An Abstract of the Thesis of

Shon S. Schooler for the degree of Master of Science in Entomology presented on May

7th, 1998. Title: Biological Control of Purple Loosestrife Lythrum salicaria By Two

Chrysomelid Beetles Galerucella pusilla and G. calmariensis

Redacted for Privacy Abstract approved: Peter B. McEv

In the first part of this study we monitored the development of biological control of purple loosestrife Lythrum salicaria over a six-year period at Morgan Lake in western

Oregon. In 1992, two beetles, Galerucella pusilla and G. calmariensis (Coleoptera:

Chrysomelidae), were released to control the wetland weed at this test site. Our purpose was to estimate quantitative performance parameters that might be generally applied in monitoring biological weed control. Our six performance measures were: 1) biological control agent establishment, 2) the rate of increase of the agents, 3) the rate of spread of the agents, 4) the effect of the agents on individual target plants, 5) the effect of the agents on the population of the target plants, and 6) the indirect impact of the biological control agents on the local plant community.

The beetles established viable populations that increased during the study with an intrinsic rate of increase (r), based on the growth rate in damage, estimated at 2.24/year.

Within six years after introduction, the beetles spread to saturate the entire purple loosestrife habitat (4100 m2) around the lake. The rate of spread, estimated by calculating a diffusion coefficient (D), was 57.5 m2/year. Adult beetles made seasonal, exploratory movements up to 30 m away from the host plant stand into surrounding crop fields, which suggests a disturbance-free buffer should be established in the habitat surrounding the

loosestrife stand. By 1997, both flowering success and median stem density (per 0.125 m2

plot) of purple loosestrife declined to zero. Mean above-ground biomass decreased to

8.4% of its 1994 level. Biomass of native plant species increased by only 3% between

1996 and 1997. Overall, G. pusilla and G. calmariensis reduced the abundance of the target plant at our site. Our monitoring methods were effective at quantitatively

measuring the establishment, increase, spread, and damage of the biological control agent,

the subsequent decline of the target plant, and the impact on the local plant community.

The second part of our study used field and greenhouse experiments to assess non-

target effects of two introduced biological control organisms (Galerucella pusilla

Duftschmid and G. calmariensis L.: Chrysomelidae) on the economically important

ornamental plant, crape myrtle (Lagerstroemia indica L: Lythraceae). Prior host

specificity tests performed in the laboratory found that beetles fed, but were unable to

complete their life cycle, on this non-target plant. However, there was concern over

damage that might occur when the two plant species existed together. This study

extended prior tests into a field environment in order to compare the physiological host

range revealed in greenhouse tests with the ecological host range revealed in the field.

We assumed, based on prior evidence, that the control agents would not complete development on the non-target plant, and therefore, when the non-target organism was

isolated from populations of the target organism the direct effects of the biological control

agents would be negligible. When the target and non-target organisms existed together, the magnitude of indirect effect of the target organism on the non-target organism via the

control agent was expected to increase with: 1) decreasing distance between the target and non-target organisms, and 2) increasing dispersal capability of the control agents. As expected from prior studies beetle feeding and oviposition occurred on crape myrtle but the beetles could not complete development on this non-target plant in our greenhouse and field tests. Leaf damage inflicted by the beetle was lower on crape myrtle than on purple loosestrife plants used as controls and extensive defoliation to the non-target plant was limited to within 30 m from the edge of the purple loosestrife stand. Biomass of crape myrtle was significantly reduced near the stand compared with plants that remained relatively untouched at greater distances. Purple loosestrife biomass exhibited a greater reduction with decreasing distance from the source of beetle colonization.

In this thesis we construct and implement strategies for quantitatively assessing success of biological control programs and risk of introduced biological agents to non- target organisms. Through these observations and experiments we hope to increase the predictability and safety of biological control programs. ©Copyright by Shon S. Schooler May 7th, 1998 All Rights Reserved Biological Control of Purple Loosestrife Lythrum salicaria By Two Chrysomelid Beetles Galerucella pusilla and G. calmariensis

by

Shon S. Schooler

A THESIS

submitted to

Oregon State University

in partial fulfillment of the requirements for the degree of

Master of Science

Presented May 7, 1998 Commencement June 1998 Master of Science thesis of Shon S. Schooler presented on May 7, 1998

APPROVED: Redacted for Privacy Major Professor, representing Entoni;dlogy Redacted for Privacy

Chair 6f Department of domology Redacted for Privacy

Dean of Grad

I understand that my thesis will become part of the permanent collection of Oregon State University Libraries. My signature below authorizes re se of my thesis to any reader upon request. Redacted for Privacy

Shon S. Schooler, Author Acknowledgements

I would like to extend thanks to some of the many that deserve recognition. My major advisor, Peter McEvoy, gets the award for patience. He managed, despite all odds, to illuminate for me the dark mathematical recesses of theoretical ecology. Eric Coombs was always ready to put things back into perspective with his sense of humor and eternal optimism. Hans Luh, the computer guru, helped with statistical analyses. Rich Guadagno, the manager at the study site, smoothed things over with the boss. Thanks to my graduate committee members: Chris Mundt, Gary DeLander, Paul Murtaugh, and Paul Jepson for their help in moving forward. A firm handshake to my former mentor, Tony Ives, for starting me down this long road. Special thanks to Andy and Alison Moldenke for letting me stay at their house while it was on the market to be sold. Thanks to all the entomology graduate students, especially Tim and Brian, who made life here so wonderful. Finally, thanks to my parents and brother who instilled within me the strength and perseverance to finish the things that I begin.

I have been fortunate in my academic wanderings to meet many caring individuals, some of which I have surely forgotten to mention here. Thanks to all those I may have overlooked. Contribution of Authors

Eric Coombs was involved in the experimental design used in both manuscripts.

Peter McEvoy was instrumental in the analysis and writing of Chapter 3 and the design, writing, and analysis of Chapter 2. Hans Luh gave valuable insight into alternative analysis methods in Chapter 2. Don Jo ley assisted in the design of Chapter 3. Table of Contents

Page

Chapter 1 Introduction 1

Chapter 2 Increase, Spread, and Ecological Effects of Two Beetle Species Galerucella calmariensis and G. pusilla Introduced to Control Purple Loosestrife Lythrum salicaria In Oregon 3

Abstract 4

Introduction 6

Study System 15

Methods 19

Data Analysis 29

Results 35

Discussion 51

Conclusions 64

Chapter 3 Field Test of Host Specificity of Two Chrysomelid Beetles (Galerucella pusilla and G. calmariensis) Introduced to North America For Biocontrol of Purple Loosestrife Lythrum salicaria 68

Abstract 69

Introduction 71

Study System 77

Methods 83

Results 87

Discussion 93

Conclusions 95 Table of Contents (continued) Pane

Chapter 4 Summary 97

Bibliography 99

Appendices 107 List of Figures

Figure Page

2.1 Design for: (a) intensive, (b) biomass, and (c) extensive sampling 21

2.2 Median percent beetle defoliation increased exponentially from 1994 to 1997 36

2.3 Log-odds of leaf damage as a function of distance from the initial beetle release site (X) from 1994 to 1997 37

2.4 Populations of G. calmariensis and G. pusilla at Morgan Lake continued to spread over the six year monitoring period 43

2.5 The biological control agents spread from a single release point to infest most purple loosestrife plants within a 4100 m2 area at Morgan Lake within 5 years 44

2.6 Beetles were found to spread from the purple loosestrife stand into adjacent fields throughout the sampling period during 1996 46

2.7 Leaf damage increased from very low levels in 1994-1995, to a range of intermediate levels in 1996, and then to very high levels in 1997 47

2.8 Flowering probability increased with stem length 49

2.9 Purple loosestrife response to herbivory by G. pusilla and G. calmariensis over a four-year study 50

2.10 Biomass of purple loosestrife declined precipitously between 1996 and 1997 52

2.11 Plant community composition at Morgan Lake during 1996 and 1997 53

3.1 Indirect effects of the target organism cause damage to the non-target organism via the control agent 72

3.2 Distribution of purple loosestrife (dots) and crape myrtle (shaded zone) in the United States and Canada (Thompson et al. 1987; Martin 1983) 79

3.3 Study Site: Morgan Lake (Baskett Slough National Wildlife Refuge: Western Oregon) 82

3.4 Experimental design of the non-target field study 85 List of Figures (Continued)

Figure Page

3.5 Chart used to train observers to quantify leaf damage of Galerucella spp. on L. salicaria 86

3.6 Larvae of G. pusilla and G. calmariensis were able to develop on the target plant L. salicaria but not the non-target plant L. indica 88

3.7 Percent leaf defoliation decreased with distance from the colonization source for both plant species 90

3.8 Total dry weight of plants at five distances from a G. pusilla and G. calmariensis infested stand of L. salicaria 91 List of Tables

Table Page

2.1 Reported values of "r" for invasive species 10

2.2 Summary of diffusion coefficients for insects 11

2.3 Reported values of "q" for biological control programs 14

2.4 Summary of sampling dates and distances 22

2.5 Summary of insect, plant, and other variables 24

2.6a Summary of spatial pattern in insect, damage, and plant variables 39

2.6b Summary of spatial pattern in insect, damage, and plant variables 40

2.7 Summary of odds ratios 41

2.8 Biomass data for plant community at Morgan Lake 54

3.1 Forty-eight plant species approved for host specificity screening with biological control agents for L. salicaria 74

3.2 Dry weight of Lythrum salicaria and Lagerstroemia indica (grams) 92 Appendix Figures

Figure Page

2.1 Study Site: Morgan Lake (Baskett Slough National Wildlife Refuge: Western Oregon) 107

2.2 Leaf damage chart used to determine leaf damage estimates for Galerucella species on L. salicaria 108

2.3 Log-odds of beetle eggs with respect to distance from the initial beetle release site (X) over a four year period 109

2.4 Log-odds of beetle eggs with respect to distance toward the lake (Y) over a four year period 110

2.5 Log-odds of beetle larvae with respect to distance from the initial beetle release site (X) over a four year period 111

2.6 Log-odds of beetle larvae with respect to distance toward the lake (Y) over a four year period 112

2.7 Log-odds of adult beetles with respect to distance from the initial beetle release site (X) over a four year period 113

2.8 Log-odds of adult beetles with respect to distance toward the lake (Y) over a four year period 114

2.9 Loge mean number of adult beetles with respect to distance from the initial beetle release site (X) over a four year period 115

2.10 Loge mean number of adult beetles with respect to distance toward the lake (Y) over a four year period 116

2.11 Log-odds of leaf damage with respect to distance from the initial beetle release site (X) over a four year period 117

2.12 Log-odds of leaf damage with respect to distance toward the lake (Y) over a four year period 118

2.13 Loge mean leaf damage estimates with respect to distance from the initial beetle release site (X) over a four year period 119

2.14 Loge mean leaf damage estimates with respect to distance toward the lake (Y) over a four year period 120 Appendix Figures (continued)

Figure Page

2.15 Log-odds of meristem damage with respect to distance from the initial beetle release site (X) over a four year period 121

2.16 Log-odds of apical meristem damage with respect to distance toward the lake (Y) over a four year period 122

2.17 Log-odds of damage from other herbivores with respect to distance from the initial beetle release site (X) over a four year period 123

2.18 Log-odds of damage from other herbivores with respect to distance toward the lake (Y) over a four year period 124

2.19 Log-odds of purple loosestrife flowering success with respect to distance from the initial beetle release site (X) over a four year period 125

2.20 Log-odds of purple loosestrife flowering success with respect to distance toward the lake (Y) over a four year period 126

2.21 Loge of stem length with respect to distance from the initial beetle release site (X) over a four year period 127

2.22 Loge of stem length with respect to distance toward the lake (Y) over a four year period 128 Biological Control of Purple Loosestrife Lythrum salicaria By Two Chrysomelid Beetles Galerucella pusilla and G. calmariensis

Chapter 1

Introduction

In this thesis we evaluate protocols for monitoring classical weed biological

control programs and for testing the host specificity of weed biological control agents. To

do this we took advantage of a current biological control program in which two species of

chrysomelid beetles, Galerucella pusilla Duftschmid and G. calmariensis L, were introduced from Europe into North America for the control of purple loosestrife Lythrum salicaria L. By combining experiments and observational studies with current ecological theories of invasions and species interactions, we studied two important aspects of biological control programs, monitoring and host specificity testing.

First we construct, implement, and evaluate a general quantitative monitoring protocol for potential use in future biological control programs. Six important stages were identified in biological control programs and for each stage we chose a quantitative performance measure. Our six performance measures were: 1) biological control agent establishment, 2) the rate of increase of the agents, 3) the rate of spread of the agents, 4) the effect of the agents on individual target plants, 5) the effect of the agents on the population of the target plants, and 6) the indirect impact of the biological control agents on the local plant community. For each performance measure appropriate, quantitative measurements were defined and estimated. Finally, each measurement was compared to those found in the literature for other organisms and biological control systems. 2

Next, we examine the issue of host specificity in biological control programs. Host

specificity testing protocols typically measure whether a proposed biological control agent will complete its lifecycle on a non-target plant using organisms in confinement. Although a biological control agent may not be able to form self-sustaining populations on a non- target plant, it may still inflict significant damage on non-target plants located near populations of the target plant. Current host specificity testing does not evaluate this potential for non-target damage. Damage to the non-target plant will be dependent upon:

1) the overlap of the geographic ranges of the target and non-target plant, 2) the local distance between the target and non-target plant populations, and 3) the dispersal ability of the biological control agents. To quantify risk, we developed and implemented an experimental protocol for evaluating the non-target effects of the biological control agents on an economically important ornamental plant, crape myrtle Lagerstroemia indica L. 3

Chapter 2

Increase, Spread, and Ecological Effects of Two Beetle Species Galerucella calmariensis

and G. pusilla Introduced to Control Purple Loosestrife Lythrum salicaria in Oregon

Shon S. Schooler, Peter B. McEvoy, Eric M. Coombs, and Hans Luh 4

Abstract

We monitored development of biological control of purple loosestrife Lythrum salicaria over a six-year period at Morgan Lake in western Oregon. In 1992, two beetles,

Galerucella pusilla and G. calmariensis (Coleoptera: Chrysomelidae), were released to control the wetland weed at this test site. Our purpose was to estimate quantitative performance parameters that might be generally applied in monitoring biological weed control. Our six performance measures were: 1) biological control agent establishment, 2) the rate of increase of the agents, 3) the rate of spread of the agents, 4) the effect of the agents on individual target plants, 5) the effect of the agents on the population of the target plants, and 6) the indirect impact of the biological control agents on the local plant community.

The beetles established viable populations that increased during the study with an intrinsic rate of increase (r), based on the growth rate in damage, of 2.24/year. Within six years after introduction, the beetles spread to saturate the entire purple loosestrife habitat around the lake. The rate of spread, estimated by calculating a diffusion coefficient (D), was 57.5 m2/year. Adult beetles made seasonal, exploratory movements up to 30 m away from the host plant stand into surrounding crop fields, which suggests a disturbance-free buffer should be established in the habitat surrounding the loosestrife stand. Median percent leaf defoliation, monitored annually within a radius of 300 m from the initial point of beetle release, increased exponentially over time. By 1997, both flowering success and median stem density (stems per 0.125 m2 plot) of purple loosestrife declined to zero.

Mean above-ground biomass decreased to 8.4% of its former level. Abundance of native species increased by only 3% of former levels. Overall, G. pusilla and G. calmariensis 5 reduced the abundance of the target plant at our site. Our monitoring methods were effective at quantitatively measuring the establishment, increase, spread, and damage of the biological control agent, the subsequent decline of the target plant, and the impact on the local plant community. 6

Introduction

Classical biological weed control programs involve the deliberate introduction of a foreign herbivorous natural enemy to control a foreign weed species that has spread outside its native range. Biological control programs develop by stages, and it is useful to develop quantitative measures of performance for each stage to analyze progress step by step and thereby create a more reliable basis for intervention, comparison, interpolation, and extrapolation. Here we develop and apply performance measures based on field observation, over a six-year period, of two leaf beetles Galerucella pusilla Duftschmid and G. calmariensis L. (Coleoptera: Chrysomelidae) introduced to control purple loosestrife Lythrum salicaria L. (Lythraceae) at a local test site in western Oregon. We monitored the progress of biological weed control from colonization by the insects, to local interactions between plants and insects, to successional development leading to replacement of the weed by other plant species. We divided the ecological evaluation of the outcome of biological control into six stages: (1) establishment of the control organism, (2) increase of the control organism, (3) spread of the control organism, (4) negative effects of the control organism on individual target organisms, (5) negative effects of the control organism on target organism populations, and (6) positive, indirect effects of the control organism on the plant community.

Establishment

The first hurdle of a biological control program is the establishment of viable populations of the control organism in the new environment. The criteria for successful establishment are not well defined. One criterion currently used by E. Coombs in Oregon 7

is recovery of a univoltine insect three years in a row or a population size of> 1000

individuals during the second year after release. Worldwide, fewer than 65% of agents

released against exotic weeds successfully establish (Julien et al. 1984). The rate is higher

in Oregon, where there is an 84% establishment rate (Coombs et al. 1992). The rate of

establishment for biological control organisms generally increases with the number of

individuals released at a site (Beirne 1985). This is because establishment is from small

populations, and small populations suffer increased extinction risk due to a number of

demographic and genetic hazards (Lande 1988). Local extinction may be due to the lack

of genetic variation necessary for adaptation to new climatic conditions (Hopper and

Roush 1993). Another cause of local extinction may be an Allee effect, which occurs

when some benefit gained through social interaction is reduced or eliminated at low

densities (Allee 1939). Group benefits sacrificed at low densities may be related to finding

mates, exploiting a resource, or defense against predators (Begon et al. 1996; Memmott et

al. 1996; Grevstad 1998). Allee effects related to mate-finding can be mitigated by initially

introducing insects under cages to prevent dispersal and therefore maintain higher

population densities (Harley and Forno 1992). A high intrinsic rate of increase (r) helps

rescue small populations from extinction risk and increases the probability of successful

establishment (Crawley 1986). In a prior study on G. calmariensis and G. pusilla

Grevstad (1998) found that establishment was positively correlated with release size and that introductions consisting of over 500 beetles resulted in establishment rates of over

80%, however, the intrinsic rate of increase for these populations was highly variable. 8

Rate of Increase

The second stage of a biological control program is the increase of the control organism populations in the new environment. In an intensive review of weed biological control programs using insects, Crawley (1986) found that the probability of the reduction of target species to low levels is positively correlated with the intrinsic rate of increase (r) of the biological control agent. Increase in r of herbivorous insects leads to increase in host plant injury, which may, in turn, lead to reduction in host plant performance or yield

(Southwood and Norton 1972; Bardner and Fletcher 1974). The relationship between the density of an herbivorous insect and the injury it causes is generally linear, as in the case of the cereal leaf beetle Oulema melanopus L. (Coleoptera: Chrysomelidae) on oats, as well as in a suite of other plant/insect-herbivore interactions studied (Southwood and Norton

1972). Crawley (1986) and Lawton and Brown (1986) attempted to relate invasion success and success in biological control to intrinsic rates of increase. Crawley found a clear correlation while Lawton and Brown did not. However, as Williamson observed

(1996), in neither paper are there actual estimates of r and the conclusions come from measurements of components (Crawley 1986) or correlations (Lawton and Brown 1986) of r, which may explain differences in their conclusions. Goeden (1983) also used components of r, namely fecundity and generations per year, for finding promising biological control agents.

We estimated the intrinsic rate of increase for G. pusilla and G. calmariensis based on increasing damage and compared our estimate to that for beetles using data from an independent source. We calculated r at 1.11/year for G. calmariensis using estimates of fecundity (176), sex ratio (1:1), generation time (0.5 years), and survivorship from egg to 9 reproductive adult (0.02) (Grevstad 1996). Estimates of the intrinsic rate of increase for some successful biological control agents and invasive insect species vary between 1.6­

4.6/year (Table 2.1).

Rate of Spread

The third stage is the spread of the control organism throughout the area of weed infestation. The rate of spread of a biological control agent helps predict how readily the control organism will suppress local weed infestations and colonize new host patches

(Andow et al. 1990). Simple diffusion is the starting point for modeling the spread of invading organisms (Skellam 1951). Stratified diffusion combines two scales of movement, and simulations show this speeds the invasion rate relative to simple diffusion

(Hengeveld 1989; Shigesada and Kawasaki 1997). The standard measure of the rate of spread of an expanding population is the diffusion coefficient (D) and can be calculated from mark-recapture experiments or by direct estimates of area colonized (Hengeveld

1989; Shigesada and Kawasaki 1997; Kareiva 1983; Skellam 1951; Okubo 1980; Andow et al. 1990).

Dispersal has been previously studied for G. calmariensis. An experimental study of colonization rates indicates that G. calmariensis adults can detect host patches within

50 m and are capable of flying at least 850 m (Grevstad and Herzig 1997). From mark- recapture data reported by Grevstad and Herzig (1997), we calculated a diffusion coefficient of 0.455 m2/day using method on page 36 in Shigesada and Kawasaki (1997).

Prior estimates of diffusion coefficients for other insects exhibit large variation among insect species (Table 2.2). Table 2.1 Reported values of "r" for invasive species

Organism Details r (year) Reference

Galerucella pusilla and G. calmariensis introduced for the control of Lythrum 1.11 Grevstad 1996 (Coleoptera: Chrysomelidae) salicaria (purple loosestrife) in North America

Longitarsus jacobaeae (Waterhouse) one of three insects that successfully 2.15 McEvoy, unpublished data ragwort flea beetle controlled Senecio jacobea (tansy (Coleoptera: Chrysomelidae) ragwort) in North America

Oulema melanopus successfully invaded North America 1.6-1.9 Shigesada and Kawasaki 1997 cereal leaf beetle (Coleoptera: Chrysomelidae)

Cactoblastis cactorum successfully controlled Opuntia species 3.6-3.8 Caughley and Lawton 1981 (Lepidoptera: Pyralidae) (prickly-pear cactus) in Australia

Lymantria dispar (L.) successfully invaded North America 4.6 Liebhold et al. 1992 gypsy moth (Lepidoptera: Lymatriidae) Table 2.2 Summary of diffusion coefficients for insects. Secondary source indicates D was calculated from prior reference.

Insect species Notes D calculated Diffusion Original data source from: coefficient (D) (secondary source) Galerucella pusilla and Galerucella introduced for target centered 0.455 m2/day Grevstad and Herzig 1997 calmariensis (Coleoptera: biological control of mark-recapture Chrysomelidae) Lythrum salicaria Pieris rapae invasive pest species area of spread 2400-6400 Andow et al. 1990 (Lepidoptera: Pieridae) m2/year Small cabbage white butterfly Oulema melanopus invasive pest species area of spread 400 m2/year Andow et al. 1990 (Coleoptera: Chrysomelidae) Cereal leaf beetle Lymantria dispar invasive pest species area of spread 340 m2/year Liebhold et al. 1992 (Lepidoptera: Lymantriidae) (Shigesada and Kawasaki Gypsy moth 1997) Phyllotreta striolata agricultural pest mark-recapture 24.6-31.9 m2/day Kareiva 1982 (Coleoptera: Chrysomelidae) (Kareiva 1983)

Phyllotreta cruciferae agricultural pest mark-recapture 35.1-45.0 m2/day Kareiva 1981 (Coleoptera: Chrysomelidae) (Kareiva 1983) Trirhabda sericotrachyla mark-recapture 0.19-0.43 m2/day Karieva 1981 (Coleoptera; Chrysomelidae) (Kareiva 1983) Epilachna sparsa orientalis beneficial insect mark-recapture 0.71 m2/day Iwao and Machida 1963 (Coleoptera: Coccinellidae) (Kareiva 1983) Publica concava (Homoptera: native treehopper mark-recapture 0.23 m2/day McEvoy 1977 Membracidae) (Kareiva 1983) 12

In our examination of spread, we were also interested in identifying spatial effects.

If biological control agents spread in a homogeneous fashion they will provide no spatial or temporal refuge for the target organism population, as opposed to spread of the agents in a heterogeneous fashion, where refugia will exist for the target organisms. In the former case, we would expect a greater impact and a greater reduction in abundance of the target organism. Spatial autocorrelation analysis is a technique currently used to determine and classify spatial effects (Legendre and Fortin 1989). We expected a spatial effect to be evident at the initial stages of the program, as the beetles colonized new host plants. As the beetles saturated the environment, we expected the spatial effect to disappear, assuming homogeneous colonization. If the spatial effect did not disappear, or reappeared later in the program, we could then conclude that the beetles were patchily distributed and may not be affecting all target plants equally.

Impact on individual target plants

The fourth stage is reached when control organisms attack the target plant and reduce individual plant performance. Estimates of target plant damage help link cause (the biological control agent) and effect (the decrease of plant abundance) in the biological control process. This is particularly important due to the difficulty in maintaining separate control and impact plots over many years in a monitoring study (Smith and DeBach 1942).

Herbivory by G. calmariensis has been found to cause significant reductions in individual plant height, leaf, shoot, root, and total plant biomass with increasing attack levels in experimental field tests (Blossey and Schat 1997). Here we provide evidence of the impact of the beetles on the target weed. 13

Impact on target plant population

The fifth stage is apparent when the control organism decreases target organism

distribution and abundance. It is one thing to show that insects decrease individual plant

performance, it is quite another matter to show that insects decrease plant population size

(Crawley 1983; McClay 1995). The degree of suppression in the average abundance of

the target organism is one measure of the success of a biological control program (Smith

and DeBach 1942; McClay 1995). It is irnportant to measure plant reduction at the

population level, expressed in plant abundance per unit area, such as mean or median stem

length, plant density, and biomass per sample plot (McClay 1995). A commonly used

estimate of the level of depression of plant populations by insect herbivores is "q", defined

as the ratio of the average abundance of the plant in the presence of the herbivore (V*) divided by the abundance of the plant in the absence of the herbivore (K):

q = V*/K

(Caughley and Lawton 1981; Beddington et al. 1978). It was predicted prior to releasing the control organisms that biological control would decrease purple loosestrife to

10% of its original level over 90% of its current range (Malecki et al. 1993) and that local effects will be visible within five years (Blossey 1995b). This leads to an expectation of q

0.10 within approximately five years. Estimates of q for highly successful biological control programs are generally below 0.01 (Table 2.3).

Impact on the local plant community

The sixth and final stage is apparent when the target organism sets in motion a successional process that leads to replacement of the target organism by more desirable Table 2.3 Reported values of "q" for biological control programs

Plant species Biological control agent(s) q Reference Opuntia inermis and 0. stricta Cactoblastis cactorum 0.002 Caughley and Lawton 1981 prickly-pear cactus (Lepidoptera: Pyralidae)

Hypericum perforatum Chrysolina quadrigemina <0.01 Huffaker and Kennett 1959 St. John's wort (and C. hyperici) (Coleoptera: Chrysomelidae)

Senecio jacobea L. Tyria jacobaeae <0.01 (local) McEvoy et al. 1991 tansy ragwort (Lepidoptera: Arctiidae) Longitarsus jacobaeae 0.07 (regional) (Coleoptera: Chrysomelidae) Hylemya senecilla (Diptera: Anthomiidae) Carduus nutans L. Rhinocyllus conicus Frolich 0.05 Kok and Surles 1975 musk thistle (Coleoptera: Curculionidae)

Chondrilla juncea L. Cystiphora schmidti Rubsaamen 0.36 Cullen 1978 rush skeletonweed (Diptera: Chironomidae)

Eichhornia crassipes (Martinus) Neochetina eichhorniae Warner 0.27 Cofrancesco et al. 1985 Solms-Laubach (Coleoptera: Curculionidae) water hyacinth 15 vegetation. Since the ultimate goal of the weed biological control program is to create or restore a desirable vegetation (or no vegetation as in some aquatic systems), some examination of the local plant community, preferably before and after the reduction of the target plant, is necessary in order to determine success (Huffaker and Kennett 1959;

McEvoy et al. 1991). In a rangeland, success may be measured as an increase in forage species whereas in a natural setting the return of native species may be the appropriate measurement (Malecki and Blossey 1994; Coombs et al. 1996; Gruber and Whytemare

1997).

Questions

We framed a set of six questions, each based on one stage of a biological control program.

1. How quickly do viable populations of the control organism become established?

2. How quickly do insect populations increase following establishment?

3. How quickly do insect populations spread from the point of release?

4. How much damage do the insects inflict on individual plants?

5. What level of suppression do the insects impose on target weed populations?

6. What changes occur in the plant community following suppression of the target weed?

Study System

The Study Site

Fieldwork was conducted at Morgan Lake in Western Oregon (Polk Co.- Lat. 45°,

Long. 121°). Morgan Lake is part of Baskett Slough National Wildlife Refuge which is 16

one of a series of refuges created to protect the flyway of the dusky Canada goose

(Branta canadensis occidentalis), an endangered subspecies of the Canada goose.

Morgan Lake was created by an earthfill dam in 1962 (Appendix 2.1). In the

autumn of each year, usually late November, the lake is partially drained in order to create

wetlands downstream for the overwintering goose population. Reed-canary grass

Phalaris arundinaceae L. and purple loosestrife dominate the strip of riparian vegetation

encircling the lake. The width of the riparian zone varies from 4 m to 130 m and averaged

29.5 m across stations spaced 100 m apart that extended around the edge of the entire

lake. Using a computer program (PV-Wave, version 6.2 Visual Numeric, Inc.) and a map

created from aerial photographs, we calculated the area of this zone to be 4100 m2 and

believe that this accurately represents the total potential habitat of purple loosestrife

around Morgan Lake. Inside this is a thin strip of creeping spike rush Eleocharis palustris

(L.) R. & S. which is bounded by open water. Outside of the riparian zone on the

terrestrial side is a narrow transitional zone to open fields. Cultivated fields surround the

lake on all sides with the exception of thin corridors of vegetation accompanying several

streams that feed the lake and one that flows out from below the dam.

The Target Plant Purple loosestrife is a tall 2 m), iteroparous, perennial wetland plant native to

Europe. It probably arrived on the East coast of the United States before 1830 in ballast

deposited by trading ships from Northern Europe (Thompson 1991). It has since spread

across the country, aided recently by road construction and irrigation channels, as well as

through the planting of seeds sold in wildflower mixes (Wilcox 1989). A mature plant can

produce as many as 2.5 million seeds annually (Malecki et al. 1993), which are dispersed 17 by water or in mud adhering to animals. Its mean rate of spread since 1940 has been

estimated at 645 km2 per year (Thompson 1991). Purple loosestrife is an invasive species that quickly displaces native wetland vegetation in riparian areas, often forming dense monospecific stands that degrade habitat quality for waterfowl and other wetland animal

species (Thompson et al. 1987; Balogh and Bookhout 1989). In the U.S. the estimated cost of infestation, in terms of wildlife and agriculture, is $45 million per year (Thompson et al. 1987). However, the negative effects of purple loosestrife on native species have not been adequately studied (Anderson 1995).

Three biological control agents: two leaf beetles (Galerucella pusilla and G. calmariensis) and a root weevil (Hylobius transversovittatus Goeze), were approved for release in June of 1992 (Malecki et al. 1993). Two other agents have since been approved, the seed weevils Nanophyes marmoratus Goeze and Nanophyes brevis

Boheman. N. marmoratus was released in 1994. N. brevis has not yet been released in the United States due to a nematode infection in all the European populations examined

(Rees et al. 1996).

The Biological Control Agents

Although the beetles are two separate species, they are similar in lifecycle and feeding preferences. In Europe, G. pusilla and G. calmariensis have similar life histories, ecological niches, geographic distributions, and the two species exist together even in populations of purple loosestrife with less than 10 plants (Blossey 1995a). Although the growth rate of G. pusilla is higher than that of G. calmariensis, both species appear to respond similarly to variation in environmental conditions, which leads to positive 18

correlation in growth rates for the two species (Grevstad 1998). In host specificity tests,

they were only able to complete their life cycle on purple loosestrife (Blossey et al. 1994).

Adult beetles emerge from their diapause sites in the fields surrounding the purple

loosestrife stand in late March to mid-April. They require greater than 15.5 hours of

daylight to break reproductive diapause (Grevstad 1998). They then migrate to the stand

to feed on the meristematic tissues of young loosestrife plants. Mating begins soon after the beetles emerge and females begin oviposition after about one week. Oviposition peaks in mid-May and tapers off around mid-July. Adult beetles typically live for 40-60 days

(Ragsdale 1996). Egg masses are oviposited on stems or on the upper surfaces of leaves.

Eggs are spherical, beige in color, and usually have a bit of frass deposited at the apex.

They are laid in masses of 1 to 15 eggs. Larvae emerge in approximately 10 days. The first instar larvae then move to the apical meristem where they burrow into the shoot tip and feed. Once the tip is consumed they move down the plant feeding on the undersurface of the leaves. After three instars, the larvae search for adequate pupation sites. If the plant is in standing water, the larvae burrow into the aerenchyma tissue (tissue produced on the lower stem of the plant to aid in respiration). If the plant is on dry soil, no aerenchyma tissue is produced and the larvae pupate in the plant's root mass or in the surrounding soil or litter.

The first teneral adults emerge from the pupation site approximately 5 weeks after the beginning of oviposition. They then either leave hosts and migrate to the nearby fields

(possibly entering a state of aseasonal quiescence) or remain on hosts to feed, mate, and oviposit, beginning a second generation. Adults do not normally mate until seven days after eclosion (Grevstad 1998). It is not known what triggers migration. Those beetles 19 that remain in the loosestrife stand produce a second generation and die in late August.

The second generation pupates in late August and teneral adults emerge in mid-

September. In late September, the second-generation adults migrate into the fields surrounding the riparian zone and enter a diapause from which, if left undisturbed, they emerge the following spring. Flooding is frequent during the winter months in the

Willamette Valley where our research site was conducted. When submerged, the beetles crawl out of the water and onto plant stems. They then seek protected sites lodged in the stem axils or within the hollow broken stems of grasses. A total of 96 individuals were found on 21 January 1997 in a single broken stem of reed-canary grass Phalaris arundinaceae L. (Shon Schooler, personal observation).

Two generations have been recorded for both species in Oregon while only one is found in the shorter growing seasons of the Midwest and Eastern regions of the United

States. Four generations per year have been recorded in Northern Italy (Batra et al.

1986). Adult beetles of the two species can only be easily distinguished after the teneral stage (which lasts approximately one week after emergence). G. pusilla is smaller and has solid golden-brown elytra while G. calmariensis is slightly larger and orange-brown with two dark stripes that run dorsal-laterally down the beetles' elytra.

Methods

In 1992, G. pusilla and G. calmariensis were released at our study site, Morgan

Lake (Baskett Slough National Wildlife Refuge, Dallas, Oregon) in the Willamette Valley of western Oregon. The two beetles, G. pusilla (800 individuals) and G. calmariensis

(250 individuals), were introduced into a 3.3 x 3.3 m enclosure at the middle of the south 20

shore of the lake on July 24 of 1992 (Appendix 2.1). The beetles that were released at

Morgan Lake originated from populations in Germany. On 14 May 1993, after one

complete beetle generation, the cage was lifted and the beetles were allowed to disperse.

Some beetles escaped from the cage and were found outside the cage in late 1992 and

early 1993.

We evaluated biological control in this system using intensive and extensive

sampling methods.

Intensive Sampling

Intensive sampling began in 1994 (two years after the release of the biological

agents, G. calmariensis and G. pusilla) and continued through 1997. We sampled at

monthly intervals from April through August (except no samples were taken April and

August 1995 and April 1997). We sampled from the target plant's habitat, identified as a

zone around the lake that was dominated by L. salicaria, P. arundinaceae, and E. palustris. A cultivated grass seed field on one side and open water on the other

(Appendix 2.1) bordered this zone. In 1995 seedlings were found outside of the terrestrial

boundary of this zone, however, they quickly disappeared, probably due to a combination

of desiccation, interspecific competition, and beetle defoliation. A transect was

established along the terrestrial edge of this zone parallel to the shore (hereafter referred

to as the X axis) and stakes were placed at varying distances from the initial beetle release

site (Figure 2.1a). In order to circumscribe the expanding beetle population, we increased

the area of the sample as the area of the beetle population increased over time by

increasing X-axis length and the intervals between sampling points (Table 2.4). We took 21

(a)

Beetle release site

Field

(b)

315m / Open water Beetle release site

Field

(c)

Field Beetle release site

Field

Figure 2.1 Design for: (a) intensive, (b) biomass, and (c) extensive sampling. Thick black line indicates transect and short perpendicular lines indicate length of transect during the sample period in figures a and b. In figure c open circles indicate sampling stations at 100 m intervals around the lake. 22

Table 2.4 Summary of sampling dates and distances Method Date Total transect Distance from Sample interval length (m) beetle release (m) (m) Intensive 26-Apr-94 60 30 5 17-May-94 60 30 5 21-Jun-94 60 30 5 25-Jul-94 60 30 5 19-Aug-94 400 200 10 up to 200, then 50 thereafter 16-May-95 300 150 10 16-Jun-95 200 100 10 19-Jul-95 500 250 10 25-Apr-96 200 100 10 28-May-96 200 100 10 20-Jun-96 200 100 10 11-Jul-96 200 100 10 23-Aug-96 600 300 20 02-Jun-97 200 100 10 24-Jun-97 600 300 20 18-Jul-97 600 300 20 21-Aug-97 600 300 20 Plant community 2-6 Sept 1996 630 315 30 3-6 Sept 1997 630 315 30 Extensive May 1992-1997 3400 (full lake) 1700 100 23

two 0.125 m2 plots (0.25 x 0.5 m) along the Y-axis perpendicular to the X-axis for each

sampling point along the X-axis. On the first sample date (April 26, 1994) these plots were located at regular intervals (4 m and 10 m) along the Y-axis toward the lake. On all

subsequent dates the width of the zone was measured and the plots were randomly chosen using a random numbers table. The random numbers chosen were integers and were measured in meters toward the lake from X-axis at the terrestrial edge of the habitat. The zone varied in width between 4 m and 25 m. Because our repeated visits caused small trails to develop, we placed the plots directly off to one side of the path. When sampling to the southeast of the beetle release point, the plot was located to the east of the path and when sampling to the northwest the frame was placed to the west of the path.

At each plot, five indicator and eleven response variables were recorded (Table

2.5). The indicator variables included were: year (1994-1997), month (April-August, with the omission of April and August 1995 and April 1997), direction (NW and SE), distance in X (distance from point of beetle release), and distance in Y (perpendicular to X, the gradient in conditions ranged from high and dry with small values of Y to low and wet with large values in Y). There were eleven response variables: three associated with the target plant, seven associated with the biological control organisms, and one associated with other herbivores. The three variables associated with the target plant were stem density, stem length, and number of stems in the flowering stage (phenology). Of the seven variables chosen to represent the biological control organisms two variables were associated with adults: adult density and number of stems with adults. Two variables were associated with juvenile insects: number of stems with eggs and number of stems with larvae. Stems with eggs were counted only if the eggs were recently oviposited, as 24

Table 2.5 Summary of insect, plant, and other variables

Method Variable Measurement Intensive sample method indicator variables year 1994, 1995, 1996, 1997 month Apr, May, June, July, Aug direction NW or SE distance from biocontrol agent release site (X) (0 - 300 m) distance from transect toward lake (Y) (0 - 25 m) response variables target plant stem density number of stems (0.125m2) stem length ground level to apex (cm) stem phenological stage vegetative or flowering biological control organisms presence of adult beetles on stem yes or no number of adult beetles number of adults on stem presence of beetle eggs on stem yes or no presence of larvae on stem yes or no presence of leaf damage on stem yes or no estimate of leaf damage 6 damage classes presence of apical meristem damage on stem yes or no other presence of damage by other herbivores yes or no

Plant community sample indicator variables distance from biocontrol agent release site (X) (0 - 315 m) distance from transect toward lake (Y) (0 - 25 m) response variables above ground biomass dry mass (g/0.25m2)

Extensive sample indicator variables distance from biocontrol agent release site (X) (0 - 1800 m) area covered (15 - 4100 m2) response variables signs of biological control agents yes or no (eggs, larvae, adults, or damage) 25 indicated by a firm appearance and orange or cream color (as opposed to the papery appearance and silver color of previously hatched eggs). The other three beetle variables were estimates of purple loosestrife plant damage caused by the beetles. They were: a visual estimation of the area of leaf damage (skeletonization of the upper surface of the leaf), the number of stems with leaf damage, and the number of stems with damage to the apical meristem. From field and laboratory observations we attribute most leaf damage to adult beetles and most apical meristem damage to larval beetles. We estimated leaf damage using pre-calculated photographic standards (Appendix 2.2). For each stem we recorded an average leaf damage level based on six classes (0 = 0%, 1 = >0-5%, 2 = >5­

25%, 3 = >25-50%, 4 = >50-75%, 5 = >75-95%, and 6 = >95-100%). Intervals of unequal widths were used to maximize sensitivity while avoiding difficulty in classifying very low and very high levels of damage. Using meristem damage as an indication of the presence of first and second instar larvae meant meristem (larval) damage and larval presence were necessarily correlated. The final variable, damage by other herbivores, was included to assess the possible confounding of control organism effects with those due to other herbivores. We were able to identify the damage from four vertebrate herbivores: black-tailed deer (Odocoileus hemionus columbianus), nutria (Myocaster coypu), beaver

(Castor canadensis), and Canada geese (Branta canadensis). These were easily separated from damage caused by biological control agents because the foliage and usually the plant stems were clipped while the beetles skeletonized the leaves and apical meristem, leaving the veination intact. Two other invertebrate herbivores were found at our site: the larvae of the sawfly (Ametastegia glaberata: Hymenoptera (inferred from Gary Piper, personal communication 1997)) and an aphid (Myzus lythri: Homoptera) (identified by Andrew 26

Jensen, personal communication 1995). Sawfly damage was easily distinguished from

beetle feeding because the sawfly tended to chew cleanly through the edges of the leaves

as opposed to chafing the leaf surface and leaving the underlying structure. Damage by

the sawfly was infrequent at our site. Aphid density was high only during July and August

of 1995 and damage by this phloem feeder bears no resemblance to damage caused by the

biological control agents.

As a part of this monitoring study, we estimated above-ground biomass of purple

loosestrife and other plant species in the local plant community in early September of 1996

and 1997 (2-6 Sept. 1996, 3-6 Sept. 1997). We used the same transect as used in the

observational studies, however, as not to interfere with concurrent monitoring efforts we began sampling 15 m to each side of the initial beetle release site and at distance of every

30 m thereafter up to 315 m (Table 2.4 and Figure 2.1b). At each location, we measured the width of the zone and randomly chose two points within that distance. Therefore, we had a total of 44 plots over a 630 m transect. At each point, we sampled a 0.25 m2 plot by

clipping the vegetation at ground level: the area of the sample unit remained the same between years but the shape of the plot varied. In 1996 we used a circular metal hoop as a

sample frame. However, because of difficulty in placing the hoop over grass stems, we used a square PVC frame (0.5 x 0.5 m) with one side open in 1997. Individual plants were sorted to species. In August of 1996 and May of 1997 we created a voucher

collection for the wetland plant species found around Morgan Lake. We sorted our

samples to species in the field using this voucher collection and three plant identification manuals (Hitchcock and Cronquist 1973; Pojar and MacKinnon 1994; Guard 1995). 27

Samples were dried to a constant weight at 60°C using the following procedure.

Three of the heaviest bags of plant material were chosen and weighed every two days.

After six days the weight no longer declined. The samples were left in the oven for an additional two days. After eight days, the bags were removed and each sample was weighed on an electrobalance (Mettler PN1210) to the nearest tenth of a gram. Trace samples were defined as those samples weighing less than 0.1 grams and were recorded as weighing 0.1 grams.

We estimated the biomass of L. salicaria prior to 1996 by extrapolation from measurements of linear dimensions. With data previously collected by others (Bernd

Blossey, personal communication 1997), we used linear regression to estimate a slope and intercept parameter for converting loge stem length to loge dry biomass (n 50, r2= 0.9918, y = 0.2336x + 0.4413, p < 0.001). To best match the sampling periods of our biomass data, one sample date was selected late in the growing season for each year (August 8,

1994 and July 19, 1995). We used this relationship to convert stem length measurements collected in 1994 and 1995 into biomass estimates for L. salicaria for those years.

Direct (1996, 1997) and indirect (1994, 1995) estimates of biomass were made using slightly different sampling procedures. In both cases a transect was centered on the site of beetle release, which extended around the terrestrial edge of the lake margin (X), and two plots were randomly sampled off of that transect toward the lake (Y). However, transect length, interval distance along X, and plot size differed among years (Table 2.4 and Figure 2.1).

The loge biomass of each stem was calculated using the slope of the regression line and the y-intercept. This was converted back to actual biomass using the antilog function 28 and summed for each plot. Plots for the estimated biomass from stem length were 0.125 m2 and those for directly estimated samples were 0.25 m2, so the sum of the estimated biomass from stem length for each plot was doubled. To make frequency distributions normal before averaging within sample dates, the total biomass of each plot was transformed back to its natural log. The mean of the natural log was then calculated for each sample date.

During our investigation, we also noted a pattern of adult beetle migration away from the L. salicaria stand and into the adjacent fields. In 1996, we quantified this migration by establishing ten transects parallel to the edge of the lake with three meters between each transect for a total of 30 m sampled in Y. We then conducted three sweep samples at each distance with 25 sweeps per sample. Each sample covered approximately twenty meters, for a total of 60 contiguous Meters, with the total distance centered on the initial beetle release point (30 m to each side). We sampled approximately biweekly over the course of the growing season (April through October).

Extensive Sampling Method

This sampling method was initiated in 1992, shortly after the initial release of the beetles. A transect was established around the entire lake and stakes were placed at 100 m intervals (Table 2.4 and Figure 2.1c). In late May of each year, we circumnavigated the lake and mapped all occurrences of beetle signs and loosestrife symptoms around each stake. These included: presence of eggs, larvae, adults, and beetle damage. If signs of beetles were found, we searched to identifj where the colonized area ended. These maps were then scanned into a computer and the area of infestation was estimated by calculating 29 the area represented by one pixel from a known area on the map and multiplying this by the total number of pixels of the infested area. From this analysis, distance and area of beetle spread were calculated for each year at Morgan Lake.

Data Analysis

We analyzed the data from the intensive sampling method using a combination of log-odds proportions, odds ratios, and spatial autocorrelation analysis. Log-odds were used to compute odds ratios that allowed comparison between variables. Spatial autocorrelation analysis was used to determine when spatial patterns existed for specific variables.

Log-odds proportions create a ratio (a/b) of those "which are" (a) over those

"which are not" (b) where the number of observations (stems) within a quadrat = a + b

(Hosmer and Lemeshow 1989; Sokal and Rohlf 1995). Note that this is mathematically equivalent to the more traditional procedure for calculating log-odds, which is the ratio of the corresponding proportions p/q: where p = a/(a+b) and q = b/(a+b). A transformation is needed for data expressed as proportions p and q because these proportions are hemmed in from both sides. The same is true for data expressed as the number of stems per quadrat, a and b, where the data are bounded by 0 and K, the latter value being the carrying capacity of the plot. The logit transformation, which is ln(a/b), was used to obtain a quantity that is approximately normally distributed, from which means were calculated. We used empirical logits where 0.5 is added to the numerator and denominator of the ratio because the logit cannot be calculated when a or b = 0 (Sokal and

Rohlf 1995). The log-odds proportion ln((a+0.5)/(b+0.5)) was plotted as a function of 30

distance from the point of beetle release. Initially, direction (NW and SE) was taken into

account but samples were combined to increase sample sizes for spatial autocorrelation

analysis when it was visually determined that there was no directional bias.

A ratio of odds is a useful way to compare the differences in outcomes. The

difference between the logits (ln(ai/bi) ln(a2/b2)) equals the logarithm of the odds ratio

(1n((ai/bi)/(a2/b2)) (Sokal and Rohlf 1995). The odds ratio o can be obtained from the log-odds ratio by the back-transformation el""adds(Sokal and Rohlf 1995). The odds ratio is a natural scale that is easy to understand, e.g. the odds of finding damage to stems was

20 times greater than the odds of finding the beetles that caused the damage. It is traditional to compare odds for the same variable, e.g. the odds of survival of organisms subjected to different treatments (Sokal and Rohlf 1995). Here we use the ratio in a non­ traditional way, to compare the odds that stems (the study subjects) exhibit different signs or symptoms, e.g. the odds of leaf damage are compared to the odds of finding adult beetles on a leaf:

To diagnose spatial effects we used a spatial autocorrelation procedure. Analysis of spatial autocorrelation is of particular use in ecology because it is specifically designed to examine the spatial heterogeneity that is often of interest but is assumed not to exist among sample observations in Many parametric statistical methods such as analysis of variance (ANOVA) and regression. A variable is said to be autocorrelated (or regionalized) "when it is possible to predict the values of this variable at some points of space [or time], from the known values at other sampling points, whose spatial [or temporal] positions are also known." (Legendre and Fortin 1989). 31

We analyzed the data by separating each sample into separate X (distance along transect from beetle release site) and Y (distance from transect toward lake along topographic/moisture gradient) components. Then the total distances for each sample (for both X and Y) were divided into five equal parts. Within each distance, a measure of spatial autocorrelation was calculated using Moran's I spatial autocorrelation procedure

(Cliff and Ord 1981). For each variable within each date, we constructed a correlogram, which is a graph of the autocorrelation value plotted against distance. A test of significance is associated with each autocorrelation coefficient. The null hypothesis of this test is that the coefficient is not significantly different from zero. We choose the standard a = 0.05 to test significance. For each variable within each date at each distance interval, we calculated a p-value (Legendre and Fortin 1989). We performed the analysis with two questions in mind. First, we wanted to identify which variables expressed spatial effects.

Second, we were interested when these effects occurred for the variables.

Log-odds of leaf damage reflected only the proportion of stems damaged within the plot sampled. To examine the intensity of damage, we calculated mean damage level per plot. We could not simply average the categorical data (0 = 0%, 1 = >0-5%, 2 = >5­

25%, 3 = >25-50%, 4= >50-75%, 5 = >75-95%, and 6 = >95-100%) because our damage classes were not of equal widths. Instead we converted the classes (0-6) back to percent damage by calculating the midpoint of each damage class (0, 2.5, 15, 37.5, 62.5, 85, and

97.5% leaf area removed) and assigning that percent damage to the respective stem. The mean of each plot was then calculated. A problem with this procedure was that the damage classes increased in width near 50% defoliation and taking the midpoint of each class tends to pull the estimate toward that central point. 32

We calculated the intrinsic rate of increase for G. calmariensis from data in prior literature (Grevstad 1996). Using a fecundity estimate of 175 eggs/female, survivorship from egg to adult of 0.02, a sex ratio of 1:1, a generation time of 0.5/year, and the following formula from Crawley (1986) we calculated the intrinsic rate of population increase:

r = (1n 110)/T where Ro is the number of reproductive females produced per female (number of eggs per female x survivorship x sex ratio, expressed as proportion female) and T is the generation time. We estimated the intrinsic rate of increase for G. calmariensis and G. pusilla damage from defoliation measurements because, as the reaction of beetles to the disturbance necessary during sampling was to fall from the plants, we could not get accurate direct estimates of beetle density. For each year, the median percent defoliation was calculated across all plant stems. We then fit an exponential function to the data. The exponent of this equation was our estimate of the intrinsic rate of increase for the beetles:

Yt = Yoe`` where yt is median percent defoliation at time t, yo is initial median percent defoliation, r is the increase in percent leaf defoliation, and t is time. Damage was recorded as a percentage and is therefore bounded and cannot grow in a true exponential fashion. Over a longer time series we would expect a logistic model to fit the data better. However, for our data the exponential model was the simplest descriptor of the initial, maximum rate of increase for leaf defoliation.

We used the data from the extensive sampling method to calculate both the area the beetles colonized and their distance of spread per year. Maps for each year (from 33

1992 through 1997) were created in Photoshop (version 3.0 Adobe Systems, Inc.) using the maps of the field observations. Then area colonized and distance traveled were calculated and graphed.

Beetle spread was measured as: 1) a distance around and across the lake 2) the radial spread as an increase in area (assuming concentric growth), and 3) a diffusion coefficient based on the total area of beetle colonization as a function of time. Distance traveled per year was determined by identifying the furthest distance traveled by the beetles for each year, based on the initial release point. This was done for both direct distance, across the lake, and maximum distance, around the edge of the lake. Linear regression was used to calculate distance traveled per year, based on direct distance measurements. The rate of radial spread was calculated, using linear regression, by estimating the slope of the square root of the area colonized (A) from 1992 to 1997, and dividing it by ic. We used estimates of area colonized (A) to calculate the diffusion coefficient (D) for the beetles at our site. First, the mean-squared radius of distribution

(r2) was calculated using the equation: (r2) = (A/ ,r) (Andow et al. 1993). Then the slope of the line (m) for the mean-squared radius of distribution (r2) by time (t) was estimated using linear regression. The slope of this line (m) divided by four equals the diffusion coefficient (D): D = m/4 (Shigesada and Kawasaki 1997).

We were able to estimate the diffusion coefficient for G. calmariensis from a previous mark-recapture study done by Grevstad and Herzig (1997). We used the sum of the distances at which the recaptured beetles were released (1841 m), the total number of beetles recaptured (582), the time since release (7 days), and the following formula from

Shigesada and Kawasaki (1997) to calculate the diffusion coefficient per day. 34

D = (sum of the distances/total number of recaptured organisms)2 it (time from release)

This study was a target-centered mark-recapture experiment, which means that the beetles were released at varying distances from the purple loosestrife stand and the number reaching the stand was recorded.

To facilitate comparison between our estimate of D from the above experiment with those of prior investigators, we converted the prior estimate of the diffusion coefficient taken on a daily time scale to a yearly estimate using the procedure of

Shigesada and Kawasaki (1997). We multiplied the life expectancy for adults (40-60 days) (Ragsdale 1996) with the number of generations per year (2) and by the daily diffusion coefficient (0.455 m2/day) (Grevstad and Herzig 1997). We reversed this procedure for our own yearly estimate of D from this study, also to facilitate comparison.

To examine the relationship between stem length and flowering probability before

(1994 and 1995) and after (1996 and 1997) control of purple loosestrife, we calculated the proportion of stems flowering within 19 stem length classes covering 10 cm each. We used linear regression to test the hypothesis that flowering probability was positively correlated with stem length. To determine whether reduction in purple loosestrife biomass was correlated with distance from the initial beetle release site we calculated the correlation coefficient for the direct biomass estimates for both years.

All data analysis using regression was carried out in Excel (Excel 97, Microsoft

Corp.). Autocorrelation analysis was performed using a spatial analysis program written by Hans Luh in PV-Wave (version 6.2 Visual Numeric, Inc.). 35

Results

Establishment and rate of beetle increase

Median percent leaf damage initially increased at an exponential rate with an estimated intrinsic rate of increase of r = 2.24/year (Figure 2.2). Damage is bounded by 0 and 100%, so it can not increase exponentially forever, however, the intrinsic rate of increase for damage approximates the maximum rate achieved shortly after colonization, provided there are not Allee effects. We assumed a linear relationship between leaf damage and herbivore density and used the rate of increase of leaf damage as the estimate of the intrinsic rate of increase for the beetle population at our site.

Rate of beetle spread

We expected insect density, odds of observing beetles, beetle damage, and odds of observing beetle damage to be maximum at the point of release and to decrease with increasing distance from that point. As the insects spread and saturated the environment, we predicted that the above variables would vary independent of the distance from the point of release. If we found continuing spatial effects, we could then conclude that the beetles were not affecting their resource, purple loosestrife, in a homogeneous fashion and causes for potential plant refuges could be determined.

The predicted spatial effect was confirmed by our observations for some of the insect variables. Leaf damage, attributed mostly to adult beetles, showed a significant pattern across all sample dates in 1994 and 1995, where log-odds of damage were negatively correlated with distance from the initial beetle release site (Figure 2.3). In 1996 36

100

90

80 Yt = 0.0i 602.2439t

70 r2 = 0.978

60

50

40

30

20

10

1 2 3 4 Year

Figure 2.2 Median percent beetle defoliation (y) increased exponentially over time (x). 26 April 1994 17 May 1994 21 Jun 1994 25 July 1994 19 August 1991

0.015 3 p 0.012 p <0.001 p 0.001 0.012 3 2

1 i

0

. -2 : 2 . : J 26 10 10 20 w 10 20 10 100 9.0 100

16 May 1995 16 Jul. 1995 19 -1131y 1995 0.0.0113 p 0.002 p <0.001 :

0

-1 -2 : :

20 4 go N 100 120 lo 20 0 92 00 100 ICC 150 293 no 25 *pi 199$ 20 Juno 1996 11 July1996 7 23 Pupal 1994 3 3 2 0 . 0 . f . . : 2 2 : . . . : 1.

1 0 1.

-2 -3 J. J. 20 0 60 60 100 n 60 N IN 20 110 60 NO 92 40 011 60 I00 10 IN IN 203 no soo

2 Juno 1967 - 24 Juno 1997 11 !say 1997 21 Au3us1 1997

2 . f

-1 -1. a. -2.

-3 4

20 0 0 b IN so In MO 200 no 300 so 100 iso 320 no 100 0 N 100 150 292 250 300

distance from initial beetle release site (m)

Figure 2.3 Log-odds of leaf damage as a function of distance from the initial beetle release site (X) over a four year period. 38 and 1997, despite increased transect length, this trend disappeared as the proportion of stems with leaf damage within the area sampled became relatively homogeneous.

Probability of meristem damage, attributed mostly to larval beetles, as well as presence of beetle eggs exhibited a similar significant trend on three of the sample dates in 1994 and

1995 (Appendix 2.13 and 2.3). Probability of beetle larvae and adult density showed the same trend on two sample dates in 1994 and 1995 (Appendix 2.5 and 2.9). The only insect variable to exhibit no spatial trend with respect to distance from the initial release site was the probability of finding adult beetles, which was low across all sample dates. A summary of significant spatial effects found among the insect, damage, and plant variables is given in Table 2.6a and b.

We also examined spatial effects with respect to the topographic/moisture gradient at the site (reflected by Y). We expected that overwintering beetles emerging in the spring would first colonize plants on land and then colonize plants in water after a delay. Thus, adult beetles and indications of adult beetle presence (eggs, larvae, and beetle damage) would be greater at the terrestrial edge of the stand (low values of Y) early in the season and become more homogeneously distributed as time progressed. Log-odds of leaf damage decreased with increasing distance in Y as expected in June of 1994 and May of

1995. However, presence of beetle eggs, beetle larvae, and meristem damage, showed the opposite pattern (Appendix 2.4, 2.6, 2.14). Beetle eggs particularly showed the opposite trend on five sample dates over the course of the study.

We calculated odds ratios for selected pairs of variables to determine which beetle variables were the most likely indicators of beetle presence (Table 2.7). Cumulative variables, such as presence of beetle damage and eggs, were found to be more indicative Table 2.6a Summary of spatial pattern in insect, damage, and plant variables

a. Frequency of responses varying with distance along lake margin from point of biological control agent release Variable Number of sample dates Comments a pattern was found (of 17) Insect Variables Adult Log odds of adult beetles 0 (0%) Loge mean density of adult beetles 2 (12%) density of adults declined with distance from release site on 2 of 3 dates in 1995 Juvenile Log odds of beetle eggs 3 (18%) probability of stems with eggs decreased with distance from release site on 3 of 8 dates in 1994 and 1995 Log odds of beetle larvae 2 (12%) probability of stems with larvae decreased with distance from release site on 2 of 8 dates in 1994 and 1995 Damage Variables Log odds of adult damage 8 (47%) probability of stems with adult damage decreased with distance from release site on all dates sampled in 1994 and 1995 Log odds of larval damage 5 (29%) probability of stems with meristem damage decreased with distance from release site on 4 of 13 dates in 1994-1996, but, in June of 1997 probability of stems with meristem damage increased with distance Log odds of damage by other herbivores 0 (0%)

Plant Variables Log odds of L. salicaria flowering 0 (0%) L. salicaria stem density 0 (0%) L. salicaria log, mean stem length 0 (0%) Table 2.6b Summary of spatial pattern in insect, damage, and plant variables b. Frequency of responses varying with distance along topographic/moisture gradient Variable Number of sample dates Comments a pattern was found (of 16) Adult Log odds of adult beetles 0 (0%) Loge mean density of adult beetles 0 (0%) Juvenile Log odds of beetle eggs 6 (38%) greater probability of stems with eggs with decreasing elevation on 5 of 14 dates in 1994, 1996, and 1997, however, in May 1995 the probability of stems with eggs decreased with decreasing elevation

Log odds of beetle larvae 1 (6%) greater probability of stems with larvae with decreasing elevation on 1 of 3 dates in 1995

Damage Variables Log odds of leaf damage 2 (13%) greater probability of stems with leaf damage with increasing elevatio on 2 of 8 dates in 1994 and 1995

Log odds of meristem damage 1 (6%) greater probability of stems with meristem damage with decreasing elevation on 1 of 4 dates in 1997 Log odds of damage by other herbivores 2 (13%) greater probability of stems with damage from other herbivores with decreasing elevation on 2 of 8 dates in 1994-1995 Plant Variables Log odds of L. salicaria flowering 2 (13%) greater probability of stems flowering with decreasing elevation on 2 of 8 dates in 1995 and 1996 L. salicaria stem density 0 (0%) L. salicaria loge mean stem length 9 (56%) stem length tended to be greater with decreasing elevation on 9 of 12 dates in 1995-1997 41

Table 2.7 Summary of odds ratios

Variables examined Odds ratio leaf damage / adult beetles 20.0 meristem damage / larval beetles 2.4 beetle eggs / adult beetles 5.1 leaf damage / meristem damage 4.5 leaf damage / beetle eggs 3.9 leaf damage / damage by other herbivores 14.8 42 of beetle presence than the transient variables, such as the organisms that created the damage or oviposited the eggs. Beetles come and go but damage, eggs, and egg shells remain longer in the field and are more likely to be detected. We found that presence of leaf damage was 20 times more likely to be found than the adult beetles that caused that damage. Finding meristem damage was 2.4 times more likely than finding larval beetles.

Finding eggs was 5.1 times more likely than finding the adults that oviposited those eggs.

Leaf damage appeared to be the most descriptive variable as it was 4.5 times more likely to be found than meristem damage and 3.9 times more likely than beetle eggs. We also found that leaf damage by the biological control agents was 14.8 times more likely than finding damage from other herbivores, indicating a biological control "signal" could easily be detected above the "noise" created by other herbivores.

We used the extensive study to track the spread of beetles around the entire lake

(Figure 2.4). Beetles not only spread by simple diffusion around the lake, as shown in the intensive study, but also spread by long-distance dispersal across the lake. This combination of short and long-distance dispersal has been termed "stratified diffusion"

(Hengeveld 1989; Shigesada and Kawasaki 1997). The increase in colonization distance of the beetles across the lake, using direct distance measurements, was approximately linear and was 147.51 m/year (Figure 2.5a). We did not calculate a rate of spread for the distance traveled around the edge of the lake because, due to the pattern of heterogeneous colonization, it was unlikely the beetles spread in this fashion. Radial spread, calculated from accumulated area colonized as a function of time, was also approximately linear and was 10.84 m/year (Figure 2.5b). Using estimates of areal spread we calculated mean- squared radius of distribution using the method of Andow et al. (1993) and the diffusion Figure 2.4 Spread of G. calmariensis and G. pusilla at Morgan Lake continued to increase over the six year monitoring period. Shaded areas indicate presence of adult beetles, beetle larvae, beetle eggs, or beetle damage. By 1997 the beetles had colonized all available host patches and uncolonized potential habitat (unshaded area) in that year corresponded to areas in which the host plant, purple loosestrife, was absent. The prevailing wind blows from west to east at this site. 44

1800 (a) +6' .p.0. 1500 0 in co 1200 0 r2 = 0.86 13 ,.. 900 y = 147.51x + 93.71 E 2 04.. 600 0 CD cC.) 300 across lake - 147.51 m/year 2 N 0 15

70 (b) 60 r2 = 0.96 y = 10.84x + 5.95 50

40

30

20 radial spread - 10.84 m/year 10­

0 I

1200

1000 800 = 0.90 = 230.14x 114.37 600 400

200

o diffusion coefficient (D) = 57.5 m2/yr

-200 I I I I I 1992 1993 1994 1995 1996 1997 Year

Figure 2.5 The biological control agents spread from a single release point to infest most purple loosestrife plants at Morgan Lake within 5 years. The minimum distance around the lake was 1800 m, across the lake was 700 m, and the total area of potential purple loosestrife habitat was 4100 m2 (=64 figure b). Mean radius of distribution and the diffusion coefficient (D) were calculated from area covered. 45

coefficient using the method of Shigesada and Kawasaki (1997) (Figure 2.5c). The diffusion coefficient for the spread of the beetles around Morgan Lake was D = 57.5 m2/year (0.48-0.72 m2/day). Our value of D is similar to values for G. calmariensis calculated from a prior mark-recapture study D = 36.4-54.6 m2/year (Grevstad and Herzig

1997).

Beetles not only dispersed around and across the lake, but also migrated out into surrounding fields. Using sweep samples, we found that beetles were detectable up to 30 m from the edge of the purple loosestrife habitat. Beetles were found throughout the growing season but were most prevalent in the fields when they emerged from diapause in the spring and during July and August after the emergence of the teneral adults from pupation sites as the first and second beetle generations (Figure 2.6).

Impact on individual plants

Damage (measured as the median percent defoliation) increased exponentially during the study (Figure 2.2). Both frequency and intensity of leaf damage increased over time.

During 1994 and 1995, estimates of leaf damage remained at less than 5% of the leaf area removed. In 1996, estimates of leaf damage increased to span the entire range of defoliation, from 0 to 100% of leaf surface removed. During 1997, damage levels remained at approximately 100% defoliation in our sample area, with occasional new seedlings exhibiting little or no damage (Figure 2.7).

Impact on target organism populations

Plant response to beetle defoliation was measured by flowering success, stem length, stem density (stems per 0.125 m2), and biomass (biomass per 0.25 m2). We expected both 3 6 9 12 distance from 15 edge of 18 purple 21 loosestrife 24 Autumn stand (m) 27 30 Summer

Figure 2.6 Beetles were found to spread from the purple loosestrife stand into adjacent fields throughout the sampling period during 1996. 200

150

100

no damage

>0-5%

>5-25% Tv >25-50%

>50-75% to 1. A 1.­ r­ Chr... °.) >75-95% co 9 9 g 18,) In ci 1 75 = Z-71 g? , >95-100% 1 2. 2 < a, cli 16?) 15. c) v 1 g ='" C1 9 a') C4 1997 Nt tif m ci 1,4 ci) t '3' 7 .­ >.. 1 uS Z2 1996 a) .­ N csi - 1995 N r!.. 1994 date sampled

Figure 2.7 Leaf damage increased from very low levels in 1994-1995, to a range of intermediate levels in 1996, and then to very high levels in 1997. 48 beetles and their impacts would decrease from the point of beetle release. We expected a decrease in flowering success, stem density, and plant biomass would start near the point of release and spread over time with the advancing wave of herbivory along increasing distance in X. We expected that mean stem length would exhibit the opposite trend due to disproportionately increased mortality among smaller plants.

Contrary to expectation, no measure of impact yielded a spatial effect at the population level. We found significant spatial effects when examining flowering success and stem length with respect to distance along the topographic/moisture gradient (Y).

Both flowering success (2 sample dates) and stem length (9 sample dates) tended to be positively correlated with increasing distance toward the lake (Table 2.6b, Appendix 2.18 and 2.20).

The proportion of plants flowering increased with plant size before control (1994 and 1995) but flowering virtually ceased, regardless of plant size, as damage increased in

1996 and 1997 (after control) (Figure 2.8). Flowering success decreased over time, declining sharply in 1996 and reaching zero in 1997 (Figure 2.9). Stem density also exhibited the expected decline and decreased from median densities of approximately 4 stems/0.125 m2 in 1996 to 0 stems/0.125 m2 in 1997. As expected, median stem length continued to increase throughout the study from a high of 50 cm in 1994 to 100 cm in

1997. This was correlated with a shift in the frequency distribution of stem lengths from favoring shorter plants when damage levels were low in 1994 and 1995 to a nearly normal distribution by the end of 1996. 49

before control (1994 and 1995)

after control (1996 and 1997) CO C 0.8 .a r2 = 0.89 a) y = 0.0067 - 0.0970 3 p < 0.001 40 U) 0.6 E 4.,O N O C 0.4 0 E 0 0. r2 = 0.19 2 0.2 0 y = 0.0005 - 0.019 p = 0.053

0 20 40 60 80 100 120 140 160 180 200 Stem length class (cm)

Figure 2.8 Flowering probability increased with stem length. Estimates for before control were from July of 1994 and 1995 while estimates for after control were from 1996 and 1997. 26 April 1994 17 May 1994 21 June 1994 25 July 1994 19 August 1994 TO median length = 20 median length = 35 median length = 35 median length = 39 a median length = 50 median density = 4 median density = 5 . median density = 8 median density = 5 11 median density = 2

P 41, 4P 4P I f 4 41. e . 4P 4. 41. $ F e

16 May 1995 16 June 1995 19 July 1995 median length = 48 r. median length = 71.5 tt median length = 56.5 median density = 1 median density = 3 . median density = 4 vegetative plant 0 flowering plant

* 4' d e * d e

26 April 1996 28 May 1996 20 June 1998 11 July 1996 23 August 1996 . . median length = 54 median length = 61 median length = 95 median length = 17 ^ median density = 2 = median density = 6 median density = 3 . median density = 0 Y. m 10

. 4P 4. SI

10 MI

14 III . 11.041111M. . * * e s e e e r * e * e e 0 4P * d . e * e ,P # d F * e *seeeeee 2 June 1997 24 June 1997 18 July 1997 2 August 1997 median length = 82 median length = 72 median length = 70.5 " median length = 100.5 ',­ median density = 0 median density = 0 median density = 0 median density = 0

pi "..ALAIbbaaaL._ * df f.fdf 41. d * f * 4P n# *.sereeee e ee*e*e Stem length (cm)

Figure 2.9 Purple loosestrife response to herbivory by G. pusilla and G. calmariensis over a four year study period. Stem length was measured in centimeters. Density estimates were based on a 0.125 m2 plot size. 51

Biomass of the target plant decreased by 91.6% over time from 1995 to 1997

(Figure 2.10). The ratio of biomass estimates for 1997 (V*) and 1995 (K) yielded a q value of q = 0.084.

Impact on the local plant community

Decrease in L. salicaria biomass had a small effect on the plant community. The dominant species Phalaris arundinaceae and Eleocharis palustris accounted 86% of biomass in 1996 and 89% in 1997. A collection of subdominant species increased by

16%. Subdominant species that increased were Salix piperi Bebb, Cirsium vulgare

Tenore, Rubus discolor Weihe & Nees, Agrostis alba L., Polygonum hydropiperoides

Michx., Holcus lanatus L., Mentha pulgium L., and Bidens cernua L. (Figure 2.11 and

Table 2.8). The decrease of L. salicaria biomass was also correlated with a slight, 3.2% increase of biomass of native species (Figure 2.11 and Table 2.8). Woody plants increased in abundance by 5% while the abundance of forbs (herbs other than grasses, sedges, or rushes) decreased by 8% (Figure 2.11).

Discussion

Establishment and rate of beetle increase

Our estimate of the intrinsic rate of increase calculated from damage was r =

2.24/year. The accuracy of our estimate depends upon the assumption of a linear relationship between leaf damage and herbivore density. In prior studies involving insect pests of crops, host plant injury has been found to increase linearly with the density of insect herbivores (Southwood and Norton 1972; Bardner and Fletcher 1974). 52

15

,.., E

-0- indirect biomass

-0- direct biomass

:15

cu 2ci)

0 1994 1995 1996 1997 Year

Figure 2.10 Biomass of purple loosestrife declined precipitously between 1996 and 1997. Indirect biomass refers to estimates calculated from stem length using a standard regression equation, while direct biomass refers to estimates from weighed samples. Confidence intervals (95%) were calculated for estimates from weighed samples (n=44). 53

Phalaris arundinaceae

Eleocharis palustris

Lythrum salicana

Said piperi

Carex unilateralis

Cirsium vulgate 1996

Rubus discolor 1997

Agrostis alba

Polygonum hydropiperoides

Holcus lanatus

Mentha pulegium

Bidens cemua 1

Solanum dulcamara 1996 1997 Centaunum umbellatum

Vicia totrasperma cu 13% C. Daucus carota native

Hypochaeris radicata 111

Cirsium atvense

Crataegus douglasii

Parentucellia viscosa 90% 87% Epilobium watsonii it,voody plants Alisma plantago-aquatica 11% forbs Convolvulus arvensis

Lotus micrantha

Ludwigia palustris

Anthoxanthum ororatum

Galium aparine 88% 91%

Lycopus amencanus

Cerastium viscosum

Ceratophyllum demersum

Naverretia intertexta

1 10 100 1000 Mean biomass (grams/0.25 m2)

Figure 2.11 Plant community composition at Morgan Lake during 1996 and 1997. 54

Table 2.8 Biomass data for plant community at Morgan Lake

Species 1996 Biomass 1997 Biomass mean (0.25 m2) percent mean (0.25 m2) percent Exotic species Agrostis alba 0.814 0.305 1.841 0.779 Anthoxanthum ororatum 0.009 0.003 0.000 0.000 Cerastium viscosum 0.005 0.002 0.000 0.000 Cirsium arvense 0.048 0.018 0.000 0.000 Cirsium vulgare 0.000 0.000 3.805 1.611 Convolvulus arvensis 0.005 0.002 0.007 0.003 Daucus carota 0.075 0.028 0.030 0.013 Holcus lanatus 0.016 0.006 1.227 0.520 Hypochaeris radicata 0.095 0.036 0.000 0.000 Lythrum salicaria 29.432 11.043 0.334 0.141 Mentha pulegium 0.175 0.066 0.796 0.337 Parentucellia viscosa 0.036 0.014 0.000 0.000 Phalaris arundinaceae 209.764 78.705 194.514 82.346 Rubus discolor 0.000 0.000 3.193 1.352 Solanum dulcamara 0.195 0.073 0.000 0.000 Total 240.668 90.300 205.746 87.102

Native species Alisma plantago-aquatica 0.016 0.006 0.000 0.000 Bidens cernua 0.000 0.000 0.582 0.246 Carex unilateralis 2.848 1.068 1.700 0.720 Centaurium umbellatum 0.152 0.057 0.000 0.000 Ceratophyllum demersum 0.005 0.002 0.000 0.000 Crataegus douglasii 0.041 0.015 0.000 0.000 Eleocharis palustris 20.665 7.754 15.439 6.536 Epilobium watsonii 0.023 0.009 0.000 0.000 Galium aparine 0.007 0.003 0.000 0.000 Lotus micranthus 0.007 0.003 0.005 0.002 Ludwigia palustris 0.011 0.004 0.000 0.000 Lycopus americanus 0.007 0.003 0.000 0.000 Naverretia intertexta 0.000 0.000 0.002 0.001 Polygonum hydropiperoides 0.145 0.055 1.695 0.718 Salix piperi 1.805 0.677 11.045 4.676 Vicia tetrasperma 0.120 0.045 0.000 0.000 Total 25.851 9.700 30.468 12.899 55

Specifically, this linear pest-injury relationship was found for the cereal leaf beetle Oulema melanopus on oats, an insect that is in the same family as the beetles we studied

(Chrysomelidae) (Southwood and Norton 1972). Therefore, we felt justified in approximating r from estimates of damage, which was probably a more accurate measurement of beetle density than direct observation, as adult beetles tended to drop from the stems when approached.

Our estimate of the intrinsic rate of increase (r = 2.24/year) was between that of a prior successful North American invader, the cereal leaf beetle Oulema melanopus, whose intrinsic rate of increase was 1.6-1.9/year (Shigesada and Kawasaki 1997) and that estimated for Cactoblastis cactorum (Bergroth) (Lepidoptera: Pyralidae) (Caughley and

Lawton 1981), r = 3.6-3.8/year, a successful biological control agent of the prickly-pear cactus (Opuntia inermis de Candolle and Opuntia stricta (Haworth)) in Australia.

Therefore, our estimate of the intrinsic rate of increase for G. pusilla and G. calmariensis is in the range predicted for successful biological control agents and for successful insect invaders (Table 2.1).

Beetle spread

We found that the beetles spread by stratified diffusion, combining simple diffusion from the point of release with longer distance dispersal across the lake. Stratified diffusion

(Hengeveld 1989) can be modeled using the coalescing-colony model proposed by

Shigesada and Kawasaki (1997) for the modeling of spread of invasive species. This proposes an acceleration of radial distance covered by the organisms over time due to the independent growth of secondary colonies. This model may be contrasted with the simple 56

diffusion model where radial growth is constant (Skellam 1951; Williamson 1996).

Despite fitting the stratified diffusion model's definition, our system exhibited a constant

radial growth and we found that the simple diffusion model fit our data well (Figure 2.5c).

This may have been due to the limited area available for colonization or possibly the shape

of the potential habitat. Most growth models assume a homogenous habitat where a

population of an invading organism can increase in concentric rings from the point of

introduction (Skellam 1951; Hengeveld 1989), although more recently effects of barriers to organism spread models have been discussed (Andow et al. 1993; Shigesada and

Kawasaki 1997). However, we were unable to find any discussion in the literature of the

effect of a roughly donut-shaped habitat, as in the riparian zone around a lake, on models

of organism spread. This difference in shape allows for a small host habitat area but

relatively large colonization distance and explains the discrepancy between our calculation

of linear distance covered by the beetles of 147.51 m/year and our estimate of radial

spread of only 10.84 m/year (Figure 2.5a and b).

The estimation of the diffusion coefficient for G. calmariensis from a previous mark-recapture experiment was 36.4 - 54.6 m2/year (Grevstad and Herzig 1997). This is surprisingly close to our own estimate of 57.5 m2/year. However, because the prior mark- recapture study used a target-centered design where the beetles were released into open fields at varying distances from the host patch, this estimate may be an overestimation of

spread as it forces beetles to move or risk starvation. On the other hand, it may underestimate dispersal, because only the area of the target patch was searched for beetles and the fate of those not recaptured is unknown, so it is possible that they may have traveled to other host patches some distance away. Although we found no studies that 57 provide estimates of D for biological weed control agents, there are estimates for invasive exotic insect species. We found the rate of spread of G. pusilla and G. calmariensis was substantially less than that estimated for the gypsy moth Lymantria dispar (340 m2/year) and the cereal leaf beetle Oulema melanopus (400 m2/year) but similar to that reported for other chrysomelid species (Table 2.2).

We expected beetles to move around the lake, but we were surprised to find that large numbers of beetles also periodically migrate away from the purple loosestrife stand into the adjacent fields. This occurred during much of the growing season and the beetles may diapause within these areas during the winter (Figure 2.6). Therefore, any disturbance in this area is likely to negatively affect beetle populations and decrease their effectiveness as biological control agents by increasing mortality and reducing population growth rates. Although we detected beetles up to 30 m of a possible 50 m from the edge of the loosestrife, sweep sampling may have failed to detect very low beetle densities at longer distances.

The best indicators of beetle presence in our study were the cumulative variables of presence of leaf damage, meristem damage, and beetle eggs. Leaf damage was 20 times more likely to be found than were the adults that caused the damage and 4.5 times more likely than the runner-up, meristem damage (Table 2.7). Leaf damage also exhibited the most consistent pattern with respect to both time and distance from the initial beetle release site (Figure 2.3). For managers that seldom have time to measure all of the variables that we examined here, we recommend leaf damage as the best single measure of beetle activity for use in monitoring biological control of purple loosestrife. 58

As expected, using spatial autocorrelation analysis, we initially found significant a spatial effect of beetles (1994 and 1995), as the beetles spread and colonized new host plants. Later, in 1996 and 1997, as the beetles saturated the riparian habitat around

Morgan Lake (300 m from the beetle release point), the effect disappeared. This spatial effect was most evident in our best indicator of beetle activity, presence of leaf damage.

Therefore, we conclude that the beetles homogeneously colonized the habitat of their resource, purple loosestrife.

A shortcoming of our intensive study was that the area of the sample did not increase fast enough to keep up with the spreading beetle population. By 1996, the beetles had spread beyond the edge of our intensive sample area, spreading along the constricted corridor around the lake and forming new colonies across the lake. The new colonies coalesced with the primary colony and nullified the spatial pattern of simple diffusion (Figure 2.4). Despite this rapid dispersal, the beetles remained at high enough densities near their initial release point to cause damage to increase from approximately

5% leaf area removed in 1995 to levels of around 100% defoliation in late 1996. These damage levels remained at around 100% throughout 1997 (Figure 2.7). This indicates that the beetles persist in high densities long enough to cause significant damage to their host plant. This aggregation may be due to conspecific attraction as documented in a field experiment by Grevstad and Herzig (1997) and could be an important factor in determining the magnitude of the reduction of the target plant abundance. 59

Impact on Individual Plants

As expected, damage by the beetles increased in both frequency and intensity over time. We found that the presence of leaf damage exhibited the clearest spatial pattern over distance as this symptom spread from the initial beetle release site outward along the edge of the lake (Figure 2.3). It also showed the most consistent temporal trend, as the difference in damage levels quickly increased and eventually became homogenous (Figure

2.3). Intensity of damage exhibited an exponential increase over time. This can be seen directly as the median percent leaf damage per year (Figure 2.2). This trend is also seen in the damage frequency distribution over time (Figure 2.7).

Impact on Target Plant Population

Massive defoliation, beginning in 1996, caused a considerable decrease in purple loosestrife flowering success, density, and biomass in 1997. Damage increased rapidly throughout our sampling area in 1996 (Figure 2.7) and therefore we did not find the expected spatial pattern of these plant traits decreasing first near the beetle release point and spreading outward from there (Table 2.6a). Instead, we recorded a simultaneous decline in these variables within our sample area. Similar effects of the beetles were found on European populations of purple loosestrife (studied near Kiel, Germany). At two of four sites examined, the beetles caused high plant mortality and reduced seed output

(flowering) to < 1% of levels prior beetle colonization (Blossey 1992).

Biomass decreased to 8.4% (q = 0.084) of its prior level (Figure 2.10). The new local level of abundance in our study is slightly lower than regional level of 10% of former abundance (q = 0.01) predicted by Malecki et al. (1993). This dramatic reduction 60 occurred approximately six years after the initial biological control agent introduction, which compares favorably with the predicted five years for visible local effect (Blossey

1995). Other biological control programs have reported a wide range of values of q

(Table 2.3). Generally the more spectacular successes, such as control of the prickly-pear cactus in Australia (Caughley and Lawton 1981) and St. John's wort (Hypericum perforatum L.) (Huffaker and Kennett 1959) and tansy ragwort (Senecio jacobaea L.)

(McEvoy et al. 1991) in the United States have been associated with q-values of less than

0.01 (99% decrease in target plant abundance). However, our value of q = 0.084 is a local estimate representing a limited sample of space and time. It is too soon to characterize the new, regional equilibrium that will unfold as control spreads to more sites over time.

Stem length increased over time but because the frequency distribution of stem heights was not normal we chose to express it as the median and not mean per plot. This increase in stem length was due to the disproportionate mortality of short stems which changed the frequency distribution from favoring shorter stems in 1994 and 1995 to an almost normal distribution in 1997 (Figure 2.9). Changes in frequency distribution in response to parasitism/herbivory have been shown in the biomass of skeleton weed

Chondrilla juncea L., another exotic weed species, due to disease pressure from the rust

Puccinia chondrillina Bubak and Syd. However, in this case, biomass shifted from a normal distribution to a skewed distribution (Burdon et al. 1984). The change in frequency distribution to favor taller plants also follows the self-thinning rule. Although, the self-thinning rule was proposed as a response to intraspecific competition and not parasitism/herbivory (Silvertown 1993), it is also derived from disproportionately 61 increased mortality among smaller and weaker plants and the initial result, the plant height frequency distribution shifting to favor taller plants, is the same.

We found three variables that exhibited significant patterns with reference to the topographic/moisture gradient along the Y-axis: stem length, flowering success, and presence of eggs (Table 2.6b). Stem length showed the strongest pattern. During 9 of the

16 sample dates (56%) stem length tended to be greater with increasing distance along the topographic/moisture gradient. Early in the season, these stems tended to be submerged, which may act as a temporary refuge from herbivory, allow plants to begin growing earlier in the season by providing protection from early spring frost damage, reduce competition from the other dominant species (reed-canary grass), or may simply force the plants to grow taller in order to reach open air in order to respire and photosynthesize.

Flowering probability also exhibited a spatial pattern across the topographic/moisture gradient. Flowering success tended to be greater with increasing distance on 2 of the 16 sampling dates. Although this constitutes a pattern during only

13% of all sample dates, since flowering is generally restricted to July and August, we can recalculate this percentage as 2 of 7 samples showing spatial pattern (29%). This is probably not surprising, as we would expect larger plants to have greater reproductive success. It may also be due to temporary refuge from beetle attack during times when plants are submerged, as discussed in the previous paragraph. We used linear regression to examine the relationship between stem length and flowering success and found that before control, 1994 and 1995, stem length was positively correlated with flowering success. As damage increased in 1996 and 1997 this correlation disappeared with the reduction of flowering stems (Figure 2.8). To see whether this increased stem height and 62 flowering success may have been caused by an early season refuge from herbivory we examined spatial patterns among probability of leaf damage. We did find that on two occasions damage appeared to occur less frequently with increasing distance along the topographic/moisture gradient (Table 2.6b and Appendix 2.12). This does provide some evidence of a refuge from herbivory. However, these patterns were only found on 2 of 16 sample dates (13%).

The third variable that showed strong patterns with reference to the topographic/moisture gradient was presence of beetle eggs (Table 2.6a and Appendix 2.4).

On 5 sample dates (31%), generally later in the growing season (1 in June, 2 in July, and 1 in August), the probability of stems having eggs increased with increasing distance along the gradient, while on one date (May) (6%) the opposite spatial trend was found. This means that increased probability of finding eggs was generally correlated with increased stem length. This indicates that beetles may locate longer stems more easily or preferentially choose taller stems for oviposition sites. It may also be an effect of the water barrier. We found that beetles migrate out of the purple loosestrife stand, on the terrestrial edge, into adjacent fields. However, the aquatic edge may hinder migration and this may cause an increase in beetle density in the second beetle generation (as the first generation generally moves in from the fields after overwintering). This would explain the greater probability of finding eggs near the water's edge later in the growing season.

Impact on local plant community

When we examined the impact of L. salicaria biomass reduction on the local plant community we found a 3.2% increase in the biomass of native plant species (Figure 2.11 63

and Table 2.8). This result is mainly due to an increase of Salix piper/. Since this is a

native species, we might conclude that the decrease in L. salicaria is facilitating the return of the native plant community. However, this increase is probably caused by a change in management practices for the control of coypu (or nutria, Myocaster coypus) at the site rather than the decrease in purple loosestrife. In a study at the William L. Finley National

Wildlife Refuge, 90 km to the south of our study site, Wentz (1971) found that Salix species were the preferred forage of coypu in the area (12.3% of all foraging observations). During 1996, the refuge manager removed approximately 25 coypu from around our study site (Richard Guadagno personal communication 1997), which probably resulted in the increase of Salix piper/ biomass in 1997.

Two other native species increased in biomass during the time between samples.

These were Polygonum hydropiperoides and Bidens cernua. However, these two species represented less that 1% of the total biomass in 1997 (Table 2.8). Therefore, we found no appreciable increase in the aboveground biomass of native species in the plant community around Morgan Lake that can unequivocally be attributed to decline and fall of the purple loosestrife population over this period. However, because the lake was initially created by an earthfill dam in 1962, it may be that there is no history of native riparian vegetation and therefore no seed bank of native species. Another possible explanation is that more time is needed for recolonization by native species. However, given the apparent competitive ability of the dominant invasive species, reed canary grass (Phalaris arundinaceae), recolonization by natives is unlikely. A study has been proposed to examine the use of herbicide treatments in conjunction with biological control agents to repress the 64

populations of the two dominant exotic species, P. arundinaceae and L. salicaria

respectively.

Human-created wetlands like that studied here may have special relevance toward wetland mitigation projects. Wetland mitigation laws are intended to allow human

development of natural wetlands with the stipulation that a new wetland will be created at

another location so that no net loss of wetlands occur and therefore ecological functions are preserved (Zelder 1996). Recent studies question our ability to create wetlands that are equivalent in form and function to those that are destroyed (Malakoff 1998). Our results indicate that, if the native species assemblage is linked with wetland ecological function, we may not be able to recreate wetlands that duplicate natural wetlands.

Instead, we may create disturbed habitat that favors colonization by exotic species where biological control efforts simply switch dominance between invasive exotic species.

Conclusions

In conclusion, we found that the beetles were effective at reducing the biomass of purple loosestrife by 91.6% (within our sampling area) at Morgan Lake within 5 years.

Concurrently, they rapidly colonized the entire host plant habitat of approximately 4100 m2. This is evidence of their potential as effective biological control agents of purple loosestrife. However, we have little knowledge of persistence of the insect-plant interaction in an environment at low densities of the target plant in North America, although field studies in Europe indicated both species were common in all areas surveyed, even in purple loosestrife populations of less than 10 plants (Blossey 1995a). Also, since our data are restricted to a single location, we have no evidence of how habitat variation 65 influences the effectiveness of these beetles or their ability to disperse larger distances between purple loosestrife stands. In Europe, it was found that the purple loosestrife populations at two of four sites were drastically reduced by the beetles, but the other two experienced medium to light defoliation and little effect was found on plant performance

(Blossey 1995a). As the beetles are common throughout Europe, we might expect this to indicate that they will disperse as well in North America. However, as purple loosestrife is still uncommon in parts of North America, the spatial pattern of host patches may be significantly different than in Europe. More study is needed to examine the questions of:

1) persistence: Is the system stable? Will the beetles continue to depress the abundance of purple loosestrife while maintaining healthy populations themselves? and 2) regional dispersal: Will the beetles be able to colonize, and potentially recolonize, all populations of purple loosestrife without human intervention?

Another pertinent subject is the role of the other two introduced biological control agents: the seed-feeding weevil Nanophyes marmoratus and the root-feeding weevil

Hylobius transversovittatus. The destruction of the plant stages invulnerable to G. pusilla and G. calmariensis, seeds and roots, may play an important role in the speed and magnitude of the reduction of purple loosestrife abundance. However, the overwhelming damage caused by G. pusilla and G. calmariensis may also negatively affect the establishment of populations of the other agents. Knowledge of these interactions will be important in future releases of the insects in terms of where and when to release what combinations of agents in order to maximize the efficiency, speed, and magnitude of reduction of target plant abundance. 66

Although the spread of the beetles at our site was most simply explained by the passive diffusion model, given a larger habitat or a longer time series, the coalescing- colony model may become a better descriptor of beetle spread. The theory behind the coalescing-colony model has implications for biological control agent release strategies.

That is, many releases at a site will tend to accelerate D, and therefore increase the colonization rate, and presumably, the speed of control of the target organism (Shigesada and Kawasaki 1997). However, this must be balanced by the need to release enough individuals at each site in order to reduce probably of extinction due to Allee effects, reduced adaptability due to lack of genetic diversity, and demographic and environmental stochasticity (Grevstad 1998). Grevstad (1998) found that in releases of 20, 60, 180, and

540 individual beetles, probability of extinction for both G. pusilla and G. calmariensis decreased with increasing release sizes, where the releases of 540 individuals exhibited a greater than 80% establishment rate. This indicates that the optimal release strategy is multiple introductions at a single site with each introduction consisting of more than 540 beetles of a particular species (assuming a 1:1 sex ratio).

We suggest three improvements on this design. First, it would be useful to more accurately quantify beetle density in order to calculate diffusion coefficients and examine alternative models of spatial spread (Shigesada and Kawasaki 1997). Second, a control treatment, using cages or insecticides, would solidify the link between cause and effect

(McClay 1995). Third, rather than estimating beetle damage using damage classes for individual leaves, we suggest estimating damage directly as a percent defoliation for entire plants. This would eliminate the difficulty we had in analyzing data from damage classes and would reduce observer bias in choosing the damage class for the "average leaf'. 67

Here we constructed and implemented a monitoring design and assessed the usefulness of specific variables for potential use in future monitoring studies. We also provided a time frame in which to predict results. These, of course, may vary for different

climates and habitat types. 68

Chapter 3

Field Test of Host Specificity of Two Chrysomelid Beetles (Galerucella pusilla and G. calmariensis) Introduced to North America for Biocontrol of Purple Loosestrife Lythrum

salicaria

Shon S. Schooler, Eric M. Coombs, Peter B. McEvoy, and Don B. Joley 69

Abstract

We used field and greenhouse experiments to assess non-target effects of two introduced biological control organisms (Galerucella pusilla Duftschmid and G. calmariensis L.: Chrysomelidae) on the economically important ornamental plant, crape myrtle (Lagerstroemia indica L: Lythraceae). The beetles were introduced to Oregon in

1992 for the control of purple loosestrife (Lythrum salicaria L: Lythraceae), an invasive exotic perennial wetland plant of Eurasian origin. Prior host specificity tests performed in the laboratory found that beetles fed, but were unable to complete their life cycle on this non-target plant. However, there was concern over damage that might occur when the two plant species existed together. This study extended prior tests into an in situ environment in order to compare the physiological host range revealed in greenhouse tests with the ecological host range revealed in the field.

We assumed, based on prior evidence, that the control agents would not complete development on the non-target plant, and therefore, when the non-target organism was isolated from populations of the target organism the direct effects of the biological control agents would be negligible. When the target and non-target organisms existed together, the magnitude of indirect effect of the target organism on the non-target organism via the control agent was expected to increase with: 1) decreasing distance between the target and non-target organisms, and 2) increasing dispersal capability of the control agents. As expected from prior studies beetle feeding and oviposition occurred on crape myrtle but the beetles could not complete development on this non-target plant in our greenhouse and field tests. Leaf damage inflicted by the beetle was lower on crape myrtle than on purple loosestrife plants used as controls and extensive defoliation to the non-target plant 70 was limited to within 30 m from the edge of the purple loosestrife stand. Mean biomass of

crape myrtle plants in the stand was 35% less than that of plants that remained relatively untouched at the distance of 50 m. Purple loosestrife biomass in the stand was 55% less than that of those plants 50 m from the edge of the stand. 71

Introduction

Host specificity tests typically measure whether a proposed biological control agent can complete its life cycle on a non-target plant isolated in a closed system (McEvoy

1996). Although a biological control agent may not be able to form self-sustaining populations on a non-target plant, it may still inflict significant damage on non-target plants located near populations of the target plant. This non-target damage by the control agent is considered an indirect effect because the damage on the non-target plant is dependent upon the presence of the third species, the target plant (Figure 3.1) (Wootton

1994). Damage to the non-target plant will be dependent upon: 1) overlap of the geographic ranges of the target and non-target plant, 2) local distance between the target and non-target plant populations, and 3) colonization ability of the biological control organisms. Using control of purple loosestrife Lythrum salicaria L. (Lythraceae) as a case study, we examined the effect of two introduced Chrysomelid beetles (Galerucella pusilla Duftschmid and G. calmariensis L.) on the economically important ornamental plant, crape myrtle Lagerstroemia indica L. (Lythraceae). We present a procedure for extending host specificity studies to include effects of distance and colonization on potential non-target effects.

A primary concern in the use of classical biological control is the possibility of non-target effects (Zwolfer and Harris 1971; Andres 1981; Turner, 1985; Howarth 1991;

Diehl and McEvoy 1990; McEvoy 1996; Simberloff and Stiling 1996; Louda et al. 1997;

Strong 1997). That is, a foreign organism purposefully introduced to control a foreign pest may itself become a pest. In weed biological control programs, this risk is potentially large due to the possibility of introducing new herbivorous pest organisms into 72

Control Agent

Target Non-target Organism Organism

Figure 3.1 Indirect effects of the target organism cause damage to the non- target organism via the control agent. Solid lines indicate strong effects while dashed lines indicate weak effects. Arrows indicate positive effects while open circles indicate negative effects. 73 economically important agricultural systems. The solution to this problem has been the implementation of host specificity testing.

Usually tests of host specificity are measured on organisms in confinement or

"closed systems" that restrict freedom of movement and choice (Zwolfer and Harris 1971;

Cullen 1990; Harley and Forno 1992; McEvoy 1996; Secord and Kareiva 1996). This approach is adequate when determining the physiological host range of a biological control organism. However, it overestimates the ecological range that is expressed in the field where biological control organisms are allowed freedom to move about and choose among plants in the field environment (Cullen 1990; Clement and Cristofaro 1994). Open- field tests allow insects the freedom to move and choose among potential host plants and therefore provide a more accurate and realistic representation of their effects on non- target plants (Cullen 1990; Clement and Cristofaro 1994; Secord and Kareiva 1996).

In this study we examined the potential non-target effects of two biological control agents, Galerucella pusilla and G. calmariensis, that were introduced in Oregon in 1992 to control the exotic wetland plant, purple loosestrife (Thompson 1991; flight and Drea

1991; Malecki et al. 1993). Prior studies have reported feeding and oviposition by adult beetles on crape myrtle (Blossey et al. 1994), which is an ornamental plant economically important to the California nursery industry that was originally introduced from Asia and

Australia (Dunmire 1979, Martin 1983). It was included in the initial host specificity screening process because it is in the same family as purple loosestrife (Lythraceae)(Table

3.1). The results of no-choice testing indicated feeding and oviposition by adults and occasional larval feeding, although the beetles could not complete their life cycle on the non-target plant (Blossey et al. 1994). The California Department of Food and Agriculture Table 3.1 Forty-eight plant species approved for host specificity screening with biological control agents for L. salicaria

A. Taxonomically associated: B. Associated plants of wildlife importance: C. Important agricultural plants: Lythraceae Typhaceae Poaceae Lythrum salicaria L. Typha latifolia L. Triticum aeslivum L. "Blue Boy" L. lineare L. Sparganiaceae Oryza sativa L. "Bluebonnet" L. alatum Pursh. Sparganium eurycarpum Engelm. Zea mays L. "Pioneer 37-44" L. californicum Ton. & Gray Alismataceae Chenopodiaceae L. hysopifilia L. Sagittaria latifolia Willd. Beta vulgaris L. "Golden Tankard" Decodon verticillatus (L.) Ell Poaceae Fabaceae Rota la ramosior (L.) Koehne Zizania aquatica L. Glycine max L. "Pella" Ammania auriculata Wild Cyperaceae Malvaceae A. coccinea Rottb Scirpus americanus Pers. Gossypium hirsutum L. "Tameot" A. robusta Heer & Regel S. acutus Muhl. Asteraceae A. lattifolia L. Carex comosa Bostt. Helianthus annuus L. "Mingren" Cuphea wrightii Gray Salicaceae C. viscosissima Jacq Salix interior Rowlee C. laminuligera Koehne Polygonaceae C. lanceolata Alton Rumex verticillatus L. C. lutea Rose R. crispus L. Lagerstroemia indica L. Polygonum coccineum Muhl. Punicaceae Chenopodiaceae Punica granatum L. Chenopodium hybridum L. Melastromataceae C. album L. Rhexia mariana L. Ranunculaceae Ranunculus sceleratus L. Table 3.1 (continued)

A. Taxonomically associated: Onagraceae Ludwiga alternifolia L. Epilobium angustifolium L. Oenothera biennis L. Gaura parviflora Dougl. G. biennis L. Circaea quadrisulcata (L.) Hara Thymelaceae Dirca palustris L.

( from Blossey et al. 1994) 76

(CDFA) denied permission to release the beetles within the state due to uncertainty about the potential impact on this non-target plant species. Concern for safety must be balanced with concern for the economic threat of encroachment by purple loosestrife on wild rice production in the Fall River Drainage (Shasta Co., CA) and the environmental threat of displacing native vegetation and wildlife.

We expected, as documented in a prior study (Blossey et al. 1994), that the control agents could not complete development on the non-target plant, and therefore, when the non-target organism was isolated from populations of the target organism the direct effects of the biological control agents would be negligible. The magnitude of indirect effect of the target organism would be dependent upon: 1) overlap of the geographic ranges of the target and non-target plant, 2) local distance between the target and non- target plant populations, and 3) colonization ability of the biological control organisms.

Our study was designed to clarify the effects of Galerucella pusilla and G. calmariensis on crape myrtle using both open-field and greenhouse experiments. We were particularly interested in how non-target plant damage decreased with respect to distance from a colonization source that was at "outbreak" density levels. The three main questions were:

1) What is the capacity of the insects to feed and develop on the non-target plant?

2) What is the capacity of the insects to colonize the plant in the field?

3) What is the influence of insect feeding on the plant's growth, survivorship, and reproduction? 77

Study System

The Target Plant

Purple loosestrife (Lythrum salicaria) is a tall (2 m), iteroparous, perennial wetland plant native to Europe. It probably arrived on the East coast of the United States before 1830, in ballast deposited by trading ships from Northern Europe (Thompson

1991). It has since been spreading across the country, aided recently by road construction and irrigation channels (Wilcox 1989), as well as through the planting of seeds sold in wildflower mixes (Figure 3.2). A mature plant can produce as many as 2.5 million seeds annually (Malecki et al. 1993) which are dispersed by water or in mud adhering to animals.

Its mean rate of spread since 1940 has been estimated at 645 km2 per year (Thompson

1991).

Purple loosestrife is an invasive species that quickly displaces native wetland vegetation in riparian areas, often forming dense monospecific stands that degrade habitat quality for waterfowl and other wetland animal species (Thompson et al. 1987; Balogh and Boohhout 1989). In the U.S. the estimated cost of infestation, in terms of wildlife and agriculture, is $45 million per year (Thompson et al. 1987). However, the negative effects of purple loosestrife on native species are poorly documneted (Anderson 1995).

The Non-target Plant Crape myrtle L. indica is an ornamental plant that is economically important to the

California nursery industry. It is typically a large shrub or small tree, although there are also dwarf varieties (Martin 1983). It was originally introduced from Asia and Australia

(Bailey 1951; Dunmire 1979) and needs hot summers and well drained soil in order to 78 flower and is therefore cultivated mainly in the southern portion (zone 7) of the United

States (Figure 3.2) (Martin 1983).

The Biological Control Agents

Galerucella pusilla L. and Galerucella calmariensis L. (Chrysomelidae) were released at our study site, Morgan Lake (Baskett Slough NWR, Western Oregon), in

1992.

Adult beetles emerge from their diapause sites in the fields surrounding the purple loosestrife stand in late March to mid-April. They then migrate to the stand to feed on the meristematic tissues of young loosestrife plants. Mating begins soon after the beetles

emerge and females begin oviposition after about one week. Oviposition peaks in mid-

May and tapers off around mid-July. Adult beetles typically live for 40-60 days (Ragsdale

1996). Larval beetles proceed through three instars, during which, they preferentially feed

on the apical meristem of the host plant. Once the tip is consumed they move down the plant feeding on the undersurface of the leaves. After three instars, the larvae search for

adequate pupation sites in the aerenchyma tissue of the host plant or in the litter

surrounding the plant's root mass.

The first teneral adults emerge from the pupation site approximately 5 weeks after the beginning of oviposition. They then either leave hosts and migrate to the nearby fields

(possibly entering a state of aseasonal quiescence) or remain on hosts to feed, mate, and

oviposit, beginning a second generation. It is not known what triggers migration. Those

beetles that remain in the loosestrife stand produce a second generation and die in late

August. The second generation pupates in late August and teneral adults emerge in mid­ 5.4 /1 ,,;P' ir -, , , 04 I

I r- cj 14., ) .... - i c ', (C% a--1' '-',\---­ ,z-..., .) ,..r7 ,Nc, -) PRA)R i£\_ ,P5 HOLE

.., .-1-4.:2 -4\y/A-TRFOveL BR EED i I4G i 1. ,­ -1,41-. -_r I. $1/4 1--- ,L cRourios -----i\ ­ ----,_,-,..--_.. -- -,----.., , i ---,'.-,-.__ .s.'..---1---,---- -.....,_....,,,.S. -%.. s----­ r-' .--- , of...... ,e, ,_.:2 k II) 4 i t 1 1 : . -- -4, q i- - --- k - . -:,..: i, / CL. ....-< . 0, "-i4c...... is.7._ 4_,"*. iTipelt I f ) --- 8: ;itt ; A., . :. --( -i- - -a; I( '\.-1...../: .-1. 1 9 ,1----- . kn ; 1----,. cr ,,,_I, '- 1-- -. _....1-=-_...... ,..:* _..--.''. a* i I el%... ! , ____...., I . ------­

.. ak lrvOi 44. 4,74

Figure 3.2 Distribution of purple loosestrife (dots) and crape myrtle (shaded zone) in the United States and Canada (Thompson et al. 1987; Martin 1983). 80

September. In late September, the second-generation adults migrate into the fields surrounding the riparian zone and enter a diapause from which, if left undisturbed, they emerge the following spring.

Two generations have been recorded for both species in Oregon while only one is found in the shorter growing seasons of the Midwest and Eastern regions of the United

States (Bernd Blossey personal communication 1997). Four generations per year have been recorded in Northern Italy (Batra et al. 1986). Adult beetles of the two species can only be easily distinguished after the teneral stage (approximately one week after emergence). G. pusilla is smaller and has solid golden-brown elytra while G. calmariensis is slightly larger and orange-brown with two dark stripes that run dorsal-laterally down the beetles' elytra.

Dispersal has been previously studied for G. calmariensis. Experimental study of movement between patches using a target-centered mark-recapture method indicated that

G. calmariensis adults can detect host patches within 50 m and are capable of flying at least 850 m within one week (Grevstad and Herzig 1997). This study also documented conspecific aggregation of G. calmariensis. We documented migration patterns of the beetles into the fields surrounding the purple loosestrife stand in 1996. We found that the beetles migrated up to 30 m from the edge of the stand and this migration coincided with the emergence of teneral and overwintering adults during summer and early spring, respectively (Chapter2). 81

The Study Site

The fieldwork was conducted at Morgan Lake in western Oregon (Polk Co.- Lat.

45°, Long. 121'). Morgan Lake is part of Baskett Slough National Wildlife Refuge, one of several refuges created to protect the flyway of the dusky Canada goose (Branta canadensis occidentalis), an endangered subspecies of the Canada goose. This site was selected because mechanical disturbance to the lake and surrounding vegetation is minimal, although the adjacent fields are cultivated for grass seed and goose forage.

Morgan Lake was created by an earthfill dam in 1962 (Figure 3.3). Reed-canary grass (Phalaris arundinaceae L.) and purple loosestrife (L. salicaria) dominate the strip of riparian vegetation encircling the lake (Chapter 2). Outside of the riparian zone on the terrestrial side is a narrow transitional zone to open fields. Cultivated fields surround the lake on all sides with the exception of thin corridors of vegetation accompanying several streams that feed the lake and one that flows out from the dam.

The two beetles that we studied, G. pusilla and G. calmariensis, were introduced in a 3.3 x 3.3 m enclosure at the middle of the south side of the lake on July 24 of 1992

(Figure 3.3). On May 14 of 1993 the cage was lifted and the beetles were allowed to disperse. By 1996 the beetles had increased in abundance to levels that were severely defoliating the loosestrife plants near the initial release site and this continued through

1997 (Chapter 2). The superabundance of herbivores gave us the opportunity to test a worst case scenario on negative effects to local non-target vegetation and the possible mitigating influence of increasing spatial separation between target and non-target organisms. Access Road Cultivated Field

-, -, ft -.--­ 4--- Earthfill --N \ ..---­ ___.--­ Dam -..:N_ .-­ -,,,,,:---____ ,---­

Morgan Lake

X Cultivated Field Beetle Release Site Open Water

Cultivated Field Riparian Zone

100m

Figure 3.3 Study Site: Morgan Lake (Baskett Slough National Wildlife Refuge: Western Oregon) 83

Methods

Larval Development Studies

To assure correct identification of species based on adult characters, larvae used in the laboratory feeding studies were reared in the greenhouse from adults collected from a local population at Baskett Slough National Wildlife Refuge in Western Oregon. Adults from the overwintering beetle population were collected using a sweep net in April of

1996. Approximately 40 individuals of each species were placed on caged purple loosestrife and crape myrtle plants and left to oviposit for three days. They were then removed and plants were checked daily for larval emergence.

Ten potted purple loosestrife plants and ten potted crape myrtle plants

(approximately 1 year old and 30 cm in height) were selected and randomly placed into a shallow pool with 5 cm of water in order to discourage larval movement to other plants.

Using a fine paintbrush five newly emerged larvae of G. pusilla were transferred to each of five potted purple loosestrife plants and each of five potted crape myrtle plants. This process was repeated for G. calmariensis larvae. Larvae placed on crape myrtle were taken from eggs oviposited on crape myrtle while larvae placed on purple loosestrife were taken from eggs oviposited on purple loosestrife. The greenhouse was kept at 24°C with a natural photoperiod of approximately 16:8 (Light:Dark). The number of larvae on each plant was counted at four-day intervals until they reached their third instar. The remaining larvae were then placed in transparent plastic containers with fresh leaf material and left to pupate. The numbers of pupae and resulting teneral adults were recorded. 84

Colonization and Impact Studies

To study the colonization potential of the beetles on crape myrtle, we transplanted four nursery grown plants at each of five distances (0, 5, 15, 30, and 50 m) from the edge of a heavily beetle-infested stand of purple loosestrife on the southern side of Morgan

Lake, Baskett Slough NWR, western Oregon. We also planted a cohort of greenhouse- grown purple loosestrife 50 m to the east at the same distances from the purple loosestrife stand edge (Figure 3.4). Colonization ability may be influenced by plant cues or by intraspecific pheromones (Visser 1986, Grevstad and Herzig 1997), so the two plant species were planted in separate locations to assure no influence of purple loosestrife in attracting the beetles to L. indica. Our block design did not allow for randomization or interspersion of species, although each individual plant was randomly assigned a location within its block.

Each week, from initial transplanting to the end of the second generation of adult beetles, every plant was examined to determine: 1) percentage leaf damage, 2) number of adults, 3) presence of eggs, 4) presence of larvae, and 5) plant condition. Approximately one minute was spent searching for each beetle stage on each plant. Percentage of leaf damage was assessed by visually identifying an average leaf, comparing it to a pre­ calculated chart (Figure 3.5), and placing it into one of six damage classes (0 = 0%, 1 =

>0-5%, 2 = >5-25%, 3 = >25-50%, 4 = >50-75%, 5 = >75-95%, 6 = >95-100%). Plant condition indicated whether the plant was alive or dead. Plants were considered alive if green tissue was present or if there was evidence of new buds.

At the end of the growing season, after the adult beetles of the second generation moved off the plants and entered diapause, all of the purple loosestrife and crape myrtle 85

= 0 0 Om ED 0 0 0 4 m 0 0 0 0 8 m 0 0 CI = 12m El a) JCD 50 m .a) LL

El Bs Bs as ss m lis Es m

im Bs m

5m 15m 30 m 50 m

I---- Z O Lythrum salicaria Ell Lagerstroemia indica

Figure 3.4 Experimental design of the non-target field study 7£,

0% 6% 15% 26% 39% 46% 51% 64% 69% 75% 80% 87% 91%

Figure 3.5 Chart used to train observers to quantify leaf damage of Galerucella spp. on L. salicaria 87

plants were harvested to estimate biomass. The plants were carefully washed through a

mesh sieve (a. 2 mm2 cell size) in order to retain as much root material as possible. They

were then divided into root, stem, and leaves and placed into separate paper bags. The

root was separated from the stem by cutting immediately above the first laterally branching

root. The samples were then placed into a drying oven at 60°C. Samples were dried to

constant weight. Three of the largest rooted plants were selected and weighed every two

days. After the weight no longer declined, they were left in the dryer for two additional

days (for a total of 8 days) and then weighed on a Mettler electrobalance (PN1210) to the

nearest tenth of a gram. If plant material existed but weighed less than 0.05 g it was

recorded as 0.1 g.

Results

Lan'al Development

In the greenhouse trials, none of the 50 beetle larvae placed on crape myrtle

survived longer than eight days, and therefore none completed development past the

second instar (Figure 3.6a). Control larvae reared on purple loosestrife showed a 78%

survival rate to the third instar and 74% reached the adult stage (Figure 3.6b).

Insect Colonization

Adults from the first generation colonized all of the crape myrtle and purple

loosestrife plants at 30 m or less from the purple loosestrife stand. Of the plants at 50 m,

two (50%) of the purple loosestrife plants were colonized while one (25%) of the crape

myrtle plants were colonized. 88

a. non-target plant (L. indica) 25

20 G. pusilla 15 0 G. calmariensis

10 5

Instar 1 Instar 2 Instar 3 Pupae Adult

b. target plant (L. salicaria)

25

. : :" :

:;.;.;.; 20 ::::. ::::...... 15 ...... ::...... ! .::::. . . . 10 ...... i ...... , ...... - 5 .. :::: : ...... 1i .:.:.:.:...... : Instar 1 Instar 2 Instar 3 Pupae Adult Stage of Development

Figure 3.6 Larvae of G. pusilla and G. calmariensis were able to develop on the target plant L. salicaria but not on the non-target plant L. indica . 89

Eggs were present on all of the plants of both species colonized by the beetles.

During the first beetle generation six separate crape myrtle plants were found with larvae

(30%) while 14 purple loosestrife plants were found infested with larvae (70%).

To determine the probability that crape myrtle would be adjacent to populations of purple loosestrife, we overlaid a map of recorded purple loosestrife infestations

(Thompson et al. 1987) with the recommended cultivation zone of crape myrtle (Martin

1983). We found that the distribution of purple loosestrife and the recommended range of crape myrtle in the United Sates and Canada exhibited minimal overlap (Figure 3.2).

Beetle Impact

Damage (% leaf area destroyed) decreased with distance from the source of the colonists (the purple loosestrife stand) at Morgan Lake. At each distance, damage was lower on the non-target plant compared to damage on the target plant (Figure 3.7).

Plant biomass increased with distance for both plant species (purple loosestrifep <

0.001, crape myrtle p < 0.006). The change in biomass with a change in distance

(indicated by the slope of the regression) was greater for crape myrtle than for purple loosestrife (Figure 3.8). Relationships between biomass and distance were similar for total plant biomass and for biomass plotted separately by plant parts (roots, stems, leaves). A complete table for individual weights is listed in Table 3.2. o Lagerstroemia indica Lythrum salicaria

CD DID

GED o CEDE) 0

CD CD

0 10 20 30 40 50 60 Distance from edge of L. salicaria stand (m)

Figure 3.7 Percent leaf defoliation decreased with distance from the colonization source for both plants. Visual estimates of damage were assessed by comparing leaf damage to a pre-calculated chart and placed into one of seven damage classes. The classes were: 0 = 0%, 1 = >0-5%, 2 = >5-25%, 3 = >25-50%, 4= >50-75%, 5 = >75-95%, 6 = >95-100% (see Figure 3.4). Lagerstroemia indica 4 0 0

Lythrum salicaria

El

2 El

0 10 20 30 40 50 Distance from edge of L. salicaria stand (m)

Figure 3.8 Total dry weight of plants at five distances from a G. pusilla and G. calmariensis infested stand of L. salicaria. For L. salicaria: n= 20, r2= 0.76, Y = 0.045x + 1.77, p< 0.001. For L. indica: n = 20, r2= 0.35, Y = 0.010x+ 3.67,p= 0.006. 92

Table 3.2 Dry weight of Lythrum salicaria and Lagerstroemia indica (grams)

Distance Lythrum salicaria Lagerstroemia indica (m) root stem leaves total root stem leaves total 0 7.9 0.1 0.1 8.1 19.7 25.3 3.5 48.5 0 5.5 0.0 0.0 5.5 22.3 22.7 6.6 51.6 0 7.6 0.8 0.1 8.5 15.7 25.2 6.1 47.0 0 2.3 0.5 0.3 3.1 10.5 18.1 1.6 30.2 5 5.0 0.7 0.0 5.7 18.7 25.1 3.1 46.9 5 6.8 0.6 0.2 7.6 17.9 28.9 2.9 49.7 5 5.0 0.5 0.6 6.1 11.5 14.8 1.0 27.3 5 10.6 1.4 0.5 12.5 14.7 11.4 2.8 28.9 15 21.3 2.5 0.9 24.7 27.3 19.1 2.6 49.0 15 4.1 0.9 0.7 5.7 17.9 11.9 4.3 34.1 15 4.5 0.5 0.6 5.6 16.4 32.5 5.4 54.3 15 13.6 1.0 2.7 17.3 21.2 30.9 3.1 55.2 30 21.2 4.4 4.4 30.0 22.6 17.2 5.6 45.4 30 10.6 1.3 2.2 14.1 24.0 32.8 9.1 65.9 30 10.7 2.3 2.0 15.0 18.6 15.6 5.8 40.0 30 43.0 4.6 3.3 50.9 22.9 24.9 6.9 54.7 50 46.9 5.9 10.2 63.9 12.0 21.3 9.9 43.2 50 24.1 7.9 2.6 34.6 32.9 40.4 16.5 89.8 50 46.7 10.3 5.6 62.6 30.7 33.5 11.0 75.2 50 41.0 9.2 13.7 63.9 23.2 30.6 11.3 65.1 mean biomass mean biomass Distance root stem leaves total root stem leaves total 0 5.8 0.4 0.1 6.3 17.1 22.8 4.5 44.3 5 6.9 0.8 0.3 8.0 15.7 20.1 2.5 38.2 15 10.9 1.2 1.2 13.3 20.7 23.6 3.9 48.2 30 21.4 3.2 3.0 27.5 22.0 22.6 6.9 51.5 50 39.7 8.3 8.0 56.3 24.7 31.5 12.2 68.3 93

Discussion

Insect feeding and development on the non-target plant

Our findings confirm the results of the previous study (Blossey et al. 1994) showing that the beetles are unable to complete their life cycle on crape myrtle. The feeding by larvae in the greenhouse caused little damage to the plant and was minor, at the densities observed, compared to the damage inflicted by the adults in the field. Despite the presence of eggs on all the plants colonized for both species and the immediate severe defoliation of purple loosestrife, we found larvae on twice as many loosestrife plants

(70%) than on crape myrtle (30%). This result was expected, as the larval tests indicated that few larvae were able to reach the second instar on crape myrtle and as first instar larvae are very small 2 mm) they are difficult to detect in the field. Larvae on purple loosestrife completed subsequent instars and were larger and more easily distinguished.

Insect colonization of the non-target plant in the field

We found that, within 30 m of an "outbreak" beetle population, the colonization potential on crape myrtle was identical to that on purple loosestrife (100%). At 50 m, the colonization potential dropped to 25% for crape myrtle and to 50% for purple loosestrife.

This ability to locate new plants may be due to random movement away from the stand or by following chemical cues associated with the host plant. A study by Grevstad and

Herzig (1997) found that, when cohorts of G. calmariensis were released within 50 m of a host patch, the proportion of beetles reaching that patch was greater than that expected by random chance. This indicates that the beetle uses host plant cues in order to locate host 94 plant patches. The similarity we found in colonization ability between purple loosestrife and crape myrtle suggests that the beetles may use similar cues to locate crape myrtle plants. However, this similarity may also be due to random migration from the purple loosestrife stand (Chapter 2).

Influence of insect feeding on plant growth, survivorship, and reproduction

We were only able to evaluate the insects' effect on plant biomass due to a lack of mortality (no plants died) and flowering (only two purple loosestrife plants flowered, one at 50 m and one at 30 m) in this experiment. Biomass of both plant species declined with decreasing distance from the source of colonizing herbivores and increasing beetle damage

(Figure 3.8). This indicates that the introduced herbivorous beetles exert a negative impact on both plant species. The greater slope and lower y-intercept of the regression line for purple loosestrife suggest that beetles exert a greater impact on their intended target plant. This is probably due the greater defoliation of purple loosestrife that we found in our visual estimates of beetle damage (Figure 3.7). However, because we did not randomize between plant species we cannot make statistical comparisons between plant species and it is possible that the difference between the slopes of the regression lines was caused by effect of location.

Three weaknesses of our study were lack of randomization, lack of appropriate control, and not fully circumscribing colonization distances. Nonetheless, it seems clear that G. calmariensis and G. pusilla caused more damage to purple loosestrife plants compared with crape myrtle plants at similar distances from the colonization source and 95 that risk of severe non-target defoliation is limited to within 30 m of a population of the herbivores at superabundant beetle densities.

Conclusions

Our results suggest that the introduced biological agents, G. pusilla and G. calmariensis will have minimal impacts on populations of crape myrtle based on evidence of overlap in geographic distributions, suitability for insect development, and decrease in impact on non-target with distance from target. Larvae of G. pusilla and G. calmariensis could not complete development on the non-target plant, L. indica, confirming that the beetles are unable to develop self - sustaining isolated populations on the non-target host.

Damage to crape myrtle in the field by G. calmariensis and G. pusilla arises by adults colonizing from local stands of purple loosestrife. We found that the probability of beetles colonizing crape myrtle within 50 m from an infested purple loosestrife stand was high, but the attack rate measured as leaf damage decreased rapidly with distance. There is little overlap in the distributions of non-target and target hosts (Figure 3.2). Crape myrtle is mainly cultivated in the southern United States (zone 7) due to the hot, dry summers necessary for flowering (Martin 1983). Purple loosestrife is mainly distributed across the northern portion of the U.S. (Thompson et al. 1987). Therefore, there is low probability that these two plant species will exist near enough to each other to facilitate severe defoliation by the introduced beetles.

Host specificity tests typically measure whether a proposed biological control agent can complete its life cycle on a non-target plant. Some investigators argue that this rule of thumb no longer applies (Secord and Kareiva 1996), principally because it neglects 96 dispersal, evolution, and indirect effects. Here, we present a procedure for extending host specificity evaluations to include the effects of biological control agent dispersal and the resultant indirect effects to non-target organisms. By accurately quantifying risks to non- target organisms we will make more informed decisions regarding release of biological control organisms and thereby increase predictability and build public support. Although we agree that the field experiment is too labor intensive to be useful at the initial stages of host specificity testing, we believe it could be a useful technique in clarifying non-target impacts of biological agents on a local scale. 97

Chapter 4

Summary

In this thesis, we investigated protocols for monitoring classical weed biological control programs and for testing the host specificity of weed biological control agents.

Our objective was to identify measurements that would be useful to both biological control practitioners in adaptive management of purple loosestrife and to biological control researchers in furthering theory and predictability by promoting a general protocol that would facilitate cross program comparisons.

In the first section of the thesis we described the construction, implementation, and evaluation of a quantitative monitoring protocol proposed for general use in biological control programs. By reducing biological control programs to six basic stages, we identified a framework that could then be used to determine appropriate measurements for each stage. We then tested this protocol in the field and determined that it was feasible and yielded a set of quantitative measurements that accurately defined the succession of events that occurred at our site over the course of a successful biological program.

In the second part of the thesis, we examined the issue of host specificity testing in biological control programs. We created, implemented, and evaluated a protocol for quantifying risk to non-target organisms located near stands of the target organism. This adds a spatial component to the host specificity testing process. It has received little previous attention because the usual release criterion is whether the biological control agent is able to complete its life cycle on the non-target plant in nature. However, occasionally concerned organizations are not assuaged by the results of these tests and 98 require more specific information in order to determine risk to non-target plants that may be the source of their livelihood. The protocol presented here effectively measured the risk of the introduced biological control agents to the non-target plant, crape myrtle

Lagerstroemia indica at increasing distances from a colonization source. The results of this experiment were used in California in order to approve the release of the biological control agents within the state. The beetles were approved for release in May of 1998.

With the development of these evaluation techniques, we stress two concerns that we feel have not been dealt with adequately in the analysis of biological control programs.

First, in our evaluation of the progress of biological control programs we indicate a need to move from a focus mainly on suppression of pest abundance toward an emphasis concerned with managing ecological succession. Although the replacement of the target organism with more desirable organisms is the stated goal of biological control, it has rarely been investigated. Second, in our protocol for the evaluation of potential non- target effects, we stress the possibility of indirect effects damaging populations of non- target organisms. Neglecting to determine risk to non-target organisms has, in the past, seriously damaged the credibility of biological control programs. Public skepticism of biological control is retained to this day, to the point of favoring pesticides, which pose a much greater risk to environmental and human health. We hope that these evaluation techniques will facilitate research and discussion that leads to safer, more effective, and more efficient biological control programs. 99

Bibliography

Al lee, W. C. 1939. The social life of animals. Heinemann, London, England.

Anderson, M. G. 1995. Interactions Between Lythrum salicaria and Native Organisms: A Critical Review. Environmental Management 19:225-231.

Andow, D. A., P. M. Kareiva, S. A. Levin, and A. Okubo. 1990. Spread of invading organisms. Landscape Ecology 4:177-188.

Andow, D. A., P. M. Kareiva, S. A. Levin, and A. Okubo. 1993. Spread of invading organisms: patterns of spread. Pages 219-242 in K. C. Kim and B. A. McPheron, editors. Evolution of insect pests: patterns of variation. John Wiley, New York, New York, USA.

Andres, L. A. 1981. Conflicting interests and the biological control of weeds. Pages 11-20 in E. S. Delfosse, editors. Proceedings of the Fifth International Symposium on Biological Control of Weeds, 22-27 July 1980, Brisbane, Australia. CSIRO, Melbourne.

Bailey, L. H. 1951. Manual of cultivated plants most commonly grown in the continental United States and Canada. Macmillan, New York, New York USA.

Balogh, G. R., and T. A. Bookhout. 1989. Purple loosestrife (Lythrum salicaria) in Ohio's Lake Erie marshes. Ohio Journal of Science 89:62-64.

Bardner, R., and K. E. Fletcher. 1974. Insect infestations and their effects on the growth and yield of field crops: a review. Bulletin of Entomological Research 64:141-160.

Batra, S. W. T., D. Schroeder, P. E. Boldt, and W. Mendl. 1986. Insects associated with purple loosestrife (Lythrum salicaria L.) in Europe. Proceedings of the Entomological Society of Washington 88:748-759.

Beddington, J. R., C. A. Free, and J. H. Lawton. 1978. Characteristics of successful natural enemies in models of biological control of insect pests. Nature 273:513­ 519

Begon, M., J. L. Harper, and C. R. Townsend. 1996. Ecology: individuals, populations, and communities. Sinauer Associates, Sunderland, Massachusetts.

Beirne, B. P. 1985. Avoidable obstacles to colonization in classical biological control of insects. Canadian Journal of Zoology 63:743-747. 100

Blossey, B. 1992. Impact of Galerucella pusilla and G. calmariensis (Coleoptera: Chrysomelidae) on Field Populations of Purple Loosestrife (Lythrum salicaria). Pages 27-31 in E. S. Delfosse and R. R. Scott, editors. Proceedings of the Eighth International Symposium on Biological Control of Weeds, 1995, DSIR/CSIRO, Lincoln University, Canterbury, New Zealand.

. 1995a. Coexistence of two leaf-beetles in the same fundamental niche. Distribution, adult phenology, and oviposition. Oikos 74:225-234.

. 1995b. A comparison of various approaches for evaluating potential biological control agents using insects on Lythrum salicaria. Biological Control 5:113-122.

Blossey, B., D. Schoeder, S. D. Hight, and R. A. Malecki. 1994. Host specificity and environmental impact of two leaf beetles (Galerucella calmariensis and G. pusilla) for biological control of purple loosestrife (Lythrum salicaria). Weed Science 42:134-140.

Blossey, B., and M. Schat. 1997. Performance of Galerucella calmariensis (Coleoptera: Chrysomelidae) on different North American populations of purple loosestrife. Environmental Entomology 26:439-445.

Burdon, J. J., R. H. Groves, P. E. Kaye, and S. S. Speer. 1984. Competition in mixtures of susceptible and resistant genotypes of Chondrilla juncea differentially infected with rust. Oecologia (Berlin) 64:199-203.

Caughley, G., and J. H. Lawton. 1981. Plant-herbivore systems. Pages 132-166 in R. M. May, editors. Theoretical ecology: principles and applications. Blackwell Scientific Publications, Oxford, England.

Clement, S. L., and M. Cristofaro. 1994. Open-field tests in host-specificity determination of insects for biological control of weeds. Biocontrol Science and Technology 5:395-406.

Cliff, A. D., and J. K. Ord. 1981. Spatial Processes: Models and Applications. Pion Press, London.

Cofrancesco, A. F., Jr., R. M. Stewart, and D. R. Sanders, Jr. 1985. The impact of Neochetina eichhorniae (Coleoptera: Curculionidae) on waterhyacinth in Louisiana. Pages 525-535 in E. S. Delfosse, editors. Proceedings of the VI international symposium on biological control of weeds, 19-25 August 1984, Vancouver, Canada. Agriculture Canada, Ottawa. 101

Coombs, E. M., D. L. Isaacson, and R. B. Hawkes. 1992. The Status of Biological Control of Weeds in Oregon. Pages 463-71 in E. S. Delfossee and R. R. Scott, editors. Proceedings of the Eighth International Symposium on Biological Control of Weeds, 1995, 7, DSIR/CSlRO Melbourne, Lincoln University, Canterbury, New Zealand.

Coombs, E. M., H. Radtke, D. L. Isaacson, and S. P. Snyder. 1996. Economic and regional benefits from the biological control of tansy ragwort, Senecio jacobaea, in Oregon. Pages 489-494 in V. C. Moran and J. H. Hoffman, editors. Proceedings of the IX International Symposium on Biological Control of Weeds, 19-26 January 1996, Stellenbosch, South Africa, University of Cape Town, South Africa.

Crawley, M. J. 1983. Herbivory: The dynamics of animal-plant interactions. University of California Press, Berkeley, California, USA.

. 1986. The population biology of invaders. Philosophical Transactions of the Royal Society of London, Series B 314:711-731.

Cullen, J. M. 1978. Evaluating the success of the programme for the biological control of Chondrilla juncea L. Pages 117-121 in T. E. Freeman, editors. Proceedings of the Fourth International Symposium on Biological Control of Weeds, 30 August - 2 September 1976, Gainesville, Florida. University of Florida, Institute of Food and Agricultural Science, Gainesville, Florida, USA.

. 1990. Current problems in host specificity screening. Pages 27-36 in E. S. Delfosse, editors. Proceedings of the VII international symposium on biological control of weeds, 6-11 March 1988, Rome, Italy. Ministero dell'Agricoltura e delle Foreste, Rome/CSIRO, Melbourne.

Diehl, J., and P. B. McEvoy. 1990. Impact of the cinnabar moth (Tyria jacobaeae) on Senecio triangularis, a non-target native plant in Oregon. Pages 119-126 in E. S. Delfosse, editors. Proceedings of the VII International Symposium on Biological Control of Weeds, 6-11 March 1988, Rome, Italy. Ministero dell'Agricoltura e delle Foreste, Rome/CSIRO, Melbourne.

Dunmire, J. R., editor. 1979. Sunset: New Western Garden Book. Lane Publishing, Menlo Park, CA.

Goeden, R. D. 1983. Critique and revision of Harris' scoring system for selection of insect agents in biological control of weeds. Protection Ecology 5:287-301. 102

Grevstad, F. S. 1996. Establishment of weed control agents under the influences of demographic stochasticity, environmental variability and Allee effects. Pages 19-26 in V. C. Moran and J. H. Hoffman, editors. Proceedings of the IX International Symposium on Biological Control of Weeds, 19-26 January 1996, Stellenbosch, South Africa, University of Cape Town, South Africa.

. 1998. The colonization ecology of two loosestrife leaf beetles, Galerucella calmariensis and G. pusilla. Dissertation Thesis. Cornell University.

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.

Gruber, E., and A. Whytemare. 1997. The return of the native? Sidalcea hirtipes in coast Oregon. Pages 121-124 in T. N. Kaye, A. Liston, R. M. Love, D. L. Luoma, R. J. Meinke and M. V. Wilson, editors. Conservation and management of native plants and fungi. Native Plant Society of Oregon, Corvallis, Oregon, USA.

Guard, B. J. 1995. Wetland Plants of Oregon and Washington. Lone Pine Publishing, Vancouver, Canada.

Harley, K. L. S., and I. W. Forno. 1992. Biological control of weeds: a handbook for practitioners and students. Inkata Press, Melbourne, Australia.

Hengeveld, R. 1989. Dynamics of biological invasions. Chapman and Hall, London, England.

Hight, S. D., and J. J. Drea, Jr. 1991. Prospect for a classical biological controlproject against purple loosestrife (Lythrum salicaria L.). Natural Areas Journal 11:151­ 157.

Hitchcock, C. L., and A. Cronquist. 1973. Flora of the Pacific Northwest. University of Washington Press, Seattle, Washington.

Hopper, K. R., and R. T. Roush. 1993. Mete finding, dispersal, number released, and the success of biological control introductions. Ecological Entomology 18:321-331.

Hosmer, D. W., and S. Lemeshow. 1989. Applied logistic regression. Wiley, New York, New York, USA.

Howarth, F. G. 1991. Environmental impacts of classical biological control. Annual Review of Entomology 36:485-509. 103

Huffaker, C. B., and C. E. Kennett. 1959. A ten-year study of vegetational changes associated with biological control of Klamath Weed. Journal of Range Management 12:69-82.

Iwao, S., and A. Machida. 1963. A marking- and recapture analysis of the adult population of a phytophagous lady-beetle, Epilachna sparsa orientalis. Research on Population Ecology 5:107-116.

Julien, M. H., J. D. Kerr, and R. R. Chan. 1984. Biological control of weeds: an evaluation. Protection Ecology 7:3-25.

Karieva, P. M. 1981. Non-migratory movement and the distribution of herbivorous insects: experiments with plant specing and the application of diffusion models to mark-recapture data. Dissertation Thesis. Cornell University.

. 1982. Experimental and mathematical analyses of herbivore movement: quantifying the influence of plant spacing and quality on foraging discrimination. Ecological Monographs 52:261-282.

. 1983. Local movement in herbivorous insects: applying a passive diffusion model to mark-recapture field experiments. Oecologia 57:322-327.

Kok, L. T., and W. W. Surles. 1975. Successful biocontrol of musk thistle by an introduced weevil, Rhinocyllus conicus. Environmental Entomology 4:1025-1027.

Lande, R. 1988. Genetics and demography in biological conservation. Science 241:1455­ 1460.

Lawton, J. H., and K. C. Brown. 1986. The population and community ecology of invading insects. Philosophical Transactions of the Royal Society of London B 314:607-617.

Legendre, P., and M. J. Fortin. 1989. Spatial Pattern and Ecological Analysis. Vegetatio 80:107-138.

Liebhold, A. M., J. A. Halverson, and G. A. Elmes. 1992. Gypsy moth invasion in North America: a quantitative analysis. Journal of Biogeography 19:513-520.

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.

Malakoff, D. 1998. Restored Wetlands Flunk Real-World Test. Science 280:371-372.

Malecki, R. A., and T. J. Rawinski. 1983. Control of purple loosestrife (Lythrum salicaria). U.S. Fish and Wildlife Research Information Bulletin Report # 83. 104

Malecki, R. A., B. Blossey, S. D. Hight, D. Schroeder, L. T. Kok, and J. R. Coulson. 1993. Biological control of purple loosestrife. BioScience 43:680-686.

Malecki, R. A., and B. Blossey. 1994. Insect biological weed control: an important and underutilized management tool for maintaining native plant communities threatened by exotic plant introductions. Transactions of the 59th North American Wildlife and Natural Resources Conference 400-404.

Martin, E. C. 1983. Landscape Plants In Design. AVI Publishing Company, Inc.,

McClay, A. S. 1995. Beyond "before-and-after": experimental design and evaluation in classical weed biocontrol. Pages 213-219 in E. S. Delfosse and R. R. Scott, editors. Proceedings of the VIII international symposium on biological control of weeds, 2-7 February 1992, Lincoln University, Canterbury, New Zealand. DS111/CSIRO, Melbourne.

McEvoy, P. B. 1977. Adaotive significance of clumped dispersion in treehopper, Publilia concava (Homoptera: Membracidae). Dissertation Thesis. Cornell University.

. 1996. Host specificity and biological pest control. BioScience 46:401-405.

McEvoy, P. B., C. Cox, and E. Coombs. 1991. Successful biological control of ragwort, Senecio jacobaea, by introduced insects in Oregon. Ecological Applications 1:430­ 442.

Memmott, J., S. V. Fowler, H. M. Harman, and L. M. Hayes. 1996. How best to release a biological control agent. Pages 291-296 in V. C. Moran and J. H. Hoffman, editors. Proceedings of the IX International Symposium on Biological Control of Weeds, 19-26 January 1996, Stellenbosch, South Africa, University of Cape Town, South Africa.

Okubo, A. 1980. Diffusion and ecological problems: mathematical models. Springer- Verlag, New York, USA.

Pojar, J., and A. MacKinnon. 1994. Plants of the Pacific Northwest Coast. Lone Pine Publishing, Vancouver, Canada.

Ragsdale, D. 1996. Rearing Galerucella: The Minnesota experience. in Purple Loosestrife Biological Control Workshop, Minnesota Department of Natural Resources and U.S. Fish and Wildlife Service, Minneapolis, Minnesota, USA.

Rees, N. E., J. P.C. Quimby, G. L. Piper, E. M. Coombs, C. E. Turner, N. R. Spencer, and L. V. Knutson, editors. 1996. Biological Control of Weeds in the West. Western Society of Weed Science, Bozeman, Montana. 105

Secord, D., and P. Kareiva. 1996. Perils and pitfalls in the host specificity paradigm. Bioscience 46:448-453.

Shigesada, N., and K. Kawasaki. 1997. Biological invasions: theory and practice. Oxford University Press, New York, New York, USA.

Silvertown, J. W. 1993. Introduction to plant population biology. Blackwell Scientific Publications, Inc., Cambridge, Massachusetts, USA.

Simberloff, D., and P. Stiling. 1996. How risky is biological control? Ecology 77:1965­ 1974.

Skellam, J. G. 1951. Random dispersal in theoretical populations. Biometrika 38:196-218.

Smith, H. S., and P. De Bach. 1942. The measurement of the effect of entomophagous insects on population densities of their hosts. Journal of Economic Entomology 35:845-849.

Smith, R. W., and J. B. Whittaker. 1980. Factors affecting Gastryphysa viridula populations (Coleoptera: Chrysomelidae) in different habitats. Journal of Animal Ecology 49:537-548.

Sokal, R. R., and F. J. Rohlf. 1995. Biometry. W.H. Freeman, New York, New York, USA.

Southwood, T. R. E., and G. A. Norton. 1972. Economic aspects of pest management strategies and decisions. Pages 168-184 in P. W. Geir, editors. International Congress of Entomology, 1, Ecological Society of Australia, Canberra.

Strong, D. R. 1997. Fear No Weevil? Science 277:1058.

Thompson, D. Q. 1991. History of purple loosestrife (Lythrum salicaria L.) biological control efforts. Natural Areas Journal 11:148-150.

Thompson, D. Q., R. L. Stuckey, and E. B. Thompson. 1987. Spread, impact and control of purple loosestrife (Lythrum salicaria) in North American wetlands. U.S. Fish and Wildlife Research 2.

Turner, C. E. 1985. Conflicting interests and biological control of weeds. Pages 203-225 in E. S. Delfosse, editors. Proceedings of the Sixth International Symposium on Biological Control of Weeds, 19-25 August 1984, Vancouver, Canada. Agriculture Canada, Ottawa.

Visser, J. H. 1986. Host odor perception in phytophagous insects. Annual Review of Entomology 31:121-144. 106

Wentz, W. A. 1971. The Impact of Nutria (Myocaster coypus) on Marsh Vegetation in the Willamette Valley, Oregon. Masters Thesis. Oregon State University.

Wilcox, D. A. 1989. Migration and control of purple loosestrife (Lythrum salicaria L.) along highway corridors. Environmental Management 13:365-370.

Williamson, M. 1996. Biological invasions. Chapman and Hall, London, England.

Wootton, J. T. 1994. The nature and consequences of indirect effects in ecological communities. Annual Review of Ecology and Systematics 25:443-466.

Zelder, J. B. 1996. Ecological Issues in Wetland Mitigation: An Introduction to the Forum. Ecological Applications 6:33-37.

Zwolfer, H., and P. Harris. 1971. Host specificity determination of insects for biological control of weeds. Annual Review of Entomology 16:159-178.. Appendix 2.1 Study Site: Morgan Lake (Baskett Slough National Wildlife Refuge: Western Oregon)

Access Road Cultivated Field

--- Earthfill \ Dam

Morgan Lake

X Cultivated Field Beetle N, Release Site Open Water

Cultivated Field

100m Appendix 2.2 Leaf damage chart used to determine leaf damage estimates for Galerucella species on L. salicaria

5% 10% 15% 25% 50% 75% Appendix 2.3 Log-odds of beetle eggs with respect to distance from the initial beetle release site (X) over a four yearperiod.

1090900 1994 26 0661 1994 17 94ey 1994 21 AEA 1994 25 .141y 1994 it< 0 661 p <0001 3 2 2

0 0 4 .: . -3 .3 t

333 ro 20 93 102 00

16 May 1995 16 Juno 1995 19 July 1995 p. 0 004

2 2 . .

. . 1 ...... : a : : .3. . e - a.

90 120 150 20 a 10 12 100 ISO WO 250

11 41/ 1096 23 00 906 1996 25 April 1996 26 Mop 1996 20 Jun 1996

2. . . . . f 1 I . 01 0 0 a

-2 .2

ILO 293 303 93 al 0 313 a 09 0 Ica SO PE 200 93 93 a/ 20 a 10 3 193 0 1 .359 1091 21 999901 1997 2 Aro 1997 24 Juno 1997 3 3. 3 8 2. 2 2. 2 . 1.

0. U. .1. .1 .2 i -2 ­ 4. : .3 a 4. .3

513 IGO 150 203 250 350 0 50 103 150 391 250 350 .0 100 00 NO IT 200 20 100

Distance from initial beetle release site (m) Appendix 2.4 Log-odds of beetle eggs with respect to distance toward the lake (Y) over a four year period.

26 Ann, 1994 17 May 1994 21 June 1994 25 July 1994 4 19 August 1994 p =0.009 p < 0.001 3 3- 3­

2 2 1 2.

t 4 0 0 0

s -2 -2 . . -2 .9 -2 2 o. -3 3 3 3 8 O . 4 4 -.4 ­ 5 10 15 20 25 0 5 10 15 25 5 10 15 20 25 10 15 20 25 5

16 May 1995 16 June 1995 19 July 1995 p < 0.001 3 3 o 2 o.. 2

Or ­ 00 . 44*. 41,40 4, S. 4 8 4 2.0. .5. .3 I .3 6 4 10 15 25 5 20 25 0 5 IC 20

4- 25 April 1996 4. 28 May 1996 4.. 20 June 1996 11 July 1996 23 August 1996 p<0.001 p <0.001 7 3 . 0.4,0 3 2 2 2: . .8 2­ 5: 1. " ,, l 0 4 0 0 . 0 0 -I, -10 $ ; -2 .2 : .2 -2 -3 1. -3 -3

-4 -4 10 IS 33 25 ID 25 5 10 20 25

21 August 1997 3 3

2

.2

-3 -4­ 20 25 0

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet) Appendix 2.5 Log-odds of beetle larvae with respect to distance from the initial beetle release site (X) over a four year period.

26 Pool 1994 17May 1994 21 .one 1994 25 .11.19 1994 19 44.9.91 1994

= 0 008 3 1

2 2 2

1

a . . -2 .2 :: t .8. . - ro 10 20 ILO 130

4 ­ 16 Mey 1995 16 .A31 1995 19 My 1995 p.0011011 3

2­ 2

0 0

-3 6:.:.:. I 8 4 118. ID 130 313 10 30 10 100 m Im Im 202 293

25 /6911996 28 1309 1996 20..6. WM 11 July 1098 23 444201 1996 . T. : . I : 2 1

0 1 .1. i : ,2 .2 : I . . 8 1

012 22 43 110 193 93 NO 90 DO 220 DO DI al 10 102 JO NI 0 112 III 102 0

18 July 1997 214410.Y 1997 2 An. 1997 24 JUNI 1997 1. 3.

2 2

1 0 I 0 11 1 1 0 .1 -1 .1

-2 .2 -2 8 a. a. - 103 130 233 293 3130 20 41 10 10 1CO 100 193 201 230 MO 0 m

Distance from initial beetle release site (m) Appendix 2.6 Log-odds of beetle larvae with respect to distance toward the lake (Y) over a four year period.

. 26 April 1994 17 May 1994 21 June 1994 25 July 1994 19 August 1994

3

2 7 2.

1

I 0 0 4 g -I .41 S. -2 g *SS It*. 4 .9 -3 . -3. 4 .0 .4 -3 11 I I e :.

10 15 20 25 0 5 10 15 20 75 0 5 10 20 25 5 i0 15 20 25 10 15

16 May 1995 4 16 June 1995 19 July 1995

3 3

2 2 2 4.41 0 ...... 0 1 :** $ o 0.*: . $ 3 1 : *1 1 A

S 10 15 20 25 10 15 20 25 0 5 W IS 20 25

25 Awn 1996 4 28 May 1996 4- 20 June 1996 11 July 1996 4 23 August 1996 p<0 001 3 3 so 3

2 2 7- 2 0 O. $

0 a I 0 T. s -1 .9.9 .2 0 I -2 So 2­ -3 3­ 4 10 IS 20 25 0 10 16 20 25 10 IS 10 70 5 15 20 10 15 20

4 2 June 1997 24 June 1997 18 July 1997 4 ­ 21 August 1997

3 3 3 I..2 2. 2­

O r 1 I 11 a.

7 2 -2. -3 , -3­ 4- 4 4 4­ 5 10 15 70 25 10 IS 20 25 10 15 20 75 0 10 IS 20

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet) Appendix 2.7 Log-odds of adult beetles with respect to distance from the initial beetle release site (X) over a four year period.

26 9019 1991 17 May 1994 21 June 1994 15 JtAy 1994 19 Auwest 1994

2.

:

.3 4 10 33 33 10 10 150 203

16 1.4y 1995 16 Jule 1995

2 2

0 I 4

4 . . 1

30 93 93 133 150 L 40 03 113 103

25 200 1996 21 Ma 1996 20 Axle 1996 13 .44001 1994

2 2 2 1 ­ 0 0 .

-1

. -2 II 24 : . . 4 : 11 : '3 : 1 ' . . I : : 4 . .4

31 a 60 a 103 33 40 10 ID NE al a a a .13 1m Ms m as

2 Jae 1997 24 Jun 1917 IS July 1997

3. 3 2 2 2

1

0 0 d. . . - . : 4 .3 .3 1 31 a ID 63 KO 94 00 150 36 393 330 3 lm 150 200 250 103 150 30 250

Distance from initial beetle release site (m) Appendix 2.8 Log-odds of adult beetles with respect to distance toward the lake (Y) over a four year period.

28 April 1994 17 May 1994 21 June 1994 25 July 1994 19 August 1994

3 3 3 2

0 0 -1

-2 -2 -2 4 s :It .3 -3 o -34 -3 0:001 o** A 4 *. A 10 20 25 a i5 1U IS 20 25 5 10 15 20 75 10 15 25 5 10 15 20 25

16 May 1995 16 June 1995

3 3

2

0 O. so 4 0 : .8 't %ie. 4 4 4 5 w 20 25 5 10 16 20 25

- 25 April 1996 28 May 1996 4- 20 Arne 1996 11 July 1996 23 August 1996

3. 3 3- 3 3

2 - 2- 2 2

1 I - I 1­

0 a 1 1 0 I 0 0

1. -1, 1 -1 - : -2..8 00 -2 0.6. . -2 I 4. -2 4111: ,01 :,' -36: :"! $ -31 .:. :02.! -3 111 : 4- J- J1 J- .. 5 10 15 20 25 o II 15 20 25 5 10 15 20 25 0 5 10 15 20 ZS 5 10 15 75

2 Aine 1997 24 June 1997 4 18 Juiy 1997 21 August 1997 3 ­ 2 2 2 ­

1 ­

1 1 t I 0

1 8 SO -2 -2 -2 -3

-4

10 15 20 25 10 15 20 25 0 10 IS 25 25 6 10

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet) Appendix 2.9 Loge mean number of adult beetles with respect to distance from the initial beetle release site (X) over a four year period.

20 apnl 1991 17 May 1994 21 June 1994 25 flay 1994 Pusyst 1991 1

20 :0 10

16 May 1995 16 June 1995 19 July 1995 2 p.0006 p <0001

- 0 0.11,.­ 0 21 A 110 110 KO 132 10 31 0 10 10 1212 so 0 co iso 012

25 M49 1996 20 May 1996 20 Juno 1390 II .0.2 1996 23 August 1096

32 A O 00 0 33 0 O 1m a 0

2 Juno 1997 2 24 ..kno 1997 19 July 1997 21 hapusl 1997

0 IC IC 102 lO IC

Distance from initial beetle release site (m) Appendix 2.10 Loge mean number of adult beetles with respect to distance toward the lake (Y) over a four year period.

26 April 1994 17 May 1994 21 June 1994 25 July 1994 19 August 1994

_ _ - ­ - I 0 0 0 10 15 20 25 5 5 10 15 25 0 10 15 25 0 IC 25 0 15 25 25

19 July 1995 2­ 16 May 1995 2 16 June 1995 2

ot. , o $ _ - 1 03-0.0.0-11..-0.11-0-1-3-44-0--1-1 01,.. .8: 6 4 IS 70 25 0 5 10 IS 20 25 0 6 10 IS

2 2 5 April 1908 2- 28 May 1996 2- 20 June 1998 11 July 1996 23 August 1996

I . 0------4---t-, 5 to I6 20 25 0 0 10 15 20 25 0 6 10 16 25 0 6 10 15 20 25

1- 2 June 1997 2 24 June 1997 2 18 July 1997 21 August 1997

, 0 1 0 5 10 15 A 26 0 10 IS 25 75 0 10 15 52 25 5 10 20 ;6

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet) Appendix 2.11 Log-odds of leaf damage with respect to distance from the initial beetle release site (X) over a four year period.

28 Apra 1994 17 Wry 1994 21 Axle 1994 25 Jay 1994 19 August 1994 p9 0 001 IT p 0 018 p 0 012 p40001 p.0012 3 3

: 12 . : 0 0 0 . s . . : .

10 33 10 33 10 20 150

18 May 1995 16 Juno 1995 19 July 1995 p = 0018 9.0002 p < 0 001 3 ;:g* ***

0 o l I . .1 ; : 2 a

120 1513 20 0 50 10 ICO O 333 150 m 250

25 44281 1996 29 Hoy 1996 20 Juno 1996 11 tuly Wee 23 9t3340 lege - . : . : . . S 1 1

0 0

.1 -2­ -2 .2 .3­ 4 4

33 0 01 40 1410 0 21 40 o 513 KO 20 40 03 10 103 33 3 22 00 22 to us 850 22) 96 VD

24 Am. 1997 11July 1997 4 .. 23 ..433341 1997 3

2 ,

.2

.3

so 1113 150 333 293 51:1 ICC 193 m 150 327 103 150 m 250

Distance from initial beetle release site (m) Appendix 2.12 Log-odds of leaf damage with respect to distance toward the lake (Y) over a four year period.

26 April 1994 17 May 1994 1- 21 June 1994 25 July 1994 3- =0.023 2. 11 : 0 t I 0 0 2 . -4 10 15 20 25 10 I5 20 25 10 15 20 25 0 10

4_ 18 May 1995 16 June 1995 4 19 July 1995 p < 0.001 3 . : 2. 2 1 ... s ! 0 0* , 0 f . . ' , 3AI 2, -7 4 . I .' . -3 A

10 15 20 25 5 10 16 20 25 0 5 10 15 20 25

25 Apre 1998 28 May 1996 20 June 1998 11 July 1998 4- 23 August 1996 : lb a g 2 2 8 14..: 1 1 1 p

0. 0 1 0 0 1 4

-2 I. -3 4 -4 6 0 15 20 36 0 10 15 20 25 10 15 70 25 5 10 15 20 25 5 10 IS 20

4 2 June 1997 . 4- 24 June 1997 4 18 July 1997 21 August 1997 3 : 4. 3.. S 21 t 2. .: 1 - 1 1

0 1 1 0 1 4 5 1 -1. -1, -2 4 4_ -4 4 10 15 20 25 10 16 25 5 5 10 15 20 25

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet) Appendix 2.13 Loge mean leaf damage estimates with respect to distance from the initial beetle release site (X) over a four year period.

26 Apnl 1994 1700.9.1991 21 June 1994 25 J., 1994 19 Ng151 1994

I. t : :

20 to

16 June 1995 19 Jun 1995

. : ;

0 o a 40 19) D m m D 100 100

25 asori1 1990 21 May 1996 20.4013 1996 11 July 1995 23 August 1996

a I. : I : . 3. 3 3 : 2

ID 10) 0 M 4p 10 K. KO 10 m KO 20 40 50 101

2 June 1997 24 June 1997 1 AA 1997 21 August 1997 . 4.

3

2

191 am 191 271 250 300 0 ea 103 NO 291

Distance from initial beetle release site (m) Appendix 2.14 Loge mean leaf damage estimates with respect to distance toward the lake (Y) over a four year period.

17 May 1994 21 June 1994 26 April 1994 25 July 1994 5 19 August 1994 5

4

2 7 0 1 41.8 g: a . .:1 .044 0 0 . 10 15 25 0 10 20 25 0 IS 15 70 25 15 25 - 10 15 16 May 1995 16 June 1995 s, 19 July 1995

3

7 .1 1 0.:11. *0 : . Ot:: 00 0 0 5 0 15 25 0 10 IS 20 0

25 A9111 1996 28May 1996 20 June 1996 11 July 1996 23 August 1996 6

4 4 _ 4 8. o: 4 . It 44. e 4141 3 4141.:0 :": 2 2

0 0 05 0 20 0 10 15 211 75 0 70 25 0 15 20 25 0

2 June 1997 24 June 1997 18 July 1997 21 August 1997 :es 6

4

s. 3 3

2 2 2

1. 1

0 10 70 76 10 15 25 0 10

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet) Appendix 2.15 Log-odds of meristem damage with respect to distance from the initial beetle release site (X) over a four year period.

16 Pool 1951 17 May 1994 21 Ane 1994 25 My 1991 19 441310 1994 6<0001 3* 3 pt 0 011 2* 2

i I! 2. : : I . : i 11 10 33 1a n 103

161409 1995 16 An* 1995 19.09 1995 p 4 0 001 3.

2. 2

1 0 1 : .3 : ell 4 S ::

33 10 03 130 93 3) 43 10 10 KO 103 203 250

23 August 1996 2541prd 1996 28 1105, 1996 20 Juno 1996 11 July 1996 0 002 3

2

e. : : .2 .2 : 8 .. : 4

33 40 03 160 03 KO ZO 3613 n 40 00 10 100 0 Zs 40 00 10 10 33 43 10 10 KO

2.092 1907 24 Jun 1997 .146 1997 21 August 1997

3 p0.015 . : 2 * I

1

0 0

.3

O t0 no 300 250 303 0 O 1m 160 303 250 3:13 33 0 10 0 KO 60 100 103 >m 3613 300

Distance from initial beetle release site (m) Appendix 2.16 Log-odds of apical meristem damage with respect to distance toward the lake (Y) over a four year period.

26 April 1994 17 May 1994 21 June 1994 25 July 1994 19 August 1994

3 3 3

2 2

1

0 0 I 0

41 ,... 46$ -1 .2 -2 :SI 11 ., .3 : -3 . -3 : . -4 10 15 20 25 0 5 10 15 70 25 10 IS 20 75 5 N 70 75 5 10 15 20

... 16 May 1995 4- 16 June 1995 ,, 19 July 1995

3­ 3 3 2- 2. 2: : 1- 1.

0 4 I 5

-1- -1

4 -2 : 111 .2-46. 1 : -3 OA' -3- -4- 4 5 10 15 20 25 10 15 20 25 20 25

25 April 1998 4- 28 May 1998 4 20 June 1998 11 July 1998 ., 23 August 1998

3- 3 3 : 9. 2 ._ 2. 2- 2 : 2. 1...8 14 I- I 0 o I I I 1 0 I I 1 I e 4 4 I '0 4 1 1 -I -1.6 -I -14 -2 111 1 .21 -2 -7 -2 : -3 -34 : -3... ': e $ -3 -3 .4- -4 .1 6 10 15 70 25 0 6 10 IS 20 25 5 10 15 20 25 10 IS 20 25 10 15 20 25

2 Ane 1997 4- 24 June 1997 18 July 1997 21 August 1997 p = 0.018 3 3

2 2 1 : .

0 a 0 0 -1

-2 -2 -2

.3 .3 -3 -4 -4- -4 10 20 25 10 16 20 25 10 15 20 76 10 15 20 25

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet) Appendix 2.17 Log-odds of damage from other herbivores with respect to distance from the initial beetle release site (X) over a four year period.

26 Apnl 1994 17 May 1954 21 Atm 1994 25 July 1994 19 409391 1994

2 2 a

0 0 . : t .

20 to 33 Itto 192

16 May 1995 16 .490 1995 19 July 1995

2 ,. : t 1 1 0

.1 e se 4 I 30 SO 93 130 20 40 99 99 to3 iso XO 293

25 Apnl 11198 28 May 1998 20 Jule 1996

2

J. J. 1 4 a t .1 4 J. 41 33 99 10 93 KO 93 10 103 33 40 00 10 KO 0 213 43 0 110 103

2 Juno 1997 2 JUn 1997 18 Jay 1997 21 Augml 1997

1 0 0 0

.1.. n J i 4. . * * : a. .1... .3. .3 4.­ J. 4 0 33 10 10 99 103 93 WO MO MO MO 302 92 101 110 MO 88 3413 0 SO or 193 113 293 M

Distance from initial beetle release site (m) Appendix 2.18 Log-odds of damage from other herbivores with respect to distance toward the lake (Y) over a four year period.

26 April 1994 17 May 1994 21 June 1994 25 July 1994 19 August 1994 p < 0.001 3. 3 3

2 2

1

0 1 0. I 0 4 . -1 :s. j 4 -2 10 -2 -2 :11 : , .. 3_ .3 4 4 ­ -1 M IS 25 0 M IS 20 25 5 10 M 25 0 2 15 25 5 $0 15 20 25

4. 16 May 1995 16 June 1995 19 July 1995 p < 0.001 3. . 3 3­

2 2 2.. a 1

0 1 1 : 1 I. 2_I;, :Ole : 0 .ar.*. In a : 1.1 4 I 41" 10 IS 20 25 10 IS 20 25 0 5 10 15 20 25

4 25 April 1996 - 28 May 1996 20 June 1998 4. 11 July 1996 23 August 1996

3. 3. 3- 3

2 2 1. 2

0 0 0 006 2 4 ..1-**1 -2 *0 41. 1 .2 .. , $ . : .3 4 oe :3. - -3 *.: ., a 4 4 -4 16 20 25 10 15 25 5 10 15 20 25 6 .111154025 0 5 10 IS 20 25 5 10

4- 2 June 1997 ._ 24 June 1997 18 July 1997 21 August 1997

2

1 0 0 .1 41 .* ,t 4 1 a .2 .3 31 .

10 IS 20 25 5 10 15 20 25 10 15 25 S 10 15 20

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet) Appendix 2.19 Log-odds of purple loosestrife flowering success with respect to distance from the initial beetle release site (X) over a four year period.

26 443111 1994 17 1.4ay 1994 21 Jun. 1994 25 July 1994 19..4.9 1994

3

I.

0 0 . a. . 3 . . 7 :

10 ID 10 20 10 103

16 May 1995 16 Jule 1995 19 Ail/ 1995

3 3 C 1 0 0 t : . : 2 . : .

92 10 133 33 40 10 10 103 103 .0 333 203

25 Apnl 1998 29 May 1996 20 Jule 1998 11 349 1998 23 469..31 1996

3 3 3 3 21 2 13. 2 1 2

0 1 : a o 0. 0 .1 1 41 -2 : : a 8 : : $ , I ! 4 : 0 .3 : .3 4 i Ji 4 4 33 10 113 10 1133 0 30 AO .3 10 KO 93 100 I93 333 293 3E0 22 40 93 03 103 0 a, a, in 03 193

18 Ay 1997 21 August 1997 2 Jule 1997 24 Aro 1997 A 4

2 2 2

1

4 4 0 1 .1 . .1

S I . . : . : 4 a a a a J_ 4 J. 93 103 193 230 NO 0 30 103 190 331 230 300 20 AO 91 ID KO 93 90 KO 99 290 320 39

Distance from initial beetle release site (m) Appendix 2.20 Log-odds of purple loosestrife flowering success with respect to distance toward the lake (Y) over a four year period.

26 April 1994 17 May 1994 21 June 1994 25 July 1994 19 August 1994

3

7. 2 2

1. _ I s. 0 , jot 46% 1 .2 -2 -2 0 -2 :t 0 I -3 S ; 4 61. 4 4t 4 5 10 15 70 26 0 5 10 20 25 5 15 20 25 10 1s 20 25 5 la 15 TO

18 May 1995 18 June 1995 19 July 1995

3­ 3IIII p = 0.010 2­

1 ­ tutu 0 0 ss II 4. e 1601 4i 4.111 :.26 ' s 4* ** Is . .3 .3

la 15 20 25 0 10 15 20 25 io is 20 25

25 April 1998 28 May 1998 20 June 1996 ._ 11 July 1998 23 August 1998 p 0.004 3 3. 3 3 2. 2 2

1 1.

0 1 1 - . I : .I -2 4, 0 .2 .:0 .0 .00 0.0 .3 4-41616: $ -31 -3 . S 20 25 0 6 10 15 20 25 5 10 A A 5 MAMA SAMA

2 June 1997 21 August 1997 3­ 3­ 3

0 0 4. 4­

6 10 5

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet) Appendix 2.21 Loge of stem length with respect to distance from the initial beetle release site (X) over a four year period.

26 906111895 17 May 1999. 21 June 1994 25 Juiy 1994 19 &qua 1994 6 6 1:1 . I 3 2 2

0 0

0 tO 20

16 easy 1995 16 June 1995 6 19 July 1996

- 6 I ; 84:!i".:". :, ; 0. 0 : . S. . . 3. : 2 2,

1

0 10 X) SO 70 Po 90 130 19 6 D 1n

Party 1996 28 My 1996 20.0M 1996 11 Aly 1996 73 e9T9t 1996 6­ 6 6 , 6J 6i $ . 6.. . . 4 4. : i 3.­ II 3

2, 2 2 2 t .

60 so so 40 loo 0 P . 0 93 ; To 260

2 June 1997 24 Jure 1997 1.09019071997 6 6 6 21 August 1997 6 6 6_ 6

4 I j 4 4 S. S. S 2 2.

1 1 1. 1

0 1 1 1 6 1 1 0 1 1 0 0 L 40 40 ea TO 0 O TO 90 20 293 96 0 io 1m 103 203 260 306 60 1m

Distance from initial beetle release site (m) Appendix 2.22 Loge of stem length with respect to distance toward the lake (Y) over a four year period.

26 April 1994 17 May1994 21 June 1994 a 25 July 1994 a 19 August 1994

5 5­ 5 0.0 ; .0 .111. 66 3 . 3 2

0 0 0 0 0 10 20 75 0 5 10 15 20 25 0 5 10 15 25 10 25 10 15 20 75

6, 16A6.1995 19 July 1995 16 May 1995 p.0.001 p = 0.004 5 is* s 4* 1.: :.so:. 06 ; "0' ' 3

2 2

1

0 0 I 0 6 1 1.5 26 0 10 70 25 6 10 IS 20 25

25 April 1996 28 May 1998 20 June 1996 I 11 July 1996 23 August 1996 v40.001 p < 0.001 p<0.001 p < 0.001 p < 0.001 6 5 6 o. :l : 0. . 4 I :8 4..s. :1.1 . 4 3 $: 8 : 3

3 2 7

1

0 0 0

0 10 15 75 0 10 0 1 10 15 20 75 0 5 IS 20 15 0 10 15

2 June 1997 e_ 24 June 1997 18 July 1997 21 August 1997 p 0.025 p =0.015 5­ 5 . : 4 .° $ .:$ 4 3 . 38: 3 2 2

1 1

0. I 0 15 25 0 10 15 70 25 10 15 20 2s 10 15 20

Distance toward lake (m) (small values indicate high and dry, large values indicate low and wet)