Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Patterns and impact of herbivory by a biological control insect on its target weed and a native nontarget… Catton, Haley Autumn 2014

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

Item Metadata

Download

Media
24-ubc_2014_spring_catton_haley.pdf [ 947.8kB ]
Metadata
JSON: 24-1.0074334.json
JSON-LD: 24-1.0074334-ld.json
RDF/XML (Pretty): 24-1.0074334-rdf.xml
RDF/JSON: 24-1.0074334-rdf.json
Turtle: 24-1.0074334-turtle.txt
N-Triples: 24-1.0074334-rdf-ntriples.txt
Original Record: 24-1.0074334-source.json
Full Text
24-1.0074334-fulltext.txt
Citation
24-1.0074334.ris

Full Text

PATTERNS AND IMPACT OF HERBIVORY BY A BIOLOGICAL CONTROL INSECT ON ITS TARGET WEED AND A NATIVE NONTARGET PLANT  by  HALEY AUTUMN CATTON  Bachelor of Science, University of Manitoba, 2001 Master of Science, University of Manitoba, 2005  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF  THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY  in  THE COLLEGE OF GRADUATE STUDIES (Biology)      THE UNIVERSITY OF BRITISH COLUMBIA (Okanagan) April 2014   ? Haley Autumn Catton, 2014  ii ABSTRACT  Classical biological control (biocontrol) of weeds involves importing foreign, host-specific insects with the intent of reducing the fitness of invasive plants in their introduced range. When anticipated risks to nontarget species are low, insects capable of feeding and developing on some native nontarget plants have been given regulatory approval for release.  In this thesis, I study patterns and impacts of herbivory by the root-feeding weevil Mogulones crucifer on its target weed Cynoglossum officinale and a native nontarget plant Hackelia micrantha.  I released large numbers of M. crucifer into naturally-occurring patches of H. micrantha growing with or without C. officinale to simulate a ?worst case? scenario of high insect density and low target plant density, and subsequently recorded herbivory patterns and plant demographic parameters for two years on release and non-release sites. Additionally, I collected oviposition data from non-experimental sites 0-4 years after M. crucifer release. Compared to the target weed, H. micrantha use by M. crucifer was temporary, rare, mild, and limited to immediately around release points, suggesting that the nontarget plant is buffered from population-level effects by spatial, temporal and probabilistic refuges from biocontrol herbivory. M. crucifer did not persist 2 years after release in the absence of C. officinale, indicating that the insect is limited to ?spillover? nontarget use. A separate mark-release-recapture experiment indicated that M. crucifer has reduced host-finding behaviours for its novel host compared to its evolutionary host. Plant demographic data indicated that outbreak densities of M. crucifer appeared to impact C. officinale populations by increasing rosette mortality. While there was some evidence of impact to individual H. micrantha plants immediately adjacent to release points (i.e., plant death or dieback of flowering shoots), these effects did not translate to the population level.  I synthesize this information to suggest why M. crucifer has been an effective biocontrol agent against C. officinale in Canada, and why H. micrantha is unlikely to incur population-level effects from the weevil. This study is a clear example of how nontarget use of individual plants can be noticeable yet not have population-level implications, and demonstrates the importance of post-release research in weed biocontrol.    iii PREFACE  The research in this dissertation was designed, performed, analyzed and written by me, Haley Catton, in collaboration with my supervisory committee Drs. Robert Lalonde, Rosemarie De Clerck-Floate, Karen Hodges, Jason Pither, and Rebecca Tyson. Data were collected with the help of field research assistants Jordan Bannerman, William Van Der Weide, Shelby McLeod and Karma Tiberg.  Statistical analysis of the demographic data in Chapter 4 was supervised by Prof. Yvonne Buckley (formerly of the University of Queensland and the Commonwealth Scientific and Industrial Research Organisation, Brisbane, Australia, and currently of Trinity College Dublin, Ireland).       iv TABLE OF CONTENTS  ABSTRACT?????????????????????????...????ii  PREFACE?????????????????????????...???..?iii  TABLE OF CONTENTS?.???????????????????????.iv  LIST OF TABLES???????????????????????????vi  LIST OF FIGURES??????????????????????????..vii  ACKNOWLEDGMENTS???...????????????????????..ix  DEDICATION????????????????????????????..xi  CHAPTER 1: Introduction: benefits, risks and challenges regarding classical  biological control of weeds?????????????????????1   1.1.  Invasive plants and biological control??????????????....1   1.2.  Nontarget use????????????????????.......??...2   1.3.  Challenges in studying impact of biocontrol agents?????????..4  1.4.  Study system????????????????????????...6  1.5. Thesis overview??????????????????????.......8  CHAPTER 2: Intensity and temporal patterns of M. crucifer herbivory on target  and nontarget plants???????????????????????..10  2.1.  Literature review and objectives????????????????..10  2.2.  Methods??????????????????????????15   2.2.1.  Study system????????????????????..15   2.2.2.  Rangeland experiment (visual use assessments and destructive  sampling)????????????????????......15 2.2.3.  Non-experimental release sites?????????????..17 2.2.4.  Data analysis????????????????????..18 2.3.  Results??????????????????????????..19  2.3.1.  Visual indications of use??????????????..?.19  2.3.2.  Destructive sampling and dissections???????????20 2.4.  Discussion????????????????????????.....22  CHAPTER 3: Differential within-patch target and nontarget use following  M. crucifer releases: spatial patterns and underlying mechanisms?...???...33 3.1.  Literature review and objectives???????????????.?.33  3.2.  Methods??????????????????????????37 3.2.1.  Study system???????????????????......37   3.2.2.  Rangeland experiment????????????????...38 3.2.3.  Mark-release-recapture experiment???????????...39 3.2.4.  Data analysis???????????????????......41  v  3.3.  Results??????????????????????????..43 3.3.1.  Rangeland Experiment????????????????..43   3.3.2.  Mark-release-recapture experiment???????????...44  3.4.  Discussion????????????????????????....47  CHAPTER 4: Individual and population-level impacts to target and nontarget plants  following M. crucifer release???????????????????...56 4.1.  Literature review and objectives????????????????..56  4.2.  Methods??????????????????????????60 4.2.1.  Study system????????????????????..60   4.2.2.  Rangeland release experiment?????????????...62 4.2.3.  Seedling emergence experiment?????????????65 4.2.4.  Analysis of weevil impact to vital rates......????????..66 4.2.5.  Transition matrix construction, parameterization, and analysis?67  4.3.  Results??????????????????????????..69   4.3.1.  C. officinale dynamics????????????????...70   4.3.2.  H. micrantha dynamics????????????????.71   4.3.3.  Effect of distance from release points??????????...72   4.3.4.  Use scars as indicators of demographic impact???????.73  4.4.  Discussion????????????????????????....74 4.4.1.  Impact of M. crucifer on C. officinale??????????...75   4.4.2.  Impact of M. crucifer on H. micrantha??????????.78    4.4.3.  Implications for impact monitoring???????????...81  CHAPTER 5: Conclusions: implications of M. crucifer herbivory patterns and  demographic effects??????..????????????????.108  5.1.  Why M. crucifer has been an effective biocontrol agent against C.  officinale in Canada???????????????????...111 5.2.  Reasons for lack of effect of M. crucifer on H. micrantha populations?.112 5.3.  Management recommendations???????????????..?113 5.4.  Future work???????????????????????....115  5.4.1.  Refining and expanding population models??????...?.115  5.4.2.  Further exploration of M. crucifer herbivory effects on  individual plants?????????.????????....117  5.4.3.  Investigating potential changes in M. crucifer host preference and performance???.??????????????....119 5.5.  Final thoughts???????????????????????.120  REFERENCES CITED????????????????????????..122  APPENDICES????????????????????????????138  APPENDIX A:  Plant fecundity calculations?????????????.138  APPENDIX B:  Calculating upper and lower parameter boundaries from         the seedling emergence experiment?????????...140    APPENDIX C:  Individual site demographic parameters?...??????...145  vi LIST OF TABLES   Table 2.1.   Numbers of M. crucifer eggs and larvae found in C. officinale and H.  micrantha plants.??????????????????????..28  Table 4.1.   Symbols and definitions for annual probability and fecundity parameters  used in matrix models.????????????...???????..84  Table 4.2.   Annual transition values for C. officinale on sites grouped by M.  crucifer releases??????????????????????...85  Table 4.3.   Annual transition values for H. micrantha on sites grouped by M.  crucifer releases and abundance of C. officinale (?target common? and  ?target rare?)????.???????????????????...86  Table A.1.  Sources of the values of the number of tetrads per bolting stems used for  calculating plant fecundity values??..?????????????139  Table B.1.   Parameter values for C. officinale and H. micrantha seed and seedling  dynamics???..????????????????????.?..143  Table C.1.  Overall C. officinale rosette and bolting transitions for each site-year combination...???????????????????????.146  Table C.2.  Overall H. micrantha rosette and bolting transitions for each site-year combination...???????????????????????.149      vii LIST OF FIGURES  Figure 2.1.   Probabilities of C. officinale and H. micrantha plants exhibiting  M. crucifer use scars 0, 1, and 2 years after weevil release in the  rangeland experiment??????????????...?????...29  Figure 2.2.   Frequency and intensity of M. crucifer colonization with varying plant  size in C. officinale and H. micrantha ?????.????????...30  Figure 2.3.   Frequency and intensity of M. crucifer colonization in H. micrantha in  relation to site-level colonization in C. officinale ???...?????...31  Figure 2.4.   Probabilities of C. officinale and H. micrantha being colonized by at  least one M. crucifer egg or larva when displaying above-ground use scars??????????????????????..????...32  Figure 3.1.   Schematic diagram of mark-release-recapture experiment....????.....52  Figure 3.2.   Within-patch spatial patterns of M. crucifer use scarring on  C. officinale and H. micrantha around release points in the rangeland  experiment??????????????????????..?.....53  Figure 3.3.   M. crucifer recaptures on C. officinale and H. micrantha situated 2 m,  10 m, and 100 m from release points over 8 days in the mark-release-recapture experiment??????????????????..?.....54  Figure 3.4.   M. crucifer recaptures on the H. micrantha clusters where they were  released in the mark-release-recapture experiment??????..?.......55  Figure 4.1.   Life cycle diagram and transition matrix used for C. officinale?..?.......90  Figure 4.2.   Life cycle diagram and transition matrix used for H. micrantha?.?.......91  Figure 4.3.   Lambda values for C. officinale and H. micrantha on M. crucifer release and non-release sites????????????..??..?.......92  Figure 4.4.   Contributions of life stage transitions to differences in C. officinale lambda values between non-release sites and M. crucifer release sites......93  Figure 4.5.   Contributions of life stage transitions to changes in C. officinale lambda values between Year 0-1 to Year 1-2?????????.........94  Figure 4.6.   Elasticity values for C. officinale life stage transitions.?.???..?.......95    viii Figure 4.7.   Survival of small C. officinale rosettes on M. crucifer release and non- release sites???????????????.??????..?.......96  Figure 4.8.   Survival of large C. officinale rosettes on M. crucifer release and non- release sites???????????????.??????..?.......97  Figure 4.9.   Sensitivity of C. officinale lambda values from non-release sites to a  reduction in rosette survival as observed on M. crucifer release sites??98  Figure 4.10. Contributions of life stage transitions to differences in H. micrantha lambda values between non-release sites and M. crucifer release sites......99  Figure 4.11. Contributions of life stage transitions to changes in H. micrantha lambda values between Year 0-1 to Year 1-2?????????.......100  Figure 4.12. Elasticity values for C. officinale life stage transitions.?.???..?.....101  Figure 4.13. The relationship between C. officinale fecundity and distance from  M. crucifer release points?????????...??????..?.....102  Figure 4.14. The relationship between survival and growth of small H. micrantha  bolters with distance from M. crucifer release points.?..???..?.....103  Figure 4.15. The relationship between small bolter H. micrantha fecundity with distance from M. crucifer release points.?..???..?..................104  Figure 4.16. Survival and growth of C. officinale rosettes with and without   M. crucifer use scars???????????..??????..?.....105  Figure 4.17. Fecundity of C. officinale plants with different M. crucifer scar ratings..106   Figure 4.18. Fecundity of H. micrantha plants with different M. crucifer scar  Ratings???????????????????????..?.....107  Figure B.1.  Seedling survival scenarios for C. officinale in the seedling emergence experiment???????????????...??????..?....144  Figure C.1.  Contributions of life stage transitions to changes in C. officinale lambda values between Year 0-1 to Year 1-2 on each site????.......158  Figure C.2.  Contributions of life stage transitions to changes in H. micrantha lambda values between Year 0-1 to Year 1-2 on each site????.......159     ix ACKNOWLEDGMENTS   From the moment I learned about this project, I knew it was a golden opportunity. My return to graduate school has been a long and winding road, and I feel extremely fortunate to have taken this journey. The joys of being able to indulge in curiosity and philosophy, to ask and answer scientific questions, and to spend my summers on knees and elbows counting plants, watching insects and smelling the soil, are matched only by the connections made with wonderful colleagues and friends along the way.   This project would not have been possible without the help, guidance, and moral support of a large group of people.  First and foremost, I have deep gratitude for my two supervisors, Dr. Rosemarie De Clerck-Floate and Dr. Bob Lalonde. Rose and Bob, your love of ecology inspires me, and your patient, wise, and generous mentorship and friendship have been a tremendous gift beyond my expectations.   I?d like to thank my supervisory committee members, Dr. Karen Hodges, Dr. Jason Pither, and Dr. Rebecca Tyson, who each provided valuable constructive guidance and encouragement in their own way. Thank you to Prof. Yvonne Buckley who hosted my research visit to the University of Queensland and provided essential statistical guidance and scientific discussion for the plant demographic analysis. Thank you to my external examiner, Dr. John Maron, for his thought-provoking questioning and fun scientific discussions during and after my thesis defence.   The science communities at UBC Okanagan, the Agriculture and Agri-Food Canada Lethbridge Research Centre, and the University of Queensland provided three stimulating environments in which to do this work. I?ve benefitted from many informal scientific discussions over the years with many people, including Dr. Kevin Floate, Dr. Brian Van Hezewijk, and Dr. Jennifer Firn, all of whom improved my science considerably. My labmates and fellow graduate students Morgan Whitehouse, Chandra Moffat, and Emily Barnewall offered helpful discussions and quality field assistance over the years. I thank my M.Sc. supervisor Dr. Bill Remphrey (University of Manitoba) for recruiting me into research and painstakingly laying my scientific foundation despite my protestations.   I am fortunate to have received extraordinary and essential technical help in this project.  Eva Pavlik was extremely instrumental in making this project happen, rearing weevils, growing plants, and tolerating my lab messes during the busiest times. Jordan Bannerman, William Van Der Weide, Shelby McLeod, and Karma Tiberg were all wonderful field research assistants, and I truly appreciate their dedication to this project and our thought-provoking discussions during long hours in the field. Each summer was an adventure with its own challenges out on the range, and was a lot more fun because of you.  Ray Wilson, Chelsey Durand, Jayden Dyck, Craig Anderson, Stephanie Erb, and Barb Lucente also provided important technical help and moral support along the way.     x A special thank-you goes to out Shane and Laurel Hansen, who graciously allowed me to use their rangeland for the main field experiment in this project. Even though they had to let some houndstongue infestations go uncontrolled for 4 years, they let me conduct this important work on their beautiful ranch in the Alberta foothills. Even during the busiest times, each day I would pause to appreciate the beauty of their land and of the mighty Chief Mountain.   I?d like to thank the people behind the funding and scholarships for this project, which came from (in alphabetical order): Agriculture and Agri-Food Canada Peer Review Program, BC Ministry of Forests, Lands and Natural Resource Operations, the Cattle Industry Development Council of BC, NSERC Alexander Graham Bell Canada Graduate Scholarship, NSERC Discovery Grant, UBC Okanagan, and the Wyoming Biocontrol Steering Committee.   Finally, I send profound thanks to my friends and family in Winnipeg, Vernon, Lethbridge, Brisbane, and Kelowna, who have supported me through thick and thin.  My karate families and table tennis and swimming friends and rivals were instrumental for camaraderie and study of mind and body. I particularly thank Ki Au, Allen Woloshyniuk, and Monika Gorzelak for sharing with me their transformative coaching and philosophies.  I thank my friends, particularly Leo King, Julie O?Donovan Mohite, Karen Mah, Brian Ohsowski, Jenny Janok and Greg Cavana for scientific discussions, social coordination and general shenanigans during this chapter of my life.  My community of longtime friends in Winnipeg has been a huge and invaluable source of support. To the Catton and Tomko families, there are no words, you have given me everything required to grow and enjoy the journey, I love you.       xi   DEDICATION    To my parents, Judy and Brand Catton, who have been the finest examples of  generosity, dedication, perseverance, and love.  1 CHAPTER 1  INTRODUCTION: BENEFITS, RISKS, AND CHALLENGES  REGARDING CLASSICAL BIOLOGICAL CONTROL OF WEEDS    1.1. INVASIVE PLANTS AND BIOLOGICAL CONTROL  Exotic species are one of the main threats to global biodiversity (Wilcove et al. 1998, Myers and Bazely 2003). Invasive plants are species of foreign origin that have been accidentally or deliberately introduced to a new area where their adaptability, prolific reproduction, or lack of specialist herbivores allow them to dominate ecosystems (Mack et al. 2000).  As such, invasive plants are a serious issue both environmentally and economically. Pimentel et al. (2005) estimated that the approximately 5,000 imported plant species naturalized in the United States of America (USA) contributed to $34 billion annually in losses and control costs, although this value is likely an underestimate due to the difficulty of quantifying ecological degradation.  On native rangeland, invasive plants are serious pests that may outcompete desirable forage species, be toxic to animals, and produce barbed seeds or stems that irritate animals and reduce market value of livestock. Annual losses and control costs from rangeland weeds in the USA are estimated at up to $6 billion (Pimentel et al. 2005).   Classical biological control (biocontrol) of weeds involves importing foreign, host-specific agents, usually insects, with the intent of reducing the fitness of the invasive plant in its introduced range (McFadyen 1998).  The aim is to create a stable insect-target plant interaction that suppresses target weed populations to low levels with acceptable ecological and economic impacts (McFadyen 1998, Syrett et al. 2000, Buckley et al. 2005).  When this goal is reached, the potential benefits of weed biocontrol are high; for example the benefit to cost ratio of weed biocontrol in Australia has been calculated at 23:1 (Page and Lacey 2006), and in some systems the ratio is estimated in the hundreds or even thousands (Culliney 2005). Canada has been active in  2 classical weed biocontrol for over five decades, with successes such as the suppression of St. John?s Wort (Hypericum perforatum L.), nodding thistle (Carduus nutans L.), knapweeds (Centaurea L. spp.), leafy spurge (Esula euphorbia L.), and Dalmatian toadflax (Lanaria dalmatica L.) (Harris 1993, Myers et al. 2009, Mason and Gillespie 2013).     1.2. NONTARGET USE  Biocontrol involves the irreversible release of an exotic organism into an ecosystem, which, as with any form of pest control, involves elements of both benefit and risk (Sheppard et al. 2003, Delfosse 2005). Because perfectly monophagous insect herbivores are rare (Fox and Morrow 1981, Odegaard et al. 2000), when expected net benefits of biocontrol are high, ?oligophagous? agents with known but relatively low propensities to use (i.e., for feeding, oviposition and larval development) close taxonomic relatives of their target weed (i.e., native nontarget plants) have been given regulatory approval for release (van Klinken and Edwards 2002, Louda et al. 2003, Sheppard et al. 2005). Therefore, risk in weed biocontrol occurs mainly in the potential of released agents to use nontarget species that are closely related to the target invasive (Pemberton 2000, Louda et al. 2003), and has been controversial (Louda et al. 2003, Hoddle 2004a, 2004b, Louda and Stiling 2004).  Approval of oligophagous agents is typically granted under the assumption that an insect?s reduced preference and performance on nontarget plants compared to target weeds in pre-release host specificity tests would minimize nontarget impact to acceptable levels post-release (Zw?lfer and Harris 1984, De Clerck-Floate and Schwarzl?nder 2002).  However, the ecological basis behind this assumption is not explicitly detailed. Recently, widespread, persistent and damaging nontarget use of rare native thistles by flower-feeding biocontrol weevils Rhinocyllus conicus Fr?lich, Larinus planus Fabricius, and the rosette weevil Trichosirocalus horridus Panzer have challenged this assumption (Louda 1998, Louda and O'Brien 2002, Takahashi et al. 2009, Havens et al. 2012). Analysis of other systems using oligophagous agents is important and necessary to determine whether these  3 documented cases of sustained nontarget use are isolated incidences, or represent a widespread trend in the field.   The regulatory trend toward an expectation of zero-risk in weed biocontrol introduction threatens the future of this valuable tool (Sheppard et al. 2003). More data on the ecology of target weeds, nontarget plants and potential agents is needed so that decision-makers can consider both benefits and risks of an agent?s potential release simultaneously and in balance (Kriticos 2003, Delfosse 2005, Raghu et al. 2007). Retrospective analyses on agents already released can be especially valuable in this pursuit (McEvoy and Coombs 1999, Louda et al. 2003). In some cases, nontarget use may be less harmful than the economic and ecological damage the targeted invasive plant would inflict if left uncontrolled, meaning the risk of releasing biocontrol agents must be considered in combination with the risk of ?doing nothing? and allowing the invasive plant to spread uninhibited (Thomas and Willis 1998, Delfosse 2005).   The level of risk posed by oligophagous agents is a function of the nature of their nontarget use.  ?Spillover? nontarget use, also known as ?associative susceptibility? or ?apparent competition?, occurs when insect density is high and agents move to less-preferred hosts (White and Whitham 2000). Spillover is a common pattern of nontarget use described in weed biocontrol research (Rand and Louda 2004, Dhileepan et al. 2006, Russell et al. 2007), and may occur as adult feeding, oviposition, larval feeding, or a combination of these behaviours.  Spillover is associated with high insect density, therefore the duration of this type of nontarget use in any one place is determined by the factors regulating local insect abundance. For effective agents, spillover can be temporary when local agent density drops quickly from dispersal due to a shortage of preferred hosts (Blossey et al. 2001). Alternatively, spillover duration can be longer when agents fail to decrease the local density of their host plant (Holt and Hochberg 2001).  For effective agents, spillover also has limited potential for causing landscape-level, nontarget population impacts, because nontarget use will be localized and temporary, meaning native plants away from high target plant (and thus insect) density will be unaffected (Cullen 1989).  In contrast, ?persistent use? occurs when biocontrol  4 insects successfully establish and sustain populations on nontarget plants in the absence of their target weed, and represents a new ecological relationship between the agent and nontarget host (Fowler et al. 2000). This type of use requires that a sufficient level of preference (measured as oviposition on a chosen host) and performance (measured here as complete larval development on a chosen host) occur on the nontarget plant to sustain viable insect populations. Persistent use of isolated nontarget patches is more ecologically significant than spillover because in that case nontarget use is not restricted by the distribution of the invasive plant (Cullen 1989). Moreover, isolated reproduction on nontarget plants presents the possibility of evolutionary changes occurring in host preference and performance (Dr?s and Mallet 2002, van Klinken and Edwards 2002, Hufbauer and Roderick 2005) that could lead to increased negative effects on nontarget plants.  Persistent use in weed biocontrol has rarely been documented, with the exception of thistle biocontrol weevils feeding on native thistles (Louda 1998, Louda and O'Brien 2002, Louda et al. 2005, Rand and Louda 2006, Takahashi et al. 2009).    1.3. CHALLENGES IN STUDYING IMPACT OF BIOCONTROL AGENTS  While much effort is justifiably expended in determining the host ranges of potential biocontrol agents through pre-release testing, the post-release impacts of agents on target and nontarget plants are rarely quantified, because of lack of funding and political and scientific motivation (McEvoy and Coombs 1999, Delfosse 2005). However, it is becoming more recognized that quantification of population-level impact of weed biocontrol agents post-release is important both for improving predictability of releases and for advocating for continued investment in the discipline (McEvoy and Coombs 1999, Blossey and Skinner 2000, Delfosse 2005, Carson et al. 2008, Morin et al. 2009).   Population-level effects from biocontrol agents on target and nontarget plants cannot occur without damage to individual plants (McClay and Balciunas 2005). Insect herbivory can result in reduced survival, growth and fecundity for individual plants, but these effects are highly dependent on multiple factors such the nature and extent of the  5 damage, the biology of the plant and herbivore, and the competitive dynamics and abiotic conditions of the growing environment (Trumble et al. 1993, Peterson 2000). Plants can react to and minimize insect herbivory damage through inducible chemical defenses (Howe and Jander 2008), and exhibit compensatory or even overcompensatory growth in response to injury (Trumble et al. 1993). In other cases, particularly under high ?outbreak? insect density, damage can reduce fecundity, growth and even survival rates for individuals (Romme et al. 1986, Strong et al. 1995, Carson and Root 2000, Coupe and Cahill 2003, Yang 2012).    Relative to what has been studied for individuals, herbivory effects on plant populations remain poorly understood and difficult to predict (Crawley 1989, Halpern and Underwood 2006, Maron and Crone 2006).  For example, depending on the frequency, severity, and distribution of herbivore damage to individual plants, population-level reductions in average survival, growth or fecundity (i.e., vital rates) may or may not occur.  Aggregated patterns of herbivory can leave ?refuge? plants that escape heavy damage and continue to survive, grow, and reproduce to maintain viable populations (Hawkins et al. 1993, Berryman et al. 2006, Johnson 2010). Even if vital rates are affected at the population level, their changes may not impact the overall population growth rate (Crawley 1989). Compensatory mechanisms in population dynamics, such as seed dormancy, density-dependence, and competitive release can buffer population growth rates from change (Garren and Strauss 2009, Swope and Parker 2010, Ortega et al. 2012).   Because of these and other factors, the impact of biocontrol agents on both target and nontarget plants is difficult to predict and to quantify (Morin et al. 2009).  The result is that many released and established weed biocontrol agents have failed to suppress their target weeds (McClay and Balciunas 2005), and uncertainty remains regarding risks of impact to nontarget hosts.  One way population vital rate data can be integrated to determine population growth rates is through the use of matrix population models (Caswell 2001, Crone et al. 2013). These models have been used for hundreds of plant species to study their population dynamics, often with a main objective of identifying influential transitions between life  6 stages that can be targeted for management purposes (Crone et al. 2011).  In weed biocontrol, matrix models have been used to inform selection of candidate biocontrol agents based on targeted plant life cycle disruption (McEvoy and Coombs 1999, Davis et al. 2006), to assess and predict population-level impact (or lack thereof) by agents (Shea and Kelly 1998, DeWalt 2006, Schutzenhofer and Knight 2007, Dauer et al. 2012), and to identify compensation mechanisms at the plant population level (Maines et al. 2013).  Conversely, matrix population studies of nontarget plants in weed biocontrol systems are rare, except for a few studies that focus on invasive thistle biocontrol in North America (Louda et al. 2005, Rose et al. 2005, but see Havens et al. 2012).    1.4. STUDY SYSTEM  Cynoglossum officinale L. (Boraginaceae), commonly known as houndstongue, is a biennial or short-lived, generally monocarpic, perennial plant native to Eurasia that has become invasive in North America. C. officinale grows in disturbed areas in western North America, including disturbed rangeland, roadsides and logged areas (Upadhyaya et al. 1988, De Clerck-Floate 2013).  Pyrrolizidine alkaloids in C. officinale kill livestock when ingested and animals grazing on infested rangeland can suffer dermatitis and fetch reduced prices at auction when coated in the plant?s adhesive burrs (Upadhyaya et al. 1988, Upadhyaya and Cranston 1991). Because C. officinale can quickly exploit disturbed habitats, the plant has patchy distributions and displays metapopulation dynamics in both its native and introduced ranges (van der Meijden et al. 1992, De Clerck-Floate 1996).  Mogulones crucifer Pallas [=Ceutorhynchus cruciger Herbst, Mogulones cruciger Herbst, Coleoptera: Curculionidae] is a European root-feeding weevil approved for release in Canada in 1997 to control C. officinale.  M. crucifer is a successful biocontrol insect in Canada, effectively and rapidly suppressing C. officinale patches and dispersing to surrounding infestations (De Clerck-Floate et al. 2005, De Clerck-Floate and Wikeem 2009). M. crucifer adults are highly mobile and can walk or fly to new host plants. The  7 weevil is adept at dispersing to new C. officinale infestations, and has found the target weed at least as far as 1.42 km from release points within 3 years (De Clerck-Floate et al. 2005).  M. crucifer is univoltine and can reach ?outbreak? population levels within two to three generations when not limited by availability of its target weed (Schwarzlaender 1997, De Clerck-Floate et al. 2005), a characteristic that may be important for achieving successful suppression of target organisms (Gassmann 1996). M. crucifer damages C. officinale mainly through larval root-feeding and reduces fecundity of bolting plants by 30-35% in their native range (Prins et al. 1992, Williams et al. 2010). In its native range, the weevil is in turn attacked by at least three specialist egg or larval parasitoids at rates estimated between 14-23% (Schwarzlaender 1997). These parasitoids are presumed to be absent in the introduced range.   Pre-release host-specificity tests revealed that the M. crucifer?s host range included several Boraginaceae species within different genera, but both its preference and performance were much stronger for C. officinale (Jordan et al. 1993, De Clerck-Floate et al. 1996, De Clerck-Floate and Schwarzl?nder 2002). Since its release, M. crucifer has sporadically fed and oviposited on several native nontarget Boraginaceae species in Canada (Andreas et al. 2008). However, any resulting individual or population-level impacts are unknown. Despite its success in Canada, M. crucifer was not approved for release in the USA because of the weevil?s potential impact on Endangered Boraginaceae in the USA, and in 2010 it was declared a federal pest (U. S. Department of Agriculture 2010).  The weevil is dispersing naturally from Canada to C. officinale infestations in the USA, and substantial M. crucifer populations in Washington and Idaho are spreading 10-15 km southward per year (M. Schwarzlaender, personal communication).  Of particular concern is the potential threat M. crucifer poses to Hackelia venusta (Piper H. St. John), a native Boraginaceae species reduced to a single population of approximately 300 plants on 16 hectares in Washington and federally listed as Endangered since 2002 (U. S. Fish and Wildlife Service 2007, 2011). More information is needed on assessing the threat that M. crucifer poses to H. venusta and specifically how to monitor and predict the weevil?s activity on and around the Endangered potential nontarget plant (U. S. Fish and Wildlife Service 2007, 2011).  8 The focal nontarget species in this study is Hackelia micrantha, commonly known as ?blue stickseed? or ?Jessica sticktight?. H. micrantha is a polycarpic perennial native to North America, occurring on mesic slopes, grasslands or shrublands in British Columbia, Alberta and the western USA, and has a global conservation status of ?secure? (NatureServe 2012). H. micrantha grows sympatrically with C. officinale in semi-forested rangeland and is similar to the invasive plant in life history, morphology and phenology. Like C. officinale, H. micrantha forms a taproot, reproduces only by seed, and produces burred nutlets. However, H. micrantha has a branched shoot architecture as a result of multiple woody caudices emerging from the taproot below ground. Plants in the genus Hackelia incurred some of the highest levels of use among nontarget species in previous M. crucifer studies (De Clerck-Floate and Schwarzl?nder 2002, Andreas et al. 2008), and H. micrantha is known to incur M. crucifer feeding and oviposition and support complete development of the insect (H. Catton and R. De Clerck-Floate, unpublished results). Information gained about M. crucifer nontarget use of H. micrantha may be relevant to conservation of the Endangered and perennial congener, H. venusta.   1.5. THESIS OVERVIEW  The ability to experimentally manipulate biocontrol agents once they have been approved for release presents a rich opportunity to study nontarget use patterns by oligophaous agents.  In this thesis, I put a ?high risk? agent in a ?high risk? situation (high insect density) with a nontarget species with ?high risk? of herbivory to test hypotheses regarding the patterns and implications of M. crucifer target and nontarget use. To create this situation, I performed a factorial field experiment where I released M. crucifer into naturally-occurring patches of a native nontarget plant growing interspersed or isolated from the target weed on native rangeland in the introduced range. I monitored visual indications of weevil host use for 2 years post-release, then destructively sampled and dissected a subset of plants from both species to test for the presence of M. crucifer. This approach is recognized as powerful for testing for persistent use and its potential  9 ecological effects (Louda 1998, De Clerck-Floate and Schwarzl?nder 2002, Briese 2005, Andreas et al. 2008), and to my knowledge has not been performed previously in the introduced range in weed biocontrol.    In Chapter 2, I test hypotheses regarding the nature of M. crucifer nontarget use (spillover or persistent), and combine the results with dissections from additional non-experimental sites to characterize and compare patterns of M. crucifer target and nontarget use in terms of temporal patterns and oviposition intensity. In Chapter 3, I examine in detail the spatial patterns of the transient nontarget use observed shortly after M. crucifer release.  I characterize host-specific spatial patterns of use at the within-patch scale, and combine the results with a mark-release-recapture experiment to propose a behavioural mechanism explaining the patterns observed.  In Chapter 4, I present demographic data and build matrix population models to describe and integrate the impact of target and nontarget herbivory from vital rates (i.e., survival, growth, and fecundity of each life stage)  to the population level. Finally, in Chapter 5, I combine the patterns of nontarget use documented in Chapters 2 and 3 with ecological refuge theory to explain the levels of target and nontarget population-level impacts detected in Chapter 4. I then integrate the information gained to hypothesize why M. crucifer has been a successful biocontrol agent against C. officinale in Canada, and why it is unlikely to cause population-level effects to H. micrantha. More generally, I provide predictions and management recommendations applicable to any biocontrol system using effective oligophagous agents, and demonstrate why nontarget use for similar biocontrol agents may be common, transient, and an indicator of high insect density and therefore forthcoming weed control.    10 CHAPTER 2  INTENSITY AND TEMPORAL PATTERNS OF  M. CRUCIFER HERBIVORY ON TARGET AND NONTARGET PLANTS   2.1. LITERATURE REVIEW AND OBJECTIVES  Biological control (biocontrol) of introduced weeds using foreign, host-specific insects (i.e., the ?classical? approach) is a valuable tool for mitigating the environmental impact of invasive plants, but the risk of nontarget use (i.e., feeding, oviposition and larval development) on native nontarget plant species has made this practice controversial (Louda et al. 2003, Hoddle 2004a, 2004b, Louda and Stiling 2004).  When expected net benefits of biocontrol are high, ?oligophagous? agents with known but relatively low propensities to use closely-related, native nontarget plants have been given regulatory approval for release (van Klinken and Edwards 2002, Louda et al. 2003, Sheppard et al. 2005). Approval is typically granted under the assumption that an insect?s reduced preference and performance on a nontarget plant minimizes nontarget impact to acceptable levels (Zw?lfer and Harris 1984, De Clerck-Floate and Schwarzl?nder 2002).  However, the extent of nontarget use by thistle biocontrol weevils Rhinocyllus conicus Fr?lich, Larinus planus Fabricius, and Trichosirocalus horridus Panzer has challenged this assumption, because in these systems insect use of nontarget species was widespread, long-term and potentially damaging to native plant populations (Louda 1998, Louda and O'Brien 2002, Takahashi et al. 2009, Havens et al. 2012). Investigating the potential for oligophagous agents to cause negative population-level effects to nontarget plants is therefore a valuable exercise, and may be especially relevant when rare plants are potential nontarget hosts (Ancheta and Heard 2011).    Damage to individual plants in the form of reduced survival, growth and fecundity can translate to population-level effects (i.e., a change in population growth rate), but this effect does not necessarily occur (Crawley 1989). Compensatory mechanisms at the  11 population level such as seed bank dynamics, density-dependence, and competitive release can buffer populations from the effects of stresses (Garren and Strauss 2009, Swope and Parker 2010, Ortega et al. 2012). Additionally, non-uniform patterns of herbivory can leave ?refuge? plants that escape heavy damage and continue to survive, grow, and reproduce, thus maintaining viable populations (Hawkins et al. 1993, Berryman et al. 2006, Johnson 2010).  For example, non-uniformity of use that is spatially or temporally correlated creates areas or times with no or minimal herbivory (Gutierrez et al. 2008). And even when use is spatially or temporally uniform, a large proportion of plants can escape (heavy) damage when herbivory is rare or shows great variance in intensity per plant. For example, ?probabilistic? refuges can be created when  oviposition is aggregated, so that the bulk of eggs occur in a small proportion of used plants.  Therefore, independently of spatial or temporal patterns, such clumped oviposition can stabilize herbivore-plant dynamics through probabilistic refuge creation (Myers 1976, Myers et al. 1981, Louda et al. 1997, 2005, Rose et al. 2005, Heard and Remer 2008, Stephens and Myers 2012). Because of these and other factors, the impact of biocontrol agents on target and nontarget plants has been difficult to predict.  The result is that many released and established weed biocontrol agents have failed to suppress their target weeds (McClay and Balciunas 2005), and substantial uncertainty remains regarding risks of impact to nontarget hosts.     While the concept of refuges for nontarget plants is intuitive and directly relevant to weed biocontrol, it has not been discussed explicitly in the literature of the discipline.  Most studies describe nontarget use that is mild (Dudley and Kazmer 2005, Andreas et al. 2008, Paynter et al. 2008, Diaz et al. 2009, Moran et al. 2009), transient (Dhileepan et al. 2006, Pratt et al. 2009), or localized (Schooler et al. 2003, Russell et al. 2007, Taylor et al. 2007, Paynter et al. 2008, Pratt et al. 2009), but do not go beyond basic description to either demonstrate or predict the effects of these use patterns on plant populations.  The only studies to focus on nontarget plants at the population level, to my knowledge, are demographic models of the impact of Rhinocyllus conicus Fr?lich and Larinus planus Fabricius, flower-head biocontrol weevils used in North America to control invasive thistles (Louda et al. 2005, Rose et al. 2005, Havens et al. 2012). Rose et al.  12 (2005) reveal the importance of variation in the number of eggs laid per plant in reducing herbivory effects on nontarget plants (i.e., egg clumping leading to probabilistic refuges).  However, the existing nontarget models assume uniform patterns of use in space and time, and overlooking the potential effects of spatial and temporal refuges can overestimate population-level impact.   Bridging the gap between how use of individual nontarget plants may or may not translate to population-level effects is instrumental for developing reliable monitoring methods, accurately predicting the scale and severity of potential impact of released agents, and assessing pre-release predictions to inform future biocontrol agent release decisions (Louda 1998, Hoddle 2004b).  The formation of refuges, and thus the risk of population-level impact to nontarget plants from oligophagous agents, is a function of the nature of nontarget use.  When two hosts co-occur within a patch, ?spillover?, also known as ?associative susceptibility? or ?apparent competition?, occurs when insect density is high and agents move to less-preferred hosts (White and Whitham 2000). Spillover is a common pattern of nontarget use described in weed biocontrol research (Rand and Louda 2004, Dhileepan et al. 2006, Russell et al. 2007), and may occur as adult feeding, oviposition, larval feeding, or a combination of these behaviours.  Spillover is directly linked with high insect density, therefore the duration of this type of nontarget use is determined by the factors regulating insect density. For example, spillover can be temporary when local agent density drops quickly from dispersal due to insect density-dependent processes or a shortage of preferred hosts (Blossey et al. 2001). Density near nontargets when target plants are locally absent may depend on the ability of the less-preferred hosts to attract and arrest the insects (Chapter 3). Alternatively, spillover can be longer in duration when agents fail to decrease the local density of their host plant (Holt and Hochberg 2001).  For effective agents, spillover has limited potential for causing landscape-level, nontarget population impacts because spatial and temporal refuges from herbivory are left for native plants away from high target plant (and thus insect) density (Cullen 1989).    In contrast to spillover, persistent use occurs when biocontrol insects successfully establish and sustain populations on nontarget plants in the absence of their target weed,  13 and represents a new ?ecological relationship? between the agent and nontarget host (Fowler et al. 2000). This type of use requires that sufficient preference (oviposition on a chosen host) and performance (complete development on a chosen host) occur on the nontarget plant to sustain viable insect populations. Persistent use of isolated nontarget patches is more ecologically significant than spillover because it does not necessarily leave spatial refuges from insect activity and therefore may generate population-level impacts on nontarget species (Cullen 1989). Moreover, isolated reproduction on nontarget plants enables evolutionary changes in host preference and performance (Dr?s and Mallet 2002, van Klinken and Edwards 2002, Hufbauer and Roderick 2005) that could lead to increased negative nontarget effects.  Persistent use in weed biocontrol has rarely been documented, with the exception of thistle biocontrol weevils feeding on native thistles (Louda 1998, Louda and O'Brien 2002, Louda et al. 2005, Rand and Louda 2006, Takahashi et al. 2009). However, the occurrence of persistent use in the introduced range has not been thoroughly monitored, or specifically tested for, in any weed biocontrol systems.   Mogulones crucifer Pallas [=Ceutorhynchus cruciger Herbst, Mogulones cruciger Herbst, Coleoptera: Curculionidae] is a European root-feeding weevil approved for release in Canada in 1997 to control houndstongue, Cynoglossum officinale L. (Boraginaceae), an invasive rangeland weed.  M. crucifer is a successful biocontrol insect in Canada, effectively and rapidly suppressing C. officinale patches and dispersing to surrounding infestations (De Clerck-Floate et al. 2005). Pre-release host-specificity tests revealed that the weevil?s host range included several Boraginaceae species, but both its preference and performance were much stronger for C. officinale (Jordan et al. 1993, De Clerck-Floate et al. 1996). M. crucifer has sporadically fed and oviposited on several native nontarget Boraginaceae species in the field in Canada (Andreas et al. 2008), although any resulting individual or population-level impacts are unknown. Despite its success in Canada, M. crucifer was not approved for release in the United States of America (USA) because of the weevil?s potential impact on Endangered Boraginaceae in the USA, and in 2010 it was declared a federal pest (U. S. Department of Agriculture 2010).  The weevil is dispersing naturally from Canada to C. officinale  14 infestations in the USA, and substantial M. crucifer populations in Washington and Idaho are spreading 10-15 km southward per year (M. Schwarzlaender, personal communication).  Of particular concern is the potential threat M. crucifer poses to Hackelia venusta (Piper H. St. John), a native Boraginaceae species reduced to a single population of approximately 300 plants on 16 hectares in Washington and federally listed as Endangered since 2002 (U. S. Fish and Wildlife Service 2007, 2011). More information is needed on assessing the threat that M. crucifer poses to H. venusta and specifically how to monitor and predict the weevil?s activity on and around the Endangered potential nontarget plant (U. S. Fish and Wildlife Service 2007, 2011).  The ability to experimentally manipulate biocontrol agents once they have been approved for release presents a rich opportunity to study nontarget use patterns by oligophaous agents. Given that preference and performance by M. crucifer on all nontarget species studied to date is lower than for C. officinale (Jordan et al. 1993, De Clerck-Floate et al. 1996, De Clerck-Floate and Schwarzl?nder 2002, Andreas et al. 2008), I was interested in testing the hypotheses that: 1) M. crucifer post-release maintains higher preference for and performance on C. officinale than on a nontarget plant, and 2) Nontarget use occurs as spillover and not as persistent use. For this study, I released M. crucifer into naturally-occurring patches of the native nontarget plant Hackelia micrantha (Eastw.) J.L. growing interspersed or isolated from C. officinale. This approach is recognized as powerful for testing for persistent use and its ecological effects (Louda 1998, De Clerck-Floate and Schwarzl?nder 2002, Briese 2005, Andreas et al. 2008), and to my knowledge has not been performed previously in the introduced range in weed biocontrol.  I monitored visual indications of weevil host use for 2 years post-release. I then destructively sampled and dissected plants on my experimental release sites and an additional 7 non-experimental release sites (0 to 4 years post-release) to test for the presence of M. crucifer at release sites and quantify M. crucifer eggs and larvae in each species.      15 2.2. METHODS  2.2.1.  STUDY SYSTEM Houndstongue, Cynoglossum officinale, is a biennial or short-lived, generally monocarpic, perennial plant native to Eurasia that has become invasive in North America.  Mogulones crucifer damages C. officinale mainly through larval root-feeding (Prins et al. 1992), and can reach high population numbers within several generations when not limited by C. officinale availability (Schwarzlaender 1997). Overwintered adults emerge in spring, feed on C. officinale foliage, and mate. Females lay eggs singly into petioles, bolting stems, or the root crown (lifetime average= 190 eggs, Schwarzlaender, 1997). Larvae mine the root crown and roots, consuming tissues until they exit as third instar larvae to pupate in the soil. Emerging adults feed and oviposit in late summer prior to overwintering in soil or leaf litter. Eggs, larvae and pupae may also overwinter. The weevil is univoltine and most adults perish within 12 months (Schwarzlaender 1997). M. crucifer adults are highly mobile and can walk or fly to new host plants. Weevil populations can disperse to C. officinale at least as far as 1.42 km from their release within 3 years (De Clerck-Floate et al. 2005).   Blue stickseed, Hackelia micrantha, is a polycarpic perennial native to North America that grows sympatrically with C. officinale. It is similar to the invasive plant and to H. venusta in life history, morphology and phenology. H. micrantha supports full development of M. crucifer (H. Catton and R. De Clerck-Floate, unpublished data). H. micrantha?s global conservation status is listed as secure (NatureServe 2012).  2.2.2. RANGELAND EXPERIMENT (VISUAL USE ASSESSMENTS AND DESTRUCTIVE SAMPLING) I conducted a field experiment on rangeland in the Foothills Fescue Natural Region (Natural Regions Committee 2006),  in southern Alberta, Canada from 2009-2011 to monitor M. crucifer use of target and nontarget plants under outbreak weevil densities. On June 4, 2009, batches of 300 M. crucifer were released on six patches of grazed rangeland under aspen, Populus tremuloides Michx., canopy. Patches contained naturally-occurring populations of 34-84 vegetative or reproductive H. micrantha plants  16 growing either (i) on ?target common? sites (n=3), where plants were interspersed with 30-97 C. officinale plants within a 13 m radius, or (ii) on ?target rare? sites (n=3), where plants were interspersed with sporadic C. officinale plants, which were generally small rosettes and were removed in July 2009.  Patches were at least 360 m from other M. crucifer releases and showed no signs of weevil activity prior to my releases. Weevils used for the releases were collected from an outdoor rearing plot near Lethbridge, Alberta, and had a female:male ratio of 2.23:1 based on a pooled subsample of 300 weevils. Weevils were released on the ground at a point approximately equidistant from potential host plants so that insects were forced to find hosts of either species. The release point was marked and used as a reference point for subsequent surveys of M. crucifer activity.   Around each field release point, individual H. micrantha and C. officinale plants (when present) were identified and marked within permanent 1x1 m quadrats. Quadrats were placed around plants nearest the release point until 30 rosettes and 20 bolting plants of each species (when possible) were sampled, to a radius of 4-13 m, depending on plant density. Plants were individually examined in July 2009 (i.e., 4-7 weeks after release), and again in July 2010 and July 2011 for visual indications of M. crucifer adult feeding and oviposition.  M. crucifer adult activity on C. officinale forms distinctive scars, appearing in the form of oval-shaped holes or raised bumps on the petioles (De Clerck-Floate et al. 2005). Weevil damage appeared similar on H. micrantha and also occurred on bolting stems of both species.    A total of 55 C. officinale plants found on ?target rare? sites were removed in July 2009 at the time of Year 0 visual damage assessments. These plants were often small and in difficult to access locations (i.e., under dense willow Salix spp. brush), and all but 10 were outside of my plant evaluation radii. A further 4 and 16 stray C. officinale plants were removed upon discovery from these sites in 2010 and 2011 and were dissected under stereoscopic microscopy in search of M. crucifer eggs and larvae.    17 To test for M. crucifer presence and evaluate colonization patterns on the sites 2 years after release, I destructively sampled 7-11 H. micrantha from all 6 sites, and 10-16 C. officinale plants from the three ?target common? sites. Plants were harvested on July 16-21, 2011, and subsequently dissected in search of M. crucifer eggs and larvae.   2.2.3. NON-EXPERIMENTAL RELEASE SITES To investigate the effect of years since release on M. crucifer use patterns and increase the sample sizes of colonized plants, in 2011 I collected additional data from non-experimental sites containing both plant species where M. crucifer releases were made 0, 3, or 4 years prior. I destructively sampled 10-18 H. micrantha and 10-16 C. officinale from seven such sites. Plants were not assessed for insect activity prior to sampling.  Sampled plants were subsequently evaluated for visual signs of M. crucifer use, and after measuring root crown diameter (RCD) with calipers, entire roots and shoots were dissected under stereoscopic microscopy to locate M. crucifer eggs and larvae.  For Year 0 plants, shoot tissues further than 10 cm above the root were not dissected as both my experience and Schwarzlaender (1997) indicated that the majority of aboveground eggs and larvae are located near the root.    Non-experimental release sites were sampled as follows. Year 0, n=3 sites: I released 300 lab-reared M. crucifer on each of these sites on June 22, 2011. On July 24, 2011, I harvested 10 plants each of C. officinale and H. micrantha closest to the release points.   Year 3, n=1 site: A release of 100 weevils was made on this site in May 2008 as part of a regional biocontrol program. On June 22, 2011, I harvested the 10 closest C. officinale plants to the release point and 10 H. micrantha within the same range of distances. Year 4, n=3 sites: Releases of 100 weevils per site were made in May 2007 as part of a regional biocontrol program. On June 23, 2011, these sites had many more H. micrantha than C. officinale, therefore I harvested the closest 15-16 target plants to the release points and their nearest H. micrantha neighbour. Paired plants ranged from 0.05 to 4 m apart.  Occasionally, C. officinale plants did not have paired H. micrantha plants.      18 2.2.4. DATA ANALYSIS Data were analyzed using JMP version 10.0.1.1 (SAS Institute Inc., Cary, NC, USA, 2012)  and R version 2.15.2 (R Core Team 2012). Patterns of M. crucifer damage observed on experimental release sites from 2009-2011 were analyzed using generalized linear mixed-effects models (GLMM, ?glmer? function in package ?lme4? version 0.999999-0) with presence or absence of M. crucifer use scars as a binary response variable, and thus binomial error distributions and a logit link function. Fixed effects were species-site category (C. officinale, H. micrantha on ?target common? sites, and H. micrantha on ?target rare? sites) and years since release. Random effects were nested as site/quadrat/plant to account for spatial hierarchy and repeated measures. In all model selection analyses, full models including interactions were generated, and non-significant interactions and explanatory variables were dropped one by one, until the most reduced model was not significantly different than the full model (using the ?2 test in the ?anova? function in R). These ?minimum adequate models? were compared to null models with identical random effects to confirm significant explanatory power. Model goodness of fit is expressed as the percentage of deviance explained (DE) by the fixed effects (i.e., after consideration of random effects), and was calculated as 100 ? (deviance in null model with random effects ? deviance in minimum adequate model with identical random effects)/(deviance in null model with random effects). For models with categorical explanatory variables, least squared means and 95% confidence intervals were calculated in the R package ?lsmeans? version 1.06-05. The standard level of significance used was ?=0.05, but when posthoc multiple comparisons were performed, P-values were adjusted using the ?Tukey? method in lsmeans.   Frequency of M. crucifer use in dissected plants was compared using GLMM with a  presence/absence of colonization as a binary response variable (with binomial error distribution and logit link function).  Fixed effects included species, ln-transformed root crown diameter (a measure of plant size), and years since release. Random effects were nested as site/plant pair number.  Colonized plants were further analyzed to describe intensity of colonization with GLMM.  The model had the number of eggs and larvae in colonized plants as the response variable (with Poisson error distribution and log link  19 function), and fixed effects of species, ln-transformed root crown diameter, and years since release. Site was included as a random effect. Plant pair number was excluded as a random effect because of low sample size.   The relationship between target and nontarget oviposition was examined with similar GLMM as above, using response variables for nontarget presence/absence of colonization (binomial errors, logit link function) and number of eggs + larvae per plant (Poisson errors, log link function). Within-site colonization of C. officinale was included as a fixed effect and was calculated as the back-transformed mean of ln (number of eggs and larvae in C. officinale+1). Release site was included as a random effect.    The pooled distributions of larval instar stages in C. officinale and H. micrantha were compared using Fisher?s exact tests. The correspondence between the presence of feeding scars and colonization was evaluated with generalized linear models (GLM) with a binary response variable of colonized/uncolonized (binomial errors and logit link function) and fixed effects of species and presence of M. crucifer scars. This model included all dissected plants from site-years where M. crucifer was confirmed to be present, including C. officinale removed from ?target rare? sites in the year of release.    2.3. RESULTS    2.3.1. VISUAL INDICATIONS OF USE Scars characteristic of M. crucifer use were observed on both plant species and on both site types (Fig. 2.1).  In Year 0, target and nontarget plants had similar chances of being used (52% for targets, 35-40% for nontargets, Z<-1.31, P>0.53). Probability of use of both species declined with time, but particularly for nontargets, as C. officinale was 25-30% likely to be used after 1 and 2 years, compared to <4% for H. micrantha on either ?target common? or ?target rare? sites.  The seven used H. micrantha plants on ?target rare? sites in Year 2 had only a single scar per plant, and the three dissected plants did not contain eggs or larvae.  20 Two years after release, M. crucifer had established in C. officinale on all three ?target common? sites, and eggs or larvae were found in 72% (26 of 36) of sampled target plants (mean?S.E.= 5.1?1.2 juveniles per sampled plant). However, no eggs or larvae were found in any H. micrantha sampled from ?target common? (n=29) or ?target rare? (n=29) sites. Moreover, none of the stray C. officinale plants (n=4 in 2010, n=16 in 2011) found within 19 m of the ?target rare? release points exhibited M. crucifer use scars or contained eggs or larvae. These C. officinale essentially acted as trap plants, and their lack of use further suggests that the weevils did not persist on the ?target rare? sites.   2.3.2. DESTRUCTIVE SAMPLING AND DISSECTIONS Of the plants dissected from release sites with both species (three ?target common? sites and seven non-experimental sites), 83% (102 of 123) C. officinale and 17% (19 of 112) H. micrantha contained at least one M. crucifer egg or larva (?2=124.1, P<0.0001). Colonization intensity ranged from 0-116 eggs+larvae for C. officinale and 1-10 eggs+larvae for H. micrantha. Plant species and root crown diameter were significant explanatory variables in the minimum adequate GLMM for probability of colonization (Fig. 2.2A). Even very small C. officinale had a >50% chance of colonization, and those with root crown diameters >10 mm were nearly all colonized. While probability of H. micrantha colonization also increased with plant size, the largest nontargets still had a probability of colonization of less than 50%, even though colonized nontargets had larger root crown diameter than colonized targets (ANOVA, F1,115=11.4, P=0.0010). Years since release was not a significant explanatory variable for probability of colonization for either species.    The minimum adequate model for number of eggs and larvae in colonized plants included significant explanatory variables of plant species, root crown diameter, years since release, and an interaction between species and years since release. Intensity of colonization in H. micrantha (but not in C. officinale) increased mildly with time; however, years since release added only 1% to the goodness of fit of the model. A model without years since release is displayed in Fig. 2.2B. The number of eggs and larvae per  21 colonized plant was significantly higher in C. officinale and significantly more positively related to root crown diameter than in H. micrantha.    When all 10 mixed species sites were considered together, 38.8 M. crucifer juveniles were found in C. officinale for every one in H. micrantha (Table 2.1). The relative abundance of larval instars trended toward more mature larvae in C. officinale, although this difference was not statistically significant (Fisher?s exact test, P=0.12). However, the low number of larvae in H. micrantha reduced the statistical power of the test; for example, there were 204 second or third instar larvae found in C. officinale plants, but only one in H. micrantha. Oviposition was aggregated in both target and nontarget plants, with half of all eggs and larvae occurring in 16-17% of used plants for each species. The distance of the 10 closest C. officinale plants to release points was significantly farther on sites after 4 years compared to 0 years after release (0 years: 4.5?3.5 m, 4 years: 15.9?8.7 m, n=60, F1,58=44.4, P<0.0001), suggesting C. officinale close to release points disappeared within 4 years.   Site-level frequency and intensity of H. micrantha colonization increased with colonization intensity in C. officinale (Fig. 2.3).  Nontargets from sites with the heaviest target use were more than 4 times as likely to contain an egg or larva than those with the lowest target use (?2=10.0, P=0.0015, Fig. 2.3A).  The number of eggs and larvae in nontargets also increased with site-level target oviposition (?2=6.29, P=0.0121, Fig. 2.3B), but the trend appeared to be driven by the high frequency of single-egg deposits on the two sites with the highest C. officinale use.  When only used nontargets were analyzed, site-level C. officinale oviposition was not a significant explanatory variable for the number of eggs and larvae in H. micrantha (?2=0.15, P=0.69).   The reliability of M. crucifer use scars as indicators of colonization differed significantly between C. officinale and H. micrantha (Fig. 2.4).  Target plants with above-ground scars were 3 times more likely to contain at least one M. crucifer juvenile than were scarred nontargets (Z ratio = -8.69, P<0.0001). Target plants with no visible scars were still colonized in 40% of cases, while only 3% of unscarred nontargets were colonized (Z  22 ratio = -8.69, P<0.0001). Scarred H. micrantha had a similar probability of colonization as unscarred C. officinale (Z ratio = 0.99, P=0.75).     2.4. DISCUSSION  The data from this study support both of my hypotheses regarding the occurrence and implications of higher preference and performance by M. crucifer on target (C. officinale) than on nontarget (H. micrantha) hosts in the field. Nontarget use occurred on all release sites, but was significantly less common, lower in intensity, shorter in duration, and less responsive to plant size compared to use of the target weed. Larvae in H. micrantha trended toward being less mature than those in C. officinale, even though both plant species were sampled at the same time. Several lines of evidence suggest that nontarget use by M. crucifer occurs as spillover and is not persistent; 1) site-level target and nontarget oviposition were positively correlated, 2) use scars on H. micrantha on experimental sites were independent of target weed density and limited almost exclusively to the year of release (i.e., when weevil density was standardized and artificially high), 3) oviposition in H. micrantha continued at least 4 years after release on non-experimental sites where both plant species were present, and 4) M. crucifer establishment was not detected on H. micrantha two years after release when C. officinale was absent from the patch.  Identifying the mechanism of nontarget use by M. crucifer as spillover as opposed to persistent use means that circumstances, patterns, and potential population-level impacts of nontarget use can be predictable.   Since spillover requires high insect density, determining when and where agents occur in large numbers facilitates prediction of patterns of nontarget host use, and thus refuges from use in the field.  As is typical for specialist insects, local density of M. crucifer is likely highly responsive to the abundance of its target weed, as C. officinale is known to be both a strong source of new weevils (Schwarzlaender 1997, De Clerck-Floate and Schwarzl?nder 2002) and an attractive sink to dispersing weevils (De Clerck-Floate et al. 2005, Chapter 3). High insect density may also occur temporarily and independently  23 of target weed density during transient phases when insects are not in equilibrium with host plants (Holt and Hochberg 2001), such as shortly after biocontrol releases, following rapid local extirpation of the target weed, or when insects are dispersing in high numbers.  In these ?worst-case? scenarios, temporal refuges may be especially strong, as high M. crucifer density appears to be unsustainable on H. micrantha alone. In contrast, in areas with both plant species, the temporal refuge is governed by how quickly a biocontrol agent suppresses its target weed, leading to a decrease of local agent density (Holt and Hochberg 2001, Russell et al. 2007).  M. crucifer severely depleted C. officinale patches within 2 years in British Columbia (De Clerck-Floate and Wikeem 2009), analogous to a weevil ?fire? quickly burning through and moving to new fuel.  On my non-experimental sites, although C. officinale was still present 4 years after release, target plants on those sites were 3 times farther away from release points than on sites where weevils were released within the year, suggesting weed control immediately around the release points.  The fact that H. micrantha remained abundant near release points on these sites also suggests that the nontarget plant does not suffer population-level impacts from M. crucifer, although demographic information is necessary to properly test for any such effect (Chapter 4).   The spillover nature of M. crucifer nontarget use also means that spatial refuges should form for nontarget plants away from any situation of high weevil density. The extent of these refuges may be a function of the insect?s reduced host-finding and arrestment behaviours for nontargets (Chapter 3), and the degree of non-overlap in plant distributions (Wiggins et al. 2010).  Detailed information on the landscape-level sympatry of H. micrantha and C. officinale is not available, but I observed patches of the nontarget plant in isolation from C. officinale for hundreds of metres in my study area. The level of spatial isolation required for refuge formation is a logical next question in exploring this phenomenon, and is the subject of Chapter 3 in this dissertation.   Finally, in spaces and times of the heaviest spillover use observed, oviposition in H. micrantha was so rare and clumped that population-level effects resulting from this nontarget use are highly doubtful.  This probabilistic refuge may be the result of several  24 factors. First, M. crucifer appears to have specialized host-finding and arrestment (i.e., tendency to remain on a host once it is encountered) behaviours for C. officinale but not for H. micrantha, meaning nontarget plants even several metres away from insects may not be used (Chapter 3). Second, weevils that do encounter and feed on nontarget plants typically depart without ovipositing, as demonstrated by the reduced likelihood of colonization in scarred H. micrantha compared to scarred C. officinale.  Specialist beetles are known to ?test-bite? potential host plants before feeding further or ovipositing (Chapman and Bernays 1989, Withers 1999, Schoonhoven et al. 2005), and nontargets may elicit sufficient stimulation in insects for feeding, but not oviposition (Ganga Visalakshy et al. 2008, Smith 2012). Therefore, these data help to identify the specific behaviours that may cause the differing interactions between M. crucifer and its two host plants in the field.   The differing relationship between scars and oviposition in target and nontarget plants is directly relevant to monitoring herbivory by ?cryptic? (i.e., endophagous) biocontrol agents, a task that is difficult to achieve non-destructively (Hunter 2001). Because visible above-ground M. crucifer use scars reliably represented oviposition in C. officinale, they are suitable as a positive indicator of weevil reproductive activity on their target weed. However, giving equal weight to scars on target and nontarget plants during monitoring would overestimate oviposition in nontargets, producing a value 3 times higher than the true level. In contrast, the absence of scars reliably indicated a lack of oviposition in nontargets, but not in target plants. Therefore, the presence of M. crucifer use scars on H. micrantha should be treated with scepticism, but the absence of scars on the nontarget plant is a dependable indicator of no oviposition.  M. crucifer is recorded as accepting North American host species from at least seven Boraginaceae genera (Jordan et al. 1993, De Clerck-Floate and Schwarzl?nder 2002, Andreas et al. 2008), with the highest preference among nontargets generally for Hackelia species. Therefore, for nontarget plants outside of this genus, I expect scars to be even less indicative of oviposition. To estimate oviposition accurately in nontargets, monitoring programs of endophagous insects should destructively sample and dissect nontarget  25 plants (when possible) to develop species-specific correlations between visual indication of insect feeding and oviposition.   The lack of observed M. crucifer population establishment on isolated nontarget patches suggests that the risks of evolutionary changes in host preference in the ecological time scale are low for this insect. In addition to the reduced oviposition preference for nontargets described above, larvae may have slowed development in H. micrantha. This reduced larval performance may occur from limited resource quality and quantity within nontarget hosts, as H. micrantha has a hollow and noticeably woodier taproot than C. officinale (R. De Clerck-Floate and H. Catton, unpublished data).  The resulting low number of F1 weevils emerging from nontarget plants reduces the chance of M. crucifer population establishment in isolated nontarget patches, due to demographic stochasticity and Allee effects (Grevstad 1999). If however, the weevil can persist solely on nontargets, several conditions are required for new host race formation. First, sufficient genetic variation in host preference and performance is necessary for selection to act upon (van Klinken and Edwards 2002). The aggregated pattern of nontarget oviposition observed in this study suggests that most eggs may be laid by a small number of females, suggesting genetic variation in host specificity in M. crucifer could exist. Second, increased preference and performance on nontargets must be heritable and may need to be linked (Gripenberg et al. 2010).  Third, insect populations on nontargets may need to be reproductively isolated from insects using target plants (Dr?s and Mallet 2002).  While the formation of new host races appears unlikely, further population-level, behavioural, spatial and genetic study of the weevil?s nontarget preference and performance are necessary to determine the likelihood and time required for such an evolutionary change to occur.    The results of this study generate several management recommendations that are applicable to any weed biocontrol system with the distinctive characteristics of the M. crucifer system, i.e., a highly effective and mobile oligophagous agent that displays limited spillover nontarget use with reduced preference and performance for a nontarget plant. Here I aim to facilitate information-based responses by managers where nontarget  26 use is discovered or anticipated, and reduce the risk of impact to nontarget species, particularly when nontarget use causes reduced survival or reproduction in individual plants.  1) Create spatial refuges to prevent spillover: Localized populations of special interest may be protected from spillover nontarget use by removing nearby target weeds to create spatial refuges from insect activity, as has been generally suggested by Harris (1988) and Ancheta and Heard (2011).   To my knowledge, this approach has not been actually implemented in a field situation, but it is an option for management of Hackelia venusta, the Endangered plant in Washington reduced to a single population and growing in close proximity to C. officinale (U.S. Fish and Wildlife Service 2007, Vance 2013). should M. crucifer be discovered near its remaining population. Given the reduced nontarget host-finding tendencies of M. crucifer, it is unlikely that dispersing insects would find and use the H. venusta patch, and the inability of M. crucifer to maintain populations on H. micrantha suggests that a strong temporal refuge would also form. 2) Expect refuges even where spillover occurs: If weed removal is not feasible and spillover occurs, population-level effects may still be buffered by temporal and probabilistic refuges. 3) Conduct evidence-based monitoring: In cryptic agents that are not easily monitored, superficially similar sign of use may represent different oviposition rates on target and nontarget plants. In these cases, species-specific monitoring indicators through destructive sampling may be necessary to avoid overestimating the frequency of nontarget oviposition.   To my knowledge, this is the first study in weed biocontrol to explore the ecological heterogeneity of target and nontarget use by a released agent and to explicitly detail how the herbivory patterns are likely to be inconsequential to nontarget populations. Logical next steps are to explore herbivory patterns of transient spillover at different spatial scales, and to specifically test for population-level effects using plant demographic data in areas with and without biocontrol insects; these are the topics for Chapters 3 and 4 in this dissertation. The current study is also the first to deliver specific data-based management recommendations to aid in preventing and mitigating nontarget herbivory through the removal of nearby target plants, and to develop non-destructive monitoring methods for deducing the extent of nontarget spillover by an endophagous agent.  The  27 resulting recommendations are relevant to all nontarget Boraginaceae, but perhaps particularly so to the Endangered H. venusta. Little can be done to prevent the ongoing movement of M. crucifer along C. officinale infestations in the USA, however M. crucifer herbivory on the rare and localized H. venusta can be minimized and more accurately monitored as a result of this study?s characterization of the nature of nontarget use by M. crucifer in the surrogate species H. micrantha. The need for this type of information is specifically expressed in the recovery plan for H. venusta (U. S. Fish and Wildlife Service 2007, 2011).     28  Table 2.1. Numbers of M. crucifer eggs and larvae found in C. officinale and H. micrantha plants harvested in June-July 2011 from sites with known weevil populations, 0-4 years after release (n=10 sites). Distributions of larval instars are not significantly different between the species (Fisher?s exact test, P=0.12).  Species Plants  Total  Eggs All   1st  2nd  3rd    Dissected Eggs+Larvae  Larvae  Instar Instar Instar  C. officinale 123  1669  853 814  610 93 111  H. micrantha 112  43  23 20  19 1 0 _______________________________________________________________________     29     Figure 2.1. Probabilities of C. officinale and H. micrantha plants exhibiting above-ground M. crucifer scars 0, 1, and 2 years after release of 300 weevils on ?target common? sites (n=3) and ?target rare? sites (n=3).  Bar heights are mean values and vertical lines represent 95% confidence intervals generated and back-transformed from a generalized linear mixed model using the presence/absence of scars as a binary response variable, year and species-site category and their interaction as fixed effects, and site/quadrat/plant as nested random effects (P<0.0001, DE= 21%). The number of target plants available for monitoring per site per year ranged from 17-99, and the number of nontarget plants 33-95.              Year 0 Year 1 Year 20.00.20.40.60.8Probability of scarringC. officinale from "target common" sitesH. micrantha from "target common" sitesH. micrantha from "target rare" sites 30         Figure 2.2. Frequency and intensity of M. crucifer colonization with varying plant size in C. officinale and H. micrantha plants sampled from 10 sites with both plant species 0-4 years after weevil release. A) Probability of colonization in all sampled plants (C. officinale n=121, H. micrantha n=103). Lines represent a generalized linear mixed-effects model (GLMM) with presence/absence of eggs or larvae as a binary response variable, plant species, and ln root crown diameter as fixed effects, and release site/pair number as nested random effects (?2=144.7, P<0.001, DE=45%). Years since release was not a statistically significant explanatory variable.  B) The number of eggs and larvae in colonized plants (C. officinale n=100, H. micrantha n=17). An outlier C. officinale plant with 116 eggs+larvae and root crown diameter of 26.64 mm is included in analysis but not displayed. Lines represent a GLMM with count of eggs+larvae as the response variable, species and ln root crown diameter as fixed effects, and release site/pair number as nested random effects (?2=790.7, P<0.001, DE=56%).   01020304050Root crown diameter (mm)Number of plants                             Probability of colonization0 10 20 30 40 500.00.20.40.60.81.0A) All plantsTarget colonizedTarget uncolonizedNontarget colonizedNontarget uncolonizedTarget modelNontarget model0 10 20 30 400102030405060No. eggs + larvaeRoot crown diameter (mm)B) Colonized plants onlyTargetNontarget 31    Figure 2.3. Frequency and intensity of M. crucifer colonization of H. micrantha in relation to site-level C. officinale colonization on 10 sites with both plant species 0-4 years after weevil release. A) Probability of colonization in H. micrantha. The line represents a generalized linear mixed-effects model (GLMM) with used presence/absence of colonization as a binary response variable, average within-site C. officinale use as a fixed effect and site as a random effect (?2=10.0, P=0.002, DE=10%). B) Number of eggs and larvae in sampled H. micrantha. The line represents a GLMM with count of eggs and larvae in plants as a response variable, average within-site C. officinale use as a fixed effect and site as a random effect (?2=6.3, P=0.012, DE=4%). Sample sizes were H. micrantha n=112, C. officinale n=123.   Years since release was not a significant explanatory variable in either model.   010203040Number of plants0 5 10 15xp0.00.20.40.60.81.0Site mean no. eggs+larvae in C. officinaleProbability of  H. micrantha colonizationA) Colonization probabilityColonizedUncolonized0 5 10 150246810Site mean no. eggs+larvae in C. officinaleNo. eggs+larvae in H. micranthaB) Colonization intensity 32      Figure 2.4. Mean (? 95% c.i.) probabilities of C. officinale and H. micrantha from sites with known M. crucifer populations being colonized by at least one egg or larva when displaying above-ground use scars. Values are generated from a generalized linear model with presence/absence of eggs or larvae as a binary response variable and species and presence of scars as explanatory variables (P<0.001, DE=43%).  Sample sizes for C. officinale and H. micrantha were ?Scars Present?: n=139 and n=73 and ?Scars Absent?: n=24 and n=45.  Scars Present Scars Absent0.00.20.40.60.81.0Probability of colonizationTargetNontarget 33 CHAPTER 3  DIFFERENTIAL WITHIN-PATCH TARGET AND  NONTARGET USE FOLLOWING M. CRUCIFER RELEASES:  SPATIAL PATTERNS AND UNDERLYING MECHANISMS    3.1. LITERATURE REVIEW AND OBJECTIVES.  Biological control (biocontrol) of introduced weeds using foreign, host-specific insects (i.e., the ?classical? approach) is a valuable tool for mitigating the environmental impact of invasive plants, but the risk of nontarget use (i.e., feeding, oviposition and larval development) on native nontarget plant species has made this practice controversial (Louda et al. 2003, Hoddle 2004a, 2004b, Louda and Stiling 2004).  Modern biocontrol insects have high host specificity, but may be prone to use less-preferred nontarget hosts in situations with high agent density (Davies and Greathead 1967, Harris 1988, Blossey et al. 2001, Schooler et al. 2003, Baker et al. 2004, Dhileepan et al. 2006, Ganga Visalakshy et al. 2008), or when nontarget plants are in close proximity to target plants (Dennill et al. 1993, Blossey et al. 2001, McFadyen et al. 2002, Schooler et al. 2003, Rand and Louda 2004, Russell et al. 2007, Taylor et al. 2007, Paynter et al. 2008). With effective agents (i.e., those that reduce the abundance of their target weed), these spillover situations are transient and occur either within-season when large numbers of agents are released, or over multiple generations when agents quickly build up on abundant supplies of the target weed (Holt and Hochberg 2001, Lynch et al. 2002).  These situations are not unlike insect outbreaks in natural or agricultural systems, which are known to cause greater impact in terms of loss of biomass or even plant mortality compared with equilibrium dynamics (Romme et al. 1986, Strong et al. 1995, Carson and Root 2000, Coupe and Cahill 2003, Yang 2012). Transient dynamics can have lasting effects on populations (i.e., for at least tens of generations, Lynch et al. 2002, Hastings 2004), and may be especially important to understand in weed biocontrol  34 systems with highly prolific and effective agents, as such herbivores are likely to have high oscillations in abundance, and therefore nontarget use, in response to local target plant density (Holt and Hochberg 2001). Despite the occurrence and potential relevance of transient spillover to nontarget populations, the spatial and temporal patterns and population-level implications of this type of nontarget use in weed biocontrol have largely been unexplored (but see Schooler et al. 2003, Baker et al. 2004), and is generally implied to be inconsequential due to its transient and localized nature (Sheppard et al. 2005).   Damage to individual plants does not necessarily translate to population-level effects (Crawley 1989). One potential reason for this disconnect is the existence of refuges, as plants that escape herbivory can support population persistence (Godfray and Shimada 1999, Berryman et al. 2006). Refuges therefore have direct relevance to weed biocontrol: their existence among target weed populations can thwart successful control (Johnson 2010), yet their presence among nontarget populations may mitigate collateral damage. Refuges can occur in space when insect use is spatially aggregated and ?prey? individuals in certain areas escape damage (Gutierrez et al. 2008). Investigating possible spatial refuges and their underlying mechanisms, such as insect host choice behaviours, can help with prediction of both efficacy and safety in weed biocontrol, particularly during transient periods of intense nontarget use.    Weed biocontrol agents require their target weed in order to maintain high densities in the field, leaving spatial refuges from herbivory for nontargets in patches without their target weed (with a few notable exceptions, Chapter 2). But what scale of spatial isolation is required to create nontarget spatial refuges? The answer likely varies with insect density and host choice behaviours ? in theory, strong host-specificity may generate spatial refuges even within mixed patches of target and nontarget plants. Previous nontarget field observations or studies have been conducted on a range of spatial scales, some along clearly defined distance gradients from target plants (Blossey et al. 2001, Schooler et al. 2003, Russell et al. 2007, Paynter et al. 2008), several as planted open-field tests (Briese et al. 2002, Schooler et al. 2003, Frye et al. 2010) and  35 others at the patch level coarsely defined as ?near? or ?far? from target plants (McFadyen et al. 2002, Rand and Louda 2004, Taylor et al. 2007). However, no study has explicitly compared spatial variation of target and nontarget use within patches, despite the potential importance of herbivory patterns at this scale in determining patch-level plant persistence (Thompson 1978). It is at this within-patch scale where specialized insect host choice behaviours operate (Bernays and Chapman 1994, Finch and Collier 2012), thus potential spatial refuges from herbivory at this fine scale should not be overlooked for predicting population-level effects of nontarget use.     Mogulones crucifer Pallas [=Ceutorhynchus cruciger Herbst, Mogulones cruciger Herbst, Coleoptera: Curculionidae] is a European root-feeding insect approved for release in Canada in 1997 for control of its target invasive rangeland weed, houndstongue, Cynoglossum officinale L. (Boraginaceae). M. crucifer is a successful biocontrol insect in Canada, effectively and rapidly suppressing C. officinale patches and dispersing to surrounding C. officinale infestations (De Clerck-Floate et al. 2005, De Clerck-Floate and Wikeem 2009). M. crucifer can reach high populations within only a few generations when not limited by C. officinale (Schwarzlaender 1997, De Clerck-Floate et al. 2005).  Pre-release and post-release host-specificity testing revealed that the weevil has strong preference for and performance on C. officinale, but that it also can feed, oviposit and develop in several European and North American nontarget Boraginaceae species, including several that are Federally listed as Threatened or Endangered in the United States of America (USA, De Clerck-Floate and Schwarzl?nder 2002). For the latter reason in particular, M. crucifer was denied approval in the USA and was eventually declared a federal pest in that country (U. S. Department of Agriculture 2010). Weevils have since naturally immigrated across the border to C. officinale infestations in the USA (M. Schwarzlaender, personal communication).  Of particular concern is the potential threat M. crucifer poses to Hackelia venusta (Piper H. St. John), a native Boraginaceae species reduced to a single population of approximately 300 plants on 16 hectares in Washington and federally listed as Endangered since 2002 (U. S. Fish and Wildlife Service 2007, 2011). More information is needed on assessing the threat that M. crucifer poses to H. venusta and specifically how to monitor and  36 predict the weevil?s activity on and around the Endangered potential nontarget plant   (U. S. Fish and Wildlife Service 2007, 2011).  Consistent with pre-release tests, M. crucifer in Canada heavily uses its target weed and sporadically consumes and oviposits in several nontarget Boraginaceae species (De Clerck-Floate and Schwarzl?nder 2002, Andreas et al. 2008), with target and nontarget use being positively correlated (Andreas et al., 2008 and Chapter 2 in this dissertation). My focal native, nontarget species in this paper is blue stickseed Hackelia micrantha (Eastw.) J.L. Gentry; plants in this genus incurred some of the highest use among nontarget species in previous M. crucifer studies (De Clerck-Floate and Schwarzl?nder 2002). H. micrantha is known to incur M. crucifer feeding and oviposition and support complete development (Catton and De Clerck-Floate, unpublished results), but requires C. officinale to sustain populations in the field (Chapter 2).   In this chapter, I explore the patterns, mechanisms and potential implications of different within-patch spatial patterns of target and nontarget host use in transient situations of high agent density. I address several key questions: 1) Do target and nontarget use during transient ?outbreak? dynamics differ in spatial pattern at the within-patch scale? 2) Does the pattern of nontarget use during these situations differ predictably with local target plant density? 3) Can within-patch spatial patterns of host use be explained by insect host selection behaviours, such as host-finding and arrestment tendencies? I used two experimental approaches to address these questions. First, I simulated ?worst case? scenarios for nontarget use by releasing large standardized numbers of M. crucifer at single release points in naturally-occurring rangeland patches of H. micrantha with varying densities of C. officinale.  I returned to the sites after 4-7 weeks to document visual indications of M. crucifer use within the patches during the outbreak period.  Second, I performed an outdoor mark-release-recapture experiment (MRRE) in a separate location to explore the behavioural mechanisms behind the patterns observed in the rangeland experiment in terms of M. crucifer host-finding and arrestment (i.e., remaining on a host once encountered) tendencies.  Given the quick spread of M. crucifer to isolated patches of C. officinale upon release in Canada (De Clerck-Floate et  37 al. 2005), and the weevil?s documented higher preference for its target plant, I hypothesized that M. crucifer would have finely-tuned abilities to locate and become arrested by its target weed compared to its novel nontarget host H. micrantha.    3.2. METHODS   3.2.1. STUDY SYSTEM Cynoglossum officinale L. (Boraginaceae), commonly known as houndstongue, is a biennial or short-lived, generally monocarpic perennial plant native to Eurasia that has become invasive in North America. C. officinale grows in disturbed areas in western North America, including disturbed rangeland, roadsides and logged areas (Upadhyaya et al. 1988, De Clerck-Floate 2013).  Pyrrolizidine alkaloids in C. officinale kill livestock when ingested and animals grazing on infested rangeland can suffer dermatitis and fetch reduced prices at auction when coated in the plant?s adhesive burrs (Upadhyaya et al. 1988, Upadhyaya and Cranston 1991). Because C. officinale can quickly exploit disturbed habitats the plant has patchy distributions and displays metapopulation dynamics in both its native and introduced ranges (van der Meijden et al. 1992, De Clerck-Floate 1996).  Mogulones crucifer Pallas [=Ceutorhynchus cruciger Herbst, Mogulones cruciger Herbst, Coleoptera: Curculionidae] is a European root-feeding weevil approved for release in Canada in 1997 to control C. officinale.  M. crucifer is a successful biocontrol insect in Canada, effectively and rapidly suppressing C. officinale patches and dispersing to surrounding infestations (De Clerck-Floate et al. 2005). M. crucifer adults are highly mobile and can walk or fly to new host plants. The weevil is adept at dispersing to new C. officinale infestations, and populations have found the target weed at least as far as 1.42 km from their release within 3 years (De Clerck-Floate et al. 2005).  M. crucifer is univoltine and can reach ?outbreak? population levels within two to three generations when not limited by availability of its target weed (Schwarzlaender 1997, De Clerck-Floate et al. 2005), a characteristic that may be important for achieving successful  38 suppression of target organisms (Gassmann 1996). M. crucifer damages C. officinale mainly through larval root-feeding and reduces fecundity of bolting plants by 30-35% in their native range (Prins et al. 1992, Williams et al. 2010). The weevil is attacked by at least three specialist egg or larval parasitoids at estimated rates between 14-23% in its native range (Schwarzlaender 1997), and these parasitoids are presumed to be absent in the introduced range.   The focal nontarget species in this study is Hackelia micrantha, commonly known as ?blue stickseed? or ?Jessica sticktight?. H. micrantha is a polycarpic perennial native to North America, occurring on mesic slopes, grasslands or shrublands in British Columbia, Alberta and the western USA, and has a global conservation status of abundant and secure (NatureServe 2012). H. micrantha grows sympatrically with C. officinale in semi-forested rangeland and is similar to the invasive plant in life history, morphology and phenology. Like C. officinale, H. micrantha forms a taproot, reproduces only by seed, and produces burred nutlets. However, H. micrantha has a branched shoot architecture as a result of multiple woody caudices emerging from the taproot below ground. Plants in the genus Hackelia incurred some of the highest levels of use among nontarget species in previous M. crucifer studies (De Clerck-Floate and Schwarzl?nder 2002), and H. micrantha is known to incur M. crucifer feeding and oviposition and support complete development of the insect (H. Catton and R. De Clerck-Floate, unpublished results). Information gained about M. crucifer nontarget use of H. micrantha may be relevant to conservation of the Endangered and perennial congener, H. venusta.  3.2.2. RANGELAND EXPERIMENT This field experiment was conducted to characterize within-season spatial patterns of transient host use by M. crucifer around a source of high weevil density (i.e., release points). On June 4, 2009, batches of 300 M. crucifer were released under aspen Populus tremuloides Michx. canopy on six patches of grazed rangeland in the Foothills Fescue Natural Region in southern Alberta, Canada. Patches contained naturally-occurring populations of 34-85 rosette (vegetative) or bolting (reproductive) H. micrantha plants  39 growing interspersed with 0-97 C. officinale rosettes or bolting plants within the radius of plants evaluated (4-14 m).  Patches were at least 360 m from other M. crucifer releases, separated by unshaded grassland (i.e., unsuitable habitat), and showed no signs of weevil activity prior to my releases. Weevils used for the releases were overwintered adults (i.e., in a dispersive state) collected from an outdoor rearing plot and had an estimated female:male ratio of 2.23:1, based on a pooled subsample of 300 weevils. Releases were made on the ground at a marked point within 1 m of host plants, approximately equidistant from nontarget and target plants (when present), so that insects had to move to host plants.   To evaluate M. crucifer use, I assessed 30 rosettes and 20 bolting plants of each species (where possible) as near to the release point as possible on each site; the radius sampled varied with plant density. There were no obvious size differences between plant species within the same life stage (rosette or bolting). Plants were individually examined and mapped in July 2009 (4-7 weeks after release) for visual indications of M. crucifer adult use.  M. crucifer adult feeding and oviposition activity on C. officinale is detectable as distinctive scars appearing as oval-shaped holes (feeding) or raised bumps (oviposition) on the petioles (De Clerck-Floate et al. 2005). Weevil use appeared similar on H. micrantha and also occurred on bolting stems of both species. The amount of cumulative insect activity per plant was categorized as absent (0 scars), mild (1-9 scars), or severe (?10 scars).  3.2.3. MARK-RELEASE-RECAPTURE EXPERIMENT A mark-release-recapture experiment (MRRE) was conducted in spring 2009 during the time of natural activity of M. crucifer in southern Alberta to assess weevil host-finding behaviour for C. officinale and H. micrantha and the effects of proximity of the target weed on dispersal behaviour.  Three pairs of parallel 100 m ?control? and ?treatment? transects were set up on grassy field edges at the Agriculture and Agri-Food Canada Lethbridge Research Centre (Fig. 3.1).  Potted H. micrantha plants were arranged in clusters of 7 randomly-assigned pots (cluster diameter = approximately 55 cm) placed on the ground at 0 m (collection clusters), 2 m, 10 m and 100 m (release clusters) along  40 the transects. Treatment transects had C. officinale plants instead of H. micrantha at collection clusters.  The distance between paired treatment and control transects was 245-255 m and transect pairs were at least 1100 m apart.  Transects faced north to standardize the effects of prevailing west winds on weevil movement and were undisturbed during the experiment.  H. micrantha and C. officinale plants were grown in the greenhouse from seed for 3-4 and 7 months respectively, were all in the rosette stage, and appeared similarly leafy and healthy. H. micrantha and C. officinale plants protruded approximately 5-10 cm and 15-20 cm vertically, respectively, above their 12-15 cm tall and 15 cm diameter pots.   M. crucifer used for the MRRE releases were obtained from the same source as the rangeland experiment.  Several days before the MRRE, weevils were each marked with a spot of Testors? enamel model paint on their pronotum or elytra using one of six colours.  This paint was durable, did not appear to impair the ability of the weevils to open their elytra for flight, and has no apparent effect on M. crucifer survival, fecundity, or behaviour (R. De Clerck-Floate, unpublished results). Two time replicates of this experiment were conducted (May 28-June 5 and June 11-19, 2009). At the beginning of each time replicate, 97-114 M. crucifer of a single colour were released at each 2 m, 10 m, and 100 m H. micrantha cluster.  After 2 days, pots in all clusters were individually inverted and gently shaken to retrieve any weevils. Recaptured weevils from the 2 m, 10 m and 100 m clusters were counted and immediately replaced on their individual plants.  Weevils found on the collection clusters (C. officinale in treatments, H. micrantha in controls) were not replaced, and instead were collected, counted, and sexed under a stereoscopic microscope in the laboratory.  This process was repeated every 2 days until 8 days post-release.  I replaced 2-3 plants per H. micrantha cluster and all C. officinale plants with fresh greenhouse-grown specimens for replicate 2, as the quality of some plants had deteriorated due to a late frost on June 3.  Six new colours were used in the second replicate on new, field-reared weevils.    During replicate 1, weevils at the 100 m cluster in one of the treatment transects were attacked by aggressive Formica spp. ants originating from a previously unnoticed ant  41 hill; therefore, recapture data on this cluster are excluded from analyses. For replicate 2, this transect was moved 26.5 m to the south, which successfully eliminated ant attacks on the released weevils.  Recapture data on 61% of release clusters from day 4 in time replicate 1 were excluded from analyses due to a known inconsistency in weevil counting effort between researchers collecting the data on that day.   3.2.4. DATA ANALYSIS Data were analyzed using JMP 10.0.1.1 (SAS Institute Inc., Cary, NC, USA) and R v2.15.1 (R Core Team 2012). Patterns of visual M. crucifer damage observed in the rangeland experiment were analyzed using generalized linear mixed-effects models with logit link functions and binomial error distributions (GLMM, glmer function in lme4). For the rangeland experiment, the response variables used were odds of presence:absence of weevil scars on all individual plants, and odds of severe:mild use among scarred plants only.  Fixed effects included plant species, life stage (rosette or bolting) and distance to the release point.  Release site was included as a random effect to account for spatial autocorrelation.   Full models including interactions were generated, with non-significant interactions and explanatory variables dropped one by one, until the most reduced model not significantly different than the full model was found (anova function in R). These minimum adequate models were compared to null models including random effects to confirm significant explanatory power. Goodness of fit of models is expressed as the percentage of deviance explained (DE) by the fixed effects (i.e., after consideration of random effects), calculated as 100 ? (deviance in null model with random effects ? deviance in minimum adequate model with random effects)/deviance in null model with random effects.  The standard level of statistical significance used was ?=0.05, and when analyses with multiple categories were significant, posthoc multiple comparisons used Bonferroni-corrected P-values.   For the MRRE, proportions of weevils recaptured on their release clusters and collection clusters were compared using similar GLMM and methods as above.  Fixed effects included collection patch species and distance from the collection patch, while random  42 effects were nested as time replicate/transect pair/transect/plant cluster to account for spatial and temporal autocorrelation and repeated measures.   In posthoc analysis, an abundance of zeroes in the data in some categories made using GLMM inappropriate. Therefore, I pooled cumulative numbers of weevils captured and not captured at collection clusters among the time replicates at the 8-day mark and conducted multiple comparisons with Fisher?s exact tests.  I generated expected numbers of weevils captured at collection clusters based on the proportion of the 360? arc that collection clusters occupied at different distances from releases (2 m = 0.0435, 10 m = 0.0088, 100 m = 0.0009). This calculation incorporates simplifying assumptions that weevils travel in a straight line, choose their angle of movement randomly, remain on encountered plants until my data collection times, and are not influenced by intermediate release clusters. Observed and expected proportions were compared using Fisher?s exact tests.  To test for a sex effect, the cumulative proportions of females:males collected on C. officinale clusters after 8 days were compared to the pre-release proportion of 2:23 females:1 male using Chi-square analysis of multiple proportions (function prop.test in R).   The effect of proximity to target plants on the number of weevils recaptured on their release patches was tested in several ways. To compare the effect of a nearby C. officinale cluster to no nearby clusters, I used GLMM to compare the odds of weevils recaptured:not recaptured between the extreme distances in treatment transects (2 m and 100 m), excluding the plant cluster attacked by ants (100 m, treatment transect). When comparing the effect of a C. officinale cluster to a nearby H. micrantha cluster, GLMM were created comparing collection cluster species (i.e., comparing between control and treatment transects) for each release distance (2 m, 10 m, 100 m).         43 3.3. RESULTS   3.3.1. RANGELAND EXPERIMENT Scars characteristic of M. crucifer use were observed on both C. officinale (when present) and H. micrantha on all release sites. Pooling all sites together, a higher proportion of target plants (106 of 192=55.2%) were used than nontargets (124 of 359=34.5%, ?2=21.9, P<0.001).   Within 1 m of release points, plants of both species and life stages were highly likely to be used by M. crucifer. The minimum adequate GLMM for probability of use indicated that frequency of use decreased significantly with distance for both C. officinale and H. micrantha (Fig. 3.2A). Plant species and life stage and their interactions were also significant in the model, but indicated primarily that the distance decline in host use was stronger for nontarget than for target plants, and bolting plants were more likely to have scars for both species, particularly close to release points. At farther distances, the probability of use of bolting plants and rosettes equalized (at ~4.5 m for H. micrantha), or even inverted (beyond ~10 m for C. officinale).  All bolting C. officinale beyond 11.5 m were unused.  However, this distance included only five bolting target plants, all occurring within meters on one site. I therefore have reduced confidence in the curve for bolting C. officinale far from release points, as the pattern observed could be simply from random properties occurring at that single site.  The model had significantly more predictive power than a null model with random effects (?2=209.7, P<0.001, DE=28%). Whereas target use was observed throughout my evaluation radii, 95% of all used nontargets (118 of 124) occurred within 4.25 m of release points, even though evaluated nontargets occurred beyond this radius.  Within 4.25 m of release points, 47.6% (118 of 248) H. micrantha were used, compared to 5.4% (6 of 111) H. micrantha between 4.26-14 m from the releases. Target plant use was more evenly spread, but was not ubiquitous, as within the 4.25 m radius 82.6% (38 of 46) C. officinale were scarred. Among scarred plants, the minimum adequate model included species, distance from release and plant size (Fig. 3.2B, ?2=46.7, P<0.001, DE=15%).  Consistent with the above pattern, severity of use declined sharply with distance from release points for both  44 species. Bolting plants were 10-40% more likely to have severe use than rosettes.  Severe use was more likely to occur on C. officinale than H. micrantha, although the effect of plant size meant that use severity of scarred target rosettes and nontarget bolting plants was not significantly different. The number of C. officinale plants within the 4.25 m nontarget use radii at the 6 sites ranged from 0-26, and was not a significant variable in explaining probability of use (?2=0.90, P=0.34) or severity of use in scarred plants (?2=0.45, P=0.50).  3.3.2. MARK-RELEASE-RECAPTURE EXPERIMENT Overall I released 3850 painted weevils and retrieved 362 (9.4%) of them on collection clusters within 8 days of release.  Weevil recapture rates on collection clusters were explained by time, distance, and collection cluster species (?2=324.0, P<0.001, DE=71%, Fig. 3.3A). The cumulative number of weevils captured on the collection clusters always increased from 2 to 8 days after release, except for weevils released 100 m from H. micrantha collection clusters, where no weevils were successful dispersers. Significantly higher percentages of released weevils (pooled from all distances and both time replicates) were collected over 8 days at the collection clusters when they contained the target weed (351 of 1918=18.3%) rather than the nontarget (11 of 1932=0.6%, P<0.001 Fisher?s exact test). Weevils were 43 to 96 times more likely to find the C. officinale rather than H. micrantha collection clusters in 8 days from each release distance (Fig. 3.3B). Recapture rates were a minimum of 9 times greater than the random expectations for C. officinale and below or equal to random expectations for H. micrantha collection clusters. On transects with C. officinale collection clusters, weevils were more likely to reach the target weed from 2 m than 10 m, and 10 m than 100 m (Fisher?s exact test, P?0.008).  On transects with H. micrantha collection clusters, 10 weevils reached the collection clusters from 2 m away, one weevil from 10 m away, and no weevils from 100 m. Notably, the proportion of weevils reaching C. officinale from 100 m away was equal to those finding H. micrantha 2 m away.  I continued to collect weevils from the first time replicate on collection clusters for up to 22 days post-release during the second time replicate (weevils from replicate 1 were  45 painted different colours from replicate 2). After the 5-day break in data collection between time replicates (days 8-14), an additional 9.9 ? 1.8% (mean ? S.E.), 4.0? 0.5%, and 2.0 ? 0.0% of released weevils found C. officinale collection patches from 2 m, 10 m, and 100 m away, respectively (n=2 transects). In contrast, no weevils were recaptured on H. micrantha collection clusters after the 5-day break (n=3 transects). By the end of the second time replicate, 52.9% (108 of 204), 21.9% (46 of 210), and 3.4% (7 of 207) of released weevils from time replicate 1 reached C. officinale collection clusters from 2 m, 10 m, and 100 m over the 22 day period, whereas only 1.4% (13 of 955) of weevils released at any distance from H. micrantha were recaptured on the nontarget collection clusters (n=3 cluster-transects) over 22 days.  The sex ratios of weevils collected on C. officinale were not significantly different than the expected values of 2.23 females:1 male (n=18, ?2=21.23, P=0.268).   Marked weevils were recaptured on other release clusters within their transect 49 times within 8 days. However, these cases may not represent 49 individuals, as weevils counted on release clusters were replaced (unlike collection clusters). Assuming these are 49 unique individuals, 1.3% (49 of 3850) weevils were found on other H. micrantha release clusters in their transects 2, 4, 6 or 8 days post-release. The most frequent within-transect movement was from 10 m to 2 m release clusters (with both collection cluster species) and from 2 m to 10 m release clusters with H. micrantha collection clusters only.  Out of 643 weevils released 2 m from H. micrantha collection clusters, 10 weevils moved to the collection clusters over 8 days, while 10 weevils moved 8 m in the opposite direction to the 10 m release clusters.  In contrast, of weevils released 2 m from C. officinale, 258 moved 2 m to the target weed and 0 moved to the 10 m H. micrantha release clusters.   There were 17 instances of long-range dispersal (225-270 m) between paired transects over 16 days. Ten weevils from control transects (i.e., with H. micrantha collection clusters) found their paired C. officinale collection clusters (3 weevils within 2 days), and 7 weevils achieved inter-transect dispersal to H. micrantha release clusters (2 weevils within 2 days). All inter-transect movements except one were in an easterly  46 direction, consistent with both the direction toward C. officinale, and the prevailing westerly winds of the experimental area.   More than 3 in 4 weevils released on nontargets were absent from their release clusters after 2 days (mean ? S.E. percentage not recaptured per cluster = 77.7 ? 1.3%, n=35 clusters). After 8 days, 92.1 ? 0.9% of released weevils (n=35 clusters) were not recovered on their release clusters.  The number of weevils recaptured on their release clusters decreased significantly with time (Fig. 3.4A, n=35, ?2=365, P<0.001, DE= 57%). There were, however, instances when the number of weevils was higher than the previous collection, indicating that some weevils left and re-found their release clusters. The minimum adequate model included statistically significant effects and interactions of collection cluster distance and species (not shown); however, no clear patterns were evident, and the extra 10 parameters increased the amount of deviance explained by the model by only 6%.  In post-hoc multiple comparisons within distances, no significant effects of collection cluster species were observed in release clusters 10 m or 100 m from H. micrantha or C. officinale collection clusters.  At 2 m, a significant interaction between time and collection cluster species was observed (Fig. 4B, ?2=15.6, P<0.001); however, the reduced number of weevils recaptured 2 m from C. officinale may be due to many weevils being collected (and thus removed from the experiment) on the nearby target collection clusters.  When data from day 2 were analyzed separately for 2 m release clusters (i.e., before collected weevils were removed from the experiment), the collection cluster species effect was not significant (?2=2.0, uncorrected P=0.16).   There was no difference between the number of weevils recaptured on H. micrantha release clusters 2 m and 100 m from C. officinale collection clusters (?2=2.5, P=0.28). Dead weevils recaptured on their release clusters revealed a minimum 8-day mortality rate per release cluster of 3.0?0.5% (range= 0-11.3%, n=36).      47 3.4. DISCUSSION  This study uncovers valuable information about differential host use at the within-patch scale during a transient ?worst-case scenario? of high agent density. While C. officinale was used throughout my 14 m evaluation radii, 95% of H. micrantha use was restricted to within 4.25 m of release points, demonstrating that nontarget spatial refuges from biocontrol herbivory occur even within mixed patches of target and nontarget plants. In contrast, the extended radius of C. officinale use around release points suggests that target plants away from high weevil density get used, a characteristic that surely contributes to the success of this insect as a biocontrol agent in Canada. H. micrantha use was independent of the abundance of the target weed, suggesting that transient nontarget use can occur anywhere there is high weevil density, such as immediately following biocontrol releases.   The host-specific distance effect observed in this study suggests that nontarget use by M. crucifer is more dependent on a high chance of random encounters than its use of the target weed.  Plant size likely also plays a part in encounter rate, as M. crucifer prefers larger C. officinale (Prins et al. 1992, Schwarzlaender 1997), and this pattern holds as well for H. micrantha (Chapter 2). The distance pattern differs from the only other study comparing target and nontarget use from a source of high insect density.  There, Schooler at al. (2003) found feeding damage by Galerucella spp. beetles (Coleoptera: Chrysomelidae) was higher on target (Lythrum salicaria L.) than nontarget plants (Lagerstroemia indica L.) and declined with distance from high densities of Galerucella spp. beetles, but the distance effect was not species-specific. In general, dispersers have a decreasing chance of encountering plants with increasing distance because the resulting larger search area leads to increased mortality and reduced random chance of finding hosts (Grevstad and Herzig 1997). Likewise, equal numbers of released insects in a smaller search area means increased insect density and thus increased chance of insect-plant encounters.  Because of the standardized number of insects released in the rangeland experiment and the consistency of the spatial patterns observed among release sites, it is possible to estimate a critical M. crucifer density for H. micrantha use.  48 Assuming the 300 weevils on each site diffused outward in random directions from their release point (Kareiva 1983, Grevstad and Herzig 1997), the maximum local instantaneous density of insects at each site would have occurred at the release point and decreased exponentially with distance and time as insects dispersed. Therefore, shortly after release in the 4-7 week period when the use scar ?footprints? were made, there were an estimated maximum of 5.3 weevils/m2 (300 weevils/57 m2) in the 0-4.25 m nontarget use radius compared to 0.5 weevils/m2 (300 weevils/616 m2) throughout the 0-14 m C. officinale use radius.   Notably, while the evidence of use remained for weeks, weevil density near the release points would have been decreasing over time during the diffusion process. Therefore, the observed weevil scarring on lower-ranked hosts may have occurred over a short time span (Albright et al. 2004), as the ?wave? of insects radiated outward from its high density release area (Landis et al. 2003, De Clerck-Floate et al. 2005).  The rapid departure of most M. crucifer from nontargets in my MRRE and the weevil?s reduced tendency to find and remain on H. micrantha supports the idea of short-term nontarget use.  Furthermore, M. crucifer that encounter and feed on H. micrantha most often depart without ovipositing (Chapter 2). Therefore, while use scars are a valuable indicator of cumulative past within-season weevil activity, they should not necessarily be interpreted as ongoing high-level use, particularly on nontarget plants.   The relatively large radius where M. crucifer used C. officinale may be explained by specialized target finding behaviours. Specifically, behaviours such as changes in the weevil?s search locomotion when a stimulus is perceived (Visser 1988), arrestment behaviour (i.e., suspension of searching movement once a host is encountered), and host acceptance for feeding or oviposition may be greater for its highly-preferred host plant than for the H. micrantha.  M. crucifer weevils disperse within and between patches of C. officinale, and efficiently locate isolated target plants, even when forced to travel through dense forest (De Clerck-Floate et al. 2005). In the MRRE, weevils were recaptured on C. officinale located 2 m and 10 m from release points at rates significantly greater than random expectation (and recaptures 100 m from releases were  49 close to statistically significant).  This pattern suggests that my assumption of random straight-line movement by the weevils is overly simplistic, and that M. crucifer weevils are likely attracted to C. officinale over some distance, or use other locomotion patterns such as random walks (walking or flying, Codling et al. 2008) to increase their chance of finding suitable hosts. Indeed, phytophagous insects are known to switch between straight-line ranging and tortuous local-searching motions depending on stimuli such as wind, or host plant volatiles (Visser 1988). Because my data represent an integration of host-finding and arrestment (and acceptance behaviours in the rangeland experiment), I cannot determine the relative contributions of these separate and important processes. However, olfactometer experiments with C. officinale suggest that target plant volatiles do not attract M. crucifer but instead arrest their searching behaviour when detected (Andreas et al. 2009). The large number of time replicate 1 weevils recaptured on C. officinale after a 5-day break in data collection is consistent with a strong arrestment mechanism on the target weed. Similarly, a study of Galerucella calmariensis L. movement by Grevstad and Herzig (1997) suggested beetles moved in random directions and stopped when target plants were encountered. Further research to distinguish between host-finding, arrestment, and host acceptance can narrow nontarget use predictions from a broad probability of encounters (based on weevil density and plant density and dispersion) to more specific probabilities of actual use (the proportion of encountered plants that get accepted for feeding or oviposition).    In contrast with C. officinale, H. micrantha appears to have only a weak ability to attract and retain dispersing M. crucifer. Consistent with nontarget behaviour in other biocontrol agents (Briese et al. 2002, Frye et al. 2010), M. crucifer recaptures on H. micrantha in the MRRE were equal to or lower than random expectation, and no weevils were recaptured on nontarget collection clusters between MRRE time replicates.  Both the phylogenetic and ecological associations of phytophagous insects and their host plants influence host-searching and arrestment behaviours (Mackay 1985). M. crucifer may therefore be more responsive to its evolutionary host, C. officinale, during such processes than to H. micrantha, a novel host encountered only since 1997 when the weevil was first released in Canada. The difference in recaptures on C. officinale  50 compared to H. micrantha in the MRRE may be at least partially attributable to the larger size of the potted target plants relative to potted nontargets, as the plant size effect was significant in the rangeland experiment. However, a lack of nontarget host-finding is consistent with olfactometer experiments where M. crucifer were arrested by odour plumes of C. officinale but were repelled by volatiles from Hackelia floribunda (Lehm.) I.M. Johnst. (Andreas et al. 2009). Another congeneric weevil, Mogulones borraginis F. (Coleoptera: Curculionidae) also distinguished C. officinale from a native confamilial plant, Hackelia californica (A. Gray) I.M. Johnst. based on olfactory and visual cues in dual choice bioassays (I. Park, unpublished data). H. micrantha produces volatiles of several compounds not detected in C. officinale (J. Runyon, unpublished data), and these compounds may contribute to dispersing M. crucifer being repelled or rarely arrested by H. micrantha. The MRRE demonstrated that more than 3 in 4 weevils leave nontargets within 2 days, and fewer than 10% remain after 8 days. This result is consistent with nontarget residence of Zygogramma bicolorata (Coleoptera: Chrysomelidae), a biocontrol weevil for Parthenium hysterophorus in India. Of weevils found on the nontarget crop plant sunflower Helianthus annuus, 87% were only 7-8 days old, and no weevils were more than 10 days old (Ganga Visalakshy et al. 2008).  A lack of arrestment supports the idea that the nontarget use observed in the rangeland experiment was of short duration.      The knowledge gained in this study contributes to our understanding of the patterns and implications of transient nontarget spillover at the within-patch scale in the M. crucifer biocontrol system. Study at this temporal and spatial scale may be especially relevant for a locally unstable system characterized by a mobile, effective and fast-acting oligophagous agent like M. crucifer.  The weevil overexploits its target host plants and does not appear to be heavily regulated by anything but C. officinale availability in the introduced range (specialized egg and larval parasitoids attack M. crucifer in its native range, Schwarzlaender, 1997), therefore local outbreaks may be common and this weed-biocontrol system may never reach equilibrium dynamics at the within-patch scale. Rather, long-term dynamics likely need to be considered at the landscape level, in the form of metapopulation dynamics (Hanski 1998). A similar dynamic exists with the  51 overexploiting weevil Hadramphus spinipennis and its host plant Aciphylla dieffenbachii in New Zealand (Sch?ps 2002).  C. officinale has metapopulation dynamics in both its native and introduced range (van der Meijden et al. 1992, De Clerck-Floate 1996). Therefore, not only can we expect localized and temporary nontarget use around recent biocontrol releases, but also during regular ?boom and bust? dynamics of C. officinale and M. crucifer.  However, because relative to nontarget plants the target weed is both a strong source of new weevils (De Clerck-Floate and Schwarzl?nder 2002, Chapter 2) and an attractive sink for dispersing weevils (MRRE experiment in this chapter), extended periods of high M. crucifer densities should be constrained to a narrow area in and around C. officinale plants.  Finally, when spillover occurs during transient outbreaks, my data demonstrate that although nearby C. officinale is not required to stimulate nontarget use, M. crucifer rarely find and use H. micrantha, even when it is only meters away within the patch. The resulting differential host use may lead to spatial refuges from nontarget use at the within-patch level that can buffer nontarget plants from population-level effects. In fact, since spillover may be a reliable indicator of high local insect density, nontarget use by effective agents such as M. crucifer may be a strong predictor that local target weed control is forthcoming.  In terms of other nontarget species, plants in the genus Hackelia were among the most used nontargets during host-specificity testing (De Clerck-Floate and Schwarzl?nder 2002); I therefore expect that most native Boraginaceae outside this genus will be even more buffered from population-level effects. Nontarget species preferred more by M. crucifer than H. micrantha may have smaller spatial refuges.  Biocontrol insects with reduced search efficiency for nontarget relative to target organisms present a low risk for causing population-level impacts to nontargets in both the short-term and long-term (Lynch et al., 2002). A logical next step is to specifically test for population-level effects from transient spillover using plant demographic data, and this is the topic for Chapter 4 in this dissertation.  52   Figure 3.1. Schematic diagram (not to scale) of a pair of transects in the mark-release-recapture experiment using M. crucifer and potted C. officinale and H. micrantha plants. The full experiment consisted of six pair of transects, three each from two time replicates.  Cluster diameters were 0.55 m.      COLLECTION CLUSTER RELEASE CLUSTER COLOUR #1 2 m 10 m 100 m = potted H. micrantha = potted C. officinale TREATMENT TRANSECT CONTROL TRANSECT >245 m Legend: North 0 m RELEASE CLUSTER COLOUR #4 RELEASE CLUSTER COLOUR #5 RELEASE CLUSTER COLOUR #6 RELEASE CLUSTER COLOUR #3 RELEASE CLUSTER COLOUR #2  COLLECTION CLUSTER  53     Figure 3.2.  Frequency and severity of M. crucifer use scarring on C. officinale and H. micrantha around M. crucifer release points in the rangeland experiment. Plants were evaluated 4-7 weeks after releases of 300 M. crucifer each on 6 sites.  A) Probability of all evaluated bolting plants and rosettes having evidence of scarring. Lines represent a generalized linear mixed model (GLMM) with presence/absence of scarring as a binary response variable, plant life stage and species and their interactions as fixed effects, and site as a random effect (?2=209.7, P<0.001, DE = 28%). B) Among scarred plants only, probability that use ?severe? (10 or more scars) or mild (1-9 scars). Lines represent a GLMM with severity as a binary response variable, and identical explanatory variables as above (?2=46.7, P<0.001, DE = 15%). C) and D) Raw data pooled among 3 sites for C. officinale and 6 sites for H. micrantha, respectively.   0.00.20.40.60.81.0Prob. of scarringA) All plantsTarget ros.Target bolterNontarget ros.Nontarget bolter0.00.20.40.60.81.0Prob. scarring is severe B) Scarred plants onlyTarget ros.Target bolterNontarget ros.Nontarget bolterProb. scarring is severe0102030405060No. plantsC) Target plantsRos. unscarredRos. mildRos. severeBolter unscarredBolter mildBolter severe0102030405060Number of plants0 2 4 6 8 10 12 14D) Nontarget plantsRos. unscarredRos. mildRos. severeBolter unscarredBolter mildBolter severeDistance from release (m) 54     Figure 3.3. Weevil recaptures on C. officinale and H. micrantha collection clusters 2 m, 10 m and 100 m from release points in the mark-release-recapture experiment. A) Probabilities that marked M. crucifer were recaptured on collection clusters over 8 days. Collection clusters consisted of potted C. officinale (n=6) or H. micrantha (n=6).  The number of insects released at each cluster ranged from 97-114.  Gray lines represent raw data from individual release clusters from 2 m (open circles), 10 m (closed circles), or 100 m (triangles) away. The thick, black lines represent a generalized linear mixed model with recaptures as a binary response variable, days since release, distance, and collection cluster species as fixed effects and replicate/transect/cluster as nested random effects (?2=324.0, P<0.001, DE = 71%).  B) Cumulative numbers of marked M. crucifer recaptured on target and nontarget collection clusters by day 8 compared to numbers expected by random chance. The total number of weevils released per species-distance category ranged from 638-645. Numbers are pooled across time and transect replicate (n=6 replicates per category).  Bars with different letters indicate significant differences in proportions (P?0.008) according to Fisher?s exact tests with Bonferroni-corrected P-values. Bars with asterisks (*) are significantly different from expected values (Fisher?s exacts tests, P?0.02). 0.00.10.20.30.40.50.60.72 4 6 8Proportion finding collection clusterDays since release C. officinale  H. micrantha2 4 6 8(A) Cumulative recaptures over timeModels2 m10 m100 m2 m 10 m 100 m020406080100240260a*c*b*cdcdNumber of WeevilsRelease Clusters(B) Totals after 8 daystargetnon-targetexpected 55    Figure 3.4. Probabilities that marked M. crucifer weevils released on H. micrantha plant clusters in the mark-release-recapture experiment were recaptured and replaced on their release clusters over 8 days post-release.  A) Thin grey lines represent raw data from individual release clusters, from two time replicates (n=35 clusters with 97-114 weevils released on each). The thick black line represents a generalized linear mixed model (GLMM) with recaptures as a binary response variable, the number of days since release as a fixed effect and time replicate/transect pair/transect/cluster as nested random effects (?2=409.9, P<0.001, DE = 57%).  B) The GLMM for 2 m release clusters, showing a significant interaction between species and distance (?2=136.6, P<0.001, DE = 67%).  0.00.10.20.30.40.52 4 6 8Proportion stayedDays since release(A) All clusters2 4 6 8Days since release(B) 2 m clusters onlyindiv. clusters, target at 0 mindiv. clusters, non-target at 0 mmodel: all clustersmodel: target at 0 mmodel: non-target at 0 m 56 CHAPTER 4  INDIVIDUAL AND POPULATION-LEVEL IMPACTS TO TARGET  AND NONTARGET PLANTS FOLLOWING M. CRUCIFER RELEASE   4.1. LITERATURE REVIEW AND OBJECTIVES  Biological control (biocontrol) of introduced weeds using foreign, host-specific insects (i.e., the ?classical? approach) is a valuable tool for mitigating the environmental impact of invasive plants, but the risk of nontarget use (i.e., feeding, oviposition and larval development) on native, nontarget plant species has made this practice controversial (Louda et al. 2003, Hoddle 2004a, 2004b, Louda and Stiling 2004).  Perfectly monophagous insect herbivores are rare (Fox and Morrow 1981, Odegaard et al. 2000), and when expected net benefits of biocontrol are high, oligophagous agents (i.e., insects with known but relatively low propensities to use closely-related, native nontarget plants) have been given regulatory approval for release (van Klinken and Edwards 2002, Louda et al. 2003, Sheppard et al. 2005). Approval is typically granted under the assumption that an insect?s reduced preference and performance on a nontarget plant minimizes nontarget impact to acceptable levels (Zw?lfer and Harris 1984, De Clerck-Floate and Schwarzl?nder 2002).  However, the extent of nontarget use by thistle biocontrol weevils Rhinocyllus conicus Fr?lich, Larinus planus Fabricius, and Trichosirocalus horridus Panzer has challenged this assumption, because in these systems insect use of nontarget species was widespread, long-term and potentially damaging to native plant populations (Louda 1998, Louda and O'Brien 2002, Takahashi et al. 2009, Havens et al. 2012).  The post-release impact of weed biocontrol agents is rarely studied (McEvoy and Coombs 1999), and further investigation is needed to define the risks biocontrol agents may pose to native species in the introduced range. Such information will aid in optimal management strategies for existing weed biocontrol systems and increase informed decision-making regarding the release of new agents  57 through an understanding of how pre-release host specificity test results translate into post-release herbivory patterns and impacts.   Insect herbivory on individual plants can result in reduced demographic parameters of survival, growth and fecundity, but this effect is highly dependent on multiple factors such the nature and extent of the damage, the biology of the plant and herbivore, competitive dynamics, and abiotic conditions (Trumble et al. 1993, Peterson 2000). Plants can react to and prevent further insect herbivory damage through inducible chemical defenses (Howe and Jander 2008), and exhibit compensatory or even overcompensatory growth in response to insect herbivory (Trumble et al. 1993). In other cases, particularly with high insect density, injury can be devastating or even lethal to individuals (Romme et al. 1986, Strong et al. 1995, Carson and Root 2000, Coupe and Cahill 2003, Yang 2012).    Herbivory effects on plant populations remain poorly understood and difficult to predict (Crawley 1989, Halpern and Underwood 2006, Maron and Crone 2006).  For example, depending on the frequency, severity, and distribution of herbivore damage to individual plants, population-level reductions in average survival, growth or fecundity (i.e., vital rates) may or may not occur.  Aggregated patterns of herbivory can leave ?refuge? plants that escape heavy damage and continue to survive, grow, and reproduce to maintain viable populations (Hawkins et al. 1993, Berryman et al. 2006, Johnson 2010). Even if vital rates are affected at the population level, their changes may not impact the overall population growth rate (Crawley 1989). Compensatory mechanisms in population dynamics, such as seed dormancy, density-dependence, and competitive release can buffer population growth rates from change (Garren and Strauss 2009, Swope and Parker 2010, Ortega et al. 2012).   Because of these and other factors, the impact of biocontrol agents on both target and nontarget plants is difficult to predict.  The result is that many released and established weed biocontrol agents have failed to suppress their target weeds (McClay and Balciunas 2005), and uncertainty remains regarding risks of impact to nontarget hosts.   58 One way population vital rate data can be integrated to determine population growth rates is through the use of matrix population models (Caswell 2001, Crone et al. 2013). Matrix models have been used for hundreds of plant species to study their population dynamics, often with a main objective of identifying influential life stage transitions that can be targeted for management purposes (Crone et al. 2011).  In weed biocontrol, matrix models have been used to inform selection of candidate biocontrol agents based on targeted disruption of plant life cycle stages (McEvoy and Coombs 1999, Davis et al. 2006), to assess and predict impact (or lack thereof) to target weed population growth rates by agents (Shea and Kelly 1998, DeWalt 2006, Schutzenhofer and Knight 2007, Dauer et al. 2012), and to identify compensation mechanisms in target plant population dynamics (Maines et al. 2013).  Conversely, matrix population studies of nontarget plants in weed biocontrol systems are rare, except for a few studies on invasive thistle biocontrol in North America (Louda et al. 2005, Rose et al. 2005, Havens et al. 2012). These studies have indicated that nontarget species Platte thistle (Cirsium canescens Nutt.) and USA federally Threatened Pitcher?s thistle (Cirsium pitcheri [Torr.] Torrey & Gray) face persistent herbivory and reduced fecundity and population growth rates as a result of the introduced seed-feeding weevils Rhinocyllus conicus Fr?lich and Larinus planus Fabricius, and have raised important questions regarding the assumptions justifying the release and distribution of oligophagous agents in biocontrol (Louda et al. 2003).  Demographic analysis of other weed biocontrol systems with oligophagous agents is important and necessary for determining whether the thistle cases are unique in their severity, and for defining how to monitor for nontarget effects. Such retrospective analyses can increase our predictive power regarding nontarget risks and inform future release decisions and management actions.    Cynoglossum officinale (Borginaceae) is a monocarpic, perennial plant species of European origin that has become invasive in disturbed rangeland, roadsides and logged areas in western North America (Upadhyaya et al. 1988, Upadhyaya and Cranston 1991, De Clerck-Floate 2013). In 1997, the Canadian government approved the release of Mogulones crucifer, an oligophagous European root-feeding weevil, with high preference and performance for C. officinale, as a biocontrol agent to control the  59 invasive plant. Since its release, M. crucifer has been highly effective at suppressing C. officinale populations in western Canada, even in some cases eliminating the target weed within 2 years after release (De Clerck-Floate et al. 2005, De Clerck-Floate and Wikeem 2009).  In its native range, M. crucifer reduces C. officinale fecundity by 30-35% without affecting survival of plants (Prins et al. 1992, Williams et al. 2010), a demographic impact that was not predicted to have major population-level consequences for C. officinale in North America (Maron et al. 2010, Williams et al. 2010).  As the impact predictions based on European data are in contrast with observed results in Canada, the question presents itself as to how the M. crucifer-C. officinale interaction differs in the introduced range to produce the dramatic weed control observed.    In addition to controlling C. officinale, M. crucifer also sporadically feeds and oviposits on native confamilial nontarget species in Canada (De Clerck-Floate and Schwarzl?nder 2002, Andreas et al. 2008), with unknown individual and population-level consequences. Nontarget use was not unexpected for this oligophagous insect, as pre- and post-release host specificity tests revealed that the weevil?s host range included several Boraginaceae species, however both its preference and performance were much stronger for C. officinale (Jordan et al. 1993, De Clerck-Floate et al. 1996, De Clerck-Floate and Schwarzl?nder 2002).  Despite its success in Canada, M. crucifer was not approved for release in the USA because of the weevil?s potential impact on Endangered Boraginaceae in the USA, and in 2010 it was declared a federal pest (U. S. Department of Agriculture 2010).  The weevil is dispersing naturally from Canada to C. officinale infestations in the USA, and substantial M. crucifer populations in Washington and Idaho are spreading 10-15 km southward per year (M. Schwarzlaender, personal communication).  Of particular concern is the potential threat M. crucifer poses to Hackelia venusta (Piper H. St. John), a native Boraginaceae species reduced to a single population of approximately 300 plants on 16 hectares in Washington and federally listed as Endangered since 2002 (U. S. Fish and Wildlife Service 2007, 2011). More information is needed on assessing the threat that M. crucifer poses to H. venusta and specifically how to monitor and predict the weevil?s activity on and around the Endangered potential nontarget plant (U. S. Fish and Wildlife Service 2007, 2011).  60  In Chapters 2 and 3, I demonstrated that relative to C. officinale, many H. micrantha plants escape use by M. crucifer through spatial, temporal, and probabilistic refuges. In this chapter, I examine the differences in vital rates and population growth rates on both plant species in relation to M. crucifer releases.  I present demographic data for C. officinale and H. micrantha plants on M. crucifer release and non-release sites for two one-year transitions (2009-2010, 2010-2011). I compare the demographic parameters of survival, growth and fecundity of rosette and flowering C. officinale and H. micrantha between release and non-release sites to test for weevil impact on vital rates. Further, I compare demographic parameters of plants within release sites based on distance from release points and visually distinctive scars representing M. crucifer feeding or oviposition on plants.  Finally, I combine the vital rates from each plant species on each site to obtain population growth rates (lambda) and assess weevil impacts on both plant species at the population level.  My hypotheses, based on previous data in Canada (De Clerck-Floate et al. 2005, De Clerck-Floate and Wikeem 2009), were that M. crucifer would have more severe demographic impacts on C. officinale than that observed in Europe (Prins et al. 1992, Williams et al. 2010), resulting in lower vital rates and population growth rates.  In contrast, my prior observations of abundant H. micrantha at separate M. crucifer release sites 1 year after weevil release led me to hypothesize that the weevil has no population-level impact on the nontarget plant.   4.2. METHODS  4.2.1.  STUDY SYSTEM Cynoglossum officinale, commonly known as houndstongue, is a biennial or short-lived, generally monocarpic perennial plant native to Eurasia that has become invasive in North America. C. officinale grows in disturbed areas in western North America, including disturbed rangeland, roadsides and logged areas (Upadhyaya et al. 1988).  Pyrrolizidine alkaloids in C. officinale kill livestock when ingested and animals grazing on infested rangeland can suffer dermatitis and fetch reduced prices at auction when  61 coated in the plant?s adhesive burrs (Upadhyaya et al. 1988, Upadhyaya and Cranston 1991). C. officinale reproduces by seed only and generally requires small-scale soil disturbance to establish in both its native and introduced ranges (Klinkhamer and de Jong 1988, Williams et al. 2010).  C. officinale seeds germinate in the spring and overwinter as a vegetative rosette and taproot (de Jong and Klinkhamer 1988b).  When a threshold size is reached (Wesselingh et al. 1997), in the second or later spring, plants flower (i.e., ?bolt?) and produce adhesive burred nutlets for dispersal by epizoochory (De Clerck-Floate 1997). Flowering is usually fatal for C. officinale, but 2-45% of flowering C. officinale in the introduced range may survive to flower a second time (Williams 2009).  C. officinale can quickly exploit disturbed habitats and therefore has patchy distributions and displays metapopulation dynamics in both its native and introduced ranges (van der Meijden et al. 1992, De Clerck-Floate 1996).  Mogulones crucifer damages C. officinale mainly through larval root-feeding (Prins et al. 1992), and can reach high population numbers within several generations when not limited by C. officinale availability (Schwarzlaender 1997). Overwintered adults emerge in spring, feed on C. officinale foliage, and mate. Females lay eggs singly into petioles, bolting stems, or the root crown (lifetime average= 190 eggs, Schwarzlaender, 1997), with a preference for large plants and for bolting plants over rosettes (Prins et al. 1992, Schwarzlaender 1997). Larvae mine the root crown and roots, consuming tissues until they exit as third instar larvae to pupate in the soil. Emerging adults feed and oviposit in late summer prior to overwintering in soil or leaf litter. Eggs, larvae and pupae may also overwinter. The weevil is univoltine and most adults perish within 12 months (Schwarzlaender 1997). M. crucifer adults are highly mobile and can walk or fly to new host plants. The weevil is adept at dispersing to new C. officinale infestations, and populations have found the target weed at least as far as 1.42 km from their release within 3 years (De Clerck-Floate et al. 2005).  M. crucifer is attacked by at least three specialist egg or larval parasitoids at estimated rates between 14-23% in its native range (Schwarzlaender 1997), but these parasitoids are presumed to be absent in the introduced range.    62 Hackelia micrantha, commonly known as ?blue stickseed? or ?Jessica sticktight?, is a polycarpic perennial native to North America, occurring on mesic slopes, grasslands or shrublands in British Columbia, Alberta and the western USA. The plant?s global conservation status is abundant and secure (NatureServe 2012). H. micrantha can grow sympatrically with C. officinale in semi-forested rangeland and is similar to the invasive plant in life history, morphology and phenology. Like C. officinale, H. micrantha forms a taproot, reproduces only by seed, and produces burred nutlets. However, H. micrantha has a branched shoot architecture as a result of multiple woody caudices emerging from the taproot. H. micrantha incurs M. crucifer feeding and oviposition and supports complete development of the insect (H. Catton and R. De Clerck-Floate, unpublished results), but does not appear to independently sustain M. crucifer populations in the field (Chapter 2). Information gained about M. crucifer nontarget use of H. micrantha may be relevant to conservation of its Endangered and fellow perennial congener H. venusta.  4.2.2. RANGELAND RELEASE EXPERIMENT A field experiment using insect releases was conducted from 2009 to 2011 to characterize demographic patterns of C. officinale and H. micrantha in the presence and absence of M. crucifer. In 2008, 12 sites with naturally-occurring patches of H. micrantha were located under aspen Populus tremuloides Michx. canopy on grazed rangeland in the Foothills Fescue Natural Region in southern Alberta, Canada (Natural Regions Committee 2006).  Sites contained 29 to 85 H. micrantha rosettes or bolting plants within a radius of 4-22 m. The exception was one linear population that occurred along a cow path for a distance of approximately 100 m (this site was used as a non- release site because of its unique shape).  On six sites, herein referred to as ?target common? sites, H. micrantha were growing interspersed with 35-98 naturally-occurring rosette or bolting C. officinale plants.   The other six sites were classified as ?target rare? sites, as they had low numbers of C. officinale (0-7) within their evaluation radii that were manually removed upon discovery beginning in July 2009 until August 2011. Patches were at least 360 m apart and separated by unshaded grassland (i.e., habitat not suitable for either plant species), and showed no signs of M. crucifer activity prior to my releases.  63  On June 4, 2009, batches of 300 M. crucifer were released at three ?target common? and three ?target rare? sites.   Weevils used for the releases were overwintered, ovipositing adults collected from an outdoor rearing plot in southern Alberta and had an estimated female:male ratio of 2.23:1, based on a pooled subsample of 300 weevils. Releases were made on the ground at a marked point within 1 m of host plants, approximately equidistant from nontarget and target plants (when present), so that insects had to move to host plants.   I identified and tracked individual plants on all sites by marking 1?1 m permanent quadrats around release or, for non-release sites, central reference points. Where possible, I sampled 30 rosettes and 20 bolting plants, and 50 seedlings of each species as near to the release or reference point as possible. The radius sampled and number of quadrats laid per site varied with plant density (8 to 38 quadrats per site, total=228 quadrats).  Plants within quadrats were mapped to the nearest 10 cm once per year during July 3-26, 2009, June 29-July 23, 2010, and July 5-August 3, 2011. Plants of both species in each quadrat were visually examined each year and assigned to one of the following development stages: 1) seedling: a plant with cotyledons or 3 or fewer true leaves; 2) small rosette: a plant with 4-9 true leaves, with all leaves <30 cm tall; 3) large rosette: a plant with 10 or more true leaves, or at least one leaf ?30 cm tall, 4) bolter: a plant with at least one flowering shoot (?bolt?). By recording the exact plant locations in my quadrat maps, I could determine survival and growth of individual small rosettes, large rosettes, and bolters from year to year.   Quadrat plants were also examined each year for visible indications of M. crucifer adult use.  M. crucifer adult feeding and oviposition activity on C. officinale is detectable as distinctive scars appearing as oval-shaped holes (feeding) or raised bumps (oviposition) on the petioles (De Clerck-Floate et al. 2005). Weevil use appeared similar on H. micrantha and also occurred on bolting stems of both species. The amount of cumulative insect activity per plant was categorized as absent (0 scars), mild (1-9 scars), or severe (?10 scars). Spatial and temporal patterns of M. crucifer use on both plant species in this  64 experiment are described in Chapters 2 and 3, but generally were widespread and persistent for C. officinale and limited to 4.25 m from release points and to Year 0 for H. micrantha.   I returned to the sites on July 27-August 12, 2009, July 27-August 6, 2010, and August 4-August 17, 2011 to estimate fecundity of bolting plants, when seed set and development of both species were complete. Both plant species produce multiple bolts per plant with each flower producing a tetrad of nutlets. The majority of bolts from both plant species were at least 75% intact (Appendix A, Table A.1), and for these bolts I counted the number of tetrads that appeared to contain at least one viable seed (where the pericarp appeared full and not white). I counted the number of bolts per plant, and rated each bolt for ?intactness?, as some had been trampled, severed, or stripped of their seeds by grazing cattle. To investigate the effect of the biocontrol weevil on plant population dynamics, I estimated the biological potential for plants to produce seeds, regardless of trampling damage.  In 2010 and 2011, I also measured the height of intact bolts and created site-specific regressions of square-root-transformed number of tetrads by bolt height. Tetrad numbers for non-intact bolts where heights were available (i.e., when seeds were stripped off of retrievable inflorescences) were estimated using the bolt?s site tetrad-height regression, or if none was available, the height regression from all sites in that year combined. For non-intact bolts in 2009, and those in 2010 and 2011 without height measurements (i.e., bolt was completely severed or damaged so that accurate measurements were not possible), potential tetrad production was estimated using site-level, or when necessary, overall (within-year) per-bolt averages.   Total viable seed counts per bolt per year were calculated by multiplying the number of tetrads by the average number of viable nutlets per tetrad per species. These values were calculated from a subsample of inflorescence branches destructively sampled between July 28-August 6, 2010 from 25 C. officinale plants from 5 study sites and 93 H. micrantha from 11 study sites. The calculated overall average numbers of viable nutlets per tetrad were: C. officinale=2.60 (n=484 tetrads), H. micrantha= 2.16 (n=2209 tetrads).   65 4.2.3. SEEDLING EMERGENCE EXPERIMENT I conducted a second experiment to parameterize rates of seed dormancy, germination, and seedling survival and growth of C. officinale and H. micrantha. Seeds were collected from 21 C. officinale and 15 H. micrantha plants on 4 sites in my study area during August 9-18, 2009, and stored in paper bags at room temperature until October 1-4 2009, when 4300 seeds were randomized within each species, counted, and placed in paper envelopes in the amounts of 10, 100, or 250 seeds. Only nutlets that appeared to contain an embryo within their pericarp, detectable as a noticeable lump upon gentle squeezing, were included in this experiment.   On October 5, 2009, five rows of eight 0.5 ? 0.5 m metal quadrats were nailed onto a grassy area in an aspen grove near one of my study sites. Care was taken to choose an area where existing adult C. officinale and H. micrantha were not present within 2 m of the plot to minimize interference from non-experimental seeds.  Quadrats were ?disturbed? with a garden rake to create open soil conditions typical of invaded areas.  Seeds were sown by hand within each quadrat to one of eight treatment levels: 1) 10 C. officinale seeds, 2) 100 C. officinale seeds, 3) 500 C. officinale seeds, 4) 10 H. micrantha seeds, 5) 100 H. micrantha seeds, 6) 500 H. micrantha seeds, 7) 250 C. officinale and 250 H. micrantha seeds, 8) No seeds.  The experiment was a randomized block design, with each row containing all eight treatments, and treatments were ordered randomly in each row. Immediately after sowing, the dried grassy vegetation within the quadrats was gently shaken to encourage seeds to fall to the soil surface to decrease the chance of seeds being displaced outside of the quadrats by water or animals. The nearest M. crucifer release was approximately 40 m away, but no signs of M. crucifer activity were ever observed on the seedlings or rosettes in this experiment.  The quadrats were revisited on June 4-7, 2010, August 11, 2010, June 9-10, 2011, and August 19-21, 2011.  At each visit, seedlings and rosettes of both study species in each quadrat were counted.  In 2011, the seedling count was separated into seedlings with and without visible cotyledons. In June, 2010, several dead aspen had fallen within the study area, and the resulting deadfall partially obstructed three quadrats. For these three  66 quadrats I visually estimated the area per quadrat covered by deadfall and calculated a ?corrected? number of seeds sown by multiplying the original number sown by the amount of unobstructed area in the quadrat.    4.2.4.  ANALYSIS OF WEEVIL IMPACT TO VITAL RATES Survival, growth, and fecundity data for both plant species were compared between release and non-release sites, and years using generalized linear mixed models (GLMM), through the ?glmer? function in the R package ?lme4?. Additional analyses were performed within release sites for both species, and between site type (?target common? and ?target rare?) for H. micrantha. Response variables of odds of success:fail within quadrats or individual binary response variables used binomal errors and logit link functions, and models with count response variables (i.e., fecundity) used poisson errors and log link functions. Random effects of site, and where possible, nested random effects of site/quadrat or site/quadrat/plant were included to account for spatial and temporal autocorrelation. Fixed effects for between-site comparisons included site release status (i.e., release or non-release site), year, density of conspecific rosettes and bolting plants in the quadrats, and all interactions. Fixed effects for within-site comparisons were distance from release points, presence/absence or rating category of M. crucifer scars. Full models including interactions were fit and were reduced to the simplest models not significantly different than full models (i.e., minimum adequate models). Models were compared to null models with identical random effects using a Chi-squared likelihood ratio test to determine significance of explanatory power. Goodness of fit was calculated as the percent deviance in a null model explained by the fixed effects of the minimum adequate model with identical random effects (DE=deviance explained).   Due to the fine-scale nature of within-site analysis, five H. micrantha plants farther than 15 m from release points were excluded from analysis as distance outliers, and three H. micrantha and one C. officinale with extremely high seed counts were excluded from analysis as fecundity outliers. Within-site fecundity analysis was limited to plants for  67 which at least 50% of bolts were intact enough to be counted (as opposed to being estimated by height regressions or site-bolt averages).   4.2.5. TRANSITION MATRIX CONSTRUCTION, PARAMETERIZATION, AND ANALYSIS Transition matrices were developed for each site in each year using the life cycle diagrams and matrices in Figures 4.1 and 4.2 and parameters in Table 4.1. In addition to site matrices, mean matrices for each site type category for each species in each year were constructed using the mean values of parameters or transitions from the n=3 sites per category.  To account for spatial variation of M. crucifer use scars on nontarget plants, separate H. micrantha matrices were calculated using pooled plants from all release sites within or outside the radius in which nontarget scarring occurred (4.25 m of release points).  Seed production of H. micrantha varied widely. To encapsulate this variance, bolting H. micrantha were separated into ?small bolters? with one or two bolts, and ?large bolters? with three or more bolts (maximum number of bolts per plant =36).  Seed count was significantly different between these two groups (Poisson regression, 95% confidence interval for small bolters = 106-155, large bolters= 725-857, F1,781= 414.4, P<0.0001), and sample sizes were similar (small bolters n=426 plant-years, large bolters n=357 plant-years).     Rates of rosette and bolter survival, transitions between life stages, and transition type were separated into binomial probabilities (i.e., growth given survival) and parameterized using overall site proportions in each year (Morris and Doak 2002, Appendix C, Tables C.1 and C.2).  Average fecundity values per plant per category-year were calculated using a generalized linear model (GLM) with poisson errors and a log link function in R v2.15.2 (R Core Team 2012). Density-dependence was not included in the matrix models to conserve the ability to analyze models analytically, and because excluding intraspecific density-dependence from models may be appropriate in situations where interspecific competition exists (Section 5.4.1).   68  The proportions of C. officinale and H. micrantha seed dormancy, germination, and seedling growth to rosettes were calculated from the seedling emergence experiment under a range of scenarios for seed dormancy and seedling overwinter survival (Appendix B). I chose biologically reasonable assumptions to parameterize seed and seedling dynamics in the transition matrices.  C. officinale has only a 1-2 year seed bank (Van Breemen 1984, Williams et al. 2010), therefore for this species I chose seed parameters under the scenario of maximum seed survival and minimum dormancy. The seed bank dynamics of H. micrantha are unknown, therefore I chose mid-level seed survival and dormancy values in relation to maximum and minimum possible values based on the emergence data in the experiment (Appendix B). Seedling overwinter survival rates are unknown for both species. Based on observations in the field, I used a rate calculated on the assumption that all seedlings without cotyledons in the seedling emergence experiment had overwintered (i.e., they were not new germinated individuals). Species-specific seed and seedling parameters calculated under these assumptions were held constant in all matrices.  However, when all other parameters were held constant, the range of different seed dormancy and seedling overwintering scenarios changed the lambda (population growth rate) value in overall mean matrices by less than 2.5% in either direction.  The proportion of seeds that land on suitable habitat is unknown, and was set consistently at 0.8.   Matrix lambdas, elasticity values, and life table response experiments (LTRE; Caswell 2001) were calculated using the package ?popbio? in R v2.15.2 (Stubben and Milligan 2007, R Core Team 2012).   Lambdas for each species and year combination were calculated for each site and were compared between release and non-release sites using t-tests. Differences in lambdas between years on each site were compared between release and non-release sites with paired t-tests. LTREs were performed for individual sites with the matrix from Year 1-2 being compared to reference matrix Year 0-1.  LTREs comparing release and non-release sites (within ?target common? and ?target rare? sites for H. micrantha) were performed on matrices constructed from the mean transition values from the n=3 sites in the corresponding site category.  The sensitivity of  69 lambda values to parameters affected by M. crucifer releases was determined by calculating the change in lambda occurring when the parameter(s) in question are varied from the values in the mean non-release matrix to those in the mean release matrix.    4.3. RESULTS  The data provided biologically plausible transition matrices, lambdas, LTREs and elasticity values for both C. officinale and H. micrantha release and non-release sites (Tables 4.2 and 4.3). No signs of M. crucifer were ever observed on non-release sites. For both species, most dormant seeds die, and survival and growth of seedlings are low.  Growth rate of surviving C. officinale (but not H. micrantha) seedlings decreased with density in the range of 100-900 seedlings m-2 (Appendix B, Fig. B.1).  Most seedlings in each species did not successfully transition to the rosette stage. Survival became much higher for both species once individuals achieved the small rosette stage. The majority of surviving small rosettes remained small rosettes, but many transitioned to large rosettes or even into bolting plants. Surviving large rosettes of either species were always more likely to transition to bolting plants than remain large rosettes, except for C. officinale on release sites in Year 0-1. Large rosettes transitioned to small rosettes less than 15% of the time. For C. officinale, bolting plants produced up to 1742 viable seeds per plant, with most going into the seed bank, and some germinating into seedlings the following year. Survival of bolting C. officinale was very low (8 of 120 over both transition years), while overall survival of bolting H. micrantha was very high (275 of 285 small bolters, 197 of 204 large bolters over both transition years). Of the 10 H. micrantha small bolters that died, equal numbers were from release and non-release sites, but all release site deaths were in Year 0-1. Notably, no large H. micrantha bolters died on release sites during the 2 transition years recorded. Bolting H. micrantha produced up to 8014 viable seeds per plant with approximately equal proportions becoming seedlings and moving into the seed bank.  Bolting H. micrantha moved regularly to the other bolting size class or to large rosette stage, reflecting a high variability in annual seed production in this species.    70  4.3.1. C. OFFICINALE DYNAMICS Overall, the lambda values for each site in each year for C. officinale (mean ? S.D. = 0.962 ? 0.324, n=6) were lower than those for H. micrantha (1.623 ? 0.335, F1,34= 31.9, P<0.0001, Fig. 4.3). Lambda values for C. officinale ranged from 0.491 to 1.525, and four values each on release and non-release sites were less than 1.0 (the value where population growth is zero).  Evidence for the hypothesis that M. crucifer impacted its target weed at the population level was mixed. C. officinale lambda values from release sites in both years tended to be lower than those on non-release sites (mean ? S.D.: release sites = 0.838 ? 0.296, n=6, non-release sites = 1.086 ? 0.326, n=6), but this difference was not statistically significant (Year 0-1: F1,4 = 1.87, P=0.2431; Year 1-2:  F1,4 = 0.22, P=0.6638). However, the differences in lambda for C. officinale from transition Year 0-1 to Year 1-2 was significantly greater for the 3 release sites than the 3 non-release sites (mean difference ? S.D.: release sites = 0.238 ? 0.090, non-release sites = -0.001 ? 0.040, t4=4.20, P=0.0136), suggesting a suppression in lambda on release sites in Year 0-1 that was not observed on non-release sites.   The lambda from mean matrices for C. officinale in Year 0-1 on non-release sites was 1.102, while on release sites it was 0.751.  LTRE analysis showed that the drop in lambda of 0.351 resulted mostly from reduced growth of small and large rosettes and reduced fecundity on release sites (Fig. 4.4). In Year 1-2, the difference in lambdas was smaller at 0.164, and was mainly a result of small rosette growth, large rosette survival and fecundity. When individual C. officinale site matrices were compared from Year 0-1 to Year 1-2 (n=3 sites per category), contributions to change in lambda in Year 1-2 were higher and more variable on release sites (Fig. 4.5).  Increased rosette growth in Year 1-2 was an important contributor to the increase in lambda in both site categories, but was higher in magnitude on release sites.  Elasticity values for individual C. officinale site matrices were fairly evenly distributed among non-zero transitions with no individual elasticities above 0.20 (Fig. 4.6).  Growth of seedlings to small rosettes, large rosettes to bolters, and fecundity consistently had high elasticity values.    71 Survival of C. officinale small rosettes decreased significantly on release sites, with within-quadrat C. officinale density, and with year on non-release sites (DE=11.9%, ?2= 24.6, P<0.0001, Fig. 4.7).  Growth of surviving small rosettes increased with conspecific density in Year 1-2 (DE=12.5%, ?2= 20.2, P=0.0001), but was not different on release and non-release sites. The bolting probability of surviving, transitioning, small rosettes was not explained by insect releases, density, or year. Survival of large C. officinale rosettes was significantly lower on release sites, but only in Year 0-1, and it was not density-dependent (Fig. 4.8, DE=9.6%, ?2= 11.3, P=0.010). Neither the probability that surviving large rosettes grew to bolters nor the survival rate of bolters were significantly predicted by release, density or year. C. officinale seed production was also not significantly different between M. crucifer release and non-release sites.  The lambda value of a matrix constructed from the mean of all C. officinale parameters on non-release sites was sensitive to decreases in small and large rosette survival (Fig. 4.9). Reducing the survival of small and large rosettes in the non-release matrix to the proportions observed on release sites significantly decreased the lambda value from ?=1.090 to ?=0.744.      4.3.2. H. MICRANTHA DYNAMICS Lambda values for H. micrantha per site per year ranged from 1.135 to 2.458, and were notably never below 1.0 (Fig. 4.3). There was no evidence to support any population-level impact to the nontarget plant by M. crucifer, as insect releases and site type (?target common? or ?target rare?) did not explain H. micrantha lambda values (?2=1.0, P=0.79), or the highly variable difference in lambda values from Year 0-1 to Year 1-2 (difference range = -0.326 - 0.828, mean difference ? S.D.: = 0.168 ? 0.361, n=12, F3,8=0.42, P=0.74). Differences in H. micrantha lambda values from mean matrices between release and non-release sites were positive or negative (Fig. 4.10).  In particular, the lambda on ?target common? sites in Year 0-1 was higher on release sites, even though it likely had the highest and most prolonged density of M. crucifer after releases (i.e., compared to ?target rare? release sites).  No consistent pattern was evident in transitions contributing to differences in H. micrantha lambda values between release  72 and non-release sites in either year.  When individual site matrices were compared from Year 0-1 to Year 1-2 (n=3 sites per category), contributions to change in lambda in Year 1-2 were highly variable with no pattern emerging between release and non-release sites (Fig. 4.11).  Elasticity values for individual H. micrantha were highest for seedling to small rosette growth, small and large rosette transitions and fecundity of large bolters to the seedling stage (Fig. 4.12). Seed bank characteristics had near-zero elasticity values.    Survival, growth and seed production by H. micrantha were never significantly different on M. crucifer release and non-release sites.  Site type (?target common? or ?target rare?) and within-quadrat H. micrantha density were also not significant predictors for H. micrantha survival or growth, except for growth of surviving small rosettes, which decreased with within-quadrat H. micrantha density and varied between years and on ?target common? versus ?target rare? sites.  Seed production by H. micrantha small bolters was not predicted by release site, but was significantly higher in 2010 and 2011 than 2009, and was generally lower on ?target common? sites (from a model with no interactions, DE=5.4%, ?2= 348.0, P<0.0001). Large bolter fecundity was also unaffected by insect releases and increased with year (DE=14.9%, ?2= 7989.7, P<0.0001), but did not vary between ?target common? and ?target rare? sites.    4.3.3. EFFECT OF DISTANCE FROM RELEASE POINTS Spatial analysis within release sites for Year 0 showed no relationship for C. officinale rosette survival or growth to distance from release points. Fecundity of the target weed, however, increased from less than 100 seeds per plant close to release points to roughly 500 seeds per plant 12 m away in the year of release, although this pattern appeared to be driven by only one of the three release sites (Fig. 4.13, ?2=2656.7, P<0.0001, DE=36.6%).  Survival and growth of both H. micrantha rosette stages did not vary significantly with distance from release points. However, there is some evidence that survival of small bolters increased with distance from release, as out of 71 small bolters on release sites, all 5 that died in Year 0-1 (from 4 sites) were within 2.1 m of release points (Fig. 4.14A,  73 DE=10.7%, ?2= 4.0, P=0.0499). Probability of bolting again (given survival and growth) of small H. micrantha bolters increased with distance from release (Fig. 4.14B, DE=20.0%, ?2= 9.6, P=0.002). In contrast, for large bolting H. micrantha, no mortality occurred on release sites, and growth did not vary with distance from release points. I observed heavy adult feeding damage and dieback of H. micrantha bolting stems near release points in Year 0, and fecundity of small bolters increased with distance from release points (Fig. 4.15, DE=3.1%, ?2= 52.5, P<0.0001). However, fecundity of large bolters decreased from release points (DE=0.3%, ?2= 6.5, P=0.0107), which is not consistent with impact from M. crucifer herbivory.    4.3.4. USE SCARS AS INDICATORS OF DEMOGRAPHIC IMPACT The presence of M. crucifer scars was not associated with negative impacts to small C. officinale rosettes, but was associated instead with increased survival, growth and bolting, with variation between the two transition years (Fig. 4.16). In fact, the probability that small rosettes bolted (given survival and growth) was more than double for scarred compared to unscarred plants (Fig. 4.16C, DE=11.3%, ?2= 6.7, P=0.010).   No relationship was found between scarring and survival or growth of large C. officinale rosettes.  Fecundity of bolting C. officinale with M. crucifer scars decreased with scar rating intensity in the year of release, but the pattern was reversed after one year (Fig. 4.17, DE=19.9%, ?2= 3942.7, P<0.0001).   The presence and intensity of M. crucifer scars were not related to survival or probability of growth for H. micrantha small and large rosettes.  However, surviving, growing H. micrantha small bolters were less likely to become large bolting plants when scarred compared to unscarred (mean probability and 95% c.i.: not scarred= 0.63 (0.37-0.84), scarred=0.18 (0.06-0.41); DE=15.1%, ?2= 7.6, P=0.006).   Scarred small bolters that transitioned to rosette stages were more likely to become large rosettes as opposed to small rosettes than unscarred small bolters (mean probability and 95% c.i.: not scarred= 0.33 (0.08-0.73), scarred=0.83 (0.59-0.94); DE=17.6%, ?2= 5.1, P=0.0237). Surviving large bolters that changed life stage were more likely to become small bolters  74 than rosettes if plants were scarred (mean probability and 95% c.i.: not scarred= 0.450 (0.036-0.947), scarred=0.958 (0.321-0.999); DE=22.6%, ?2= 4.1, P=0.0433). Fecundity of H. micrantha small bolters in the year of release was lower for scarred plants than unscarred plants (mean and 95% c.i.: not scarred= 92.7 (54.7-157.2), scarred=77.9 (46.0-131.8); DE=1.75%, ?2= 28.8, P<0.0001), and dropped with severe scar rating (Fig. 4.18A, DE=7.1%, ?2= 118.5, P<0.0001). Fecundity of large H. micrantha was higher in scarred plants (mean probability and 95% c.i.: not scarred= 351.2 (246.6-500.3), scarred=426.1 (299.5-606.2); DE=3.0%, ?2= 58.1, P<0.0001) and increased with scar intensity rating (Fig. 4.18B, DE=3.3%, ?2= 64.1, P<0.0001).     4.4. DISCUSSION  For a biocontrol agent to cause reductions in population vital rates or growth rates (lambda), the incidence and severity of herbivory must be connected with changes in plant survival, growth or fecundity (McClay and Balciunas 2005, Morin et al. 2009). In this chapter I measured and compared demographic parameters of C. officinale and H. micrantha on M. crucifer release sites and non-release sites to detect impacts of biocontrol herbivory on individual vital rates and lambda values. Although there was no statistically significant difference in C. officinale lambdas on release and non-release sites, release sites showed significantly higher recoveries in lambda after one year, suggesting growth rate suppression in the year of release. M. crucifer releases increased mortality of small and large C. officinale rosettes, and these parameters are influential in determining C. officinale population growth rate. Therefore, there is evidence that this biocontrol weevil can be damaging to its target weed at the population level, particularly at high weevil densities. In contrast, there is only weak evidence for mild effects of M. crucifer herbivory on the heaviest-used H. micrantha individuals immediately adjacent to release points on release sites. Notably, this rare and mild damage did not translate to population-level differences in vital rates or lambdas between release and non-release sites or years. To my knowledge, this dissertation is the first study to simultaneously  75 examine demographic impacts of a weed biocontrol agent to target and nontarget plants in its introduced region.     4.4.1. IMPACT OF M. CRUCIFER ON C. OFFICINALE Survival of both small and large C. officinale rosettes was lower on M. crucifer release sites than non-release sites.  Population growth rates generated from matrix models were sensitive to these survival parameters, suggesting that rosette mortality may lead to reductions in C. officinale population size. This rosette mortality effect was not observed in Europe by Williams et al. (2010) and helps explain why M. crucifer has been a fast-acting and effective biocontrol insect for C. officinale in Canada. For example, De Clerck-Floate and Wikeem (2009) found substantial reductions of C. officinale populations in British Columbia only 2 years after release, a time period too short for differences in fecundity to impact population abundance, although in that study, weevil effects were likely interacting with drought.    It is unknown if C. officinale rosettes died because of M. crucifer adult feeding or larval feeding. Root borers damage plants by disrupting vascular and support tissues, even at low levels of infestation (Blossey and Hunt-Joshi 2003, Preisser and Bastow 2005).  Adult and larval M. crucifer feeding may have a greater per capita demographic impact on smaller plants because of a simple mathematical reason; compared to larger plants, at any given amount of root or stem boring, small plants have a lower ratio of cross-sectional area of healthy conductive tissue to area injured by feeding.   M. crucifer preferentially oviposits into larger C. officinale plants, and specifically prefers bolting plants to rosettes (Prins et al. 1992, Schwarzlaender 1997).  However, small plants may be used more often under high insect density, when plants are more likely to be encountered by wandering insects that are likely to be motivated by crowding and hunger to use less acceptable hosts. Insects behave differently under high density (Behmer and Joern 2012). Therefore, ?outbreaks? of M. crucifer may be especially devastating to C. officinale not only because of higher oviposition levels per plant, but also because smaller plants (rosettes) are more likely to be used.  M. crucifer  76 has a high population growth rate when not limited by C. officinale availability, especially in areas with high soil nitrogen content (Van Hezewijk et al. 2008), and may also benefit from escape from its specialized egg and larval parasitoids from the native range (Schwarzlaender 1997).  Therefore, introducing M. crucifer into North America may not fulfill the traditional biocontrol goal of restoring the natural level of herbivory pressure from the native range (Hoddle 2004a), but instead creates a more unstable and responsive insect-plant interaction, with the weevil acting as a ?self-persisting herbicide? (Keane and Crawley 2002).  Such a system may be characterized by transient outbreak dynamics at the patch level, and metapopulation dynamics at the landscape level, dynamics that have been documented in other overexploiting insects (Myers and Post 1981), including the specialist weevil Hadramphus spinipennis and its host plant Aciphylla dieffenbachii in New Zealand (Sch?ps 2002).  C. officinale exhibits metapopulation dynamics in both the native and introduced ranges (van der Meijden et al. 1992, De Clerck-Floate 1996). Further investigation of the differences between M. crucifer impacts at outbreak and low or ?equilibrium? densities in the introduced range is necessary, particularly to predict the stability of the weed-weevil dynamics once the abundance of the invasive plant has been reduced.   The population-level effect of increased rosette mortality by M. crucifer herbivory may be in addition to potential decrease in fecundity of the target weed. M. crucifer causes a 30-35% decrease in per capita C. officinale seed production in its native range (Prins et al. 1992, Williams et al. 2010), an effect that was not detected in this study. My data showed no decrease in C. officinale fecundity on M. crucifer release sites compared to non-release sites. However, on release sites, heavily scarred C. officinale did have reduced fecundity compared to unscarred plants. Regardless, this short-term study may not have fully detected the effect of M. crucifer on C. officinale seed production. The weevils were released on June 4, 2009, approximately 4-5 weeks after established populations of overwintered adult M. crucifer would have emerged to begin feeding and ovipositing in their target plants. It is possible that my releases were made too late for adult or larval feeding to affect within-season fecundity in bolting plants.  In comparison, eggs or larvae in established M. crucifer populations may overwinter in C.  77 officinale roots, and plants can incur M. crucifer damage for several years before bolting, resulting in a cumulative effect of herbivory over multiple years. Therefore, further work in established populations of C. officinale and M. crucifer is necessary to quantify the effect of weevil herbivory on seed production in North America. Increased rosette mortality may impose the main challenge to conducting such an experiment, since finding populations where bolting C. officinale plants exist years after M. crucifer release is difficult (R. De Clerck-Floate, personal communication).     This study adds valuable information to previous demographic and population-modeling work by Williams et al. (2010) that demonstrated a decrease in fecundity, but not survival, in C. officinale infested with M. crucifer in its native range. The resulting predictions by Maron et al. (2010) and Williams et al. (2010) that M. crucifer would not be an effective biocontrol agent in North America were based on the assumption that demographic impacts by the weevil would be identical in the native and introduced ranges. Here I demonstrate that such an assumption is not reliable. There are many variables in the two environments that may influence plant and insect populations and their interaction, such as plant competitive dynamics, climate, and soil nutrient levels (van Hezewijk et al. 2008, Williams et al. 2010) and microbial communities (Klironomos 2003).  It is possible a more realistic model for the introduced range would use the C. officinale survival and growth parameters from this study from release sites and fecundity levels from non-release sites decreased 30% (Prins et al. 1992, Maron et al. 2010, Williams et al. 2010). However, a comparison of C. officinale data from sites with and without naturally-occurring M. crucifer populations is required to test this idea.    Although lambda values in non-release and release sites were not significantly different for C. officinale, the data trended in the direction of impact, and the power to detect an effect was low with only 3 sites per group per year.  Further study with larger sample sizes will be necessary to better determine any difference in C. officinale population growth rates with and without M. crucifer.  Interestingly, the recovery level of C. officinale from transition years 0-1 to 1-2 was significantly higher on release sites and was driven mainly by increased rosette growth.  This pattern may have been a result of a  78 drop in peak weevil density (and thus feeding and oviposition) 1 year after release (Chapter 2), meaning M. crucifer herbivory was more severe in the year of release. My release size of 300 weevils per site was 3 times higher than the standard release size for M. crucifer biological control (De Clerck-Floate et al. 2005, De Clerck-Floate and Wikeem 2009), and may have exceeded the equilibrium level the C. officinale on sites could support. Additionally, Year 0 was relatively dry, with 46-60% of April-July precipitation as Years 1 and 2 (192.5, 414.2, and 321.5 mm in 2009, 2010, and 2011 respectively, Agriculture and Agri-Food Canada Lethbridge Research Station weather data).  C. officinale is water-limited, especially early in its life cycle (de Jong and Klinkhamer 1988a), and while 2009 was not a severe drought year (mean ? S.D. April-July precipitation from 1998-2009 = 227.3 ? 87.7, n=12), the stress imposed by root-mining damage may be lessened with increasing availability of moisture (Blossey and Hunt-Joshi 2003).  This scenario seems less likely as C. officinale plants in the present study were in competition with rangeland grasses and Canada thistle (Cirsium arvense (L.) Scop), both of which were noticeably larger or more abundant in 2010 and 2011 compared to 2009.  The relative effects of the combination of increased moisture and interspecific competition in Years 1 and 2 compared to a drop in M. crucifer density after Year 0 are unknown, but either or both could be responsible for the increase in C. officinale population growth rate on release sites.  It is interesting that small C. officinale rosettes with M. crucifer scars were more likely to bolt than unscarred plants. This result may represent a direct compensatory response by the plant to M. crucifer herbivory damage through phenotypic plasticity (Wesselingh et al. 1997).  Herbivory by M. crucifer may be an important selection pressure for smaller C. officinale threshold flowering size in the native range (Wesselingh et al. 1997, Williams 2009), but further research is necessary to determine if individual C. officinale rosettes respond directly within-season to herbivory damage.   4.4.2. IMPACT OF M. CRUCIFER ON H. MICRANTHA The data did not reveal any population-level differences in H. micrantha vital rates or lambda values between M. crucifer release sites and non-release sites.  H. micrantha  79 lambdas were highly variable but always above 1.0, and their values and rates of change were not explained by insect releases.  No clear patterns emerged in terms of transition contributions to changes in lambda, except perhaps fecundity on ?target common? sites. The variability observed may have been the result of site characteristics such as microclimate, competitive dynamics, and grazing regimes, as the amount of disturbance and trampling damage could affect growth rates. H. micrantha lambda has high elasticity to growth and fecundity transitions.   As described in Chapters 2 and 3, M. crucifer use of H. micrantha on my release sites during the 3 years of the experiment was almost totally limited to the year of release and within 4.25 m of release points. This spatial variation in use within release sites allowed me to compare demographic characteristics with distance from release. There is some evidence for a reduction in H. micrantha survival, bolting, and seed production near release points in the year of release (Figs. 4.14 and 4.15). In contrast to the situation with C. officinale, the reduction in H. micrantha fecundity near field release points appeared to be a direct result of heavy feeding by M. crucifer adults on the tips of flowering stems. Particularly, all 5 small bolters that died on release sites in Year 0-1 were within 2.1 m of release points (Fig. 4.14).  Perhaps small bolters were large enough to be found by a number of wandering insects thus leading to heavy feeding and oviposition, but small enough to be vulnerable from the effects of the herbivory.  However, any individual effects that occurred in H. micrantha close to release points did not translate to the population-level estimates of vital rates or lambda values. Furthermore, the lambda corresponding to a transition matrix calculated using only H. micrantha plants within 4.25 m of release points pooled from 6 release sites was not lower than those calculated for non-release sites (Fig. 4.3G).    The lack of impact of M. crucifer releases to H. micrantha plants may indicate that the nontarget plant?s current traits and adaptations make it well-suited to withstand this type of herbivory.  Plants with longer adult life spans can have sufficient energy reserves to enable compensatory re-growth after herbivory damage compared to shorter-lived species (Maron and Crone 2006). The architecture and morphology of the native plant  80 appears to be well suited for compensatory growth in response to M. crucifer damage. Similar to C. officinale, H. micrantha has a taproot, however, I observed during dissections that the taproot of the native plant is hollow and noticeably woodier than that of C. officinale. The bulk of M. crucifer larval feeding in the native plant is thus restricted to the fleshy base of shoots and within the pith of the woody caudices emerging from the root. The difference in availability of fleshy tissue in H. micrantha may mean that most disruption of conductive tissues from larval feeding occurs at the level of individual shoots rather than impacting the entire plant (R. De Clerck-Floate and H. Catton, unpublished data).  Scaling down the idea of refuges from herbivory to the within-plant scale, the modular architecture of H. micrantha may provide refuges from M. crucifer impact that buffer against whole-plant mortality or fecundity losses. .  I observed two specialist native herbivores on H. micrantha during my studies. H. micrantha experiences damage at the shoot level by an unnamed microlepidopteran (Lepidoptera: Choreutidae, Caloreas spp.).  Bud-feeding larvae of native herbivore prevented growth and flowering of individual reproductive H. micrantha shoots (rough estimate= 1-10% per site), and therefore H. micrantha populations have evolved to be robust to some seed loss. Alternatively, larvae of a second native herbivore, the Police Car Moth (Lepidoptera: Noctuidae, Gnophaela vermiculata Grote), can completely defoliate H. micrantha plants and resulted in the disappearance of a whole study population two years after the end of the rangeland experiment (H. Catton, unpublished results).  The effects of existing plant population dynamics, and individual morphological and architectural characteristics on tolerance to herbivory by biocontrol agents warrants further investigation, and should be being considered when predicting potential impacts to both target and nontarget plants in biocontrol.   In summary, the lack of translation of damage to individual H. micrantha plants from M. crucifer use to differences in population vital rates and lambda values is not surprising for several reasons. First, there were three types of refuges from M. crucifer use for nontarget plants. H. micrantha use was temporary and localized, leading to temporal and spatial refuges from insect use (Chapters 2 and 3). Oviposition into H. micrantha was less frequent and less severe than for C. officinale thus generating probabilistic refuges  81 from heavy use (Chapter 2). These refuges served to ?dilute? the influence of damaged plants on calculated population-level H. micrantha vital rates. Second, H. micrantha may have morphological, architectural and population traits that are well-suited to withstand the type and level of herbivory damage exhibited by M. crucifer, minimizing the effect of the biocontrol insect on the nontarget plant at both the individual plant and population level.   4.4.3. IMPLICATIONS FOR IMPACT MONITORING The activities of endophagous insects are difficult to monitor non-destructively (Hunter 2001). Previous efforts for M. crucifer have used distinctive petiole scars as a non-destructive indication of feeding and ovipostion in C. officinale (De Clerck-Floate et al. 2005). However, with no information available on non-destructively monitoring M. crucifer activity on other plant species, it could be assumed that using a similar method on nontarget plants would be as reliable as for the target weed.  My data suggests otherwise. The presence of M. crucifer use scars on C. officinale is a reliable indication of oviposition in target plants, and the absence of scars on H. micratha is a reliable indication of no oviposition in the nontarget plant (Chapter 2).  Scars on H. micrantha may be a reliable within-season indicator of high weevil density (Chapter 3), but not oviposition in scarred plants (Chapter 2).    In this Chapter, I demonstrated that the presence or absence of M. crucifer scars do not predict survival and growth of individual plants of either species. Scars on C. officinale had different effects on fecundity depending on the year; for example heavily scarred C. officinale had reduced fecundity in the year of release but increased fecundity one year later (Fig. 4.17).  This mixed effect is likely the result of the confounding effects of plant size and weevil density.  M. crucifer has a preference for large plants (Prins et al. 1992, Schwarzlaender 1997), and these plants produce high numbers of seed.  When weevil density is low, the weevil?s plant size preference is expressed through the herbivory pattern in the patch, therefore plants with higher fecundities may also have higher levels of scarring.  My scar ratings were simply a count per plant, therefore the larger the plant (of either species), the more leaf and stem material there was to assess for scars.  82 Additionally, my scar assessments were made in July, which would not capture weevil activity occurring after those times; for example newly emerged F1 M. crucifer often oviposit in C. officinale rosettes in the fall before hibernation (Schwarzlaender 1997). Collectively, use scars appear to be informative for estimating M. crucifer occurrence, dispersal, and maximum density at the patch-level (Chapters 2 and 3), but should not be use to predict demographic impacts in target or nontarget plants.   An important distinction must be made between identifying the occurrence of population-level impact, and studying the mechanisms causing such impact. While monitoring herbivory patterns using scars is valuable for studying mechanisms, the occurrence of impact should be monitored at the population level through plant abundance and demographic data (McClay 1995, Delfosse 2005, Morin et al. 2009). Given the elasticity values of the transition matrices calculated in this study, practitioners would be best served to monitor C. officinale seedling growth, rosette survival and bolting plant fecundity to predict or assess population-level impacts (Fig. 4.6). It is perhaps tempting to focus only on large plants, but the results in this chapter indicate that seedling growth and rosette survival are both very influential in C. officinale population dynamics.  For H. micrantha, I recommend evaluation of the same life stages, along with large bolter survival, as similar influential transitions exist for the nontarget plant.    This study demonstrates the importance of considering spatial scale when collecting demographic data. For example, in a situation where herbivory negatively affects individual plants and use is spatially aggregated (i.e., around a release point), calculated vital rates for the population depend very much on the radius of plants assessed around release points. One important assumption in matrix models is that vital rates are uniform within the modeled patch or population of plants.  Therefore in this case, the larger the evaluation radius, the more ?diluted? the effect of spatially-aggregated individual impacts.  Had H. micrantha incurred drastic individual injury from M. crucifer use in this study, a smaller evaluation radius during data collection could have led to the conclusion that H. micrantha vital rates or even lambda values were being impacted  83 heavily by M. crucifer. The evaluation radii I chose at each site were limited by the logistical restraints of the time-consuming nature of demographic data collection, and I was fortunate in that the chosen spatial scale captured the outer bound of H. micrantha use around all my release points.  Therefore, a key message emerging from this chapter is that studies of within-patch dispersal use patterns by a biocontrol agent are important for determining an appropriate spatial scale for demographic data collection that accounts for heterogeneity in spatial herbivory patterns.   84 Table 4.1. Symbols and definitions for annual probability and fecundity parameters used in C. officinale and H. micrantha matrix models.  Symbol      Definition  sF   probability a new seed survives hF   probability a surviving new seed lands on suitable habitat gF   probability a surviving new seed on suitable habitat germinates sD  probability a seed in the seed bank survives gD  probability a surviving seed in the seed bank grows j  probability a surviving, germinating seed jumps to small rosette sG   probability a seedling survives gG   probability a surviving seedlings grows lG  probability a surviving, growing seedling grows to a large rosette sSR  probability a small rosette survives gSR  probability a surviving small rosette grows bSR  probability a surviving, growing small rosette bolts sLR  probability a large rosette survives gLR   probability a surviving large rosette transitions to another stage bLR   probability a surviving, transitioning large rosette bolts lLR   probability a surviving, transitioning, bolting large rosette grows to a large    bolter     C. officinale sB   probability a bolter survives gB   probability a surviving bolter reverts to a large rosette F   number of viable seeds produced per bolter  H. micrantha sSB   probability a small bolter survives gSB   probability a surviving small bolter transitions to another stage bSB   probability a surviving transitioning small bolter bolts lSB   probability a surviving transitioning, non-bolting small bolter grows to large    rosette FSB   number of viable seeds produced per small bolter sLB   probability a large bolter survives gLB   probability a surviving large bolter transitions to another stage bLB   probability a surviving transitioning large bolter bolts FLB   number of viable seeds produced per large bolter    85 Table 4.2. Mean (? standard deviation) transition values for C. officinale categorized by year and presence or absence of a M. crucifer release in Year 0 (n=3 sites per category).    Non-release, Year 0-1   seed bank seedling small rosette large rosette bolter  seed bank  0.0022 0 0 0 209.0 ? 158.9  seedling  0.2707 0.1173 0 0 84.0 ? 63.9  sm. rosette  0.0003 0.0193 ? 0.0026 0.3734 ? 0.1299 0.0893 ? 0.0341 0.0841 ? 0.0640  large rosette  0 0.0029 ? 0.0026 0.3117 ? 0.2025 0.3047 ? 0.1606 0  bolter  0 0 0.0609 ? 0.0267 0.3957 ? 0.2724 0.0256 ? 0.0444   Release, Year 0-1   seed bank seedling small rosette large rosette bolter  seed bank  0.0022 0 0 0 129.5 ? 134.9  seedling  0.2707 0.1173 0 0 52.1 ? 54.2  sm. rosette  0.0003 0.0214 ? 0.0013 0.2430 ? 0.0690 0.0222 ? 0.0385 0.0521 ? 0.0543  large rosette  0 0.0007 ? 0.0013 0.0733 ? 0.0609 0.2833 ? 0.1041 0  bolter  0 0 0.0701 ? 0.0356 0.1833 ? 0.2128 0.0303 ? 0.0525   Non-release, Year 1-2   seed bank seedling small rosette large rosette bolter  seed bank  0.0022 0 0 0 166.7 ? 146.9  seedling  0.2707 0.1173 0 0 67.1 ? 59.1  sm. rosette  0.0003 0.0195 ? 0.0011 0.2249 ? 0.0911 0 0.0671 ? 0.0591  large rosette  0 0.0027 ? 0.0011 0.4296 ? 0.2273 0.3516 ? 0.2416 0.0222 ? 0.0385  bolter  0 0 0.0500 ? 0.0608 0.5055 ? 0.2884 0.0667 ? 0.1155   Release, Year 1-2   seed bank seedling small rosette large rosette bolter  seed bank  0.0022 0 0 0 106.3 ? 54.7  seedling  0.2707 0.1173 0 0 42.7 ? 22.0  sm. rosette  0.0003 0.0178 ? 0.0059 0.3279 ? 0.1959 0 0.0428 ? 0.0220  large rosette  0 0.0044 ? 0.0059 0.2552 ? 0.0889 0.2037 ? 0.2625 0  bolter  0 0 0.0452 ? 0.0783 0.5463 ? 0.3220 0.0513 ? 0.0888    86 Table 4.3. Mean (? standard deviation) transition values for H. micrantha separated by year, type of site (?target common? or ?target rare?) and presence or absence of a M. crucifer release in Year 0 (n=3 sites per category).    Non-release, ?Target Common?, Year 0-1   seed bank seedling small rosette large rosette small bolter large bolter  seed bank  0.0643 0 0 0 21.0 ? 5.9 84.4 ? 43.9  seedling  0.0690 0.3082 0 0 20.1 ? 5.6 80.6 ? 42.0  small rosette  0.0001 0.1394 0.4781 ? 0.1255 0.1333 ? 0.2309 0.0784 ? 0.0682 0.0807 ? 0.0420  large rosette  0 0 0.3064 ? 0.1217 0.4452 ? 0.3042 0.3083 ? 0.1876 0.0303 ? 0.0525  small bolter  0 0 0.0303 ? 0.0525 0.1976 ? 0.0536 0.5000 ? 0.1250 0.5212 ? 0.4226  large bolter  0 0 0 0.2000 ? 0.2000 0.1333 ? 0.0144 0.3485 ? 0.3027   Release, ?Target Common?, Year 0-1   seed bank seedling small rosette large rosette small bolter large bolter  seed bank  0.0643 0 0 0 34.3 ? 25.1 288.4 ? 380.0  seedling  0.0690 0.3082 0 0 32.7 ? 23.9 275.4 ? 362.4  small rosette  0.0001 0.1394 0.5958 ? 0.0360 0.0351 ? 0.0608 0.0750 ? 0.0178 0.2757 ? 0.3628  large rosette  0 0 0.1428 ? 0.1117 0.3830 ? 0.1097 0.2192 ? 0.1072 0.0556 ? 0.0962  small bolter  0 0 0.0739 ? 0.0397 0.3450 ? 0.0203 0.4282 ? 0.0627 0.3651 ? 0.1755  large bolter  0 0 0 0.2018 ? 0.1180 0.1679 ? 0.1754 0.5794 ? 0.0836     87 Table 4.3. (continued) Mean (? standard deviation) transition values for H. micrantha separated by year, type of site (?target common? or ?target rare?) and presence or absence of a M. crucifer release in Year 0 (n=3 sites per category).    Non-release, ?Target Common?, Year 1-2   seed bank seedling small rosette large rosette small bolter large bolter  seed bank  0.0643 0 0 0 36.2 ? 3.9 121.6 ? 68.5  seedling  0.0690 0.3082 0 0 34.6 ? 3.7 116.1 ? 65.5  small rosette  0.0001 0.1394 0.4028 ? 0.1684 0.1421 ? 0.1656 0.0346 ? 0.0037 0.1162 ? 0.0655  large rosette  0 0 0.2569 ? 0.1147 0.3564 ? 0.1786 0.1833 ? 0.2754 0.0476 ? 0.0825  small bolter  0 0 0.0486 ? 0.0434 0.2475 ? 0.2176 0.4056 ? 0.2275 0.3810 ? 0.5408  large bolter  0 0 0 0.2222 ? 0.1925 0.3528 ? 0.2381 0.5238 ? 0.5017    Release, ?Target Common?, Year 1-2   seed bank seedling small rosette large rosette small bolter large bolter  seed bank  0.0643 0 0 0 26.7 ? 6.0 272.9 ? 272.6  seedling  0.0690 0.3082 0 0 25.5 ? 5.7 260.6 ? 260.3  small rosette  0.0001 0.1394 0.5820 ? 0.2257 0.0556 ? 0.0962 0.0671 ? 0.0690 0.2608 ? 0.2605  large rosette  0 0 0.2145 ? 0.0849 0.4899 ? 0.2094 0.3182 ? 0.1720 0.1071 ? 0.1288  small bolter  0 0 0.0332 ? 0.0033 0.3687 ? 0.1862 0.4735 ? 0.0883 0.2857 ? 0.2575  large bolter  0 0 0 0.0581 ? 0.0504 0.1667 ? 0.1909 0.6071 ? 0.3763    88  Table 4.3. (continued) Mean (? standard deviation) transition values for H. micrantha separated by year, type of site (?target common? or ?target rare?) and presence or absence of a M. crucifer release in Year 0 (n=3 sites per category).    Non-release, ?Target Rare?, Year 0-1   seed bank seedling small rosette large rosette small bolter large bolter  seed bank  0.0643 0 0 0 32.5 ? 3.2 166.7 ? 60.4  seedling  0.0690 0.3082 0 0 31.1 ? 3.1 159.2 ? 57.6  small rosette  0.0001 0.1355 ? 0.0067 0.3697 ? 0.2028 0.0238 ? 0.0412 0.0311 ? 0.0031 0.1594 ? 0.0577  large rosette  0 0.0039 ? 0.0067 0.1880 ? 0.0118 0.3883 ? 0.1590 0.1418 ? 0.0567 0.1389 ? 0.1273  small bolter  0 0 0.0222 ? 0.0385 0.3407 ? 0.0769 0.3025 ? 0.1317 0.1806 ? 0.0636  large bolter  0 0 0 0.1722 ? 0.1516 0.4859 ? 0.1062 0.6389 ? 0.1273    Release, ?Target Rare?, Year 0-1   seed bank seedling small rosette large rosette small bolter large bolter  seed bank  0.0643 0 0 0 36.1 ? 17.3 115.6 ? 100.9  seedling  0.0690 0.3082 0 0 34.5 ? 16.5 110.4 ? 96.3  small rosette  0.0001 0.1394 0.5858 ? 0.2975 0.0957 ? 0.0835 0.1853 ? 0.0114 0.1105 ? 0.0964  large rosette  0 0 0.1401 ? 0.0446 0.4603 ? 0.1989 0.2698 ? 0.1997 0.1923 ? 0.2692  small bolter  0 0 0.0142 ? 0.0246 0.2201 ? 0.1761 0.3492 ? 0.0727 0.1026 ? 0.1776  large bolter  0 0 0 0.1650 ? 0.2574 0.1746 ? 0.1074 0.3718 ? 0.3271   89 Table 4.3. (continued) Mean (? standard deviation) transition values for H. micrantha separated by year, type of site (?target common? or ?target rare?) and presence or absence of a M. crucifer release in Year 0 (n=3 sites per category).    Non-release, ?Target Rare?, Year 1-2   seed bank seedling small rosette large rosette small bolter large bolter  seed bank  0.0643 0 0 0 50.7 ? 18.7 215.9 ? 73.1  seedling  0.0690 0.3082 0 0 48.4 ? 17.9 206.1 ? 69.8  small rosette  0.0001 0.1239 ? 0.0268 0.4591 ? 0.2072 0.0905 ? 0.1014 0.0485 ? 0.0179 0.2063 ? 0.0699  large rosette  0 0.0155 ? 0.0268  0.0706 ? 0.0286 0.2702 ? 0.1382 0.2134 ? 0.1077 0.0417 ? 0.0722  small bolter  0 0 0.0139 ? 0.0241 0.4179 ? 0.2127 0.2618 ? 0.1771 0.1806 ? 0.0636  large bolter  0 0 0 0.1976 ? 0.0536 0.5248 ? 0.2670 0.7407 ? 0.0160    Release, ?Target Common?, Year 1-2   seed bank seedling small rosette large rosette small bolter large bolter  seed bank  0.0643 0 0 0 39.3 ? 8.6 219.9 ? 149.0  seedling  0.0690 0.3082 0 0 37.5 ? 8.2 210.1 ? 142.3  small rosette  0.0001 0.1394 0.6146 ? 0.1042 0.0357 ? 0.0619 0.0746 ? 0.0564 0.2102 ? 0.1424  large rosette  0 0  0.2170 ? 0.1591 0.4816 ? 0.1576 0.3177 ? 0.1580 0.? 0  small bolter  0 0 0.0133 ? 0.0231 0.3044 ? 0.0254 0.3038 ? 0.2244 0.3689 ? 0.1396  large bolter  0 0 0 0.1426 ? 0.1656 0.3415 ? 0.2156 0.6311 ? 0.1396    90       Figure 4.1. Life cycle diagram and transition matrix used for C. officinale. Solid lines indicate plant transitions, dotted lines indicate fecundity transitions.  Seedling((G)(Bolter((B)( Large(Rose4e((LR)(Small(Rose4e((SR)(Seed(Bank((D)(Seed(Bank((D)( Seedling((G)( Small(Rose4e((SR)( Large(Rose4e((LR)( Bolter((B)(Seed(Bank((D)( sD((17gD)( 0( 0( 0( F(sF(hF((17gF)(Seedling((G)( sD(gD((17jD)( sG((17gG)( 0( 0( F(sF(hF(gF((17jF)(Small(Rose4e((SR)( sD(gD(jD( sG(gG((17lG)( sSR((17gSR)( sLR(gLR((17bLR)( F(sF(hF(gF(jF(Large(Rose4e((LR)( 0( sG(gG(lG( sSR(gSR((17bSR)( sLR((17gLR)( sB(gB(Bolter((B)( 0( 0( sSR(gSR(bSR( sLR(gLR(bLR( sB((17gB)( 91       Figure 4.2. Life cycle diagram and transition matrix used for H. micrantha. Solid lines indicate plant transitions, dotted lines indicate fecundity transitions.  Seedling((G)(SBoll(trlLea((St)(Roage(sr4eme((Rs)(SBoll(sr4eme((Ss)(Seed(tonk((D)(Roage(trlLea((Rt)(Seed(Bank((D)( Seedling((G)( Small(Rose5e((SR)( Large(Rose5e((LR)( Small(Bolter((SB)( Large(Bolter((LB) ((Seed(Bank((D)( sD((18gD)( 0( 0( 0( FSB(sF(hF((18gF)( FLB(sF(hF((18gF)(Seedling((G)( sD(gD((18j)( sG((18gG)( 0( 0( FSB(sF(hF(gF((18j)( FLB(sF(hF(gF((18j)(Small(Rose5e((SR)( sD(gD(j( sG(gG(18lG)( sSR((18gSR)( sLR(gLR((18bLR)( sSB(gSB((18bSB)((18lSB)(+(FSB(sF(hF(gF(j(FLB(sF(hF(gF(j(Large(Rose5e((LR)( 0( sG(gG(lG( sSR(gSR((18bSR)( sLR((18gLR)( sSB(gSB((18bSB)(lSB( sLB(gLB((18bLB)(Small(Bolter((SB)( 0( 0( sSR(gSR(bSR( sLR(gLR(bLR(18lLR)( sSB((18gSB)( sLB(gLB(bLB(Large(Bolter((LB)(( 0( 0( 0( sLR(gLR(bLR(lLR( sSB(gSB(bSB( sLB((18gLB)( 92     Figure 4.3.  Lambda values for C. officinale and H. micrantha in year 0-1 and year 1-2 on non-release (control) sites, and M. crucifer release sites. Average matrices were generated from the mean values for each transition in the n=3 sites per category. C. officinale and H. micrantha on ?target common? sites were on the same sites, while H. micrantha on ?target rare? sites were on separate sites.   0.61.01.41.82.22.60-1 1-2LambdaA) C. officinaleNon-release sitesindividual sitesaverage0-1 1-2B) C. officinaleRelease sites0-1 1-2C) H. micranthaTarget CommonNon-release sites0-1 1-2D) H. micranthaTarget CommonRelease sites0-1 1-2E) H. micranthaTarget RareNon-release sites0-1 1-2F) H. micranthaTarget RareRelease sites0-1 1-2G) H. micranthaAllRelease sitesoutside 4.25minside 4.25mTransition Year 93     Figure 4.4. Contributions of transitions to differences in C. officinale lambda values from non-release to release sites, as found by a life table response experiment. Matrix elements are the means from the 3 sites within the release and year categories.  Lambda values for the average matrices were: Year 0-1: Non-release=1.102, Release=0.751, Difference=0.351.  Year 1-2: Non-release=1.112, Release=0.948, Difference=0.164.  Code for life stages: D=seed bank (i.e., dormant), G=seedling (i.e., germinated), SR=small rosette, LR=large rosette, B=bolter.     Year 0-1-0.15-0.10-0.050.000.05G-SRG-LRSR-SRSR-LRSR-BLR-SRLR-LRLR-BB-DB-GB-SRB-LR B-BContributions to changes in lambdaLambda nonrel= 1.102, rel=0.751Year 1-2G-SRG-LRSR-SRSR-LRSR-BLR-SRLR-LRLR-BB-DB-GB-SRB-LR B-BHT LTRE of average non-release (ref) and release (trt) matrices in each yearTransitionsLambda nonrel= 1.112, rel=0.948 94      Figure 4.5. Mean ( ?  standard deviation) contributions of each transition type to the change in C. officinale lambda from year 0-1 to 1-2 on 3 release and non-release sites. Values were obtained from a life table response experiment. Code for life stages: D=seed bank (i.e., dormant), G=seedling (i.e., germinated), SR=small rosette, LR=large rosette, B=bolter.    Non-release sites-0.10-0.050.000.050.100.150.20G-SRG-LRSR-SRSR-LRSR-BLR-SRLR-LRLR-BB-DB-GB-SRB-LR B-BRelease sitesG-SRG-LRSR-SRSR-LRSR-BLR-SRLR-LRLR-BB-DB-GB-SRB-LR B-BContributions to changes in lambdaLTRE results for HT changes from yr0-1 to 1-2 95    Figure 4.6. Mean ( ?  standard deviation) elasticity values for C. officinale transitions on release sites and non-release sites in Year 0-1 and Year 1-2 after release of M. crucifer on release sites (n=3 sites per category). Code for life stages: D=seed bank (i.e., dormant), G=seedling (i.e., germinated), SR=small rosette, LR=large rosette, B=bolter.    D-G G-G G-SRG-LRSR-SRSR-LR SR-BLR-SRLR-LR LR-B B-D B-G B-BNon-release, Year 0-10.00.10.20.3D-G G-G G-SRG-LRSR-SRSR-LR SR-BLR-SRLR-LR LR-B B-D B-G B-BRelease, Year 0-10.00.10.20.3D-G G-G G-SRG-LRSR-SRSR-LR SR-BLR-SRLR-LR LR-B B-D B-G B-BNon-release, Year 1-20.00.10.20.3D-G G-G G-SRG-LRSR-SRSR-LR SR-BLR-SRLR-LR LR-B B-D B-G B-BRelease, Year 1-20.00.10.20.3Elasticities for HT on release and non-release sites each yearTransitionsElasticities 96      Figure 4.7. Survival of small C. officinale rosettes (4-9 leaves) in 1x1 m quadrats with various densities of C. officinale in transition year 0-1 and 1-2 after M. crucifer releases.  Symbols are raw proportion values for each quadrat and are jittered slightly to reduce overlay. Lines represent a generalized linear mixed model with survival as the response variable, insect release, year, within-quadrat C. officinale density, and the interaction between release site and year as fixed effects, and site/quadrat nested as random effects (?2=24.6, P<0.0001, DE=11.9%). The number of quadrats for each category were:  Release sites Year 0-1 n=48, Non-release sites Year 0-1 n=26, Release sites Year 1-2 n=31, Non-release sites Year 1-2 n=27.     0.00.20.40.60.81.00 5 10 15 20 25Survival probabilityC. officinale m-2Release sites Year 0-1Non-release sites Year 0-1Release sites Year 1-2Non-release sites Year 1-2 97    Figure 4.8. Mean ( ?  95% c.i.) survival of large C. officinale rosettes (>10 leaves or longest leaf >30 cm) in 1x1 m quadrats, in transition years 0-1 and 1-2 after M. crucifer releases.  Values are back-transformed predictions from a generalized linear mixed model with survival as the response variable, insect release, year, and their interaction, as fixed effects, and site/quadrat nested as random effects (?2=11.3, P=0.010, DE=9.6%). The number of quadrats for each category were:  Release sites Year 0-1 n=26, Non-release sites Year 0-1 n=29, Release sites Year 1-2 n=17, Non-release sites Year 1-2 n=26.    Year 0-1 Year 1-20.00.20.40.60.81.0Survival probabilityNo ReleaseRelease 98         Figure 4.9. Sensitivity of C. officinale lambda values of a mean matrix of 3 non-release sites in Year 0-1 to a reduction in small and large rosette survival rates observed on M. crucifer release sites in that year. All vital rates were held constant while small and large rosette survival were varied singly from the value observed on non-release sites to the value observed on release sites in Year 0-1.  The dotted line represents small and large rosette survival simultaneously being decreased through their range of values. Models specifications: small varied, large constant: y= 0.443x + 0.763; large varied, small constant: y= 0.331x + 0.829; small varied, large decreased through its range: y= 0.624x + 0.626.  0.70.80.91.01.10.2 0.3 0.4 0.5 0.6 0.7 0.8Varied parameter valueLambdasmall varied, large=0.79large varied, small=0.746small varied, large decreasing from 0.79 to 0.5 99     Figure 4.10. Contributions of each transition type to the change in H. micrantha lambda value from non-release to release sites. Matrices for this analysis were calculated using mean transitions on release or non-release sites within each ?target common? or ?target rare? category for each year (n=3 sites per category).  Lambda values for the ?averaged? matrices were: Year 0-1, Target common, Release=1.789, Non-release=1.522,  Difference=0.267.  Year 0-1, Target rare: Release=1.402, Non-release=1.679,  Difference=-0.277. Year 1-2, Target common: Release=1.661, Non-release=1.682, Difference=-0.021.  Year 1-2, Target rare: Release=1.748, Non-release=1.806, Difference=-0.058.  Code for life stages: D=seed bank (i.e., dormant), G=seedling (i.e., germinated), SR=small rosette, LR=large rosette, SB=small bolter, LB=large bolter.    Target Common Year 0-1-0.2-0.10.00.10.20.3G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBTarget Rare Year 0-1-0.2-0.10.00.10.20.3G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBTarget Common Year 1-2-0.2-0.10.00.10.20.3Target Rare Year 1-2-0.2-0.10.00.10.20.3LTRE for release and non-release (ref) average matrices by yearTransitionsContributions to difference in lambda 100          Figure 4.11.  Mean ( ?  standard deviation) contributions of each transition type to the difference in H. micrantha lambda from year 0-1 to 1-2 on release and non-release ?target common? and ?target rare? sites (n=3 sites in each category). Values were obtained from a life table response experiment (LTRE). Code for life stages: D=seed bank (i.e., dormant), G=seedling (i.e., germinated), SR=small rosette, LR=large rosette, SB=small bolter, LB=large bolter.    Non-release, Target Common-0.3-0.2-0.10.00.10.20.3G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBRelease, Target Common-0.3-0.2-0.10.00.10.20.3G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNon-release, Target Rare-0.6-0.5-0.4-0.3-0.2-0.10.00.10.20.3Release, Target Rare-0.6-0.5-0.4-0.3-0.2-0.10.00.10.20.3LTRE averages for BS on each site ref=09-10, treatment 10-11TransitionsContributions to difference in lambda 101       Figure 4.12.  Mean ( ?  standard deviation) elasticity values for H. micrantha transitions in transition years 0-1 to 1-2 on release and non-release and ?target common? and ?target rare? sites (n=3 sites in each category). Code for life stages: D=seed bank (i.e., dormant), G=seedling (i.e., germinated), SR=small rosette, LR=large rosette, SB=small bolter, LB=large bolter.    D-D D-G D-SR G-G G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LB SB-DSB-GSB-SR SB-LRSB-SB SB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNon-release, Target Common, Year 0-10.00.10.2D-D D-G D-SR G-G G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LB SB-DSB-GSB-SR SB-LRSB-SB SB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBRelease, Target Common, Year 0-10.00.10.2D-D D-G D-SR G-G G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LB SB-DSB-GSB-SR SB-LRSB-SB SB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNon-release, Target Rare, Year 0-10.00.10.2D-D D-G D-SR G-G G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LB SB-DSB-GSB-SR SB-LRSB-SB SB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBRelease, Target Rare, Year 0-10.00.10.2D-D D-G D-SR G-G G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LB SB-DSB-GSB-SR SB-LRSB-SB SB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNon-release, Target Common,Year 1-20.00.10.2D-D D-G D-SR G-G G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LB SB-DSB-GSB-SR SB-LRSB-SB SB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBRelease, Target Common, Year 1-20.00.10.2D-D D-G D-SR G-G G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LB SB-DSB-GSB-SR SB-LRSB-SB SB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNon-release, Target Rare, Year 1-20.00.10.2D-D D-G D-SR G-G G-SRG-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LB SB-DSB-GSB-SR SB-LRSB-SB SB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBRelease, Target Rare, Year 1-20.00.10.2Elasticities for BS on Target Common and Target Rare release and non-release sites each yearTransitionsElasticities 102    Figure 4.13. The relationship between seed count of n=25 C. officinale plants and distance from M. crucifer release points on three sites in the year of release. Symbols correspond to sites.  The line corresponds to a generalized linear mixed model with site as a random effect (?2=2656.7, P<0.0001, DE=36.6%). Four plants were excluded from analysis as outliers.  One plant with 2751 seeds (7.8 m from release) was overly influential on the model, and 3 plants >20 m from release points were overly influential on the distance effect.  Inclusion of the outliers did not change model significance.    Distance from release (m)Seed count0 2 4 6 8 10 12050010001500 103       Figure 4.14. Probabilities of H. micrantha small bolter A) survival, and B) bolting given survival and transition to a different life stage, with distance from M. crucifer release points on n=6 sites in the year of release. The 5 plants that died were on 4 different release sites. Generalized linear mixed models included site as a random effect (survival: n=70, DE=10.7%, ?2= 3.8, P=0.0499, bolting: n=38, DE=20.0%, ?2= 9.6, P=0.002).    Distance from release (m)Number of plants                                  Survival probability051015200.00.20.40.60.81.00 2 4 6 8 10 12DiedSurvivedDistance from release (m)Number of plants                                  Bolting probability051015200.00.20.40.60.81.00 2 4 6 8 10 12BoltedDid not bolt 104      Figure 4.15. Fecundity of small bolting H. micrantha on 6 M. crucifer release sites, with distance from release points in year of release. Symbols correspond to different sites. The generalized linear mixed model includes site as a random effect (n=39, DE=3.1%, ?2= 52.5, P<0.0001).   Distance from release (m)Seed count0 2 4 6 80100200300400500600 105      Figure 4.16. Mean ( ?  95% c.i.) survival, growth and bolting of small C. officinale rosettes (4-9 leaves) with and without M. crucifer use scars on three sites in transition years 0-1 and 1-2 after weevil release.  Values are back-transformed fixed effects parameter estimates from minimum adequate generalized linear mixed model with success as the response variable, scar presence, year, or their interaction, as fixed effects, and site/plant (or just site for bolting plants) as random effects. Specifics for each model were: A) Survival: DE= 4.6%, ?2=13.3, P=0.004, Year 0-1 n=126, Year 1-2 n=84;  B) Growth given survival: DE= 10.8%, ?2=15.4, P=0.0004, Year 0-1 n=54, Year 1-2 n=51; C) Bolting given survival and growth: DE= 11.3%, ?2=6.7, P=0.010, Year 0-1 n=21, Year 1-2 n=34.   A) Survival0.00.20.40.60.81.0Year 0-1    Year 1-2Probability of survivingB) Growth0.00.20.40.60.81.0Year 0-1    Year 1-2Probability of growth, given survivalC) BoltingBoth years0.00.20.40.60.81.0Probability of bolting, given survival and growthNot scarredScarred 106     Figure 4.17. Mean ( ?  95% c.i.) fecundity values for C. officinale with different M. crucifer scar ratings in Years 0, 1, and 2 after release of 300 M. crucifer at 3 sites. A scar rating of 0 signified no M. crucifer scars were visible, while a rating of 1 corresponds to 1-9 scars, and a rating of 2 signifies 10 or more scars. The generalized linear mixed model contains site as a random effect (?2= 3942.7, P<0.0001, DE=19.9%). The number of plants per scar rating per year were: Year 0, rating 0, n=10, rating 1, n=8, rating 2, n=9; Year 1, rating 0, n=4, rating 1, n=8, rating 2, n=4; Year 2, rating 0, n=12, rating 1, n=15, rating 2, n=1 (therefore data not shown).  Year 0 Year 1 Year 20100200300400500600700Seed countNo scarsMild scarringSevere scarring 107   A) Small bolters   B) Large bolters        Figure 4.18. Mean ( ?  95% c.i.) H. micrantha fecundity values for A) small bolters and B) large bolters with M. crucifer scar rating in the year of release.   Sample sizes were Small bolters: No scars, n=13; Mild scarring, n=16; Severe scarring, n=10; Large bolters: n=10; Mild scarring, n=2; Severe scarring, n=9.   050100150200Seed countNo scarsMild scarringSevere scarring0100200300400500600700Seed countNo scarsMild scarringSevere scarring 108 CHAPTER 5  CONCLUSIONS: IMPLICATIONS OF  M. CRUCIFER HERBIVORY PATTERNS AND DEMOGRAPHIC EFFECTS   In this dissertation, I used a novel approach to examine the within-patch patterns and potential implications of herbivory by a biocontrol insect on its target weed and a nontarget plant. Experimentally releasing biocontrol agents into naturally-occurring nontarget patches in the introduced range with and without the target weed simulates a ?worst case? scenario and addresses a key question regarding weed biocontrol among scientists and non-scientists alike: ?What will the insects do when their target weed is gone?? Whenever a biocontrol agent can fully develop on a nontarget plant species, the potential for a novel ecological association exists, and the question of whether nontarget use will remain a localized and temporary phenomenon (spillover) or become a widespread persistent problem becomes crucial. Previous efforts in characterizing post-release nontarget use in weed biocontrol as spillover or persistent use have been limited to post-hoc observational studies of established systems with unknown local insect densities (i.e., Louda 1998, McFadyen et al. 2002, Russell et al. 2007, Taylor et al. 2007).  This study is, to my knowledge, the first example of an experimental post-release test for persistent nontarget use in weed biocontrol.  The approach used is simple and repeatable, and could be used in many biocontrol systems. This dissertation therefore advances knowledge and possibilities in both the theoretical and applied avenues of weed biocontrol.    As expected based on pre-release host-specificity tests (Jordan et al. 1993, De Clerck-Floate et al. 1996), in this field study M. crucifer demonstrated greater preference for and performance on its target weed, C. officinale, relative to the nontarget native host, H. micrantha.  M. crucifer fed, oviposited, and at least partially developed on H. micrantha, but did not sustain its own populations on the nontarget species in the absence of C. officinale. This pattern is strong evidence that nontarget use by M. crucifer is limited to  109 spillover, i.e., situations with high local insect density.  This distinction immediately separates the nontarget risk posed by M. crucifer from the persistent and widespread use by the notorious thistle biocontrol weevils Rhinocyllus conicus, Larinus planus and Trichosirocalus horridus (Louda and O'Brien 2002, Rand and Louda 2006, Takahashi et al. 2009).   However, spillover must not be automatically dismissed as inconsequential to plant populations simply because it is temporary. Transient effects can have lasting effects on populations (i.e., for at least tens of generations, Lynch et al. 2002, Hastings 2004), particularly so in biocontrol where predator and prey abundances are dynamic and likely to be out of equilibrium (Lynch et al. 2002, Faria et al. 2008). Instead of the intuitive question ?What will the insects do when their target weed is gone??, a more appropriate question for M. crucifer appears to be ?What do the insects do when their target plant is abundant??  As spillover is presumably the most common form of nontarget use occurring in weed biocontrol, effort should be invested into understanding its patterns and potential consequences to nontarget plants. To address this question, I characterized the different within-patch, temporal and spatial patterns of M. crucifer herbivory on target and nontarget plants, and bridged them with the ecological concept of refuge theory (Hawkins et al. 1993).  Refuge theory has been used to help understand the dynamics of pest insect and pathogen biocontrol systems (Hawkins et al. 1993, Johnson 2010), but so far has been only implied or mentioned in passing (Myers and Post 1981) in biocontrol of weeds.  Interpreting the herbivory patterns documented in this study through temporal, spatial and probabilistic refuges contributes to explanations of why M. crucifer has been an effective biocontrol agent against C. officinale in Canada, and why even the most severe spillover nontarget use observed is unlikely to affect H. micrantha populations. Understanding the mechanisms causing the refuges for nontargets, such as reduced M. crucifer host-finding and arrestment behaviours, infrequent oviposition into encountered plants and potentially poorer performance of developing larvae helps explain and predict future herbivory patterns and impacts.    110 For a herbivore to affect plant population vital rates or lambda values, the incidence and severity of use must be connected with changes in individual plant survival, growth or fecundity (McClay and Balciunas 2005, Morin et al. 2009). I measured demographic parameters of C. officinale and H. micrantha on M. crucifer release sites and non-release sites to detect impacts to vital rates and population growth rates.  When in high density, M. crucifer increased mortality rates of small and large C. officinale rosettes, both of which are sensitive life stages for the invasive plant.  This mortality helps to explain the rapid collapse of C. officinale populations after M. crucifer releases in this study and on other Canadian release sites (De Clerck-Floate and Wikeem 2009), and differs from predictions of no effects on population growth rate made based on European data, where the weevil reduces C. officinale fecundity but not survival (Prins et al. 1992, Williams et al. 2010).  In contrast to target plant effects, for H. micrantha there was only weak evidence for mild, within-release site effects of M. crucifer herbivory, in terms of decreased survival of small bolting plants near release points, and potentially decreased fecundity.  Notably, these individual effects did not translate into population-level effects for the nontarget plant, most likely because of the spatial, temporal and probabilistic refuges from nontarget herbivory that occurred.    In this thesis, I examined patterns and effects of herbivory by a biocontrol agent on its target weed and one nontarget plant species. This work is not intended to be a statement that all nontargets will be unaffected by M. crucifer or other biocontrol agents, but instead is a clear example of how nontarget use can appear severe (i.e., dieback of flowering shoots) and yet not translate to population-level declines. I propose that two main mechanisms buffer H. micrantha from population-level effects, both of which are functions of the weevil?s reduced preference for and performance on the nontarget plant; 1) temporal, spatial, and probabilistic refuges from nontarget herbivory, and 2) weak or no impact of herbivory on individual H. micrantha survival, growth or fecundity. Collectively, this study demonstrates that M. crucifer should not be considered a detriment to the reputation of weed biocontrol, but instead presents an opportunity to study the patterns, mechanisms and population-level effects of both target use and spillover nontarget use in a highly effective and mobile oligophagous agent. As most  111 biocontrol agents likely have the potential to use nontarget plants to some degree (Sheppard et al. 2005), studying the specifics of post-release herbivory patterns and impact by released oligophagous agents is necessary to provide information for optimal management of weeds, agents and nontargets in existing systems. Furthermore, understanding how pre-release host specificity test results translate into post-release herbivory patterns and impacts for oligophagous insects is instrumental for making informed decisions regarding the future release of similar candidate agents.     5.1. WHY M. CRUCIFER HAS BEEN EFFECTIVE AGAINST C.  OFFICINALE IN CANADA  The success of M. crucifer as a biocontrol agent for C. officinale in Canada appears to be the outcome of several factors.  First, the weevil demonstrated a high frequency and intensity of oviposition in C. officinale and successful larval development (De Clerck-Floate et al. 2005). This prolific reproduction combined with the lack of specialist parasitoids from the native range means that M. crucifer populations can build to outbreak levels within only two to three generations when not limited by availability of C. officinale (Schwarzlaender 1997, De Clerck-Floate et al. 2005), a characteristic that may be important for successful biocontrol (Gassmann 1996). Second, the weevil has specialized host-finding and arrestment behaviours for its target weed (De Clerck-Floate et al. 2005, Chapter 3), meaning that plants at a wide range of distances are found by dispersing insects.  Typically, large bolting plants are preferred by ovipositing females (Prins et al. 1992, Schwarzlaender 1997), but when insect density is high, small and large rosettes incur feeding and oviposition, decreasing their survival rates. Third, the population growth rate of C. officinale is sensitive to a decrease in rosette survival (Chapter 4), meaning that widespread, persistent rosette herbivory on the target weed has an impact at the population level. This effect may be in addition to a potential 30-35% decrease in fecundity caused by the insect in C. officinale in its native range (Prins et al. 1992, Williams et al. 2010). Therefore, introducing M. crucifer into North America may not fulfill the goal of restoring the natural level of herbivory pressure from the  112 native range (Hoddle 2004a), but instead creates a more unstable and responsive insect-plant interaction, with the overexploiting weevil acting as a ?self-persisting herbicide? (Keane and Crawley 2002).  Such a system may be characterized by transient colonization, outbreak, and extinction dynamics for both C. officinale and M. crucifer at the patch level, and metapopulation dynamics at the landscape level, which C. officinale exhibits in both the native and introduced ranges (van der Meijden et al. 1992, De Clerck-Floate 1996).    5.2. REASONS FOR LACK OF EFFECT OF M. CRUCIFER ON H.  MICRANTHA POPULATIONS  Spillover nontarget use by M. crucifer is restricted to situations when and where the insects are in high density, and in contrast with target use, nontarget use is transient, localized, low intensity, and unlikely to be consequential to H. micrantha populations.  The coupled ?boom and bust? dynamics of M. crucifer and C. officinale, a property that likely contributes to the effectiveness of this agent, means that there may be regular but transient situations of high insect density, and thus nontarget use, in patches across the landscape.  This phenomenon may be particularly prevalent in the initial phases of M. crucifer release or invasion, when C. officinale occurs in high local densities.  However, the weevil?s low preference for and poor performance on the nontarget mean that dispersing insects are unlikely to encounter, become arrested on, and oviposit in nontargets, thus creating spatial and temporal refuges from herbivory even in patches where the target and nontarget plant species grow sympatrically (Chapters 2 and 3). When oviposition does occur, it is aggregated and low-level. This pattern means that the vast majority of H. micrantha plants escape heavy use, and creates a substantial probabilistic refuge for the nontarget plant even under severe spillover situations (Chapter 2).   Finally, even the most severe cases of M. crucifer use had only a marginal impact on individual plant survival, growth, and fecundity, and these impacts occurred only immediately adjacent to release points, where weevil density was presumed to be highest (Chapter 4). Interestingly, the decrease in H. micrantha fecundity near release  113 points appeared to be from heavy adult feeding of flowering shoot tips, as opposed to larval feeding damage within roots. Adult M. crucifer feeding damage should therefore not be dismissed in assessing impacts to nontarget plants. However, in summary, this study detected no population-level effects of M. crucifer herbivory on H. micrantha.     5.3.  MANAGEMENT RECOMMENDATIONS  Although no detectable population-level effects were found for H. micrantha in this study, of course this result does not mean that no other nontarget Boraginaceae species are affected by M. crucifer use. The weevil is recorded as accepting North American host species from at least seven Boraginaceae genera (Jordan et al. 1993, De Clerck-Floate and Schwarzl?nder 2002, Andreas et al. 2008), but among these nontarget hosts, M. crucifer has generally exhibited the highest preference for Hackelia species. Therefore, the conclusions and management recommendations from this study are likely to be conservative for nontargets in other genera.   For some nontarget species (i.e., Threatened and Endangered plants), any signs of M. crucifer use, damaging or not, may impose unacceptable risk from a political or conservation standpoint. The recommendations in this thesis are therefore particularly relevant to Hackelia venusta, a federally-listed Endangered plant in Washington that has been reduced to a single population and that grows in close proximity to C. officinale (U.S. Fish and Wildlife Service 2007, 2011, Vance 2013).  Little can be done to prevent the ongoing movement of M. crucifer from Canada into the USA, as the weevil follows the trail of C. officinale infestations that transverse the Canada-USA border. However, potential impacts of M. crucifer on the rare and localized H. venusta may be minimized and properly monitored as a result of what has been learned in the current study involving the surrogate congeneric species, H. micrantha. Specifically, I recommend the following:    114 1) Create spatial refuges to prevent spillover. Localized populations of a plant species of special interest may be protected from spillover nontarget use by removing nearby target C. officinale to create spatial refuges from insect activity, as has been generally suggested by Harris (1988) and Ancheta and Heard (2011).   To my knowledge, this approach has not been actually implemented in a field situation, but it is an option for H. venusta management should M. crucifer be discovered near the last remaining known population of the Endangered plant. Given the reduced nontarget host-finding tendencies of M. crucifer, it is unlikely that dispersing insects would find and use the H. venusta patch, and the inability of M. crucifer to maintain populations on H. micrantha suggests that a strong temporal refuge would also form.  However, even if weevil density was temporarily high near H. venusta, transient adult feeding of flowering shoot tips could cause reduced fecundity for that season (Chapter 4). Although seed production may be a limiting factor in the population growth rate of H. venusta (U. S. Fish and Wildlife Service 2007), as a perennial species, the population would likely be able to tolerate a temporary drop in fecundity.  Nevertheless, nontargets with low population growth rates, such as H. venusta, may be especially vulnerable to impact from herbivory (Holt and Hochberg 2001), and it is thus best to manually reduce any nearby populations of M. crucifer as a safeguard. 2) Conduct evidence-based monitoring: Endophagous insect agents, such as root-feeders, are difficult to monitor (Hunter 2001), and signs of use that appear similar superficially may represent different oviposition frequencies on different species. I recommend establishing species-specific relationships between use scars and oviposition levels through destructive sampling to avoid overestimating larval feeding damage to nontarget plants. If destructive sampling is not possible (i.e., for Endangered species), it is important to at least remain aware that M. crucifer scars and oviposition frequency are not necessarily tightly correlated in lower-ranked (i.e., nontarget) host plants.        115 5.4. FUTURE WORK  Any detailed study will generate questions for future research, and this work is no exception. I see at least three main areas for future investigation toward a better understanding of the dynamics of the C. officinale ? M. crucifer ? H. micrantha system, in both the short- and long-term.    5.4.1. REFINING AND EXPANDING POPULATION MODELS Several possibilities exist to refine the basic matrix models I constructed.  The models were deterministic, meaning they do not take into account year-to-year variation. While the deterministic growth rates calculated may be appropriate for understanding the dynamics occurring during the measured years (Crone et al. 2013), they are likely to be inaccurate in predicting future levels of infestation. Stochastic growth rates and population viability analysis can be used to predict extinction rates (Morris and Doak 2002) using transition matrices from both years after insect release. However, these methods may not be appropriate in this case because the differences in vital rates between the first and second transition years for both plant species were likely a result of the large increase in precipitation in Year 1-2 (environmental stochasticity), and a dramatic drop in insect density one year after release (an artefact of not having high enough C. officinale density on sites to retain or increase population levels).   The matrix models developed in this study were density-independent because they are simpler to construct and interpret, and may be appropriate when plants are growing in competition with other species and interspecific interactions may be more important than intraspecific competition (Crone et al. 2011). However, there is evidence that several C. officinale vital rates were density-dependent (seedling growth, Appendix B, and rosette survival, Chapter 4), and since early life stages of C. officinale generally require disturbances to reduce interspecific competition, density-dependent models might improve accuracy of population projections for this species.  Furthermore, I used the same seed and seedling vital rates for all site transition matrices, which is likely a source of error in the models. Seedling vital rates are likely to be highly variable, and for C.  116 officinale are known to depend on environmental conditions and competitive interactions at the micro-site level (de Jong and Klinkhamer 1988a).  The effect of M. crucifer herbivory on C. officinale and H. micrantha seedlings is unknown.  I observed heavy feeding damage on C. officinale cotyledons at the Lethbridge M. crucifer rearing site where weevils were deliberately at outbreak densities, and occasionally saw mild seedling feeding on C. officinale (but not H. micrantha) seedlings on my experimental release sites. Therefore, future investigation into the density-dependence and effect of M. crucifer on C. officinale seedling dynamics may be warranted because seedling recruitment has a large impact on populations (de Jong and Klinkhamer 1988b).   I calculated asymptotic population growth rates to assess the population-level effects of herbivory on both plant species.  However, as mentioned previously, the C. officinale ? M. crucifer interaction in any patch may be in a continuous state of transient dynamics.  Interpreting transient interactions with equilibrium metrics may underestimate or overestimate growth rates in the near-term, depending on the difference in the true stage distribution in the field compared to the stable stage distribution (Williams et al. 2011).  Transient dynamics can have lasting effects on populations (i.e., for at least tens of generations, Lynch et al. 2002, Hastings 2004), and therefore population projections for the M. crucifer ? C. officinale ? H. micrantha system should begin with accurately estimated starting stage distributions (Maron et al. 2010). This task would be a challenge, however, due to the difficulties in determining the number of dormant seeds in the soil. Quantifying the H. micrantha seed bank would be particularly difficult because the native species has small, fragile seeds, and unknown dormancy tendencies.  The models presented in this study consider only the plant populations and ignore M. crucifer demography, except for implying that insect density was highest during the year of release as an artefact of inadequate C. officinale abundances on experimental sites to maintain weevil populations.  Buckley et al. (2005) created a coupled plant-herbivore model to predict stable suppression of the target weed Echium plantagineum L. by the weevil Mogulones larvatus Schultze.  A similar model could be constructed for the C. officinale ? M. crucifer ? H. micrantha system to predict insect outbreaks and nontarget  117 use.  A model for this system would likely have to be spatially explicit to account for the metapopulation dynamics exhibited by the herbivore and target weed, and techniques for spatial demographic models have been developed (Neubert and Caswell 2000, Jongejans et al. 2008).  Sufficient data on M. crucifer dispersal, reproduction, and impact on C. officinale may already exist to parameterize such a model, although an important parameter, the density-dependence of larval competition within plants (Buckley et al. 2005), is unknown.    5.4.2. FURTHER EXPLORATION OF M. CRUCIFER HERBIVORY  EFFECTS ON INDIVIDUAL PLANTS The observed patterns of individual effects of M. crucifer on C. officinale and H. micrantha generate several questions for future study. It is unknown if increased mortality of C. officinale rosettes on M. crucifer release sites was the result of adult feeding, larval feeding, or an interaction between the two. Root borers heavily damage plants by disrupting vascular and support tissues, even at low levels of infestation (Blossey and Hunt-Joshi 2003, Preisser and Bastow 2005).  Adult and larval M. crucifer feeding may have a greater per capita demographic impact with decreasing plant size because of the decrease in cross-sectional ratio of area of conductive tissue (i.e., phloem and xylem) to area injured by feeding. The effect of adult feeding could be teased apart from the effects of larval feeding by comparing C. officinale demography on sites with releases of sterilized or all male M. crucifer to both non-release sites and sites with regular M. crucifer releases.   In contrast to findings in Europe (Prins et al. 1992, Williams et al. 2010), the rangeland release experiment showed no decrease in C. officinale fecundity on M. crucifer release sites compared to non-release sites. However, within release sites, heavily scarred C. officinale did have reduced fecundity compared to unscarred plants. Regardless, this short-term study may not have fully detected the effect of M. crucifer on C. officinale seed production. The weevil releases were made on June 4, 2009, approximately 4-5 weeks after established populations of overwintered adult M. crucifer would have emerged to begin feeding and oviposition in their target plants. It is possible that releases  118 were made too late for adult or larval feeding to affect within-season fecundity in bolting plants.  In comparison, eggs or larvae in established M. crucifer populations may overwinter in C. officinale roots, and plants can incur M. crucifer damage for several years before bolting, resulting in a cumulative effect of herbivory over multiple years. Therefore, further work in established populations of C. officinale and M. crucifer is necessary to quantify the effect of weevil herbivory on seed production in North America. Increased rosette mortality may explain the main challenge to conducting such an experiment, which is finding populations where bolting C. officinale plants exist years after M. crucifer release (R. De Clerck-Floate, personal communication).     In contrast to the situation with C. officinale, the reduction in H. micrantha fecundity near field release points appeared to be a direct result of heavy feeding by M. crucifer adults on the tips of flowering stems. Otherwise, it is likely that the weakness of the individual effect of M. crucifer on H. micrantha is at least partly the result of the low levels of oviposition in the nontarget plant.  It is also possible that the architecture and morphology of the native plant is well-suited for compensatory growth in response to this type of herbivore damage. Similar to C. officinale, H. micrantha has a taproot.  However, I observed during dissections that the taproot of the native plant is hollow and noticeably woodier than that of C. officinale, suggesting it is less of a carbohydrate-storing organ than that of the invasive plant. The bulk of M. crucifer larval feeding in the native plant is thus restricted to the fleshy base of shoots and within the pith of the woody caudices emerging from the root. The difference in availability of fleshy tissue in H. micrantha may mean that disruption of conductive tissue from larval feeding occurs at the individual shoot level, rather than the plant level as in C. officinale (R. De Clerck-Floate and H. Catton, unpublished data).  Thus, there may be refuges from M. crucifer damage and impact even within individual H. micrantha plants. This hypothesis warrants further investigation, and would lead to plant architecture and morphology being considered when predicting potential impacts to target and nontarget plants in biocontrol.     119 5.4.3. INVESTIGATING POTENTIAL CHANGES IN M. CRUCIFER  HOST PREFERENCE AND PERFORMANCE In addition to potential immediate nontarget effects, a second concern regarding using oligophagous insects in biocontrol is the potential for evolutionary increases in agent preference for and performance on nontarget species (van Klinken and Edwards 2002, Hufbauer and Roderick 2005). In theory, agent ?host races? favouring nontarget plants can evolve under certain conditions when insects reproduce and maintain reproductively isolated populations on the alternative hosts (Dr?s and Mallet 2002). Post-release development of new host races have not been documented in weed biocontrol systems to date (van Klinken and Edwards 2002, Hufbauer and Roderick 2005). Nevertheless, rapid evolution of traits has occurred in insects (Singer et al. 1993, Thompson 1998), including in at least one biocontrol agent (i.e., a decrease in juvenile development time described in McEvoy et al. 2012).   The lack of observed M. crucifer population establishment on isolated nontarget patches suggests that the risks of evolutionary changes in host preference are low for this insect within a time scale relevant to human society (i.e., within hundreds of years).  If, however, the weevil can persist solely on nontargets, several conditions are required for new host race formation. First, sufficient genetic variation in host preference and performance is necessary for selection to act upon (van Klinken and Edwards 2002). The aggregated pattern of nontarget oviposition observed in this study suggests that most eggs may be laid by a small number of females, suggesting genetic variation in host specificity in M. crucifer could exist. Quantification of variability in female oviposition in nontarget species is necessary, and possible in a laboratory setting. Second, increased preference for and performance on nontargets must be heritable and may need to be linked (Gripenberg et al. 2010); such genetic information is unknown for M. crucifer.  Third, insect populations on nontargets may need to be reproductively isolated from insects using target plants (Dr?s and Mallet 2002).  While the formation of new host races of M. crucifer appears unlikely, further population-level, behavioural, spatial and quantitative genetic study of the weevil?s nontarget preference and performance are  120 necessary to determine the likelihood and time required for such an evolutionary change to occur, and would be a natural extension of this work.    Strong larval performance is crucial for the maintenance of insect populations on a host plant. However, accurately assessing this aspect for the endophagous M. crucifer in H. micrantha is difficult for several reasons. First, H. micrantha is difficult to grow under artificial conditions, therefore wild plants must be harvested for performance assays, where variability in plant age, physiology and morphology is unavoidable. Moreoever, oviposition in nontargets is sporadic and variable, and so far it is impossible to get an accurate estimate of the number of eggs in a plant without destructively sampling and dissecting the sampled plant.  Possible future investigations of M. crucifer larval performance in nontargets may need to use egg or larval transfer techniques, and a different nontarget plant species. A good candidate for this work is Hackelia floribunda (Lehm.) I. M. Johnston, a species much more amenable for greenhouse cultivation (E. Pavlik, personal communication), and known to support full development of M. crucifer (De Clerck-Floate and Schwarzl?nder 2002).      5.5. FINAL THOUGHTS  A detrimental imbalance exists between pre-release and post-release research efforts in weed biocontrol.  While abundant resources are justifiably invested in reliably identifying the nontarget host range of proposed agents, comparatively little effort has been expended to monitor the mechanisms and patterns of impact to target weeds and nontarget plants within an insect?s host range after release (McEvoy and Coombs 1999, Morin et al. 2009).  In this dissertation, I address this gap in knowledge using a highly effective but oligophagous biocontrol insect, its target weed, and a native nontarget plant species. By combining experimental field research with ecological refuge theory and plant population modeling, I characterize and interpret how differential preference and performance on plant species are expressed in terms of herbivory patterns in the field. I then describe how the resulting herbivory translated into differential effects to vital rates  121 and population growth rates in a target and nontarget plant. This type of post-release analysis contributes valuable information to a relevant but understudied area of weed biocontrol, and simultaneously increases our understanding of the risks and benefits of a controversial agent. I recommend that similar post-release studies in other systems be conducted to improve the predictability of target and nontarget impacts and to aid in making informed decisions on future releases of potentially effective but oligophagous biocontrol insects.   122 REFERENCES CITED  Albright, M. F., W. N. Harman, S. S. Fickbohm, H. Meehan, S. Groff, and T. Austin. 2004. Recovery of native flora and behavioral responses by Galerucella spp. following biocontrol of purple loosestrife. The American Midland Naturalist 152:248?254. Ancheta, J., and S. B. Heard. 2011. Impacts of insect herbivores on rare plant populations. Biological Conservation 144:2395?2402. Andreas, J. E., M. Schwarzl?nder, H. Ding, and S. D. Eigenbrode. 2009. Post-release non-target monitoring of Mogulones cruciger, a biological control agent released to control Cynoglossum officinale in Canada. Pages 75-82 in M. H. Julien, R. Sforza, M. C. Bon, H. C. Evans, P. E. Hatcher, H. L. Hinz, and B. G. Rector, editors. Proceedings of the XII International Symposium on Biological Control of Weeds, CAB International, Wallingford, UK. Andreas, J., M. Schwarzl?nder, and R. De Clerck-Floate. 2008. The occurrence and potential relevance of post-release, nontarget attack by Mogulones cruciger, a biocontrol agent for Cynoglossum officinale in Canada. Biological Control 46:304?311. Baker, J. L., N. A. P. Webber, and K. K. Johnson. 2004. Non-target impacts of Aphthona nigriscutis, a biological control agent for Euphorbia esula (leafy spurge), on a native plant Euphorbia robusta. Pages 247?251 in J. M. Cullen, D. T. Briese, D. J. Kriticos, W. M. Lonsdale, L. Morin, and J. K. Scott, editors. CSIRO Entomology, Canberra, Australia. Behmer, S. T., and A. Joern. 2012. Insect herbivore outbreaks viewed through a physiological framework: insights from Orthoptera. Pages 3?29 in P. Barbosa, D. K. Letourneau, and A. A. Agrawal, editors. Insect Outbreaks Revisited. Wiley-Blackwell, Chichester, UK. Bernays, E. A., and R. Chapman. 1994. Host-Plant Selection by Phytophagous Insects. Chapman & Hall, New York, New York. Berryman, A. A., B. A. Hawkins, and B. A. Hawkins. 2006. The refuge as an integrating concept in ecology and evolution. Oikos 115:192?196.  123 Blossey, B., and L. Skinner. 2000. Design and importance of post-release monitoring. Pages 693?706 in N. R. Spencer, editor. Proceedings of the X International Symposium on Biological Control of Weeds. Montana State University, Bozeman, Montana. Blossey, B., and T. R. Hunt-Joshi. 2003. Belowground herbivory by insects: influence on plants and aboveground herbivores. Annual Review of Entomology 48:521?547. Blossey, B., R. Casagrande, L. Tewksbury, D. Landis, R. Wiedenmann, and D. Ellis. 2001. Nontarget feeding of leaf-beetles introduced to control purple loosestrife (Lythrum salicaria L.). Natural Areas Journal 21:368?377. Briese, D. 2005. Translating host-specificity test results into the real world: the need to harmonize the yin and yang of current testing procedures. Biological Control 35:208?214. Briese, D., M. Zapater, A. Andorno, and G. Perez-Camargo. 2002. A two-phase open-field test to evaluate the host-specificity of candidate biological control agents for Heliotropium amplexicaule. Biological Control 25:259?272. Buckley, Y. M., M. Rees, A. W. Sheppard, and M. J. Smyth. 2005. Stable coexistence of an invasive plant and biocontrol agent: a parameterized coupled plant-herbivore model. Journal of Applied Ecology 42:70?79. Carson, W. P., and R. B. Root. 2000. Herbivory and plant species coexistence: community regulation by an outbreaking phytophagous insect. Ecological Monographs 70:73?99. Carson, W. P., S. M. Hovick, A. J. Baumert, D. E. Bunker, and T. H. I. Pendergast. 2008. Evaluating the post-release efficacy of invasive plant biocontrol by insects: a comprehensive approach. Arthropod-Plant Interactions 2:77?86. Caswell, H. 2001. Matrix Population Models: Construction, Analysis and InterpretationSecond Edition. Sinauer Associates, Inc., Sunderland, Massachusetts. Chapman, R., and E. Bernays. 1989. Insect behavior at the leaf surface and learning as aspects of host plant selection. Cellular and Molecular Life Sciences 45:215?222. Codling, E. A., M. J. Plank, and S. Benhamou. 2008. Random walk models in biology. Journal of The Royal Society Interface 5:813?834.   124 Coupe, M. D., and J. F. J. Cahill. 2003. Effects of insects on primary production in temperate herbaceous communities: a meta-analysis. Ecological Entomology 28:511?521. Crawley, M. 1989. Insect herbivores and plant population dynamics. Annual Review of Entomology 34:531?562. Crone, E. E., E. S. Menges, M. M. Ellis, T. Bell, P. Bierzychudek, J. Ehrl?n, T. N. Kaye, T. M. Knight, P. Lesica, W. F. Morris, G. Oostermeijer, P. F. Quintana-Ascencio, A. Stanley, T. Ticktin, T. Valverde, and J. L. Williams. 2011. How do plant ecologists use matrix population models? Ecology and Society 14:1?8. Crone, E. E., M. M. Ellis, W. F. Morris, A. Stanley, T. Bell, P. Bierzychudek, J. Ehrl?n, T. N. Kaye, T. M. Knight, P. Lesica, G. Oostermeijer, P. F. Quintana-Ascencio, T. Ticktin, T. Valverde, J. L. Williams, D. F. Doak, R. Ganesan, K. McEachern, A. S. Thorpe, and E. S. Menges. 2013. Ability of matrix models to explain the past and predict the future of plant populations. Conservation Biology 27:968?978. Cullen, J. 1989. Current problems in host-specificity screening. Pages 27?36 in E. Delfosse, editor. Proceedings of the VII International Symposium on Biological Control of Weeds, Rome, March 1988. CSIRO Publishing, Melbourne. Culliney, T. W. 2005. Benefits of classical biological control for managing invasive plants. Critical Reviews in Plant Sciences 24:131?150. Dauer, J. T., P. B. McEvoy, and J. Van Sickle. 2012. Controlling a plant invader by targeted disruption of its life cycle. Journal of Applied Ecology 49:322?330. Davies, J. C., and D. J. Greathead. 1967. Occurrence of Teleonemia scrupulosa on Sesamum indicum Linn. in Uganda. Nature 213:102?103. Davis, A. S., D. A. Landis, V. Nuzzo, B. Blossey, E. Gerber, and H. L. Hinz. 2006. Demographic models inform selection of biocontrol agents for garlic mustard (Alliaria petiolata). Ecological Applications 16:2399?2410. De Clerck-Floate, R. 1996. The role of pre-release studies in developing a biocontrol strategy for hound's-tongue in Canada. Pages 143?148 in V. C. Moran and J. H. Hoffmann, editors. Proceedings of the IX International Symposium on Biological Control of Weeds, University of Capetown, South Africa.   125 De Clerck-Floate, R. 1997. Cattle as dispersers of hound's-tongue on rangeland in southeastern British Columbia. Journal of Range Management:239?243. De Clerck-Floate, R. A. 2013. Cynoglossum officinale (L.), Houndstongue (Boraginaceae). Pages 309?314 in P. G. Mason and D. R. Gillespie, editors. Biological Control Programmes in Canada 2001-2012. CABI Publishing, Wallingford, UK. De Clerck-Floate, R. A., B. Wikeem, and R. S. Bourchier. 2005. Early establishment and dispersal of the weevil, Mogulones cruciger (Coleoptera: Curculionidae) for biological control of houndstongue (Cynoglossum officinale) in British Columbia, Canada. Biocontrol Science and Technology 15:173?190. De Clerck-Floate, R., and B. Wikeem. 2009. Influence of release size on establishment and impact of a root weevil for the biocontrol of houndstongue (Cynoglossum officinale). Biocontrol Science and Technology 19:169?183. De Clerck-Floate, R., and M. Schwarzl?nder. 2002. Host specificity of Mogulones cruciger (Coleoptera: Curculionidae), a biocontrol agent for houndstongue (Cynoglossum officinale), with emphasis on testing of native North American Boraginaceae. Biocontrol Science and Technology 12:293?306. De Clerck-Floate, R., D. Schroeder, and M. Schwarzlaender. 1996. Supplemental Information to the Petition (Can-93-4 and TAG 93-06) to release Ceutorhynchus (Mogulones) cruciger for the biological control of hound's-tongue (Cynoglossum officinale, Boraginaceae) in Canada. Agriculture and Agri-Food Canada report. de Jong, T. J., and P. G. L. Klinkhamer. 1988a. Seedling establishment of the biennials Cirsium vulgare and Cynoglossum officinale in a sand-dune area: the importance of water for differential survival and growth. Journal of Ecology 76:393?402. de Jong, T. J., and P. G. L. Klinkhamer. 1988b. Population ecology of the biennials Cirsium vulgare and Cynoglossum officinale in a coastal sand-dune area. Journal of Ecology 76:336?382. Delfosse, E. S. 2005. Risk and ethics in biological control. Biological Control 35:319?329.    126 Dennill, G. B., D. Donnelly, and S. L. Chown. 1993. Expansion of host-plant range of a  biocontrol agent Trichilogaster acaciaelongifoliae (Pteromalidae) released against the weed Acacia longifolia in South Africa. Agriculture 43:1?10. DeWalt, S. J. 2006. Population dynamics and potential for biological control of an exotic invasive shrub in Hawaiian rainforests. Biological Invasions 8:1145?1158. Dhileepan, K., M. Trevi?o, and S. Raghu. 2006. Temporal patterns in incidence and abundance of Aconophora compressa (Hemiptera: Membracidae), a biological control agent for Lantana camara, on target and nontarget plants. Environmental Entomology 35:1001?1012. Diaz, R., W. A. Overholt, J. P. Cuda, P. D. Pratt, and A. Fox. 2009. Host specificity of Ischnodemus variegatus, an herbivore of West Indian marsh grass (Hymenachne amplexicaulis). BioControl 54:307?321.Dr?s, M., and J. Mallet. 2002. Host races in plant-feeding insects and their importance in sympatric speciation. Philosophical Transactions of the Royal Society B: Biological Sciences 357:471?492. Dudley, T. L., and D. J. Kazmer. 2005. Field assessment of the risk posed by Diorhabda elongata, a biocontrol agent for control of saltcedar (Tamarix spp.), to a nontarget plant, Frankenia salina. Biological Control 35:265?275. Faria, L. D. B., J. Umbanhowar, and K. S. McCann. 2008. The long-term and transient implications of multiple predators in biocontrol. Theoretical Ecology 1:45?53. Finch, S., and R. Collier. 2012. The influence of host and non-host companion plants on the behaviour of pest insects in field crops. Entomologia Experimentalis et Applicata 142:87?96. Fowler, S. V., P. Syrett, and R. L. Hill. 2000. Success and safety in the biological control of environmental weeds in New Zealand. Austral Ecology 25:553?562. Fox, L. R., and P. A. Morrow. 1981. Specialization: species property or local phenomenon? Science 211:887?893. Frye, M., E. Lake, and J. Hough-Goldstein. 2010. Field host-specificity of the mile-a-minute weevil, Rhinoncomimus latipes Korotyaev (Coleoptera: Curculionidae). Biological Control 55:234?240.    127 Ganga Visalakshy, P. N., K. P. Jayanth, S. K. Ghosh, and M. Chaudhary. 2008. Post introductory risk assessment studies on Zygogramma bicolorata (Coleoptera: Chrysomelidae), a classical biological control agent of Parthenium hysterophorus (Asteraceae) in India. Biocontrol Science and Technology 18:1083?1086. Garren, J. M., and S. Y. Strauss. 2009. Population-level compensation by an invasive thistle thwarts biological control from seed predators. Ecological Applications 19:709?721. Gassmann, A. 1996. Classical biological control of weeds with insects: a case for emphasizing agent demography. Pages 171?175 in V. C. Moran and J. H. Hoffmann, editors. Proceedings of the IX International Symposium on Biological Control of Weeds, University of Capetown, South Africa. Godfray, H. C. J., and M. Shimada. 1999. Parasitoids as model organisms for ecologists. Researches on Population Ecology 41:3?10. Grevstad, F. S. 1999. Factors influencing the chance of population establishment: implications for release strategies in biocontrol. Ecological Applications 9:1439?1447. Grevstad, F. S., and A. L. Herzig. 1997. Quantifying the effects of distance and conspecifics on colonization: experiments and models using the loosestrife leaf beetle, Galerucella calmariensis. Oecologia 110:60?68. Gripenberg, S., P. Mayhew, M. Parnell, and T. Roslin. 2010. A meta-analysis of preference?performance relationships in phytophagous insects. Ecology Letters 13:383?393. Gutierrez, A. P., K. M. Daane, L. Ponti, V. M. Walton, and C. K. Ellis. 2008. Prospective evaluation of the biological control of vine mealybug: refuge effects and climate. Journal of Applied Ecology 45:524?536. Halpern, S., and N. Underwood. 2006. Approaches for testing herbivore effects on plant population dynamics. Journal of Applied Ecology 43:922?929. Hanski, I. 1998. Metapopulation dynamics. Nature 396:41?49. Harris, P. 1988. Environmental impact of weed-control insects. BioScience 38:542?548. Harris, P. 1993. Effects, constraints and the future of weed biocontrol. Agriculture, Ecosystems and Environment 46:289?303.  128 Hastings, A. 2004. Transients: the key to long-term ecological understanding? Trends in Ecology & Evolution 19:39-45. Havens, K., C. L. Jolls, J. E. Marik, P. Vitt, A. K. McEachern, and D. Kind. 2012. Effects of a non-native biocontrol weevil, Larinus planus, and other emerging threats on populations of the federally threatened Pitcher's thistle, Cirsium pitcheri. Biological Conservation 155:202?211. Hawkins, B. A., M. B. Thomas, and M. E. Hochberg. 1993. Refuge theory and biological control. Science 262:1429?1432. Heard, S. B., and L. C. Remer. 2008. Travel costs, oviposition behaviour and the dynamics of insect?plant systems. Theoretical Ecology 1:179?188. Hoddle, M. 2004a. Restoring balance: using exotic species to control invasive exotic species. Conservation Biology 18:38?49. Hoddle, M. S. 2004b. The strength of biological control in the battle against invasive pests: a reply. Conservation Biology 18:61?64. Holt, R. D., and M. E. Hochberg. 2001. Indirect interactions, community modules and biological control: A theoretical perspective. Pages 13?37 in E. Wajnberg, J. Scott, and P. C. Quimby, editors. Evaluating Indirect Ecological Effects of Biological Control. CABI Publishing, CABI International, Wallingford, Oxon. Howe, G. A., and G. Jander. 2008. Plant immunity to insect herbivores. Annual Review of Plant Biology 59:41?66. Hufbauer, R., and G. Roderick. 2005. Microevolution in biological control: mechanisms, patterns, and processes. Biological Control 35:227?239. Hunter, M. D. 2001. Out of sight, out of mind: the impacts of root-feeding insects in natural and managed systems. Agricultural and Forest Entomology 3:3?9. Johnson, K. B. 2010. Pathogen refuge: a key to understanding biological control. Annual Review of Phytopathology 48:141?160. Jongejans, E., O. Skarpaas, and K. Shea. 2008. Dispersal, demography and spatial population models for conservation and control management. Perspectives in Plant Ecology, Evolution and Systematics 9:153?170.    129 Jordan, T., M. Schwarzl?nder, I. Tosevski, and A. Freese. 1993. Final Report: Ceutorhynchus cruciger Herbst (Coleoptera: Curculionidae): A candidate for the biological control of hound's-tongue (Cynoglossum officinale L., Boraginaceae) in Canada. Ministry of Forests, Province of British Columbia, Canada. Kareiva, P. 1983. Local movement in herbivorous insects: applying a passive diffusion model to mark-recapture field experiments. Oecologia 57:322?327. Keane, R. M., and M. J. Crawley. 2002. Exotic plant invasions and the enemy release hypothesis. Trends in Ecology & Evolution 17:164?170. Klinkhamer, P. G. L., and T. J. de Jong. 1988. The importance of small-scale disturbance for seedling establishment in Cirsium vulgare and Cynoglossum officinale. Journal of Ecology 76:383?392. Klironomos, J. N. 2003. Variation in plant response to native and exotic arbuscular mycorrhizal fungi. Ecology 84:2292?2301. Landis, D. A., D. C. Sebolt, M. J. Haas, and M. Klepinger. 2003. Establishment and impact of Galerucella calmariensis L. (Coleoptera: Chrysomelidae) on Lythrum salicaria L. and associated plant communities in Michigan. Biological Control 28:78?91. Louda, S. M. 1998. Population growth of Rhinocyllus conicus (Coleoptera: Curculionidae) on two species of native thistles in prairie. Environmental Entomology 27:834?841. Louda, S. M., and C. W. O'Brien. 2002. Unexpected ecological effects of distributing the exotic weevil Larinus planus (F.), for the biological control of Canada thistle. Conservation Biology 16:717?727. Louda, S. M., and P. Stiling. 2004. The double-edged sword of biological control in conservation and restoration. Conservation Biology 18:50?53. Louda, S. M., D. Kendall, J. Connor, and D. Simberloff. 1997. Ecological effects of an insect introduced for the biological control of weeds. Science 277:1088?1090. Louda, S. M., R. W. Pemberton, M. T. Johnson, and P. A. Follett. 2003. Nontarget effects - the Achilles' heel of biological control? Retrospective analyses to reduce risk associated with biocontrol introductions. Annual Review of Entomology 48:365?396.  130 Louda, S., T. Rand, A. Arnett, A. McClay, K. Shea, and A. McEachern. 2005. Evaluation of ecological risk to populations of a threatened plant from an invasive biocontrol insect. Ecological Applications 15:234?249. Lynch, L. D., A. R. Ives, J. K. Waage, M. E. Hochberg, and M. B. Thomas. 2002. The risks of biocontrol: transient impacts and minimum nontarget densities. Ecological Applications 12:1872?1882. Mack, R. N., D. Simberloff, W. M. Lonsdale, H. Evans, M. Clout, and F. A. Bazzaz. 2000. Biotic invasions: causes, epidemiology, global consequences, and control. Ecological Applications 10:689?710. Mackay, D. A. 1985. Prealighting search behavior and host plant selection by ovipositing Euphydryas editha butterflies. Ecology 66:142?151. Maines, A., D. Knochel, and T. Seastedt. 2013. Biological control and precipitation effects on spotted knapweed (Centaurea stoebe): empirical and modeling results. Ecosphere 4:1?14. Marohasy, J. 1998. The design and interpretation of host-specificity tests for weed biological control with particular reference to insect behaviour. Biocontrol News and Information 19:13N?20N. Maron, J. L., and E. Crone. 2006. Herbivory: effects on plant abundance, distribution and population growth. Proceedings of the Royal Society B: Biological Sciences 273:2575?2584. Maron, J. L., C. C. Horvitz, and J. L. Williams. 2010. Using experiments, demography and population models to estimate interaction strength based on transient and asymptotic dynamics. Journal of Ecology 98:290?301. Mason, P. G., and D. R. Gillespie (Eds.). 2013. Biological Control Programmes in Canada 2001-2012. CABI Publishing, Wallingford, UK. McClay, A. S. 1995. Beyond ?before-and-after:? Experimental designs and evaluation in classical weed biological control. Pages 213?219 in E. S. Delfosse and R. R. Scott, editors. Proceedings of the XIII International Symposium on the Biological Control of Weeds. Melbourne, Australia.    131 McClay, A. S., and J. K. Balciunas. 2005. The role of pre-release efficacy assessment in selecting classical biological control agents for weeds - applying the Anna Karenina principle. Biological Control 35:197?207. McEvoy, P. B., and E. M. Coombs. 1999. Biological control of plant invaders: regional patterns, field experiments, and structural population models. Ecological Applications 9:387?401. McEvoy, P. B., K. M. Higgs, E. M. Coombs, E. Kara?etin, and L. A. Starcevich. 2012. Evolving while invading: rapid adaptive evolution in juvenile development time for a biological control organism colonizing a high-elevation environment. Evolutionary Applications 5:524?536. McFadyen, R. 1998. Biological control of weeds. Annual Review of Entomology 43:369?393. McFadyen, R. E. C., M. Vitelli, and C. Setter. 2002. Host specificity of the rubber vine moth, Euclasta whalleyi Popescu-Gorj and Constantinescu (Lepidoptera: Crambidae: Pyraustinae): field host-range compared to that predicted by laboratory tests. Australian Journal of Entomology 41:321?323. Moran, P. J., C. J. Deloach, T. L. Dudley, and J. Sanabria. 2009. Open field host selection and behavior by tamarisk beetles (Diorhabda spp.) (Coleoptera: Chrysomelidae) in biological control of exotic saltcedars (Tamarix spp.) and risks to non-target athel (T. aphylla) and native Frankenia spp. Biological Control 50:243?261. Morin, L., A. Reid, N. M. Sims-Chilton, Y. M. Buckley, K. Dhileepan, G. T. Hastwell, T. L. Nordblom, and S. Raghu. 2009. Review of approaches to evaluate the effectiveness of weed biological control agents. Biological Control 51:1?15. Morris, W. F., and D. F. Doak. 2002. Quantitative Conservation Biology: Theory and Practice of Population Viability Analysis. Sinauer Associates, Sunderland, Massachusetts. Myers, J. H. 1976. Distribution and dispersal in populations capable of resource depletion. Oecologia 23:255?269. Myers, J. H., and B. J. Post. 1981. Plant nitrogen and fluctuations of insect populations: A test with the cinnabar moth?tansy ragwort system. Oecologia 48:151?156.  132 Myers, J. H., and D. R. Bazely. 2003. Ecology and Control of Introduced Plants. Cambridge University Press, Cambridge, UK. Myers, J. H., C. Jackson, H. Quinn, S. R. White, and J. S. Cory. 2009. Successful biological control of diffuse knapweed, Centaurea diffusa, in British Columbia, Canada. Biological Control 50:66?72. Myers, J. H., J. Monro, and N. Murray. 1981. Egg clumping, host plant selection and population regulation in Cactoblastis cactorum (Lepidoptera). Oecologia 51:7?13. Natural Regions Committee. 2006. Natural Regions and Subregions of Alberta. Compiled by D.J. Downing and W.W. Pettapiece. Government of Alberta. NatureServe. 2012. NatureServe Explorer: An online encyclopedia of life [web application]. Version 7.1. NatureServe, Arlington, Viriginia. Available http://www.natureserve.org/explorer. (Access: August 14, 2012). Neubert, M. G., and H. Caswell. 2000. Demography and dispersal: calculation and sensitivity analysis of invasion speed for structured populations. Ecology 81:1613?1628. Odegaard, F., O. H. Diserud, S. Engen, and K. Aagaard. 2000. The magnitude of local host specificity for phytophagous insects and its implications for estimates of global species richness. Conservation Biology 14:1182?1186. Ortega, Y. K., D. E. Pearson, L. P. Waller, N. J. Sturdevant, and J. L. Maron. 2012. Population-level compensation impedes biological control of an invasive forb and indirect release of a native grass. Ecology 93:783?792. Page, A. R., and K. L. Lacey. 2006. Economic impact assessment of Australian weed biological control. Pages 1?151 Technical Series. CRC for Australian Weed  Management, Glen Osmond, Australia. Paynter, Q., N. Martin, J. Berry, S. Hona, P. Peterson, A. H. Gourlay, J. Wilson-Davey, L. Smith, C. Winks, and S. V. Fowler. 2008. Non-target impacts of Phytomyza vitalbae a biological control agent of the European weed Clematis vitalba in New Zealand. Biological Control 44:248?258. Peterson, R. K. D. 2000. Illuminating the black box: the relationship between injury and yield. Pages 1?12 in R. K. D. Peterson and L. G. Higley, editors. Biotic Stress and Yield Loss. CRC Press, Boca Raton, FL.  133 Pimentel, D., R. Zuniga, and D. Morrison. 2005. Update on the environmental and economic costs associated with alien-invasive species in the United States. Ecological Economics 52:273?288. Pratt, P. D., M. B. Rayamajhi, T. D. Center, P. W. Tipping, and G. S. Wheeler. 2009. The ecological host range of an intentionally introduced herbivore: A comparison of predicted versus actual host use. Biological Control 49:146?153. Preisser, E. L., and J. L. Bastow. 2005. Plant damage from and defenses against ?cryptic? herbivory: A guild perspective. Journal of Plant Interactions 1:197?210. Prins, A. H., H. W. Nell, and P. G. L. Klinkhamer. 1992. Size-dependent root herbivory on Cynoglossum officinale. Oikos 65:409?413. R Core Team. 2012. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http://www.R-project.org/. Rand, T. A., and S. M. Louda. 2004. Exotic weed invasion increases the susceptibility of native plants to attack by a biocontrol herbivore. Ecology 85:1548?1554. Rand, T. A., and S. M. Louda. 2006. Invasive insect abundance varies across the biogeographic distribution of a native host plant. Ecological Applications 16:877?890. Romme, W. H., D. H. Knight, and J. B. Yavitt. 1986. Mountain pine beetle outbreaks in the Rocky Mountains: regulators of primary productivity? American Naturalist:484?494. Rose, K. E., S. M. Louda, and M. Rees. 2005. Demographic and evolutionary impacts of native and invasive insect herbivores on Cirsium canescens. Ecology 86:453?465. Russell, F. L., S. M. Louda, T. A. Rand, and S. D. Kachman. 2007. Variation in herbivore-mediated indirect effects of an invasive plant on a native plant. Ecology 88:413?423. Schooler, S. S., E. M. Coombs, and P. B. McEvoy. 2003. Nontarget effects on crepe myrtle by Galerucella pusilla and G. calmariensis (Chrysomelidae), used for biological control of purple loosestrife (Lythrum salicaria). Weed Science 51:449?455.   134 Schoonhoven, L., J. van Loon, and M. Dicke. 2005. Insect-Plant Biology. Oxford University Press. Sch?ps, K. 2002. Local and regional dynamics of a specialist herbivore: overexploitation of a patchily distributed host plant. Oecologia 132:256?263. Schutzenhofer, M. R., and T. M. Knight. 2007. Population-level effects of augmented herbivory on Lespedeza cuneata: implications for biological control. Ecological Applications 17:965?971. Schwarzlaender, M. 1997. Bionomics of Mogulones cruciger (Coleoptera: Curculionidae), a below-ground herbivore for the biological control of hound's-tongue. Environmental Entomology 26:357?365. Shea, K., and D. Kelly. 1998. Estimating biocontrol agent impact with matrix models: Carduus nutans in New Zealand. Ecological Applications 8:824?832. Sheppard, A. W., R. D. van Klinken, and T. A. Heard. 2005. Scientific advances in the analysis of direct risks of weed biological control agents to nontarget plants. Biological Control 35:215?226. Sheppard, A. W., R. Hill, R. A. DeClerck-Floate, A. McClay, T. Olckers, P. C. Quimby Jr, and H. G. Zimmermann. 2003. A global review of risk-benefit-cost analysis for the introduction of classical biological control agents against weeds: a crisis in the making? Biocontrol News and Information 24:91N?108N. Singer, M., C. Thomas, and C. Parmesan. 1993. Rapid human-induced evolution of insect?host associations. Nature 366:681-683. Smith, L. 2012. Host plant oviposition preference of Ceratapion basicorne (Coleoptera: Apionidae), a potential biological control agent of yellow starthistle. Biocontrol Science and Technology 22:407?418. Stephens, A. E. A., and J. H. Myers. 2012. Resource concentration by insects and implications for plant populations. Journal of Ecology 100:923?931. Strong, D. R., J. L. Maron, P. G. Connors, A. Whipple, S. Harrison, and R. L. Jeffries. 1995. High mortality, fluctuation in numbers, and heavy subterranean insect herbivory in bush lupine, Lupinus arboreus. Oecologia 104:85?92. Stubben, C., and B. Milligan. 2007. Estimating and analyzing demographic models using the popbio package in R. Journal of Statistical Software 22:1?23.  135 Swope, S. M., and I. M. Parker. 2010. Widespread seed limitation affects plant density but not population trajectory in the invasive plant Centaurea solstitialis. Oecologia 164:117?128. Syrett, P., D. T. Briese, and J. H. Hoffmann. 2000. Success in biological control of terrestrial weeds by arthropods. Pages 189?230 in G. Gurr and S. Wratten, editors. Biological Control: Measures of Success. Kluwer Academic Publishers, Dordrecht, Australia. Takahashi, M., S. M. Louda, T. E. X. Miller, and C. W. O'Brien. 2009. Occurrence of Trichosirocalus horridus (Coleoptera: Curculionidae) on native Cirsium altissimum versus exotic C. vulgare in North American tallgrass prairie. Environmental Entomology 38:731?740. Taylor, D. B. J., T. A. Heard, Q. Paynter, and H. Spafford. 2007. Nontarget effects of a weed biological control agent on a native plant in Northern Australia. Biological Control 42:25?33. Thomas, M. B., and A. J. Willis. 1998. Biocontrol ? risky but necessary? Trends in Ecology & Evolution 13:325?329. Thompson, J. N. 1978. Within-patch structure and dynamics in Pastinaca sativa and resource availability to a specialized herbivore. Ecology 59:443?448. Thompson, J. N. 1998. Rapid evolution as an ecological process. Trends in Ecology & Evolution 13:329?332. Trumble, J. T., D. M. Kolodny-Hirsch, and I. P. Ting. 1993. Plant compensation for arthropod herbivory. Annual Review of Entomology 38:93?119. U. S. Department of Agriculture. 2010. Pest Alert: Mogulones cruciger. Animal and Plant Health Inspection Service, APHIS 81-35-014. U. S. Fish and Wildlife Service. 2007. Recovery Plan for Hackelia venusta (Showy Stickseed). U.S. Fish and Wildlife Service, Portland, Oregon. U. S. Fish and Wildlife Service. 2011. Hackelia venusta (Showy Stickseed) 5-year review summary and evaluation. U.S. Fish and Wildlife Service, Lacey, Washington. Upadhyaya, M. K., and R. S. Cranston. 1991. Distribution, biology, and control of hound's-tongue in British Columbia. Rangelands 13:103?106.  136 Upadhyaya, M. K., H. R. Tilsner, and M. D. Pitt. 1988. The biology of Canadian weeds. 87. Cynoglossum officinale L. Canadian Journal of Plant Science 68:763?774. Van Breemen, A. M. M. 1984. Comparative germination ecology of three short-lived monocarpic Boraginaceae. Acta Botanica Neerlandica 33:283?305. van der Meijden, E., P. Klinkhamer, T. De Jong, and C. Van Wijk. 1992. Meta-population dynamics of biennial plants: how to exploit temporary habitats. Acta Botanica Neerlandica 41:249?270. van Hezewijk, B. H., R. A. De Clerck-Floate, and J. R. Moyer. 2008. Effect of nitrogen on the preference and performance of a biological control agent for an invasive plant. Biological Control 46:332?340. van Klinken, R. D., and O. R. Edwards. 2002. Is host-specificity of weed biological control agents likely to evolve rapidly following establishment? Ecology 5:590?596. Vance, J. M. 2013, June. An examination of the soils supporting Hackelia venusta, Washington State's most endangered species. MSc Thesis, University of Washington. Visser, J. H. 1988. Host-plant finding by insects: Orientation, sensory input and search patterns. Journal of Insect Physiology 34:259?268. Wesselingh, R. A., P. G. L. Klinkhamer, T. J. de Jong, and L. A. Boorman. 1997. Threshold size for flowering in different habitats: effects of size-dependent growth and survival. Ecology 78:2118?2132. White, J. A., and T. G. Whitham. 2000. Associational susceptibility of cottonwood to a box elder herbivore. Ecology 81:1795?1803. Wiggins, G. J., J. F. Grant, P. L. Lambdin, J. W. Ranney, J. B. Wilkerson, and F. T. van Manen. 2010. Spatial prediction of habitat overlap of introduced and native thistles to identify potential areas of nontarget activity of biological control agents. Environmental Entomology 39:1866?1877. Wilcove, D. S., D. Rothstein, J. Dubow, A. Phillips, and E. Losos. 1998. Quantifying threats to imperiled species in the United States. BioScience 48:607?615. Williams, J. L. 2009. Flowering life-history strategies differ between the native and introduced ranges of a monocarpic perennial. The American Naturalist 174:660?672.  137 Williams, J. L., H. Auge, and J. L. Maron. 2010. Testing hypotheses for exotic plant success: parallel experiments in the native and introduced ranges. Ecology 91:1355?1366. Williams, J. L., M. M. Ellis, M. C. Bricker, J. F. Brodie, and E. W. Parsons. 2011. Distance to stable stage distribution in plant populations and implications for near-term population projections. Journal of Ecology 99:1171?1178. Withers, T. 1999. Examining the hierarchy threshold model in a no-choice feeding assay. Entomologia Experimentalis et Applicata 91:89?95. Yang, L. H. 2012. The ecological consequences of insect outbreaks. Pages 197?218 in P. Barbosa, D. K. Letourneau, and A. A. Agrawal, editors. Insect Outbreaks Revisited. Wiley-Blackwell, Chichester, UK. Zw?lfer, H., and P. Harris. 1984. Biology and host specificity of Rhinocyllus conicus (Froel.) (Col., Curculionidae), a successful agent for biocontrol of the thistle, Carduus nutans L. Zeitschrift der Angewandte Entomologie 97:36?62.   138 APPENDICES  APPENDIX A  PLANT FECUNDITY CALCULATIONS  139 Table A.1. Sources of the values of number of tetrads per bolting stems used for calculating plant fecundity values.   Year Number of Bolts Proportion counted (i.e., bolts at least 75% intact) Proportion of bolt fecundity values estimated from within-site height regressions Proportion of bolt fecundity values estimated from within-year site averages Proportion of bolt fecundity values estimated from within-year overall height regressions or averages       H. micrantha      2009 694 0.572 0.000 0.415 0.013 2010 895 0.505 0.302 0.193 0.000 2011 1242 0.546 0.241 0.187 0.027       C. officinale      2009 148 0.872 0.000 0.128 0.000 2010 117 0.872 0.034 0.094 0.000 2011 166 0.964 0.012 0.024 0.000     140 APPENDIX B  CALCULATING UPPER AND LOWER PARAMETER BOUNDARIES FROM THE SEEDLING EMERGENCE EXPERIMENT   In June 2011, 3.4% of C. officinale seeds and 10.2% of H. micrantha seeds sown in October 2009,  appeared as seedlings without cotyledons, making it impossible to distinguish whether these plants had overwintered as seedlings or were new germinating individuals. These scenarios produced different estimates of seed survival and dormancy, and seedling survival rates. I modelled both scenarios.  I used proportions pooled among all relevant sowing quadrats from the June 2010 and June 2011 counts to define relationships between counts and parameters with the equations below, using with separate calculations for the two seedling overwintering scenarios:  N seedlings June 2010/Corrected N seeds sown = sF  ? gF ? (1- jF)    N new seedlings June 2011/Corrected N seeds sown =  sF ? (1- gF) ? sD ? gD  N small rosettes June 2010/Corrected N seeds sown = sF ? gF ? jF    N overwintered seedlings June 2011/N seedlings June 2010) = sG ? (1-gG)  N small rosettes June 2010/N seedlings June 2010 = sG ? gG  Separately, I calculated upper or lower boundaries for some parameters. The lower boundaries for seed survival were the minimum number of individuals that germinated under both overwintering scenarios using June and August counts in 2010 and 2011. Minimum and maximum germination (i.e., seeds that did not go dormant) under the different overwintering scenarios were calculated by dividing the number of new germinators in 2011 by the maximum (1.0) and minimum seed survival rates. The  141 minimum probability a dormant seed germinates in the second year was calculated as the number of new seedlings in June 2011 divided by the number of ungerminated seeds in 2010. Seedling survival and growth were determined from the number of seedlings and rosettes in June 2010 and June 2011.    The minimum germination rate of sown seeds was 48-52% for C. officinale and 36-46% for H. micrantha, depending on the overwintering scenario. As I was unable to determine if ungerminated seeds had died or remained dormant in the soil, I examined 5 scenarios for seedling dynamics: 1) maximum survival with minimum dormancy, 2) maximum survival with maximum dormancy, 3) mid-level survival with mid-level dormancy, 4) minimum survival with minimum dormancy, and 5) minimum survival with maximum dormancy.  I calculated separate matrices (each with and without seedling overwintering) for scenarios 1-3 for C. officinale as it has only a 1-2 year seed bank (Van Breemen, 1984; Williams et al., 2010), using mean rosette and bolter parameter values from all sites in both years. The seed bank dynamics of H. micrantha are unknown, therefore I calculated separate matrices for scenarios 1-5 for the nontarget plant, using mean rosette and bolter parameter values from all sites in both years.  I chose biologically reasonable scenarios of maximum survival with minimum dormancy and successful seedling overwintering for C. officinale, and mid-level survival with mid-level dormancy and successful seedling overwintering for H. micrantha to enter in the overall species transition matrices. The different seed and seedling scenarios changed the lambda values by less than 2.5% in either direction.  All seed and seedling parameters were tested for density-dependence using generalized linear models (GLM) with binary response variables (i.e., germinated, ungerminated), binomial errors and logit link functions in the program R (R Core Team 2012). When models indicated overdispersion, I re-fit the data with quasibinomial errors, and compared models with the F ratio test. When density was not a significant predictor, the GLM returned values equal to the overall pooled proportion of all quadrats.  Control quadrats (i.e., those with no seeds sown) had up to 2 C. officinale and 3 H. micrantha seedlings throughout the experiment. Because of these extraneous seeds, I excluded the  142 10 seeds/quadrat treatments from my analysis because even such a low number of extra germinators would be very influential to proportions in these quadrats. Analyses presented for this experiment are therefore based on quadrats with at least 100 seeds sown.    143 Table B.1. Parameter values for C. officinale and H. micrantha seed and seedling dynamics.  C. officinale seedling overwintering C. officinale       no seedling overwintering H. micrantha seedling overwintering H. micrantha     no seedling overwintering  Total seeds sown 4248 4248 4091 4091  Proportion of seeds that may have overwintered seedling to seedling 0.0337 0 0.1022 0 Germination rate June 2010 0.2870 0.2870 0.3315 0.3315  Proportion of seeds sown that appeared as new seedlings in June 2011 0.1933 0.2269 0.0240 0.1261  Proportion of seeds sown that appeared as small rosettes in June 2011 0 0 0 0  Proportion of seedlings from June 2010 that overwintered to June 2011 0.1173 0 0.3083 0  Proportion of seedlings from June 2010 growing to rosettes in June 2011 0.0221 0.0221 0.1394 0.1394  Proportion of seeds surviving (sF) 0.4826-1 0.5162-1 0.3566-1 0.4588-1  Proportion of surviving seeds on suitable habitat that germinated (gF) 0.2870-0.5946 0.2870-0.5559 0.3315-0.9294 0.3315-0.7224  Proportion of seeds surviving in the seed bank (sD) 0.2731-1 0.3204-1 0.0373-1 0.1902-1  Proportion of surviving seeds in the seed bank that germinate (gD) 0.2715-1 0.3188-1 0.0358-1 0.1887-1  Proportion of seedlings that survive to the next year (sG) 0.1395 (density-dependent) 0.0222 (density-dependent) 0.4476 0.1394  Proportion of surviving seedlings growing to rosettes (gG) 0.1588 1 0.3114 1 Proportion of germinating seeds jumping to small rosette (j) 0 0 0 0            144           Figure B.1. Probability that C. officinale (target) seedlings from June 2010 survived to June 2011 as either seedlings (if seedlings overwinter), or small rosettes (no seedling overwintering). Values are calculated from 30 0.5m2 quadrats where 100, 250, or 500 C. officinale seeds were sown in October 2009.  Lines represent generalized linear models for scenarios with seedling overwintering (top line) and no overwintering (bottom line). Seedling density in June 2010 was a significant explanatory variable in the models.  E) HT with density dependence0.00.20.40.60 200 400 600 800 1000Seedling survival (proportion)C. officinale seedlings m-2E) HT with density dependence0.00.20.40.60 200 400 600 800 1000Seedling survival (proportion)C. officinale seedlings m-2 145 APPENDIX C  INDIVIDUAL SITE DEMOGRAPHIC PARAMETERS   146 Table C.1. Overall C. officinale rosette and bolting transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories. Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion surviving, transitioning seedlings to large rosette Proportion bolting, given survival and transition Non-release Year 0-1         Site NY116         Seedlings n/a n/a 6 0 n/a n/a 0.1667 0 Small rosettes 58 33 14 3 0.5690 0.4242  0.2143 Large rosettes 25 19 12 9 0.7600 0.6316  0.7500 Bolters 15 0 0 15 0 0            Site NY138         Seedlings n/a n/a 2 0 n/a n/a 0 0 Small rosettes 25 19 6 1 0.7600 0.3158  0.1667 Large rosettes 21 15 5 3 0.7143 0.3333  0.6000 Bolters 7 0 0 7 0 0            Site NY139         Seedlings n/a n/a 9 0 n/a n/a 0.2222 0 Small rosettes 11 10 7 1 0.9091 0.7000  0.1429 Large rosettes 19 17 14 13 0.8947 0.8235  0.9286 Bolters 13 1 0 13 0.0769 0            Release Year 0-1         Site YY127         Seedlings n/a n/a 10 0 n/a n/a 0 0 Small rosettes 21 8 3 2 0.3810 0.3750  0.6667 Large rosettes 5 2 0 0 0.4000 0.0000  0.0000 Bolters 8 0 0 8 0 0    147 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion surviving, transitioning seedlings to large rosette Proportion bolting, given survival and transition          Site YY136         Seedlings n/a n/a 1 0 n/a n/a 0 0 Small rosettes 34 8 2 1 0.2353 0.2500  0.5000 Large rosettes 15 6 3 2 0.4000 0.5000  0.6667 Bolters 11 1 0 11 0.0909 0            Site YY137         Seedlings n/a n/a 20 0 n/a n/a 0.1000 0 Small rosettes 70 38 16 6 0.5429 0.4211  0.3750 Large rosettes 12 8 5 5 0.6667 0.6250  1.0000 Bolters 15 0 0 15 0 0            Non-release Year 1-2         Site NY116         Seedlings n/a n/a 7 0 n/a n/a 0.1429 0 Small rosettes 31 11 7 1 0.3548 0.6364  0.1429 Large rosettes 19 16 11 11 0.8421 0.6875  1.0000 Bolters 13 0 0 13 0 0            Site NY138         Seedlings n/a n/a 15 0 n/a n/a 0.0667 0 Small rosettes 29 22 13 0 0.7586 0.5909  0 Large rosettes 16 13 3 3 0.8125 0.2308  1.0000 Bolters 4 0 0 4 0 0                     Table C.1. (continued) Overall C. officinale rosette and bolting transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   148 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion surviving, transitioning seedlings to large rosette Proportion bolting, given survival and transition          Site NY139         Seedlings n/a n/a 13 0 n/a n/a 0.1538 0 Small rosettes 17 17 13 2 1.0000 0.7647  0.1538 Large rosettes 12 11 9 9 0.9167 0.8182  1.0000 Bolters 15 4 1 15 0.2667 0.2500            Release Year 1-2         Site YY127         Seedlings n/a n/a 3 0 n/a n/a 0 0 Small rosettes 16 10 3 0 0.6250 0.3000  0 Large rosettes 4 2 2 2 0.5000 1.0000  1.0000 Bolters 2 0 0 2 0 0            Site YY136         Seedlings n/a n/a 55 0 n/a n/a 0.0909 0 Small rosettes 9 6 2 0 0.6667 0.3333  0 Large rosettes 4 3 1 1 0.7500 0.3333  1.0000 Bolters 4 0 0 4 0 0            Site YY137         Seedlings n/a n/a 20 0 n/a n/a 0.5000 0 Small rosettes 59 35 29 8 0.5932 0.8286  0.2759 Large rosettes 18 18 16 16 1.0000 0.8889  1.0000 Bolters 13 2 0 13 0.1538 0    Table C.1. (continued) Overall C. officinale rosette and bolting transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   149  Table C.2.  Overall H. micrantha rosette and bolter transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories. Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion bolting, given survival and transitioning Proportion becoming large rosettes, given survival, transitioning, no bolting Proportion becoming large bolters, given survival, transitioning, bolting           Non-release sites with C. officinale Year 0-1         Site NY116          Seedlings n/a n/a 1 0 n/a n/a 0 0  Small rosettes 6 3 1 0 0.5000 0.3333 0.0000   Large rosettes 5 5 4 2 1.0000 0.8000 0.5000  0.5000 Small bolters 8 8 3 1 1.0000 0.3750 0.3333 0.5000  Large bolters 10 7 2 2 0.7000 0.2857 1.0000             Site NY138          Seedlings n/a n/a 1 0 n/a n/a 0 0  Small rosettes 18 17 7 0 0.9444 0.4118 0.0000   Large rosettes 14 13 2 2 0.9286 0.1538 1.0000  0.0000 Small bolters 8 8 5 1 1.0000 0.6250 0.2000 1.0000  Large bolters 1 1 1 1 1.0000 1.0000 1.0000                                  150 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion bolting, given survival and transitioning Proportion becoming large rosettes, given survival, transitioning, no bolting Proportion becoming large bolters, given survival, transitioning, bolting Site NY139          Seedling n/a n/a 1 0 n/a n/a 0 0  Small rosettes 11 11 5 1 1.0000 0.4545 0.2000   Large rosettes 20 20 13 13 1.0000 0.6500 1.0000  0.6154 Small bolters 20 20 10 3 1.0000 0.5000 0.3000 0.8571  Large bolters 11 11 5 4 1.0000 0.4545 0.8000             Non-release sites with no C. officinale Year 0-1         Site NN123          Seedling n/a n/a 0 0 n/a n/a 0 0  Small rosettes 17 13 3 0 0.7647 0.2308 0.0000   Large rosettes 7 7 3 3 1.0000 0.4286 1.0000  0.0000 Small bolters 11 11 6 4 1.0000 0.5455 0.6667 1.0000  Large bolters 6 6 2 1 1.0000 0.3333 0.5000             Site NN129          Seedling n/a n/a 0 0 n/a n/a 0 0  Small rosettes 16 6 3 0 0.3750 0.5000 0.0000   Large rosettes 13 11 7 7 0.8462 0.6364 1.0000  0.4286 Small bolters 13 11 8 7 0.8462 0.7273 0.8750 1.0000  Large bolters 8 7 1 1 0.8750 0.1429 1.0000   Table C.2. (continued) Overall H. micrantha rosette and bolter transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   151 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion bolting, given survival and transitioning Proportion becoming large rosettes, given survival, transitioning, no bolting Proportion becoming large bolters, given survival, transitioning, bolting Site NN134          Seedling n/a n/a 12 0 n/a n/a 0 0.0833  Small rosettes 15 9 4 1 0.6000 0.4444 0.2500   Large rosettes 14 13 9 8 0.9286 0.6923 0.8889  0.5000 Small bolters 18 17 13 10 0.9444 0.7647 0.7692 1.0000  Large bolters 8 8 4 2 1.0000 0.5000 0.5000             Release sites with C. officinale Year 0-1         Site YY127          Seedling n/a n/a 3 0 n/a n/a 0 0  Small rosettes 37 26 5 4 0.7027 0.1923 0.8000   Large rosettes 19 17 11 9 0.8947 0.6471 0.8182  0.2222 Small bolters 20 18 10 7 0.9000 0.5556 0.7000 0.6667  Large bolters 8 8 4 4 1.0000 0.5000 1.0000             Site YY136          Seedling n/a n/a 5 0 n/a n/a 0 0  Small rosettes 33 27 6 1 0.8182 0.2222 0.1667   Large rosettes 6 6 4 4 1.0000 0.6667 1.0000  0.5000 Small bolters 13 12 7 2 0.9231 0.5833 0.2857 0.8000  Large bolters 7 7 3 3 1.0000 0.4286 1.0000   Table C.2. (continued) Overall H. micrantha rosette and bolter transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   152 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion bolting, given survival and transitioning Proportion becoming large rosettes, given survival, transitioning, no bolting Proportion becoming large bolters, given survival, transitioning, bolting Site YY137          Seedling n/a n/a 9 0 n/a n/a 0 0  Small rosettes 12 11 4 1 0.9167 0.3636 0.2500   Large rosettes 12 12 6 6 1.0000 0.5000 1.0000  0.3333 Small bolters 4 3 1 0 0.7500 0.3333 0.0000 1.0000  Large bolters 6 6 2 1 1.0000 0.3333 0.5000             Release sites with no C. officinale Year 0-1         Site YN122          Seedling n/a n/a 6 0 n/a n/a 0 0  Small rosettes 9 9 1 0 1.0000 0.1111 0.0000   Large rosettes 13 12 9 7 0.9231 0.7500 0.7778  0.8571 Small bolters 14 14 8 4 1.0000 0.5714 0.5000 0.5000  Large bolters 6 6 3 0 1.0000 0.5000 0.0000             Site YN125          Seedling n/a n/a 3 0 n/a n/a 0 0  Small rosettes 17 7 2 0 0.4118 0.2857 0.0000   Large rosettes 12 12 5 5 1.0000 0.4167 1.0000  0.0000 Small bolters 14 14 10 1 1.0000 0.7143 0.1000 0.7778  Large bolters 13 13 5 4 1.0000 0.3846 0.8000   Table C.2. (continued) Overall H. micrantha rosette and bolter transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   153 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion bolting, given survival and transitioning Proportion becoming large rosettes, given survival, transitioning, no bolting Proportion becoming large bolters, given survival, transitioning, bolting Site YN131          Seedling n/a n/a 10 0 n/a n/a 0 0  Small rosettes 47 38 11 2 0.8085 0.2895 0.1818   Large rosettes 30 27 10 6 0.9000 0.3704 0.6000  0.1667 Small bolters 6 5 3 1 0.8333 0.6000 0.3333 0.5000  Large bolters 0 0 0 0 0.0000 0.0000 0.0000             Non-release sites with C. officinale Year 1-2         Site NY116          Seedling n/a n/a 2 0 n/a n/a 0 0  Small rosettes 8 4 1 0 0.5000 0.2500 0.0000   Large rosettes 3 3 2 1 1.0000 0.6667 0.5000  1.0000 Small bolters 8 7 5 1 0.8750 0.7143 0.2000 1.0000  Large bolters 7 6 2 1 0.8571 0.3333 0.5000             Site NY138          Seedling n/a n/a 2 0 n/a n/a 0 0  Small rosettes 12 12 5 1 1.0000 0.4167 0.2000   Large rosettes 22 22 10 9 1.0000 0.4545 0.9000  0.0000 Small bolters 6 6 2 2 1.0000 0.3333 1.0000 0.0000  Large bolters 1 1 1 1 1.0000 1.0000 1.0000   Table C.2. (continued) Overall H. micrantha rosette and bolter transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   154 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion bolting, given survival and transitioning Proportion becoming large rosettes, given survival, transitioning, no bolting Proportion becoming large bolters, given survival, transitioning, bolting Site NY139          Seedling n/a n/a 1 0 n/a n/a 0 0  Small rosettes 16 10 6 1 0.6250 0.6000 0.1667   Large rosettes 21 19 15 14 0.9048 0.7895 0.9333  0.5000 Small bolters 20 19 13 12 0.9500 0.6842 0.9231 1.0000  Large bolters 17 17 0 0 1.0000 0.0000 0.0000             Non-release sites with no C. officinale Year 1-2         Site NN123          Seedling n/a n/a 4 0 n/a n/a 0 0  Small rosettes 14 10 1 0 0.7143 0.1000 0.0000   Large rosettes 10 10 6 4 1.0000 0.6000 0.6667  0.5000 Small bolters 9 9 5 2 1.0000 0.5556 0.4000 1.0000  Large bolters 8 8 2 2 1.0000 0.2500 1.0000             Site NN129          Seedling n/a n/a 3 0 n/a n/a 0 0.3333  Small rosettes 3 1 0 0 0.3333 0.2963* 0.0000   Large rosettes 8 8 7 7 1.0000 0.8750 1.0000  0.2857 Small bolters 8 8 6 5 1.0000 0.7500 0.8333 1.0000  Large bolters 16 16 4 2 1.0000 0.2500 0.5000   Table C.2. (continued) Overall H. micrantha rosette and bolter transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   155 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion bolting, given survival and transitioning Proportion becoming large rosettes, given survival, transitioning, no bolting Proportion becoming large bolters, given survival, transitioning, bolting Site NN134          Seedling n/a n/a 4 0 n/a n/a 0 0  Small rosettes 24 14 2 1 0.5833 0.1429 0.5000   Large rosettes 14 13 9 8 0.9286 0.6923 0.8889  0.2500 Small bolters 11 11 10 8 1.0000 0.9091 0.8000 1.0000  Large bolters 18 16 3 3 0.8889 0.1875 1.0000             Release sites with C. officinale Year 1-2         Site YY127          Seedling n/a n/a 10 0 n/a n/a 0 0  Small rosettes 34 32 5 1 0.9412 0.1563 0.2000   Large rosettes 12 12 5 3 1.0000 0.4167 0.6000  0.0000 Small bolters 24 24 15 3 1.0000 0.6250 0.2000 0.7500  Large bolters 14 14 6 5 1.0000 0.4286 0.8333             Site YY136          Seedling n/a n/a 9 0 n/a n/a 0 0  Small rosettes 28 25 8 1 0.8929 0.3200 0.1250   Large rosettes 11 11 4 4 1.0000 0.3636 1.0000  0.2500 Small bolters 11 11 5 0 1.0000 0.4545 0.0000 1.0000  Large bolters 8 8 6 4 1.0000 0.7500 0.6667   Table C.2. (continued) Overall H. micrantha rosette and bolter transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   156 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion bolting, given survival and transitioning Proportion becoming large rosettes, given survival, transitioning, no bolting Proportion becoming large bolters, given survival, transitioning, bolting Site YY137          Seedling n/a n/a 20 0 n/a n/a 0 0  Small rosettes 29 19 9 1 0.6552 0.4737 0.1111   Large rosettes 12 11 8 8 0.9167 0.7273 1.0000  0.1250 Small bolters 8 8 4 3 1.0000 0.5000 0.7500 1.0000  Large bolters 6 6 0 0 1.0000 0.0000 0.0000             Release sites with no C. officinale Year 1-2         Site YN122          Seedling n/a n/a 0 0 n/a n/a 0 0  Small rosettes 27 22 3 0 0.8148 0.1364 0.0000   Large rosettes 9 9 6 6 1.0000 0.6667 1.0000  0.5000 Small bolters 8 8 7 3 1.0000 0.8750 0.4286 1.0000  Large bolters 13 13 5 5 1.0000 0.3846 1.0000             Site YN125          Seedling n/a n/a 5 0 n/a n/a 0 0  Small rosettes 10 9 4 0 0.9000 0.4444 0.0000   Large rosettes 17 17 6 6 1.0000 0.3529 1.0000  0.1667 Small bolters 13 13 10 7 1.0000 0.7692 0.7000 1.0000  Large bolters 9 9 2 2 1.0000 0.2222 1.0000   Table C.2. (continued) Overall H. micrantha rosette and bolter transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   157 Starting stage No. plants No. surviving No. growing No. bolting Proportion surviving Proportion transitioning, given survival Proportion bolting, given survival and transitioning Proportion becoming large rosettes, given survival, transitioning, no bolting Proportion becoming large bolters, given survival, transitioning, bolting           Site YN131          Seedling n/a n/a 8 0 n/a n/a 0 0  Small rosettes 50 41 9 2 0.8200 0.2195 0.2222   Large rosettes 28 25 12 9 0.8929 0.4800 0.7500  0.1111 Small bolters 9 9 4 1 1.0000 0.4444 0.2500 0.6667  Large bolters 2 2 1 1 1.0000 0.5000 1.0000    Table C.2. (continued) Overall H. micrantha rosette and bolter transitions for each site-year combination. Letter codes before site numbers refer to yes/no for M. crucifer release and ?target common? site categories.   158           Figure C.1. Life table response experiment results for contributions in changes to C. officinale lambda values on each site from year 0-1 to year 1-2.    n116 Non-release-0.10-0.050.000.050.100.150.20G-SRG-LRSR-SRSR-LRSR-BLR-SRLR-LRLR-BB-DB-GB-SR B-LRB-Bn138 Non-releaseG-SRG-LRSR-SRSR-LRSR-BLR-SRLR-LRLR-BB-DB-GB-SR B-LRB-Bn139 Non-releaseTransitionsG-SRG-LRSR-SRSR-LRSR-BLR-SRLR-LRLR-BB-DB-GB-SR B-LRB-By127 Release-0.10-0.050.000.050.100.150.20y136 Release y137 ReleaseTransitionsContribution to change in lambdaLTRE of HT on each site (ref yr0-1, trt yr1-2) 159    Figure C.2. Life table response experiment results for contributions in changes to H. micrantha lambda values on each site from year 0-1 to year 1-2.   NY116 Non-release, Target Common-0.6-0.5-0.4-0.3-0.2-0.10.00.10.20.3G-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNY138 Non-release, Target CommonG-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNY139 Non-release, Target CommonG-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBYY127 Release, Target Common-0.6-0.5-0.4-0.3-0.2-0.10.00.10.20.3G-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBYY136 Release, Target CommonG-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBYY137 Release, Target CommonG-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNN123 Non-release, Target Rare-0.6-0.5-0.4-0.3-0.2-0.10.00.10.20.3G-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNN129 Non-release, Target RareG-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBNN134 Non-release, Target RareG-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBYN122 Release, Target Rare-0.6-0.5-0.4-0.3-0.2-0.10.00.10.20.3G-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBYN125 Release, Target RareG-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBYN131 Release, Target RareG-SR G-LRSR-SRSR-LRSR-SBLR-SRLR-LRLR-SBLR-LBSB-DSB-GSB-SRSB-LRSB-SBSB-LB LB-DLB-GLB-SRLB-LRLB-SBLB-LBLTRE for BS on each site ref=09-10, treatment 10-11TransitionsContributions to difference in lambda

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0074334/manifest

Comment

Related Items