fire-bark beetle interactions: exploring links between ... · responses after bole charring, but...

Fire-Bark Beetle Interactions: Exploring Links Between Fire Injury, Resin Defenses, and Beetle-Induced Mortality in Ponderosa Pine Forests

Daniel D. B. Perrakis

A dissertation submitted in partial fulfillment of the requirements of the degree of

Doctor of Philosophy

University of Washington

2008

Program Authorized to Offer Degree: College of Forest Resources

UMI Number: 3345747

Copyright 2008 by

Perrakis, Daniel D. B.

All rights reserved.

INFORMATION TO USERS

The quality of this reproduction is dependent upon the quality of the copy

submitted. Broken or indistinct print, colored or poor quality illustrations and

photographs, print bleed-through, substandard margins, and improper

alignment can adversely affect reproduction.

In the unlikely event that the author did not send a complete manuscript

and there are missing pages, these will be noted. Also, if unauthorized

copyright material had to be removed, a note will indicate the deletion.

®

UMI UMI Microform 3345747

Copyright 2009 by ProQuest LLC.

All rights reserved. This microform edition is protected against

unauthorized copying under Title 17, United States Code.

ProQuest LLC 789 E. Eisenhower Parkway

PO Box 1346 Ann Arbor, Ml 48106-1346

University of Washington Graduate School

This is to certify that I have examined this copy of a doctoral dissertation by


and have found that it is complete and satisfactory in all respects, and that any and all revisions required by the final examining committee have been made.

Chair of the Supervisory Committee:

(_J James K. Agee

Reading Committee:

C ) James K. Agee*

Robert I. Gara

Douglas G. Sprugel (

Donald McKenzie

Date: T W - 1 , 100$

In presenting this dissertation in partial fulfillment of the requirements for the doctoral degree at the University of Washington, I agree that the Library shall make its copies freely available for inspection. I further agree that extensive copying of the dissertation is allowable only for scholarly purposes, consistent with "fair use" as prescribed in the U.S. Copyright Law. Requests for copying or reproduction of this dissertation may be referred to ProQuest Information and Learning, 300 North Zeeb Road, Ann Arbor, MI 48106-1346, 1-800-521-0600, to whom the author has granted "the right to reproduce and sell (a) copies of the manuscript in microform and/or (b) printed copies of the manuscript made from microform."

Signature: ^ 7 / j L ^ ^ 5 ^

Date: j)l6.. 7 Z£>0$

University of Washington

Abstract

Fire-Bark Beetle Interactions: Exploring Links Between Fire Injury, Resin Defenses, and Beetle-Induced Mortality in Ponderosa Pine Forests


Chair of the Supervisory Committee: Professor Emeritus James K. Agee

College of Forest Resources

Following a century of fire exclusion, ponderosa pine {Pinus ponderosa

Laws., TP') forests of Crater Lake National Park, Oregon, are in a condition of poor

forest health. Prescribed fire has been used to restore old PP stands, but often with

counterproductive effects, including widespread bark beetle (Coleoptera: scolytidae)

attacks observed following fires. A series of experiments was designed to assess the

effects of fire on PP oleoresin flow (OF), believed to be the main defense against

beetle attacks. The restoration problem provided the experimental setting, while

findings were discussed in the context of a modern plant defense theory, the Growth-

Differentiation Balance (GDB).

Monitoring of large PP mortality after spring and fall burns suggested that

beetle activity and PP mortality were related to fire intensity and tree vigor. Both

mortality and OF were greatest in hot fall burns, lower in cool spring burns, and least

in controls; OF was also positively related to vigor. Additional experiments on pole-

sized individuals nearby tested the effects of mechanical injuries on resin. OF

responses after bole charring, but not trenching or pruning, were similar to those after

burning. The post-fire response consisted of a short-term (2-4 weeks) post-fire OF

decrease, followed by OF recovery by ~4 weeks and possible increase above pre-

treatment levels within ~8 weeks. Mid-summer OF remained elevated after fire for 2-

3 years in pole-sized PP and 4+ years in old-growth trees. There was also equivocal

evidence after burning for increases in resin volatile emissions.

Physiological influences on OF were discussed in the context of carbon

allocation. The GDB hypothesis appeared overly simplistic for explaining the short-

and long-term variability in OF in Pinus. Increased OF after fire appears to be a

generalized response to bole injury. This study suggests that carbon allocation to

resin defenses is complex and not easily assessed by simple OF measurements.

Commonly accepted ideas regarding PP resin defenses are invalid in the case of fire-

injury, as both OF and mortality increased among burned trees despite heightened

resin defenses. Induced or enhanced attraction of beetles following fire is suggested

as the main mechanism for explaining post-fire beetle responses.

TABLE OF CONTENTS

Page

List of Figures iii List of Tables iv

1. Introduction and Background 1.1. Introduction 1

1.1.1. Context and project history 1 1.1.2. Fire management at Crater Lake... 3

1.2. Literature review 6 1.2.1. Bark beetle ecology and fire ecology 6 1.2.2. Defense physiology 1: oleoresin production and composition 10 1.2.3. Defense physiology 2: resin response to wounding 15 1.2.4. Physiology of resin defenses: challenges 17 1.2.5. Growth-differentiation balance: a framework for

understanding resin response? 18 1.2.6. Fire effects on tree structures and carbohydrate relations 22

1.3. Experimental questions and hypotheses 26

2. Experimental Methods and Results 2.1. Experiment 1: Mortality and vigor in spring- and fall-burned trees 29

2.1.1. Objectives 29 2.1.2. Methods 29 2.1.3. Data analysis 31 2.1.4. Results 32

2.2. Experiment 2: Resin flow monitoring in spring- and fall-burned trees 38 39


2.3. Experiment 3: Seasonal resin flow and fire surrogates 45 2.3.1. Objectives 45 2.3.2. Methods 48 2.3.3. Data analysis 55 2.3.4. Results 59

2.4. Experiment 4: Isolated bole charring and resin flow at different heights 71


2.5. Experiment 5: Fire effects on resin chemistry 85

i

2.5.1. Objectives 85 2.5.2. Methods and data analysis 85 2.5.3. Results 89

2.6. Experiment 6: Methods testing between resin sampling techniques 93 2.6.1. Objectives 93 2.6.2. Methods 94 2.6.3. Data analysis 97 2.6.4. Results 98

3. Discussion 3.1. Old ponderosa pines at Crater Lake NP 101

3.1.1. Crown vigor and radial growth 101 3.1.2. Burning treatments 102 3.1.3. Resin flow monitoring at Crater Lake 104

3.2. Fire injury, surrogate treatments, and resin flow 106 3.2.1. Burning and bole charring treatments (Experiments 3, 4) 107 3.2.2. Pruning and root trenching treatments (Experiment 3) I l l

3.3. Carbon allocation, GDB, and response to wounding 114 3.4. Resin chemistry and host-beetle interactions 117 3.5. Beetle evolution, resin volatiles, and kairomones 122 3.6. Methodological considerations 124 3.7. Summary, management recommendations, and conclusions 126

References 132

Appendices Appendix 1: Latin and common names of species used in text 144 Appendix 2: Variations in surface fuels following burning 145

A2.1 Objectives, Methods, and Data analysis 145 A2.2 Results 146 A2.3 Discussion 149 A2.4 References 152

Appendix 3: Use of automatic dendrometers 155 A3.1 Dendrometer overview 155 A3.2 Methods 157 A3.3 Problems and challenges encountered 159 A3.4 Results and discussion 162 A3.5 References 167

Appendix 4: Southern Oregon Palmer Drought Severity Index values 169

u

LIST OF FIGURES

Figure Number Page

1. Conceptual Relationships between Resource Availability, Net Assimilation, Relative Growth, and Secondary Metabolism 21

2. Hypothesized Effects of Fire Injury on Crown, Bole, and Roots 25 3. Greater Crater Lake NP Study Area Overview 32 4. Crater Lake: Mortality by Crown Class and Treatment 34 5. Crater Lake: Mortality by Year and Treatment 36 6. Crown Vigor Class by Rings/cm 37 7. Resin Flow by Year and Treatment 44 8. Sun Pass Study Area Photograph 49 9. Increment Core Dissection for 13C Analysis 54 10. Sun Pass Resin Flow, Control Samples 62 11. Sun Pass Resin Flow Boxplot, All Years and Samples 64 12. Bole Charring Experiment: Resin Sampling 74 13. Bole Charring Experiment: Char Treatment 75 14. Photograph of Turpentine Beetle Attacks in Sampling Scars 78 15. Turpentine Beetle Attacks in Sampling Scars: Proportions 79 16. Bole Charring: Resin Volume by Height, Time, and Treatment 80 17. Bole Charring: Resin Volume Means 83 18. Monoterpene Proportions by Treatment and Site 92 19. Photograph of Resin Sampling Methods 96 20. Xylem Area Analysis of Resin Sampling Methods 100 Al. Changes to Fuel Loading 147 A2. Changes to Fuelbed Components 148 A3. Photograph of Installed Dendrometer.. 158 A4. Growth Correlations Between Dendrometers and Increment Cores 164 A5. Dendrometer Trace Examples 166-7

i i i

LIST OF TABLES

Table Number Page

1. Spring/Fall Burning: Ponderosa Pine Mortality 33 2. Spring/Fall Burning: Logistic Regression Model Parameters 35 3. Post-Hoc Comparisons, Crown Vigor 38 4. Spring/Fall Burning: Resin Flow Analysis 43 5. Spring/Fall Burning: Resin Flow Change Analysis 44 6. Fire Surrogates: Resin Analysis 1 Parameters 61 7. Fire Surrogates: Resin Analysis 2 Parameters 63 8. Fire Surrogates: Resin Analysis 3, Step 1 Parameters 65 9. Fire Surrogates: Resin Analysis 3, Step 2 Parameters 65 10. Fire Surrogates: Resin Analysis 3, Step 3 Parameters 66 11. Fire Surrogates: Resin Analysis 4 Parameters 68 12. Fire Surrogates: Resin Analysis 5, Step 1 Parameters 68 13. Fire Surrogates: Resin Analysis 5, Step 2 Parameters 69 14. Fire Surrogates: Treatment effects on BAI 70 15. Bole Charring: Pre-Treatment Analysis 81 16. Bole Charring: Resin Change Analysis, Full Factorial. 82 17. Bole Charring: Resin Change Analysis, Parameters and Estimates 83 18. Resin Chemistry: Analysis of Monoterpene Proportions 91 19. Resin Chemistry: Monoterpene Ranks and Proportions 91 20. Methods Comparison: Parameters 99 Al. Latin and Common Species Names 144 A2. Fuel Loading Analysis 147 A3. Palmer Drought-Severity Index Values 169

IV

ACKNOWLEDGEMENTS

This project would not have been completed without the generous support and assistance of many individuals. Back in 2003, Michael Keim and Rob Banes helped with the initial resin sampling and tinkering with the dendrometer apparatus. Dallas Anderson was an invaluable assistant from 2004 to 2006, going well beyond the call of duty in terms of working alone, working on weekends and holidays, and quietly accomplishing the tasks at hand while I panicked. Carson Sprenger eagerly ventured to the field site in mid-winter and competently assembled the analysis of the increment cores from Crater Lake. Measurement of the Sun Pass ring widths was accomplished cheerfully by Tania Taipale. I greatly appreciated the company and advice of Phil Monsanto, my fellow lab-mate, who assisted in many ways in both the field and the office, not least of which was moral support. Assistance with various portions of the analysis was provided by Mike Keim, as well as my father, Stylianos Perrakis. Crater Lake National Park employees Michael Murray and Scott Girdner offered support and assistance with the logistics of working in the park. Ed DeBlander did the same for the Oregon Department of Forestry, and provided the fire crew who lit the Sun Pass burn treatment. ODF forester Steve Jones was invaluable in assisting us with the bole charring treatment, and made sure we didn't start a wildfire (which we nearly did). Matt Dahlgreen from the Wenatchee National Forest assisted with site selection in the Entiat area, while Deb Shepherd (unknowingly, at the time) provided the same assistance on the Winema NF. USFS entomologist Andy Eglitis happily ventured out to the Crater Lake site every fall to monitor for new beetle activity in the burned area; this work did not go unnoticed.

University of Washington faculty Bob Gara, Doug Sprugel and Don MacKenzie assisted as part of my committee, guiding this project steadily towards completion and offering helpful comments on draft versions of this document. Renata Burn gladly offered advice on the arcane topic of resin chemistry analysis. Perhaps greatest of all, I am greatly thankful to the support and kindness of Jim Agee, my chair, who encouraged me throughout this project, with advice, financial backing, and ideas. I particularly appreciated his support and recommendation when I applied for a job position in Canada, knowing that it would delay my graduation.

On the personal front, my parents, Drs. Phyllis and Stylianos Perrakis, offered continual encouragement over the many years of work. My friends and skiing partners Michell daLuz Barry, Ralph Bodenner, Mike Keim, Crystal Raymond, Tonja Campbell, Reid Holmes and particularly Joshua Stern generously offered me a place to stay during my many visits to Seattle after I moved to Canada. Laura Mercer, my lovely partner, has been the most recent victim of this affliction, and her understanding and support over the past two years have been wonderful.

Finally, I would like to thank my current employer, Parks Canada Agency, and in particular managers Gail Fitzmartyn and Salman Rasheed for granting me the flexibility and leave from work that allowed me to finish.

-DP, December 2008

v

1

1. Introduction and Background

1.1. Introduction

1.1.1. Context and project history

Fire is recognized as a defining disturbance agent in dry forests of the western

US. The fire exclusion policy has been shown to be flawed, and active management

treatments are frequently applied to forest lands as part of restoration and community

protection efforts. Such treatments, including prescribed burning and mechanical fuel

reduction, have often appeared promising in the short term. However, long-term success

with this strategy remains uncertain, and treatment details are still being adjusted in many

areas to match restoration objectives. Such has been the case in the ponderosa pine

forests of Crater Lake National Park, Oregon.

At Crater Lake, fire restoration efforts have faced recurring problems related to

post-burn insect outbreaks, undermining the stand-level objectives. The main difficulties

lie with poor forest health conditions and low individual tree resilience. Prescribed burns

designed to maintain dominant trees have nonetheless resulted in high mortality among

such trees in forests being restored. For those individuals that survive the treatments in

the short term, weakened defenses and complex interactions with insects and pathogens

make long-term survival uncertain. Despite considerable study, there are still many

knowledge gaps relating to physiological interactions between fire injuries, tree defense

mechanisms, insect host-selection, and recovery through time.

2 Bark beetles1 are known to interact in various ways with fire (McCullough et al.

1998), although some of the interaction mechanisms are only beginning to be understood.

In ponderosa pine forests, attacks by insects such as Dendroctonus beetles appear to

increase after prescribed burns or wildfires (Miller and Keen 1960, Ganz et al. 2003,

McHugh et al. 2003, Perrakis 2004, Breece et al. 2008). Burning treatments are typically

patterned after historic fire effects and designed to protect dominant ponderosa pines

from competing fire-intolerant vegetation. High beetle-related mortality after fire in these

trees is therefore a critical management issue in many areas (Kolb et al. 2007).

Fire disturbances, beetles, and trees all interact in these forests, producing a

system that has proved challenging to study. A comprehensive understanding of the

disturbances and responses necessitates merging literature from the fields of forest

ecology, entomology, and at the individual tree and tissue level, plant physiology and

even organismal and molecular biology. Interest in understanding these systems and

interactions is not new (e.g. Miller and Keen 1960, Wagener 1961). Early work

conducted decades ago yielded important breakthroughs in many areas, especially in the

areas of bark beetle host-selection mechanisms and pheromones. However, the search for

broad similarities between different species (of both beetles and host trees, see Wood

1972, Christiansen et al. 1987) has revealed some fundamental differences in interaction

mechanisms; despite researchers' hopes, bark beetle-host tree relationships are not all

created equal. Furthermore, differences in research methods, overly confident and

1 Common names are used for most biota in this text. A key to the Latin and common names for all species mentioned is provided in Appendix 1.

3 sweeping conclusions, and inconsistent use of terminology have led to confusion and

interpretation errors.

Fire restoration of ponderosa pine forests in Crater Lake National Park began in

the late 1970s. Following a few years of experimental prescribed burning, problems with

post-fire tree mortality attributed to beetles were identified in the mid 1980s (Thomas

1982, Thomas and Agee 1986). A few research projects attempted to answer some of the

larger questions related to mortality mechanisms. These studies took the form of

empirical approaches based on comparatively simple hypotheses (e.g. Swezy and Agee

1991), but did not adequately explain interactions between prescribed fire and beetle

attacks. For the present study, I attempted a more focused approach, using a longer time

scale and multiple physiology-based experiments informed by emerging autecology

theories. The purpose of this project was to examine fire effects on bark beetle

susceptibility from a defense physiology perspective, with a number of related

experiments examining ponderosa pine mortality after fire, resin defenses, and the

physiological links between the two.

1.1.2. Fire management at Crater Lake

The modern era of prescribed burning at Crater Lake National Park, Oregon,

began over 25 years ago (Swezy 1988). Restoration efforts, including prescribed fire,

were initiated after a commissioned fire history study suggested that the park's ponderosa

pine stands were at risk of disappearing completely; the absence of periodic fire

promoted encroachment by white fir and other species, which were then able to

4 outcompete the ponderosa pines (McNeil and Zobel 1980). The southeastern region of

the park consists of old-growth mixed-conifer forests dominated by old ponderosa pines,

with a historic low-severity fire regime (McNeil and Zobel 1980). Such stands are

representative of widely distributed mixed-conifer forests elsewhere in Oregon and

California. At Crater Lake, no fires burned in these forests between park establishment in

the early 20th century until prescribed burning began in the late 1970's. Lightning-caused

or accidental human ignitions during that time were all promptly extinguished. Goals of

the modern fire program include killing post-settlement understory trees, maintaining the

dominant pine cohort, and reducing fuel loads to prevent crown fire initiation. In the past

decade, these objectives have been refined and quantified, and now include forest

structure and fuels specifications designed to meet restoration timelines and post-

treatment wildfire intensity targets (Anonymous 2002).

Following the first decade of burning projects, the aforementioned problems with

the burning treatments became apparent. Canopy-dominant ponderosa and sugar pines

were dying in the first few years following these fires. Mortality of these trees continued

for several years after burning and appeared to be partly caused by bark beetle attacks,

especially the western pine beetle (attacking ponderosa pine) and mountain pine beetle

(attacking sugar pine) (Thomas and Agee 1986, Swezy and Agee 1991, Agee 2003). A

recent study on seasonal burning at Crater Lake observed a similar response, noting

higher mortality of large ponderosa pines following hot fall burns than following cool

spring burns, continued mortality each year after prescribed fire treatments, and post-fire

beetle attacks playing a dominant role in ponderosa pine mortality (Perrakis 2004,

5 Perrakis and Agee 2006). A major component of that study examined ponderosa pine

resin defenses, seeking to identify whether fire might reduce resin flow or pressure; no

evidence of such a mechanism was detected, with many burned trees showing stronger

resin defenses than unburned controls one year after treatment. These findings suggested

some notable inconsistencies in the state of knowledge of fire-beetle interactions. In the

absence of an obvious restoration strategy, and facing a lack of understanding of the

processes at work, fire restoration in the Crater Lake panhandle was indefinitely

suspended (M. Murray, Crater Lake NP Terrestrial Ecologist, personal communication,

2007). These factors provided the impetus for the current project.

6 1.2. Literature Review

1.2.1. Bark beetle ecology and fire ecology

Bark beetle life history characteristics and ecology are well described elsewhere

in the literature (see Miller and Keen 1960, Rudinsky 1962, Wood 1972,1982,

Christiansen et al. 1987, Schowalter and Filip 1993). A brief synopsis of a typical

western pine beetle attack sequence will be useful for the discussion that follows. Wood

(1972) defined the four stages of attack as dispersal, host selection, concentration, and

establishment.

Dispersal is accomplished by flight and appears temperature-dependent (Gara and

Vite 1962, Peck et al. 1997), although little is known about distances covered or the

duration of travel by individual beetles. Host selection (discussed below) is believed to be

accomplished via either random landings independent of species and host suitability, or

by chemotaxis - attraction to chemical odors released by suitable hosts. Beetles also use

visual cues (tree silhouettes) to orient towards suitable trees (Strom et al. 2001). Hosts

consist of large (40+ cm dbh), old, low-vigor ponderosa pines. Upon landing on a

suitable host, beetles immediately begin feeding and nesting activities in the bark. This

activity initiates the concentration phase, whereby powerful attractant pheromones are

released that attract other beetles to the focus tree in a mass attack. During the

establishment phase, beetles dig egg galleries and oviposit in the phloem. In the spring,

emerging larvae feed on the phloem (briefly) and outer bark before emerging and

dispersing, completing the life cycle. The host tree is killed due to phloem girdling from

7 the larval galleries as well as disruption of water transport by blue-stain fungi (symbiotic

with the beetles) that grow in the sapwood tracheids (Miller and Keen 1960; Wood

1972). Although many of these steps are similar for other species of Scolytidae, others

are not, and thus inference to the Crater Lake-ponderosa pine case from studies on

different host or insect species should be viewed with some caution.

Initial host-selection by bark beetles therefore occurs via one of two methods:

random dispersal, or primary attraction (Wood 1982). The latter refers to the

phenomenon of beetles being attracted from afar to chemical cues emitted by potential

host trees (kairomones; Wood 1982), and was first observed and suggested many decades

ago (Pearson 1931, according to Wood 1982). Ponderosa pine kairomones have been

confirmed to attract red turpentine beetles (Hobson et al. 1993, Erbilgin and Raffa 2000),

although attacks by this species are usually not fatal (Furniss and Carolin 1977, McHugh

and Kolb 2003, but see Ganz et al. 2003). Kairomone effects have been demonstrated

between many other host tree and beetle species, and studies have shown hints of

kairomonal effects between ponderosa pine and western pine beetle under certain specific

conditions (Goheen et al. 1985). One of the constituents of ponderosa pine resin, the

monoterpene compound myrcene, was identified early on as a powerful co-attractant

when its vapors were combined with beetle pheromones exo-brevicomin and frontalin

(Bedard et al. 1969). Although a widely-cited study in the 1980's found no evidence of

attraction between beetles and freshly cut wood samples (Moeck et al. 1981), more recent

work suggests that western pine beetles show antennal responses to a variety of host and

non-host volatiles (Shepherd et al. 2007), giving renewed credibility to the primary

8 attraction hypothesis. At this date, primary attraction remains unconfirmed as a host

selection mechanism for this species, although the opinions of entomologists remain

divided on the subject.

Excluding the possibility of primary attraction, the default host-selection

mechanism is random landing, independent of tree species or host suitability (Wood

1972, Raffa et al. 1993). In Sierra Nevada mixed-conifer forest, Moeck et al. (1981)

calculated landing rates of one western pine beetle per tree per day, a sufficiently high

visitation rate to ensure initial attack on all trees in a stand within a few years. A

modeling experiment also confirmed that chance landings among a mixed stand of

potential hosts and non-hosts would be sufficiently effective as a selection method for

western pine beetles (Byers 1996). None of these studies specifically targeted recently

burned stands. In general, initial host selection has proven to be a challenge to study (due

to the rarity of observing adult beetles in flight, and the confounding effects of attraction

pheromones following establishment), leading some entomologists to believe that the

lack of evidence for primary attraction merely reflects the failures of researchers thus far

to design conclusive experiments on the topic.

As suggested above, numerous studies have observed elevated conifer mortality

attributed to bark beetles following fire, compared with unburned controls. However, the

mechanism behind this mortality increase is not obvious, with at least two obvious

possibilities. The first possibility is an elevated number of initial beetle attacks, and no

reduction in tree defenses; a second possible mechanism is the converse - no increase in

9 beetle attacks, but a reduction in tree defenses. The two mechanisms might also occur

together: both increased initial attacks and reduced host tree defenses.

The increased initial attacks hypothesis assumes that primary attraction is either

induced or augmented by non-lethal fire effects. However, even less is known about this

possible interaction than about primary attraction alone, although anecdotal evidence

from forest managers supports it (A. Eglitis, US Forest Service Entomologist, personal

communication; Lorraine Maclauchlan, British Columbia Ministry of Forests

Entomologist, personal communication). Since no studies have directly studied primary

attraction of western pine beetles to fire-damaged ponderosa pines, it cannot be either

assumed to occur nor entirely ruled out as a possible explanation for increased post-fire

pine mortality. Other kairomone effects might also be responsible for western pine beetle

attraction, such as pheromones from sympatric insects (e.g. turpentine beetles) already

present in host trees. Difficulties in experimental design •associated with this type of

investigation may have deterred their occurrence.

The second explanation for increased host mortality is a reduction in host tree

defenses. This mechanism assumes that regardless of whether initial attacks are

augmented by fire damage to trees or neutral to it, the rate of beetle attack success will be

increased on fire-damaged trees. This mechanism has received increased attention

recently in the context of fire effects, and was explored in greater detail in the course of

this project.

10 1.2.2. Defense physiology 1: oleoresin production and composition

The primary conifer defense against bark insects is from oleoresin (or simply

resin, Phillips and Croteau 1999). Resin is considered a secretion, or a metabolic end-

product substance that is not remobilized for other purposes. Accordingly, its only

purposes appear to be defense against pathogens or desiccation, or wound healing (Fahn

1979, Himejima et al. 1992, Kozlowski and Pallardy 1997b). Higher resin exudation

volume is associated with greater beetle resistance in ponderosa and other pines, other

factors being equal (Smith 1975, Hodges et al. 1979, Smith 2000, Strom et al. 2002,

Boyle et al. 2004, Wallin et al. 2008). Resistant trees repel beetles by physically ejecting

or smothering them in resin, through resin toxicity, or by disrupting nesting activities

(construction of galleries, ovipositing, etc.) and pheromone emittance, preventing mass

attack (Miller and Keen 1960, Smith 1975, Raffa et al. 1993, Smith 2000). More resistant

ponderosa pines therefore tend to have greater overall resin flow volume, resin flow that

persists for longer following wounding, and particular resin chemistry traits, as discussed

below.

Techniques for measuring pine resin defenses are a topic of early and recent

debate. Older studies focused on oleoresin exudation pressure (OEP), the pressure of the

resin inside the canals (Vite 1961, Vite and Wood 1961, Mathre 1964). However,

subsequent research has questioned the predictive value of OEP (Stark 1965, Lorio 1994,

P. Lorio, personal communication 2001), and the method of choice in recent times has

been to directly measure the volume or mass of oleoresin flow (sometimes termed

oleoresin exudation flow, or OEF). Resin flow is typically assessed by funneling resin

11

that emerges from drilled or cut bole wounds into test tubes over a certain period of time,

such as 24 hours (e.g. Feeney et al. 1998, Strom et al. 2002, Wallin et al. 2003, Perrakis

2004, Wallin et al. 2008).

The physiology of resin production and storage are important factors in its

defensive properties. In most pines, resin is produced and stored in pre-formed networks

of sapwood and phloem canals (Busgen and Munch 1929, Bannan 1936, Lin et al. 2002).

The resin canals consist of an interconnected elongated network in which epithelial

parenchyma cells form the canal walls. Resin is synthesized in epithelial cells and

secreted into the intercellular space (the resin canal), where it is stored under

considerable pressure (Busgen and Munch 1929, Vite 1961). Resin pressure appears

strongly and positively correlated with cellular turgor pressure and soil moisture

availability (Vite 1961, Hodges and Lorio 1971). Functional canals in pines have been

observed to connect different tissue types and structures within a tree, and individual

epithelial cells can live for several years; thus, canals can be continuous between vascular

tissues and needles, or between separate growth rings within the xylem (Lapasha and

Wheeler 1990, Lin et al. 2002). However, chemical composition differences have been

found in the resin properties between different morphological structures, such as between

cortex, sapwood and needle resin in pines (Latta et al. 2000, Smith 2000).

When a duct is severed, such as during bark beetle attack, the internal pressure

forces the resin out at the wound site. The epithelial cells expand and continue exerting

pressure on the resin as it is depleted, collapsing the duct in the process. After the wound

has sealed (via monoterpene volatization and crystallization of resin acids; see below),

12 the epithelial cells slowly renew the resin supply, returning the ducts to their full

cylindrical shape (Busgen and Munch 1929, Fahn 1979).

The primary structures producing resin in pine sapwood are vertical canals

(Barman 1936, Blanche et al: 1992). These can range from a few millimeters in

continuous length to much longer, up to a meter in some species (Busgen and Munch

1929, Barman 1936, Werker and Fahn 1969). Bannan (1936) suggested that resin canals

in Pinus species are mostly vertical, sometimes radial, and occasionally (rarely)

tangential (or axial), thus forming an extensive 3-dimensional network in the xylem

throughout which resin can be produced and relocated. Lin et al. (2002) also suggested

that both vertical and horizontal canals are present in pine phloem, but no distinctions are

made between species nor between radial or tangential directions. Fahn (1979), working

on P. halepensis, denied the existence of tangential canals and suggested that the

networks of interconnected canals were entirely two-dimensional in nature, including

vertical and radial canal directions, and therefore did not constitute a true network across

the entire volume of sapwood or phloem tissues. It is very likely that inter-specific

differences exist in resin canal architecture, despite the generalizations (by genus or even

family) suggested in some of the older studies (e.g. Busgen and Munch 1929, Bannan

1936). Little information is available on the specific resin canal characteristics of

ponderosa pine. More recent observations of large within-tree variation in resin flow

(such as on different sides of a single tree) support the notion of resin canals being

primarily interconnected in vertical and radial orientations, and absent in tangential

13 directions (e.g. Lombardero et al. 2000, 2006, Perrakis 2004, D. Wood, personal

communication 2001; C. Nelson, personal communication 2005).

Chemically, resin is composed of resin acids (diterpenes, C20 molecules)

commonly known collectively as rosin, dissolved in a liquid mixture of monoterpenes

(C10) and sesquiterpenes (C15). The latter two components constitute natural turpentine,

which is an easily volatized solvent (Phillips and Croteau 1999, Seybold et al. 2000).

Thus, when resin is exposed to air, the turpentine fraction evaporates, leaving the solid

resin acids behind, resulting in the commonly observed drops of crystallized resin

covering wound sites on physically damaged conifers. In ponderosa pine, the

monoterpenes comprise about 25% of the total resin weight (Smith 2000).

The monoterpene fraction dominates the volatile portion of oleoresin and has

therefore been extensively studied in the context of beetle host selection (but see also

Billings et al. 1976, Kelsey and Joseph 2003, Shepherd et al. 2007). The main

monoterpene compounds in Pinus resin are a-pinene, p-pinene, 3-carene, sabinene,

myrcene, limonene, and P-phellandrene, but considerable differences in monoterpene

composition have been documented between species, geographic locations, and even

individual trees (Smith 2000), complicating research efforts. Monoterpenes are

considered toxic to beetles (Byers 1995, Smith 2000) and other pathogens (Himejima et

al. 1992, Hofstetter et al. 2005). Limonene is the resin constituent believed to be the most

toxic to attacking western pine beetles, followed by 5-3-carene, myrcene, p-pinene and a-

pinene (Smith 1975, 2000). However, several of the monoterpenes are also used by

beetles as precursors for the synthesis of pheromone attractants, with various

14 combinations being used by different beetle species for chemical signaling purposes

(Seybold et al. 2000, 2006).

Sturgeon (1979) studied resin chemistry in ponderosa pines in northern California

and southern Oregon, finding that limonene content in the resin of individual trees was

correlated with historic local bark beetle presence. He further suggested that the beetles

were exerting selection pressures on these trees for higher limonene content in their resin.

This co-evolution theory was expanded by other researchers, with evidence that the

development of ponderosa pine's elaborate resin defenses was countered by the beetles'

development of attractant pheromones out of oleoresin components (Mitton and Sturgeon

1982, Seybold et al. 2000). Recent reviews of chemical ecology and resin biochemistry

(Byers 1995, Seybold et al. 2000, 2006) have offered some generalized observations on

beetle-host chemical evolution. In particular, they noted 1) the sophisticated yet relatively

common scolytid adaptations to conifer defense chemicals that involve using them as

chemical precursors to pheromone attractants, and 2) the remarkable ability in beetles to

occasionally produce the same compounds de novo - without actually any use of host

tree compounds. These findings suggests an elaborate variety of redundant attack

mechanisms in beetles, with profound implications for beetle-conifer interactions.

Fire adds an additional level of complexity to the picture, one that has received

little attention until very recently. Fire can have various effects on trees, including direct

effects on resin production, exudation, and storage, and possibly changes to chemical

composition. Thus, the next discussion pertains to the topic of resin responses to

wounding and fire injury.

15

1.2.3. Defense physiology 2: resin response to wounding

Several recent studies have reported increases in the resin flow of pines following

fire (Feeney et al. 1998, Ryan 2000, Santoro et al. 2001, Perrakis and Agee 2006, Knebel

and Wentworth 2007), or more generally, increased resin flow or resin canal density

following physical injury or other types of acute stress (Barman 1936, Fahn and Zamski

1970, Lewinsohn et al. 1991, Ruel et al. 1998, Phillips and Croteau 1999, Lin et al.

2002). Given the previous discussion, and the general consensus that resin flow is

positively correlated with beetle resistance, this response seems counterintuitive and

merits further investigation. An increase in tree defenses following fire does not match

with the frequently reported post-fire increases in beetle attacks and tree mortality.

Investigating this disconnect proved to be a major topic of this study.

Resin exuding from plants has been broadly separated into two types based on

mechanism of storage and origin: pre-formed (constitutive) resin, and traumatic

(induced) resin (Berryman 1972, Phillips and Croteau 1999, Franceschi et al. 2005).

Pines, spruces, larches, and Douglas-fir are somewhat distinct from most other conifers

in having the elaborate resin canal networks described above. Constitutive resin is stored

in such structures, in contrast with induced resin, which is rapidly formed during or

immediately following wounding or pathogen attack. Ponderosa pine does not appear to

depend on short-term induced resin, unlike many conifer species (including some pines)

(Lewinsohn et al. 1991, Ruel et al. 1998, Wallin et al. 2003, Hudgins et al. 2004). Older

laboratory studies on other Pinus species have reported increased resin duct formation

16 following wounding, but at longer time scales (months) in seedling-size individuals

(Barman 1936, Fahn and Zamski 1970). A survey of previous research suggests that resin

exudation from the bole or branches generally increases following physical injury, with

many variations based on factors such as species, distance from wound location, and

season of injury (Biisgen and Munch 1929, Barman 1936, Fahn and Zamski 1970,

Cheniclet 1987, Croteau et al. 1987, Lewinsohn et al. 1991).

Resin flow is also responsive to stand-level silvicultural treatments that cause

injury at the individual-tree level. Working on loblolly pine, both Nebeker and Hodges

(1983) and Fredericksen et al. (1995) observed increased resin flow following

mechanical injury treatments to sample trees. In the latter study, however, resin flows

were significantly reduced the following season (7-11 months after injury) in treated trees

compared with controls, suggesting a dynamic response to injury over longer terms. Few

studies of the sort have been done over longer time periods, or on ponderosa pine.

Silviculture researchers (e.g. Sartwell and Stevens 1975, Hessburg et al. 1994) have

noted that ponderosa pines tended to be more susceptible to beetle attacks at higher stem

densities, likely due to decreased vigor from high levels of competition for water and

other resources. This theory was reinforced by Mason (1971) and Kolb et al. (1998), who

both found significantly higher OEF in thinned stands than in unthinned controls, and

negative correlations between resin flow and stand basal area. In contrast, Feeney et al.

(1998) noted no significant difference in resin flow one year after thinning in old-growth

ponderosa pine in Arizona, suggesting that changes to resource budgets may not provoke

such short-term responses, compared to physical injuries to sample trees.

17

1.2.4. Physiology of resin defenses: challenges

Attempting to compare the findings from laboratory and stand-level experiments

exposes the deficiencies that still exist (or existed recently) in the understanding of tree

responses to fundamental physical and environmental stimuli. Physiological responses to

experimental treatments are usually complex, with numerous confounding factors and

negative feedbacks affecting study populations. For example, stand thinning should

increase resource availability by decreasing competition from other vegetation (Oliver

and Larson 1996, Feeney et al. 1998, McDowell et al. 2003). However, thinning can also

change microclimatic conditions, affecting environmental variables such as wind speed,

solar radiation, and evaporation rate. Resin defenses may be affected differently by these

two types of environmental changes, with potentially confounding effects. Low-severity

fire (from prescribed burning or wildfire) can have similar stand effects (decreasing

competition, increasing available resources to remaining trees, affecting microclimate),

while also adding the variable effects of fire injury to the equation.

In conifer resin research, additional complications arise due to inconsistent use of

the terminology of sampling techniques, with little information on correlations between

methods (Lorio 1994 - e.g., compare the meaning of'OEF' in Cobb et al. 1968, Mason

1971, Feeney et al. 1998, Smith 2000, Santoro et al. 2001). Different scales and contexts

are also a problem. Resin defenses have previously been studied in laboratory bioassays

(e.g. Lewinsohn et al. 1991, Seybold et al. 2000 and references therein), via dissection

and microscopic techniques (Bannan 1936, Lin et al. 2002), chemical analysis of foliage

18 (Johnson et al. 1997) or other tissues (Latta et al. 2000), direct resin tapping in forest

stand experiments (Feeney et al. 1998, Kolb et al. 1998, Wallin et al. 2003), and

integration into whole-tree autecology models (Waring and Pitman 1985, Lorio 1986,

Lombardero et al. 2000). Each of these diverse types of analysis has its respective

strengths, as the feedbacks and interaction mechanisms transcend any single approach or

scale. However, since findings and conclusions using different experimental approaches

have sometimes been contradictory, there is a need to integrate these multiple scales of

analysis to achieve adequate understanding of processes and interactions (such as for

designing management treatments).

Some background discussion is still needed of the processes by which trees build

resin defenses. For several decades now, internal carbohydrate allocation has been

discussed in the context of defense investments by trees. Various conceptual models of

plant defense have been proposed to explain variability in resin defenses in response to

environmental change.

1.2.5. Growth-differentiation balance: a hypothesis for explaining resin response?

Although the Growth-Differentiation Balance (GDB) hypothesis was first

proposed over 50 years ago (Loomis 1953), it was reemphasized in recent decades by the

work of P.L.J. Lorio (1986,1993) in the context of modern bark beetle research (see also

Lorio and Sommers 1986, Lorio et al. 1990). Following an extensively cited review that

defined the GDB as the most comprehensive plant defense theory to date (Herms and

Mattson 1992), many other studies successfully used GDB principles to explain plant

19 defense responses to various experimental treatments. According to a recent discussion

paper, the GDB remains the most current theory of plant defense, incorporating previous

concepts such as the Carbon-Nitrogen Balance and others theories (Stamp 2003).

The hypothesis is as follows: during photosynthesis, plants produce carbohydrate

energy that is stored in tissues or allocated directly to plant structures based on

source/sink relationships that vary by environmental conditions and phenology. Growth

processes and cellular differentiation (specialization) are competing sinks for

carbohydrates, but growth is usually prioritized over differentiation (Herms and Mattson

1992). Environmental conditions that constrain growth, however, tend to be less limiting

to photosynthesis (Hsiao 1973, Luxmoore 1991). Thus, when resource availability (e.g.

water, nutrients) and environmental conditions are optimum, growth of shoots, roots, or

cambial tissues occurs, rapidly depleting resources and energy stores. When sub-optimal

environmental conditions occur (such as mild or moderate water stress, Hsiao 1973),

growth is immediately reduced, but photosynthesis can often continue. This leads to a

surplus in available carbohydrate energy. Carbohydrates are then used in processes

requiring fewer environmental resources - cell differentiation and secondary metabolism

(Lorio 1986; Herms and Mattson 1992). The conceptual model is illustrated in Figure 1,

demonstrating that carbon assimilation is partitioned between growth and secondary

metabolism at various levels of resource availability. At intermediate to high levels of

resources, reductions in resource availability will reduce overall growth, but increase

secondary metabolism. At more extreme levels of environmental limitation,

photosynthesis is also compromised. In such cases (such as prolonged drought), both

20 carbohydrates and environmental resources are limiting, and consequently both growth

and differentiation are reduced; this corresponds to the left side of the graph in Figure 1.

In this discussion, 'growth' refers generally to irreversible cell division and

expansion in a general sense (particularly growth of shoots, roots, or the cambium).

'Differentiation' refers to the production of complex compounds and structures within

young cells, forming specialized forms such as secretory tissues, defensive structures and

compounds, vascular tissue, certain reproductive tissues, and so forth (Loomis 1953,

Herms and Mattson 1992, Taiz and Zeiger 1998). Both of these processes require

carbohydrate energy inputs in the form of sucrose and other sugars (Kozlowski and

Pallardy 1997).

Originally developed based on studies on herbs and crop plants (Loomis 1953,

Herms and Mattson 1992), GDB principles are now supported, at least in part, by

evidence from a considerable body of experimental work on conifers. For example,

studies on ponderosa pine needles suggested that water stress increased needle toughness

(McMillin and Wagner 1996) and terpenoid concentrations (Johnson et al. 1997) at the

expense of growth, compared with unstressed controls. Studies in many other Pinaceae

species have also found results in general agreement with GDB predictions (e.g. Lorio

1986, Lorio et al. 1990, Johnson et al. 1997, McKinnon et al. 1998, Wilkens et al. 1998,

Lombardero et al. 2000, Turtola et al. 2003).

21

Figure 1. The theoretical relationships between resource availability and assimilation (photosynthesis), growth rate, and secondary metabolism (cell differentiation). From Herms and Mattson (1992).

A key to the logic of the GDB theory in the current context is the notion that

certain levels of resource availability exist that will restrict growth, but still allow

photosynthesis to continue (Herms and Mattson 1992). Specific resource levels will

clearly be species- and site-dependent, but water (Johnson et al. 1997, Lombardero et al.

2000, McDowell et al. 2003) and nutrients (McKinnon et al. 1998, Wilkens et al. 1998)

are typical candidates, where responses to varying resource levels have at least partly

confirmed GDB predictions. Water deficit is the most frequent source of environmental

stress in plants (Kramer 1983) and has been suggested as an influential factor on tree

survival at Crater Lake (Swezy and Agee 1991, Agee 2003).

22 The GDB hypothesis is relevant to the present context because epithelial cells

forming pine resin canals and oleoresin production require considerable investment in

cellular differentiation. If growth is prioritized over differentiation processes at high

levels of resource availability, the GDB hypothesis suggests that lower levels of resource

availability will result in increased investment in resin defenses (Lorio 1986). This

concept was the basis for the design of Experiment 3 (see Section 3), which was partly

aimed at assessing the potential effects on resin of induced water deficit. In addition, the

GDB hypothesis was used in this study as a general concept for interpreting (or

attempting to interpret) various observed responses following stress or injury in

ponderosa pine.

To date, research on conifer trees has shown mixed agreement with GDB theory -

many studies appear to support it, in broad terms. However, a few reports have emerged

since the initiation of the present study that have questioned the relevance of the GDB

hypothesis in the context of resin flow. These will be discussed later in this text.

1.2.6. Fire effects on tree structures and carbohydrate relations

Fire may be important to these relationships due to its effects on tree carbon

budgets, water and nutrient uptake capacity, and perhaps physical injury and tissue death.

Based on these processes, an initial hypothesis for the observed responses and

interactions occurring at Crater Lake was developed based on some of the GDB concepts.

The main effects of fire on conifers is injury via bole charring, crown scorch, and

root damage (Agee 1993). At Crater Lake, root losses appeared particularly important

23 following prescribed fire, because of the slow combustion of thick duff layers and high

root density in the duff (Swezy and Agee 1991), a process that has been well observed

and documented in large, fire-excluded ponderosa pine stands (Ryan and Frandsen 1991).

Crown scorch was also a factor on some burn sites at Crater Lake and was positively

correlated with bark beetle attacks (Swezy and Agee 1991), although was less of an issue

in other burn experiments (Perrakis 2004).

Root losses from burning can obviously provoke water stress. GDB principles

suggest that at some moderate level of water deficit, growth is compromised while most

photosynthesis continues. The loss of carbohydrate sinks from growing meristems would

then be expected to increase allocation to cell differentiation, including production of

resin canals. This may explain the increase in resin production that was observed in the

season following burning at the Crater Lake sites (Perrakis 2004). In the long term,

however, severe injury to the root system would unbalance the root/shoot ratio

(Kozlowski and Pallardy 1997), severely limiting water availability until roots were

repaired (or foliage was dropped), perhaps over several seasons. Expected effects could

include shortened growing seasons, ultimately reducing both growth and photosynthesis,

reduced tree vigor, and increased susceptibility to insects and other pathogens.

Crown scorch has also been associated with successful beetle attacks in

ponderosa pine at Crater Lake and elsewhere (Swezy and Agee 1991, Harrington 1993,

McHugh and Kolb 2003). While effects of crown reductions on water relations are less

obvious due to internal feedback effects (Ryan 2000, Wallin et al. 2003), losses of

photosynthetic foliage necessarily reduces carbohydrate production, and should therefore

24 decrease allocations to both growth and resin defenses, according to the GDB hypothesis

(e.g. Wallin et al. 2003).

Such a pattern of dynamic response of tree defenses to environmental conditions

and injury can potentially explain observed beetle-related overstory mortality at Crater

Lake. In previous studies, the peak of successful beetle attacks occurred between 2 and 8

years following burning (Thomas and Agee 1986, Agee 2003). Continued monitoring of

burned trees from the 2002 prescribed fires (Perrakis 2004) in this study was designed to

explore the duration of resin flow response: based on GDB predictions, increased resin

flow was hypothesized to be short-lived, perhaps similar to the findings of Fredericksen

et al. (1995), and consistent with observed bark beetle infestation patterns (Figure 2).

In typical wildfire and prescribed fire situations, the effects of root damage, bole

charring and crown scorch occur together. However, their effects on physiological

processes may be different in different regions of a tree, confounding efforts to

understand whole-tree responses. Several of the experiments in this study were designed

to isolate these specific effects on tree structures, while comparing individual effects to

actual fire effects and unburned controls.

25

U Photosynthetic capacity

Wound effect (induced response)

J} water/ nutrient uptake

•O-Growth rate JJ Resin production

Resin production

(Lower severity fire)

Water deficit

H Overall growth t Resin

(Higher severity fire)

Time since fire

II Overall growth J} Resin

U- Resin tf Shoot growth ("repair")

fl Resin tfCambial growth ("repair")

U. Resin 0 Root growth ("repair")

Short term (•» 1 year?)

Medium term (1-3 years?)

Long term (2-5 years?)

Figure 2. The hypothesized impacts of fire injury to a tree's crown, bole, and roots on growth and resin production.

26 1.3. Experimental questions and hypotheses

The above review demonstrated the complexity of the management problem in

Crater Lake ponderosa pine stands. Understanding the interactions between the trees,

prescribed fire effects, and bark beetles in park stands involves untangling a considerable

body of existing research at multiple scales of analysis, from tree physiology to bark

beetle chemical ecology to fire response. Many knowledge gaps still exist regarding the

particular host tree (ponderosa pine) and beetle (western pine beetle) interactions in this

system, and it is not obvious that findings on other taxonomic groups, such as other

conifers or beetle species, are applicable. With these issue firmly in mind, I combined

experiments at varying scales to integrate a study on resin ecophysiology into the larger

context of prescribed fire and bark beetle management.

Based on the dynamic interactions described above between fire, host tree

response, and beetles, the experiments in this study were designed to answer the

following questions related specifically to fire-bark beetle interactions in southern

Oregon ponderosa pine forests:

• Previous studies reported elevated post-fire mortality of old ponderosa pines

compared with unburned stands (Agee 2003, Perrakis 2004). For how long is the

elevated mortality signal detectable after spring and fall burning treatments in this

ecosystem?

27

• Previous work also observed increased post-fire (Perrakis 2004) resin flow in

ponderosa pines at Crater Lake NP. What is the duration of that response, and

how does it related to different burning treatments and tree vigor indices?

• Can post-fire resin response in ponderosa pine be explained by the effects of

injury to tree structures? That is, can root damage, crown scorch, or bole charring

alone properly imitate observed fire effects on resin defenses?

• Are carbohydrate allocation tradeoffs between resin production (as measured by

resin flow) and cambial growth detectable in ponderosa pines? How does fire

injury affect these tradeoffs, and can they be explained by the Growth-

Differentiation-Balance hypothesis?

• What is the seasonal pattern of resin flow in ponderosa pines? Does fire injury

affect this pattern?

• Is the elevated post-fire resin response dependent on measurement height on the

bole? Is elevated post-fire resin response a localized phenomenon?

• Following fire and fire injury, do ponderosa pines produce different resin

monoterpenes?

In addition, this study incorporated some methods-testing questions:

• Are xylem resin yields from the cambium similar to yields from drilled holes in

the sapwood? More specifically, what is the relationship between the "arch-

punch/outer xylem" resin flow measurement method (Feeney et al. 1998, Strom et

al. 2002, Karsky et al. 2004) and the "drilled scoop/sapwood" resin flow

28 measurement method (Perrakis 2004)? Are the two methods comparable, or do

they assess different properties of sample trees?

• The existence of tangential resin canals, and the degree to which ponderosa pines

trees can translocate resin horizontally, is in question. Is there any evidence for

tangential or circumferential resin movement within the sapwood of ponderosa

pines?

The overarching question related back to the management problem at Crater Lake

NP:

• What are the implications of the study findings for forest managers? Do any clear

solutions emerge regarding the problem of perpetuating old ponderosa pines?

Individual experiments were designed to address these questions in this order.

29

2. Experimental Methods and Results

2.1. Experiment 1: Mortality and vigor in spring- and fall-burned trees

2.1.1. Objectives

This experiment was designed to follow up previous observations of post-fire

ponderosa pine mortality with some longer-term monitoring data. As previously

described (Perrakis 2004), prescribed fire was applied in spring and fall treatments to a

-70 ha area in the southern end of Crater Lake National Park with known concerns

related to immediate and extended ponderosa pine mortality. The present experiment

involved monitoring the survivorship and beetle presence in the old growth ponderosa

pine population for 3 additional years following the 2003 monitoring and analysis session

(which was summarized in Perrakis 2004). This longer-term evaluation was intended to

provide additional insights into the issue of extended post-fire pine mortality and the role

of bark beetles (c.f. Kolb et al. 2007). In addition, ring-width data were collected from

certain trees to test the correlation between radial growth indices and Keen's vigor class,

which had previously been strongly associated with probability of mortality (Keen 1943,

Swezy and Agee 1991, Perrakis 2004).

2.1.2. Methods

This experiment took place in the 'panhandle' region of Crater Lake National

Park, near the southern park boundary (Figure 3). As described in Perrakis (2004), a

series of prescribed fire treatments was conducted in spring and fall 2002. Diameter and

30 crown vigor class were initially assessed on 1725 large ponderosa pines within the study

boundary before treatment in 2001 and 2002, with crown vigor class being based on the

'California Pine Risk-Rating System', as described by F.P. Keen (1943). This method

involves a qualitative assessment of overall tree vigor based on crown appearance and

approximate age. The entire sample (1725 trees) was monitored annually for mortality,

including an assignment of cause-of-death for those trees that died, during the 2002-2006

period. Although diameter at breast height data was collected from all trees, this factor

was insignificant in predicting mortality in previous analyses (Perrakis 2004), and so was

not incorporated into the models. Likewise, 96 of the trees were subjected to resin flow

sampling (see Experiment 2) and previous resin exudation pressure measurements

(described in Perrakis 2004). These sampling efforts were believed to have had negligible

effects on the trees and were not factored into the mortality analysis.

Increment cores of the outer xylem rings were collected in summer 2005 from 10

trees per experimental unit. In each unit, cores were taken at breast height (1.4 m) from

the north sides of 5 A- or B-class trees, and from 5 C- or D- class trees (Keen 1943); trees

were randomly selected within each unit out of the pool of available A/B and C/D class

trees, with only trees alive at the time of coring being selected. Since only the outer rings

were analyzed, no effort was made to collect complete cores to the pith. Once collected,

standard storage and mounting techniques were used. On all cores, Mahoney's Periodic

Growth Ratio (PGR) and the number of rings in the outer centimeter were measured and

computed. Previous studies have shown these indices to be significant predictors of

31 mortality, with higher Mahoney's PGR and lower numbers of rings/cm in beetle-resistant

trees (Raffa and Berryman 1982, Stuart 1984).

2.1.3. Data Analysis

This analysis used logistic regression on individual tree mortality data (Dalgaard

2002). Treatment (spring burn, fall burn, or control) and Keen's crown class (class A, B,

C, or D) were the categorical predictor variables. The model was fitted twice: once for

the entire sample (N=1725), and once excluding trees that were believed to have died

without evidence of insect attack (i.e. direct mortality from fire, windthrow or other

causes; N=1701). The second run was designed to isolate the major factor of interest to

this study, that being post-fire attacks by bark beetles and other insects or pathogens.

Increment core data were analyzed using a simple one-way analysis of variance.

Mahoney's PGR and the rings per outer cm were both evaluated separately as a function

of Keen's crown vigor class. Post-hoc pairwise differences were evaluated using Tukey's

HSD test (Zar 1999) at the a =0.05 level of significance.

All analyses described in this study (including Experiments 2-6) were performed

using R version 2.7 (R-Project Collective 2008)2, with differences evaluated at the a

=0.05 level of significance.

2 Available online: http://www.r-project.org/

http://www.r-project.org/

32

Figure 3. Overview of the southern boundary area of Crater Lake National Park, with the main study areas for several experiments shown highlighted. The park boundary is approximately represented by the green line. CL PH-3 represents the Crater Lake NP Panhandle-3 unit, where the 2002 prescribed burns were conducted (Experiments 1,2); SPSF is the approximate location of the Sun Pass State Forest site (Experiments 3, 4), and ACW is the Annie Creek/Winema National Forest site (Experiment 5). Most of the lower right section of the figure is contained in Sun Pass State Forest, but a small area (not shown) adjacent to the 'panhandle' formation is part of Winema National Forest.

2.1.4. Results

By fall 2006, 139 large ponderosa pines had died, representing 8.1 % of the

original population of 1725. Of those that died, 24 were estimated to have been killed

directly by burning or windthrow in 2002 or early 2003. Thus, mortality for 115 trees out

33 of 1701 (6.8 %) was at least partly related to post-treatment insect or pathogen attack. Of

the 115, the majority were found in spring burn or fall burn treatment units (Table 1).

The main effects logistic regression model was significant at the a=0.05 level

when run with 1-all trees and 2- only live trees and those killed by secondary mortality

(Table 2). The null model (indicated by the Intercept line in Table 2) represents an A-

class tree and a control treatment. The burn treatment factor was significant, with spring

burning increasing the probability of mortality, and fall burning increasing it further yet.

The crown class factor was also significant, with class B trees more likely than class A

trees to die during the study, and C and D classes each further increasing that probability.

While these terms were significant predictors of mortality, the overall predictive power

of the logistic regression was low, explaining only 10-12 % of observed mortality based

on residual deviance (Table 2).

Table 1. Mortality of large ponderosa pines by treatment and Keen's crown vigor class, 2003-2006, excluding direct mortality from fire and windthrow. SB and FB refer to spring and fall burn treatments, respectively. Cell values indicate numbers of trees, with percentages of group totals in parentheses. Dead ponderosa pines - secondary mortality

Treatment:

Crown Class: A B C D

Total dead

Control

0(0) 3(1.1) 7 (2.9) 3 (8.8)

13

SB

1 (5.3) 11 (4.8) 15(5.7) 5 (8.8)

32

FB

5(5.1) 31 (12.3) 23(15.0) 11 (42.3)

70

Total dead

6 45 45 19 115

Total

181 746 657 117

Total 599 572 530 1701

34

Figure 4. Post-treatment mortality of large ponderosa pines between spring 2003 and fall 2006, separated by Keen's (1943) class (A through D refer to crown vigor classes; see text) and burn treatment type (Control: unburned; SB: spring burn treatment; FB: fall burn treatment).

The overall mortality pattern is evident in Figure 4: probability of mortality was

inversely proportional to crown class, and increased from controls (lowest probability) to

spring burns and fall burns (highest probability). The apparent exception to this is the

higher than expected mortality rate among A-class trees in spring burn units; this is

probably due to a small sample size - only 1 tree died in that group out of 19,

representing 5.3% of the initial group size. The pattern of mortality inversely related to

crown class is evident from the remaining crown classes, B through D.

35

Table 2: Parameters of the logistic regression models. The model form is

= exp(/?0 + /3JXJ + p2x2 +... + p x), where/? is the probability of mortality, xj \-p through xg are predictor variables, fio is the intercept, and /?; through/?? are the coefficient estimates. All dead trees: Coefficients:

Factor Estimate Std. Error z P

Secondary mortality only:

Estimate Std.Error z P (Intercept)

TrtSB TrtFB

ClassB ClassC ClassD

-4.9332 0.8408 2.2325 0.9875 1.2792 2.4703

0.4767 0.3249 0.2979 0.4166 0.422 0.467

-10.349 2.588 7.493 2.37 3.031 5.29

<0.001 0.0096 <0.001 0.0178 0.0024 <0.001

-4.9231 0.8389 2.046 0.9094 1.2341 2.2609

0.5055 0.3369 0.3118 0.4487 0.4533 0.5038

-9.739 2.49 6.562 2.027 2.722 4.488

<0.001 0.0128 <0.001 0.0427 0.0065 <0.001

Null deviance: 966.63 on 1724 degrees of freedom Residual deviance: 849.52 on 1719 degrees of freedom AIC: 861.52

Null deviance: 841.67 on 1700 d.f. Residual deviance: 759.28 on 1695 d.f. AIC: 771.28

Figure 5 summarizes the total numbers of trees killed by year and treatment type.

Mortality in spring and fall burn units was clearly highest within the first year after

burning, and then dropped off steadily in the following years. By the end of the study,

2.3%, 6.1%, and 16.4% of trees in control, spring burn, and fall burn treatments,

respectively had died.

6

5

I 4

"8 2

A

2002-3

36

- •— Controls

A Sp. Bum

- • — F . Bum

2004 2005 2006

Year

Figure 5. Proportion (%) of large ponderosa pines killed (proportion of initial live population) by year and treatment, considering all mortality sources.

Two hundred and thirty nine viable cores were collected from trees in all 24 units

(10 per unit, minus one core that was lost during analysis). The number of cores from

each crown class was 16, 97, 110, and 16 from classes A, B, C, and D, respectively -

proportions very similar to the study area population as a whole. Mahoney's PGR was

not significantly related to Keen's crown classes (p = 0.471). However, the relationship

between number of rings per outer centimeter and Keen's crown class was highly

significant (p < 0.001), with increasing number of rings/cm between classes A to D

(along with high variability within groups; Figure 6). The summary of the analysis of

variance and Tukey's HSD analyses are given in Table 3. As the pairwise contrasts show,

the differences between class A and B trees and between class C and D trees were not

significant, but all other differences were significant at the a =0.05 level. Mean numbers

of rings per outer cm (and standard deviations) for the 4 classes were A: 21.8 (11.9); B:

37 26.4 (13.2); C: 35.8 (15.3); D: 42.3 (17.4). The grand mean from all samples was 31.5

rings/cm.

Keerfs Cro/wn Classes Figure 6. Boxplots of the number of rings per outer centimeter, grouped by Keen's (1943) crown classes. Center lines and boxes represent median and interquartile range values, respectively. Circles represent outlier values in each crown class. Numbers of tree cores analyzed in each class were 16, 97, 110, and 16, for classes A through D, respectively.

38 Table 3. Anova table and Tukey test summary for the increment core analysis (Rings per outer cm as a function of Keen's crown vigor class). 'Contrast' refers to the crown classes being compared. Factor

CrClass

Residuals

Tukey's HSD Contrast: B-A C-A D-A C-B D-B

D-C

DF

3

235

(0.95): difference

4.693 14.014 20.563 9.320

15.869 6.549

SumSq 7871 48866

lower bound

-5.375 4.030 7.371 4.123 5.801

-3.435

Mean Sq

2624 208

upper bound.

14.761 23.997 33.754 14.517 25.937 16.532

F

12.618

padj

0.624 0.002

<0.001 <0.001 <0.001

0.327

P <0.001

39

2.2. Experiment 2: Resin flow monitoring in spring- and fall-burned trees

2.2.1. Objectives

This experiment involved additional post-burn resin flow monitoring in large

ponderosa pines from the Crater Lake prescribed burn site. As resin flow was previously

found to increase following burn treatments (Perrakis 2004), the present experiment was

designed to test the duration of this effect, as well as assess whether there was evidence

that post-fire resin flow followed the energy allocation patterns predicted in Section 1.2.

Spring burn, fall burn, and control treatments as well as high and low vigor (see

Experiment 1) were contrasted for this purpose, with resin measurements taken for an

additional 3 years beyond the previous (Perrakis 2004) dataset.

2.2.2. Methods

This experiment took place in the same study area as Experiment 1 (Crater Lake

NP 'panhandle'; Figure 3). Resin flow was measured on 96 large (40-170 cm DBH,

average 90 cm) ponderosa pines before and after prescribed fire treatments (Perrakis

2004); 4 trees (2 high vigor, 2 low vigor) were monitored in each of 24 experimental

units (8 units each in spring burn, fall burn, control treatments). Half of the trees in each

unit (1 high, 1 low vigor) were also subjected to a 'light raking' treatment prior to

burning to evaluate the effects of this technique on resin flow. However, the raking

treatment was not significant in any prior analyses on resin flow or pressure (Perrakis

2004) and is not further considered here. Trees were selected randomly for resin

40 sampling based on Keen's crown vigor classes, the high vigor group consisting of A-

class trees (or B-class, when no A-class trees existed) and the low vigor group consisting

of C-class trees.

Resin sampling involved drilling angled holes into each tree bole at

approximately breast height (1.2-1.5 m), with brass funnels directing resin into centrifuge

tubes, and resin volumes measured 24 hours after drilling (see Perrakis 2004). Two holes

were drilled in each tree at each measurement time. Because of logistical difficulties with

the early OEF sampling, reliable pre-treatment resin flow data were collected only from

the fall burn experimental units and half of control treatment units. As a result, the full

dataset of resin flow values are compared only in terms of total resin volume at each time

period; a separate analysis based on differences from pre-burn values was done only for

the fall burn units (see Data Analysis below). Pre-treatment sampling on fall treatments

(burns and 4 control units) was conducted in early September, 2002. Measurements in

2003-2006 were conducted in mid-August of each year for consistency, with all

measurements conducted within a few days of stable weather to avoid bias. Although

2002-2003 data were previously presented (Perrakis 2004), results from all years are

included for completeness.

2.2.3. Data Analysis

The structure of treatments and randomization at the scale of experimental units

suggested a split-plot design linear model (Oehlert 2000, Pinheiro and Bates 2000).

Replication (the burn treatments) and tree grouping were at the level of experimental

41 unit, with resin flow measured in the 4 sample trees per unit. As in the previous

experiments in this study, resin volumes were square-root transformed to stabilize

variances. Categorical predictors included Crown class (high or low, based on Keen's

classes A or C), treatment (Control, Spring burn, or Fall burn), and year (2003 - 2006).

The second analysis involved only the fall burns and 4 control units. For this

model run, pre-treatment resin flow values were subtracted from each year's measured

values (post - pre); differences could therefore be either positive or negative. Fixed

factors (treatment, crown class, and year) and all interactions were initially included in

the model; interaction terms were included and evaluated based on the a=0.05 level of

significance.

2.2.4. Results

Prior to burning, resin flow was not significantly different between trees in fall

burn and control units (Perrakis 2004). In the first post-treatment year (2003), there were

several instances when sampling test tubes overflowed with resin in the highest flow

samples. This was noted in 7 trees in fall burn treatments (4 high, 3 low vigor), 4 trees in

spring burn treatments (3 high, 1 low vigor), and 1 tree in control treatments (high vigor).

In following years, a switch to larger centrifuge tubes (50 mL) eliminated this problem.

Thus, resin flow was slightly higher in 2003 than the values presented, particularly in fall

and spring burn treatments.

Post-treatment analysis results are presented in Tables 4 and 5. Trt and Cclass

refer to treatment type (Control, spring burn (SB), or fall burn (FB)) and Keen's crown

42 class (high vigor or low vigor) respectively. The intercept represents the default case of a

high vigor tree (A class) in a control (unburned) treatment unit. Over the course of the

experiment, two of the 96 trees died - both low vigor trees, one in a spring burn unit in

2003, one in a fall burn unit in 2004. Both trees were killed by post-fire western pine

beetle attacks.

None of the 2-way or 3-way interaction terms was significant, so these were

dropped from subsequent model runs. Since the TrV.Year term was not significant (p =

0.84 in a full-factorial run (not shown)), this suggests that mean treatment effects were

not significantly different between the years considered in the study. All fixed-effect

factors were significant at the 0.05 level (Table 4). Both the SB and FB terms were

significantly different from zero, suggesting that trees in both spring burn and fall burn

treatments had significantly higher resin flow than in controls. Low vigor (ClassC) trees

had significantly lower resin flow than high vigor trees. The overall Year effect is

somewhat harder to interpret. Although the fixed factor term was significant, none of the

coefficients for individual years was different from zero at the 0.05 level. Year-to-year

variation in resin flow likely affected all trees, but these results do not suggest that they

were a significant factor in any individual year during this experiment.

The separate analysis of resin differences between fall treatments shows similar

results (Table 5). Trees in fall burn units had greater resin flow differences than those in

control units, further suggesting a post-fire increase in resin flow. In this analysis, Cclass

was not significant, suggesting that the trees' induced resin response was similar in the

43 high and low crown vigor classes. The Year factor was also not significant in this

analysis, indicating that relative resin flow change was constant across the study timeline.

The overall resin flow trend over the 5 years of the study is presented in Figure 7

(presented as means of all trees in each treatment group). The post-treatment response of

increased resin flow in burned trees is apparent from the figure, with both spring and fall

burn treatments apparently peaking in the second year after burning (2004). By 2006,

resin flow from spring burn and fall burn treatment groups has declined to levels nearly

approaching control treatments.

Table 4. Anova table and summary from post-treatment resin flow measurements, August 2003-2006. Numbers shown represent summary of transformed data, as outlined in the Data Analysis section. DF represents degrees of freedom. The value column shows estimates of the linear model coefficients (based on transformed data); the intercept represents Control treatments and A-class (high vigor) trees; SB, FB, and Class C model terms and statistics are contrasted with high vigor control trees.

Factor (Intercept) Trt Year Cclass

DF 1,349 2,21

3,349 1,349

F 568.961

5.731 4.278

21.877

P O.001 0.0103 0.0055 <0.001

(Intercept) SB FB Year2004 Year2005 Year2006 ClassC

Random effect:

Value 3.576

0.8925 1.349 0.479 0.412

-0.398 -0.922

StdDev I

Std.Error 0.350 0.410 0.409 0.278 0.278 0.278 0.197

Residual

DF t-value 349 10.2286

21 21

349 349 349 349

2.178 3.294 1.724 1.482

-1.431 -4.677

p-value O.001

0.041 0.0035 0.0856 0.1393 0.1533 <0.001

Unit 0.663 1.911

44 Table 5. Data summary from fall treatment resin difference analysis. Pre-treatment (early September 2002) resin flow values were subtracted from measurements in 2003-2006 for each tree. Abbreviations are as in Table 5, above. Burn treatments were conducted on 8 experimental units; these are contrasted with 4 control units (4 sample trees per unit).

Factor DF Value Std. Error DF (Intercept) Trt Year Cclass

1,173 1,10

3,173 1,173

2.091 7.287 2.063 2.591

0.15 0.022 0.107 0.109

(Intercept) Trt2 Yeary04 Yeary05 Yeary06 CclassC

Random effect:

-1.742 2.413 0.315 0.122

-1.051 -0.699

StdDev

0.854 0.906351 0.609014 0.609125 0.609125

0.43451

Residual

173 10

173 173 173 173

-2.040 2.662 0.517 0.201

-1.725 -1.610

0.0429 0.0238 0.6061 0.841

0.0864 0.1093

Unit 1.278 2.967

2002 2003 2004 2005 2006

Figure 7. Untransformed resin flow means and standard errors (of unit means) by year and treatment group; n = 4 trees per unit. Dashed line indicates approximate timing of burn treatments (not to scale); SB: spring burn treatments; FB: fall burn treatments. *Note: unbiased data from spring burn treatment trees in 2002 (pre- and post-treatment) was not collected and is not presented here. Control treatment mean in 2002 is a combination of early August and early September measurements, while 2002 fall burn (pre-treatment) data was collected in early September. Measurements in all subsequent years were taken approximately synchronously between treatment groups.

45

2.3. Experiment 3: Seasonal resin flow and fire-injury surrogates

2.3.1. Objectives:

This experiment was designed to assess some of the physiological and seasonal

controls of resin defense response in pole-sized ponderosa pines. In brief, the objective of

this experiment was to subject trees to fire or various "fire-surrogate" treatments

(described below), and measure their resin flow response repeatedly over time. The

treatments were structured around simulating fire injury to various tree structures, while

the general hypothesis was that treated trees would respond to injuries according to the

Growth Differentiation Balance (GDB) hypothesis, as described in Section 1. The

implementation and analysis related to this experiment involved considerable complexity

over the four years of study.

As previously discussed, resin defenses in ponderosa pines consist of an

interconnected network of vertical and radial resin canals. Canals are formed by

epithelial parenchyma cells that secrete resin into the tube-like intercellular space.

Previous studies have suggested that the formation of the specialized epithelial cells is

energy-intensive, and that their abundance may be partly controlled by carbohydrate and

resource budgets, as predicted by the GDB hypothesis.

The hypothesis suggests that resin canal production should be favored under

suboptimal conditions, when growth is restricted but carbon assimilation can continue.

The most common cause of suboptimal growing conditions is limited water availability.

The GDB hypothesis suggests that mild or moderate limits on available water should

46 decrease overall growth and favor resin production, while more severe drought

conditions decrease both growth and resin (Lorio 1986). Losses of foliage should

decrease the rate of carbohydrate production, reducing energy availability for both

growth and resin production, despite the improved water relations. The treatments in this

experiment, although designed around simulating fire injury effects to various tree

structures (roots, crown), were tested against these GDB hypotheses in terms of resin

flow and radial growth. Radial expansion in conifers is considered a low priority growth

process for energy allocation (Waring and Pitman 1985, Kozlowski and Pallardy 1997),

and so was assumed to be appropriate for testing of tradeoffs between carbohydrate sink

processes.

Additional experimental objectives included assessing the seasonal pattern of

resin response over several years and evaluating the sensitivity to injury (fire or

mechanical) of pole-sized ponderosa pines. Given the long-term nature of post-fire

mortality in this system (Agee 2003, this study) it seemed desirable to carry on the

experiment for several years after treatments. Practical considerations placed an upper

limit on the study duration, and in the end 4 years of data were collected, including the

shortened initial season of treatment (2003).

The experiment was conducted in a stand of pole-size ponderosa pines close to

the previous experiments' study area in Crater Lake NP. Chosen treatments included 2

levels of root disturbance (achieved by mechanical trenching) and one level of crown

reduction (pruning) to simulate fire damage to these structures, in the sense of emulating

root mortality from soil heating and crown mortality from scorching, respectively. These

47

treatments were compared with a sample treated with prescribed fire, to compare fire-

surrogate effects to actual fire effects, and with a control sample to reduce confounding

due to background variability.

To evaluate the effects of the treatments on the trees' water balance, carbon-13

(13C) discrimination was analyzed in xylem growth rings taken from increment cores.

Naturally occurring atmospheric carbon dioxide primarily contains 12C, but also contains

about 1.1 % of the 13C isotope (Smith 1972a). Because of its heavier molecular weight,

C carbon dioxide is slower to diffuse across stomatal boundaries and less reactive to the

photosynthetic catalyst Rubisco (ribulose 1,5 biphosphate carboxylase-oxygenase)

(O'Leary 1981). Thus, plants discriminate against the heavier C molecules during

photosynthesis, and less of this type of carbon ultimately becomes incorporated into plant

tissues. However, the degree of C-isotope discrimination is related to stomatal

conductance and ultimately, plant water balance: as water stress increases, so does

stomatal closure; stomatal conductance decreases accordingly, and CO2 remains trapped

in intercellular mesophyll tissues (O'Leary 1981). Thus, a greater proportion of 13C

becomes assimilated into plant tissues. 13C-discrimination in xylem rings has therefore

been positively correlated with soil water potential under water-limited growing

conditions (McNulty and Swank 1995, McDowell et al. 2003); based on this mechanism,

13C was used in this experiment as an indicator of overall tree moisture stress.

The design of this experiment initially involved using high-resolution automatic

dendrometers to track diurnal variations in stem diameter much more precisely than using

increment cores. Unfortunately, difficulties with the equipment resulted in little usable

48 data. Ring width data were ultimately used in the analyses, providing a much simpler

measure of annual radial growth. Further details on the dendrometer work are provided in

Appendix 3.

2.3.2. Methods:

This experiment was conducted in a stand of approximately 50 year-old

ponderosa pines (35-55 cm DBH) in Sun Pass State Forest, OR (Figure 3). The site was

adjacent to the Crater Lake National Park southern 'panhandle' area, close to where

previous prescribed burning treatments have taken place (see Thomas and Agee 1986;

Swezy and Agee 1991; Perrakis 2004). According to Oregon Department of Forestry

(ODF) personnel, the stand was pre-commercially thinned during the late 1980's,

approximately 15 years previous to the start of the experiment.

The main tree species at the site was ponderosa pine; white fir was also present

though much less abundant. Other tree species present in smaller amounts included red

fir, lodgepole pine, trembling aspen, sugar pine, and Douglas-fir. Compared to the

adjacent area in Crater Lake NP (see Perrakis 2004, Experiments 1 and 2 above), the

stand contained relatively more ponderosa pine and less white fir and other conifer

species. This may have been due to the site's position slightly further east of the Cascade

divide - a region of rapidly changing environmental gradients (Franklin and Dyrness

1988) - and possibly due to previous silvicultural activities (logging and replanting or re-

seeding) in the mid- 20th century. Twenty five pole-sized ponderosa pine trees were

selected for this experiment within an area of about 1 ha. At the start of the study, all

49 trees appeared in good health, with high crown vigor and no visible scars or other defects

(Figure 8).

Figure 8. The study area in wintertime, showing pole-sized ponderosa pines with high crown vigor (sensu Keen 1943).

50 Experimental treatments and sampling

Sample trees were split into 5 treatment groups of 5 trees each. The first group

(treatment 'P') had the lower half of each tree's crown pruned. Pruning was done by

hand, removing live branches until a predetermined point representing half the crown

volume was reached. A second group was subjected to mechanical root damage by

trenching to a depth of approximately 50-60 cm at a distance of approximately 1.5 m

from the bole. Trenches were dug in three separate segments around each tree to add up

to an estimated 1/3 of the tree's circumference ("light root trenching", treatment 'r'). A

third group was subjected to the same trenching treatment, but dug around 2/3 of the each

tree's circumference ("heavy root trenching", treatment 'R'). The fourth treatment group

was subjected to a mid-summer prescribed burn (treatment 'F'). The five remaining

sample trees were untreated controls (treatment 'C'). The burn treatment was assigned to

a group of trees clustered in a ~0.2 ha area, based on operational requirements related to

the prescribed burn; other treatments were assigned to trees randomly. Pruning and

trenching treatments were conducted on June 26-27 and July 2-4, 2003, respectively. The

prescribed burn treatment was conducted on August 2, 2003, with isolated smoldering

until August 4. Prior to ignition (in mid-July), surface litter fuels were aerated using hand

tools to reduce the duration of smoldering for fire control purposes.

Resin flow

Resin flow was measured on all trees twice per month from mid-spring to early

fall between 2003 and 2006. Measurements were timed as follows: 2003 (7 repeats): June

51 22 to Sept. 25; 2004 (8 repeats): May 16 (only 1 measurement in May) to Sept. 13; 2005

(8 repeats): May 27 to Sept. 14; 2006 (7 repeats): June 13 to Sept. 19. The 12-14 day

delay between sampling periods was chosen as a compromise between the benefits of

frequent sampling and the drawbacks of causing frequent injuries (albeit minor) to

sample trees. The period between sampling sessions was roughly estimated to be a

minimum time delay for resin ducts to recharge (Busgen and Munch 1929).

Resin flow measurements were conducted as described in Perrakis (2004), a

method very similar to numerous previous studies (Latta and Linhart 1997, Smith 2000).

Briefly, at each resin sampling time, 2 holes of 5.56 mm (7/32 inch) diameter were

drilled on opposite sides of the lower bole region of the trees. Holes were drilled at an

angle of approximately 30 degrees below horizontal and to a depth of 3.8 cm (1.5 inches,

estimated) into the sapwood. A funnel "scoop" consisting of 6.35 mm (1/4 inch)

diameter brass tubing, filed down on one side (see Perrakis 2004) was inserted into each

hole to collect the resin and direct it into a 50 mL test tube. Resin yield in test tubes was

measured after 24 hours. In about 5% of holes, a certain quantity of resin was obviously

leaking around the outside of a scoop (in addition to flowing into the test tube); in these

cases, 0.5 mL was added to the test tube value as a very conservative correction factor.

After each sampling repeat, tubes and scoops were removed and small sections of clean

hardwood dowel were hammered into the holes to seal the opening and end the outflow

of resin.

This method for measuring OEF was selected because of the small area of injury

it creates for each resin sample, and the ease of blocking the sampling holes following

52 measurement (see Experiment 6). Over the course of the experiment, 60 holes were

drilled in the lower bole region of each tree, roughly between about 0.8 m and 1.5 m

height.

The variations in treatment timing (see above) resulted in differences between

treatment groups in the numbers of pre-treatment and post-treatment resin flow

measurements collected. Thus, out of the 7 resin flow measurements in 2003, only the

first was a pre-treatment measurement in the pruned tree group (the other 6 were

conducted at various times post-treatment); in the two trenched tree groups, 2 sets of

resin flow measurements were pre-treatment measures; in the burned tree group, 3 of

these were pre-treatment measurements. The variable timing of treatments and resin

sampling had serious implications for data analysis (see below).

Basal area and foliar mass

Values of annual radial growth used in modeling resin flow were calculated from

increment core xylem rings. Standard increment boring techniques were used to obtain

outer sapwood cores for all trees in early spring 2007. Cores were taken on the northeast,

north, or northwest sides of each tree at approximately breast height (1.37 m) and stored

in plastic straws for protection and transport. At least three cores were taken from each

tree; two cores were mounted, sanded, and used for measuring radial growth3 and one

was dissected for carbon isotope discrimination (see below). Growth ring widths were

measured manually using a digitally projected light microscope with a movable stage.

3 In some trees, the first two cores had notable differences between measured same-year ring widths due to quality problems with the core or apparent real discrepancies in rings. In these cases, up to two additional cores were taken and measured, with ring widths analyzed as the mean width of the ring from all cores.

53 Basal area increment (BAI) was calculated as the difference between successive basal

area values (starting with 2003 DBH)4, using ring width values to increase each tree's

radius every year. Average 1998-2002 BAI was also calculated for all trees. Foliar mass

was estimated based on equations from Gholz (1982), which were derived from studies

on northern Arizona ponderosa pines between 15 and 80 cm DBH. Based on visual

estimates, the pruning treatment reduced crown volume by 50% in those trees;

accordingly, foliar mass values were divided by 2 to obtain post-treatment values for

these trees.

Carbon-13 discrimination

Carbon isotope discrimination was analyzed based on growth ring xylem in order

to evaluate treatment effects on water balance. As mentioned above, cores were collected

in early 2007 using standard techniques, and core handling and contact were minimized

to avoid contamination. Annual rings were separated by hand using a sharp knife and

packaged individually in sterile centrifuge tubes for transport to the laboratory (Figure 9).

Carbon isotope analysis was conducted at the Colorado Plateau Stable Isotope

Laboratory (Northern Arizona University, Flagstaff, AZ). Samples were processed using

a Thermo-Finnigan Delta-Plus Advantage gas isotope ratio mass spectrometer (Thermo

Fisher Scientific, Waltham, MA) interfaced with a Costech Analytical ECS4010

elemental analyzer (Costech Analytical Technologies, Valencia, CA). 813C, the

4 The 2003 DBH value for one tree was lost; it was subsequently estimated based on 2007 DBH and dendrometer data (see Appendix 3).

54 difference in carbon-13 per mille between laboratory standards (Pee Dee Belemnite) and

experimental samples, was the value used in analyses.

Figure 9. Annual xylem rings were carefully separated from increment cores using a 1 ^

sharp knife; rings were then individually analyzed for C discrimination.

55 2.3.3. Data analysis:

Because of the complexity of this dataset, several different analysis approaches

were used. Ultimately, resin volume was modeled as a function of several categorical and

continuous predictor variables, including treatment, season, year, basal area increment,

813C, and modeled foliar mass. Raw resin volumes were square-root transformed in order

to stabilize variances and meet the assumptions of linear modeling. The 'season'

predictor was a categorical variable based on sampling time, designed to separate the

mid-summer resin flow peak (Harper and Wyman 1936, Smith 2000) from earlier- and

later-season periods of lower flow. The variable was designed as follows: Season 1 (si):

May and June; s2: July to late August; s3: August 30 through September.

Resin flow was measured once in all trees prior to treatment (22 June, 2003), and

was initially modeled using the above predictors (except treatment, season, and year) to

ensure there was no bias between treatment groups (Analysis 1). This was done using a

simple linear regression model, with transformed resin volume (Volpre) as the response

variable and treatment (Trt), average BAI (avBAI), S13C, and Foliar mass as predictors:

•JVolpre = p0+phTrt + p2BAIav + p3(8nC) + p4(Foliarmass) + p5J[Tree] + eiJk

i = {Control, Burning, Pruning, Light trenching, Heavy trenching} (1)

j = {Tree 1...25 (Random)}

Post-treatment resin volume was initially analyzed in terms of mid-summer

average values in all treatments across all years (Analysis 2), to compare this dataset to

the findings in the older Crater Lake trees (see Experiments 1 and 2). This was done by

averaging the resin flow volume by tree in early and mid-August. Values in 2003

56 therefore represent the transformed average of resin volumes measured 1 to 1.5 months

following trenching and pruning treatments, and 3 days and 2 weeks following the burn

treatment. Measurement dates included in this first analysis were as follows - 2003: 05

and 18 Aug.; 2004: 02 and 15 Aug.; 2005: 01 and 16 Aug.; 2006: 01 and 21 Aug.5 Resin

volume was modeled using a mixed-effects multiple linear regression model, initially as

a full factorial model using Treatment and Year categorical predictors. Data were

grouped by the random effect Tree, since resin measurements were not taken randomly,

but rather repeatedly on the same sample trees (Pinheiro and Bates 2000). The model-

fitting process included several iterations, as described in the Results.

yjVolume = P0 + PuTrt + /32JSeason + PuYear + fiMj (Trt x Season) + P5ik (Trt x Year)

+P6jk (Season x Year) + filiJk (Trt x Season x Year) + fi8l[Tree] + ejjU

i - {Control, Burning, Pruning, Light trenching, Heavy trenching}

j = {sl,s2,s3}

£ = {2004,2005,2006}

/ = {Tree 1...25 (Random)}

Analysis on the larger dataset (Analysis 3) across all seasons, which included

measurements from May to September, was done on 2004-2006 data, 1-3 years post-

treatment. Data from 2003 were excluded from this analysis to avoid model-fitting errors

due to nonexistent early 2003 post-treatment data. Model-fitting proceeded as before,

with several continuous predictor variables added to the previous model (Eq. 2). Factors

were successively eliminated based on the a = 0.05 level or significance.

5 For comparison, equivalent sampling dates at the Crater Lake site were between August 09 and August 21,1 to 4 years after burning. See Experiments 1 and 2 for details.

57

yjVolume = p0+ PXBAI + p2 (<513C) + P3Foliarmass + PMTrt + p5JSeason + p6kYear +

P7iJ (Trt x Season) + P8ik (Trt x Year) + P9Jk (Season x Year) +

Pwijk (Trt x Season x Year) + pni[Tree] + eijldm


j = {51,52,53}

£ = {2004,2005,2006}

/ = {Tree 1...25 (Random)}

A fourth analysis (Analysis 4) was done on the differences from pre-treatment

data. This was a simpler linear mixed-effects model that focusing on the burned trees,

taking advantage of the three pre-burn resin measurements that were taken on this

treatment group (22 June, 02 July, and 17 July 2003). The response values for the model

(Voldif) were calculated for burned and control trees as the post- minus pre-treatment

differences (untransformed) from mid-June, late June/early July, and mid-July for 2004,

2005 and 2006. Predictors consisted of treatment (burn or control), year, and the

interaction between the two.

Voldif = P0+ PuTrt + P2jSeason + PuYear + P4iJ (Trt x Season) + P5ilc (Trt x Year) +

P6Jk (Season x Year) + Plijk (Trt x Season x Year) + Psl[Tree] + eijklm


j = {51,52,53}

£ = {2004,2005,2006}

/ = {Tree 1...25 (Random)}

At the end of these analyses, some questions remained regarding the resin

response during the first year, specifically in the weeks that followed the August

response. An additional analysis (Analysis 5) was performed on the last two

measurements of 2003: 28 August and 25 September. As in previous runs, this was a

58 mixed-effects linear model, using treatment and month (August or September) to model

resin flow (square-root transformed).

yJVolume = P0+ fiuTrt + p2jMonth + P3jj (Trt x Month) + fi4k [Tree] + ei]k

i = {Control, Burning, Pruning, Light trenching, Heavy trenching}

j = {August 2003, September 2003}

k = {Tree 1...25 (Random)}

Two final relationships in the data were also investigated. The first of these was

an analysis of treatment effects on radial growth. This would help evaluate whether, other

than from their effects on resin flow, treatments were sufficiently intense to impact trees.

This was done using a simple one-way analysis of variance on the differences between

treatments in average basal area increment (B AI) values before and after treatment.

Average BAI values from 2004 to 2006 were subtracted from average BAI values from

1998 to 2002; this difference (Diff) was analyzed between groups:

Diff = BAIim:2002 — BAI2004:2006 (6)

Finally, a potentially confounding effect of fire injury on increment cores was

investigated. It was recognized that if the fire intensity was sufficient to kill portions of

the cambium (create fire scars), then cores taken in these regions would be missing

growth rings, and rings from previous years would erroneously be assumed to reflect

2004-2006 growth. For this reason, total radial growth was also analyzed using starting

(2003) and ending (2007) DBH values, and related (via simple linear regression) to total

resin extracted from each tree (again, square-root transformed). This would serve as an

overall test of the hypothesized inverse relationship between radial growth and resin

flow.

59

2.3.4. Results:

Xylem ring width among all sample groups at the start of the experiment (1997-

2002) was 2.8 ±0.60 mm (mean ± standard deviation), or 3.6 ± 0.86 rings/cm. Before any

treatments were applied, resin flow was not significantly different between treatment

groups (Analysis 1; Table 6). There was no evidence that average BAI (BAIav), 813C, or

Foliar mass were significantly related to mid-June resin flow volume. Changing the

order of the model terms also failed to reveal any significant probability values for any

predictors (not shown).

Results from the first post-treatment analysis (Analysis 2, concerning August

values, 2003 - 2006) are shown in Table 7A. The model intercept represents the control

group in 2003, and TrtF, TrtP, Trtr and TrtR represent Fire (burning), Pruning, light root

trenching and heavy root trenching treatments, respectively. The Anova table (Table 7A,

left side) shows that the treatment main effect factor (Trt) was not significant, but the

Year factor and the Trt.-Year interaction factor were both highly significant. The model

summary (Table 7A, right side) contains one point of significant interest: immediately

after burning, burned trees had significantly lower resin flow than controls (TrtF

variable). This was followed in the subsequent 3 years by significant resin increases in

the burned group compared to 2003 (TrtF.-Year2 004, TrtF: Year2 005, and TrtF-.Year2006

variables), particularly in 2004; however, with the terms in this order, the model does not

establish whether this increase results in significantly higher values than in control

treatment trees for the last 3 years, merely that resin flow was significantly higher under

60 these treatment-year combinations than was accounted for by default treatment and year

effects (control treatment and 2003 year). In addition, the Year2006 variable was

significant, suggesting that overall resin flows were lower in that year than in previous

years. None of the other terms were significant in the model, suggesting that neither

pruning nor the trenching treatments affected resin flow measurably.

To further clarify these results, the data (still August averages, 2003-2006) were

analyzed with a slightly simpler model structure, this time with only the Year and

TrV.Year predictor terms (since the main treatment effect, Trt, was non-significant in the

first iteration). The results are presented in Table 7B. The findings are essentially the

same as the full factorial model, with one key difference: in this form, the Year2004:TrtF

variable was non-significant. Although the burned trees still had reduced resin flow

immediately after burning in 2003, subsequent increases were too variable to be

significant in the model. The Year2006 term remained significant, showing an overall

decrease in resin flow that year, as in the previous analysis.

The full dataset included measurements from early snowmelt to late summer.

Figure 10 shows the resin volume values for control trees over the duration of the

experiment - presented as a baseline of variability in untreated trees. The seasonal pattern

is apparent, with one or two modal peaks in mid-summer, and much lower values early in

the spring and early autumn. Analysis of the full seasonal dataset included all 2004-2006

resin measurements (post-treatment years) from all treatment groups (Analysis 3). The

range of values produced is summarized in Figure 11. Several points are immediately

apparent: if the control group (group ' C in Figure 11) is assumed to be the default, it is

61 clear that the frequency distributions of resin volume in most other treatments and trees

are very similar. Only the burning treatment group (group 'F') stands out, with two trees

apparently producing much higher resin volumes at certain times; one tree in the pruning

group also stands out with somewhat higher resin flow. Focusing on the burning group,

the two high resin flow trees are accompanied by two trees with average resin yields, and

one tree with the lowest resin flow of the entire experiment. This outlier has a maximum

resin volume that is lower than the median resin flow of nearly every other tree in the

experiment. Before burning, the tree yielded 0, 0.5, and 0 mL of resin in early 2003, the

lowest values of the entire sample. After burning and in subsequent years, the tree yield

up to 8 mL of resin, still much lower than the maximum yield of the next-lowest tree (17

mL in a control tree).

Table 6. Analysis 1: Pre-treatment (June 2003) summary. DF, SumSq and MeanSq represent degrees of freedom, sum of squares, and mean squares, respectively. See text for further details. Factor Trt BAI av c13.current Foliarmass Residuals

DF 4 1 1 1 17

SumSq 2.161 1.177 0.103 1.328

31.136

MeanSq 0.540 1.177 0.103 1.328 1.832

F 0.295 0.643 0.056 0.725

P 0.877 0.434 0.816 0.406

62

40

35

30

J 2D

i 15

10

5

-2003

-2004

-2005

-2006

0 30-Apr 31-May 01-Jul 01-Ajg 01-Sep 02-Oct

Figure 10. Resin flow for the control trees, 2003-2006.

63 Table 7A. Analysis 2: Post-treatment Parameters:

August resin flow; see text for details. Coefficient estimates (square-root transformed):

Factor DF Variable Value Std.Error DF t

(Intercept)

Trt Year

Trt:Year

AIC:

353.87

Random

effects:

Tree

Table 7B. Factor

(Intercept)

Year

Year:Trt

AIC:

353.87

Random

effects:

Tree

1,60

4,20

3,60

12,60

Analysis DF

1,56

3,56

16,56

355.813

0.314

7.566

2.932

StdDev

1.672

<0.0001

0.865

0.000

0.003

Residual

1.005

2: alternate form, F

355.813

7.566

2.277

StdDev

1.672

P <0.0001

0.000

0.012

Residual

1.005

(Intercept)

TrtF

TrtP

Trtr

TrtR Year2004

Year2005

Year2006 TrtF:Year2004

TrtP:Year2004

Trtr:Year2004

TrtR:Year2004

TrtF:Year2005

TrtP:Year2005

Trtr:Year2005

TrtR:Year2005 TrtF:Year2006

TrtP:Year2006

Trtr:Year2006

TrtR:Year2006

7.312

-2.919

-0.309

-1.483

-1.177

-0.705

0.154

-1.363 4.665

1.017

1.147

1.750

2.523

-0.044

0.101

0.855 2.444

1.164

1.207

1.760 same data as above.

Variable

(Intercept)

Year2004

Year2005

Year2006 Year2003:TrtF

Year2004:TrtF

Year2005:TrtF

Year2006:TrtF

Year2003:TrtP

Year2004:TrtP

Year2005:TrtP

Year2006:TrtP

Year2003:Trtr Year2004:Trtr

Year2005:Trtr

Year2006:Trtr

Year2003:TrtR

Year2004:TrtR

Year2005:TrtR

Year2006:TrtR

Value

7.312

-0.705

0.154

-1.363 -2.919

1.746

-0.395

-0.474

-0.309

0.707

-0.354

0.855

-1.483

-0.336 -1.382

-0.276

-1.177

0.573

-0.322

0.582

0.872

1.234

1.234

1.234

1.234

0.636

0.636

0.636 0.899

0.899

0.899

0.899

0.899

0.899

0.899

0.899

0.899

0.899

0.899

0.899

Std.Error

0.872

0.636

0.636

0.636 1.234

1.234

1.234

1.234

1.234

1.234

1.234

1.234

1.234

1.234 1.234

1.234

1.234

1.234

1.234

1.234

60

20

20

20

20

60

60

60

60

60

60

60

60

60

60

60

60

60

60

60

DF

56

56

56

56

56

56

56

56

56

56

56

56

56

56 56

56

56

56

56

56

8.382

-2.366

-0.251

-1.202

-0.954

-1.109

0.242

-2.144

5.188

1.130

1.276

1.946

2.806

-0.049

0.113

0.951

2.718

1.295

1.342

1.957

t

8.382

-1.109

0.242

-2.144

-2.366

1.416

-0.320

-0.384

-0.251

0.573

-0.287

0.693

-1.202

-0.272 -1.120

-0.224

-0.954

0.464

-0.261

0.472

O.0001

0.028

0.805

0.243

0.351

0.272

0.810

0.036

<0.0001

0.263

0.207

0.056 0.007

0.961

0.911

0.346

0.009

0.200

0.185

0.055

P 0.000

0.272

0.810

0.036 0.022

0.162

0.750

0.702

0.803

0.569

0.775

0.491

0.234

0.786 0.268

0.824

0.344

0.644

0.795

0.639

8 ^

Treatment Figure 11. Medians, interquartiles, and ranges of resin volume by tree and treatment for all post-treatment data. C, F, P, r, and R refer to Control, Fire, Pruning, root trenching (light) and Root trenching (heavy), respectively.

In Analysis 3, the model-building process quickly eliminated most predictors as

non-significant (Table 8). The only continuous covariate that looked significant was 813C

('cl3.current'). The main effects factors Season and Year were highly significant at the

0.05 level, as was as the Season:Year interaction factor. The TrtiYear factor was nearly

significant (p = 0.052). All other variables and interactions were eliminated from further

model iterations.

For the second run, only S13C, Treatment, Season, and Year and the 2-way

interaction terms between categorical variables were included. Although the Anova table

shows that S13C ('cl3.current') was significant, (Table 9), the coefficient estimate (not

65

shown) was not significantly different from 0 (/? = -0.049, p = 0.779). That term was

then excluded from the analysis. The Treatment main effect factor and Trt.Season

interaction term were also excluded as non-significant at this point.

Table 8. Analysis 3, Step 1: Anova table of the initial full model of post-treatment resin volumes. Parameters:

Factor (Intercept) BAI d 3.current Foliarmass_post Trt Season Year Trt:Season Trt:Year Season: Year Trt:Season:Year

DF 1,508 1,508 1,508 1,19 4,19 2,508 2,508 8,508 8,508 4,508 16,508

F 271.611 0.994 4.689 0.246 0.179

104.139 15.234 1.065 1.944 2.734 0.376

P O.0001

0.319 0.031 0.626 0.946

<0.0001 <0.0001

0.386 0.052 0.028 0.988

Table 9. Analysis 3, Step 2. Anova table for a simplified post-treatment resin volume model. Parameters:

Factor (Intercept) c13.current Trt Season Year Trt:Season Trt:Year Season:Year

DF 1,525 1,525 4,20 2,525 2,525 8,525 8,525 4,525

F 286.982 4.606 0.127

102.760 19.490 1.086 2.082 2.792

P <0.0001

0.032 0.971

<0.0001 <0.0001

0.371 0.036 0.026

66 By the third iteration, the model contained little of analytical interest (Table 10).

Even the Year: Treatment interaction, the last term containing an experimentally

manipulated variable, was not significant at the 0.05 level. This last model shows that

transformed resin volume was only affected by measurement season and year, and

potentially their interaction.

Table 10. Analysis 3, Step 3. Anova table and summary for the larger post-treatment resin volume model. As before, resin volume (square-root transformed) was the response variable.

Factor

(Intercept)

Season

Year

Year:Trt Season: Year

AIC:

1751.70

Random

effects:

Tree

DF

1,530

2,530

2,530

12,530

4,530

LL

290.741

101.682

20.984

1.485

2.781

StdDev

1.211

P <0.0001

<0.0001

<0.0001

0.126

0.026

Residual

0.988

Variable

(Intercept) S2

S3

Year2005

Year2006

Year2004:TrtF

Year2005:TrtF

Year2006:TrtF

Year2004:TrtP

Year2005:TrtP

Year2006:TrtP

Year2004:TrtR30

Year2005:TrtR30

Year2006:TrtR30 Year2004:TrtR60

Year2005:TrtR60

Year2006:TrtR60

S2:Year2005

S3:Year2005

S2:Year2006

S3:Year2006

Value

3.312

1.387

0.677

0.401

-0.187

0.888

0.428

-0.148

0.371

0.196

0.408

-0.259

-0.313

-0.300

0.268

0.248

-0.024

-0.035

-0.737

0.050

-0.530

Std.Erroi

0.571

0.161

0.180

0.255

0.306

0.797

0.797

0.802

0.797

0.797

0.802

0.797

0.797

0.802

0.797

0.797

0.802

0.228

0.255

0.273

0.302

• DF

530

530

530

530

530

530

530

530

530

530

530

530

530

530

530

530

530

530

530

530

530

t

5.799

8.597

3.755

1.572

-0.610

1.113

0.537

-0.185

0.465

0.245

0.509

-0.325

-0.393

-0.374

0.337

0.311

-0.030

-0.153

-2.889

0.183

-1.757

P O.0001

<0.0001

<0.0001

0.117

0.542

0.266

0.591

0.854

0.642

0.806

0.611

0.745

0.694

0.708

0.737

0.756

0.976

0.878 0.004

0.855

0.080

67 The fourth model examining these data (Analysis 4) was fitted on the early season

differences in the burned and control treatment groups. Table 11 shows the Anova table

and summary for this model. The intercept coefficient was not significantly different

from zero, suggesting that control trees' resin volumes did not vary significantly from

pre-burn values in 2003. Non-significant probability values for the Year2005 and

Year2006 variables confirm that subsequent years' values for controls were also not

significantly different from zero. Only the treatment effect was significant, and this time

the interpretation is quite clear: burned trees had significant resin increases after

treatment compared with unburned controls. Interaction variables showing differences

among burned trees in subsequent years (a further increase in 2005 and a decrease in

2006) were not significantly different from zero (Table 11), suggesting that the overall

increase in resin volume remained relatively constant across years 2004-2006.

The last analysis of the resin data (Analysis 5) examined late summer resin flow

in 2003, the same year the treatments were performed. The previous models suggested

that only the burning treatment had a significant effect on resin flow, with a decrease

detectable 1 -2 weeks after burning and significant increases in subsequent years. This last

analysis was then designed to identify the duration of the resin reduction and the timing

of the resin increase. Table 12 shows the Anova table and summary for the initial late

2003 model. In the full model, neither the Trt nor Month factors or their interaction was

significant at the 0.05 level. This indicates that less than one month after the burning

treatment (which was implemented 02 August), resin flow among burned trees

(3c = 22.7)had returned to similar levels as controls (3c = 18.2).

68 To test whether the relationship was changing over the two time periods, the Trt

factor was then eliminated from the model. Table 13 shows that, even in this simplified,

model, none of the predictors were significant at the 0.05 level. However, the

MonthSept: TrtF estimate was nearly significant (p = 0.052), with a positive coefficient.

This suggests some evidence that resin flow was increasing among burned trees

(3c = 27.8) compared to controls (3c = 9.7) approximately 7.5 weeks (54 days) after

burning, although the difference remains statistically insignificant at the 0.05 level.

Table 11. Analysis 4. Anova table and summary for resin difference model, burned and control trees only. Factor

(Intercept)

Trt

Year

TitYear

Random effects:

Tree

DF

1,76

1,8

2,76

2,76

F-value

8.822

8.381

0.568

1.066

StdDev

6.037

p-value

0.004

0.020

0.569

0.349

Residual

' 10.647

Variable

(Intercept)

TrtF

Yeary2005

Yeary2006

TrtF:Yeary2005

TrtF:Yeary2006

Value

-0.433

13.533

0.367

1.433

2.900

-5.033

Std.Error DF

3.853

5.449

3.888

3.888

5.498

5.498

76

8

76

76

76

76

t-value

-0.112

2.484

0.094

0.369

0.527

-0.915

p-value

0.911

0.038

0.925

0.713

0.599

0.363

Table 12. Analysis 5, Step 1. Anova table, full model, for the late 2003 resin flow model. Resin volume (square-root transformed) was the response variable. Factor DF F-value p-value

(Intercept) 1,20 264.356 O.0001

Trt 4,20 1.122 0.374

Month 1,20 3.023 0.097

Trt: Month 4,20 1.868 0.156

69 Table 13. Analysis 5, Step 2. Anova table and summary for a simplified late 2003 resin flow model. As before, resin volume (square-root transformed) was the response variable. Factor

(Intercept)

Month

TrtMonth

Random effects:

Tree

DF

1,24

1,16

8,16

F-value

264.356

3.023

1.495

StdDev

0.940

p-value

<0.0001

0.101

0.235

Residual

0.998

Variable

(Intercept)

MonthSept

Month Aug :TrtF

MonthSept:TrtF

Month Aug :TrtP MonthSept:TrtP

Month Aug :Trtr

MonthSept:Trtr

MonthAug:TrtR

Value

4.212

-1.111

0.180

1.812

0.657

0.228

-0.994

0.308

-0.566

Std.Error DF

0.613

0.631

0.867

0.867

0.867

0.867

0.867

0.867

0.867

24

16

16

16

16

16

16

16

16

t-value

6.871

-1.761

0.207

2.090

0.758

0.263

-1.146

0.355

-0.653

p-value

0

0.097

0.838

0.053

0.460

0.796

0.269 0.727

0.523

MonthSeptTrtR 0.030 0.867 16 0.034 0.973

The analysis on treatment effects on B AI showed that trees in the burning

treatment group had the lowest overall growth, which was significantly different from the

control and pruning treatment groups at the a=0.05 level (Table 14). All other group

differences were non-significant. Thus, the two levels of root trenching were not

significantly different from any of the other treatments in terms of radial growth, as

represented by BAI.

In the last investigation of this experiment, the effects of total radial growth (DBH

change) on total resin flow was not significant (p = 0.808). Although the trees that

produced the highest resin flows (the burned group) had the lowest rates of growth,

overall there was no evidence supporting an inverse relationship between radial growth

and resin flow.

70 Table 14. Anova table and Tukey test for the treatment vs. BAI test. Trt signifies treatment, while C, F, P, r, and R represent the treatment levels Control, Fire (burning), Pruning, root trenching (light) and Root trenching (heavy), respectively. Factor DF Sum Sq Mean Sq F p

0.011 Trt Residuals

Tukey's HSD

Contrast: F-C

P-C r-C R-C P-F r-F R-F r-P R-P R-r

4

20

(0.95):

Difference

14.368 2.182 4.009 4.450

-12.186 -10.359 -9.918 1.827

2.268 0.441

609.89 705.57

Lower bound

3.127 -9.059 -7.232 -6.791 -23.426 -21.600 -21.158 -9.414 -8.973 -10.799

152.47 35.28

Upper Bound. 25.608 13.423 15.249 15.691 -0.945 0.882 1.323

13.067 13.509 11.682

4.322

padj 0.008 0.976 0.821 0.760 0.030 0.080 0.100 0.988 0.973 1.000

71

2.4. Experiment 4: Isolated bole charring and resin flow at different heights

2.4.1. Objectives:

Following the first 2 seasons of monitoring in Experiment 3, it had become

apparent that some of the prescribed burn samples were exuding much more resin than

the fire-surrogate treatments (pruning or either of the trenching treatments; see

Experiment 3). This second experiment was then designed to further test the possibility

that bole heating and cambial injury were causing the observed resin increases. This

required the development of a bole charring treatment that avoided any crown scorch or

root damage effects. An additional objective involved evaluating resin response at

different heights above the bole char site. Resin responses were measured before and

after charring, on treated trees and controls at three heights on each sample tree, in order

to reduce individual tree variability and test the vertical extent of the hypothesized

response.

2.4.2. Methods

A pilot study was first conducted in October 2005 near Entiat, WA to refine

treatment methods. The main experiment was held in Sun Pass State Forest, OR, adjacent

to the Experiment 3 site. Pre-treatment sampling and the bole heating treatment were

conducted in August 2006 (sampling August 4) and post-treatment data were collected

one month (September 2, 2006) and slightly over one year (September 6,2007)

afterwards.

72 Twenty two6 high vigor trees (sensu Keen 1943) were selected with diameters

(DBH) between 30 and 55 cm. Half of these were randomly assigned to the bole heating

treatment, while the other half were untreated controls. Prior to sampling and treatment,

dead branches in the lower bole area were removed to allow safe access to the bole for

attaching the resin collection apparatus. Disruption of live crown branches was minimal

on all trees.

Resin flow was measured in all trees at three heights: 1.0 m, 2.75 m, and 4.5 m

(±0.2 m), with two measurements (on approximately opposite sides of the bole) at each

height level. Resin sampling for this experiment was performed according to Lorio et

al.'s (1990) methods, using the resin samplers described by Karsky et al. (2004). Briefly,

at each measurement location, tree bark was shaved and smoothed using a Log Wizard

planing tool (Spruce Grove, Alberta, Canada; Figure 12A) attached to a small chainsaw

body. Within the smoothed area, a circular disc of the remaining bark and phloem was

removed using a 1.27 cm diameter Osborne arch punch tool (Campbell Bosworth

Equipment, TX), which leaves a circular scar of cut bark and phloem tissues but does not

damage xylem tissue (Figure 12B). This OEF sampling method has been suggested to be

more representative than other methods of resin defenses encountered by an attacking

bark beetle, since it does not cause injury to the xylem tissues (Lorio et al. 1990). The

exuded resin was captured by enclosed oleoresin sampler funnels (Karsky et al. 2004)

6 An initial sample size of 24 trees (12 controls, 12 char-treatment trees) was selected for this experiment. However, two samples (one in each of the treatment and control groups) were abandoned following errors: in one case, a fire escaped the confines of the platform and burned some of the forest floor with the tree's drip line; in the second case, one side of a sample tree was accidentally forgotten during pre-treatment resin sampling.

73 directing resin into 50 mL centrifuge tubes. Samplers and attached centrifuge tubes were

placed over the exposed xylem and screwed into place, forming a tight seal (Figure 12C).

Subsequent sampling locations at the three heights were offset vertically and

horizontally from previous sampling wounds in order to minimize the potential effects of

previous sampling in the same vicinity on the outer bole. At each measurement date,

tubes were revisited twice to record total resin volume: 1 and 2 days after phloem

removal (24 hours and 42 or 48 hours)7.

The bole charring treatment was conducted on August 6, 2006, following the pre-

treatment resin sampling. Small platforms (~ lm2) were erected around the 11 treatment

trees at a height of approximately 0.4 m, using wire fencing and commercially available

tile backing board covered in metal foil (Figure 13 A). Charcoal briquettes were piled on

these platforms against the boles of sample trees on two opposing sides, in order to

isolate heating effects to specific areas of boles; treated trees were thus heated along two

quarters of their respective circumferences. Quantities of charcoal were estimated to be

roughly equal per unit of tree circumference. Sections of damp canvas cloth were

wrapped around tree boles beneath the platforms to catch any burning briquettes that

might fall through the platforms to the ground. Briquettes were ignited and left to burn

for approximately 1-1.5 hours following ignition (Figure 13B, C); at that point, any

remaining coals were scraped away. The resulting fire scar was clearly visible, extending

through the bark and into the sapwood xylem (Figure 13D).

7 During the pre-treatment and 1-month post-treatment sampling sessions the second day's measurements were taken at 42 ±1 hours; bark shaving was done first on all trees, and arch-punching and sampler attachment were completed afterwards. During the 1-year post-treatment sampling session, however, for safety reasons (to avoid having to move and reset ladder twice for every tree), resin was measured at approximately 48 hours. This explains the need to reduce the resin flow for the 2007 measurements.

Figure 12. Resin sampling. In clockwise order: A-Use of the Log Wizard planing tool; saw chain was de-toothed prior to use for safety purposes; B- a close-up of a sampling location (photo was taken after resin sampling was completed; visible resin is dripping from sampler screw-holes on either side of the exposed xylem scar, as well as from the scar itself) C: a resin sampler and centrifuge tube installed on a tree, containing a small quantity of resin.

In September 2007 it was noted that many of the previous year's sampling scars

had been attacked by red turpentine beetles (Figure 14). The number of attacks was

75 counted at each height level for further analysis to determine differences by treatment

and height level.

Figure 13. Bole charring treatment photos. Clockwise: A-Platforms and charcoal briquettes assembled and ignited; B-Close-up of burning briquettes; C-Position of briquettes around bole during combustion; D-Close-up of the burn scar on one side of a treated tree.

76 2.4.3. Data Analysis:

Resin flow data were analyzed using linear mixed-effects models with a split-plot

design. Several versions of the model were fitted: the first iteration used pre-treatment

resin volume as the response variable in order to test for inherent (e.g. genetic, site)

differences between individuals in the two treatment groups. The second iteration tested

the differences from pre-treatment values in a repeated measures design (comparing the

two post-treatment measurement times) to test the effects of the charring treatment and

variations in sampling height over two time periods.

Thus, the models were initially constructed with resin flow Volume (24 hour total

from both sides of the tree at each height level) or volume difference (Voldif) fitted using

treatment (Trt, 2 levels: control, burn) and Height (fitted as continuous variable) as the

main fixed factors. Volume was square-root transformed, as before, to stabilize sample

variances. The post-treatment model for Voldif also included the Time factor, set as a

fixed-effect categorical variable since the two post-treatment measurement times were

chosen deliberately and with no expectation of a unidirectional response over time (2

levels: 1 month and 1 year post-treatment). Of greater interest was the Trt x Time

interaction, which would assess whether the two treatments (char and control) had similar

effects at the two time periods. Since multiple measurements were made on individual

trees (measurements at different heights were not independent), the random factor Tree

was needed to represent the nested structure.

Thus, the two mixed-effects model structures were initially designed as follows:

77

•jVolpre = p0 + PuTrt + p2Height + p3i (Trt x Height) + p4j [Tree] + eiJk

i = {Control, Char}

j = {Tree 1...25 (Random)}

Height - {Actual Measurement height, m}

Voldif = p0+ puTrt + P2Height + PyTime + p4i (Trt x Height) + P5ij (Trt x Time)

+P6J (Height x Time) + P7jJ (Trt x Height x Time) + Pu[Tree] + eiJkl

i = {Control, Char}

j - {l month post-trt, 1 year post-trt}

k = {Tree 1 ...25 (Random)}

Height = {Actual Measurement height, m}

Volpre and FbM/*represent pre-treatment resin volume, and the post-treatment

minus pre-treatment resin volume difference, respectively. Trt and Time are the

categorical predictors described above, Height is a continuous covariate, and e represents

the error term. Final models were chosen based on significance of the terms, model

simplicity, and lowest Akaike's Information Criterion (AIC) values (Pinheiro and Bates

2000). Differences were evaluated at the a =0.05 level of significance.

2.4.4. Results:

As noted above, by September 2007, many sampling scars from the previous year

had been attacked by red turpentine beetles, as evidenced by their large and unmistakable

pitch tubes (Figure 14). However, the beetles attacked the majority of scars irrespective

of height or treatment (Figure 15). Because of the potentially confounding effects of

having intermixed charred trees and controls in close proximity (as well as very similar

78 mean attack proportions, irrespective of treatment or height level), no further analysis of

the beetle attack data was undertaken.

Figure 14. By late 2007, most of the resin sampling scars from the previous year had been attacked by red turpentine beetles.

79

Figure 15. Mean proportion of sampling scars attacked by turpentine beetles; error bars represent the range of values by tree and height level. Green and red symbols represent control trees and charred trees, respectively. Height levels are horizontally offset for display only, as all resin samplers were placed at one of the three height levels indicated.

In terms of resin flow, high variability was encountered between different trees

and measurements, as expected from previous trials and experiments on oleoresin flow.

The sampling methodology was also not free of problems, with several instances of

observed resin leakage between the bark-sampler interface, as well as a few cases where

resin was suspected to have drained into bark crevices at the bottom of the punched scar

area instead of draining into the resin sampler. As in Experiment 3, instances where

leaking was identified had 0.5 mL added to their resin volume quantities.

Initial model fitting was attempted as described above. In total, 396 samples were

taken (22 trees, 3 time periods, 3 height levels, 2 sides measured per height level);

measurements on opposite tree sides at each height level were totaled, giving 198 data

80 points for analysis. Mean resin volumes shown by treatment group and time period are

represented in Figure 16. Pre-treatment resin flow was not significantly different between

control trees (x = 6.3 mL) and char treatment trees (x - 4.5 mL), as expected with

randomly assigned treatments among trees of relatively homogenous vigor (Table 15). In

addition, pre-treatment resin flow was not significantly affected by Height or the

TrtHeight interaction.

12

10

6 6

2

Pre-treatment: Aug.2006

<V 4>

1-month post

Sampling height (m)

1 -year post

l Ctrl

I Char

Figure 16. Summary of mean resin volume at the three time periods in the two treatment groups.

81 Table 15. ANOVA table for the initial pre-treatment resin flow model.

Factor DF F p (Intercept) 1,42 227.812 <0.0001 Trt 1,20 2.040 0.169 Height 1,42 2.447 0.125 Trt:Height 1,42 0.290 0.593

As previously described, post-treatment resin flow was modeled using differences

in resin volume from pre-treatment levels at the 1-month and 1-year time periods. In the

initial full factorial model, the Trt:Height:Time 3-way interaction, the Trt:Height

interaction and Height main effect were all non-significant; of the terms containing the

Height variable, only the Height .Time term was significant (Table 16). The pattern is

evident from Figure 16 (above): the relationship between post-treatment resin volume

and height is clearly different over the two post-treatment time periods - at 1 month post-

treatment, resin volume increased with height, and at 1 year post-treatment, resin volume

decreased with height. Since this resin-height relationship is equally evident in control

trees and treatment trees (as indicated by the non-significant 3-way interaction term), it

was likely caused by factors external to the question of interest here. Based on non-

significance of these terms, the Height variable was dropped from subsequent model-

fitting.

The parameters and estimates of the simplified post-treatment resin model are

given in Table 17. The Trt main effect factor was not significant at the a =0.05 level,

indicating similar overall changes in resin flow volume averaged over the two post-

treatment times. However, both the Time main effect and the Trt: Time interaction were

significant, reflecting both overall differences in resin volume change at the two time

82 periods and the opposing effects of the Time factor on the two different treatments.

Because the two remaining predictor variables are categorical, parameter estimates can

be easily added together to give group means. Thus, control trees had slightly increased

resin flow at time 1 (after 1 month: x = +1.7 mL, p = 0.019) and time 2 (following

year: x = +1.742 - 0.576 = +1.2 mL, not significantly different from +1.7 mL);

meanwhile, charred trees had a slight decrease after 1 month ( x = 1.742 - 2.273 = —0.5

mL; p = 0.041, significantly different from control group at time 1) and a stronger

increase the following year (x = 1.742 - 2.273+4.409 = +3.3 mL; p < 0.001, significantly

different from charred trees at time 1), compared to the increase in the control group. See

Table 17 for analysis details. Figure 17 shows the summary of resin flow at the three

measurement times, organized by group (excluding height differences).

Table 16. Resin difference (post-treatment - pre-treatment): Anova table of the initial full factorial model.

Factor DF F 9 (Intercept) 1,104 12.005 0.001 Trt 1,20 0.0069 0.935 Height 1,104 0.124 0.725 Time 1,104 7.308 0.008 TrtHeight 1,104 1.285 0.260 Trt:Time 1,104 13.387 <0.001 Height:Time 1,104 14.370 <0.001 Trt:Height:Time 1,104 0.514 0.475

83 Table 17. Parameters and estimates for bole-char experiment, post-treatment resin difference model. See Methods section for model term descriptions. Default cases represented by the Intercepts for variable estimates are the control treatment (Trtl) for the pre-treatment model; and the control treatment (Trtl) and 1-month time period (Timel) for the post-treatment resin difference model. Thus, Trt2 refers to the bole char treatment and Time2 refers to the 1-year post-treatment time period. Post-treatment/Resin difference analysis (24hr resin volume): Parameters: Coefficient estimates:

Factor DF Variable Value Std. Error DF t

(Intercept)

Trt

Time Trt:Time

1,108 12.005 <0.001

1,20 0.007 0.935

1,108 6.561 0.012

1,108 12.019 0.001

(Intercept) Trt2

Time2

Trt2:Time2

Random effects:

1.742

-2.273

-0.576

4.409

StdDev I

0.734

1.038

0.899

1.272

Residual

108

20

108

108

2.375

-2.190

-0.640

3.467

0.019 0.041

0.523

0.001

Tree 1.214 3.653

co H

c 'to 01

* H

o H

Pre 1 month 1 year

Time

Figure 17. 24-hour resin volumes on all trees (ignoring height effects), by time and treatment group. Symbols represent treatment means (green: control; red: charred) and whiskers are standard errors.

84 Other iterations of these models were fitted, all yielding similar results. Variations

included using Height in the model as a categorical variable, and using second day (42-

or 48-hour) resin volume as the response (ignoring differences in resin flow duration

between years). The latter was considered after comparing the proportional differences

between 24- and 42-hour quantities in pre-burn and 1-month sampling periods: across all

heights and both treatment groups, 2006 measurements had a mean proportion of 85.8%

of 42-hour resin volume exuded in the first 24 hours (ignoring values of 0); comparing

values in the 1-year sampling period (24-hour volume and 48-hour volume) gave a

proportion of 81.4%, with a similar variance around the mean. The 4.4% difference in

proportions (unbiased between treatment groups) was negligible compared to the

between-tree variability, and the original model was used to fit the second-day (42/48-

hour) resin volume. As in the model results for 24-hour resin samples (Table 17), both

the Time main effect and the Trf.Time interaction were significant at the a =0.05 level;

coefficient estimates and other model statistics varied only slightly, causing no changes

to overall results or interpretation. Compared to control trees and pre-treatment

measurements, bole-charred trees had significantly lower resin flow approximately 1

month after treatment, and significantly higher resin flow approximately 1 year after

treatment.

85

2.5. Experiment 5: Fire effects on resin chemistry

2.5.1. Objectives:

As described in Section 1.2, oleoresin plays an important role in conifer defenses,

with evidence suggesting that both resin quantity and composition affect tree defenses.

This experiment focused on potential changes to resin composition following prescribed

fire, focusing on the monoterpenes - the primary resin fraction known to be volatile,

toxic (to beetles), and highly important in beetle semiochemistry. Techniques for

separating and identifying monoterpenes are well developed and dependable using

modern laboratory techniques.

The present analysis was relatively simple and exploratory in nature. Although

resin chemistry is a complex topic that could be explored in depth - examining

monoterpene enantiomers separately, for instance (Pureswaran et al. 2004) - this

experiment consisted of a simple collection of resin from burned and unburned trees at

two separate study sites. Monoterpene analysis was done using gas-liquid

chromatography. Special emphasis was placed on limonene, 3-carene, and myrcene due

to the importance of these three monoterpenes in western pine beetle chemical ecology

and toxicity.

2.5.2. Methods and data analysis:

Monoterpene composition changes were assessed based on resin collected from

46 trees at two locations, each with very different site characteristics: The first site was

86 Annie Creek, within Winema National Forest, OR, approximately 1 km from the

Experiment 3 location, and subjected to a spring prescribed burn in 2006; the second site

was in the Preston Creek drainage, in the Entiat River watershed near Ardenvoir, WA,

and subjected to a fall prescribed fire in 2006. The Annie Creek site was very close to the

previously described Sun Pass area in Experiments 3 and 4, although the stand selected

for this experiment was composed of larger, mature size ponderosa pines located in the

thin land section managed by Winema National Forest adjacent to Crater Lake NP and

Sun Pass SF (see Figure 3, ACW site). Trees at Annie Creek varied between about 60

andllOcmDBH.

The Preston Creek site was located in the eastern Washington Cascades in an area

with a history of at least two large wildfires in recent memory - the 1970 Entiat fire and

the 1994 Tyee fire, both being landscape-level high intensity events; the stand was also

treated with a lower-intensity prescribed fire in October 2005. The specific stand selected

for this experiment was uneven-aged (30-65 cm DBH), including younger pine

individuals (ponderosa and lodgepole) that likely germinated following the 1970 event as

well as older veterans of larger size- and age-classes (many with fire scars); other tree

species observed in the drainage included Douglas-fir and trembling aspen. Although no

fire behavior information was available from either Annie Creek or Entiat, fire effects

appeared more severe at Entiat, with all litter and duff consumed around sample trees,

and several patches of complete crown consumption within the surrounding forest stand.

The fire at Annie Creek was reported as within the burning prescription (Deb Shepherd,

87 Winema National Forest, personal communication), and showed signs of having burned

at lower intensity levels than Entiat.

Resin was collected from burned and unburned trees in both sites in September

2007. At the Sun Pass site, usable samples were obtained from 25 trees (13 burned, 12

unburned). Two unburned trees failed to produce any resin at all during the sampling

period. At the Entiat site, 24 trees were originally selected for sampling; 3 unburned

individuals yielded no resin, resulting in viable samples from 21 trees (12 burned, 9

unburned).

Methodology for resin collection was similar to Experiments 2 and 3 with minor

differences: in each tree, a 5.56 mm (7/32 inch) hole was drilled close to the ground (10 -

20 cm height) at an upward angle into the sapwood. A brass tubing funnel directed

emergent resin into a 4 mL glass vial (Entiat) or centrifuge tube (Sun Pass). Resin was

exposed to open air for 1-5 hours, as required to obtain a sample of at least 1 g.

Following collection, vials were capped and kept on ice for 1-2 days before being

transferred to a laboratory freezer for storage prior to analysis. Unfortunately, the Sun

Pass samples had to be transferred from centrifuge tubes to glass vials prior to analysis.

This involved allowing them to warm to room temperature before transferring the resin

(pouring and scooping) from the tubes to vials. At this time it was noted that considerable

resin crystallization had occurred.

Samples were sent to the Colorado Plateau Stable Isotope Laboratory (Northern

Arizona University, Flagstaff, AZ) for chromatographic analysis. Resin samples were

dissolved in hexane, vortexed, and then left to stand for 24 hours to ensure resin

88 dissolution. A known volume of sample (30-75 uL) was measured using a gastight SGE

syringe and transferred into a 5 ml volumetric flask containing 50 uL of standard

fenchone. The sample and internal standard were diluted with n-hexane to the 5 ml mark.

Samples were transferred to analysis vials using a glass Pasteur pipette.

Monoterpenes analysis was performed on an HP5890 gas chromatograph (HP-

Agilent Technologies, Santa Clara, CA) equipped with a DB-1 capillary column (J&W

Scientific, Folsom, CA) and an HP Automatic Liquid Sampler. On-column inlet

temperature was 275°C and the helium flow pressure was set to 10 PSI. The following

temperature program was used: 40°C for 10 minutes, increase to 165°C at a rate of

2°C/minute, hold for 0.5 minutes, increase to 300°C at a rate of 20°C/minute, hold 2

minutes. Eluting was measured with a flame ionization detector (FID) set at a

temperature of 300°C. Peaks were identified using retention time on the chromatogram

and quantified with a calibration curve developed from standard laboratory chemicals.

PeakSimple chromatography software (SRI Instruments, Torrance, CA) was used to

measure and record data.

Data analysis in this experiment was relatively simple: the first comparison was

between the total proportion of monoterpenes (as a fraction of total resin mass) between

burn and control treatment groups. The difference between means was analyzed using a

two way analysis of variance (factors Site and Treatment). Individual monoterpene data

were analyzed using the same 2-way ANOVA model on the differences in monoterpene

proportion (per mille (%o) of total monoterpene mass) between sites and treatment

groups.

89

2.5.3. Results

Overall, the mean monoterpene fraction in resin samples was 246.8 mg/g resin,

with a standard deviation of 66.8 mg/g; thus, approximately one quarter of the resin

weight from all samples consisted of the monoterpene component. Despite the short

exposure time during sampling (~l-5 hours total), resin crystallization was observed on

the walls of several sampling vials during collection and storage, suggesting that resin in

these samples was reacting with ambient air, with monoterpenes volatizing. Among the

control trees at both sites, monoterpene proportions were similar (p=0.582). However, the

two-way ANOVA (factors Site and Treatment) had a significant Site: Treatment

interaction effect: at Entiat, burned trees (307.21 mg/g resin) had higher monoterpene

proportions than control trees (235.33 mg/g resin), while at Sun Pass burned trees

(220.17 mg/g resin) had slightly lower (N.S.) monoterpene proportions than controls

(223.22 mg/g resin) (Table 18).

The GLC analysis tested for and detected 7 monoterpenes in various proportions

in the samples. In terms of ranking, the most abundant monoterpene was 5-3-carene,

followed by myrcene, followed by (3-pinene (absent from a few samples), followed by a

tie between a-pinene, limonene and terpinolene (Table 19); fewer than half of the

samples (mostly those from the Sun Pass site) also had trace levels of longifolene.

Relative ranking of monoterpene compounds was affected by some of them being absent

(or below detectable limits) from some samples.

90 Proportions of individual monoterpenes (as a fraction of total monoterpene mass)

varied considerably between individuals, with few patterns standing out. This experiment

did not control for tree size, age, or pre-treatment resin properties - all of which could

arguably confound real treatment effects. In any case, none of the tested differences was

significant at the a=0.05 level. Limonene proportions were not significantly different

between burned and unburned trees at Entiat (p = 0.363) or at Sun Pass (p = 0.151).

Mean limonene proportions actually moved in opposite directions at the two sites, with

slightly lower (N.S.) proportions in burned trees at the Entiat site, and slightly higher

(N.S.) proportions in burned trees at Sun Pass. Likewise with myrcene - burned trees at

the Entiat site had slightly lower (N.S.) myrcene levels than controls, while at Sun Pass

burned trees had slightly higher (N.S.) myrcene levels than controls (Entiat: p=0.137;

Sun Pass: p=0.359). Differences in 3-carene proportions, the most abundant

monoterpene, were similarly non-significant (Entiat: p = 0.467; Sun Pass: p = 0.788);

slight decreases (N.S.) in mean 3-carene proportion in burned trees at both sites were lost

amidst the high within-group variability (Figure 18).

91 Table 18. Linear model results for total monoterpene proportion (mg monoterpene/ g of resin) as a function of Site, treatment (Trt), and the Site:Trt interaction. Default case {Intercept in model) refers to control treatment at Entiat site; SP refers to Sun Pass site, and Trt2 refers to the burn treatment.

Factor DF SumSq MeanSq F p Site 1 34086 34086 10.118 0.003 Trt 1 10803 10803 3.207 0.081 Site:Trt 1 15829 15829 4.699 0.036 Residuals 42 141494 3369

Variable Value Std.Error t p (Intercept) 235.33 19.35 12.164 <0.001 SiteSP -12.11 25.17 -0.481 0.633 Trt2 71.88 25.59 2.809 0.0075 SiteSP:Trt2 -74.93 34.57 -2.168 0.036

Table 19. Mean proportions and ranking of monoterpenes at both sites (mg monoterpene/ g resin). Rank refers to the average rank within each sample group of each monoterpene.

a-pinene p-pinene Myrcene 5-3-carene Limonene Terpinolene %o Rank %o Rank %o Rank %o Rank %o Rank %o Rank

Entiat Control 9.36 4.3 11.49 4.6 42.94 2.2 149.94 to 10.84 4.7 10.02 4.1 Bum 13.06 4.6 33.49 3.7 47.98 2.4 190.09 1.0 7.73 5.3 13.83 4.3

Sun Pass Control 11.47 4.3 33.61 3.6 36.80 2.6 124.15 1.0 7.95 5.2 7.85 4.6 Burn 13.70 4.6 23.03 3.9 40.75 2.3 120.43 1.0 14.19 4.2 7.91 5.0

92

M o n o t e r P e n e

200

150

100

50

0 ^ { \ 1

® A n

i A t • j _

T

A

Site: Entiat •^«^— linionene p : 0.363 rayrcene p : 0.137 3-carene p : 0.467

?(

Sun Pass • 0.151 0.359 0.788

u

r 800

700

600

500

400

300

200

100

0

Figure 18. Normalized mean proportions (mg monoterpene/ g total monoterpenes) and standard errors (whiskers) of limonene, 3-carene, and myrcene by site and treatment. Proportion-values refer to t-test differences between treatments - control (open symbols) or burned (closed symbols).

93

2.6. Experiment 6: Methods testing between resin sampling techniques

2.6.1. Objectives:

This experiment was designed to test and quantify some of the variability in

sampling methodology, as well as test certain assumptions regarding resin sampling in

general. Since attacking bark beetles do not penetrate deeper than the outer face of the

xylem (Miller and Keen 1960), there is some question as to the utility or validity of

measuring oleoresin properties deeper in the sapwood (Lorio 1994). As discussed

previously, OEF methods that involved removing larger sections of bark have the

disadvantage of causing much larger scars, potentially causing cumulative and severe

injuries to sample trees. However, using an arch punch tool to cut through the bark and

phloem has the advantage of avoiding injury to the xylem, since it is a relatively easy

task for the operator to feel when the tool has penetrated the bark and phloem. In

addition, the arch punch method only measures resin exuded from the outer face of the

xylem (or phloem), which is potentially more representative of defenses encountered by a

bark beetle than deeper in the sapwood (see Section 1.2). Thus, it has been suggested that

this technique might provide a more realistic index of beetle susceptibility than drilling

into the sapwood xylem (Lorio 1993,1994).

The counter argument is that vertical and radial resin canals are interconnected in

continuous networks, and therefore, resin mobilized to wound sites can originate from

deeper in the sapwood (Section 1.2.2). Furthermore, resin pressure should be fairly

constant within a tree (Vite 1961) as water can travel between sapwood xylem cells in all

directions (Vite 1961, Tyree and Zimmerman 2002). However, resin pressure and resin

94 flow or volume are not equivalent, and as mentioned in previous sections, several authors

have observed wide variations in intra-tree resin flow, particularly in the case of trees

experimentally treated on only one side. This experiment was designed to test the

assumption of equivalency between resin flow exuded from the xylem face and from

deeper in the sapwood; this was evaluated by measuring OEF using two methods

simultaneously in individual trees. Based on the logic of interconnected vertical and

radial resin canals, resin flow measurement using these two methods can also be thought

of as exposing a certain surface area of sapwood xylem to the air, in either a cylindrical

(drilling) or circular (arch-punch) shape. Therefore, one of the analyses (see below)

tested this simple geometric area model as a predictor of resin flow using both

techniques.

2.6.2. Methods

Sampling for this experiment was conducted in two separate sampling periods

and locations. The first iteration took place 24-25 September 2005 on 14 mature

ponderosa pines widely scattered near Leavenworth, WA, along the Icicle Creek and

Wenatchee River valleys. The arch-punch sampling method was previously described in

Experiment 4. This first iteration and data set involved some protocol differences from

those described previously, however: instead of using the Log Wizard tool to smooth the

bark at the sampling location, bark thickness was shaved by hand using chisels, similar to

the methods described by Lorio (1993). Additionally, a 2.54 centimeter (1-inch) diameter

arch-punch tool was used rather than the smaller 1.27 cm in previously described

95 experiments. Also, rather than using the enclosed resin samplers, resin was tunneled from

each circular scar to the centrifuge tube using small sections of sliced plastic tubing

material that were pinned to the bark below the scar (Figure 19). The protocol for the

drilled OEF method being contrasted was identical to that described in Experiments 2

and 3. Two arch-punch and two drilled resin samples were taken per tree, and resin

volumes were verified the following morning (14-15 hours later).

The second iteration of this experiment took place September 15, 2007 in the

Entiat River Valley, ~ 1 km from the location where monoterpene samples were collected

(see Experiment 5). This time the Experiment 4 protocols were closely followed for the

arch punch technique: bark shaving was accomplished using the Log Wizard tool, a 1.27

cm (0.5 inch) diameter arch punch was used, and plastic resin sampler attachments

(Karsky et al. 2004) were screwed in place over the sampling scar. At this site, the trees

being sampled consisted of 10 pole-sized ponderosa pines that had been burned the year

before in a fall prescribed fire (see Experiment 5). Again, the drilled OEF protocol being

contrasted was identical to the methodology described in Experiments 2 and 3. However,

the sampling time period was very short - only 2.5-3.5 hours elapsed between the

equipment installation times and the resin volume measurement times. To avoid any time

bias between the two methods, drilling and phloem removal (with the arch-punch tool)

were performed sequentially from tree to tree (two drilled samplers and two arch punch

samplers installed on one tree before moving to the next tree).

This experiment section also describes the findings from a brief microstudy of

ponderosa pine resin canals. Three small (< 20 cm dbh) ponderosa pines were cut from

96 the Sun Pass site, and discs from each were sanded smooth with 800 grit sandpaper.

Blocks from two of the trees' discs were cut so as to expose radial and transverse sections

of the xylem, and then sanded as before. The sanded discs and blocks were examined for

approximately 1 hour each at up to 200x magnification, with a particular focus on

searching for any evidence of transverse resin canals.

Figure 19. Example of measuring resin flow using two different methods on one tree at Leavenworth, WA. The exposed xylem of the arch-punch scar is visible on the right.

97 2.6.3. Data Analysis

Due to the differences in experimental protocols (primarily the 1.27 cm vs. 2.54

cm arch punch diameters used at the two sites), the two experiment iterations were first

analyzed separately. Linear mixed-effects models were used in both cases in a split-plot

design (Pinheiro and Bates 2000). Resin Volume measured at each individual sampler

was the response variable since the objective was to test the difference between methods

rather than between trees. Because 4 measurements were taken from each tree, Method

was the main fixed factor, with measurements grouped by the random factor Tree. As

before, raw resin volumes were square-root transformed to stabilize variances.

Differences were evaluated at the a=0.05 level of significance.

The surface area analysis used simple geometric equations to compare resin flow

collecting in terms of mL of resin collected per cm2 of xylem exposed. Thus, the 1.27 cm

arch punch tool exposed a circle of % • r2 {r is radius of the circle) or n x ^1.27cw^2

about 1.3 cm . The value for the 2.54 cm arch punch is about 5.1 cm . Modeled as a

simple one-sided cylinder with height of 3.8 cm (based on an estimated hole depth of 1.5

inches, as described in Experiment 2 and 3 methods) and diameter of 0.56 cm, the drilled

hole method exposes n • r2 + c • h (c is circumference, h is cylinder height), or

n • (0.556 /2)2 +;r-0.556 -3.81 = about 6.9 cm2 of xylem area. These values do not

account for the slight curvature on the face of tree boles, nor for the angled outer cylinder

end for the drilled method. Considering the limited precision of these techniques, these

approximations are probably adequate. Based on this geometric model, resin yields using

98 3 different techniques (drilling, 1.27 cm arch punch, 2.54 cm arch punch) were compared

using a simple bivariate correlation; to reduce within-tree error and simplify the analysis,

resin quantities were evaluated as totals per tree using each method.

2.6.4. Results

Between-tree variability differed between the two iterations of this experiment,

with the Leavenworth site trees being much more variable than the Entiat trees. Table 20

gives the summary statistics from both sites. The main difference between sites is that the

sampling with the smaller arch punch tool (Entiat) yielded significantly lower resin

volume than the drilling method (p = 0.0063). At the Leavenworth site, the difference

between the resin yield from the larger arch punch sampler and the drilled method was

not significantly different (p = 0.800).

The exposed xylem model area produced a strong correlation between the two

resin collection methods (slope = 0.926, p < 0.001, r = 0.855; Figure 20). Although some

unexplained error remains, particularly around the Entiat sample values, this model

supports the concept of resin flow measurements being largely determined by exposed

xylem surface area. The brief microstudy of tree discs revealed images very similar to

those in existing literature (e.g. Fahn 1979). Vertical resin canals were widespread, radial

canals were common but somewhat less abundant, and there was no evidence of

transverse canals.

99 Table 20. Parameters and estimates for resin method correlation experiment. Default cases represented by the Intercept terms represent the arch punch method (Methodl), while MethodD refers to the drilling/brass funnel method. Estimated values represent square-root transformed means. DF indicates degrees of freedom (numerator, denominator). Entiat site (1.27 cm arch punch):

Parameters: Coefficient estimates:

Factor DF F rj Variable Value Std.Error DF t p

(Intercept) 1,29 115.721 <0.0001 (Intercept) 1.725 0.227 29 7.611 <0.0001 Method 1,29 8.683 0.0063 MethodD 0.701 0.238 29 2.947 0.006

Random effects: StdDev Residual

Tree 0.480 0.752

Leavenworth site (2.54 cm arch punch):

Parameters: Coefficient estimates:

Factor DF F rj Variable Value Std.Error DF t p

(Intercept) 1,41 31.868 O.0001 (Intercept) 3.216 0.579 41 5.558 O.0001

Method 1,41 0.065 0.800 MethodD 0.040 0.158 41 0.256 0.800

Random effects: StdDev Residual

Tree 2.124 0.591

100

0 1 y=0.9262x+1.1162

R2^ 0.7303

Figure 20. Xylem surface area model showing correlation between resin sampling methods. The Entiat sample (pink squares) used a 1.27 cm diameter arch punch on burned trees, and the Leavenworth sample (blue triangles) used a 2.54 cm arch punch on unburned trees. Each scatter point represents the resin volume for one tree. The green line is the line of perfect correlation (y = x) between methods, and the black line is a fitted linear regression using data from all trees.

101

3 . Discussion

3.1. Oldponderosa pines at Crater Lake NP

Monitoring tree populations over time can help assess the long-term effects and

recovery following management treatments. Experiment 1 demonstrated that post-

treatment mortality following prescribed fire in these forests can continue for several

years. By the end of the monitoring period (4 years after burning), ponderosa pine

mortality had declined in spring burn units to match the level associated with control

units (in 2006, 2 trees died in each treatment type in, 0.3% of pre-treatment populations).

Mortality in fall units was declining steadily as well, although was still elevated

compared to controls (13 trees died in 2006, 2.4% of pre-treatment population).

3.1.1. Crown vigor and radial growth

Continuing the pattern found in previous surveys at this site (Perrakis 2004),

mortality was strongly associated with Keen's (1943) crown class. Although this type of

risk-rating system is very simple, involving only a visual estimate of crown vigor, it was

also significantly (inversely) correlated with the number of rings in the outer centimeter

in these trees. Thus, trees in A and B classes had higher radial growth in recent years than

those in C or D classes, and the former were more likely to have survived for 4 years

after burning, regardless of burn season or control treatment.

102 It is unsurprising that higher vigor should correspond with higher survivorship, as

trees with well developed root and crown systems tend to have high carbohydrate

reserves and are more resilient to environmental stresses and pathogen attacks (Waring

1987, Kozlowski and Pallardy 1997). The present findings corroborate this, with high

crown vigor class being positively correlated with wider growth rings, higher resin flow

(see below), and lower mortality rates. Several previous studies in ponderosa (Larsson et

al. 1983) and other pines (Raffa and Berryman 1982, Stuart 1984) have also found higher

radial growth indices to be positively related to tree survival following bark beetle

attacks.

3.1.2. Burning treatments

The mortality levels observed in this study (2.3 %, 6.1 %, and 16.4 % in control,

spring burn, and fall burn treatments, respectively, after four years) are similar to or

slightly lower than levels reported in other studies on ponderosa pine survival after fire.

Thies et al. (2005) studied mortality in uneven-aged ponderosa pine stands (7.5 to >75

cm DBH) in eastern Oregon. They noted between 0-0.6 % annual mortality in control

plots, and mortality in burn plots that peaked one year after burning (spring burns: 4.3 %,

fall burns: 11.8 %) and leveled off (matching control levels) after four years of post-burn

monitoring - a nearly identical pattern to that observed in the present study. Their fall

burn treatments caused higher mortality overall (29 % of trees dead after four years), but

the uneven-aged population structure suggests that many of the killed trees may have

103 been much smaller than those in the present study, and therefore susceptible to mortality

from severe crown scorch (Harrington 1993).

As reported previously (Perrakis 2004), earlier studies at Crater Lake observed

higher mortality in spring burns than in fall burns (Swezy and Agee 1991, Agee 2003).

However, a review of the literature on ponderosa pine mortality reveals a mix of findings

on seasonal fire effects. Although phenology may suggest that early season burn effects

are most likely to be lethal due to reduced carbohydrate stores (Kozlowski 1992),

prescribed burns later in the season tend to have higher fire intensities and proportional

coverage (Ganz et al. 2003, Perrakis 2004, Knapp and Keeley 2006, Schwilk et al. 2006).

A recent study comparing spring and fall burning in a Sierra Nevada mixed-conifer forest

found that ponderosa and sugar pine mortality were significantly related to fire intensity,

while mortality was not significantly different overall between the two burning seasons

(Schwilk et al. 2006). In northern Arizona, McHugh and Kolb (2003) compared spring

and early summer wildfires to a fall prescribed fire. They observed that the most

significant predictors of mortality in logistic regression models were crown damage

(scorch and consumption) and bole char - factors closely related to fire intensity. In the

present study, with spring burns having significantly lower coverage and visibly lower

intensities than fall burns (Perrakis 2004), a similar pattern emerges. Although season of

burn and tree phenology should not be discounted, the intensity of a particular fire is

probably a more important variable overall in predicting mortality.

For management purposes, it is promising that mortality has declined steadily

every year after post-burn year 1, even in the intense fall burn treatments. Thirteen years

104 after a series of very similar burns at another Crater Lake site, Agee (2003) found no

significant difference between large ponderosa pine mortality in burned units and

unburned controls. Ideally, post-burn mortality should be below control unit mortality

levels after a few years. Since these sites have been identified as suffering from low

forest health due to large increases in tree density and competition in the fire-exclusion

era (McNeil and Zobel 1980, Agee 2003, Perrakis, unpublished vegetation data),

restoration treatments are intended to promote increased survivorship among remaining

old pines (Kolb et al. 2007). Reaching a point of equal mortality levels between burn

treatments and controls after a few years is a good start, but is not indicative of long-term

success according to the restoration criteria. Due to the generally low levels of mortality

among untreated stands, even among stressed trees, identifying enhanced post-treatment

survivorship will require longer-term monitoring efforts.

3.1.3. Resin flow monitoring at Crater Lake

The Crater Lake burning and resin monitoring experiment presents one of the

longest annually repeated post-fire resin monitoring efforts in the literature. The pattern

of increased post-fire resin flow identified in numerous other studies (Santoro et al. 2001,

Perrakis 2004, Knebel and Wentworth 2007) is hereby confirmed to continue for up to

four years after fire in old-growth ponderosa pines. The results suggest that four years

after burning, bole resin flow (likely representative of a combination of constitutive and

induced resin) may still be higher in burned trees than those in unburned controls. The

relative increase due to fire injury appears to be declining in the spring burns but remains

105 elevated among trees in fall burn treatments. Although the analyses suggest that resin

flow peaked in year two after burning for both spring and fall burn treatments, between-

year differences were not statistically significant; some problems with overflowing test

tubes in year 1 may also have masked higher resin flow values in that year. The analysis

of differences between post- and pre-treatment resin flow among fall burns and controls

followed the same pattern of a greater resin increase among burned trees than controls.

Absolute resin flow was also greater in trees of high crown vigor class. As crown

vigor class was inversely correlated with rings per cm (Experiment 1), this suggests a

positive correlation between constitutive resin flow and sapwood xylem production

(although this was not tested in the same trees; cored trees and resin-tapped trees were

different samples). While this does not support the GDB predictions (see below), it is in

agreement with the findings of McDowell et al. (2007), who found positive correlations

between basal area increment (BAI) and resin flow in a survey of several different studies

in southwestern ponderosa pine stands. In Experiment 3, the BAI factor was not a

significant predictor of resin flow in any of the analyses. In that experiment, burned trees

had the lowest rates of growth post-treatment, but before treatment (and in the overall

analysis), there was no significant relationship between BAI and resin flow. In the survey

study by McDowell et al. (2007), the authors remarked that the positive BAI-resin flow

relationship showed variability within narrower BAI ranges (e.g. Zausen et al. 2005), and

that only across a wider range of growth values did the relationship become significant.

In this study, the overall relationship may have been masked due to small sample sizes or

similar pre-treatment BAI values between trees at the Sun Pass site.

106

3.2. Fire injury, surrogate treatments, and resin flow

Experiments 3 and 4 attempted to reproduce prescribed fire effects on resin flow

via controlled injury treatments to tree structures. Root trenching and lower crown

pruning were the injury surrogate treatments in Experiment 3, and bole charring was the

surrogate treatment in Experiment 4, with various tree growth and size covariates also

tested as potential predictors affecting resin flow. The mixed-effects models had many

terms and interactions, and had the potential to be complex to untangle and interpret. In

retrospect, Experiments 3 and 4 had much in common with another recently published

study, Lombardero et al. (2006), which studied the effects of bole heating and mechanical

wounding on resin flow in red pine.

Overall, the analyses in Experiment 3 were more straightforward than expected.

Many model terms were non-significant, resulting in reasonably clear model

interpretations and findings. Based on 30 sampling sessions over 3 years after treatment,

there was no evidence that any of these mechanical injury mechanisms had significant

effects on resin flow, as evaluated against untreated controls and a prescribed burn. Of

the four treatments contrasted with controls in the various analyses, only the burning

treatment showed significant deviation from default resin flow patterns. None of the

physiological covariates related to tree or crown size, growth rate, or carbon-13

discrimination was a significant predictor of resin flow in any of the analyses, before or

after treatment. Seasonal patterns of resin flow in this study (Experiment 3) were similar

to previous research in pine species (Harper and Wyman 1936, Lombardero et al. 2000,

107 Wallin et al. 2004), with the highest resin volumes obtained in mid-summer (August),

and lower yields earlier or later in the season.

3.2.1. Burning and bole charring treatments (Experiments 3. 4)

With respect to the burned trees at Sun Pass in Experiments 3 and 4, there was

evidence of increased post-fire resin response one or more years after burning. This

response has been previously documented in ponderosa pines subjected to prescribed

burning (Kolb et al. 1998, Perrakis 2004). In this study, there was also evidence of a

short-term reduction in resin flow measured a few weeks after burning - a response that

was hypothesized, but not confirmed, in the previous work at Crater Lake (Perrakis

2004).

In the present study, the period of reduced defenses was detected as significant at

3 and 16 days after burning (Experiment 3) and 27 days after bole charring (Experiment

4). This pattern is a slightly slower response than that of a recent study in red pine:

Lombardero et al. (2006) observed that bole wounding or heating (with a blowtorch) led

to a resin flow reduction that was significant after 1-3 days, had recovered after 7-9 days,

and increased strongly after 55 days, compared to controls.

The mechanism behind these reductions is unclear. Lombardero et al. (2006)

suggested that short-term resin flow reduction may be a sort of 'shock response' to the

treatment, similar to previous reports of lightning-struck trees readily succumbing to

beetle attacks (Blanche et al. 1985). Physiologically, bole charring may have interrupted

sap flow in bole tissues, ultimately leading to reduced resin pressure and exudation.

108 Water potential has been found to be positively related to resin pressure (Vite 1961,

Hodges and Lorio 1971), and fire damage to sapwood tissue might have reduced water

potential in the lower bole, reducing resin exudation pressure and volume in the days

following burning. This mechanism assumes a positive correlation between resin flow

and resin pressure at any one time, a correlation that was previously believed to be

significant based on capillary flow modeling (Bourdeau and Schopmeyer 1958, Vite

1961) and found significant (although weak) in separate OEP and OEF measurements at

Crater Lake (Perrakis and Agee 2006), but has been previously disputed (Lorio 1994).

Another mechanism for explaining the short-term resin flow reduction may

simply be resinosis at injury sites. Bole charring can sever and expose sapwood resin

canals (Ryan 2000), and resin loss from these wounds sites may have reduced resin

volume and pressure within the canal network surrounding char wounds, and less

constitutive resin remaining for sampling. Resinosis on the lower bole areas of burned

trees was widespread in this experiment (Experiment 3) as well as in the burning

experiments at Crater Lake (Experiments 1,2).

It was previously hypothesized that a reduction in resin defenses (flow or

pressure) following fire might provide a 'window of opportunity' for beetles to attack

burned trees (Geiszler et al. 1984, McCullough et al. 1998, Perrakis 2004). Although

Experiments 3 and 4 suggest that this reduction does occur, the timing of the resin

decline did not appear to match with observed patterns of beetle attacks in this study.

Mortality attributed to post-fire beetle attacks peaked 1-2 years after burning, as

discussed above (Section 3.1.2) and continued for several years thereafter. This is

109 considerably longer than the period of reduced defenses: Experiment 3 (Analysis 2)

suggested depressed resin defenses measured at 3 and 16 days after burning, while

Experiment 4 suggested reduced resin defenses approximately one month after burning.

Analysis 5 of the Experiment 3 data suggested that resin flow in burned trees had

recovered to the level of controls by 26 days after burning, while even greater differences

at 54 days suggested that burned trees were already beginning to demonstrate the post-

injury increase identified in subsequent years. Therefore, the period of reduced resin flow

in these experiments appears to have lasted for only a few weeks (< 4). At Crater Lake,

assuming that the older ponderosa pines responded similarly to the pole-sized individuals

tested in Experiments 3 and 4, these few weeks of reduced resin flow did not match the

timing of the increased post-fire beetle attacks observed there (Experiment 1), and appear

minor to the post-fire tree mortality processes.

Experiment 4 confirmed that bole wounding alone, with no accompanying forest

floor disruption, root mortality, or crown scorch, caused a similar dynamic response in

resin defenses. As in Experiment 3, charred trees in Experiment 4 had short-term

reductions in resin flow approximately one month following the treatment, and strong

increases the following year. This experiment also highlights the importance of

incorporating pre-treatment data into resin flow measurements, as the second year

increases in resin flow were only significantly different from controls in terms of change

from pre-burn values - not in terms of absolute resin flow volume.

Measurement height on the bole, which varied between 1 m and 4.5 m in

Experiment 4, did not have a consistent effect on resin flow, nor did it have a significant

110 interaction effect with experimental treatments. The overall effect of charring the bole

appeared similar across height levels; the hypothesis of increased resin flow being a

localized response to wounding was not supported by the data from this experiment.

Although previous studies in other pine species have described smaller-scale (< 1 m)

localized effects, including resin-soaked lesions in phloem and outer xylem tissues

surrounding physical wounds (Cheniclet 1987, Lombardero et al. 2006), these data

suggest a more generalized response to lower bole charring in ponderosa pine.

The Height: Time interaction remains a significant and unexplained factor. The

difference between years is visually striking, with resin flow positively related to height

in September 2006, and resin flow inversely related to height in September 2007 (Figure

16). Since the pattern was consistent across both treatments (char and control), it was not

investigated further in this study. The most obvious difference between the two years is

climate, with the variations in precipitation levels between years creating a sort of natural

experiment. According to the southern Oregon Palmer Drought Severity Index values

(see Appendix 4), 2006 was a year of greater than average precipitation during the

growing season, while 2007 was a year with a strong drought signal during the same

months. Further investigation is needed to confirm any possible interaction effect on

resin flow between PDSI and measurement height. It is unclear why resin flow would

increase with height during a wet year and decrease with height during a drought year.

I l l

3.2.2. Pruning and root trenching treatments (Experiment 3)

The results of Experiments 3 and 4 suggest that post-fire resin flow increases are

a response to bole injury, rather than to root or lower crown damage. This assumes that

mechanical treatments effectively simulated fire effects to roots and crowns, an

assumption that merits some discussion.

Pruning treatments were simple to implement and assess, and likely replicated

lower crown scorch effectively. As in the case of a vigorous surface fire, the pruning

treatment removed significant portions of the live leaf area in the lower crown, while

leaving the upper crown unaffected. Growth-differentiation balance principles (see

below) suggest that greater leaf area will provide a greater source of carbohydrates

(assuming unchanging sapwood area and root mass), which is then available for growth

or metabolic processes depending on other source-sink relationships. However, the lower

crown, which would be the most likely to be scorched by fire and was the target of

pruning in the present study, is considered photosynthetically inefficient due to internal

shading, compared to the upper and middle crown areas (Helms 1970, Kozlowski and

Pallardy 1997). Pruning of the lower crown may have even increased the trees' xylem

water potential (by decreasing transpiration sinks from foliage) without significantly

reducing carbohydrate production. However, Ryan (2000) conducted a similar

experiment (scorching lower crown foliage) on young (14-25 cm DBH) ponderosa pines,

and noted no significant differences on transpiration or stomatal conductance following

scorching treatments applied to 0%, 40% or 80% of crowns. He further suggested that his

sample trees were likely resilient to foliage reductions due to carbohydrate stores in roots,

112 and had apparently succeeded in regenerating foliage in the next few years. A similar

resiliency may have existed in the pruned trees in the present study (root stores

compensating for foliage reduction), with little measurable effect on either overall

carbohydrate allocation or mid-summer water balance. Confirming this premise would

require further testing. In the case of fire, young ponderosa pines are generally highly

resistant to crown scorch (Dieterich 1979, Harrington 1993), although post-fire bark

beetles attacks undermine that resistance in older stands, as previously discussed. In this

study, data from Experiment 3 do not suggest any consistent effects of pruning on resin

flow. There is no evidence that crown scorch effects include measurable changes to resin

defenses that would affect bark beetle attack success.

Effects of root trenching were more difficult to assess. Prescribed fire effects in

fire-excluded ponderosa pine stands can be severe, as thick duff and poor soil moisture

retention promote dense root growth in the upper soil and forest floor (Swezy and Agee

1991). Although the soil parent material (pumice from the Mount Mazama eruptions)

was identical at the Sun Pass site, duff layers appeared thinner at Sun Pass than at Crater

Lake's 'panhandle' sites (based on unpublished Sun Pass soil pit data and Swezy 1988).

This may be due to forest floor disturbance during recent management activities, such as

entries into the stand for pre-commercial thinning (Ed Deblander, Sun Pass State Forest,

personal communication 2003). Alternatively, rates of organic material accumulation

may have been lower in the second growth stands at Sun Pass, following logging of older

trees in the mid-20th century, than in the old growth stands at Crater Lake. Thus, root

113 growth at Sun Pass may have been less concentrated at the surface compared to the

Crater Lake sites.

Aside from digging a soil pit in the stand (which confirmed that the upper mineral

soil layers had abundant roots), pre-treatment root structure was not assessed. It was

recognized that the effects of 'light' or 'heavy' trenching around the circumference of

trenched trees would be highly variable. Such variability likely occurs in the case of

actual fire effects on roots, but in this case (perhaps compounded by the small sample

sizes) made the effects of treatments uncertain. The two levels of trenching (roughly 1/3

and 2/3 of treated trees' circumference) were chosen to attempt to induce two levels of

moisture stress, potentially prompting imbalances in root -to-shoot ratios and perhaps

reducing photosynthesis and carbohydrate availability. Although there was great

uncertainty over the degree of root reduction required to simulate various levels of

moisture stress, it seemed unwise to completely encircle trees with root trenches, lest

they become unstable and topple over. The analysis of treatment effects on BAI

(Experiment 3, following Analysis 5) suggested that the trenching treatments had not

significantly or consistently affected radial growth, as compared to untreated controls.

Although the heavily trenched trees, in particular, had a greater mean growth reduction

following treatments compared to controls, it appears that at least one outlier in each

treatment group existed: one control tree grew much less in 2004-2006 than in 1998-

2002, and one heavily trenched tree was little affected by trenching, with nearly equal

BAI levels between 1998-2002 and 2004-2006. Other trees in the heavily trenched

sample behaved as expected, with much reduced BAI among trenched trees after

114 treatment. This may be a simple result of small sample sizes, but it also suggests that the

young ponderosa pines at Sun Pass had deeper root systems than the much older (and

denser) stands within Crater Lake National Park. Aside from effects on radial growth, the

other analyses in Experiment 3 suggested that the trenching treatments were not a

significant predictor of resin flow in any of the models.

The lack of significance of the carbon-13 discrimination factor suggests that

moisture balance during the growing season was not a principal factor affecting resin

flow. Previous studies have found strong relationships between 813C in xylem rings and

moisture availability, due to the previously discussed stomatal responses and differential

assimilation between 12C and 13C in photosynthetic tissue (see Experiment 3, Objectives).

Even if trenching treatments did not cause consistent effects in terms of induced moisture

stress, 813C should still have emerged as a significant covariate in the resin flow models

if it were a major resin flow predictor. Although Experiment 3 suffered from small

sample sizes and uncertain treatment effects, it does suggest that post-fire resin flow

increases are not well simulated by mechanical damage to roots.

3.3. Carbon allocation, GDB, and response to wounding

How then do the present findings from Experiments 3 and 4 relate to the carbon

allocation discussion? Although the GDB hypothesis (as reviewed in Section 1) was not

tested extensively in this study, it formed a unifying concept for explaining and

discussing tree responses to treatments at a physiological level. Although the GDB

115 concept has been quite successful at predicting conifer defense responses in recent

studies, it may not be very relevant to the responses observed in these experiments.

Several emerging studies on resin flow in conifers have also attempted to use GDB

concepts to explain resin response patterns to experimental treatments or manipulations.

Coupling these recent findings with the results of my experiments reveal a mix of

findings both confirming and questioning GDB predictions, amidst a complex web of

carbon allocation feedbacks depending on many environmental and genetic factors.

As reviewed in Section 1, recent studies on conifer defenses have reported many

examples of tradeoffs between growth processes and investments in defense compounds,

including monoterpenes, phenolics, alkaloids, tannins, and so forth. However, studies

focusing specifically on resin flow measurements have proved more difficult to interpret

from a carbon allocation standpoint. One large reason for this may be due to the

architecture of resin defenses and difficulties associated with resin measurements. In

Pinaceae species, resin canals may continue to function for several years (see Section

1.2.2). Thus, measurements made in one year do not necessarily reflect the most recent

environmental conditions or carbon investment patterns, and there is an inherent problem

with trying to relate resin flow at medium time intervals (e.g. monthly) with rapidly

varying growth processes (an issue shared by Experiment 3 of the present study)

(Gaylord et al. 2007). Ponderosa pine resin is mostly constitutive (c.f. Phillips and

Croteau 1999, Franceschi et al. 2005), and conditions measured at any one time can

reflect several years of investment in resin defenses. Comparisons between species can

be invalid due to the differences between species that primarily depend on induced resin

116 production (such as grand fir or loblolly pine, Lewinsohn et al. 1991, Ruel et al. 1998)

and species with primarily constitutive resin production, such as ponderosa pine.

Resource partitioning, such as between growth and defense, may be confused with

localized response to injury.

Little is known in ponderosa pine about the sensitivity of resin canal development

and oleoresin production to variable growing conditions or injury. In loblolly pine, there

is some evidence that canals may not be filled to capacity during normal growing

conditions; following wounding, canals are rapidly filled with induced resin from

surrounding epithelial cells, with production most vigorous in the fastest-growing trees

with the largest crowns (Blanche et al. 1992, Lombardero et al. 2000). A recent study on

fire-surrogate (bole heating and physical injury) effects on red pine suggested that a

similar mechanism may occur in that species as well, albeit on a slower time scale

(Lombardero et al. 2006). As discussed above (Section 3.2.1), ponderosa pine appears to

respond less rapidly to wounding than loblolly pine does in the short term, but has a

similar pattern of response as red pine at a slightly longer time scale: decreased resin

flow for several days, recovery after approximately a month, and an increase potentially

detectable 7-8 weeks after wounding. In this study (Experiments 2, 3, 4), increased post-

fire resin flow persisted for one to several years. Whether this is due to increased

production of resin flow within existing resin canals (as is believed to occur in loblolly

pine), growth of additional vertical canals, as suggested by earlier work on eastern white

pine and aleppo pine (Barman 1936, Fahn and Zamski 1970), or a different mechanism

altogether, is not known.

117 These uncertainties makes the physiological interpretation of ecological stand

experiments more difficult. Thus, recent studies have both purportedly confirmed (Baier

et al. 2002, Zausen et al. 2005) and rejected (Latta and Linhart 1997, Rosner and

Hannrup 2004, Gaylord et al. 2007, McDowell et al. 2007) the GDB hypothesis as a

useful conceptual model for interpreting controls on resin flow. The results of this study

tend to support the latter view: the processes described by the GDB hypothesis may be

important in determining seasonal investment by trees into resin defenses, but these do

not appear to be critical in determining either bole resin flow or beetle responses

following fire in ponderosa pine stands. Instead, the evidence suggests that a one-time

wounding response best explains the resin increases observed at Crater Lake. The

observations that more vigorous trees (higher BAI and fuller crowns) have greater resin

flow both before and after wounding, and that no tradeoff is evident between radial

growth and resin production up to 3 years after wounding, do not match the fairly

simplistic predictions of the GDB model.

3.4. Resin chemistry and host-beetle interactions

In Experiment 5, fire effects on ponderosa pine resin monoterpenes were studied

at the Entiat and Annie Creek locations, sites at the northern and southern end,

respectively, of the eastern Cascades. Based on Smith's (1977, 2000) studies of resin

chemistry distribution, these sites are located in the Cascade Northern ecoregion (Entiat)

and Sierra Pacific ecoregion (Annie Creek). Monoterpene compositions in the two

118 regions are mostly similar, except that the average limonene content (percentage of total

monoterpenes) of individuals in the Sierra Pacific region is about 10%, while for the

Cascade Northern it is only 1%.

Much of the career of R. H. Smith was spent on the genetics and breeding of pest-

resistant pine trees. His findings and observations suggested that limonene content was of

major importance in determining the insect- and pathogen-resistance of individual

ponderosa pines, and much effort was invested into studying, and ultimately breeding,

trees with 'high-limonene' alleles (Smith 1972b, 2000). His work emphasized the

insecticidal effects of resin, and particularly monoterpenes, on colonizing beetles,

suggesting that the most toxic monoterpenes (to western pine beetles) were limonene and

8-3-carene, and the least toxic myrcene and a-pinene. However, this focus ignores the

critically important role (see below) of monoterpenes in ponderosa pine chemical

ecology (Wood 1982, Byers 1995, Seybold et al. 2006). It is also unclear to what extent

individual monoterpenes properties (toxicity or otherwise) are additive. Monoterpene

combinations can elicit different responses from beetles than exposure of lone

monoterpenes in laboratory settings (Byers 1995), and additional classes of compounds

in natural oleoresin may also act as toxins, or as semiochemicals, for other forest insects.

The role of monoterpenes as semiochemicals in bark beetle-host interactions has

been known and recognized for many years (see reviews by Wood 1982, Seybold et al.

2006) and may be more important at Crater Lake than resin toxicity. Myrcene, despite

being a relatively minor oleoresin constituent (Smith 1977, Hobson et al. 1993, Smith

2000), appears to be the most important monoterpene in western pine beetle attraction,

119

with the most attractive western pine beetle pheromone consisting of a mixture of beetle-

synthesized exo-brevicomin and frontalin, along with myrcene from oleoresin (Bedard et

al. 1969, Vite and Pitman 1969, Wood 1972, Gaylord et al. 2006). Alpha-pinene,

considered the least toxic of the major monoterpenes (Smith 2000), was found to be a

precursor to the western pine beetle anti-aggregant pheromones trans-verbenol and

verbenone (Byers 1983), pheromones released by male beetles to reduce intra-specific

competition on mass-attacked trees.

Monoterpenes are also significant semiochemicals for other beetles parasitizing

ponderosa pine. Red turpentine beetles are highly attracted to several resin monoterpenes

alone (a true primary attraction effect), including certain enantiomers of a-pinene, P-

pinene, and 8-3-carene (Hobson et al. 1993, Erbilgin and Raffa 2000). Other scolytids,

including mountain pine beetle and various Ips species, have also demonstrated

responsiveness (such as antennal responses and preferential landing) to various

ponderosa pine monoterpenes, often in combination with pheromone emissions.

However, the particular monoterpenes that are detected vary widely by beetle species,

and so no individual monoterpene or combination stands out as especially important for

either resistance or susceptibility (Trapp and Croteau 2001). Volatiles other than

monoterpenes have also shown promise as scolytids co-attractants in ponderosa pine,

including acetone (Billings et al. 1976) and ethanol (Kelsey and Joseph 2003).

Mechanistic explanations for the effects of these compounds are lacking, however.

In this study, the average proportion of monoterpenes as a fraction of resin mass

was about 24.7%, consistent with Smith's (2000) estimate of 25% monoterpene

120 composition in ponderosa pine oleoresin across the continent. Differences between

monoterpene proportions were significant between treatment groups at the Entiat site, but

not at Annie Creek. This discrepancy is possibly indicative of the differences in tree sizes

and injury effects: at Annie Creek, larger and older trees may have been less injured by

fires than the generally smaller individuals at Entiat. Fire intensity was also likely higher

at Entiat, and both fire injury and potential fire-induced changes to resin production

likely more pronounced than at Annie Creek.

The implications and mechanism behind the increased monoterpene proportion

measured at Entiat are not apparent. The wound-induced resin effect has been studied

intensely in recent years on grand fir (Steele et al. 1998, Trapp and Croteau 2001),

showing differential timing of synthesis between various monoterpenes and resin acids,

but is probably not relevant to ponderosa pine, as previously discussed. Smith (2000)

argues strongly in favor of genetic rather than environmental control of monoterpene

composition, noting very little variation between measurements made up to 9 years apart

in individual trees. However, this does not preclude the possibility of resin composition

being affected by an induction process that changes resin composition following fire or

physical damage. A few studies have shown increases in the overall monoterpene

fraction of pines following drought stress (Hodges and Lorio 1975, Llusia and Penuelas

1998, Turtola et al. 2003), but none following injury. Increases in total resin

monoterpenes (in addition to total resin volume) may represent a general response to

injury in ponderosa pine, but the results from this study are not conclusive. If increased

121 monoterpenes do indeed represent an effect of fire injury, this represents further support

for the notion of induced or enhanced beetle attraction in this species following fire.

In terms of specific monoterpenes, neither of the two sites in this study offered

any indication that monoterpene composition (proportion of myrcene, 3-carene or

limonene) was affected by fire injury. Previous studies have previously noted various

effects of stress on resin monoterpene composition, although none has directly studied

ponderosa pine resin composition following wounding or fire injury. Smith (2000) found

that ponderosa pine saplings subjected to grafting, constrained root growth in containers,

or heavy herbicide application responded with decreased levels of myrcene and

limonene, and increases in a- and |3-pinene, as measured by proportions extracted from

bole resin. This combination of effects has not been identified to cause a net increase or

decrease in either tree resistance or beetle attraction. A few other studies have reported

mostly inconclusive relationships between environmental variables affecting growth and

resin flow (Latta and Linhart 1997, Thoss and Byers 2006). Although several studies

have shown effects of wounding on foliage chemistry, (including monoterpenes), such as

due to herbivory or defoliation (Johnson et al. 1997, Litvak and Monson 1998, Thoss and

Byers 2006), these are not considered directly relevant to bark beetle resistance due to

known differences in composition between resin from different structures in individual

trees - particularly between foliage and bole tissues (Latta et al. 2000, Smith 2000).

Future research on the differences between constitutive and induced resin

(Phillips and Croteau 1999) will be of greatest relevance to this question. As previously

mentioned, there is some debate as to whether traumatic (induced) resin canals actually

122 exist in ponderosa pine. In several other pine species, the canals that form around injury

sites are indistinguishable from constitutive resin canals and only appear after delays of

weeks or months (Bannan 1936, Fahn and Zamski 1970). In ponderosa pine, there is no

short-term increase in monoterpene cyclase activity (Lewinsohn et al. 1991) or

measurable resin production (Wallin et al. 2003) following bole wounding. Exuded resin

sampled several months or longer following stress or injury will typically include both

constitutive and induced resin, complicating the study of the induction process in this

species.

3.5. Beetle evolution, resin volatiles and kairomones

The results of this study suggest that beetle attacks were not associated with

reduced resin defenses; on the contrary, fire appeared to increase resin quantity measured

at the bole, and possibly the monoterpene proportion among affected trees. We must

therefore reexamine the possibility of post-fire beetle attraction. Although primary

attraction, as previously discussed, has not been confirmed in western pine beetles, the

use of monoterpene co-attractants (particularly myrcene) has been discussed as greatly

increasing the potency of beetle pheromones in this species. In addition, even if

monoterpenes do not act directly as kairomones for western pine beetle, they may do so

for other beetle species. There is evidence for inter-specific pheromonal attraction in

several species of scolytids present at these sites. Recent studies in Arizona have

observed that several beetle species (including turpentine beetles) were attracted to

123 western pine beetle pheromone, and the authors suggested that the collective attacks by

beetles in various positions on trees' boles overwhelmed tree defenses (Gaylord et al.

2006, Wallin et al. 2008). However, the specific study of lure preferences by various

beetle species noted that western pine beetles were attracted almost exclusively to their

species-specific pheromone/kairomone combination (exo-brevicomin, frontalin, and

myrcene), and not to the semio-chemicals of other species (Gaylord et al. 2006).

In their timely review of the state of knowledge of beetle-monoterpene

interactions, Seybold et al. (2006) remarked that the complex responses of dispersing

bark beetles to the forest airspace around them has only begun to be explored. A recent

study on antennal responses to host and non-host volatiles by western pine beetles

identified over 40 different compounds that elicited antennal responses by beetles,

including 24 compounds from ponderosa pine. This does not confirm primary attraction

per se, but rather suggests that beetles navigate through the forest in part by chemotaxis

(positive or negative) to aid in foraging behavior - both finding hosts and avoiding non-

hosts (Shepherd et al. 2007). Not surprisingly, there is also evidence that volatile

emissions from forest trees can be stimulated by forest disturbances. Mechanical thinning

treatments in ponderosa pine forests were noted in one study to increase monoterpene

fluxes within the canopy by 40-fold compared to pre-treatment conditions (Schade and

Goldstein 2003). Another recent study conducted in northern Arizona and California

demonstrated significant short-term increases in beetle attacks immediately following

various thinning treatments; attacks by various beetle species were generally highest

immediately following late spring treatments, compared to late summer treatments and

124 untreated stands. Searching for an explanation, the authors conducted a separate

laboratory experiment on volatile emissions from thinned and chipped biomass. They

measured significant short-term pulses in resin monoterpene emissions from lopped and

(especially) chipped biomass samples, and suggested a causal link between ambient

monoterpene levels and beetle activity (Fettig et al. 2006).

While kairomonal relationships between conifer resin volatiles and beetles have

been recognized for decades, the effects of stand management activities on volatile

emissions have been very poorly recognized and understood. No studies have yet

documented the effects of fire on ponderosa pine monoterpene emissions, but it would be

very surprising if emissions were not augmented significantly by fire. Such volatiles

would inevitably include myrcene, the known monoterpene component of the western

pine beetle pheromone. Even if host selection in this species is not accomplished via

attraction to host kairomones alone, it could certainly be enhanced via volatization of

pheromone co-attractants from existing beetle attacks within treated stands. In the

absence of other mechanisms explaining how increased resin defenses can occur

synchronously with heightened beetle attacks, it seems very plausible that a large pulse of

fire-related resin volatiles released into the air could be responsible.

3.6. Methodological considerations

The last experiment involved a comparison of methods and assumptions behind

resin flow measurements in this study. Based on the two trials in Experiment 6, the

125 drilling method used in this study produced values closely resembling those of a 2.54 cm

diameter arch punch. The geometric model based on the estimated xylem surface area

exposed by each method also had a reasonably strong correlation across both sites and

arch punch sizes. Thus, there does not appear to be a qualitative difference between resin

flow sampled from the sapwood to resin flow sampled at the cambium.

With respect to the xylem surface area model, some unexplained variability or

error remains, particularly with the Entiat data. This variability may originate from

differences in vertical canal density across the xylem, vertical canals being considered

the main reservoirs of pine resin, their density positively correlated with resin flow

(Blanche et al. 1992 and references therein). Resin canal density seems to vary during

normal growing conditions based on environmental conditions (Blanche et al. 1992), but

also wounding, pressure, and other stresses, including fire injury. Thus, the larger arch

punch may be better suited to measuring an average resin yield over a somewhat larger

region of the bole that might be vary in resin production capability. Such variability

might exist due to normal variations in environmental conditions during growth (Blanche

et al. 1992) or, perhaps in this case, to heterogeneous stimulation due to fire injury.

As mentioned previously, the phloem removal/arch punch method has the

advantage of clearly showing if tissue is alive at the collection site, and perhaps also of

measuring resin at the same location as that encountered by bark beetles (the outer

xylem). On the other hand, the drilling method creates a much smaller outer wound, and

is simple to seal following sampling with wooden dowel. This experiment shows a clear

and positive relationship between the collection of resin at the cambium and from the

126 sapwood. For the purposes of this study, the benefit of being able to reseal the drilled

holes after sampling was very beneficial to avoid attracting chemically-attuned beetles or

other insects, and to minimize the damage to the phloem caused be frequent resampling.

The microstudy involved a very brief look at the xylem of three small ponderosa

pines. Vertical and radial canals were readily observed and abundant, but there was no

evidence of transverse canals. Although this was by no means an exhaustive or

systematic analysis, it would appear that any transverse resin movement is much rarer, if

even possible, compared with vertical or radial resin movement. This is in agreement

with Fahn's (1979) generalized observations that there are no transverse resin network in

pines, and that resin canals exist only in vertical and radial 2-dimensional planes.

Variable resin properties measured from different sides of individual trees (as reported in

Perrakis 2004, Lombardero 2006) are thus readily explained.

3.7. Summary, management recommendations, and conclusions

This study set out to examine some of the mechanisms and physiology behind

secondary mortality of ponderosa pines following prescribed fire. The experimental

context consisted of a real-life management challenge in the form of a fire restoration

project in old growth ponderosa pine - mixed-conifer stands at Crater Lake NP, Oregon.

The present study involved a multi-disciplinary investigation into possible mechanisms

behind elevated bark beetle response following prescribed fire treatments, and the

127 findings can be summarized in the form of a suggested explanation for the observations

at Crater Lake.

The initial post-fire ponderosa pine monitoring experiment showed that mortality

of the old trees was synchronous with resin response patterns over time (higher mortality

in treatments producing higher resin flows - fall burning and spring burning). This

undermines the conventional belief that resin defenses are indicative of survivorship in

fire-injured ponderosa pines. Various experiments confirmed that fire induced increased

resin flow in this species. In the Crater Lake old-growth stand, resin flow appeared to

peak in years 2 and 3 following fire, and to begin tapering off by year 4 after fire

treatment. The greatest effects were observed in more intense fall burn treatments

compared with patchy spring burn treatments.

In the parallel experiment in younger trees at Sun Pass, changes from pre-burn

resin flow were greatest one year after burning, and had returned to control levels by

post-burn year three. A brief period of reduced resin flow immediately after burning

appeared unlikely as an explanation for increased bark beetle attacks, as it was quickly

reversed by resin flow increases within a few weeks. An additional experiment confirmed

that much or all of the observed post-fire resin flow increases could be explained by bole

charring, rather than any effects on roots, crown, or other tree structures. An investigation

into fire effects on resin chemistry was mostly inconclusive, although there was some

evidence at one of the two sites that fire increased overall proportions of monoterpenes,

the volatile portion of pine resin. Additionally, a methods investigation confirmed that

128 resin volume collected by drilling into the sapwood is well correlated with resin collected

at the outer xylem.

Integrating these findings with results from previous studies suggests that effects

of restoration treatments on resin flow should be interpreted with caution. Resin flow

measurements appear sensitive both to factors that affect whole-tree vigor (constitutive

resin) as well as to disturbances or treatments that only cause bole injury. Historically,

the effects of frequent low-intensity fire probably contributed to both reduced

competition (and higher vigor in surviving trees) and bole char injuries - both of which

apparently can promote greater resin flow.

The importance of the complex mixture of airborne chemicals, including resin

volatiles, on insect movements in forest ecosystems is only beginning to be understood.

Although primary attraction in western pine beetles is not confirmed, monoterpene co-

attraction (with beetle pheromones) has long been known to occur. Primary attraction has

also been observed many times in red turpentine beetles, which may have contributed to

ponderosa pine mortality at Crater Lake, as well as in other bark beetles present in the

area. Large releases of monoterpenes into the air have been observed following

silvicultural manipulations and tree damage in other conifer species. Continued

ponderosa mortality (as observed at Crater Lake) as well as heightened post-fire resin

flow suggest that the potential benefits of increased resin defenses may be confounded by

vast release of attractive resin compounds into the air. Thus, while both trees and beetles

appear to benefit from the effects of fire, in this case, the beetles seem to win by being

able to locate potential hosts more easily. The generally low vigor of the ponderosa pine

129 population means that their defenses remain insufficient to protect them against the

increased attacks, despite enhanced resin flow.

The growth-differentiation balance was neither confirmed nor refuted by the

results of this study. The experiment on root and crown reduction may have been

undermined by small sample sizes and uncertain treatment effects, but provided no

evidence for any effects on resin flow. However, if we reject resin production and flow as

the key determinants of tree resistance or susceptibility, then resource allocation tradeoffs

become secondary in determining post-fire mortality. When combined with previous

findings, the results from this study agree with a newer conceptual model of fire-beetle

interaction that is based on both attractive and defensive roles of oleoresin. Ponderosa

pine survival may be enhanced by minimizing local releases of resin volatiles, avoiding

fire injury (or other types of tree damage), and promoting high constitutive resin

production (for defenses in the case of attack).

Management implications can mostly be deduced from other reviews (Seybold et

al. 2006, Kolb et al. 2007). Pine survivorship appears most closely related to high radial

growth/BAI and high crown vigor. Thus, treatments that improve individual tree vigor,

such as removal of competing vegetation surrounding old pines, are likely beneficial to

survival (Feeney et al. 1998, Latham and Tappeiner 2002, McDowell et al. 2003). This

study has shown that vigor is positively related to resin flow, both with and without fire

injury. Additionally, resin flow monitoring in ponderosa pine is not a reliable indicator of

the likelihood of survival after fire and is not suggested as a post-treatment monitoring

technique. Although this notion awaits testing, there may also be merit to minimizing the

130 release of resin volatiles in stand treatments. Mitigative measures could include avoiding

injuring trees of interest, ensuring that thinning treatments include off-site disposal of

thinned biomass, and conducting thinning treatments only during periods of low beetle

activity (fall or winter).

This study suggested that bole charring or scorching was an important driver of

post-fire resin increases. Assuming that bole char is also related to the hypothesized

attraction of beetles, this suggests that raking fuels from the bases of individual trees, or

otherwise protecting their lower boles, may be worth further investigation to mitigate

injury (Kolb et al. 2007). The current record of success of raking treatments is mixed,

with some studies suggesting it can effectively mitigate fire injury and mortality in

ponderosa pine stands (Sackett and Haase 1998, Kolb et al. 2001, Moore et al. 2003), but

others suggesting that its effects may be either neutral or detrimental to survival of

treated trees (Fule et al. 2002). Previous raking experiments at Crater Lake suggested that

raking around old pines may have negatively impacted surface roots, resulting in equal,

or higher mortality of raked trees compared to unraked trees (Swezy and Agee 1991,

Perrakis 2004). Attempts to minimize fire injury via pre-burn raking should be

approached with caution in these stands, and preferably done one or more years before

burn treatments are applied to permit recovery.

Overall, the findings of this study support the notion of an incremental approach

to fire restoration in ponderosa pine forests at Crater Lake. Restoration treatments,

including burning, and possibly thinning or other mechanical fuel manipulations, should

be approached gradually. Although such treatments can eventually benefit these stands, it

131 may be necessary to improve the vigor of individual trees prior to implementing all but

the least intense prescribed fires, lest bark beetle populations benefit more than the trees

of concern. Further studies and monitoring are needed, but desired stand conditions may

be more achievable using repeated lower intensity burns rather than fewer higher

intensity burns.

132

REFERENCES

Agee, J. K. 1993. Fire ecology of Pacific Northwest forests. Island Press, Washington, D.C.

Agee, J. K. 2003. Monitoring post-fire tree mortality in mixed-conifer forests of Crater Lake, Oregon. Natural Areas Journal 23:114-120.

Anonymous. 2002. Crater Lake National Park fire management plan - draft environmental assessment. Crater Lake National Park, Crater Lake, OR.

Baier, P., E. Fuhrer, T. Kirisits, and S. Rosner. 2002. Defence reactions of Norway spruce against bark beetles and the associated fungus Ceratocystis polonica in secondary pure and mixed species stands. Forest Ecology and Management 159:73-86.

Bannan, M. W. 1936. Vertical resin ducts in the secondary wood of the Abietineae. New Phytologist 35:11-46.

Bedard, W. D., P. E. Tilden, D. L. Wood, R. M. Silverstein, R. G. Brownlee, and J. O. Rodin. 1969. Western pine beetle: field response to its sex pheromone and a synergistic host terpene, myrcene. Science 164:1284-1285.

Berryman, A. A. 1972. Resistance of conifers to invasion by bark beetle-fungus associations. Bioscience 22:598-602.

Billings, R. F., R. I. Gara, and B. F. Hrutfiord. 1976. Influence of ponderosa pine resin volatiles on the response of Dendroctonus ponderosae to synthetic trans-verbenol. Environmental Entomology 5:171-179.

Blanche, C. A., J. D. Hodges, and T. E. Nebeker. 1985. Changes in bark beetle susceptibility indicators in a lightning-struck loblolly pine. Canadian Journal of Forest Research 15:397-399.

Blanche, C. A., P. L. J. Lorio, R. A. Sommers, J. D. Hodges, and T. E. Nebeker. 1992. Seasonal cambial growth and development of loblolly pine: xylem formation, inner bark chemistry, resin ducts, and resin flow. Forest Ecology and Management 49:151-165.

Bourdeau, P. F., and C. S. Schopmeyer. 1958. Oleoresin exudation pressure in slash pine: its measurement, heritability, and relation to oleoresin yield. Pages 313-319 in K. V. Thimann, editor. The physiology of forest trees. The Ronald Press Co., New York.

133 Biisgen, M., and E. Munch. 1929. The structure and life of forest trees. Translated by T.

Thomson. John Wiley & Sons, New York, NY.

Byers, J. 1983. Bark beetle conversion of a plant compound to a sex-specific inhibitor of pheromone attraction. Science 220:624-626.

Byers, J. 1995. Host-tree chemistry affecting colonization in bark beetles.//? R. T. Carde and W. J. Bell, editors. Chemical ecology of insects 2. Chapman and Hall, New York.

Byers, J. 1996. An encounter rate model of bark beetle populations searching at random for susceptible host trees. Ecological Modelling 91:57-66.

Cheniclet, C. 1987. Effects of wounding and fungus inoculation on terpene producing systems of maritime pine. Journal of Experimental Botany 38:1557-1572.

Christiansen, E., R. H. Waring, and A. A. Berryman. 1987. Resistance of conifers to bark beetle attack: searching for general relationships. Forest Ecology and Management 22:89-106.

Dalgaard, P. 2002. Introductory statistics with R. Springer, New York, NY.

Dieterich, J. H. 1979. Recovery potential of fire-damaged southwestern ponderosa pine. Research Note RM-379, USDA Forest Service, Rocky Mountain Forest and Range Experiment Station, Fort Collins, CO.

Erbilgin, N., and K. F. Raffa. 2000. Opposing effects of host monoterpenes on responses by two sympatric species of bark beetles to their aggregation pheromones. Journal of Chemical Ecology 26:2527-2548.

Fahn, A. 1979. Secretory tissues in plants. Academic Press, New York.

Fahn, A., and E. Zamski. 1970. The influence of pressure, wind, wounding and growth substances on the rate of resin duct formation in Pinus halepensis wood. Israel Journal of Botany 19:429-446.

Feeney, S. R., T. E. Kolb, W. W. Covington, and M. R. Wagner. 1998. Influence of thinning and burning restoration treatments on presettlement ponderosa pines at the Gus Pearson Natural Area. Canadian Journal of Forest Research 28:1295-1306.

Fettig, C. J., J. D. McMillin, J. A. Anhold, S. M. Hamud, R. R. Borys, C. P. Dabney, and S. J. Seybold. 2006. The effects of mechanical fuel reduction treatments on the activity of bark beetles (Coleoptera: Scolytidae) infesting ponderosa pine. Forest Ecology and Management 230:55-68.

134 Franceschi, V. R., P. Krokene, E. Christiansen, and T. Krekling. 2005. Anatomical and

chemical defenses of conifer bark against bark beetles and other pests. New Phytologist 167:353-376.

Franklin, J. F., and C. T. Dyrness. 1988. Natural vegetation of Oregon and Washington, 2nd edition. Oregon State University Press, Corvallis.

Fredericksen, T. S., R. L. Hedden, and S. A. Williams. 1995. Susceptibility of loblolly pine to bark beetle attack following simulated wind stress. Forest Ecology and Management 76:95-107.

Fule, P. Z., W. W. Covington, H. B. Smith, J. D. Springer, T. A. Heinlein, K. D. Huisinga, and M. M. Moore. 2002. Comparing ecological restoration alternatives: Grand Canyon, Arizona. Forest Ecology and Management 170:19-41.

Furniss, R. L., and V. M. Carolin. 1977. Western Forest Insects. Miscellaneous Bulletin 1339, USDA Forest Service, Washington, D.C.

Gara, R. I., and J. P. Vite. 1962. Studies on the flight patterns of bark beetles (Coleoptera: Scolytidae) in second growth ponderosa pine forests. Contributions from the Boyce Thompson Institute 21:275-289.

Gaylord, M. L., T. E. Kolb, K. F. Wallin, and M. R. Wagner. 2006. Seasonality and lure preference of bark beetles (Curculionidae: Scolytinae) and associates in a northern Arizona ponderosa pine forest. Environmental Entomology 35:37-47.

Gaylord, M. L., T. E. Kolb, K. F. Wallin, and M. R. Wagner. 2007. Seasonal dynamics of tree growth, physiology, and resin defenses in a northern Arizona ponderosa pine forest. Canadian Journal of Forest Research 37:1173-1183.

Gholz, H. L. 1982. Environmental limits on aboveground net primary production, leaf area, and biomass in vegetation zones of the Pacific Northwest. Ecology 63:469-481.

Goheen, D. J., F. W. J. Cobb, D. L. Wood, and D. L. Rowney. 1985. Visitation frequencies of some insect species on Ceratocystis wageneri infected and apparently healthy ponderosa pines. Canadian Entomologist 117:1535-1543.

Harper, V. L., and L. Wyman. 1936. Variations in naval-stores yields associated with weather and specific days between chippings. USDA Technical Bulletin 510.

Harrington, M. G. 1993. Predicting Pinus ponderosa mortality from dormant season and growing season fire injury. International Journal of Wildland Fire 3:65-72.

Helms, J. A. 1970. Summer photosynthesis of ponderosa pine in its natural habitat. Photosynthetica 4:234-253.

135 Herms, D. A., and W. J. Mattson. 1992. The dilemma of plants: to grow or defend.

Quarterly Review of Biology 67:283-335.

Hessburg, P. F., R. G. Mitchell, and G. M. Filip. 1994. Historical and current roles of insects and pathogens in eastern Oregon and Washington forested landscapes. General Technical Report PNW-GTR-327, USDA Forest Service Pacific Northwest Research Station.

Himejima, M., K. R. Hobson, T. Otsuka, D. L. Wood, and I. Kubo. 1992. Antimicrobial terpenes from oleoresin of ponderosa pine tree Pinus ponderosa: a defense mechanism against microbial invasion. Journal of Chemical Ecology 18:1809-1818.

Hobson, K. R., D. L. Wood, L. G. Cool, P. R. White, T. Ohtsuka, I. Kubo, and E. Zavarin. 1993. Chiral specificity in responses by the bark beetle Dendroctonus valens to host kairomones. Journal of Chemical Ecology 19:1837-1846.

Hodges, J. D., and P. L. J. Lorio. 1971. Comparison of field techniques for measuring moisture stress in large loblolly pines. Forest Science 17:220-223.

Hofstetter, R. W., J. B. Mahfouz, K. D. Klepzig, and M. P. Ayres. 2005. Effects of tree phytochemistry on the interactions among endophloedic fungi associated with the southern pine beetle. Journal of Chemical Ecology 31:539-560.

Hsiao, T. C. 1973. Plant responses to water stress. Annual Review of Plant Physiology 24:519-570.

Johnson, R. H., B. L. Young, and D. N. Alstad. 1997. Responses of ponderosa pine growth and volatile terpene concentrations to manipulation of soil water and sunlight availability. Canadian Journal of Forest Research 27:1794-1804.

Karsky, D., B. L. Strom, and H. Thistle. 2004. An improved method for collecting and monitoring pine oleoresin. 0434-2306-MTDC, USDA Forest Service Missoula Technology and Development Center.

Keen, F. P. 1943. Ponderosa pine tree classes redefined. Journal of Forestry 41:249-253.

Kelsey, R. G., and G. Joseph. 2003. Ethanol in ponderosa pine as an indicator of physiological injury from fire and its relationship to secondary beetles. Canadian Journal of Forest Research 33:870-884.

Kolb, T. E., J. K. Agee, P. Z. Fule, N. G. McDowell, K. Pearson, A. Sala, and R. H. Waring. 2007. Perpetuating old ponderosa pine. Forest Ecology and Management 249:141-157.

136 Kolb, T. E., P. Z. Fule, M. R. Wagner, and W. W. Covington. 2001. Six-year changes in

mortality and crown condition of old-growth ponderosa pines in different ecological restoration treatments at the G.A. Pearson Natural Area. Pages 61-66 in R.K. Vance, C.B. Edminster, W.W. Covington, J.A. Blake (compilers), Ponderosa Pine Ecosystems Restoration and Conservation: Steps Towards Stewarship (conference proceedings). RMRS-P-22, USDA Forest Service Rocky Mountain Research Station, Ogden, UT.

Kolb, T. E., K. M. Holmberg, M. R. Wagner, and J. E. Stone. 1998. Regulation of ponderosa pine foliar physiology and insect resistance mechanisms by basal area treatments. Tree Physiology 18:375-381.

Kozlowski, T. T. 1992. Carbohydrate sources and sinks in woody plants. Botanical Review 58.

Kozlowski, T. T., and S. G. Pallardy. 1997. Growth control in woody plants. Academic Press, New York.

Kramer, P. J. 1983. Water relations of plants. Academic Press, San Diego.

Lapasha, C. A., and E. A. Wheeler. 1990. Resin canals in Pinus taeda: longitudinal canal lengths and interconnections between longitudinal and radial canals. International Association of Wood Anatomists Bulletin 11:227-238.

Larsson, S., R. Oren, R. H. Waring, and J. W. Barrett. 1983. Attacks of mountain pine beetle as related to tree vigor of ponderosa pine. Forest Science 29:395-402.

Latham, P., and J. Tappeiner. 2002. Response of old-growth conifers to reduction in stand density in western Oregon forests. Tree Physiology 22:137-146.

Latta, R. G., and Y. B. Linhart. 1997. Path analysis of natural selection on plant chemistry: the xylem resin of ponderosa pine. Oecologia 109:251-258.

Latta, R. G., Y. B. Linhart, L. Lundquist, and M. A. Snyder. 2000. Patterns of monoterpene variation within individual trees in ponderosa pine. Journal of Chemical Ecology 26:1341-1357.

Lewinsohn, E., M. Gijzen, and R. Croteau. 1991. Defense mechanisms of conifers: differences in constitutive and wound-induced monoterpene biosynthesis among species. Plant Physiology 96:44-49.

Lin, J., Y. Hu, X. He, and R. Ceulemans. 2002. Systematic survey of resin canals in Pinaceae. Belgian Journal of Botany 135:3-14.

137 Lombardero, M. J., M. P. Ayres, and B. D. Ayres. 2006. Effects of fire and mechanical

wounding on Pinus resinosa resin defenses, beetle attacks, and pathogens. Forest Ecology and Management 225:349-358.

Lombardero, M. J., M. P. Ayres, P. L. J. Lorio, and J. J. Ruel. 2000. Environmental effects on constitutive and inducible resin defenses of Pinus taeda. Ecology Letters 3:329-339.

Loomis, W. E. 1953. Growth and differentiation in plants. Iowa State College Press, Ames, LA.

Lorio, P. L. J. 1986. Growth-differentiation balance: a basis for understanding southern pine beetle-tree interactions. Forest Ecology and Management 14:259-273.

Lorio, P. L. J. 1993. Environmental stress and whole-tree physiology. Pages 81-101 in T. D. Schowalter and G. M. Filip, editors. Beetle-pathogen interactions in conifer forests. Academic Press., London, England.

Lorio, P. L. J. 1994. The relationship of oleoresin exudation pressure (or lack thereof) to flow from wounds. Journal of Sustainable Forestry 1:81-93.

Lorio, P. L. J., and R. A. Sommers. 1986. Evidence of competition for photosynthates between growth processes and oleoresin synthesis in Pinus taeda L. Tree Physiology 2:301-306.

Lorio, P. L. J., R. A. Sommers, C. A. Blanche, J. D. Hodges, and T. E. Nebeker. 1990. Modeling pine resistance to bark beetles based on growth and differentiation balance principles. Pages 402-409 in R. K. Dixon, G. A. Meldahl, G. A. Ruark, and W. G. Warren, editors. Process modeling of forest growth responses to environmental stress. Timber Press, Portland, OR.

Luxmoore, R. J. 1991. A source-sink framework for coupling water, carbon, and nutrient dynamics of vegetation. Tree Physiology 9:267-280.

Mason, R. R. 1971. Soil moisture and stand density affect oleoresin exudation flow in a loblolly pine plantation. Forest Science 17:170-177.

McCullough, D. G., R. A. Werner, and D. Neumann. 1998. Fire and insects in northern and boreal forest ecosystems of North America. Annual Review of Entomology 43:107-127.

McDowell, N., J. R. Brooks, S. A. Fitzgerald, and B. J. Bond. 2003. Carbon isotope discrimination and growth response of old Pinus ponderosa trees to stand density reductions. Plant, Cell and Environment 26:631-644.

138 McDowell, N. G., H. D. Adams, J. D. Bailey, and T. E. Kolb. 2007. The role of stand

density on growth efficiency, leaf area index, and resin flow in southwestern ponderosa pine forests. Canadian Journal of Forest Research 37:343-355.

McHugh, C. W., and T. E. Kolb. 2003. Ponderosa pine mortality following fire in northern Arizona. International Journal of Wildland Fire 12:7-22.

McKinnon, M. L., D. T. Quiring, and E. Bauce. 1998. Influence of resource availability on growth and foliar chemistry within and among young white spruce trees. Ecoscience 5:295-305.

McMillin, J. D., and M. R. Wagner. 1996. Season and intensity of water stress effects on needle toughness of ponderosa pine. Canadian Journal of Forest Research 26:1166-1173.

McNeil, R. C , and D. B. Zobel. 1980. Vegetation and fire history of a ponderosa pine-white fir forest in Crater Lake National Park. Northwest Science 54:30-46.

McNulty, S. G., and W. T. Swank. 1995. Wood [delta-13]C as a measure of annual basal area growth and soil water stress in aPinus strobus forest. Ecology 76:1581-1586.

Miller, J. M., and F. P. Keen. 1960. Biology and control of the western pine beetle. USDA, Washington, D.C.

Mitton, J. B., and K. B. Sturgeon. 1982. Bark beetles on North American conifers: a system for the study of evolutionary biology. University of Texas Press, Austin.

Moeck, H. A., D. L. Wood, and K. Q. J. Lindahl. 1981. Host selection behavior of bark beetles (coleoptera: scolytidae) attacking Pinus ponderosa, with special emphasis on the western pine beetle, Dendroctonus brevicomis. Journal of Chemical Ecology 7:49-83.

Moore, K., B. Davis, and T. Duck. 2003. Mt. Trumbull ponderosa pine ecosystem restoration project. Pages 117-132 in P.N. Omi, L.A. Joyce (technical eds.), Fire, Fuel Treatments, and Ecological Restoration: Conference proceedings; 16-18 April 2002, Fort Collings, CO. RMRS-P-29, USDA Forest Service Rocky Mountain Research Station, Fort Collins, CO.

Nebeker, T. E., and J. D. Hodges. 1983. Influence of forestry practices on host susceptibility to bark beetles. Zeitschrift Fur Angewandte Entomologie 96:194-208.

O'Leary, M. H. 1981. Carbon isotope fractionation in plants. Phytochemistry 20:553-567.

139 Oehlert, G. W. 2000. A first course in design and analysis of experiments. W. H.

Freeman, New York.

Pearson, H. L. 1931. Theory in explanation of the selection of certain trees by the western pine beetle. Journal of Forestry 29:696-699.

Peck, R. W., A. Equihua-Martinez, and D. W. Ross. 1997. Seasonal flight patterns of bark and ambrosia beetles (Coleoptera: Scolytidae) in northeastern Oregon. Pan-Pacific Entomologist 73:204-212.

Perrakis, D. D. B. 2004. Seasonal fire effects on mixed-conifer forest structure and pine resin properties. Master of Science Thesis. University of Washington, Seattle, WA.

Perrakis, D. D. B., and J. K. Agee. 2006. Seasonal fire effects on mixed-conifer forest structure and ponderosa pine resin properties. Canadian Journal of Forest Research 36:238-254.

Phillips, M. A., and R. B. Croteau. 1999. Resin-based defenses in conifers. Trends in Plant Science 4:184-190.

Pinheiro, J. C , and D. M. Bates. 2000. Mixed-effects models in S and S-PLUS. Springer-Verlag, New York, NY.

Pureswaran, D. S., R. Gries, and J. H. Borden. 2004. Quantitative variation in monoterpenes in four species of conifers. Biochemical systematics and ecology 32:1109-1136.

Raffa, K. F., and A. A. Berryman. 1982. Physiological differences between lodgepole pines resistant and susceptible to the mountain pine beetle and associated microorganisms. Environmental Entomology 11:486-492.

Raffa, K. F., T. W. Phillips, and S. M. Salom. 1993. Strategies and mechanisms of host colonization by bark beetles. Pages 103-128 in T. D. Schowalter and G. M. Filip, editors. Beetle-pathogen interactions in conifer forests. Academic Press, San Diego.

Ruel, J. J., M. P. Ayres, and P. L. J. Lorio. 1998. Loblolly pine responds to mechanical wounding with increased resin flow. Canadian Journal of Forest Research 28:596-602.

Ryan, K. C. 2000. Effects of fire injury on water relations of ponderosa pine. Pages 58-66 in Innovative Silviculture and Vegetation Management, Tall Timbers Fire Ecology Conference Proceedings, No. 21, Tall Timbers Research Station, Tallahassee, FL.

140 Ryan, K. C , and W. H. Frandsen. 1991. Basal injury from smoldering fires in mature

Pinusponderosa Laws. International Journal of Wildland Fire 1:107-118.

Sackett, S. S., and S. M. Haase. 1998. Two case histories for using prescribed fire to restore ponderosa pine ecosystems in northern Arizona. Pages 380-389 in T.L. Pruden, L.A. Brennan (eds.), Fire in Ecosystem Management: Shifting the Paradigm from Suppression to Prescription. Tall Timbers Fire Ecology Conference Proceedings, No. 20, Tallahassee, FL.

Sartwell, C , and R. E. Stevens. 1975. Mountain pine beetle in ponderosa pine: prospects for silvicultural control in second-growth stands. Journal of Forestry 73:136-140.

Schade, G. W., and A. H. Goldstein. 2003. Increase of monoterpene emissions from a pine plantation as a result of mechanical disturbances. Geophysical Research Letters 30:1380-1383.

Schwilk, D. W., E. E. Knapp, S. M. Ferrenberg, J. E. Keeley, and A. C. Caprio. 2006. Tree mortality from fire and bark beetles following early and late season prescribed fires in a Sierra Nevada mixed-conifer forest. Forest Ecology and Management 232:36-45.

Seybold, S. J., J. Bohlmann, and K. F. Raffa. 2000. Biosynthesis of coniferophagous bark beetle pheromones and conifer isoprenoids: evolutionary perspective and synthesis. Canadian Entomologist 132:697-753.

Seybold, S. J., D. P. W. Huber, J. C. Lee, A. D. Graves, and J. Bohlmann. 2006. Pine monoterpenes and pine bark beetles: a marriage of convenience for defense and chemical communication. Phytochemistry Reviews 5:143-178.

Shepherd, W. P., D. P. W. Huber, S. J. Seybold, and C. J. Fettig. 2007. Antennal responses of the western pine beetle, Dendroctonus brevicomis (Coleoptera: Curculionidae), to stem volatiles of its primary host, Pinus ponderosa, and nine sympatic nonhost angiosperms and conifers. Chemoecology 17:209-221.

Smith, B. M. 1972a. Natural abundance of the stable isotopes of carbon in biological systems. Bioscience 22:226-231.

Smith, R. H. 1972b. Xylem resin in the resistance of the Pinaceae to bark beetles. General Technical Report PSW-1, USDA Forest Service Pacific Southwest Forest and Range Experiment Station, Berkeley, CA.

Smith, R. H. 1975. Formula for describing effect of insect and host tree factors on resistance to western pine beetle attack. Journal of Economic Entomology 68:841-844.

141 Smith, R. H. 1977. Monoterpenes of ponderosa pine xylem resin in western United

States. Technical Bulletin 1532, USDA Forest Service, Washington, D.C.

Smith, R. H. 2000. Xylem monoterpenes of pines: distribution, variation, genetics, function. General Technical Report PSW-GTR-177, USDA Forest Service Pacific Southwest Research Station, Berkeley, CA.

Stamp, N. 2003. Out of the quagmire of plant defense hypotheses. Quarterly Review of Biology 78:23-55.

Stark, R. W. 1965. Recent trends in forest entomology. Annual Review of Entomology 10:303-324.

Steele, C. L., S. Katoh, J. Bohlmann, and R. Croteau. 1998. Regulation of oleoresinosis in grand fir (Abies grandis). Plant Physiology 116:1497-1504.

Strom, B. L., R. A. Goyer, and P. J. Shea. 2001. Visual and olfactory disruption of orientation by the western pine beetle to attractant-baited traps. Entomologia Experimentalis et Applicata 100:63-67.

Stuart, J. D. 1984. Hazard rating of lodgepole pine stands to mountain pine beetle outbreaks in southcentral Oregon. Canadian Journal of Forest Research 14:666-671.

Sturgeon, K. B. 1979. Monoterpene variation in ponderosa xylem related to western pine beetle predation. Evolution 33:803-814.

Swezy, D. M. 1988. Mortality and stress of ponderosa pine after low intensity prescribed fire at Crater Lake, Oregon. M.S. Thesis. University of Washington, Seattle, WA.

Swezy, D. M., and J. K. Agee. 1991. Prescribed-fire effects on fine-root and tree mortality in old-growth ponderosa pine. Canadian Journal of Forest Research 21:626-634.

Taiz, L., and E. Zeiger. 1998. Plant physiology, second edition. Sinauer Associates, Inc., Sunderland, MA.

Thies, W. G., D. J. Westlind, and M. Loewen. 2005. Season of prescribed burn in ponderosa pine forests in eastern Oregon: impact on pine mortality. International Journal of Wildland Fire 14:223-231.

Thomas, T. L. 1982. Forest restoration of Sun Creek, Crater Lake National Park. M. S. Thesis. University of Washington, Seattle, WA.

142 Thomas, T. L., and J. K. Agee. 1986. Prescribed fire effects on mixed conifer forest

structure at Crater Lake, Oregon. Canadian Journal of Forest Research 16:1082-1087.

Thoss, V., and J. A. Byers. 2006. Monoterpene chemodiversity of ponderosa pine in relation to herbivory and bark beetle colonization. Chemoecology 16:51-58.

Trapp, S., and R. Croteau. 2001. Defensive resin biosynthesis in conifers. Annual Review of Plant Physiology and Plant Molecular Biology 52:689-724.

Tyree, M. T., and M. H. Zimmerman. 2002. Xylem structure and the ascent of sap. Springer, Berlin.

Vite, J. P. 1961. The influence of water supply on oleoresin exudation pressure and resistance to bark beetle attack in Pinusponderosa. Contributions from the Boyce Thompson Institute 21:37-66.

Wagener, W. 1961. Guidelines for estimating the survival of fire-damaged trees in California. Miscellaneous Paper 60, U.S.D.A. Forest Service Pacific Southwest Forest and Range Experiment Station, Berkeley, CA.

Wallin, K. F., T. E. Kolb, K. R. Skov, and M. R. Wagner. 2003. Effects of crown scorch on ponderosa pine resistance to bark beetles in northern Arizona. Environmental Entomology 32:652-661.

Wallin, K. F., T. E. Kolb, K. R. Skov, and M. R. Wagner. 2008. Forest management treatments, tree resistance, and bark beetle resource utilization in ponderosa pine forests of northern Arizona. Forest Ecology and Management 255:3263-3269.

Waring, R. H. 1987. Characteristics of trees predisposed to die. Bioscience 37:569-574.

Waring, R. H., and G. B. Pitman. 1985. Modifying lodgepole pine stands to change susceptibility to mountain pine beetle attack. Ecology 66:889-897.

Wilkens, R. T., M. P. Ayres, P. L. J. Lorio, and J. D. Hodges. 1998. Environmental effects on pine tree carbon budgets and resistance to bark beetles. Pages 591-616 in R. A. Mickler and S. Fox, editors. The productivity and sustainability of southern forest ecosystems in a changing environment. Springer-Verlag, New York.

Wood, D. L. 1972. Selection and colonization of ponderosa pine by bark beetles. Pages 101-117 in H. F. van Emden, editor. Insect/Plant Relationships. Halsted Press, New York.

143 Wood, D. L. 1982. The role of pheromones, kairomones, and allomones in the host

selection and colonization behavior of bark beetles. Annual Review of Entomology 27:411-446.

Zar, J. H. 1999. Biostatistical analysis. 4th edition. Prentice Hall, Upper Saddle River, N.J.

Zausen, G. L., T. E. Kolb, J. D. Bailey, and M. R. Wagner. 2005. Long-term impacts of stand management on ponderosa pine physiology and bark beetle abundance in northern Arizona: a replicated landscape study. Forest Ecology and Management 218:291-305.

144

APPENDICES

Appendix 1: Latin and common names of biota mentioned in text

Table Al. Latin and common names of species used in text Common name Latin name

Bark beetle Western pine beetle Mountain pine beetle Red turpentine beetle Ips species

Ponderosa pine Loblolly pine Lodgepole pine Western white pine Sugar pine Eastern white pine Aleppo pine Red pine Scots pine

White fir Grand fir

Norway spruce White spruce Larch species Douglas-fir Trembling aspen

Coleoptera (Scolytidae) Dendroctonus brevicomis D. ponderosae D. valens Ips spp.

Pinus ponderosa P. taeda P. contorta P. monticola P. lambertiana P. strobus P. halepensis P. resinosa P. sylvestris

Abies concolor A. grandis

Picea abies Picea glauca Larix spp. Pseudotsuga menziesii Populus tremuloides

Blue-stain fungus Ceratocystis sp.

145

Appendix 2: Variations in surf ace fuels following burning1

A2.1. Objectives, methods and data analysis:

Fuels reduction is a major topic of interest in modern fire management, and

prescribed burns are frequently-used treatment tools for modifying fuel loading in many

types of forest ecosystems (Fernandes and Botelho 2003). This experiment consisted of

an extension of a previous study on fuels changes following the previously-described

spring and fall prescribed burns in 2002 in Crater Lake NP. The objective of this

experiment was to monitor post-fire loading of dead woody fuels for two additional years

following prescribed fire. Since the duration of treatment effectiveness of prescribed

burning is uncertain and potentially short-lived, there is interest in tracking post-

treatment changes to fuel loading over time.

Study area, prescribed burn treatments, and plot layout were described previously

(Perrakis 2004). Surface fuels were measured along identical transects using standard

methods on 3 occasions: in summer 2001/2002 (pre-treatment), in summer 2003 (1 year

post-treatment), and in summer 2005 (3 years post-treatment). Standard planar-intersect

methods were used (Brown 1974), with equation coefficients for appropriate species

from van Wagtendonk et al. (1996). Litter and duff measurements were taken at 30

locations per experimental unit; fuel loading of these components was calculated from

equations in Agee (1973). The total length of transect surveyed per treatment type (spring

burn, fall burn, or control) was 1600 m (200 m per unit).

1 Although related to fire management at Crater Lake, this experiment was not related to the primary fire-bark beetle interaction topic and is therefore included separately.

146 Data were analyzed using analysis of variance (ANOVA), followed by Tukey's

HSD test for post-hoc comparisons (Zar 1999). Values were analyzed as the differences

from pre-treatment values at the two post-treatment times: 2003 and 2005. Differences

were compared separately between fine fuel classes (1, 10, 100 hr time-lag classes plus

litter) and total fuel loading (all layers) at the a=0.05 level of significance. Results from

all years are included for completeness.

A2.2. Results

As previously reported, fine and total fuel loading averaged 42.5 Mg/ha and 151.4

Mg/ha, respectively, before treatment, and differences between treatment groups were

not significant. Analysis results are summarized in Table A2 and shown in Figure Al.

Fine fuel loading on control units showed slight fuel increases in both 2003 and 2005,

suggesting some measurement variability or error between years. Both spring burns and

fall burns had significant reductions in fine fuels in both 2003 and 2005 compared to

controls; mean reductions were more limited in spring burn than fall burn units, but

differences between burn seasons was not significant at the 0.05 level. Total fuel loading

in control units indicated slight increases in 2003 and slight decreases in 2005.

Both spring and fall burning treatments also significantly reduced total fuel

loading (fine fuel + coarse wood + duff) in both years. Mean reductions in fall burn units

were greater than in spring burn units in both years, but differences between burn seasons

were only significant in 2003.

147 Table A2. Summary of changes to fuel loading, based on re-measurements in 2003 and 2005. Changes were calculated as Post - Pre; negative values therefore indicate fuel reduction. Different letters indicate significant differences between treatment means at the 0.05 level. ANOVA: fuel load differences AFuel loading (mean Mg/ha)

Year DF Controls Spr. Burns Fall Burns Fine fuels: 2003

2005 2 2

18.209 14.152

O.001 O.001

12.25a 7.18a

-2.90b -6.69b

-12.31b -17.44b

Total fuels: 2003 2 33.135 <0.001 16.85a -27.04b -72.81c 2005 2 7.781 0.003 -11.31a -45.49b -62.52b

Fine fuel-2003

9

8 a o

3 ° LL '

8

3 e ,

Total fuel -2003

8 -

o -

8-

8_

i

_ i _ i

o

a i r

Cbntrds Spring turns Fall burns

8

o A

8

Fine fuel - 2005

1

1

T $ i i

i

i

Total fuel - 2005

• i

i i

T r Cbntrols Spring bums Fall bums

Figure Al. Boxplots showing variations in fuel loading changes between treatments at two post-treatment times. Pre-treatment measurements were conducted in 2001/2002 and data presented are differences from pre-treatment values. Center lines and boxes represent median and interquartile range values, respectively; circles depict extreme values.

148

Duff Litter Coarse Wood Fine Wood

Pre-trt 2003 2005

Figure A2. Average surface fuel loading at the three sampling times, broken down by fuel component. Dotted lines represent timing of burn treatments.

Figure A2 shows the variation over time between the various components of the

dead surface fuels. As the figure shows, the majority of the fuel mass was contained in

149 the coarse woody fuels and forest floor components (litter and duff). In addition,

variability in fuel loading in control units over the 3 measurement years suggests

considerable measurement error or variability within experimental units. The most

difficult and subjective fuelbed components to measure were the depths of litter and duff

layers.

In terms of overall mean percentage change and recovery, by 2005, spring burn

units contained 84% and 71% of pre-burn fine and total fuel levels, respectively;

comparable figures for fall burns are 60% (fine) and 55% (total), while fuels measured on

controls were 117% (fine) and 93% (total) of 2001 levels.

A2.3. Discussion

The burn treatments successfully reduced both fine and total fuels, and

differences were still statistically significant by 2005. However, the benefits in terms of

reduced fuel loading were relatively limited, particularly in spring burn units, and likely

short-lived as fire-killed biomass progressively becomes part of the surface fuels layer in

the immediate post-fire years. As previous studies have suggested, a first post-exclusion

restoration burn will often kill but not consume vegetation, reducing the fuel-reduction

effectiveness of the treatment (Thomas and Agee 1986, Agee and Huff 1987); in the case

of a dense forest with many trees killed, dead surface fuel loads can be expected to

surpass pre-burn levels within a few years after fire, if the treatment is not renewed.

Standing fire-killed snags will eventually fall and contribute to coarse surface fuel

loading, with larger-diameter snags generally lasting longer than smaller-diameter ones

150 (as reviewed by Everett et al. 1999). Loading of finer materials tends to increase after fire

as well, and may peak and then drop if decomposition surpasses inputs of new material,

in the absence of another fire event. In the Colorado front range, Hall et al. (2006) found

that total woody debris levels in ponderosa pine forests peaked approximately 10 year

after individual (primarily high severity) fire events before declining for several decades

afterwards. For litter and duff layers, the response after a single fire appears asymptotic,

with no clear peak discernible even after many years. Mackenzie et al. (2004) found that

approximately 2/3 of the forest floor had returned 10-15 years after fires in ponderosa

pine/Douglas-fir forests in western Montana, with no evidence of peaking even 130 years

after fire (severity or consumption levels of fires were not specified, unfortunately).

Fernandes and Botelho (2003), citing several articles from locations around the world,

suggest that litter fuels tend to recover even quicker - approximately 2-5 years after

typical prescribed fire treatments in a number of different ecosystems.

A recent publication of long-term fuels monitoring work in the Sierra Nevada

range is probably the most comparable to this study (Keifer et al. 2006). Pre-burn fuel

levels (and tree species composition) were quite similar to conditions at Crater Lake, with

163-191 Mg/ha of total fuels measured before burning. Total fuel loading in ponderosa

pine forests, initially reduced by 99 % by prescribed burning, had recovered to 85 % of

pre-burn fuel levels after 10 years. A separate series of plots in different ponderosa pine

vegetation associations was monitored after burning, and had surpassed pre-burn levels -

by 57 %, 80 %, and 50 % - 31 years after prescribed fire. In white fir/mixed-conifer

forests, fuel loading had returned to 83 % of pre-burn levels 10 years after burning,

151 despite being reduced by 85 percent immediately after burning. Post-fire fuel

composition tended to be different than pre-treatment conditions, with higher levels of

woody fuels and lower levels of duff post-burn (Keifer et al. 2006).

In terms of operational effectiveness, anecdotal evidence has suggested for many

years that prescribed burning, with or without mechanical thinning, could be effective in

mitigating subsequent wildfire severity and crown fire likelihood in ponderosa pine

forests (reviewed by Finney et al. 2005). Several rigorous studies have now confirmed

that suggestion in various western forest types (Pollet and Omi 2002, Finney et al. 2005,

Raymond and Peterson 2005). Of particular relevance is the study by Finney et al.

(2005), which found that fuel treatments, including prescribed burns, conducted up to 9

years prior effectively reduced the severity of the 2002 Rodeo-Chediski fire complex -

the largest wildfire in Arizona in modern history. However, Southwestern ponderosa pine

forests tend to have lower overall surface fuel levels than those in this study (e.g. Fule et

al. 2002); annual productivity of these forests are also quite different. It should not be

assumed that the findings of Finney et al. (2005) apply equally well to the ponderosa pine

forests at Crater Lake.

In sum, by 2005 the 2002 prescribed fires at Crater Lake still appear somewhat

successful at lowering fuels from fire exclusion levels. However, it is clear that biomass

inputs from killed and scorched trees are undermining the gains achieved thus far.

Managers can expect total fuel loads to return quickly to pre-burn levels, probably within

about 10 years or less. In the long term, dead fuels in burned areas will likely surpass

152 levels in unburned areas if additional burns are not conducted to further restore and

maintain low severity fire.

The findings in this component of this study have focused on dead surface fuels,

which are only one part of the fuel complex that determines wildfire behavior. Canopy

fuels and structure are equally or more influential in determining the intensity of crown

fires (van Wagner 1977, Scott and Reinhardt 2001), although surface and canopy fuel

production are related, since post-fire increases in dead surface fuels are withdrawn from

existing canopy fuel loads. Additional modeling efforts on changes to forest structure and

implications for fire behavior would be needed to properly assess the dynamic changes

that occur following restoration or fire-safe treatments over time. While fuel management

objectives may only require additional prescribed burns, managers will also have to

balance fuel reduction with the goal of maintaining old ponderosa pines at these sites, as

discussed in the main body of this study. Fire regime restoration at Crater Lake will

require continued commitment and long-term investment of time and resources if it is to

succeed.

A2.4. References

Agee, J. K. 1973. Prescribed fire effects on physical and hydrologic properties of mixed-conifer forest floor and soil. Contributions of the University of California Water Resource Center No. 143.

Agee, J. K., and M. H. Huff. 1987. Fuel succession in a western hemlock/Douglas-fir forest. Canadian Journal of Forest Research 17:697-704.

Brown, J. K. 1974. Handbook for inventorying downed woody material. Gen. Tech. Report INT-16, U.S.D.A. Forest Service, Ogden, Utah.

153 Daudet, F.-A., T. Amegio, H. Cochard, O. Archilla, and A. Lacointe. 2005. Experimental

analysis of the role of water and carbon in tree stem diameter variations. Journal of Experimental Botany 56:135-144.

Ducrey, M., F. Duhoux, R. Hue, and E. Rigolot. 1996. The ecophysiological and growth responses of Aleppo pine {Pinus halepensis) to controlled heating applied to the base of the trunk. Canadian Journal of Forest Research 26:1366-1374.

Everett, R. L., J. F. Lehmkuhl, R. Schellhaas, P. Ohlson, D. Keenum, H. Riesterer, and D. Spurbeck. 1999. Snag dynamics in a chronosequence of 26 wildfires on the east slope of the Cascade Range in Washington state, USA. International Journal of Wildland Fire 9:223-234.

Fernandes, P. M., and H. S. Botelho. 2003. A review of prescribed burning effectiveness in fire hazard reduction. International Journal of Wildland Fire 12:117-128.

Finney, M. A., C. W. McHugh, and I. C. Grenfell. 2005. Stand- and landscape-level effects of prescribed burning on two Arizona wildfires. Canadian Journal of Forest Research 35:1714-1722.

Fule, P. Z., W. W. Covington, H. B. Smith, J. D. Springer, T. A. Heinlein, K. D. Huisinga, and M. M. Moore. 2002. Comparing ecological restoration alternatives: Grand Canyon, Arizona. Forest Ecology and Management 170:19-41.

Hall, S. A., I. C. Burke, and N. T. Hobbs. 2006. Litter and dead wood dynamics in ponderosa pine forests along a 160-year chronosequence. Ecological Applications 16:2344-2355.

Keifer, M., J. W. van Wagtendonk, and M. Buhler. 2006. Long-term surface fuel accumulation in burned and unburned mixed-conifer forests of the central and southern Sierra Nevada, CA (USA). Fire Ecology 2:53-72.

Lassoie, J. P. 1973. Diurnal dimensional fluctuations in a Douglas-fir stem in response to tree water status. Forest Science 19:251-255.

Mackenzie, M. D., T. H. DeLuca, and A. Sala. 2004. Forest structure and organic horizon analysis along a fire chronosequence in the low elevation forests of western Montana. Forest Ecology and Management 203:331 -343.

Perrakis, D. D. B. 2004. Seasonal fire effects on mixed-conifer forest structure and pine resin properties. Master of Science Thesis. University of Washington, Seattle, WA.

Scott, J. H., and E. D. Reinhardt. 2001. Assessing crown fire potential by linking models of surface and crown fire behavior. Research Paper RMRS-RP-29, USDA Forest Service Rocky Mountain Research Station.

154 Thomas, T. L., and J. K. Agee. 1986. Prescribed fire effects on mixed conifer forest

structure at Crater Lake, Oregon. Canadian Journal of Forest Research 16:1082-1087.

van Wagner, C. E. 1977. Conditions for the start and spread of crown fires. Canadian Journal of Forest Research 7:23-34.

van Wagtendonk, J. W., J. M. Benedict, and W. M. Sydoriak. 1996. Physical properties of woody fuel particles of Sierra Nevada conifers. International Journal of Wildland Fire 6(3): 117-123.

Zar, J. H. 1999. Biostatistical analysis. 4th edition. Prentice Hall, Upper Saddle River, N.J.

155 Appendix 3: Use of automatic dendrometers

The initial design of Experiment 3 involved the use of automatic dendrometers for

measuring radial tree growth. The high resolution (temporal and radial) of these

instruments allows for the tracking of diurnal variations in tree stem diameters, as well as

the seasonal timing of events such as radial growth. Dendrometer data can also provide

insight into the moisture status of a tree. Unfortunately, serious problems were

encountered with the equipment and experimental setup, and no dendrometer data was

used in any analyses in this study. This section summarizes the rationale, difficulties, and

some results involved with the dendrometer work.

A3.1. Dendrometer overview

The technique of using automatic sensors to record short-term stem diameter

fluctuations is not new. Diurnal patterns of diameter variation were documented as early

as the 1920's, with several investigations of the phenomenon in the late 1960's and early

1970's (see references in Lassoie 1973). Working on pole-sized Douglas-fir, Lassoie

(1973) noted correlations between branch water potential and diameter. The article

explains that high midday transpiration decreases stem (or branch) water potential. Daily

minima of stem water potential are directly correlated to times of greatest stem shrinkage,

sometimes lagged by up to 2-3 hours. Other factors were later found to contribute to

diameter fluctuations, including radial growth, thermal expansion, and tension levels in

dead conducting tissues (Daudet et al. 2005). Since diameter of a stem or branch

156 measured at any particular time is affected by these numerous predictor variables, the

interpretation of dendrometer data requires an understanding of vascular physiology and

hydraulic processes.

Most studies using automatic dendrometers on conifers have been exploratory in

nature, using only one or very few samples (Lassoie 1973, Ducrey et al. 1996). The

original intent of this study was to have dendrometers on all 25 trees in Experiment 3.

The stem diameter traces were to be used to identify the periods of active growth,

determine the moisture status of each tree (using the daily amplitude fluctuations), and

determine annual growth by calculating the difference between spring diameter and fall

diameter each year. None of these steps proved simple to derive from the data, and other

methods proved more workable for determining most of these factors. Thus, increment

cores were taken to determine annual growth and 13C discrimination was used to

determine average moisture status during a growing season. The one great strength of the

dendrometer data is in measuring the timing of growth events. Apart from manually

taking daily DBH measurements, no other technique can match the temporal resolution

of automatic dendrometers, with their ability to continuously log diameter variability

(e.g. every hour or even more frequently). There may still be an opportunity to salvage

some of the dendrometer data from this study for this purpose, but for the present

document, this was not feasible.

157 A3.2. Methods

The dendrometer package purchased for this project consisted of point

dendrometers (25), automatic data loggers (2), solar panels, batteries, and cables

assembled as a custom package by Agricultural Electronics Inc. (Tucson, AZ). Point

dendrometers were of the Linear Variable Differential Transformer (LVDT) type2 with a

rated precision of ±2 urn of radius. Point dendrometers were chosen over band

dendrometers to avoid known problems with the latter due to temperature expansion

and/or band corrosion. The dendrometers were fixed in place on sample trees in

Experiment 3 with a sensor head held in constant pressure against a clean section of bark.

Installation on sample trees involved drilling two thin holes (~ 1.6 mm diameter) into the

heartwood and screwing in two sections of threaded rod, to which the dendrometer was

bolted (Figure A3). One dendrometer was installed on the north side of each sample tree,

at approximately breast height (1.4 m).

Dendrometer cables ran from the base of each dendrometer to one of two 'Pod'

data loggers (model AN3, Agricultural Electronics Inc., Tucson, AZ) in the study area.

Each Pod was a weatherproof unit with dendrometer and other cables emerging from its

base, powered by an internal battery recharged by a solar panel. Solar panels were

installed on posts in small clearings on the edges of the study area, approximately 1 m off

the ground. Due to the location of the study trees at the site, 15 trees were wired to one

Pod, and 10 to another.

2 See online diagrams at http://www.phytogram.com/dendrometer2.htm

http://www.phytogram.com/dendrometer2.htm

158

Figure A3. A dendrometer installed on a sample tree. The arrow indicates the location of the measurement sensor at the bark interface.

Pod data loggers and dendrometers were installed and initiated at the beginning of

the experiment, in late June 2003. The Pods were programmed to read dendrometer

values every hour, and contained sufficient internal memory to store about 6 months of

data before overwriting earlier values. Initially, the Pods were dismantled and stored off-

site for the winter in fall 2003. This was extremely labor-intensive, requiring the

unplugging (in the fall) and rewiring (the following spring) of all dendrometers, solar

panels, and grounding cables, and re-installation of the assembly at the site the following

season. Following the second year of data collection (fall 2004), the Pods were left

running onsite. This was also problematic (see below), and in subsequent seasons, the

159 dendrometers were removed annually over winter, but the Pods and wiring were left on

site.

Retrieving data from the Pods involved a cumbersome series of computing steps.

The initial file download involved connecting a serial port cable from a Pod to a laptop

computer and running a communication utility (Microsoft Windows Hyperterminal;

Microsoft Corp., Redmond, WA) that downloaded raw data files. This was a workable

although awkward process that involved about 5 minutes of holding the laptop at Pod

height while each month of data was transferred. Data files downloaded in this manner

were in an unreadable format; a simple Microsoft DOS-based program had to be run to

convert files to a text format. Text files were then processed by a Microsoft Excel macro

to organize them into columns and make them readily visible in a spreadsheet. This

process yielded values in millivolt-eighths. Another simple calculation using individual

dendrometer calibration values (provided by Agricultural Electronics Inc. based on

laboratory testing of each dendrometer) finally produced values in mm. Although the

calibration values for each unit varied, the conversion was on average 200 mV per mm of

radius. These radial increases or decreases could be compared as changes between

treatment groups, or differenced from initial diameter values to give overall cambial

growth over a period of time.

A3.3. Problems and challenges encountered

One of the major drawbacks of the Pod design was that there was no simple

means of assessing whether or not the dendrometers (and Pods) were functioning

160 properly. The Pod user interface was very rudimentary, with a very simple text menu

accessed on a laptop computer via the Hyperterminal protocol described above. Live

testing of individual dendrometer channels was possible in the field, although there was

no graphic capability of any sort, merely a single value shown at any one time.

Continuous testing showed a value in millivolts varying continuously within a narrow

range around the mean. Thus, a test of a functioning dendrometer might show on-screen

fluctuations between 2110 and 2120 mV over the course of a few seconds. Through trial

and error, it was discovered that a steady value of 4096 mV represented a short-circuit

error (such as if cables had been shredded by animals).

A more difficult problem (as yet unsolved) was encountered later, involving a

fluctuating value centered around ~2500 mV. For some reason, a disconnected or cleanly

severed (but not short-circuited) dendrometer cable would cause the reading to fluctuate

around this value, with a short-term pattern of variability that was frustratingly similar to

a 'real' value (ie., ± 10-15 mV). However, this fault value would not vary by time of day,

or over the course of several days - it would continue to jump up and down seemingly at

random around the value of 2500 mV. If a value of 2500 mV represented an obviously

outlying value, this problem would have been more easily diagnosed. However, the range

of common values measured during the course of the experiment was between about

1500 mV (starting values) and about 3200 mV (finishing values), representing a

difference of up to about 8 or 9 mm in total radial change. So by years 2 or 3 (2004/5),

error values fluctuating around 2500 mV were not easily diagnosed as errors at all. Only

by running through the data processing steps described above, and failing to see the tell-

161 tale sinusoidal hourly fluctuation pattern could such a fault be diagnosed. This problem

was the single biggest source of lost data, as it was not diagnosed until the 2005 season,

and was not easily detected or fixed even at that time.

As implied above, using dendrometers in the field setting also proved to be a

major challenge. The first problems were encountered in 2003. One of the dendrometers

was accidentally bent while digging one of the treatment trenches. The device was sent

back to the manufacturer, who repaired it and returned it 2 months later. This resulted in

nearly a full season of lost data for one tree (from the 'light trenching' group). Another

data gap resulted from burn preparations for the prescribed burn treatment group. Fire

management personnel from the Oregon Department of Forestry advised the field team in

early July that the burning window was imminent, and that we should make sure our burn

area was ready for the treatment. Dendrometers and cables were removed from the site

on July 11. In the end the burn was not lit until August 2 and smoldered until August 5;

dendrometers were finally reattached on August 8, resulting a data gap of nearly a month.

(After all the delay, an unfortunate misunderstanding with the ODF fire crew made our

research crew miss the actual burn. The crew ignited the unit one day earlier than

expected, and we were greeted the following morning by smoke and charred trees.)3

Many other incidents occurred that resulted in data losses over the four years of

this experiment. In several instances, some type of animal apparently chewed through the

instrument cables. These faults were sometimes only discovered several days later, as

3 This is not meant as a criticism of ODF personnel, who were very gracious and helpful throughout the experiment. Prescribed fire planning is a notoriously uncertain process, and delays are common. The clarification is offered here to explain the cause of a large data gap in the burn treatment group during the first year of the experiment (2003), as well as the lack of photos of the burning operation.

162 several other experiments were taking place concurrently, and the Pods were not tested

exhaustively every day (as described above, to diagnose these problems required

downloading the most recent data and carefully examining it to ensure it matched the

correct diurnal pattern). Other data losses resulted from battery failures, the seizing of

sensors due to arthropod cocoons inside them, and the gradual loosening on several

dendrometers of the constant-pressure mechanism that held the sensor against the tree.

A3.4. Results and discussion

Over 4 years of study, over 300,000 hourly radius measurements were collected

from the 25 trees. Data collection occurred between 01 July and 24 September 2003

(2039 hours); 21 May and 09 October 2004 (3384 hours); 01 May and 31 October 2005

(4416 hours); and 22 May and 20 September 2006 (2905 hours). Overall, about 10% of

the total data points were discarded after being diagnosed as as error values, with errors

variably distributed between all 25 trees.

Annual radial growth values measured by the dendrometers correlated only

weakly with the radial growth measured by ring widths (see Experiment 3). Figure A4A

and B show the correlations between the dendrometer data (by year and tree) and the

Experiment 3 increment core widths. Dendrometer data was summed up for each year by

subtracting the final diameter in the fall (average of last 24 hrs) from the second day

spring starting value (average of second 24 hr period). In the event that deleted error

values existed within these ranges, the closest available 24 hr period was chosen instead.

Figure A4A shows all years and trees (negative growth values graphed as '0'), and

163 Figure A4B shows only the data pairs with relatively clean dendrometer data: deleted

tree-year combinations include all values for 2003 (since the dendrometers were installed

after much of the growing season had occurred), all values with negative net growth for a

given year, and years when more than 20% of the data was faulty on a given tree.

The correlations highlight one of the problems with point dendrometers - they are

highly prone to confounding when the growth rate around a tree is uneven. This was

recognized as an issue with the burn treatment trees (note all the 'Burn' group trees with

0 dendrometer growth in Figure A4A), but may have been an issue with other samples as

well. With the known data gaps removed, the correlation between methods improves

considerably, but is still rather weak.

The other intended function of the dendrometers was to track the diurnal

variations in radius, in order to infer moisture stress between trees and treatment groups.

Although a diurnal pattern was evident, it was quite ragged, appearing highly out of

phase at certain times compared to published reports. Figure A5 A - D shows some

examples of dendrometer traces. In three of the examples (A, B, D), radial traces appear

exactly out of phase, with maxima occurring in mid-afternoon (13:00 to 17:00). Many

studies have shown that radii are least in the heat of the afternoon, as sap tension levels

increase in response to evapotranspiration, and greatest at dawn when humidity is higher

and water demands are less. In the third example (C), the diurnal amplitude is much

lower, and the fluctuation matches the known variation (dawn maxima, afternoon

minima).

164

A:

£ I s 3 3

fl

*+

3 . 5

3

2 . 5

2

1.5

1

0.5

O

• Control I A Burn

| • Prune • R30 • R60

y" A

^~ A :

•

• A

A A ^ B * *

•

•

/*

• •

•

• • • •

•

«

•

*

• •

/

•

•

• " • •

+ s

*

•

•

• •

-T#-*W* •—

«

•

• /

m>

• •

—•—A-

A

A /

* .

•

•

• j <

•

A

•

A

A—

o 0 . 5 1.5 2 2 .5

Core (mm) 3 . 5

3 . 5

I 2 . 5

1.5

0 . 5

• Control ; A Burn • Prune • R30 • R60

0 . 5 1.5 2 2.5 Core width (mm)

3 . 5

B : l • —

Figure A4 A and B. Correlations between total annual radial growth as measured by dendrometers and by ring widths. Each point represents one year of growth and one tree. A: all data. B: known faulty data deleted (see text for details). The pink line represents perfect correlation. For graph A, r = 0.398; for graph B, r = 0.594.

165 The results as such (out of phase radial fluctuations) are clearly nonsensical for a

certain proportion of the dataset. A fairly extensive search was done to identify whether

there was an obvious explanation (e.g. clock error in the Pods, multiplication by a

negative number required, etc.), but none was found. Mid-afternoon 'spikes', either

upwards or downwards (Figure A5D), were also prominent at certain times and remain

unexplained.

With huge variability apparent from day to day, the severe data gaps mentioned

above, and great uncertainty as to the integrity of the remaining data, this portion of

Experiment 3 was abandoned in favor of alternate methods. Future investigations using

these types of dendrometers are advised to begin by having Pods and dendrometers tested

and recalibrated by the manufacturer prior to installation. In addition, dendrometers

should be checked daily to ensure that the data being produced is reasonable.

Unfortunately, because of the very primitive user interface, the latter task is very onerous.

Future users may want to enlist the services of a programmer to produce a software

application for monitoring the data more easily. At the least, such an interface should,

say, permit a quick perusal of the previous few days of measurements, preferably in

graphical form.

166 A: 15-25 June, 2004. The bottom two lines are from burn treatment trees, and top line is from a pruned tree.

( J "nlTTITlTI

8 8 8 8 8 8 8 8 8 CD CO CM CM

B: 06 - 16 August, 2004. Three control trees.

4.5

4

3.5

3

2.5

2

1.5

1

0.5

8 8 8 8 8 8 8 8 ™™ ^ j,™ ^ j^j pp £j3 0 3 (j5 U5~-

Figure A5 A, B. Examples of diurnal variations in dendrometer traces.

167 C: 27 June- 05 July, 2005. Two control trees. Note different vertical scale compared to previous figures.

5.5

5 i

4.5

3.5

n i i i i i i i i i i i i i i i i n i i i

8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 O O O N O f f l N d M I N d W N d f f l N d o J N d f l d N d d O N

D: 05 - 10 July, 2003. Two control trees, each attached to a different Pod data logger.

- 1 h'rrnTi 11 11 M 11 rTTrrnTTrrrrrn "T I I I I I I I I i-n-rrr'nTrrrrrrTn r i m 11 m irrr;rrrrrrrr|"rn-HTTrT,rr,i,iT"fi,TT,),ri,i'[ ITT'I M M M I I I I I I I

8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8

Figure A5 (continued) C, D. More examples of diurnal variations in dendrometer traces.

A3.5. References

Daudet, F.-A., T. Amegio, H. Cochard, O. Archilla, and A. Lacointe. 2005. Experimental analysis of the role of water and carbon in tree stem diameter variations. Journal of Experimental Botany 56:135-144.

168 Ducrey, M., F. Duhoux, R. Hue, and E. Rigolot. 1996. The ecophysiological and growth

responses of Aleppo pine (Pinus halepensis) to controlled heating applied to the base of the trunk. Canadian Journal of Forest Research 26:1366-1374.

Lassoie, J. P. 1973. Diurnal dimensional fluctuations in a Douglas-fir stem in response to tree water status. Forest Science 19:251-255.

169

Appendix 4: Southern Oregon Palmer Drought Severity Index (PDSI)

Table A3. PDSI Anomaly (1971-2000), South-central Oregon, 2002-07 Year Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 2002 -3.50 -3.65 -3.77 -3.65 -3.95 -4.24 -4.81 -4.97 -5.26 -5.07 -5.44 -4.68 2003 -4.98 -5.13 -5.23 -4.10 -3.51 -4.12 -4.68 -4.85 -4.69 -4.84 -4.56 -3.38 2004 -3.21 -2.44 -2.99 -3.06 -2.05 -1.77 -1.87 -1.41 -1.82 -0.94 -1.57 -1.26 2005 -2.17 -2.74 -3.20 0.51 2.06 1.97 1.39 0.46 -0.17 0.12 0.58 1.95 2006 2.32 2.53 2.79 2.85 2.74 3.20 3.07 -0.25 -0.88 -0.99 -1.19 -0.93 2007 -1.90 -2.08 -3.01 -3.03 -3.90 -4.11 -4.88 -5.06 -4.87 -3.74 -3.70 -3.55

Source: NOAA (USA) Climate Data Center online: http ://www. cdc.noaa. gov/cgi-bin/Timeseries

170 VITA

Daniel Perrakis was born in Ottawa, Ontario, Canada. He holds a Bachelor of Science degree from the University of British Columbia (1999) and completed a Master of Science from the University of Washington in 2004. Since 2006 he has worked as a Fire Ecologist with Parks Canada Agency out of Calgary, Alberta. When not thinking about fire science, he enjoys hiking, climbing, and skiing adventuring in the mountains as well as playing the mandolin.

fire-bark beetle interactions: exploring links between ... · responses after bole charring, but...

Documents