ANALYSIS, MODELLING AND ESTIMATION OF WILDFIRE FUEL LOAD IN
NORTH-CENTRAL BC FORESTS
by
Andrew Peacosh, RPF
B.S.F., University of British Columbia, 2008

THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
IN
NATURAL RESOURCES AND ENVIRONMENTAL STUDIES

UNIVERSITY OF NORTHERN BRITISH COLUMBIA
April 2024
© Andrew Peacosh, 2024

ABSTRACT
Improving our ability to assess fuel loads across forested landscapes will aid our ability to
manage and mitigate wildfire risk in the short- and long-term. Forest stands exhibit complex
spatial variation in fuel load amount and characteristics across landscapes. Conventional standlevel inventory methods often fail to adequately capture this variability, particularly for fuel load
specific metrics such as canopy fuel load (CFL), canopy bulk density (CBD), canopy base height
(CBH) and coarse woody debris (CWD). Airborne laser scanning (ALS) presents an opportunity
to model and predict these metrics at fine resolution, over landscape-level areas.
In this study, I assessed fuel loading in four basic forested stand types common in northcentral BC: live conifer, dead conifer, mixedwood and deciduous stands. I secondly evaluated
the potential for using ALS to model CFL, CBD, CBH, CWD and other pertinent fuel metrics
across stand types. ALS derived models were then used to project the above-noted metrics
across the entire study area, and to compute critical surface fire intensity (CSI) at a ten-meter
resolution for the 1,800 square kilometer land base that comprises the Chinook Community
Forest and surrounding lidar coverage.
Results demonstrate that live conifer, mixedwood, deciduous leading, and MPB killed
pine stands differ considerably with respect to CFL, CBD, crown length and CWD, and have
similar CBH Live conifer stands had the highest canopy load, CFL and CBD. Fuel loads in
mixedwood stands were not statistically different from live conifer stands for most fuel load
metrics. Dead pine stands did not constitute a significant canopy fuel hazard; however, their
surface fuel loads were very high.
My results demonstrate that many of the forest metrics important for fuel load
characterization can be modeled with high accuracy and mapped at a fine grain (i.e. 100 m2)
using ALS derived data. My work demonstrates that modelling fuel loads using ALS is broadly
ii | P a g e

applicable across the stand types in the study area, enabling the projection of fuel load metrics
without being constrained by different stand types.
This thesis contributes to the science of wildfire fuel load management by establishing a
replicable framework for quantifying fine-resolution fuel load metrics using ALS data across
diverse stand types and providing valuable insights for effective wildfire risk assessment and
mitigation strategies in forest management.

iii | P a g e

TABLE OF CONTENTS
ABSTRACT ................................................................................................................................................................................. II
TABLE OF CONTENTS ......................................................................................................................................................... IV
LIST OF TABLES .................................................................................................................................................................... VI
LIST OF FIGURES ................................................................................................................................................................ VII
ACKNOWLEDGEMENT ....................................................................................................................................................... XI
DEDICATION ......................................................................................................................................................................... XII
CHAPTER 1 .................................................................................................................................................................................1
Introduction to the research topic .......................................................................................................................................................... 1
Overview of the thesis structure ............................................................................................................................................................. 5
CHAPTER 2 .................................................................................................................................................................................6
Current fuel treatments to mitigate fire risk ................................................................................................................................ 9
Characterizing fuel load components........................................................................................................................................... 10
Research Objective ............................................................................................................................................................................. 23
Materials and Methods ............................................................................................................................................................................23
Study Area .............................................................................................................................................................................................. 23
Field Sampling ..................................................................................................................................................................................... 26
Stand Metrics ........................................................................................................................................................................................ 29
Canopy Fuel Load Calculations and Estimations: Biomass, CFL, CBD ........................................................................ 31
Statistical Analysis .............................................................................................................................................................................. 39
Results ...........................................................................................................................................................................................................40
Basic Tree and Stand Metrics ......................................................................................................................................................... 40
Canopy Fuel Metrics .......................................................................................................................................................................... 50
Surface Fuel Metrics .......................................................................................................................................................................... 68
Discussion ....................................................................................................................................................................................................70
Stand structure ..................................................................................................................................................................................... 70
Fuel loading by stand type ............................................................................................................................................................... 71
Fuel Loading in Dead Conifer Stand fuel load ......................................................................................................................... 78
Conclusions ........................................................................................................................................................................................... 81
CHAPTER 3 .............................................................................................................................................................................. 82
Introduction .................................................................................................................................................................................................82
Fuel hazard mapping ......................................................................................................................................................................... 86
Fuel mitigation ..................................................................................................................................................................................... 88
Materials and Methods ............................................................................................................................................................................89
Statistical Metrics ................................................................................................................................................................................ 90
Study Area .............................................................................................................................................................................................. 92
Field Data .............................................................................................................................................................................................. 93
ALS Data collection ............................................................................................................................................................................ 93
ALS data processing ........................................................................................................................................................................... 93
Predictive modelling .......................................................................................................................................................................... 94
Results ...........................................................................................................................................................................................................96
Tree and Stand Metrics ..................................................................................................................................................................... 96
Canopy Fuel Metrics ....................................................................................................................................................................... 103
Canopy Base Height Metrics ........................................................................................................................................................ 107
Canopy Length Metrics .................................................................................................................................................................. 111
Surface Fuel Metrics ....................................................................................................................................................................... 116
ABA Wall-to-wall results ............................................................................................................................................................... 119
Discussion ................................................................................................................................................................................................. 133

iv | P a g e

Fuel metrics ........................................................................................................................................................................................ 134
Coarse Woody Debris ..................................................................................................................................................................... 136
Landscape results ............................................................................................................................................................................. 137
Conclusion .......................................................................................................................................................................................... 138
CHAPTER 4 SYNTHESIS ................................................................................................................................................. 139
REFERENCES ...................................................................................................................................................................... 145
APPENDICES ........................................................................................................................................................................ 162
Appendix 1: Plot attributes summary .............................................................................................................................................. 162
Appendix 2: Plot Empirical summary ............................................................................................................................................. 163
Appendix 3: Predicted metrics by Community forest compartment .................................................................................... 164
Critical surface fire intensity ........................................................................................................................................................ 164
Fuel metrics ........................................................................................................................................................................................ 166
Appendix 4: Variable Importance. ................................................................................................................................................... 170
Basal area variable importance plot ......................................................................................................................................... 170
CBH (height to live crown) variable of impotance plot ...................................................................................................... 170
CFL variable of importance plot ................................................................................................................................................ 171
CBD variable of importance plot ................................................................................................................................................ 171
Crown length variable of importance plot ............................................................................................................................... 172
CWD variable of importance plot............................................................................................................................................... 172

v|Page

LIST OF TABLES
Table 1 Declay class reduction factors by species (n represents unidentifiable species). ......................................................29
Table 2. Summary of height to live crown relationships derived using multiple regression equations with the
intercept forced through zero. .........................................................................................................................................................31
Table 3. Biomass equations (from Ung et al. (2009), where DBH is diameter at breast height in cm and H is total
tree height in meters. ..........................................................................................................................................................................36
Table 4 Plot summary by basal area stand type....................................................................................................................................41
Table 5 Stand metrics from empirical data plots collected in 2018 and 2019. Tree Height and DBH are reported for
individual trees, while BA, SPH, Volume and Conifer Volume represent plot level data. Red font indicates sp
value of <0.05 for ANOVA. ............................................................................................................................................................43
Table 6 One way ANOVA using Kruskal-Wallis with Wilcoxon rank sum pairwise compartison of means for
volume per hectare. .............................................................................................................................................................................48
Table 7 Canopy fuel load metric results. ................................................................................................................................................52
Table 8 Canopy base height metric results.............................................................................................................................................59
Table 9 Canopy length metrics. ..................................................................................................................................................................64
Table 10 Coarse woody debris (CWD) fuel load metric results. ....................................................................................................69
Table 11 Power and effect analysis for CWD comparison. Sample types 1 and 2 are the samples in each of the stand
type pairs. Refer to Champely (2022) for details. The power attained by current samples available in the pairs
for all but CL-ML is sufficient for the purposes of determining a statistical difference at 80% (power = 0.8)
chance of correcly rejecting a false null hypothesis. ...............................................................................................................76
Table 12. List of ALS metrics used to develop fuel models and maps. .......................................................................................90
Table 13. Evaluation of model accuracy for individual tree and forest stand metrics. For each random forest model
run the ‘out of bag’ r2 (OOB r2), cross-validated r2 (CV r2), and averaged mean square error from the k-fold
cross validation runs are reported. .................................................................................................................................................97
Table 14. Evaluation of model accuracy for canopy fuel metrics. For each random forest model run the ‘out of bag’
r2 (OOB r2), cross-validated r2 (CV r2), and averaged mean square error from the k-fold cross validation runs
are reported......................................................................................................................................................................................... 103
Table 15. ALS canopy base height metrics random forest model results. ............................................................................... 108
Table 16. ALS canopy length metrics random forest model results. ......................................................................................... 112
Table 17. ALS surface fuel load (CWD) metrics random forest model results. .................................................................... 117
Table 18. Plot summary table showing stand type, vri stand type, and descriptive statistics for each plot. ............... 162

vi | P a g e

LIST OF FIGURES
Figure 1. Study area overview map. There are four management units (pink - Babine, green - Hannay, blue Roselake and purple - Southside) within the study area . The extent of LiDAR coverage (grey crosshatch) and
plot locations by year of data collection are shown across the study area. .....................................................................25
Figure 2. Study area with VRI characterized stand types indicated. .............................................................................................26
Figure 3. Estimation of small branch biomass. Proportion of small branch <0.5 cm for DBH values 4.5 to 50 cm
using Johnson et al. (1990) biomass equations for lodgepole pine and white spruce. ................................................34
Figure 4 One plot vertical fuel distribution example where the biomass of each tree is a separate colour and the peak
of the vertical distribution is the CBD. Some research identifies 0.037 as the beginning of the effective canopy
load (Van Wagner, 1977). This line is marked on the plot as a dashed vertical line. ..................................................38
Figure 5. Tree height (with quartiles and range included as box and vertical line) for each of the four stand types are
shown (CL Conifer Live, CD conifer Dead, ML Mixedwood Live, DL Deciduous Live). Plots are divided by
tree component with the left panel showing all trees (purple), only live conifer in the center left panel (green),
only dead conifers in the center right panel (grey), and only deciduous individuals shown in the right panel
(yellow). Statistical analysis (Table 5) was performed using “All” trees and reflects observed differences in the
left panel (purple). ...............................................................................................................................................................................44
Figure 6. DBH (with quartiles and range included as box and vertical line) for each of the four stand types are shown
(CL Conifer Live, CD conifer Dead, ML Mixedwood Live, DL Deciduous Live). Plots are divided by tree
component with the left panel showing all tree (purple), only live conifer in the center left panel (green), only
dead conifers in the center right panel (grey), and only deciduous individuals shown in the right panel
(yellow). Statistical analysis (Table 5) was performed using “All” trees and reflects observed differences in the
left panel (purple). ...............................................................................................................................................................................45
Figure 7. Basal area per hectare (median with quartiles and range included as box and vertical line) for each of the
four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood live, DL deciduous live). Plots
are divided by tree component with the left panel showing all tree (purple), only live conifer in the center left
panel (green), only dead conifers in the center right panel (grey), and only deciduous individuals shown in the
right panel (yellow). Statistical analysis (Table 5) was performed using “All” trees and reflects observed
differences in the left panel (purple).............................................................................................................................................46
Figure 8. Stand density (SPH) (median with quartiles and range included as box and vertical line) for each of the
four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood live, DL deciduous live) for
stems >4 cm in diamater. Plots are divided by tree component with the left panel showing all tree (purple),
only live conifer in the center left panel (green), only dead conifers in the center right panel (grey), and only
deciduous individuals shown in the right panel (yellow). Statistical analysis (Table 5) was performed using
“All” trees and reflects observed differences in the left panel (purple). ..........................................................................47
Figure 9. Commercial Conifer volume per hectare (median with quartiles and range included as box and vertical
line) for each of the four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood live, DL
deciduous live) for stems >4 cm in diamater. Plots are divided by tree component with the left panel showing
all tree (purple), only live conifer in the center left panel (green), only dead conifers in the center right panel
(grey), and only deciduous individuals shown in the right panel (yellow). Statistical analysis (Table 5) was
performed using “All” trees and reflects observed differences in the left panel (purple). .........................................49
Figure 10. Conifer volume per hectare (combined live & dead) (median with quartiles and range included as box and
vertical line) for each of the four stand types (CL conifer live, CD conifer dead, ML mixedwood live, DL
deciduous live). ....................................................................................................................................................................................50
Figure 11. Canopy fuel load (median with quartiles and range included as box and vertical line) for each of the four
stand types are shown (CL conifer live, CD conifer dead, ML mixedwood live, DL deciduous live). Plots are
divided by tree component with the left panel showing all tree (purple), only live conifer in the center left
panel (green), only dead conifers in the center right panel (grey), and only deciduous individuals shown in the
right panel (yellow). Statistical analysis (Table 7) was performed on the “All” and “Live Conifer” trees, and
reflects measured differences in the left (purple) and center left (green) panels. .........................................................54
Figure 12. Canopy bulk density (median with quartiles and range included as box and vertical line) for each of the
four stand types are shown (CL conifer Live, CD conifer Dead, ML Mixedwood Live, DL Deciduous Live).
Plots are divided by tree component with the left panel showing all tree (purple), only live conifer in the center
left panel (green), only dead conifers in the center right panel (grey), and only deciduous individuals shown in

vii | P a g e

the right panel (yellow). Statistical analysis (Table 7) was performed on the “All” and “Live Conifer” trees
and reflects measured differences in the left (purple) and center left (green) panels. .................................................55
Figure 13. Height of maximum CBD (median with quartiles and range included as box and vertical line) for each of
the four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood live, DL deciduous Live).
Plots are divided by tree component with the left panel showing all trees (purple), only live conifer in the
center left panel (green), only dead conifers in the center right panel (grey), and only deciduous individuals
shown in the right panel (yellow). Statistical analysis (Table 7) was performed on the “All” and “Live
Conifer” trees and reflects measured differences in the left (purple) and center left (green) panels. ....................57
Figure 14. Height of Max CBD profile for four stand types and tree types. Maximum CBD (all tree types) is
indicated with the blue arrow for each stand type. This point is markedly different in mixedwood and
deciduous stand types when considering only conifer tree types, indicating some complexities in fuel load
distribution.............................................................................................................................................................................................58
Figure 15. Height to live crown (median with quartiles and range included as box and vertical line) for each of the
four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood live, DL deciduous live). Plots
are divided by tree component with the left panel showing all trees (purple), only live conifer in the center left
panel (green), only dead conifers in the center right panel (grey), and only deciduous individuals shown in the
right panel (yellow). Statistical analysis (Table 8) was performed on the “All” and “Live Conifer” trees and
reflects measured differences in the left (purple) and center left (green) panels. Note that dead conifer trees do
not have live crowns...........................................................................................................................................................................60
Figure 16. Canopy base height using a 0.011 kg/m 3 CBD threshold (median with quartiles and range included as box
and vertical line) for each of the four stand types are shown (CL conifer live, CD conifer dead, ML
mixedwood live, DL deciduous live). Statistical analysis (Table 8) was performed on the “All” and “Live
Conifer” trees and reflects measured differences in the left (purple) and center left (green) panels. Note the
dead conifer panel shows no canopy base height for dead conifer trees due to there being no dead conifer
effective canopy fuel. .........................................................................................................................................................................62
Figure 17. Canopy base height at the 0.037 kg/m 3 CBD threshold (median with quartiles and range included as box
and vertical line) for each of the four stand types are shown (CL conifer Live, CD conifer dead, ML
mixedwood Live, DL deciduous live). Statistical analysis (Table 8) was performed on the “All” and “Live
Conifer” trees and reflects measured differences in the left (purple) and center left (green) panels. Note the
dead conifer panel shows no canopy base height for dead conifer trees due to there being no dead conifer
effective canopy fuel. .........................................................................................................................................................................63
Figure 18. Crown length (median with quartiles and range included as box and vertical line) for each of the four
stand types are shown (CL conifer live, CD conifer dead, ML mixedwood live, DL deciduous live). Plots are
divided by tree component with the left panel showing all trees (purple), only live conifer in the center left
panel (green), only dead conifers in the center right panel (grey), and only deciduous individuals shown in the
right panel (yellow). Statistical analysis (Table 9) was performed on the “All” and “Live Conifer” trees, and
reflects measured differences in the left (purple) and center left (green) panels. Note the dead conifer panel
shows no canopy length for dead conifer trees due to there being no dead conifer effective canopy fuel. .........65
Figure 19. Canopy length using a 0.011 kg/m 3 CBD threshold (median with quartiles and range included as box and
vertical line) for each of the four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood
live, DL deciduous live). Statistical analysis (Table 9) was performed on the “All” and “Live Conifer” trees
and reflects measured differences in the left (purple) and center left (green) panels. Plots are divided by tree
component with the left panel showing all trees (purple), only live conifer in the center left panel (green), only
dead conifers in the center right panel (grey), and only deciduous individuals shown in the right panel
(yellow). Note the dead conifer panel shows no canopy length for dead conifer trees due to there being no
dead conifer effective canopy fuel. ...............................................................................................................................................67
Figure 20. Canopy length using a 0.037 kg/m 3 CBD threshold (median with quartiles and range included as box and
vertical line) for each of the four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood
live, DL deciduous live). Statistical analysis (Table 9) was performed on the “All” and “Live Conifer” trees
and reflects measured differences in the left (purple) and center left (green) panels. Plots are divided by tree
component with the left panel showing all trees (purple), only live conifer in the center left panel (green), only
dead conifers in the center right panel (grey), and only deciduous individuals shown in the right panel
(yellow). Note the dead conifer panel shows no canopy length for dead conifer trees due to there being no
dead conifer effective canopy fuel. ...............................................................................................................................................68

viii | P a g e

Figure 21. Coarse woody debris (CWD) amount (median with quartiles and range included as box and vertical line)
for each of the four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood live, DL
deciduous live). Statistical analysis (Table 10) was performed on the “All” trees (purple). ....................................70
Figure 22. Observed versus predicted mean tree height for each sample plot. Plots are color coded by stand type, and
the full model cross-validated r2 value is reported. .................................................................................................................98
Figure 23. Observed versus predicted mean tree 90th percentile of tree height for each sample plot. Plots are color
coded by stand type, and the full model cross-validated r2 value is reported.................................................................99
Figure 24. Observed versus predicted mean tree DBH for each sample plot. Plots are color coded by stand type, and
the full model cross-validated r2 value is reported. ............................................................................................................... 100
Figure 25. Observed versus predicted mean basal area per hectare for each sample plot. Plots are color coded by
stand type, and the full model cross-validated r2 value is reported. ................................................................................ 101
Figure 26. Observed versus predicted mean canopy crown length for each sample plot. Plots are color coded by
stand type, and the full model cross-validated r2 value is reported. ................................................................................ 102
Figure 27. Observed versus predicted crown fuel load for each sample plot. Plots are color coded by stand type, and
the full model cross-validated r2 value is reported. ............................................................................................................... 104
Figure 28. Observed versus predicted canopy bulk density for each sample plot. Plots are color coded by stand type,
and the full model cross-validated r2 value is reported........................................................................................................ 106
Figure 29. Observed versus predicted height of maximum CBD for each sample plot. Plots are color coded by stand
type, and the full model cross-validated r2 value is reported. ............................................................................................ 107
Figure 30. Observed versus predicted mean height to live crown for each sample plot. Plots are color coded by stand
type, and the full model cross-validated r2 value is reported. ............................................................................................ 109
Figure 31. Observed versus predicted canopy base height using the 0.011 kg/m 3 threshold for each sample plot. Plots
are color coded by stand type, and the full model cross-validated r2 value is reported............................................ 110
Figure 32. Observed versus predicted canopy base height using the 0.037 kg/m 3 threshold for each sample plot. Plots
are color coded by stand type, and the full model cross-validated r2 value is reported............................................ 111
Figure 33. Observed versus predicted mean crown length for each sample plot. Plots are color coded by stand type,
and the full model cross-validated r2 value is reported........................................................................................................ 113
Figure 34. Observed versus predicted canopy length using the 0.011 kg/m 3 CBD threshold for each sample plot.
Plots are color coded by stand type, and the full model cross-validated r2 value is reported................................. 115
Figure 35. Observed versus predicted Canopy Length using the 0.037 kg/m 3 CBD threshold for each sample plot.
Plots are color coded by stand type, and the full model cross-validated r2 value is reported................................. 116
Figure 36. Observed versus predicted coarse woody debris for each sample plot. Plots are color coded by stand type,
and the full model cross-validated r2 value is reported........................................................................................................ 118
Figure 37. Observed versus predicted log CWD for each sample plot. Plots are color coded by stand type, and the
full model cross-validated r2 value is reported. ...................................................................................................................... 119
Figure 38. Predicted basal area per hectare for entire study area at 10 m resolution. Blue represents high basal area
(>46 m2/ha) and yellow represents very low (<12 m 2/ha). Wildland Urban Interface polygons are also shown.
................................................................................................................................................................................................................ 121
Figure 39. Babine compartment predicted basal area per hectare. The area defined by the smaller red square is
shown in higher resolution in Figure 40................................................................................................................................... 122
Figure 40. Red square inset (1km square) from previous figure showing predicted basal area per hectare. ............... 123
Figure 41. Predicted canopy bulk density for Southside compartment at 10 m resolution. CBD varies from 0 to
approximately 0.3 kg/m 3................................................................................................................................................................ 124
Figure 42. Predicted canopy bulk density for Roselake compartment at 10 m resolution. CBD varies from 0 to
approximately 0.3 kg/m 3. The majority of areas with CBH higher than 0.10 kg/m 3 is along the southwest of the
compartment although there are predicted CBDs higher than 0.10 throughout the area. ........................................ 125
Figure 43. Predicted canopy bulk density for Babine compartment at 10 m resolution. CBD varies from 0 to
approximately 0.2 kg/m 3 with large portions of the area predicted as having CBD above 0.10 kg/m 3. ............ 126
Figure 44. Predicted canopy bulk density for Hannay compartment at 10m resolution. CBD varies from 0 to
approximately 0.3 kg/m 3 with concentrations predicted to be higher than 0.10 primarily in the northern
portions of this compartment. ...................................................................................................................................................... 127
Figure 45. Predicted CBD in the Southside compartment showing wildland urban interface (WUI) areas (pink
polygons). Predictions indicate high CBD within WUIs in this area. ............................................................................ 128
Figure 46. Predicted CBD from previous figure, zoomed showing WUI boundary and predicted CBD inside and
outside of the WUI. Note contiguous nature of CBD within WUI. ................................................................................ 129

ix | P a g e

Figure 47. Predicted CWD in kg/m2 for an area within the Southside compartment. Note that anything yellow,
orange or red is considered high CWD fuel load. ................................................................................................................. 130
Figure 48. Critical surface intensity (CSI) calculated from predicted conifer canopy base height with constant FMC
of 95 for the entire study area. Red indicates that the critical surface fire intensity is predicted to be low (i.e.
surface fire can easily transition into crown fire) based on low predicted canopy base heights. The yellow
square (at the bottom of the figure) shows the areas that is presented at a higher resolution in Figure 49. ...... 131
Figure 49. Close up of yellow inset from previous figure, showing calculated CSI from predicted conifer canopy
base height. ......................................................................................................................................................................................... 132
Figure 50. Roselake CSI ............................................................................................................................................................................ 164
Figure 51. Babine CSI. ............................................................................................................................................................................... 165
Figure 52. Hannay CSI............................................................................................................................................................................... 165
Figure 53. Soutside CSI. ............................................................................................................................................................................ 165
Figure 54 Variable importance plot for the random forest model of basal area metric, displaying only the top 20 out
of 65 ALS metrics included in the model. The vertical axis represents the predictor variable ranked by
importance, with the highest variable being the most significant to model predictive power. The horizontal
axis shows the importance score, an indicator of the predictive utility of variables, determined using the
varImpPlot() function in the randomForest package (Liaw & Wiener, 2002). Variable importance scores are
calculated based on the mean decrease in accuracy when each predictor variable is permuted, reflecting its
impact on model performance. This plot shows zmean as the highest variable of importance along with several
height percentile metrics. ............................................................................................................................................................... 170
Figure 55 Variable importance plot for the random forest model of CBH metric, showing higher point height
percential metrics (70, 75, 80 and 85) and several cummula ........................................................................................... 170
Figure 56 Variable importance plot for the random forest model of CFL metric, showing two height percentile
metrics (zq60 and zq55), as well as the vertical complexity index metric (VCI). ..................................................... 171
Figure 57 Variable importance plot for the random forest model of CBD metric showing alsvlad (coeeficient of leaf
area density metric) as the higheset als metric along with two low-to-ground cummulative point metrics
(zpcum1 and zpcum2). Also showing pabove zmean which is the percentage of poinst above the mean. ..... 171
Figure 58 Variable importance plot for the random forest model of crown length metric showing primarily zq ALS
metrics (height percentile metrics). ............................................................................................................................................ 172
Figure 59 Variable importance plot for CWD showing cummulative height metrics for the first and second meter
above ground. .................................................................................................................................................................................... 172

x|Page

ACKNOWLEDGEMENT
I would like to express my deepest gratitude to my wife and partner Leanne Elliott for her
boundless patience and support on this longer-than anticipated academic journey. Through covid
and two children your continued support helped me somehow see the end. Also to my children
Ryleigh and Nathan. You have never known a father who was not always wondering with some
part of his mind how he would ever get his thesis completed so he could spend more time with
his kids on a sunny weekend. Prepare to meet dad 2.0.
Special thanks are also due to my supervisor Dr. Che Elkin for his invaluable guidance,
encouragement, expertise, and understanding as I navigated the demands of student, work, life
and family. I am grateful for his mentorship and dedication to my academic and professional
development. Also, to my committee for their encouragement through this long journey.
I extend my appreciation to the members of the Mixedwood Ecology Team, past and
present, for their collaboration, insights, and contributions to this research endeavor and to my
employer Sinclar Group Forest Products Ltd. who supported my research the whole journey by
allowing me have a flexible work schedule to complete academic studies while working full
time.
I would also like to thank the Chinook Community Forest, represented by Ken Nielsen
for the generous financial support, assistance with LiDAR acquisition, provision of field data,
and for their extreme patience.
Furthermore, I wish to acknowledge the six First Nations whose traditional territory
encompasses the study area: Burns Lake Band, Cheslatta Carrier Nation, Lake Babine Nation,
Nee Tahi Buhn Indian Band, Skin Tyee Band, and Wet’suwet’en First Nation. I am deeply
respectful of their stewardship of the land and grateful for their enduring connection to and
protection of these landscapes. I hope that this research contributes to fostering safer and more
wildfire-resilient landscapes and communities for the benefit of all.

xi | P a g e

DEDICATION
This thesis is dedicated to the memory of Changru Li (1982-2023) whose unwavering
support, encouragement, and friendship sustained me throughout my academic journey. It is my
hope that with more access to remote sensing derived modelling of fuel and forest attributes,
fewer forestry and wildfire workers will be put in harm’s way. Although he is no longer with us
to witness its completion, his impact on my life and work will always be cherished and
remembered.

xii | P a g e

CHAPTER 1
Introduction to the research topic
In 2011 Jerry Williams, the former National Director of Fire and Aviation Management
for the United States Forest Service, presented a paper at the International Association of
Wildland Fire in which he explored the onset of high-impact mega-fires through a forest land
management prism (Williams, 2013). Williams identified some differences and commonalities
among a number of large fires which he has been credited with calling mega-fires. This term has
commonly been used to describe a fire greater than 10,000 ha, but it has also been used to
describe fires of significant behavior or impact (Linley et al. 2022). While the fires he examined
were from regions all over the world, contributing factors in all the fires included an
accumulation of biomass and fuels dominating the landscape, and forest land management
practices supporting the occurrence of the mega-fires (Williams, 2013). Williams also noted a
reduction of extreme fire behaviour when fires encountered areas that had been previously
treated to reduce fuels (Williams, 2013).
Williams argued that forest management objectives that aim to limit the accumulation of
fuels and diversify of stands may be the best protection against the threat of mega-fires. He also
suggested that the spatial arrangement of less susceptible stands, interspersed across the
landscape, may also act to reduce the threat of mega-fires and provide protection to values that
specifically require dense fuel types (Williams, 2013). While the fires Williams evaluated were
not in British Columbia (B.C.), this province has since experienced significant occurrences of
large fires. In particular, the Elephant Hill fire, the Plateau fire complex, and Hanceville-Riske
Creek fires in 2017 are noteworthy. The Shovel Lake, Nadina Lake, Verdun Mountain, Alkali
Lake Lutz Creek and Tweedsmuir complex fires in 2018, and the Donnie Creek fire in 2023

1|Page

were all megafires. While not as large, other fires that occurred during this time were devastating
due to their proximity to communities and their impacts on people’s lives. The Okanagan
Mountain Park and the McLure fires in 2003, and the Lytton fire in 2021 were particularly
devastating. In 2017 and 2018 B.C. experienced two of the most challenging wildfire seasons in
recorded history, breaking all records for area burned and number of evacuees. . The 2017 fire
season resulted in 1.2 million hectares burned, 568 million dollars expended in suppression, and
65,000 people evacuated.
Two of the largest fires of the 2017 season, the Elephant Hill fire near Ashcroft, BC, and
the Plateau Complex fire in the Cariboo region west of Quesnel were record breaking fires by
themselves. The Elephant Hill fire reached 191,865 ha, while the Plateau Complex, which started
as 20 separate fires that combined into one, formed the largest fire in BC’s recorded history to
that date, at 545,151 ha (Cariboo Regional District Emergency Operations Centre, 2018). The
following year resulted in over 1.3 million hectares of area burned spanning the breadth of the
province. In 2023, an even larger fire, the Donnie Creek fire burned 575,511 ha in northeastern
B.C. There were 66 evacuation orders, thousands of properties impacted, and over 615 million
dollars spent on suppression, a total area burned of 2.8 million hectares
(https://www2.gov.bc.ca/gov/content/safety/wildfire-status/about-bcws/wildfire-history/wildfireseason-summary). The Northwest Fire Center and Prince George Fire Center experienced a
disproportionate amount of area burned compared to other regions of the province, topping 2.2
million ha burned in the Prince George Fire Centre and 174,000 hectares in the Northwest. The
prevalence of these significant fires is supported in part by contiguous fuel loads from dead
mountain pine beetle-killed stands, lightning and hot dry weather causing drought in many parts
of the province. Stephens et al. (2014) described a mega-fire triangle that identifies climate

2|Page

change, fire exclusion and antecedent disturbance as three major contributing factors to megafires.
It is not only becoming more expensive and challenging to fight wildfires, but
communities are becoming increasingly at risk. We must therefore look to long-term solutions
for wildfire mitigation, and solutions must be at the scale of the problem: the landscape level. It
is also argued that solutions need to include a silviculture component that incorporates a strategic
design of forests at the landscape level.
The challenge of creating a landscape that is resilient and able to generate and recycle
fuels at a natural rate (Williams, 2013) is exasperated by the warming climate and all the
associated impacts on disturbances. While the science of projecting how disturbance regimes
will change in the future is an evolving field (Haughian et al., 2012), it is evident that the climate
is changing and that B.C. landscapes can expect a significantly altered disturbance regime.
According to some models, British Columbia is projected to experience a temperature increase of
2.3 degrees Celsius, and an 8% increase in annual precipitation by 2050 (Werner, 2011). In
addition, lightning strikes, a major cause of ignition of wildfires, are predicted to increase by
approximately 12% for every temperature increase of 1 degree Celsius (Romps et al., 2014).
Flannigan et al. (2016) examined the sensitivity of fuel moisture to temperature and precipitation
changes under future climates and projected that for every degree of warming precipitation must
increase by more than 15% for the Canadian Forst Fire Danger Rating System (CFFDRS) fine
fuel moisture Code (FFMC), 10% for the CFFDRS duff moisture code (DMC) and 5% for the
drought code (DC) in order to compensate for drying resulting from increased temperatures.
They also found that large increases in precipitation have little ability to compensate for small
increases in temperature when the FFMC is above the critical threshold of 91 (Flannigan et al.,

3|Page

2016, p63). Although precipitation is also expected to increase with climate change (Werner,
2011), the Flannigan et al. (2016) results suggest that precipitation alone cannot moderate the
drying that should be expected by increased temperatures, particularly on fine fuels.
Disturbances in ecosystems occur at various scales, and come in many types (Picket and
White, 1985). When considering fire disturbance cycles, an evaluation of historic periods can
help in identifying how vegetation characteristics have influenced the occurrence, size and
spatial arrangement of fires (Picket & White, 1985). However, as a result of a policy of fire
suppression, and compounded by forest pest epidemics, changing climates and silvicultural
practices, historic patterns of fire disturbance may differ from current conditions.
Fuels are the primary factor that contribute to fire behaviour, and it should be expected
that fire behaviour will change in response to shifts in fuel loads and composition. While
management of fuels is the largest lever available for land managers to modify fire occurrence
and behavior, it is not the only mechanism. Along with surface fuel load mitigation, it has been
suggested that fire hazard in general might be reduced by the introduction of more broadleaves in
conifer stands (i.e. mixedwoods), and the promotion of pure broadleaf stands. Increasing species
diversity and stand resilience and resistance to fire may also occur from increased broadleaves on
the landscape (Harper & Roach, 2014).
Common metrics used to assess fuel hazard are canopy fuel load (CFL) in kg/m 2, canopy
bulk density (CBD) in kg/m3, canopy base height (CBH) in meters, and canopy length in meters.
CFL is the total amount of fuel available to be immediately burned in the flaming front of a fire.
CBD is the density of fuel immediately available to be consumed in the flaming front of a fire
within the volume of the canopy. CBH is the height at which the canopy begins (either the
average height to live crown, or the height at which the CBD exceeds a critical density). Canopy

4|Page

length is the length in meters from the CBH to the top of the canopy, an important metric
because it reflects the density of fuel and CBD.
Coarse woody debris (CWD) is also an important metric, with high CWD often indicating high
surface fuel loads. For this research I did not collect fine surface fuel data but did collect CWD
data. This research aims to evaluate how canopy fuel load (CFL), canopy bulk density (CBD),
canopy base height (CBH), canopy length, and coarse woody debris (CWD) differ among four
common stand types in north-central BC.
In the second part of my thesis, I use Aerial Laser Scanning (ALS) with an Area-Based
Approach (ABA) to model these fuel metrics at fine resolution across an 1,800 km2 land base
situated around the village of Burns Lake, BC. Analysis of the differences in fuel loads among
forest stand types as well as the projection of fuel metrics for the landscape will provide context
to four wildfire fuel load and mitigation questions: First, does fuel load differ with respect to
stand type? Second, do some stand types, such as mixedwoods, offer potential as silvicultural
options for reducing fuel hazard? Third, how hazardous are dead pine stands in relation to other
stand types? Fourth, does modelling fuel metrics with ALS provide some landscape-level context
as to whether fuel loads are hazardous within and adjacent to the study area?
Overview of the thesis structure
This thesis is organized into four chapters. Chapter 1 provides an introduction to the
research topic and an overview of the thesis structure. Chapter 2 is an empirical evaluation of
forest fuel loads in different stand types in north-central BC. I specifically focus on mixed
deciduous/conifer stands and stands impacted by mountain pine beetle (MPB). This chapter
addresses four questions: 1) Do four common northern BC forested stand types (live conifer,
dead conifer (primarily MPB-impacted), mixedwood and deciduous) exhibit similar stand

5|Page

structure? 2) Do these different stand types exhibit significant differences in fuel load and fuel
hazard conditions? 3) Do dead conifer stands represent a higher fuel hazard compared to other
stand types due to their high fuel load and higher probability of ignition? and finally, 4) Do
mixedwood stands represent a more balanced alternative to deciduous stands with respect to
achieving timber value while mitigating fuel hazard, due to their lower fuel load and higher
timber value relative to deciduous?
Chapter 3 evaluates if and how Aerial Laser Scanning (ALS) can be used to estimate
forest fuel loads. This chapter tests if accurate estimates for CFL, CBD, CBH, and CWD can be
developed using ALS predictor variables. I also evaluate if forest fuel estimated using an areabased approach (ABA) can produce an accurate representation of hazardous fuel loading
conditions at the landscape scale. Using this approach, I also evaluate if fuel metrics and
projected fuel loads can be used to identify hazardous fuel conditions.
Chapter 4 provides a synthesis of the two data chapters and offers conclusions and
recommendations for future research. This chapter is divided into several sections: a summary of
the main findings, implications for theory and practice, limitations of the study, future research
directions, and concluding remarks.
CHAPTER 2
Wildfires are a stand replacing disturbance agent in many terrestrial ecosystems and serve
as one of three major pathways for recycling organic matter (Pausas & Bond, 2020). Despite a
downward trend in global area burned in recent decades, fire activity, size, and severity have
increased in many parts of the world (Doerr & Santı́n, 2016). In Canada, there are approximately
8,000 wildfires per year. There is an average of 2.5 million hectares burned per year (Canadian
Forest Service (CFS), n.d.) with wildfire frequency and severity expected to rise (Coogan et al.,
6|Page

2019). It is anticipated that fire suppression and management will continue to become
increasingly challenging and expensive (Tymstra et al., 2020).
British Columbians are familiar or experienced with effects of severe wildfires, having
recently experienced some of the largest and most destructive fires in recorded history. In 2003,
the Okanagan Mountain Park Wildfire burned 25,600 ha, much of it in a wildland-urban
interface (WUI) area. In the 2017 wildfire season 1.2 million hectares burned, over 649 million
dollars were spent on suppression activities, and 65,000 people were evacuated. The following
year, a similar number of hectares burned, and similar costs and evacuations were incurred, with
approximately 837,379 hectares burned in the Northwest Fire Centre alone in three major fire
complexes: the Babine, Stikine, and Nilkitkwa. From 2019 to 2022 there was less fire activity in
terms of area burned, but not a reduction in cost. In 2021, the Lytton Creek Fire destroyed the
village of Lytton, BC, highlighting the need for greater awareness and mitigation of fuels in
interface areas. In 2023, the province again experienced large fires. In particular, the Donnie
Creek wildfire in northeast BC is now the largest wildfire BC’s recorded history at 574,511 ha.
Rising temperatures and other climatic conditions are likely to produce similar wildfire seasons
in the future.
Wildfire occurrence and severity is principally influenced by climatic conditions,
topography, and the amount of available fuel (Walker et al., 2020). In their study on climate and
fire season, Flannigan et al. (2013) used Cumulative Severity Rating (CSR) to assess the possible
climate change-related impacts on fire season and suggested that fire seasons will be more severe
relative to the 1971-2000 baseline particularly in northern high latitudes. Jolly et al. (2015) found
that fire seasons have become longer across 25% of the planet’s vegetated surface. Wang et al.
(2015) modeled extreme fire weather events across Canada between 1970 and 2090 and argued

7|Page

that fire spread days will increase 35 to 400% by 2050. BC experiences large-scale, regular
climate variations, dominated by the El Nino Southern Oscillation (ENSO) on both a seasonal
and inter-annual basis (Haughian et al., 2012). The Pacific Decadal Oscillation (PDO) and
Pacific North American Pattern (PNA) dominate on decadal timescales (Haughian et al., 2012).
These natural climatic patterns can create seasonal and regional variations, such as warm, dry
winters or the opposite. While these are natural, cyclical processes, global climate model
projections indicate the potential for these extremes to occur more frequently (Lapp et al., 2012),
which can facilitate drought conditions in some areas or more extreme precipitation and weather
in others. All of these factors must be considered when projecting future changes to forest
disturbance regimes. With warmer and drier conditions, occurrence and severity of wildfire has
the potential to increase (Haughian et al., 2012).
While weather conditions and topography are key determinants of fire risk and fire
behaviour, fuel availability and distribution also play an important role. Of these three
components (fuel, weather, and topography) fuel is the only one that forest managers can
influence. Recent research into the interactions between fuel availability and fire weather
indicate that in boreal forests, fuel availability rather than fire weather controls the severity of
wildfires (Walker et al., 2020). Understanding the amount and distribution of forest fuels, in both
the horizontal and vertical spaces, is therefore critical in mitigating the hazard caused by fuels,
particularly where crown fires are a risk.
Crown fires are a destructive type of wildfire. They are difficult to suppress and control,
spread quickly, and can cause high or total stand mortality, high smoke production, and loss of
soil nutrients (Scott & Reinhardt, 2001). Managing the hazards associated with canopy fuel load,
canopy bulk density and canopy base height requires an understanding of the conditions that lead

8|Page

to crown fire initiation and spread (Cruz & Alexander, 2010).
Current fuel treatments to mitigate fire risk
In BC, the provincial government considers the goal of fuel treatment to prevent the
transition of surface fires into crown fires (BC Wildfire Service (BCWS), 2023). To achieve this,
prescriptions for hazard mitigation are to be designed to reduce surface loading below the point
at which potential surface fire intensity levels reach a critical surface fire intensity of 2,000
kW/m. This threshold is crucial because it allows ground crews to operate safely and minimizes
transition of surface fires into the crown where fires often become more dangerous (Wotton,
Flannigan, & Marshall, 2017). The concept of a critical surface fire intensity, introduced by Van
Wagner (1977) is an important component of BC’s approach to fuel mitigation. Van Wagner’s
modified formula for CSI is represented as CSI = 0.001 x CBH^1.5 + (460 + 25.9 * FMC)^1.5
and provides a method for estimating the energy required for a surface fire to transition into a
crown fire. The formula incorporates the Foliar Moisture Content (FMC), reflecting its
flammability, and considers the distance between the ground surface and the base of the crown
vegetation (CBH). CSI depends on two variables, the CBH and the Foliar Moisture Content
(FMC). CBH is described above, and FMC is a metric used in forest fire danger models to
describe the moisture content in conifer needles or leaves (Danson & Bowyer, 2004). FMC
varies across species, foliage ages, and seasons, with percentages ranging from 73% to 480%
(Keyes, 2006). The formula is relatively insensitive to changes in FMC. For this reason, fuels
planning focuses mainly on canopy base height. However, when surface fire intensities and
flame lengths are higher, FMC becomes more important (Keyes, 206). For FMC levels of
between 95 and 120 %, a CBH between 4 and 5 m is enough to keep the CSI below 2,000 kW/m.
If FMC falls below 95%, depending on flame length, a higher CBH may be required. As the

9|Page

height to live crown increases, a higher threshold for critical surface fire intensity is required
(BCWS, 2023). Crown closure and CBD should also be kept low to decrease crown fire potential
and minimize crown fire spread rate and spotting. A high hazard fuel condition associated with
CBD is considered to be where CBD exceeds 0.10 kg/m3 (Agee, 1996).
Fine fuels (less than 7 cm in diameter or thickness) contribute the most to rate of spread
(Government of British Columbia, 2012). The hazard associated with fine fuels depends on
slope, aspect, proximity to interface values, and base height of the crown and the flame length.
Surface fire can create enough energy in certain circumstances to preheat and combust fuels in
the canopy (Agee & Skinner, 2005). The process where surface fire begins to involve the crown
of individual canopy trees is referred to as torching (Van Wagner, 1977). Torching occurs when
the length of the surface flames exceeds a certain value determined by the fuel moisture content
and the canopy base height (Agee & Skinner, 2005).
Avoiding crown fire initiation requires maintaining surface fire intensity below the
critical surface fire intensity. This can be achieved by reducing fine surface fuels or by ensuring
that initial intensity (Io) required for crowning is large by maintaining high canopy base height or
by maintaining canopy bulk density below 0.10 kg/m3 (Agee, 1996). High fuel hazard conditions
exist where either or both of these conditions are not met.
Characterizing fuel load components
Van Wagner (1977) identified criteria for crown fire initiation and spread. These criteria
have been further developed and incorporated into many wildfire behaviour models
(Xanthopoulos, 1990; Alexander et al., 1998; Cruz et al., 2002; Reinhardt et al., 2003).
Modelling and assessment of canopy fuel loads and associated hazards requires estimations of
the fuel load and density of fuel in the canopy as well as the height of the canopy above the

10 | P a g e

ground and surface fuels. These key metrics are described as canopy base height (CBH), canopy
length, canopy fuel load (CFL), canopy bulk density (CBD) and coarse woody debris (CWD).
These metrics are challenging to describe and quantify, and are often estimated in different ways,
reflecting the different objectives, regional data availability and modelling inputs (RuizGonzález & Álvarez-González, 2011; Carey & Schumann, 2003; Cruz et al., 2003; Reinhardt et
al., 2006. In the following sections I provide an overview of these five metrics and how they
have been historically estimated.
Canopy base height
The crown base height is the height to live crown of an individual tree. It is an important
metric describing the connectivity between surface fuels and canopy (Ex et al., 2016). However,
canopy base height (CBH) is often described in two different ways. Sando and Wick (1972)
assumed that CBH was the height above ground at which fire could propagate vertically into the
canopy and estimated it as the lowest 0.3 m section of canopy where the amount of available
canopy fuel exceeded 0.037 kg/m3 (Sando & Wick, 1972). Reinhardt et al. (2006) applied a
similar threshold methodology to describe CBH, using 0.012 kg/m3as their critical threshold.
Others, including Van Wagner (1977), Cruz et al., (2003), McAlpine and Hobbs (1994),
Beukema et al. (1997), Phelps & Beverly (2022) and others describe CBH as the mean height to
live crown taken as the average height of the lowest live branch in a plot or stand.
Canopy length
Crown length, sometimes referred to as crown depth, is typically measured from tree top
height to the lowest green branch or complete green whorl (Kohyama et al., 1990). At the plot
level canopy length can be calculated as the average of all crown lengths or can also be
interpreted as the vertical length at which the canopy density profile exceeds a certain density

11 | P a g e

threshold, often either 0.011 kg/m3 (Scott & Reinhardt, 2001) or 0.037 kg/m3 (Van Wagner,
1977). Canopy length is a metric influencing the behaviour of wind in the canopy (Massman et
al., 2017; Russell et al., 2016) and is an important component used for calculating CBD.
Canopy fuel load
Canopy fuel load (CFL) is defined as the weight of canopy biomass per unit area that is
available for combustion during the flaming front of a fire (Cruz et al., 2003). It can be measured
through destructive sampling or estimated with allometric equations based on previous
destructive sampling (Cameron et al., 2021). Many researchers and models include only live
foliage as a component of CFL, while others include live foliage and half of the 0 to 6 mm
branch biomass (Alexander, 1988; Finney, 1998; Reinhardt et al., 2006; Scott & Reinhardt,
2001; Van Wagner, 1977). Conifer foliage and small branches less than 6 mm are usually
considered the available or effective portion of the canopy. This is not always the case.
Flammability, defined as the capacity of forest fuels to ignite and combust, has many
determinants, not simply size or chemical properties (Varner et al., 2015). For example, conifer
needles and twigs may readily ignite and burn at low moisture contents but not when moisture
content is high. The same can be said for deciduous foliage. In spring, deciduous stands may be
more volatile than at other times (Alberta, 2012; Burton, 2023; Quintilio et al., 1991). As fire
seasons change, obtaining a comprehensive understanding of moisture content of foliage and
small branches, regardless of the species, becomes important for making informed management
decisions.
Biomass of available canopy fuel components can be estimated using appropriate
species-specific equations and adding the sum of the foliage and small branch. However, in
many regions equations for all components are not always available. Brown’s (1978) biomass

12 | P a g e

equations are commonly used in studies in the United States to estimate crown fuel load,
however they are localized and specific to certain species. Using local allometries was found by
Keyser and Smith (2010) to produce higher estimates and is generally recommended over nonlocal relationships. In Canada, equations from Lambert et al. (2005) and Ung et al. (2008) have
been used to calculate biomass components, but these do not differentiate subgroups within the
branch component, leading many to use only the foliage (e.g. Phelps & Beverly, 2022). In both
the Lambert et al. (2005) and Ung et al. (2008) equations foliage is defined as live foliage and
associated twigs, while branch biomass refers only to live branches, and neither include dead
branches (Aldred & Alemdag, 1988). Using local allometries has been shown to be more
accurate than regional or national level allometries that are prone to either systematically under
or over-estimate biomass components. Keyser and Smith (2010) found that using local rather
than generic equations for ponderosa pine (Pinus ponderosa) in the US accounted for a 47%
increase in CBD in their study. Differences were also found by Xing et al. (2019) when
comparing regionally fitted allometric equations for three boreal tree species from Alberta to
commonly used Canadian national equations for aboveground biomass (Lambert et al., 2005;
Ung et al., 2008).
Once crown biomass is estimated for individual trees, the sum of the foliage and small
branch components in the stand is considered the canopy fuel load as measured in kilograms per
square meter (kg/m2).
Canopy bulk density
The canopy fuel load must then be distributed within the canopy, considering tree height
and canopy length, to derive the density of available fuel within the volume of canopy space.
This density metric is referred to as the canopy bulk density (CBD). It is the mass of the

13 | P a g e

available CFL per unit of canopy volume measured in kg/m3. It is a measure of how closely
packed the needles, leaves and twigs are within the volume of the canopy and tells us how likely
it will be for fire to move from crown to crown within the canopy (Keyser & Smith, 2010) and is
needed to predict canopy fire spread, initiation (Van Wagner, 1977) and intensity (Keely, 2009).
The most common method for estimating CBD is to use the allometrically derived CFL and
mathematically distribute it vertically through the canopy volume either uniformly or nonuniformly. Uniform distribution is done by dividing the CFL by the canopy length or depth (load
over depth) (Reinhardt et al., 2006). The assumption that canopy biomass is uniformly
distributed is unlikely, even for simple stands (Reinhardt et al., 2006) and so methods have been
developed to distribute the load in a more realistic manner. For non-uniform distribution, a
distribution such as Weibull or beta distribution can be used to simulate crown shape and
distribute the CFL weight in a manner that imitates natural form (Affleck et al., 2013). To
calculate CBD, CFL for all trees is summed in incremental height bins and then smoothed with a
running mean. Using this framework CBD is taken to be the maximum value of the running
mean (Reinhardt et al., 2006). Commonly CFL is summed in 1 foot (0.3 m) height bins and
smoothed with a 15 foot (4.5 m) running mean although these methods vary. Hunter et al. (2011)
used 1 m vertical height bins and calculated CBD as the maximum of a 3 m running mean of the
vertical layers. Stocks et al. (2004) used 1 m bins in their assessment of canopy fuel load.
Most fire behaviour models rely on CBD, but these have been found to be very sensitive
to the vertical distribution of crown fuels (Ex et al., 2015). A uniform vertical distribution is
considered unrealistic and has a tendency to predict less severe fire behaviour (Ex et al., 2015).
Reinhardt et al. (2006) compared uniform vertical distribution outcomes for CBD to empirically
calculated CBD values and found the former to be inaccurate for the stands measured (Keyser &

14 | P a g e

Smith, 2010; Reinhardt et al., 2006). It has also been suggested that crown biomass can be
represented using a skewed normal distribution (Gillespie et al., 1994), with mass concentrated
in the center of the crown. The skewed normal distribution is assumed to be influenced by stand
density, which acts to shift biomass up or down depending on tree social position (Keyser and
Smith, 2010). Misrepresentation of vertical distribution (i.e., applying a uniform or some other
assumption) can result in over- or underestimation- of both CBD or CBH, causing fuel
treatments to be ineffective or cost ineffective. While the use of local biomass allometries rather
than regional or national formulas makes the most sense ecologically speaking, the methodology
for fuel accounting may not necessarily align perfectly with fire behaviour models and caution
should be exercised in implementation and interpretation. Affleck et al. (2013) found that profile
curves were highly species-specific and that there was no difference in bias between Weibull and
Beta equations for subalpine fir (Abies lasiocarpa). However, Ex et al. (2015) suggested that
species-specific models are unnecessary to estimate CBD. Instead, they proposed opting for a
positively skewed distribution rather than assuming uniform fuel distribution within crowns to
improve the accuracy of characterizing canopy fuels (Ex et al., 2015).
Several different thresholds for critical CBD have been proposed. Agee (1996) suggested
that for a stand with CBD below 0.10 kg/m3 canopy fire spread would be very limited. This
value is supported by Powell (2010) and Cruz and Alexander (2014). Scott and Reinhardt (2001)
initially defined CBD as the maximum 4.5 m vertical running mean bulk density, then changed it
to a 3 m running mean but gave no reason (Cruz & Alexander, 2010). This method of CBD
calculation is a distinct departure from the way Van Wagner (1977) calculated CBD and likely
will lead to higher CBD values and therefore lower critical spread rate values required for active
crowning to take place (Cruz & Alexander, 2010).

15 | P a g e

Coarse woody debris
Surface fuel loads and characteristics are also very important for understanding fuel
hazard and differences in fuel hazards between stands. Surface fuel loads are defined as all
biomass between the ground surface and two meters in height and include litter, fine and coarse
woody debris, understory plants, lichen and moss (Labenski et al., 2022). Fine particle sized
materials are more readily ignited, while coarser materials take longer to contribute to a fire.
Coarse woody debris has been defined in several ways in the literature. In BC, the
definition of CWD varies. Some define it as non-self-supporting pieces with diameters >7.5 cm
at crossing point of a transect (Resources Information Standards Committee (RISC), 2008;
Forest Analysis and Inventory Branch (FAIB), 2018). Others define it as pieces greater than 7.62
cm in diameter at a transect crossing point (Hyde et al., 2011; Griffin, Simard, and Turner, 2013;
Brown et al., 1974), this value being equivalent to the 3-inch limit used in the Unites States for
the lower limit of the 1,000 hour fuel category. In the context of wildfire fuel, the 1,000-hour
category refers to dead branches, logs or other combustible material that take up to 1,000 hours
to adjust to atmospheric moisture content changes (Woodall & Leutscher, 2003), considered
commonly to be greater than or equal to 3 inches in diameter (7.62 cm). The next smallest
category of fuel is the 100-hour fine fuels which are 1” to 2.9”, the upper limit of which
translates to 7.59 cm (2.9” inclusive of 2.90” to 2.99”).
Small woody debris less than 7.62 cm diameter has a large influence on spread rate and
surface fire intensity. However, CWD has less influence on fire intensity and rate of spread but
can contribute to the development of large and high severity fires (Brown, 2003). Though most
components of CWD are not immediately available as fuel in the flaming front of a fire, some
pieces may burn, increasing the severity of the fire (Hanes et al., 2021). CWD is also an
important part of forest ecosystems and surface fuel load contributions to fire spread rates,
16 | P a g e

transition of fire into the canopy and support of the crown phase of an active fire (Hanes et al.,
2021). Establishing a balance between fire hazard and ecological benefits of CWD can be
challenging, but Brown (2003) suggested that 5 to 10 tons per acre (1 to 2.2 kg/m2) in warm dry
ponderosa pine and Douglas-fir (Pseudotsuga menziesii) stands is optimal for ecosystem
function and 2 to 5.3 kg/m2 for lodgepole pine (Pinus contorta) and subalpine fir stands is an
optimal load for balance (Brown, 2003). CWD loads can vary in natural stands from 0.11 kg/m2
in boreal forests (Krankina & Harmon, 1995) to 14 kg/m2 in old-growth US Pacific Northwest
mixed conifer forests (Spears et al., 2003).
CWD fuel load differentiation is important because smaller fuels have less surface area,
burn more quickly and contribute less to the impact on a site than a larger piece would, while
large pieces remain burning longer. In the Canadian Forest Fire Danger Rating System
(CFFDRS) (Canadian Forestry Service 1987) CWD is merged with other surface fuels and is
represented as an average single layer for the fuel type. To date classes of surface fuels have not
been differentiated in Canada’s CFFDRS, (Canadian Forestry Service 1987), however a project
is being developed to address this issue (Hanes et al., 2021). Recent analyses have shown that
fuel consumption is sensitive to fuel load variation (Kennedy et al., 2020), therefore a better
understanding of local and regional surface fuel loads can improve fuel modelling.
Climate-driven disturbances are contributing to tree mortality globally (Goodwin et al.,
2021). Along with the increased available dead biomass, climate-related deficits and temperature
increases have been found to lead to reduced amount of moisture stored in large dead fuels in
some areas (Goodwin et al., 2021). Reduced fuel moisture content (FMC) can increase
flammability in CWD and standing dead biomass, which in some cases can lead to more energy
being released during fires. Increased heat flux in these circumstances has the potential for

17 | P a g e

positive feedback where fires are able to create their own weather (Goodwin et al., 2021). Fire
intensity refers to the energy released from a fire (Keeley, 2009). As more energy is released by
inclusion of more standing dead and 1,000-hour fuels, wildfires may reach tipping points leading
to catastrophic rank 5 and 6 fire behaviour scenarios (Adams 2013; Goodwin et al., 2021).
Therefore, CWD is an important metric to consider, though not considered immediately
combustible in the flaming front of a fire.
Evaluating fuel load profiles across heterogeneous stand types
The fuel metrics described above are often combined into fuel type groups based on stand
types or on how groups of vegetation behave in fire behaviour in fuel models. In general terms a
fuel type is a fuel complex that is similar enough in attributes and over an area large enough to
produce a particular fire behaviour (Forestry Canada Fire Danger Group, 1992). In the CFFDRS
there are 16 fire behaviour prediction (FBP) fuel types described as “an identifiable association
of fuel elements of distinctive species, form, size, arrangement, and continuity that will exhibit
characteristic fire behaviour under defined burning conditions” (Merrill & Alexander, 1987). The
CFFDRS and the FBP subsystem is a fire behaviour prediction system used commonly in
Canada for wildfire behaviour modelling.
There are 5 major groups of CFFDRS fuel types: Coniferous, Deciduous, Mixedwood,
Slash and Open. Within these 5 groups there are 16 discrete fuel types: C1 to C7, D1, M1 to M4,
S1 to S3 and O1. They are described qualitatively using standard structural and compositional
terminology as well as forest floor thickness and organic layer content. Within each fuel type
stands are assumed to exhibit similar structural and species composition and have the same
representative forest floor thickness and organic layer content (Taylor, Pike, & Alexander, 1996).
In British Columbia’s Northwest Fire Centre (where this study area is located), the fuel types

18 | P a g e

applied to provincial Vegetation resources Inventory (VRI) types are C-1, C-2, C-3, C-4, M3/65, D-1/2, M-1 and M-2 (Perrakis et al., 2018). C-1 is defined by open spruce-lichen
woodlands and is limited to Boreal White and Black Spruce (BWBS) and Spruce-Willow-Birch
(SWB) biogeoclimatic zones where the vegetated density class is sparse. The C-2 boreal black
and white spruce fuel type is applied to mid-elevation interior white (Picea glauca (Moench)
Voss) and hybrid spruce (Picea engelmannii Parry ex Engelmann x Picea glauca (Moench)
Voss) stand types (Perrakis et al., 2018, 9). The C-3 type includes several combinations of fully
stocked mature lodgepole pine stand types but is also assigned to pure and mixed density
Douglas-fir stands from 4 to 12 m in height, open stands of pure Engelmann or interior spruce
(Picea engelmannii or P. engelmanii x glauca), dense pure or mixed western redcedar (Thuja
plicata Donn ex D. Don), western hemlock (Tsuga heterophylla (Raf.) Sarg) or yellow-cedar
(Callitropsis nootkatensis) with specific height and age criteria. C-4 (immature jack or lodgepole
pine) is assigned to forested conifer stands between 4 and 12 m in height and more than 8,000
stems per ha or to stands with more than 60% crown closures, 4 to 12 m in height and more than
34% dead stems. For MPB-killed stands, Perrakis et al. (2018) used C-3 for stands greater than 6
years since mortality and less than or equal to 50% dead and they used C-2 if mortality was
greater than 50% for stands >5 years since mortality. Although they include stands with 0-5
years since attack in their algorithm, these are not currently present in the study area. The M-2
and M-2 fuel types are a blend of C-2 and D-1 fuel types and the D-1 and D-2 fuel types are
attributed to broadleaf leafless and green (Perrakis et al., 2018). These are all fit to vegetation
type polygons using the existing and often out of date provincial VRI data. The fuel type
assignment process is limited to the resolution of the VRI polygons. In addition, vegetation types
not included in the original CFFDRS studies are potentially incorrectly assigned to the wrong

19 | P a g e

fuel type class. Because of the limits of VRI, and the cost of mitigation, there is a need for finer
resolution and more accurate estimates of fuel loading and fire risk within different stand types
common to forest managers in north-central BC and these must be specific to local species and
stand types. Greater accuracy in estimating fuel loading would allow locally relevant fire risk
assessments, allow for the legacy of mountain pine beetle killed pine stands to be better
represented in fire behaviour models, and would allow variability of fuel loads to better represent
real forest structure and the impact of recent disturbances.
Determining how different stand types impact fuel hazard and fire risk is a pressing
challenge due to the increasing prevalence of catastrophic rank 5 and 6 fires. Mitigating fire
hazard must incorporate site- and stand-level interventions and also consider wider landscape
fuel connectivity. In addition, longer-term silvicultural planning should include an evaluation of
landscape-level fuel hazard to alleviate the potential consequences of silvicultural decisions at
stand establishment and onwards. Promoting broadleaf- and mixedwood stands is a potential fire
risk mitigation tool, but more quantitative analysis of the fuel characteristics of these stand types
is needed.
Vegetation classification is a tool used to describe, qualify or quantify and compare
variations and responses particular to areas of interest and application of management regimes
(De Cáceres et al., 2019). Forest stands are commonly distinguished by species composition and
age categories (age classes) in BC but can also be characterized by numeric metrics such as
merchantable volume, basal area, or biomass per unit area. When considering the attributes of
vegetation in a wildfire context, often it is the behaviour of the vegetation complex when
exposed to fire and fire weather that is of interest. In Canada the CFFDRS fuel type designation
is a qualitative categorization of vegetation types that have analogs for fire behaviour. A default

20 | P a g e

CFL and CBH value is applied for each vegetation type based on their expected fire behaviour
(Forestry Canada Fire Danger Group, 1992). In north-central BC, there are many stand types
which may or may not be similar to CFFDRS fuel types, but commercial forestry in this region
focuses traditionally on a limited group of mature stand types when prescribing management.
These are conifer stands, deciduous stands and mixed stands. Conifer stands are either pure
stands of one species or a mixture of lodgepole pine, subalpine fir., white spruce (or hybrid
spruce, and to some extent, Douglas-fir and black spruce (Picea mariana). Deciduous stands are
primarily trembling aspen (Populus tremuloides) but may contain paper birch (Betula
papyrifera) and black cottonwood (Populus trichocarpa). Mixed stands have varying proportions
of aspen, birch, cottonwood, and some conifer(s). A relatively new stand type category is what
would have been lodgepole pine and aspen mixedwoods but are characterized by almost 100%
mortality of lodgepole pine in varying stages of collapse, mixed with aspen co-dominants and a
recently invigorated conifer understory. Despite their substantial global coverage—
approximately 16% of forest area in Canada (Statistics Canada, 2018) and approximately 23% in
Europe (De Cáceres et al., 2019)—quantification of mixed forest attributes is limited (Bieng et
al., 2006).
Mixedwood stands have been touted as better for biodiversity than pure conifer stands
(Cavard et al., 2011), however there are mixed results on whether they have a positive or
negative effect on fuel hazard (Hély, Bergeron, & Flannigan, 2000). Some mixed stands have
been shown to have intermediate values for predicted rate of spread, head fire intensity, and area
burned (Hély et al., 2000), but it is uncertain if these are comparable to stands consisting of dead
pine with aspen and dense conifer. Small differences in fuel hazard metrics can significantly
change fire behaviour model outcomes (Penman et al., 2022).

21 | P a g e

Stand type categories differ from region to region, even within central BC. Some studies
split stand types into four distinct categories: Deciduous, deciduous-leading mixedwood,
conifer-leading mixedwood and conifer, with a percent conifer of <25%, 25-50, 51 to 75% and
>75% respectively (Hély et al., 2000; Kabrick et al., 2017). Others lump stands having less than
75% deciduous or conifer dominance into a mixedwood category. There has been interest in
potential benefits of mixedwood stands, such as their potential resilience to disturbance
(Gonçalves et al., 2017), high foliar moisture content, and value for biodiversity and climate
change mitigation and adaptation, particularly in urban interface areas (Kabrick et al., 2017).
Mixedwood stand definitions have also relied on a site’s potential to support multiple tree
species rather than a percentage-based definition of mixedwood (MacDonald, 1996; Andison &
Kimmins, 1999).
The most recent outbreak of mountain pine beetle (Dendroctonus ponderosae; MPB) in
interior BC has created large swaths of stands that may also differ with respect to fuel loading
and the contribution that they make to wildfire risk. Stands attacked by mountain pine beetle may
have different CWD volumes and overstories that still contain live and dead standing lodgepole
pine. Because MPB targets mature lodgepole pine trees, the stands that have been attacked will
also have modified species and age distributions. These MPB-induced changes in the stand
structure, and the development trajectory that these stands will take, also have the potential to
alter the fuel load characteristics of those stands. Bright et al. (2017) studied similar stands in
four different time-since mortality cohorts, and assessed fuel load within them but indicated that
more research was needed in long dead (i.e. ‘grey’ stage) stand fuel metrics (Bright et al., 2017).

22 | P a g e

Research Objective
The objective of this chapter is to compare and contrast the fuel load characteristics
(CFL, CBD, CBH, canopy length) and CWD of four commonly occurring mature forest stand
types in north-central British Columbia. Stand types I assessed were live conifer (Interior hybrid
spruce, subalpine fir, and lodgepole pine), dead conifer (primarily dead mountain pine beetlekilled lodgepole pine >5 years since mortality), deciduous (trembling aspen), and mixedwood
stands (mixtures of live or dead conifer and deciduous).
This research investigates the following four questions:
1.

Are the four stand types fundamentally different with regard to their basic stand structure
and composition (height, diameter at breast height (DBH), basal area, tree density (in
stems per hectare, SPH),

2.

Do the four different stand types exhibit differences with regard to fuel load as
represented by CFL, CBD, CBH, crown length, and CWD?

3.

Do dead pine stands represent a very high fuel hazard compared to other stand types?

4.

Do mixedwood stands represent a balanced alternative to deciduous stands with respect
to maintaining timber value while also reducing fuel hazard?

Materials and Methods
Study Area
The study area for this research included a 900 km2 community forest near Burns Lake,
British Columbia, Canada, as well as a buffer surrounding the community forest encompassing a
total area of approximately 1,800 km2 (Figure 1). The Chinook Community Forest was
established in 2016 as a partnership between six First Nations and two municipal governments
and is managed as an area-based forest tenure (https://chinookcomfor.ca/). The majority of the
23 | P a g e

study area is within the provincial timber harvesting land base (THLB) and is managed under an
area-based license by the Community Forest.
North-central BC is in the Montane Cordillera ecozone, experiencing a mixed-severity
fire regime (natural disturbance type 2, NDT2), with high-severity stand-initiating fires at
intervals averaging approximately 250 years (BC MOF & BCMELP 1995). The study area
forests are dominated by hybrid spruce and Engelmann spruce, lodgepole pine, subalpine fir and
trembling aspen and are ecologically classified as primarily Sub-Boreal Spruce dry cool
(SBSdk), and moist cold (SBSmc) with a small component of Engelmann Spruce Subalpine fir
moist cold (ESSFmc) biogeoclimatic subzones. In the late 1990’s a mountain pine beetle
outbreak occurred which affected the interior of the province by the 2010’s, impacting an
estimated 18 million hectares (close to 20% of BC’s land area) and killing 723 million cubic
meters of merchantable pine volume (Corbett et al., 2016). The lodgepole pine in the study area
(Figure 1) were infested in the 2010’s and mortality in most mature pine was complete by
approximately 2015. The stand types characterized in this study were delineated using VRI
(Figure 2).

24 | P a g e

Figure 1. Study area overview map. There are four management units (pink - Babine, green Hannay, blue - Roselake and purple - Southside) within the study area . The extent of LiDAR
coverage (grey crosshatch) and plot locations by year of data collection are shown across the
study area.

25 | P a g e

Figure 2. Study area with VRI characterized stand types indicated.
Field Sampling
Field sampling took place in 2018 and 2019. Sites were sampled within the study area in
four stand types: Conifer, Deciduous, Mixed and Dead Conifer. Sampling in 2018 was conducted
on a grid, where 83 sites were sampled. Sampling in 2019 was designed to capture less well
represented stand types than were sampled in the grid sample.
To estimate species composition of stands for sampling, the Vegetation Resources
Inventory (VRI) data were used to stratify stands into four categories based on species cover
percent. Stands with >70% conifer and <50% stand dead % were classified as ‘Conifer Live’
(CL). Stands with >70% deciduous were classified as ‘Deciduous’ or deciduous live (DL).
26 | P a g e

Stands with <=70% conifer and <=70% deciduous and less than 50% dead percent were
classified as ‘Mixedwood Live’ (ML). Stands with >50% dead percent and conifer >50% were
classified as ‘Conifer Dead’ (CD). There were very few dead deciduous or dead mixed stands by
area and so these were grouped with Deciduous and Mixedwood, respectively. Over half of the
767 dead standing trees were dead lodgepole pine. The remainder were a mixture of subalpine
fir black spruce, and interior hybrid spruce. Willow was not included in the basal area stand
categorization process because for the most part this species consisted of numerous small
diameter stems similar to alder. Fallen trees were included in the basal area stand type
categorization. An overview of the stands selected is given in Table 4, and plot level summaries
are shown in appendices 1 and 2.
Plot location
There were 101 plots captured for ALS ground truthing purposes by the Community
Forest in the study area prior to the commencement of this project. These ground plots were
assessed for use in this study and sixteen were dropped from the analysis for either having no
trees or due to errors in the data, leaving 85 plots. After stratifying the study area and
surrounding buffer by VRI stand types, it was determined that 68 plots were located in live
conifer stand types, 8 were located in dead conifer stands, 3 were located in mixedwood stands
and 6 were located in deciduous stands. Using statistical sample size calculations based on plot
basal area per hectare I determined that more plots would be required in the non-live conifer
stand types. Five additional plots were sampled in dead conifer dead conifer stands, 11 in
mixedwood stands and 8 in deciduous stands. In the 2018 samples, no CWD data was collected.
For this reason, an additional 9 plots with CWD transects, were collected in live conifer stands.
Plot selection was buffered to be within 300m of road access. Within the buffered area plot

27 | P a g e

location was randomly selected using the random point generator function in the Cran-R package
SP (Bivand, Pebesma, & Gomez-Rubio, 2013).
Tree data collection
Field measurements for standing trees were collected as per Change Monitoring
Inventory Ground sampling procedures (Forest Analysis and Inventory Branch, Ministry of
Forests, Government of British Columbia, 2018) in 10 m radius circular plots (314.16 m2 in
area). In each 10 m radius plot and for all trees >4 cm DBH I recorded species, DBH, tree height,
height-to-live crown, mortality (live or dead), standing or fallen. The total number of plots used
in analysis was 118 with 76 Conifer Live (CL) plots, 13 Conifer Dead (CD) plots, 15
Mixedwood (ML) plots, and 14 Deciduous (DL) plots.
Tree heights and height-to-live crown were measured using a VERTEX IV-360
electronic tree measuring device and DBHs were measured using calipers.
CWD data collection and biomass estimation
CWD data was collected from 33 plots in 2019. CWD surface fuels greater than 7.5 cm
DBH were collected following the line intersect method described by Van Wagner (1968). Two
orthogonal 10 m transects were established per plot, with the bearing of the first transect
randomly determined.
All pieces greater than 7.5 cm in diameter intersecting the transects were sampled.
Species, intersect diameter, and diameter and elevation at both ends were collected, as were
piece length and decay class (1-5). CWD volume and biomass were calculated as per Waddell
(2002) according to size and decay class for pieces not self-supporting (Waddell 2002). Specific
gravity was assigned as per Miles (2009). Volume per unit area was calculated as per Vries
(1973):

28 | P a g e

Volume (m)ଷ ⋅ (ha)ିଵ =

ߨ ଶ (‫ܦ‬ௌଶ + ‫ܦ‬௅ଶ )
16‫ܮ‬

Eq. 1

In this equation, volume in m3 represents the coarse woody volume per unit area, π
represents the mathematical constant pi, DS represents the small-end diameter of the log, DL
represents the large-end diameter of the log, and L represents the total horizontal length of the
transect.
Biomass is calculated as:
Biomass ൫kg ⋅ (ha)ିଵ ൯ = ܸ ⋅ 1000 kg ⋅ mଷ ⋅ SpG ⋅ DCR

Eq.2

In this equation, biomass represents the coarse woody biomass per unit area, where V is
volume in cubic meters per hectare, SpG is the specific gravity by species, and DCR is the decay
class reduction factor of 1 to 4 (1.00, 0.84, 0.71, and 0,45) for softwoods and 1.00, 0.78, 0.45
and 0,42 for hardwoods (Waddell, 2002) (Table 1). Unidentifiable species were assigned an SpG
of 25.4.
Table 1 Declay class reduction factors by species (n represents unidentifiable species).
Decay Class

Species Group

Reduction Factor

1

Pl,Sx,At,Ahw,n

1.00

2

Pl,Sx,n

0.84

2

At,Ahw

0.78

3

Pl,Sx,n

0.71

3

At,Ahw

0.41

4

Pl,Sx,n

0.45

4

At, Ahw

0.42

Stand Metrics
Mean tree height and DBH were calculated as the mean height of all trees and the mean
DBH of all trees in the plot respectively. Basal area per hectare was calculated as the sum of tree

29 | P a g e

basal areas in the plot multiplied by 31.83. The stems per hectare metric was calculated by
multiplying the stem count by the same per hectare value.
The merchantable volume per hectare (Eq.3) was calculated for commercial species using
the interior B.C. equations in Nigh et al. (2016) and based on DBH and total tree height
subtracting the volume below stem height of 0.3 m and above the point where stem becomes less
than 4 cm diameter (Nigh et al., 2016).
Eq.3

‫ ݁ = ݒ‬௕బ × DBH௕భ × ht௕మ

where ܾ଴ , ܾଵ , and ܾଶ are species-and-region-specific constants from Nigh et al. (2016)
and summed to the plot level then converted to a per hectare value.
‾ , is calculated as:
The mean crown length, denoted as ‫ܮܥ‬
‾ = ଵ ∑௡௜ୀଵ൫ℎ௜ − ℎlive,௜ ൯
‫ܮܥ‬
௡

Eq.4

‾ represents the mean crown length, ݊ is the total number of trees with
In this equation, ‫ܮܥ‬
live crowns, ℎ௜ represents the tree height of the ݅-th tree, and ℎlive,௜ represents the height-to-live
crown of the ݅-th tree.
All analyses were performed using R Statistical Software (R Core Team, 2021). There
were 31 live trees with missing height to live crown measurements and 767 dead non-pine trees
with missing CBH measurements. Rather than drop the trees, simple species-specific linear
models were developed using DBH and total height to estimate height-to-live crown. The linear
models were fit using a zero intercept (Table 2). The data used for fitting the model was the live
trees with recorded height-to-live crown. Including a stand density metric in the linear models
was considered; however, at plot level, no discernable difference in CBH or CFL was found.
Both stems per hectare (SPH) and stand density index (SDI)were assessed as additional variables
and compared using a likelihood ratio test (Zeileis & Hothorn 2002). There was no difference in
30 | P a g e

the linear models with or without SPH or SDI for lodgepole pine or aspen models. There were
significant differences found (p < 0.05) for the subalpine fir and spruce linear models, but upon
comparing final plot-level results for CFL and CBH, no differences were found.
Table 2. Summary of height to live crown relationships derived using multiple regression
equations with the intercept forced through zero.
Species Name

Height1
DBH2 Coefficient
Coefficient

Adjusted RF Statistic & DF
squared

subalpine fir

0.277

0.131

0.86

2355 on 2 and 751 DF

spruce

0.466

-0.210

0.80

4371 on 2 and 2168 DF

black spruce

0.690

-0.330

0.72

104.5 on 2 and 78 DF

pine

0.688

-0.229

0.89

6084 on 2 and 1431 DF

aspen

0.781

-0.061

0.97

1.069e+04 on 2 and 676
DF

poplar

0.596

0.000

0.94

544.4 on 2 and 60 DF

willow and alder

0.494

-0.070

0.78

339.5 on 2 and 186 DF

1.

Height measured in meters; 2. DBH measured in centimeters

The crown length used to calculate CBD distribution was derived from total tree heights
and the modeled height-to-live crown values. For the most important species in the study area, r2
values of >0.8 were achieved.
Canopy Fuel Load Calculations and Estimations: Biomass, CFL, CBD
Available crown fuel, canopy bulk density and canopy base height were estimated using
allometric biomass models from Ung et al. (2008).
Canopy fuel load (CFL) kg/m2
Canopy fuel load is the amount of fuel in the canopy that is available in the flaming front
of a fire. Although a large proportion of plant biomass is in the canopy, not all of it is considered
available. Wood such as tree boles and large branches may make up as much or more than 80%
of total canopy weight (Keane et al., 2015), it is the finer material that is considered immediately
available to burn (Van Wagner, 1977). This finer material is foliage and small branches. A
31 | P a g e

common method to estimate available fuel load is to sum all of the foliage and half of the 0 to
0.6 cm branch biomass (Scott & Reinhardt, 2001). Brown (1978) biomass equations are often
used for fuel load studies in the US; however, Brown’s equations are unsuited to BC species and
this study because the sample data DBH ranges in that study do not include all ranges for DBHs
in this study. In addition, the sampled trees were in Utah, much further south than the study area
for this research. Ung et al. (2008) biomass equations are a better fit and are inclusive of BC
species and data was collected relatively recently and merged with Canadian national allometric
equations (Lambert, et al., 2005) however they do not break down the branch component into
subgroups representing average moisture time-lag classes as Brown’s equations do. Biomass
equations for subsets of branch components are available for Canadian tree species such as
Jenkins et al. (2004), and Chojnacky, et al. (2014), however these are not ‘additive’ equations
and lack consistency across species and regions. Standish et al. (1985) presented linear equations
for BC commercial tree species and provided three subcategories of branches: <0.5 cm, 0.5 to
2.5 cm, and >2.5 cm. The closest to the small branch category used by Scott and Reinhardt
(2001) (< 0.6 cm) would be the <0.5 cm category used by Standish et al. (1985). A comparison
of estimates between Ung et al. (2008) and Standish et al. (1985) equations indicated similar
total biomass for a given DBH and tree height, with foliage slightly higher in the Standish et al.
estimate, and total branch biomass approximately 1/3 higher than Ung et al. estimates. With
small branch category available it appears as though the Standish et al. equations estimate this
category to be about 34% of total crown, understanding that total crown is higher than estimates
by Ung et al. due to both foliage and branch biomass estimates being higher. Alternatively, using
interpolation on Brown’s (1978) figure 11 for Douglas-fir the proportion of the live crown in
Brown’s study for small branch is approximately 20% for a 20 cm DBH Douglas-fir, and the

32 | P a g e

foliage plus small branch is approximately 60% of total crown. Johnson, Woodard, and Titus
(1990) created regression equations for branch classes similar to Standish et al. (1985) for
lodgepole pine and white spruce in the Boreal Forest Region in Alberta. The equation used for <
0.5 cm diameter estimates approximately 45% of total crown weight and 29% of live crown
weight (~29%) for a given 28.6 cm lodgepole pine tree, comparable to Standish et al. (1985). For
a white spruce, the proportion is much higher for a given spruce tree of the same diameter, with
small branch <0.5 making up 78% of live crown and 65% of total crown. Determining what
proportion of Ung’s branch to allocate as ‘small’ is challenging, but assigning half of it (0.5) as
small branch to include as effective crown fuel seems to be a reasonable and conservative
estimate based on both Brown (1978) proportions and those of Standish et al. (1985) and of
Johnson et al. (1990), particularly since Ung does not include a dead foliage component, which
as demonstrated by Johnson’s inclusion of total versus live crown weights, makes up a
considerable amount of biomass in the crown of lodgepole pine and white spruce in similar stand
types to the Burns Lake study area. For DBH values between 4.5 and 50 cm, the average
proportion of small branch <0.5 cm diameter to live crown biomass has been estimated to be 44
% (Figure 3). In addition, Call and Albini (1997) suggested that approximately 65% of fuels less
than 6 mm at 100 percent moisture content would be consumed in the flaming front of a fire, and
Sando and Wick (1972) included all foliage as well as branch less than 2.5 inches (6.35 cm) in
diameter. Thus, for this study, total effective crown biomass includes all foliage and one-half of
the branch biomass estimated using Ung et al. (2008) allometric equations. Although some
research considers dead non-pine as effect canopy fuel, (Curtis, 2008) for this research, non-pine
dead trees were considered to have no effective canopy fuel (Cruz, 2003). For dead pine trees,
crown biomass is assumed to be negligible due to time-since-mortality of greater than 5 years

33 | P a g e

(Bright et al., 2017) and so both foliage and branch estimates are negated if species is pine and
status is dead. Figure 3 shows the proportion of small branch biomass for lodgepole pine (left)
and white spruce (right) using Johnson et al. (1990) biomass equations. For lodgepole pine, the
proportion of the total crown is much less than for white spruce as shown by the blue line.

Figure 3. Estimation of small branch biomass. Proportion of small branch <0.5 cm for DBH
values 4.5 to 50 cm using Johnson et al. (1990) biomass equations for lodgepole pine and white
spruce.
Loomis and Roussopoulos (1978) assessed branch biomass by size classes for trembling
aspen in northeastern Minnesota and proposed a relationship based on DBH. The proportion of
branch biomass for the 0 to 0.6 cm size class is for an average diameter aspen tree in our study is
approximately 24% and goes down as DBH increases.
Canopy fuel load was estimated as the sum of the plot foliage biomass and half of the
branch biomass. Biomass for each tree was estimated using the equations from Ung et al. (2008)
for foliage, branch, stem and wood. The sum of the foliage and half of the branch biomass were
considered the available crown fuel load. Using a framework that approximates the methods used
34 | P a g e

by Scott and Reinhardt (2001) and Bright et al. (2017), I used half of the total branch biomass
from Ung et al. (2008) to estimate available crown fuel load for all live trees and recently intact
dead trees. Most pine trees were dead as a result of an MPB attack approximately 7 to 10 years
prior to field sampling. Because of time-since-mortality, dead pine trees were assumed to contain
no available crown fuels (Bright et al., 2017).
௡

Eq. 5

1
1
Canopy Fuel Load (kg/m²) = ൭෍ ൬foliage௜ + ⋅ branches௜ ൰൱ ×
2
ߨ × 10ଶ
௜ୀଵ

In this equation, canopy fuel load (kg/m²) represents the total canopy fuel load per unit
area for a 10 m radius plot. The foliage and branch biomass of each tree in the plot are calculated
using the following equations:

ܾ
⋅ ൫‫ ܦ‬bfoliageଶ ൯ ⋅ ൫‫ ܪ‬bfoliageଷ ൯, if ‫ == ܦܮ‬′݈′
foliage = ൜ foliageଵ
0,
otherwise 0

Eq. 6

ܾ
⋅ ൫‫ ܦ‬bbranchesଶ ൯ ⋅ ൫‫ܪ‬bbranchesଷ ൯, if ‫ == ܦܮ‬′݈′
branches = ൜ branchesଵ
0,
otherwise 0

Eq. 7

where LD represents live or dead, and l represents live. The coefficients ܾfoliageଵ , ܾfoliageଶ ,
ܾfoliageଷ , ܾbranchesଵ , ܾbranchesଶ, ܾbranchesଷ are obtained from Ung et al. (2008). Biomass equation
coefficients and equations are shown in Table 3.

ଵ

The term గ×ଵ଴మ is the conversion factor to adjust the total sum of foliage and half branch
biomass for the sum of all trees in a 10m radius circular plot, converted to a per m2 basis.

35 | P a g e

Table 3. Biomass equations (from Ung et al. (2009), where DBH is diameter at breast height in
cm and H is total tree height in meters.
Tree Species

Foliage expression

Branch expression

Black spruce

0.2078(DBH)^(2.5517)H^(-1.3453)

0.0405(DBH)^(3.1917)H^(-1.3674)

Lodgepole pine

0.0769(DBH)^(2.6834)H^(-1.2484)

0.0285(DBH)^(3.3764)H^(-1.4395)

Black cottonwood

0.0224(DBH)^(1.8368)H^(0)

0.0131(DBH)^(2.576)H^(0)

Subalpine fir

0.0509(DBH)^(2.9909)H^(-1.2271)

0.0265(DBH)^(3.6747)H^(-1.5958)

Trembling aspen

0.0284(DBH)^(1.602)H^(0)

0.015(DBH)^(2.9068)H^(-0.6306)

Paper birch

0.1361(DBH)^(2.2978)H^(-1.0934)

0.0253(DBH)^(3.1518)H^(-0.9083)

Engelmann spruce

0.1832(DBH)^(2.4144)H^(-1.0948)

0.0322(DBH)^(2.8961)H^(-0.9203)

Ahw (Willow and Alder)

0.0869(DBH)^(1.8541)H^(-0.5491)

0.0448(DBH)^(2.6855)H^(-0.5911)

Canopy bulk density (CBD) kg/m3
The individual tree crown fuel load is distributed in vertical 1-meter bins within the
respective tree crown using the measured height to live crown according to a species groupspecific distribution method (deciduous or conifer) similar to Sando and Wick (1972). There are
two basic methods of calculating CBD: the maximum running mean method and the box method
(Scott, 2012). In this research, the maximum of the running mean was used because this provides
finer detail and because it is more suitable when there is a non-uniform vertical distribution of
fuel load within the canopy. Below I describe the method used in detail. A cumulative beta
distribution function (CDF beta) was then applied for all species as per Affleck et al. (2013). For
elevation bin i, the cumulative probability pi is calculated as:
‫݌‬௜ = ݂(‫݊݅ܤ ݊݋݅ݐܽݒ݈݁ܧ‬௜ ) = ‫ܨܦܥ‬஻௘௧௔ (

Eq.8

‫݊݅ܤ ݊݋݅ݐܽݒ݈݁ܧ‬௜
, ߙ, ߚ)
40

where Elevation Bini ranges from 0 to 40 and α and β are parameters of the beta
distribution. This framework returns the percent of the total canopy biomass that is expected to
36 | P a g e

occur in each elevation bin. These values are then multiplied by each tree’s total canopy
biomass in order to estimate foliage biomass within each height bin.
For all spruce, subalpine fir, lodgepole pine and deciduous trees, the available crown fuel
was distributed through the crown length using a beta distribution (Eq. 8) in 1 m intervals
beginning from the base of each height bin (Affleck et al., 2013). Others have used speciesspecific distribution methods such as a uniform vertical distribution for deciduous and pine
species and conical for others. However species-specific vertical distributions have been found to
be more representative of fuel load distributions. The parameters for the beta distribution were α
= 1.12 and β = 1.7 (Affleck, et al 2013). When each tree has had its CFL distributed within its
crown, these individual crown bulk density bins were summed into plot level canopy bins (i.e. all
of the 1 m bins from each tree are added to make a 1 m canopy bin showing the CBD from 1 to 2
m and so on upwards in height (Figure 4). A 3 m running mean was then calculated using the
plot level 1 m canopy bins. The maximum value of all bins is considered the canopy bulk density
in kg/m3 for the plot (Schoennagel et al., 2012). This is a slightly different method than others
have used (Scott & Reinhardt, 2001), where others use a 4.5 m running mean of 0.3 m height
bins. Both methods (4.5 m running mean and 0.3 m height bins, or 3 m running mean and 1 m
height bins) involve smoothing the data, but they differ in the size of the moving average
window and the initial granularity of the data (bin size). Larger windows and bin sizes will result
in smoother representation of the data. The CBD is then taken to be the maximum value of fuel
load of the running mean. It is taken at the maximum point because crown fire spread is best
approximated by the most dense canopy layer rather than the average density (Scott, 2012).
Figure 4 demonstrates the accumulation of tree biomass in a plot into 1 m elevational bins. Each
colour represents a separate tree in the plot with available crown fuel, and the dashed vertical

37 | P a g e

line represents the 0.037 threshold discussed below. The CBD for this plot is 0.04 kg/m3, the
peak of the curve.

Figure 4 One plot vertical fuel distribution example where the biomass of each tree is a separate
colour and the peak of the vertical distribution is the CBD. Some research identifies 0.037 as the
beginning of the effective canopy load (Van Wagner, 1977). This line is marked on the plot as a
dashed vertical line.

Height of maximum CBD (m)
The height of the maximum CBD was also calculated as the height in meters of the CBD
calculation described above. This is the height bin where the maximum CFL occurs in the
vertical profile, or the peak in fuel in the vertical profile.
Canopy base height metrics (CBH, CBH-0.011, CBH-0.037) (m)
Three canopy base height metrics were calculated: Mean plot crown base height and two
threshold-based points which are the points at which the fuel load in the canopy reaches either
38 | P a g e

0.011 kg/m3 or 0.037 kg/m3. The mean crown base height was calculated as the average height to
live for the stand type. The other two metrics were calculated from the CFL height bin values
where the fuel in the vertical profile of height bins exceeds either 0.011 kg/m3 or 0.037 kg/m3.
Canopy Length Metrics (CL, CL-0.011, CL-0.037) (m)
Three canopy length estimates were calculated. The first was based on tree data and
calculated as the mean tree height - mean height to live crown. The other two are based on the
0.011 and 0.037 CBD thresholds, where CBH 0.011 is the length in meters between where the
stand begins to be > 0.011 kg/m3 and the last point at which CBD is > 0.011 (start of 0.011 and
end of 0.011), and the same for the 0.037 kg/m3 threshold.
Statistical Analysis
I used the R statistical programming language to test for differences in the estimates of
CFL, CBD, CBH, and CWD fuel loads among the four stand types via Welch’s analysis of
variance (ANOVA; Welch 1947) with pairwise comparison of means using the Games-Howell
non-parametric post-hoc test (Lee & Lee 2018). R packages used for statistical analyses were
Agricolae (de Mendiburu, 2021), rstatix (Kassambara, 2023), and PMCMRplus (Pohlert, 2023).
Welch’s ANOVA is a test for multiple comparison of means when the assumption of
homogeneity of variances is violated (Delacre et al., 2019). It’s a modified one-way ANOVA
that is robust to the assumption of equal variances. Non-parametric methods may also be used for
this situation but are not as powerful as Welch’s ANOVA. I selected Welch’s ANOVA method
because variances were not equal. This analysis also controls the nominal type I error the best
when data are heterogeneous, normal, and balanced (Liu 2015). Welch’s ANOVA has also been
shown to be effective for a small number of groups and was designed to test the equality of
group means in situations where there are more than two groups and where sample sizes are

39 | P a g e

small (Liu 2015). To conduct post-hoc analysis I used a Games-Howell non-parametric post-hoc
test. The Games-Howell test does not assume equal variances or sample sizes (Lee & Lee 2018).
Results
Basic Tree and Stand Metrics
Summary statistics for stand structure of the four stand types was calculated (Table 4).
This table shows the mean, standard deviation and ranges of basal areas in each stand type for
each tree component and lists the plots in each type. The age of the tallest tree in each plot (Top
height tree age)was also collected for the majority of plots. This is shown in the table as well, in
addition to the VRI projected age for the leading species.

40 | P a g e

Table 4 Plot summary by basal area stand type.
Stand type

Plot Ids

N plots

Total ba

Live
pine ba

Dead
pine ba

Live
other
conifer
ba

Dead
other
conifer
ba

Deciduous
ba

Willow and
Alder ba

Live
pine %

Live
other
conifer
%

Dead
pine %

Dead
other
conifer
%

Deciduous
%

VRI Age
(years)

Top
height
Age
(years)

Mean (SD) Range: min-max

Conifer Live

CD123-A, CD337, CD77, CD93,
M1, M15, M439A, M472, MN3-2,
N10, N101, N102, N103, N107,
N108, N109, N110, N111, N112,
N113, N115, N120, N121, N131,
N133, N135, N136, N139, N15,
N16, N28, N29, N30, N35, N39,
N40, N41, N44, N45, N46, N47,
N48, N50, N52, N54, N55, N56,
N57, N61, N62, N63, N65, N66,
N67, N68, N70, N72, N73, N76,
N77, N78, N80, N82, N83, N84,
N85, N86, N88, N89, N9, N90,
N94, N95, N96, N98, N99

Conifer Dead

CD10, CD56, CD62A, M13, M41,
N104, N105, N118, N137, N24,
N27, N53, N93

Mixed

CD445A, D348, M26-A, M28,
M348A, M369, M433A, M462a,
M486, MN9-517A, MN9-518,
N122, N134, N42, N51

15

42.33
(15.19),
Range:
11.2 71.62

Deciduous

CD151A, CD383, D101-A, D11,
M23, M449, M476, N123, N124,
N36, N43, N58, N75, M355

14

32.27
(16.5),
Range: 8.8
- 61.69

76

27.82
(20.01),
Range: 0.2
- 89.17

6.5
(8.88),
Range:
0 - 34.2

3.02
(5.86),
Range: 0
- 27.58

14.83
(15.63),
Range: 0
- 80.79

2.05
(4.22),
Range: 0
- 28.53

0.91
(1.98),
Range: 0 8.63

0.5 (2.21),
Range: 0 18.53

29.09
(34.43),
Range: 0
- 100

47.77
(29.98),
Range: 0
- 96.98

8.48
(14.26),
Range: 0
- 48.97

8.4
(17.52),
Range: 0
- 95.15

3.73
(7.46),
Range: 0 28.71

86 (62)
Range:
9-295

66 (55)
Range:
9-262

13

24.17
(17.9),
Range:
0.53 62.14

2.63
(3.41),
Range:
013.22

15.98
(12.24),
Range:
0.53 40.96

4.85
(5.3),
Range: 0
- 19.6

0.15
(0.27),
Range: 0
- 0.9

0.57
(0.86),
Range: 0 2.78

0 (0.01),
Range: 0 - 0.06

14.18
(15.46),
Range: 0
- 41.07

14.69
(11.14),
Range: 0
- 31.55

68.95
(14.8),
Range:
52.44 100

0.4
(0.76),
Range: 0
- 2.42

1.76
(2.65),
Range: 0 8.46

88 (48)
Range:
9-188

60 (31)
Range:
3-117

2.47
(4.08),
Range:
014.25

3.99
(7.32),
Range: 0
- 23.83

14.16
(7.6),
Range:
2.22 28.12

1.75
(2.82),
Range: 0
- 9.49

18.72
(9.25),
Range:
4.33 43.94

1.24 (2.52),
Range: 0 - 9.2

5.7
(8.35),
Range: 0
- 28.7

35
(17.52),
Range:
4.47 68.54

8.5
(13.4),
Range: 0
- 36.88

4.69
(7.53),
Range: 0
- 25.86

43.4
(9.46),
Range:
30.47 61.51

110 (45)
Range:
11-168

79 (32)
Range:
52-125

0.3
(0.76),
Range:
0 - 2.74

0 (0),
Range: 0
-0

3.07
(4.41),
Range: 0
- 13.62

0.09
(0.24),
Range: 0
- 0.86

26.31
(15.83),
Range: 0 57.84

2.52 (4.07),
Range: 0 10.32

1 (2.2),
Range: 0
- 6.87

6.98
(8.67),
Range: 0
- 23.47

0 (0),
Range: 0
-0

0.23
(0.65),
Range: 0
- 2.32

76.87
(28.42),
Range: 0 100

80 (35)
Range:
11-128

68 (29)
Range:
32-105

41 | P a g e

Based on the results of a Welch’s ANOVA, there is a statistically significant difference in
the mean tree height among the four stand types (F(3,922) = 130.9, p < 0.05, Table 5, Figure 5).
Games-Howell post-hoc analysis indicated the mean tree height in Conifer Live (CL) stands was
lower than the other stands (Figure 5). However, there were no significant differences in tree
height between Conifer Dead (CD) and Mixedwood Live (ML) stand types (p = 0.11), or
Mixedwood Live (ML) and Deciduous Live (DL) stand types (p = 0.57). Trees in the DL stands
were also significantly smaller than trees in CD stands (p < 0.01,).
Additionally, mean tree diameter at breast height (DBH) displayed variations across stand
types, with CD and ML stands having the largest mean DBH, while CL stands exhibited the
smallest mean DBH (Figure 6). Welch’s ANOVA confirmed significant differences in mean
DBH among all stand types (Welch’s F: 127, p: < 0.05; Table 5), except for CD and ML stands,
which showed no statistical difference (Table 5). These findings emphasize distinct patterns in
mean tree DBH among the stand types, underscoring the impact of stand type on tree size and
distribution.

42 | P a g e

Table 5 Stand metrics from empirical data plots collected in 2018 and 2019. Tree Height and DBH are reported for individual trees,
while BA, SPH, Volume and Conifer Volume represent plot level data.

43 | P a g e

Figure 5. Tree height (with quartiles and range included as box and vertical line) for each of the
four stand types are shown (CL Conifer Live, CD conifer Dead, ML Mixedwood Live, DL
Deciduous Live). Plots are divided by tree component with the left panel showing all trees
(purple), only live conifer in the center left panel (green), only dead conifers in the center right
panel (grey), and only deciduous individuals shown in the right panel (yellow). Statistical
analysis (Table 5) was performed using “All” trees and reflects observed differences in the left
panel (purple).

44 | P a g e

Figure 6. DBH (with quartiles and range included as box and vertical line) for each of the four
stand types are shown (CL Conifer Live, CD conifer Dead, ML Mixedwood Live, DL Deciduous
Live). Plots are divided by tree component with the left panel showing all tree (purple), only live
conifer in the center left panel (green), only dead conifers in the center right panel (grey), and
only deciduous individuals shown in the right panel (yellow). Statistical analysis (Table 5) was
performed using “All” trees and reflects observed differences in the left panel (purple).
Basal area
The analysis of basal area per hectare (m2/ha) revealed a statistically significant
difference among the four stand types (F(3,11.89) = 4.45, p = 0.01; see Table 5). The overall
mean was 28.8 (SD = 19.13), with CL and CD stands having lower values and ML and DL
having higher values than the mean. Post-hoc analysis showed that the basal area per hectare of
ML stands was significantly higher than that of CL stands (p = 0.01) and CD stands (p = 0.04).
No other significant differences were found between CL and CD (p = 0.98), CL-DL (p = 0.42),
CD-ML (p = 0.04), CD-CL (p = 0.42), ML-DL (p = 0.65). These results suggest a pattern where
45 | P a g e

CL and CD stand types exhibit similar basal areas, ML and DL stand types share comparable
basal areas, and there are significant differences in basal area between CL/ML and CD/ML stand
types however there is overlap between CD, CL and DL (Figure 7).

Figure 7. Basal area per hectare (median with quartiles and range included as box and vertical
line) for each of the four stand types are shown (CL conifer live, CD conifer dead, ML
mixedwood live, DL deciduous live). Plots are divided by tree component with the left panel
showing all tree (purple), only live conifer in the center left panel (green), only dead conifers in
the center right panel (grey), and only deciduous individuals shown in the right panel (yellow).
Statistical analysis (Table 5) was performed using “All” trees and reflects observed differences
in the left panel (purple).
Stand density
There was a significant difference in the average stems per hectare (SPH) among the
stand types (F(3,35) = 8.9, p < 0.05; see Table 5, Figure 8). CD stands exhibited a statistically
lower SPH compared to CL stands (p < 0.05) and ML stands (p = 0.07) (Figure 8). Conifer
stands had the highest SPH, while Dead Conifer stands had the lowest SPH. Both ML and
46 | P a g e

Deciduous stands averaged about the same at about 1500 SPH. These findings suggest that CD
stands have a significantly lower SPH than CL and ML stands, while the differences between
CL, ML, and DL stands are not statistically significant.

Figure 8. Stand density (SPH) (median with quartiles and range included as box and vertical
line) for each of the four stand types are shown (CL conifer live, CD conifer dead, ML
mixedwood live, DL deciduous live) for stems >4 cm in diameter. Plots are divided by tree
component with the left panel showing all tree (purple), only live conifer in the center left panel
(green), only dead conifers in the center right panel (grey), and only deciduous individuals
shown in the right panel (yellow). Statistical analysis (Table 5) was performed using “All” trees
and reflects observed differences in the left panel (purple).
Stand Volume
There was a significant different in volume per ha between stand types (Welch (F(3,36) =
3, p < 0.03). CL and DL were significantly different (p <0.05), as were CD and DL (p <0.05) and
ML and DL (p<0.05) (Table 5, Figure 9).
Despite Welch’s ANOVA result of less than 0.05, the Games-Howell post-hoc test found
47 | P a g e

no significant differences between the pairs for the volume per hectare metric. This may be
because for this particular metric, the assumptions required for the welch test were violated or
that homogeneity of variances was violated despite the Welch’s test being fairly robust to
violations of these assumptions. A subsequent non-parametric one-way ANOVA test (KruskalWallis with pairwise comparison using Wilcoxon rank sum test and a Benjamini-Hochberg
adjustment, significant differences were found between the CL-ML, ML-DL, and DL-CL pairs at
p = 0.0098 (Table 6).
Table 6 One way ANOVA using Kruskal-Wallis with Wilcoxon rank sum pairwise comparison of
means for volume per hectare.

Nigh’s (2016) equations consider merchantable volume to be total tree volume minus the
volume of the stump and top for conifer trees only. When specifically analyzing conifer volume
for both live and dead trees, a significant difference was observed between the four stand types
(Welch's F(3,29) = 25, p < 0.05) (Table 5, Figure 10). No significant differences were found
between CL and CD (p = 0.29), CL and ML (p = 0.98), or CD and ML (p = 0.46). Additionally,
no significant difference was found between CD and DL stands (p = 0.65). CL and DL were
statistically different, as were ML and DL (p < 0.05).
The average conifer volume for all age classes was 192 m3/ha. Although no statistical
differences were found among the three non-deciduous leading stand types, the ML mean for
these samples was lower than both the CL and CD types. It is important to note that the
assessment did not include the potential merchantability of dead fallen logs (with combined
standing and fallen volumes shown in Figure 10) which can be considered available volume
based on current salvage practices in BC.
48 | P a g e

Figure 9. Commercial Conifer volume per hectare (median with quartiles and range included as
box and vertical line) for each of the four stand types are shown (CL conifer live, CD conifer
dead, ML mixedwood live, DL deciduous live) for stems >4 cm in diameter. Plots are divided by
tree component with the left panel showing all tree (purple), only live conifer in the center left
panel (green), only dead conifers in the center right panel (grey), and only deciduous individuals
shown in the right panel (yellow). Statistical analysis (Table 5) was performed using “All” trees
and reflects observed differences in the left panel (purple).

49 | P a g e

Figure 10. Conifer volume per hectare (combined live & dead) (median with quartiles and range
included as box and vertical line) for each of the four stand types (CL conifer live, CD conifer
dead, ML mixedwood live, DL deciduous live).
While these four stand types occupy similar growing spaces, subtle variations in
composition and structure distinguish them. The differences identified above underscore the
unique characteristics of each stand type, contributing to a nuanced understanding of the
ecological dynamics within the study area. These variations, which have been explored through
tree height, diameter at breast height (DBH), basal area per hectare, stems per hectare (SPH), and
volume per hectare, set the stage for an in-depth analysis of the wildfire fuel characteristics
within each stand type.
Canopy Fuel Metrics
In this section, I analyze key canopy fuel metrics—canopy fuel load (CFL), canopy bulk
density (CBD), and height of maximum CBD. The detailed results are systematically presented
in Table 7, offering a comprehensive overview of the variations in canopy fuel metrics across the
50 | P a g e

studied stand types.
In the context of wildfire science, the predominant focus often lies on live conifer fuels.
However, as highlighted in Chapter 1, deciduous fuels can influence fire dynamics under specific
conditions. To account for this variability, the analysis encompasses 'All' fuel components (tree
type), incorporating both conifer and deciduous elements, alongside 'Live Conifer' fuel
components, excluding dead conifer and deciduous materials. This dual approach ensures a
thorough understanding of the distinct contributions of different fuel types to wildfire behaviour.

51 | P a g e

Table 7 Canopy fuel load metric results.

52 | P a g e

Canopy Fuel Load (CFL)
When considering all standing trees contributing to available canopy fuel, encompassing
both conifer and deciduous, living and dead individuals (referred to as CFL 'all'), a significant
difference in CFL among the four stand types emerged (F(3, 34) = 17.29, p < 0.05; Table 7).
Notably, the CD stand type exhibited a much lower mean fuel load of 0.59 kg/m2, statistically
differing from the other three stand types. Among the four stand types, ML displayed the highest
total CFL, followed by DL, CL, and then CD, which had the lowest. However, when focusing
specifically on canopy fuels from live conifers, the dynamics shift. Excluding the deciduous
component, the conifer CFL in the CL and ML stand types experience a substantial decrease,
rendering them much more similar (p = 0.43 now instead of 0.07; Figure 11). There is no longer
a distinction between CD and DL (p = 0.21), but ML and DL are now statistically different (p <
0.05), as shown in Table 7 and Figure 11. When considering only CFL from live conifers, CL
has the highest load, ML the second highest, followed by CD and DL. It is notable that conifer
live stands exhibit several outliers, with plot N70 particularly standing out, hosting 165 trees,
mainly tall spruce or subalpine fir with large DBH.

53 | P a g e

Figure 11. Canopy fuel load (median with quartiles and range included as box and vertical line)
for each of the four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood
live, DL deciduous live). Plots are divided by tree component with the left panel showing all tree
(purple), only live conifer in the center left panel (green), only dead conifers in the center right
panel (grey), and only deciduous individuals shown in the right panel (yellow). Statistical
analysis (Table 7) was performed on the “All” and “Live Conifer” trees and reflects measured
differences in the left (purple) and center left (green) panels.
Canopy Bulk Density (CBD)
A statistically significant difference was observed among the four stand types when
Canopy Bulk Density (CBD) was evaluated for all tree types (F(3, 36) = 7.52, p < 0.05; Table 7).
The lowest CBD was exhibited in CD stands (0.05 kg/m3), which was statistically lower than in
CL and ML stands (see Figure 12). In contrast, the total CBD was highest in CL stands and was
found to be statistically different from the total CBD in DL stands. No statistical differences
were found in the total CBD for the CL-ML, , CD-DL or ML-DL pairs.
When considering only CBD from live conifers, a significant difference persisted among
54 | P a g e

stand types (F(3, 38) = 28, p < 0.05; see Table 7, Figure 12). Significantly higher CBD in conifer
fuels was demonstrated by CL stands compared to all other stand types. No significant difference
was found in the conifer CBD between CD and ML or DL stands. However, a significantly
higher amount of conifer CBD was exhibited by ML stands compared to DL stands (see Figure
12).
To summarize CBD results, the highest average CBD was found in CL stands, with the
second-highest CBD observed in ML stands. Some conifer fuels at low density were still present
in deciduous stands and dead conifer stands averaged less than half of conifer live stands.

Figure 12. Canopy bulk density (median with quartiles and range included as box and vertical
line) for each of the four stand types are shown (CL conifer Live, CD conifer Dead, ML
Mixedwood Live, DL Deciduous Live). Plots are divided by tree component with the left panel
showing all tree (purple), only live conifer in the center left panel (green), only dead conifers in
the center right panel (grey), and only deciduous individuals shown in the right panel (yellow).
55 | P a g e

Statistical analysis (Table 7) was performed on the “All” and “Live Conifer” trees and reflects
measured differences in the left (purple) and center left (green) panels.
Height of Maximum CBD
The height of the maximum CBD, considering all tree types, differed among the four
stand types (F(3, 26) = 4.32, p = 0.01; Table 7, Figure 13). DL stands recorded the greatest
height to maximum CBD (Figure 11), showing statistical differences from CL and CD stands but
not from ML stands (Figure 11). When only fuels from live conifers are considered there was no
significant difference in height to maximum CBD evident among the four stand types (F(3, 25) =
1.42, p = 0.26; Table 7, Figure 13). Notably, the height of the peak bulk density is higher above
the ground in certain stands compared to others, while being indistinguishable between some
specific stand types for conifer-only tree components. Figure 13 shows height of maximum
CBD as boxplots for all stand types and tree types and Figure 14. Height of Max CBD profile for
four stand types and tree types. Maximum CBD (all tree types) is indicated with the blue arrow
for each stand type. This point is markedly different in mixedwood and deciduous stand types
when considering only conifer tree types, indicating some complexities in fuel load distribution

56 | P a g e

Figure 13. Height of maximum CBD (median with quartiles and range included as box and
vertical line) for each of the four stand types are shown (CL conifer live, CD conifer dead, ML
mixedwood live, DL deciduous Live). Plots are divided by tree component with the left panel
showing all trees (purple), only live conifer in the center left panel (green), only dead conifers in
the center right panel (grey), and only deciduous individuals shown in the right panel (yellow).
Statistical analysis (Table 7) was performed on the “All” and “Live Conifer” trees and reflects
measured differences in the left (purple) and center left (green) panels.

57 | P a g e

Figure 14. Height of Max CBD profile for four stand types and tree types. Maximum CBD (all
tree types) is indicated with the blue arrow for each stand type. This point is markedly different
in mixedwood and deciduous stand types when considering only conifer tree types, indicating
some complexities in fuel load distribution.
Canopy Base Height Metrics
I evaluated six canopy base height metrics for the four stand types, including the mean
height to the live crown, the height at which crown CBD surpasses 0.011 kg/m3, the height at
which crown CBD exceeds 0.037 kg/m3, and the corresponding metrics focusing exclusively on
conifer components. The detailed results are presented in Table 8.

58 | P a g e

Table 8 Canopy base height metric results.

59 | P a g e

There is a statistically significant difference in the mean height to live crown among the
four stand types for all species and components (F(3, 647) = 172, p < 0.05; Table 8). Post-hoc
analysis using Games-Howell comparison revealed significant differences between all stands .
However, when focusing solely on the live conifer component, there is no statistical difference
observed between stand types (F(3, 211) = 0.76, p = 0.52), with the mean height to live crown
remaining consistently around 3 meters for all stands (Figure 15).

Figure 15. Height to live crown (median with quartiles and range included as box and vertical
line) for each of the four stand types are shown (CL conifer live, CD conifer dead, ML
mixedwood live, DL deciduous live). Plots are divided by tree component with the left panel
showing all trees (purple), only live conifer in the center left panel (green), only dead conifers in
the center right panel (grey), and only deciduous individuals shown in the right panel (yellow).
Statistical analysis (Table 8) was performed on the “All” and “Live Conifer” trees and reflects
measured differences in the left (purple) and center left (green) panels. Note that dead conifer
trees do not have live crowns.
60 | P a g e

The crown base height (CBH) at which CBD reached 0.011 kg/m³ showed no significant
differences among stand types (F(3,26) = 0.79, p = 0.51; Figure 14). All stand types surpassed
the 0.011 kg/m³ CBH threshold within the height range of 1 to 3 meters (mean = 1.24, sd = 1.17,
Figure 16). Similarly, there were no significant variations between stand types using the CBH
threshold of 0.037 kg/m³ (F(2,25) = 1.8, p = 0.17; Figure 17). Despite this, greater variability
was observed among stands with this higher CBD threshold, and the height at which the
threshold was reached was higher (mean = 2.3, sd = 2.4). In addition, the canopy base heights for
both the 0.011 kg/m³ and 0.037 kg/m³ thresholds were lower than the CBH derived from the
mean height to live crown method that incorporates all trees (F(2,193)= 69, p <0.05) and conifer
only (F(2,117)= 342, p <0.05). For all stand types, the base height for conifer fuel only is
approximately 3 meters, while the peak height of fuel density is around 5 meters, indicating that
fuel is concentrated at between 3 and 5 meters above ground in all stands, though for some
stands, the peak of fuel does not exceed a critical threshold (either 0.011 kg/m3 or 0.037 kg/m3),
see Figure 14.

61 | P a g e

Figure 16. Canopy base height using a 0.011 kg/m3 CBD threshold (median with quartiles and
range included as box and vertical line) for each of the four stand types are shown (CL conifer
live, CD conifer dead, ML mixedwood live, DL deciduous live). Statistical analysis (Table 8) was
performed on the “All” and “Live Conifer” trees and reflects measured differences in the left
(purple) and center left (green) panels. Note the dead conifer panel shows no canopy base height
for dead conifer trees due to there being no dead conifer effective canopy fuel.

62 | P a g e

Figure 17. Canopy base height at the 0.037 kg/m3 CBD threshold (median with quartiles and
range included as box and vertical line) for each of the four stand types are shown (CL conifer
Live, CD conifer dead, ML mixedwood Live, DL deciduous live). Statistical analysis (Table 8)
was performed on the “All” and “Live Conifer” trees and reflects measured differences in the
left (purple) and center left (green) panels. Note the dead conifer panel shows no canopy base
height for dead conifer trees due to there being no dead conifer effective canopy fuel.
Canopy Length Metrics
I examined three key metrics related to canopy length across the four stand types: mean
crown length, canopy length at the 0.011 kg/m³ threshold, and canopy length at the 0.037 kg/m³
threshold. These metrics were assessed for both all tree types and conifer-only components. The
section below provides detailed results of the variations in these Canopy Length metrics and
results are detailed in Table 9.

63 | P a g e

Table 9 Canopy length metrics.

64 | P a g e

I observed a significant difference in mean crown length among the four stand types
when considering all tree types collectively (F(3,956) = 67, p < 0.05; Table 9, Figure 18). Posthoc analysis revealed significant differences between all stand type combinations (p < 0.05),
except for CL and ML (p = 0.07). Focusing solely on the live conifer component, there was a
statistical difference among stand types (F(3, 213) = 15.2, p < 0.05; Figure 18). Notably,
differences emerged between CL and ML (p < 0.05) and between CL and CD (p = 0.01, Table 9.
Crown lengths were found to be the smallest in CL and DL stands, while CD stands exhibited
longer crown lengths, reflecting both the canopy height of remnant un-killed conifers and the
regeneration of trees under open canopy conditions (Figure 18).

Figure 18. Crown length (median with quartiles and range included as box and vertical line) for
each of the four stand types are shown (CL conifer live, CD conifer dead, ML mixedwood live,
65 | P a g e

DL deciduous live). Plots are divided by tree component with the left panel showing all trees
(purple), only live conifer in the center left panel (green), only dead conifers in the center right
panel (grey), and only deciduous individuals shown in the right panel (yellow). Statistical
analysis (Table 9) was performed on the “All” and “Live Conifer” trees and reflects measured
differences in the left (purple) and center left (green) panels. Note the dead conifer panel shows
no canopy length for dead conifer trees due to there being no dead conifer effective canopy fuel.
Canopy length at 0.011 kg/m3 and 0.037 kg/m3 thresholds.
When I evaluated the canopy length as the length of the canopy which has a crown bulk
density that exceeds 0.011 kg/m3 (Reinhardt et al., 2003) the ML (p <0.05), and CD and ML
stands (p = 0.05) were found to have statistically different mean effective crown lengths (Table
9, Figure 19). If only conifer crowns are considered, the conifer fuels in the deciduous stand are
statistically different than the other three. When using CBD that exceeds 0.037 kg/m3 to evaluate
canopy length, the CL and CD are statistically different (F(3,30) = 24.4, p = 0.01; Figure 20).
Also, the ML stand canopy length differs from the CL and CD (p <0.05), and from the DL (p <
0.05, Figure 19 and Figure 20).

66 | P a g e

Figure 19. Canopy length using a 0.011 kg/m3 CBD threshold (median with quartiles and range
included as box and vertical line) for each of the four stand types are shown (CL conifer live, CD
conifer dead, ML mixedwood live, DL deciduous live). Statistical analysis (Table 9) was
performed on the “All” and “Live Conifer” trees and reflects measured differences in the left
(purple) and center left (green) panels. Plots are divided by tree component with the left panel
showing all trees (purple), only live conifer in the center left panel (green), only dead conifers in
the center right panel (grey), and only deciduous individuals shown in the right panel (yellow).
Note the dead conifer panel shows no canopy length for dead conifer trees due to there being no
dead conifer effective canopy fuel.

67 | P a g e

Figure 20. Canopy length using a 0.037 kg/m3 CBD threshold (median with quartiles and range
included as box and vertical line) for each of the four stand types are shown (CL conifer live, CD
conifer dead, ML mixedwood live, DL deciduous live). Statistical analysis (Table 9) was
performed on the “All” and “Live Conifer” trees and reflects measured differences in the left
(purple) and center left (green) panels. Plots are divided by tree component with the left panel
showing all trees (purple), only live conifer in the center left panel (green), only dead conifers in
the center right panel (grey), and only deciduous individuals shown in the right panel (yellow).
Note the dead conifer panel shows no canopy length for dead conifer trees due to there being no
dead conifer effective canopy fuel.
Surface Fuel Metrics
Transitioning from canopy metrics, the analysis below shifts to surface fuel results, with
a specific emphasis on coarse woody debris (CWD) across the four stand types.
The analysis of Coarse Woody Debris (CWD) involved 33 plots, comprising 9 Conifer
Live (CL), 5 Conifer Dead (CD), 11 Mixedwood Live (ML), and 8 Deciduous Live (DL). Across
all stand types, the average CWD was 5.7 kg/m2 (sd: 5.1). Conifer Dead (CD) stands exhibited
the highest CWD amounts, averaging 12.9 kg/m2, followed by Conifer Live (CL) stands at 6.4
68 | P a g e

kg/m2. Mixedwood stands had a mean CWD of 4.9 kg/m2, while Deciduous stands contained the
lowest CWD at an average of 1.6 kg/m2.
I observed a statistically significant difference in coarse woody debris (CWD) fuel load
among the four stand types (F(3,12.1) = 26.5, p < 0.05; Figure 19). Games-Howell post-hoc tests
revealed no significant difference between CL and CD stands, CL and ML stands, CD and ML
stands, or ML and DL stands. The CWD load in DL stands was lower than in CD and CL stands.
It is worth noting that statistical differences were not detected between CL, CD, and ML stand
types, possibly due to a CD plot with very low CWD (standing dead versus down dead). Visual
inspection of the box plot (Figure 21) indicates a high CWD amount at most Conifer Dead (CD)
sites, with the exception being the one outlier. Additionally, there was substantial variance in
CWD amounts in Mixedwood Live (ML) stands, suggesting a transition from dead conifer stands
to another state, where the dead conifer basal area is on the ground rather than standing (Table
10, Figure 21).
Table 10 Coarse woody debris (CWD) fuel load metric results.

69 | P a g e

Figure 21. Coarse woody debris (CWD) amount (median with quartiles and range included as
box and vertical line) for each of the four stand types are shown (CL conifer live, CD conifer
dead, ML mixedwood live, DL deciduous live). Statistical analysis (Table 10) was performed on
the “All” trees (purple).
Discussion
Stand structure
The four evaluated stand types (conifer live, conifer dead, mixedwood, and deciduous)
displayed distinct differences in their above-ground structural composition. These distinctions
encompassed variations in species composition, average heights, average diameter at breast
height (DBH), basal area, stems per hectare, and volume per hectare. Limited research has
directly compared the fundamental stand structures of these specific stand types. For instance,
Hély et al. (2000) reported similar basal area per hectare for boreal mixedwoods in deciduous
and mixed stands, while Freedman et al. (1996) observed similar basal areas in mixed and

70 | P a g e

coniferous stands. Despite differences in species composition and the proportion of standing
dead trees among the four stand types in this study, they demonstrated similarities in certain
attributes. Mixedwood, deciduous, and pine-beetle killed stands exhibited similar mean DBH and
mean tree height. The live conifer stands on the other hand, featured smaller trees at a higher
density on average. This is primarily attributed to the inclusion of young conifer plantations in
this stand type, reflecting the age distribution of conifer leading stands in the region. Mature
conifer live stands in the sample were characterized by large trees, evident in the observed range
of DBH and tree heights.
Stand structure and fuel structure strongly affect fire behaviour (Graham, McCaffrey, &
Jain 2004). The Canadian Forest Fire Danger Rating system uses stand type as a surrogate for
fuel type and loading (CIFFC 2003). This information is then used to inform forest management
decisions and assist wildland fire fighters by describing how a fire in a particular vegetation
complex will behave under different topography and weather conditions. Fuel types are defined
as “an identifiable association of fuel elements of distinctive species, form, size, arrangement,
and continuity that will exhibit characteristic fire behaviour under defined burning conditions”
(CIFFC 2003) and are qualitatively described (Perrakis et al., 2018). Based on a number of actual
burns in representative fuel types, the CFFDRS categories describe how a particular vegetation
complex will behave if it catches on fire, given weather and topography.
Fuel loading by stand type
The four stand types I evaluated exhibited large differences in fuel load, fuel structure,
and vertical distribution of fuels in the forest canopy. Mixedwood stands had the highest CFL of
all four stand types when both conifer and deciduous components were considered, while
deciduous stands had the second highest. When only conifer fuels are considered, live conifer

71 | P a g e

stands had the highest fuel load, mixedwood stands had the second highest, and deciduous stands
had the lowest. Of particular importance, mixedwood stands had only slightly lower average
conifer CFL than conifer live stands, and these two stand types were not statistically different.
The range of values for conifer CFL was wider than mixedwood stands, but the medians were
approximately the same. In contrast, dead conifer stands in the study had significantly lower
conifer CFL than live conifer or mixedwood stands. In fact, all dead conifer stand type fuel loads
were below the lower quantiles for both CL and ML stand types. These results indicate that
mixedwood stands are comparable to live conifer stands in their fuel load. However, under
extreme fire weather situations where differences between conifer and deciduous fuels becomes
less relevant (Cumming, 2001; Hély et al., 2001), mixedwood stands have higher CFL than even
live conifer stands. Often, mixedwood stands in the study area have a component of dead pine as
well, which would have no canopy fuel load, but would have high surface fuel loads. Cruz et al.
(2003) found similar CFL in their mixed conifer stands. In a study that focused on mountain pine
beetle impacted stands, Bright et al. (2017) found higher CFLs in dead conifer stands; however,
their study included higher components of <5 years dead and so would contain more canopy
foliage than our study for dead pine.
Within my study area conifer live stands had the highest CBD when both coniferous and
deciduous fuel components were considered. Mixedwood stands were found to have the second
highest, followed by deciduous stands and then dead conifer stands. For the live conifer fuels,
live conifer stands have the highest mean CBD, and mixedwood stands had the second highest
live conifer fuel CBD. Dead conifer stands had a lower mean CBD than mixedwood stands,
while deciduous stands have only a small Conifer CBD. CBD in conifer live stands vary between
almost none to above 0.4 kg/m 3, reflecting the wide range of CBD in conifer stands in the study

72 | P a g e

area. The observed range of CBD in the conifer leading stands likely reflects the age class
distribution of samples. Mid-range age classes (>2 and <6, 41 years to 100 years of age) in
general have more CBD as a function of stand density. The samples in this research include
different age classes which provide a representative description of the distribution of loads across
stands in the B.C. Central Interior.
CBD fuel loads exceeding 0.05 to 0.1 kg/m3 are often considered the threshold for
identifying risky stands (Agee, 1996). Within this research study area approximately half of the
conifer live CBD samples exceed the 0.10 kg/m3 level, and over 75% exceed 0.05 kg/m3. Most
mixedwood samples fell between the 0.05 and 0.1 kg/m3 levels, and all but one outlier of dead
conifer stand samples fall below 0.05 kg/m3 CBD. CBD is a complicated metric and several
methods have been devised for estimating it. The load over depth CBD calculation method,
which is simply CLF divided by canopy length, is the simplest. However, this method is prone
to underestimating critical CBD load for crowning (Cruz, Alexander, & Wakimoto, 2003; Scott,
2012). An alternative is taking the maximum value of the running mean of the stand vertical
profile, but there are also differences in how a CBD running mean has been calculated. Keyser
and Smith (2010) examined the influence of local versus general biomass allometries, as well as
non-uniform versus uniform vertical distribution. They found that when using CBD calculation
promoted in the Fire and Fuels Extension to the Forest Vegetation Simulator (FFE-FVS
Reinhardt et al. 2003) and described in Scott and Reihhardt (2001) and undertaken in this
research (maximum of the running mean of vertically distributed CBD) with default (non-local)
and uniform vertical distribution, only 2 of their 16 plots exceeded 0.10 kg/m3. When
implementing local crown biomass and non-uniform distribution, they found the majority of their
sites exceeded the 0.10 threshold. From this result they hypothesized that the FFE-FVS method

73 | P a g e

has the potential for misdiagnosing fire hazard (Keyser & Smith, 2010). My CBD calculations
incorporated both local biomass equations and assumed a non-uniform distribution, and therefore
should be less susceptible to underestimation. Keyser and Smith (2010) reported an average
CBD of 0.12 kg/m3 for ponderosa pine in the Black Hills National Forest in South Dakota. In
this study I found the conifer component of CBD in conifer leading stands was slightly lower at
0.10 kg/m3. Ex et al. (2016) found that variation in CBD can occur due to overlap between tree
crowns. They suggest that non-uniform distribution could prove to be more unpredictable for
stands with complex structure. Reinhardt et al. (2006) found that CBD was not well predicted in
their study using the load over depth method using the average canopy length. They suggest that
average CBD is not a useful metric unless the stand is very simple and uniform (Reinhardt et al.,
2006). Instead, they suggest using the method proposed by Albini (1996) where the 90th and 10th
percentiles of biomass (Height where 90 % and 10% of biomass is below). A proxy for this
method is using the 90th percentile tree height and the median crown base height.
When CBD is calculated as the maximum of the running mean of the vertical fuel
distribution in the stand, the height at which the maximum occurs can be used as a metric of fire
risk. Within my study area all fuel components in CL, ML and CD stand types the maximum fuel
load occurs at between 5 and 7 m, and for DL stand types, 11 m. When I considered only conifer
fuel components all peak fuel loading occurred at between 5 and 7 m as well. Few studies have
evaluated the height of the peak load in the vertical density fuel profile (i.e. height of max CBD).
Keyser & Smith (2010), in their South Dakota study of ponderosa pine canopy fuels, found
values of approximately 14 m. Only two of their samples had peaks of fuel around 5 m; the rest
were between 10 and 20 m. For comparison, the mean overall mean for all 118 samples in this
study was 6.1 m (median = 5.0, sd = 5.0). Our CBD values peaked at between 4.5 m and 11.3 m.

74 | P a g e

Van Wagner (1977) found a CBD of 0.14 kg/m3 in subalpine fir stands studied and from 0.12 to
0.26 m in jack pine (Pinus banksiana) and red pine (P. resinosa) stands.
I found that canopy base height (CBH) differed between stand types, when all fuel types
are considered. However, when only conifer live fuels were assessed, there were no significant
differences among stand types in the mean height to live crowns. Mean height to live crown for
live conifer components in all stands start at about 3.1 m for deciduous and reaches 3.8 m for the
other three types. There were no differences between stand types when using CBH as the 0.011
or 0.037 kg/m3 threshold values (FFE-FVS method; Reinhardt et al., 2003). The base height is
approximately 1.5 to 2.5 m for both methods; however, these are both several meters lower than
the mean height to live crown. Phelps and Beverly (2022) found similar CBH values in their
study in Alberta. CBH is also a main driver of Van Wagner’s critical surface intensity (CSI)
metric (Van Wagner, 1977) and is used in BC along with FMC in fuel mitigation prescriptions.
Using Van Wagner’s modified formula CSI = 0.001 x CBH^1.5 + (460 + 25.9 * FMC)^1.5, we
can estimate that the four stand types would have a CSI of approximately 1,100 kW/min using
mean height to live crown as the canopy base height.
Crown lengths for conifer fuels were similar across stand types, with crown lengths
ranging from 7 to 9 m. However, mixedwood and deciduous stand types exhibited more variance
in crown length. Conifer canopy lengths in mixedwood stands never dropped below 5m and were
as high as 16 m. Conifer stand type crown lengths were lower than CD types.
Coarse Woody Debris
I found that CWD loads were highest in dead conifer stands . Although only 5 plots were
sampled in CD stands, a power analysis (Table 11) using the pwr package in R and the pwr.t.test
function (Champely, 2022) indicated a high likelihood of detecting differences between most

75 | P a g e

stand type pairs. However, the small sample size limited my ability to detect differences in CWD
loads between conifer and mixedwood stands. Table 11 presents the results of a two-sample
power analysis at a 0.05 significance level, aiming for a power of 0.8.
As shown in the CL-ML row of Table 11, more samples in both stand types (samples 1
and 2) are needed to achieve an 80% chance of correctly rejecting the null hypothesis of no
difference between CL and ML stands in terms of CWD while sufficient power was achieved in
the remaining pairs comparisons.
The mixedwood stands were diverse with respect to CWD volume. Of the 11 plots
sampled in ML types, the loads varied from almost no coarse wood to volumes that were
comparable to what was sampled in the mountain pine beetle killed conifer stands. Deciduous
stands were found to have on average very little CWD load, but some did have volumes
equivalent to what was found in some mixedwood stands. It is important to note that stand type
categorization is a snapshot in time. As dead conifer stands decay, they will transition to other
stand types based on succession pathways (Donato, Campbell & Franklin, 2012).
Table 11 Power and effect analysis for CWD comparison. Sample types 1 and 2 are the samples
in each of the stand type pairs. Refer to Champely (2022) for details. The power attained by
current samples available in the pairs for all but CL-ML is sufficient for the purposes of
determining a statistical difference at 80% (power = 0.8) chance of correctly rejecting a false
null hypothesis.

Mean CWD load in the SBSdk has been reported to be on average 55 m3/ha, with a wide
range from 2 to 473 m3/ha (Feller, 2003). Feller (2003) has suggested that CWD load increases
76 | P a g e

with increasing stand productivity. I found much higher average CWD amounts, and a wider
range of CWD, than Phelps and Beverly (2022) in all stand types except deciduous. The CWD
loads that I recorded in dead pine stands exemplifies the fuel hazard situation related to this stand
type. CWD loads were at the high end or higher than the previously recorded maximum for the
SBSdk, indicating a high fuel hazard associated with this class of fuel in dead conifer stands but
also in mixedwood stands. Hély, Bergeron, and Flannigan (2000) found about double the CWD
loads in deciduous, about the same in mixed conifer, while I found about triple the loads the
conifer stands studied.
Balancing fire hazard with other values is an increasingly important forest management
goal. Brown (2003) suggested that for cool and lower subalpine forests, a range of between 1.0
and 3.0 kg/m2 is optimum for ecosystems similar to those in his study. All of our stand types
except deciduous exceeded the maximum thresholds in Brown’s 2003 study. A study located in
the Chilcotin, southeast of my study area near Quesnel, B.C., assessed CWD fuel loads in
historic mountain pine beetle impacted stands (Hawkes et al., 2005). Hawkes et al. (2005) found
fuel loading for >7.5 cm diameter CWD was approximately 2.0 kg/m2 in 2001 for MPB
impacted stands and only about 0.5 kg/m2 for non-MPB impacted stands, which is much lower
than the amounts I found in MPB impacted stands.
Within my study area, the mixedwood stands displayed the highest basal area, featuring a
conifer component comparable to that of conifer-leading stands and a substantial deciduous basal
area component. Numerous studies have demonstrated increased forest growth and biodiversity
in mixed-species forest ecosystems (Fichtner et al., 2018). From a forestry or ecosystem
productivity perspective, this heightened basal growth in mixedwood stands is often perceived
positively. However, from a fuel loading perspective, my research not only highlights that these

77 | P a g e

mixedwood stands can maintain a large percentage of more flammable conifer trees (Popovic et
al., 2021; Varner et al., 2015; Parisien et al., 2023), with low crown base heights, but their fuel
load and distribution characteristics also add complexity to fuel load estimation. They don’t have
the highest CFL or CBD, but they have the second highest. They don’t have the highest CWD
load, but their load is variable and complex. Several types of classifications are used for forest
stands. Often a stand is classified by the percentage of species and also by seral stage. The
mixedwood stands described in this study contain a mixture of dead standing pine, live
broadleaf, and live conifers of other species. In this study area, and indeed in many parts of B.C.,
MPB-killed stands are being re-classified as stands dominated by the remaining living trees,
effectively disappearing from the inventory data as the stands transition to live stands, however a
legacy of high-risk fuel load components will remain.
Fuel Loading in Dead Conifer Stand fuel load
Dead conifer stands, which in my study region are primarily mountain pine beetle-killed
lodgepole pine stands, were found to have canopy fuel loads that were half or less that of live
conifer stands. MPB killed stands had the lowest SPH, BA, and the lowest volume per ha. In
contrast, MPB killed stands had the second highest conifer volume per hectare and the highest
average height. I found that MPB killed stands had the same CBH, CBH at 0.011 kg/m3 and
CBH at 0.037 kg/m3 as the other stand types, and generally similar crown lengths, but they had
lower amounts of canopy fuel and high CWD fuel loads.
In a study in the Cariboo region of B.C., Day et al. (2010) suggested that surface fuel
loads less than 7.6 cm diameter should be maintained below 4 kg/m2 and never exceed 8 kg/m2.
Dead conifer stands in my study area had an average of 12.9 kg/m2 of >7.5cm diameter class
surface fuel. Although the sample size was small, the CWD loads in dead conifer stands were

78 | P a g e

double those sampled in live conifer stands. It is also notable that the recorded CWD volumes in
mixedwood stands was highly variable, indicating that the definition of ‘mixedwood’ based on
standing basal area by species can include some stands that would be classified as dead conifer if
the dead stems were still standing. Hood et al. (2017) reported 1.27 kg/m2 of 1000-hr fuels in
untreated dead pine stands in Colorado, a strikingly lower amount than what was found in this
study, likely related to dead trees perhaps being still in a standing condition in their study, though
this was not specified. In a study in Alberta, Phelps and Beverly (2022) calculated CWD load in
five stand types and found none of the stand types with CWD loads over 2 kg/m2. Bright et al.
(2017), in their study of mountain pine beetle killed stands in Colorado found a mean of 1.8
kg/m2 and a maximum of only 9.7 kg/m2.
In a meta-analysis of CWD volume, Feller (2003) found CWD in old growth SBSdk
stands to have a mean of 55 m3/ha and a range between 2 and 473 m3/ha. In my study area CWD
volume was 232 m3/ha for conifer live stands, 424 m3/ha for dead conifer, 187 m3/ha for
mixedwood, and 80 m3/ha for deciduous stands. The CWD volumes I found were within Feller’s
report range, but the stands that I sampled would not be characterized as old growth.
Dead conifer stands in this study have very little canopy fuel load, but their surface loads
are extremely high. Despite the lower canopy fuel loads, these stands have been known to
produce unexpected fire behaviour. As Moriarty et al. (2019) notes, crown initiation and
propagation was observed in gray phase mountain pine beetle attacked stands and active fire
behaviour should be expected irrespective of fire weather and fuel moisture conditions in all
beetle attack phases including grey (Moriarty et al., 2019).
Fuel loading in mixedwood stands
Although I found that canopy loads in mixedwoods were not quite as high as live conifer

79 | P a g e

stands, their CBD was similar to live conifer stands, indicating that these stands can exhibit high
canopy fuel loads. Canopy base heights in mixedwood stands were also not different than the
other stand types and ranged between 1 m and about 4 m in height. Their height to maximum
CBD is slightly higher than dead conifer and live conifer stands, and not statistically different
from any other stand type. CWD load, as noted above, has a wide range from almost none to as
high as most dead conifer stands. This suggests that mixedwood stands classified by species
basal area in the study area, and which have 50% or less dead trees, also have fuel load attributes
similar to dead conifer stands.
It is important to note that managed mixedwood stands, created through planting or
thinning, may exhibit different fuel load characteristics than those evaluated in this study.
However, my results suggest that caution should be applied when designating mixedwood areas
as reserves because conifer canopy fuel loads in these areas can still be hazardous.
Hély, et al. (2000) compared surface loads between deciduous, mixed and conifer stands in the
Canadian boreal and found both mixed-deciduous and mixed-coniferous had CWD loads higher
than conifer stands.
Finally, my results indicate that mixedwood stands have CBH and CFL averages that do
not approximate the CFFDRS FBP fuel types found in the study area (Perrakis et al., 2018).
Forestry Canada Fire Danger Group (1992) suggested that CBH and CFL estimates for fuel type
C-2 Boreal spruce should be 3 m and 0.8 kg/m2 respectively. In contrast, I found that, for conifer
live stands, which are the most comparable stand type, had CBH and CFL values of 3.8 m and
1.3 kg/m2. Mixedwood M1 to M4 in the FBP shows CBH as 6 m and CFL as 0.8 kg/m2. In
contrast, I found the conifer component of mixedwoods to have a CBH of 3.8 m and a CFL of
1.06 kg/m2. CFL loads are 25 to 40% higher than those referenced for these fuel types and CBH

80 | P a g e

for mixedwoods is 3 m lower, indicating a higher hazard for crown fire initiation. These
comparisons indicate high variability within types that may warrant more study.
Conclusions
My results demonstrate that live conifer, mixedwood, deciduous leading, and MPB killed
pine stands all differ considerably with respect to CFL, CBD, and CBH, crown length and CWD.
Canopy bulk density and canopy base height are two of the most important metrics used when
assessing fuel load hazard. Low canopy base height allows heat transfer from surface fires to
transition into the crown, and high fuel density in the crown facilitates the spread of fire through
the forest canopy. I found that both live conifer and mixedwood stands can display hazardous
canopy fuel load characteristics, and all stands but deciduous leading stands exhibited high
surface fuel loads.
The four stand types in this study have different attributes that create different fuel loads.
Live conifer stands had the highest canopy load, CFL and CBD. Mixedwoods were slightly
lower. Both were above the 0.05 to 0.1 kg/m3 threshold for hazardous canopy fuel. Dead pine
stands did not constitute a significant canopy fuel hazard; however, their surface loads were very
high. Although studies that aim to quantify fuel loads in stands are fairly common, particularly in
North America, there are no studies to my knowledge that have quantified canopy fuel loads in
north-central B.C., particularly in mountain pine beetle impacted stands and stands adjacent to
and intermixed with these.
It is evident that ‘Mixedwoods’ as they were defined in this study, are not necessarily a
middle option when considering alternatives to highly volatile live conifer stands. Fuel loads in
mixedwood stands in the study area were not statistically different from live conifer stands when
considering most fuel load metrics. The mixedwood CBD conifer fuel component was lower

81 | P a g e

than the values observed in conifer stands but was still above the 0.05 kg/m3 threshold that has
been used to identify hazardous stands.
Advocates for more mixedwood representation on the land base should reflect on these
findings, particularly when considering that many current mixedwood stands may be considered
late-phase mountain pine beetle attack stands, with a legacy of dead conifer surface fuel loads.
‘Mixedwood forests’ can either be intermixed mixedwoods where broadleaf and conifers exist in
close proximity, or mixedwoods can be realized as interspersed patches of pure conifers and pure
broadleaf trees. My results suggest that a patchwork or mosaic of conifer and broadlead stands
may be better from a fire mitigation perspective. It is clear that deciduous stands offer the lower
fuel load profile, particularly when more volatile conifer fuel components are considered.
However, as noted by Alexander (2010), there are seasonal and weather conditions where
deciduous forests and broadleaf derived fuels can contribute to fire spread and fire risk.
Fuel hazard has two key facets: the fuel load itself, and its spatial context. In this chapter
I have evaluated the quantitative fuel load and its vertical distribution, and assess the differences
among four common stand types in north-central B.C. In chapter 3, I will look at the utility of
airborne laser scanning (ALS) for modeling these metrics using an area-based-approach and
predicting the basic and fuel metrics at 10 x 10 m resolution for 1,800 square kilometers.
CHAPTER 3
Introduction
Reducing the risk of large catastrophic fires in the mid- to long-term will depend on our
ability to develop landscape level forest management plans that promote more diverse and
resilient stand types, and which aim to limit the accumulation of forest fuels across the landscape
(Williams, 2011).

82 | P a g e

Within this context, accurately assessing the spatial arrangement of high fire risk stands
and less susceptible stands at the landscape scale becomes crucial for effective forest
management. The first step in this landscape-level approach is the evaluation of current fuel
loads. Quantifying and spatially locating forest fuel loading is essential for assessing forest
resistance and resilience to wildfires (DeRose & Long, 2014) and for identifying critical
management areas. Quantifying fuel loads informs strategies to mitigate fire risk by identifying
hazardous fuel strata gaps, while understanding expected fire severity and rate of spread
enhances forest resilience strategies.
Landscape mosaics, which incorporate stands of different ages and species composition
(Hély et al., 2000) or integrate pure deciduous stands into a conifer matrix, can play a significant
role in reducing fire spread potential and contagion. This reduction occurs both within the
vertical stand profile and across stands.
Building on this understanding, Ager and Vaillant (2010) conducted a study
demonstrating that treating fuels on a relatively small percentage of a landscape could result in
approximately a 70% reduction in the loss of large (>53.3 cm DBH) trees in a 16,000-ha
landscape. This underscores the potential effectiveness of targeted fuel management in
promoting landscape resilience and reducing the risk of mega-fires in the mid- to long-term. By
mapping both the location and vertical arrangement of fuels, additional management options may
emerge, further contributing to keeping fire out of the crown and sustaining a manageable fire
regime.
The planning for wildfire mitigation necessitates a broad perspective, extending to the
landscape scale, although the execution of forest management and fire mitigation strategies
typically occurs at the stand scale. Rather than treating an entire landscape simultaneously, forest

83 | P a g e

managers can strategically focus on high-risk areas, aiming to minimize costs and achieve
consistent returns from fuel mitigation efforts. The effectiveness of wildfire risk mitigation on a
landscape scale hinges not only on precise fuel load data at individual sites but also on
understanding how these sites contribute to the overall landscape. In cases where resources are
limited, targeted investment in fuel mitigation becomes crucial. Without strategic targeting,
there's a risk of inefficient resource allocation and potential waste of financial resources.
The use of airborne laser scanning (ALS) technology, also known as light detection and
ranging (LiDAR) has become increasingly popular in forest attribute analysis. It provides a
highly accurate and efficient means of collecting forest and landscape data that can assist in the
planning of forest operations and can also provide estimates of forest fuels (Akay et al., 2009).
LiDAR data is collected using an airborne platform, such as an airplane or helicopter, equipped
with a LiDAR instrument. The LiDAR instrument emits a series of laser pulses, which are
reflected off the ground and vegetation surfaces. The time it takes for each pulse to return to the
instrument is then used to calculate the distance between the instrument and the ground (White et
al., 2013). This technology allows for the creation of detailed 3D point clouds representing the
forest canopy and the underlying ground topography (Akay et al., 2009).
One of the main reasons for using ALS in forest analysis is that it allows for the gathering
of detailed information on the distribution and quantity of forest structure, such as trees,
branches, and leaves as well as their density. This information is becoming essential for
understanding how forest stands are structured but also for wildfire science (Calvo et al., 2023).
Remote sensing derived three-dimensional models of forests can provide insight into forest fuel
structure, such as the vertical and horizontal distribution of different fuel types (GarciasCimarras, 2023). ALS provides fine-resolution data at site level, but can also cover large portions

84 | P a g e

of the landscape, providing context on fuel continuity and overall hazard to communities and
other values. Spriggs et al. (2017), Sherrill et al. (2008), and Wagner et al. (2004), for example,
have investigated the use of ALS for estimating forest structure and basic stand structure. Others
have focused on fuel structure and related metrics (Andersen et al., 2005; Arroyo, et al., 2008;
Popescu & Zhao 2008; González-Ferreiro et al., 2017; Riano et al., 2003). Riano et al. (2003)
used ALS to spatially predict critical forest parameters in fire behaviour modeling. Alexander et
al. (2004) used ALS to estimate canopy bulk density and canopy base height, while Inan et al.
(2017) used ALS to assess forest fire fuel load potential. Bright et al. (2017) used ALS to
investigate the fuel loads present in bark beetle impacted stands and González-Ferreiro et al.
(2017) modeled the vertical distribution of canopy fuel loads using ALS. In B.C., Kelley et al.
(2022) combined area-based and individual tree metrics for improving volume estimates in
Coastal Douglas-fir biogeoclimatic zone stands. Ahmed et al. (2014) integrated ALS and Landsat
satellite data to estimate forest cover in coastal B.C. De Assis Barros (2019) used ALS to assess
old growth attributes in forest stands. Common to the majority of these is the use of an areabased approach.
The area-based approach (ABA) is a method of collecting and analyzing ALS data
(Næsset, 2002). The ABA method focuses on the spatial distribution of ALS returns within a
specific area and involves dividing the study area into smaller, discrete regions. ALS data are
then used to calculate metrics such as canopy height and density within each region or raster.
This method is distinctly different from an individual tree-based method which focuses on
analysis of the point cloud specific to one tree. Once the ALS data has been collected, it is
processed using specialized software to generate detailed 3D point clouds of the light returns
captured by the sensor. Field data also needs to be collected, and attributes are calculated or

85 | P a g e

estimated from the field data. The field data is then used to train ALS based models. The models
derived from this process are then used to project the metrics across the remainder of the point
cloud not sampled. This ‘wall-to-wall’ processing allows maps of individual metrics to be
modeled across the entire ALS coverage area. These projected raster layers can then be used for
a variety of applications, such as forest management and wildfire risk assessment.
Fuel hazard mapping
The recent catastrophic fires in British Columbia have underscored the critical need for
landscape-scale estimations of fuel loads. These estimates are essential for assessing risk,
reducing fuel hazards, and minimizing expenses. By providing a comprehensive overview of
various fuel metrics at fine texture and large scale, this study aims to offer managers and
government officials greater flexibility and cost-effective opportunities to mitigate hazards
associated with fuels.
Fuel mapping is not a new concept. It has commonly been employed to assess the
potential rate of spread of fires (Sandberg et al., 2001) and develop effective suppression
strategies. A crucial aspect of fire mitigation decision-making is the knowledge of the spatial
distribution and characteristics of fuels (Keane et al., 2001). Traditionally, generalized fuel
models have been used that provide a coarse measure of fuel volume and composition that can
then be applied, offering a general understanding of fuel volume and composition for use in fire
behaviour models. However, advancements in technology, particularly in the availability and
processing of ALS data, enable more refined, detailed, and spatially accurate fuel models.
In Canada, the 16 Canadian Forest Fire Danger Rating System (CFFDRS) fuel classes
have been spatialized using British Columbia’s Vegetation Resources Inventory (VRI) along
with other mapped data to generate the Provincial Strategic Threat Analysis (PSTA; Perrakis et

86 | P a g e

al., 2018). While this coarse data is useful at a provincial level, its applicability for developing
site-specific fire mitigation treatments at an operational level is challenging. In addition, these
models lack the quantification of fuel metrics necessary for the development of tailored sitespecific treatments (Keane et al., 2001). Amid these advancements, it is notable that while
vegetation inventory data is typically the key input in fuel models, Price and Gordon (2016)
observed that vegetation polygons averaging >5ha do not provide accurate enough fuel data to
predict fire behaviour. VRI polygons used in the PSTA mapping can be much larger than 5
hectares.
Recent advancements in computer software and hardware have made it feasible to
generate spatially explicit, high-resolution maps of fuel loads for use in fire behaviour modeling
programs (Keane et al., 2001; Marino et al., 2016). However, the availability of these maps is
limited due in part to the cost of ALS and a shortage of expertise in processing and fuel load
calculations. As ALS becomes more accessible, land managers will need established processes to
implement fuel load mapping, and this research serves as a starting point for such comparisons.
ALS has been used to evaluate fuel loading in different forest types around the world.
Marino et al. (2016) modeled fuel types using some basic ALS derived metrics across a large
area in Canary Islands, Spain. Surface fuel metrics were also modelled by Mutlu et al. (2008)
using ALS and multispectral data to create surface fuel maps for parts of Texas. Price and
Gordon (2016) used ALS to map fuel load and hazard over 144 km2 in Australia and Riano et al.
(2003) estimated fuel metrics for a small area in Germany. In North America, Andersen et al.
(2005) have applied similar ALS/ABA methods to model fuel; similar approaches have been
advanced by Erdody and Moskal (2010), Ramirez et al. (2018), and others.

87 | P a g e

Fuel mitigation
Long-term forest planning should aim to incorporate some form of fire mitigation
strategy, often focusing on managing forest fuels. Keyes and O’Hara (2002) discussed reducing
fuel load hazard through silvicultural methods that aim to reduce crown bulk density (CBD)
below target amounts. Day et al. (2010) specified methods and targets for achieving crown base
heights (CBH) and crown fuel loads (CFL) that increase the resistance of stands to crown fire.
They suggested that surface loads, comprised of fuels less than 7.6 cm in diameter, must be
maintained below 4 kg/m2, and they should never exceed 8 kg/m2. If they exceed these values,
the CBH must be greater than or equal to 3 m (Day et al., 2010). They also suggest that the risk
of crown fires can be reduced in forest stands with high basal area if advance regeneration and
understory trees are removed. Specifically, they suggest that risk is reduced if CBH is greater
than 2 m and surface fuels are maintained below 4 kg/m2.
Previous research has indicated that crown fire spread is greatly reduced when crown
bulk density is below 0.05 kg/m3 (Alexander, 1988; Keyes & O'Hara, 2002). Keyes & Varner
(2006) cautioned that there are often side effects of silviculture treatment that focus on reducing
CBD such as increased slash, reduced duff moisture content and increased wind penetration into
the sub-canopy. At the landscape level Ager et al. (2010) found that fuel reduction treatments
that are applied to only 10% of the landscape can result in a 70% reduction in expected wildfire
loss of large trees. In addition, Wei (2012) developed a method for optimizing the location of
fuel breaks on a landscape, but the application of the methods requires fine grain fuel maps. This
past work indicates fine grain, spatially explicit, estimates of fuel loading are an important
component of targeted and efficient fuel mitigation strategies at a landscape level.
The aim of this chapter is to test the ability of ALS data to develop models of a range of
fuel metrics across different forest types. I then evaluate the landscape level projections of these
88 | P a g e

fuel models and assess their utility with regard to developing landscape level fuel mitigation
plans. Specifically, my research addresses the following questions:
1.

Can ALS be used to model critical fuel metrics (i.e. CBD, CBH, CWD) across a large
land area at fine resolution?

2.

How accurate are the resulting fuel models across different forest types (conifer leading,
mixedwood, broadleaf leading)?

3.

How suitable are ALS projections of fuel loading (e.g. CBD (i.e. CBD >0.1kg/m3) for
landscape level hazard assessment?

4.

Can modeled CBD and CBH values be combined to produce an effective wildfire hazard
map?

5.

How does ALS-modelled CWD inform our understanding of fuel hazards?
Overall, my research aims to provide a comprehensive understanding of the factors that

contribute to wildfire risk and how these can be effectively modeled and used to inform
management.
Materials and Methods
In this section, I describe the application of ABA using ALS data, as outlined by Næsset
(2002). The methodology adheres to established best practices, as outlined in White et al. (2013),
for employing ALS data in forestry analysis. This includes the following steps:
1.

ALS data acquisition and processing (normalized, classified and denoised)

2.

Field data collection from circular 10 m radius plots as per chapter 2.

3.

Compilation of field data into plot level attributes as per chapter 2 above.

4.

Clipping the ALS point cloud to coincide with field plots.

5.

Generation of plot level attributes from the ALS data. In this case, 56 statistical metrics
were calculated on the plot ALS data using the cloud metrics function in the lidR package
(Roussel & Auty, 2022). These metrics, referred to as standard metrics, are predefined in
89 | P a g e

the lidR package and are summarized in Table 12. An additional set of 9 non-standard
metrics were calculated from the plot ALS data (Table 12).
6.

Developing predictive models for fuel types using a random forest modelling framework
using the ‘ranger’ (Wright & Ziegler, 2017) and ‘tidymodels’ (Kuhn et al., 2020) R
packages.

7.

Tuning the hyper-parameters associated with the predictive models to increase the
accuracy of the models. Hyper-parameters in the context of machine learning are settings
for the model that are assigned prior to the learning process. For this research, hyperparameters included number of trees, maximum depth of trees, minimum samples split
(Kuhn & Wickham, 2021),

8.

Generating full landscape coverage maps (wall-to-wall) of metrics from ALS data using
functions in the lidR package.

Statistical Metrics
The following table lists the predictive metrics used to model fuel load metrics. Fifty-six
of these are standard metrics, built into the lidR package for analysis of ALS data. An additional
10 were included from other sources (Table 12).
Table 12. List of ALS metrics used to develop fuel models and maps.
Metric

Description

Zmax

Values used

Equation

Source

maximum height

max(z)

Roussel et
al. 2023

Zmean

mean height

mean(z)

stdmetrics
r-lidar lidR
Wiki
GitHub

Zsd

standard deviation of height
distribution

sd(z)

Zskew

skewness of height distribution

skew(z)

Zkurt

kurtosis of height distribution

kurt(z)

90 | P a g e

Metric

Description

Values used

Equation

zentropy

entropy of height distribution.
Quantifies the diversity and
evenness of the elevational
distribution of las points.
Normalized Shannon
Index(Shannon 1948)

pzabovezmean

percentage of returns above the
mean of z

pzabove2

percentage of returns above 2
meters

zqX

xth percentile (quantile) of height
distribution

5, 10, 15, 20, 25,
30, 35, 40, 45, 50,
55, 60, 65, 70, 75,
80, 85, 90, 95

zpcumX

cumulative percentage of return
in the ith layer according to
Woods, Lim, & Treitz, (2008).

1,2,3,4,5,6,7,8,9

itot

sum of intensities for each return

sum(i)

imax

maximum intensity

max(i)

imean

mean intensity

mean(i)

isd

standard deviation of intensity

sd(i)

iskew

skewness of intensity distribution

skew(i)

ikurt

kurtosis of intensity distribution

kurt(i)

ipground

percentage of intensity returned
by points classified as "ground"

ipcumzqX

percentage of intensity returned
below the kth percentile of height

10,30,50,70,90

p1th

percentage xth returns

1,2,3,4,5

pground

percentage of returns classified
as "ground"

n

number of points

area

approximate actual area of a
raster

Zs

Height of point

Z

entropy

A normalized Shannon vertical
complexity index that quantifies
diversity & evenness of point
cloud heights. by = 1 partitions
point cloud in 1 m tall horizontal
slices. ranges from 0-1, with 1
being more evenly distributed
points.

entropy.metric <- entropy(Zs, by =
1).

Source

Pretzsch, H.
(2008);
Roussel et
al 2022

Woods et
al. 2008

Pretzsch, H.
(2008)

91 | P a g e

Metric

Description

Values used

Equation

Source

LADen

Leaf Area Density, assesses the
leaf area in the canopy volume. k
= 0.5, standard extinction
coefficient for foliage. dz = 1
partitions point cloud intp 1m
horizontal slices. z0 = 0.01 is set
to a reasonable height based on
the age and height of the study
sites

LAD(Zs, dz = 1, k=0.5, z0=0.01)

Larue &
O'Leary
2021

LADen.dct

LADen filtered for density of >0.1
and height of >3m. Assumes
crown only start above 3m
(bellow that too much noise with
veg and ladder fuels)

LADen.dct = which(LADen$lad >
0.1 & LADen$z > 3)

Larue &
O'Leary
2021

alscbh

potential estimator for crown
base height

alscbh = min(LADen.dct)

Larue &
O'Leary
2021

alsch

potential estimator for crown
height.

Alsch = max(LADen.dct)

Larue &
O'Leary
2021

alscd

potential estimator for crown
depth.

Alscd = length(LADen.dct)

Larue &
O'Leary
2021

alscvlad

Coefficient of Variation for LAD.

Alscvlad = cv(LADen.dct$lad)/100

Larue &
O'Leary
2021

lad.veg

Sum of leaf area density less
than 1m.

Lad.veg =
sum(LADen$lad[LADen.dct$z < 1])

this study

lad.ladder

Sum of leaf area density between
1 and 3 m.

Lad.ladder =
sum(LADen$lad[LADen$z > 1 &
LADen$z < 3])

this study

lad.canopy

Sum of leaf area density above
3m.

Lad.canopy =
sum(LADen$lad[LADen$z >3])

this study

VAI

Vegetation Area Index. Sum of
the leaf area density values for
all horizontal slices assessed in
previous line.

VAI = sum(LADen$lad)

Beland et
al., 2011

VCI

Vertical Complexity Index. Fixed
normalization of entropy metric.
by = 1 assesses the metric
based on 1m of horizontal slices
in the canopy. zmax set at 100 to
be comfortably above max
canopy height.

VCI = VCI(Zs, by = 1, zmax =
100).

van Ewijk et
al 2011

CSI

Critical Surface Intensity (CSI)

CSI =
0.001⋅CBH^1.5⋅(460+25.9⋅"FMC"
)^1.5

Beck &
Simpson
(2007)

Study Area
See chapter 2, Study Area section.

92 | P a g e

Field Data
Field data was collected as described in chapter 2. As noted by White et al. (2013) the use
of circular 10 m radius field plots falls within the recommended dimensions for the ABA
approach. Ground measurement of forest features was collected from a minimum of 110 plots of
10 m radius (many of these were included in the empirical plots). All trees with a DBH greater
than 4 cm were measured to obtain LiDAR metrics estimates for disturbed and/or young forests
(Keränen et al., 2015). The inventory followed the Change Monitoring Inventory (CMI)
procedures (Forest Analysis and Inventory Branch, 2018). High precision GPS was used to
obtain two measurements of ± 2m accuracy from the center of each plot.
ALS Data collection
Airborne LiDAR was collected during a leaf-on condition between July and August of
2017.Data was collected with a minimum density of 2 pulses/m2 with a half-scan angle of 12.5
degrees from NADIR and 50% overlap. There is an average density of 8.4 points per m2 and 5.5
pulses per m2, with a total of 15.22 billion points covering 1811 km2. The ALS footprint had a
range of between 30 to 70 cm. Empirical training plots in 2018 were designed to capture the
variation observed from the airborne LiDAR cloud points.
ALS data processing
ALS processing was done using the lidR package in R (Roussel & Auty, 2022), which
contains tools for processing and analyzing ALS data. Tiled ALS data was imported into R and
indexed to increase processing speed. The data was normalized using the normalize height
function in lidR and ground classification was done using a triangulated irregular network (TIN)
model. Normalization is a process of adjusting the elevations of points in an ALS point cloud to
a common reference surface. Noise points were then classified with the lidR noise classification
function, then reprocessed with the removal of points with elevations above 50 m or below 0 m.
93 | P a g e

Predictive modelling
Using the clipped ALS data from the field plots, a random forest non-parametric
approach was used to model the estimated plot fuel attributes using ALS derived independent
variables from the clipped ALS data. Random forest is a non-linear regression tree model that
functions by aggregating the results of groups of decision trees that have been fit to a subsample
of the data. It uses bootstrap aggregation and randomization of predictors to gain a high degree
of predictive accuracy (Rigatti, 2017). It is a useful method because it corrects for overfitting to
training sets, and for the most part outperforms decision trees (Breiman, 2001). In addition, it
does not make a priori assumptions regarding the underlying distribution or structure of data and
can theoretically be applied to new ALS data if the metrics are within the range of the original
data (White et al., 2013). A k-fold cross-validation technique was used to assess the performance
of the random forest models, using 5-folds (Anguita et al., 2012).
Random forest models were fit using the 65 ALS predictive metrics with the ranger
engine within the tidymodels framework in R (Kuhn & Wickham, 2020). The data was split into
two parts: a training set and a test set. The training set, comprising 90% of the data, was used to
develop the random forest model, while the test set was used to evaluate its performance. The
ranger engine used a regression tree framework with 10,000 trees, with 5 variables selected at
each split and at least 3 samples needed to create a terminal node.
The effectiveness of random forest models is presented in terms of out of bag r-squared
(OOB r2), cross-validation r-squared (cv-r2), and cross-validation root mean squared error (cv
RMSE) statistics. The OOB r2 is calculated by assessing the model’s performance on the out-ofbag data, which are data points not used in the decision trees in the random forest. For each
decision tree the OOB data are used to calculate the R-squared as a measure of how well the
model’s predictions match the actual target value for those OOB data points (Schonlau & Zou,
94 | P a g e

2020). The CV r2 is a metric used to assess the generalization performance of the model. Crossvalidation is a process of splitting the data into multiple subsets (folds) and training and
evaluating the model on different subsets iteratively (Yates et al., 2022). For the CV r2, the value
is calculated for each fold, and the average R-squared value across all of the folds is taken as the
CV R-squared. This metric estimates how well the model is expected to generalize to new,
unseen data by assessing its performance on multiple validation subsets. The CV RMSE is a
metric that quantifies the accuracy of the model’s performance, as it is an estimate of the average
prediction error of the model. The coefficient of determination (R2 or sometimes Rsquared)
resulting from a random forest model may or may not be negative, as opposed to the results of a
linear regression model (Plervis et al, 2022).
Landscape level projections of fuel loading
Landscape level projections were created for each fuel metric using the ‘predict’ function
in the raster R package (Hijmans, 2018) using the tidymodels model framework (Kuhn &
Wickham, 2020) at a 10 m by 10 m pixel size resolution. Landscape level fuel models were
prepared for: CBH (for all trees and for conifers only), CFL, CBD (all and conifer only), and
CWD. In addition, critical surface intensity (CSI) was calculated using modelled predictions for
CBH. CSI was calculated for each raster and a wall-to-wall landscape model was produced. Nonforested areas were then masked to remove out-of-scope values. The mask was set as tree height
>3m and CSI was calculated as per the formula in Beck and Simpson (2007) and formed into a
function in R as:
CSI = 0.001 ⋅ ‫ܪܤܥ‬ଵ.ହ ⋅ (460 + 25.9 ⋅ FMC)ଵ.ହ

Eq. 9

In this equation, CSI represents the critical surface intensity in kilowatts per meter
(kW/m). CBH is the canopy base height in meters, and FMC is foliar moisture content set at a

95 | P a g e

constant value of 95 (95th percentile foliar moisture content) which represents a drought
condition (BCWS, 2021).
Maps showing the spatial distribution of CFL, CBD, CBH and CSI were produced. These
maps were then evaluated to assess the potential for reducing crown fire hazard. These maps can
be used to identify high risk areas and forest management units that are suitable for treatments to
reduce CBD below 0.1 kg/m3 and/or increase CBH (Gonzalez-Ferreiro et al., 2017).
Results
Tree and Stand Metrics
ALS-derived models of tree and stand characteristics performed well across the metrics I
evaluated, with cross-validated r2 values ranging from 0.30 to 0.83 (Table 13). Models of
individual tree height estimates performed very well (i.e. mean tree height and 90th percentile
height, Table 13), while estimates of tree diameter were slightly lower (i.e. mean DBH and
quadratic mean diameter, Table 13). Stand basal area and volume were estimated with
reasonably high model accuracy (CV r2 of 0.65 and 0.52, respectively), but stems per hectare
estimates had poor accuracy (CV r2 of 0.30).

96 | P a g e

Table 13. Evaluation of model accuracy for individual tree and forest stand metrics. For each
random forest model run the ‘out of bag’ r2 (OOB r2), cross-validated r2 (CV r2), and averaged
mean square error from the k-fold cross validation runs are reported.
Response variable

OOB r2

CV r2

CV RMSE

Mean tree height (m)

0.69

0.71

2.30

90th Percentile of Tree height (m)

0.82

0.83

2.78

Mean DBH (cm)

0.51

0.56

4.05

Basal Area (m2/ha)

0.58

0.65

11.54

Density, stems per hectare (sph)

0.28

0.30

1,292.53

Quadratic Mean Diameter (cm)

0.60

0.63

4.11

Volume per hectare (m3/ha)

0.44

0.52

181.54

Commercial volume per hectare (m3/ha)

0.30

0.40

88.41

The predictive model for average tree height in the stand types achieved a crossvalidation R-squared (cv-r2) value of 0.71 and a root-mean-square error (RMSE) of 2.3 (Table
13). These results indicate that this model has good predictive power for stand tree height
estimation based on the ALS variables used: most predicted values falling within +/- 2 meters of
the observed values. However, the height of taller trees in the data set was generally
underestimated. There are notable differences in the performance of the model across stand
types, with CL stand type showing the most accurate predictions and CD stand type showing the
least accurate predictions. Canopy height, a stand level attribute, measured as the 90th percentile
of the recorded tree heights was modeled with higher accuracy (Table 13, CV r2 = 0.83).

97 | P a g e

Figure 22. Observed versus predicted mean tree height for each sample plot. Plots are color
coded by stand type, and the full model cross-validated r2 value is reported.

98 | P a g e

Figure 23. Observed versus predicted mean tree 90th percentile of tree height for each sample
plot. Plots are color coded by stand type, and the full model cross-validated r2 value is reported.
The random forest model used to predict mean DBH achieved a cross-validation Rsquared (cv-r2) value of 0.56 and a root-mean-square error (RMSE) of 4.0 (Table 13). These
results suggest that the model has good predictive power for DBH estimation based on the ALS
input variables used (Figure 24). The predicted versus observed plot for DBH shows some
evidence of bias, with smaller average DBH values being overestimated and more variability for
larger average DBH samples. The variability in the plot is good relative to the metric, with most
99 | P a g e

predicted values falling within +/- 4 centimeters of the observed values. Overall, the model
captures the general trend in the data but overestimates the size of smaller trees.

Figure 24. Observed versus predicted mean tree DBH for each sample plot. Plots are color
coded by stand type, and the full model cross-validated r2 value is reported.

The random forest model prediction for basal area achieved a cross-validation R-squared
(cv-r-squared) value of 0.58 and a root-mean-square error (RMSE) of 11. These results suggest
100 | P a g e

that our model has good predictive power for basal area per hectare estimation based on the ALS
input variables used. The variability in the plot is good relative to the metric, with most predicted
values falling within +/- 11 m2 per ha of the observed values (Figure 25). Overall, the model
captures the general trend in the data but underestimates the basal area of larger samples and
overestimates most smaller samples.

Figure 25. Observed versus predicted mean basal area per hectare for each sample plot. Plots
are color coded by stand type, and the full model cross-validated r2 value is reported.
101 | P a g e

Crown & Canopy Length
Mean Crown Length
The model of mean crown length (Figure 26) had a lower cv-r2 accuracy but captured
differences in crown length across the measured stand types.

Figure 26. Observed versus predicted mean canopy crown length for each sample plot. Plots are
color coded by stand type, and the full model cross-validated r2 value is reported.

102 | P a g e

Canopy Fuel Metrics
Models of canopy fuel amounts were reasonably accurate (cross-validated r2 ranging
from 0.25 to 0.48; Table 14). Estimates of total canopy fuel load (CFL) were more accurately
modeled, while models of canopy bulk density (CBD) had lower accuracy.
Table 14. Evaluation of model accuracy for canopy fuel metrics. For each random forest model
run the ‘out of bag’ r2 (OOB r2), cross-validated r2 (CV r2), and averaged mean square error
from the k-fold cross validation runs are reported.
OOB r2

CV r2

CV RMSE

CFL (kg/m2)

0.38

0.48

0.78

Conifer CFL (kg/m2)

0.18

0.29

0.79

CBD (kg/m3)

0.21

0.25

0.06

Conifer CBD (kg/m3)

0.21

0.25

0.06

Height of Max CBD (m)

0.51

0.51

3.53

Response variable

The random forest models used to predict CFL achieved a cross-validation R-squared
(CV r2) value of 0.45 and a root-mean-square error (RMSE) of 0.89 (Figure 27). For conifer
fuels only, the models cross-validated r2 was about half this at 0.28 (RMSE 0.89). Bias is evident
when considering dead conifer stands. The majority of dead conifer CFL estimates are
underestimated. Estimates for conifer stands are well balanced from 0 to approximately 3 kg/m2;
above 3 kg/m2. The variability in the plot is good relative to the metric, with most predicted
values falling within +/- 0.8 kg of the observed values. Overall, the model captures the general
trend in the data but struggles to predict a few outliers at larger fuel loads.

103 | P a g e

Figure 27. Observed versus predicted crown fuel load for each sample plot. Plots are color
coded by stand type, and the full model cross-validated r2 value is reported.
The random forest models used to predict CBD achieved a cross-validation r2 value of
0.24 and a root-mean-square error (RMSE) of 0.12 (Figure 28). For conifer fuels only, the
model’s cross-validation r2 was only slightly less at 0.21 (RMSE 0.11). Model bias was similar
to CFL; the majority of conifer dead CFL estimates are overestimated, smaller conifer live stand
fuel loads are overestimated, and mixedwood stand CBD are overestimated. Similar to CFL, the
104 | P a g e

models generally underestimate CBD above 0.17 kg/m3.
The four stand types differed with respect to CFL, CBD, and CBH and crown length.
As shown in Figure 28 live conifer stands have the highest CBD and the same CBH as
the other stands (conifer component). Mixedwood stands have the second highest CBD and the
same CBH as the other stands. Dead conifer stands have the second lowest conifer CBD, and the
same CBH as the other stands. This essentially tells us that the average CBH of the four stand
types is about the same, while CBD varies.

105 | P a g e

Figure 28. Observed versus predicted canopy bulk density for each sample plot. Plots are color
coded by stand type, and the full model cross-validated r2 value is reported.
The height of the maximum CBD in a plot was modeled with good accuracy and
achieved a 0.53 cross-validation r2 value (RMSE: 3.85). The model results suggest that conifer
live, conifer dead and deciduous stands are accurately estimated across the range of observed
CBD values, while CBD estimates for mixedwood are less accurate for high and low amounts.
Most conifer stands, both live and dead, have low CBD peaks due to crown form, while
106 | P a g e

deciduous have high CBD peaks.

Figure 29. Observed versus predicted height of maximum CBD for each sample plot. Plots are
color coded by stand type, and the full model cross-validated r2 value is reported.
Canopy Base Height Metrics
Six canopy base height metrics were modelled: Mean crown base height, mean conifer
crown base height, canopy and conifer base height at 0.011 kg/m3 CBD threshold and at 0.037
kg/m3 CBD threshold. Other than mean height to live crown, the model performance was poor
107 | P a g e

for these metrics (Table 15). Mean crown base height achieved a cross-validation r2 of 0.75
(Figure 30), while conifer only fuels achieved a cross-validation r2 of 0.28 (RMSE: 1.86).
Canopy base height using a threshold of 0.011 had a cross-validation r2 of 0.16 (RMSE: 0.89),
and 0.15 (RMSE: 1.02) for all and conifer respectively. Canopy base height using a threshold of
0.037 threshold was slightly better with a cross-validation r2 of 0.20 (RMSE: 2.31), and 0.14
(RMSE: 2.04) for all and conifer respectively (Table 15). The negative out-of-bag r-squared
results for CBH at 0.011 and 0.037 indicate the initial model (without cross-validation) results in
a worse fit than a simple intercept as a result of overfitting or insufficient data. The assumption
that r2 cannot be negative doesn’t apply when models are non-linear and are evaluated separately
on train and test data (Plervis et al., 2022). Cross-validation is one technique to improve model
performance in this situation. Indeed, once cross-validated, the random forest model was able to
capture approximately 16% for the variation for the first metric and 21% for the second.
Table 15. ALS canopy base height metrics random forest model results.
OOB r2

CV r2

CV RMSE

Mean Height to Live Crown (m)

0.72

0.75

1.60

Conifer Mean Height to Live Crown (m)

0.27

0.25

1.84

CBH at 0.011 kg per m3 (m)

-0.09

0.16

0.90

CBH at 0.037 kg per m3 (m)

0.01

0.21

2.27

Conifer CBH at 0.011 kg/m3 (m)

-0.05

0.16

1.02

Conifer CBH at 0.037 kg/m3 (m)

0.04

0.14

2.04

Response variable

108 | P a g e

Figure 30. Observed versus predicted mean height to live crown for each sample plot. Plots are
color coded by stand type, and the full model cross-validated r2 value is reported.
Canopy Base Height at 0.011 and 0.037 kg/m3 CBD thresholds.
The random forest models achieved a cross-validation r2 of 0.16 (RMSE 0.89) for canopy
base-height based on a 0.011 kg/m3 CBD threshold (Figure 31). For conifer fuels only, the crossvalidation r2 was slightly less at 0.14 (RMSE 2.04). For the 0.037 threshold (Figure 32), the
models did slightly better at 0.2 cross-validation r2 (RMSE 2.31).
109 | P a g e

Figure 31. Observed versus predicted canopy base height using the 0.011 kg/m3 threshold for
each sample plot. Plots are color coded by stand type, and the full model cross-validated r2 value
is reported.

110 | P a g e

Figure 32. Observed versus predicted canopy base height using the 0.037 kg/m3 threshold for
each sample plot. Plots are color coded by stand type, and the full model cross-validated r2 value
is reported.
Canopy Length Metrics
The random forest model used to predict mean crown length achieved a cross-validation
r2 value of 0.36 and a root-mean-square error (RMSE) of 2.3. These results suggest that our
model has poor to moderate predictive power for average canopy length based on the ALS input
variables (Table 16, Figure 33). At crown lengths below 5 m, the models overestimated the
111 | P a g e

response variable, and above lengths of 10 m they underestimated. There was evidence of bias in
deciduous stand types, as all but one deciduous plot was overestimated. The remaining stand
types are well distributed above and below the one-to-one line. Overall, the model captures the
general trend in the data between 5 and 10m but struggles to predict below and above these
values. For conifer fuels only, the models do slightly better with a cross-validation r2 of 0.41, but
slightly higher RMSA (2.8) (Table 16).
Table 16. ALS canopy length metrics random forest model results.
OOB r2

CV r2

CV RMSE

Average Crown Length (m)

0.26

0.36

2.26

Average Conifer Crown Length (m)

0.34

0.41

2.78

Canopy Length at 0.011 kg/m3 CBD (m)

0.73

0.74

3.52

Canopy Length at 0.037 kg/m3 CBD (m)

0.55

0.61

4.17

Response variable

112 | P a g e

Figure 33. Observed versus predicted mean crown length for each sample plot. Plots are color
coded by stand type, and the full model cross-validated r2 value is reported.
When considering canopy length as the point at which a 0.011 or 0.037 kg/m3 CBD
threshold starts, the models do much better at predicting canopy length (Figure 34, Figure 35).
For the 0.011 threshold value, a cross-validation r2 of 0.72 (RMSE 3.7) was achieved, and for the
0.037 threshold, a cross-validation r2 of 0.54 (RMSE 4.5) was achieved. These results suggest
that our model has good predictive power for CBD threshold-based canopy length using ALS
113 | P a g e

input variables. Conifer live stand types are distributed well along the line, whereas dead conifer
stands are almost all overestimated. Shorter deciduous stand type canopy lengths were
overestimated while longer deciduous stand type canopy lengths were underestimated, indicating
bias. The model captures the general trend in the data between 5 and 10 m but struggles to
predict dead conifer and deciduous stand types (Figure 34, Figure 35).

114 | P a g e

Figure 34. Observed versus predicted canopy length using the 0.011 kg/m3 CBD threshold for
each sample plot. Plots are color coded by stand type, and the full model cross-validated r2 value
is reported.

115 | P a g e

Figure 35. Observed versus predicted Canopy Length using the 0.037 kg/m3 CBD threshold for
each sample plot. Plots are color coded by stand type, and the full model cross-validated r2 value
is reported.
Surface Fuel Metrics
Model accuracy for surface fuel loads was generally good (Table 17), but the number of
empirical plots measured was small (n = 33).
When modeling the coarse woody debris biomass (kg/m2), the random forest models
achieved a cross-validation r2 of 0.38 (RMSE 4.15). These results suggest that our model has
116 | P a g e

moderate predictive power for CWD using the selected ALS input variables (Figure 36). For
deciduous and mixedwood stands the models did well at lower biomass range, from about 5 to
7.5 kg, however above this almost all samples are underestimated. The model captures the
general trend in the data but not at high biomass levels. Transforming the CWD values to log
CWD improves the fit (Figure 37), producing a cross-validation r2 of 0.47 (RMSE: 0.89) (Table
17).
Table 17. ALS surface fuel load (CWD) metrics random forest model results.
OOB r2

CV r2

CV RMSE

CWD kg/m2

0.28

0.38

4.15

Log CWD (kg/m2)

0.32

0.47

0.89

Response variable

117 | P a g e

Figure 36. Observed versus predicted coarse woody debris for each sample plot. Plots are color
coded by stand type, and the full model cross-validated r2 value is reported.

118 | P a g e

Figure 37. Observed versus predicted log CWD for each sample plot. Plots are color coded by
stand type, and the full model cross-validated r2 value is reported.
ABA Wall-to-wall results
The models described above were used to develop wall-to-wall projections for stand
characteristics and the modeled forest fuel components. The following maps showcase these
landscape level fuel load models (Figure 38 to Figure 49). I have selected only a subset of
metrics for display below including basal area, CBD, conifer CBH and calculated CSI.
119 | P a g e

The figures below (Figures 38-49) showcase several critical facets of ALS-derived fuel
maps. First and foremost, ALS enables a significantly finer-scale differentiation of forest
structure, demonstrated by metrics like basal area (m2/ha), compared to conventional tools like
VRI. This heightened resolution enhances our ability to discern more nuanced details of forest
structure as well as landscape trends.
In addition, the fine-grained characterization of fuel loading, particularly represented by
canopy bulk density (CBD) and canopy base height (CBH), stands out as an advantage of ALS.
It allows for a detailed understanding of fuel distribution across a large landscape extent as well
as vertical complexity, providing valuable insights for fire management particularly where crown
involvement may arise.
The versatility of ALS-derived fuel estimates is evident in their applicability to both large
and small-scale decision-making. Figures 41, 42, and 48 demonstrate how ALS data can inform
management decisions at a landscape level, while also offering detailed insights for finer-scale
considerations (Figure 46) such as the seamless integration of ALS layers into existing
management and planning frameworks. For instance, the mapping of wildland-urban interface
(WUI) locations (Figure 45-47) is enhanced, contributing to more robust planning processes.
In summary, the figures illustrate the effectiveness of ALS-derived fuel data in enhancing
landscape-level forest planning, providing consistent and detailed information for identifying
areas of varying risk levels and supporting informed decision-making.

120 | P a g e

Figure 38. Predicted basal area per hectare for entire study area at 10 m resolution. Blue
represents high basal area (>46 m2/ha) and yellow represents very low (<12 m2/ha). Wildland
Urban Interface polygons are also shown.

121 | P a g e

Figure 39. Babine compartment predicted basal area per hectare. The area defined by the
smaller red square is shown in higher resolution in Figure 40.

122 | P a g e

Figure 40. Red square inset (1km square) from previous figure showing predicted basal area per
hectare.

123 | P a g e

Figure 41. Predicted canopy bulk density for Southside compartment at 10 m resolution. CBD
varies from 0 to approximately 0.3 kg/m3.

124 | P a g e

Figure 42. Predicted canopy bulk density for Roselake compartment at 10 m resolution. CBD
varies from 0 to approximately 0.3 kg/m3. The majority of areas with CBH higher than 0.10
kg/m3 are along the southwest of the compartment although there are predicted CBDs higher
than 0.10 throughout the area.

125 | P a g e

Figure 43. Predicted canopy bulk density for Babine compartment at 10 m resolution. CBD
varies from 0 to approximately 0.2 kg/m3 with large portions of the area predicted as having
CBD above 0.10 kg/m3.

126 | P a g e

Figure 44. Predicted canopy bulk density for Hannay compartment at 10m resolution. CBD
varies from 0 to approximately 0.3 kg/m3 with concentrations predicted to be higher than 0.10
primarily in the northern portions of this compartment.

127 | P a g e

Figure 45. Predicted CBD in the Southside compartment showing wildland urban interface
(WUI) areas (pink polygons). Predictions indicate high CBD within WUIs in this area.
128 | P a g e

Figure 46. Predicted CBD from previous figure, zoomed showing WUI boundary and predicted
CBD inside and outside of the WUI. Note contiguous nature of CBD within WUI.
129 | P a g e

Figure 47. Predicted CWD in kg/m2 for an area within the Southside compartment. Note that
anything yellow, orange or red is considered high CWD fuel load.
130 | P a g e

Figure 48. Critical surface intensity (CSI) calculated from predicted conifer canopy base height
with constant FMC of 95 for the entire study area. Red indicates that the critical surface fire
intensity is predicted to be low (i.e. surface fire can easily transition into crown fire) based on
low predicted canopy base heights. The yellow square (at the bottom of the figure) shows the
areas that is presented at a higher resolution in Figure 49.
131 | P a g e

Figure 49. Close up of yellow inset from previous figure, showing calculated CSI from predicted
conifer canopy base height.
132 | P a g e

Discussion
My results indicate that many of the forest metrics important for fuel load
characterization can be modeled with reasonably high accuracy and mapped at a fine grain (i.e.
100 m2). To achieve this, I used 65 predictor variables, 54 of which are standard metrics
produced by the lidR package and nine of them are custom metrics describing leaf area
characteristics.
The models did well at predicting some basic stand metrics but performed poorly for
others. Average tree height cv-r2 was 0.71 and basal area per hectare had a cv-r2 of 0.65. The
result for the 90th percentile of tree height was very good at 0.82 cv-r2. CV-r2 values for mean
DBH prediction was 0.56 and stems per hectare was surprisingly quite low at 0.30. Predicted
volume per hectare CV-r2 was 0.51, and conifer volume per hectare was at 0.41 cv-r2.
The basic stand metrics primarily used z (point height) statistics for prediction. The
quantiles of the height distribution dominated the top 5 variables of importance for all basic
metrics with the exception of stems per hectare (See appendix for variable importance plots),
which relied on ALS- derived leaf area estimates, cumulative point density data and return
intensity. Although I achieved good predictive power for height-based metrics and those closely
correlated with height, the models fell short compared to some other researchers. For instance,
Dalponte et al. (2018) applied regression models and ALS in similar stands in Norway, reporting
r2 values for stems per hectare between 0.51 and 0.78. Means et al. (1999) achieved r2 values
exceeding 0.9 for height, basal area, total biomass, and leaf biomass. LaRue et al. (2020) found
similar r2 values for mean height, although they did not use a random forest method. Some
researchers have adopted an approach that involves removing predictor variables with low
correlation to enhance predictive power in their models (Sherrill et al., 2008). However, in my
analysis, I observed very little advantage in implementing such a strategy. Although there might
133 | P a g e

be potential benefits in creating further custom metrics, the ones I utilized rarely emerged among
the top 5 variables most important for predicting the basic forest stand metrics. This means that
although the custom metrics used did add value, they may be correlated with the standard metrics
and so might be redundant in the models. The standard metrics seemingly do a good job of
capturing the relevant attributes for the majority of the fuel load metrics studied here.
While individual tree methods, as opposed to ABA, may produce better results for small
areas and specific metrics (Yu et al., 2010), it is important to acknowledge that these methods
require considerably increased computing power, and have historically been restricted to
developing models for smaller land areas. In addition, it is worth noting that the accuracy of
ALS-based models is good for evaluating upper canopy structure, but accuracy decreases as you
descend through the canopy.
Fuel metrics
The canopy fuel load (CFL) was accurately modeled by the explanatory variables up to
approximately 3 kg/m2. However, in dead conifer (CD) stands, CFL was slightly overestimated
by the models, while for live conifer stands above 3 kg/m2, the models underestimated CFL.
Canopy bulk density (CBD) was generally modeled accurately, but at lower values of
CBD, the models overestimated for all stand types. The height of the maximum CBD was well
predicted, with similar overestimation of lower values and underestimation of higher values.
There was very little bias evident for the different stand types considered. Although all results
showed some slight bias, the relationship was very good for most metrics.
These findings align with the accuracy of canopy fuel models reported in past studies.
For instance, Engelstad et al. (2019) achieved an r2 of 0.48 (RMSE: 0.09 kg/m3) using lowdensity LiDAR and a random forest model in pine, spruce, birch, and aspen forests in Minnesota.

134 | P a g e

Bright et al. (2017) found r2 values of 0.56 for CFL, 0.46 for CBD, 0.28 for CBH, and 0.66 for
CH, along with 0.3 for 1,000h surface fuel. Hermosilla et al. (2013) reported adjusted r2 values
of CFL (0.79), CBD (0.67), CBH (0.78), CH (0.79), and BA (0.76).
Canopy base height (CBH) was accurately estimated, although differences were observed
between the stand types, with canopy base height of deciduous stands consistently being
underestimated. Similar findings have been reported by various researchers. For instance,
Botequim et al. (2019) found r2 for ALS prediction of CBH in pine and oak stands in Spain of
between 0.75 and 0.98. Luo et al. (2018) assessed the ability of ALS to capture the CBH as well
(height to live branch) using an individual tree method and reported r2 values of between 0.82
and 0.9 for conifer species, showing general applicability across mixed and uniform stands.
Kelly et al. (2017) found plot level CBH in mixed pine in California of r2 = 0.51. Using ALS
metrics, Hevia et al. (2016) achieved an r2 of 0.94 in pure and even aged pine in northwest Spain
(Hevia et al., 2016) while Stefanidou et al. (2020) found r2 of 0.61 (RMSE of 18.19%).
Many have attempted to model CBH defined as the average height to live crown, but to
my knowledge, only Engelstad et al. (2019) modeled the threshold value canopy base heights
proposed by Scott and Reinhardt (2002) using the 0.011 kg/m3 threshold, and to my knowledge,
no research has modelled the threshold proposed by Van Wagner (1977) which uses the 0.037
kg/m3 threshold using similar methods.
The models used in this study exhibited limitations in predicting CBH at both the 0.011
and 0.037 kg/m3 CBH thresholds. In contrast, Engelstad et al. (2019) achieved an r2 of 0.78 for
CBH (RMSE: 1.10m). The majority of stands in this study had CBH 0.011 kg/m3 height values
of between 1 and 3 m; for the wider canopy base height range for the 0.037 kg/m3 threshold,
most stands had 0.037 kg/m3 threshold heights of between 0 and 5 m, and some dead conifer and

135 | P a g e

deciduous stands exhibiting CBH of 5 to 15 m. The generally low-to-ground values of thresholdbased CBH exceeding the specified thresholds suggest the possibility that all stands in this region
may have canopy fuel loads close enough to the ground to raise concerns. However, further
research is needed to investigate and confirm this observation.
Modeling average crown length in areas where the crown was shorter than 5 m presented
challenges. However, for plots with a crown length between 5 and 10 m there was better fit.
Average crown lengths in deciduous stands were overestimated. Crown length using the 0.011
kg/m3 threshold value has a much better fit with a cv-r2 of 0.74, exhibiting little bias across stand
types. A similar result was found for crown lengths between the 0.037 kg/m3 threshold values,
though with a lower cv-r2. When using the CBD threshold of 0.011 kg/m3 the start and end
points are more easily detectable with the ALS metrics than the at the 0.037 kg/m3 points, which
were more ‘buried’ in the noise of the point clusters. The majority of the variables of importance
for the two threshold-based canopy length metrics are related to the end or top of the canopy,
metrics related to leaf area and pulse intensity metrics, intensity being the ratio of strength of the
returned laser pulse to that of the emitted pulse, which characterizes the reflectivity of the
material the light bounced off of (Song et al., 2002).
Coarse Woody Debris
The models for coarse woody debris (CWD) in this study yielded higher accuracy than
expected, especially considering that CWD is often concealed under the forest. Surface fuels,
such as CWD, pose challenges for ALS due to the limited penetration of light through foliage
(Véga, 2016). The random forest models achieved a cv-r2 of 0.40 using only 33 sample plots and
the logarithm CWD a cv-r2 of 0.47, suggesting the possibility of some relationships were not
fully captured by the random forest model. Deciduous stands were overestimated, while dead

136 | P a g e

conifer stands were underestimated. Log CWD exhibited a better fit at larger values of CWD,
with conifer dead (CD) points only slightly underestimated. Mixedwood and live conifer stands
showed increased model accuracy using Log of CWD. The substantial amounts of CWD in dead
conifer stands may have impacted the model’s fit. The model for CWD relies principally on
cumulative point densities data at 2, 3, and 4 m above the ground. In a similar context, Shokirov
et al. (2021) used terrestrial laser scanning to model CWD over a landscape scale and found that
digital surface model (DSM), surface roughness and topographic position were variables of
importance in models for their study area, achieving a variable response ranging from 20 to 96%.
In the Canadian boreal forest, Lopes Queiroz et al. (2020) achieved an r2 value of 0.62 (RMSE:
0.224 m3/100 m2). These results demonstrate that fuel loading metrics can be estimated quite
well using ABA, but differences may exist based on the height in the canopy and ability to be
captured by ALS. The models in this study demonstrated good performance across all forest
types with little systematic bias, suggesting that these plot level estimates can be used to make
projections at the landscape level for this case study region.
Landscape results
Fire risk and fuel load hazard assessment has typically been associated with inventory
polygons and field collected data (Perrakis et al., 2018). Using this coarse data results in the BC
Provincial government PSTA layer being a strategic tool rather than a tactical or operational tool.
However, operational and site level assessments lack the landscape picture provided by the
PSTA and are costly due to the necessity for field data collection and lack accuracy within and
across stand types. As shown in the landscape fuel load maps above, fuel loads appear to be
lower in some WUI buffer areas (e.g. Figure 46 ) but there are areas with high CBD and low
CBD within and adjacent to WUI buffers.

137 | P a g e

Conclusion
I have demonstrated that the ABA approach is broadly applicable across the stand types
in the study area, enabling the projection of fuel load metrics without being constrained by
different stand types. The ABA approach applied here facilitates a finer resolution evaluation of
fuel loads than has traditionally been possible. Previously, forest types were often grouped
together due to the coarse scale of stand differentiation within the Canadian fuel classification
system. In contrast, the area-based ALS system that I have advanced provides a more accurate
and comprehensive estimate of forest fuels and hazard by allowing for the assessment of the
quantity and connectivity of risky stands.
In summary, I have demonstrated that ALS can be employed to provide more detailed
landscape-scale estimates of fuel loading, aiming to enhance and inform forest management
strategies for wildfire mitigation in the short and long term. Moreover, by separately estimating
specific stand and fuel load metrics, we gain a more nuanced set of options and management
actions for wildfire mitigation. This approach enables a targeted focus on areas with high canopy
bulk density (CBD), low canopy base height (CBH), and extensive, but irregular distribution of
areas of coarse woody debris (CWD) accumulation. Additionally, we can implement thinning in
areas with high basal area, pruning where CBH is low, and address surface loads where they
exceed safe maximums. This nuanced approach holds particular importance for fostering greater
understanding among land managers, the public, and First Nations communities.

138 | P a g e

CHAPTER 4 SYNTHESIS
In chapter 2 of this thesis, I have demonstrated that there are differences in fuel
characteristics between different forest types, and that these stand characteristics should be
incorporated into fire risk assessment and mid- to long-term risk mitigation planning. Secondly
in chapter 3, I have advanced a method of using remote sensing to provide high resolution
estimates of individual fuel components, providing land managers with more tools to quantify,
understand, and manage wildfire risks.
Wildfires are a major threat to communities (Erni et al., 2023). Climate change is
compounding wildfire risk (Haughian et al., 2012), and we are seeing more communities and
lives impacted every year. It is no longer unprecedented to see towns and cities completely
overrun by fire or evacuated (Tepley et al., 2022). Effective, and operationally feasible,
management of wildfire risk is hampered by an incomplete understanding of fuel loads in this
province. More understanding of how much fuel there is and how it is distributed is critical to
address and manage the hazard fuels create. Current forest inventories lack the details required
to make informed decisions. New technologies are becoming available that can quantify both
surface and canopy fuel loads in greater detail (Labenski et al., 2023), and provide both high
resolution detail and landscape-level context to how hazard affects us and our communities and
things we value. As more of the province is covered with ALS, there is an opportunity to use this
data for wildfire mitigation and fuel management. Every year we spend millions of dollars
fighting fires; that money could be put into strategic fuel mitigation using ALS data to quantify
and spatially locate hazardous fuel loads and contiguous fuel corridors.
Chapter 2 of my thesis addresses questions about our current stand types and how they
compare with each other regarding fuel loading and forest structure, as well as whether they
139 | P a g e

should be considered hazardous with regard to wildfire risk. In north-central BC, forest stands
are dominated by live and dead lodgepole pine and hybrid spruce with trembling aspen. The
mature lodgepole pine component was heavily impacted by mountain pine beetle in the first and
second decades of the 21st century (Talucci et al., 2022). Where these mature stands once stood,
there are now either standing dead pine, dense plantations of either spruce, pine, or in many
cases, dead standing and dead fallen pine trees with ingress of pine, other conifers, or to some
extent deciduous species such as aspen. It is a complex environment not easily categorized into
neat fire-risk assessment boxes.
In this thesis I collected data from four stand types and evaluated basic stand structure
metrics and fuel load metrics for each. This evaluation included tree height, DBH, basal area,
stand density, merchantable volume per hectare, and CFL, CBD, CBH, CWD, and CL. The main
underlying questions for this chapter were whether these stands differed with regard to their fuel
load structure, whether dead conifer stands constituted high fire risk hazards, and whether
mixedwood stands represented a reduced fire risk. I found that there are considerable differences
between the four stand types with respect to both basic stand structure and fuel loading
characteristics. In terms of basic stand metrics, there are the expected differences in average tree
metrics, species, tree and stand structure and tree densities that nonetheless impact fuel load.
Live conifer stands were found to be shorter than the other stand types but have the highest stem
densities, resulting in more foliage and lower live crown base heights, two characteristics that
can facilitate a ground fire transitioning into the canopy (Van Wagner, 1977). Mixed stands had
the highest basal area, and mid-range densities. However, in the mixedwood stands some of the
stems were either dead with no crown foliage or deciduous with much higher crown base
heights. These tree and stand structure differences tie directly to differences in fuel loads, and the

140 | P a g e

likelihood of a crown fire initiating and being propagated.
Importantly, my results demonstrate that live conifer stands have the highest CFL, CBD
and the lowest CBH, indicating that they are the most hazardous stand type. Dead conifer stands
have low CFL, CBD, but very high CWD, suggesting that although their canopy fuel load is
negligible, they still should be considered as having a hazardous fuel load due to their surface
fuels. Fuel load attributes by themselves cannot be regarded without spatial context. The
intermixed nature of the high canopy load and ‘no canopy load but high surface load’ stands,
evident by projecting fuel loads across space vertically and horizontally, provides a more
complete estimate of how different fuel hazards combine into an emergent landscape hazard.
Mixedwood stands have slightly lower canopy fuel loads than live conifer stands and comparable
CWD and CBH. If only conifer fuels are considered, deciduous stands have the lowest fuel
loads of all stand types. However, if all components are being considered, their fuel loads are
also considerable. Live conifer stands had canopy bulk densities of conifer fuels averaging 0.10
kg/m3. Mixedwood stands averaged conifer CBD of 0.06 kg/m3, but there was no significant
difference between them. In addition, many of the mixedwood stands also had high CWD loads
and both live conifer and mixed stands had comparable canopy base heights. This suggests that
both conifer live and mixedwood stand types are hazardous fuel types, due in part to canopy fuel
density and in part due to surface fuel loads and low base heights.
In Chapter 3, I employed a method to model and predict fuel loads at a fine resolution
across a large area. While this method has been previously utilized in other locations, its
application to north-central B.C. forests was novel. The study encompassed various fuel load
metrics, revealing that, irrespective of stand type, the models effectively predicted fuel loads
with a reasonable degree of accuracy for the majority of the diverse fuel component metrics. By

141 | P a g e

incorporating numerous metrics, the study underscores the utility of this method over traditional
VRI-based hazard mapping, which offers a coarse spatial assessment and lacks the capacity to
evaluate individual components of fuel loading.
It is also important to note the differences between the CFFDRS fuel load attributes and
the estimated loads found in this study area. Mismatches in fuel types to stand types occur, as
noted by Baron et al. (2024). My results reinforce these mismatches, showing notable
differences in canopy base heights and canopy fuel loads when compared to the expected
CFFDRS fuel types in BC. This suggests that using ALS instead of VRI may provide more
detailed and informed fuel load estimates, supporting better fuel management. In western
Canada, there is a need to incorporate wildfire risk management into forest and landscape
management plans. This may include managers empirically evaluating fuel loading within stands
and developing stand level fuel reduction strategies. It may also include evaluating wildfire risk
at a broader scale and managing fuel continuity across the landscape. In order to effectively
accomplish fuel management at both these scales, land managers, government and communities
need to have access to high quality forest fuels data that is developed in a transparent way and
presented in a format (normally maps) that can be used to develop operational plans. In addition,
how these fuel load and fire risk maps are prepared needs to be transparent in order to facilitate
dialogue of how best to address hazards associated with fuel loads surrounding communities,
across the province, and between neighbors and communities. The results of this chapter
demonstrate that ALS data can provide a framework for developing operational fuel load maps
for all types of users.
The accuracy and predictive performance of the fuel models that I developed was good
both for forest structural attributes as well as for many of the fuel components. While some

142 | P a g e

metrics exhibit stronger predictive capabilities and lower errors, others may require further
refinement to enhance their accuracy and explanatory power.
The mean height to live crown, a method used to describe the canopy base height,
demonstrated effective modeling when considering all forest stand types. However, the accuracy
decreased considerably when restructuring the analysis to conifer fuels alone. Canopy base
height at specific thresholds, such as 0.011 or 0.037 kg/m3 CBD, was poorly captured,
suggesting that while these thresholds might be suitable for certain fire behaviour models
mathematically, they are inadequately modeled with ALS data. This consideration is essential in
the development of fire behaviour models, particularly when the mean height to the live crown
may not align precisely with the initiation of hazardous bulk density. In contrast, the height of
the maximum CBD was more accurately modeled, representing the peak bulk density. Thus,
employing a combination of the height of the peak bulk density and metrics like canopy fuel load
(CFL) or mean height to live crown could offer a more comprehensive understanding of fire
hazard.
Canopy fuel metrics, specifically CFL and CBD, were not exceptionally well modeled in
this study, with CFL exhibiting better performance compared to CBD. The challenge in
accurately predicting CBD, as a density metric, using ALS data, a 3D point cloud, is an
important consideration, particularly as CBD is often a key metric in fire behaviour models.
However, there is potential for improvement through model refinement or adjustments in CBD
calculations. Addressing issues, particularly those related to conifer live estimates, may enhance
the accuracy of CBD predictions in future modeling efforts.
The predictive models that I developed are in raster format, where results are pixels at a
given resolution and each pixel contains a predicted value based on the ALS point cloud

143 | P a g e

structure. I used mean height to live crown as the best proxy for CBH and calculating the critical
surface intensity (BCWS, 2021), thereby providing a map of CSI at the same resolution. This
framework can be used to calculate CSI for a range of different fuel moisture content values,
therefore showing land managers locations where CSI is hazardous for any moisture scenario.
Overall, the research aims were to provide a comprehensive understanding of the factors
that contribute to wildfire risk and how these can be effectively modeled and used to inform
management. As provincial scale LiDAR availability is fast approaching, we should be prepared
to rapidly apply the ABA approach for wildfire fuel metrics in order to manage fire severity and
impacts in the short and long term at multiple scales.
This study aims to evaluate the differences in basic and fuel-related metrics between
stand types used commonly by forest managers and may not speak directly to the existing
CFFDRS fuel type analogs currently in use by local wildfire managers. There are several other
limitations to this study. Results obtained here are specific to SBSdk and SBSmc forests, and to
some extent ESSF forests having subalpine fir, interior or Engelmann spruce, lodgepole pine and
trembling aspen forest types. The samples were collected in all age classes; however, some may
be underrepresented. Unequal variances can lead to inflated type II error rates, indicating that the
study may have limited statistical power to detect real differences in some comparisons of
metrics. Due to uneven sample sizes, estimates may be biased. Type I errors may occur due to
violation of the assumption of equal variances and may lead to a biased representation of the
sample population.
Recent wildfires in BC have highlighted the broad impact that fires can have on
communities and the province. Mitigation and adaptation strategies are urgently needed to
protect communities and ecosystems, and to maintain the long-term viability of the forest sector.

144 | P a g e

The research that I have advanced in this thesis will hopefully contribute to some aspects of
improved forest management planning. As remote sensing methods improve and become more
widely available, discussions among those affected by wildland fire will be greatly informed by
applying the methodology in this paper to large areas of the BC landscape.
REFERENCES
Adams, M. A. (2013). Mega-fires, tipping points and ecosystem services: managing forests and
woodlands in an uncertain future. Forest Ecology and Management, 294, 250–261.
Adnan, S., Maltamo, M., Mehtätalo, L., Ammaturo, R. N. L., Packalen, P., & Valbuena, R.
(2021). Determining maximum entropy in 3D remote sensing height distributions and
using it to improve aboveground biomass modelling via stratification. Remote Sensing of
Environment, 260, 112464.
Affleck, D., Seielstad, C., Goodburn, J., Queen, L., & Keane, R. (2013). Characterizing crown
biomass and crown profiles in conifer forests of the interior northwest. JFSP Research
Project Reports, 36. http://digitalcommons.unl.edu/jfspresearch/36
Agee, J. K. (1996). The influence of forest structure on fire behavior. In: Proceedings of the 17th
Annual Forest Vegetation Management Conference, January 16–18, 1996, Redding, CA
(pp. 52–68).
Agee, J. K., & Skinner, C. N. (2005). Basic principles of forest fuel reduction treatments. Forest
Ecology and Management, 211(1-2), 83–96.
Ager, A. A., Vaillant, N. M., & Finney, M. A. (2010). A comparison of landscape fuel treatment
Strategies to mitigate wildland fire risk in the urban interface and preserve old forest
structure. Forest Ecology and Management, 259(8), 1556–1570.
Ahmed, O. S., Franklin, S. E., Wulder, M. A., et al. (2014). Integration of LiDAR and Landsat
data to estimate forest canopy cover in coastal British Columbia. Photogramm. Eng.
Remote Sens., 80(10), 953–961.
Akay, A. E., Oğuz, H., Karas, I. R., & Aruga, K. (2009). Using LiDAR technology in forestry
activities. Environmental Monitoring and Assessment, 151, 117–125.
Albini, F. A. (1996). Iterative solution of the radiation transport equations governing spread of
fire in wildland fuel. Combustion, Explosion and Shock Waves, 32(5), 534–543.
Aldred, A. H., & Alemdag, I. S. (1988). Guidelines for forest biomass inventory. Canadian
Forest Service, Petawawa, National Forestry Institute. Information Report. PI-X-77.
134p.

145 | P a g e

Alexander, M. E. (1988). Help with Making Crown Fire Hazard Assessments. USDA For. Serv.
Gen. Tech. Rep. INT-251, 147–53.
Alexander ME (1998) Crown Fire Thresholds in Exotic Pine Plantations of Australasia.
Australian National University, PhD thesis, Canberra.
Alexander, M. E. (2010). Surface fire spread potential in trembling aspen during summer in the
Boreal Forest Region of Canada. The Forestry Chronicle, 86(2), 200–212.
Andersen, H.-E., McGaughey, R. J., & Reutebuch, S. E. (2005). Estimating forest canopy fuel
parameters using LiDAR data. Remote Sensing of Environment, 94(4), 441–449.
Andison, D. W., & Kimmins, J. P. (1999). Scaling up to understand British Columbia’s boreal
mixedwoods. Environmental Reviews, 7(1), 19–30.
Arroyo, L. A., Pascual, C., & Manzanera, J. A. (2008). Fire models and methods to map fuel
types: the role of remote sensing. Forest Ecology and Management, 256(6), 1239–1252.
Baron, J. N., Hessburg, P. F., Parisien, M. A., Greene, G. A., Gergel, S. E., & Daniels, L. D.
(2024). Fuel types misrepresent forest structure and composition in interior British
Columbia: a way forward. Fire Ecology, 20(1), 15.
British Columbia Ministry of Forests (BC MOF) and British Columbia Ministry of Environment,
Lands and Parks (BC MELP). 1995. Forest Practices Code of British Columbia:
Biodiversity Guidebook. Queens Printer, Victoria, B.C
.http://www.for.gov.bc.ca/tasb/legsregs/fpc/fpcguide/biodiv/biotoc.htm
BC Wildfire Service. (2021). BCWS fuel management prescription guidance 2021. British
Columbia. https://www2.gov.bc.ca/assets/gov/public-safety-and-emergencyservices/wildfire-status/prevention/fire-fuel-management/fuelsmanagement/2021_fuel_management_prescription_guidance.pdf
BCWS, (n.d.). Wildfire season summary, 2023,
https://www2.gov.bc.ca/gov/content/safety/wildfire-status/about-bcws/wildfirehistory/wildfire-season-summary
Beck, J., & Simpson, B. (2007). Wildfire threat analysis and the development of a fuel
management strategy for British Columbia. Proceedings of Wildfire 2007, 1–12.
Béland, M., Widlowski, J. L., Fournier, R. A., Côté, J. F., & Verstraete, M. M. (2011).
Estimating leaf area distribution in savanna trees from terrestrial LiDAR measurements.
Agricultural and Forest Meteorology, 151(9), 1252-1266.
Beukema, S. J., Greenough, J. A., Robinson, C. E., Kurtz, W. A., Reinhardt, E. D., Crookston, N.
L., ... & Stage, A. R. (1997). An introduction to the Fire and Fuels Extension to FVS. In
Proceedings of the Forest Vegetation Simulator Conference (pp. 191–195). USDA For.
Serv. Gen. Tech. Rep. INT-GTR-373.

146 | P a g e

Bieng, M. A. N., Ginisty, C., Goreaud, F., & Perot, T. (2006). Forest typology of oak and scots
pine mixed stands in Orleans Forest (France), Based on the Canopy Spatial Structure.
New Zealand Journal of Forestry Science, 36(2/3), 325.
Bivand, R., Pebesma, E., & Gomez-Rubio, V. (2013). Applied Spatial Data Analysis with R (2nd
ed.). Springer. https://asdar-book.org/
Botequim, B., Fernandes, P. M., Borges, J. G., González-Ferreiro, E., & Guerra-Hernández, J.
(2019). Improving silvicultural practices for Mediterranean forests through fire behaviour
modelling using LiDAR-derived canopy fuel characteristics. International Journal of
Wildland Fire, 28(11), 823-839.
Breiman, L. (2001). Random forests. Machine Learning, 45(1): 5–32.
Bright, B. C., Hudak, A. T., Meddens, A. J. H., Hawbaker, T. J., Briggs, J. S., & Kennedy, R. E.
(2017). Prediction of forest canopy and surface fuels from LiDAR and satellite time
series data in a bark beetle-affected forest. Forests, 8(9), 322.
Brown, J. K. (1974). Handbook for Inventorying Downed Woody Material (General Technical
Report INT-16, U.S. Forest Service), Ogden, UT.
Brown, J. K. (1978). Weight and Density of Crowns of Rocky Mountain Conifers. USDA Forest
Service Research Paper INT (USA), No. 197.
Brown, J. K. (2003). Coarse Woody Debris: Managing Benefits and Fire Hazard in the
Recovering Forest. Gen. Tech. Rep. RMRS-GTR-105. USDA For. Serv., Ogden, UT.
Burton, P. J. (2023). Understanding spring wildfires in Canada's northern forests. Global change
biology, 29(21), 5983-5985.
Call, P. T., & Albini, F. A. (1997). Aerial and surface fuel consumption in crown Fires.
International Journal of Wildland Fire, 7(3), 259–264.
Calvo, R. C., Varo Martínez, M. Á., Ruiz Gómez, F., Ariza Salamanca, A. J., & NavarroCerrillo, R. M. (2023). Improvements of fire fuels attributes maps by integrating field
inventories, low density ALS, and satellite data in complex Mediterranean forests.
Remote Sensing, 15(8), 2023.
Cameron, H. A., Dı́az, G. M., & Beverly, J. L. (2021). Estimating canopy fuel load with
hemispherical photographs: a rapid method for opportunistic fuel documentation with
smartphones. Methods in Ecology and Evolution, 12(11), 2101–2108.
Canadian Forest Inventory Committee. (2002). National Forest Inventory Ground Sampling
Guidelines, Version 1.1. Retrieved from http://nfi.cfs.nrcan.gc.ca/index_e.html [accessed
11 May 2020].
Canadian Forestry Service (CFS). (1987). Canadian Forest Fire Danger Rating System - Users’
Guide. Canadian Forestry Service Fire Danger Group, Ottawa, Ont.

147 | P a g e

Canadian Forest Service (CFS). (n.d.). Canadian National Fire Database – Agency Fire Data.
Natural Resources Canada, Canadian Forest Service, Northern Forestry Centre,
Edmonton, Alberta. Retrieved from http://cwfis.cfs.nrcan.gc.ca/en_CA/nfdb.
Carey, H., & Schumann, M. (2003). Modifying wildfire behavior-the effectiveness of fuel
treatments. The Forest Trust, 16.
Cariboo Regional District, Emergency Operations Centre. (2018). 2017 Wildfires After Action
Report. Cariboo Regional District. Williams Lake, BC.
Cavard, X., Macdonald, S. E., Bergeron, Y., & Chen, H. Y. H. (2011). Importance of
mixedwoods for biodiversity conservation: evidence for understory plants, songbirds, soil
fauna, and ectomycorrhizae in northern forests. Environmental
Champely, S. (2022). pwr: Basic Functions for Power Analysis.
R package version 1.4-1. https://CRAN.R-project.org/package=pwr
Chojnacky, D. C., Heath, L. S., & Jenkins, J. C. (2014). Updated generalized biomass equations
for North American tree species. Forestry, 87(1), 129–151.
CIFFC, Canadian Interagency Forest Fire Centre. (2003). Glossary of Forest Fire Management
Terms. Winnipeg, Manitoba, Canada, 61.
Coogan, S. C. P., Robinne, F.-N., Jain, P., & Flannigan, M. D. (2019). Scientists’ warning on
wildfire—a Canadian perspective. Canadian Journal of Forest Research, 49(9), 1015–
1023.
Corbett, L. J., Withey, P., Lantz, V. A., & Ochuodho, T. O. (2016). The economic impact of the
Mountain Pine Beetle infestation in British Columbia: provincial estimates from a CGE
analysis. Forestry: An International Journal of Forest Research, 89(1), 100–105.
Cruz, M. G., and Alexander, M. E. (2010). Assessing crown fire potential in coniferous forests of
western North America: a critique of current approaches and recent simulation studies.
International Journal of Wildland Fire, 19(4), 377–98.
Cruz, M. G., & Alexander, M. E. (2014). The start, propagation, and spread rate of crown fires.
Fire Management Today, 73(4), 17–23.
Cruz, M. G., Alexander, M. E., & Wakimoto, R. H. (2003a). Definition of a fire behavior model
evaluation protocol: A case study application to crown fire behavior models. In P. N.
Omi & L. N. Joyce (Tech. Eds.), Proceedings of the Fire, Fuel Treatments, and
Ecological Restoration Conference (pp. 49–67). USDA Forest Service, Rocky Mountain
Research Station, Proceedings RMRS-P 29. Fort Collins, CO.
Cruz, M. G., Alexander, M. E., & Wakimoto, R. H. (2003). Assessing canopy fuel stratum
characteristics in crown fire prone fuel types of western North America. International
Journal of Wildland Fire, 12(1), 39-50.
Cruz, M.G. (1999). Modeling the Initiation and Spread of Crown Fires. Master’s Thesis,
University of Montana, Missoula, MT, USA, 1999.
148 | P a g e

Cumming, S. G. (2001). Forest type and wildfire in the Alberta boreal mixedwood: what do fires
burn? Ecological Applications, 11(1), 97–110.
Curtis, P. S. (2008), Estimating aboveground carbon in live and standing dead trees. In C.M.
Hoover (Ed.), Field Measurements for Forest Carbon Monitoring., Springer, Netherlands
(2008), pp. 39-44,
Dalponte, M., Ene, L. T., Gobakken, T., Næsset, E., & Gianelle, D. (2018). Predicting selected
forest stand characteristics with multispectral ALS data. Remote Sensing, 10(4), 586.
Danson, F. M., & Bowyer, P. (2004). Estimating live fuel moisture content from remotely sensed
reflectance. Remote Sensing of Environment, 92(3), 309-321.
Day, K., Blackwell, B., & Wildeman, S. (2010). Harvesting and thinning guidance for
treatments in wildland-urban interface areas of TSA 29. FIA Project SOTSA299093008
Contract Report to Tolko Industries, Ltd. Unpublished Report.
De Assis Barros, L. (2019). Assessing Set Aside Old-Growth Forests with Airborne LiDAR
Metrics. University of Northern British Columbia: Prince George, BC, Canada.
De Cáceres, M., Martı́n-Alcón, S., González-Olabarria, J. R., & Coll, L. (2019). A general
method for the classification of forest stands using species composition and vertical and
horizontal structure. Annals of Forest Science, 76, 1–19.
De Mendiburu, F. (2021). Agricolae: Statistical procedures for agricultural research. R package
version 1.3.7. https://CRAN.R-project.org/package=agricolae.
Delacre, M., Leys, C., Mora, Y. L., & Lakens, D. (2019). Taking parametric assumptions
seriously: Arguments for the use of Welch’s f-Test instead of the classical f-test in oneway ANOVA. International Review of Social Psychology, 32(1), 1–12.
Doerr, S. H., & Santı́n, C. (2016). Global trends in wildfire and its impacts: Perceptions versus
realities in a changing world. Philosophical Transactions of the Royal Society B:
Biological Sciences, 371(1696), 20150345.
Donato, D. C., Campbell, J. L., & Franklin, J. F. (2012). Multiple successional pathways and
precocity in forest development: Can some forests be born complex? Journal of
Vegetation Science, 23(3), 576-584.
De Vries, P. G. (1973). A general theory on line intersect sampling with application to logging
residue inventory. Mededelingen Landbouwhogeschool, No. 73–11, Wageningen, The
Netherlands: Wageningen University & Research.
Engelstad, P. S., Falkowski, M., Wolter, P., Poznanovic, A., & Johnson, P. (2019). Estimating
canopy fuel attributes from low-density LiDAR. Fire, 2(3), 38.
Erdody, T. L., & Moskal, L. M. (2010). Fusion of LiDAR and imagery for estimating forest
canopy fuels. Remote Sensing of Environment, 114(4), 725–737.

149 | P a g e

Erni, S., Wang, X., Swystun, T., Taylor, S. W., Parisien, M. A., Robinne, F. N., ... & Flannigan,
M. D. (2023). Mapping wildfire hazard, vulnerability, and risk to Canadian communities.
International Journal of Disaster Risk Reduction,101, 104221.
Ex, S. A., Smith, F. W., Keyser, T. L., & Rebain, S. A. (2016). Estimating canopy bulk density
and canopy base height for interior western US conifer stands. Forest Science, 62(6),
690–697.
Ex, S., Smith, F. W., & Keyser, T. L. (2015). Characterizing crown fuel distribution for conifers
in the interior western United States. Canadian Journal of Forest Research, 45(7), 950–
957.
Feller, M. C. (2003). Coarse woody debris in the old-growth forests of British Columbia.
Environmental Reviews, 11(S1), S135–S157.
Finney, M. A. (1998). FARSITE, Fire Area Simulator–Model Development and Evaluation (No.
4). US Department of Agriculture, Forest Service, Rocky Mountain Research Station.
Fichtner, A., Härdtle, W., Bruelheide, H., Kunz, M., Li, Y., & Von Oheimb, G. (2018).
Neighbourhood interactions drive overyielding in mixed-species tree communities.
Nature Communications, 9(1), 1144.
Flannigan, M. D., Wotton, B. M., Marshall, G. A., De Groot, W. J., Johnston, J., Jurko, N., &
Cantin, A. S. (2016). Fuel moisture sensitivity to temperature and precipitation: Climate
change implications. Climatic Change, 134, 59–71.
Flannigan, M., Cantin, A. S., De Groot, W. J., Wotton, M., Newbery, A., & Gowman, L. M.
(2013). Global wildland fire season severity in the 21st Century. Forest Ecology and
Management, 294, 54–61.
Forest Analysis and Inventory Branch (FAIB), Ministry of Forests, government of British
Columbia. (2018). Change Monitoring Inventory – British Columbia. Ground Sampling
Procedures, Version 2.5. Forest Analysis & Inventory Branch, Ministry of Forests,
Government of British Columbia; Resources Information Standards Committee, Province
of British Columbia. Retrieved from
https://www2.gov.bc.ca/assets/gov/environment/natural-resource-stewardship/nr-lawspolicy/risc/cmi_ground_sampling_procedures_2018.pdf.
Forestry Canada Fire Danger Group. (1992). Development and Structure of the Canadian Forest
Fire Behavior Prediction System. Forestry Canada, Headquarters, Fire Danger Group and
Science and Sustainable Development Directorate, (Ottawa), 64. Retrieved from
https://www.frames.gov/documents/catalog/forestry_canada_fire_danger_group_1992.pd
f
Garcıá , M., Riaño, D., Chuvieco, E., Salas, J., & Danson, F. M. (2011). Multispectral and
LiDAR data fusion for fuel type mapping using support vector machine and decision
rules. Remote Sensing of Environment, 115(6), 1369–1379.

150 | P a g e

García-Cimarras, A., Manzanera, J. A., & Valbuena, R. (2023). LiDAR scan density and spatial
resolution effects on vegetation fuel type mapping. Croatian Journal of Forest
Engineering: Journal for Theory and Application of Forestry Engineering, 44(1), 189201.
Gillespie, A. R., Allen, H. L., & Vose, J. M. (1994). Amount and vertical distribution of foliage
of young Loblolly pine trees as affected by canopy position and silvicultural treatment.
Canadian Journal of Forest Research, 24(7), 1337–1344.
Gonçalves, A. C., et al. (2017). Multi-species stand classification: Definition and perspectives.
Forest Ecology and Conservation, 3–23.
González-Ferreiro, E., Arellano-Pérez, S., Castedo-Dorado, F., Hevia, A., Vega, J. A., VegaNieva, D., ... & Ruiz-González, A. D. (2017). Modelling the vertical distribution of
canopy fuel load using national forest inventory and low-density airborne laser scanning
data. PLoS One, 12(4), e0176114.
Goodwin, M. J., Zald, H. S. J., North, M. P., & Hurteau, M. D. (2021). Climate-driven tree
mortality and fuel aridity increase wildfire’s potential heat flux. Geophysical Research
Letters 48 (24): e2021GL094954.
Graham, R. T., McCaffrey, S., & Jain, T. B. (2004). Science Basis for Changing Forest Structure
to Modify Wildfire Behavior and Severity. U.S. Department of Agriculture, Forest
Service, Rocky Mountain Research Station.
Griffin, J. M., Simard, M., & Turner, M. G. (2013). Salvage harvest effects on advance tree
regeneration, soil nitrogen, and fuels following Mountain Pine Beetle outbreak in
lodgepole pine. Forest Ecology and Management, 291, 228–239.
Government of Alberta. (2012). How Different Tree Species Impact the Spread of Wildfire.
Environment; Sustainable Resource Development. Retrieved from
https://open.alberta.ca/dataset/48d6f355-413b-44bb-9aee07859ca26ed6/resource/c859881b-dab2-44f7-b283-a62a7f212baf/download/2012bragg-creek-how-different-tree-species-impact-pread-wildfire-2012-08-03.pdf.
Government of British Columbia, Ministry of Forest, Lands, and Natural Resource Operations,
Wildfire Management Branch (FLNRO). (2012). A Guide to Fuel Hazard Assessment
and Abatement in British Columbia. Retrieved from
https://www2.gov.bc.ca/assets/gov/public-safety-and-emergency-services/wildfirestatus/prevention/fire-fuel-management/hazard-assessmentabatement/bcws_hazard_assessment_abatement_guide.pdf
Hall, R. J., Skakun, R. S., et al. (2007). Mapping forest inventory attributes across coniferous,
deciduous and mixed wood stand types in the Northwest Territories from high spatial
resolution QuickBird satellite imagery. Proceedings of Our Common Borders–Safety,
Security and the Environment Through Remote Sensing, American Society for
Photogrammetry and Remote Sensing (ASPRS), 28.

151 | P a g e

Hanes, C. C., Wang, X., & de Groot, W. J. (2021). Dead and down woody debris fuel loads in
Canadian Forests. International Journal of Wildland Fire, 30(11), 871–885.
Harper, G., & Roach, J. (2014). The role of broadleaf trees: Impacts of managing boreal and subboreal mixedwood forests in British Columbia. Extension Note-British Columbia
Ministry of Forests and Range, (no. 110).
Haughian, S. R., Burton, P. J., Taylor, S. W., & Curry, C. (2012). Expected effects of climate
change on forest disturbance regimes in British Columbia. Journal of Ecosystems and
Management, 13(1), 1-24.
Hawkes, B., Taylor, S., Stockdale, C., Shore, T., Beukema, S., & Robinson, D. (2005).
Predicting mountain pine beetle impacts on lodgepole pine stands and woody debris
characteristics in a mixed severity fire regime using PrognosisBC and the Fire and Fuels
Extension. In L. Lagene, J. Zelnik, S. Cadwallader, & B. Huges (Eds.), Mixed Severity
Fire Regimes: Ecology and Management (pp. 123–135). Washington State University
Cooperative Extension Service/The Association for Fire Ecology, AFE MISC03.
Washington State University, Pullman, Washington.
Hély, C., Bergeron, Y., & Flannigan, M. D. (2000). Effects of stand composition on fire hazard
in mixed-wood Canadian boreal forest. Journal of Vegetation Science, 11(6), 813–824.
Hély, C., Flannigan, M., Bergeron, Y., & McRae, D. (2001). Role of vegetation and weather on
fire behavior in the Canadian mixedwood boreal forest using two fire behavior prediction
systems. Canadian Journal of Forest Research, 31(3), 430–441.
Hevia, A., Álvarez-González, J. G., Ruiz-Fernández, E., Prendes, C., Ruiz-González, A. D.,
Majada, J., & González-Ferreiro, E. (2016). Modelling canopy fuel and forest stand
variables and characterizing the influence of thinning in the stand structure using airborne
LiDAR. Revista de Teledetección,. Special Issue, 41–55.
Hijmans, R. J. (2018). raster: Geographic data analysis and modeling. R package version, .2.8.
https://CRAN.R-project.org/package=raster.
Hood, P. R., Nelson, K. N., Rhoades, C. C., & Tinker, D. B. (2017). The effect of salvage
Logging on surface fuel loads and fuel moisture in beetle-infested lodgepole pine forests.
Forest Ecology and Management, 390, 80–88.
Hunter, M. E., Iniguez, J. M., & Lentile, L. B. (2011). Short-and long-term effects on fuels,
forest structure, and wildfire potential from prescribed fire and resource benefit fire in
southwestern forests, USA. Fire Ecology, 7(3), 108–121.
Hyde, J. C., Smith, A. M. S., Ottmar, R. D., Alvarado, E. C., & Morgan, P. (2011). The
combustion of sound and rotten coarse woody debris: A review. International Journal of
Wildland Fire, 20(2), 163–174.
Inan, M., Bilici, E., & Akay, A. E. (2017). Using airborne LiDAR data for assessment of forest
fire fuel load potential. In Proceedings of the 4th International Geoadvances Workshop -

152 | P a g e

Geoadvances 2017: ISPRS Workshop on Multi-Dimensional & Multi-Scale Spatial Data
Modeling (pp. 255-258). Copernicus Gesellschaft Mbh.
Jakubowksi, M. K., Guo, Q., Collins, B., Stephens, S., Kelly, M., et al. (2013). Predicting surface
fuel models and fuel metrics using LiDAR and CIR imagery in a dense, mountainous
forest. Photogrammetric Engineering and Remote Sensing, 79(1), 37–49.
Jenkins, J. C., Chojnacky, D. C., Heath, L. S., & Birdsey, R. A. (2004). Comprehensive
Database of Diameter-Based Biomass Regressions for North American Tree Species.
USDA Forest Service, Northeastern Research Station, Gen. Tech. Rep. NE-GTR-319.
Johnson, A. F., Woodard, P. M., & Titus, S. J. (1990). Lodgepole pine and white spruce crown
fuel weights predicted from diameter at breast height. The Forestry Chronicle, 66(6),
596–599.
Jolly, W. M., Cochrane, M. A., Freeborn, P. H., Holden, Z. A., Brown, T. J., Williamson, G. J.,
& Bowman, D. M. J. S. (2015). Climate-induced variations in global wildfire danger
from 1979 to 2013. Nature Communications, 6(1), 7537.
Kabrick, J. M., Clark, K. L., D’Amato, A. W., Dey, D. C., Kenefic, L. S., et al. (2017).
Managing hardwood-softwood mixtures for future forests in eastern North America:
assessing suitability to projected climate change. Journal of Forestry, 115(3), 190–201.
Kassambara, A. (2018). Machine learning essentials: Practical guide in R. Sthda. CreateSpace
independent publishing platform.
Kassambara, A. (2023). Rstatix: Pipe-friendly framework for basic statistical tests. R Package
Version 0.7.2. Retrieved from https://CRAN.R-project.org/package=rstatix.
Keane, R. E. (2015). Wildland fuel fundamentals and applications (No. 11904). Cham,
Switzerland: Springer International Publishing.Keane, R. E. (2019). Canopy Fuel. In
Encyclopedia of Wildfires and Wildland-Urban Interface (WUI) Fires, edited by Samuel
L. Manzello, pp. 1–12. Cham: Springer International Publishing.
Keane, R. E., Burgan, R., & van Wagtendonk, J. (2001). Mapping wildland fuels for fire
management across multiple scales: Integrating remote sensing, GIS, and biophysical
modeling. International Journal of Wildland Fire, 10(4), 301–319.
Keeley, J. E. (2009). Fire Intensity, Fire Severity and Burn Severity: A brief review and
suggested usage. International Journal of Wildland Fire, 18(1), 116–126.
Kelley, J., Trofymow, J. A., & Bone, C. (2022). Combining area-based and individual tree
metrics for improving merchantable and non-merchantable wood volume estimates in
Coastal Douglas-Fir Forests. Remote Sensing, 14(9), 2204.
Keränen, J., Peuhkurinen, J., Packalen, P., & Maltamo, M. (2015). Effect of minimum diameter
at breast height and standing dead wood field measurements on the accuracy of ALSbased forest inventory. Canadian Journal of Forest Research, 45(10), 1280-1288.

153 | P a g e

Kennedy, M. C., Prichard, S. J., McKenzie, D., & French, N. H. F. (2020). Quantifying how
sources of uncertainty in combustible biomass propagate to prediction of wildland fire
emissions. International Journal of Wildland Fire, 29(9), 793–806.
Keyes, C. R. (2006). Role of foliar moisture content in the silvicultural management of forest
fuels. Western Journal of Applied Forestry, 21(4), 228-231.
Keyes, C. R., & O’Hara, K. L. (2002). Quantifying stand targets for silvicultural prevention of
crown fires. Western Journal of Applied Forestry, 17(2), 101–9.
Keyes, C. R., & Varner, J. M. (2006). Pitfalls in the silvicultural treatment of canopy fuels. Fire
Management Today, 66(3).
Keyser, T., & Smith, F. W. (2010). Influence of crown biomass estimators and distribution on
canopy fuel characteristics in ponderosa pine stands of the Black Hills. Forest Science,
56(2), 156–165.
Kohyama, T., Hara, T., & Tadaki, Y. (1990). Patterns of trunk diameter, tree height and crown
depth in crowded Abies stands. Annals of Botany, 65(5), 567–574.
Krankina, O. N., & Harmon, M. E. (1995). Dynamics of the dead wood carbon pool in
northwestern Russian boreal forests. Water, Air, and Soil Pollution, 82, 227–238.
Kuhn, M., & Wickham, H. (2020). Tidymodels: A collection of packages for modeling and
machine learning using tidyverse principles. https://www.tidymodels.org.
Labenski, P., Ewald, M., Schmidtlein, S., & Fassnacht, F. E. (2022). Classifying surface fuel
types based on forest stand photographs and satellite time series using deep learning.
International Journal of Applied Earth Observation and Geoinformation, 109, 102799.
Labenski, P., Ewald, M., Schmidtlein, S., Heinsch, F. A., & Fassnacht, F. E. (2023). Quantifying
surface fuels for fire modelling in temperate forests using airborne lidar and Sentinel-2:
Potential and limitations. Remote Sensing of Environment, 295, 113711.
Lambert, M. C., Ung, C. H., & Raulier, F. (2005). Canadian national tree aboveground biomass
equations. Canadian Journal of Forest Research, 35(8), 1996–2018.
Lapp, S. L., St. Jacques, J. M., Barrow, E. M., & Sauchyn, D. J. (2012). GCM projections for the
Pacific Decadal Oscillation under greenhouse forcing for the early 21st Century.
International Journal of Climatology, 32(9), 1423–1442.
LaRue, E. A., Wagner, F. W., Fei, S., Atkins, J. W., Fahey, R. T., Gough, C. M., & Hardiman, B.
S. (2020). Compatibility of aerial and terrestrial LiDAR for quantifying forest structural
diversity. Remote Sensing, 12(9), 1407.
LaRue, E., and O’Leary, D. (2021). Calculating forest structural diversity metrics from NEON
LiDAR data. NSF NEON. Published May 2021. Retrieved from
http://www.neonscience.org/resources/learning-hub/tutorials/structural-diversity-discretereturn
154 | P a g e

Lee, S., and Lee, D. K. (2018). What is the proper way to apply the multiple comparison test?
Korean Journal of Anesthesiology, 71(5), 353–360.
Liaw, A., and Wiener, M. (2002). Classification and Regression by randomForest. R News 2(3),
18-22.
Linley, G. D., Jolly, C. J., Doherty, T. S., Geary, W. L., Armenteras, D., Belcher, C. M., ... &
Nimmo, D. G. (2022). What do you mean,‘megafire’? Global Ecology and
Biogeography, 31(10), 1906-1922.
Liu, H. (2015). Comparing Welch ANOVA, a Kruskal-Wallis Test, and Traditional ANOVA in
Case of Heterogeneity of Variance. (Master's thesis, Virginia Commonwealth
University).
LM Forest Resource Solutions Ltd. (2020). Forest Fuel Treatment Efficacy in BC: Two Case
Studies: Thinning and Broadcast Burning. BC Wildfire Service, Ministry of Forests,
Lands, Natural Resource Operations; Rural Development.
Loomis, R. M., & Roussopoulos, P. J. (1978). Estimating aspen crown fuels in northeastern
Minnesota. USDA Forest Service, North Central Forest Experiment Station, Research
Paper NC-156. Saint Paul, MN.
Lopes Queiroz, G., McDermid, G. J., Linke, J., Hopkinson, C., & Kariyeva, J. (2020). Estimating
coarse woody debris volume using image analysis and multispectral LiDAR. Forests,
11(2), 141.
Luo, L., Zhai, Q., Su, Y., Ma, Q., Kelly, M., & Guo, Q. (2018). Simple method for direct crown
base height estimation of individual conifer trees using airborne LiDAR data. Optics
Express, 26(10), A562–A578.
MacDonald, G. B. (1996). Mixedwood management research and practice in Ontario. In P. G.
Comeau & K. D. Thomas (Eds.), Silviculture of Temperate and Boreal Broadleaf–
Conifer Mixtures. Province of BC, Ministry of Forests Research Program, Victoria, B.C.,
102-113.
Marino, E., Ranz, P., Tomé, J. L., Noriega, M. Á., Esteban, J., & Madrigal, J. (2016). Generation
of high-resolution fuel model maps from discrete airborne laser scanner and Landsat-8
OLI: A low-cost and highly updated methodology for large areas. Remote Sensing of
Environment, 187, 267–280.
Massman, W. J., Forthofer, J. M., & Finney, M. A. (2017). An improved canopy wind model for
predicting wind adjustment factors and wildland fire behavior. Canadian Journal of
Forest Research, 47(5), 594–603.
Mc Alpine, R. S., & Hobbs, M. W. (1994). Predicting the height to live crown base in plantations
of four boreal forest species. International Journal of Wildland Fire, 4(2), 103–106.
Means, J. E., Acker, S. A., Harding, D. J., Blair, J. B., Lefsky, M. A., Cohen, W. B., Harmon, M.
E., & McKee, W. A. (1999). Use of large-footprint scanning airborne lidar to estimate
155 | P a g e

forest stand characteristics in the western Cascades of Oregon. Remote Sensing of
Environment, 67(3), 298–308.
Merrill, D. F., & Alexander, M. E. (1987). Glossary of Forest Fire Management Terms.
Canadian Committee on Forest Fire Management, National Research Council of Canada:
Ottawa, ON, Canada.
Miles, P. D. (2009). Specific Gravity and Other Properties of Wood and Bark for 156 Tree
Species Found in North America (Vol. 38). US Department of Agriculture, Forest
Service, Northern Research Station.
Ministry of Water & Land, and Resource Stewardship. (2023). Free, Accessible Data Will Help
Communities Manage, Adapt, Plan. News Release: Retrieved from
https://news.gov.bc.ca/releases/2023WLRS0010-000512.
Moriarty, K., Cheng, A. S., Hoffman, C. M., Cottrell, S. P., & Alexander, M. E. (2019).
Firefighter observations of ‘surprising’ fire behavior in mountain pine beetle-attacked
lodgepole pine forests. Fire, 2(2), 34.
Mutlu, M., Popescu, S. C., Stripling, C., & Spencer, T. (2008). Mapping surface fuel models
using lidar and multispectral data fusion for fire behavior. Remote Sensing of
Environment, 112(1), 274–285.
Næsset, E. (2002). Predicting forest stand characteristics with airborne scanning laser using a
practical two-stage procedure and field data. Remote Sensing of Environment, 80(1), 88–
99.
Nigh, G. D. (2016). Total and merchantable volume equations for common tree species in British
Columbia by region and biogeoclimatic zone. Technical Report 106, B.C. Ministry of
Forests, Lands and Natural Resource Operations, Victoria, British Columbia.
Parisien, M. A., Barber, Q. E., Flannigan, M. D., & Jain, P. (2023). Broadleaf tree phenology and
springtime wildfire occurrence in boreal Canada. Global Change Biology, 29(16), 6106–
6119.
Pausas, J. G., & Bond, W. J. (2020). Alternative biome states in terrestrial ecosystems. Trends in
Plant Science, 25(3), 250–263.
Pebesma E, Bivand R (2005). “Classes and methods for spatial data in R.” R News, 5(2), 9–13.
https://CRAN.R-project.org/doc/Rnews/.
Penman, T. D., McColl-Gausden, S. C., Cirulis, B. A., Kultaev, D., Ababei, D. A., & Bennett, L.
T. (2022). Improved accuracy of wildfire simulations using fuel hazard estimates based
on environmental data. Journal of Environmental Management, 301, 113789.
Perrakis, D. D. B., Eade, G., & Hicks, D. (2018). British Columbia Wildfire Fuel Typing and
Fuel Type Layer Description. Information Report BC-X-444. Canadian Forest Service,
Natural Resources Canada, Victoria, B.C.

156 | P a g e

Phelps, N., & Beverly, J. L. (2022). Classification of forest fuels in selected fire-prone
ecosystems of Alberta, Canada—implications for crown fire behaviour prediction and
fuel management. Annals of Forest Science, 79(1), 40.
Plevris, V., Solorzano, G., Bakas, N. P., & Ben Seghier, M. E. A. (2022, November).
Investigation of performance metrics in regression analysis and machine learning-based
prediction models. In 8th European Congress on Computational Methods in Applied
Sciences and Engineering (ECCOMAS Congress 2022). European Community on
Computational Methods in Applied Sciences.
Pohlert, T. (2023). PMCMRplus: Calculate pairwise multiple comparisons of mean rank sums
extended. R package version 1.9.10. Retrieved from https://CRAN.Rproject.org/package=PMCMRplus.
Popescu, S. C., & Zhao, K. (2008). A voxel-based lidar method for estimating crown base height
for deciduous and pine trees. Remote Sensing of Environment, 112(3), 767–781.
Popović, Z., Bojović, S., Marković, M., & Cerdà, A. (2021). Tree species flammability based on
plant traits: A synthesis. Science of the Total Environment, 800, 149625.
Powell, D. C. (2010). Estimating crown fire susceptibility for project planning. Fire
Management Today, 70(3), 8.
Pretzsch, H. (2009). Forest dynamics, growth, and yield (Vol. 684). Berlin: Springer.
Price, O. F., & Gordon, C. E. (2016). The potential for LiDAR technology to map fire fuel
hazard over large areas of Australian forest. Journal of Environmental Management, 181,
663–73.
Quintilio, D., Alexander, M. E., & Ponto, R. L. (1989). Spring fires in semimature trembling
aspen and stand in central Alberta. Information Report NOR-X-323. Edmonton, AB:
Forestry Canada. 39 p.
R Core Team. (2021). R: A language and environment for statistical computing (version 4.3.0).
Vienna, Austria: R Foundation for Statistical Computing. https://www.R-project.org/
Ramirez, F. J. R., Navarro-Cerrillo, R. M., Varo-Martı́nez, M. Á., Quero, J. L., Doerr, S., &
Hernández-Clemente, R. (2018). Determination of forest fuels characteristics in
mortality-affected Pinus forests using integrated hyperspectral and ALS data.
International Journal of Applied Earth Observation and Geoinformation, 68, 157–167.
Reinhardt, E., Scott, J., Gray, K., & Keane, R. (2006). Estimating canopy fuel characteristics in
five conifer stands in the western United States using tree and stand measurements.
Canadian Journal of Forest Research, 36(11), 2803–2814.
Reinhardt, E., Lutes, D., & Scott, J. (2006). FuelCalc: A method for estimating fuel
characteristics. In: Andrews, Patricia L.; Butler, Bret W., comps. 2006. Fuels
Management-How to Measure Success: Conference Proceedings. 28-30 March 2006;
Portland, OR. Proceedings RMRS-P-41. Fort Collins, CO: US Department of
Agriculture, Forest Service, Rocky Mountain Research Station. p. 273-282 (Vol. 41).
157 | P a g e

Reinhardt, E., Crookston, N. L., Beukema, S. J., Kurz, W. A., Greenough, J. A., Robinson, D. C.
E., & Lutes, D. C. (2003). The Fire and Fuels Extension to the Forest Vegetation
Simulator. USDA Forest Service, Rocky Mountain Research Station, General Technical
Report RMRS-GTR-116, 209.
Riaño, D., Meier, E., Allgöwer, B., Chuvieco, E., & Ustin, S. L. (2003). Modeling airborne laser
scanning data for the spatial generation of critical forest parameters in fire behavior
modeling. Remote Sensing of Environment, 86(2), 177–186.
Rigatti, S. J. (2017). Random Forest. Journal of Insurance Medicine, 47(1), 31–39.
Resources Information Standards Committee (RISC). (2008). Vegetation resources inventory
ground sampling procedures. B.C. Min. For. Range, For. Analysis and Inventory Br. for
the Terrestrial Ecosystems Task Force, Resources Information Standards Committee,
Victoria, B.C.
Romps, D. M., Seeley, J. T., Vollaro, D., & Molinari, J. (2014). Projected increase in lightning
strikes in the United States due to global warming. Science, 346(6211), 851–854.
Roussel, J.-R., & Auty, D. (2022). Airborne LiDAR data manipulation and visualization for
Forestry Applications (R Package Version 3.1). Retrieved November 5, 2023, from
https://cran.r-project.org/package=lidR
Roussel, J.-R., Auty, D., Coops, N. C., Tompalski, P., Goodbody, T. R. H., Sánchez Meador, A.,
... Achim, A. (2020). lidR: An R package for analysis of airborne laser scanning (ALS)
data. Remote Sensing of Environment, 251, 112061.
Ruiz-González, A. D., & Álvarez-González, J. G. (2011). Canopy bulk density and canopy base
height equations for assessing crown fire hazard in Pinus radiata plantations. Canadian
Journal of Forest Research, 41(4), 839–850.
Russell, E. S., Liu, H., Gao, Z., Lamb, B., & Wagenbrenner, N. (2016). Turbulence dependence
on winds and stability in a weak-wind canopy sublayer over complex terrain. Journal of
Geophysical Research: Atmospheres, 121(19), 11–502.
Sandberg, D. V., Ottmar, R. D., & Cushon, G. H. (2001). Characterizing fuels in the 21st
Century. International Journal of Wildland Fire, 10(4), 381–387.
Sando, R. W., & Wick, C. H. (1972). A method of evaluating crown fuels in forest stands (Tech.
Rep. No. NC-84). United States Department of Agriculture Forest Service, North Central
Forest Experiment Station, St. Paul, MN, USA.
Schoennagel, T., Veblen, T. T., Negron, J. F., & Smith, J. M. (2012). Effects of mountain pine
beetle on fuels and expected fire behavior in lodgepole pine forests, Colorado, USA. PloS
One, 7(1), e30002.
Schonlau, M., & Zou, R. Y. (2020). The random forest algorithm for statistical learning. The
Stata Journal, 20(1), 3-29.

158 | P a g e

Scott, J. H. (2012). Introduction to wildfire behavior modeling. National Interagency Fuel, Fire,
& Vegetation Technology Transfer. Retrieved from http://pyrologix.com/wpcontent/uploads/2014/04/Scott_20121.pdf
Scott, J. H., & Reinhardt, E. D. (2001). Assessing crown fire potential by linking models of
surface and crown fire behavior. U.S. Department of Agriculture, Forest Service, Rocky
Mountain Research Station.
Scott, J. H., & Reinhardt, E. D. (2002). Estimating canopy fuels in conifer forests. Fire
Management Today, 62(4), 45–50.
Sherrill, K. R., Lefsky, M. A., Bradford, J. B., & Ryan, M. G. (2008). Forest structure estimation
and pattern exploration from discrete-return LiDAR in subalpine forests of the Central
Rockies. Canadian Journal of Forest Research, 38(8), 2081–2096.
Shokirov, S., Schaefer, M., Levick, S. R., Jucker, T., Borevitz, J., Abdurahmanov, I., &
Youngentob, K. (2021). Multi-platform LiDAR approach for detecting coarse woody
debris in a landscape with varied ground cover. International Journal of Remote Sensing,
42(24), 9324–9350.
Song, J. H., Han, S. H., Yu, K. Y., & Kim, Y. I. (2002). Assessing the possibility of land-cover
classification using lidar intensity data. International archives of photogrammetry remote
sensing and spatial information sciences, 34(3/B), 259-262.
Spears, J. D. H., Holub, S. M., Harmon, M. E., & Lajtha, K. (2003). The Influence of
decomposing logs on soil biology and nutrient cycling in an old-growth mixed coniferous
forest in Oregon, USA. Canadian Journal of Forest Research, 33(11), 2193–2201.
Spriggs, R. A., Coomes, D. A., Jones, T. A., Caspersen, J. P., & Vanderwel, M. C. (2017). An
alternative approach to using LiDAR remote sensing data to predict stem diameter
distributions across a temperate forest landscape. Remote Sensing, 9(9), 944.
Standish, J. T., Manning, G. H., Demaerschalk, J. P., et al. (1985). Development of Biomass
Equations for British Columbia Tree Species. Information Report BC-X-264. Canadian
Forest Service, Victoria, B.C.
Statistics Canada. (2018). Human activity and the environment: Forests in Canada. In Human
Activity and the Environment. https://www150.statcan.gc.ca/n1/en/pub/16-201-x/16-201x2018001-eng.pdf?st=xR-4zIQA
Stefanidou, A., Gitas, I. Z., Korhonen, L., Stavrakoudis, D., & Georgopoulos, N. (2020).
LiDAR-based estimates of canopy base height for a dense uneven-aged structured forest.
Remote Sensing, 12(10), 1565.
Stephens, S. L., Burrows, N., Buyantuyev, A., Gray, R. W., Keane, R. E., Kubian, R., Liu, S., et
al. (2014). Temperate and boreal forest mega-fires: Characteristics and challenges.
Frontiers in Ecology and the Environment, 12(2), 115–22.

159 | P a g e

Stocks, B. J., Alexander, M. E., Wotton, B. M., Stefner, C. N., Flannigan, M. D., Taylor, S. W.,
Lavoie, N., et al. (2004). Crown fire behaviour in a northern jack pine black spruce
forest. Canadian Journal of Forest Research, 34(8), 1548–1560.
Talucci, A. C., Meigs, G. W., Knudby, A., & Krawchuk, M. A. (2022). Fire severity and the
legacy of mountain pine beetle outbreak: high-severity fire peaks with mixed live and
dead vegetation. Environmental Research Letters, 17(12), 124010.
Taylor, S. W., Pike, R. G., & Alexander, M. E. (1996). Field guide to the Canadian Forest Fire
Behaviour Prediction (FBP) System.
Tepley, A. J., Parisien, M. A., Wang, X., Oliver, J. A., & Flannigan, M. D. (2022). Wildfire
evacuation patterns and syndromes across Canada's forested regions. Ecosphere, 13(10),
e4255.
Tymstra, C., Stocks, B. J., Cai, X., & Flannigan, M. D. (2020). Wildfire management in Canada:
Review, challenges and opportunities. Progress in Disaster Science, 5, 100045.
Ung, C.-H., Bernier, P., & Guo, X.-J. (2008). Canadian national biomass equations: New
parameter estimates that include British Columbia data. Canadian Journal of Forest
Research, 38(5), 1123–1132.
Van Ewijk, K. Y., Treitz, P. M., & Scott, N. A. (2011). Characterizing forest succession in
Central Ontario using LiDAR-derived indices. Photogrammetric Engineering & Remote
Sensing, 77(3), 261–269.
Van Wagner, C. E. (1977). Conditions for the start and spread of crown fire. Canadian Journal
of Forest Research, 7(1), 23–34.
Varner, J. M., Kane, J. M., Kreye, J. K., & Engber, E. (2015). The flammability of forest and
woodland litter: A synthesis. Current Forestry Reports, 1, 91–99.
Véga, C., Renaud, J. P., Durrieu, S., & Bouvier, M. (2016). On the interest of penetration depth,
canopy area and volume metrics to improve Lidar-based models of forest parameters.
Remote Sensing of Environment, 175, 32-42.
Waddell, K. L. (2002). Sampling coarse woody debris for multiple attributes in extensive
resource inventories. Ecological Indicators, 1(3), 139–153.
Wagner, W., Eberhöfer, C., Hollaus, M., & Summer, G. (2004). Robust filtering of airborne laser
scanner data for vegetation analysis. International Archives of Photogrammetry, Remote
Sensing and Spatial Information Sciences, XXXVI(8/W2), 56–61.
Walker, X. J., Rogers, B. M., Veraverbeke, S., Johnstone, J. F., Baltzer, J. L., Barrett, K.,
Bourgeau-Chavez, L., et al. (2020). Fuel availability, not fire weather controls boreal
wildfire severity and carbon emissions. Nature Climate Change, 10(12), 1130–1136.
Wang, X., Thompson, D. K., Marshall, G. A., Tymstra, C., Carr, R., & Flannigan, M. D. (2015).
Increasing frequency of extreme fire weather in Canada with climate change. Climatic
Change, 130, 573–586.
160 | P a g e

Wei, Y. (2012). Optimize landscape fuel treatment locations to create control opportunities for
future fires. Canadian Journal of Forest Research, 42(6), 1002–1014.
Welch, B. L. (1947). The generalization of ‘STUDENT’s’ problem when several different
population variances are involved. Biometrika, 34(1-2), 28–35.
Werner, A. T. (2011). BCSD downscaled transient climate projections for eight select GCMs
over British Columbia, Canada. Hydrologic Modelling Project Final Rep.(Part I),
Pacific Climate Impacts Consortium, University of Victoria, Victoria, BC, Canada 63.
White, J. C., Wulder, M. A., Varhola, A., Vastaranta, M., Coops, N. C., Cook, B. D., Pitt, D., &
Woods, M. (2013). A best practices guide for generating forest inventory attributes from
airborne laser scanning data using an area-based approach. The Forestry Chronicle,
89(6), 722–723.
Williams, J. (2013). Exploring the onset of high-impact mega-fires through a forest land
management prism. Forest Ecology and Management, 294, 4–10.
Woods, M., Lim, K., & Treitz, P. (2008). Predicting forest stand variables from LiDAR data in
the Great Lakes St. Lawrence Forest of Ontario. The Forestry Chronicle, 84(6), 827-839.
Woodall, C. W., & Leutscher, B. (2003). Extending and intensifying the FIA inventory of down
forest fuels: boundary waters canoe area and pictured rocks National Lakeshore. In
Proceeding of the Fifth Annual Forest Inventory and Analysis Symposium, 18–20.
Wotton, B. M., Flannigan, M. D., & Marshall, G. A. (2017). Potential climate change impacts on
fire intensity and key wildfire suppression thresholds in Canada. Environmental Research
Letters, 12(9), 095003.
Wright, M. N., & Ziegler, A. (2017). ranger: A fast implementation of Random Forests for high
dimensional data in C++ and R. Journal of Statistical Software, 77(1), 1-17.
Xanthopoulos, G. (1990). Development of a Wildland Crown Fire Initiation Model (Doctoral
dissertation). University of Montana.
Xing, D., Bergeron, J. A. C., Solarik, K. A., Tomm, B., Macdonald, S. E., Spence, J. R., & He, F.
(2019). Challenges in estimating forest biomass: Use of allometric equations for three
boreal tree species. Canadian Journal of Forest Research, 49(12), 1613–1622.
Yates, L. A., Aandahl, Z., Richards, S. A., & Brook, B. W. (2023). Cross validation for model
selection: a review with examples from ecology. Ecological Monographs, 93(1), e1557.
Yu, X., Hyyppä, J., Holopainen, M., & Vastaranta, M. (2010). Comparison of area-based and
individual tree-based methods for predicting plot-level forest attributes. Remote Sensing,
2(6), 1481–1495.
Zeileis, A., & Hothorn, T. (2002). Diagnostic checking in regression relationships. R News, 2(3),
7–10. https://CRAN.R-project.org/doc/Rnews/.

161 | P a g e

Basal Area
stand type

CD10 Conifer Dead
CD123-A Conifer Live
CD151A Deciduous
CD337 Conifer Live
CD383 Deciduous
CD445A
Mixed
CD56 Conifer Dead
CD62A Conifer Dead
CD77
Conifer Live
CD93
Conifer Live
D101-A Deciduous
D11
Deciduous
D348
Mixed
M1
Conifer Live
M13 Conifer Dead
M15
Conifer Live
M23
Deciduous
M26-A
Mixed
M28
Mixed
M348A
Mixed
M355
Deciduous
M369
Mixed
M41 Conifer Dead
M433A
Mixed
M439A Conifer Live
M449
Deciduous
M462a
Mixed
M472 Conifer Live
M476
Deciduous
M486
Mixed
MN3-2 Conifer Live
MN9-517A
Mixed
MN9-518
Mixed
N10
Conifer Live
N101
Conifer Live
N102
Conifer Live
N103
Conifer Live
N104 Conifer Dead
N105 Conifer Dead
N107
Conifer Live
N108
Conifer Live
N109
Conifer Live
N110
Conifer Live
N111
Conifer Live
N112
Conifer Live
N113
Conifer Live
N115
Conifer Live
N118 Conifer Dead
N120
Conifer Live
N121
Conifer Live
N122
Mixed
N123
Deciduous
N124
Deciduous
N131
Conifer Live
N133
Conifer Live
N134
Mixed
N135
Conifer Live
N136
Conifer Live
N137 Conifer Dead
N139
Conifer Live
N15
Conifer Live
N16
Conifer Live
N24 Conifer Dead
N27 Conifer Dead
N28
Conifer Live
N29
Conifer Live
N30
Conifer Live
N35
Conifer Live
N36
Deciduous
N39
Conifer Live
N40
Conifer Live
N41
Conifer Live
N42
Mixed
N43
Deciduous
N44
Conifer Live
N45
Conifer Live
N46
Conifer Live
N47
Conifer Live
N48
Conifer Live
N50
Conifer Live
N51
Mixed
N52
Conifer Live
N53 Conifer Dead
N54
Conifer Live
N55
Conifer Live
N56
Conifer Live
N57
Conifer Live
N58
Deciduous
N61
Conifer Live
N62
Conifer Live
N63
Conifer Live
N65
Conifer Live
N66
Conifer Live
N67
Conifer Live
N68
Conifer Live
N70
Conifer Live
N72
Conifer Live
N73
Conifer Live
N75
Deciduous
N76
Conifer Live
N77
Conifer Live
N78
Conifer Live
N80
Conifer Live
N82
Conifer Live
N83
Conifer Live
N84
Conifer Live
N85
Conifer Live
N86
Conifer Live
N88
Conifer Live
N89
Conifer Live
N9
Conifer Live
N90
Conifer Live
N93 Conifer Dead
N94
Conifer Live
N95
Conifer Live
N96
Conifer Live
N98
Conifer Live
N99
Conifer Live

Plot ID

Year of Comfor
BEC
Basal
collectio Compart Zone/Sub
Area
n
ment
zone
(m2/ha)
2019
SBSdk
20.7
2019
Babine SBSmc
37.3
2019 Southside SBSdk
35.2
2019 Roselake SBSdk
18.5
2019 Roselake SBSdk
58.0
2019 Roselake SBSdk
71.5
2019 Roselake SBSdk
9.5
2019 Southside SBSmc
27.9
2019
Babine SBSmc
37.6
2019
Babine SBSmc
15.0
2019
Hannay SBSdk
61.7
2019 Roselake SBSdk
37.1
2019 Roselake SBSdk
41.8
2019 Southside SBSdk
40.9
2019 Roselake SBSdk
43.0
2019 Roselake SBSdk
29.4
2019
Hannay SBSmc
39.8
2019
Hannay SBSmc
49.6
2019 Roselake SBSdk
52.1
2019 Roselake SBSdk
71.4
2019 Roselake SBSdk
62.8
2019
Hannay SBSdk
47.0
2019 Roselake SBSdk
29.7
2019 Southside SBSdk
31.6
2019 Roselake SBSdk
34.3
2019
Babine
SBSdk
44.7
2019 Roselake SBSdk
45.0
2019 Roselake SBSdk
18.5
2019 Southside SBSdk
36.8
2019 Roselake SBSdk
35.4
2019
Hannay SBSdk
19.4
2019
Babine SBSmc
44.0
2019
Babine SBSmc
38.8
2018
Babine SBSmc
6.4
2018
SBSdk
24.5
2018 Southside SBSdk
1.1
2018
SBSmc
61.9
2018 Southside SBSdk
22.6
2018
SBSdk
5.9
2018
SBSdk
25.6
2018 Roselake SBSmc
19.8
2018
SBSdk
47.9
2018
Hannay SBSmc
48.4
2018 Southside SBSdk
29.8
2018 Southside SBSdk
36.4
2018
SBSdk
21.8
2018
Babine SBSmc
18.4
2018
Hannay SBSdk
23.4
2018
SBSdk
26.6
2018
SBSdk
31.5
2018
SBSdk
42.5
2018
SBSdk
8.8
2018 Southside SBSdk
25.3
2018 Southside SBSmc
54.3
2018 Roselake SBSdk
36.4
2018
Babine
SBSdk
29.8
2018 Southside SBSdk
21.1
2018 Roselake ESSFmc
87.1
2018
Babine
SBSdk
61.0
2018 Roselake SBSdk
11.9
2018
Babine SBSmc
11.8
2018
Babine SBSmc
0.3
2018
Hannay ESSFmc
10.2
2018
Hannay SBSmc
18.9
2018
Hannay SBSmc
44.1
2018
Hannay SBSdk
8.6
2018
Hannay SBSdk
23.2
2018
SBSdk
10.3
2018 Roselake SBSdk
15.2
2018 Roselake SBSdk
37.0
2018 Roselake SBSdk
32.3
2018 Roselake SBSdk
5.4
2018
SBSdk
31.4
2018
SBSdk
21.9
2018 Roselake SBSdk
40.2
2018 Roselake SBSdk
23.7
2018
SBSdk
50.7
2018 Roselake SBSdk
8.9
2018 Roselake SBSdk
33.8
2018
SBSdk
16.7
2018 Roselake SBSdk
9.9
2018 Southside SBSmc
42.1
2018 Southside SBSmc
37.0
2018 Roselake SBSdk
17.1
2018 Roselake SBSdk
22.8
2018
SBSdk
7.4
2018
SBSdk
9.8
2018
SBSdk
13.4
2018 Roselake SBSmc
15.1
2018 Roselake SBSmc
17.4
2018 Southside SBSdk
5.5
2018 Roselake ESSFmc
33.7
2018 Roselake ESSFmc
1.4
2018 Roselake ESSFmc
53.0
2018
SBSdk
0.4
2018 Roselake ESSFmc
84.2
2018 Roselake SBSmc
40.5
2018 Southside SBSdk
6.6
2018
SBSdk
20.0
2018 Southside SBSdk
4.0
2018 Southside SBSdk
50.4
2018 Southside SBSdk
13.4
2018
SBSdk
0.7
2018 Southside SBSdk
33.9
2018
SBSdk
6.6
2018 Southside SBSdk
23.9
2018
SBSdk
8.5
2018
SBSdk
0.2
2018 Southside SBSmc
29.8
2018 Southside SBSmc
24.0
2018
Babine
SBSdk
79.5
2018 Southside SBSmc
9.7
2018 Southside SBSmc
3.0
2018 Southside SBSmc
13.0
2018 Southside SBSdk
25.5
2018 Southside SBSdk
11.3
2018 Southside SBSmc
35.4
2018 Southside SBSmc
55.5

Live
Conifer
BA
7.7
21.8
0.4
11.9
13.6
11.4
4.5
5.1
17.9
10.4
3.9
5.2
19.1
25.8
9.0
20.5
6.5
16.5
34.7
21.4
4.0
20.8
7.5
9.9
11.8
10.5
18.9
15.5
0.0
13.8
12.4
18.7
8.7
5.7
23.8
1.1
47.3
11.5
2.8
25.1
19.8
40.1
35.5
28.5
29.5
21.8
18.3
14.3
19.8
22.3
32.8
0.0
0.4
51.4
32.5
10.2
20.1
72.9
21.0
11.4
10.6
0.3
7.1
6.5
42.3
8.0
13.1
8.0
0.2
22.1
26.8
5.4
17.3
2.4
34.4
17.3
43.6
4.6
32.5
15.0
4.2
29.2
11.1
16.8
22.7
7.3
6.9
0.8
15.0
17.4
1.9
30.8
1.3
34.3
0.0
75.8
23.9
5.6
0.0
0.3
50.4
13.4
0.7
30.3
6.2
23.4
8.1
0.2
29.4
24.0
59.4
9.0
2.2
13.0
21.3
10.4
35.3
32.3

Dead Deciduous Live Con % Dead Con
conifer
BA
%
BA
13.0
0.0
37%
63%
14.9
0.6
58%
40%
0.0
34.8
1%
0%
6.7
0.0
64%
36%
0.0
44.4
23%
0%
26.4
33.7
16%
37%
5.0
0.0
48%
52%
22.8
0.0
18%
82%
19.8
0.0
47%
53%
4.6
0.0
69%
31%
0.0
57.8
6%
0%
0.9
31.0
14%
2%
0.0
22.6
46%
0%
15.0
0.2
63%
37%
32.2
1.8
21%
75%
8.9
0.0
70%
30%
0.2
33.1
16%
1%
13.4
19.8
33%
27%
0.0
17.4
67%
0%
6.1
43.9
30%
9%
0.0
58.8
6%
0%
7.3
18.9
44%
16%
22.0
0.3
25%
74%
4.4
17.3
31%
14%
14.7
7.7
35%
43%
0.0
34.2
23%
0%
1.2
24.9
42%
3%
0.1
2.9
84%
0%
0.0
36.8
0%
0%
6.5
15.1
39%
18%
4.1
3.0
64%
21%
1.2
24.0
42%
3%
14.3
15.8
22%
37%
0.1
0.6
89%
1%
0.7
0.0
97%
3%
0.0
0.0
96%
0%
0.3
14.3
76%
0%
10.1
1.0
51%
45%
3.1
0.0
48%
52%
0.2
0.4
98%
1%
0.1
0.0
100%
0%
7.6
0.2
84%
16%
9.7
3.2
73%
20%
1.3
0.0
96%
4%
4.9
2.0
81%
14%
0.0
0.0
100%
0%
0.0
0.0
100%
0%
8.4
0.7
61%
36%
0.0
6.8
74%
0%
0.5
8.6
71%
2%
0.3
9.4
77%
1%
0.0
8.8
0%
0%
0.0
24.9
1%
0%
2.6
0.3
95%
5%
0.0
3.9
89%
0%
6.0
13.5
34%
20%
0.1
0.9
95%
0%
14.2
0.0
84%
16%
40.0
0.0
34%
66%
0.0
0.5
96%
0%
1.2
0.0
90%
10%
0.0
0.0
100%
0%
1.5
1.6
69%
15%
12.0
0.4
35%
63%
1.8
0.0
96%
4%
0.6
0.0
93%
7%
3.5
6.6
57%
15%
0.0
2.3
78%
0%
0.0
15.0
1%
0%
12.3
2.7
60%
33%
5.5
0.0
83%
17%
0.0
0.0
99%
1%
0.0
14.1
55%
0%
0.0
19.5
11%
0%
0.0
5.8
86%
0%
2.0
4.5
73%
8%
0.1
7.0
86%
0%
3.2
1.1
52%
35%
0.8
0.4
96%
2%
0.2
1.5
90%
1%
1.4
4.3
42%
14%
12.9
0.0
69%
31%
23.2
2.8
30%
63%
0.1
0.2
98%
1%
0.0
0.0
100%
0%
0.0
0.1
98%
0%
0.0
2.9
70%
0%
0.0
12.6
6%
0%
0.0
0.1
99%
0%
0.0
0.0
100%
0%
0.0
3.6
35%
0%
2.9
0.0
91%
9%
0.1
0.0
93%
7%
18.7
0.0
65%
35%
0.2
0.2
0%
47%
8.4
0.0
90%
10%
16.6
0.0
59%
41%
1.0
0.0
86%
14%
0.0
20.0
0%
0%
3.6
0.2
6%
89%
0.0
0.0
100%
0%
0.0
0.0
100%
0%
0.0
0.0
100%
0%
3.7
0.0
89%
11%
0.0
0.4
94%
0%
0.4
0.0
98%
2%
0.4
0.0
95%
5%
0.0
0.0
100%
0%
0.4
0.0
99%
1%
0.0
0.0
100%
0%
20.1
0.0
75%
25%
0.7
0.0
93%
7%
0.8
0.0
72%
28%
0.0
0.0
100%
0%
2.3
1.9
84%
9%
0.9
0.0
92%
8%
0.0
0.1
100%
0%
23.3
0.0
58%
42%
0%
2%
99%
0%
77%
47%
0%
0%
0%
0%
94%
84%
54%
0%
4%
0%
83%
40%
33%
62%
94%
40%
1%
55%
23%
77%
55%
16%
100%
43%
15%
55%
41%
10%
0%
4%
23%
5%
0%
1%
0%
0%
7%
0%
5%
0%
0%
3%
26%
27%
22%
100%
99%
0%
11%
45%
4%
0%
0%
4%
0%
0%
16%
2%
0%
0%
28%
22%
99%
7%
0%
0%
45%
89%
14%
19%
14%
13%
1%
9%
44%
0%
8%
1%
0%
2%
30%
94%
1%
0%
65%
0%
0%
0%
53%
0%
0%
0%
100%
4%
0%
0%
0%
0%
6%
0%
0%
0%
0%
0%
0%
0%
0%
0%
7%
0%
0%
0%

Dec %

VRI Label

19693726\SxAtPli\320-22/0\he,sl\$
93551870/L\Sx(Bl)\831-8/0\sl,he\$
72906527\At\523-13/0\hf,sl
32100443/L\PliSx\831-16/0\sl,he,by\$
{ 881\28660605\PliSxAt\111-19/19\$L07
23220248/L\SxPli(At)\732-13/0\sl,he\$
85934369/L\Sx\730-12/0\he,sl\$
89320062/L\SxAt\320-17/0\sl,he\$
90991534/L\Sx(Bl)\831-9/0\he,sl\$
94441526/L\Sx(Bl)\831-10/0\he\$
90419375\AtSx(Pli)\534-19/0\$
{513\88912230\At(PliSx)\223-21/19\sl,hf\$L79
18770459\At(AcSx)\836-17/0
4900416\AtPliSx\633-14/0\sl,he
76910304/L\AtSxPli\831-16/0\sl,he\$
76910304/L\AtSxPli\831-16/0\sl,he\$
23431688\AcSxPliAt\534-16/0\$
28741390/L\AtSxPli\532-15/0\sl,he\$
43487466\AtSx\433-20/0\hf\$
72306432\AtSx\535-18/0\hg
{ 569\46275981\PliSx\211-20/20\$L97
37508124\AtSx(Pli)\733-14/0\$
59827179\AcAt(Sx)\735-15/0\hf
12120419\Sx(At)\633-13/0\sl,hf
51246997\AtSx(Pli)\731-16/0\hf,rz,sl
3825025\AtSx(Bl)\734-16/0
51417177/L\PliSx\720-12/0\hf,sl\$
90822193\AtSx(Bl)\834-15/0\sl,he
11220319\AtPli\421-14/0\hg,gp,sl
90822193\AtSx(Bl)\834-15/0\sl,he
58218714/L\SxAtPli\533-15/0\$
{813\76583103\SxAt(BlPli)\834-8/0\sl
{813\76583103\SxAt(BlPli)\834-8/0\sl
{815\23741818\PliSx\216-18/18\$L94
{10\28837466\Pli(Sx)\225-25/25\sl\$L84
44385715\PliSxAt\111-24/24\$L02
65651241\PliSx\324-19/0\hf\$
82189483/L\PliSx\423-14/0\sl,hf\$
96717677/L\SxAt\523-14/0\hf\$
48279796\SxPli\323-19/0\hf,sl
{509\78757902\SxPli\325-20/18\he,sl\$L61
76622178/L\Sx(PliSb)\524-13/0\sl\$
33570576\Sx(SbPli)\533-14/0\$
49188007\SxSb(Pli)\424-17/0\sl,hf
{55\30935542\SxPli(AtBl)\333-26/0\sl,hf\$N60
58102317\SxPli\534-15/0\$
{812\93773014\PliSx\111-19/19\he\$L98
52069577/L\AtSx(Pli)\531-14/0\he,st\$B18
49787433\AtSx\222-19/19\hf,sl\$
80224753\AtPli(Ep)\224-20/20\sl,hg\$
91092230\AtSxPli\325-16/18\sl,hf\$
40288537\At\633-14/0\hf,sl
76597027\At(Sx)\733-18/0\sl,he
{42\67479316\Sx\525-12/0\sl,hf\$L32
98868814\SxPli(At)\326-24/0\he
93526767\Sx(PliBlAtEp)\633-11/0\sl
15022820/L\Sx\733-11/0\sl,he\$
85393199/L\BlSe\833-9/0\hf,sl\$
61826958/L\SxBl(Pli)\842-11/0\$
76910304/L\AtSxPli\831-16/0\sl,he\$
{943\73693844\PliSx\111-18/18\he,es\$L11
{ 940\88942410\Pli(Sx)\111-18/18\$L10
49091732/L\PliAtSe\522-13/0\st\$
15352063/L\SxPli\532-15/0\he,sl\$
28360048\Sx\424-15/0\hg
46098575\SxAtPli\532-15/0\he\$
12338631\Sx\634-14/0
77225457\SxAt\431-20/0\hf,sl
89163903
12759443/L\SxAt\732-12/0\hf,sl\$
27527603/L\SxPli(Sb)\732-12/0\sl,hf\$
{ 569\46275981\PliSx\211-20/20\$L97
59614049/L\AtAc\534-17/0\hg
73622184\At\320-18/0\he
93408018\SxAt(Pli)\734-15/0\sl,hg\$
9286174\Sx\833-10/0\he,sl
26084311\At(Sx)\221-19/19\hf
40352437\PliSx(At)\112-20/20\$L05
53920648\PliSx\227-17/17\$L88
8582694\SxAt\223-30/0\sl,hf\$
{ 881\28660605\PliSxAt\111-19/19\$L07
22137783\SxBl\736-13/0\$
33885964/L\PliSx\531-16/0\sl,hf\$
58034687\Sx(At)\743-16/0\st,hg\$
75422934\SxBl\833-13/0\st,hf
87737771\Sx\321-23/0\sl,he
3696092/L\SxPli\211-18/18\sl,hf\$
17874306\At\420-16/0\hg,sl
{505\69429789\Sx\214-19/19\br,hf,sl\$L84
{803\1456130/L\Pli(Sx)\212-18/18\sl,hf\$L94
54228169\At(Sx)\430-17/0\st,he
70656333\BlSe(Pli)\312-11/0\hf,sl
78434674/L\BlSe\834-10/0\$
99812791\BlSe\823-5/0\sl
36336619\Sx\320-14/0\st,he
67163036\SeBl\923-3/0\sl,hf
15257526/L\PliSx\532-18/0\sl,he\$
{61\2936848\PliSx(At)\214-18/18\sl\$L90
69897157\AtSx\430-16/0\hg
85745341\SxAc\842-18/0\sl
2143519\Sx\532-16/0\sl
{34\17681896\Pli(AtSx)\223-21/21\sl,he\$L94
{1005\53005823\PliSx\111-18/18\hf,sl\$L06
83002065/L\Sb(Sx)\527-9/0
97550296\SxSb\630-12/0\sl,hg
{23\12598421\PliSx\225-17/18\sl,hf\$B18
{ 949\26306618\$B18
{1085\19715867\PliSx(Bl)\111-18/18\$L05
{41\49892215\Pli\227-17/17\hf\$L86
{112\65740522\Pli\224-21/21\sl\$L91
92685595/L\SxPli(Bl)\833-8/0\$
80818732\sl,hf\$
16852618/L\Pli(AtSx)\520-10/0\hg,sl\$
{142\31340652\PliSx(Bl)\212-18/18\sl,hf\$L97
{182\48199007\Sx\524-11/0\he,st\$
61567245\Sx\832-8/0\sl
{558\83312862\Pli(SxBl)\321-16/0\st,hf\$L68
{515\310963/L\PliSx\432-16/0\hf,sl\$I44

Mixed
Conifer Dead
Deciduous
Conifer Dead
Conifer Live
Conifer Dead
Conifer Dead
Mixed
Conifer Dead
Conifer Dead
Mixed
Mixed
Deciduous
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Conifer Live
Mixed
Deciduous
Conifer Live
Mixed
Mixed
Conifer Dead
Mixed
Mixed
Mixed
Mixed
Mixed
Mixed
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Mixed
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Mixed
Mixed
Mixed
Mixed
Deciduous
Deciduous
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Dead
Mixed
Conifer Live
Conifer Live
Mixed
Conifer Dead
Conifer Live
Mixed
Conifer Live
Mixed
Deciduous
Conifer Dead
Conifer Live
Conifer Live
Deciduous
Deciduous
Mixed
Conifer Live
Deciduous
Conifer Live
Conifer Live
Mixed
Conifer Live
Conifer Live
Conifer Dead
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Deciduous
Conifer Live
Conifer Dead
Deciduous
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Mixed
Mixed
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Deciduous
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Deciduous
Conifer Dead
Conifer Live
Conifer Live
Conifer Live
Conifer Live
Conifer Dead

VRI Stand type

VRI Age
Field
CWD
(years) recorded age (m3/ha)
(years)
48
338
188
0
88
49
148
274
11
133
138
521
133
459
58
595
148
137
168
18
88
90
40
0
148
143
108
236
148
162
148
345
88
15
88
139
78
186
83
35
22
60
138
62
128
508
108
0
138
0
128
90
138
82
148
137
78
105
148
402
98
119
168
11
168
217
22
26
30
22
16
10
48
54
78
74
83
42
58
41
41
28
83
82
88
65
63
71
47
49
82
62
20
16
88
58
33
45
33
24
43
52
118
60
128
100
84
82
48
53
113
78
138
43
213
142
188
117
148
153
9
64
9
16
88
64
88
53
72
49
98
14
118
105
68
80
NA
70
138
61
138
115
22
48
83
62
48
105
123
123
187
100
33
105
15
54
31
36
38
27
11
125
128
56
85
83
138
141
148
262
48
30
33
26
68
32
36
22
25
19
73
38
53
96
148
71
227
192
58
10
295
240
90
70
26
155
78
42
158
18
88
145
26
18
14
9
83
55
108
18
34
26
NA
44
12
9
34
24
25
23
189
82
NA
108
98
49
22
16
84
71
188
181
44
34
79
67
110653
0
11838
70676
23221
115032
149274
181130
49818
7333
16327
0
24659
52904
39614
80886
3651
48864
31904
5902
10347
10581
156222
0
0
24411
27363
44784
22196
126384
43402
1856
63562
-

CWD
(kg/ha)

APPENDICES

Appendix 1: Plot attributes summary

Table 18. Plot summary table showing stand type, VRI stand type, and descriptive statistics for each plot.

162 | P a g e

BA stand type

CD10
Conifer Dead
CD123-A Conifer Live
CD151A Deciduous
CD337
Conifer Live
CD383
Deciduous
CD445A Mixed
CD56
Conifer Dead
CD62A
Conifer Dead
CD77
Conifer Live
CD93
Conifer Live
D101-A Deciduous
D11
Deciduous
D348
Mixed
M1
Conifer Live
M13
Conifer Dead
M15
Conifer Live
M23
Deciduous
M26-A
Mixed
M28
Mixed
M348A
Mixed
M355
NA
M369
Mixed
M41
Conifer Dead
M433A
Mixed
M439A
Conifer Live
M449
Deciduous
M462a
Mixed
M472
Conifer Live
M476
Deciduous
M486
Mixed
MN3-2
Conifer Live
MN9-517A Mixed
MN9-518 Mixed
N10
Conifer Live
N101
Conifer Live
N102
Conifer Live
N103
Conifer Live
N104
Conifer Dead
N105
Conifer Dead
N107
Conifer Live
N108
Conifer Live
N109
Conifer Live
N110
Conifer Live
N111
Conifer Live
N112
Conifer Live
N113
Conifer Live
N115
Conifer Live
N118
Conifer Dead
N120
Conifer Live
N121
Conifer Live
N122
Mixed
N123
Deciduous
N124
Deciduous
N131
Conifer Live
N133
Conifer Live
N134
Mixed
N135
Conifer Live
N136
Conifer Live
N137
Conifer Dead
N139
Conifer Live
N15
Conifer Live
N16
Conifer Live
N24
Conifer Dead
N27
Conifer Dead
N28
Conifer Live
N29
Conifer Live
N30
Conifer Live
N35
Conifer Live
N36
Deciduous
N39
Conifer Live
N40
Conifer Live
N41
Conifer Live
N42
Mixed
N43
Deciduous
N44
Conifer Live
N45
Conifer Live
N46
Conifer Live
N47
Conifer Live
N48
Conifer Live
N50
Conifer Live
N51
Mixed
N52
Conifer Live
N53
Conifer Dead
N54
Conifer Live
N55
Conifer Live
N56
Conifer Live
N57
Conifer Live
N58
Deciduous
N61
Conifer Live
N62
Conifer Live
N63
Conifer Live
N65
Conifer Live
N66
Conifer Live
N67
Conifer Live
N68
Conifer Live
N70
Conifer Live
N72
Conifer Live
N73
Conifer Live
N75
Deciduous
N76
Conifer Live
N77
Conifer Live
N78
Conifer Live
N80
Conifer Live
N82
Conifer Live
N83
Conifer Live
N84
Conifer Live
N85
Conifer Live
N86
Conifer Live
N88
Conifer Live
N89
Conifer Live
N9
Conifer Live
N90
Conifer Live
N93
Conifer Dead
N94
Conifer Live
N95
Conifer Live
N96
Conifer Live
N98
Conifer Live
N99
Conifer Live

Plot_ID

16.3 (11.7)
16.2 (10.6)
11.6 (3.2)
9.6 (5.9)
11.8 (5.6)
12.9 (9)
11.1 (8.8)
17.2 (8.3)
14.2 (5.5)
12.1 (7.4)
21 (5.6)
12.8 (4.1)
18.7 (8.1)
11.4 (5.6)
15.3 (7.8)
15 (8.7)
11.6 (7.3)
14.1 (5.5)
18.3 (5.9)
23.6 (4.6)
19.6 (6.2)
14.2 (7.1)
17.9 (7.9)
11.2 (6)
20.5 (7.6)
15.6 (7.4)
15.1 (4.3)
8.4 (3.5)
15.1 (6)
10.6 (5.4)
8.8 (5.4)
13.3 (5.4)
13.9 (6)
6.6 (2.2)
10.5 (2.2)
4.3 (0.8)
12.6 (4.6)
12.9 (5.3)
8.6 (3.8)
9.3 (4.1)
8.3 (2)
11.6 (4.1)
13.7 (6.1)
7.9 (2.5)
14.3 (5.6)
12.5 (7.2)
7.3 (1.2)
14.2 (5.5)
10 (4.3)
10.2 (3.3)
9.6 (4.1)
7.4 (2.8)
10 (6.9)
7.5 (3.5)
13.1 (6.3)
11.3 (6.4)
11.8 (7.1)
15 (10.1)
19.9 (9.5)
7.8 (5.7)
6.4 (3.9)
2.7 (0.4)
11.8 (5)
12.3 (4.4)
11.2 (4.6)
7.1 (1.7)
9.7 (5.9)
13.5 (3.5)
16.7 (6.3)
12.6 (5.7)
10.8 (6.8)
6.6 (1.8)
15.3 (6)
10.1 (4.6)
21.8 (10.2)
14.1 (7.5)
15.2 (6.2)
7.1 (3.9)
10.8 (2.5)
11 (3.4)
8.5 (4.7)
7.8 (3.7)
12.7 (5)
7.9 (3.6)
9.6 (5.5)
6.4 (2)
8 (2.5)
8.5 (2.8)
6 (1.5)
7.5 (1.4)
7.1 (1.8)
7.1 (3.4)
4 (1.2)
10.4 (6.3)
5.3 (0.7)
7.5 (5.5)
13.3 (5.5)
7.6 (6)
12.1 (4.4)
9.4 (10.4)
19 (9.7)
6.7 (1.5)
3.9 (0.6)
14 (5)
7.4 (2.6)
11.4 (2.6)
11.3 (6.3)
4.4 (0.4)
9.8 (1.4)
9 (2.1)
13.5 (5.4)
12.2 (13)
7.4 (2.8)
5.7 (1.1)
9 (3.5)
18.9 (2.4)
14.7 (3.4)
10.5 (5)

30.6
29.7
15.3
18.2
19.8
25.5
23.8
26.9
20.0
23.3
26.4
17.7
25.4
17.0
25.3
26.1
22.0
20.2
24.5
28.3
23.7
24.1
25.1
19.6
27.3
23.2
19.2
12.8
22.2
16.9
17.4
20.9
21.3
9.6
12.5
5.4
19.0
18.6
12.5
14.8
11.1
16.2
20.8
11.0
21.2
24.1
8.9
20.0
15.8
14.7
15.8
10.6
20.8
11.9
22.3
20.0
21.8
28.7
29.4
9.5
12.3
3.0
18.7
17.0
18.9
9.3
16.6
16.7
23.5
20.4
23.2
8.7
22.6
17.4
31.4
24.8
21.7
11.5
13.6
15.0
13.6
13.2
21.0
12.2
17.3
9.4
10.6
11.8
8.1
9.3
9.3
11.9
5.5
18.6
6.0
15.7
20.0
15.3
17.9
17.4
30.5
8.5
4.4
19.6
11.0
13.6
21.2
4.6
11.3
11.0
21.1
24.7
9.9
7.1
13.3
21.1
17.1
18.0

18.3 (10.3)
23.1 (12.4)
13.4 (4.5)
11 (6.1)
13.9 (11.6)
18.8 (14.6)
15.9 (6.5)
20.1 (7.9)
15.7 (4.6)
15.3 (10.2)
22.4 (7)
8.5 (3.2)
31.9 (13.8)
12.7 (6)
18.3 (9.3)
19.2 (10.3)
13.4 (9.3)
13.1 (5.9)
24.7 (11.7)
26.2 (7)
29.3 (6.4)
19.3 (12.4)
20.4 (9.1)
19.8 (12.8)
30.2 (15.8)
19.8 (9.8)
16.4 (5.5)
8.3 (3.4)
28.4 (3.4)
14.2 (6.9)
10.1 (7.6)
21.4 (10.6)
21.8 (11.3)
7.4 (2.8)
10.7 (3.5)
6 (1.6)
13 (7.3)
14.5 (7)
13.4 (9.7)
11.5 (7.1)
10.9 (4.1)
11.8 (5.1)
18.3 (8.9)
8.3 (3.2)
14.3 (7.8)
13.4 (8.6)
7.7 (3.1)
13.6 (5.4)
10.5 (6.3)
10.2 (4.6)
9.8 (8.2)
15 (7.1)
23.2 (16.1)
7.8 (4.3)
14.2 (8.6)
13.3 (7.3)
15.3 (11)
19.3 (13.6)
25.1 (11.7)
9 (8.7)
10.2 (5.2)
5.5 (1.1)
19 (10.2)
15.6 (7.2)
9.3 (5)
7.1 (2.8)
11.3 (9.2)
17.6 (6)
16.9 (7.4)
15 (10.7)
12.3 (8.7)
8.2 (2.9)
16.8 (8.9)
12.6 (7.1)
24.3 (13.6)
18.1 (10.4)
19.5 (10.8)
11.4 (5.8)
7.8 (3.2)
13.6 (6.8)
10 (7.8)
8.3 (4)
17.1 (6.1)
9.1 (5.9)
10.2 (6.4)
7.7 (3.6)
9.7 (4)
10.4 (4.1)
7.4 (2.8)
8.4 (3)
8.1 (3)
14.3 (10.6)
5.4 (1.6)
15.5 (8.2)
4.8 (0.9)
11.2 (9.2)
15.2 (7.1)
9.6 (6.4)
15.2 (6.7)
9.9 (12.5)
26.6 (19.4)
8.5 (3.4)
5.1 (1)
16.1 (7)
9.4 (4.7)
11 (4)
14.6 (8.8)
6.3 (0.3)
9.2 (2.9)
11.7 (4.1)
11.6 (5.6)
21.8 (25.8)
13.8 (3.1)
7.5 (2.9)
10.2 (4.5)
29.5 (6.5)
19.9 (5.8)
10.9 (5.2)

8.7 (8.3)
3.6 (5.9)
9.8 (2.5)
3.3 (4.8)
7.5 (4.1)
5.9 (7.6)
7.4 (8.3)
0.4 (0.7)
7.5 (5.4)
2.3 (2.3)
17.4 (5.4)
9.7 (4.2)
10.2 (7.4)
5.4 (3.4)
6.1 (7.2)
4 (4)
8.4 (5.3)
8.1 (4.5)
5.2 (4.8)
15.6 (6.7)
15.6 (5)
6.3 (5.2)
9.4 (6.6)
5.7 (4.3)
8.5 (6.7)
9.4 (7.5)
7.6 (5.3)
3.4 (2.6)
12.6 (1.4)
5.5 (4.9)
3.3 (3)
5.1 (4.6)
6.6 (5.6)
2.2 (1)
3.4 (0.9)
0.4 (0.7)
5.9 (2.4)
5.1 (3.1)
1.9 (1.2)
3 (1.9)
2.7 (1.4)
6.8 (2.6)
5.1 (3.8)
3.7 (1.2)
6.7 (3.1)
5 (3.4)
2 (0.7)
5.6 (3.8)
3.8 (3)
4.4 (3)
4.2 (2.6)
1.1 (0.3)
4.5 (7.3)
3.4 (1.8)
4.9 (3.1)
5 (4.3)
3.1 (3.2)
8.9 (6)
7.2 (4.5)
1.9 (1)
2.9 (1.6)
1 (0)
4.9 (4.1)
2.6 (1.9)
5.3 (2.5)
1.1 (0.6)
3.4 (3.7)
4 (3.6)
11.1 (5)
4.2 (2.9)
4.4 (3.9)
1.3 (0.7)
8.3 (6.4)
5.2 (3.7)
9 (6.9)
4.8 (4.6)
6.1 (3.6)
2.6 (3.1)
5.6 (1.6)
2.5 (1.9)
3.5 (3.4)
4.2 (2.4)
2.4 (1.8)
2.5 (1.5)
3 (2.9)
1.4 (0.5)
2.2 (1.8)
3.6 (1.4)
1.5 (0.8)
1.8 (0.8)
1 (0)
2.4 (1.5)
2.3 (0.8)
4.6 (3.3)
3 (0)
3.5 (2.2)
6.1 (3.4)
2.3 (3.5)
4.3 (1.6)
1.3 (0.5)
7.3 (4.1)
1 (0)
0 (0)
6.3 (2.8)
1.6 (0.9)
4.6 (1.4)
2.6 (2.7)
1 (0)
4.3 (1)
2.6 (1.3)
6.2 (4.6)
2 (2.6)
1.8 (1.3)
1 (0)
2.4 (1.9)
4.5 (1.3)
5.8 (2.5)
4.9 (3.6)

225.5
431.1
165.2
117.9
399.4
955.4
58.8
255.0
268.4
114.9
550.1
820.2
338.7
341.7
371.0
250.1
1742.7
380.3
460.6
682.0
484.7
498.1
280.2
160.7
366.2
342.3
291.3
111.0
207.0
199.6
104.5
352.2
329.7
244.9
128.3
34.1
1307.4
188.2
30.1
328.1
77.4
320.0
611.8
114.3
293.1
172.0
99.8
176.7
272.1
162.0
206.5
413.8
395.4
306.3
369.4
505.9
259.6
848.6
656.2
316.7
48.0
0.4
85.3
182.9
303.4
33.4
173.4
58.2
151.8
306.4
249.2
16.9
234.6
914.2
453.5
182.5
557.8
36.9
291.5
278.3
59.0
201.3
219.2
146.1
142.6
118.9
98.0
143.8
104.3
63.2
514.6
129.4
3.0
320.6
127.7
514.1
290.5
42.1
109.7
78.4
457.5
46.0
1.5
232.8
184.1
137.7
59.7
0.4
145.4
107.3
614.2
99.0
10.9
37.2
206.8
80.2
275.2
352.3

75.1
206.6
1.5
57.6
78.9
63.9
33.1
22.5
122.9
73.1
28.4
25.0
173.4
152.5
51.9
174.2
55.3
119.1
287.6
178.8
29.4
177.0
44.4
43.5
118.2
63.0
128.3
66.5
0.0
65.6
76.4
123.8
46.8
22.0
125.8
2.2
319.8
72.8
13.1
136.9
77.2
252.3
251.8
111.3
231.0
172.0
68.0
108.7
109.6
120.6
155.7
0.0
0.6
266.0
242.1
65.1
161.3
722.1
171.7
93.9
46.2
0.4
46.6
35.2
294.9
31.5
113.0
43.2
0.5
148.5
196.7
16.8
129.2
12.5
389.6
131.0
337.0
21.4
192.7
86.7
16.9
143.3
64.6
82.4
142.5
23.4
24.7
3.2
40.7
63.2
5.4
117.8
2.9
210.8
0.0
477.9
167.3
32.2
0.0
0.5
457.5
46.0
1.5
204.7
24.8
136.2
58.1
0.4
144.0
107.3
453.2
96.4
7.3
37.2
98.1
74.5
243.3
188.9

150.4
97.0
0.0
60.3
0.0
236.7
25.6
232.5
145.5
41.8
0.0
5.0
0.0
125.5
303.8
75.9
0.3
103.6
0.0
67.6
0.0
4.0
235.0
32.8
148.4
0.0
1.9
0.3
0.0
57.7
5.1
4.4
120.6
0.1
2.5
0.0
0.8
81.1
17.0
0.3
0.1
35.9
73.4
3.0
54.3
0.0
0.0
64.6
0.0
3.2
0.7
0.0
0.0
8.5
0.0
40.8
0.1
126.6
484.5
0.0
1.8
0.0
6.9
83.4
8.5
1.9
30.5
0.0
0.0
106.2
52.5
0.1
0.0
0.0
0.0
19.2
0.2
8.9
3.3
0.6
8.4
58.0
141.1
0.1
0.1
0.0
0.0
0.0
0.0
0.0
0.0
11.5
0.1
109.8
0.4
36.2
123.2
9.8
0.0
45.8
0.0
0.0
0.0
28.1
0.0
1.4
1.6
0.0
1.4
0.0
161.0
2.6
3.5
0.0
7.8
5.7
0.0
163.4

Plot tree
Plot
Plot dbh CBH (height Volume/ha Live conifer
Dead
height Q90th tree mean(sd)
to live
(all species) m3/ha
conifer
mean(sd) height
(cm)
crown)
(m3/ha)
(m3/ha)
(m)
(m)
mean(sd)
0.0
127.5
163.7
0.0
320.5
654.8
0.0
0.0
0.0
0.0
521.7
790.3
165.3
63.7
15.3
0.0
1687.0
157.6
173.0
435.6
455.3
317.1
0.7
84.4
99.6
279.2
161.0
44.2
207.0
76.2
22.9
224.0
162.3
222.8
0.0
31.8
986.8
34.3
0.0
191.0
0.0
31.8
286.5
0.0
7.8
0.0
31.8
3.4
162.5
38.2
50.2
413.8
394.8
31.8
127.3
400.1
98.2
0.0
0.0
222.8
0.0
0.0
31.8
64.3
0.0
0.0
29.9
15.0
151.3
51.7
0.0
0.0
105.4
901.6
63.9
32.3
220.6
6.6
95.5
191.0
33.7
0.0
13.6
63.7
0.0
95.5
73.3
140.6
63.7
0.0
509.3
0.0
0.0
0.0
127.3
0.0
0.0
0.0
109.7
32.1
0.0
0.0
0.0
0.0
159.3
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
100.9
0.0
31.8
0.0

Deciduous
(m3/ha)

0.4
1.2
1.2
0.9
3.2
2.6
0.3
0.4
1.1
0.9
2.2
1.4
2.3
1.4
0.7
1.3
2.6
1.3
2.8
2.9
2.8
4.2
0.5
2.1
1.1
2.1
1.8
1.2
1.7
1.6
1.4
2.9
1.4
0.4
1.1
0.1
3.8
0.7
0.2
1.6
1.5
2.4
2.6
2.2
1.4
1.2
1.1
0.6
1.6
1.5
4.6
0.7
1.9
3.1
2.2
1.8
1.3
5.2
1.8
0.8
1.2
0.1
0.6
0.4
2.0
0.4
1.0
0.6
0.5
1.5
1.7
0.5
1.6
1.3
1.9
1.3
3.0
0.4
1.2
1.0
0.6
1.9
0.9
1.3
1.3
0.7
0.7
0.5
1.5
1.2
0.4
3.6
0.1
4.2
0.0
7.9
1.6
0.4
0.7
0.1
3.6
0.9
0.1
1.9
0.5
1.0
0.5
0.0
1.3
1.4
2.6
0.6
0.2
1.0
1.6
0.7
1.8
1.7

CFL
(kg/m2)

0.4
1.2
0.0
0.9
1.3
0.9
0.3
0.4
1.0
0.7
0.2
0.3
1.2
1.3
0.6
1.2
0.3
0.6
2.0
1.2
0.3
1.3
0.5
1.2
0.8
0.8
1.0
1.1
0.0
1.0
0.7
1.2
0.7
0.3
1.1
0.1
2.8
0.7
0.2
1.6
1.5
2.3
2.2
2.1
1.4
1.2
1.1
0.6
1.3
1.3
4.3
0.0
0.1
3.0
1.8
0.7
1.3
4.7
1.8
0.7
1.1
0.1
0.5
0.4
2.0
0.4
0.8
0.6
0.0
1.4
1.7
0.5
1.1
0.1
1.7
1.1
2.7
0.4
1.2
0.9
0.3
1.8
0.8
1.3
1.3
0.6
0.6
0.0
1.5
1.2
0.2
3.6
0.1
3.2
0.0
7.4
1.4
0.4
0.0
0.0
3.6
0.9
0.1
1.8
0.5
1.0
0.5
0.0
1.3
1.4
2.6
0.6
0.2
1.0
1.4
0.7
1.8
1.6

Conifer
live CFL
(kg/m2)

605
700
2260
1496
2260
1623
414
764
1783
573
1432
5666
446
2642
1305
796
1910
3056
891
1241
891
1146
764
732
382
1178
1910
2928
573
1814
1560
987
828
1305
2451
382
3533
1114
286
1783
1878
3692
1496
4838
1751
1114
3438
1401
2260
3215
3310
414
414
8722
1687
1655
764
2005
1019
987
1146
127
286
828
5093
1910
1401
382
573
1401
1814
923
1114
1337
668
700
1305
700
6016
923
796
6334
1432
1878
2005
1305
1146
1369
3088
2769
923
1369
573
2196
191
5093
1846
637
923
223
605
2037
350
1401
764
2228
382
64
4043
2005
6112
127
191
2546
2610
159
1050
4870

SPH

20.9
26.0
14.1
12.6
18.1
23.7
17.1
21.6
16.4
18.3
23.4
9.1
34.5
14.0
20.5
21.7
16.3
14.4
27.3
27.1
29.9
22.9
22.3
23.4
33.8
22.0
17.3
9.0
28.6
15.8
12.6
23.8
24.4
7.9
11.3
6.1
14.9
16.1
16.2
13.5
11.6
12.9
20.3
8.9
16.3
15.8
8.2
14.6
12.3
11.2
12.8
16.5
27.9
8.9
16.6
15.1
18.7
23.5
27.6
12.4
11.5
5.6
21.3
17.1
10.5
7.6
14.5
18.5
18.4
18.4
15.1
8.7
18.9
14.4
27.7
20.8
22.2
12.8
8.5
15.2
12.6
9.2
18.1
10.8
12.0
8.5
10.5
11.2
7.9
8.9
8.7
17.7
5.6
17.5
4.9
14.5
16.7
11.5
16.6
15.2
32.6
9.1
5.1
17.6
10.5
11.7
16.9
6.3
9.7
12.3
12.9
31.2
14.1
8.1
11.2
30.1
20.7
12.1

QMD (cm)

1
3
8
1
1
1
1
1
1
1
16
3
1
2
1
1
3
3
1
5
1
1
3
1
4
2
3
1
1
1
1
1
1
2
2
0
3
4
0
1
1
3
1
2
3
3
1
8
1
1
1
1
1
1
2
1
2
2
6
2
1
0
8
3
3
1
2
3
14
2
2
1
3
3
9
2
3
4
5
1
2
1
2
1
1
1
1
3
1
1
1
1
2
1
0
1
3
1
5
0
5
1
0
3
1
3
3
0
3
2
1
6
1
1
1
5
4
1

CBH at
0.037kg/
m3 (m)

2
22
14
7
25
29
2
5
14
13
25
17
22
13
7
16
23
19
22
27
23
22
9
16
21
22
18
11
18
15
15
19
18
6
10
0
18
13
0
12
9
16
19
10
17
19
7
14
16
13
15
7
23
15
20
18
17
28
22
7
10
0
13
7
18
5
16
10
21
17
21
5
21
17
28
20
20
6
12
12
19
13
11
11
15
6
8
9
7
7
5
12
3
18
0
22
17
3
13
0
25
6
0
17
6
11
5
0
10
9
20
7
2
6
12
14
15
15

CH at
0.037
kg/m3
(m)

Conifer Canopy CBH at
CBH at length at 0.011kg/
0.037kg/
0.037
m3 (m)
m3 (m)
kg/m3
(m)
0
1
1
3
19
1
0
6
1
1
6
1
1
24
1
1
28
1
1
0
1
1
4
1
1
13
1
1
12
1
0
9
3
3
14
1
4
21
1
2
11
1
1
6
1
1
15
1
15
20
1
6
16
1
1
21
1
5
22
3
0
22
1
4
21
1
3
6
1
1
15
1
4
17
1
2
20
1
3
15
1
1
10
1
0
17
1
1
14
1
2
14
1
2
18
1
1
17
1
2
4
1
2
8
1
0
0
1
3
15
1
4
9
2
0
0
1
1
11
1
1
8
1
4
13
1
2
18
1
2
8
1
3
14
1
3
16
1
1
6
1
8
6
2
1
15
1
1
12
1
1
14
1
0
6
1
0
22
1
1
14
1
3
18
1
1
17
1
2
15
1
3
26
1
6
16
2
2
5
1
1
9
1
0
0
1
11
5
1
3
4
1
3
15
2
1
4
1
2
14
1
3
7
1
0
7
11
2
15
1
2
19
1
1
4
1
3
18
1
0
14
2
9
19
2
2
18
1
3
17
1
4
2
1
5
7
1
1
11
1
2
17
1
1
12
1
2
9
1
1
10
1
1
14
1
1
5
1
1
7
1
0
6
2
1
6
1
1
6
1
1
4
1
1
11
1
2
0
1
1
17
1
0
0
0
1
21
1
3
14
1
1
2
1
0
8
3
0
0
1
5
20
3
1
5
1
0
0
1
3
14
2
1
5
1
3
8
2
3
2
1
0
0
0
3
7
2
2
7
1
1
19
1
0
1
4
1
0
1
1
5
1
1
11
1
5
9
3
4
11
1
1
14
1
24
24
15
11
27
31
6
9
18
20
26
18
26
15
20
23
26
20
25
29
23
28
15
19
27
23
19
13
21
17
19
21
20
8
12
3
19
16
8
14
10
18
20
12
20
22
8
17
18
14
21
9
25
17
22
21
26
30
24
22
12
2
16
12
20
8
24
16
22
21
23
8
22
19
31
23
22
14
13
13
21
14
15
15
17
8
10
10
8
9
7
13
4
20
0
27
19
11
16
2
28
8
2
19
9
13
17
0
11
10
21
27
6
7
13
18
16
17

CH at
0.011
kg/m3
(m)
0.05
0.08
0.21
0.12
0.21
0.16
0.04
0.06
0.08
0.08
0.35
0.19
0.25
0.15
0.07
0.08
0.28
0.11
0.18
0.23
0.42
0.76
0.06
0.23
0.08
0.22
0.20
0.17
0.17
0.13
0.18
0.28
0.14
0.07
0.18
0.03
0.39
0.07
0.02
0.18
0.25
0.34
0.22
0.37
0.14
0.08
0.22
0.05
0.16
0.17
0.83
0.13
0.31
0.38
0.18
0.12
0.10
0.40
0.23
0.07
0.17
0.02
0.07
0.05
0.20
0.07
0.07
0.07
0.06
0.12
0.12
0.09
0.10
0.19
0.11
0.08
0.27
0.05
0.19
0.12
0.06
0.21
0.09
0.16
0.10
0.14
0.12
0.08
0.31
0.23
0.09
0.53
0.04
0.36
0.01
0.65
0.17
0.05
0.09
0.01
0.26
0.20
0.02
0.22
0.09
0.14
0.05
0.01
0.26
0.25
0.20
0.04
0.04
0.25
0.18
0.07
0.23
0.17

1
6
11
3
1
3
2
2
2
2
21
14
18
8
2
3
16
16
15
22
18
1
5
2
17
18
14
3
13
12
2
5
3
3
5
1
9
9
4
6
4
10
11
5
11
13
3
10
5
9
3
2
3
5
12
1
5
18
9
3
6
1
10
4
6
2
9
4
19
8
4
3
17
5
24
9
9
5
8
4
18
7
6
4
2
3
2
6
2
3
2
5
3
8
3
8
11
2
7
2
14
2
1
10
3
7
4
2
6
4
8
6
2
2
5
7
9
4

0.04
0.07
0.00
0.12
0.11
0.10
0.04
0.06
0.07
0.07
0.02
0.05
0.07
0.15
0.07
0.08
0.06
0.06
0.15
0.09
0.02
0.11
0.06
0.13
0.04
0.07
0.10
0.17
0.00
0.10
0.07
0.13
0.10
0.06
0.18
0.03
0.29
0.07
0.02
0.18
0.25
0.33
0.22
0.37
0.13
0.08
0.22
0.05
0.14
0.12
0.83
0.00
0.02
0.37
0.15
0.08
0.10
0.37
0.23
0.06
0.17
0.02
0.04
0.05
0.20
0.07
0.05
0.07
0.01
0.12
0.12
0.09
0.08
0.02
0.10
0.07
0.25
0.05
0.19
0.11
0.05
0.21
0.08
0.15
0.10
0.13
0.11
0.01
0.31
0.23
0.04
0.53
0.04
0.27
0.00
0.60
0.16
0.05
0.00
0.01
0.26
0.20
0.02
0.21
0.08
0.14
0.05
0.01
0.26
0.25
0.20
0.03
0.04
0.25
0.18
0.06
0.23
0.16

0
0
0
3
0
0
2
2
0
0
0
0
0
0
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
5
1
0
0
4
6
4
0
0
0
0
13
3
0
0
0
0
0
0
0
0
0
0
0
9
0
6
1
0
0
0
2
0
4
0
0
4
3
0
0
0
0
0
5
0
0
0
0
0
0
2
0
0
0
0
3
0
5
0
0
0
0
0
2
0
0
14
2
1
0
0
7
4
2
6
4
0
0
2
2
0
0
9
0

338
49
274
133
521
459
595
137
18
90
143
236
162
345
15
139
186
35
61
62
508
90
82
137
105
402
119
11
217
-

CBD
CBD max Conifer Conifer
CWD
(kg/m3) height
CBD
CBD max (m3/ha)
(m)
(kg/m3) height
(m)
110653
11838
70676
23221
115032
149274
181130
49818
7333
16327
24659
52904
39614
80886
3651
48864
31904
5902
10348
10581
156222
24411
27363
44784
22196
126384
43402
1856
63562
-

CWD
(kg/ha)

Appendix 2: Plot Empirical summary

163 | P a g e

Appendix 3: Predicted metrics by community forest compartment
Critical surface fire intensity

CSI Roselake compartment (kW/m)

6040000

CSI (kW/m)

6030000

UTM Northing

<250
250-499

500-999
1000-1499
>1499

6020000

NA

6010000

280000

290000

300000

310000

UTM Easting

Figure 50. Roselake CSI

CSI Babine compartment (kW/m)

6035000

UTM Northing

CSI (kW/m)
<250
250-499

6030000

500-999
1000-1499
>1499

NA

6025000

335000

340000

345000

350000

UTM Easting

164 | P a g e

Figure 51. Babine CSI.

CSI Hannay compartment (kW/m)

6015000

UTM Northing

CSI (kW/m)
<250

6010000

250-499

500-999
1000-1499
>1499

NA
6005000

6000000

340000

345000

350000

355000

UTM Easting

Figure 52. Hannay CSI.

CSI Southside compartment (kW/m)
5990000

UTM Northing

CSI (kW/m)
<250
5980000

250-499

500-999
1000-1499
>1499

5970000

NA

280000

300000

320000

UTM Easting

Figure 53. Southside CSI.
165 | P a g e

Fuel metrics
Roselake Compartment and surrounding ALS coverage

166 | P a g e

Babine compartment and surrounding ALS coverage

167 | P a g e

Hannay compartment and surrounding ALS coverage

168 | P a g e

Southside compartment and surrounding ALS coverage

169 | P a g e

Appendix 4: Variable Importance.
Basal area variable importance plot

Figure 54 Variable importance plot for the random forest model of basal area metric, displaying only the top 20 out of 65 ALS metrics
included in the model. The vertical axis represents the predictor variable ranked by importance, with the highest variable being the
most significant to model predictive power. The horizontal axis shows the importance score, an indicator of the predictive utility of
variables, determined using the varImpPlot() function in the randomForest package (Liaw & Wiener, 2002). Variable importance
scores are calculated based on the mean decrease in accuracy when each predictor variable is permuted, reflecting its impact on
model performance. This plot shows zmean as the highest variable of importance along with several height percentile metrics.
CBH (height to live crown) variable of importance plot

Figure 55 Variable importance plot for the random forest model of CBH metric, showing higher point height percentile metrics (70,
75, 80 and 85) and several cumulative metrics.
170 | P a g e

CFL variable of importance plot

Figure 56 Variable importance plot for the random forest model of CFL metric, showing two height percentile metrics (zq60 and
zq55), as well as the vertical complexity index metric (VCI).
CBD variable of importance plot

Figure 57 Variable importance plot for the random forest model of CBD metric showing alsvlad (coefficient of leaf area density
metric) as the highest ALS metric along with two low-to-ground cumulative point metrics (zpcum1 and zpcum2). Also showing pabove
zmean which is the percentage of points above the mean.

171 | P a g e

Crown length variable of importance plot

Figure 58 Variable importance plot for the random forest model of crown length metric showing primarily zq ALS metrics (height
percentile metrics).
CWD variable of importance plot

Figure 59 Variable importance plot for CWD showing cumulative height metrics for the first and second meter above ground.

172 | P a g e