MANAGING OLD-GROWTH FORESTS FOR MULTIPLE ECOSYSTEM
SERVICES
by
Luizmar de Assis Barros
B.Sc., Federal Rural University of Rio de Janeiro, 2017

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
February 2020
© Luizmar de Assis Barros, 2020

TABLE OF CONTENTS
Abstract ................................................................................................................................. VI
List of Figures: ..................................................................................................................... VII
List of Tables:........................................................................................................................ IX
List of Acronyms: ................................................................................................................. IX
Acknowledgement: ............................................................................................................... XI
Dedication:............................................................................................................................ XII
1. GENERAL REVIEW ON OLD-GROWTH FORESTS AND THEIR PROVISION
OF ECOSYSTEM SERVICES ............................................................................................... 1
1.1. Introduction: .................................................................................................................. 1
1.2. Old-growth ..................................................................................................................... 3
1.3. Ecosystem Services ....................................................................................................... 9
1.4. Forest state and ecosystem services ............................................................................. 10
1.5. Airborne Laser Scanning (ALS): ................................................................................. 12
1.6. Systematic Conservation Planning .............................................................................. 18
1.7. Thesis Objectives ......................................................................................................... 20
2. DEFINING OLD-GROWTH FORESTS WITH AIRBORNE LiDAR .................... 22
2.1. Introduction: ................................................................................................................ 23
2.2. Material and Method: .................................................................................................. 26
2.2.1.
Study Area: .............................................................................................................. 26
2.2.2.
Data: ......................................................................................................................... 27
2.2.3.
Old-growth attributes: .............................................................................................. 28
2.2.4.
Plot-level definitions of old-growth: ....................................................................... 29
2.2.5.
Lidar processing: ..................................................................................................... 31
2.2.6.
Statistical analysis:................................................................................................... 33
2.3. Results: ........................................................................................................................ 36
2.3.1.
Fieldwork data: ........................................................................................................ 36
2.3.2.
Comparison between “random forest” models: ....................................................... 41
2.3.3.
Comparison between old-growth maps: .................................................................. 43
2.3.4.
Assessing Set-Aside Old-Growth Forests ............................................................... 47
2.4. Discussion:................................................................................................................... 50
2.4.1.
The old-growth index: ............................................................................................. 50
2.4.2.
Application: ............................................................................................................. 51
2.4.3.
Pros and Cons of the old-growth index: .................................................................. 53
2.5. Conclusion: .................................................................................................................. 55
3. SYSTEMATIC CONSERVATION PLANNING OF OLD-GROWTH VALUES . 56
3.1. Introduction: ................................................................................................................ 56
3.2. Material And Method: ................................................................................................. 60
3.2.1.
Study Area: .............................................................................................................. 60
3.2.2.
Estimating Ecosystem Services ............................................................................... 60
3.2.3.
Systematic Conservation for Old-Growth Values ................................................... 62
3.3. Results: ........................................................................................................................ 64
II

3.3.1.
Ecosystem Services ................................................................................................. 64
3.3.2.
Systematic conservation planning applied to OGMAs ............................................ 67
3.4. Discussion:................................................................................................................... 74
3.4.1.
Ecosystem services reserves .................................................................................... 76
3.5. Conclusion: .................................................................................................................. 80
4. CONCLUSIONS:........................................................................................................... 81
4.1. Main Findings and contributions: ................................................................................ 82
4.1.1.
General review on old-growth forests and their provision of ecosystem services: . 82
4.1.2.
Defining Old-Growth Forests With Airborne LiDAR............................................. 83
4.1.3.
Systematic Conservation Planning Of Old-Growth Value ...................................... 85
4.2. Future research: ........................................................................................................... 86
4.3. Concluding Remarks ................................................................................................... 89
5.

REFERENCES .............................................................................................................. 91

6. APPENDICES .............................................................................................................. 108
6.1. Data collection information ....................................................................................... 108
6.2. Allometric Equations for plot-level estimates of old-growth attributes: ................... 111
6.2.1.
Volume-Height, DBH ............................................................................................ 111
6.2.2.
Biomass — Jenkins’s equation .............................................................................. 112
6.2.3.
Carbon.................................................................................................................... 115
6.1. Additional Informal For The Methods In Chapter 2: ................................................ 116
6.2. Distribution of Old-growth attributes based on different old-growth definitions ..... 120
6.3. Water value: ............................................................................................................... 128
6.4. Random forest results: ............................................................................................... 128
6.5. Fragmented versus contiguous reserves: ................................................................... 129
6.6. PrioritizR scenarions for increasing areas ................................................................. 133

III

The UNBC Institutional Repository NON-EXCLUSIVE
DISTRIBUTION LICENSE
Summary: The University of Northern British Columbia (UNBC) maintains an open access digital repository for published and
unpublished material created by members of the UNBC community. By signing this License, you give UNBC the non-exclusive
right to copy, distribute, and transmit your Submission over the Internet through the UNBC Institutional Repository on the terms
set out below. Visitors to the UNBC Institutional Repository site will be permitted to use your Submission for non-commercial
purposes, as set out in greater detail below. You may assign different licensing terms to govern how the UNBC Institutional
Repository visitors may use your Submission. The rights you grant to UNBC under this License will last in perpetuity.
References to the UNBC Institutional Repository include any successor repository designated by UNBC.
This License grants UNBC significant rights and contains important limitations on your rights to your work. Please read it
carefully.
1.

Your Submission. Your Submission is the material described below, and includes the abstract, associated

Submission Information
Title: Managing Old-growth Forests For Multiple Ecosystem Services
Author: Luizmar de Assis Barros
Date:
(that the Submission was made final/made ready for submission to the UNBC Institutional Repository)
Copyright Owner(s): Luizmar de Assis Barros
Publication Details (if Submission previously published):

2.

Representations and Warranties. You represent and warrant that: [check each box]
all information you have provided about the Submission is, to the best of your knowledge and belief, accurate;
if the Submission has more than one author, all co-authors have approved the Submission and have read and
agreed to the terms of this License;

if the Submission is based upon work that has been sponsored or supported by an organization other than UNBC,
you have fulfilled any right of review and any other obligations required by the agreement with such sponsor or
supporter;
the Submission is not subject to any publication delays, except those approved in accordance with the UNBC
Institutional Repository Digital Repository Policies;

to the best of your knowledge, the use and distribution of the Submission as contemplated in this License will not
infringe the copyright or any other intellectual property rights of any other person;
if the Submission contains material for which you do not hold copyright and which is not in the public domain,
I.
you are validly exercising an exception to copyright infringement in accordance with the Copyright
Act, or you have obtained the permission of the copyright owner(s) to use the material as
contemplated in this License and retained records of such permission; and
II.
that such third-party owner material is clearly identified and acknowledged within the text or content
of the Submission;
the Submission does not contain any confidential or proprietary information belonging to others; and the
Submission does not, to the best of your knowledge and belief, contain any libelous or other unlawful matter and
does not violate anyone’s right to privacy.
You acknowledge that the UNBC Institutional Repository staff have not and will not independently review the Submission and that UNBC
is relying upon the representations and warranties you have made above.
If, at any time hereafter, you learn that any of the foregoing statements are inaccurate, you must contact the UNBC Institutional
Repository immediately.

IV

3.

Grant of License. In consideration of UNBC preserving your Submission in perpetuity in the UNBC Institutional Repository,
you hereby grant to UNBC the non-exclusive, world-wide, perpetual, and royalty-free right to use, reproduce, display, convert,
sublicense and distribute, in any format and any medium, your Submission, for the following purposes:
a) uploading and storing your Submission;
b) publishing the Submission electronically Internet via the UNBC Institutional Repository, which means making your
Submission available to the public by telecommunication so that they may view and download your Submission, subject to
the public access terms described in Section 5;
c) converting your Submission to any medium or format for the purposes of security, back-up, preservation and use of the
submission in accordance with this License; and
d) authorizing third-parties to perform any of UNBC’s rights described in this Section, on UNBC’s behalf. If you are
submitting a graduate thesis or dissertation, your grant of rights described above is irrevocable.
If you are submitting a graduate thesis or dissertation, your grant of rights described above is irrevocable and includes, for
greater certainty, UNBC authorizing Library and Archives Canada to copy, display, publish, communicate, archive, convert into
any format, and make your thesis or dissertation available to the public online or in paper form, all for noncommercial purposes.

4.

Publication of your Submission. This License does not affect the ownership of copyright—you maintain whatever rights you
have in the Submission. In addition, this License does not contain a waiver of your moral rights to your Submission. However,
you have granted important rights to UNBC and visitors to the UNBC Institutional Repository, and, therefore, you agree that if
you grant others copyright ownership or a license, it will be consistent with the rights you have granted under this License. For
example, if you wish to publish your Submission in a journal, you will inform the publisher of the rights you have granted under
this License and not agree to terms that restrict those rights.

5.

Terms of Access by the Public. UNBC will make a digital copy of your Submission available in the UNBC Institutional
Repository for the public to view and copy for non- commercial purposes only, without prior permission or charge, provided that
your Submission is not altered in any way and is properly acknowledged, including citing the author(s), title and full
bibliographic details. (Note: this corresponds to the Attribution-Non Commercial-No Derivatives, or CC BY-NC-ND Creative
Commons license) UNBC may, in its discretion, make your Submission available pursuant to a Creative Commons license (or
other licensing platform), on similar terms. If you wish to assign different license terms to your Submission (for example, a
different Creative Commons license), you must inform the UNBC Institutional Repository as part of the submission process to
the UNBC Institutional Repository, and the terms of that license shall apply to your Submission. Public access to your
Submission may be withheld for an agreed upon period of time: see the UNBC Institutional Repository Digital Repository
Policies for applicable policies and processes.

6.

Attribution. UNBC will attribute the Submission as you have indicated in Section 1 or as you may notify UNBC from time to
time, and not alter the Submission except as allowed by this License or with your prior permission.

7.

The UNBC Institutional Repository Digital Repository Policies. You acknowledge and agree that you have read and
understood the UNBC Institutional Repository Digital Repository Policies, and that your Submission complies with those
policies. You acknowledge that the UNBC Institutional Repository Digital Repository Policies may be amended from time to
time and that the then current policies will apply to your Submission. In particular, you acknowledge and agree to the policies
regarding UNBC’s right to withdraw public access to your Submission.

8.

Release and Indemnification. You agree that you will not hold UNBC responsible for any use or misuse of your Submission by
third- parties who access the Submission through the UNBC Institutional Repository. You agree to indemnify and hold UNBC
harmless against any loss, damage, claim, liability, settlement cost or expense (including legal fees) incurred by UNBC and
arising out of or in connection with UNBC’s use of your Submission pursuant to this License, and, for greater certainty, any of
the statements in Section 2 being inaccurate.

I have read, understand and agree to the terms set out in this License. This License may be delivered and/or stored by electronic means, and
such electronic copy delivered to and stored by UNBC will be deemed to be the official record of this License.

Date: ______________________
Signature: _________________________ Print Name: ____________________________________

V

Abstract
Old-growth forest reserves offer the potential to facilitate the maintenance of multiple
ecosystem services (ES), such as carbon storage, water and recreation, in managed
landscapes. However, substantial challenges exist with regard to defining and identifying oldgrowth forests, and suitably locating priority areas for old-growth conservation. To address
these issues, I developed a structure-based old-growth index using field and LiDAR metrics
that allowed old-growth values to be estimated at a fine grain across a landscape. I then used
a spatial prioritization tool to simulate old-growth reserves for multiple ESs. Using this
framework I evaluated trade-offs between forest ESs including timber. This thesis contributes
to the management of old-growth forests by providing a quantitative and repeatable
framework to identify, assess and monitor old-growth values while indicating the scope for
the establishment of old-growth reserves for multiple ESs.

VI

List of Figures:
Figure 2.1 Location of Chinook community forest tenure areas and distribution of Oldgrowth management areas (OGMAs). ............................................................................. 27
Figure 2.2 Method design for Random forest model generated to create old-growth maps
from fieldwork classification of forest succession and old-growth indexes.................... 35
Figure 2.3 Histograms of old-growth attributes best correlated with age, displaying the
distribution of the data. Distributions are not normal, which indicates the need for nonparametric models. ........................................................................................................... 37
Figure 2.4 Age distribution according to each different old-growth definition developed in
this study. The names on top of each graph represent the definition and are listed and
described in Table 2.2. ..................................................................................................... 41
Figure 2.5 Spatial correlations and constraints between old-growth models with red
representing underestimation and blue super estimations. .............................................. 46
Figure 2.6 Forest was classified into I) four different forest successions (classification) and
II) gradient of the abundance of old-growth attributes (regression). Fourteen locations
were visited to evaluate the effectiveness of these two classifications. Four of those sites
are depicted here: a) depicts low level of old-growth attributes (young forest), b) bare
rock c) intermediate level of old-growth attributes. a) Model Accuracy: 54.4% and
“Very-high” class error = 63,0%, b) R2: 71.5%, MSE= 0.55. ........................................ 48
Figure 2.7 Summary of old-growth assessment in the study area, utilizing the old-growth
model 7, where a) displays the percentage cover of each old-growth value class for the
whole landscape and only for the areas within OGMAs, and b) depicts the number of
OGMAs per percentage forest cover of “Very-high” old-growth value. ........................ 49
Figure 3.1 Ecosystem Services and landscape features utilized in the development of
“prioritizR” scenarios. Section 2.2.5 describes the framework applied to generate the a)
old-growth index (Figure 2.2). Same framework was applied to estimate b) carbon, c)
tree diversity, e) timber, and in Chinook Community Forest. The d) water value and
landscape features f) to i) were developed with surface analysis in ArcMap. ................ 62
Figure 3.2 Correlation analyses of ecosystem services and landscape variables a) in the
whole landscape, and b) within OGMAs. Values in the grid correspond to the
correlation direction and strength between the variables, and the color scale, in the x
axis, represents the significance level (p-value<0.01) ..................................................... 66
Figure 3.3 Ecosystem service provisioning represented (%) in current OGMAs network and
each “prioritizR” model. Percentages are the amount of the ecosystem services within
the reserve network in relation to the total amount of the ecosystem in the whole
landscape.......................................................................................................................... 68
Figure 3.4 Comparison between increase in OGMAs' areas starting from current OGMAs
(Fixed) and OGMAs designed to prioritize each individual ecosystem services and all
features together (Not-fixed). Old-growth prioritization is represented in figures from a)
to c), where a) represents current OGMAs areas priority old-growth reserves with 0%
increament b) are the reserves with 1% area increase and c) 10% respectively. The
reserves with the prioritization of all ESs are represented in the images from d) to f),
where d) are the reserves with 0% increment in area, e)1%, and f) 10%. ....................... 70

VII

Figure 3.5 Trade-off between provision of ecosystem services and timber for harvesting.
Images depict the effect of timber availability on the amount of representation of each
ecosystem service that is removed together with timber. A) old-growth value removed
from the landscape for harvesting scenarios for harvesting yields of 100% , 75%, and
50%, B) Carbon storage, C) Tree Diversity, and D) Water values. ................................ 72
Figure 3.6 Trade-off curve between timber harvesting and setting aside areas for the
provision of a) old-growth, b) carbon, c) tree diversity, d) water value, and e) all features
together. Scenarios were designed to have a starting point at the current OGMAs
network. Thus, 1% increase means that an additional of 1% of the landscape was setaside and included in the current OGAs’ network........................................................... 74
Figure 6.1 Illustration of the a) referential tree, b) plot center with the high precision gps, and
c) tree core. .................................................................................................................... 110
Figure 6.2 Pipeline of LiDAR processing using LASTools from the raw LiDAR files (.las)
to the raster outputs (.bil) used to create ecological meaningful metrics for old-growth
attributes. ....................................................................................................................... 117
Figure 6.3 Distribution of old-growth attributes per forest succession defined by age classes
(AGE_CLASS). ............................................................................................................. 120
Figure 6.4 Distribution of old-growth attributes per forest succession defined by k.means
classification of all old-growth attributes, including AGE (KOGA+AGE). ................. 121
Figure 6.5 Distribution of old-growth attributes per forest succession defined by k.means
classification of all old-growth attributes, except AGE (KOGA). ................................ 122
Figure 6.6 Distribution of old-growth attributes per forest succession defined by k.means all
proximity values from unsupervised random forest of old-growth attributes, including
age (RFOGA+AGE). ..................................................................................................... 123
Figure 6.7 Distribution of old-growth attributes per forest succession defined by k.means all
proximity values from unsupervised random forest of all old-growth attributes, except
AGE (RFOGA). ............................................................................................................. 124
Figure 6.8 Distribution of old-growth attributes against age as a continuous variable (all plot
ages were divided by the maximum age found during inventory and multiplied by five,
since age classification was divided into five classes. ................................................... 125
Figure 6.9 Distribution of old-growth attributes against old-growth index OGA+AGE (sum
of all attributes). ............................................................................................................. 126
Figure 6.10 Distribution of old-growth attributes against the old-growth index OGA (sum of
all attributes, except age). .............................................................................................. 127
Figure 6.11 ArcMap pipeline for the development of a wetness index as a proxy for water
provision. In this model, wetter areas are expected to have higher potential for water
provision. ....................................................................................................................... 128
Figure 6.12 Test with multiple boundary length penalties to visually select between OGMAs
designs that are extremely fragmented or overly contiguous. ....................................... 130
Figure 6.13 Comparison between three different levels of clumpness, from more fragmented
to more contiguous reserves networks, and three different reserves network sizes (1X=
current OGMA size, 2X= double the current size, and 3X=triple). .............................. 131
Figure 6.14 Comparison between current OGMAs' network and OGMAs designed to priority
a) Old-growth, b) Water value, c) Tree diversity, d) carbon, e) all features. The
representation of how many times one planning unit was taken as priority is represented
in f)................................................................................................................................. 133
VIII

Figure 6.15 Comparison between increase in OGMAs' areas starting from current OGMAs
(Fixed) and OGMAs designed to prioritize each individual ecosystem services and all
features together (Not-fixed). Old-growth prioritization is represented in figures from a)
to c), with areas increase of 1%, 5%, and 10% respectively; carbon prioritization from d)
to e), tree diversity from f) to h); water values from j) to l), and all feature from m) to
o). ................................................................................................................................... 135

List of Tables:
Table 1.1 Method utilized to identify and map old-growth forests. ......................................... 7
Table 1.2 Airborne LiDAR delivered metrics for old-growth forest attributes with area-based
approach (ABA) and individual tree detection (ITD)...................................................... 15
Table 2.1 Field measurement of Old-growth attributes. ......................................................... 29
Table 2.2 Plot level definitions of old-growth forests. ........................................................... 31
Table 2.3 LiDAR metrics utilized in the random forest models. ............................................ 32
Table 2.4 Plot level classification of old-growth values ......................................................... 38
Table 2.5 Summary of statistics for the five old-growth classification random forest
models .............................................................................................................................. 42
Table 2.6 Statistical summaries of old-growth index models ................................................. 43
Table 2.7 Comparison between models to evaluate and locate differences between classes. 47
Table 3.1 Conservation planning scenarios for OGMA design with “prioritizR” .................. 64
Table 3.2 Summary of results from cross-validated random forest models (fours fold
stratifications with 10 replicates) ..................................................................................... 65
Table 6.1 List of attributes measured for each plot............................................................... 108
Table 6.2 Volume parameters (Penner et al. 1997, Standish et al. 1985) ............................. 111
Table 6.3 Parameters aboveground biomass and component biomass ratio (Jenkins) ......... 114
Table 6.4 Woody specific gravity and carbon content.......................................................... 115
Table 6.5 Correlation of old-growth attributes with stand age to select the best ten
correlations of old-growth attributes with estimated stand age. CV% = Coefficient of
variation, Std = Standard Deviation, dbh = diameter at the breast height. .................... 118
Table 6.5 Ranking of importance of LiDAR metrics............................................................ 128

List of Acronyms:
ABA - Area-Base Approach
ALS - Airborne Laser Scanning
BEC - Biogeoclimatic
CCF - Chinook Community Forest
CMI - Change Monitoring Inventory
CWD - Coarse Woody Debris
dGPS - differential Global Positioning System
DSM - Digital Surface Model
IX

DTM - Digital Terrain Model
ES - Ecosystem Service
GIS - Geographic Information System
InVEST - Integrated Valuation of Ecosystem Services and Tradeoffs
LiDAR - Light Detection and Ranging
LULC - Land Use/ Land Cover
MFLNRORD - Ministry of Forests, Lands, Natural Resource Operations & Rural
Development
OGMA - Old-Growth Management Area
TIN - Triangulated Irregular Networks

X

Acknowledgement:
First, I find it essential to thank God for His blessings that kept me strong during this
journey despite the many obstacles and challenges encountered throughout.
It would not be possible to write this master’s thesis without the support of many kind
people that surrounded me in the past two years and a half, and only some of who is possible
to give a special acknowledgement.
I want to thank my supervisors Dr. Ché Elkin and Dr. Oscar Venter, for the trust and
guidance. They offered me the so rare opportunity to have the freedom to err and learn with
my mistakes, providing numerous opportunities for the person and professional growth. I
thank the committee members Dr. Roger Wheate and Dr. Joseph Shea, for the feedback that
contributed to the improvement of this thesis. Also, a special thanks to Dr. Michelle Venter,
who kindly provided comments and feedback that significantly, improved this manuscript. I
would like to acknowledge the University of Northern British Columbia for providing such
an excellent environment for learning. The Chinook Community Forest, represented by Ken
Nielsen, for the financial support and Free Growing Forestry for the data collection in Burns
Lake.
Finally, I would like to acknowledge with gratitude the Brazilians in Prince George
for the friendship and shared experiences. The Sunday service in the Grace Anglican Church
and its congregants made my Sundays meaningful and warm. My family that even so far was
always supporting and listening to my struggles on an almost daily bases. For each of the
members of the two labs that I had the enormous privilege to be part of, especially, Kristen
H. Pearson, Lu Gah, Luiz Terezan, Rafael De Camargo. My roommates and sisters Jessica
Chaves Cardoso and Aimee Irene Apsay, who became my Prince George family, I thank for
all the support and experiences we shared in those years. A special thanks to my supportive
and understanding partner Simon Shen for all the sweet distractions.

XI

Dedication:
Dedicated to my mom, Lucia Helena de Paiva Barros (1960-2017), a strong and gentle
soul who with love and patience taught me to believe in hard work and to appreciate the
small things in life as not much is needed for one to be happy.

XII

1. GENERAL REVIEW ON OLD-GROWTH FORESTS AND THEIR PROVISION
OF ECOSYSTEM SERVICES

1.1. Introduction:
“Old-growth” is a term associated with forests in the advanced development stage, with
specific structures, natural processes, and no significant anthropogenic interference (Mosseler
et al. 2003c, Spies 2004, Hilbert and Wiensczyk 2007). When a forest is allowed to reach the
latest stages of development, it attains attributes critical for the maintenance of biodiversity
in the landscape (DellaSala et al. 1996, Spies 2004). Forests with these characteristics are
valuable and rare resources declining rapidly worldwide (Watson et al. 2016, 2018).
Furthermore, the location and abundance of old-growth attributes enable the provision of a
range of ecosystem services (ES) that includes carbon storage and sequestration (Luyssaert et
al. 2008, Maxwell et al. 2019), water provision (Bithell and Brasington 2009), indigenous
cultural values, and the maintenance of human health (Wirth 2009, Watson et al. 2018).
Therefore, the value of retaining old-growth forests in the landscape goes beyond old-growth
conservation per se. It may offer the opportunity to simultaneously maintain landscape
biodiversity and ESs essential for people’s wellbeing.
Old-growth forests have been historically valued as wildlife habitat (Mosseler et al.
2003c). Strategies to promote the retention of old-growth forests in the landscape do exist,
such as the Old-growth Management areas (OGMAs) in British Columbia, Canada
(Arsenault 2003, Gillis et al. 2003, Environmental Law Centre 2013). Notwithstanding, the
selection of such areas is a difficult task due to the lack of a standard definition for what
constitutes an old-growth forest (Hilbert and Wiensczyk 2007). Old-growth forest types vary
1

in terms of stand age, disturbance frequency, anthropogenic interference, and abundance of
specific forest structures such as the abundance of large and old tree and coarse woody debris
(Mosseler et al. 2003c, Spies 2004, Bauhus et al. 2009). Moreover, the management of
landscapes often involves conflicting objectives such as timber to one group of people and
recreation to another (Ninan and Inoue 2014). Also, the relationships between ESs are
complex, and our knowledge of their response to management strategies is limited (Kremen
and Ostfeld 2005). As human populations continue to grow, demand for ecosystem services
is also increasing (MA 2005, United Nations 2018), and so are the conflicts between multiple
objectives.
The identification of strategies for the retention of old-growth values and multiple
ecosystem services, while leaving opportunities for timber harvesting, is a complex spatial
optimization problem (Schröter and Remme 2016, Snäll et al. 2016). Two steps are required
to address this problem. First, it is necessary to define and locate old-growth forests and ESs
in the landscape. Second, we need to apply a thoughtful strategy to allow old-growth
conservation while leaving opportunities for timber harvesting in the landscape. A systematic
conservation-planning tool can be used to identify management strategies that can cope with
multiple objectives landscapes. In the following sections, I outline the background and
motivation for this thesis.
A general review divided into six sections provides information to understand the
problem and identify a framework for its solution. The first section, “Old-growth” highlights
the diversity of old-growth forest types and the multiple definitions available. A general
review of the ESs literature is provided in section two, “Ecosystem Services.” Section three,
“Forest state and ecosystem services,” touches on the relationship between forest succession
and the provision of multiple ESs. Section four, “Airborne Laser Scanning (ALS),”
2

summarizes some of the literature on the use of Light Detection and Ranging (LiDAR) to
evaluate forest attributes and succession, as well as its potential for ESs estimation. The fifth
section, “ Systematic Conservation Planning,” introduces the systematic conservation
planning theory and spatial prioritization tools developed to solve conservation problems.
Finally, the sixth section, “Thesis Objective,” outlines the research gaps identified in the
previous section and the objectives and structure of this thesis.

1.2. Old-growth
In previous years, the public concerns about the decline of older forests have pushed
policymakers to place great importance in the old-growth forest conservation (Mosseler et al.
2003c). However, the public concept of old-growth is usually of tall and large trees, whereas
old-growth forests are much more complex (Wirth et al. 2009b). Spies (2004) defines an oldgrowth forest as mature or senescent forest, associated with specific structures (ex. snags,
coarse woody debris) and processes (ex. gap dynamic, natural regeneration). These forests
differ in character and degree depending on the forest region, and are not necessarily primary
forests (Spies 2004). Old-growth types vary in terms of longevity of dominant species, return
period of natural disturbances, human intervention, shade tolerance, and presence of specific
structures such as coarse woody debris. Therefore, multiple regional specific definitions of
old-growth are more desirable than one standard definition of old-growth, given the diversity
of forests (Spies 2004).
Forest composition, stand age, diameter, snags and vertical diversity are attributes
commonly evaluated by managers to define old-growth (McElhinny et al. 2006a, Bauhus et
al. 2009), but which levels of those attributes are considered old-growth? Age is commonly
used to define old-growth, a useful proxy in an even-aged forest, but less valuable in multi3

age forests (Spies 2004). For example, according to (MFLNRORD 2003), BC’s coastal
forests are considered old growth if trees are more than 250 years old. In the province’s
interior, where the longevity of trees are shorter and disturbances more frequent than in the
coastal forests, old growth is defined as more than 120 years of age for forests dominated by
lodgepole pine or broadleaf species. As well, other forests such as Englemann spruce, white
spruce and Interior Douglas-fir reach old-growth status with 140 years.
Structural complexity is also another characteristic utilized to define the old-growth
forest. Compared with old forests, young natural forests or intensively managed forest
plantations have less complex structures. Old-growth attributes are dynamic, which means
that one stand classified as old-growth may not display old-growth attributes after
disturbances (Spies 2004). On the other hand, young stands can develop old-growth attributes
over time (Mosseler et al. 2003c). The authors recognize that for regions where fire-rotation
and other disturbances occur over exceptionally long periods, it is easier to designate an oldgrowth stand, whereas areas more affected by disturbances are less clearly defined. Based on
the results of past efforts, it is clear that there are challenges to providing a universal
definition of old-growth forests.
Policymakers and managers need specific definitions of old-growth to better track and
map old-growth forests within their management areas, and to avoid conflicts with other
types of management. Multiple local definitions have been developed (Table 1.1), where the
most common are definitions based on the abundance of some forest structures (Wirth et al.
2009b, Bauhus et al. 2009). Spies (2004) points out the importance of defining and mapping
old-growth forest for each forest type using measurable structural features and biophysical
site conditions, considering the continuous nature of forest structure development. The
combination of the structural attributes, species composition, and ESs to provide an index for
4

a particular forest type of ecological region could clarify the old-growth identification, and
consequently, conservation (Mosseler et al. 2003c). To determine which variables are most
relevant for an old-growth index requires a strong understanding of the ecology of each forest
type. For example, the index could also consider animal associations since many of the oldgrowth structures have a close association with biodiversity and species habitat (Mosseler et
al. 2003b, 2003c, Spies 2004, Bauhus et al. 2009).
Different researchers have endorsed the development and use of an index that could
track old-growth values in the landscape, rather than only using stand age classes (Mosseler
et al. 2003c, Spies 2004, Hilbert and Wiensczyk 2007). Compared with old forests, young
natural forests or intensively managed forest plantations have a simpler structure (Spies 2004,
McElhinny et al. 2006a). Thus, the abundance of old-growth attributes (e.g. large trees,
snags, and accumulated woody debris), which contribute to the structural complexity in oldgrowth forest, may be used as a proxy for old-growth forest mapping. Much work has been
conducted using traditional field-based measurement of forest attributes to classify forest
succession and assess the quality of old-growth forests (McElhinny et al. 2005, 2006, Bauhus
et al. 2009) (See more in Table 1.1). Even though field-based methods are essential for most
forest studies, they are less applicable for landscape-scale evaluation.
Broad scale definitions of forest succession using remote sensed imagery have also been
developed (Table 1.1). For example, Hansen et al. (2019) utilized forest structural complexity
as a proxy for forest quality and capacity to support biodiversity and provide ESs. The study
was undertaken for tropical forests in a broader scale utilizing optical sensors, and validated
with airborne LiDAR. Other similar studies are listed in Table 1.1. However, none of these
studies have focused on identifying old-growth per se. Moreover, while it is interesting to
have a broad scale definition and mapping of old-growth, it might not be desirable in a
5

ecological perspective. As discussed by many authors (Mosseler et al. 2003c, Hilbert and
Wiensczyk 2007, Wirth et al. 2009b, Bauhus et al. 2009), old-growth conditions are
particular to local environment conditions. The conditions for a forest to became old-growth
include, in addition to time, the presence of multiple forest attributes that may not be capture
by optical sensors (Lefsky et al. 2002). In addition, such broad definitions can underestimate
the value of old-growth forests types in landscapes more prone to natural disturbances where
old-growth attributes are less prominent than in coastal temperate and tropical forests (Spies
et al. 2006). Thus, multiple definitions of old-growth based on local conditions might be
more desirable, and LiDAR offers the potential to accurately measure these conditions and
aid in the mapping of old-growth values in the landscape.

6

Table 1.1 Method utilized to identify and map old-growth forests.

Method
Field metrics

Description
Classification of forest succession in a stand level based on age thresholds defined for
each Biogeoclimatic (BEC) zone and fire return interval, utilizing outdated forest cover
maps.

Reference
(MFLNRORD 1995)

Use a series of forest attributes associated with old-growth forest to create an old-growth
index succession and evaluate old-growth forest reserves in different BEC zones.

(Braumandl and Holt
2000, Holt 2000, Holt
et al. 2001, 2002,
DeLong et al. 2004)
Reviewed and listed a series of forest attributes strongly associated with old-growth forest (Bauhus et al. 2009)
to identify silvicultural approaches that could promote old-growth forests.
Evaluated different plot sizes to determine the minimum plot that still captures oldgrowth indicators (e.g. number of living trees, trees with DBH >50cm, dead wood
volume, etc).

(Lombardi et al. 2015)

Review of old-growth static and dynamic attributes and use of cohort basal area ratio
(understory cohort/post-disturbance cohort) as a proxy for old-growth forests in boreal
forest, simultaneously addressing the dynamic nature of forest.

(Kneeshaw and
Burton 1998,
Kneeshaw and
Gauthier 2003)
(Chen and Bradshaw
1999, Hao et al. 2007)

Mapping of individual stems and their respective features (e.g. height, crown area) in
temperate old-growth forests to study forest structure and dynamics.

Optical
Sensors

Developed a stand-scale index of structural complexity combining a core set of forest
structural attributes.
Forest succession model (ZELIG) and a canopy reflectance model (GORT) were applied
to compare with forest succession from Landsat TM and test the potential of remote
sensing on mapping successional stages.

(McElhinny et al.
2006a)
(Song et al. 2007)

7

Landsat ETM+ combined with ecological land unit classifications.
Landsat TM image resampled to 25m cell size and 106 ground reference stands of tree
density, basal area.

LiDAR

(Bergen and Dronova
2007)
(Cohen et al. 1995)

Mapping of structural stage classes with Landsat TM data through ISODATA analysist
technique.

(Miller et al. 2003)

Spatial manifestation of forest succession in optical imagery (Landsat TM) through three
types of model.

(Song and Woodcock
2002)

Use of Lidar delivered metrics (e.g. height percentiles and statistics, % of vegetation
returns, % of first returns, and etc.) with Random Forests statistical analysis to identify
seven stages of forest succession.

(Falkowski et al.
2009)

Documented increasing vertical structure complexity along five development stages in (Kane et al. 2010b)
western coastal forests with five field, six LiDAR metrics, and their combination.
Use of LiDAR-delivered tree height variance to distinguish between single-story (young (Zimble et al. 2003)
forests) and multistory vertical structural classes (old forests).
Use of two principal components (PCA) of the Integration of airborne LiDAR (canopy (Hill and Thomson
height model) and spectral data (12 wavebands of HyMap) to perform an unsupervised 2005)
classification of forest classes.
Estimated stand age across 158 plots in managed Boreal forest with forest structures and
site attributes delivered from LiDAR.

(Racine et al. 2014)

8

1.3. Ecosystem Services
Costanza et al. (1997) describe ESs as the goods and services derived from ecosystem
functions that benefit human well-being directly or indirectly. MA (2005) divided ESs into
four main categories: provisioning services (e.g food, water, and raw materials), regulating
services (e.g. erosion control, carbon sequestration, water and climate regulation), cultural
(e.g. aesthetics values, the spiritual experiences, scientific endeavors, recreation
opportunities), habitat services (life cycle of all materials and the genetic reservoirs).
Forested ecosystems are responsible for the provision of a large amount of these services
(Pearce 2001, Watson et al. 2018). Much discussion has been given to the theory behind ESs
(Costanza et al. 1997, Kremen and Ostfeld 2005, Kremen et al. 2007, Isbell et al. 2011,
Crossman et al. 2013), and methods to measure and evaluate them (MA 2005, Balvanera et
al. 2006, Nelson et al. 2009, Polasky et al. 2011, Kareiva 2011, Schirpke et al. 2016). As
human populations continue to grow, demand for ecosystem services is also increasing (MA
2005, United Nations 2018). The increasing demand for ecosystem services may increase
biodiversity loss in various ecosystems and the services derived from them if the landscape is
managed by one or few services (Coomes et al. 2008, Lindenmayer et al. 2012, Gaston et al.
2013). Despite the gravity of the issue, the management of ecosystem services is rarely
incorporated into decision-making.
As discussed in the Millennium Ecosystem Assessment (MA 2005), the focus on timber,
agriculture and other provisioning services has driven changes in landscapes, often
simplifying their structure and reducing their species diversity (Coomes et al. 2008, Polasky
et al. 2011). In general, these working landscapes are typified by high production yields and
low per unit production costs, yet provide reduced stocks and flows of other non-target ESs
9

(MA 2005). Managing for multiple services is an alternative to reduce these tradeoffs and
increase synergies between ESs. Also prioritizing regulating services (Bennett et al. 2009)
and biodiversity (Polasky et al. 2008) has shown increasing provision of multiple ESs. For
example, the Catskill Mountains watershed experienced improvements in the provision of
ESs such as flood control, wildlife habitat, and recreation value after being managed to
improve water quality (Lubchenco 1998). Wendland et al. (2010) identified areas in
Madagascar, which protect biodiversity and offer multiple ES, for projects involving
payment for ESs. Many researchers have also modeled conservation plans for multiple ESs
(Naidoo and Ricketts 2006, Venter et al. 2009b, 2009a, Nelson et al. 2009, Schirpke et al.
2016). According to the works above, conserving for biodiversity, especially in forested
ecosystems, seems well related to the provision of many other ESs in the landscape.

1.4. Forest state and ecosystem services
Forest loss is often associated to a subsequent decline in ESs provision, such as declining
climate regulation, carbon storage, water quality and quantity, disease control, recreation
opportunities (Pearce 2001). In places, forests are increasing due to forest regeneration in
abandoned croplands and afforestation for commercial or environmental purposes, mostly in
the northern hemisphere (FAO 2016). However, a degraded forest ecosystem may only
partially recover its potential for biodiversity and ESs provision after restoration (Chazdon
2008). Also, primary forests and old-growth forests provide many of these ESs (MA 2005,
FAO 2016, Watson et al. 2018), yet are under high pressure from logging, agriculture and
urban expansion (Pearce 2001, Mosseler et al. 2003c). Thus, it is crucial to understand the
behaviour of ESs in different forest states.

10

The different levels of forest structures development provide different levels of ESs.
Early-successional forest ecosystems, developed post-disturbances, are biodiverse and
abundant in biological legacies such as individuals resistant to the disturbance and organic
matter resultant from the fallen trees. It can also provide an opportunity for recharging the
nutrient pool of soil through the mineralization of organic material (Swanson et al. 2011).
Forest regeneration and forest plantations established for commercial and restoration
purposes can improve ecosystem services and enhance biodiversity conservation (Chazdon
2008), even not reaching original levels (Hobbs et al. 2006, Chazdon 2008).The restoration
of watershed successfully restore water regulation service (Chichilnisky and Heal 1986) and
consequently promoted a series of associated ESs (Lubchenco 1998). Despite the value of
early forest successions and second growth forest, it is on the primary and old growth forest
that the provision of multiple services seems to peak.
Various studies highlighted the importance of old-growth forest for the provision of
ESs (e.g. Mosseler et al. 2003, Luyssaert et al. 2008, Wirth 2009, Keenan et al. 2015, Watson
et al. 2018). Old-growth forests are not only massive carbon storages but also carbon sinks
(Luyssaert et al. 2008), as these forests provide an environment for increasing the
accumulation of carbon into the soil (Zhou et al. 2006). If disturbed, the carbon stored both in
the biomass above ground and in the soil that risks being released back into the atmosphere is
much higher than first expected (Maxwell et al. 2019). Other benefits of these mature and
old-growth forests are the high biodiversity of trees and animals (Kremen et al. 2007, Isbell
et al. 2011). Old-growth forest populations can serve as gene pool and seed sources for
forests to adapt to future environmental conditions, as long as they are large enough to avoid
inbreeding and genetic drift (Mosseler et al. 2003b). In addition to adaptive capacity, genetic
resources of old-growth forests might also be the source of new medical discoveries
11

(Costanza et al. 1997, Keenan et al. 2015, FAO 2016). Isbell et al. (2011) demonstrated the
importance of high plant diversity to ESs provision, where the extinction of species may
mean the reduction of ESs. For example, many pollinators and pest controllers, crucial for
agriculture and other services, inhabit these forests (Kremen et al. 2007). The works cited
above indicated direct or indirect relations between forest state and ESs. However, ESs
modelling requires improvements to increase confidence in the predictions of their amount
and location in the landscape.
Improving the resolution, extent and quality of ESs estimates has the potential to
improve evaluations of the link between forest state and ESs. Some authors have already
pointed out this need for increased reliability on ESs estimations (Polasky et al. 2011, Naidoo
et al. 2008). For example, Naidoo et al. (2008) observed that areas that maximize biodiversity
did not provide more ESs than randomly selected places. However, when utilizing data with
higher resolution, they identified areas with simultaneous high biodiversity and ESs. Thus,
different resolutions can lead to complete different conclusions. The lack of high-resolution
data is also an issue for old-growth forest definition and mapping (Holt 2000, Mosseler et al.
2003c, Hilbert and Wiensczyk 2007). Airborne LiDAR may offer the resolution needed for
an accurate assessment of old-growth and ESs.

1.5. Airborne Laser Scanning (ALS):
The rapid emergence of remote sensing technologies has enabled researchers to, for
instance, identify forest succession in broad scales (Hermosilla et al. 2015) (See also Table
1.1). Nevertheless, remotely sensed images are often two-dimensional (x and y), which
cannot fully represent the nuances of the three-dimensional (3D) structures present in old-

12

growth forests (Lefsky et al. 2002). On the other hand, ALS has been proven to be an
effective technique to estimate 3D forest attributes.
ALS is active remote sensing, which emits high pulse frequencies in the near-infrared
wavelength (e.g. 1064 nm), and further captures the intensity of returns reflected by different
surfaces (Goodwin et al. 2006). The surface location is acquired from an onboard differential
global positioning system (dGPS) (Hofton et al. 2000, Gaveau and Hill 2003). This
technology performs accurate mapping of terrain and the 3-dimensional attributes of
vegetation (Lefsky et al. 1999, Hyyppä et al. 2008, Bater et al. 2009). Starting with civil
engineering applications (Meng et al. 2010), airborne LiDAR has been rapidly incorporated
in forest management (Reutebuch et al. 2005, Wulder et al. 2008), wildlife habitat assessment
(Hyde et al. 2006, Martinuzzi et al. 2009), evaluating of the effect of pests (Bright et al.
2013), and other applications (see Table 1.1). For forests, ALS has been widely used to
measure tree height and forest biomass (Næsset 2002, Hyde et al. 2006), along with a variety
of other old-growth forest attributes (Table 1.2). Furthermore, forest attributes can be
assessed for an individual tree and area-base approach (ABP) (Reutebuch et al. 2005). These
abilities make airborne LiDAR an effective way of mapping old-growth forests and a useful
means for the estimation of ESs on a landscape scale.
Although the use of ALS for the assessment of multiple ESs is still in its infancy (Ayanu
et al. 2012), there are examples of the successful use of this remote sensing technique to
estimate a wide range of services. The most common ESs estimated with ALS are timber
volume (Reutebuch et al. 2005, Wulder et al. 2008) and carbon storage (Mascaro et al. 2011).
Researchers have also used ALS to estimate water-related services, such as the potential
water storage capacity (Lang and McCarty 2009) and understory inundation (Lane and
D’Amico 2010). Moreover, Müller and Vierling (2014) discussed the use of ALS for the
13

assessment of biodiversity, while Simonson et al. (2014) reviewed the application of ALS on
the estimation of animal diversity. Even cultural services have been recently estimated with
ALS (Dade 2018, Van Berkel et al. 2018). These studies highlighted the potential of ALS to
estimate individual ESs. However, the potential of ALS for the management of multiple ESs
has been rarely tested.
ALS is a remote sensing technique that proved to be an efficient means for estimating
forest values and a wide range of ESs in a landscape-scale. The use of ALS can, therefore,
allow for more transparent management of forested landscapes for multiple ESs. However,
ensuring that forest management actions retain multiple ecosystem services while providing
opportunities for timber extraction is a complex spatial optimization problem. The next
section offers an introduction to Systematic Conservation Planning and the use of spatial
prioritization tools as a means to design conservation areas.

14

Table 1.2 Airborne LiDAR delivered metrics for old-growth forest attributes with area-based approach (ABA) and individual tree detection
(ITD).

Old-Growth
Attribute*

Lidar Estimators

Scale

Reference

Tree height

Treetops were detected with the highest return of point cloud from each tree and
compared with high precision field measured of treetops.

ITD

(Andersen et al. 2006)

Estimation of plot based height measurements (e.g. average, maximum, standard
deviation) with LiDAR delivered metrics, indicating correlation close to 1:1.

ABA

(Hopkinson et al. 2006,
Goodwin et al. 2006)

Basal area

Random Forest models were developed with LiDAR delivered metrics with and
without intensity metrics to predict total, live and dead basal area.

ABA

(Bright et al. 2013)

Number of
dead standing
trees (snags)

Filtering algorithm based on density and intensity statistics to remove points
associated with living trees, followed by an individual tree detection procedure.

ITD

(Wing et al. 2015)

ABA

(Bater et al. 2009)

ABA

(Martinuzzi et al. 2009)

Correlation of height metrics with field observed frequency of snags to estimate
snag frequency for the landscape.
Median absolute deviation of height was associated with the abundance of snags
in different DBH classes, as well as different other canopy and topography
metrics, using Random forest algorithm

15

Structural
Complexity

Biomass

Understory
Density

Intensity, density and height statistics were used to estimate basal area (BA) of
live, dead trees, and total BA through Random Forest models

ABA

(Bright et al. 2013)

Canopy volume profile estimates and leaf area index (LAI)

ABA

(Lefsky et al. 1999,
Coops et al. 2007)

Complexity of vertical forest structure was estimated with LiDAR derived height
variance

ABA

(Zimble et al. 2003)

Indicates that the 95th height percentile, rumple (ratio of canopy outer surface
area to ground surface area), and canopy density had the strongest correlation
with field measured stand complexity.

ABA

(Kane et al. 2010b,
2010a)

LiDAR height percentile (h80) and crown width (CW) measurement were the
best metrics for aboveground biomass (AGB) estimates using a multilinear model

ITD

(Wan-Mohd-Jaafar et
al. 2017)

Quantiles and full returns against field measurement, and other simple LiDAR
metrics were tested against field estimates with correlation analysis and
multilinear models

ABA

First returns in a specific range of intensity in lower strata of the canopy was
utilized to estimate live understory distribution

ABA

Proportion of ground returns, vegetation return between 1 and 2.5 m in height
and percent slope times cosine of aspect were fed to a random forest model to
predict presence and absence of understory vegetation

ABA

(Næsset 2011, Ahmed
et al. 2013, Næsset et
al. 2013)
(Koukoulas and
Blackburn 2004a,
Vepakomma et al.
2008, Wing et al. 2012,
White et al. 2018)
(Martinuzzi et al. 2009)
16

LiDAR delivered metrics were proven more accurate predictors of coarse woody
debris (CWD) than field measurement of living trees, and indicated as important
auxiliaries in the prediction of CWD in the landscape

Canopy Gap

ABA

(Seielstad and Queen
2003, Pesonen et al.
2008, 2009)

Canopy gap was measured based on a canopy height mode (CHM) with a height
threshold measured during field (4-5m) and gap area of 5m2. Slope from CHM
was also an important feature to map canopy gaps.

ABA

(Koukoulas and
Blackburn 2004b,
Vepakomma et al.
2008)

Applied fixed and variable height thresholds to a 1m resolution CHM to detect
gaps, further filtered by area. Gap areas <5m2 and >2ha were excluded.

ABA

(White et al. 2018)

17

1.6. Systematic Conservation Planning
Systematic conservation planning was developed to ensure the efficient use of limited
resources for conservation, aiming to direct conservation investments to priority areas
(Margules and Pressey 2000, Wilson et al. 2009). The most straightforward conservation
problem formulation involves the binary decision of whether to select a planning unit for
prioritization or not. Many conservation planning problems are incredibly complex,
involving thousands of planning units and a myriad of features to be prioritized (e.g. species'
habitats and ecosystem services). Chan et al. (2006) found that systematic planning provides
a framework useful for identifying valuable synergies and potential trade-offs between
conservation for biodiversity and ESs. This framework can, thus, be applied to identify
priority areas for the conservation of old-growth forests, while simultaneously targeting
multiple ESs. However, in some forested landscape, setting aside old-growth forest for oldgrowth conservation can conflict with other management actions. Ensuring that management
actions retain multiple landscape values to target levels while meeting the objectives of
different stakeholders is a complex spatial optimization problem (Schröter and Remme 2016,
Snäll et al. 2016).
The minimum set problem in conservation prioritization is designed to secure a target
level of each conservation feature with the smallest possible set of areas, often referred to as
planning units (Wilson et al. 2009). There are two main approaches to solving the minimum
set problem, the integer linear programming (ILP) and heuristics, most commonly simulated
annealing (Beyer et al. 2016). Marxan is the most common conservation planning software
that uses simulated annealing (Ball et al. 2009, Watts et al. 2009). To promote connectivity
and management effectiveness, Marxan includes a mechanism to control the degree of
18

aggregation among selected units. Instead of enforcing hard constraints, Marxan utilizes a
"shortfall penalty" function, which assumes that even configurations of planning units that do
not meet all of the targets may still have value. The "shortfall penalty" provides a way of
finding reasonable solutions where all targets cannot be met (Beyer et al. 2016). The main
concern with the simulated annealing approach is that there is no measure of how far from
optimal the solutions is (Underhill 1994, Önal and Briers 2002). While simulated annealing
approaches stochastically explore the decision variables providing a range of solutions,
problems solved with ILP have only one optimum solution or a solution with a specific
distance from optimum.
Although very similar to Marxan, "PrioritizR" utilizes integer linear programming (ILP)
and is implemented in an "R package.” PrioritizR provides a flexible interface for building
and solving conservation planning problems (Hanson et al. 2019). PrioritizR also supports a
broad range of functions and modifiers to attend to the particularities of a conservation
planning exercise. Once built, PrioritizR problems are solved using a solver algorithm, such
as Gurobi (V.7.0) and “SYMPHONY in R” (Harter et al. 2017). ILP is recommended over
simulated annealing whenever this is possible because it offers higher quality solutions in
less processing time over a wide range of problem sizes and for both linear and quadratic
models (Beyer et al. 2016). It also facilitates the development of trade-off curves and
comparison between multiple scenarios.
The systematic conservation planning solutions are flexible and defensible, allowing the
critical review of multiple management decisions (Margules and Pressey 2000). "PrioritizR"
can simulate the allocation of different management actions (e.g. harvesting and old-growth
protection), allowing practitioners to identify solutions that meet multiple objectives.

19

Therefore, "PrioritzR" will be utilized as a conservation planning tool for the identification of
priority areas for old-growth and multiple ESs in Chinook Community Forest.

1.7. Thesis Objectives
The location and abundance of "Old-growth" forest attributes enable the provision of a
range of ecosystem services (ESs) (Hendrickson 2003, MA 2005, McElhinny et al. 2005,
Isbell et al. 2011). These forests are also the habitat of different animal species (Mosseler et
al. 2003a). Therefore, it might be possible that identifying and retaining old-growth forests
can simultaneously retain essential ESs in the landscape. Nevertheless, there is a great
diversity of old-growth forests, as well as the definitions and approaches to classify them,
which makes the selection of such areas a difficult task (Hilbert and Wiensczyk 2007). The
primary strategy utilized to define and locate old-growth forests in the landscape is the "age"
classification (Environmental Law Centre 2013). Although "age" is an important metric,
"age" by itself may fail to capture some old-growth features (Braumandl and Holt 2000).
This thesis addresses two main objectives to approach these knowledge gaps.

1. Develop an old-growth index through the use of LiDAR to enable the definition
and location of old-growth forests in a landscape scale.
2. Evaluate the opportunity for multiple ESs provision in a complex landscape
managed manly for timber through the spatial prioritization of old-growth values.

This thesis addresses the two objectives in the Chinook Community Forest (CCF),
located in Burns Lake, British Columbia, CA. A community forest is an area-based tenure for
"any forestry operation managed by a local government, community group, or First Nation
20

for the benefit of the entire community" (MFLNRORD 2017a). The community forest is
located within the Skeena region and overlaps with six First Nations' and Bands' territories:
Cheslatta Carrier Nation; Lake Babine Nation; Burns Lake Band; Wet 'suwet 'en First
Nation; Skin Tyee Nation; and Nee Tahi Buhn Band. The diversity of stakeholders adds
greater complexity to the management of community forest and makes it a strong case study.
While the main objective of the community forest is to manage the landscape for timber
extraction, it should not affect the provision of other ESs for the local communities. The Oldgrowth Management areas (OGMAs) might be a strategy to maintain essential ESs in the
landscape while providing opportunities for timber extraction.
Two data chapters follow this general review. In the first data chapter (Chapter 2), I
tackle the first objective utilizing a combination of traditional field-based definitions of oldgrowth reviewed in this chapter (Chapter 1) with airborne LiDAR. For the second data
chapter (Chapter 3), I utilize old-growth and ESs layers as inputs for a spatial prioritization
tool, "PrioritizR." Different prioritization scenarios were tested to identify trade-off and
synergies between old-growth and ESs and the possibility to design OGMAs for multiple
ESs.

21

2. DEFINING OLD-GROWTH FORESTS WITH AIRBORNE LiDAR

Abstract: In this work, we developed a structure-based old-growth forest index tailored to
map old-growth value in managed landscapes. Forests in their later stages of development
attain attributes that support biodiversity and provide a variety of benefits to human
populations. Despite their irreplaceable value, old-growth forests are declining worldwide
due to anthropogenic pressures. A definition of old-growth is needed to facilitate mapping
and delineation of old-growth in the landscape and old-growth reserves. LiDAR-derived
metrics were utilized with a random forest (RF) modeling framework to develop an oldgrowth index across the landscape. Using this old-growth index, we found that forests with
“Very-high” old-growth values cover 14.7% of the study area and only 24.9% of the current
designated old-growth management areas (OGMAs). However, the set-aside forest with
“Very-high” old-growth value is mostly fragmented between the OGMAs, as only one
OGMA has more than 50% of its area covered by forests with “Very-high” old-growth value.
This research brings light to old-growth, and OGMAs’ definition and their assessment
through the use of fine-scale remotely sensed data, LiDAR. While the index developed is
specific to the study site, the framework, however, is generic enough to be adapted to other
forest types and ecosystems. More importantly, the identification of the amount and location
of old-growth forests over the landscape can aid in the conservation of this rare resource and
its services.
Keywords: Community Forests, Conservation, Old-growth forests, remote sensing

22

2.1. Introduction:
“Old-growth” is a term associated with forests in the advanced development stage
(Mosseler et al. 2003c, Spies 2004, Hilbert and Wiensczyk 2007). Forests with these
characteristics are rare, and in decline worldwide, despite the multiple benefits they provide
for the maintenance of human wellbeing (Wirth 2009, Bithell and Brasington 2009, Watson
et al. 2016, 2018, Maxwell et al. 2019). Strategies to promote the retention of old-growth
forests are often incorporated into landscape level forest planning (Arsenault 2003, Gillis et
al. 2003, Environmental Law Centre 2013). Yet the identification of old-growth is a difficult
task due to the lack of a standard definition for what constitutes an old-growth forest (Hilbert
and Wiensczyk 2007). Old-growth forest types vary in terms of longevity of dominant
species, return period of natural disturbances, human intervention, shade tolerance, and
abundance of specific structures such as the number of large trees, snags, accumulated woody
debris (Mosseler et al. 2003b, 2003c, Spies 2004, Bauhus et al. 2009). For example, coastal
Douglas-fir forests may grow for centuries without disturbances, whereas Ponderosa pine
forest is frequently disturbed by fires (Spies 2004, Spies et al. 2006). These ecological
differences across forest types pose a significant methodological challenge to the
characterization of these forests only based on disturbance frequency.
There exists multiple definitions and approaches to define and locate old-growth
forests (Wirth et al. 2009b). Age is a proxy that has commonly been used to define and locate
old-growth forests. For example, according to MFLNRORD (2003), BC’s coastal forests are
considered old-growth if trees are more than 250 years old. For forests dominated by
lodgepole pine or broadleaf species in the northern interior, old-growth are forests with more
than 120 years of age. In these landscapes, the longevity of trees tends to be shorter, and
disturbances more frequent. Although age is a useful proxy, its measurement with traditional
23

field methods is costly (Racine et al. 2014). As well, important structural elements of oldgrowth can be omitted using only an age threshold (Arsenault 2003, Gillis et al. 2003, Holt et
al. 2008). More importantly, forest cover maps currently used to locate old-growth forests
often do not accurately reflect the age class distribution in the landscape (Holt et al. 2008).
This inaccuracy can lead to management that underrepresents old-growth forest in the
landscape, or incorrectly identifies forests as being old-growth even though they do not
exhibit the desired characteristics. As a result, in many areas, it may be prudent to move
away from a simple age threshold for old-growth definition towards a more ecologically
based representation of forest structures.
Different authors have pointed out the need to develop an index that could be used to
track old-growth forest in the landscape, rather than only using stand age (Mosseler et al.
2003c, Spies 2004, Hilbert and Wiensczyk 2007). Compared with old forests, young natural
forests or intensively managed forest plantations have a simpler structure (Spies 2004,
McElhinny et al. 2006a). Thus, the abundance of old-growth attributes (e.g. large trees,
snags, and accumulated woody debris), which contributes to the structural complexity in the
old-growth forest, can be used as a proxy for old-growth forest mapping (Mosseler et al.
2003a, 2003c, Bauhus et al. 2009). A myriad of work has been conducted using a traditional
field-based measurement of forest attributes to classify forest succession and assess the
quality of old-growth forests (McElhinny et al. 2006a) (See also Table 1.1). Even though
field-based methods are essential for most forest studies, they are less applicable for
landscape-scale evaluation.
The rapid emergence of new technologies has allowed the development of highly
precise measures of forest condition across broad areas (Hansen et al. 2019), which exceed
what has been possible based on traditional field-based and areal interpretation methods
24

(Cohen et al. 1995, Song and Woodcock 2002, Hyyppä et al. 2008, Kane et al. 2010b).
Typical applications of passive or active optical sensors have proven to be useful for a variety
of ecological studies, enabling researchers to, for instance, identify forest succession in broad
scales (Song and Woodcock 2002). Nevertheless, remotely sensed images are often
constrained to two-dimensional (x and y) interpretation, which cannot fully represent the
nuances of the three-dimensional (3D) structures present in old-growth forests (Lefsky et al.
2002). On the other hand, airborne LiDAR has been proven to be an effective technique to
estimate 3D forest attributes, particularly for height and biomass (Næsset and Økland 2002,
Hyde et al. 2006). Starting with civil engineering applications (Meng et al. 2010), airborne
LiDAR has been rapidly incorporated in to forest management (Reutebuch et al. 2005,
Wulder et al. 2008), wildlife habitat assessment (Hyde et al. 2006, Martinuzzi et al. 2009),
evaluation the effect of pests (Bright et al. 2013), and other applications. In addition to height
and biomass, a variety of other old-growth forest attributes can be accurately estimated with
airborne LiDAR (White et al. 2018) (See also Table 1.2). Thus, airborne LiDAR has the
potential to be an effective way of generating an old-growth index to effectively map oldgrowth forests.
The definition and mapping of old-growth forests with measurable structural and
biophysical features, considering the continuous nature of forest structure, is imperative for
their conservation and maintenance in managed landscapes. Throughout the years, different
authors have attempted to map forest succession in the landscape (Table 1.1). However, few
have used old-growth attributes to create an index for old-growth value, and none has done it
in landscape scale. In this work, we aim to: (1) develop an old-growth index based on forest
structures measured with traditional field methods; (2) extrapolate the old-growth index to
the landscape utilizing LiDAR-derived metrics; and (3) evaluate the amount and quality of
25

old-growth forest for the study site, simultaneously evaluating the set-aside old-growth
forests currently present in the landscape.

2.2. Material and Method:
2.2.1. Study Area:
A community forest is an area based tenure meant for “any forestry operation
managed by a local government, community group, or First Nation for the benefit of the
entire community”(MFLNRORD 2017a). The Chinook Community Forest (CCF), located
within the Skeena region, overlaps with six First Nations’ and Bands’ territories: Cheslatta
Carrier Nation; Lake Babine Nation; Burns Lake Band; Wet‘suwet‘en First Nation; Skin
Tyee Nation; and Nee Tahi Buhn Band. The forests in the study site are categorized into two
biogeoclimatic zones (BEC), the Englemann Spruce – Subalpine Fir (ESSF) and Sub-Boreal
Spruce (SBS). The tenure area for CCF operations is approximately 123,695 ha, currently
encompassing around 40 set-aside old-growth forests or old-growth management areas
(OGMAs). The total set-aside old-growth forests area is ~ 8,618 ha, 6.96% of the tenure area.
The CCF area has five different management blocks (Figure 2.1). Figures in the results and
discussion depict only block 04 to facilitate visualization. However, the analysis and numeric
results are reported for the whole land base.

26

Figure 2.1 Location of Chinook community forest tenure areas and distribution of Old-growth
management areas (OGMAs).

2.2.2. Data:
Empirical measurement of forest composition and structural attributes were collected
from 99 plots of 10 m radius (Table 2.1; 5222 trees sampled). All trees greater than 4 cm
DBH were measured such that forest structure in disturbed and young forests was recorded,
increasing correlation with ALS metrics (Keränen et al. 2015). In addition, trees with a
diameter smaller than 4cm were tallied to obtain the density of small trees and seedlings. The
inventory followed the Change Monitoring Inventory (CMI) procedures used by B.C.
Ministry (MFLNRORD 2017b). High precision GPS was used to obtain two measurements
of ± 2m accuracy from the plot center. Geographic information system (GIS) exercises

27

incorporate Airborne LiDAR and ground survey into old-growth maps and evaluate oldgrowth forests.

2.2.3. Old-growth attributes:
From the list of thirteen old-growth attributes indicated by Bauhus et al. (2009), I was
able to estimate eleven of them for the Chinook Community Forest (Table 2.1). In addition to
that, I also included maximum tree height as it has a strong correlation with age in oldgrowth assessment (Kneeshaw and Burton 1998, Hao et al. 2007). Tree diversity was also
included, as old-growth is expected to have higher biodiversity (Mosseler et al. 2003b,
McElhinny et al. 2006b). However, we only have information on the diversity of trees. Here I
utilized these attributes as basis for forest classification and the development of an oldgrowth index. Aboveground biomass estimates were based only on the value of DBH in the
form of an exponential curve developed by (Jenkins et al. 2003). DBH and height are the
base for volume estimates calculated using volume equations developed by Penner et al.
(1997) and Standish et al. (1985). A description of the equations and associated parameters
utilized in this study are included in Appendix 6.2.

28

Table 2.1 Field measurement of Old-growth attributes.

Old-growth Attribute

Mean

Range

1

Large trees density (number of trees - dbh>40cm/ha)

14.47

0 - 222.82

2

Presence of regeneration (number of trees <1.3m/ha);

4,564

0 – 37,179

3

Biomass of late succession species (Spruce, Balsam fir, tons/ha)

54.4

0 - 384.6

4

Coefficient of variation of DBH (Horizontal Complexity)

47.52

0 - 125.84

5

Coefficient of variation of height (Vertical Complexity)

37.08

0 - 110.79

6

Basal area of dead standing trees (m2/ha)

180.38

0 – 1,655.21

7

Volume of dead fallen trees (m3/ha)

13.08

0 - 115.16

8

Wide decay class distribution (Std of decay class)

1.32

0 – 3.56

9

Total Volume (m3/ha)

170.44

0 – 614.62

10

Biomass (tons/ha)

3,297

0 – 14,322

11

Basal area (m2/ha)

0.61

0 - 2.38

12 Abundance of special attributes (broken top, fork, scars, etc.)

3.19

0 – 14.49

13 Age (year)

63.81

0 - 262

2.2.4. Plot-level definitions of old-growth:
I developed eight indices of old-growth forests using the empirically measured forest
attributes (Table 2.2). The first five of these indices were categorical and divided the forest
into old-growth classes. The first index was based only on estimated stand age and divided
the forest into four provincially defined forest age classes (MFLNRORD 1995): initiation (0
– 40 year), young (40 – 70 year), mature (70 – 140 year) and old-growth (>140 years). Four
additional forest classification indices were developed including stand structural attributes
(Table 2.1). For these other four classifications, stand structural attributes were delimitated
into classes using unsupervised k-means classification (indices 2 and 3) or by using a
stepwise procedure that used a random forest routine to reduce the dimensionality of the data

29

prior to applying k-means classification (indices 4 and 5)(Shi and Horvath 2006, Afanador et
al. 2016).
I also created three different old-growth indexes to capture the continuous nature of
the forest structures. To allow the combination of multiple measures to form a single index, I
scaled all old-growth attributes (Table 2.1) to a range from 0 to 1, such that each attribute had
the same weight. Age was utilized as a continuous variable to create the first continuous oldgrowth index (index 6). The other two indexes were created utilizing all old-growth
attributes, with and without age.

30

Table 2.2 Plot level definitions of old-growth forests.

Model

Old-Growth

Type

Definition

Method

Response
Variable

0 = Very-low; >0 and <= 40 = Initiation;
Age Classes

>40 and <= 70 = Young; >70 and <= 140

AGE_CLASS

= Mature; >140 Old-growth
(Discrete Variable)

Classification

Old-growth attributes
+ Age
Old-growth attributes
Old-growth attributes
+ Age

Unsupervised K-means classification, 5
classes.
Unsupervised K-means classification, 5
classes.

KOGA+AGE
KOGA

Reduction of dimensionality with random
forest + Unsupervised K-means

RFOGA+AGE

classification, 5 classes.
Reduction of dimensionality with random

Old-growth attributes

forest + Unsupervised K-means

RFOGA

(Continuous Variable)

Regression

classification, 5 classes.
Age
Old-growth attributes
+ Age

Age as a continuous variable ranging from
0-1
Sum of all old-growth attributes, each as a
continuous variable ranging from 0 - 1

AGE
OGA+AGE

Sum of all old-growth attributes, except
Old-growth attributes

age, each as a continuous variable ranging

OGA

from 0 - 1

2.2.5. Lidar processing:
Airborne LiDAR was collected in a leaf-on condition with a minimum density of 2
pulses/m2, a half-scan angle of 12.5o from nadir, with a 50% overlap. The footprint is
estimated to be from 30 to 70 cm. LAStools (version 161114) was the software utilized to
process the LiDAR’s point cloud. A pipeline for LiDAR processing was illustrated in the
31

Appendix 6.3. Tree height is one of the most fundamental measurements in the forest
industry and has a critical role in the quantitative assessment of forest biomass, carbon
stocks, growth, and site productivity (Andersen et al. 2006). Tree height is highly variable
throughout forest succession, and it is considered an important old-growth attribute (Spies
2004, McElhinny et al. 2006a). Tree height was extracted from the difference between the
Digital Surface Model (DSM) and DTM, where DSM is derived from the first returns and
DTM from the last (Hopkinson et al. 2006, Andersen et al. 2006, Aryal et al. 2017). A list
and description of the LiDAR metrics, mostly derived from height returns, are available in
Table 2.3.
Table 2.3 LiDAR metrics utilized in the random forest models.

Metric name

Metric Description

AHR_Avg

Average of all height returns

AHR_Kur

Kurtoses of all height returns

AHR_Max

Max of all height returns

AHR_Qva

Average of squared height of all height returns

AHR_Ske

Skewness of all height returns

AHR_Std

Standard Deviation of all height returns

AHR_Dns

Number of all points above 1.3m / number of all returns.

H10PercT

Height 10th Percentile

H25PercT

Height 25th Percentile

H50PercT

Height 50th Percentile

H75PercT

Height 75th Percentile

H90PercT

Height 90th Percentile

H95PercT

Height 95th Percentile

STH1_Com

Coefficient of variation of returns of height >0.2m and <1.0m

STH1_Den

Density of points for returns >0.2m and <1.0m / Density of ground returns

STH1_Ske

Skewness of all height returns

STH1_Kur

Kurtoses of all height returns
32

STH1_Cov

Canopy cover (First returns at height > 3.0m/ all first returns*100)

STH2_Com

Coefficient of variation of returns of height >1.0m and <2.0m

STH2_Den

Density of points for returns >1.0m and <2.0m / Density of ground returns

STH2_Ske

Skewness of all height returns

STH2_Kur

Kurtoses of all height returns

STH2_Cov

Canopy cover (First returns at height > 3.0m/ all first returns*100)

STH3_Com

Coefficient of variation of returns of height >2.0m and <3.0m

STH3_Den

Density of points for returns >2.0m and <3.0m / Density of ground returns

STH3_Ske

Skewness of all height returns

STH3_Kur

Kurtoses of all height returns

STH3_Cov

Canopy cover (First returns at height > 3.0m/ all first returns*100)

STH4_Com

Coefficient of variation of returns of height >3.0m

STH4_Den

Density of points for returns >3.0m / Density of ground returns

STH4_Ske

Skewness of all height returns

STH4_Kur

Kurtoses of all height returns

STH4_Cov

Canopy cover (First returns at height > 3.0m/ all first returns*100)

UNDEN

Density of points for returns > 0.2m and < 3.0m / Density of ground returns

VERCOMP

Coefficient of variation of all height returns

2.2.6. Statistical analysis:
In this study, I used the machine learning technique called Random forest (RF) as it is
a powerful classification technique and has been successfully utilized for forest succession
classification (Belgiu and Drăguţ 2016, Cutler and Wiener 2018). RF is a machine learning
method that adds randomness by randomly selecting subsets of the data with the same
distribution without replacement, which increases the diversity of decision trees ("Regression
Trees"). RF combines decision trees, considering the values of an independent random
sample, for all the trees in the forest (Breiman 2001). Thus, each decision tree (regression
tree) is built with not only a random subset of the response variable but also the predicting
33

variables. This structure prevents overfitting and increases the robustness of the model. I
applied the random forest (RF), statistical model, using the "randomforest" package (Cutler
and Wiener 2018) in the R (R Development Core Team 2018) programming environment to
connect field delivered metrics to LiDAR metrics.
Eight random forest models were generated, one for each of the old-growth indices
described in Table 2.2. The models developed here utilized the plot-level classification and
old-growth indexes delivered from fieldwork data as response variable. The predicting
variables are the set of LiDAR metrics listed in Table 2.3. Figure 2.2 depicts the overall
structure of the models. Each random forest model generated 10,000 decision trees to ensure
the stabilization of the model. For each tree, "random forest" utilized a subset of 12 out of 36
predicting variables as suggested by (Breiman and Cutler 2003). In addition, I applied a kfold (k=4) procedure with the r package "Caret" to divide the data into training and validation
data set. Thus, each random forest model was generated with a subsample of 75% of the
available data and validated with the remaining 25%. This procedure was repeated ten (10)
times for each model. Mean accuracy, kappa and balanced class accuracy was reported as a
means for comparison. Similarly, the random forest was applied in its regression mode for
the old-growth indices (6-8 in Table 2.2). For the regression, we reported the means and
standard deviation of the r-squared and the mean square error of the ten repetitions.

34

Figure 2.2 Method design for Random forest model generated to create old-growth maps from
fieldwork classification of forest succession and old-growth indexes, where the response variable are
the plot-level old-growth definitions (Table 2.2) and the predicting variable the LiDAR derived
metrics (Table 2.3).

The R package “raster” was used to generate old-growth maps from the different oldgrowth models developed in this work. The five Random forest models ran with the plotlevel forest succession classification were compared in terms of out of bag error and the oldgrowth misclassification error (Belgiu and Drăguţ 2016). The three Random forest models
generated from the two old-growth indexes and age estimates were compared in terms of
mean squared error. The most robust models, one from classification and one from regression
(old-growth index), were used to generate old-growth maps for the whole study area. To
make comparisons between the categorical and continuous models, I had first to break the
continues model into classes. I utilized the natural breaks (Jenks) option from the ArcGIS
35

classification method to create five classes analogous to the other categorical definitions of
old-growth.
2.3. Results:
2.3.1. Fieldwork data:
I identify ten variables that were best correlated with age (Figure 2.3). None of the
attributes displayed a normal distribution, which highlights the importance of choosing a noparametric statistical framework, “random forest”, for the development of old-growth
models. Nine out of the ten variable included in the development of the old-growth indices
(Table 2.2) were identified and listed by Bauhus et al. (2009) as important old-growth
attributes. From the old-growth attributes included in this study to Bauhus’ list (Table 2.1),
only maximum height was among the highest correlated.

36

Figure 2.3 Histograms of old-growth attributes best correlated with age, displaying the distribution of
the data. Distributions are not normal, which indicates the need for non-parametric models.

I found that for the classification models that utilized old-growth attributes without
transformation (KOGA and KOGA+AGE), the number of plots classified as high and veryhigh old-growth value decreased when compared with the age classes (Table 2.4). Age was
not an essential attribute for KOGA and KOGA+AGE as it did not change the classification.
RFOGA and RFOGA+AGE had a closer distribution of plots in each old-growth value class
as AGE_CLASS compared to the remaining models. However, RFOGA and RFOGA+AGE
have a higher number of plots classified with very-high value for old-growth than any other
classification.
37

Table 2.4 Plot level classification of old-growth values

AGE_CLASS KOGA+AGE

KOGA

RFOGA+AGE

RFOGA

Very-low

7

35

35

13

14

Low

29

26

26

15

17

Intermediate

29

23

23

19

18

High

22

6

6

22

25

Very-high

11

8

8

29

24

Age was not successfully differentiated into age classes in none of the classification
methods developed here, except Age_Class (Figure 2.4). For the distribution of the other ten
old-growth attributes according to the classes and old-growth indices developed in this study,
refer to the Appendix 6.3. I found that, although the oldest stands were mostly classified into
“High” and “Very-high” old-growth values in all classifications, there were still important
misclassifications. The plot where the oldest tree was sampled was classified as either “Low”
(Figure 2.4 b and c) or “High” (Figure 2.4 d and e) old-growth value depending on the
classification method, where it is expected to be in the “Very-high” old-growth value. Similar
to age, the other old-growth attributes were not well differentiated into classes in none of the
classifications proposed here. However, for the age classification (AGE_CLASS), all oldgrowth attributes seem to follow a trend from very low abundance of old-growth attributes
for the “Very-low” and “Low” classes to high abundance for the “High” and “Very-high”
classes (Appendix 6.3 – Figure 6.3). This reinforces the importance of age as a proxy for
forest succession and old-growth value. Similar trends were also present for the other
classifications, although more clear only for the old-growth attributes “maximum tree
height”, “Biomass” and “Basal Areas”. In addition, the old-growth class (“very-high”) would
be better differentiated by some old-growth attributes than to others. For example,
38

“maximum tree-height” and “biomass” separated well the “very-high” class from the
remaining classes in all models, except for AGE_CLASS. “Horizontal complexity” and
“Volume” separated well the “Very-low” and “Low” old-growth value classes from the
remaining classes. Yet, while one attribute separated one class well from the others, there
were still overlaps between at least two classes in all classification for all attributes
(Appendix 6.3). In addition, no single attribute provided a clear differentiation of all classes.
This is mostly because there is not a clear threshold to indicate when a forest enters the oldgrowth stage since stand development is continuous (Wirth et al. 2009b). The continuous oldgrowth index might be a better way to demonstrate the continuous accumulation of oldgrowth attributes suggested in the classification methods.
AGE, OGA+AGE, and OGA represented the old-growth value in a continuous way
instead of arbitrary age classes. In this case, instead of looking at only age or other arbitrary
classifications, I focused on the overall abundance of old-growth attributes. Old-growth
attributes that were well related to AGE were the “Maximum Tree height” and “Biomass,”
both with r-squared >40% (Appendix 6.2 – Figure 6.8). No attributes had r-squared greater
than 45% with AGE. Thus, the old-growth index purely based on age poorly represented
most old-growth attributes, except “Maximum Tree Height,” “Biomass,” and “Age” itself.
The Age-based old-growth index failed to represent other important features, such as the
vertical and horizontal complexity, coarse woody material, dead standing trees, volume and
etc. OGA+AGE and OGA, on the other hand, lost a bit on the age representation, r-squared
55.6% and 45.4% respectively (Figure 2.4 f and g), to gain a greater representation of all
other old-growth attributes (Appendix 6.3 – Figure 6.9 and 6.10). For example, “Dead
Standing Trees” r-squared increased from 8.0% in AGE definition to over 34.6% and 37.1%
for OGA+AGE and OGA, respectively. Similarly, the r-squared for “Late succession
39

species” increased from 9.7% to over 30% for OGA+AGE and OGA. Finally, even
“Maximum Tree Height” and “Biomass,” the best-represented old-growth attributes in the
age-based old-growth index (AGE), were significantly increased in OGA+AGE and OGA.
“Maximum Tree Height” increased from 44.9% to 71.0% and 69.2% for OGA+AGE and
OGA, respectively. For “Biomass” the r-squared was increased from 43.9% in the AGE
model to 79.3% for OGA+AGE and 78.6% for OGA.

40

Figure 2.4 Age distribution according to each different old-growth definition developed in this study.
The names on top of each graph represent one old-growth definition and are listed and described in
Table 2.2.

2.3.2. Comparison between “random forest” models:
The AGE_CLASS model, which utilizes only age classes, performed well for the lowest
levels of old-growth value (Table 2.5). The classes “very-low” and “Low” had the best class
accuracy of all models, 95 and 86%, respectively. However, AGE_CLASS had a weak
41

performance in the remaining classes and overall accuracy. KOGA and KOGA+AGE, which
utilizes k-mean classes of old-growth attributes as response variables, did not perform better
than AGE_CLASS. The stepwise classification applied for the RFOGA and RFOGA+AGE
models, which involves the use of “random forest”, substantially improve the model fit.
RFOGA and RFOGA+AGE had not only the best overall accuracy but also the best class
accuracy for “Intermediate,” “High,” and “Very-high” classes. While RFOGA+AGE has
greater predicting power for class “very-low” and “Intermediate,” RFOGA performed better
in class “Low,” “High,” and “Very-high.” RFOGA based models were better at identifying
forests with “High” and “Very-high” old-growth value than any categorical definition
utilized here. However, the stepwise framework of the RFOGA based models involved the
use of two classification methods (“random forest” and K-means) to reach the final
classification. Using random forest to reduce dimensionality of the data added randomness to
the plot classification, which was stabilized after multiple k-means classification on the data
results of the unsupervised random forest. Thus, while the final RFOGA classifications have
better performance than the others, the classification accuracy for pure age increased from
0.54 in pure age definition to 0.69.

Table 2.5 Summary of statistics for the five old-growth classification random forest models

Accuracy Mean

Kappa Mean

Very-

(+/- SD)

(+/- SD)

Low

AGE_CLASS

0.54 (+/- 0.03)

0.39 (+/- 0.04)

KOGA+AGE

0.53 (+/- 0.03)

KOGA

Classification

Very-

Low

Intermediate

High

0.95

0.86

0.61

0.53

0.63

0.34 (+/- 0.05)

0.86

0.58

0.62

0.50

0.52

0.55 (+/- 0.02)

0.37 (+/- 0.03)

0.86

0.61

0.64

0.51

0.53

RFOGA+AGE

0.69 (+/- 0.02)

0.60 (+/- 0.03)

0.92

0.71

0.83

0.73

0.84

RFOGA

0.68 (+/- 0.02)

0.59 (+/- 0.03)

0.89

0.75

0.74

0.75

0.86

42

High

The regression model generated with age only (AGE) did not perform as well as the
old-growth index models (OGA+AGE and OGA) (Table 2.6). Besides, there was no
difference between the performances of the old-growth index models when age was absent as
an old-growth attribute. Additionally, the difference between the performance of the linear
regressions between OGA and OGA+AGE with old-growth attributes were minimum
(Appendix 6.3 – Figure 6.9 and 6.10). From all eight models developed, I selected three to
predict old-growth values for the whole study site. First was AGE_CLASS, because it is the
method currently used for defining and location old-growth forests (MFLNRORD 1995).
Second, RFOGA because this was the model with the highest accuracy among the categorical
models for the “high” and “very-high” old-growth values. The third and last model was
OGA, because the model has a high performance when compared with the pure age model
(AGE), and it did not include age, which is a costly attribute to be measured.

Table 2.6 Statistical summaries of old-growth index models

Adj. R-squared

Residual standard error

(+/- SD)

(+/- SD)

AGE

0.35 (+/- 0.04)

31.19 (+/- 1.84)

OGA+AGE

0.71 (+/- 0.01)

0.68 (+/- 0.02)

OGA

0.71 (+/- 0.01)

0.55 (+/- 0.01)

Regression

2.3.3. Comparison between old-growth maps:
The RFOGA map underestimated, by at least 1 class difference, 67.5% of the pixels
in the AGE_CLASS map (Table 2.7). This percentage went up to 85.4% when comparing
AGE_CLASS with the OGA map. The high percentage of underestimation suggests that

43

AGE_CLASS maps overestimated the old-growth value, especially for the “Very-high”
class. The “Very-low” class was the one with the least difference between the models. In
addition, the OGA map is the most conservative regarding old-growth mapping, as it mostly
underestimated the “high” and “very-high” old-growth value classification comparing to both
AGE_CLASS and RFOGA models. Besides, the AGE_CLASS map has a higher number of
pixels classified as “Very-high” old-growth value than OGA. However, those pixels are more
scattered and do not follow patterns observed in both RFOGA and OGA (Figure 2.5).
I noticed that AGE_CLASS mapping seems not to classify the classes with higher
old-growth value correctly since the RFOGA and OGA models diverged from AGE_CLASS
in a similar way. The OGA map captures patterns of aggregation between “High” and “Veryhigh” old-growth value, while AGE_CLASS has scattered pixels with very high values
surrounded by pixels with intermediate and low values. I expected that clumps of the oldgrowth forest would intertwine with mature, and that old-growth would be the core of clumps
of mature forests. The lack of pattern observed in the AGE_CLASS map suggests that age
alone is a poor indicator of forest structures, which has also been observed with the empirical
data (Appendix 6.3). Thus, when I tried to force random forest to predict age with LiDAR,
the resultant model was weak and with a high error. Also, when we compare the RFOGA
and OGA maps, RFOGA over-represented “intermediate,” “high,” and “very-high” oldgrowth value classes. However, the patterns of distribution of OGA correlate with the ones
from RFOGA as overestimations were systematic throughout the old-growth value classes
(Figure 2.5 and Table 2.7).
Overall, the results suggest that AGE_CLASS is not capturing the real pattern of
distribution of old-growth values. Additionally, RFOGA is overly representing “high” and
“very-high” old-growth classes. The use of such a map can wrongly indicate an abundance of
44

high old-growth value forests. Setting definition thresholds by age or structure influences the
perception of how much old-growth is in a landscape. If rarity is the impetus to conserving
old-growth and if the old-growth definition is arbitrary, old-growth should be defined in a
way that keeps it rare but not too rare. If old-growth is defined more broadly so that it is
relatively common, it may be less valuable from a biodiversity perspective. Even though
RFOGA is a robust model with high accuracy for “Very-high” old-growth value, its
definition is too broad. Thus, we opted for a more conservative old-growth mapping, OGA.
Finally, we also compared the difference between model OGA+AGE and OGA to make sure
the exclusion of “age” as one of the old-growth attributes did not alter the patterns of
distribution of old-growth values. We found that no pixel had a difference higher than +/0.66 (13%), which is less than one class difference.

45

Figure 2.5 Spatial correlations and constraints between old-growth models with red representing
underestimation and blue overestimations.

46

Table 2.7 Comparison between models to evaluate and locate differences between classes.

RFOGA – AGE_CLASS

OGA – AGE_CLASS

OGA - RFOGA

Class Error (%)

Class Error (%)

Class Error (%)

< -1

>1

< -1

>1

< -1

>1

Very-low

0.0

0.1

0.0

0.0

0.0

0.2

Low

17.8

0.4

16.9

0.7

0.6

0.0

Intermediate

6.3

45.8

16.5

10.6

51.7

0.1

High

8.2

0.0

37.5

0.0

60.3

0.0

Very-high

67.5

0.0

85.4

0.0

76.6

0.0

Class

2.3.4. Assessing Set-Aside Old-Growth Forests
The images collected during field visits supported the “classification” according to the
OGA model (Figure 2.6). From the 14 sites visited, four plots were classified as “Very-high”
old-growth value by OGA, against one by AGE_CLASS. As well, OGA classified only one
plot as “Intermediate” old-growth value against seven by AGE_CLASS. AGE_CLASS
classified most of the plots as “Intermediate” old-growth value, which supports the idea that
age classification masks old-growth forest attributes. Additionally, the old-growth index
(OGA) captured better the accumulation of old-growth attributes in comparison with
AGE_CLASS. However, wetlands received low old-growth values in both models. While
wetlands do not attain the characteristics we were tracking as important old-growth attributes,
this environment usually represent significant pools of carbon and biodiversity (Adhikari et
al. 2009, Kayranli et al. 2010). Despite its importance for biodiversity, carbon and water
resources, these areas often do not have high value for timber and are difficult to access with
machinery. Also, in the study site, wetlands already have legal protections, and have to be
avoided during forest operations (MFLNRORD 1995). Although not accurately depicted in
the map, wetlands are less likely to be affected by forest operations then old-growth forests.
47

In addition, all result suggested that OGA is the model to better capture old-growth values in
the landscape. Thus, I utilized the OGA map to track old-growth values in the landscape and
assess the current set-aside old-growth forests (OGMAs).

Figure 2.6 Forest was classified into a) four different forest successions (classification) and b)
gradient of the abundance of old-growth attributes (regression). Fourteen locations were visited to
evaluate the effectiveness of these two classifications. Four of those sites are depicted here: I) depicts
low level of old-growth attributes (young forest), II) bare rock, III) intermediate level of old-growth
attributes, and IV) high level of old-growth attributes. a) Model Accuracy: 54.4% and “Very-high”
class error = 63,0%, b) R2: 71.5%, MSE= 0.55.

According to our old-growth index model, 14.7% of the tenure area has "Very-high"
old-growth value (Figure 2.7). If we add the "High" old-growth value to it, it can go up to
48

41.6% of potential old-growth. However, from the 14.7% of the landscape that is covered by
forest with "Very-high" old-growth value, only 13.5% is "protected" inside OGMAs.
Additionally, only 2.55% of the OGMAs have "Very-high" old-growth value cover greater
than 50%, and more half of OGMAs had less than 25%. These results suggest a high
fragmentation of "Very-high" old-growth value forest within OGMAs. There is no
specification regarding the percentage cover of old-growth forests one OGMA has to contain
to be considered adequate. It is not the goal of this study to tell which OGMA is effective or
not. However, it concerns that only 13.5% of all forests with "Very-high" old-growth value is
inside OGMAs. Also, these forests are fragmented between the 40 existing OGMAs, since
more than half of the OGMAs have less than 25% of the forest with "Very-high" old-growth
value. Yet, areas with "High" old-growth value were generally surrounding areas with "Veryhigh" old-growth value, in a gradient. This pattern indicates that one OGMA while retaining
a small percentage of "Very-high" old-growth value forests, may simultaneously contain
"High" old-growth value forests.

Figure 2.7 Summary of old-growth assessment in the study area, utilizing the old-growth index OGA,
where a) displays the percentage cover of each old-growth value class for the whole landscape and
only for the areas within OGMAs, and b) depicts the number of OGMAs per percentage forest cover
of “Very-high” old-growth value.
49

2.4. Discussion:
2.4.1. The old-growth index:
The results suggest that the structure-based old-growth definition better captures the
continuous nature of the forest than pure age. The exclusion of age from the models did not
affect the overall models' performances. Moreover, age, although a useful indicator of oldgrowth, is not considered as valuable as some other indicators because the dominant species
in old-growth forests are often multi-aged (Gillis et al. 2003). As well, forests of the northern
interior of British Columbia, Canada, are under a regime of frequent large-scale disturbances,
such as wildfire (DellaSala et al. 1996, Spies et al. 2006). In such a scenario, age estimates
can become quite tricky due to the legacies (e.g. surviving trees, woody debris) left in the
landscape after stand-replacing disturbances. The inclusions of these legacies during
inventories could lead to the overestimation of stand age and drive wrong conclusion about
the forest succession. Therefore, for forests that experience more frequent events of natural
disturbances, the structure-based old-growth definition developed in this study is more
indicated than age-based definitions.
The forest structure-based model OGA not only had better overall statistical
performances than age-based models but also better captured old-growth attributes.
Consequently, the old-growth index has a better chance to capture the functional old-growth
forests in the landscape, as opposed to a forest with few old-trees. This is important because
evidence suggests that the occurrence of many, but not all, species typically found in oldgrowth is linked to specific structural attributes and not to old-growth as such (e.g.
McElhinny et al. 2006, Lonsdale et al. 2008). Due to environmental conditions created by
old-growth forest structures, these forests are often the habit for many wildlife species
50

(Mosseler et al. 2003a, McElhinny et al. 2006b). Thus, the strict separation of forested
landscapes in old-growth and non-old-growth may not represent an optimal species
conservation strategy concerning the provision of habitats in the landscape (Bauhus et al.
2009). Stand development is continuous, and hence there is not a clear threshold to indicate
when a forest enters the old-growth stage (Spies and Franklin 1991, Wirth et al. 2009b).
Nevertheless, a classification of forest succession, either age- or structure-based, is still
arbitrary (Hilbert and Wiensczyk 2007, Wirth et al. 2009b). Therefore, the use of the oldgrowth index (OGA) instead of binary classes is essential in dealing with the continuous
nature of forest structural development and capturing functional old-growth forests.

2.4.2. Application:
Despite the advances in the old-growth definition, there are still important
methodological challenges to be overcome. The old-growth index may be an alternative to
the current landscape scale definitions and deal with some of these challenges. For example,
due to the diversity of old-growth forests, a consensus on a single ecological definition of
old-growth may never be reached, and may not be desirable given the diversity of forests
(Spies 2004, Wirth et al. 2009b). Thus, it is impossible to use the same definitions and
management practices everywhere. The framework utilized in this study can be the means for
multiple local definitions based on measurable structural features and biophysical site
conditions. Such multiple definitions have been long supported by different authors
(Braumandl and Holt 2000, Mosseler et al. 2003c, McElhinny et al. 2006a, Hilbert and
Wiensczyk 2007, Wirth et al. 2009b). Since the index is developed based on local abundance
of old-growth attributes, the framework may facilitate the development of multiple local
definitions. In addition, the degree to which old-growth forests and old-growth structures
51

should be maintained or restored at the landscape level is a complex political question
(MFLNRORD 1995, Environmental Law Centre 2013). The methods current used to select
set-aside old-growth forests might be increasing the risk of old-growth loss as they are
mostly based on stand-age (MFLNRORD 1995, Gillis et al. 2003, Environmental Law Centre
2013). The old-growth index allows for a holistic view of the landscape’s old-growth values,
which indicate the level of abundance or rarity of old-growth in a specific location. It may
serve as the basis for setting local or regional targets for conservation and management
activities.
The old-growth index could be used to design new old-growth reserves, assess current
reserves, and monitor old-growth values as an alternative to the current age-based methods.
However, set-aside forests may be prone to natural disturbances, especially in boreal and subboreal regions (Spies et al. 2006). Thus, as discussed by (Burton et al. 1999), protecting oldgrowth forests should be only part of the strategy to maintain old-growth value in the
landscape. Also, while old-growth is the most vulnerable forest succession, there is not a
forest succession considered as the most essential (Swanson et al. 2011). Areas outside
reserves facilitate gene flow and migration of populations as well as provide complementary
habitat (Lindenmayer and McCarthy 2002). Thus, the old-growth index can aid in the
planning of long-term strategies to accommodate a dynamic population of old-growth stands
in multiple stages of development within landscapes subjected to wildfire, pathogens, and
climate change (Spies 2004). Since the index can track old-growth values throughout the
whole landscape, set-aside forests can be complemented with managed forests that also retain
key attributes of primary and old-growth forests. Such a strategy has been previously
proposed by Beese et al. (2003), who suggested that set-aside old-growth should be
combined with uneven-aged stand management to maintain late-successional forest
52

attributes. This strategy is particularly important in areas where reserves are too small to
ensure the occurrence of natural disturbances within their boundaries or to accommodate all
developmental stages of forest succession (Kneeshaw and Gauthier 2003).

2.4.3. Pros and Cons of the old-growth index:
Despite the vast literature on the topic (Braumandl and Holt 2000, Holt et al. 2002,
McElhinny et al. 2006a, 2006a, Hilbert and Wiensczyk 2007, Wirth et al. 2009b)(see also
Table 1.1), this work is the first to create a landscape-scale definition of old-growth value
based on field-measured old-growth attributes and ALS metrics. The advantages of this
structure-based definition are that data from most forest inventories contain many of the
attributes included in the index and the ease of measurement of those attributes (Mosseler et
al. 2003c, McElhinny et al. 2006a). Additionally, the management objective can dictate the
old-growth threshold. For example, in Chinook community forest forests are managed for
timber and under the frequent regime of large-scale disturbances. In these conditions, it is
expected that stands with high old-growth value to be rare (DellaSala et al. 1996, Spies et al.
2006). Thus, a threshold for the definition of old-growth based on the index can be set such
that set-aside old-growth forests include forest with intermediate old-growth value. The
definition of a threshold is as arbitrary as any other classification method. However, with the
old-growth index the information of the distribution of old-growth values in the whole
landscape is known, and the rarity of old-growth may be used to obtain desirable levels of
old-growth.
The old-growth index allows the measurement of old-growth relative to the
abundance of attributes present in the target landscape. A local old-growth index instead of
broad definitions developed for entire regions can avoid misinterpretations of forest state
53

(Braumandl and Holt 2000, Holt 2000). However, while this framework gains in local
representation, it loses in portability. Daniels (2003) pointed out that it is challenging to
extrapolate structural definitions from one forest type to another. Indeed, while the
methodology of this study may apply to other environments, the characteristic of the oldgrowth forest may change drastically. Also, old-growth includes many distinctive structural
components that do not all change at the same rate and may not all be present in every stand
(Spies 2004).
In addition to the lack of portability, the old-growth index and old-growth mapping have
other caveats. The index includes the classification of wetlands as areas of low old-growth
value. The wetlands are environments that play a significant role in the provision of ESs and
biodiversity (Adhikari et al. 2009, Kayranli et al. 2010). Despite the importance of wetland
ecosystems, these areas often do not have high value for timber and have difficult access to
machinery. Also, in the study site, wetlands have already legal protections, and so have to be
avoided during forest operations (MFLNRORD 1995). The inclusion of a wetness metric to
the old-growth index could lead to overestimation of areas with low old-growth value that
also are not valuable for wetland protection. Thus, while old trees might be present in
wetlands, that should not be a reason enough to modify the old-growth index and classify
them as old-growth forests. The other problem with the old-growth index is that it does not
include any estimates of human or natural disturbance. In the interior forest, where natural
distubances are more frequently, old-growth attributes can be less distinctive (DellaSala et al.
1996). Similar to natural disturbances, human disturbances often reduce the structural
variability that is typical of many naturally developed older forests (Spies 2004).
Nevertheless, Spies (2004) discuss that the source of the disturbance (e.g., human or fire)
would be irrelevant if stands are defined primarily based on structural development.
54

2.5. Conclusion:
In this work, we developed an old-growth index based on forest structures measured with
traditional field methods. While the index is specific to the study site, the methodology is
adaptable to other forest types and ecosystems. We utilized the old-growth index to map oldgrowth value in the landscape and evaluate the amount and quality of the old-growth forest
for the study site. This assessment showed that less than 15% of the forests in the study area
have “very-high” old-growth value. Inside OGMAs, forests with “very-high” represent ~25%
of the total set-aside forests. The main limitations of the index are in the identification of oldgrowth values in wetlands and other environments that limit tree growth (e.g. high elevation),
and the portability of the index to other forest environments. The old-growth index map
allows for a holistic assessment and monitoring of old-growth value in the landscape, which
can aid managers to track the amount and quality of old-growth in the landscape and set
targets for old-growth retention in a transparent manner. Most importantly, mapping oldgrowth value can aid in the conservation of this rare resource and its services in managed
landscapes.

55

3. SYSTEMATIC CONSERVATION PLANNING OF OLD-GROWTH VALUES

Abstract: A systematic conservation-planning tool was applied to design and evaluate oldgrowth reserves that simultaneously provide multiple ecosystem services (ESs). Old-growth
forests play an essential role in the provision of many ESs, such as stocks of carbon and
habitat for many species. Current conservation policies, such as the old-growth management
areas (OGMAs) may not be well situated to protect old-growth while ensuring the provision
multiple services. Thus, identifying priority areas for old-growth forests conservation and
multiple ESs can aid to the maintenance of these values in the landscape. First, ecosystem
services were mapped using LiDAR and field measurements, then an spatial optimization
tool, “PrioritizR,” was utilized to identify optimum reserves’ networks for alternate
conservation scenarios. We discovered that ESs provisioning of current OGMAs are similar
to designed optimum reserves, even though current and designed reserves have minimum
shared territory. In addition, the synergies among the ESs and old-growth were increased
when water was removed from scenarios. Finally, we observed that an increase in OGMAs
areas are not likely to affect timber harvesting until 28% of the study site is set-aside, more
than five times the current OGMAs’ area. The information obtained from "PrioritizR" can be
used to indicate the scope for altering forest reserves locations and to guide the establishment
of new reserves while ensuring the provision of multiple ESs.

3.1. Introduction:
Ecosystem Services (ESs) are the mental and physical benefits obtained by human
populations from ecosystems. Despite recent advances in science, ESs have rarely been
56

incorporated into management decisions (Chan et al. 2006). Old-growth forests have high
provisioning of ESs, such as biodiversity (Spies 2004, Bauhus et al. 2009), ecotourism (FAO
2016), genetic resources (Mosseler et al. 2003b), and carbon storage and sequestration
(Luyssaert et al. 2008, Maxwell et al. 2019). Therefore, using ESs provisioning as a way to
manage older forests could provide a unique opportunity for their sustainability. However,
land management decisions for multiple-user landscapes, such as old-growth forest
management, can be complex if the users have different values. For example, the same parcel
of land can be valued for timber extraction, recreation, biodiversity conservation, food
production, or cultural values by different users (Coomes et al. 2008, Ninan and Inoue 2014).
These conflicting values can lead to management problems. Therefore, thoughtful strategies
are required to manage forest resources for the provision of multiple ESs.
The improvement of human well-being through strategies that promote ESs provisioning
is one of the goals of many environmental policy initiatives (MA 2005, Raudsepp-Hearne et
al. 2010, Guerry et al. 2015). In Canada, Quebec and British Columbia governments
developed strategies to retain potential forest ecosystems for old-growth areas (MFLNRORD
1995, Arsenault 2003, Mosseler et al. 2003c). Canada has pledged, in the International
Convention on Biological Diversity, to develop its forests sustainably, which requires close
attention to old-growth (Mosseler et al. 2003c). Moreover, initiatives such as the United
Nation's Sustainable Development Goals (SDGs) and the Reduced Emission from
Deforestation and Environmental Degradation (REDD+) focus on managing multiple ESs
(Alexander et al. 2011, Griggs et al. 2013). As well, the relationships between different ESs
are complex, and our knowledge of the effect of the multiple drivers of change (e.g. timber
harvesting) on ESs is limited (Nelson et al. 2009, Bennett et al. 2009, Kremen and Ostfeld
2005). Ensuring that forest management actions retain multiple ESs to target levels while
57

providing opportunities for timber extraction is a complex spatial optimization problem
(Schröter and Remme 2016, Snäll et al. 2016).
Unsustainable management strategies can cause unexpected declines in ESs provisioning
and human wellbeing (Lindenmayer et al. 2012, Gaston et al. 2013). For example, logging
can lead to increased areas for cattle grazing and meat production but also lead to decreased
biomass carbon storage (Coomes et al. 2008). The loss of forests can also reduce habitat for
bushmeat species, an important ES for some local communities (Damania et al. 2005). On the
other hand, lower timber densities might be favourable for other bushmeat species (Swanson
et al. 2011). Lastly, timber can be an essential part of local economies. If all logging
operations were banned, local communities could be negatively affected. Consequently, a
community dependent on timber extraction revenues might not value the other ESs provided
by preserved forests compared to those of forest extraction. The participation of communities
and smallholders in the management of forests is essential to increase society's access and
recognition of ESs (FAO 2016). Thus, an alternative to protected areas, such as community
management areas, might be an alternative conservation tool for managed forests (Rodrigues
et al. 2004). A community forest is a notable example of the management of multiple ESs
since timber is not the only target in these communities (MFLNRORD 2017a).
Several studies have previously focused on using systematic conservation planning (SCP)
tools, such as Marxan and Zonation, to spatially optimize ESs provision (Chan et al. 2006,
Nelson et al. 2009, Dade 2018)(see also Luck et al. 2012). Systematic conservation planning
describes the process of identifying and preserving areas of conservation value (Margules
and Pressey 2000, Wilson et al. 2009). It utilizes a spatial analysis of quantitative data to
identify locations for conservation investment. For example, Chan et al. (2006) identified the
priority areas for ecosystem service provisioning across a multi-functional region. Law et al.
58

(2017) identified the effect of different land-use strategies on achieving ecosystem service
targets within a multi-use region of Borneo, Indonesia. More recently, (Dade 2018) utilized
spatial prioritization in an urban setting to identify management strategies to enhance ESs
provisioning in parks while promoting social equality. However, our knowledge on spatially
optimization of the provisioning of multiple ecosystem services is still limited (Snäll et al.
2016). In addition, the concept of systematic conservation planning and spatial prioritization
has not yet being applied to the conservation of old-growth values and multiple ESs in
landscapes managed for timber.
In the conservation realm, old-growth forests have been traditionally valued as wildlife
habitat (Mosseler et al. 2003c). However, conservation of old-growth forests should go
beyond the protection of wildlife habitat as these forests offer vast ESs (Wirth et al. 2009a,
FAO 2016). In contrast, the focus of landscape management has been the provisioning of
timber, food and other raw materials, which can negatively affect other ESs (Monfreda et al.
2008, Ramankutty et al. 2008, Bennett et al. 2009). Therefore, finding a balance in
management that retains key areas for the maintenance of the provision of ESs in old-growth
forests is crucial if maintaining multiple values is a goal. This work aims to (1) analyze the
spatial relationships between ESs and old-growth values, (2) identify strategies for the
management of multiple ESs using a systematic conservation planning tool to, and (3)
evaluate the trade-off between ESs' habitat protection versus timber harvesting. We perform
this analysis in the Chinook Community Forest (CCF), located within the Skeena region,
which management area overlaps with six First Nations' territories.

59

3.2. Material And Method:
3.2.1. Study Area:
(See section 2.2.1)
3.2.2. Estimating Ecosystem Services
A machine learning approach, Random forest (RF), was utilized to extrapolate plot-level
estimates of ESs to the whole study area with LiDAR metrics (Wulder et al. 2008, Dade
2018) (See also Table 1.2). These services were timber volume (m3), carbon storage (Mg),
tree diversity (Shannon diversity index), and water values (wetness index) (Figure 3.1). The
Shannon diversity was generated following the same procedures as DeJong (1975). The
major difference is that we utilized aboveground biomass estimates of individual species to
calculate the index. The allometric equations utilized to estimate timber volume, carbon and
above-ground biomass at the plot level are described in Appendix 6.2.
RF is a powerful classification technique and has been successfully utilized for forest
succession classification (Falkowski et al. 2009, Belgiu and Drăguţ 2016, Cutler and Wiener
2018). I applied the random forest (RF), statistical model, using the "randomforest" package
(Therneau et al. 2011, Cutler and Wiener 2018) in the R (R Development Core Team 2018)
programming environment to connect field delivered metrics to LiDAR metrics. RF is a
machine learning method that adds randomness by randomly selecting subsets of the data
without replacement, which increases the diversity of decision trees ("Regression Trees"). RF
combines decision trees, considering the values of an independent random sample, with the
same distribution, for all the trees in the forest (Breiman 2001). Thus, each decision tree
(regression tree) is built with not only a random subset of the response variable but also the

60

predicting variables. This structure prevents overfitting and increases the robustness of the
model.
The prediction variables are the same set of LiDAR delivered metrics listed in Table 2.3.
Since there are 36 predicting variables, 12 of them are randomly utilized in each division as
indicated by Breiman and Cutler (2003). The response variables are the plot-level estimates
of carbon, timber, tree diversity, and an index for old-growth value (see Chapter 2 – Section
2.2.5). The random forest model produced 10,000 decision trees to ensure the stabilization of
the model. Then, a k-fold (k=4) procedure with the r package "Caret" divided the data into
training and validation data set. Thus, a random forest model is generated with a subsample
of 75% of the available data and validated with the remaining 25%. This procedure was
repeated ten times for each model. The results are reported in terms of means and standard
deviation of the r-squared and mean square error of the ten repetitions.
It was also the objective of this study to represent water value as an ES since old-growth
play an essential role in the landscape hydrology (Wirth et al. 2009a). For that, the soil
moisture index, or wetness index, might be a relevant proxy for water value (Lang and
McCarty 2009). The wetness index from ArcMap 10.1, was derived from a LiDAR highresolution digital elevation model (DEM), was utilized as a proxy for water values (Appendix
6.5). Biodiversity, although a critical ES, was only partially measured in this study. Since the
actual representation of the variation of biodiversity within or between regions is not likely to
be captured through neither fieldwork nor remote sensing, surrogates are still necessary to
represent this ES. Old-growth values and tree diversity are the proxies for biodiversity value.
As indicated by Wilson et al. (2018), there is divergence regarding the performance of
surrogates. Regardless, the objective of this study is to test the use of the conservation

61

prioritization tool “prioritizR” as an OGMAs’ designing tool. Thus, while the accuracy of the
individual inputs is important, they do not limit the study.

Figure 3.1 Ecosystem Services and landscape features utilized in the development of “prioritizR”
scenarios. Section 2.2.5 describes the framework applied to generate the a) old-growth index (Figure
2.2). The ame framework was applied to estimate b) carbon, c) tree diversity, e) timber, and in
Chinook Community Forest. The d) water value and landscape features f) to i) were developed with
surface analysis in ArcMap.

3.2.3. Systematic Conservation for Old-Growth Values
As a conservation planning tool, I utilized “PrioritizR” to simulate optimum reserves’
networks for the provision of old-growth and multiple ecosystem services (Table 3.1).
PrioritizR is an “R package,” which utilizes integer linear programming (ILP) techniques to
62

provide a flexible interface for building and solving conservation planning problems (Hanson
et al. 2019). Once built, conservation-planning problems were solved using the exact
algorithm solver Gurobi (V.7.0). Our conservation-planning problems were the simulation of
reserves that prioritize each ecosystem service individually (old-growth, carbon, water value,
and tree diversity) and all of them together with the same area current set-aside for OGMAs
in the study site. For that, we utilized the “maximum utility objective” function, which
allocates the maximum of the target features into a limited number of planning units (limited
area). High clumpness penalty was also applied to the problems to obtain contiguous
reserves’ design. More information on the selection of clumpness is available in the
Appendix 6.7. We generated five OGMAs’ networks as an alternative to the current one,
where the location might be partially or entirely shifted while preserving the current
OGMAs’ extension. We also simulated increases in OGMAs areas having current OGMAs
and the five optimum OGMAs’ networks as a starting point (Table 3.1).
Timber harvesting has the potential to modify forest patterns in the landscape and
negatively affect other ESs provisioning. However, the study site is primarily managed for
timber values. In this conjecture, in order to prioritize areas for the conservation of multiple
ES, harvesting scenarios have to be taking into account if we want to understand the tradeoffs between protecting and harvesting. For that, we set “prioritizR” to simulate areas that are
“priority” for harvesting, which are the areas with high value for timber, low elevation, flat
slopes, and great proximity to existing roads (Table 3.1). The results of harvesting scenarios
were plotted against OGMAs simulations to evaluate the trade-off between increasing
protection and timber harvesting. It is worth noting that we did not discriminate forest stand
that has not reached the harvesting stage, or timber that is no longer viable. Thus, the total
estimated timber volume was corrected by harvesting yields (100, 75, and 50%) to account
63

for losses in timber volume due to harvesting practices, bole rot, tree size, decay class, age,
and other factors. The deductions were applied uniformly throughout the landscape.

Table 3.1 Conservation planning scenarios for OGMA design with “prioritizR”

Prioritizing
Feature

PrioritizR Models

Same area Scenarios

Increase Area Scenarios

Percentage Measure Percentage Measure

Current OGMAs

5.39%

3541 ha

1 - 50%

1% = 669 ha

Old-growth OGMAs

5.39%

3541 ha

1 - 50%

1% = 669 ha

Carbon

Carbon OGMAs

5.39%

3541 ha

1 - 50%

1% = 669 ha

Tree diversity

Tree Div. OGMAs

5.39%

3541 ha

1 - 50%

1% = 669 ha

Water value

Water OGMAs

5.39%

3541 ha

1 - 50%

1% = 669 ha

All Features

All Features
OGMAs

5.39%

3541 ha

1 - 50%

1% = 669 ha

Timber 100%

Landscape Features* -

-

1 - 50%

1% = 148,047 m3

Timber 75%

Landscape Features* -

-

1 - 50%

1% = 148,047 m3

Timber 50%

Landscape Features* -

-

1 - 50%

1% = 148,047 m3

Old-growth

*High value for timber, proximity with existing roads, low elevation, and flat slopes
3.3. Results:
3.3.1. Ecosystem Services
Random forest models for the estimation of old-growth index and timber had higher rsquares the other estimated services (Table 3.2). The high standard deviation is expected due
to the diversity of plots measured in the field. We encountered plots placed in lakes, barerock
till plots in very high value old-growth forests. The estimated tree diversity was the least
robust of the models. The inclusion of hyperspectral imagery metrics such NDVI could

64

improve the estimates by aiding to the differentiation of species. For timber and carbon, the
pixel value corresponds to per hectare measurement of timber volume (m3) and megagram of
carbon (Mg). Old-growth and tree diversity are indices, where the former ranges from 0-5
and the latter 0-1. There was no measure of accuracy for the water value, as it was directly
derived from a DEM.

Table 3.2 Summary of results from cross-validated random forest models (four fold stratifications
with 10 replicates)

Adj. R-squared

Residual standard

(+/- SD)

error (+/- SD)

2.18 (+/-1.36)

0.71 (+/- 0.01)

0.55 (+/- 0.01)

Timber (m3/ha)

169.17 (+/-153.30)

0.70 (+/- 0.03)

2.06 (+/- 0.07)

Carbon (Mg/ha)

0.04 (+/-0.03)

0.58 (+/- 0.01)

0.01 (+/- 0.00)

Tree Diversity

0.43 (+/-0.39)

0.35 (+/- 0.03)

0.19 (+/- 0.01)

RF Model

Mean (+/- SD)

Old-growth

As expected, carbon and timber had the greatest correlation, as carbon is a function of
timber. Nonetheless, there was a strong correlation between old-growth and carbon compared
to old-growth and timber, 0.91 and 0.78 respectively. The smaller correlation suggests that
there is room for management of the landscape for old-growth and carbon values while
maintaining areas for timber harvesting. The correlations between timber and water values
with old-growth did not change substantially comparing OGMAs (Figure 3.2 b) with
landscape (Figure 3.2 a). On the other hand, tree diversity has a much lower correlation with
old-growth inside OGMAs compared to the landscape, 0.10 and 0.49 respectively. It both
indicates that the current OGMA network does not effectively capture tree diversity and a
need to develop strategies to promote such value both inside and outside OGMAs. The

65

correlation between old-growth and elevation increased from 0.08 in the landscape to 0.18
for OGMAs. It suggests some bias in protecting higher altitude old-growth. However, other
landscape factors such as slope, aspect and road distance did not differ from landscape
correlation to OGMAs’ correlations. Water values had a weak negative correlation with all
other ecosystem services and old-growth, which was not expected as wetter environment are
usually positively correlated carbon. However, it is worth noting that correlation does not
mean causation. The correlation can give us an indication of which services are likely to have
synergies or originate higher trade-off with one another. Yet, we still need to evaluate
management scenarios to evaluate these relationships.

a)

b)

Figure 3.2 Correlation analyses of ecosystem services and landscape variables a) in the whole
landscape, and b) within OGMAs. Values in the grid correspond to the correlation direction and
strength between the variables, and the color scale, in the x axis, represents the significance level (pvalue<0.01)

66

3.3.2. Systematic conservation planning applied to OGMAs
Only a fraction of the priority areas for the conservation of old-growth and other
ecosystem services are currently inside OGMAs (Figure 3.4 a and d) (See also Appendix
6.8). The greatest overlapping between OGMAs and priority areas was for old-growth
prioritization (7.68%), and the smallest for tree diversity (0.44%). Besides, only 1.05% of the
area is shared among all alternative scenarios developed here. However, the difference
between the percentages of each ES reserved in each alternative OGMAs’ network was quite
small (Figure 3.3). For instance, the highest percentage of old-growth representation was
achieved by the old-growth prioritization (7.16%) and the smallest by water value
prioritization (4.96%). The difference between the two scenarios is less than 30%. In
addition, the current OGMAs’ network has 6.44% of total old-growth represented within its
area. The difference between current OGMAs to alternative reserves is also quite small. This
is likely due to the areas constraint utilized. Even prioritizing areas for individual ESs
provision, there is only a small room for improvement of current level of ESs provision in
OGMAs. These results suggest that while multiple ecosystem services can be simultaneously
reserved, the optimum areas for the provision of each ecosystem service are not aligned.
Then, altering the location of OGMAS would no make significant difference in the amount of
old-growth reserved and ESs provision. However, the quality of the forests selected in the
prioritizing scenarios might be greater then in the forests currently set-aside (OGMAs).

67

Old-growth

Timber

Carbon

Tree Diversity

Water Value

10.00%
8.00%
6.00%

4.00%
2.00%
0.00%

Current

Old-growth

Carbon

Tree Diversity

Water Value

All Features

PrioritizR models for OGMAs Design

Figure 3.3 Ecosystem service provisioning represented (%) in current OGMAs network and each
“prioritizR” model. Percentages are the amount of the ecosystem services within the reserve network
in relation to the total amount of the ecosystem in the whole landscape.

The results in Figure 3.4 also suggest that current OGMAs do not need to be
relocated. If the same prioritization strategy is utilized to set-aside new areas for
conservation, it does not matter whether the current OGMAs are relocated to priority areas or
not. The overlap between alternative reserves for old-growth starting from current OGMAs
(Fixed) and priority old-growth areas (Not-fixed) increased from 7.68% to 33.25% with only
1% increase in total reserve areas (Figure 3.4 b). Tree diversity prioritization that had the
smallest overlapping between current OGMAs and priority areas increased from 0.44% to
13.06% in the first 1% area increase (Appendix 6.8 – Figure 6.15). For a 10% area increase,
the overlapping between “Fixed” and “Not-Fixed” scenarios was greater than 60% for oldgrowth prioritization, more than 70% for the prioritization of all ESs simultaneously (Figure
3.4 f). Not only that, the difference between ESs representation from “Fixed” and “NotFixed” scenarios are minimum, and rapidly decreased. For example, in a scenario of a 10%
68

area increase in OGMAs areas, the difference between the ESs provisioning inside “Fixed”
and “Not-Fixed” reserves is less than 10% for all services analyzed. For old-growth
prioritization scenarios, the fixed scenario had a slightly better ES representation. This
suggests that there is a lack of scope for altering OGMAs location to priority areas if there is
intension for OGMAs area increment.

69

Figure 3.4 Comparison between increase in OGMAs' areas starting from current OGMAs (Fixed) and OGMAs designed to prioritize each
individual ecosystem services and all features together (Not-fixed). Old-growth prioritization is represented in figures from a) to c), where a)
represents scenarios with current OGMAs size (0% increment) b) and c) are the reserves with area increment of 1% and 10%, respectively.
The reserves with the prioritization of all ESs are represented in the images from d) to f), where d) are the reserves with 0% increment in area,
e)1%, and f) 10% increment. Each 1% increment equals to an addition of 669ha to the current OGMAs area.
70

The effect of timber harvesting on the provision of ESs is greater than 1:1, where for
each units of timber, one unit of ESs provision would be affected. When we considered
different yields, this relationship became even more detrimental to the ecosystem provision.
The most linear effect of timber removal is on carbon storage loss, since in this study carbon
is a function of timber. The relationship between harvesting and carbon storage loss was
mostly linear, reaching up to a 1:2 relationship, considering the lowest harvesting yield
(50%) (Figure 3.5 b). Besides, despite the low correlation between timber and water (-13%)
and timber and tree diversity (42%) (Figure 3.2), these two values were strongly affected by
timber harvesting (Figure 3.5 c and d). For water values, every 1% of the timber extracted
from the landscape could affect up to 8% of the total water value provision (Figure 3.5 d).
Similarly, the effect of timber removal on tree diversity was up to 1:4.6%. For old-growth
value, that had a strong correlation with timber value (78%), the relationship between timber
harvesting and old-growth value loss was up to 1:3.7%. For example, to harvest 148,047 m3
(1% of the total) affects 2% of the old-growth value assuming 100% yield, 3% for 75%, and
4% for a 50% yield (Figure 3.5 a). The effect of harvesting yields on the extent of the
landscape affected by forest operations, and thus the ESs provided in it, are significantly
increased considering lower harvesting yields. As well, the results suggested that correlations
analysis alone can mislead to the actual ESs relationships, as we noticed a much greater
trade-off between timber and water value than we would expect considering such a small
correlation.

71

Figure 3.5 Trade-off between provision of ecosystem services and timber for harvesting. Images
depict the effect of timber availability on the amount of representation of each ecosystem service that
is removed together with timber. A) old-growth value removed from the landscape for harvesting
scenarios for harvesting yields of 100% , 75%, and 50%, B) Carbon storage, C) Tree Diversity, and
D) Water values.

The combination of curves of prioritizing scenarios enables the evaluation of tradeoffs between protection and timber harvesting and identifying a threshold where protection
would affect timber supply (Figure 3.6). For example, we observed that the increase in
OGMAs areas is not likely to affect timber harvesting before 45% of the study site is setaside for old-growth reserves, assuming that 100% of the non-protected timber is available
for extraction with zero loss during the process (6,660,633 m3). When we assume a loss of
72

25% of the non-protected timber (75% yield), up to 38% of the landscape could be set-aside
for protection without affecting the availability of timber for forestry operations. Even for the
most conservative scenario, where we assume that only 50% of the non-protected timber is
viable for extraction, 28% of the landscape could be set-aside for protection without affecting
harvesting operations (4,143,835 m3 available) (Figure 3.6 a). Also assuming the most
conservative scenarios (50% yield), reserves designed for the provisioning of all ESs could
cover up to 30% of the study site before timber supply was affected (Figure 3.6 e). It is worth
noting that the size of the OGMAs evaluated in this study cover less than 6% of the study
site. Thus, a five-fold increment in OGMAs could occur without affecting forestry operations
considering the most conservative scenario. Notwithstanding, even setting aside the
maximum possible for protection, as indicated by the thresholds between protections and
harvesting scenarios, does not mean that all "non-protect" timber is available for harvesting.
In addition, a five-fold increment in OGMAs areas is not likely to happen. An alternative to
clear-cut, such as partial-cut and selecting logging, could be implemented in the nonprotected landscape to reduce the effect of harvesting on the ESs reserves.

73

Current OGMAs
Timber

Not-available
100% Yield
75% Yield
50% Yield

Figure 3.6 Trade-off curve between timber harvesting and setting aside areas for the provision of a)
old-growth, b) carbon, c) tree diversity, d) water value, and e) all features together. Scenarios were
designed to have a starting point at the current OGMAs network. Thus, 1% increase means that an
additional of 1% of the landscape was set-aside and included in the current OGMAs’ area network.

3.4. Discussion:
This study provides a new insight into the conservation of old-growth forest values in
landscapes managed for timber. It also provides scope for the decision of whether or not to
reallocate current set-aside old-growth forest (OGMAs), and if and how OGMAs can be
designed for the provision of multiple ecosystem services in the landscape. Our results

74

demonstrated that OGMAs could be a strategy to cope with the loss of old-growth (Watson et
al. 2018), while maintaining the provision of multiple of other ESs, including opportunity for
timber harvesting. With the use of a spatial optimization, it was possible to identify potential
thresholds for the conservation of multiple ESs while leaving opportunities for timber
harvesting in the landscape. This information is particular important for community forests
and other land-base tenures that manage the landscape for timber. Since timber is the main
ES that is directly harnessed from these forests (MFLNRORD 2017a), the idea of setting
aside large areas from the land-base to maintain multiple ESs provisioning can be a
conflicting subject. However, even in our most conservative scenario, assuming that only
50% of the non-protected timber would be available for harvesting, OGMAs areas could
cover up to 28% (five-fold current areas) of the land base without compromising timber
supply. The non-protected landscape play an important role in the connectivity of reserves,
gene flow, animal habit etc. Thus the management of the non-protected timber has also to be
strategically planned to retain multiple values, especially considering that unsustainable
management strategies (e.g. clear-cut) can cause unexpected declines in ESs provisioning and
human wellbeing (Lindenmayer et al. 2012, Gaston et al. 2013).
We noticed a much greater trade-off between timber and water value, as we would expect
considering the small correlation between those values. This was similar to a study by Dade
(2018), who also suggested that correlations analysis alone can be misleading regarding the
relationships between ESs. The same was also observed to tree diversity. In addition, when
timber extraction scenarios were evaluated, we found that even for the most correlated ES,
carbon, the removal of timber had a much greater impact on the service provisioning them a
1:1 relationship. Considering that there is other multiple ESs simultaneously provided in
forested landscapes, the indirect effect of unsustainable timber removal may have greater
75

effect on other services. Some studies have pointed out that old-growth forests also support
local economies by providing renewable resources and by attracting tourism and are
important for cultural and religious values (Cronon 1995, MA 2005, Watson et al. 2018).
Cultural values are considered the ES most difficult to be replaced (MA 2005). In addition,
healthier forests are better habitats for some game species, which offers the human
population both recreation opportunities and game meat (Damania et al. 2005). However, the
harvesting scenarios considered here were all clear-cut, assuming the complete removal of
forest structures. A diversity of forest management strategies may, however, reduce the
impact of timber removal on the landscape provision of ESs (Duncker et al. 2012, Schwenk
et al. 2012). In addition, diversifying silvicultural approaches in the non-protected landscape
may also improve the overall representation of tree diversity in the landscape. Our results
showed that not only tree-diversity had a low correlation with old-growth and other
ecosystem values, but also it was poorly represented in all prioritization scenario, except
where tree diversity was the service prioritized. This poor correlation was expected as tree
diversity had a positive response in both early succession and late succession forest types
(Schwenk et al. 2012). The diversification of strategies for timber management can create
landscape pockets with different environmental conditions, which promote the diversity of
tree species and biodiversity in general (McElhinny et al. 2006b, Isbell et al. 2011, Schwenk
et al. 2012).

3.4.1. Ecosystem services reserves
Some reserves protect recreational and scenic values. Others protect ESs such as the
delivery of clean water or the supply of timber or mitigate the expected adverse effects of
over-clearing (Grove 1992). OGMAs were originally idealized to set-aside areas with high
76

old-growth value in the landscape. However, over time, other features started to be
incorporated, and OGMAs were selected for other values such as biodiversity and wildlife
habitat (MFLNRORD 1995, Mosseler et al. 2003a, Environmental Law Centre 2013). Thus,
it is possible to use the OGMA strategy to promote the provision of multiple ecosystem
services. However, OGMAs are not protected areas per se (MFLNRORD 1995,
Environmental Law Centre 2013). Their location can be changed for innumerous
management reasons, including road building and savage of beetle infected trees. However,
their sizes have to be respected, even if they are completely shifted to other regions of the
landscape. This study demonstrated that spatial prioritization could be utilized to demonstrate
whether or not OGMAs should be relocated to better capture the values they were design for.
For the study site, for example, we observed that there would be only a small gain in the
provision of ESs if OGMAs were shifted to areas found as priority for old-growth
conservation. Even when ESs were prioritized individually, their provision were not
substantially higher than in current OGMAs network to advocate for a complete or partial
shift of OGMAs areas. However, the ecosystem features that sustain the provision of those
services may change over time.
Natural disturbances such as wildfire and insect outbreaks, frequent in the landscape
focus of this study (DellaSala et al. 1996, Spies et al. 2006), have the potential to change
planning unit values on a large scale, which could drastically change solutions that were once
considered optimum. Thus, the possibility of partially shifting OGMAs can also be used to
maintain original reserves’ levels of the provision of services. The framework utilized in this
study offers a holistic view of old-growth and ESs values in the landscape, which provides
the opportunity to set targets for their conservation relative to the landscape provision. All
ecosystem services layers generated for this work were derived from field measurements and
77

LiDAR surveys, which can offer a great insight into forests and ecosystem services (Andrew
et al. 2014, Campbell et al. 2017, Aryal et al. 2017). Even though the ESs estimates are
surrogates or partial measures of the actual ecosystem services and old-growth values, it does
not prevent their utilization in this work (Margules and Pressey 2000). Moreover, the
conservation prioritization follows the principle of complementarity, which means that each
reserve contributes to achieving the set of objectives of a prioritization problem for a reserve
network (Margules and Pressey 2000, Wilson et al. 2009). Thus, the representation of oldgrowth values and ecosystem services may change rapidly for individual OGMAs, but less so
for the OGMAs’ network. New OGMAs could, then, account for the multiple ESs oldgrowth forests can provide, reducing management conflicts. However, OGMAs should not be
the only strategy to promote ESs provisioning in the landscape.
Regulating services, such as water and carbon sequestration, usually have synergies with
a few other ESs, such as recreation and habitat quality (Bennett et al. 2009). In this study,
however, the prioritization of water values had the most significant trade-off with the other
ESs evaluated, and was the most affected by timber harvesting, despite the low correlation
between water value and timber. Water value was the liming factor for the simultaneous
multiple ES representation, which means that the old-growth conservation by itself would not
simultaneously protect water-values. These wetter environments tend to restrict tree growth
(Adhikari et al. 2009, Kayranli et al. 2010). Thus, all values related to trees (e.g. old-growth,
tree diversity, and above-ground carbon storage) were not well represented when water
values were the focus of the optimization scenarios. As well, it might also be an effect of the
water value surrogate utilized, the wetness index. The wetness index has the greatest values
assigned to areas with poor drainage, and thus areas with limiting conditions to tree growth
and old-growth characteristics (Lang and McCarty 2009, Lane and D’Amico 2010).
78

Even though wetlands had low correlation with other ESs evaluated in this study, these
are important ecosystem that also provide a variety of ESs (Adhikari et al. 2009, Lang and
McCarty 2009, Kayranli et al. 2010, Lane and D’Amico 2010, Stutter et al. 2012). Moreover,
similarly to agricultural landscapes (Stutter et al. 2012), riparian zones and wetlands are also
often unsuitable for forestry. In addition, there are legal restrictions to forest operations in
riparian areas for the study site (MFLNRORD 1995). Thus, there is little competition
between forestry and riparian ESs. In a scenario where water values were independently
reserved, the overall ESs representation increased from 5.88% to 7.11% for the same reserve
size. Implementing riparian protection in conjunction with OGMAs could increase water
value representation in the landscape while playing an essential role in the total carbon pool
and other ES. Reserves designed for multiple ESs could focus more on the services that have
higher potential for synergies (e.g. carbon and old-growth).
Some assumptions were made in the reserve selection with systematic conservation
planning problems. For this study, the most critical assumption is that the benefits associated
with the selection of a planning unit are guaranteed, are not dynamic, and are independent of
what happens in other planning units (Margules and Pressey 2000). In addition, the problems
addressed in this study are simplified versions of real-world problems. The degree to which
the optimal solution to the simplified problem also represents a good solution to the complex,
real-world problem is generally not known and not evaluated (Langford et al. 2011). Then,
future research on the topic should include to the ESs evaluated here some social aspects of
the landscape values, such as cultural services. Due to the partnership that created the
community forest focus of this study, Indigenous cultural values may play an important role
in the management decisions. Thus, involving the Indigenous groups in an interdisciplinary
study of the landscape values can offer a better insight on the relationships between ESs and
79

the use of OGMAs for the maintenance of multiple ESs in the landscape. Building the
relationship with local community and involving them with the research process might also
be crucial to bridging the research-implementation gap so often mentioned (Knight et al.
2008, Beyer et al. 2016), and aid to validating the effectiveness of conservation plans through
monitoring during and following implementation.

3.5. Conclusion:
Spatial prioritization was successfully utilized to simulate optimum networks of oldgrowth and ESs reserves. While current OGMAs are not placed in optimum areas, the ESs
provisioning in optimum reserves are not substantially different from current OGMAs. Also,
the differences between current OGMAs and optimum reserves decreased rapidly as new setaside areas were added to the current and alternative reserves’ network. These suggest a lack
of scope for altering the location of current OGMAs. We also found that water value was the
services that displayed the greatest trade-off among all scenarios. Since there is little
competition between timber harvesting and water values, specific water conservation
strategies should be implemented simultaneously to multiple ESs OGMAs. Lastly, the results
suggested that an increase in OGMAs areas is not likely to affect timber harvesting before
28% of the study site is set-aside for protection. The information obtained from the spatial
prioritization of old-growth and multiple ESs can be used to indicate the scope for altering
OGMAs’ locations or guiding the establishment of new OGMAs in the landscape. The
spatial prioritization can be the means for identifying priority areas for ESs provisioning,
designing OGMAs for multiple ESs, and the evaluation of trade-off between ESs due to
management objectives.

80

4. CONCLUSIONS:
An underlying premise of my thesis is that identifying and retaining old-growth
forests can aid to their conservation and the maintenance of other essential ESs in the
landscape. Old-growth forests are ecosystems often associated with the provision of a range
of ecosystem services (ESs) such as biodiversity, water provision, carbon storage and
sequestration, cultural values and the maintenance of human well-being. Although oldgrowth forests are a valuable ecological resource, they are also rare and in a rapid decline
worldwide. The Old-Growth Management Areas (OGMAs) in British Columbia, Canada is
one example of strategy to protect and retain old-growth forests and their ESs in managed
landscapes. However, the selection of such areas is a difficult task due to two critical
problems. First, there are many definitions and approaches to define and locate old-growth
forests due to their diversity across ecological types. Second, the relationships between
ecosystem services and old-growth values are complex, and our knowledge of how ESs
responds to forest management is limited.
To address these gaps, I studied the role of old-growth reserves on the provision of
multiple ecosystem services (ESs), using the Chinook community forest, located in Burns
Lake, British Columbia, Canada, as a case study. To bridge the knowledge gaps, I used
review of literature on old-growth and ESs, a machine learning technique, field and ALS data
on forest structures, and a conservation planning method. Specifically, I addressed three
major objectives: (1) identify definitions of forest succession, structural attributes and ESs
commonly associated with old-growth forests (Chapter 1); (2) develop a landscape scale
definition of old-growth forest that is both quantitative and repeatable (Chapter 2); (3)
understand and apply the synergies and trade-off between ESs and old-growth values to
promote the use of OGMAs as strategies for the maintenance and provision of multiple ESs
81

in the landscape.
In this concluding chapter, I synthesize the main findings from each of the chapters of
my thesis while discussing their implications for the management of multiple ecosystem
services in forests managed for timber, and recommend future research directions.

4.1. Main Findings and contributions:
4.1.1. General review on old-growth forests and their provision of ecosystem
services:
Chapter 1 of this thesis presents a general review of the literature for the multiple
definitions of old-growth forest and the structural attributes commonly associated with oldgrowth forests. Attribute-based definitions were found to be the most common and
transparent approaches used in the literature. In addition, most of the old-growth attributes
are already part of traditional field inventories, which aids their utility for identifying oldgrowth. However, methods to accurately assess these attributes across a landscape at an
appropriately grain size have not been available. This has meant that spatially explicit
estimates of old-growth forests have often been low accuracy, thereby limiting land managers
ability to accurately identify and manage old-growth areas. Advances in remote sensing
technology, such as Airborne LiDAR, potentially provide the ability to evaluate the 3D
structural attributes of forest ecosystems, develop modeled estimates of old-growth
characteristics, and incorporate this information into landscape planning that aims to
maintain old-growth forests while also promoting multi-functional forestry that includes
timber productions. The selection of old-growth is a difficult task due to the lack of a
standard definition for what constitutes an old-growth forest (Hilbert and Wiensczyk 2007,
Wirth et al. 2009b). In temperate forests, obtaining a definition of old-growth forest is
82

somewhat easier as the fire-rotation periods are exceptionally long in such forests. However,
in the other regions, a clear decision on when to designate a given stand "old-growth" is more
difficult due to disturbances and successional changes that maintain the system in a dynamic,
non-climax state (Mosseler et al. 2003a, Wirth et al. 2009b). As a result, developing a
continuous index of old-growth attributes, instead of a categorical and somewhat arbitrary
designation, have the potential to improve our ability to identify old-growth forests for
conservation and incorporate them into landscape planning.

4.1.2. Defining Old-Growth Forests With Airborne LiDAR
In this work, we developed an old-growth index based on forest structures measured
with traditional field methods. Old-growth attributes were utilized to develop a structurebased old-growth index tailored to map old-growth value in the Chinook Community Forest
landscape. LiDAR-derived metrics were modeled using a “random forest” (RF) framework to
estimate the old-growth index across the landscape. To obtain the landscape-scale old-growth
index I first needed to evaluate means for assessing old-growth, and OGMAs, and develop
quantitative definitions that could be used in conjunctions with fine-scale remotely sensed
data such as ALS. While the index developed is specific to the study site, the framework, is
generic enough to be adapted to other forest types and ecosystems. More importantly, the
identification of the amount and location of old-growth forests in the landscape can aid in the
conservation of this rare resource and its services.
I used the old-growth index OGA developed to map old-growth value in the
landscape and evaluate the amount and quality of the old-growth forest for the study site. The
old-growth index map developed in this study allows for a quantitative assessment and
monitoring of old-growth value in the landscape. It can help managers to not only track the
83

amount and quality of old-growth in the landscape but also set targets for old-growth
retention in a transparent manner. Most importantly, the old-growth mapping value can aid in
the conservation of this resource and its services in managed landscapes. The old-growth
policy in Canada came from a public concern about the decline in the amount and
distribution of older forests, especially those with large, old trees (Environmental Law Centre
2013). Moreover, Canada has pledged, in the International Convention on Biological
Diversity, to develop its forests sustainably, which requires close attention to old-growth as a
distinct component of those forests (Mosseler et al. 2003c). The old-growth index can serve
as an alternative to the current old-growth definitions utilized (Gillis et al. 2003)
(MFLNRORD 1995, Arsenault 2003). The old-growth index developed here, while specific
for the study site, it can be adapted for other forest types.
This methodology offers a way to evaluate old-growth values in set-aside old-growth
forests, OGMAs. Since the index can track old-growth values throughout the whole
landscape, set-aside forests can be complemented with managed forests that also retain key
attributes of primary and old-growth forests. Such a strategy has been previously proposed by
Beese et al. (2003), who suggested that set-aside old-growth should be combined with
uneven-aged stand management to maintain late-successional forest attributes. Other
researchers also indicate the use of multiple silvicultural approaches simultaneously in the
non-protected landscape, as a mean for retaining forest structures and ESs while extracting
timber values (Duncker et al. 2012, Schwenk et al. 2012). This strategy is particularly
important in areas where reserves are too small to ensure the occurrence of natural
disturbances within their boundaries or to accommodate all developmental stages of forest
succession (Kneeshaw and Gauthier 2003).

84

4.1.3. Systematic Conservation Planning Of Old-Growth Value
The management of landscapes often involves conflicting objectives such as timber to
one group of people and recreation to another (Ninan and Inoue 2014). As human
populations continue to grow, demand for ecosystem services is also increasing (MA 2005,
United Nations 2018), and so are the conflicts between multiple objectives. Using the spatial
optimization package “PrioritizR,” I was able to evaluate how prioritizing different
objectives can alter the spatial location of set-aside areas, and alter old-growth maintenance
and ES provisioning. Current OGMAs locations are not optimum for any of the ESs
evaluated. Although current OGMAs and priority ESs reserves barely overlapped, the
provision of ecosystem services had a very small difference between scenarions. This is
likely due to the size restriction. Since I was comparing current OGMAs, the size of current
OGMAs was utilized in all scenarios. Thus, even using a spatial prioritaztion tool, the
priority areas selected were still too small to provide a more substantial difference. I also
simulated increases in OGMAs areas having current OGMAs and optimum ESs' reserves as a
starting point. The differences between current OGMAs and optimum reserves decreased
rapidly as new set-aside areas were placed and overlapping between scenarios increased.
However, for a better comparison, I would have to select areas to include in current OGMAs
in the same way they are currently selected. While the way the analyses were conducted in
this thesis points to an interesting result, it does not offer the scope to say which method is
more robust for OGMA designing.
As well, the smallest overall ESs representation was achieved in scenarios where
water value is prioritized, indicating that water values is a limiting factor for the prioritization
of multiple ESs. However, when water value is reserved separated from other ESs, the
synergies between the remaining services are increased. Since there is little competition
85

between timber harvesting and riparian areas where water values are often higher, removing
water values from the prioritization scenarios can substantially increase other ESs,
representation. A conservation plan that includes strategies to protect riparian and wetlands in
conjunction with multiple ESs OGMAs can substantially increase synergies and overall
representation of ESs.
Regarding increments in set-aside areas, OGMAs would not affect timber harvesting before
receiving a five-fold increment in their current areas. This information can be used to indicate
the scope for altering OGMAs' locations and guide the establishment of new OGMAs in the
landscape. These findings suggest that the use of spatial prioritizing tools can be a new
means for identifying priority areas for ESs provisioning, designing OGMAs for multiple
ESs, and the evaluation of trade-off between ESs due to management objectives.

4.2. Future research:
Future research for the old-growth definition would require testing the portability of the
framework developed in this study to other forest types, evaluating other old-growth
attributes, and which ones are better suited for which forest type. In addition, forest
inventories have to be tailored for forest succession studies, including direct measurements of
forest floor and coarse woody debris, likens and other attributes (Wirth et al. 2009b), if there
is the intension of improving current definitions. The idea of the index developed here is that
it can be repeated and applied in other regions, since most of the old-growth attribute utilized
are present in tradition forest inventories for timber supply purposes (Gillis et al. 2003).
However, improvement on the representation of old-growth could be achieved by a forest
inventory with an ecological purpose, with data collection focused to capture old-growth
attributes, biodiversity metrics, likens and fungi communities, ecosystem services and other
86

relevant landscape values. Not only for the development of the old-growth index, this
inventory can also support the production of better estimates of ESs estimates and include
other services such as biodiversity, hunting opportunities, soil carbon, and cultural and
recreational values.
Old-growth forests and other important ecosystems sustain critical ecological processes,
e.g. water filtration, carbon sequestration, and nitrogen fixation) that constitute Earth’s life
support systems (e.g. provision of clean water and air, climatic stability) (MA 2005, Watson
et al. 2018). They also support local economies by providing renewable resources and by
attracting tourism and are important for cultural and religious values (Cronon 1995, FAO
2015). Culture and religion may play an essential role in the relationship of ESs in the
landscape. Cultural services refer to the non-material benefits people obtain from ecosystems,
such as aesthetic appreciation, different forms of recreation, and cultural identity (Daniel et
al. 2012). I did not have any direct measure of culturally related ESs. However, it is possible
to state that those values were partially prioritized by surrogates such as old-growth value.
Such assumption is often made, but there is little research that accounted for the ecosystems
and ecosystems functions that sustain the provision of those services (Milcu et al. 2013,
Schirpke et al. 2016, Watson et al. 2018, Dade 2018). Cultural values are considered the ES
most difficult to be replaced once lost (MA 2005). Thus, evaluating the role of old-growth for
the local culture is an important relation that needs to be studies. The study site offers a
unique opportunity for such research since it overlaps with six first nations territories
(Chinook Community Forest 2017). Although the participation of local communities in the
development of the study would offer an interesting and crucial perspective on the project,
building such relations is out of the scope of this master’s thesis.

87

Other services that are also simultaneously provided by old-growth values are game
meat and recreation. Healthier forests are better habitats for some game species, which offers
the human population both recreation opportunities and game meat (Damania et al. 2005).
Old-growth forests are winter habitat for some of the these species (Mosseler et al. 2003c).
As well, large mammals can also beneficiate from the conservation of such habitat (bears
(Ursus spp.), wolves (Canis lupus), and wolverines (Gulo)) (Mosseler et al. 2003c). Not only
that, old-growth forests are a unique environment with high eco-tourism value (MA 2005).
The recreation aspect is already partially exploited by the community forest by means such as
trails, sightseen spots, mountain biking, and camping sites (Chinnok Community Forest
2017b). However, those recreation opportunities still need to be inventoried, mapped and
evaluated in terms of usage, surrounding environments, and other proxies for their quality.
Therefore, the conservation of old-growth have potential to simultaneously provide multiple
of other services not evaluated in this study. Further research on the provision of these other
services has to be conducted before drawing conclusion about the importance of old-growth
their provision.
Management plans often include the idea of uncertainty. Old-growth values, as well
as all ESs, evaluated here were only estimated. Different harvesting yields were attempted to
input an uncertainty measure for timber harvesting and its effect on ESs. The evaluation of
all ESs and conservation plans should include a similar idea. Decision-makers have to know
the risk of setting-aside areas and the uncertainty on the returns of their investment. GrêtRegamey et al. (2013) discussed the idea of risk assessment for ESs provisioning and applied
a Bayesian network for its assessment. Thus, future research may include a risk assessment to
ESs provision on the prioritizing scenarios.

88

Finally, monitoring before and after the implementation of conservation plans for
multiple ESs is a crucial step in future research (Margules and Pressey 2000, Wilson et al.
2009). This thesis offers a valuable framework to measure ESs and old-growth value in the
landscape through ALS. Repetitive ALS expeditions are already expected to happen in the
study site and might be the means for monitoring on a landscape scale the effects short and
long-term management practices. Thus, old-growth value and ESs could be re-measured and
used to the evaluation of the effect of timber harvesting implemented in the year following
the first ALS expedition. As well, the degree to which the spatial prioritization solutions
represents a good solution to the complex, real-world problem is generally not known and not
evaluated (Langford et al. 2011). This is an important conservation planning shortcome
(Margules and Pressey 2000, Wilson et al. 2009). Then, ALS expedition prior and after the
implementation of spatial prioritization solution could aid to evaluate how systematic
conservation problems correspond to the real-world problem, the effect of implementing
them on the provision of ESs. To better capture the real world complexity, future research
should include the ESs evaluated here with some social aspects of the landscape values, such
as cultural services. Thus, building the relationship with the local community and involving
them with the research process might also be crucial to bridging the research-implementation
gap so often mentioned (Knight et al. 2008, Beyer et al. 2016), and help validate the
effectiveness of conservation plans through monitoring during and following
implementation.

4.3. Concluding Remarks
Due to the vital importance and rapid decline of old-growth forests, it is imperative
advancement in polices for their conservation. Old-growth forests are often associated with
89

the provisions of multiple ecosystem services. In the current scenario of increasing
population and demand for ecosystem services, policies that promote the provision of
multiple ecosystem services can be the means for the protection of old-growth forests. In that
regard, my thesis advances our understanding on the multiple definitions of old-growth
forests, provides a quantitative and repeatable framework for identifying old-growth values
in the landscape, and recommends the use of spatial prioritization tool as the means to
manage the landscape and design OGMAs for the provision of multiple ecosystem services.

90

5. REFERENCES
Adhikari, S., R. M. Bajracharaya, and B. K. Sitaula. 2009. A Review of Carbon Dynamics
and Sequestration in Wetlands. Journal of Wetlands Ecology:42–46.
Afanador, N. L., A. Smolinska, T. N. Tran, and L. Blanchet. 2016. Unsupervised random
forest: a tutorial with case studies. Journal of Chemometrics 30:232–241.
Ahmed, R., P. Siqueira, and S. Hensley. 2013. A study of forest biomass estimates from lidar
in the northern temperate forests of New England. Remote Sensing of Environment
130:121–135.
Alexander, S., C. R. Nelson, J. Aronson, D. Lamb, A. Cliquet, K. L. Erwin, C. M. Finlayson,
R. S. de Groot, J. A. Harris, E. S. Higgs, R. J. Hobbs, R. R. R. Lewis, D. Martinez,
and C. Murcia. 2011. Opportunities and Challenges for Ecological Restoration within
REDD+. Restoration Ecology 19:683–689.
Andersen, H.-E., S. E. Reutebuch, and R. J. McGaughey. 2006. A rigorous assessment of tree
height measurements obtained using airborne lidar and conventional field methods.
Canadian Journal of Remote Sensing 32:355–366.
Andrew, M. E., M. A. Wulder, and T. A. Nelson. 2014. Potential contributions of remote
sensing to ecosystem service assessments. Progress in Physical Geography: Earth and
Environment 38:328–353.
Arsenault, A. 2003. A note on the ecology and management of old-growth forests in the
Montane Cordillera. The Forestry Chronicle 79:441–454.
Aryal, R. R., H. Latifi, M. Heurich, and M. Hahn. 2017. Impact of Slope, Aspect, and
Habitat-Type on LiDAR-Derived Digital Terrain Models in a Near Natural,
Heterogeneous Temperate Forest. PFG – Journal of Photogrammetry, Remote
Sensing and Geoinformation Science:1–13.
Ayanu, Y. Z., C. Conrad, T. Nauss, M. Wegmann, and T. Koellner. 2012. Quantifying and
Mapping Ecosystem Services Supplies and Demands: A Review of Remote Sensing
Applications. Environmental Science & Technology 46:8529–8541.
Ball, I. R., H. P. Possingham, and M. E. Watts. 2009. Marxan and Relatives: Software for
Spatial Conservation Prioritization:12.

91

Balvanera, P., A. B. Pfisterer, N. Buchmann, J.-S. He, T. Nakashizuka, D. Raffaelli, and B.
Schmid. 2006. Quantifying the evidence for biodiversity effects on ecosystem
functioning and services. Ecology Letters 9:1146–1156.
Bater, C. W., N. C. Coops, S. E. Gergel, V. LeMay, and D. Collins. 2009. Estimation of
standing dead tree class distributions in northwest coastal forests using lidar remote
sensing. Canadian Journal of Forest Research 39:1080–1091.
Bauhus, J., K. Puettmann, and C. Messier. 2009. Silviculture for old-growth attributes. Forest
Ecology and Management 258:525–537.
Beese, W. J., B. G. Dunsworth, K. Zielke, and B. Bancroft. 2003. Maintaining attributes of
old-growth forests in coastal B.C. through variable retention. The Forestry Chronicle
79:570–578.
Belgiu, M., and L. Drăguţ. 2016. Random forest in remote sensing: A review of applications
and future directions. ISPRS Journal of Photogrammetry and Remote Sensing
114:24–31.
Bennett, E. M., G. D. Peterson, and L. J. Gordon. 2009. Understanding relationships among
multiple ecosystem services. Ecology Letters 12:1394–1404.
Bergen, K. M., and I. Dronova. 2007. Observing succession on aspen-dominated landscapes
using a remote sensing-ecosystem approach. Landscape Ecology 22:1395–1410.
Beyer, H. L., Y. Dujardin, M. E. Watts, and H. P. Possingham. 2016. Solving conservation
planning problems with integer linear programming. Ecological Modelling 328:14–
22.
Bithell, M., and J. Brasington. 2009. Coupling agent-based models of subsistence farming
with individual-based forest models and dynamic models of water distribution.
Environmental Modelling & Software 24:173–190.
Braumandl, T., and R. Holt. 2000. Refining definitions of old growth:4.
Breiman, L. 2001. Random Forests. Machine Learning 45:5–32.
Breiman, L., and A. Cutler. 2003. Setting Up, And Understanding Random Forests V4.0.
Pages 5–32.
Bright, B. C., A. T. Hudak, R. McGaughey, H.-E. Andersen, and J. Negrón. 2013. Predicting
live and dead tree basal area of bark beetle affected forests from discrete-return lidar.
Canadian Journal of Remote Sensing 39:S99–S111.
92

Burton, P. J., D. D. Kneeshaw, and K. D. Coates. 1999. Managing forest harvesting to
maintain old growth in boreal and sub-boreal forests. The Forestry Chronicle 75:623–
631.
Campbell, L., N. C. Coops, and S. C. Saunders. 2017. LiDAR as an Advanced Remote
Sensing Technology to Augment Ecosystem Classification and Mapping. Journal of
Ecosystems and Management 17.
Chan, K. M. A., M. R. Shaw, D. R. Cameron, E. C. Underwood, and G. C. Daily. 2006.
Conservation Planning for Ecosystem Services. PLOS Biology 4:e379.
Chazdon, R. L. 2008. Beyond Deforestation: Restoring Forests and Ecosystem Services on
Degraded Lands. Science 320:1458–1460.
Chen, J., and G. A. Bradshaw. 1999. Forest structure in space: a case study of an old growth
spruce-fir forest in Changbaishan Natural Reserve, PR China. Forest Ecology and
Management 120:219–233.
Chichilnisky, G., and G. M. Heal. 1986. The Evolving International Economy. Cambridge
University Press.
Chinnok Community Forest. 2017a. Ownership Structure. http://chinookcomfor.ca/.
Chinnok Community Forest. 2017b. Chinook Community Forest: Recreation Management
Plan. http://chinookcomfor.ca/chinook-community-forest-recreation-managementplan/.
Cohen, W. B., T. A. Spies, and M. Fiorella. 1995. Estimating the age and structure of forests
in a multi-ownership landscape of western Oregon, U.S.A. International Journal of
Remote Sensing 16:721–746.
Coomes, O. T., F. Grimard, C. Potvin, and P. Sima. 2008. The fate of the tropical forest:
Carbon or cattle? Ecological Economics 65:207–212.
Coops, N. C., T. Hilker, M. A. Wulder, B. St-Onge, G. Newnham, A. Siggins, and J. A.
(Tony) Trofymow. 2007. Estimating canopy structure of Douglas-fir forest stands
from discrete-return LiDAR. Trees 21:295.
Costanza, R., R. d’Arge, R. de Groot, S. Farber, M. Grasso, B. Hannon, K. Limburg, S.
Naeem, R. V. O’Neill, J. Paruelo, R. G. Raskin, P. Sutton, and M. van den Belt. 1997.
The value of the world’s ecosystem services and natural capital. Nature
387:387253a0.

93

Cronon, W. 1995. Uncommon Ground: Toward Reinventing Nature.
https://repository.library.georgetown.edu/handle/10822/879958.
Crossman, N. D., B. Burkhard, S. Nedkov, L. Willemen, K. Petz, I. Palomo, E. G. Drakou, B.
Martín-Lopez, T. McPhearson, K. Boyanova, R. Alkemade, B. Egoh, M. B. Dunbar,
and J. Maes. 2013. A blueprint for mapping and modelling ecosystem services.
Ecosystem Services 4:4–14.
Cutler, F. original by L. B. and A., and R. port by A. L. and M. Wiener. 2018. randomForest:
Breiman and Cutler’s Random Forests for Classification and Regression.
Dade, M. C. 2018, December 20. Managing multiple ecosystem services. PhD Thesis, The
University of Queensland.
Damania, R., E. J. Milner-Gulland, and D. J. Crookes. 2005. A bioeconomic analysis of
bushmeat hunting. Proceedings of the Royal Society of London B: Biological
Sciences 272:259–266.
Daniel, T. C., A. Muhar, A. Arnberger, O. Aznar, J. W. Boyd, K. M. A. Chan, R. Costanza,
T. Elmqvist, C. G. Flint, P. H. Gobster, A. Grêt-Regamey, R. Lave, S. Muhar, M.
Penker, R. G. Ribe, T. Schauppenlehner, T. Sikor, I. Soloviy, M. Spierenburg, K.
Taczanowska, J. Tam, and A. von der Dunk. 2012. Contributions of cultural services
to the ecosystem services agenda. Proceedings of the National Academy of Sciences
109:8812–8819.
Daniels, L. D. 2003. Western redcedar population dynamics in old-growth forests:
Contrasting ecological paradigms using tree rings. The Forestry Chronicle 79:517–
530.
DeJong, T. M. 1975. A Comparison of Three Diversity Indices Based on Their Components
of Richness and Evenness. Oikos 26:222–227.
DellaSala, D. A., J. R. Strittholt, R. F. Noss, and D. M. Olson. 1996. A Critical Role for Core
Reserves in Managing Inland Northwest Landscapes for Natural Resources and
Biodiversity. Wildlife Society Bulletin (1973-2006) 24:209–221.
DeLong, S. C., P. J. Burton, and M. Harrison. 2004. Assessing the relative quality of oldgrowth forest 4:16.
Duncker, P. S., K. Raulund-Rasmussen, P. Gundersen, K. Katzensteiner, J. De Jong, H. P.
Ravn, M. Smith, O. Eckmüllner, and H. Spiecker. 2012. How Forest Management
affects Ecosystem Services, including Timber Production and Economic Return:
Synergies and Trade-Offs. Ecology and Society 17.
94

Environmental Law Centre. 2013. An Old Growth Protection Act for British Columbia.
University of Victoria.
Falkowski, M. J., J. S. Evans, S. Martinuzzi, P. E. Gessler, and A. T. Hudak. 2009.
Characterizing forest succession with lidar data: An evaluation for the Inland
Northwest, USA. Remote Sensing of Environment 113:946–956.
FAO. 2015. Global forest resources assessment 2015.
FAO. 2016. State of the World’s Forests 2016 | FAO | Food and Agriculture Organization of
the United Nations. http://www.fao.org/publications/sofo/en/.
Gaston, K. J., M. L. Ávila‐Jiménez, and J. L. Edmondson. 2013. REVIEW: Managing urban
ecosystems for goods and services. Journal of Applied Ecology 50:830–840.
Gaveau, D. L. A., and R. A. Hill. 2003. Quantifying canopy height underestimation by laser
pulse penetration in small-footprint airborne laser scanning data. Canadian Journal of
Remote Sensing 29:650–657.
Gillis, M. D., S. L. Gray, D. Clarke, and K. Power. 2003. Canada’s National Forest
Inventory: What can it tell us about old growth? The Forestry Chronicle 79:421–428.
Goodwin, N. R., N. C. Coops, and D. S. Culvenor. 2006. Assessment of forest structure with
airborne LiDAR and the effects of platform altitude. Remote Sensing of Environment
103:140–152.
Grêt-Regamey, A., S. H. Brunner, J. Altwegg, and P. Bebi. 2013. Facing uncertainty in
ecosystem services-based resource management. Journal of Environmental
Management 127:S145–S154.
Griggs, D., M. Stafford-Smith, O. Gaffney, J. Rockström, M. C. Öhman, P. Shyamsundar,
W. Steffen, G. Glaser, N. Kanie, and I. Noble. 2013. Sustainable development goals
for people and planet. Nature 495:305–307.
Grove, R. H. 1992. Origins of Western Environmentalism. Scientific American 267:42–47.
Guerry, A. D., S. Polasky, J. Lubchenco, R. Chaplin-Kramer, G. C. Daily, R. Griffin, M.
Ruckelshaus, I. J. Bateman, A. Duraiappah, T. Elmqvist, M. W. Feldman, C. Folke, J.
Hoekstra, P. M. Kareiva, B. L. Keeler, S. Li, E. McKenzie, Z. Ouyang, B. Reyers, T.
H. Ricketts, J. Rockström, H. Tallis, and B. Vira. 2015. Natural capital and ecosystem
services informing decisions: From promise to practice. Proceedings of the National
Academy of Sciences 112:7348–7355.

95

Hansen, A., K. Barnett, P. Jantz, L. Phillips, S. J. Goetz, M. Hansen, O. Venter, J. E. M.
Watson, P. Burns, S. Atkinson, S. Rodríguez-Buritica, J. Ervin, A. Virnig, C. Supples,
and R. De Camargo. 2019. Global humid tropics forest structural condition and forest
structural integrity maps. Scientific Data 6:1–12.
Hanson, J., R. Schuster, N. Morrell, M. Strimas-Mackey, M. Watts, P. Arcese, J. Bennett,
and H. P. Possingham. 2019. Systematic conservation prioritization in R. R, prioritizr.
Hao, Z., J. Zhang, B. Song, J. Ye, and B. Li. 2007. Vertical structure and spatial associations
of dominant tree species in an old-growth temperate forest. Forest Ecology and
Management 252:1–11.
Harter, R., K. Hornik, S. Theuss, C. Szymanski, and F. Schendinger. 2017. SYMPHONY in
R.
Hendrickson, O. 2003. Old-growth forests: Data gaps and challenges. The Forestry Chronicle
79:645–651.
Hermosilla, T., M. A. Wulder, J. C. White, N. C. Coops, and G. W. Hobart. 2015. Regional
detection, characterization, and attribution of annual forest change from 1984 to 2012
using Landsat-derived time-series metrics. Remote Sensing of Environment 170:121–
132.
Hilbert, J., and A. Wiensczyk. 2007. Old-growth definitions and management: A literature
review. Journal of Ecosystems and Management 8.
Hill, R. A., and A. G. Thomson. 2005. Mapping woodland species composition and structure
using airborne spectral and LiDAR data. International Journal of Remote Sensing
26:3763–3779.
Hobbs, R. J., S. Arico, J. Aronson, J. S. Baron, P. Bridgewater, V. A. Cramer, P. R. Epstein,
J. J. Ewel, C. A. Klink, A. E. Lugo, D. Norton, D. Ojima, D. M. Richardson, E. W.
Sanderson, F. Valladares, M. Vilà, R. Zamora, and M. Zobel. 2006. Novel
ecosystems: theoretical and management aspects of the new ecological world order.
Global Ecology and Biogeography 15:1–7.
Hofton, M. A., J. B. Blair, J.-B. Minster, J. R. Ridgway, N. P. Williams, J. L. Bufton, and D.
L. Rabine. 2000. An airborne scanning laser altimetry survey of Long Valley,
California. International Journal of Remote Sensing 21:2413–2437.
Holt, R. F. 2000. Inventory and Tracking of Old Growth Conservation Values for Landscape
Unit Planning.

96

Holt, R. F., D. J. MacKillop, and T. Braumandl. 2001. Definitions of old-growth in the MSdk
BEC unit in the Nelson Forest Region. Nelson Forest Region.
Holt, R. F., D. J. MacKillop, and T. Braumandl. 2002. Defining Old-Growth Forest in the
ICHWK1 BEC Variant in the Nelson Forest Region. Nelson Forest Region.
Holt, R., K. Price, L. Kremsater, A. MacKinnon, and K. Lertzman. 2008. Defining old
growth and recovering old growth on the coast: discussion of options:16.
Hopkinson, C., L. Chasmer, K. Lim, P. Treitz, and I. Creed. 2006. Towards a universal lidar
canopy height indicator. Canadian Journal of Remote Sensing 32:139–152.
Hyde, P., R. Dubayah, W. Walker, J. B. Blair, M. Hofton, and C. Hunsaker. 2006. Mapping
forest structure for wildlife habitat analysis using multi-sensor (LiDAR, SAR/InSAR,
ETM+, Quickbird) synergy. Remote Sensing of Environment 102:63–73.
Hyyppä, J., H. Hyyppä, D. Leckie, F. Gougeon, X. Yu, and M. Maltamo. 2008. Review of
methods of small‐footprint airborne laser scanning for extracting forest inventory data
in boreal forests. International Journal of Remote Sensing 29:1339–1366.
Isbell, F., V. Calcagno, A. Hector, J. Connolly, W. S. Harpole, P. B. Reich, M. SchererLorenzen, B. Schmid, D. Tilman, J. van Ruijven, A. Weigelt, B. J. Wilsey, E. S.
Zavaleta, and M. Loreau. 2011. High plant diversity is needed to maintain ecosystem
services. Nature 477:199.
Jenkins, J. C., D. C. Chojnacky, L. S. Heath, and R. A. Birdsey. 2003. National-Scale
Biomass Estimators for United States Tree Species. Forest Science 49:12–35.
Kane, V. R., J. D. Bakker, R. J. McGaughey, J. A. Lutz, R. F. Gersonde, and J. F. Franklin.
2010a. Examining conifer canopy structural complexity across forest ages and
elevations with LiDAR data. Canadian Journal of Forest Research 40:774–787.
Kane, V. R., R. J. McGaughey, J. D. Bakker, R. F. Gersonde, J. A. Lutz, and J. F. Franklin.
2010b. Comparisons between field- and LiDAR-based measures of stand structural
complexity. Canadian Journal of Forest Research 40:761–773.
Kareiva, P. 2011. Natural Capital: Theory and Practice of Mapping Ecosystem Services.
Oxford University Press.
Kayranli, B., M. Scholz, A. Mustafa, and Å. Hedmark. 2010. Carbon Storage and Fluxes
within Freshwater Wetlands: a Critical Review. Wetlands 30:111–124.

97

Keenan, R. J., G. A. Reams, F. Achard, J. V. de Freitas, A. Grainger, and E. Lindquist. 2015.
Dynamics of global forest area: Results from the FAO Global Forest Resources
Assessment 2015. Forest Ecology and Management 352:9–20.
Keränen, J., J. Peuhkurinen, P. Packalen, and M. Maltamo. 2015. Effect of minimum
diameter at breast height and standing dead wood field measurements on the accuracy
of ALS-based forest inventory. Canadian Journal of Forest Research 45:1280–1288.
Kneeshaw, D. D., and P. J. Burton. 1998. Assessment of functional old-growth status: A case
study in the Sub-Boreal Spruce zone of British Columbia, Canada. Natural Areas
Journal 18.
Kneeshaw, D., and S. Gauthier. 2003. Old growth in the boreal forest: A dynamic perspective
at the stand and landscape level. Environmental Reviews 11:S99–S114.
Knight, A. T., R. M. Cowling, M. Rouget, A. Balmford, A. T. Lombard, and B. M.
Campbell. 2008. Knowing But Not Doing: Selecting Priority Conservation Areas and
the Research–Implementation Gap. Conservation Biology 22:610–617.
Koukoulas, S., and G. A. Blackburn. 2004a. Quantifying the spatial properties of forest
canopy gaps using LiDAR imagery and GIS. International Journal of Remote Sensing
25:3049–3072.
Koukoulas, S., and G. A. Blackburn. 2004b. Quantifying the spatial properties of forest
canopy gaps using LiDAR imagery and GIS. International Journal of Remote Sensing
25:3049–3072.
Kremen, C., and R. S. Ostfeld. 2005. A call to ecologists: measuring, analyzing, and
managing ecosystem services. Frontiers in Ecology and the Environment 3:540–548.
Kremen, C., N. M. Williams, M. A. Aizen, B. Gemmill-Herren, G. LeBuhn, R. Minckley, L.
Packer, S. G. Potts, T. Roulston, I. Steffan-Dewenter, D. P. Vázquez, R. Winfree, L.
Adams, E. E. Crone, S. S. Greenleaf, T. H. Keitt, A.-M. Klein, J. Regetz, and T. H.
Ricketts. 2007. Pollination and other ecosystem services produced by mobile
organisms: a conceptual framework for the effects of land-use change. Ecology
Letters 10:299–314.
Lamlom, S. H., and R. A. Savidge. 2003. A reassessment of carbon content in wood:
variation within and between 41 North American species. Biomass and Bioenergy
25:381–388.

98

Lane, C. R., and E. D’Amico. 2010. Calculating the Ecosystem Service of Water Storage in
Isolated Wetlands using LiDAR in North Central Florida, USA. Wetlands 30:967–
977.
Lang, M. W., and G. W. McCarty. 2009. Lidar intensity for improved detection of inundation
below the forest canopy. Wetlands 29:1166–1178.
Langford, W. T., A. Gordon, L. Bastin, S. A. Bekessy, M. D. White, and G. Newell. 2011.
Raising the bar for systematic conservation planning. Trends in Ecology & Evolution
26:634–640.
Law, E. A., B. A. Bryan, E. Meijaard, T. Mallawaarachchi, M. J. Struebig, M. E. Watts, and
K. A. Wilson. 2017. Mixed policies give more options in multifunctional tropical
forest landscapes. Journal of Applied Ecology 54:51–60.
Lefsky, M. A., W. B. Cohen, S. A. Acker, G. G. Parker, T. A. Spies, and D. Harding. 1999.
Lidar Remote Sensing of the Canopy Structure and Biophysical Properties of
Douglas-Fir Western Hemlock Forests. Remote Sensing of Environment 70:339–361.
Lefsky, M. A., W. B. Cohen, G. G. Parker, and D. J. Harding. 2002. Lidar Remote Sensing
for Ecosystem StudiesLidar, an emerging remote sensing technology that directly
measures the three-dimensional distribution of plant canopies, can accurately estimate
vegetation structural attributes and should be of particular interest to forest,
landscape, and global ecologists. BioScience 52:19–30.
Lindenmayer, D. B., K. B. Hulvey, R. J. Hobbs, M. Colyvan, A. Felton, H. Possingham, W.
Steffen, K. Wilson, K. Youngentob, and P. Gibbons. 2012. Avoiding bio-perversity
from carbon sequestration solutions. Conservation Letters 5:28–36.
Lindenmayer, D., and M. A. McCarthy. 2002. Congruence between natural and human forest
disturbance: a case study from Australian montane ash forests. Forest Ecology and
Management 155:319–335.
Lombardi, F., M. Marchetti, P. Corona, P. Merlini, G. Chirici, R. Tognetti, S. Burrascano, A.
Alivernini, and N. Puletti. 2015. Quantifying the effect of sampling plot size on the
estimation of structural indicators in old-growth forest stands. Forest Ecology and
Management 346:89–97.
Lonsdale, D., M. Pautasso, and O. Holdenrieder. 2008. Wood-decaying fungi in the forest:
conservation needs and management options. European Journal of Forest Research
127:1–22.

99

Lubchenco, J. 1998. Entering the Century of the Environment: A New Social Contract for
Science. Science 279:491–497.
Luck, G. W., K. M. Chan, and C. J. Klien. 2012. Identifying spatial priorities for protecting
ecosystem services. F1000Research 1.
Luyssaert, S., E.-D. Schulze, A. Börner, A. Knohl, D. Hessenmöller, B. E. Law, P. Ciais, and
J. Grace. 2008. Old-growth forests as global carbon sinks. Nature 455:213.
MA. 2005. Millennium Ecosystem Assessment.
https://www.millenniumassessment.org/en/Condition.html#download.
Margules, C. R., and R. L. Pressey. 2000. Systematic conservation planning. Nature 405:243.
Martinuzzi, S., L. A. Vierling, W. A. Gould, M. J. Falkowski, J. S. Evans, A. T. Hudak, and
K. T. Vierling. 2009. Mapping snags and understory shrubs for a LiDAR-based
assessment of wildlife habitat suitability. Remote Sensing of Environment 113:2533–
2546.
Mascaro, J., M. Detto, G. P. Asner, and H. C. Muller-Landau. 2011. Evaluating uncertainty
in mapping forest carbon with airborne LiDAR. Remote Sensing of Environment
115:3770–3774.
Maxwell, S. L., T. Evans, J. E. M. Watson, A. Morel, H. Grantham, A. Duncan, N. Harris, P.
Potapov, R. K. Runting, O. Venter, S. Wang, and Y. Malhi. 2019. Degradation and
forgone removals increase the carbon impact of intact forest loss by 626%. Science
Advances 5:eaax2546.
McElhinny, C., P. Gibbons, and C. Brack. 2006a. An objective and quantitative methodology
for constructing an index of stand structural complexity. Forest Ecology and
Management 235:54–71.
McElhinny, C., P. Gibbons, C. Brack, and J. Bauhus. 2005. Forest and woodland stand
structural complexity: Its definition and measurement. Forest Ecology and
Management 218:1–24.
McElhinny, C., P. Gibbons, C. Brack, and J. Bauhus. 2006b. Fauna-habitat relationships: a
basis for identifying key stand structural attributes in temperate Australian eucalypt
forests and woodlands. Pacific Conservation Biology 12:89–110.
Meng, X., N. Currit, and K. Zhao. 2010. Ground Filtering Algorithms for Airborne LiDAR
Data: A Review of Critical Issues. Remote Sensing 2:833–860.
MFLNRORD. 1995. Biodiversity guidebook. Forest Practices code, Victoria, BC.
100

MFLNRORD. 2003. BC Forests Geographical Snapshot.
https://www.for.gov.bc.ca/scripts/hfd/pubs/hfdcatalog/index.asp.
MFLNRORD. 2017a. Community Forest Agreements. https://www.for.gov.bc.ca/hth/timbertenures/community/index.htm.
MFLNRORD. 2017b. Change Monitoring Inventory (CMI) - Ground Sampling Procedures.
https://www.for.gov.bc.ca/hts/risc/pubs/teveg/cmi_sampling_procedure_2017/CMI_
Ground_Sampling_Procedures_2017.pdf.
Milcu, A., J. Hanspach, D. Abson, and J. Fischer. 2013. Cultural Ecosystem Services: A
Literature Review and Prospects for Future Research. Ecology and Society 18.
Miller, J. D., S. R. Danzer, J. M. Watts, S. Stone, and S. R. Yool. 2003. Cluster analysis of
structural stage classes to map wildland fuels in a Madrean ecosystem. Journal of
Environmental Management 68:239–252.
Monfreda, C., N. Ramankutty, and J. A. Foley. 2008. Farming the planet: 2. Geographic
distribution of crop areas, yields, physiological types, and net primary production in
the year 2000. Global Biogeochemical Cycles 22:GB1022.
Mosseler, A., J. A. Lynds, and J. E. Major. 2003a. Old-growth forests of the Acadian Forest
Region. Environmental Reviews 11:S47–S77.
Mosseler, A., J. E. Major, and O. P. Rajora. 2003b. Old-growth red spruce forests as
reservoirs of genetic diversity and reproductive fitness. Theoretical and Applied
Genetics 106:931–937.
Mosseler, A., I. Thompson, and B. A. Pendrel. 2003c. Overview of old-growth forests in
Canada from a science perspective. Environmental Reviews 11:S1–S7.
Müller, J., and K. Vierling. 2014. Assessing Biodiversity by Airborne Laser Scanning. Pages
357–374 in M. Maltamo, E. Næsset, and J. Vauhkonen, editors. Forestry Applications
of Airborne Laser Scanning: Concepts and Case Studies. Springer Netherlands,
Dordrecht.
Næsset, E. 2002. Predicting forest stand characteristics with airborne scanning laser using a
practical two-stage procedure and field data. Remote Sensing of Environment 80:88–
99.
Næsset, E. 2011. Estimating above-ground biomass in young forests with airborne laser
scanning. International Journal of Remote Sensing 32:473–501.

101

Næsset, E., T. Gobakken, O. M. Bollandsås, T. G. Gregoire, R. Nelson, and G. Ståhl. 2013.
Comparison of precision of biomass estimates in regional field sample surveys and
airborne LiDAR-assisted surveys in Hedmark County, Norway. Remote Sensing of
Environment 130:108–120.
Næsset, E., and T. Økland. 2002. Estimating tree height and tree crown properties using
airborne scanning laser in a boreal nature reserve. Remote Sensing of Environment
79:105–115.
Naidoo, R., A. Balmford, R. Costanza, B. Fisher, R. E. Green, B. Lehner, T. R. Malcolm, and
T. H. Ricketts. 2008. Global mapping of ecosystem services and conservation
priorities. Proceedings of the National Academy of Sciences 105:9495–9500.
Naidoo, R., and T. H. Ricketts. 2006. Mapping the Economic Costs and Benefits of
Conservation. PLOS Biology 4:e360.
Nelson, E., G. Mendoza, J. Regetz, S. Polasky, H. Tallis, Dr. Cameron, K. M. Chan, G. C.
Daily, J. Goldstein, P. M. Kareiva, E. Lonsdorf, R. Naidoo, T. H. Ricketts, and Mr.
Shaw. 2009. Modeling multiple ecosystem services, biodiversity conservation,
commodity production, and tradeoffs at landscape scales. Frontiers in Ecology and
the Environment 7:4–11.
Ninan, K. N., and M. Inoue. 2014. Valuing forest ecosystem services: what we know and
what we don’t. Valuing Ecosystem Services.
Önal, H., and R. A. Briers. 2002. Incorporating spatial criteria in optimum reserve network
selection. Proceedings of the Royal Society of London. Series B: Biological Sciences
269:2437–2441.
Pearce, D. W. 2001. The Economic Value of Forest Ecosystems. Ecosystem Health 7:284–
296.
Penner, M., K. Power, C. Muhairwe, R. Tellier, and Y. Wang. 1997. Canada’s Forest
Biomass Resources: Deriving Estimates from Canada’s Forest Inventory. Canadian
Forest Service, Pacific Forestry Centre, Victoria, B.C.
Pesonen, A., O. Leino, M. Maltamo, and A. Kangas. 2009. Comparison of field sampling
methods for assessing coarse woody debris and use of airborne laser scanning as
auxiliary information. Forest Ecology and Management 257:1532–1541.
Pesonen, A., M. Maltamo, K. Eerikäinen, and P. Packalèn. 2008. Airborne laser scanningbased prediction of coarse woody debris volumes in a conservation area. Forest
Ecology and Management 255:3288–3296.
102

Polasky, S., E. Nelson, J. Camm, B. Csuti, P. Fackler, E. Lonsdorf, C. Montgomery, D.
White, J. Arthur, B. Garber-Yonts, R. Haight, J. Kagan, A. Starfield, and C. Tobalske.
2008. Where to put things? Spatial land management to sustain biodiversity and
economic returns. Biological Conservation 141:1505–1524.
Polasky, S., E. Nelson, D. Pennington, and K. A. Johnson. 2011. The Impact of Land-Use
Change on Ecosystem Services, Biodiversity and Returns to Landowners: A Case
Study in the State of Minnesota. Environmental and Resource Economics 48:219–
242.
R Development Core Team. 2018. R: The R Project for Statistical Computing. https://www.rproject.org/.
Racine, E. B., N. C. Coops, B. St-Onge, and J. Bégin. 2014. Estimating Forest Stand Age
from LiDAR-Derived Predictors and Nearest Neighbor Imputation. Forest Science
60:128–136.
Ramankutty, N., A. T. Evan, C. Monfreda, and J. A. Foley. 2008. Farming the planet: 1.
Geographic distribution of global agricultural lands in the year 2000. Global
Biogeochemical Cycles 22:GB1003.
Raudsepp-Hearne, C., G. D. Peterson, and E. M. Bennett. 2010. Ecosystem service bundles
for analyzing tradeoffs in diverse landscapes. Proceedings of the National Academy
of Sciences 107:5242–5247.
Reutebuch, S. E., H.-E. Andersen, and R. J. McGaughey. 2005. Light Detection and Ranging
(LIDAR): An Emerging Tool for Multiple Resource Inventory. Journal of Forestry
103:286–292.
Rodrigues, A. S. L., H. R. Akçakaya, S. J. Andelman, M. I. Bakarr, L. Boitani, T. M. Brooks,
J. S. Chanson, L. D. C. Fishpool, G. A. B. Da Fonseca, K. J. Gaston, M. Hoffmann, P.
A. Marquet, J. D. Pilgrim, R. L. Pressey, J. Schipper, W. Sechrest, S. N. Stuart, L. G.
Underhill, R. W. Waller, M. E. J. Watts, and X. Yan. 2004. Global Gap Analysis:
Priority Regions for Expanding the Global Protected-Area Network. BioScience
54:1092–1100.
Schirpke, U., F. Timmermann, U. Tappeiner, and E. Tasser. 2016. Cultural ecosystem
services of mountain regions: Modelling the aesthetic value. Ecological Indicators
69:78–90.
Schröter, M., and R. P. Remme. 2016. Spatial prioritisation for conserving ecosystem
services: comparing hotspots with heuristic optimisation. Landscape Ecology 31:431–
450.
103

Schwenk, W. S., T. M. Donovan, W. S. Keeton, and J. S. Nunery. 2012. Carbon storage,
timber production, and biodiversity: comparing ecosystem services with multi-criteria
decision analysis. Ecological Applications 22:1612–1627.
Seielstad, C. A., and L. P. Queen. 2003. Using Airborne Laser Altimetry to Determine Fuel
Models for Estimating Fire Behavior.
Shi, T., and S. Horvath. 2006. Unsupervised Learning With Random Forest Predictors.
Journal of Computational and Graphical Statistics 15:118–138.
Simonson, W. D., H. D. Allen, and D. A. Coomes. 2014. Applications of airborne lidar for
the assessment of animal species diversity. Methods in Ecology and Evolution 5:719–
729.
Snäll, T., J. Lehtomäki, A. Arponen, J. Elith, and A. Moilanen. 2016. Green Infrastructure
Design Based on Spatial Conservation Prioritization and Modeling of Biodiversity
Features and Ecosystem Services. Environmental Management 57:251–256.
Song, C., T. A. Schroeder, and W. B. Cohen. 2007. Predicting temperate conifer forest
successional stage distributions with multitemporal Landsat Thematic Mapper
imagery. Remote Sensing of Environment 106:228–237.
Song, C., and C. E. Woodcock. 2002. The spatial manifestation of forest succession in optical
imagery: The potential of multiresolution imagery. Remote Sensing of Environment
82:271–284.
Spies, T. A. 2004. Ecological Concepts and Diversity of Old-Growth Forests. Journal of
Forestry 102:14–20.
Spies, T. A., and J. F. Franklin. 1991. The Structure of Natural Young, Mature, and OldGrowth Douglas-Fir Forests in Oregon and Washington.
Spies, T. A., M. A. Hemstrom, A. Youngblood, and S. Hummel. 2006. Conserving OldGrowth Forest Diversity in Disturbance-Prone Landscapes. Conservation Biology
20:351–362.
Standish, J. T., G. H. Manning, and J. P. Demaerschalk. 1985. Development of biomass
equations for British Columbia tree species. Development of biomass equations for
British Columbia tree species.
Stutter, M. I., W. J. Chardon, and B. Kronvang. 2012. Riparian Buffer Strips as a
Multifunctional Management Tool in Agricultural Landscapes: Introduction. Journal
of Environmental Quality 41:297–303.
104

Swanson, M. E., J. F. Franklin, R. L. Beschta, C. M. Crisafulli, D. A. DellaSala, R. L. Hutto,
D. B. Lindenmayer, and F. J. Swanson. 2011. The forgotten stage of forest
succession: early-successional ecosystems on forest sites. Frontiers in Ecology and
the Environment 9:117–125.
Therneau, T. M., E. J. Atkinson, and M. Foundation. 2011. An Introduction to Recursive
Partitioning Using the RPART Routines. Page 67. Mayo Clinic, Rochester (MM).
Underhill, L. G. 1994. Optimal and suboptimal reserve selection algorithms. Biological
Conservation 70:85–87.
United Nations. 2018. The Convention on Biological Diversity.
https://www.cbd.int/sp/targets/.
Van Berkel, D. B., P. Tabrizian, M. A. Dorning, L. Smart, D. Newcomb, M. Mehaffey, A.
Neale, and R. K. Meentemeyer. 2018. Quantifying the visual-sensory landscape
qualities that contribute to cultural ecosystem services using social media and
LiDAR. Ecosystem Services 31:326–335.
Venter, O., W. F. Laurance, T. Iwamura, K. A. Wilson, R. A. Fuller, and H. P. Possingham.
2009a. Harnessing Carbon Payments to Protect Biodiversity. Science 326:1368–1368.
Venter, O., E. Meijaard, H. Possingham, R. Dennis, D. Sheil, S. Wich, L. Hovani, and K.
Wilson. 2009b. Carbon payments as a safeguard for threatened tropical mammals.
Conservation Letters 2:123–129.
Vepakomma, U., B. St-Onge, and D. Kneeshaw. 2008. Spatially explicit characterization of
boreal forest gap dynamics using multi-temporal lidar data. Remote Sensing of
Environment 112:2326–2340.
Wan-Mohd-Jaafar, W., I. Woodhouse, C. Silva, H. Omar, and A. Hudak. 2017.
MODELLING INDIVIDUAL TREE ABOVEGROUND BIOMASS USING
DISCRETE RETURN LIDAR IN LOWLAND DIPTEROCARP FOREST OF
MALAYSIA. Journal of Tropical Forest Science 29:465–484.
Watson, J. E. M., T. Evans, O. Venter, B. Williams, A. Tulloch, C. Stewart, I. Thompson, J.
C. Ray, K. Murray, A. Salazar, C. McAlpine, P. Potapov, J. Walston, J. G. Robinson,
M. Painter, D. Wilkie, C. Filardi, W. F. Laurance, R. A. Houghton, S. Maxwell, H.
Grantham, C. Samper, S. Wang, L. Laestadius, R. K. Runting, G. A. Silva-Chávez, J.
Ervin, and D. Lindenmayer. 2018. The exceptional value of intact forest ecosystems.
Nature Ecology & Evolution 2:599–610.

105

Watson, J. E. M., D. F. Shanahan, M. Di Marco, J. Allan, W. F. Laurance, E. W. Sanderson,
B. Mackey, and O. Venter. 2016. Catastrophic Declines in Wilderness Areas
Undermine Global Environment Targets. Current Biology 26:2929–2934.
Watts, M. E., I. R. Ball, R. S. Stewart, C. J. Klein, K. Wilson, C. Steinback, R. Lourival, L.
Kircher, and H. P. Possingham. 2009. Marxan with Zones: Software for optimal
conservation based land- and sea-use zoning. Environmental Modelling & Software
24:1513–1521.
Wendland, K. J., M. Honzák, R. Portela, B. Vitale, S. Rubinoff, and J. Randrianarisoa. 2010.
Targeting and implementing payments for ecosystem services: Opportunities for
bundling biodiversity conservation with carbon and water services in Madagascar.
Ecological Economics 69:2093–2107.
White, J. C., P. Tompalski, N. C. Coops, and M. A. Wulder. 2018. Comparison of airborne
laser scanning and digital stereo imagery for characterizing forest canopy gaps in
coastal temperate rainforests. Remote Sensing of Environment 208:1–14.
Wilson, K. A., M. Cabeza, and C. J. Klein. 2009. Fundamental Concepts of Spatial
Conservation Prioritization. Pages 16–27 Spatial conservation prioritization:
quantitative methods and computational tools. Oxford University Press, New York.
Wilson, S., R. Schuster, A. D. Rodewald, J. R. Bennett, A. C. Smith, and P. Arcese. 2018.
Prioritize diversity or declining species? Trade-offs and synergies in spatial planning
for the conservation of migratory birds. bioRxiv:429019.
Wing, B. M., M. W. Ritchie, K. Boston, W. B. Cohen, A. Gitelman, and M. J. Olsen. 2012.
Prediction of understory vegetation cover with airborne lidar in an interior ponderosa
pine forest. Remote Sensing of Environment 124:730–741.
Wing, B. M., M. W. Ritchie, K. Boston, W. B. Cohen, and M. J. Olsen. 2015. Individual snag
detection using neighborhood attribute filtered airborne lidar data. Remote Sensing of
Environment 163:165–179.
Wirth, C. 2009. Old-Growth Forests: Function, Fate and Value – a Synthesis. Pages 465–491
in C. Wirth, G. Gleixner, and M. Heimann, editors. Old-Growth Forests: Function,
Fate and Value. Springer Berlin Heidelberg, Berlin, Heidelberg.
Wirth, C., G. Gleixner, and M. Heimann. 2009a. Old-Growth Forests: Function, Fate and
Value – an Overview. Pages 3–10 Old-Growth Forests. Springer, Berlin, Heidelberg.
Wirth, C., C. Messier, Y. Bergeron, D. Frank, and A. Fankhänel. 2009b. Old-Growth Forest
Definitions: a Pragmatic View. Pages 11–33 in C. Wirth, G. Gleixner, and M.
106

Heimann, editors. Old-Growth Forests: Function, Fate and Value. Springer Berlin
Heidelberg, Berlin, Heidelberg.
Wulder, M. A., C. W. Bater, N. C. Coops, T. Hilker, and J. C. White. 2008. The role of
LiDAR in sustainable forest management - The Forestry Chronicle. http://pubs.cififc.org/doi/abs/10.5558/tfc84807-6.
Zhou, G., S. Liu, Z. Li, D. Zhang, X. Tang, C. Zhou, J. Yan, and J. Mo. 2006. Old-Growth
Forests Can Accumulate Carbon in Soils. Science 314:1417–1417.
Zimble, D. A., D. L. Evans, G. C. Carlson, R. C. Parker, S. C. Grado, and P. D. Gerard. 2003.
Characterizing vertical forest structure using small-footprint airborne LiDAR. Remote
Sensing of Environment 87:171–182.

107

6. APPENDICES

6.1. Data collection information
Free Growing Forestry Company has conducted the field data collection in Burns Lake
with teams of two members. Each team collects an average of 2 plots per day due to the
walking distance to plot, site conditions, and the amount of data collected (Table 6.1).

Table 6.1 List of attributes measured for each plot
Data Collected

Description

Tree # :
Species (2 Letter Code):

e.g. Pl =Lodgepole Pine, At= Trembling Aspen, Sx= Hybrid Spruce

Diameter:

DBH (cm)

Height:

Tree Length (m)

Loss Factor Information:

Tree Class; Conk; Blind Conk; Scar; Fork/ Crook; Frost Crack;
Mistletoe; Rotten Branch; Dead/ Br. Top; Root Rot Code; Insect
Code; Fire Code; and Blowdown Code

Live or Dead:
Standing or Fallen:
Crown Class (D, C, I, S):

D= dominant, C= Codominant, I= Intermediate, S= Suppressed

Site Tree Ages:

Age at DBH Counted - Field
Age at DBH Counted - Office

# of small tree (DBH<4cm):

Species code; Length class: 10-30cm, 31cm-1.3m, >1.3m

Stumps >= 4cm DIB and

Species code, frequency, DIB(cm), length(m), and %Sound

length <1.3m:

The steps for the data collection consisted in:


Identify a referential tree and mark it with tape and ink (Figure 1 a));

108



Find plot center with high precision GPS (Figure 1 b));



Delimitate the work sector. Each plot is divided into 8 working sector;



Each sector is sub-divided into 2.5 m and 5.64 m. From 2.5m circle, I measure trees
with DBH smaller than 4 cm (saplings) and stumps. From the 5.64 circle, I collected
the tree cores from the biggest specimen of the leading species;



In each sector, I obtained species, DBH, height, tree features (scars, crooks, forks,
broken top, etc), status (live or dead, standing or fallen), height to live crown,
competition (dominant, co-dominant, intermediate and suppressed), etc.



Trees on the ground are also measured if greater than 17.5 for spruce or 14.5 for pine,
and if wood is still sound. When logs are rotten, they are not measured.

109

Figure 6.1 Illustration of the a) reference tree, b) plot center with the high precision gps, and
c) tree core.


After measuring all tree down to 4 cm in each of the 8 sectors in the plot, I selected
the largest specimen of the leading and second leading species, extract on core from
each (Figure 1 c), and count the growth rings in the plot (they are counted again in the
office);



Cored trees are located by their bearing and distance to the plot center;



A second point is collected from the plot center with the high precision GPS before
leaving the plot to improve the location accuracy (Figure 1 b));

110



The plot location is signalized with tape, and path directions are transcribed into the
document;

6.2. Allometric Equations for plot-level estimates of old-growth attributes:
6.2.1. Volume-Height, DBH
Volume equations (Standish et al. 1985, Penner et al. 1997) are based on the DBH
and height.
(2)

𝑣 = 𝑝1 × 10−5 × 𝑑 × 𝑒 𝑝2 × ℎ × 𝑒 𝑝3

Where
v = tree volume (m3)
h = tree height (m)
d = diameter at breast height (cm)
p1,p2,p3 = parameters of volume calculation, Parameters are below (table 6.2)

Table 6.2 Volume parameters (Penner et al. 1997, Standish et al. 1985)
Species Code

Species

Latin Name

Volume parameters (m3)
p1

p2

p3

Bl

SubalpineFir

Abies,las

5.106002228

1.87293

0.998274

Alder

GreenAlder

Alnus,crispa

NA

NA

NA

Ep

Birch

Betula,pap

3.60460765

1.90956

1.0525

NA

NA

genx,x

5.106002228

1.87293

0.998274

Sx

Spruce

Picea,gl*eng

5.079336672

1.85859

1.00779

Sb

BlackSpruce

Picea,mariana

5.079336672

1.85859

1.00779

Pl

lodgepole

Pinus,contorta

4.47194033

1.82276

1.10812

111

Acb

BalsamPopular

Pop,balsamifera

2.246823719

1.73518

1.35601

At

Aspen

Pop,trem

3.804275847

1.89476

1.05373

Fd

DouglasFir

Pseudotsuga,menziesii

4.139024528

1.74294

1.15641

Ww

Salix

Salix

NA

NA

NA

Hw

WesternHemlock

Tsuga,heterophyl

4.030574937

1.9429

0.990275

Dmaple

DouglasMaple

Acer,glabrum

NA

NA

NA

6.2.2. Biomass — Jenkins’s equation
Aboveground biomass is calculated based only on the value of DBH (Jenkins et al. 2003) in
the form of exponential curve. Component biomass, including foliage, root, stem bark and
stem wood, is calculated by the ratio of the component and the total aboveground biomass.
𝑎𝑏 = exp⁡(𝑝1 + 𝑝2 𝑙𝑛𝑑)

(3)

Where

ab = aboveground biomass
exp = exponential function
d = DBH
ln = log base e (2.718282)
p1,p2 = parameters of aboveground biomass (table 6.4)

𝑟𝑎𝑡𝑖𝑜 = ⁡exp⁡(𝑝1 + 𝑝2 /𝑑)

(4)

Where

112

ratio = ratio of component biomass to total aboveground biomass
exp = exponential function
d = DBH
p1,p2 = parameters of component biomass (table.3)

113

Table 6.3 Parameters aboveground biomass and component biomass ratio (Jenkins)

Species

Aboveground

Component Biomass

biomass(kg)

Foliage

Root

Stem bark

Stem wood

p1

p2

p1

p2

p1

p2

p1

p2

p1

p2

Subalpine Fir

-2.5384

2.4814

-2.9584

4.4766

-1.5619

0.6614

-2.098

-1.1432

-0.3737

-1.8055

Green Alder

-2.5384

2.4814

-4.0813

5.8816

-1.6911

0.816

-2.0129

-1.6805

-0.3065

-5.424

Birch

-1.9123

2.3651

-4.0813

5.8816

-1.6911

0.816

-2.0129

-1.6805

-0.3065

-5.424

NA

-2.5384

2.4814

-2.9584

4.4766

-1.5619

0.6614

-2.098

-1.1432

-0.3737

-1.8055

Spruce

-2.0773

2.3323

-2.9584

4.4766

-1.5619

0.6614

-2.098

-1.1432

-0.3737

-1.8055

Black Spruce

-2.0773

2.3323

-2.9584

4.4766

-1.5619

0.6614

-2.098

-1.1432

-0.3737

-1.8055

lodgepole

-2.5356

2.4349

-2.9584

4.4766

-1.5619

0.6614

-2.098

-1.1432

-0.3737

-1.8055

Balsam Popular

-2.22094

2.3867

-4.0813

5.8816

-1.6911

0.816

-2.0129

-1.6805

-0.3065

-5.424

Aspen

-2.22094

2.3867

-4.0813

5.8816

-1.6911

0.816

-2.0129

-1.6805

-0.3065

-5.424

Douglas Fir

-2.2304

2.4435

-2.9584

4.4766

-1.5619

0.6614

-2.098

-1.1432

-0.3737

-1.8055

Salix

-2.2094

2.3867

-4.0813

5.8816

-1.6911

0.816

-2.0129

-1.6805

-0.3065

-5.424

Western Hemlock

-2.5384

2.4814

-2.9584

4.4766

-1.5619

0.6614

-2.098

-1.1432

-0.3737

-1.8055

Douglas Maple

-1.9123

2.3651

-4.0813

5.8816

-1.6911

0.816

-2.0129

-1.6805

-0.3065

-5.424

114

6.2.3. Carbon
Carbon is calculated based on the volume (m3), the woody specific gravity (g/cm3)
(Jenkins et al. 2003) and the carbon content, which is generally around 50% (Lamlom and
Savidge 2003).

(7)

𝑐 = 𝑔 × 103 × 𝑣 × 𝑐𝑐 × %

Where
c = carbon (kg)
g = woody specific gravity (g/cm3)
v = volume (m3)
cc = carbon content (%), parameters are below (table.3)

Table 6.4 Woody specific gravity and carbon content
Species
Code

Species

Latin Name

Woody specific
gravity (g/m3)

Carbon
Content
(%)

Bl

Subalpine Fir

Abies,las

0.4

50.08

Alder

Green Alder

Alnus,crispa

0.4

50.08

Ep

Birch

Betula,pap

0.43

48.37

NA

NA

genx,x

0.4

50.08

Sx

Spruce

Picea,gl*eng

0.36

50.39

Sb

Black Spruce

Picea,mariana

0.38

50.39

Pl

lodgepole

Pinus,contorta

0.38

50.32

Acb

Balsam Popular

Pop,balsamifera

0.32

47.09

At

Aspen

Pop,trem

0.34

47.09

115

Fd

Douglas Fir

Pseudotsuga,menziesii

0.4

50.5

Ww

Salix

Salix

0.46

49.05

Hw

Western Hemlock

Tsuga,heterophyl

0.43

50.6

Dmaple

Douglas Maple

Acer,glabrum

0.43

49.64

6.1. Additional Informal For The Methods In Chapter 2:
Gaps’ size, area and shape differ greatly depending on the forest stage (White et al. 2018)
and thus could be used to differentiate forest succession. For this work, a LiDAR metric for
canopy cover using only the vegetation point >3m high and further normalized with ground
returns were meant to capture gap differences (STH4_Cov in Table 2.3). While this metric
does not measure the canopy gap directly, canopy closure depicts the differences in canopy
openness in different forest succession, especially when used together with other old-growth
attributes developed for this work, such as vertical complexity. We utilized the coefficient of
variation (CV) of ALS-derived tree heights. It has a high correlation with the number and
complexity of canopy strata or vertical complexity (Zimble et al. 2003). Old-growth forests
are expected to have higher complexity not only in the crown height but also in the
understory. Thus, a metric for vertical complexity was calculated in multiple strata in the
forest (0.2 – 1m, 1 – 2m, 2 – 3m, and >3m). The pipeline for these and other metrics is
depicted in the Figure 6.2.

116

Figure 6.2 Pipeline of LiDAR processing using LASTools from the raw LiDAR files (.las) to the raster outputs (.bil) used to create ecological
meaningful metrics for old-growth attributes.

117

Table 6.5 summarizes the correlation analyses done with a set of attributes indicated
by different authors as important old-growth attributes (McElhinny et al. 2006b, 2006a, Spies
et al. 2006, Bauhus et al. 2009). The idea of this correlation was to remove variables with a
low correlation with age. We found that nine out of the ten variables best correlated with age
were listed by Bauhus et al. (2009). Also, from the variables that we included in addition to
Bauhus’s list, only maximum height was among the ten best correlated. Early succession
species had a negative correlation with age, which is ecologically logical. However, it was
not among the ten best correlates. Figure 2.4 depicts the histograms of the ten variables best
correlated with age. Maximum height, vertical and horizontal complexity was the only
variable with distribution closer to normal. All other variables skewed towards small values.
In order to use more common statistical models, such as multilinear regression, I would have
to normalize the variable. We understand that such a procedure can induce a false conclusion.
Therefore, we opted to use the original data format and chose a robust statistical approach
that could deal with non-normal data distribution. Random forest is a powerful machine
learning technique that learns with the data input. The random nature of this model avoids
overfitting and generates robust classification and regression models. Forest succession
classification and the old-growth indices (Table 2.2) utilized the ten old-growth attributes
best correlated with age (Table 2.4).

Table 6.5 Correlation of old-growth attributes with stand age to select the best ten correlations of oldgrowth attributes with estimated stand age. CV% = Coefficient of variation, Std = Standard
Deviation, dbh = diameter at the breast height.

Old-Growth Attributes
Age

Correlation

p-value

Maximum Height (m)

0.670

4.34E-14

Above Ground Biomass (tons/ha)

0.662

1.12E-13

Vertical Complexity (CV% of height)

0.588

1.95E-10
118

Basal Area (m2/ha)

0.521

3.86E-08

Volume (m3/ha)

0.509

8.49E-08

Horizontal complexity (CV% of dbh)

0.409

2.91E-05

Special Features (count #)

0.317

1.49E-03

Late succession species (tons/ha)

0.311

1.84E-03

Dead standing trees (m2/ha)

0.284

4.58E-03

Coarse woody debris (m3/ha)

0.262

9.04E-03

Tree class (std)

0.225

2.62E-02

Early succession species (tons/ha)

-0.223

2.74E-02

Basal areas of large trees (m2/ha, dbh>40cm)

0.192

5.84E-02

Tree diversity (Shannon Index)

0.172

8.96E-02

Density of trees (#trees/ha)

0.172

8.98E-02

Density of small trees (# of trees dbh < 4cm/ha)

0.155

1.28E-01

119

6.2. Distribution of Old-growth attributes based on different old-growth definitions

Figure 6.3 Distribution of old-growth attributes per forest succession defined by age classes
(AGE_CLASS).

120

Figure 6.4 Distribution of old-growth attributes per forest succession defined by k.means
classification of all old-growth attributes, including AGE (KOGA+AGE).

121

Figure 6.5 Distribution of old-growth attributes per forest succession defined by k.means
classification of all old-growth attributes, except AGE (KOGA).

122

Figure 6.6 Distribution of old-growth attributes per forest succession defined by k.means all
proximity values from unsupervised random forest of old-growth attributes, including age
(RFOGA+AGE).

123

Figure 6.7 Distribution of old-growth attributes per forest succession defined by k.means all
proximity values from unsupervised random forest of all old-growth attributes, except AGE
(RFOGA).

124

Figure 6.8 Distribution of old-growth attributes against age as a continuous variable (all plot ages
were divided by the maximum age found during inventory and multiplied by five, since age
classification was divided into five classes.

125

Figure 6.9 Distribution of old-growth attributes against old-growth index OGA+AGE (sum
of all attributes).

126

Figure 6.10 Distribution of old-growth attributes against the old-growth index OGA (sum of
all attributes, except age).

127

6.3. Water value:

Figure 6.11 ArcMap pipeline for the development of a wetness index as a proxy for water provision.
In this model, wetter areas are expected to have higher potential for water provision.

6.4. Random forest results:
In Table 6.5, I listed the five most important prediction variables, LiDAR metrics, for
each old-growth model. For the model chosen and utilized in this study, the most important
variables were all height-derived metrics, where the maximum height, proxy for dominant
tree height, was the most important metrics for old-growth definition. The standard deviation
of height, appointed as one important metric for the differentiation between forest
successions, was also in between the five most important metrics.
Table 6.6 Ranking of importance of LiDAR metrics
First

Second

Third

Fourth

Fifth

AGE_CLASS

AHR_Std

STH4_Com

AHR_Max

VERCOMP H95PercT

KOGA+AGE

STH4_Den

AHR_Dns

H95PercT

AHR_Max

STH4_Cov

KOGA

STH4_Den

AHR_Dns

AHR_Max

H95PercT

STH4_Cov

RFOGA+AGE

AHR_Max

H95PercT

AHR_Std

AHR_Qva

STH4_Cov

RFOGA

AHR_Std

STH4_Cov

AHR_Dns

AHR_Avg

AHR_Qva

AGE

AHR_Max

AHR_Std

AHR_Qva

H75PercT

H95PercT

OGA+AGE

AHR_Max

AHR_Std

AHR_Qva

H75PercT

H95PercT

OGA

AHR_Max

AHR_Qva

H75PercT

H90PercT

AHR_Std

128

6.5. Fragmented versus contiguous reserves:
Planning unit selections resulting from simple objective functions often result in
solutions that are highly fragmented and widely dispersed, yet spatial aggregation of planning
units may be desirable for both ecological and management reasons. The ecological
justification for aggregation often relates to the ‘single large or several small’ (SLOSS)
debate (Diamond 1975), species-area relationship, and population viability. Researches
indicate that bigger reserves, more circular with a shorter distance between each other and
with habitat corridor links are better than otherwise (Diamond 1975). Determining the
strength of the aggregation of compactness effect is a subjective decision that can be usefully
visualized by trade-off curves (Beyer et al. 2016). Thus, different clumpiness levels were
visually tested to determine the clumpiness level that better approximates the current
OGMA’s design (Figure 6.10). Also, timber extraction is the primary management strategy in
the landscape. Thus, timber harvesting may occur in all areas outside OGMAs.

129

Figure 6.12 Test with multiple boundary length penalties to visually select between OGMAs designs
that are extremely fragmented or overly contiguous.

In Figure 6.10, I compare scenarios of increased clumpness (reduction in
fragmentation) and increase in area sizes with the current OGMAs network. I assumed that
doubling and tripling the current OGMAs areas would result in double and triple of
ecosystem services representation in order to get a baseline for comparison. The results in the
image are the percentage different between proposed scenarios and the baseline created here.
The scenarios demonstrated that more fragmented networks can target better the areas that
offer the greatest amount of ecosystem services. Thus, for low clumpness scenarios, all
ecosystem services provisioning were mostly above baseline, except for tree diversity and
water prioritization. Scenarios with medium and high clumpness were closer to baseline
values, which indicate a more linear relationship with area. Most scenarios have higher tree
diversity representation than the current OGMA network. In addition, area and feature

130

representation are not linear. In other words, doubling area of OGMA does not mean,
representing the double of ESs.

Figure 6.13 Comparison between three different levels of clumpness, from more fragmented to more
contiguous reserves networks, and three different reserves network sizes (1X= current OGMA size,
2X= double the current size, and 3X=triple).

131

A higher level of clumpiness limits the fragmentation and edge effect. However, low
levels of clumpiness (more fragmented networks) achieve higher representation of ecosystem
services and old-growth values than high clumpiness for the area. The boundary penalties for
high clumpiness forces “prioritizR” to achieve contiguous reserves’ networks at the expense
of selecting low-value pixels. As a consequence, contiguous networks have less ESs
representation with the same amount of set-aside planning units. Even with the smaller
feature representation than highly fragmented scenarios, an intermediate clumpiness is more
relevant in CCF for two reasons: it better corresponds to the current OGMA design in terms
of shape and area, and it allows for a reduced edge effect. Also, bigger OGMAs have a
higher chance to accommodate disturbances than small fragmented ones. Bigger reserves
allow for multiple different forest succession to occur within the same OGMA and reduce the
risk of succession being reset throughout by a single event such as a wildfire (Pickett and
Thompson 1978). The reduced edge-to-area ratio may incur more viable populations and
ecological processes, crucial for biodiversity and other ecosystem services provisioning. In
general, edges between priority areas and cleared or degraded areas are unfavourable
ecologically, although, for some species of conservation concern, edges are favourable
(Fahrig 2002). On the other hand, when highly clumped, OGMAs’ network might not be
separated from an appropriate geographic distance to protect species in multiple places,
which might increase the risk of extinction due to a catastrophic event (e.g. wildfires, disease
outbreaks)(Game et al. 2008).

132

6.6. PrioritizR scenarions for increasing areas

Figure 6.14 Comparison between current OGMAs' network and OGMAs designed to priority a) Oldgrowth, b) Water value, c) Tree diversity, d) carbon, e) all features. The representation of how many
times one planning unit was taken as priority is represented in f).

Managers are not allowed to reduce the current sizes of Old-growth management
areas (OGMAs), only shift to a different location or increase their sizes (MFLNRORD 1995).
133

Thus, I designed increments in OGMAs’ areas from 1-50% of the total landscape starting
from current OGMAs’ area (5.39% of the landscape or 3,541 ha) to observe possible
synergies and trade-off between ESs. Figure 6.15 depicts the comparison between the
increments in OGMAs’ areas starting from current OGMAs’ design and the five alternative
networks for ESs’ reserves developed in this study. Scenarios starting from current OGMAs’
design were called “Fixed” because the current OGMAs’ locations were unchanged. The
alternative scenarios called “Not-Fixed” because they were individually designed to represent
priority areas for the provision of each ES and them all together. The idea of this analysis is
to evaluate if, for future OGMA increment, the starting point affects the final ES provision.

134

Figure 6.15 Comparison between increase in OGMAs' areas starting from current OGMAs (Fixed)
and OGMAs designed to prioritize each individual ecosystem services and all features together (Notfixed). Old-growth prioritization is represented in figures from a) to c), with areas increase of 1%,
5%, and 10% respectively; carbon prioritization from d) to e), tree diversity from f) to h); water
values from j) to l), and all feature from m) to o).

135