or estimation error. 2.5.3 DISTURBANCE AND ESTIMATION ERROR Plant pathogens, as well as herbivory by large ungulates, small mammals or insects, can all reduce plant cover. No obvious signs of any of these were observed in the present study, but it is possible that herbivory or diseases occurred during the winter or early in the growing season, or that individual plants could have been completely removed or decayed. By late summer the evidence of such disturbance could be masked, but cover could still be affected. Changes in vegetation can be recorded even where none exist as a result of measurement error. Variations in vegetation cover observed in undisturbed areas are often smaller than successional changes in disturbed areas, increasing the chance that measurement errors could be equal to, or exceed, the actual variation. On some of the photographs of the vegetation, the cover of abundant species appeared similar, but recorded values varied. The exact extent of these errors could not be determined for any growth form. Kent and Coker (1992) and Grieg-Smith (1964) recognized that estimates of plant cover are subjective, varying from one observer to the next, and stated that observers tend to overestimate cover of species that are in flower, attractive, conspicuous and known to the observer. In the present study, plants of most species tended to occur dispersed throughout the plots intermixed with other species, rather than occupying the whole space within a clearly defined outline, which complicated cover determination. Plant covers in the present study were assessed by different people each year, and in some cases two different groups of people estimated plant cover on different plots at the same site during the same year. The effect of using different observers on plant cover estimates was studied by Sykes et al. (1983), who found that total vegetation cover estimates made in the same 4 m2 plots by different observers were up to 20% higher or lower than the mean of all observers, and that variability of estimates was higher in larger plots. Individual observers showed some bias toward either overestimating or underestimating percent cover relative to other observers. Repeated cover estimates of the same plot made a few days apart by the same observer were also different from one another, but this variability was smaller than the between-observer variability. This suggests that there is an advantage to using only a single observer. Sykes et al. (1983) also found that plant cover estimates by different observers, or by 26 the same observer on subsequent occasions, were more consistent for plant species with large leaves than for fine-leaved species such as grasses. Leps and Hadincova (1992) noted that when two trained observers examined 5 X 5m plots, each observer missed several of the rarer plant species found by the other observer. Dropping a series of pins from a frame provides a more objective assessment of cover (Kent and Coker, 1992), but even if it were possible to locate the frame at exactly the same spot on successive occasions, the plant leaves could be slightly displaced. Even a displacement of less than 1 cm could change whether or not a small or narrow-leaved species is detected. Kent and Coker (1992) believed that problems of subjectivity of cover estimates may have been overestimated, and considered estimation of cover a valid sampling method with the advantage of being more rapid to use than more objective methods, such as pin frames. Leps and Hadincova (1992) did not consider data from pin frames reliable for any but the most frequent species at a site. The number of species detected also varied from year to year, with more species being observed in 1995 and 1996 than in 1994. This may be due to either improvement of plant identification skills of the observers, a more thorough search effort in the later years, growth of plant individuals near plot margins into the plot, or colonization by new species. 2.6 CONCLUSIONS AND RECOMMENDATIONS All plant growth forms exhibited year-to-year fluctuations in cover in at least one of the four stands, but these trends were not consistent among growth forms, suggesting that environmental changes favoured some growth forms over others. At Twin Lakes, where the differences between stands in overstory and edaphic (soil) factors were less than at Tulip Creek, the vegetation trends observed in the two stands were generally also more similar. Only forbs showed similar trends at all sites; this was essentially no change in cover during the first yearly interval and a cover increase in the second interval; however, this trend was only marginally significant. At both sites, grarrurioid cover was higher in the stand with lower canopy cover at all sampling periods. This suggests that graminoids are most sensitive to the levels of light and other resources, which differed among the two adjacent stands. No clear relationships between the cover of any plant growth form and the recorded weather variables were observed, and higher precipitation did not usually result in higher cover. Vegetation development was not necessarily similar in the closely adjacent stands, 27 which received very similar weather. This suggests that either estimation error or other factors that were not measured in this study influenced plant cover. The following recommendations could improve studies of vegetation change over time: • experiment with expressing the weather data over different time periods preceding sampling, e.g. one month, two months, to determine if they relate to vegetation cover. Also, alternative measures of temperature data, that have been shown to be related to plant growth, such as growing degree days, could be used. • sample soil moisture to determine the amount of water actually available to the plants, and examine how soil moisture is related to temperature and precipitation. • test whether temperature or precipitation in winter or in the previous year is related to vegetation abundance in the current year. • place a weather station on-site for accurate weather measurements. • sample at the same phenological time every year to minimize variation due to differing stages of phenological development. • gather more data on herbivory, and/or use exclosures for large mammals. • use as few different observers as possible, and have all observers assess vegetation in several plots to obtain an estimate of variability caused by different observers. If each observer assesses a few plots several times, an estimate of the precision of each observer could also be obtained. 28 3. RESPONSE OF UNDERSTORY VEGETATION TO BURNING AND THINNING 3.1 INTRODUCTION Prescribed burning in forests can be used to manipulate stand structure and understory vegetation. Fire affects vegetation in two ways: a) it reduces the available growing stock by Idlling all or portions of some plants and b) it changes the growth environment by affecting temperature and the availability of plant resources, such as radiation, water and nutrients. These changes in the vegetation and the environment can facilitate or retard vegetative regeneration by sprouts, stolons or fragments, or germination from seeds or spores. 3.1.1 MEASURES OF FIRE SEVERITY To relate plant response to fire severity, it is necessary to quantify fire severity. Ryan and Noste (1985) proposed a method of classifying fire severity using four ground char classes (unburned, and low, moderate, and deep ground char) to characterize the heat pulse down. These ground-char classes are based on the quantity and appearance of any remaining forest floor and on the degree to which the mineral soil is altered. This system has been used in several American studies to quantify fire severity (e.g. Brown and DeByle, 1989; Geier-Hayes, 1989; Stickney, 1985). Brown and DeByle (1989) found that the percentage of the forest floor mass and fine fuels consumed was related to Ryan and Noste's (1985) char classes, but no statistical test was performed. Schimmel and Granstrom (1996) assessed the severity of small experimental fires differently, using four classes based on available fuel and duration of smoldering combustion. These classes were related to the duration of lethal temperatures above 60° C, but again this was not statistically tested. Ryan and Noste's ground char classes (1985) constitute a discontinuous variable, which is also somewhat subjective. Wade (1986) recommended using the depth of the forest floor burned, a continuous variable, as a measure of fire severity. However, Yearsley (1993) found that depth of burn was not a reliable indicator of the duration of temperatures of either 60, 70, 80, or 100° C, or maximum temperatures at several forest floor depths. This may be due to variability in the depth, moisture gradient and packing ratio of the remaining forest floor, or the porosity, water and organic matter content of the soil, which affect below-ground transfer of heat (Feller, 1982; Hungerford et al, 1990; Hartford and Frandsen, 1992; Wade, 1986). If the forest floor, which acts as a barrier to downward heat flow, is completely removed, the heat 29 generated by combustion of surface fuels becomes an important determinant of the downward heat pulse (Wade, 1986). Amour et al. (1984), observed a greater impact of fire on graminoids in areas of greater duff consumption, suggesting that the depth of duff removed may be an indicator of plant response, especially if a large proportion of plant regenerative organs are contained within this layer. However, no studies have systematically determined if depth of burn can be used to predict plant response in dry Douglas-fir forests. 3.1.2 COMMUNITY RESPONSE TO DISTURBANCE .The plant community as a whole integrates the response of many different species to fire, as estimated by measures of species richness and turnover rates. While there may be an initial drop in the number of species present above-ground, the number of new species appearing has been found to exceed the number of species lost within a few years, resulting in a plant community with a higher species richness (e.g. Halpern, 1989; Lyon, 1971). An increase in the number of niche spaces can result from the generally patchy nature of fires, which depends on topography, climate, and ecosystem (Agee, 1993; Malanson, 1987; Rowe, 1983; Wein and El-Bayoumi, 1983; Wright and Bailey, 1982). Local extinctions combined with coionization of new species has resulted in species turnover after burning (e.g. Brown and DeByle, 1989; Lyon, 1971) and after beetle kill of overstory trees (e.g. Stone and Wolfe, 1996). In undisturbed forests, a small amount of apparent species turnover has also been observed (Chapter 2, 0kland and Eilertsen, 1996), but this species turnover in undisturbed sites has not been compared to the gain and loss of species after fire. To provide a framework for the analysis of the changes in the plant community during succession, Halpern (1989) separated the species in the post-treatment community into two groups. Residuals are species that survived the disturbance, or recolonized the immediate area where they were found before, and invaders are species that were not present above ground in the pre-disturbance community, or restricted to locally disturbed microsites. The term "residuals " could result in confusion with statistical residuals; therefore residual species are referred to as persisters hereafter. Plants regenerate after fire either by sprouting, from seed, or from vegetative propagules, which are favoured by different factors. Each of these modes of regeneration is important mostly for either persisters or invaders (Halpern, 1989). The factors affecting each of these regeneration methods are discussed below. 30 3.1.3 REGENERATION BY SPROUTING Resprouters often compose the majority of persisters (e.g. Halpern, 1989). Partially top-killed shrubs maintain apical dominance by secretion of growth hormones, while complete killing of the crown stimulates resprouting (Noste and Bushey, 1987). Most sprouting species in Douglas-fir ecosystems regenerate from root crowns (e.g. Amelanchier alnifolia, Rosa spp.) and rhizomes (e.g. Mahonia repens, Spirea betulifolia, Symphoricarpos albus, Arnica cordifolia, Aster conspicuus, Calamagrostis rubescens), (Fisher and Bradley, 1987; McLean, 1969; Noste and Bushey, 1987), but some species can sprout from dormant buds on stolons (e.g. Arctostaphylos uva-ursi, Fragaria spp.), from taproots (e.g. Shepherdia canadensis, Castilleja miniata, Astragalus miser) (McLean, 1969), root networks (e.g. Populus tremuloides) and stumps (e.g. Betula papyrifera) (Brown and DeByle, 1987; Coates et al, 1990). The extent of fire damage depends on the amount of forest floor and soil heating caused, by the fire and the fire susceptibility of the plant. Fire susceptibility depends mostly on the depth of its perennating organs (Flinn and Wein, 1977; McLean, 1969; Noste and Bushey, 1987; Rowe, 1983), but the size of these organs, and the size, age, vigour, and phenological state of the plant also influence fire susceptibility (Daubenrnire, 1968; Kauffman, 1990; Noste and Bushey, 1987). 3.1.4 REGENERATION FROM SEEDS AND SPORES After slashburning, as well as after small experimental fires, plants regenerated from seeds and spores were absent from, or a minor part of, the persisters, but all important invading species originated from these propagules (Halpern, 1989; Schimmel and Granstrom, 1996). Seed can be either from the seedbank in the soil or dispersed into the area by wind or animals. The rate of germination of seeds in the seed bank depends on burn severity and the ability of the seeds to survive long burial periods without predation and decomposition, while germination rates of off-site seeds depend on the timing of the fire in relation to seed dispersal and the period in which the seeds are viable, proximity of a seed source, and suitability of the seed bed (Halpern, 1989; Hamilton, 1988). Information on seedbanks in the IDF in BC is scanty, but some species found there, such as Mahonia repens, Achillea millefolium, Arnica cordifolia, Fragaria virginiana, Hieracium albiflorum, and Aster spp. were present in the seedbank in central Idaho (Kramer and Johnson, 1987). Species with windborne seeds, such as Epilobium spp. and Tragopogon dubius (Crane et al, 1987; Lyon, 1971; Stickney, 1985), have been reported as off-site colonizers in dry Douglas-fir forests. 31 Bryophyte spores are often dispersed by wind (Schofield, 1985), but a spore bank has also been documented for boreal forests (Jonsson, 1993). The heat tolerances and optimal germination conditions of many species are not well known (Yearsley, 1993), but seeds generally have higher heat tolerances than vegetative plant tissue (Wright and Bailey, 1982). Seeds of some species, such as Mahonia repens, Shepherdia canadensis and Symphoricarpos albus require cold stratification for germination (Noste and Bushey, 1987), while others (e.g. Ceanothus spp.) require scarification by heating (Kauffman, 1990). Lodgepole pine has serotinous cones, which will open at temperatures of 45° C to 50° C to release seeds (Fisher and Bradley, 1987; Wright and Bailey, 1982). As for many pioneer bryophytes and herbaceous plants, mineral soil is a better seedbed than duff for seedlings of many tree species. For forest regeneration, it is sometimes best to have partial exposure of mineral soil to avoid overstocking (Walstad and Seidel, 1990). 3.1.5 REGENERATION BY FRAGMENTS AND OTHER VEGETATIVE STRUCTURES Vegetative reproduction other than by sprouting has not been reported for vascular plants in dry Douglas-fir forests, but it is important for bryophytes, and lichens. Both of these groups can regenerate from fragments (Schimmel and Granstrom, 1996; Vitt et al., 1988), or from small specialized dispersal bodies containing a few cells called gemmae for mosses, and soredia and isidia for lichens (Vitt et al., 1988). For fruticose and foliose lichens, asexual reproduction is probably more important than sexual, because only the fungal partner produces spores sexually and these spores have to germinate where they can combine with algae from the environment in a process called lichenization (Vitt et ah, 1988). Documentation of the importance and dispersal distances of vegetative propagules is poor, but theoretically bryophytes and lichens originating from these can be either persisters or invaders. 3.1.6 THE POST-FIRE ENVIRONMENT All plants require adequate moisture, nutrients, favourable temperatures and photosynthetically active radiation. The availability of these resources is to some extent determined by the ecosystem, but can be altered by fire either directly or indirectly by the reduction of the overstory and understory competiton often associated with fire. Post-burn weather fluctuations also affect temperature and moisture. It is difficult to assess the relative importance of the direct and indirect effects of fire, the ecosystem, and the post-burn weather on plant resources, and this has not been attempted in dry Douglas-fir ecosystems. 32 Of the ecosystem attributes affecting plant response, moisture regime seems to be the most important and well documented. For example, Festuca idahoensis and Amelanchier alnifolia decreased after burning in drier ecosystems, but generally increased in moister ones (Clark and Starkey 1990; Wright, 1972). Percent cover of shrubs and herbs increased much faster in ravines than on drier upland areas after a wildfire in a dry Douglas-fir ecosystem in Montana, and cover of seeded grasses increased faster in moister than in drier ravines (Crane et al., 1987). The effect of post-disturbance weather patterns has been poorly documented, but herbs seem more sensitive to changes in precipitation than shrubs. In Douglas-fir and subalpine-fir ecosystems in Montana and Idaho, drought years with 33% to 56% of normal precipitation resulted in reduced herb cover after years of post-fire increase, but only some shub species were affected (Crane et al, 1987; Geier-Hayes, 1989; Stickney, 1980). Mortality of trees and other vegetation after fire decreases shading and increases surface albedo, which results in higher daytime and lower nighttime temperatures (Hungerford et al, 1990; Yearsley, 1993). Shading was not related to density of germinants from the seedbank in moist interior cedar/hemlock and subalpine forests (Yearsley, 1993), but may play more of a role in drier forests. Grasses, which require elevated temperatures to initiate growth, often start to grow earlier in the spring if shading is removed (DauberLmire, 1968). Shade removal by burning increased the availability of photosynthetically active radiation in tallgrass prairie (Knapp, 1984), and the higher leaf water potentials observed in Andropogon gerardii on burned prairie were attributed to warmer soils leading to faster root growth. Hungerford et al. (1990) also considered warmer soils to be beneficial to root growth, except in very hot, dry sites, where root mortality can occur. Manual thirvning can be used to reduce the overstory canopy cover in an effort to mimic the tree mortality occurring during fire. Thinning without fire also results in decreased shading, increased precipitation throughfall, and reduced root competition (McLaughlin, 1978). Riegel et al. (1995) found that, in ponderosa pine forests, tlrinning affected light levels much more than air or soil temperatures or belowground resources, such as soil moisture and nutrients. However, cutting of tree roots growing into the plots by trenching resulted in higher moisture and nutrient levels, indicating that complete removal of the canopy would increase soil moisture and nutrients as well. After tHnning, the production or cover of understory species increased in various ecosystems (Crouch, 1986, McConnell and Smith, 1970). The direction, magnitude and importance of changes in soil moisture regime after a burn depend on the interactions between burn severity, vegetation cover and thermal regime 33 and are difficult to predict (Hamilton, 1988). Much of the effects of burning on soil moisture is due to the effects of overstory reduction. Losses of water by foliar interception and transpiration are reduced, but evaporation rates are increased (Hungerford, 1990). Clark and Starkey (1990) attributed the increased production of burned grasses partly to decreased consumption of water by overstory shrubs. However, burning can also affect soil moisture directly. Increases and decreases of soil moisture in the rooting zone have both been found after the formation of a hydrophobic layer, depending on its depth (Hungerford et al, 1990). Soil porosity and water infiltration rates were reduced after a burn near Kamloops in the Very Dry, Hot IDF subzone (Strang and Johnson, 1988). Canopy reduction could reduce nutrient uptake by trees, and increase decomposition rates due to warmer daytime temperatures. Feller and Kimmins (1984) demonstrated that more nutrients were leached after clearcutting than in an undisturbed coastal forest, but that leaching was drastically increased after clearcutting followed by slashburning. In dry ecosystems, fires are often the primary mechanism of decomposition of woody debris (Kauffman, 1990; McNabb and Cromack, 1990). Nitrogen is volatilized at temperatures above 200° C, but nitrogen that remains on the site may become more available (Hungerford et al, 1990, McNabb and Cromack, 1990). Cations, which are found mainly in the mineral soil, are generally not lost to a great extent and also become more soluble (McNabb and Cromack, 1990) resulting in soil pH increases (Hungerford et al, 1990; McNabb and Cromack, 1990; Strang and Johnson, 1988). 3.1.7 VARIABLES RELATED TO POST-BURN RESPONSE The predominant methods of regeneration differ for persisters and invaders; therefore any one set of post-burn conditions should favour them differentially, and each group should be most sensitive to a different set of variables descriptive of the disturbance and pre- and post-burn conditions. Schimmel and Granstrom (1996) hypothesized that with increasing depth of burn, rhizomatous species decrease linearly, species that depend on postfire seed dispersal increase linearly, and seed bank species increase at low depth of burn, but decrease at high depth of burn where the forest floor containing the seeds has been removed. However, they proposed no quantitative units for either depth of burn, plant response, or the time period under consideration. Large areas of the IDF zone in BC fall within the Fraser Dry, Cool IDF variant (IDFdk3), where feathermosses, such as Pleurozium schreberi, are important components of the climax 34 understory on mesic sites (Steen and Coupe, 1997). While a well-developed moss layer has been reported in other dry Douglas-fir forests prior to burning (e.g. Crane et al, 1987), the vegetation response to fire in moss dominated communities has not been compared to that in herb dominated communities. 3.2 OBJECTIVES The present study was performed to characterize the effects of fire on understory vegetation in the IDF, and to determine the influence of the pre-existing vegetation and the reduction of overstory competition on plant response to disturbance and on plant community development over time. Specifically, the following objectives were addressed: 1. To determine if changes in abundance and species richness Of persisting and invading species during the first two growing seasons after disturbance are related to fire severity. 2. To determine if the understory vegetation after fire is related to the composition of the pre-treatment understory vegetation, site, overstory attributes or the reduction of overstory competition by thinning, and whether these variables affect the response of understory plants to fires of different severities. 3. To compare the effects of fire and overstory reduction on understory growth forms and abundant species in two different plant communities (moss and herb) in a descriptive manner comparable to other studies of fire succession in dry interior Douglas-fir forests. 4. To describe the extent to which the response of the understory after a reduction in overstory competition alone is similar to the response after fire combined with overstory reduction. The first two objectives were addressed by correlation analysis and multiple regression analysis of factors potentially related to the response of persisting and invading species. The last two objectives were met by graphical comparison of the response of understory plants to the different treatments. 3.3 METHODS 3.3.1 STUDY SITES The two study sites were located in the Knife Creek section of the University of British Columbia Research Forest near Williams Lake, BC at 52° 2' N, 121° 50' W. The soils at both 35 sites were moderately well drained Orthic Gray Luvisols with a gravelly loam texture developed on moderately calcareous morainal parent materials (Lord, 1984). The KC-1 site was a level area at 980 m elevation within the Fraser Dry, Cool IDF variant (IDFdk3) (Steen and Coupe, 1997). The overstory was dominated by Douglas-fir with scattered lodgepole pine and paper birch, and larger diameter trees had been selectively logged in 1969. The site had a mesic moisture regime and an intermediate nutrient regime, and the vegetation closely resembled that of the FdPl-Pinegrass-Feathermoss climax plant association of the zonal site series (Steen and Coupe, 1997). A broadcast burn of KC-1 was attempted on April 23, 1995, but the fire did not spread well, possibly due to a lack of wind and fine fuel. The KC-2 site was located at 760 m elevation in the Very Dry, Mild IDF subzone (IDFxm), which is warmer and drier than the IDFdk3. This site is represented by climate data from 150 Mile House, where the mean annual temperature was 4.2 °C and the mean annual precipitation was 414 mm (Meidinger and Pojar, 1991). The overstory consisted of uneven-aged Douglas-fir. The terrain at KC-2 was undulating, and while dominated by the zonal Fd-Pinegrass-Feathermoss site series, slightly moister and slightly drier microhabitats were also frequent (Steen and Coupe, 1997). The understory at both sites was a mosaic of more or less continuous feathermoss mats interspersed with shrubs, forbs, graminoids and lichens, and patches dominated by forbs and grarrunoids with scattered shrubs, mosses and lichens. The feathermoss mats at both sites were predominantly composed of Pleurozium schreberi. Calamagrostis rubescens was the most abundant graminoid at both sites, and at KC-2 Oryzopsis asperifolia was also important. The dominant shrub at both sites was Rosa acicularis; at KC-1, Spirea betulifolia, Vaccinium caespitosum and Linnaea borealis were also prominent, while at KC-2 Amelanchier alnifolia and Symphoricarpos albus were abundant. The most abundant forbs were Arnica cordifolia, Fragaria virginiana, and Lathyrus ochroleucus at KC-1, and Aster conspicuus, Aster ciliolatus, Antennaria neglecta and Lathyrus ochroleucus at KC-2. While the herb dominated communities occurred throughout the KC-2 site, at KC-1 they were largely restricted to skid trails and other areas where trees had been removed in 1969. Therefore, at KC-1 herb communities were likely more common than in the climax vegetation for this site. Cattle grazed on both sites during all years of the study, and deer pellets were also found. The Knife Creek area is an important winter range for mule deer, which feed on many of the vascular plant species present, depending on snow depth (Waterhouse et al. 1994). 36 3.3.2 EXPERIMENTAL DESIGN Repeatedly sampling the same area is superior to a one-time sample of ecologically equivalent sites, which received presumably similar disturbances at different times in the past, because it avoids the inherent variability in initial composition and stand history as well as the fragmentary representation of time Stickney (1985). This method also allows the randomization and manipulation of the disturbance, and therefore it was used in this study. Six different treatments were applied in this study, three in openings and three in the forest thickets. Forest openings were defined as areas where the nearest trees were at least 2 m distant from the plot centre, while thickets were defined areas with numerous small trees (< 8.5 cm in diameter at breast height) within a 2 m radius around the plot centre. The three treatments in forest openings, unburned control, light burn and moderate burn (Table 3.1), were applied to assess the effect of burn severity (Objective 1). Burn severity was manipulated by using different fuel loads. To reduce the risk of fire escape, the moderate burn treatment could be applied only to larger forest openings. Because there were few large openings at KC-2, the moderate burns were randomly assigned among larger openings at this site. The other two treatments were then assigned randomly among the smaller openings and remaining large openings. An unavoidable consequence was a disproportionate number of moderate burns in large openings, but this was accounted for in the equations by variables quantifying overstory attributes. The three treatments in forest thickets, control, thinning, and tliinning + light burn, together with the treatments in openings (Table 3.1), were used to determine if the effects of canopy reduction were similar to the effects of canopy reduction combined with burning (Objective 4). At KC-1 these treatments were assigned randomly to the plots, but at KC-2 the thiruning was done primarily to reduce fuel hazard, and all forest thickets in a contiguous area were thinned. Therefore the control plots had to be located in an area adjacent to the thinned area, in microhabitats that were as similar as possible to the thinned area. The thinned plots at KC-2 were therefore pseudoreplicates. The thmning only treatment served to separate the effects of burning and thinning from those of thinning alone. All six treatments at both sites were applied to similar experimental units, which were square plots with an area of 1 m . Two corners of each plot were marked with wire or steel • rebar stakes to facilitate relocation. In total, 196 plots were established, 96 at KC-1 and 100 at KC-2. At each site, half of the plots were located in forest openings, and the other half in 37 Table 3.1 Treatments applied to the plots in a) forest openings and b) forest thickets at the two Knife Creek sites. The codes are used instead of full treatment descriptions in subsequent figures and tables. Treatment description Number of plots KC-1 KC-2 Code a) Openings 1. No burn (opening control) 2. Light burn • 2-3 kg dry mass of fuel/m2 • up to 45% of forest floor depth consumed 8 herb 8 moss 8 herb 8 moss 8 herb 8 moss 8 herb 8 moss hOC mOC hOL mOL 3. Moderate burn • 15-20 kg dry mass of fuel/ m2 • 46% to 100% of forest floor depth consumed 8 herb 8 moss 9 herb 9 moss hOM mOM b) Thickets 4. No thinning or burning (forest control) 8 herb 8 moss 8 herb 8 moss hFC mFC 5. Thinning only 6. Thinning and light burn • 2-3 kg dry mass of fuel/ m 2 • up to 38% of forest floor depth consumed 8 herb 8 moss 8 herb 8 moss 8 herb 8 moss 9 herb 9 moss hFT mFT hFTL mFTL Total number of plots 96 100 196 38 thickets. Within forest openings as well as thickets, half of the plots were located in feathermoss conrmunities, where the cover of mosses exceeded 50%, and the other half in herb communities, where moss cover was less than 35% and forbs and graminoids were more prominent. 3.3.3 TREATMENT DESCRIPTIONS For the thinning treatment, some trees < 8 cm dbh were cut so that the remaining trees were spaced 2 m to 3 m from one another within at least a 4 m radius around the plots. Thinning at both sites was done between July 25 and August 4,1995. The exact number of trees removed in and around each plot varied. A fuel load of 2 kg/m^ (dry mass) for the light burns was chosen because it reflected natural fuel loads, and a fuel load of 15 to 20 kg/m 2 (dry mass) was chosen for the moderate burns because it approximates that of operational piling and burning. Wood moisture content was determined from the wet and dry mass of cuttings taken at the KC-2 site. This moisture content was used to determine the amount of freshly cut material needed to achieve the desired fuel load. For the first two plots of each burn severity class, the wood was weighed using field scales to calibrate the eyes, and for the following plots the fuel mass was visually estimated. The fuef on the light burn plots consisted mainly of foliage and branch sections < 2.5 cm in diameter obtained from thinning, while the moderate severity burn plots received larger woody fuel as well. The piles projected approximately 30 cm beyond the edge of all burn plots. When several plots close together were assigned to receive a burning treatment of the same severity, they were covered with a single larger pile. Burning was conducted from October 19 to 24,1995. To prevent fire escape, conditions too moist for the spread of a natural fire were chosen. Heat was applied to the piles with a propane torch until the fire was self-sustaining, fn the case of the low severity burns, it was often necessary to add some dry twigs to ignite the cuttings on the piles. The fires burned until they extinguished themselves. After burning, any large partially burned wood fragments were removed to increase uniformity among burned plots. 39 3.3.4 DATA COLLECTION 3.3.4.1 Vegetation The percent cover of each understory plant species present was recorded in all plots in June 1995 before the treatments were applied. Post-treatment assessments were done on three occasions: 1) in late June of 1996 (eight months after treatment), 2) in late July 1996 (nine months after treatment), and 3) in July of 1997 (twenty-one months after treatment). As in Chapter 2, all herbs, mosses, liverworts, terrestrial lichens, and seedlings and shrubs < 2 m in height were defined as understory plants. Because small cover percentages were difficult to judge, all cover values less than one percent were denoted as 0.5%. Nomenclature of all plants followed Qian and Klinka (1998). 3.3.4.2 Environmental Variables The pre-treatment attributes of the overstory canopy, as well as the change caused by the thinning treatments, were measured in several ways. The percentage of overstory cover directly over the plots was estimated visually through a cardboard tube 2.5 cm in diameter and 1.8 cm in length prior to the treatments in summer 1995 and again after the treatments in the summer of 1996. To obtain additional measures of overstory attributes and thinning severity, which were more closely related to the reduction in belowground competition, the basal diameter (the diameter just above the flare of the roots) was recorded for all trees and stumps within a variable radius from the plot centre. This radius increased with increasing basal diameter of the trees. This sampling scheme was chosen to account for the more far-reaching competitive effects of larger trees, while limiting the cost of sampling large numbers of small trees too far from the plot to affect the understory vegetation. The sampling radius was 2 m for basal diameters from 2-4 cm, 3 m for diameters from 4 -10 cm, 4 m for diameters from 10-20 cm, 5 m for diameters from 20-50 cm, 7 m for diameters from 50-80 cm, and 9 m for diameters > 80 cm. The pre-treatment basal area, as well as the basal area reduction, was calculated for each plot on a per hectare basis. Since the diameter of cut trees could not be measured at breast height, the basal areas were calculated from diameters at the tree base, and therefore represent a slightly different measure from basal areas reported in the literature. For each plot, the number of trees found within the above radii before the treatments and the number of trees cut were also recorded. These numbers were not adjusted to density per hectare. 40 To assess depth of burn, six depth of burn pins were placed in each plot underneath the fuel piles prior to burning as in McRae et al. (1979), except that the top of the crossbar of the pin was placed at the top of the F-horizon or at the interface of live and dead feathermoss. If the arrangement of large fuel logs on the plot prohibited regular spacing, the pins were spread out in the plot as much as possible. After the burns, but before snow covered the plots, the depth of burn to the bottom of the char layer, as well as the depth of remaining forest floor, was measured with a ruler to the nearest millimetre. Because the rooting depth of many plant species is related to the depth of the forest floor (Rowe, 1983), the percentage of the forest floor depth burned was considered to be a more biologically relevant predictor of plant response than absolute depth of burn. In some cases, areas at the plot edges escaped burning, and for each plot the percentage of the plot area where all plants were top killed was recorded. 3.3.5 DATA ANALYSIS In each plot, each species found was designated either as a persister or an invader similar to Halpern (1989). However, to assess small-scale species turnover on the plot level, only species found in the actual plot prior to the disturbance (rather than in any undisturbed area of the stand) were considered persisters. It .was impossible to determine if species found for the first time after the treatment were actual colonizers, or if they were overlooked during the initial sampling in 1995. To be certain that all true colonizers were designated as such, species of doubtful pre-treatment presence were included in the invader class. Therefore some species, which were essentially late serai species, but which were not detected in a plot by the pre-treatment sampling, were considered invaders. Using this classification scheme, the exact subset of species contained in each class varied from plot to plot, and it was possible for a species to be a persisting species in some plots and an invader in others. The total percent cover and species richness (number of species) were determined for both species classes in each plot, for a total of four plant response variables. Treatment type and severity, overstory canopy density and moss cover were measured for each plot, and while they could be represented by categorical factors, a large amount of variation was found within each treatment class, and often values overlapped considerably between treatment classes. Therefore, correlation analysis and multiple regression analysis were used to prevent the loss of information. Correlation analysis was performed to determine which independent variables were related to plant response at each post-treatment sampling 41 occasion, and which were related to each other. Regression analysis was used to determine which combinations of variables were best related to the plant response variables. The independent variables considered for the correlation analysis and for entry into the regression equations are listed in Table 3.2 together with the codes used in the text and tables. The variables were descriptive of the burn (to address Objective 1) or of the pre-treatment vegetation, the pre-treatment overstory, site, or the severity of the thinning (Objective 2). The data for the percent cover and number of species of both persisters and invaders failed to meet the requirement for normality and/or homogeneity of variances. Therefore, for the correlation analysis, the Spearman rank correlation coefficient (p) was used. For the regression analysis, the four vegetation response variables were ranked within each time period using Systat 7.0, and normalized using Blom's (1958) method. This transformation prevented the prediction of exact values of percent cover or number of species, but permitted testing whether a variable was related to a high or low value of a particular plant response at each of the sampling occasions. The Spearman rank correlations between all continuous independent variables and each of the four plant response variables were determined for June 1996, July 1996 and July 1997. The correlations between the independent variables were also determined. Two-tailed significance tests were performed using tables of critical values from Conover (1980). The a-level was set at 0.05, and, because the experiment-wise error level was not of interest, it was not corrected for the number of comparisons. For the multiple regression analysis the REGRESSION procedure in Systat 7.0 was used. To limit the inclusion of variables and interactions whose relationship with the dependent variables would likely have been spurious with no obvious biological explanation, the stepwise procedure was not used, but the variables were manually entered one by one in the order of a decreasing relationship with the understory vegetation as determined graphically. Two-way interactions between independent variables, but not higher order interactions, were entered into the equations if they were considered potentially biologically meaningful. The variables were entered into the equation in the following order: a) variables descriptive of the burns (group 1), b) pre-treatment vegetation variables (group 2), c) interaction terms between burn variables and pre-treatment vegetation, d) variables describing thinning severity (group 3) and overstory attributes (group 4), e) interactions of overstory density or thinning severity with burn variables and pre-treatment vegetation variables, f) SITE (group 5), and g) interactions of SITE with variables in any of the other groups. At each step, any variables whose coefficients 42 Table 3.2 Independent variables available for inclusion in the regression equations predicting cover and number of species of persisters and invaders. The codes are used instead of full variable names in subsequent text figures and tables. Group Variable Code 1 descriptive of burning Burn (0 = un burned, 1 = burned) BURN1 Depth of forest floor burned (%) DOB Plot area burned (%) PAB 2 descriptive of pre- Total cover of all plants combined (%) TC treatment vegetation2 Cover of moss (%) MC Cover of vascular plants (%) VC Total number of species TNS Number of moss species MNS Number of vascular species VNS 3 descriptive of thinning Number of trees cut around plot NTC severity Basal area of trees cut around plot BAC Cover of seedlings cut accidentally (%) SCC3 Seedlings cut (0 = no seedlings cut, SC3 1 = at least one seedlings cut) 4 descriptive of pre- Number of trees around plot NT treatment overstory Basal area of trees around plot BA attributes Canopy cover (%) CC 5 separates the two sites Site (0 = KC-2,1 = KC-1) SfTE not included in the correlation analysis because it separates burned from unburned plots. 2 independent variables descriptive of vegetation cover were only used in correlations with, or regressions predicting, cover of persisters, and variables descriptive of number of species were only included in correlations with or regressions predicting, number of species of persisters. In correlations with, or regressions predicting, both the cover or number of species of invaders only pre-treatment vegetation cover variables were considered because they were believed to be more indicative of seedbed and competitive conditions than the pre-treatment number of species. 3 variables not included in the correlation analysis because they affect only a few plots. 43 were no longer significantly different from zero at oc= 0.05 after the addition of new variables were removed. To assess the relative influence of the variables in the multiple regression equations, the predicted values were plotted separately against each variable, fn each plot, the scatter or "noise" was the variation accounted for by the other variables in the equation, and if the influence of a variable was small relative to that of the other variables, the relationship with the dependent variable was poor on these plots. The residuals of each of the regression analyses predicting cover or number of species of persisters or invaders were tested for normality with a Kolmogorov-Smirnov test using Minitab (Release 10.51,1995). The equality of variances and the fit of the equations were assessed by visually inspecting plots of the residuals vs. predicted values. The significance of all effects was determined at a = 0.05. 3.4 RESULTS 3.4.1 ENVIRONMENTAL VARIABLES 3.4.1.1 Overstory Attributes Canopy covers in openings ranged from 0% to 80% and basal areas from 0 m 2 /ha to 36 m 2/ha, while all canopy covers in thickets ranged from 20% to 90% and basal areas from 1.3 m 2 /ha to 37.9 m 2 /ha (Appendix 3.1). The plots in openings and thickets showed less overlap in the number of trees. This number ranged from 0 to 21 in openings, and from 4 to 36 in thickets, but only 6% of plots in thickets had less than 11 trees, and only 10% of plots in openings had more than 10. The three pre-treatment overstory variables were significantly correlated with each other. Thinning severity in thinned plots ranged from 0.8 m 2 /ha to 11.2 m 2 /ha and from 2 to 24 trees (Appendix 3.2). The two thinning variables were correlated with each other (Appendix 3.3). The estimated canopy cover was not a reliable measure of thirming, because it increased for some thinned plots, and decreased for some unthinned plots, which was believed to be due to measurement error. However, it was considered useful as a rough measure of the light conditions experienced by the understory. Because unthinned plots in openings were sometimes located within 4 m of thinned plots, trees were often cut within the radii used for basal area measurements. However, this accidental thinning never exceeded 3.3 m 2 /ha of basal 44 area or 7 trees (Appendix 3.2). Two of the control plots in forest thickets at KC-1 received light thiru-Ling and were deleted from the graphic analysis. 3.4.1.2 Burn Severity The mean depth of burn for all community types at both sites was > 1.8 cm in the moderate burn plots, and < 0.8 cm for the light burns (Appendix 3.4). The percentage of forest floor depth burned best separated the different burning severities, because the values for the treatments did not overlap. The light burn plots had their forest floor depth reduced by 0% to 45%, while the reduction for moderate burn plots ranged from 47% to 100%. The moderate burn plots fell within the moderate ground char class proposed by Ryan and Noste (1985). 3.4.2 VARIABLES RELATED TO UNDERSTORY RESPONSE Many more single variables showed a simple correlation with the response of both persisters and invaders (Tables 3.3 and 3.4) than were sigriificant in the regressions (Tables 3.5 and 3.6), which can be explained by the significant correlations between some of the independent variables. Significant (p < 0.05) regression equations predicting cover or the number of species of persisters or invaders were found for all three sampling periods. For cover and number of persisting species, the coefficients of determination (R2) ranged from 0.83 to 0.90, indicating that the variables measured accounted for a large proportion of the variation (Table 3.5). For both cover and number of invading species the predictability was quite low (R2 ranging from 0.28 to 0.47, Table 3.6). Even after transformation, the variances of the residuals for both cover and number of persisting and invading species at all three sampling occasions were not homogeneous (Appendix 3.5). The violation of the assumption of normality of the residuals for cover of persisters in June 1996, for the number of persisters in June and July 1996, and for the number of invaders in June 1996 (Appendix 3.6), may have resulted in erroneous p-values, which could have led to an unjustified inclusion or exclusion of predictor variables with p-values near the cutoff of 0.05. However, no lack of fit substantial enough to invalidate the models was found. 3.4.2.1 Persisting Species 3.4.2.1.1 Pre-existing Vegetation and Burn Severity As expected, the cover and number of persisting species in unburned plots were closely related to the total pre-treatment cover (TC) and number of species (TNS), respectively, (Tables 45 Table 3.3 Spearman rank correlations between a) percent cover and b) number of species of persisters and the continuous variables available for inclusion in the regression equations in June 1996, July 1996 and July 1997 for burned and unburned plots. The variable codes are explained in Table 3.2. Variable Group Variable Unburned Burned June 1996 July 1996 July 1997 June 1996 July 1996 July 1997 a) Cover of persisting species 1 DOB _ _ -0.59** -0.49** -0.44** PAB - - - -0.55** -0.44** -0.39** 2 TC 0.75** 0.74** 0.74** -0.20* -0.21* -0.14 MC 0.75** 0.63** 0.63** -0.34** -0.37** -0.33** VC -0.33** -0.15 -0.10 0.35** 0.39** 0.42** 3 NTC -0.06 -0.13 -0.18 0.34** 0.26** 0.11 BAC -0.04 -0.11 -0.17 0.32** 0.24* 0.10 4 NT -0.03 -0.21* -0.22* 0.25* 0.20* 0.04 BA 0.09 -0.07 -0.09 -0.06 -0.04 -0.14 CC -0.16 -0.28** -0.26** 0.07 0.08 0.01 5 SITE -0.04 0.01 -0.06 0.23* 0.34** 0.16 b) Number of persisting species 1 DOB _ -0.56** -0.56** -0.53** PAB - - - -0.52** -0.48** -0.49** 2 TNS 0.95** 0.96** 0.94** 0.25* . 0.30** 0.36** MNS 0.41** 0.41** 0.49** -0.05 -0.02 -0.02 VNS 0.83** 0.82** 0.82** 0.45** 0.48** 0.57** 3 NTC -0.08 -0.08 -0.08 0.17 0.22* 0.16 BAC -0.07 -0.08 -0.09 0.17 0.23* 0.17 4 NT -0.43** -0.43** -0.43** 0.09 0.14 0.10 BA -0.43** -0.45** -0.45** -0.12 -0.06 -0.10 CC -0.32** -0.32** -0.33** -0.04 0.01 0.03 5 SITE -0.37** -0.36** -0.39** 0.02 0.04 -0.09 significant at p<0.05 significant at p<0.01 4 6 Table 3.4 Spearman rank correlations between a) percent cover and b) number of species of invaders and the continuous variables available for inclusion in the regression equations in June 1996, July 1996 arid July 1997 for burned and unburned plots. The variable codes are explained in Table 3.2. Variable Unburned Burned Group Variable June 1996 July 1996 July 1997 June 1996 July 1996 July 1997 a) Cover of invading species b) 1 DOB - - - 0.02 0.45** 0.63** PAB - - - -0.08 0.21* 0.30** 2 TC -0.15 -0.05 0.05 -0.02 0.07 0.06 VC 0.48** 0.48** 0.43** 0.12 0.02 0.16 MC -0.37** -0.31** -0.20 -0.12 0.00 -0.08 3 NTC 0.03 -o:oi 0.00 -0.10 -0.34** -0.35** BAC 0.01 -0.04 -0.02 -0.06 -0.32** -0.33** 4 NT -0.31** -0.29** -0.38** -0.07 -0.33** -0.41** BA -0.37** -0.48** -0.48** -0.05 -0.25* -0.33** CC -0.24* -0.26* -0.34** -0.14 -0.36** -0.40** 5 SITE -0.19 -0.08 -0.14 0.07 0.19 0.10 Number of invading species 1 DOB _ 0.02 0.47** 0.69** PAB - - -0.09 0.27** 0.38** 2 TC 0.11 0.01 0.06 0.03 0.19 0.08 VC 0.49** 0.47** 0.44** 0.08 -0.07 0.06 MC -0.36** -0.28** -0.30** -0.10 0.13 -0.01 3 NTC -0.06 -0.08 0.03 -0.07 -0.37** -0.38** BAC -0.08 -0.10 0.01 -0.03 -0.35** -0.36** 4 NT -0.32** -0.29** -0.30** -0.03 -0.30** -0.41** BA -0.37** -0.48** -0.47** -0.03 -0.11 -0.27** CC -0.25* -0.29** -0.29** -0.11 -0.30** -0.33** 5 SITE -0.22* -0.10 -0.15 0.08 0.22* 0.01 * significant at p<0.05 ** significant at p<0.01 47 Table 3.5 Coefficient of determination (R2), standard error of the estimate (SEE) and coefficients of the variables in the multiple linear regression equations predicting a) percent cover and b) number of persisting species, in June 1996, July 1996 and July 1997. The variable codes are explained in Table 3.2. Coefficients Variable June 1996 July 1996 July 1997 a) Cover of persisting species R2 0.89 0.85 0.82 SEE 0.33 0.38 0.42 Constant -0.5488** -0.6183** -0.6057** DOB -0.0053** -0.0071** -0.0062** TC 0.0186** 0.0171** 0.0176** VC -0.0045**- - -BURN X TC -0.0172** -0.0195** -0.0196** BURN X VC 0.0180** 0.0141** 0.0155** DOB X VC -0.0002** -0.0001** -0.0001** BAC 0.0296* - -SCC -0.0163** -0.0205** -0.0147** CC -0.0055** -0.0051** -0.0054** SITE 0.1780** 0.2758** 0.1279* b) Number of persisting species R2 0.90 0.90 0.87 SEE 0.31 0.32 0.35 Constant -0.9257** -1.1744** -1.4870** DOB - . -0.0144** PAB -0.0076** -0.0064** TS 0.1137** 0.1254** 0.1378** BURNXTS - -0.1075** -0.1123** -0.1430** BURNXVS 0.1195** 0.1255** 0.1442** DOBXVS -0.0009** -0.0011** CC -0.0038** -0.0035** -0.0029* BAC X SITE 0.0294* 0.0371** * significant at p<0.05 ** significant at p<0.01 48 Table 3.6 Coefficient of determination (R2), standard error of the estimate (SEE) and coefficients of the variables in the multiple linear regression equations predicting a) percent cover and b) number of invading species, in June 1996, July 1996 and July 1997. The variable codes are explained in Table 3.2. Variable June 1996 Coefficients July 1996 July 1997 a) Cover of invading species R2 0.39 SEE 0.72 0.25 0.83 0.45 0.71 Constant DOB PAB MC BURN X MC DOB X MC CC DOB X SITE 1.3028** -0.0149** -0.0105** 0.0081** -0.0059** 0.9533** 0.0074* -0.0132** -0.0075** 0.0083** -0.0085** 0.0092** 0.6266** 0.0098** -0.0081** -0.0001** 0.0001** -0.0089** 0.0099** b) Number of invading species R2 0.41 SEE 0.71 0.35 0.77 0.39 0.75 Constant DOB PAB MC BURN X MC SITE BURN X SITE DOB X SITE NT X SITE 1.1866** -0.0168** -0.0099** 0.0081** -0.3255* -0.0252** 0.7203** 0.0079** -0.0176** -0.0068** 0.0112** 0.4380* 0.0086* -0.0379** 0.6549** 0.0171** -0.0151** -0.0075** 0.0085** 0.0066* -0.0216** * significant at p<0.05 ** significant at p<0.01 49 3.3 and 3.5, Figs. 3.1b and 3.2b), with other variables having only minor effects on predictability. In burned plots, however, the pre-treatment cover and number of species of vascular plants (VC and VNS) were better predictors of the response of persisters than TC and TNS, illustrating that vascular plants were the major contributors to both the cover and number of persisting species. Pre-treatment moss cover (MC) or number of moss species (MS) was negatively correlated with cover and number of species of persisters, respectively, at all three time periods (Table 3.3). However, when differences in the pre-treatment vascular flora were accounted for by the inclusion of VC or VNS in the equations, MC was not significant. This indicates that the response of persisters in moss plots was not directly influenced by the presence of a large moss cover, but that plots with high pre-treatment moss cover generally had a lower cover and number of vascular species. The expected negative relationship between fire severity, as measured by the percentage of the forest floor depth burned (DOB), and the cover and number of persisting species was also observed (Table 3.3, Figs. 3.1a and 3.2a). However, the relationship between the pre-treatment vegetation and the response of persisters depended on fire severity (DOB X VC and DOB X VNS interactions, Table 3.5). In moderate burn plots, both the cover and number of species of persisters were generally low, regardless of the cover and number of vascular species present before the burn. However, for light burn plots there was a clear relationship between the response of persisters and the pre-treatment cover and number of species of vascular plants (Figs. 3.1c and 3.2c). The only instance where this type of interaction was not observed was for the number of persisting species in July 1997, where some species in moderate burn plots had resprouted after a time lag. This resulted in a positive relationship between the number of persisting species with number of vascular species present before the treatments regardless of DOB (Table 3.5, Fig. 3.2c). As more plant species resprouted between June 1996 and July. 1997, the difference in the predicted number of persisters between burned and unburned plots decreased, except for plots with high DOB and low VC, where the predicted number of persisting species was always low (Fig. 3.2c). Although the percentage of the plot area burned (PAB) was significantly negatively correlated with both plant responses at all three sampling occasions, PAB was a significant predictor only of the number of persisting species in June and July of the first post-treatment year (Table 3.6). Of the 100 burned plots, 97 had more than 90% of their area burned; therefore the variable PAB 50 a) DOB June 1996 July 1997 X I N X I 0) N 75 E o c > o o o X I 0 20 40 60 80 Depth of forest floor burned (%) Treatment • burned o unburned 100 b) TC X I CD N "ro o X I -1.2h o X I m 0 30 60 90 120 150 180 210 Pre-treatment cover of all plants (%) 0 20 40 60 80 100 Depth of forest floor burned (%) 2.4 1.8 1.2 0.6 0.0 -0.6 -1.2 -1.8 • • _ • • 9 _l_ _1_ Treatment • burned o unburned 0 30 60 90 120 150 180 210 Pre-treatment cover of all plants (%) c) VC o X J 0 25 50 75 100 125 Pre-treatment cover of vascular plants (%) 2.4 1.8 1.2 0.6 H 0.0 -0.6 -1.2 -1.8 0 ' 0 * \ \ ± M ? » * £** * * * * * * • * Treatment o unburned • light burn * moderate burn 0 25 50 75 100 125 Pre-treatment cover of vascular plants (%) Figure 3.1 Influence of a) depth of burn, b) pre-treatment cover of all plants, and c) pre-treatment cover of vascular plants on the predicted normalized ranks of cover of persisters in June 1996 (left side) and July 1997 (right side). 51 a) DOB June 1996 July 1997 Treatment • burned H o unburned 20 40 60 80 Depth of forest floor burned (%) 20 40 60 80 Depth of forest floor burned (%) 100 b) TNS 2.4 1.8 1.2 0.6 0.0 w o> o CD Q . CO 75 - 0 6 o XI CD -1.2 -1.8 -2.4 ~\ 1 1 r , 8 o • : • . t 11 . . > • . CD E 3 5 10 15 20 25 30 °- Pre-treatment number of all plant species c) VNS 2.4 1.8 1.2 0.6 0.0 co CD o CD CL £ - 0 . 6 -1.2 -1.8 £ -2.4 CD O 1 o~ §i . . . o° • . : - • • t • • Treatment • burned o unburned 10 15 20 25 30 Pre-treatment number of all plant species CO c ra XI CD N 75 E CO CD O CD CL CO CD _Q E u c o XI CD -0.6 h -1.2 -1.8 h Treatment o unburned • light burn H * moderate burn 12 15 18 21 °- Pre-treatment number of vascular plant species °- Pre-treatment number of vascular plant species Figure 3.2 Influence of a) depth of burn, b) pre-treatment number of all plant species, and c) pre-treatment number of vascular plant species on the predicted normalized ranks of the number of persisting species in June 1996 (left side) and July 1997 (right side). 52 was very similar to the variable BURN, which discriminated between burned and unburned plots in the interaction terms in the equations. 3.4.2.1.2 Other Variables In several thinned plots large tree seedlings, which had been inventoried as part of the understory, were accidentally cut during the thinning process. The variable SCC was created to account for the missing seedlings. In most cases, seedlings of only one species, Pseudotsuga menziesii, were cut, therefore, the number of species cut (SC) was not a significant predictor of the number of persisting species. Several other variables were also significantly correlated with the response of persisters, or were significant in the regression equations (Tables 3.3 and 3.5). However, in the regressions, the relationships between these variables and the predicted values were masked by the more pronounced relationships with burn severity and the pre-treatment vegetation (Appendices 3.7 and 3.8). The correlations between the three overstory variables and the cover of persisting species were relatively low and inconsistent over time both in burned or unburned plots. However, NT, BA and CC were all correlated significantly with the number of persisting species in unburned plots, but never in burned plots (Table 3.3). In the regressions, CC was related to both cover and number of species of persisters at all sampling occasions, regardless of whether the plots were burned or not (i.e. CC was significant, but the BURN X CC interaction was not, Table 3.5). The three overstory variables were correlated with each other (Appendix 3.3), and when entered into the regression equations individually, each of them was related to the number of persisters. However, the equation with CC had the highest predictive ability. In unburned plots, the cover of persisters was similar at the two sites, but in burned plots in June and July 1997, cover was higher at KC-1 (positive correlation with SITE, Table 3.3). However, when other variables were in the regression equations, the cover of persisters was higher at KC-1 at all three sampling occasions, regardless of whether the plots were burned. While in burned plots SITE was not significantly associated with the number of persisting species at any of the three sampling occasions, unburned plots at KC-1 had a lower number of persisting species than at KC-2 at all three post-treatment time periods (Table 3.3). Since the pre-treatment number of species was also lower in unburned plots at KC-1, the variable TNS accounts for this variation in the regression equation (Table 3.5), and SITE was no longer significant. 53 While both the basal area cut (BAC) and the number of trees cut (NTC) were significantly correlated with the cover of persisters in burned plots in June and July 1996, and with their number of species in July 1996, BAC was significant only in the equation predicting cover of persisters for June 1996. In both June and July 1996, BAC was positively related to the number of persisting species at KC-1, but unrelated at KC-2 (BAC X SITE interaction). 3.4.2.2 Invading Species 3.4.2.2.1 Fire Severity Neither the cover nor the number of species of invaders could be predicted from the the depth of forest floor burned as a percentage of the pre-burn forest floor thickness (DOB) in June 1996, but the number of invading species was higher at KC-1 (negative BURN X SITE interaction). DOB became a progressively more important predictor with a positive coefficient thereafter (Table 3.6). By July 1997, it was the most important predictor variable of the response of invaders in burned plots, with a larger cover and number of species of invaders in moderate burn plots than in unburned or light burn plots (Figs. 3.3 and 3.4). In July 1996 and July 1997, the relationships between DOB and both cover and number of species of invaders differed at the two sites (DOB X SITE interactions). At low DOB, the predicted cover and number of species of invaders were similar at both sites, but at higher DOB, higher cover and number of species of invaders were predicted for KC-1 than for KC-2 (Figs. 3.3 and 3.4). For June 1996 these relationships were not yet observed, but more species had invaded the burned plots at KC-1 than at KC-2 regardless of burn severity. In the correlations for July 1996 and July 1997, the plot area burned (PAB) accounted mainly for the higher cover and number of species of invaders in burned plots (i.e. PAB functioned similarly to the variable BURN) (Table 3.4). When this function of PAB was performed by interactions of BURN with other variables in the equations, PAB was still a significant predictor for both cover and number of species of invaders at all sampling occasions, but the relationship was always negative (Table 3.5). This indicates that with other variables held constant, the cover and number of invaders actually tended to be lower in plots with a larger fraction of burned area. By July 1997, this relationship was not observed anymore for cover of invaders or for number of invading species in moderate burn plots (Figs. 3.3 and 3.4). 54 a) DOB June 1996 July 1997 0) N "55 E > O o & t5 T3 (D a. Treatment * KC-2 burned A KC-2 unburned • KC-1 burned o KC-1 unburned 20 40 60 80 Depth of forest floor burned (%) b) PAB 2.4 "D N O T3 CL 1.8 1.2 0.6 0.0 -0.6 -1.2 100 _ l _ 20 40 60 Plot area burned (%) 80 3 2.4 1.8 1.2 0.6 0.0 i -0.6 f-100 -1.2 20 40 60 80 Depth of forest floor burned (%) 100 Treatment • burned o unburned 20 40 60 80 Plot area burned (%) 100 C) MC <D N "ra E o Q) -0 -1 ,8° 0. I I I I I _-| _2 I 1 2.4 c 1.8 i_ •o ^ 1.2 E 0.6 ' o CO ^>c t4 Treatment o unburned • light burn • moderate burn 0 20 40 60 80 100 Pre-treatment moss cover (%) 0 20 40 60 80 100 Pre-treatment moss cover (%) Figure 3.3 Influence of a) depth of burn, b) plot area burned, and c) pre-treatment moss cover on the predicted normalized ranks of the cover of invading species in June 1996 (left side) and July 1997 (right side). 55 a) DOB June 1996 July 1997 Treatment *• KC-2 burned A KC-2 unburned • KC-1 burned o KC-1 unburned 20 40 60 80 Depth of forest floor burned (%) 0 20 40 60 80 100 Depth of forest floor burned (%) b) PAB -I Treatment • burned o unburned 20 40 60 Plot area burned (%) 20 40 60 80 Plot area burned (%) 100 C) MC in c T l ID N "55 E in o <D CL 1.5 1.0 0.5 0.0 TS 0) -0.5 -1 .0h -1.5 1U • : o o ° e° - ° i i • • o_ a. ^  0 20 40 60 80 100 Pre-treatment moss cover (%) m c ro ro E o in 0) o CL w • E C T3 fi !» 1.5 1.0 0.5 0.0 -0.5 W -1.0 -1.5 o 0 o " ° ° a - ® Treatment • burned o unburned 0 20 40 60 80 100 Pre-treatment moss cover (%) Figure 3.4 Influence of a) depth of burn, b) plot area burned, and c) pre-treatment moss cover on the predicted normalized ranks of the number of invading species in June 1996 (left side) and July 1997 (right side). 56 3.4.2.2.2 Pre-existing Vegetation In June and July 1996, the cover of invading species was correlated with the pre-treatment vegetation variables MC (moss cover) and VC (vascular cover) in unburned plots, but in burned plots the cover of invaders was not related to the pre-existing vegetation (Table 3.4). A significant BURN X MC interaction in the equations effectively cancelled the effect of MC in burned plots (Table 3.6). Unlike persisters, invaders had higher cover in plots with higher pre-treatment vascular cover or lower moss cover. In July 1997, high pre-treatment moss cover was associated with lower cover of invaders in unburned and light burn plots, but in moderate burn plots, a higher cover of invaders was predicted for plots with a higher moss cover (DOB X MC interaction, Table 3.6, Fig. 3.3c). As with cover, the number of invading species was negatively correlated with pre-treatment moss cover in unburned plots, but not in burned plots, at all three time periods (Table 3.4). 3A2.2.3 Other Variables The relationships between each of the overstory variables, NT (number of trees), BA (basal area) and CC (canopy cover), and the cover or number of invading species were always negative, and they were more consistent over time in unburned than in burned plots (Table 3.4). The overstory measure best related to the response of invaders in the regression equations differed for cover and number of species. The best overstory predictor variable for cover was CC, which was significant at all three sampling occasions, and this relationship was the same for burned and unburned plots (i.e. the BURN X CC interaction was not significant) (Table 3.6, Appendix 3.9a). Of the three overstory variables, NT was the best predictor of the number of invading species, but it was important only at KC-1 (NT X SITE interaction, Table 3.6, Appendix 3.9b). Cover of invaders was similar at both KC-1 and KC-2 at all time periods (i.e. no correlation with SfTE, Table 3.4). The number of species was only sporadically correlated with SITE (Table 3.4), and SITE was only a significant predictor in July 1996. The basal area cut (BAC) and the number of trees cut (NTC) were significantly correlated with both the cover and number of species in July 1996 and July 1997, but these relationships between thinning and invader response were not found when other variables were included in the regression equations. 57 3.4.3 RESPONSE OF INDIVIDUAL PLANT GROWTH FORMS AND SPECIES TO BURNING AND REDUCED OVERSTORY COMPETITON To provide detailed information on the response of growth forms and individual plant species, a list of species and the average cover for each treatment at each of the three sampling occasions grouped by community type (herb or moss) and site (KC-1 and KC-2) is shown in - Appendix 3.10. 3.4.3.1 Response to Fire There was no overlap in the percentage of forest floor depth burned was found between light burn and moderate burn plots. Therefore, when the plots were grouped by treatment class, both the negative relationship between DOB and the cover and number of species of persisters, and the positive relationship between DOB and the cover and number of species of invaders are readily apparent (Figs. 3.5-3.8). However, the responses differed depending on the plant growth form. 3.4.3.1.1 Cryptogams Over the course of this study, lichens and bryophytes were the growth forms most severely affected by burning. Both cover and the number of species of lichens were reduced to nearly zero after even light burns (Fig. 3.5 and 3.6). On unburned plots, fluctuations of the total cover and species richness of lichens were quite small. Some lichen species disappeared, and some new species were observed, but cover fluctuations were low (Appendix 3.11). In a few plots, lichens survived in unburned plot areas, and colonization of burned areas by new lichens was found only in three plots during the duration of the study. As with lichens, the cover of bryophytes was drastically reduced in herb and moss plots at both sites after both light and moderate burns (Fig. 3.5). None of the original species of liverwort survived even light burning, or recolonized the plots in the first two post-treatment years. In July 1996, some persisting mosses, such as Pleurozium schreberi, Eurynchium pulchellum and Dicranum polysetum, began to recover on light burn plots, either from tissue surviving in the charred moss mats, or by the establishment of loose moss fragments blown in from the surroundings. However, their cover was low (Fig. 3.5). As early as June 1996, several species of invading bryophytes had appeared in moderate burn plots at both sites (Fig. 3.7). At KC-1, bryophyte cover in moderate burn plots exceeded the cover of all other plant growth forms combined by July 1996, while at KC-2 the period of 58 a) Herb plots b) Moss plots Treatment Figure 3.5 Percent cover of residual species of different growth forms grouped by treatment at KC-1 and KC-2 in a) herb plots and b) moss plots. The four adjacent bars are values for June 1995, June 1996, July 1996 and July 1997 (pre-treatment, and 8,9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 59 a) Herb plots OC FC FT OL FTL OM Treatment b) Moss plots 25 T 20 4-'KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 OC FC FT OL FTL OM Treatment Figure 3.6 Number of residual species of different growth forms grouped by treatment at KC-1 and KC-2 in a) herb plots and b) moss plots. The four adjacent bars are values for June 1995, June 1996, July 1996 and July 1997 (pre-treatment, and 8,9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 60 a) Herb plots 8 T Treatment b) Moss plots 8 T KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 OC FC FT OL FTL OM Treatment Figure 3.7 Number of invading species of different growth forms grouped by treatment at KC-1 and KC-2 in a) herb plots and b) moss plots. The three adjacent bars are values for June 1996, July 1996 and July 1997 (8,9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 6 1 a) Herb plots 30 25 20 H > o o c 0) o 1_ Q- 10 15 + • Lichens • Bryophytes • Graminoids • Forbs m Woody plants KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 OC FC FT OL FTL Treatment KC-1 KC-2 OM b) Moss plots 30 > o o c o o CL KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 KC-1 KC-2 OC FC FT OL FTL OM Treatment Figure 3.8 Percent cover of invading species of different growth forms grouped by treatment at KC-1 and KC-2 in a) herb plots and b) moss plots. The three adjacent bars are values for June 1996, July 1996 and July 1997 (8,9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 62 rapid expansion in cover occurred between July 1996 and July 1997 (Fig. 3.8). In July 1996 and July 1997, the cover of bryophytes in moderate burn plots was more than twice as high at KC-1 than at KC-2 (Fig. 3.8). The colonizers at both sites were chiefly Ceratodon purpureus, Pohlia nutans, Funaria hygrometrica and Marchantia polymorpha, which were found in various combinations. A Bryum species was also abundant in one plot at KC-2. 3.4.3.1.2 Vascular Plants The fluctuations in the cover of persisting and invading vascular species in control plots were not consistent, but varied depending on the growth form, site, and plant community type. Woody, forb, and graminoid persisters generally increased in cover from June 1996 to July 1996 . on control plots, as well as on thinned or burned plots (Fig 3.5). After light burns, cover of woody persisters was near, or above, pre-treatment values by July 1997, but the relatively large gains in woody cover made on control plots were not realized on burned plots. After moderate burns, the cover of woody plants always remained below pre-burn levels during the first two growing seasons (Fig. 3.5). Most persisting shrubs and Populus tremuloides recovered by resprouting. Arctostaphylos uva-ursi and Linnaea borealis extended new stolons into the burned plot area. In some cases, burning top-killed shrubs just outside the plot and stimulated the growth of shoots inside the plot. The largest numbers of woody invaders were found in moderate burn plots at KC-1. These included Pseudotsuga menziesii, Pinus contorta, a Salix species and Betula papyrifera, which colonized the burned areas from seed. These species were observed only on a few light burn plots. Pinus contorta and Betula papyrifera were not found at KC-2, where they were not present in the overstory. The three common shrub species investigated in detail, Rosa acicularis, Spirea betulifolia and Vaccinium caespitosum, had recovered to levels similar to before treatment by July 1997 after light burns (Figs 3.9 and 3.10). However, response to moderate burns varied depending on site and community type. For Vaccinium caespitosum, which was common only in herb plots in openings at KC-1, recovery after moderate burns by July 1997 was only minimal. After moderate burns at KC-1, the cover of Rosa acicularis in moss plots and Spirea betulifolia in herb plots was still depressed below half of their pre-treatment cover by July 1997. At both sites, the cover of both persisting and invading f orbs combined in moderate burn plots remained below pre-burn forb cover even in the second growing season, but for plots receiving less severe fire, patterns of forb abundance were different at the two sites 63 A Rosa acicularis KC-1 KC-2 a) herb plots b) herb plots June 95 Treatments June July 96 Time July 97 June 95 Treatments June July 96 Time July 97 June 95 Treatments June July 96 Time July 97 June 95 Treatments June July 96 Time July 97 B Spirea betulifolia e) herb plots 5 T ^4 + June 95 I June July 96 July 97 Treatments Time Figure 3.9 Mean percent cover of Rosa acicularis and Spirea betulifolia in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 64 June 95 I June July 96 July 97 Treatments Time Figure 3.10 Mean Percent cover of Vaccinium caespitosum in herb plots at KC-1 in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 65 (Fig. 3.5). Most of the forb cover on both light and moderate burn plots was comprised of persisters. At KC-1, the cover of persisting forbs had exceeded 1995 levels after all treatments except light burns in moss plots in openings by June 1996, and it had increased further by July. At KC-2, persisting forbs were slower to develop in 1996, and by July cover was still below pre-treatment levels in tWnned + light burn herb plots and in light burn moss plots in openings. While the cover of persisting forbs remained stable or increased under most treatment regimes from July 1996 to July 1997, it declined on thinned +light burn moss plots at KC-1 due to a sharp drop in the abundance of Arnica cordifolia on three plots. At both sites, species richness of forbs in light burn plots was similar to, or even above, pre-burn levels by July 1996, with few species being lost, but some invaders appearing (Fig. 3.8). While by July 1997 the number of invading forbs found in moderate burn plots was similar to, or higher, than that in light burn plots, the number of persisting forbs was still lower after the more severe fires in the second post-burn growing season, especially in moss plots. Individual forb species showed a wide range of responses to burning. Some invading species, including Geranium bicknelli, Dracocephalum pdrviflorum, and a Cirsium species, were found only on burned plots. Most common forbs, such as Arnica cordifolia, Aster conspicuus, Fragaria virginiana, Lathyrus ochroleucus, and Taraxacum officinale, were not eliminated by light or moderate burns, but recovery rates in the two post-burn growing seasons varied depending on burn severity. Fragaria virginiana and Lathyrus ochroleucus both recovered slowly from moderate burns, but had regained pre-burn cover on light burn plots by July 1997 (Fig. 3.11). Cover of Arnica cordifolia and Aster conspicuus had recovered to pre-burn levels after both light and moderate burns by this time (Fig. 3.12). Cover of Taraxacum officinale showed a quick recovery on light burn plots, which was sometimes followed by a drop in cover between July 1996 and July 1997. On moderate burn plots at KC-1, it exceeded pre-burn levels by July 1996, but the greatest increases had occurred at KC-2 by July 1997 (Fig. 3.13). Arnica cordifolia was observed to flower in several burned plots but not in unburned plots. The recovery of persisting graminoids after moderate burns was even slower than that of forbs at both sites, with cover still below 50% of pre-burn levels in July 1997 (Fig. 3.5). At KC-1, the cover of persisting graminoids in light burn plots was close to or above that found in control or thinned plots by July 1996. At KC-2, cover of persisting graminoids was below pre-treatment levels throughout 1996 regardless of treatment regime; perhaps the weather was not 66 A Fragaria virginiana KC-1 a) herb plots 8 T KC-2 June 95 Treatments June July 96 Time June 95 Treatments c) moss plots 2 June 95 Treatments June July 96 Time July 97 d) moss plots 0.8 June 95 I Treatments June July 96 Time July 97 B Lathyrus ochroleucus e) herb plots June 95 Treatments June July 96 Time July 97 June 95 Treatments June July 96 Time July 97 f) herb plots June 95 Treatments June July 96 Time July 97 June 95 Treatments June July 96 Time July 97 Figure 3.11 Mean percent cover of Fragaria virginiana and Lathyrus ochroleucus in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in operiings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-tfunned in thickets, and FTL-thinned + light burn in thickets. 67 A Arnica cordifolia KC-1 KC-2 a) herb plots b) herb plots Treatments Time Treatments Time c) moss plots d) moss plots Treatments Time Treatments Time B Aster conspicuus e) herb plots 3 2.5 2 1.5 1 0.5 0 June 95 Treatments June July 96 Time July 97 f) herb plots 8 June 95 Treatments June July 96 Time July 97 g) moss plots 2.5 T June 95 Treatments June July 96 Time July 97 h) moss plots 6 5 4 3 2 1 0 June 95 Treatments June July 96 Time July 97 Figure 3.12 Mean percent cover of Arnica cordifolia and Aster conspicuus in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 68 KC -1 K C - 2 a) herb plots b) herb plots Treatments Time Treatments Time c) moss plots June 95 I June July 96 July 97 Treatments Time Figure 3.13 Mean percent cover of Taraxacum officinale in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Due to very low cover, data for moss plots at KC-1 is not shown. Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 69 favourable for graminoid growth. A similar spurt of graminoid recovery occurred after light burns at KC-2 during the following growing season, while at KC-1 the graminoid cover remained relatively stable or decreased over this time period. The cover or number of species of invading graminoids was always low, regardless of treatment (Figs. 3.7 and 3.8). However, in a few plots colonization by Carex rossii, which was rare before burriing, was observed. Cover of Calamagrostis rubescens on light burn plots followed a pattern very similar to that in control plots as early as June 1996, showing that this species was neither harmed nor enhanced to a great extent by light burning (Fig. 3.14). However, in moderate burn plots at both sites, Calamagrostis rubescens had regained less than half of its pre-treatment cover by July 1997. Flowering was common in burned plots in June and July 1996. 3.4.3.2 Response to Overstory Reduction In thinned plots at both sites, both the cover and species richness of persisting bryophytes and lichens showed only small changes, which did not exceed the fluctuations found in control plots, and which were unlike the almost complete elimination of these growth forms by burning (Figs. 3.5 and 3.6). In thinned herb plots at KC-1, a small decrease in cover of persisting bryophytes was observed between June and July 1996, while cover increased in the control plots in thickets over the same time period. However, a larger decrease in the cover of persisting bryophytes was observed in control plots in openings. In thinned plots at KC-1, where some large Douglas-fir seedlings inside the plots had been accidentally cut, the cover of woody persisters in June 1996 had dropped sharply from pre-treatment levels. After this initial reduction, the cover of woody plants in thinned plots fluctuated to a similar degree as in unthinned control plots. In herb plots in thickets, Rosa acicularis increased more after thinning than in control plots, but this was not observed for Spirea betulifolia (Fig. 3.9). In dunned +light burn moss plots, the cover of woody persisters recovered more slowly from fire than on light burn moss plots in openings. However, the plots in thickets had contained a larger proportion of Douglas-fir seedlings prior to the treatments. This species cannot resprout, giving the impression that the slower recovery in burned thickets was directly due to the canopy reduction. Thinning alone did not seem to affect the cover or the number of species of persisting forbs, which were similar on thinned and control plots in thickets, at both sites in both herb or moss plots (Figs. 3.5 and 3.6). Large fluctuations were observed for individual forb species even in undisturbed control plots, and these fluctuations were often not consistent among the 7 0 KC-1 KC-2 a) herb plots 40 b) herb plots 25 T June 95 Treatments June July 96 Time July 97 June 95 Treatments d) moss plots 5 c u June July 96 Time July 97 June 95 Treatments June July 96 Time July 97 June 95 Treatments June July 96 Time July 97 Figure 3.14 Mean percent cover of Calamagrostis rubescens in herb and moss plots at KC-1 (left side) and KC-2 (right side) in June 95 (pre-treatment), June 96, July 96, and July 97 (8, 9, and 21 months post-treatment, respectively). Treatments were: OC-control in openings, OL-light burn in openings, OM-moderate burn in openings, FC-control in forest thickets, FT-thinned in thickets, and FTL-thinned + light burn in thickets. 71 two sites, in openings or thickets, or in herb or moss plots. Fluctuations in cover of the forbs examined in detail were also not consistent after thinning, but large decreases in cover between June 1995 and June 1996 similar to those often found in burned plots were not observed. The patterns of change in cover or number of species of persisting graminoids in thinned plots were virtually indistinguishable from those in control plots, and graminoids also responded very similarly to light burns in openings and light burns in thinned thickets. 3.5 DISCUSSION 3.5.1 VARfABLE COMBINATIONS BEST RELATED TO PLANT RESPONSE 3.5.1.1 Fire Severity Feller (1996) stated that more severe fires favour colonizing species and retard persisters. The present study shows that this principle applied for both the cover and species richness in Douglas-fir forests when the percentage of forest floor depth burned (DOB) was used as a measure of fire severity. Links between fire severity and a reduction in the cover of some plant species characteristic of the IDF have been found by some workers (e.g. Amour et al, 1984; Brown and DeByle, 1989). However, in the former study, fire severity was only qualitatively assessed and the effects of fire severity were confounded by burning in different ecosystems, and in the latter study only the effects of fire intensity, but not fire severity, were statistically tested. Only in a few studies have burns of different fire severity been conducted in the same ecosystem, and assessments of fire severity been made by measuring forest floor consumption. The results of these studies indicate that the greater reduction of resprouting persisters related to higher fire severity is not an isolated phenomenon, but that the effect of fire severity on invaders vary greatly depending on the seedbank present at the site. For example, Schimmel and Granstrom (1996) observed that after small plot burns in Swedish boreal forests only a small number of species colonized from the seedbank and species established from windblown propagules dominated, while Whittle et al. (1997) found that the majority of species invading a jack pine forest in eastern Ontario after slashburning originated from the seedbank. However, in the study by Whittle et al. (1997), cryptogams were not sampled. In the present study, burning was done under very moist conditions. It is possible that depths of burn similar to those in the present study, but under drier conditions, would have resulted in more soil heating and higher mortality of sprouting organs and seeds in the seedbank. 72 While a relationship between cover of persisters and DOB was already established eight months after the fires, the low cover of invaders observed at the beginning of the first growing season was independent of fire severity, and a relationship between cover of invaders and DOB was only observed later. It is possible that this time lag in colonization was due to the toxicity of fresh ash. Fresh ash is more toxic to some species (e.g. Pinus banksiana) than others (e.g. Betula spp. and Salix spp.) (Thomas and Wein, 1994), which suggests that seedbed suitability for sensitive species can improve over time. However, a time lag of a year before colonization from dispersed propagules has been observed if burning occurred in August or later in the fall after seeds and spores had been dispersed (Halpern, 1989; Schimmel and Granstrom, 1996); therefore a lack of propagules is a more likely explanation. The seedbed conditions on burned plots seemed to be closely related to burn severity and invaders appeared to be more common and abundant on exposed mineral soil. Exposed mineral soil was commonly found in high DOB plots, but rarely in low DOB plots, but the fraction of the plot area with exposed mineral soil was not directly measured. Invaders were less abundant on the undisturbed forest floor and moss mats in unburned plots, and on the charred forest floor or moss mats, which dominated low DOB plots. Unfortunately, edge effects due to plot margins remaining unburned or being burned with a lesser severity could not be avoided. Some of this effect was quantified by the variable PAB (plot area burned). The positive correlation between PAB and DOB indicated that light burn plots tended to have larger fractions of unburned plot areas than moderate burn plots. In all cases PAB was a better predictor of the number of persisting species than BURN, which suggests that areas where the soil surface was not charred were important, possibly because they provided small refuge patches where some species were not burned. By July 1997, the number of persisting species could no longer be predicted from the amount of unburned plot area, most likely because the number of persisting species recovering in burned plot areas had increased and masked the number of surviving species in unburned plot areas. The cover of persisters was not affected by PAB, which indicates that the resprouters on burned plot areas contributed more to the cover of persisters than did the survivors on unburned plot areas. In the regression equations predicting the cover or number of invading species, PAB had a negative coefficient, even in July 1996 and July 1997, when the simple correlations were positive. This may be explained by the strong correlation of PAB with DOB, which had a positive coefficient in the regressions, and perhaps by correlations of PAB with other variables in the equations. 73 3.5.1.2 Pre-existing Vegetation Both the correlation analyses and regression analyses indicate that variables descriptive of the vegetation prior to the treatments were related to the cover and number of species of both persisters and invaders. However, the variables related to the response of persisters were different from those related to the response of invaders, and the importance of the pre-treatment vegetation variables depended on whether the plot was burned or on fire severity. As expected, in unburned plots where the existing growing stock of plants was not reduced by disturbance, the cover of persisters was strongly related to the total pre-treatment vegetation cover (TC) and the number of persisting species to the total pre-treatment number of species (TNS). However, in burned plots where some plants, especially bryophytes and lichens, were killed, the relationships between the pre-treatment cover and number of species of all plants and the response of persisters were weaker. The persisting species were predominantly vascular plants (Figs. 3.5-3.6), which was also illustrated by the significant positive interactions of BURN with the pre-treatment cover of vascular plants (VC) and number of vascular plant species (VNS). On plots with high DOB where VNS was low, most persisting species recolonized rather quickly or did not recover at all during the duration of the study, but in high DOB plots with a richer pre-treatment vascular flora, species recolonizing after a time lag were common. Perhaps when a greater number of vascular species is present it is more likely that some of them are delayed resprouters. Clearly, knowledge of the vascular component of the vegetation prior to treatment was important for prediction of post-burn response of persisters, especially when only a small fraction of the forest floor depth was removed. The pre-treatment vegetation affected invaders much differently than persisters. Both the cover and number of species of invaders were not significantly correlated with TC, but were positively correlated with VC and negatively correlated with MC, with MC being a slightly better predictor in the regression equations (i.e. the standard error of the estimate was lower). It is likely that many of the invaders detected on unburned plots were not true invaders, which resulted from buried or dispersed propagules that found a suitable seedbed, but that they were present, but overlooked or misidentified, during the pre-treatment sample. It is possible that a more or less continuous moss cover inhibited colonization, but it is also likely that small bryophytes and lichens, which were often found as invaders on unburned areas, were more visible on the moss mats. A high cover of vascular plants might make it more difficult to detect 74 these small species. However, as described in Chapter 2, measurement errors are common and impossible to avoid. Most invaders on burned plots were likely true invaders. The small bryophyte and lichen species often found as invaders on unburned plots were seldom detected after burning, when even small plants were highly visible on the blackened plots. None of the pre-existing vegetation variables were significantly correlated with either the cover or number of species of invaders in burned plots, and in the regression equations the positive coefficients of the BURN X MC or DOB X MC interactions counteracted the inverse relationship between MC and the cover of invaders in unburned plots. This indicates that the pre-burn moss cover did not have an effect on the ability of colonizers to become established on burned plots from windblown propagules or soil stored seeds. 3.5.1.3 Other Variables Some moderate correlations were found between variables descriptive of the overstory, canopy reduction, and site, and the response of persisters (Table 3.3). However, the contributions of these variables to the prediction of cover and number of species of persisters were small compared to the variables descriptive of fire severity or the pre-treatment vegetation. The relationships between some of these variables and the response of invaders seemed clearer (Figs. 3.3 and 3.4), but the equations contained fewer variables and their predictive ability was lower overall. Likely some of these relationships were spurious, because there seem to be no plausible explanations. Both the correlation analyses and regression analyses indicated that the cover and number of species of persisters and invaders were related to at least some of the overstory variables, or interactions between an overstory variable and site. In all cases, the coefficients in the regressions were negative, indicating that more open areas are more favourable for both the growth of existing plants and for the establishment and growth of new species. While the simple correlations between overstory variables and response of persisters and invaders often differed for burned and unburned plots, after accounting for the effects of other variables in the regressions, overstory canopy affected both burned and unburned plots in the same way (i.e. no interactions between the overstory variables and BURN were found). It is possible that light or moisture could be limiting plant cover in thickets, and that plants in more open areas can capitalize more on favourable weather conditions. However, there does not seem to be a plausible explanation as to why a larger proportion of persisting 75 species should be detected in more open areas. It could not be discerned if the higher cover of invaders in plots with lower pre-treatment canopy cover (CC) was the result of better growing conditions in these areas, greater propagule availability, or a combination of both. Perhaps in openings around the plots, the air could better circulate resulting in more windblown propagules to arrive. It is also possible that the relatively large error in the CC measurement could have resulted in a nonsensical relationship. The number of invading species was negatively related to the number of surrounding trees (NT) at KC-1 but not at KC-2. This relationship seems to be confined to moss plots, but the MC X NT X SITE interaction was not tested. While the simple correlations between SITE and the cover of persisters were not significant in unburned plots, and only significant in burned plots in June and July 1996, SITE was a significant predictor in the regression equations for all three sampling occasions and did not interact with BURN. This suggests that when other variables influencing the cover of persisters were held constant, a greater cover would be found at the moister KC-1 site in both burned and unburned plots. The site effect was prominent in burned plots, but was overshadowed by the effects of other variables in unburned plots (Appendix 3.7). The simple correlations found between basal area cut (BAC) or number of trees cut (NTC) and the response of persisters, as well as the coefficients of BAC or BAC X SITE in the regressions predicting response of persisters were positive, but due to the overriding influence of other variables, the estimated cover and number of species in more heavily thinned plots were generally similar, or slightly lower, than those in unthinned plots. Although large increases in cover of forbs and. graminoids occurring after tlrinning were sometimes observed, Crouch (1986) found this only after the removal of a large fraction of the trees, and even then the effect of thinning during the first year was small. Therefore, it is difficult to explain why the level of overstory removal should be a significant predictor only in the first year after treatment. There also seems to be no plausible explanation why the number of persisting species detected in June and July 1996 should be related to BAC at KC-1, but not at KC-2. A large proportion of the variance of the variation of both cover and number of species of invaders remained unexplained by the variables in the regression equations. This is likely due to variations in the amount of mineral soil exposure, as well as local differences in propagule availability, which were not measured in this study. Many of the factors affecting propagule availability, such as the nature of the seed bank, proximity of a seed source, seed 76 viability in the time after burning, seed predation and the effectiveness of dispersal agents, are difficult to assess and may be very site-specific. 3.5.2 RESPONSE OF INDIVIDUAL PLANT GROWTH FORMS AND SPECIES 3.5.2.1 Response to Fire 3.5.2.1.1 Cryptogams Severe impacts of fire on lichens and slow recovery have been observed by several workers. Viereck (1983) stated that the first Peltigera species usually become established 6-25 years after fires in boreal forests. In lodgepole pine forests originating from stand-replacing fires in the Montane Spruce zone in the Chilcotin area, terrestrial lichens, such as Cladonia, Cladina or Peltigera spp., were not found in stands younger than 50 years (Brulisauer et al., 1996). However, in a moister and warmer Douglas-fir/ninebark ecosystem in Montana, Peltigera species were found in small amounts in the 4th year after a wildfire on upland sites (Crane et al., 1987). Lichens have no belowground regenerative organs, and therefore are very fire sensitive. Perhaps the size and patchiness of a fire play a role in the speed of recovery by influencing the availability of spores or vegetative fragments; the small fires in this study probably represented more opportunities for recolonization than would a larger burn. The response of bryophytes to fire in this study was similar to that observed by Crane et al. (1987) after a severe wildfire in a Douglas-fir habitat type in Montana. The existing bryophyte flora was nearly eliminated, but trace amounts of climax species, such as Eurynchium pulchellum, Dicranum scoparium, and Brachythecium albicans, were detected again in a few plots in the 4th post-burn year. Crane et al. (1987) also observed a similar group of pioneering bryophytes, albeit in different proportions. While in the present study Funaria hygrometrica, Marchantia polymorpha and Pohlia nutans were most abundant, in Montana they were low in cover, but Ceratodon purpureus and Bryum caespiticium were dominant, reaching a combined cover of over 70% in the 4th post-burn year in several stands. Post-burn bryophyte colonization appears to be similar elsewhere in the Northern Hemisphere. Ceratodon purpureus and Marchantia polymorpha were common after fire in Canadian boreal forests (Viereck, 1983), in Swedish boreal forest Ceratodon purpureus and Polytrichum spp. were abundant, and Ceratodon purpureus and Funaria hygrometrica were observed on burned heathlands in France (Clement and Touffet, 1990). All of these species prefer mineral soil (Parish et al, 1996), and in this study 77 they almost completely covered all areas where mineral soil had been exposed. Since extensive sporophyte formation was found by July 1996, it is likely that in 1997 the amount of available mineral soil was more limiting than spore availability. Recovery of feathermosses occurred by new shoots emerging from moss mats that appeared to be completely dead, and by estabhshment of windblown fragments. These mechanisms of bryophyte recovery were also observed after small plot burns in boreal forests in Sweden (Schimmel and Granstrom, 1996). 3.5.2.1.2 Vascular Plants The widespread increase in vascular cover occurring between June and July 1996 on control as well as treated plots indicated that the maximum vascular plant cover in this area was not reached in June, but later in the season (Figs. 3.5-3.8). While at the warmer KC-2 site, most of the spring growth of graminoids and, to a smaller extent, of forbs, was believed to have preceded the June sampling date in 1996, while at the cooler KC-1 site vegetation developed later, and a large portion of the seasonal increase occurred between the June and July samples. With respect to the response of vascular plants to burriing, few meaningful comparisons with the literature can be made, because the nature of the disturbance and the initial species composition were different for each study examined, resulting in different plant responses. Some areas were logged before burning (Lyon, 1971, Geier-Hayes, 1989), non-native grass species were sown to control erosion (Crane et al., 1987; Lyon, 1984), and a record of the pre-burn vegetation was available only for a few studies (Lyon, 1971, Brown and DeByle, 1989). In the present study, many of the vascular species disappeared from some, but not all, burned plots. Some of these species, such as Amelanchier alnifolia, Spirea betulifolia, Symphoricarpos albus, Achillea millefolium, Arnica cordifolia, Aster conspicuus, Fragaria virginiana, Galium boreale, Viccia americana, Poa spp. and Calamagrostis rubescens, have been observed within a few years after fires of moderate to high severity in similar ecosystems (e.g. Crane et al., 1987; Lyon, 1971; Geier-Hayes, 1989; Amour et al., 1984). It is urdikely that either wildfires or prescribed burns would elirriinate these species from a stand, especially if some unburned or lightly burned patches remain as refugia for recolonization. The number of vascular persisters had increased at each post-burn sampling occasion, indicating that many species will reappear in the plots with a time lag. A time lag of one or more years before resprouting or recolonization were also observed by Crane et al. (1987) and Stickney (1985) for a variety of vascular species. The time lag was relatively short for some 78 stoloniferous plants growing back into the plots, but in larger burned areas the distances from source plants would be increased and more time would be required for recolonization. Amour et al, 1984 detected a relationship between the depth of remaining duff and the grarrtinoid cover, which was dorninated by Calamagrostis rubescens. However, this was not statistically tested. Fire stimulated flowering of Calamagrostis rubescens after burning has been commonly reported, e.g. by Crane et al. (1987) and Stickney, (1981) in Montana, and by Brown and DeByle (1989) in Idaho and Wyoming. Riegel et al. (1995), observed flowering of Calamagrostis rubescens in unburned experimental plots where both the above- and below-ground competition had been reduced, and suggested that increased water and nutrients stimulated flowering. The abundances of invader species varied widely among the burns surveyed in the literature. For example, Epilobium angustifolium, which did not show a large post-burn increase in the present study, has often been found in small quantities after fire (Brown and DeByle, 1989; Lyon, 1984; Crane et al, 1987), but in some cases, this species was the dominant colonizer (Lyon, 1971). Lyon (1971) observed an invasion of 45 new herbaceous species, especially annuals, and Amour et al. (1984) found colonization by a large number of seedlings of four annual forb species. However, in the present study the number of colonizing species was relatively low and only mosses were very abundant colonizers. The forbs Dracocephalum parviflorum and Geranium bicknellii, which were found in low abundance in the present study, have also often been observed after fires (Brown and Debyle, 1989; Crane et al, 1983; Stickney, 1985), but not always (Lyon, 1984; Amour et al, 1984). Where they appeared after fire they were often not present in the pre-fire community (Brown and DeByle, 1989; Lyon, 1971). Other workers have found a high abundance Ilamnia rivularis (Lyon, 1971; Brown and DeByle, 1989; Stickney, 1985) or Ceanofhus velutinus (Brown and Debyle, 1989; Geier-Hayes, 1989; Lyon, 1966; Stickney, 1980), which were not detected in the present study. The seeds of many of the forbs that were found before and after burning, such as Fragaria virginiana, Achillea millefolium, Arnica cordifolia, Hieracium albiflorum, Viola adunca and Taraxacum officinale, have been found in seedbanks (Kramer and Johnson, 1986; Whittle et al, 1997), making it difficult to discern the source of a persisting plant. This illustrates that the seedbank is one of the most important factors in post-fire succession. 3.5.2.2 Response to Overstory Reduction Unlike burning, which was followed by at least temporary reductions in cover and 79 number of species of all growth forms, thirLning resulted in most of the vegetation remaining intact, and fluctuations in cover and number of species were very similar to those in control plots. In plots that were both thinned and burned, the response of vegetation cover was in most cases similar to that on light burn plots in openings. This indicates that overstory removal of the severity performed in this study did not result in an understory response comparable to that after fire. None of the thirLning studies reviewed examined the effect of decreased overstory competition on bryophytes and lichens. The effects of overstory removal on vascular plants have been documented, although not in dry Douglas-fir forests. Woody plants seem to be particularly sensitive to damage by thinning operations; which may explain the decreases in shrub cover observed by Crouch (1986) after thinning in lodgepole pine forests. Crouch (1986) and Riegel (1995) observed that thinning did not affect different species or different vascular growth forms equally, but if a response was observed, this was most often an increase in cover over several years following treatment. Similar to the present study, neither of these two workers found an increase in shrub cover, even after levels of thinning more severe than in this study. Forbs and graminoids showed an increase after thinning (Crouch, 1986), but only if canopy removal exceed the thinning levels used in this study. 3.5.2.3 Complications due to Herbivory Both sites were grazed by cattle during all three years of the study, but not all plots were affected to the same degree. In general, at the KC-1 site more evidence of cattle use was observed, but cattle damage was not quantified. Some species not otherwise present on the site were found on cattle droppings in one plot, indicating that cattle can be a vector of species colonization. The moss cover of a few plots was disrupted slightly by trampling. Taller forbs and grasses, such as Aster conspicuus, Epilobium angustifolium, Castilleja miniata, Calamagrostis rubescens and Poa species were preferentially eaten. The forbs were generally bitten off above their basal leaves, having little impact on the immediate cover and identification of the species, but perhaps it could have affected the cover of the plants in the next growing season. Cattle grazing has been shown to result in changes in species composition over time, and is difficult to quantify (Fleischner, 1994; Irwin et al, 1994). Plants on overgrazed grasslands are more often harmed by fire than on lightly grazed ones (Vogl, 1974). Direct evidence of deer browsing was not found, but deer pellets were observed on the study site. Ideally, cattle and wildlife would be fenced out to avoid confounding of the burning experiment, but cattle and wildlife are 80 ubiquitous in the IDF, and relative to the effects of burning, the impact of herbivores seemed to be small. 3.6 CONCLUSIONS AND RECOMMENDATIONS 3.6.1 CONCLUSIONS In this study, the effect of fire was quite prominent. Fire severity and the pre-treatment vegetation were important variables to predict the cover and number of species of both persisters and invaders on burned plots. The cover and number of species of invaders was positively related to fire severity, but the cover and number of species of persisters was inversely related to fire severity. After light burns, the cover and number of vascular plants before the treatments were also an important predictor of cover and number of persisting species. Other variables, such as canopy cover, number of trees surrounding the plot, site, and basal area cut were also significant as predictors, but the amount of the variation explained by these variables was small and often overshadowed by other variables. However, the number of significant variables and interactions suggests that a simple picture of causality is not adequate to explain change in plant communities. It was sometimes not readily apparent how the variables in the equations were related to the quantity of resources available to plants, i.e. if the treatments resulted in increases in soil moisture, nutrients or light. The equations predicting the response of persisting species explained a much larger portion of the variation present than did the equations for invaders, indicating that many of the factors governing propagule availability and establishment and growth of colonizer species were not measured in this study. The different plant growth forms, and individual species within growth forms, varied greatly in their response to fire. Lichens and bryophytes were eliminated on nearly all burned plots and showed only minimal recovery over the duration of the study. However, an important exception occurred in moderate burn plots, where a small number of pioneer bryophyte species often colonized large areas. Vascular plants exhibited the widest range of responses. Some species were completely eliminated even after burns of light severity, while others were favoured by burns of moderate severity. Some species appeared only after burning from seeds found in the seedbank or from wind dispersed seed. After light burns, the cover of persisting vascular species had recovered to pre-burn levels by the second growing season, but an average of 10 to 33% of the species found in each plot before burning remained absent at that time. After moderate burns the mean of both the cover and number of persisting vascular 81 species per plot remained below 50% of pre-burn values. Compared to the effects of burning, the impact of thiruiing on most growth forms was small. The reduction of the above ground plant cover associated with burning did not take place, and the cover and number of species in thinned plots were usually more similar to values in control plots than to values in burned plots. A substantial reduction in woody cover resulted only in cases where some seedlings were accidentally cut during the thinning operation. While manual overstory reduction can mimic some of the tree mortality caused by fire, the effects of the overstory removal of the severity performed in this study on the understory was not comparable to that after fire. 3.6.2 MANAGEMENT IMPLICATIONS The rank transformations permitted testing of the contribution of independent variables to the percent cover and number of species, but the predicted values could not be back transformed to obtain useful predictions in the original units. Therefore, the equations cannot be used to predict the percent cover or number of species expected to be present after any other burn, nor can they be tested on data collected from another burn. While the significant variables could be used to construct equations using the original units, it is not known how valid any quantitative predictions would be in slightly different ecosystems, in other years with different weather, and for burns performed in different seasons. When deciding whether or not a prescribed burn is appropriate, or which fire severity should be used, the effects on many species have to be weighed. Even light burning resulted in an almost complete loss of the existing cover of bryophytes and lichens for at least two years, and the vascular cover was reduced as well. The reduction of plant cover was related to fire severity, and the time to complete recovery is not known. Unlike most of the pre-existing species at the site, new colonizer species were favoured by burns of higher severity. During the first two post-burn growing seasons, the number of persisting vascular plant species was also reduced on the plot scale. It is possible that at least some of the fire sensitive species found at Knife Creek prior to the treatments were present at the site only because the preceding fire free interval exceeded historical averages. This would imply that the re-introduction of a fire regime resembling the historical one would result in the permanent loss of these species due to the insufficient time for them to recolonize between fires. Since the present plant community could be labeled somewhat unnatural, it might not even be desirable to protect the element of diversity represented by these species. 82 Often burning is performed to enhance the food base for wildlife, but at Knife Creek it would be difficult to justify burning on those grounds. Shrubs that were decadent or too tall for ungulates to reach were not found, and after burning with any severity, shrub, forb and graminoid cover were temporarily reduced. Impacts on wildlife would depend very much on the size of the burned area and the availability of food elsewhere. 3.6.3 RECOMMENDATIONS To give a more complete picture of the effects of fire, and to further separate the effects of fire from those caused by other factors, the following recommendations are made: • Ideally, this experiment would have been a complete factorial design, including light and moderate burns in thickets without prior thinning, and a moderate severity burn in thinned thickets. VVTiile some additional precautions to prevent fire escape, such as coating trees with fire retardant, would be necessary, the use of these treatments would have been possible under the conditions found at the time of burning. • Burning affects both the pre-existing vegetation and alters the environment. To separate the effects of plant damage from those of a change in the plant environment, it would be desirable to include treatments such as clipping of the aboveground parts of the plants, combined with removal of various depths of forest floor. • It may be possible that some of the effects of piling and burning were actually caused by the plants being deprived of light under a pile for several months, rather than by the fire. To test whether this would be a concern, a separate piling control could be added as a treatment. These plots would receive a pile of fuel, which would be removed rather than burned. • Most species that were misidentified or missed on one or more of the sampling occasions were small. To reduce the likelihood of the true invasion rates being confounded by errors, it would be helpful to only count species that reach a certain cover in the plot, e.g., 5 cm X 5 cm (0.25 % cover). • Subdividing each plot and then recording rooting frequency would provide additional information about plant distribution, which is not given by cover estimates. This would require more time spent per plot, but perhaps this could reduce the chances of small species being missed during sampling. • Consumption of plants by cattle and wildlife may have contributed to variations in plant growth that were not accounted for by any of the variables measured. It is difficult to 83 measure the level of browsing or to distribute it evenly among plots; ideally, cattle and wildlife would be fenced out of the research area. Larger fuel piles would be desirable to avoid unburned areas near the plot edges and to achieve more consistent depth of burn throughout the plots. For management purposes, it would be desirable to generate quantitative predictions using untransformed dependent variables and the independent variables found significant in this study. It could then be determined if these equations are applicable for other burns in similar ecosystems. 84 4. MORTALITY OF OVERSTORY CONIFERS 4.1 INTRODUCTION 4.1.1 FIRE DAMAGE AND TREE MORTALITY To aid in planning prescribed fires or to mark burned trees for salvage logging, it is often desirable to be able to predict which trees are likely to die from fire caused injuries. Mortality prediction can also be used to increase knowledge about the biological functioning and fire sensitivity of different tree species. Tree species differ greatly in the location of the heat sensitive parts crucial for survival and in adaptations to reduce heating. All of the conifers found in the fnterior Douglas-fir zone have their regenerative organs high above the ground surface and lack the capacity to resprout (Kauffman, 1990) and will die if they suffer excessive crown, bole, or root damage (Reinhardt and Ryan, 1988b). Damages to the three main tree organs will be discussed in turn. Crown damage is easiest to measure and is often an important indicator of tree mortality (Harrington, 1993; Peterson and Arbaugh, 1989; Reinhardt and Ryan, 1988b; Ryan and Reinhardt, 1988; Ryan et al, 1988; Wyant et al, 1986). Crown damage depends on the scorch height and the height of the foliage above the ground. Rising hot gases kill needles, and the height of needle scorch is related to frontal fire intensity, wind speed, and the tree species (Saveland, 1982; VanWagner, 1973). Larger trees generally have higher crowns, and are more likely to survive (Reinhardt and Ryan, 1988b). Fisher and Bradley (1987) reviewed information on species specific features of tree architecture that contribute to fire resistance for conifers found in the IDF (Table 4.1). Ponderosa pine, lodgepole pine and western larch all have very to moderately high and open crowns, while crowns of Douglas-fir are lower and more dense. Species with larger buds and twigs are more resistant to fire-caused crown injury (Reinhardt and Ryan, 1988b). The buds of ponderosa pine and western larch are more resistant to fire than their foliage, and western larch can initiate new foliage from surviving buds shortly after burning (Ryan and Reinhardt, 1988). Harrington (1993) observed that some ponderosa pines with 100% needle scorch were still alive ten years after burning, especially if the fires occurred in the fall when the trees were dormant, fn contrast, almost no lodgepole pine or Douglas-fir trees survived when more than 90% of the live crown volume was scorched (Peterson and Arbaugh, 1989). 85 Table 4.1 Relative fire resistance of conifers found in the IDF (from Fisher and Bradley, 1987). Species Thickness Root Resin in old Branch habit Lichen Degree of bark of habit bark growth of fire old trees resistance western larch very thick deep ponderosa very thick deep pine Douglas-fir very thick deep very little abundant moderate high and very open moderately high and open moderately high and dense medium heavy most resistant medium to very light resistant heavy to medium very resistant lodgepole very thin deep pine abundant moderately high and open light medium 86 Resistance to cambium injury theoretically increases with the square of bark thickness, and bark thickness of a given species increases with tree diameter (Ryan et al, 1988). Bark thickness varies greatly among species, with western larch, ponderosa pine and Douglas-fir having very thick bark at maturity, and lodgepole pine having thin bark (Fisher and Bradley, 1987). Thermal conductivity of bark also varies among species, but not nearly to the same extent as bark thickness (Kauffman, 1990). Ponderosa pine develops very fire resistant bark at a younger age than western larch or Douglas-fir, which has thin bark with resin blisters at the sapling stage (Fisher and Bradley, 1987; Kauffman, 1990). The extent of cambium damage can be estimated directly by chemical assessment of extracted tissue samples, or indirectly by examining external charring of the bark. Ryan et al (1988) extracted four tissue samples per tree at a height of 1.4 m, and found the number of samples containing dead cells to be the best indicator of fire-caused mortality in Douglas-fir. Mortality was high even if only two cambium samples were dead; it is possible that the cambium was damaged in a larger fraction of the circumference closer to the ground, where most of the fuel was located. However, Peterson and Arbaugh (1986) measured the percentage of the tree circumference charred at 0.5 m above the ground, depth of bark char, and the ratio of depth of bark char to bark thickness, and found that crown scorch was the most important predictor of mortality two years after a wildfire, and basal scorch was an important additional variable only for lodgepole pine. For Douglas-fir, none of the bark char variables were important. While time consuming and expensive, the extraction of cells seems to possess superior predictive value. Root damage is most difficult to assess due to the tediousness of root system excavation (Wade, 1986). Trabaud (1987) hypothesized that tree roots suffer little damage, since they are protected by a corky outer layer and the soil itself, with the exception of feeder roots near the surface. However, Wade (1986) believed that root damage is more important than commonly thought. Geiszler et al. (1984) reported that the probability of some root kill in lodgepole pine was near 100% when 66% of the basal circumference was charred, and that even if the fire did not reach the trunk, 33% of trees had some killed roots. Ryan et al. (1988) found that 7% of the trees in their study died without detected crown or bole damage; this may be due to background mortality, or perhaps root damage. According to Fisher and Bradley (1987), ponderosa pine, Douglas-fir, lodgepole pine and western larch are all described as deep rooted, which theoretically makes them more resistant to root damage. In old-growth stands, especially ponderosa pine, thick layers of needles and duff can be found at the bases of trees; if 87 consumed by severe fires, enough heat is generated at the root collar to kill trees (Walstad and Seidel, 1990). Tree roots should be affected by fire severity similarly to the understory vegetation discussed in the previous chapter; therefore, root damage should be somewhat predictable from depth of burn or fire severity class. However, trees have spreading and irregular root systems whose exact extent and location is not visible. Therefore, many depth of burn or fire severity class assessments would be necessary to completely cover the area where roots of a given tree could be found. The relative importance of crown, bole, and root damage is often unknown, and it depends on the type of fire as well as on the morphology of the tree. According to Ryan et al. (1988), crown scorch may be the most important factor affecting mortality of Douglas-fir after wildfires, which spread freely, but bole and root damage may be relatively more important in prescribed fires, where weather conditions are selected to control fire intensity. Fires that burn primarily in the litter or duff can cause severe cambial or root damage, especially if they have a long residence time, without causing any crown damage (Peterson and Arbaugh, 1986). This type of mortality is common in stands composed of thinner barked species, such as lodgepole pine, especially where surface fires creep at night when relative humidity is higher (Wright and Bailey, 1982). Trees with low vigour and sparse foliage may die at a lower crown scorch percentage (Ryan and Reinhardt, 1988); this may apply to other damage as well. For surviving trees fire damage can be a pronounced stress; with a substantial portion of the photosynthetic capacity removed carbohydrate production is reduced and resistance to insects and drought is lower (Peterson and Arbaugh, 1986). Increased litterfall one year after has been found both in mixed jack pine/hardwood stands in Minnesota (Grigal and McColl, 1975) and a mixed conifer stand in Montana (Kilgore, 1985). Fire scars can act as infection courts for butt rot fungi, especially if close to the ground (Thies, 1990). 4.1.2 EFFECTS OF FIRE ON SURVIVING TREES Amman and Ryan (1991) found that fire-damaged Douglas-firs were very susceptible to attacks of Douglas-fir beetle (Dendroctonus pseudotsugae), with infestations soon spreading to over half of the healthy, undamaged trees in the stand. While burned lodgepole pines were not found to be especially attractive to mountain pine beetles (Dendroctonus ponderosae), pine engravers (Ips pini) and wood borers (Burpestidae and Cerambycidae) did cause some damage (Amman and Ryan, 1991). Mitchell (1990) noted that ponderosa pine with more than 50% crown scorch invited attacks by engravers, mountain pine beetle, and western pine beetle 88 (Dendroctonus brevicomis). They also found that the beetles quickly increased the population by secreting an aggregation pheromone; this attracted more beetles, which later infested uninjured trees. Sometimes fire-injured lodgepole pines need to be weakened further by root diseases for years before becoming susceptible to attack by bark beetles (Geiszler et al., 1984). The effect of fires on the productivity of surviving trees varies. Miller and Seidel (1990) summarized the results of several studies that examined underburns in ponderosa pine stands. In dense stands, where competition was a factor, the crop trees showed variable responses, ranging from increased to decreased growth, as some likely benefited from the death of other individuals in the stand. In open stands, where the trees probably were not competing, both diameter and height growth were decreased (Miller and Seidel, 1990). fn a Montana Douglas-fir/ western larch stand, Reinhardt and Ryan (1988a) found that surviving western larch trees in a burn grew larger annual radial increments (relative to before the burn) than did trees in adjacent undisturbed areas. This species is a shade intolerant species, and this growth increase was also attributed to the reduction of competition. There was no relationship between increase of radial increment and crown scorch, but the average crown scorch was only 5%. Variable, mostly positive correlations were found between the number of dead cambial samples and relative radial increment, perhaps this was also due to a reduction in competition (Reinhardt and Ryan, 1988a). 4.1.3 MODELLING OF POST-FIRE TREE MORTALITY Models to predict mortality from tree size and indicators of fire injury have been developed for Douglas-fir, ponderosa pine, lodgepole pine, and various other conifer species occurring in the IDF (e.g., Harrington, 1993; Peterson and Arbaugh, 1986; Ryan and Reinhardt, 1988; Ryan et al, 1988; Wyant et al., 1986). These models used discrirninant analysis, logistic regression, or a combination of the two to determine which variables best relate to tree mortality after fires. Bevins (1980) developed a model predicting probability of tree survival for Douglas-fir from scorch height and diameter at breast height outside of the bark (DBH). A later model by Ryan and Reinhardt (1988) predicting tree mortality from DBH and percentage of crown volume scorched, was based on a large number of trees of seven different species in coastal and wet and dry interior forests. It should have wider applicability than models developed using fewer fires. This model did not include ponderosa pine, however, but Savelartd (1982), Harrington (1993) and Regelbrugge and Conard (1993) developed models specifically for this species. 89 If any of the models predictiting post-fire tree mortality developed to date were tested on data from fires other than the ones used to generate the models, this information was not published. Also, all of these studies were performed in slightly different ecosystems than those found in BC, and the applicability of these results in BC is not known. Many tree species vary greatly in growth habit and productivity over their range, depending on climatic and edaphic factors (Krajina, 1969), and it is possible that this pattern extends to fire tolerance as well. 4.2 OBTECTIVES The objectives of the present study were: 1. To develop equations to predict the probability of tree mortality after fire for Douglas-fir, ponderosa pine and lodgepole pine in IDF zone in BC. 2. To assess the applicability of the American tree mortality models to the IDF zone in BC. 4.3 METHODS 4.3.1 STUDY SITES This study was conducted in the southern Rocky Mountain Trench in the Nelson Forest Region on two sites in the Kootenay Dry, Mild IDF Variant (IDFdm2). These sites were burned as part of the Ecosystem Maintenance Burning Evaluation and Research (EMBER) Project conducted jointly by the BC Ministry of Forests and the Canadian Forest Service. Details of site selection and sampling procedures can be found in Braumandl et al (1995). The Picture Valley (PIC) site was located about 8 km south of Fort Steele, and consisted of two different stands, one with a J-shaped diameter distribution curve, and the other with a smaller number of stems < 4 cm in diameter and many standing dead trees. Both stands were dominated by Douglas-fir, but some ponderosa pine was present. The second site (FIN) was located about 17.5 km west of Highway 93/95 in the Findlay Creek valley. The stand contained Douglas-fir, lodgepole pine, ponderosa pine, and western larch (Braumandl et al, 1995). Soils at both sites were Eutric Brunisols situated on glacial till, with Findlay Creek soils belonging to the Elko silt loam series and those at Picture Valley belonging to the Wycliffe silt loam series (Kelley and Sprout, 1956). 4.3.2 PLOT LAYOUT Plots were established prior to burning and sampled (pre and post burn) by a subcontractor of the BC Ministry of Forests. Twenty plots were established in each stand at 90 Picture Valley, and 30 plots were established at Findlay Creek. Plot centres were located on a systematic grid with plot spacing varying from 50 to 100 m depending on site and stand size and orientation, and were marked with electrical conduit or rebar. Plot centres were a minimum of 30 m from any boundary. Trees < 12.5 cm diameter outside of the bark at a height of 1.3 m (DBH) were assessed at all plots, and trees > 12.5 cm DBH were assessed at every second plot. A sweep with a prism of basal area factor 4 m /ha was used to select the trees in each plot. At Findlay Creek, two transects were also established to include more small trees. A total of 180 and 239 trees were tagged at the Picture Valley and Findlay Creek sites, respectively. 4.3.3 DATA COLLECTION For all tagged trees, DBH, total height, and average height to live crown were measured before the burn. The Picture Valley site was burned on May 11,1994, and the Findlay Creek site on April 29,1994. The weather conditions and components of the Canadian Fire Weather Index (FWI) system on the days of the fires are given in Table 4.2. About one year after the burn on each site, the scorch height was measured for each tree. This was defined as the average height where needles were browned. If all needles were consumed or browned by the fire, the scorch height was considered to be equal to the height of the tree. In some cases, live branches were present below the average height to the live crown, and if these were scorched, a scorch height less than the height of live crown resulted. It is possible that some dead trees that were recorded as 100% crown scorched actually had lower crown scorch, and that browned needles, which were due to death of the tree, were interpreted to be crown scorch. The maximum height of charcoal formation on the trunk and the circumference of the tree charred at the base were also assessed. Each tree was recorded as being alive or dead. A tree was considered dead if it had no green foliage. In August 1996,1 assessed any additional tree mortality. 4.3.4 DATA ANALYSIS Trees of species other than Douglas-fir or pine, as well as unburned trees, for which every fire damage variable had a value of zero, were excluded from any statistical procedures. A total of 159 Douglas-firs, 82 lodgepole pines and 102 ponderosa pines were used for analysis. 91 Table 4.2 Weather and FWI System values at noon on the day of the burn. Site and Temp. R H 1 Wind FFMC2 DMC 3 DC 4 BUI5 ISI6 FWI7 date of (OC) (%) speed burning (km/h) Findlay 13.7 Creek (April 29, 1994) 35 • 18 89 31 75 31 8 15 Picture 23.0 Valley (May 11, 1994) • 33 2 93 64 204 72 7 20 1 R H = relative humidity 2 FFMC = Fine Fuel Moisture Code 3 DMC = Duff Moisture Code 4 DC = Drought Code 5 BUI = Build Up Index 6 ISI = Initial Spread Index 7 FWI = Fire Weather Index 92 4.3.4.1 Mortality Prediction Logistic regression models are appropriate for data sets where the outcome is binary (0 or 1). Urdike linear regression models, they are constrained to predict a probability of obtaining one of the outcomes (in this case either tree mortality or survival) falhng between 0 and 1, and are often used to predict tree mortality (Hamilton and Edwards, 1976). The general form of the logistic regression model is: P(m) = 1/ [l+exp( - (b0+bixi+.. .b nXn))] or equivalently, P( m ) = exp (bo+biXi+...bnXn)/[l+exp (b0+blXl+...bnXn)] where P(m) is the probability of mortality, bo is the constant, and bi.. .bn are the regression coefficients. The SYSTAT v. 7.0 Logistic Regression procedure based on likelihood ratios (Norussis., 1997) was used to generate four models for the Picture Valley and Findlay Creek data. One of these was a simple model for all tree species combined containing only the most important tree damage variable (percentage of the live crown length scorched), and the three other models (one for each tree species) contained all independent variables selected as significant by the regression procedure as outlined below. A model for all species combined, which uses dummy variables to represent species, could also be generated, but the three species were considered different enough to warrant separate models. The following variables had been shown to be related to tree mortality after fire (see 4.1.1 FIRE DAMAGE AND TREE MORTALITY) and were entered manually one by one in the following order of importance indicated by preliminary graphical analysis: a) the percentage of the live crown length scorched (PSH); b) diameter at breast height in cm (DBH); c) maximum height of charring on the bark of the bole in m (HBC); d) the percentage of the circumference of the tree base where smoldering was evident (PBS); e) interactions between any two of the above variables; and f) the square of the percentage of the live crown length scorched (PSH2). If a variable was no longer significant (p > 0.05) after the addition of new variables, it was deleted from further analysis. For ponderosa pine, the effect of using the tenth power and fifteenth power of the percentage of the live crown length scorched (PSHiU and PSH ) on the 93 shape and location of the 0.5 probability of mortality curve was also investigated graphically, but these equations were not tested further. Preliminary analysis indicated that mortality varied at the different sites, but since the use of SITE as a predictor eliminates the portability of the model to fires other than the ones used to generate the equation, this variable was excluded from the final list of potential predictors. Chi-square goodness of fit tests and the percentage of trees classified correctly as dead or alive can both be used to assess the fit of a mortality model and each of these methods has its advantages and drawbacks. Correct classification percentages can only be used to compare the relative performance of different models on the same data set, and decisions on whether a model is adequate are arbitrary. Chi-square tests reveal significant differences between the observed and expected numbers of dead trees caused by over- or underpredictions of mortality, and chi-square values can be compared across data sets. However, it cannot be determined whether the actual dead trees are the trees predicted to be dead by the equation, ff dead and surviving trees are misclassified in close to equal numbers, differences between observed and expected numbers of dead and live trees are not significant. In this case, classification percentages have to be examined in addition to the chi-square tests to reveal this type of error. 4.3.4.2 Variable Cutoffs fn this study, both correct classification percentages and Chi-square goodness of fit tests were used to compare the expected number of dead trees calculated using the models to the number of trees that actually died. Since the models predict a probability of mortality between 0 and 1, and not the actual status of the tree, which can only be 0 (alive) or 1 (dead), the predicted and observed values cannot be directly compared. All of the reviewed papers which presented logistic regressions used a probability of mortality of 0.5 as the cutoff value above which a tree was expected to be dead. To explore the effect of altering this value, the correct classification percentages for the three best fit species specific equation were compared using 0.2, 0.3, 0.4, 0.5, 0.6 and 0.7 and 0.8 as cutoff values. 4.3.4.3 Validation Data Sets Validation of a model on a data set independent of the data used to create each model is desirable; but while the validation of models by independent data (trees not used in the generation of the models) from the same fires can be used to investigate the consistency of the relationships between tree mortality and the measured variables, this method cannot be used to 94 assess the portability of the models to different fires. Since the consistency of the relationships between the predictor variables and tree mortality was of some concern, validation data sets were used in this study. The data set for each species (the 159 Douglas-firs, the 82 lodgepole pines, and the 102 ponderosa pines) was systematically split into 3 groups by sorting the trees by site, plot and tag number, assigrting a number from 1 to 3 in that order, and combining the trees with the same number in a group. This ensured representation from all areas of the burn. For the data set of each species, the trees from two groups were combined to create a new equation with all variables selected as significant in the equation from the prior analysis. The third group was used to test these equations. This was repeated three times, each time using a different group as the validation data set. Each time a chi-square test was performed on the observed values from the validation data set and the values predicted from the equation. Also, the percentage of trees in the validation data set that was correctly classified was calculated for each equation. 4.3.4.4 Evaluation of Models from the Literature Six logistic regression models predicting the probability of mortality (P(m)) or survival (P(S)) using fire damage and tree size variables were tested. The probability of survival is 1 -P(m). Graphic representations of these models are shown in Figs. 4.1 -4.2. for Douglas-fir and lodgepole pine: P ( m ) = l/[l+exp( - (1.466-1.91 BT+0.1775 BT2+0.000542 PCS2))] where BT = bark thickness in cm and PCS = percentage of crown volume scorched (Ryan and Reinhardt, 1988) for Douglas-fir: P(S) = l/[l+exp( - (-0.1688+0.1250 DBH-0.30564 SH))] where DBH = diameter at breast height in cm and SH = scorch height in m (Bevins, 1980) for ponderosa pine: P(s) = V[l+exp( - (-2.33+0.37 DBH-0.36 SH))] where DBH = diameter at breast height in cm and SH = scorch height in m (Saveland, 1982) 95 a) Ryan and Reinhardt model for Douglas-fir and lodgepole pine b) Bevins model for Douglas-fir Figure 4.1 Logistic regression models predicting the probability of mortality using various predictors created by a) Ryan and Reinhardt (1988), b) Bevins, (1980), c) Saveland, (1982) and d) Regelbrugge and Conard(1993). 96 c) Saveland model for ponderosa pine a) Harrington dormant season model percent of crown volume scorched b) Harrington growing season model percent of crown volume scorched Figure 4.2 Logistic regression models created by Harrington (1993) predicting the probability of mortality for ponderosa pine burned during the a) dormant season and b) growing season. Discrete classes of crown scorch and diameter were used to create the models. 98 P(m) = l/[l+exp( - (1.0205-0.0999 DBH+0.2858 HBC))] where DBH = diameter at breast height in cm and HBC is the height of bark char on the bole (Regelbrugge and Conard, 1993) P ( m ) = l/[l+exp( - (1.16-1.04 S-1.94 L-0.12 H-0.14 D))] where S = -1 for growing season and S = 1 for dormant season, 20-89% crown volume scorched L = 1 and H = 0,90-99% scorch L = 0 and H = 1 and 100% scorch L = -1 and H = -1, and D = diameter class midpoint (7.5,15.0, 22.5, or 30.0 cm) (Harrington, 1993) The Harrington model for growing season burns (S = -1) was tested separately from the model for dormant season burns (S = 1). The percentage of the live crown volume scorched was used as a predictor variable by Harrington (1993) and Ryan and Reinhardt (1988). This variable was not estimated directly in the present study and had to be calculated. The paraboloid was found to be the most common crown shape for lodgepole pine and Douglas-fir (Peterson, 1985), and was assumed for all trees in this study. The following equation from the FIRESUM model (Keane et al, 1989) was used to calculate percentage of crown volume scorched (PCS): PCS = (CLS (2CL - CLS)/CL2)100 where CLS = crown length scorched = (height of crown scorch - height to live crown) and CL = live crown length = (total height - height to live crown) Bark thickness, which was a variable used in the Ryan and Reinhardt (1988) equations, was calculated for the different species with the following equations: BT (Fd) = 0.065 DBH (Ryan and Reinhardt, 1988) BT( PI) = 0.0688+0.0143 DBH (Ryan and Reinhardt, 1988) BT (Py) - - 0.0954+0.0584 DBH (Faurot, 1977) where BT = bark thickness (cm) and DBH = diameter outside bark at breast height (cm). To ensure comparability of the relative performance of the four models, each model was tested on how well it predicted the mortality/survival two years after fire. The Saveland model was also tested on the mortality one year after fire, since that was the observational time span used to create the model. Bevins (1980) also assessed mortality one year after fire, but based it 99 on the death of three out of four cambium samples instead of on the lack of green foliage as in the other studies. Ryan et al. (1988) found death of cambial tissue to be a good estimation of long term mortality, but many trees with three dead cambium samples were not yet dead the first year after fire. Regelbrugge and Conard (1993) observed mortality during the second post-burn growing season. The mortality data used to generate the Ryan and Reinhardt model ranged from three to eight years post-burn, depending on the fire, and the Harrington model was created from mortality observations ten years after the fires. All 343 ponderosa pine, lodgepole pine and Douglas-fir at Picture Valley and Findlay Creek were used to test the Ryan and Reinhardt (1988) equation. The Regelbrugge and Conard (1993) model was assessed using all 102 ponderosa pines. For the test of the Saveland (1982) and Bevins (1980) equations a measure of scorch height was required, which could not be determined on trees where no foliage was scorched. Therefore, only the 75 ponderosa pines and 135 Douglas-firs showing foliage scorch were used, respectively. Only 36 ponderosa pines were used to test the Harrington equations, because trees with less than 20% crown volume scorched or with a DBH smaller than 3.8 cm or larger than 33.8 cm were not accommodated by the classes used in the model. The probability of mortality was calculated for each tree in the appropriate data set, and if it exceeded 0.5, the tree was considered dead. For models predicting tree survival, trees with a probability of survival of 0.5 or above were considered alive. Chi-square tests were conducted on the expected frequencies of dead and alive trees generated by each of the logistic equations and the observed frequencies. For each of these partial and full data sets, the percentage of trees correctly classified by the equations mentioned above was calculated. 4.3.4.5 Composite Models For both salvage marking and planning applications a decision needs to be made about trees for which the Bevins (1980), Saveland(1982) and Harrington (1993) equations cannot be used, and therefore new composite models with several decision rules incorporating all trees of the species were created. The models generated in this study, as well as other models that included trees throughout the whole range of diameters and fire damages (e.g., Ryan and Reinhardt, 1988), predicted large trees and trees with no or low crown scorch to be alive, and small trees, especially those with crown damage, to be dead. Therefore, the following rules were chosen to constitute the composite equations: 100 a) for the Bevirts and Saveland models: 1) If no live foliage was scorched, count the tree as alive. 2) If live foliage was scorched, and the scorch height could be measured, use the equation with a 0.5 cutoff value. b) for the Harrington dormant season and growing season models: 1) If < 20% of the live crown height was scorched, count the tree as alive. 2) If 20% to 99% of the live crown height was scorched, and a) the DBH of the tree is > 33.8 cm, count the tree as alive. b) the DBH of the tree is between 3.8 cm and 33.8 cm, use the equation. c) the DBH of the tree is < 3.8 cm, count the tree as dead. 3) If 100% of the live crown height was scorched, and a) the DBH of the tree is> 33.8 cm, count the tree as dead. b) the DBH of the tree is between 3.8 cm and 33.8 cm, use the equation. c) the DBH of the tree is < 3.8 cm, count the tree as dead. The performance of these composite models on the two-year tree mortality data was assessed using a chi-square test and classification results. The composite model based on the Bevins (1980) model, the Ryan and Reinhardt (1988) model, and the PSH and species specific models produced in this study were compared in how well they classified the 159 Douglas-firs. The composite models based on the Saveland (1982) and Harrington (1993) models, the models created by Ryan and Reinhardt, and Regelbrugge and Conard (1993), as well as the PSH and species specific models from the present study were compared in how well they classified the 102 ponderosa pines. The Ryan and Reinhardt model was not designed to be used with ponderosa pine, but was applied to this species in the FIRESUM model (Keane et ah, 1989), and, therefore, was also compared to the models created specifically for ponderosa pine. 4.4 RESULTS Of the 343 trees used in the analyses, 158 were dead after one year and 175 had died by the second post-burn growing season. Trees that had died two years after the fire tended to have a larger percentage of the live crown height scorched, higher bark char, higher scorch heights, and, except for lodgepole pine, smaller DBH, tree height and lower percentage of the basal circumference scorched (Table 4.3). However, the values of most of these variables 101 Table 4.3 Means and standard deviations of tree dimensions and fire damage characteristics of . dead and surviving Douglas-fir (Fd), lodgepole pine (PI) and ponderosa pine (Py). Standard deviations are shown in brackets. Species Status N DBH1 HT2 HBC1 PBS4 SH5 PSH6 of tree (cm) (m) (m) (m) (m) Fd dead 78 10.4 8.2 2.9 94 7.8 90.7 (10.1) (4.4) ' (4.0) (20) (4.4) (24.2) alive 81 21.8 14.0 1.9 72 6.0 16.8 (11.6) (5.7) (2.0) (35) (4.9) (22.4) PI dead 76 9.3 7.6 0.9 81 6.7 84.1 (8.1) (5.8) (1.2) (35) (5.51 (32.3) alive 6 10.2 8.2 0.4 92 3.2 22.2 (3.2) (2.9) (0.5) (20) (2.3) (19.6) Py dead 22 11.9 8.6 2.7 98 8.1 82.6 (7.61) (4.23) (2.58) (11) (4.52) (33.19) alive 80 29.3 16.6 1.7 88 5.1 9.9 (14.3) (6.2) (2.1) (25) (4.4) (18.1) 1 DBH = diameter at 1.3 m 2HT = height of tree 3 HBC = maximum height of bark char 4 PBS = percentage of basal circumference smouldered 5 SH = scorch height 6 PSH = percentage of the height of the live crown scorched 102 overlapped for dead and surviving trees. Most Douglas-fir and ponderosa pine in the lower percent scorch height classes survived, while most lodgepole pine died regardless of the level of crown scorch (Fig. 4.3). 4.4.1 MORTALITY PREDICTION The best fit simple equation for all species using only PSH as a predictor was (Fig. 4.4): P(m) = l/[l+exp (- (-2.8198+0.0591 PSH))] where PSH = percentage of live crown height scorchedWhen using a cutoff of 0.5, this equation correctly predicted the post-burn status of 88.6% of the trees two years after the fire. The best fit species specific equations each included a different set of significant predictors. The only variable common to all species was the square of the percentage of the live crown height scorched (Table 4.4). The equations were (Fig. 4.5): a) Douglas-fir P(m) = l/[l+exp( - (-0.4157 + 0.0006 PSH2 - 0.2384 DBH +0.8567 HBC))] where PSH2 = square of the percentage of live crown height scorched, DBH = diameter at breast height (cm), and HBC = maximum height of bark char (m) b) lodgepole pine P ( m ) = l/[l+exp( - (0.5173 + 0.0005 PSH2))] where PSH2 = square of the percentage of live crown height scorched c) ponderosa pine P(m) = l/[i+exp( - (-0.9291 + 0.0006 PSH2 - 0.1116 DBH))] and for higher powers of PSH: P(m) = l/[l+exp( - (0.1047 + 1.0332 X 10"19 PSH 1 0 - 0.1496 DBH))] P(m) = l/[l+exp( - (0.1114 + 1.6517 X 10~29 PSH15 - 0.1489 DBH))] where PSHZ, PSH , and PSH15 = the square, tenth power, and fifteenth power of the percentage of live crown height scorched, respectively, and DBH = diameter at breast height (cm). The P(m) = 0.5 probability of mortality curves for the three ponderosa pine equations are shown in Fig. 4.6. As the power of PSH was increased, at high PSH values the equation became more and more sensitive to PSH relative to DBH. Figs. 4.7 to 4.9 illustrate at which values of PSH, DBH and HBC the incorrect predictions 103 a) Douglas-fir 70 j 60 -in 50 --CD CD i_ 40 -o CD 30 -n E 20 -3 C 10 --0 --• dead • alive 1-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 percent scorch height class 80-90 90-99 b) lodgepole pine 60 50 4-8 40 01 o 30 cu •a | 20 + ' 10 4-• dead • alive 1-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-99 100 percent scorch height class c) ponderosa pine 60 50 4-8 40 at o 30 at Si E 20 10 0 • dead • alive 1-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 percent scorch height class 80-90 90-99 Figure 4.3 Distribution of dead and live a) Douglas-fir, b) lodgepole pine and c) ponderosa pine with respect to percent scorch height class. 104 0 20 40 60 80 100 crown height scorched (%) Figure 4.4 Probability of tree mortality as predicted from the equation for all trees using percentage of the live crown height scorched as the only predictor. 105 Table 4.4 Regression coefficients and diagnostic statistics for each equation predicting probability of mortality for the equation generated using all species and the species specific equations. Variable codes as defined in Table 4.3. Equation N Variable Diagnostics coefficient % Correct1 -2 LL 2 PSH only, all species 343 constant PSH -2.8198* 0.0591** 88.6 193.6 Douglas-fir 159 constant PSH2 DBH HBC -0.4157 0.0006** -0.2384** 0.8567* 93.7 53.6 Lodgepole pine 82 constant PSH2 0.5173 0.0005* 92.7 25.9 Ponderosa pine 102 constant PSH2 DBH -0.9291 0.0006** -0.1116* 92.2 34.6 significant at p =0.05 using the likelihood ratio test significant at p = 0.01 using the likelihood ratio test 1 % correct = the percentage of trees correctly classified 2-2 LL = -2 log likelihood 106 a) Douglas-fir at HBC = 0 m d) ponderosa pine at all HBC b) Douglas-fir at HBC = 4 m e) lodgepole pine at all HBC and DBH c) Douglas-fir at HBC = 16 m Figure 4.5 Probability of mortality as a function of diameter at breast height (DBH) and percentage of the live crown height scorched (PSH) for the species specific equations for Douglas fir at maximum height of bark char (HBC) values of a) 0 m, b) 4 m, and c) 16 m, and for d) ponderosa pine and e) lodgepole pine at all HBC values. 107 Figure 4.6 Probability of mortality of 0.5 for ponderosa pine from equations using the square, tenth power and fifteenth power of the percentage of live crown height scorched (PSH) and diameter at breast height (DBH). Below the line the probability of mortality is > 0.5 and above the line it is < 0.5. 108 Species specific equation PSH only equation 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched 70 60 50 tn 40 i(O OJ 4—' co <D E ro b •S 20 10 0 ~ \ — r " i — i — r o o o 30|- o ° ' cP„ o o °n J L T r J L Status, Prediction ± dead, false • alive, false A dead, correct o alive, correct 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched Status, Prediction - dead, false • alive, false A dead, correct o alive, correct 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched Figure 4.7 Distribution of dead and alive Douglas-fir correctly and incorrectly predicted by the species specific equation and the all species, PSH only equation with respect to the percentage of crown height scorched, diameter at breast height and maximum height of bark char. 109 Species specific equation PSH only equation Jf Status, Prediction ± dead, false • alive, false A dead, correct o alive, correct 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched " i — r ~ i — i — i — r -ii 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched Status, Prediction ± dead, false • alive, false A dead, correct o alive, correct 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched Figure 4.8 Distribution of dead and alive lodgepole pine correctly and incorrectly predicted by the species specific equation and the all species, PSH only equation with respect to the percentage of crown height scorched, diameter at breast height and maximum height of bark char. 110 Species specific equation PSH only equation Status, Prediction - dead, false • alive, false A dead, correct o alive, correct 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched Status, Prediction - dead, false • alive, false ^ dead, correct o alive, correct 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched Figure 4.9. Distribution of dead and alive ponderosa pine correctly and incorrectly predicted by the species specific equation and the all species, PSH only equation with respect to the percentage of crown height scorched, diameter at breast height and maximum height of bark char. Ill by both the PSH only model and the species specific model occurred for each species. Trees with a calculated probability of mortality exceeding 0.5 were counted as dead. For Douglas-fir, both models underpredicted mortality for trees with a DBH of < 20 cm and HBC of < 3 m at scorch heights < 50% of the live crown. Both models also overpredicted mortality for trees with scorch heights from 50 to 90%, but this overprediction of mortality occurred more often for trees of larger DBH and HBC with the PSH model. The species specific lodgepole pine model predicted all trees of this species to be dead (Fig. 4.5e), and overpredicted mortality for trees with < 60% PSH, while the PSH only model underpredicted mortality of trees with the same range of crown scorch. For ponderosa pine, the incorrect predictions were confined to smaller trees with a low height of bark char DBH < 25 cm and HBC < 4 m, and both models performed about equally well. Similarly to Douglas-fir, overpredictions of mortality occurred at PSH < 55% and underpredictions at PSH > 55. 4.4.2 VARIABLE CUTOFFS When the probability of mortality cutoff value used to classify a tree as dead was at or near 0.5, the Douglas-fir equation classified dead trees about equally well as live trees (Fig. 4.10), mirroring a nearly equal distribution of dead and live trees in the sample. At this value, the total number of correct classifications was highest. As the cutoff value was raised, more of the live trees were classified correctly as such, but more dead trees were predicted to be alive, resulting in an underprediction of mortality. The reverse was true when the cutoff was lowered, and the number of dead trees was overpredicted. However, the overall percentage of trees classified correctly varied only slightly over the range of cutoff values tested. For lodgepole pine and ponderosa pine, where the sample contained predominantly dead or live trees, the best overall balance of classification of dead and live trees was not achieved when the cutoff was near 0.5. Only when the 0.8 cutoff was used with the lodgepole pine equation was the number of predicted dead and alive trees significantly different from the observed values as determined with a chi-square goodness of fit test for a = 0.05 (Appendix 4.1). 4.4.3 VALIDATION DATA SETS The coefficients of the equations can be found in Table 4.5. Correct classifications by the species-specific equations ranged from 90.6 to 94.3% for Douglas-fir, from 85.7 to 96.3% for lodgepole pine, and from 88.2 to 94.1% for ponderosa pine (Table 4.6). The observed numbers 112 a) Douglas-fir n E II M i l l 0.2 0.3 0.4 0.5 0.6 0.7 0.8 b) lodgepole pine P robab i l i t y of mo r t a l i t y cu to f f va lue 80 60 -• " SO 20 10 o J — M M — i — M M — i — • & — , — B H — i — B S ® — , — ! I I 0.2 0.3 0.4 0.5 0.6 0.7 0.8 P robab i l i t y of m or ta l i ty cu to f f c) ponderosa pine 100 80 I 60 o E 40 20 1 0.2 0.3 0.4 0.5 0.6 P robab i l i t y of m or ta l i ty cu to f f • alive, correct • a l i v e , predicted dead Hdead, predicted alive 0.7 0.8 I dead, correct Figure 4.10 Number of trees correctly and incorrectly predicted dead or alive when different probability of mortality cutoff values (above which a tree is considered dead) are used. Values are from the species specific equations for a) Douglas-fir, b) lodgepole pine, c) ponderosa pine. 113 Table 4.5 Regression coefficients and diagnostic statistics for each of three species-specific equations predicting probability of mortality. For each species, the data were split into three equal groups (A, B, and C) and two of these groups were used to fit the model the process was repeated three times. Variable codes as in Table 4.3. Coefficient Equation Diagnostic/Variable Groups B+C Groups A+C Groups A+B Douglas-fir -2 LL 1 38.6 35.0 30.5 N 106 106 106 constant -0.8386 -0.7789 113685 PSH2 0.0006 0.0006 0.0006 DBH -0.2121 -0.2184 -0.3681 HBC 0.8543 0.7686 0.9563 Lodgepole pine -2LL 10.0 19.9 20.6 N 54 ' 5 5 55 constant 1.1311 0.2988 0.3288 PSH2 0.0007 0.0005 0.0005 Ponderosa pine -2 LL 25.8 15.2 25.3 N 68 68 68 constant -0.4454 -2.1604 -0.3150 PSH2 0-0006 0.0007 0.0005 DBH -0.1130 -0.1139 -0.1339 x-2 LL = -2 log likelihood 114 Table 4.6 Classification of trees and Chi-square statistic for each of the three test groups (A, B, and C) using the species specific equations for a) Douglas-fir, b) lodgepole pine and c) ponderosa pine generated from 67% of the data. Percentage of trees predicted correctly test group NT surviving dead total ^2 a) Douglas-fir, species specific equation A 53 96.3 92.3 94.3 0.1 B 53 96.2 85.2 90.6 0.7 C 53 92.9 88.0 90.6 0.1 b) Lodgepole pine, species specific equation A 28 0 100 85.6 B 27 0 100 96.3 C 27 . 0 100 96.3 c) Ponderosa pine, species specific equation A 34 92.9 100 94.1 0.65 B 34 100 66.7 91.2 1.82 C 34 92.6 71.4 88.2 0 The observed frequencies of dead and surviving trees were not significantly different from the values predicted by any of the equations for a = 0.05 using a cutoff value of 0.5. 115 of dead trees did not differ significantly from the numbers predicted by the equations for any of the test data sets as determined with a chi-square goodness of fit test for a = 0.05 (Table 4.6). 4.4.4 EVALUATION OF MODELS FROM THE LITERATURE The distribution of correctly and incorrectly predicted Douglas-fir from the Bevins (1980) equation and Douglas-fir and lodgepole pine from the Ryan and Reinhardt (1988) equation are shown in Fig. 4.11. Overall, the Bevins equation overpredicted tree mortality in the reduced data set excluding trees with an undefined scorch height, especially for trees with DBH ranging from 10 cm to 30 cm and scorch heights (SH) from 5 m to 15 m. Some large trees with a scorch height above 20 m were incorrectly predicted to be alive. The expected frequency of dead and surviving Douglas-fir calculated with the Bevins equation differed significantly from the observed frequency at a = 0.05 The Ryan and Reinhardt (1988) model also overpredicted tree mortality for both Douglas-fir and lodgepole pine with < 100% of the crown volume scorched. A chi-square test showed the observed frequencies of dead and live Douglas-fir to be significantly different from the expected at p < 0.05, which means the model did not predict the overall number of dead trees well. The Ryan and Reinhardt model correctly classified only 73.0% of the 159 Douglas-fir but 92.7% of the 82 lodgepole pine (Table 4.7). All lodgepole pines in the present study were predicted to be dead by this model. Similar to the Bevins (1980) model, the Saveland (1982) model estimated the probability of survival using scorch height and DBH as major predictors, and was tested on reduced data sets limited to trees with a definite scorch height. The distributions of the correctly and incorrectly predicted ponderosa pine from this equation are shown in Figure 4.12. The expected numbers of dead and live trees differed significantly from the observed frequency at a = 0.05. The model overestimated mortality of trees with a DBH < 10 cm and SH < 6 m, and underpredicted mortality of trees with a SH > 13 m. The chi-square test did not detect a significant difference between the expected frequency of surviving ponderosa pine calculated by the Saveland model and the observed frequency. The correctly and incorrectly predicted trees from the Harrington (1993) dormant season and growing season models are shown in Figure 4.12. The observed data fit the growing season model better than the dormant season model, but both Harrington models underpredicted mortality for trees with < 90% crown volume scorched. The dormant season model also underpredicted mortality for trees with crown scorch from 90% to 100%. A chi-116 a) Douglas-fir, Bevins (1980) model 70 r £ 60 o Status, Prediction - dead, false • alive, false A dead, correct o alive, correct 5 10 15 20 Scorch height (m) b) Douglas-fir, Ryan and Reinhardt (1988) model 25 Status, Prediction • alive, false A dead, correct f o alive, correct 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown volume scorched c) Lodgepole pine, Ryan and Reinhardt (1988) model Status, Prediction • alive, false A dead, correct 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown volume scorched Figure 4.11 Distribution of dead and alive a) and b) Douglas-fir and c) lodgepole pine correctly and incorrectly predicted by the a) Bevins (1980) model, b) and c) Ryan and Reinhardt (1988) model with respect to the independent variables used to create each model. 117 Table 4.7 Classification of trees using a 0.5 cutoff value and chi-square statistic for each of the models tested with the reduced and full data set of each of the three tree species. Classification percentages should only be compared within numbered categories. The 2 n d year post-burn mortality was used unless otherwise noted. Models in bold were developed in the present study. Percentage of trees predicted correctly Model N Surviving Dead Total y} 1) Douglas-fir, 2) 3) a) Bevins 135 59.7 96.2 80.7 14.9* b) Fd-species specific 159 95.1 92.3 93.7 0.1 c) PSH-all species 159 86.4 92.3 89.3 0.6 d) Bevins composite 159 71.6 96.2 83.7 10.6* e) Ryan and Reinhardt 159 46.9 100.0 73.0 63.9* odgepole pine, a) Pl-species specific 82 0 100 92.7 N/A b) Ryan and Reinhardt 82 0 100 92.7 N/A c) PSH-all species 82 83.3 85.5 85.4 7.8* jonderosa pine a) Harrington growing season 36 100 80.0 88.9 1.8 b) Harrington dormant season 36 100 63.2 80.5 6.1* c) Saveland 2 n d yr. mort. 75 88.9 71.4 84.0 0.0 d) Saveland 1st yr. mort. 75 86.0 72.2 82.7 0.6 e) Harrington growing season 102 100.0 77.3 95.1 1.8 composite f) Py-species specific 102 92.2 93.8 93.0 0.1 g) Harrington dormant season 102 100.0 63.6 92.2 5.3* composite h) PSH-all species 102 92.5 81.8 90.2 1.1 i) Saveland composite 102 92.5 68.2 87.3 0.1 j) Regelbrugge and Conard 102 77.5 81.8 78.4 8.4* k) Ryan and Reinhardt 102 73.8 81.8 75.5 12.0* * the observed frequencies of dead and surviving trees differ significantly from the expected values for a = 0.05. 118 a) Ponderosa pine, Saveland (1982) model b) Ponderosa pine, Regelbrugge and Conard (1993) model Status, Prediction * dead, false • alive, false A dead, correct o alive, correct 5 10 15 Scorch height (m c) Ponderosa pine, Harrington (1993) growing season model 40 2 4 6 8 10 12 Maximum height of bark char (m) d) Ponderosa pine, Harrington (1993) dormant season model Status, Prediction A dead, false A dead, correct o alive, correct 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown height scorched e) Ponderosa pine, Ryan and Reinhardt (1988) model 3.5 r Status, Prediction * dead, false • alive, false A dead, correct o alive, correct 0 10 20 30 40 50 60 70 80 90 100 Percentage of live crown volume scorched Figure 4.12 Distribution of dead and alive ponderosa pine correctly and incorrectly predicted by the a) Saveland (1982), b) Regelbrugge and Conard (1993), c) Harrington (1993) growing season, d) Harrington dormant season and e) Ryan and Reinhardt (1988) models with respect to the independent variables used to create each model. 119 square test revealed that the observed numbers of dead trees were not significantly different from the numbers predicted by Harrington's growing season model (p > 0.05), but they differed significantly from the predictions by Harrington's dormant season model. The growing season and dormant season models correctly classified 89.2% and 81.1% of the trees in the reduced ponderosa pine data set (N=36), respectively. No trees were incorrectly predicted alive. The Regelbrugge and Conard (1993) model overpredicted mortality for small diameter trees (< 12 cm) with low bark char (HBC), and also for trees of 15-25 cm DBH with a bark char height > 2 m (Figure 4.11). At DBH from 20-25 cm and bark char heights of 1-3 m mortality was underpredicted. The observed number of dead and surviving trees was significantly different from the expected value calculated by the equation. Overall, 78.4% of trees were classified correctly. The Ryan and Reinhardt (1988) model performed quite poorly for ponderosa pine. A few trees without crown scorch were incorrectly predicted to be alive. The model predicted that trees with high crown scorch would die, even larger trees with higher bark thickness. However, of the trees with > 50% of the crown volume scorched (PCS), many remained alive, resulting in an overprediction of mortality. The chi-square test indicated that the observed number of dead and surviving trees was significantly different from the expected value calculated by the equation. 4.4.5 COMPOSITE MODELS The results of using the models, including composite models, on the full Douglas-fir, lodgepole pine and ponderosa pine data sets can be found in Table 4.7. For Douglas-fir, the species specific equation performed best, correctly classifying 93.7% of the trees, followed by the PSH and the Bevins (1980) equation. The Ryan and Reinhardt (1988) equation showed the poorest performance, classifying only 73.0% of the trees correctly. The species specific equation and the Ryan and Reinhardt model each classified 92.7% lodgepole pine correctly, but a chi-square test could not be performed, since the expected number of live trees was zero. The PSH equation performed more poorly, classifying only 85.4% of the trees correctly. For ponderosa pine, the best model was the composite model involving the Harrington (1993) growing season equation, which classified 95.1% of the trees correctly, followed by the species specific equation, the Harrington dormant season model, the PSH model, the Saveland (1982) model, and the Regelbrugge and Conard (1993) model. The Ryan and Reinhardt model again classified the 120 trees most poorly, with 75.5% correct. The Harrington growing season composite model was the only model from the literature that classified the trees of the full ponderosa pine data set better than any of the models generated in the present study. 4.5 DISCUSSION 4.5.1 MORTALITY PREDICTION Percentage of the live crown height scorched, or its square, was the one variable that was significant in all equations. When used as the only predictor variable, the equation correctly classified a large percentage of the trees (Table 4.7). This agreed with the results of studies by Peterson and Arbaugh (1986) and Wyant et al. (1986). Crown damage can be quantified in several ways. The percentage of the crown volume scorched is most directly related to the amount of photosynthetic capacity removed (Peterson and Arbaugh, 1986), but is difficult to accurately estimate or calculate. Any equations used to calculate the percentage of the crown volume scorched rely on subjective assessment of crown shape, and make assumptions about the basal area of the crown if this quantity is not measured. Peterson (1985) obtained consistent differences between visually estimated and calculated percentages of crown volume scorched, with calculated volumes being larger than visually estimated ones, especially for trees with lower crown scorch. Therefore, it was thought preferable to use the percentage of the live crown height scorched, which was also used by Wyant et al. (1986) and Harrington (1993), instead of the percentage of crown volume scorched. Scorch height combined with DBH only indirectly measure crown damage, and the relationship between the two and tree mortality would be expected to differ for trees with different crown architecture. The high percentage of trees correctly classified by all equations by the present study was to a large extent due to the high proportion of trees with 100% crown scorch, which were invariably dead, or with no crown scorch, most of which survived. The relationship between crown scorch and probability of tree mortality was poorest for lodegepole pine; this was likely due to the low number of lodgepole pines that survived the fires, and the occurrence of these survivors over a range of crown scorch percentages. The Douglas-fir and ponderosa pine models also predicted a higher probability of mortality for trees with smaller diameter at breast height, and in the Douglas-fir model the maximum height of bark char was also important (Fig. 4.5). Bark, thickness, which is an indicator of resistance to cambium mortality, varies with the species and DBH of the tree (Ryan 121 and Reinhardt, 1988). Lodgepole pine is only moderately fire resistant due to its thin bark, while Douglas-fir and ponderosa pine are both thick barked and very fire resistant (Fisher and Bradley, 1987). For trees with low crown scorch and without bark char (HBC = 0 m), the models predict that lodgepole pine of all diameters will die, but that larger diameter ponderosa pine or Douglas-fir will survive (Fig. 4.4). This suggests the cambium of thin barked trees could be damaged without visible signs of charring. The ponderosa pine model did not include bark char as a predictor, but the HBC term was significant for Douglas-fir, which suggests ponderosa pine could likely be charred without sustaining cambium damage, while in Douglas-fir the maximum height of charring would be an indicator of damage. The percentage of the basal circumference showing signs of smouldering was also thought to be an indicator of cambium damage near the root collar. Peterson and Arbaugh (1986) found this to be a significant predictor of the probability of mortality for lodgepole pine, but in the present study, this variable had no significant predictive value and was therefore excluded from all of the equations. The Douglas-fir model, and to a lesser extent the ponderosa pine model, predicted probabilities of mortality near zero for large diameter trees with low bark char, even if the percentage of the crown scorched was near or at 100%. This is an unrealistic artifact of the low number of trees in this class. When a higher power of the percentage of the live crown height scorched (PSH) was used, this problem was corrected to some extent by decreasing the relative influence of the other variables on the predicted probability of mortality for trees with a high PSH. However, using an equation in which PSH was raised to a lower power would not present a practical problem because large diameter trees with low bark char and high crown scorch percentages are rare. Only three Douglas-fir trees with a DBH larger than 40 cm in the present data set had a crown scorch percentage exceeding 70%, and for all of these trees the bark char heights were above 12 m. 4.5.2 VARIABLE CUTOFFS If either over- or underprediction of mortality needs to be avoided, the cutoff value could be shifted from 0.5 without drastically reducing the total number of trees classified correctly. Shifting of cutoff values could also be used to adjust a tree mortality model to a stand burned under different conditions. For example, if one was considering using the Harrington (1993) model, but bole heating was believed to be more severe than in Harrington's study, the 122 cutoff point could be lowered. This manipulation would have the greatest impact if many trees with probabilities of mortality near 0.5 were encountered. 4.5.3 VALIDATION DATA SETS Where the models were generated using 67% of the complete data set, the values of the coefficients varied slightly from equation to equation, but all equations classified the trees of the validation data sets quite well. This shows that, under the given fire conditions, the relationships between tree mortality and the measured variables were quite consistent for this study, especially for Douglas-fir. The lower variability of the coefficients among the Douglas-fir models as compared to the lodgepole pine and ponderosa pine models points to the importance of the critical region in variable space, where both dead and surviving trees of similar size with similar damages coexisted in near equal proportions, and, therefore, the probability of mortality was near 0.5. If this region contains a low number of trees, the removal of 33% of the trees as validation data is more likely to result in an unequal distribution of dead and surviving trees from this region in that portion of the data remaining to generate the equation. This causes a greater shift in the location of the region, and therefore also in the values of the 0.5 probability of mortality decision cutoff, and in the coefficients of the equation. For each species, two of the options are a) to use a model generated from a smaller data set and which has been validated, and b) to use a model that was generated from all trees of the species, which could only be validated on trees used in the generation of this model (such as those in Table 4.4). Since the validated models were reasonably similar to each other, and to the model using all trees of a given species, it is possible to use the model generated from all trees, especially the one for Douglas-fir. Most likely both types of models would perform more poorly on data from different fires. 4.5.4 EVALUATION OF MODELS FROM THE LITERATURE It was not possible to compare all equations found in the literature, because often different indicators of damage or fire resistance were measured and the observations were often made over different time intervals. The tree mortality models created by Peterson and Arbaugh (1986,1989), Wyant et al. (1986), or Ryan etal. (1988) could not be tested with, the present data set, because some damage variables used in their equations were not measured in the present study. Peterson and Arbaugh measured the upslope and downslope depth of bark char 50 cm above the ground (1986,1989), and counted the number of bark beetle entry holes (1986). Wyant et al. (1986) measured the height of bark char in four quadrants for each tree. 123 Ryan et al. (1988) and Peterson and Arbaugh (1989) extracted cambium samples and chemically determined whether they were dead or alive. While these variables might give more accurate indications of fire damage, they would be too time consuming and expensive to measure operationally. This would limit the applicability of the model, and therefore these variables were not included in the present study. None of the models reviewed were formulated to be directly comparable to others found in the literature except that of Saveland (1982), who compared the results of his study to Bevins' (1980) model; Generally, for a Douglas-fir of a given diameter from the Bevins (1980) data set a lower scorch height would be sufficient to kill it than for a ponderosa pine of the same diameter from the Saveland (1982) data set. The ponderosa pine model created in this study only predicted this relationship for trees with high bark char. However, the fact that HBC was not a predictor of mortality for ponderosa pine, and the coefficients of the variables common to both equations also differed, suggested that some differences existed between Douglas-fir and ponderosa pine with respect to fire susceptibility. Most likely the Saveland model for ponderosa pine and the Bevins model for Douglas-fir are different due to a combination of several reasons; they were constructed using trees of different species in different ecosystems, which were burned by different fires. The Saveland (1982) model underpredicted mortality for ponderosa pine with DBH > 10 cm and scorch heights > 6 m, and overpredicted mortality for trees < 10 cm DBH and with scorch heights < 6 m, but these errors were moderate and are balanced enough that the chi-square was not significant. The Bevins (1980) model generally overpredicted mortality for Douglas-fir < 45 cm DBH with scorch heights < 15 m, while underprediction of mortality was rarer and evident mostly for large diameter trees at high scorch heights. The overpredictions and underpredictions were not balanced, resulting in a significant chi-square. Possibly the trees used to generate the Bevins equation received more bole heating than those in the present study, or that the mortality criterion used (three dead cambium samples out of four) better approximated the long-term mortality than tree mortality two years after fire. Both of these models include scorch height as a predictor, but, similar to the height of bark char, scorch height by itself is not a direct estimate of fire damage to tree tissue. Since tree height and the height to live crown are correlated with DBH (Braumandl et ah, 1995), crown damage can be indirectly estimated by using scorch height and DBH together in an equation predicting tree mortality. However, this assumes that the DBH/crown parameter relationships are relatively constant, and predictions generated with a model using scorch height and DBH 124 could be inaccurate for stands where the relationship between DBH and crown parameters differs from that in the stand(s) used to generate the model. The Ryan and Reinhardt (1988) model also overpredicted mortality for trees with < 100% of crown volume scorched. This model was reformulated as a series of nomograms that practicing foresters could use to predict expected tree mortality after fires of various intensities (Reinhardt and Ryan, 1988b). The model represents an average of data from 43 fires, and actual tree mortality after several of the fires used to create the model was substantially different from the values predicted by the model (Ryan and Reinhardt, 1988). The only two fires in this data set where mortality was overpredicted nearly as much as for the trees from the present study did not cover the treatment area unif ormly, and many of the trees received no bole scorch. However, in the present study, tree mortality was assessed after only two years, as opposed to eight years in the Ryan and Reinhardt (1988) study. In studies examining the trend of mortality over time, most mortality was observed in the first two years after fire, but at least some additional mortality occurred every year at least until ten years after fire (Harrington, 1993; Ryan et al., 1988). Also, the percentage of the crown volume scorched in the present study was calculated, and not visually estimated as in the Ryan and Reinhardt study, and according to Peterson (1985) calculated, scorch percentages are larger than estimated ones, which would lead to excessively high predictions of mortality. Tree mortality in the Harrington (1993) study was observed ten years after the fire, but both the growing season and dormant season model, as well as the composite models based on them, underpredicted mortality for the trees in the present data set. None of the fires in Harrington's study were performed in the spring before budbreak similar to those in the present study, therefore none of the Harrington models is strictly applicable. However, the growing season model seems more appropriate, because likely dormancy was already broken. Harrington (1993) reported flame lengths averaging 1 m, and thought that cambium injury was not important in his study, but this may not be true for the present study, where many trees showed basal smouldering and/ or bole charring. The good performance of the Harrington composite models is quite surprising, because Harrington grouped the predictors of crown volume scorched and DBH into classes, which resulted in some loss of information. However, if any additional mortality occurs in the future, the percentage of trees correctly classified by both of the Harrington composite models will decline, since the number of trees predicted dead by these models had already died after two years. 125 Drought may also affect tree mortality after fires. The Regelbrugge and Conard (1993) model overpredicted overall mortality of ponderosa pine for the trees in the present data set, but they mentioned that a period of drought occurred during the course of their study. However, they did not specify the duration of the drought, or which proportion of normal precipitation fell during this time. It is possible that the overprediction of mortality for smaller diameter trees, and for larger trees with higher bark char, occurred due to a more severe effect of the drought on these trees. Because crown scorch is difficult to assess two years after a fire, Regelbrugge and Conard (1993) did not include this variable in their model. They hypothesized that the percentage of the live crown length scorched was correlated with the height of stem bark char, but the poor performance of this model for the present data set suggests that this relationship may not exist in all cases. It cannot be ruled out that ecotypic and site related differences could influence levels of tree mortality, because most of the studies either included coastal and interior wetbelt sites (Ryan and Reinhardt, 1988), or were located far from British Columbia. The Harrington (1993) study was performed in southwestern Colorado at much higher elevations than the present study, and the Regelbrugge and Conard (1993) study was conducted in central California. It seems advisable to use a model generated in an ecosystem as similar as possible to that of the site under consideration. Models predicting probability of fire-caused tree mortality do not directly take into account background mortality caused by factors other than fire damage, which could be expected to vary with tree density and growing conditions. Trees of very small and very large size and low vigour due to competition or other causes, are most likely to die (Buchman, 1985). In a fire situation, such trees would be observed to die even with minimal or no injuries. The rates of non-catastrophic mortality were found to be < 1 % annually for larger diameter ponderosa pine and lodgepole pine in northern Idaho (Hamilton and Edwards, 1976), but would probably be higher for small, suppressed trees found in thickets. It should be noted that the mortality observed in the present study probably was not entirely due to fire damage, although the numbers of trees affected would likely be too small to make a practical difference. Mortality resulting from insect attacks on fire weakened trees has been observed for Douglas-fir in some areas (Amman and Ryan, 1991; Peterson and Arbaugh, 1986), but not in others (Peterson and Arbaugh, 1989), depending on insect population levels, genetic subtypes of the tree species, and post-fire moisture stress. Mortality caused by insects is often confounded with mortality from fire injury, and is very difficult to predict. If insect damage is 126 correlated with fire damage, and incidence of insect infestation is highest several years after the fire (as observed by Amman and Ryan (1991) for both Douglas-fir and lodgepole after the Yellowstone fires), longer term mortality models implicitly include mortality caused by insects. 4.5.5 APPLICATION OF MORTALITY MODELS To decide which of the mortality models tested would be most useful, one has to determine what application the model would be used for, if the conditions of the fire and the stand characteristics resemble any of those from which the models were created, and what magnitude and direction of error is acceptable. Assuming that trees with equal amounts of damage are equally likely to die, models in which all variables are measurements of actual damage, such as percentage of crown volume scorched, beetle damage or cambium death, would be more applicable under fire conditions differing from those used in model creation. However, with the exception of crown scorch, such variables are time consuming and expensive to measure. Models using variables reflecting fire resistance and/or scorch height would be cheaper to apply, but could produce larger misclassification rates if applied when the intensity and severity of the fire are different from those of the fire(s) used to create the models. In most studies, fire characteristics were qualitatively described, but not quantified in a way that permitted comparison with other studies. Since bark thickness and crown form, which are related to fire resistance, depend on tree species, mortality models should only be applied to the species used to create the models unless the equation includes direct estimates of both bark thickness and crown scorch. Mortality prediction for planning and modeling applications will have quite a coarse resolution. Detailed stand characteristics and the nature of the fire need to be estimated, and a estimates will also have to be made about variables that will not be uniform throughout the stand, such as scorch height or tree crown characteristics. Generally, for these types of variables a single value is estimated for the whole stand (e.g. the FfRESUM model by Keane et al. (1989)). The more variables are estimated rather than measured, the greater will be the potential for errors. Equations predicting scorch height from fire intensity have been developed for different wind speeds, temperatures and species (VanWagner, 1973; Saveland, 1982), and if models using scorch height and DBH as variables are used, the number of estimations is kept at a minimum. The Ryan and Reinhardt (1988) model can be applied using the nomograms, but estimates of tree height and the ratio of crown length to tree height are required (Reinhardt and 127 Ryan, 1988b). The nomograms can also be used to predict percentage of crown volume scorched, which is required by the Harrington (1993) model. No equations exist to link fire intensity or fire severity to other fire damage variables, such as bark char, and to include a component relating to cambium damage in a mortality model one has to rely on variables related to fire resistance, such as DBH. To use one of the models developed in the present study, one would need to choose one that does not include the height of bark char as a variable, and a new nomogram predicting the percentage of live crown length scorched from crown ratio and percentage of the tree length scorched would need to be developed. For salvage marking, the fire has already occurred, resulting in fewer unknown quantities. Fire damage variables, such as scorch height and height of bark char, can be measured or visually estimated, rather than being calculated using estimates. Models developed from fires that were clearly different than the burn under consideration can be avoided, reducing the potential for errors. For example, if heavy charring of most tree boles is encountered in a ponderosa pine stand, the Harrington (1993) models, which were developed from fires where bole heating was limited, would not be appropriate unless the cutoff value was modified. Models containing only two predictor variables can be represented as simple two-dimensional graphs (see Bevins, 1980) that can be used in the field to mark a tree as dead or alive. The Douglas-fir model generated in the present study will require a three-dimensional graph, or many two-dimensional graphs, to represent all possible combinations of the variables. For all applications of mortality models, one has to decide if over- or underprediction of mortality is associated with greater costs. For example, if a salvage operator is interested in being able to cut more volume, and it is acceptable to cut some trees that would have survived, a model likely to overpredict mortality under the present conditions could be chosen, or the cutoff value could be lowered from 0.5. If it was necessary to minimize cutting of trees that would survive, and leaving some snags in the stand was desirable, a model that underpredicts mortality would miriimize risk. Both the Bevins (1980) composite model and the Ryan and Reinhardt (1988) model overpredicted mortality for Douglas-fir to a large extent under the conditions found in. the present study, with error rates that were approximately 10% and 20% higher than for the Douglas-fir equation generated in this study. Of these two, the Bevins composite model would seem preferable unless bole charring was much more severe than that observed during the present study. For lodgepole pine, neither the Ryan and Reinhardt and the lodgepole pine 128 model from this study predict surviving trees very well, but, in the absence of better models, both would be appropriate. For ponderosa pine all models, except those created by Ryan and Reinhardt and Regelbrugge and Conard (1993), performed quite well. If the use of one of the Harrington models were considered, the growing season model would be preferable for spring burns. Of all the models reviewed, the Ryan and Reinhardt (1988) model was generated from data sets including the greatest number and variety of fires and species, and therefore it might be the most useful one if a model created for the specific ecosystem, season, time after fire, and fire conditions in question is not available. However, if this model were chosen, estimating the crown scorch percentage visually, rather than calculating it, might improve the performance of this model. If the Bevins (1980), Saveland (1982) or Harrington (1993) models were chosen, one could either use the composite models proposed here, or different composite models containing a mechanism to classify the trees with indeterminate scorch heights, or low crown scorch, that would most likely be encountered. 4.6 CONCLUSIONS AND RECOMMENDATIONS Several logistic regression models were developed to predict tree mortality two years after a prescribed fire from tree size variables and fire damage. A model for all tree species and the model for lodgepole pine used only the percentage of the crown length scorched as an independent variable, but the others included this variable together with DBH, and, in the case of Douglas-fir, the maximum height of char on the bark. These models performed generally well on the data set from which they were created, with the species specific models being superior to the PSH only model for all species. With the exception of ponderosa pine, the models created by the present data set performed as well, or better than, any equation from the literature. When using a model to make predictions about tree mortality after a different fire, one has to be prepared for more misclassification error, which increases as the stand and fire conditions deviate more from those used to generate the equations. Except for lodgepole pine, the Ryan and Reinhardt (1988) model generally did not classify the trees very well, but its performance may be improved if crown scorch was visually estimated. When choosing a model, it is generally wise to use one that was developed under stand conditions that closely approximate the stand in question, and to measure variables in a similar way. 129 The following recommendations are made for additional studies that would increase understanding of the relationships between fire damage and tree mortality: • repeat mortality assessment for the trees in the present data at least five years after the burn to obtain equations predicting probability of long-term mortality. • measure depth of bark left after burrung in the most deeply charred area to better quantify stem damage. • examine the relationship between bark char and cambium death for trees of different species and bark thickness. • test whether fire intensity is correlated with either height of bark char or the number of dead cambium samples. • compare the performances of the Ryan and Reinhardt (1988) model using both visually estimated and calculated crown scorch percentages from the same trees. • isolate the effects of crown, bole, and root damage on mortality by experimentally applying heat to a) only the crown, b) only the bole, c) only the roots, d) both the crown and bole, e) both the crown and the roots, f) both the bole and the roots, and g) the crown, the bole and the roots combined. • determine whether fire severity as assessed with burn pins systematically spaced around the tree base can be used to predict the probability of tree mortality • test the models developed in the present study on similar prescribed burns in the IDF, perhaps incorporating the new information into better prediction models. A larger proportion of large trees should be tagged, since larger trees, especially those with a high crown scorch percentage, were underrepresented in the present data set. • repeat this study in a pure lodgepole pine forest, or tag a greater number of lodgepole pines, because surviving trees of this species were underrepresented in the present sample. • Develop models from a greater number of fires. Then stand level variables, such as elevation or stand density, could be added to improve the models. 130 5. GENERAL CONCLUSIONS ABOUT PRESCRIBED FIRE IN THE INTERIOR DOUGLAS-FIR ZONE The understanding of fire effects on tree mortality and on understory vegetation in the IDF in British Columbia has been increased by this study. In Chapter 2, the dynamics of undisturbed Douglas-fir ecosystems were described, and year-to-year variations were found in one or more of the stands for all plant life forms. These changes did not show a simple relationship with any of the weather variables that could be easily summarized from fire weather data. Burning in standing forest resulted in tree mortality, which was higher for trees of all species when a greater fraction of the crown length was scorched. Ponderosa pine and Douglas-fir with a lower DBH were also more likely to die than those with a greater DBH, and a greater maximum height of bark char was also related to greater probability of mortality for Douglas-fir. All lodgepole pine in this study were predicted to die. Several other tree mortality models were also appropriate to predict mortality for Douglas-fir, lodgepole pine and ponderosa pine, but the Ryan and Reinhardt model, which was developed for seven species using data from many fires, was not adequate for Douglas-fir or ponderosa pine. By July of the first post-burn growing season fire severity, as expressed by the percentage of the forest floor depth burned, was a good predictor of the cover and species richness of both residual and invading plants. Higher fire severity resulted in lower cover and number of persisting species, but higher cover and number of invading species. When fire severity was low, the cover and number of species of persisters was strongly dependent on the cover and number of vascular plant species before the burn, but when fire severity was high the pre-existing vegetation was less important as a predictor. The pre-treatment cover of mosses was an important predictor of the cover and number of species of invaders in unburned plots, but was less important in burned plots, suggesting that charred forest floor and moss are both unsuitable seedbeds. Several other variables related to the overstory canopy, site, or reduction of overstory competition were also related to vegetatal cover, but in most cases their effect was small relative to that of fire severity. Pre-existing bryophytes and lichens were reduced to a greater extent by fire and recovered more slowly than vascular plants, but a few pioneer mosses increased greatly on moderate burn plots. Within each growth form group some species benefited from fire, while others were harmed or eliminated. 131 Fire managers deciding whether a prescribed burn is appropriate to achieve vegetation management goals have to be aware of the different options and that all of these options, including burning with any level of fire severity, may have both beneficial and detrimental effects. The steps in planning prescribed fires include: 1) evaluation the existing and potential vegetation and ecosystem, 2) determination of the management objectives and target species, 3) evaluation of the fire environment of the proposed treatment area, 4) prediction of occurrence, abundance and growth rates of plants following fire and alternative vegetation management treatments, and 5) choice of the appropriate treatment (Feller, 1996). The results of the studies discussed in this thesis contribute to the information needed in step 4) prediction of the characteristics of the vegetation after disturbance, as well as in undisturbed areas. Managers need data that integrates the complexities of causality and that allows judgement of the applicability of study results to specific situations. Therefore, an attempt was made to include as many variables as was practical to measure in the prediction equations and analyses of variance. These analyses in most parts of this thesis were multiple correlation rather than specific tests of the effects of a single variable under controlled conditions. Therefore, many of the conclusions must be carefully evaluated in light of established biological principles. The necessity for drastic data transformations and the distribution patterns of the residuals suggest that the linear and logistic models used in this thesis were not always adequate to model the patterns of causality found in complex biological systems. The information provided by the studies described in Chapters 2 and 3 was quite general with respect to the effects on the vegetation, but very specific with respect to location, season of burning, and the ranges of fire and thinning severity. Therefore, if more specific information about the response of a single species, or about plant response under different burning conditions is required, an exhaustive literature review or even new studies will be necessary. In the case of choosing a treatment to counteract the effects of tree encroachment due to fire suppression in the IDF, prescribed fire and manual tree removal could both be treatment alternatives. Prescribed fires in standing forests resulted in the differential mortality of smaller trees, especially those of fire-susceptible species, such as lodgepole pine. This effect is not mimicked by diameter limit logging, where all trees above a certain diameter are cut. However, thinning from below can be used to selectively remove the smaller, more susceptible trees to any desired density. Thinning, especially if combined with pruning of dead lower 132 branches and pile burning of the woody residues as was done at Knife Creek, may reduce the likelihood of crown fires, and in overstocked stands it would also result in the added benefit of concentrating timber growth on fewer trees. Cover of vascular understory species, which provide wildlife browse, was not greatly affected by the amount of overstory reduction performed in this study, but may increase if all trees in the thickets were completely removed. A more severe level of thinning would also be required to preserve the aesthetic value of open stands. It is possible that the severity of budworm outbreaks could also be reduced by removing affected or susceptible trees, but this needs to be investigated. Change of the tree species composition may be possible if trees of undesirable species were mechanically removed. However, a mineral seedbed for more desirable species, such as ponderosa pine, may need to be created by mechanical scarification. The other alternative, prescribed fire, may also not be 100% effective in reversing all effects of fire suppression. After the experimental burns performed in the southern Rocky Mountain Trench, the distribution of dead trees was patchy, and many of the small trees marked for the study survived, some showing no evidence of burning. This partial retention of thickets results in some of the problem persisting in the stand. However, fire can be combined with manual thinning if costs warrant. Fire also resulted in the death of some larger trees, which represents a loss of harvestable timber unless salvage logging is done. To create a mineral seedbed for desirable species such as ponderosa pine, a fire of at least moderate severity would be necessary. However, burning, especially with higher severity, wouid likely result in a long-term loss of the moss and lichen layers, and a temporary reduction of cover and species diversity of vascular plants. Managers have to decide whether the reductions in understory cover are acceptable or not. While in this thesis the relationship between tree size, fire damage and mortality of different conifers and the effects of burning on understory vegetation were examined, several areas remain where additional research is necessary to establish the link between appropriate prescriptions to produce the desired effects. The equations by Van Wagner (1973) assume constant fire intensities throughout a stand. However, the patchy stand structure in the IDF results in patchy fuel loads, which means that fire intensity and flame heights, and therefore scorch heights, vary throughout the stand. Scorch height distributions for the two burns where tree mortality was assessed in the present study were outlined in Braumandl et al. (1995), and after both fires many unscorched trees were found. A model predicting the likelihood of crown scorch or bark char from the quantity, arrangement and patchiness of surface fuels and the 133 proximity and size of neighbouring trees (ladder fuels) would permit the prediction of tree mortality for a stand level rather than for individual damaged trees. While the present study established that the post-burn vegetation cover is somewhat predictable when the pre-burn vegetation and fire severity are known, a link between a given fire prescription or the indices and codes of the Canadian Fire Weather Index (FWI) System could not be tested. Forest floor depth of burn is theoretically related to the Duff Moisture Code (DMC) and the Buildup Index (BUI), but correlations between these codes and indices and depth of burn have been found to be poor for slashburns in British Columbia by a number of studies reviewed by Feller (1996). However, all of these studies involved slashburns in moister ecosystems than the IDF. The Canadian Forest Fire Behaviour Prediction (FBP) System predicts forest floor consumption in the C-7 (Ponderosa Pine-Douglas-fir) fuel type with a nonlinear regression equation from the Fine Fuel Moisture Code (FFMC) (Forestry Canada Fire Danger Group, 1992), but it is doubtful that this relationship would be applicable in pure Douglas-fir stands, where the forest floor wOuld likely be thicker and fine fuel more scarce. The failure of a prescribed burn in the interior of a stand at Knife Creek indicated that at the stand edge and in large openings fine fuel loads were adequate.and appropriate burning conditions were predictable from the FWI codes, while this was not true in areas with greater canopy closure. While fuel distribution was controlled in the present study, patchiness of the natural fuel complex would also result in heterogeneous forest floor reduction and burn severity. More research into the relationships between the codes and indices of the FWI system, fuel distribution, and depth of burn is needed, especially for areas that do not strictly represent either the C-7 fuel type, or any of the other recognized fuel types. 134 L I T E R A T U R E C I T E D Agee, J.K. 1993. Fire ecology of Pacific Northwest forests. Island Press, Washington, DC. 493 pp. Agee, J.K. and M. H. Huff. 1986. Structure and process goals for vegetation in wilderness areas In Lucas, R.C. (ed.). Proceedings: National Wilderness Research Conference, current research, held July 23-26,1985. USDA For. Serv. Gen. Tech. Rep. INT-212. pp. 17-25. Amman, G.D. and K G Ryan. 1991. Insect infestation of fire-injured trees in the Greater Yellowstone Area. USDA For. Serv. Res. Note INT-398. Anderson, R.C., O.L. Loucks, and A.M. Swain. 1969. Herbaceous response to canopy cover, light intensity, and throughfall precipitation in coniferous forests. Ecology 50:255-263. Armour, CD., S.C. Bunting and L.F. Neuenschwander. 1984. Fire intensity effects on the understory in ponderosa pine forests. J. Range Manage. 37:44-49. Arno, S.F. 1976. The historical role of fire on the Bitterroot National Forest. USDA For. Serv. Res. Paper INT-187. Arno, S.F., J.H. Scott, and M.G. Hartwell. 1995. Age-class structure of old-growth ponderosa pine/Douglas-fir stands and its relationship to fire history. USDA For. Serv. Res. Paper INT-481. Barrett, S.W. 1988. Fire suppression's effects on forest succession within a central Idaho wilderness. West. J. Appl. For. 3:76-80. Barrett, S.W., S.F. Arno and C.H. Key. 1991. Fire regimes of western larch-lodgepole pine forests in Glacier National Park, Montana. Can. J. For. Res. 21:1711-1720. Bevins, CD. 1980. Estimating survival and salvage potential of fire-scarred Douglas-fir. USDA For. Serv. Res. Note INT-287. Blom, G. 1958. Statistical estimated and transformed beta variables. New York: John Wiley and Sons Inc. Braumandl, T.F. and M.P. Curran (comps. and eds.). 1992. A field guide for site identification and interpretation for the Nelson Forest Region. BC Min. For. Land Manage. Handbook No. 20. Nelson, BC. Braumandl, T.F., S. Taylor, CF. Thompson, A.J. Stock and D Gayton. 1995. Ecosystem Maintenance Demonstration Burns Monitoring Project. General work plan and establishment report for Picture Valley and Findlay Creek sites. Internal report. BC Min. For. Research Branch. Nelson Forest Region, Nelson, BC. Brown, J.K. and N.V. Debyle. 1987. Fire damage, mortality and suckering in aspen. Can. J. For. Res. 17:1100-1109. 135 Brown, J.K. and N.V. DeByle. 1989. Effect of prescribed fire on biomass and plant succession in western aspen. USDA For. Serv. Gen. Tech. Rep. INT-412. Brulisauer, A.R., G.E. Bradfield, and J. Maze. 1996. Quantifying organizational change after fire in lodgepole pine forest understorey. Can. J. Bot. 74:1773-1782. Buchman, R.G. 1985. Mortality assessment: What are the essentials. In Van Hooser, D.D. and N. Van Pelt (eds.) Proceedings-Growth and yield and other mensurational tricks: A regional technical conference, held Nov. 6-7,1984. Logan, Utah. USDA For. Serv. Gen. Tech. Rep. fNT-193. pp. 21-24. Clark, R.G. and E.E. Starkey. 1990. Use of prescribed fire inrangeland ecosystems. In J.D. Walstad, S.R. Radosevich, and D. V. Sandberg (eds.). Natural and prescribed fire in Pacific Northwest forests. Oregon State University Press, Corvallis, Oregon, pp. 39-52. Clary, W.P. 1975. Range management and its ecological basis in the ponderosa pine type in Arizona: the status of our knowledge. USDA For. Serv. Res. Paper RM-158. Clement, B. and J. Touffet, 1990. Plant strategies and secondary succession on Brittany heathlands after severe fire. J. Veg. Sci. 1:195-202. Coates, D., S. Haeussler, and J. Mather. 1990. A guide to the response of common plants in British Columbia to management treatmetns. For. Can. and BC Min. For. FRDA Handbook 8,154 pp. Conover, W.J. 1980. Practical nonparametric statistics. Wiley, New York. 493 pp. Crane, M.F., J.R. Habeck, and W.C. Fisher. 1987. Early postfire revegetation in a western Montana Douglas-fir forest. USDA For. Serv. Res. Paper FNT-319. Crouch, G.L. 1986. Effects of thinning pole-sized lodgepole pine on understory vegetation and large herbivore activity. USDA For. Serv. Res. Paper RM-268. Daniels, L.D., J. Dobry, and K. KjjLnka. 1995. Fire history of two Douglas-fir stands in the Alex Fraser Research Forest: A Pilot Study. Internal report to the Cariboo Forest Region, BC Ministry of forests. Daubenmire, R. 1968. Ecology of fire in grasslands. Adv. Ecol. Res. 5:209-266. Dodd, C.J.H., A McLean, and V.C. Brink. 1972. Grazing values as related to tree-crown covers. CanJ.For.Res. 2:185-189. Faurot, J.L. 1977. Estimating merchantable volume and stem residue in four timber species: ponderosa pine, lodgepole pine, western larch, Douglas-fir. USDA For. Serv. Res. Pap. INT-196. Feller, M.C. 1982. The ecological effects of slashburning with particular reference to British Columbia: A literature review. Ministry of Forests Land Management Report 13. Victoria, BC. 136 Feller, M.C. 1996. Use of prescribed fire for vegetation management. In P.G. Comeau, G.J. Harper, M.E. Blache, J.O. Boateng and L.A. Gilkeson (eds.). Integrated forest vegetation management: Options and applications: proceedings of the fifth BC Forest Vegetation Management Workshop. Canadian Forest Service and BC Min For. FRDA Rep. 251, Victoria, BC. pp 17-34. Feller, M.C. and J.P. Kimnuns. 1984. Effects of clearcutting and slashburning on streamwater chemistry and watershed nutrient loss in southwestern British Columbia. Water Resour. Res. 20:29-40. Fisher, W.C. and A..F. Bradley. 1987. Fire ecology of western Montana habitat types. USDA For. . Serv. Gen. Tech. Rep. INT-223. Fleischner, T.L. 1994. Ecological costs of livestock grazing in western North America. Cons. Biol. 8:629-644. Flinn, M.A. and R.W. Wein. 1977. Depth of underground plant organs and theoretical survival during fire. Can J. Bot. 55:2550-2554. Flinn, M.A. and R.W. Wein. 1988. Regrowth of forest understory species following seasonal burning. Can. J. Bot. 66:150-155 Forestry Canada Fire Danger Group. 1992. Development and structure of the Canadian Forest Fire Behaviour Prediction System. Information Report ST-X-3. Forestry Canada, Ottawa. Geier-Hayes, K. 1989. Vegetation response to helicopter logging and broadcast burning in Douglas-fir habitat types at Silver Creek, Idaho. USDA For. Serv. Res. Paper INTA05. Geiszler, D.R., R.I. Gara amd W.R. Littke. 1984. Bark beetle infestations of lodgepole pine following a fire in south-central Oregon. Zeit. Angw. Entomol. 39:389-394. Grelen, H.E. and R.E. Lohrey. 1978. Herbage yield related to basal area and rainfall in a thinned longleaf plantation. USDA For. Serv. Res. Note SO-232. Grieg-Smith, P. 1964. Quantitative plant ecology. Butterworth and Co, London, Great Britain 256 pp. Grigal, D.F. and J.G. McColl. 1975. Litter fall after wildfire in virgin forests of northeastern Minnesota. Can. J. For. Res. 5:655-661. Habeck, J.R. 1990. Old-growth ponderosa pine-western larch forests in western Montana: Ecology and management. Northwest Environ. J. 6:271-292. Halpern, C.B. 1989. Early successional patterns of forest species: Interactions of life history traits and disturbance. Ecology 70:704-720. Hamilton, D.A. Jr., and B.M. Edwards. 1976. Modelling the probability of individual tree mortality. USDA For. Serv. Res. Pap. PNT-185 137 Hamilton, E.H. 1988. Impacts of prescribed burning on soil-vegetation relationships in the Sub-Boreal Spruce zone. In M.C. Feller and S.M. Thompson (eds.). Wudlife and range prescribed burning workshop proceedings, held on October 27 and 28,1987, Richmond, BC. Faculty of Forestry, University of BC, Vancouver, BC. pp. 171-184. Harrington, M.G. 1993. Predicting Pinus ponderosa mortality from dormant season and growing season fire injury. Int. J. Wildland Fire 3:65-72. Hartford, R.A. and W.H. Frandsen. 1992. When it's hot, it's hot..or maybe it's not! (Surface flarruhg may not portend extensive soil heating). Int. J. of Wildland Fire 2:139-144. Hungerford, R.D., M.G. Harrington, W.H. Frandsen, KCRyan, and G.J. Niehoff. 1990. Influence of fire on factors that affect site productivity. In A.E. Harvey and L.F. Neunschwander (Compilers) Proceedings-Management and productivity of western forest soils. USDA For. Serv. Gen. Tech. Rep. INT-280. pp. 29-31. Huynh, H. and L.S. Feldt. 1976. Estimation of the Box Correction for Degrees of Freedom from Sample Data in Randomized Block and Split-Plot Designs. J. Ed. Stat. 1:69-82. Irwin, L.L., J.G. Cook, R.A. Riggs and J.M. Skovlin. 1994. Effects of long-term grazing by big game and livestock in the Blue Mountains forest ecosystems. USDA For. Serv. Gen. Tech. Rep. PNW-325. Jonsson, B.G. 1993. The bryophyte spore bank and its role after small-scale disturbance in a boreal forest. Journal of Vegetation Science 3:729-832. Jungen, J.R. 1980. Soil Resources of the Nelson Map Area. R.A.B. Bull. 20. (Soil Survey Report No. 28.) Resource analysis Branch, BC Min. Environ., Kelowna, BC. Kauffman, J.B. 1990. Ecological relationships of vegetation and fire in Pacific Northwest forests. In J.D. Walstad, S.R. Radosevich, and D.V. Sandberg (eds.). Natural and prescribed fire in Pacific Northwest forests. Oregon State University Press, Corvallis, Oregon, pp. 39-52. Keane, R.E., S.F. Arno and J.K. Brown. 1989. FIRESUM-an ecological process model for fire succession in western conifer forests. USDA For. Serv. Gen. Tech. Rep. INT-266. Kelley, and P.N. Sprout. 1956. Soil Survey of the upper Kootenay and Elk River valleys in the East Kootenay District of British Columbia. Report No. 5 of the British Columbia Soil Survey. BC Dept. of Agriculture, Victoria, BC. Kent, M. and P. Coker. 1992. Vegetation description and analysis: a practical approach. CRC Press, Boca Raton, Florida. 363 pp. Kilgore, B. 1985. Evaluating direct response to understory burning in a pine-fir-larch forest in Glacier National Park. In Lucas, R.C. (ed.). Proceedings: National Wilderness Research Conference, current research,, held July 23-26,1985. USDA For. Serv. Gen. Tech. Rep. INT-212. pp.17-25. Knapp, A.K. 1984. Post-burn differences in solar radiation, leaf temperature and water stress influencing production in a lowland tallgrass prairie. Amer. J. Bot. 71:220-227. 138 Krajina, V.J. 1969. Ecology of forest trees in British Columbia. Ecol. West. N. Amer. 2:1-146. Kramer, N.B. and F.D. Johnson. 1987. Mature forest seed banks of three habitat types in central Idaho. Can. J. Bot. 65:1961-1966. Lacelle, L. 1990. Biophysical resources of the East Kootenay area: soils. British Columbia Soil Survey Report No. 20. Habitat Inventory Section, Victoria, BC. Leps, J. and V. Hadincova. 1992. How reliable are our vegetation analyses. Journal of Vegetation Science, 3:119-124. Lord, T.M. 1984. Soils of the Horsefly Area, BC. British Columbia Soil Survey Report No. 32. Research Branch, Agriculture Canada, Ottawa, ON. Low, D.J. 1988. Effects of prescribed burning on non-target wildlife species associated with fire-prone ecosystems in the southern Interior of British Columbia. In M.C. Feller and S.M. Thompson (eds.). Wildlife and range prescribed burning workshop proceedings, held on October 27 and 28,1987, Richmond, BC. Faculty of Forestry, University of BC, Vancouver, BC. pp. 185-196 Lyon, L.J. 1984. The Sleeping Child Burn-21 years of postfire change. USDA For. Serv. Res. Pap. fNT-330. Lyon, L.J. 1971. Vegetal development following prescribed burning of Douglas-fir in South-Centreal Idaho. USDA For. Serv. Res. Pap. INT-105. Malanson, G.P. 1987. Diversity, stability and resilience: Effects of fire regime. In L. Trabaud (ed.). The role of fire in ecological systems. SPB Academic Publishing. The Hague, The Netherlands, pp. 49-63. McConnell, B.R. and J.G. Smith. 1970. Response of understory vegetation to ponderosa pine thinnng in eastern Washington. J. Range Manage. 23:208-212. McLaughlin, S.P. 1978. Overstory attributes, light, throughfall, and the interpretation of overstory-understory relationships. For. Sci. 24:550-553. McLean, A. 1969. Fire resistance of forest species as influenced by root systems. J. Range Manage. 22: 120-122. McNabb, D.H.and K. Cromack. 1990. Effects of prescribed fire on nutrients and soil productivity. In J.D. Walstad, S.R. Radosevich, and D.V. Sandberg (eds.). Natural and prescribed fire in Pacific Northwest forests. Oregon State University Press, Corvallis, Oregon, pp 39-52. McRae, D.J., M.E. Alexander and B.J. Stocks. 1979. Measurement and description of fuels and fire behaviour on prescribed burns, a handbook. Canadian Forest Service Information Report O-X-287. Meidinger, D. and J. Pojar, (eds). 1991. Ecosystems of British Columbia. Special Report Series No. 6, BC Min. For. Victoria, BC. 139 Miller, R.E. and K.W. Seidel. 1990. Effects of prescribed fire on timber growth and yield. In J.D. Walstad, S.R. Radosevich, and D.V. Sandberg (eds.). Natural and prescribed fire in Pacific Northwest forests. Oregon State University Press, Corvallis, Oregon, pp 39-52. Minitab Inc. 1995. Minitab Reference Manual, Release 10 Xtra. Minitab Inc. Mitchell, R.G. 1990. Effects of prescribed fire on insect pests. In J.D. Walstad, S.R. Radosevich, and D.V. Sandberg (eds.). Natural and prescribed fire in Pacific Northwest forests. Oregon State University Press, Corvallis, Oregon, pp. 111-11.6 Norusis, M. J. 1997. SPSS v. 7.5. Professional Statistics. SPSS Inc. Chicago, Illinois. Noste, N.V. and C.L. Bushey. 1987. Fire responses of shrubs in dry forest habitat types in Montana and Idaho. USDA For. Serv. Gen. Tech. Rep. INT-239. Noste, N.V., E.D.Reinhardt and R.A. Wilson Jr. 1988. Fire effects on Amelanchier alnifolia shrubs during phenological development. In A. Wallace, E.D. McArthur and M.R. Hafercamp (Comp.) Proceedings - Symposium on shrub ecophysiology and biotechnology, held on June 30 - July2,1987. Logan, Utah. USDA For. Serv. Gen. Tech. Rep. INT-256. pp. 158-163. 0kland ,R.H, and O. Eilertsen. 1996. Dynamics of understory vegetation in an old-growth boreal coniferous forest, 1988-1993. J. Veg. Sci. 7:747-762. Parish, R., R. Coupe and D. Lloyd (eds.). 1996. Plants of Southern Interior British Columbia. Lone Pine Publishing, Vancouver, BC. 463 pp. ParTranter, J. 1992. Typical historic patterns of wMfire disturbance by biogeoclimatic zone. Pamphlet. BC Min. For. Protection Branch, Victoria, BC. Peterson, D.L. 1985. Crown scorch volume and scorch height: estimates of postfire tree condition. Can. J. For. Res. 15:596-598. Peterson, D.L. and M.J. Arbaugh. 1986. Postfire survival in Douglas-fir and lodgepole pine: Comparing the effects of crown and bole damage. Can. J. For. Res. 16:1175-1179. Peterson, D.L. and M.J. Arbaugh. 1989. Estimating postfire survival of Douglas-fir in the Cascade Range. Can. J. For. Res. 19:530-533. Qian, H. and K. Klinka. 1998. Plants of British Columbia: Scientific and common names of vascular plants, bryophytes, and lichens. University of British Columbia Press, Vancouver, BC. 534pp. Rabotnov,T.A. 1974. Differences between fluctuations and successions. In: Knapp, R. (ed.) Vegetation dynamics. Junk, The Hague, pp. 21-24. Regelbrugge, J.C. and S.G. Conard. 1993. Modeling tree mortality following wuclfire in Pinus ponderosa forest in the central Sierra Nevada of California, hit. J. Wildland Fire 3:139-148. Reinhardt, E.D. and K.C. Ryan. 1988a. Eight-year tree growth following prescribed under-burning in a western Montana Douglas-fir/western larch stand. USDA For. Serv. Res. Note INT-387. 140 Reinhardt, E.D. and K.C. Ryan. 1988b. How to estimate tree mortality resulting from under-burning. Fire Manage. Notes 49 (4): 30-36. Riegel, G.M., R.M. Miller and W.C. Krueger. 1995. The Effects of aboveground and belowground competition on understory species composition in a Pinus ponderosa forest. For. Sci. 41:864-889. Rowe, J.S. 1983. Concepts of fire effects on plant individuals and species. In R.W. Wein and D.A. McLean (eds.). The role of fire in northern circumpolar ecosystems. SCOPE 18. John Wiley and Sons, New York. pp. 135-154. Ryan, K G and N. V. Noste. 1985. Evaluating prescribed fires. In J.E. Lotan, B.M. Kilgore, W.C. Fisher and R.W. Mutch (tech. coords.). Proceedings-symposium and workshop on wilderness fire, Missoula, Montana. USDA For. Serv. Gen. Tech. Rep. INT-182. pp. 239-238. Ryan, K.C., D.L. Peterson, and E.D. Reinhardt. 1988. Modeling long-term fire-caused mortality of Douglas-fir. Forest Sci. 34:190-199. Ryan, K G and E.D. Reinhardt. 1988. Predicting postfire mortality of seven western conifers. Can. J. For. Res. 18:1291-1297. Saveland, J.M. 1982. Predicting mortality and scorch height in ponderosa pine from understory prescribed burning. M.S. thesis, Forest Resources, University of Idaho. Moscow, Idaho. 36 pp. Schirnmel, J. and A. Granstrom. 1996. Fire severity and vegetation response in the boreal Swedish forest. Ecology. 77:1436-1450. Schofield, W.B. 1985. Introduction to Bryology. Macmillan Publishing Co. New York, New York. 431 pp. SPSS Inc. 1997. Systat 7.0 for Windows: Statistics. SPSS Inc. Chicago, fllinois. 751 pp. Steen, O. A., and R. A. Coupe. 1997. A Field Guide to Forest Site Identification and Interpretation for the Cariboo Forest Region. Research Branch, BC Min. For. Williams Lake, BC. Stickney, P.F. 1980. Database for post-fire succession, first 6 to 9 years, in Montana larch-fir forests. USDA For. Serv. Gen. Tech. Rep. INT-62. Stickney, P.F. 1981. Vegetative recovery and development. In DeByle, N.V. (ed.). Clearcutting and fire in the Larch/Douglas-fir forests of Western Montana: A multifaceted research summary. USDA For. Serv. Gen. Tech. Rep. INT-99. pp. 33-44. Stickney, P.F. 1985. Data base for early postfire succession on the Sundance burn, northern Idaho. USDA For. Serv. Gen. Tech. Rep. INT-189. Stone, W.E. and M.L. Wolfe. 1996. Response of understory vegetation to variable tree mortality following a mountain pine beetle epidemic in lodgepole pine stands in northern Utah. Vegetatio 122:1-12. 141 Strang, R.M. and A.H. Johnson. 1988. Prescribed burning trialson the Dewdrop Ranges, Kamloops-an mterirn report. In M.C. Feller and S.M. Thompson (eds.). Wudlife and range prescribed buming workshop proceedings, held on October 27 and 28,1987, Richmond, BC. Faculty of Forestry, University of BC, Vancouver, BC. pp. 134-138. Sykes, J.M., A.D. Horrill and M.D. Mountford. 1983. Use of visual cover assessments as quantitative estimators of some British woodland taxa. J. Ecol. 71:437-450. Thies, W.G. 1990. Effects of prescribed fire on diseases of conifers. In J.D. Walstad S.R. Radosevich, and D.V. Sandberg (eds.). Natural and prescribed fire in Pacific Northwest forests. Oregon State University Press, Corvallis, Oregon pp. 117-121. Thomas, P. A. and R.W. Wein. 1994. Amelioration of wood ash toxicity and jack pine establishment. Can. J. For. Res. 24:748-755. Tisdale, E.W. 1950. Grazing of forest lands in British Columbia. J. For. 48 (12): 856-860. Tisdale, E.W. and A. McLean. 1957. The Douglas-fir zone of southern British Columbia. Ecol. Monogr. 27:246-266. Trabaud, L. 1987. Fire in ecological systems. SPB Academic Publishers. The Hague, Netherlands. 157 pp. Van Wagner, C.E. 1973. Height of crown scorch in forest fires. Can J. For. Res.3: 373-378. Viereck, L.A. 1983. The effects of fire in spruce ecosystems of Alaska and northern Canada. In R. W. Wein and D.A. McLean (eds.), The role of fire in northern circumpolar ecosystems. SCOPE 18. John Wiley and Sons. pp. 201-220 Vitt, D.H., J.E. Marsh and R.B. Bovey. 1988. Mosses, Lichens and Ferns of Northwest North America. Lone Pine, Edmonton, AB. Vogl, R. 1974. Effects of fire on grasslands, pp. 139-194 in T.T. Kozlowski and CE. Ahlgren (eds.). Fire and ecosystems. Academic press, New York. Wade, D.D. 1986. Linking fire behavior to its effect on living plant tissue. In Forests, the world and the profession: Proceedings of the Society of American Foresters National Convention, held October 5-8,1986. Birmingham, Alabama. Soc. Amer. For. Bethesda, Maryland, pp. 112-116. Walstad, J.D. and K.W. Seidel. 1990. Use and benefits of prescribed fire in reforestation. In J.D. Walstad S.R. Radosevich, and D.V. Sandberg (eds.). Natural and prescribed fire in Pacific Northwest forests. Oregon State University Press, Corvallis, Oregon, pp. 67-80. Waterhouse, M.J., H.M Armleder, and R.J. Dawson. 1994. Winter Food Habits of Mule Deer on the Central Interior of British Columbia. BC Min. For. Research Note no. 113. BC Min. For, Research Branch, Williams Lake, BC. 142 Wein, R.W. and M.A. El-Bayourni. 1983. Lirnitations to predictability of plant succession in northern ecosystems. In R.W. Wein, R.R. Riewe and I.R. Methven (eds.). Resources and dynamics of the boreal zone: proceedings of a conference held at Thunder Bay, ON August 1982. Association of Canadian Universities for Northern Studies, Ottawa, ON. pp. 214-225. Whittle, C.A., L.C. Duchesne and T. Needham. 1997. The Impact of broadcast burning and fire severity on species composition and abundance of surface vegetation in a jack pine (Pinus banksiana) clear cut. For. Ecol. Manage. 94:141-148. Wikeem, N.M. and R.M. Strang. 1983. Prescribed burning on BC rangelands: the state of the art. J. Range Manage. 36:3-8. Wright, H.A. 1972. Shrub response to fire. In CM. McKell, J.P. Blaisdell and J.R. Goodin (eds.) Wildland Shrubs-Their Biology and Utilization. USDA For. Serv. Gen. Tech. Rep. INT-1. pp. 204-217. Wright, H.A. and A.W. Bailey. 1982. Fire Ecology: United States and southern Canada. John Wiley and Sons, New York. Wyant, J.G., P.N. Omi, and R.D. Laven. 1986. Fire induced tree mortality in a Colorado ponderosa pine/Douglas-fir stand. For. Sci. 32:49-59. Yearsley, H.K. 1993. Forest floor seed banks and their response to slashburning in some ecosystems in south central British Columbia. M.Sc. Thesis, University of BC, Vancouver, BC. 136 pp. 143 Appendix 2.1 Mean percent cover values for species found in the open- and closed-canopy stands at Tulip Creek and Twin Lakes in July 1994,1995 and 1996. Species are arranged alphabetically within growth forms. A "+" denotes a cover of < 0.1%. Tulip Creek Twin Lakes closed open closed open 1994 1995 1996 1994 1995 1996 1994 1995 1996 1994 1995 1996 TREE SEEDLINGS Pseudotsuga menziesii 0.7 0.1 0.2 0.4 0.3 0.6 0.4 0.2 0.2 3.2 1.3 0.7 SHRUBS Amelanchier alnifolia 0.3 0 + 0.8 0.3 0.4 • 0 0 + Arctostaphylos uva-ursi 2.0 1.8 1.4 7.3 3.0 5.3 Holodiscus discolor 6.50 3.25 4.75 1.82 1.40 0 Lonicera ciliosa/L. utahensis1 0.4 0.2 0.4 0.1 0.1 0.2 Mahonia aquifolium + M. repens 1.4 1.4 1.8 14.4 14.0 11.6 9.3 3.8 10.1 11.6 4.5 7.4 Paxistima myrsinites 0 0.1 + Philadelphus lewisii 0.1 0.2 0.2 Physocarpus malvaceous 1.1 2.3 2.8 2.5 1.5 2.4 Rosa acicularis 1.2 1.1 1.1 1.4 0.6 0.7 0.4 0.2 0.5 Spirea betulifolia 2.4 1.7 1.9 2.0 2.3 1.8 9.3 3.7 5.7 7.3 3.2 3.9 Symphoricarpos albus 3.1 2.5 3.1 8.3 9.2 8.9 1.8 0.6 2.2 0 + 0 Vaccinium sp. 1.9 1.1 2.3 0.9 0.1 0.6 TOTAL WOODY PLANTS 15.8 11.1 15.1 30.6 30.0 26.7 26.7 12.1 22.8 30.7 12.3 18.3 FORBS Achillea millefolium 1.6 1.2 1.3 0.1 0.1 0.1 1.0 0.5 0.9 Agoseris glauca 0 0 + Allium cernuum 0.1 0.3 0.3 0.4 0.6 0.5 Anemone multifida + 0 + Antennaria microphylla 3.0 0.5 0.6 Antennaria neglecta 0.5 0.4 0.2 2.7 2.1 3.5 Antennaria racemosa 0. 0.1 0.1 Armaria serpyllifolia 0 0.1 0.1 Aralia nudiacaulis 6.0 8.0 2 Arnica cordifolia 0 0.3 0.7 0 0 0.1 1.4 2.7 5.7 1.1 2.7 6.7 Aster ciliolatus 0 0 0.1 Aster conspicuus 0.2 0.4 0.3 0.3 0.2 0.2 0.4 0.2 0.2 0 0.1 0.1 Aster/Erigeron sp. 1 0 0.3 0 Aster/Erigeron sp. 2 + + 0 Calochortus apiculatus + + 0.1 0 0.1 0.3 Castilleja miniata 0.4 0.4 0.2 Centaurea maculata 0.6 2.1 1.4 Claytonia perfoliata 0 0 + Collinsia parviflora 0 0 0.2 + 0.4 0.2 144 Tulip Creek Twin Lakes closed open closed open 1994 1995 1996 1994 1995 1996 1994 1995 1996 1994 1995 1996 Disporum trachycarpum 0 0 0.1 Epilobium brachycarpum 0 0 + Epilobium sp. 1 0 0 + Epilobium spp. 0 0.1 0.2 Erigeron sp. 1 0.1 0 0 Erigeron sp. 2 0 0 0.3 Erigeron/Aster sp. 0.1 0 0 Filago arvensis 0 0 + Fragaria vesca 0.2 0.6 0.7 0.4 1.4 1.4 0 0 0.2 Fragaria virginiana 1.1 0.6 0.5 0.7 0.6 0.8 1.7 0.7 2.0 Galium boreale 0.3 0.2 0.5 + 0.1 + Gentianella amarella 0 0 0.1 Goodyera oblongifolia 0.3 0.1 0.2 0 0 + 0.2 0 + + + 0.1 Hedysarum sulphurescens 0.2 0.1 0.5 Heuchera cylindrica 0 0.1 0.1 0.1 0.1 0.2 Hieracium albiflorum 0 1.8 1.1 0.1 0.1 0.1 0.3 0.1 0.2 Hieracium umbellatum + 0 0 Lathyrus ochroleucus 0.1 0.1 0.4 Lithospermum ruder ale 0.1 + 0 Lupinus sericeus. 0.5 1.0 1.4 0.9 0.8 0.7 0.5 0.4 1.1 Maianthemum racemosum 0.5 + 0.1 0 0 + Maianthemum stellatum 0.2 0.1 0.1 Medicago lupulina + Trifolium aureum 2 6.9 4.2 8.9 Orthilia secunda 0 0 0.1 Osmorhiza berteroi 0 0.2 0.4 Penstemon confertus 1.4 0.8 0.8 Plantago lanceloata 0 0.1 0.2 Platanfhera sp. + 0 + Polygonum douglasii 0 0.1 0 Potentilla arguta 0 0.1 0.1 Potentilla gracilis 0.8 2.0 2.9 Pyrola chlorantha 0 + + Rumex acetosella 0 0 + Taraxacum officinale 0 + 0 0 0.3 0.8 0 0.2 0.1 Tragopogon dubius + + 0 0 0.2 0 Trifolium arvense 0 + 0.1 Trifolium pratense + 0 0 Trifolium repens 3.8 0.5 0.2 Veronica serpyllifolia 0 0 0.1 Vicia americana 0.2 0.1 0.1 Vicia sp. 0 0 0.1 Viola adunca 0.1 0.2 0.1 + 0.1 + forb sp. 1 0 0 0.1 145 Tulip Creek Twin Lakes closed open closed open 1994 1995 1996 1994 1995 1996 1994 1995 1996 1994 1995 1996 forb sp. 2 0 0 + forb sp. 3 0 0 + forb sp. 4 0 0 0.1 forb sp. 5 0 0 0.1 forb sp. 6 0.1 0 0 forb sp. 7 0 + 0 forb sp. 8 0 0 + TOTAL FORBS 7.3 11.7 13.7 17.4 15.3 21.1 4.8 5.3 9.4 11.8 9.3 18.3 GRAMINOIDS Bromus sp. 0 0 0.1 Bromus tectorum 0 + + Calamagrostis rubescens 7.6 8.1 6.9 16.0 12.8 10.0 6.5 4.3 6.9 24.2 17.3 19.8 Carex concinnoides 0 0 + + 0.6 + 0.8 0.5 0.4 1.0 1.1 1.2 Carex rossii 0 0 0.7 0 0.2 0.1 Carex sp. 1 0 0 + Carex sp. 2 0 0 + Carex sp. 3 0 0 0.2 Elymus glaucus 0.2 0.1 0.1 1.0 1.6 1.8 0.3 0.1 + Elymus lanceloatus + + 0 Elymus trachycaulus 0 0 0.2 Festuca campestris + 0.1 0.3 Festuca occidentalis + F. saximontana 2 0.6 0.7 3.1 0 0.3 0.6 0.2 0.6 0.2 0. 0 Festuca sp. 1 0 0.4 0.3 Festuca sp. 2 1.3 + 0.4 Festuca sp. 3 0 0 2.1 Koeleria macrantha 0.2 0.3 + Leymus innovatus 0 0 0.3 Poa pratensis 0 0 0.4 + 0 0 Poa sp. 0.9 0.2 0.1 Pseudoroegneria spicata 0 0 0.5 0 0.3 0.6 Stipa occidentalis 0 0 + Trisetum cernuum var. canescens 0 0 + Trisetum cernuum var. cernuum 0 1.5 0.6 0.7 1.4 0.3 grass spp. 0 0.6 0.5 0 + + . 0.2 0.3 + TOTAL GRAMINOIDS 8.0 10.2 8.1 23.1 18.2 15.9 8.2 5.2 8.0 25.6 19.1 23.7 MOSSES Atrichum tenellum 0.1 0.6 0.1 Aulacomnium androgynum 0.2 0.4 0.2 0 0.1 0.1 0.3 0.2 0.3 0 0 0.5 Barbula convoluta + + 0 146 Tulip Creek Twin ^akes closed open closed open 1994 1995 1996 1994 1995 1996 1994 1995 1996 1994 1995 1996 Brachythecium leibergii 0.1 0.1 0.2 Brachythecium hylotapetum 0.2 0.1 0 Brachythecium sp. 1.6 1.2 1.4 22.9 9.8 5.5 5.3 3.0 3.1 8.8 2.4 3.7 Bryum caespiticium + 0.1 + Bryum flaccidum + 0.8 + Ceratodon purpureus + 0.1 0.1 0.1 0 0.2 0.2 0.1 0.1 Dicranum fuscescens 0 0.5 0.1 0 0.1 0 Dicranum polysetum + + + 0 + + Dicranum scoparium 0.6 0 0.3 0.1 0 0 Dicranum tauricum 0.2 0.4 0.4 + + 0.1 0.3 0.2 0.2 0.4 0.1 0.1 Didymodon sp. 0 0 0.2 Eurhynchium pulchellum 0 0 + Fissidens bryoides 0 + 0 Hylocomium splendens 0 + 0 Hypnum revolutum Mnium spinulosum 0.1 + 0 0.3 0.4 + + + 0 Pleurozium schreberi 0.8 0.5 0.5 0.9 0.7 1.2 1.1 1.2 1.0 Pohlia nutans + 0.1 0.1 + + 0.1 0.1 0.4 0.2 0.1 0.3 0.2 Polytrichum juniperinum 0 0.2 0.1 1.5 1.0 0.3 + 0.1 + 0.5 0.7 0.6 Polytrichum piliferum + + + 0 0.1 0.1 Racomitrium heterostichum + + + 0 + + Rhytidiadelphus triquetrus 1.0 0.4 0.1 0.1 0 0.3 Rhytidiopsis robusta 0.1 0.1 0.1 0 0.1 + 0 0 + Sanionia uncinata + + 0 0.1 0.1 0.1 Tortula ruralis 0.2 0.7 0.5 0 + 0 + 0.1 + moss sp. 0 0 + LIVERWORTS Barbilophozia hatcheri 0 0 + 0 0.2 0.2 Lophozia sp. 0 + 0 0 + 0 Manniafragrans 0 + 0 Ptilidium ciliare 0 0 + Ptilidium pulcherrinnum 0 0.1 0.1 0. 0 + TOTAL BRYOPHYTES 3.1 3.6 3.0 26.5 13.6 7.7 7.3 5.3 5.3 11.5 5.1 6.8 LICHENS Caloplaca sp. + + 0 Cladonia botrytes 0 + 0 Cladonia cariosa 0 + + 0.2 + + Cladonia cenotea 0 0 + 0 + 0 0 0.1 0.1 Cladonia cervicornis 0.1 + 0.1 Cladonia chlorophaea 0.1 0.2 0.3 0.6 0.2 0.2 0.1 0.1 0.2 0.6 0.3 0.3 Cladonia deformis 0 0 + Cladonia ecmocyna 0.1 0.2 0.2 147 Tulip Creek Twin Lakes closed open closed open 1994 1995 1996 1994 1995 1996 1994 1995 1996 1994 1995 1996 Cladonia fimbriata 0.3 0.2 0.2 + 0.2 0.1 0.1 0.1 + Cladonia gracilis + + + 0.2 0.1 0 Cladonia maccilenta var baccillaris + C. ochrochlora 0.1 0.2 0.3 0.2 + 0.2 0.2 0.2 0.3 0.2 0.1 + Cladonia multiformis 0 + 0.1 Cladonia phyllophora 0 + 0 Cladonia pyxidata 0.6 0.2 + Cladonia subulata 0 + 0 0 0 + 0 0 + 0.1 + + Cladonia spp. 0.4 0.2 0.10 0.1 0.2 0.1 1.7 1.4 1.5 Peltigera aphthosa + P. leucophlebia 3.0 2.8 1.8 0 0 + Peltigera canina 2.6 3.5 3.1 1.2 1.0 1.0 2.4 2.6 2.1 Peltigera didactyla var. extenuata + 0.1 0.1 Peltigera malacea 0.2 0.1 0.1 Peltigera membranacea 0 0.5 + 0 + 0 Peltigera neckeri + 0.2 0.3 0.2 0.2 0.1 0.1 + 0 Peltigera ponojensis 0.1 0 0.3 0.1 0.1 0.1 Peltgera praetextata 0.1 + 0.1 0.5 0.2 0.1 0 0.1 0 0.4 0.3 0.1 Peltigera rufescens + 0.2 0.2 3.0 2.6 2.0 Peltigera venosa + + 0.1 TOTAL LICHENS 0.6 0.6 0.8 4.4 5.3 4.6 4.8 4.4 3.6 9.9 7.8 6.5 "taxon a/ taxon b" denotes a single taxon that could not be identified definitively to either a or b. The most likely taxon was placed first; in this case Lonicera ciliosa, but it could also have been L. utahensis. 2 "sp. a + sp. b" denotes the combined percent cover of several species. •148 Appendix 2.2 Normal probability plots and P-values from the Kolmogorov-Srrurnov normality tests for the residuals of the repeated measures ANOVAs comparing change in vegetation cover at Twin Lakes and Tulip Creek for a) bryophytes in 1994 and b) lichens in 1994 where the residuals were not distributed normally. a) Bryophytes, 1994. p < 0.01 149 Appendix 2.3 Daily precipitation (in mm) at Castlegar Airport in a) 1994, b) 1995, and c) 1996. a) Castlegar Airport in 1994 1 4 7 10 13 16 19 22 25 28 31 3 6 9 12 15 18 21 24 27 30 3 6 9 12 15 18 May June July Date b) Castlegar Airport in 1995 30 25 E E 20 •B 15 • f 10 a) £ 5 0 I I I I I I I I I I M I I I I I I i i H H i ^  i n n i i i m n i I 1 4 7 10 13 16 19 22 25 28 31 3 6 9 12 15 18 21 24 27 30 3 6 9 12 May June July Date c) Castlegar Airport in 1996 E 20 c .2 15 .9- 10 0. 5 -i i n i I I i i i i i i i i i 1 4 7 10 13 16 19 22 25 28 31 3 6 9 12 15 18 21 24 27 30 3 6 9 May June July Date 150 Appendix 2.4 Daily precipitation (in mm) at a) Twin Lakes in 1994, and at Kilcornxm Creek Park b) in 1994, c) in 1995, and d) in 1996. a) Twin Lakes in 1994 30 f 25| c 2 0 •B 15 •S : § • 1 0 H i i i H i i H 1 4 7 10 13 16 19 22 25 28 31 3 6 9 12 15 18 21 24 27 30 3 6 9 12 May June July Date b) Kikomun Creek Park in 1994 30 | 25 c 20 •B 15 I I I i i w I J ^ - H - f c - i - H - + - n B i H i -^-1^- B i i i i t - i i H i i i i n i i i i i i n i i i i i i 1 4 7 10 13 16 19 22 25 28 31 3 6 9 12 15 18 21 24 27 30 3 6 9 12 May June July Date c) Kikomun Creek Park in 1995 ~ 30 E E 25 "c 20 O £ 15 Co Q. 10 cu 5 '•M i H i i i i —i—i— j l . j L . 1 4 7 10 13 16 19 22 25 28 31 3 6 9 12 15 18 21 24 27 30 3 6 9 May June July d) Kikomun Creek Park in 1996 Date ? 3 0 c 20 O += 15 w Q. 10 <D 5 MBII I I I I IMIIBI I t-JLfc-^ B-1 4 7 10 13 16 19 22 25 28 31 3 6 9 12 15 18 21 24 27 30 3 6 9 May June July II i i i i i i i i i i Date 151 Appendix 3.1 Density of the overstory in a) forest openings and b) forest thickets for the various treatments at Knife Creek. Percent Canopy Basal area (m2/ha) Number of trees Treat- Plot Cover ment a Site type N Mean Range Mean Range Mean Range a) Forest Openings OC KC-1 herb 8 12.0 0-48 6.3 0.5 - 25.9 5.3 1 -10 moss 8 43.0 5-70 21.0 12.0 - 34.2 10.5 8 -15 KC-2 herb 8 31.6 13-60 14.2 7.0 - 21.4 4.4 2 - 8 moss 8 23.4 10-55 7.7 3.6 -12.8 3.5 1 - 8 OL KC-1 herb 8 15.0 0-40 9.6 0.9 - 24.4 3.8 1 - 7 moss 8 49.6 15-70 25.2 14.5 - 35.9 11.8 4 -21 KC-2 herb 8 41.4 2-78 12.7 0.5-25.2 3.9 1 - 8 moss 8 27.6 5-70 9.0 2.8 - 20.6 4.9 1 -11 OM KC-1 herb 8 24.8 7-55 9.5 2.0 - 20.0 4.4 1 - 8 moss 8 30.3 7-50 13.8 0 - 24.1 4.5 0 - 7 KC-2 herb 9 21.4 2-40 4.8 0 -12.0 1.3 0 - 3 moss 9 30.1 8-60 8.3 0 - 26.2 • 3.3 0 -10 b) Forest Thickets FC KC-1 herb 8 59.1 27-90 18.0 4,8 - 37.9 21.5 8 -36 moss 8 68.3 55-80 28.3 12.8 - 35.9 27.0 13 -32 KC-2 herb 8 63.5 30-88 11.8 3.7-20.4 16.1 11 -22 moss 8 61.8 . 45-87 17.6 5.6 - 27.1 20.1 13 -32 FT KC-1 herb 8 54.1 31-82 13.2 1.3 - 29.1 13.6 4 -22 moss 8 70.5 50-80 22.8 9.3 - 35.0 24.6 10 -34 KC-2 herb 8 50.4 20 - 80 10.1 2.1 -14.7 16.4 11 -24 moss 8 58.8 38-77 12.7 1.8-31.2 18.4 9 -26 FTL KC-1 herb 8 55.5 30-80 8.3 3.3-19.6 12.8 9 -17 moss 8 55.0 35-80 22.8 12.2 - 34.2 21.4 12 -29 . KC-2 herb 9 62.9 50-77 13.8 5.5 - 25.9 14.0 6 -25 moss 9 53.3 48 - 62 11.5 6.3 -16.4 16.0 11 -29 a OC-open control, OM-moderate burn, OL-light burn, FC-forest thicket control, FT-thLnning only, FTL-thirming + light burn 152 Appendix 3.2 Severity of overstory thinning in a) forest openings and b) forest thickets for the various treatments at Knife Creek. Treat- Plot Basal area reduction (m2/ha) Number of trees cut ment a Site type N Mean Range Mean Range a) Forest Openings OC KC-1 herb 8 0.1 0-1.1 0.4 0-3 moss 8 0.0 0-0.2 0.1 0-1 KC-2 herb 8 0.7 0-1.0 0.4 0-2 moss 8 0.2 0-1.3 0.4 0-2 OL KC-1 herb 8 0.1 0-0.8 0.3 0-2 moss 8 0.7 0-3.0 1.9 0-6 KC-2 herb 8 0.1 0-0.9 0.6 0-4 moss 8 0.7 0-3.1 1.5 0-7 OM KC-1 herb 8 0.2 0-0.7 0.1 0-1 moss 8 0.1 0-0.6 0.1 0-1 KC-2 herb 9 0.4 0-3.3 0.2 0-1 moss 9 0.5 0-2.5 1.2 0-6 b) Forest Thickets • FC KC-1 herb 8 0.0 0-0.2 0.1 0-1 moss 8 0.1 0-0.7 0.1 0-1 KC-2 herb 8 0.0 0.0 _ moss 8 0.0 0.0 -FT KC-1 herb 8 2.2 0.8 - 3.5 6.9 3-15 moss 8 4.8 0.9 - 7.2 12.5 7-17 KC-2 herb 8 1 3.8 1.5- 6.4 11.1 7-18 moss 8 5.4 1.8 -10.6 12.1 5-16 FTL KC-1 herb 8 3.8 2.7- 6.9 10.3 7-16 moss 8 5.6 1.4-11.2 12.0 2-18 KC-2 herb 9 4.5 2.0 - 8.5 9.1 4-15 moss 9 4.0 2.4 - 9 1 10.7 4-24 a OC-open control, OM-moderate burn, OL-light burn, FC-forest thicket control, FT-thinning only, FTL-thinrung + light burn 153 Appendix 3.3 Spearman rank correlations between the independent variables used in the correlation and regression analyses in a) unburned plots and b) burned plots. The variable codes are explained in Table 3.2. DOB PAB TC VC MC TNS VNS MNS NTC BAC NT BA CC a) Unburned plots * PAB TC - -VC - - 0.01 MC - - 0.69 -0.64 TNS - - -0.21 0.50 -0.43 VNS - - -0.18 0.65 -0.51 0.80 MNS - - -0.03 -0.05 -0.05 -0.48 0.05 NTC _ _ -0.05 -0.09 0.03 -0.05 -0.07 0.06 BAC - - -0.07 -0.13 0.05 -0.04 -0.08 0.08 0.99 NT _ -0.09 -0.39 0.19 -0.38 -0.44 0.01 0.28 0.27 BA - - 0.02 -0.43 0.21 -0.40 -0.45 -0.13 0.01 0.03 0.63 CC - - -0.05 -0.29 0.14 -0.27 -0.34 -0.06 0.19 0.19 0.66 0.55 SITE - - -0.09 0.01 -0.12 -0.33 -0.18 -0.37 -0.02 -0.03 0.17 0.24 0.09 b) Burned plots ** PAB 0.67 TC -0.03 -0.17 VC 0.13 0.01 -0.13 MC -0.13 -0.13 0.79 -0.63 TNS 0.16 0.17 -0.13 0.36 -0.28 VNS -0.07 -0.09 -0.13 0.57 -0.42 0.76 MNS 0.06 0.09 0.10 -0.17 0.26 0.42 -0.02 NTC -0.42 -0.42 -0.01 -0.15 0.13 -0.19 -0.26 0.06 BAC -0.41 -0.42 -0.04 -0.17 0.13 -0.17 -0.23 0.07 0.96 NT -0.47 -0.38 0.01 -0.30 0.18 -0.21 -0.32 0.05 0.82 0.79 BA -0.22 0.02 0.03 -0.39 0.20 -0.16 -0.22 0.00 0.23 0.25 0.64 CC -0.36 -0.17 -0.14 -0.23 0.04 -0.21 -0.22 0.08 0.55 0.52 0.72 0.64 SITE 0.00 -0.04 -0.08 0.08 -0.19 -0.11 -0.05 -0.34 -0.05 -0.06 0.20 0.24 -0.03 for unburned plots n=96 and correlations with p>0.201 are significant at p<0.05 and p>0.264 are significant at p<0.001 for burned plots n=100 and correlations with p>0.197 are significant at p<0.05 and p>0.259 are significant at p<0.001. 154 Appendix 3.4 Burn severity statistics for the various treatments at Knife Creek. Depth of forest Percentage of forest Percentage of plot Treat- Plot floor burned (cm) floor burned area burned ment a Site type N Mean D Range c Mean D Range c Mean Range OM KC-1 herb 8 2.1 1.4-3.0 89.4 79.4 -100 99.6 97 -100 moss 8 3.6 2.4 - 4.9 88.4 77.6 -100 99.6 97 -100 KC-2 herb 9 1.7 1.0-4.0 76.9 52.5 -100 99.3 94 -100 moss 9 3.0 1.6-4.5 79.2 47.0 -100 100 -OL KC-1 herb 8 0.4 0.1-0.8 19.3 2.9-45.0 98.0 90 -100 moss 8 0.2 0-0.5 6.4 0-15.9 90.6 . 50 -100 KC-2 herb 8 0.5 0-1.2 14.0 0-31.0 96.3 85 -100 moss 8 0.5 0-1.1 9.8 0 - 20.6 86.3 70 -100 FTL KC-1 herb 8 0.5 0.1-1.0 12.8 3.2 - 23.1 91.3 80 -100 moss 8 0.3 0 -1.1 7.5 0-31.0 84.9 80- 98 KC-2 herb 9 0.3 0-0.6 12.1 0 - 32.5 93.0 50 -100 moss 9 0.3 0-0.7 13.0 0-38.1 85.9 30 -100 a OM-moderate burn in openings, OL-light burn in openings, FTL-thinning + light burn in thickets D Mean of plot means, each plot mean was calculated from up to six depth of burn pins. c Range of plot means 155 Appendix 3.5 Residuals from the equations vs. predicted normalized ranks of cover and number of persister species, as well as normalized ranks of cover and number of invaders for June 1996, July 1996, and July 1997. C o v e r of Pers i s t ing S p e c i e s -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 Predicted cover of persisting species July 1996 •1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 Predicted cover of persisting species -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 Predicted cover of persisting species 2.0 2.5 Treatment l mTL o mT mOC o mM • mL 0 mFC 0 hTL V hT A hOC + hM X hl_ o hFC 156 Number of Pers i s t ing S p e c i e s -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2,5 Predicted number of persisting species July 1996 1.8 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 Predicted number of persisting species Treatment I mTL o mT mOC o mM • mL mFC < hTL V hT A hOC + hM X hL o hFC July 1997 1.8 1.2 g 0.6 1 o.o cu * -0.6 -1.2 -1.8 o o < 3 0 < I A v 3*L a „ v / o • 0|_ Q- i .j^y ^ •<,«&> o* ^  % r A ^ v o + < + + o _L -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 Predicted number of persisting species 157 C o v e r of Invading S p e c i e s June 1996 -2.4 -1.2 -0.9 -0.6 -0.3 0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 Pred ic ted cover of invading spec i es July 1996 -1.2 -0.9 -0.6 -0.3 0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 Pred ic ted cover of invading spec i es July 1997 -1.2 -0.9 -0.6 -0.3 0.0 0.3 0.6 0.9 1.2 1.5 1.8 2.1 2.4 Pred ic ted cover of invading spec ies Treatment i mTL o mT a- mOC o mM • mL > mFC < hTL v hT A hOC + hM x hL o hFC 158 Number of Invading S p e c i e s June 1996 ZJ •g to CD DC July 1996 -1.0 -0.5 0.0 0.5 1.0 Predicted number of invading species -1.0 -0.5 0.0, 0.5 1.0 Predicted number of invading species 1.5 1.5 Treatment i mTL o mT •ft- mOC o mM • mL > mFC .< hTL v hT A hOC + hM x hL o hFC July 1997 -1.0 -0.5 0.0 0.5 1.0 Predicted number of invading species 1.5 159 Appendix 3.6 Normal probability plots and results of the Kolmogorov-Smirnov normality test for the residuals of the equations predicting a) cover of persisting species in June 1996, the number of residual species in b) June 1996 and c) July 1996, and e) the number of invading species in June 1996. b) Number of persisters in J une 1996. p = 0.027 -1 0 1 Residual 160 c) Number of persisters in July 1996. p < 0.01 -1 0 1 R e s i d u a l 161 Appendix 3.7 Influence of a) basal area cut, b) pre-treatment canopy cover, and c) site on the predicted normalized ranks of cover of persisters in June 1996 (left side) and July 1997 (right side). June1996 2.4 te o Xi CD 2 4 6 8 10 Basal area cut (m /ha) 12 1.8 1.2 0.6 0.0 -0.6 - 1 . 2 * -1.8 July 1997 ~~i i r o o « , o ° o ° o o fe Treatment • burned o unburned 0 2 4 6 8 10 Basal area cut (m /ha) 12 b) CC 2.4 | 1.8 (Q "§ 1.2 N "c5 £ 0.6 o 0.0 -0.6 -1.2 CD > o o XI & o XI CD CL -1.8 o o o o o o o o o o 0o o o o o ca ( o • 8 • „ 0 o 88 "<P 8 ° • ° o 0 0 O o° _.. • •: • « . •.. • : ! ! ! • - J • • • * i i _ ! i i_ C) 0 20 40 60 80 Overstory canopy cover (%) SITE CD § O T3 CD •s CD 100 2.4 1.8 1.2 0.6 0.0 -0.6 h -1.2 -1.8 o o oo t o o° ° g ° 8 o E T o o * o c?oo *. . . . . t. Treatment • burned o unburned 0 20 40 60 80 100 Overstory canopy cover (%) CD > o o XI CD -*-» o X) CD 2.4 1.8 1.2 0.6 0.0 -0.6 -1.2 -1.8 I I 88o ooo ooo IL ooo - ooo° 8888 oo • i o • - ooo tJ** oo 88 • • • -tuts - lit. I ••• « c CO l _ XI CD N "ra E o c CD > o o X J CD CJ XI 9> KC-1 KC-2 Site 2.4 1.8 1.2 0.6 0.0 -0.6 -1.2 -1.8 -I oo I --88 o 888 ooo --ooooo CO ooo ooooo -88 *•• ooo ss°. oo * -oo It. oo ttt. -«•" I I * • Treatment • burned o unburned KC-1 KC-2 Site 162 Appendix 3.8 Influence of a) basal area cut, b) pre-treatment canopy cover, and c) plot area burned on the predicted normalized ranks of the number of persisting species in June 1996 (left side) and July 1997 (right side). a) BAC June1996 July 1997 "1 Treatment A KC-2 burned A KC-2 unburned • KC-1 burned o KC-1 unburned Treatment • burned H o unburned 100 Treatment • burned o unburned 20 40 60 80 Plot area burned (%) 20 40 60 80 Plot area burned (%) 100 163 Appendix 3.9 Influence of a)pre-treatment canopy cover on the predicted normalized ranks of the cover of invaders and b) number of trees surrounding the plot on the predicted normalized ranks number of invading species in June 1996 (left side) and July 1997 (right side). a) CC June1996 July 1997 2.4 1.8 .3 1.2 0.6 8* 0.0 -0.6 -1.2 o « o o o § o n o e * • • • 2.4 1.8 T3 H N 1.2 P 0 20 40 60 80 Overstory canopy cover (%) 0.6 f o ! 0) s o.o -0.6 -1.2 n ^ O p * O o n • «J S"0 O • . ° 0 0 o § • * ° 5 a g a g • • „• • •• J L Treatment • burned o unburned 100 0 20 40 60 80 Overstory canopy cover (%) 100 a) NT 1.5 1.0 0.5 in q> o a CL in 0.0 o T J CD -0.5 -1.0 -1.5 ~i 1 1 1 1 r A A • A A A A A & A A A 0 A A A A A A o o OQ O O o o , • • • 1 . 4 . * o . i A A * * * * J I I L O O O o o 9 o oo A O n J L 5 10 15 20 25 30 35 Number of trees surrounding plot 40 1.5 « 1 .oh (0 E in <j) o <D D . in 0.5 0.0 -0.5 E 3 C T J I T J -1.0 -1.5 AA. A A A A A A 0® 6 S A A A A A A A A A . " A A 0 ° § 0 § S A A 4 o o o ° A A A O . A O 8 „ ° A A * A 4 A A A » O T A A A 6 . O A * A A A A OO A A 4 * A O . . J r» * • 0 O A AV*. 4» it A . « • °0°° J I I L Treatment ^ KC-2 burned A KC-2 unburne • KC-1 burned o KC-1 unburne 5 10 15 20 25 30 35 Number of trees surrounding plot 40 164 "3 0) NO C N CD T3 CO cu is co -t-» O 'a, u O) u CN ON o oo q r H T f CO ON CO ON r H 13.6 r H ON CO r H t-i r H r H d d d CO d T f d 13.6 d I N d O NO ON Tt< r H CO CO oo T f r H LO r H 00 LO T f r H d d d d CN oi d LN d ON d d FT] Jn96 FT] Jn96 r H d r H d r H d r H d NO d ON d oo r-i r H d O ON CN d CO d NO r H r H NO NO CO ON oq ON CO 14.1 r H LN T f r H P-, r - i CN d d d d r H r H d 14.1 d T f d d tN ON LO LO T f 00 12.1 r H O r H LN CO r H CO CO r H r H d d r - i 12.1 d CO d O CM d d 00 r-i NO ON 00 r H r H CN CN CO ON r H NO CO CO r H FT If 96 d r H CO d CO d CO r H CN 00 T f r H r H NO d r H 11.6 13 d r H r H I N O O oi r H r H d r H d d r-i NO d r - i d 11.6 13 d d L N r - i d in oq r H CO ON r H NO r H CO CO CO q T f cu d od r H d d d NO r H d OQ CN d L N r-i d tN ON r H r H r H CN T f r H CO CN NO r H CN r H ON 00 d d CN LN CO d ?i d d ON T f d NO ON o r H r H r H co T f r H 15.8 18.1 17.2 T f r H NO CO o FC Pre Jn96 JI 5.8 9.1 8. r H d 0.1 0.1 0. 1.6 1.9 2. 0.1 0. 5.4 5.4 5. 2.8 1.4 1. 0.1 0.1 0. 15.8 18.1 17.2 0.3 0.4 0. 0.1 0. 5.3 8.1 7. 2.4 2.8 4. 0.6 0.6 2. tN ON q r H T f CO r H NO LO T f CO T f r H r H r H d d d d CN d d in CN d d NO ON CO r H r H CN r H r H CJN CO r H 2 ci d d d d d d T f d o Jn96 r H d r H d r H d CN O LO r H a O r H T f NO CO oq r H 14.1 co ON T f L N CO CH CN d r H d CN r H NO 14.1 d d T f d d tN ON q r H CN NO r H T f -* r H NO NO CN r H r H d d NO oi T f r t rH d d 00 r-i r - i NO ON T f r H r H CO r H T f T f ON NO r H LO CO NO L N OL d d d d d CO r H d NO d d NO d d OL Jn96 00 d CO d NO d NO 1-i r H d CO d CO NO r H d CN d CN r H r H q CO ON CO IN CN ON C O co q CH d d d CO r i CN d CO rH rH d d CN d r H IN ON T f r H T f ON Q\ o NO ON CN CO r H 12.5 CO CO CN d CO d d CO LN d LN d d 12.5 NO oi d NO ON NO NO CO 00 T f 00 T f 00 17.6 T f r H T f o L N u d CN d d cd LO d CO 17.6 d oci NO CO CN 0 NO ON r H NO CO NO ON ON CO oo CO ON NO r H oq JH d CN d d CN CO d T f LO rH d CN L N CN r H 01 JH NO CN r H q ON o CO ON 12.0 CO CN CO NO LO c d - CN d r - i d CO d CO 12.0 d T f CO r-i r H TREE SEEDLINGS Betula papyrifera Pz'nus contorta Populus tremuloides Pseudotsuga menziesii Sa/ix sp. SHRUBS Amelanchier alnifolia Ardostaphylos uva-ursi Linnaea borealis Mahonia aquifolium Rosa adcularis Shepherdia canadensis Spirea betulifolia Vacdnium caespitosum TOTAL WOODY | PLANTS | FORBS Achillea millefolium Allium cernuum Antennaria negleda Arnica cordifolia Aster dliolatus Aster conspicuus 165 JI 97 VO d r H r H vq r - i Ov d d 10.1 r H d Ol d Ov d d 24.5 26.4 r H d CO d FTL Pre Jn96 JI 96 0.1 0.1 0.3 0.8 1.2 1.4 1.9 0.8 1.1 0.4 1.1 1.4 r H d 0.1 0.1 0.1 0.3 2.2 1.5 4.8 0.1 0.1 0.1 0.1 0.1 r H d 0.4 0.4 1.0 0.4 0.1 0.3 11.9 14.8 21.8 37.9 14.6 24.1 0.4 0.1 JI 97 r H oi oo in ro d CO d r H d r - i r H d d d VO d 21.3 18.1 r H d r H d f—' t t . Pre Jn96 JI 96 0.8 1.0 2.0 2.8 5.3 8.0 0.4 0.5 0.5 0.1 0.2 0.1 0.1 0.1 r H d 0.8 0.8 1.9 0.1 0.1 0.1 0.5 0.4 0.6 0.2 0.2 0.5 0.3 0.8 0.8 14.8 18.3 24.4 28.8 11.1 25.0 0.3 0.5 0.4 0.4 0.6 2.6 JI 97 r H d ON CO 00 d CO d r H d 00 d in d d 21.8 25.4 oi U vo ON d r H d CO vo d , r H d r H d Ov d r H d vo d d 21.0 24.1 r H r - i m oi Pu Pre Jn96 0.3 0.3 0.1 0.1 2.3 2.6 0.5 0.4 0.1 0.1 0.4 0.4 r H d 0.9 0.6 ol d 0.3 0.6 13.5 17.1 31.5 13.1 0.1 0.1 2.0 1.1 JI 97 ON r - i r H r i CO d ol d CO r - i r H d in d r H d d r H d in 00 r H S 96 If o\ d in r - i Ol d Ol d Ol d r H d CO d CO d r H d ol 00 CN oi o Pre Jn96 r H d 0.1 0.1 3.8 0.1 0.7 0.1 r H d r H d 2.8 0.3 r H d r H d ol d 0.2 0.3 Ov d 15.3 2.6 27.3 0.8 CN d in oi JI 97 H r H r H oi d r H d p r - i r H d co oo r H d r H d r H d p r - i CO d 26.2 22.3 r H d 96 If r H d Oj r - i in r - i CO d CO d VO d vo r H d r H d r H d r H d CO d Ov r - i r H d 19.4 28.4 r H d r H d 0 Jn96 IN d q r - i CO d ol d •* d vq r - i r H d r H d r H d r H d CN d 12.1 13.0 r H d Pre r H d d IN r - i co d Ov d CO d r H d " * oi r H d r H d r H d r H d r H d d 12.1 26.5 vq r H CO d JI 97 Ov d OV oo d d r H d r H d o oo d d r H d Ol r H r H r - i 40.6 27.1 Ol d CO r H oc 96 If CO r H d OV vd IN d d r H d o vd r H d co d r H d vo d r H d T - i r H 40.5 34.3 Ov d m d CO d oc Jn96 VO d CO d Ov m' vo d r H d r H d VO oi r H d r H d vo d r H d Ov d OV d 27.4 20.9 CO d CO d CO d Pre r H d d d CO CO vo d r H d in oi r H d r H d vo d in d m d 19.8 28.8 ol d r H d IN d Astragalus miser Castilleja miniata Cirsium sp. Clematis columbiana Coeloglossum viride Epilobium angustifolium Fragaria virginiana Galium boreale Gentianella amarella Geranium bicknellii Geranium viscosissimum Goodyera oblongifolia Hieracium umbellatum Lathyrus ochroleucus Lilium columbianum Melampyrum lineare Moehringia laterifolia Orthilia secunda Pyrola chlorantha Solidago spathulata Taraxacum officinale Tragopogon dubius Trifolium repens Vj'a'a americana Vio/fl adunca TOTAL FORBS GRAMINOIDS Calamagrostis rubescens Carex concinnoides Carex rossii Festuca sp. Oryzopsis asperifolia 166 JI 97 d 26.9 CO d r H d r H d r H d in d FTL Pre Jn96 JI 96 0.1 0.3 38.3 14.8 24.4 in d r H d r H d r H r H d r H d 5.0 0 0 JI 97 22.7 CO r - i r H d 0 0 r - i r H d r H d m 0 0 r H d r H d r H d r H d r H d 12.1 r H d r H d H H-Pre Jn96 JI 96 29.5 12.2 28.1 0.5 0.7 0.8 1.3 1.1 0.9 0.3 0.1 0.1 7.7 6.7 6.3 r H d 0.1 0.1 0.1 r H d 0.9 0.1 r H d 0.1 0.3 0.1 r H d 10.8 8.9 8.6 0.1 0.1 0.1 0.1 JI 97 27.8 d CO oi r H d co CO r H d f r H d 11.8 r H d 0 0 d r H d r H d r H d 18.6 r H d U Pre Jn96 JI 96 33.6 14.4 27.8 r H d 2.2 2.3 2.9 0.1 0.1 1.9 2.6 2.8 r H d 0.1 0.2 0.1 0.1 0.1 .0.1 5.6 5.7 5.6 r H d 0.1 0.1 0.1 0.1 0.2 0.3 0.1 0.1 0.1 0.1 0.1 0.1 0.1 10.1 11.6 12.2 0.1 0.1 0.1 JI 97 CO d CO ** Ov oi 13.5 0 0 r H 0 0 CO 21.9 OM Pre Jn96 JI 96 r H d 30.4 0.8 2.8 r H d r H r H tN r - i vq r H 0.1 2.1 in d 3.4 0.1 3.6 r H d r H d JI 97 22.3 oo d r H d r H d r H d r H d r H OL Pre Jn96 JI 96 28.4 13.1 28.6 0.3 0.1 0.1 0 0 oi r H d r H d Ov in Ol d 9.5 0.1 0.1 r H d 0.1 0.1 JI 97 28.6 vq r H CO r H CO t< r H d r H d U JI 96 36.0 Ov d r H r - i in CO r H d in L O r H d r H d o Pre Jn96 29.7 21.7 0.3 0.7 0.5 1.6 3.2 5.8 4.0 8.1 r H d r H d Phleum pratense Poa pratensis TOTAL GRAMINOIDS MOSSES Aulacomnium palustre Bfachythecium hylotapetum Brachytheaum sp. Ceratodon purpureus Dicranum fuscescens Dicranum polysetum Dicranum scoparium Dicranum tauricum Eurynchium pulchellum Funaria hygrometrica Hylocomium splendens Mnium spinulosum Pleurozium schreberi Pohlia nutans Polytrichum juniperinum Ptilium crista-castrensis Rhytidiadelphus triquetrus Sanionia uncinata LIVERWORTS Barbilophozia hatdieri Lophozia sp. Marchantia polymorpha Ptilidium ciliare TOTAL BRYOPHYTES LICHENS Cetraria ericetorum Cladina mitis 167 FTL Pre Jn96 JI 96 JI 97 r H d CN d CN d IO d r H d r H d r H d 0.2 0.1 0.1 0.1 CN d CN d CN d 2.1 0 0 0.1 f-B . I N ON NO ON NO ON j : !H PH r H d r H d r H d r H d CO d r H d r H d r H d r H d CO d CO d CO d CN d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d T f CN oq r H NO r H oq r H T f d T f d co d T f d L N d T f d T f d T f d 0 0 d T f d r H d r H d r H d r H d r H d r H d r H d r H d r H d Tf Lfj •<* CO CO CN ro FC t N NO ON NO ON JH PH d r H d CN d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d CN d CO d CM d r H d t N d T f d LO d CO d T f d T f d r H d r H d CN d r H d r H d r H d CO CN r H CN CO CN CO r H r H d r H d r H d r H d r H d -* T# o T* LO CO Tf oi OM [ N ON NO ON NO PH T-H d r H d r H d r H d r H d T f d ON d 0 0 CO r H d o o o LN LO OL t N ON NO ON NO ON PH r H d CN d r H ' d CO d r H d r H d r H d r H d CN d r H CN oo d NO CN CN d o rH O O LN OC t N ON NO ON NO ON E PH r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d Tt< d CN d CN d CO d r H d r H d CN d r H d r H d r H d r H d r H d CN d CN d CO d r H d r H d r H d r H d LO d LO d CO d CO d CO d NO d LO d r H d CO r H 0 0 d r H r H 0 0 d r H d r H d r H d r H d T f d CO d r H d CO d NO CO CO CO rH CO T* CN C « K a .K *S ts Q C <tf > .tS ts s '11 0 .§ tu o •2 "S o 1 3 a 0 '5 8 1 •2 "5 o ts D tS •n O o ? ,ts "S o a U es •S s .ts "5 o ts D .Ul •"3 « •2 "5 o ts <s 0 + m 1 § •3 t D U S "3 S '5 O ts ts U ts o ? o •5 o •2 '5 o ts <3 0 U + ts ts S 3 •s * ? ri •s > -2 4 3 "S G •3 « D S ts o f *Si • n -2 '5 o ts ts 0 ' -S •2 '5 a ts ts U « 13 3 Ul •2 "5 o ts ts 0 f i PH m •2 "5 o ts ts 0 ts tft O HS §-ts •SP )^ O H \ ts § « HS .2 -a •J? g ^S ts •5 5 S 1 •S? CH •2 tu t 8 3 tu tS fe. CK) ^ l ts 8 ts C ts *H s ts fe CH "£ IS tu K ts fe £ ^ 1 O H .JS JS S-i tu tu ts RL ts fe •2? O H Ul Ul 3 ^ »H ts fe ^ i O H 1X1 Z w X U HH < H 0 H 1 68 L N N O L O o O N o r l O N r H N O C O o\ d r H L O d oo' d C O d r i d N O C O O N N O C O rH r H C N r l L O r l Tf Tf r H C N J [ , o\ d d C N d «* d N O d d d d r i d d tr U H Pre Jn96 C N C O r H d 2.4 0.1 0.5 0.1 3.1 0.2 1.1 0.2 15.4 0.6 0.1 0.1 3.4 5.4 . 0.1 0.1 r H d 0.3 0.3 1.3 0.4 0.1 0.1 r l d L N r H oo r l N O N O rH r l C O O N C O 0 0 C N O N d C N C N d d N O d N O d d d d N O C O C O 0 0 L O N O Tf r l o L O r H C O N O C O O N d r H r i d d Tf d Tf d d d d d f—' ^ P H N O O N C r H d C O r i N O d C O d N O d 00 CN r l d O N Tf C O d r l d co d Tf d C O d Pre 0 0 N O C O Tf O N O N r l o Tf r l r l Tf C N Pre L N r H d d d d O N rH d C N d d d d d L N r H Tf O r l C N C N O r l r H o O N N O r l N O r l O N d d r i d r i d CO d d C N d d d d d N O r H oo r H 0 0 C N r l Tf C O N O r l r l N O L O N O r l Tf r l O N d d d d d d d d CN d d r i d d d d d U P H N O r H Tf r H L O r H r l N O C O rH r l r H 0 0 C O Tf r l Tf r l O N c d d d d d d d d CN d d r i d d d d d Pre r H N O Tf oo r l L O r H N O r l r l L N C N Tf r l C O r l Pre d d d d d d d CN d d d d d d d d L N C N r H Tf C O r H r l N O r l N O C N r l Tf C N Tf C N O N r i d d d d r-i r i d Tf Tf d d d d d N O Tf C N r l r H O N O N Tf r l Tf O N d d d d d C O d d d s 0 N O o\ c r H d r H d rH d L N d r l d r H d Pre o C O N O Tf L O N O C N o r l r H N O Tf N O Tf C N Pre Tf d r H C O C N r H d Tf rH d d Tf d d r i d L N C N r H C O O N Tf r l O N C N C N O N r l I N r l Tf r l O N d d r H L N Tf d CO rH d d Tf d d r i d d N O r H Tf N O C N CO C N r l r l C N N O L N Tf N O O N d d C O C N NO- d d Tf d d d d d 1—1 ~^ o N O O N c r H d C O d r l d N O d C N d C O oi r l d r H d Tf d C O d C O d Pre Tf C O r H Tf r l Tf oo r l L O C N r l O N N O Tf 0 0 C O Pre d r H d d C N C O C N d d rH d d r i d d. d d L N r H oo r l O N C O L O rH C O 0 0 r l C O Tf L N C O O N d C O C N C O r l C O d Tf CN d N O r i oi d r i d N O C N to O N C O o O N C O r l r H O O N o r l r l L N C O u O N d C N r-i C N r l C O d d d L N d C N d d r i d 0 N O O N C r H d o C N N O r H oo L O Tf C N r H r-i CO rH r l d r l d C O N O Tf d L O d r l d q r i C O d Pre ; r H N O O N Tf Tf oo rH r l C O L O O N O N C N Pre ; d C N d T)i r i d d rH d C O d d d d TREE SEEDLINGS Betula papyrifera Pinus contorta Populus tremuloides Pseudotsuga menziesii Sah'x sp. SHRUBS Amelanchier alnifolia Ardostaphylos uva-ursi Linnaea borealis Rosa adcularis Shepherdia canadensis Spirea betulifolia Symphoricarpos albus Vacdnium caespitosum 1 TOTAL WOODY 1 | PLANTS 1 FORBS Achillea millefolium Antennaria negleda Arnica cordifolia Aster dliolatus Aster conspicuus Cirsium sp. Clematis columbiana Disporum trachycarpum Epilobium angusHfolium Fragaria virginiana Galium boreale Geranium bicknellii 169 JI 97 r H d NO r - i d r H d r H d rH 0 0 11.6 d 12.0 r H d O l d LN d LO o i FTL NO ON NO r H CO d r H d r H d r H d ON o" rH CO oi r H r H d -* CN rH r H d ON d FTL NO ON j : CD u OH H d r H d o o d NO d CO d r H d r H d r H d r H d r H d r H d CN IN LO NO 00 CO o CN CO d r H d NO d 0 0 CO ON IN NO r - i r H d r H d 00 CO 00 d 00 NO IN ON 00 d CN d r H d CO d r H d CO d d NO d rH IN CN r H d q r H oo oo' NO r - i - * d r H d T f H u. NO ON NO ON d d o o d r H d r H d CO d r H d r H d r H d r H d r H d CO d CO d NO IN NO LN (N NO o CO r H d CN d d CN d CN NO' r f CO r H T-i • * r - i r H d ro d r H d o CN IN OO LN LN OH CO d r H d r H d r H d r H d r H d r H d CO d tn •^ i o LO r H d CO d LO r H d CO r - i r H d r H d 00 K CN ON r H d r H d rH d r H d r H d r H d 0 0 NO CO d CN NO CN r - i LO ON CO LO NO FC NO ON NO ON o. r H d r H d r H d r H d r H d r H d ro d CO d r H d r H d r H d r H d •«* NO CO r f CN 00 LO ON IN 00 CO r H d r H d r H d CO d CN d CN d rH NO rH ro rH rt< r H d ON r H r - i r H o 00 r H d r H NO o i r H r H d r H d 00 NO NO 00 ON NO NO NO NO IN ON CO d ON d CO d oo NO r H NO rH VO 00 r H CO r H OM NO ON NO ON _C HI u Cu d CO d r H d r H d r H d 00 r - i r H d r H d r H d CO d r H d r H d r H d •* LO o r-i O O rH NO CN 00 d o q r - i r H CN d NO oi 0 0 d CN rH 00 d r H d o i r H d LO CN CN IN ON ON r - i CO d r H d r H d r H d rH d rH LO r H d NO LO r H d 00 T-i NO ON rH d CO r - i CN d r H d CO d IN oo NO CN r H d NO LN CO d OL NO ON a OH r H d d CN d CN d r H r H r H d CN d r H d r H d CN d r H d NO rH to ' CN CN LO CO d LO d CN oi ro LO O l T-i CO o i r H CO d o o o o NO IN ON r H d CO r - i d r H d r H d CO d rH LO rH ON o i r H CO d r - i NO s LO d O l ro o o id o d [N OC NO' ON NO ON r H d r H d CO r - i CO d LO d LO d r H d r H d r H d r H d r H d CO d CN d NO rH ON ON LO r H r H NO LO co d CO d CO r - i NO d o CO rH LO NO r H d r H d r H o i o l o i r H d r H LO ON CO r H d o o si L0 NO 2 OH r H d r H d d CN d r H d r H d r H d CN d o IN o o o CO d q r - i ro ON ^ p o i o o o o d CN £ 3 s _Vi Vi 8 .CO £ . 3 5 u CS SS o o CS fe Q O u £ 3 S " 3 £ 3 s 3 "Q cs fe iE 8 •s o to 3 ^ J S cs •-J £ 3 K .cs s 3 8 £ . 3 '•3 cu cs CU • 5 s 3 a. £ 5 SS 8 CU CO •S O "5 JS fe J S .N i s s Vi o ca cs 8 o <s fe J S SS «s C3 s; CS J S SS CS o ? s "S v. "« SS £ 8 CS X CS E2 to § £ . 3 ft CS 5S a fe £ CS CS " 0 ' £ 3 c •S CS CS "S P-H CO -a 0 to S 0 PH < H 0 H c/) D 0 Z HH 0 to £o tu J i 3 to "43 to fe <S s tS 3 to o R C "Q SS 8 K CS U .« i l fe sx to CS .to to § -o C/l Q HH 0 z HH 0 < H 0 CO w CO CO 0 s Cu CO s 3 "Q cu HS •s CS CQ to g e-3 a. SS o •a CS fe U to to 8 v i s 3 SS <s b Q £ •3 tu to ,>> "S SA. s 3 SS cs b 5 •2 CS^  to £ 3 SS CS b 5 £ _ 3 ^) £ 3 SS ? 3 s cu £ o fe S) J ; .cs CS SS to § "CS § "a. to £ 3 £ 8 o £ 3 to .O * 3 , S "a. to £ •2 '5 fe •8 r-. VI £ •2 O g 170 FTL Pre Jn96 JI 96 JI 97 T-H d 1.8 0.3 0.3 0.6 75.7 1.1 1.3 4.1 T-H d T-H d r H d CN d 2.1 0.1 0.1 0.1 NO d T-H d FT Pre Jn96 JI 96 JI 97 0.1 0.5 0.2 0.1 77.3 79.8 78.6 81.4 0.1 0.1 0.1 0.1 0.1 T-H d T-H d 0.1 0.1 0.1 0.2 0.1 0.1 0.1 0.1 0.1 0.2 0.1 0.2 r H d 0.1 0.1 0.1 0.3 0.3 0.4 0.1 1.3 1.4 1.3 1.8 0.1 0.2 0.1 0.1 0.1 0.1 FC Pre Jn96 JI 96 JI 97 0.1 0.1 0.1 0.1 d 0.1 0.1 0.1 3.1 3.1 2.5 1.9 84.0 84.6 79.4 78.4 T-H d 0.1 0.1 0.1 r H d T-H d 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.2 0.2 0.1 0.1 0.1 0.2 0.1 r H d 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.2 0.1 0.5 0.6 0.8 0.4 0.1 0.3 0.1 0.1 0.5 0.6 0.6 0.9 OM Pre Jn96 JI 96 JI 97 0.1 1.3 0.9 CN d 0.5 7.3 79.9 0.1 2.6 24.4 T~< d T-H d T-H d T-H d r H d CO d T-H d CO d CN d r H T-i CO d ON d OL Pre Jn96 JI 96 JI 97 CO d VI 83.5 0.3 0.3 1.9 T-H d r H d T-H d r H d T-H d r H d LO TH CN d oq T-i OC Pre Jn96 JI 96 JI 97 -6.9 1.6 4.1 3.5 1.5 1.7 2.3 1.9 0.1 0.1 CO d T-H d r H d 89.9 86.0 88.7 85.2 0.1 0.1 0.1 r H d 0.3 0.2 0.1 0.1 T-H d 0.1 0.1 0.1 0.1 0.1 0.1 0.8 1.2 1.1 1.5 0.1 0.1 0.2 0.4 0.4 0.6 Pohlia nutans Polytrichum juniperinum Ptilium crista-castrensis Rhytidiadelphus triquetrus Sanionia uncinata LIVERWORTS Barbilophozia hatcheri Lophozia sp. Marchantia polymorpha Ptilidium ciliare TOTAL BRYOPHYTES LICHENS Cetraria ericetorum Cladina mitis Cladina rangiferina Cladonia bacillaris Cladonia hotrytes Cladonia cervicornis Cladonia chlorophaea Cladonia fimbriata Cladonia gracilis Cladonia multiformis Cladonia ochrochlora Cladonia ochrochlora + C. macillenta var. bacillaris Cladonia subulata Cladonia spp. Peltigera aphthosa Peltigera canina Peltigera leucophlebia Peltigera malacea 171 CN T-I ON d 1 — • NO TH ON d FT Jn96 TH d Si en cu IN. NO ON '—> NO ON oi ' ' tu NO »t ON tN T-I d ri CU CN tn ON tN 1 ' NO ON en u ' ' Cu- NO TH ON oi CN Si r-i CU CN o ON 1—> NO o ON S !=. o Jn96 d Si 00 en CU I N o ON ' — ' NO o ON > — ' 0 Jn96 o OJ SH ON en Cu CN ON oi 1—> NO ON ON T-i u 0 Jn96 ON T-i Si r(l T-i Cu to Z *TH w £ IS IC cu IC K h-1 I5 -+H OTA Cu CO HUJ O "EH •s 0) CN i U rH rH CN Ol CO CN rH rH o . oo Ol Ol rH d d CO d d oi rH NO TH d d oi d d d d d T-H Ol rH ol NO TH rH L O NO rH d 0.2 1. 0.1 0. 0.1 0. 2.8 7. 3.2 9. 0.1 0. 0.3 0. 0.4 0. 0.3 0. d rH CO Ol Ol NO oo Ol rH Ol Ol Ol CO rH ol d CO CO T-i T-i TH d d CO d d d d Th 00 q o NO TH Ol rH NO ON NO CO T-i tN d d r-i CN r i IN rH d d d rH oi CO d T-H CO CO I N ON CO LO ol rH rH 00 00 00 rH LO T-i tN d d d NO T-i NO rH d d d d r i oi oi d ** 00 CO NO ON LO TH Ol Ol rH CO ON LO CO T-i d d d LO d oi TH d d d d oi r-i co d rH CO ol CO ON CO o Ol rH LO LO Ol o oi CO d d d CO r-i ON d d d r-i oi oi d T-H NO rH ^fi rH rH LO NO 00 ol rH rH 00 co Ol NO d d T-i d d d oi rH oi T - i CM d d oi d oi NO rH d T-I NO NO rH rH Ol o oq oo CO o LO LO Ol d d d d d d tN oi rH LN TH d oi oi rH CN rH d T-H NO LO rH T-H LO oq CO ro 00 •NO ON NO q CO d d d d d d CN rH T-i oi rH d d oi d T-i d T-H LO 00 NO rH ON oo O vo CO rH co CO IN rH LO d d d d d 00 oi oi LfS T-I d d rH oi oi rH d rH Ol rH rH o NO T-I rH rH Ol Ol oo NO d d d d oi Tti IN d d d d d oi d rH rH Ol ON CO rH rH rH rH CN ro d d d T-i oi d d d d d d d rH rH Ol en rH rH rH rH rH rH d d d d d d d d d d ON q ON CO oo en CN q CO to. CN ro rH d r-i T-i d rH NO oi TH d rH oi r-i LO oi r-i CO rH 00 rH 00 d rH d CO CO III rH d 18.4 rH d Ol d rH d co d CN d o LO rH CO CO CO CN O rH NO CO rH LO ro rH d 0.1 0. 0.3 0. rH 3.5 7. 0.1 0. rH d 4.0 9. 0.3 0. 0.2 0. 0.3 0. 0.2 0. 1.4 3. 0.7 2. rH rH Ol rH CO NO CO NO LO CO ON CO oq CO ro rH oi d oo od d Ol d d d T-i T-i r-i d Ol rH CO CO rH NO NO CN rH rH O LO LO NO LO CO oi d oo LO ol d d d CO CO oi ON rH rH T-H CN rH CO LO en oo rH o CO NO NO CO CO oi d d NO d d d oi CO oi T-i NO rH Ol CO Ol ON rH rH LO NO rH rH rH •<*! ON O CO oi oi d d LO d co LN TH d d d oi oi r-i oi rH NO NO rH oo ON NO rH ro rH rH LO CO CO NO oi r-i d d CO ON TH r-i d d rH rH T-i TREE SEEDLINGS Populus tremuloides Pseudotsuga menziesii Salix sp. SHRUBS Amelanchier alnifolia Ardostaphylos uva-ursi Linnaea borealis Mahonia aquifolium Prunus virginiana Rosa adcularis Spirea betulifolia •Symphoricarpos albus TOTAL WOODY 1 PLANTS | FORBS Achillea millefolium Allium cernuum Anemone multifida Antennaria microphylla Antennaria neglecta Arnica cordifolia Aster dliolatus Aster conspicuus Astragalus miser Castilleja miniata Cirsium vulgare 172 I N CN rH LO rH rH rH CN Tf T-H L N LO CO rH rH ON d rH T - i d d CN d d d T-i r-i ON d d rH NO I N I N T-H rH L N rH TH L N rH L N L N NO rH ON d d d d d d d d d d d LO L N d - J K> H P . NO LO T f rH TH TH L N rH TH CO TH CO Tf CO ON d d d d d d d d d d d d CO T f T H rH CN L N LO ON TH NO o VO NO O N CO u C H d d rH d d d d d T - i ON od d d I N CN oo L N CN CN TH ON LO L N rH ON NO CO CO ON d d d d d d d T-i d d d d Tf CN d rH TH 1—1 NO rH rH CN ON T H T H CN LO T f oo T H ON rH T f NO ON d d rH d d d d T H d d d d CO T - i d _ H rH rH E - ' ' NO rH NO 0 0 T-H T H CO ON CN ON T H T f NO ON CO T-H ON d d d d d d d d d d d CO CO d d c rH rH CO T f NO TH CO CO CN ON NO NO LO ON d d d d d d T-i d d d d od d 0. rH T-H c— rH ON NO ON T-H co ON NO T-H T H T H oo O LO NO rH ON d d d d d d d d d d d d ON LO d d , _ H rH rH 1—1 * NO rH T f oo T H TP L N NO T H T H rH L N Tf CO LO ON rH d d d d d d d d d d d T f d CN rH u 1 ' p . NO T f LO 0 0 T-H CN NO T f rH rH NO CN rH T f ON d d d d d d d d d d Tf rH d c rH rH —' ON NO 0 0 TH ON NO CO ON NO ON 0 0 d d d d d d d d L N d d c rH CN LN. CN rH rH rH T f T H TH oq CN CN T H CN rH rH ON d d d d d d d TH d CN d d d LO rH 1 ' NO rH rH T-H NO rH ON TP T H O CN ON d d d d T - i T-i d d NO CO "-'| o NO rH T H Tf TH T H TH CN LO ON d d d d d d r-i d rH L N LO L N L N CO T H T f ON rH L N NO NO NO TH Pn d CN T-i d CN d d d d d d NO rH T - i d L N NO ON CO Tf CO NO T-H T f ON oq NO CN ON d CN CN d T-i d d d d CO r-i d ^ H rH rH 1 ' NO rH CO T H CN 0 0 TH ON rH T f ON L N rH ON d CN T-i d d d d d T-H d CO NO d rH ' ' 0 NO CO oo CO TH 0 0 rH Tf T f T f LO CO rH ON d d T - i d d d d d d L N CO d CO ON ON T H T f ON T - H LO T H Tf CO T - H T f SH d rH T-H d d T-i T - i d r-i LO d L N rH CH rH rH L N rH rH L N ON T-H NO o TH T H T-H ON CO CO LN. rH ON d d rH T-i d d CO T-i d d d CO T - i d 1—' NO rH rH ON NO T H NO oo O N rH T H CO oo rH T-H CN rH rH ON d d rH T-i d d CN d d d T-H ON d rH CN d d u rH TH 1 ' 0 NO rH NO T H CO CO CN TH rH 0 0 O rH CN T f ON d r-i T-i d d CN T-i d d d L N d d rH C rH rH 1 ' rH rH O L N p CN CO T H T H T-H rH CO L N NO T H u d d CN T-H rH T f T - i d d T - i T-i CN CN d d C H CN rH £ S S »H cn Disporum trachycarpm Epilobium angustifoliu Epilobium sp. Erigeron/'Aster sp. Fragaria virginiana Galium boreale Gentianella amarella Geranium bicknellii Geranium viscosissimu Goodyera oblongifolia Hieracium umbellatum Lathyrus ochroleucus Melampyrum lineare Moehringia laterifolia Dracocephalum parvifli Orthilia secunda Osmorhiza berteroi Pyrola chlorantha Solidago spathulata Taraxacum officinale Tragopogon dubius Trifolium repens Vicia americana Viola adunca TOTAL FORBS GRAMINOIDS Bromus ciliatus Calamagrostis rubescer Carex condnnoides Carex rossii Elymus trachycaulus Festuca sp. 173 JI 97 CN c-i 10.8 rH d CO d TP d FTL Pre Jn96 JI 96 9.1 0.1 0.2 18.9 4.4 7.8 CO CN rH d rH d oo CN rH d TP d ON ON rH d LO d rH d JI 97 10.2 22.5! rH d ON TP CN rH CO CN CO d TP d 10.3 rH d r-i FT Jn96 J196 L1.3 12.7 21.5 25.1 4.6 4.9 0.1 0.2 2.1 2.4 0.9 1.4 rH d 0.1 0.3 0.1 0.1 0.1 0.1 6.1 6.1 0.1 0.1 0.7 1.3 rH d rH d 0.1 0.1 0.1 0.1 Jn96 J196 21.5 25.1 Pre TP ON 27.5 ON CN CO r-i NO CN rH d CN d rH d TP NO 00 d rH d JI 97 NO d 16.8 ON CN rH d rH d O TP rH d CO d rH d rH d 13.9 rH d rH d rH d rH d rH d rH d FC Jn96 J196 0.5 0.8 L2.0 15.6 1.4 2.9 0.1 0.1 2.9 4.4 0.4 0.4 0.1 0.3 0.1 0.1 6.4 10.4 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 Jn96 J196 Pre CO r-H 22.9 NO CN rH . d rH CO CN d rH d CO tx rH d rH d rH d tx. ON rH d rH in CO d rH d CO d TP d rH d rH d TP CN rH d CO r-i OM Jn96 J196 0.5 3.2 0.1 0.1 co d rH d 0.1 0.1 0.1 0.1 0.1 0.3 Jn96 J196 Pre oq r-i NO d 20.6 rH d tx. r-i CN d ON CN rH d rH d rH d CN LO CN d rH d rH d JI 97 rH rH rH d 12.9 CN d rH d CO d rH d CO d NO ON CO d rH tx rH d rH d OL Jn96 JI CN d in 00 rH d rH d Jn96 JI Pre o CO 21.8 in rH CO d O CN rH d tx d CO NO JI 97 rH CN TP d CO d 12.7 rH TP CO d ON CN CO d rH d tx NO LO d OC Jn96 J196 1.4 1.2 0.4 0.4 L3.5 15.1 1.1 1.3 0.1 0.1 0.2 0.3 1.9 2.6 rH d 0.3 0.1 rH d 0.1 0.1 4.6 5.8 0.2 0.2 0.1 0.1 0.8 0.3 Jn96 J196 Pre TP CN oo d 18.9 rH rH d rH d CO CN rH d TP TP rH d oo d rH d CO d Oryzopsis asperifolia Poa palustfis Stipa occidentalis TOTAL GRAMINOIDS MOSSES \Barbula convoluta + Pohlia nutans Brachytheaum leibergii Brachytheaum sp. Brachytheaum velutinum Bryum argenteum Bryum flaccidum s s >> K Ceratodon purpureus Dicranum fuscescens Dicranum polysetum Dicranum scoparium Dicranum tauricum Eurhynchium pulchellum Funaria hygrometrica Hylocomium splendens Mnium spinulosum Pleurozium schreberi Pohlia nutans Polytrichum juniperinum Ptilium crista-castrensis Rhytidiadelphus triquetrus Sanionia uncinata Schistidium heterophyllum Thuidium recognitum Tortula ruralis moss sp. 174 FTL Pre Jn96 JI 96 JI 97 d 16.4 0 0 0.8 r H d C O d C N d C O d r H d C O d r H d T f d C N d L N d T f d r l O u . L N C \ V O O N V O O N B P -r l d r H d r H d N O d CM L O VO rH CM • * CN rH r H d r H d r H d r H d r H d r H d r H d C N r H C N r H r H r H O N d C N d r H d r H d r H d T f d r H d C O d r H d r H d C O d C N d r H d r H d T f d C O d C O d C N d L O d T f d C O d r H d C N d r H d r H d O N d r H r H 0 0 d q r H O N d 0 0 d 0 0 d T f d o o d 0 0 d N O d r H r i T f d C N d r l d r l d FC L N O N V O O N V O O N QJ H. PH 0 0 d vo d L N d L O d O N T * O N rH N O CN rH rH s C O d C O d Tt< d T f d N O d N O d L O d N O d C O d C N d C N d C O d r H d r H d r H d r H d r H d r H d r H d r H d C N d C N d C O d r H d r H d C N d C O d C O d T f d C N d r H d C N d V O d T f d T f d T f d T f ' d C O d C O d C N d r H d r H O r H d r H d r H d r H d r H d r H d C O d C N d r H d r H d r H d C O d c o d r H d L O r H 0 0 d 0 0 d T f d L N d T f d c o d C O d C N d T f d T f d VO d OM L N O N N O O N N O O N JH 01 U PH L N d 0 0 L O 0 0 o CN d N O d rH r H d r H d r H d r H d r H d r H d r H O C O d r H d r H d vo d C N d L N d C N r H C N d C N d OL I N O N VO O N VO O N JH 8 PH CJN d rH © rH d oo d rH r H d L O d r H d r H d r H d C O d r H d L O d rH d r H d r H d r H d r H d o q r i r H d OC L N O N VO O N V O O N PH r H d r H d 0 0 s oo d rH to O N L N O N r H d r H d r H d C N d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d r H d C O d C O d C O d C N d r H d r H d vo d T f d vo d T f d C O d C O d C O d r H d r H d T f d C O d L O d C O d r H d r H d r H d r H d C N d C N d C N d r H d 0 0 d r H d T f d r H d o T f r H C O C N C N C O C N r H d 0 0 r H T f r H T f r H T f r H r H r H T f r H O N r i T f r i ' vq r i T f d vo d C O d C / j i > P H to tS 8 •n o ta ~ R e-o IS ,>> "o a. ta • « R « •5 co *H •3 s 3 •*§ 42 O H C/3 w H >H K CH 0 PQ < H 0 H to Z w I! U HH r J a. tn s ta 3 s s D B -S 3 cn S ta R •s ta 0 « R R ts >H ts R •3 ts U JH CtS > ts H-* ts s §1 •§•"3 u .§ ts CO •2 3 •S "5 o T S ta 0 ta o •S "5 •§ ta D .tn "5 .§ •S '£ o ta CJ s ts f r -o ? •S ' R o T S ta U ta H— ta a. tn "B ,ta "5 T S ts D •3 -S .a s .ta "5 o T S ta U .tn "Q ta oc .ta " R o T S ta U + .tn 1 g ts o fe-S. ta 5 " 1 >> ta U U cn 1 r . & "3 £ .ta " R o T S ta U ts *-•5 .ts '5 T S ta U ts o f t .ta "5 o T S ta U •3 ta •*S r t .ta " R o T S ta 0 .ta "3 .o s CD .ta "5 O T S ta 0 CH PH tn .ta '5 T S ta U ta cn o H « a. ta ts fe -SP -42 ' 3 O H ts -5 S ta fe s o. ta £ ta T S *S ta fe -SP " 3 O H .ta •S to 8 s to <a fe -SP OH 1 ta £ ta fe • S ? OH 'B iS CO R ta fe •SP "55 OH 175 CN ON NO ON H Cu f— UU CN ON NO ON let I CN ON t N d U uu CN ON NO ON o CN ON NO ON i—I O I N ON NO ON U 0 t N d CO d LO d LO d TP d LO d NO d LO d TP d t N d CO CO 00 LO tN NO T)i NO NO LO LO LO Tf o o NO NO CN d TP d CO d TP d CO d Hi i l l 00 CO CM (N o ON CO d 00 LN CD Z i w K U HH i - l HJ «3 H 0 H co d LO d oo d co d CN d O E H CO to o £ CN u CO u z HH i-l Q w w co w TP CN ON CN CO | o CN d CN NO I TP CN TP CN NO CO TP d oo 3 |3 co g K to CN id CN CO GO d uo CO LO CO CN CN oo CN d NO d I 00 d CN CN ON d TP d CO d CN d TP CO CN d o NO CN TP oo CN LO d LO d o CN ON TP LO TP CN LO CN d CN d NO d NO d CN d CN d LO d §-1 •P NO CN NO d , 5^ Ko CN co < Z O J CN d TP d CO d NO d ON d I N d NO d TP d Tf d NO d 00 d TP d CN d LO d TJ< d CO d CN d ON d co d CN d CN d NO d oo d l$ l ' f o b co 0 UU LO d rcNi NO d CN d ON d oo d 1^  NO d TP d CO d co d CN d 176 JI 97 CO d NO d rH d CN d ON d rH d CN d rH d ON rH 10.1 rH CN rH d rH d ON d rH co FTL NO ON tN d Tf d rH d rH d Tf d rH O Tf d L O d ON ** NO T-i rH d r l d rH d oo rH FTL NO ON •-0. rH O CN d CO d CO d rH d rH d rH d CO d CN d rH d rH d rH d CN d Tf d CN d 00 d CO CN rH LO 00 d Tf r-i CO d r l d ON CN ON d NO Tf ON CN d CO d rH d rH d rH d CO d rH d rH d Tf d VO LO fO T-i rH d rH d oq r i CO CO NO ON CO d CO d rH d rH d CN d rH d rH d Tf d oo LO oo d CN d rH d oq r i 00 CN H UH NO ON 01 u CH rH d tN d Tf d CO d rH d rH d rH d rH d CO d CN d rH d rH O rH d rH d Tf d Tf d ON T f IN T f 00 d rH rH rH d CN d rH d rH d Tf r i Tf CN CO CN Tf CO CN ON rH d Tf d co d rH d Tf d rH d NO d ON CN rH rH CN rH d Tf d LO LO FC NO ON NO ON CH rH d rH d rH d rH d Tf d CO d Tf d co d CO d CO d rH d rH d CO d Tf d NO d rH d rH d rH d rH d rH d LN d Tf d CO r-i CO r-i rH ON IN T f d rH CN T-i rH rH CO CN T-H d rH d rH d rH d CO d CO d 00 d rH Tf rH CO rH LN CN ON TH d rH d rH d LO d rH d rH d rH d LN d rH CN oo d CN d ON NO Tf d rH d rH d NO d OM NO ON NO ON d CO d rH d CO d rH d CO d CN d rH rH rH d LO d rH d CO d CN d rH d LO T f rH d NO d CN d rH d rH d NO d CN d 01 L. CH NO d 00 d CN d CO r-i rH d rH d rH d rH d co d CN d CN d rH oo' rH d ON T-i NO d oq r i CO Tf IN ON CN d CN d rH d Tf d CN d CO d CN d T f LO 00 T-i r l d rH d CJN r i NO ON rH d CN d CO d rH d rH d CO d CO d rH d CO T f CN T-i rH O rH d rH d r l d NO T-i OL NO ON rH d CO d rH d CO d rH d CN d rH d O CO 00 d r l d 00 d o> »H CH CN d CN d rH d rH d CO d NO d rH d rH d Tf d rH d Tf d IN NO rH d Tf CN LN d r l d CO CO CN ON oo d rH rH rH d rH d rH rH rH d NO d rH r-i q r-i NO T-i rH O LO CN d Tf CN NO LN OC NO ON NO ON SH CH CO d CO d Th d 00 d oo d CO rH rH d rH d CN d rH d rH d LO d NO d rH rH rH d rH d rH d rH d LO d LN d Tf d rH d NO d CN d rH d 00 d LO d NO d ON oo' oo oo' CO od rH d rH d 00 CO rH CO ON Tf CN d CO d NO d r l rH 00 d q T-i rH LO CN Tf in NO £ •2 Ul fe R ta £ •2 .S o s-OH CO fe •4-4 Ul fe w £ SS r . 1 Ul « R •2 '5 & S .ts §b ts SH tL, tu 1 £ •2 "3 U *3 V. ts £ ts _cs c .ts 1 O £ •2 5 ts fe o £ a £ Ul ,U1 Ul 8 .cn £ •2 5 ts fe u .ts fe c .o 3 o cs fe •£? Q O U £ •3 .« "tu £ a £ a "C ts fe 5 .cn *3 H-i 5 ts "C u o "a 1 tn 2 o cn 3 v. S i H5 ts »—1 t j »H ts CU •2 £ a *H i t £ .ts fe .a ts '5b •2 •5 o 5 £ a B ts a . £ ,3 "3 f 8 Q ts t s R g tu v> .ts HS o ts HS R ts o ? ts ts .a "s ts a. cn 41 *S CO tu "3 K £ 3 ts s-: ts »H E2 tn .3 IS •S K & 1 ft; tS R 3 fe £ ts ts 'Cj 5 R a ts ts ^ts "S to 3 0 u-< E-H 0 H to D HH 0 z HH 0 to -2 . t s ""a to 3 £ o CQ CO to CU rO 3 .cn "co D fe ts £ ts 3 to "CU ts 5 R R "C R 8 X tu c3 CO cn o X tu »H ts U CH CO s 3 tn cu tL, . t s i t fe a. CO ts .cn CO a. o N r? o .cn to 1 ts a . ts o OH to D 0 Z HH 0 H-i < E-H 0 E-H 177 o tN 178 tx ON NO ON H PL tx ON NO ON tL, U tL, 2 o tx ON NO ON tx ON ho , ON , 2 tx ON NO ON O tx ON NO ON U o Cu d CO d •2 -2 K ts U ts fe .ts '5 •3 ts ID | 1 u 0 ol d CO d NO oi CO d Tt< d Ol d LO d o o CO oi CO d CO d CO d ol d ON LO ON CO LO "o~ TP d oi d CO d-ol d tx d NO d NO d TP d TP d oi d TP d CO d LO LO o T t 00 CO 11 U I + ts o s a o J ; | -J ; Ci-ts "to-l a l l c o ts ID l l a. ts ts I S to CL, I to C ts ••a in Z w K U HH _ l < 0 to cu •TH CJ 0) PH tn 3^ L. > cn SH CU > o CJ -t-> Pi 0) CJ SH QJ PH .6 1 O CJ o; CO CU O PI cu PH CO + cC PH CO 1 79 Ponderosa pine equation, N=102 X 2 Appendix 4.1 Percentages of surviving, dead, and all trees correctly classified by the single species equations using varying probability of mortality cutoff values. Percentage of trees classified correctly Cutoff value Surviving Dead Total Douglas-fir equation, N=159 0.2 87.7 94.9 91.2 0.9 0.3 93.8 92.3 93.1 0.0 0.4 93.8 92.3 93.1 0.0 0.5 95.1 92.3 93.7 0.0 0.6 96.3 89.7 93.1 0.6 0.7 96.3 87.2 91.8 1.3 0.8 97.5 87.2 92.5 1.6 Lodgepole pine equation, N=82 0.2 0 100 92.7 N/A 0.3 0 100 92.7 N/A 0.4 0 100 92.7 N/A 0.5 0 100 92.7 N/A 0.6 0 100 92.7 N/A 0.7 33.3 88.2 84.1 2.6 0.8 85.4 85.5 85.4 7.8 * 0.2 91.3 81.8 89.2 0.5 0.3 95.0 81.8 92.2 0.0 0.4 95.0 81.8 92.2 0.0 0.5 96.3 77.3 92.2 0.3 0.6 98.8 77.3 94.1 1.1 0.7 100 77.3 95.1 1.8 0.8 100 72.7 94.1 2.7 the predicted values are significantly different from the observed values for a = 0.05. 180 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0089392/manifest

Comment

Related Items