Danaus Plexippus) Movement Ecology to Inform Conservation Strategies

Iowa State University Capstones, Theses and Graduate Theses and Dissertations Dissertations

2021

Investigating monarch butterfly (Danaus plexippus) movement ecology to inform conservation strategies

Kelsey Elizabeth Fisher Iowa State University

Follow this and additional works at: https://lib.dr.iastate.edu/etd

Recommended Citation Fisher, Kelsey Elizabeth, "Investigating monarch butterfly (Danaus plexippus) movement ecology to inform conservation strategies" (2021). Graduate Theses and Dissertations. 18493. https://lib.dr.iastate.edu/etd/18493

This Dissertation is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State University Digital Repository. It has been accepted for inclusion in Graduate Theses and Dissertations by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected]. Investigating monarch butterfly (Danaus plexippus) movement ecology to inform conservation strategies

Kelsey Elizabeth Fisher

A dissertation submitted to the graduate faculty

in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

Major: Entomology

Program of Study Committee: Steven P. Bradbury, Major Professor James Adelman Brad Coates Richard Hellmich Amy Toth

The student author, whose presentation of the scholarship herein was approved by the program of study committee, is solely responsible for the content of this dissertation. The Graduate College will ensure this dissertation is globally accessible and will not permit alterations after a degree is conferred.

Iowa State University

Ames, Iowa

2021

DEDICATION

This dissertation is dedicated to those near and far who supported me during my graduate work: my husband, my family, my friends, and associated pets.

iii

TABLE OF CONTENTS

Page

LIST OF FIGURES ...... vii

LIST OF TABLES ...... xi

ACKNOWLEDGMENTS...... xiii

ABSTRACT...... xv

CHAPTER 1. GENERAL INTRODUCTION...... 1 Outline and Objectives of the Dissertation Chapters...... 5 Chapter 2: Estimates of common milkweed (Asclepias syriaca) utilization by monarch larvae (Danaus plexippus) and the significance of larval movement ...... 5 Chapters 3, 4, 5, and 6: Analyzing monarch movement ecology at a landscape-scale with aid from handheld and automated VHF radio telemetry systems ...... 6 Chapter 7: Prediction of mitochondrial genome-wide variation through sequencing of mitochondrion-enriched extracts ...... 11 Chapter 8: General Conclusion...... 14 Contributions of this Dissertation ...... 14 References ...... 14

CHAPTER 2. ESTIMATES OF COMMON MILKWEED (Asclepias syriaca) UTILIZATION BY MONARCH LARVAE (Danaus plexippus) AND THE SIGNIFICANCE OF LARVAL MOVEMENT...... 20 Abstract ...... 20 Introduction...... 21 Methods ...... 24 Plants ...... 24 Insects...... 25 Experimental Design ...... 25 Statistical Analysis ...... 27 Results ...... 29 Survival and Developmental Endpoints ...... 29 Natal Ramet Abandonment ...... 29 Larval Movement Behavior ...... 30 Plant Utilization ...... 30 Discussion...... 32 Data Accessibility...... 39 Acknowledgements ...... 39 References ...... 39 Figures and Tables...... 44 Appendix. Impact of Intraspecific Competition on Larval Monarch (Danaus plexippus) Movement Behavior and Fitness ...... 48 Introduction...... 48 iv

Methods ...... 48 Preliminary Results and Discussion...... 50

CHAPTER 3. EMPLOYING VERY HIGH FREQUENCY (VHF) RADIO TELEMETRY TO RECREATE MONARCH BUTTERFLY (Danaus plexippus) FLIGHT PATHS...... 52 Abstract ...... 52 Introduction...... 53 Methods ...... 56 Field Sites ...... 56 Insects...... 57 Transmitter Attachment ...... 57 Effects of Transmitter Attachment...... 57 Tracking Female Monarch Butterflies with VHF Radio Telemetry ...... 58 Space-Use Analysis ...... 59 Results ...... 61 Effects of Transmitter Attachment...... 61 Space-Use Analysis ...... 61 Discussion...... 63 Data Accessibility...... 69 Acknowledgements ...... 69 References ...... 69 Figures and Tables...... 74 Appendix. Supplemental Information ...... 78

CHAPTER 4. LOCATING LARGE INSECTS USING AUTOMATED VHF RADIO TELEMETRY WITH A MULTI-ANTENNAE ARRAY ...... 79 Abstract ...... 79 Introduction...... 80 Materials and Methods...... 85 Overview ...... 85 System Design...... 86 Conceptual Model to Estimate Transmitter Location...... 89 Determining fdistance (d)...... 89 Determining fangle () ...... 92 Random Effects of Receiving Stations and Antennae ...... 93 Estimating Location...... 94 Estimating location of stationary transmitters at random locations ...... 96 Evaluating the Method ...... 97 Results ...... 99 Estimating locations of a moving transmitter ...... 99 Estimating location of a mobile butterfly ...... 100 Discussion...... 102 Data Accessibility...... 107 Acknowledgements ...... 107 References ...... 108 Appendix. Supplemental Information ...... 113 v

CHAPTER 5. ESTIMATING PERCEPTUAL RANGE OF FEMALE MONARCH BUTTERFLIES (Danaus plexippus) TO VEGETATIVE COMMON MILKWEED (Asclepias syriaca) AND NECTAR RESOURCES ...... 119 Abstract ...... 119 Introduction...... 120 Methods ...... 122 Insects...... 122 Perceptual Range of Experimental Monarchs to Potted Resources ...... 123 Perceptual Range of Non-experimental Monarchs to Potted Resources ...... 124 Statistical Analyses...... 124 Results ...... 126 Perceptual Range of Experimental Monarchs to Potted Resources ...... 126 Perceptual Range of Non-experimental Monarchs to Potted Resources ...... 128 Discussion...... 128 Data Accessibility...... 134 Acknowledgements ...... 134 References ...... 135 Figures and Tables...... 137

CHAPTER 6. INFLUENCE OF RESOURCE ABUNDANCE AND HABITAT CONFIGURATION ON FEMALE MONARCH BUTTERFLY (Danaus plexippus) MOVEMENT AND HABITAT UTILIZATION PATTERNS AT A LANDSCAPE-SCALE. 143 Abstract ...... 143 Introduction...... 144 Methods ...... 147 Field Sites ...... 147 Monarchs ...... 148 Tracking Female Monarch Butterflies ...... 149 Statistical Analyses...... 150 Results ...... 153 Discussion...... 160 Movement among Habitat Classes...... 161 Orientation of Movement ...... 162 Step Lengths and Directionality ...... 163 Space Use ...... 164 Data Accessibility...... 165 Acknowledgements ...... 166 References ...... 166 Figures and Tables...... 170 Appendix. Supplemental Information: Automated Radio Telemetry System (ARTS) ...... 180 System and Field Design...... 180 Monarchs ...... 182 Looking Forward ...... 182

CHAPTER 7. PREDICTION OF MITOCHONDRIAL GENOME-WIDE VARIATION THROUGH SEQUENCING OF MITOCHONDRION ENRICHED EXTRACTS ...... 184 Abstract ...... 184 Introduction...... 185 Results ...... 187 Optimization of Differential Centrifugation and DNA Extraction ...... 187 Quantification of Mitochondrion Enrichment at Optimized Parameters...... 187 De novo Mitochondrial Genome Assembly, Annotation and Variant prediction ...... 188 Comparison between enriched and unenriched reads for de novo mitochondrial genome assembly ...... 191 Discussion...... 192 Methods ...... 199 Specimens and sampling ...... 199 Optimization of Differential Centrifugation and DNA Extraction ...... 199 Quantification of Mitochondrion Enrichment at Optimized Parameters...... 201 De novo Mitochondrial Genome Assembly, Annotation, and Variant prediction ...... 204 Comparison between enriched and unenriched reads for de novo mitochondrial genome assembly ...... 206 Data Accessibility...... 207 Acknowledgements ...... 207 References ...... 208 Figures and Tables...... 214

CHAPTER 8. GENERAL CONCLUSION...... 218 References ...... 226 vii

LIST OF FIGURES

Page

CHAPTER 2

Figure 1. Experimental cages with common milkweed ramets...... 44

Figure 2. Relative time monarch larvae were observed on portions of milkweed ramets...... 45

Figure 3. Average frequency of instar specific ramet abandonments...... 45

Figure 4. Larval milkweed consumption at natal ramet abandonment...... 46

Figure 5. Proportion of larval observations on milkweed ramet ...... 46

Figure 6. Representative common milkweed ramet after a 4th instar abandonment...... 47

CHAPTER 3

Figure 1. Monarch butterflies with a 220 mg LB-2X radio telemetry transmitter (a) or sham-transmitter (b; 300 mg watch battery) attached to the ventral side of the abdomen with superglue ...... 74

Figure 2. Field site in Boone, IA, U.S.A., where flight paths of monarch butterflies were tracked with VHF radio telemetry...... 74

Figure 3. Proportion of time sham-tagged (300 mg watch battery) and sham-untagged monarchs were observed engaging in various behavior classifications...... 75

Figure 4. Example recreated flight path of one monarch butterfly...... 75

Figure 5. Characterization of 13 female monarch butterfly flight steps (distances between two estimated locations) tracked with radio telemetry in a restored prairie...... 76

Figure 6. Frequency of turn angles (angle created from 3 consecutive, estimated locations) for 13 female monarch butterflies tracking with radio telemetry in a restored prairie...... 76 viii

Figure 7. Occurrence model created using the ctmm package in RStudio for nine radio- tagged monarch butterflies...... 77

Figure 8. Individual recreated straight-line flight paths for all 13 VHF radio-tagged monarch butterfly females tracked in July 2016 at the field site in Boone, IA, U.S.A...... 78

CHAPTER 4

Figure 1. Each receiving station in our automated telemetry system consisted of four 3- element Yagi antennae attached to the top of a 7m pole, supported by a guy wire system that was augured 0.5m into the ground (a). Yagi antennae were placed in vertical polarization (b) separated by 60 around the pole's axis (c)...... 87

Figure 2. We found a clear relationship between received power and the distance between transmitter and antenna (left panel), and a more variable relationship between received power and the angle between antenna and transmitter (right panel)...... 91

Figure 3. The system accurately estimated locations of six of the ten test locations of stationary transmitters on a sod field at Blue Grass Enterprises...... 97

Figure 4. The system accurately captured the path of an investigator walking from the southeast to northwest receiving station at a pace of 50m/min at a harvested maize field...... 101

Figure 6. Power was recorded at 13 locations along a 100 m circle surrounding one receiving station at three gain settings: 50, 65, and 75...... 114

Figure 7. Variability in power for three configurations of transmitting antenna ...... 115

Figure 8. Locations used to calibrate the model for power as a function of distance and angle from the transmitter to the receiving antenna...... 115

Figure 9. Theoretical power vs angle from a 3 element Yagi antenna...... 116

Figure 10. Using thin-plate regression smoothing provided an effective alternative to calculating the median power recorded over time...... 116

Figure 11. Received power as a function of distance for a transmitter worn by an investigator (in hat) and one mounted on a butterfly in a plant...... 117 ix

Figure 12. Received power over time at four antennae when a monarch was allowed to move freely through a field...... 118

CHAPTER 5

Figure 1. Example experimental set-up at a 32-ha sod field in central Iowa in which radio- tagged, sham tagged, or untagged experimental monarchs were released on a vegetative stage Monarda fistulosa plant 5 m downwind from two milk crates containing vegetative stage Asclepias syriaca and one milk crate containing blooming Liatris sp...... 137

Figure 3. Histograms displaying the similarity of effective and total distance traveled for 148 experimental monarchs released on vegetative stage Monarda fistulosa 5, 25, 50, or 75 m from potted vegetative stage Asclepias syriaca and blooming Liatris sp in sod fields of 4-32 ha...... 139

Figure 5. Histograms with 25 m bins of all 418 flight steps (a), 372 flight steps within the sod field (b), and 46 flight steps associated with exiting the sod field (c) taken by 155 monarchs released on vegetative stage Monarda fistulosa 5, 25, 50, or 75 m from potted vegetative stage Asclepias syriaca and blooming Liatris sp in sod fields of 4-32 ha...... 140

Figure 6. Histogram of opportunistically observed direct flights upwind to potted Asclepias syriaca and Liatris sp. by 49 non-experimental monarchs in 4-32 ha sod fields in 2017 and 2018...... 141

CHAPTER 6

Figure 2. The mean number of times female monarchs (n=88) left release habitat and crossed a habitat boundary during the observation period of 75 to 7,500 s...... 171 x

Figure 3. Histograms of step lengths for 114 female monarchs released in all habitat classes; 0-100 m (a) and 100-2000 m (b)...... 172

Figure 4. The time elapsed prior to initiating a large step (> 50 m) by habitat class and habitat configuration...... 173

Figure 8. Summaries of raster cells with a non-zero probability of occurrence by habitat class...... 177

CHAPTER 7

Figure 2. Estimates of mitochondrion enrichment of Ostrinia nubilalis from homogenized thorax tissue through differential fractionation prior to DNA extraction...... 215

Figure 3. In-frame and frameshift mutations predicted among assembled Ostrinia mitochondrial genomes...... 216

Figure 4. Alignment of 60 C-terminal residues from representative NADH Dehydrogenase subunit 4 (ND4) protein sequence accessions...... 217 xi

LIST OF TABLES

Page

CHAPTER 2

Table 1. Description of milkweed ramet condition ranking...... 47

Table 2. Number (and percent) of individuals that survived to developmental stages by year, number of ramets, and pooled across years and number of ramets ...... 47

CHAPTER 3

Table 1. Number of female and male wild-caught individuals observed with and without sham-transmitters (300 mg watch batteries) in 2016, 2017, and 2018 to determine if there was an effect of weight attachment on behavior or flight ability...... 77

Table 2. Percent of time observed monarchs occurred in various landcover types, as estimates with ctmm occurrence models ...... 77

CHAPTER 4

Table 1. Receiver settings...... 114

CHAPTER 5

Table 1. The number of individuals that flew or were unresponsive (UNR; no movement within 10 minutes of release) based on attachment treatment (χ2=0.067975, df=2, p=0.9666)...... 141

Table 2. The number of individuals that flew or were unresponsive (UNR; no movement within 10 minutes of release) based on the number of days held in captivity prior to an experimental trial (χ2=3.6211, df=4, p=0.4597)...... 141

Table 3. The number of individuals that flew or were unresponsive (UNR; no movement within 10 minutes of release) based on the release distance to the potted resources (χ2=7.1975, df=3, p=0.06586)...... 142

Table 4. The unimodal movement patterns displayed at 5, 25, and 50 m releases were not oriented toward the resources...... 142

Table 5. The unimodal movement patterns displayed at 5, 25, and 50 m releases were in the same direction as the wind...... 142

xii

CHAPTER 6

Table 1. Terms and definitions for field sites, release sites, and habitat types. Terms are described further in subsequent tables...... 178

Table 2. Categorization of milkweed ramet and blooming species abundance estimated within 100 m transect surveys to classify resource plant density as zero, low, moderate, or high...... 178

Table 3. Combination of milkweed and bloom classifications (see Table 2) to identify habitat type as “crop field,” “resource-poor,” or “resource-rich.”...... 178

Table 4. The number of radio-tagged and untagged female monarchs released in the mosaic and roadside corridor focal areas that flew or were unresponsive (UNR; no movement within 15 minutes of release)...... 179

Table 5. The number of total female monarchs released in mixed land use and roadside corridor habitat configurations that flew or were unresponsive (UNR; no movement within 15 minutes of release) based on days captive...... 179

Table 6. The number of female monarchs released in the mosaic configuration and the roadside corridor configuration that flew or were unresponsive (UNR; no movement within 15 minutes of release)...... 179

Table 7. The number of large steps (≥ 50 m) by step length categories and habitat of step initiation combined across all habitat configurations...... 180

CHAPTER 7

Table 1. Ostrinia nubilalis mitochondrial genome sequence assembly accessions ...... 217

xiii

ACKNOWLEDGMENTS

This work was supported by the Agriculture and Food Research Initiative Pollinator

Health Program (Grant 2018-67013-27541) from the USDA National Institute of Food and

Agriculture, Garden Club of America Board of Associates Centennial Pollinator Fellowship received in 2017, Holohil Grant Program received in 2018, ISU College of Agriculture and Life

Sciences, the Iowa State University Foundation, and the Iowa Monarch Conservation

Consortium. Without help from the United States Department of Agriculture, Agriculture

Research Service, Corn Insects and Crop Genetics Research Unit this work would not be possible. Specifically, thank you to Rick Hellmich, Kieth Bidne, and Royce Bitzer for maintaining monarch butterfly colonies, providing support, materials, and larvae used in these experiments.

Thank you, Daniel Borchardt, Pheasants Forever, and Quail Forever for aid in identifying potential field sites. I have a special appreciation for the property managers and landowners that allowed me to work on their property: Mike Loan and Sarah Nolte at Blue Grass Enterprises,

Justin Stensland at Stensland Sod, Adam Thomes and Ben Pease at the ISU Turf Research

Station, David Bruene at the ISU Beef Teaching Farm, Kent Berns at the ISU Dairy Farm,

Daniel Gudahl, Rob Davis, Penny Perkins, and Liz Garst at Whiterock Conservancy, the Mull and Gambaiani families, Jesse Randell, and Brab Licklider, as well as Brian and Cheryl Fuller for providing field housing, bonfires, and s’mores. Additional thanks to the many owners of properties that were trespassed across in order to relocate our radio-tagged monarch butterflies.

Thank our tireless research technicians for their enthusiasm, support, assistance, and willingness to learn: Cody Acevedo, Kevin Anderson, Jacqueline Appelhans, Matthew

Dollenbacher, Alec Euken, Cory Haggard, Signey Hilby, Anthony Kerker, Michael Roth, Riley xiv

Nylin, and Julia Pfeiffer, Jenna Nixt, Kayla Shepard, Samantha Shimota, Brooklyn Snyder, and

Kara Weber. Thank you to the undergraduate students who participated in independent research experiences and enabled me to grow as a mentor: Hectar Arbuckle, Karly Jans, Audrey Jenkins,

Taylor Rouse, Morgan Ubbelohde, Kayla Wernsing, Carolyn White, and Elke Windschitl.

Thank you to my colleagues and lab mates for your support, collaboration, and friendship: Seth Appelgate, Teresa Blader, James Dupuie, Emily Ernst, Brett Kelly, Nirinjana

Krishnan, Pat McGovern, Jake Olbrich, Victoria Pocius, Nancy Shryock, Chris Sullivan, and

Grace Vaziri. Thank you to my committee: James Adelman, Brad Coates, Rick Hellmich, and

Amy Toth for pushing me to reach my potential and never letting me take the easy way out.

Importantly, thank you to my major professor, Steven P. Bradbury, for helping me during every step of this project and beyond. I am thankful to have a wonderful adviser that leads by example and values his students as colleagues. I have learned so much from you and will always remember my time at Iowa State University fondly because of you.

Finally, this dissertation is dedicated to those near and far who supported me during my graduate work: family, friends, and pets. I was lucky to always have caring people to provide perspective when work became challenging. Moving 1,100 miles from my childhood home was difficult, but the constant love being sent through phone calls, video call, text messages, and snail mail, as well as the people I met in Iowa helped make this place home. My Ph.D. journey was made infinitely more enjoyable and manageable by my loving husband, Michael A. Roth

(and our dog Sulley). I will always be grateful for your courage to start a new chapter of life together in a new place with new people. Your active involvement in my data collection each summer, interest in understanding my projects, and consideration when work had to come first are only a few reasons I value our relationship. Thank you for helping me achieve my goals. xv

ABSTRACT

The eastern North American monarch butterfly is at risk of quasi-extinction and determined to be a candidate species for listing under the Endangered Species Act due, in part, to the loss of breeding habitat in agricultural landscapes of the Midwest United States. Because the monarch is a vagile species, the spatial arrangement of milkweed in the landscape is expected to influence movement patterns, habitat utilization, and reproductive output. To help obtain the biggest impact of newly established habitat, monarch butterfly movement behavior should be considered. This dissertation aims to advance understanding of monarch movement ecology at the milkweed patch, landscape, and continental scales and provide information that supports conservation planning. In addition, the work presented here propels the field of insect movement ecology with the development of new insights and implementation of novel methods and techniques.

At the milkweed patch scale explored in chapter 2, because of seemingly innate milkweed abandonment behavior and biomass consumption requirements, our results suggest milkweed patches containing at least two to four ramets of closely spaced common milkweed would provide sufficient biomass for development and increase the likelihood that larvae moving in random directions would encounter non-natal ramets to support complete development. Larval movement behavior and biomass requirements are critical aspects of monarch larval biology that should be considered in habitat restoration and maintenance plans, monitoring survey designs and protocols, and population modeling.

Through five years of landscape-scale movement experiments (Chapters 3-6), we released 393 radio-tagged, sham-tagged, and untagged monarch butterflies in various land cover types. With the use of radio telemetry, we attained more information about monarch butterfly xvi movement than was possible by visual observations only. We developed a new data collection system capable of estimating the location of a monarch butterfly in a sod field with a frequency

12 times faster than currently employed methods. We analyzed movement characteristics, including step length, directionality, and perceptual range. We employed novel approaches, specifically continuous-time movement models, to analyze monarch butterfly occurrence.

Finally, we conclude by reporting that 50 m perceptual range and 30 m step length are appropriate parameter selections in a spatially explicit agent-based model reported by Grant et al.

(2018). We found that turn angle was variable in relation to milkweed density, and therefore could be further explored for model parameterization. Additionally, parameters to simulate large dispersal steps should be included as this will greatly influence how monarchs move across the landscape.

Chapter 7 addresses the development of a method to improve the means to quantify mitochondrial DNA (mtDNA) variation, which is foundational to exploring haplotype variation and understand migratory movements of monarch butterflies. Although mtDNA haplotype variation is often used to estimating population dynamics and phylogenetic relationships, economical and generalized methods for entire mtDNA genome enrichment prior to high- throughput sequencing are not readily available. Our method of differential centrifugation showed a significant mean increase in mtDNA in comparison to traditional methods. These methods can be employed to evaluate spatial-temporal mtDNA variation and natal origins of monarchs across the breeding range in comparison to the overwintering population in Mexico.

This would significantly advance fundamental knowledge of continental-scale dispersal ecology and genetic fluctuation in monarch populations, which is essential for evaluating potential resiliency and developing effective conservation strategies. xvii

This dissertation successfully addressed gaps in the knowledge of monarch butterfly biology that are instrumental to creating responsive conservation strategies. A variety of experimental techniques, including VHF radio telemetry and next-generation sequencing, were utilized to investigate larval and adult monarch movement behavior at multiple spatial scales.

Results will inform conservation plans with recommendations for the spatial arrangement of restored habitat patches to create a connected landscape for monarch butterflies in their summer breeding range. In addition to providing valuable insight, advancements were made by developing new methods, employing technology with novel approaches, and applying the latest analyses in spatial statistics. 1

CHAPTER 1. GENERAL INTRODUCTION

The monarch butterfly (Danaus plexippus) is a holometabolous lepidopteran species that is known for its charismatic orange and black appearance, congregation-like overwintering behavior, and expansive distribution across the United States of America during the summer months. Monarchs have an obligate relationship with their host plant, milkweed (Asclepias sp.).

Because of its relative abundance, common milkweed (Asclepias syriaca) is generally associated with monarchs in the northcentral states (Malcolm et al. 1993, Geest et al. 2019), but other species within Asclepiadaceae (e.g., Asclepias tuberosa and Asclepias incarnata) and

Apocynaceae elicit oviposition and support larval development. After eggs hatch, larvae feed exclusively on milkweed, develop through five instars, pupate, and eclose as adults. Monarch development is temperature dependent, but larval development typically takes 14 days and pupation lasts approximately nine days. Non-migratory adults have a lifespan of approximately six weeks (Nail et al. 2015). Migratory adults can live up to nine months and travel thousands of kilometers (Oberhauser et al. 2017).

The eastern and western North American monarch butterfly subpopulations are assumed to be panmictic (Pyle 2015), despite differences in migratory behavior and overwintering sites.

The overwintering generation of monarchs in the eastern subpopulation migrates from Mexico to the southern U.S. in late winter/early spring. The subsequent generation migrates to the northcentral and northeastern U.S. in late spring, and the 4th generation migrates back to Mexico in the fall. During summer months, 2nd and 3rd generations in northcentral and northeastern U.S. are not migratory (Oberhauser et al. 2017). The western subpopulation overwinters along coastal

California and has a shorter northern migration to its breeding range in the Pacific Northwest

(Pelton et al. 2019). Other non-migratory monarch subpopulations reside in Mexico, Central 2

America, Florida, the Caribbean, Hawaii, Australia, New Zealand, and Southern Europe (Brower and Jeansonne 2004, Lyons et al. 2012, Pierce et al. 2014, Servín-Garcidueñas and Martínez-

Romero 2014, Zhan et al. 2014, Pfeiler et al. 2017).

Both the eastern and western North American subpopulations have experienced population declines. Changes in forest area occupied by overwintering monarch colonies in

Mexico provides an indirect estimate of changes in population size for the eastern subpopulation.

Based on this metric, there has been an 80% decline since 1994 (Brower et al. 2012, Pleasants and Oberhauser 2013, Jepsen et al. 2015, Semmens et al. 2016, Martínez-Meza et al. 2020). A long-term average of six-hectares of overwintering monarchs in Mexico is needed to support a resilient population and reduce the risk of losing the eastern North American migration in the next 10 to 20 years (Semmens et al. 2016). Likewise, the western population has declined by

99%, as estimated by annual citizen science counts of overwintering roosts (Pelton et al. 2019).

With the magnitude of this decline, it is unknown if the western subpopulation will be able to recover (Pelton et al. 2019). Given these declines, the U.S. Fish and Wildlife Service deemed the monarch as a candidate species for listing under the Endangered Species Act in December 2020

(USFWS 2020).

There is no single cause for the monarch butterfly declines; rather, a combination of factors influence the population size, such as weather conditions, overwintering, and breeding habitat loss, and pesticide use (Brower et al. 2012, Flockhart et al. 2015, Pelton et al. 2019).

Eastern subpopulation conservation is currently approached from multiple perspectives based on regional needs (Flockhart et al. 2015, Oberhauser et al. 2017). Because stable isotope analyses suggest a large fraction of overwintering monarchs originate in the northcentral states

(Wassenaar and Hobson 1998, Flockhart et al. 2017), conservation actions in this region 3 emphasize the establishment of milkweed to provide oviposition sites and larval forage.

Southcentral states are restoring floral resources to provide nectar for migrating monarchs, while

Mexico emphasizes forest conservation to protect overwintering sites. Overall, sustainable migratory monarch populations require the maintenance of genetic diversity that increases the ability to adapt to changing climatic conditions (Pauls et al. 2013, Schierenbeck 2017).

A 20-year landscape-scale, habitat conservation, and milkweed establishment strategy for the northcentral region was published in 2018 (Midwest Association of Fish and Wildlife

Agencies 2018). To support a resilient eastern monarch population, an estimated 1.6 billion additional stems of milkweed need to be established in the monarchs’ breeding range, which transcends the northcentral states; Iowa is responsible for approximately 140,185,076 stems

(Midwest Association of Fish and Wildlife Agencies 2018). To meet this goal, milkweed will need to be established in every sector, including protected areas, conservation reserve program, urban/suburban, rights of way, and agriculture (Thogmartin et al. 2017). Because of the enormity of this task and logistical restrictions, it will take time to reach the habitat establishment goal.

To help obtain the largest impact of newly established habitat, monarch butterfly movement behavior should be considered. Discerning animal movement patterns in fragmented landscapes is foundational to understanding population dynamics, population persistence, and population distribution (Hanski 1998, Barton et al. 2009, Stevens et al. 2012, MacDonald et al.

2019). The size of the overwintering population results from the number of monarch eggs laid in the landscape and their survival – more eggs laid and survival of more individuals to the adult stage will result in a larger overwintering population size (Flockhart et al. 2015, Oberhauser et al.

2017). Understanding monarch movement at the host plant, landscape, and continental-scale are critical to employing a biologically relevant conservation strategy. 4

There are three issues that need to be considered to increase the monarch’s realized fecundity: 1) milkweed is not distributed evenly across the landscape; 2) monarch females are not patch residents; and 3) milkweed is not utilized evenly by monarchs for oviposition.

Common milkweed is a clonal plant species, meaning that in addition to reproducing through seed, it reproduces by sending new shoots from its rhizomes. Often, a field has dense patches of milkweed; however, isolated milkweed ramets also can be found (Blader 2018, Pitman et al.

2018). Non-migratory monarch females are vagile and estimated to travel up to 15 km/d between milkweed patches during the breeding season (Zalucki and Kitching 1984, Zalucki and Lammers

2010, Zalucki et al. 2016). During their lifetime, female monarchs lay 300-400 eggs singly but do not oviposit equally in milkweed patches and on isolated ramets. Eggs typically accumulate on the edge of milkweed patches leaving ramets in the middle underutilized, while isolated milkweed ramets can contain up to ten monarch eggs at one time (Zalucki and Kitching 1982a).

Additionally, the survival of eggs on isolated stems and in milkweed patches is uneven, with greater survival rates suggested on isolated stems because of reduced predation (Zalucki and

Kitching 1982b, Nail et al. 2015).

Given these behaviors, the spatial arrangement of milkweed in the landscape is expected to influence movement patterns, habitat utilization, and reproductive output (Bergin et al. 2000,

Misenhelter and Rotenberry 2000, Grant et al. 2018, Pitman et al. 2018, Grant and Bradbury

2019). This dissertation aims to understand monarch butterfly movement ecology to provide information that supports conservation planning. Research studies were undertaken to advance understanding of monarch movement ecology at multiple spatial scales. In addition to providing valuable insight for monarch butterfly conservation, the work presented here also propels the 5 field of insect movement ecology with the development of new insights and the implementation of novel methods and techniques.

Outline and Objectives of the Dissertation Chapters

Chapter 2: Estimates of common milkweed (Asclepias syriaca) utilization by monarch larvae (Danaus plexippus) and the significance of larval movement

Chapter 2 addresses larval monarch butterfly movement ecology at the milkweed patch scale. Results of this research were published in January 2020 in the Journal of Insect

Conservation (Fisher, Hellmich, et al. 2020). This study explored the effect of milkweed ramet density on larval search behavior, milkweed utilization, and survival without predation, parasitism, or competition. Under experimental greenhouse conditions, monarch larvae abandoned their natal milkweed ramet and subsequent ramets prior to the pre-pupal wandering stage and before all available leaf biomass on a ramet was consumed. This finding is consistent with previous field observations (Urquhart 1960, Rawlins and Lederhouse 1981, Borkin 1982).

Larvae consumed biomass from three or four milkweed ramets that totaled the approximate biomass of a single 10–35 cm ramet. Larval movement behavior suggests that isolated ramets may not support development through pupation, even though an isolated ramet could provide enough biomass.

Our results suggest milkweed patches containing at least two to four ramets of closely spaced common milkweed would provide sufficient biomass for development and increase the likelihood that larvae moving in random directions would encounter non-natal ramets to support complete development. Larval movement behavior and biomass requirements are critical aspects of monarch larval biology that should be considered in habitat restoration and maintenance plans, monitoring survey designs and protocols, and population modeling. An appendix of this chapter summarizes a follow-up study concerning the influence of intraspecific competition on larval 6 movement, natal plant abandonment, and fitness under controlled greenhouse conditions in the absence of predation.

Chapters 3, 4, 5, and 6: Analyzing monarch movement ecology at a landscape-scale with aid from handheld and automated VHF radio telemetry systems

The next four chapters advance understanding of monarch butterfly movement at a landscape scale. Spatially-explicit, landscape-scale population models are often used in conservation planning to understand population dynamics (Lima and Zollner 1996). Agent-based modeling suggests establishing small habitat patches at distances within the monarch’s perceptual range will support increased egg densities in the landscape (Zalucki et al. 2016, Grant et al. 2018, Grant and Bradbury 2019). Simulations of female movement in a spatially-explicit

Iowa, USA landscape are, however, highly dependent on an assumed perceptual range between

50 and 400 m (Grant et al. 2018). The research reported in Chapters 3 – 6 seeks to refine current agent-based model inputs such as perceptual range, directionality, and step lengths to predict realistic output (Zalucki et al. 2016, Grant et al. 2018). This research developed and implemented

VHF radio telemetry techniques to provide empirical evidence of monarch movement behavior.

Additionally, work in these chapters provides baseline recommendations for implementing radio telemetry to track insects as technology improves.

Based on a review by Kissling et al. (2014) that addresses the strengths and limitations of a variety of tracking technologies for insects, Chapter 3 reports the development of VHF radio telemetry methods to reduce uncertainty in estimated flight paths of non-migratory female monarch butterflies at distances beyond human visual limits. This chapter was published in

Environmental Entomology in March 2020 (Fisher, Adelman, et al. 2020). We reported: 1) a

220 mg transmitter does not significantly affect monarch butterfly behavior and flight capability,

2) a technique using handheld antennae to collect location data every minute, which is a time- 7 scale relevant for monarch flight behavior, and 3) preliminary insights on monarch butterfly movement with a straight-line flight path analysis over distances larger than previously reported with visual data collection. Thirteen radio-tagged monarchs were released in a restored prairie, and locations were estimated every minute for up to 39 min by simultaneous triangulation from four technicians. Monarchs that left the prairie were tracked and relocated at distances up to 250 m. Assuming straight flights between locations, most steps within the prairie were below 50 m. Steps associated with exiting the prairie exceeded 50 m with high directionality. Because butterflies likely do not fly in straight lines between stationary points, we also illustrate how occurrence models can use location data obtained through radio telemetry to estimate movement within a prairie and over multiple land cover types.

In Chapter 4, the design of an automated radio telemetry system (ARTS) is described.

The ARTS was developed to increase scalability and frequency of data collection intervals while potentially reducing errors in location estimates for insects (Kissling et al. 2014). This chapter was published in Methods in Ecology and Evolution in September 2020 (Fisher, Dixon, et al.

2020). Several research teams have described ARTS for tracking mammals and birds (Cochran et al. 1965, Larkin et al. 1996, Kays et al. 2011, Ward et al. 2013, Lenske and Nocera 2018, Crewe et al. 2019, Knight et al. 2019). Most notable Kays et al. (2011) designed and utilized an ARTS to track rainforest birds and mammals on Barro Colorado Island, Panama. This system incorporated seven custom, automated receiving units on 40 m receiving stations to cover an area of roughly 1,500 ha. Each receiving station housed six, six-element directional Yagi antennae pointing in known bearing directions. The advances of Kays et al. (2011) and others to track animals ranging from 21 to 10,000 g in size provide the framework we used to develop an ARTS 8 designed for tracking 0.5 g insects, carrying low-powered transmitters, over spatial scales of 4–6 ha.

We designed the ARTS, using commercially available equipment, to estimate fine-scale locations of large insects (e.g., 0.5 g butterflies) within a 200 m2 study area. The ARTS consisted of four receiving stations. Each receiving station had four 3-element, directional Yagi antennae

(separated by 60°) connected to an automated receiver that records detected power sequentially from each antenna. To develop and evaluate the ARTS performance, four receiving stations were installed in the corners of 4 and 6.25-ha square fields with varying heights of vegetative cover.

The location of a 0.22 g transmitter was estimated with a statistical method implementing both distance-and angle-power relationships. Calibrated model parameters were based on power detected from transmitters at known locations. Using independently collected data, model performance was evaluated based on estimated locations of a georeferenced stationary transmitter, a moving transmitter with a known georeferenced path, and a transmitter attached to a monarch butterfly. Estimated locations were calculated as frequently as every 5 s, which is at least 12 times greater than the sampling frequency previously reported for tracking insects with handheld equipment. When sufficient power data were received, the median estimated locations of a transmitter attached to an investigator's hat were <16 m from the true location. The effective median radius of the 95% confidence ellipse was 18.3 m for stationary targets and 15.9 m for a moving transmitter. The median distance between the estimated and true locations of radio- tagged monarchs was 72 m. Greater error in location estimation was expected when the transmitter was attached to a monarch butterfly due to interference from vegetation and variability in antenna orientation and transmitter height. After applying a correction factor for the effect of vegetation, median location error was reduced by 12 m. While our ARTS reflects the 9 limit of current technology, the system is a substantial methodological advancement for locating butterflies and other large insects. Our efforts provide a performance benchmark as technology improves.

Chapter 5 describes a study to estimate female perceptual range to milkweed and nectar resources. This chapter was submitted to Environmental Entomology in December 2020 and is under review (Fisher and Bradbury 2021). While collecting data required to develop and calibrate the ARTS described in Chapter 4, we conducted experiments to estimate the perceptual range of monarch butterflies at a landscape scale. As insects disperse across a landscape to locate oviposition and forage resources, they are guided by olfactory and visual cues. Dispersal and movement behaviors are directly impacted by perceptual range: the maximum distance individuals are able to detect stimuli with their sensory organs (Wiens 1989). Given its influence on movement, and therefore population dynamics, perceptual range is a key input parameter in many landscape-scale population models (Lima and Zollner 1996). Agent-based modeling suggests establishing small habitat patches at distances within the monarch’s perceptual range will support increased egg densities in the landscape (Zalucki et al. 2016, Grant et al. 2018).

Simulations of female movement in a spatially-explicit Iowa, USA landscape are, however, highly dependent on an assumed perceptual range between 50 and 400 m (Grant et al. 2018).

We released 188 monarch females 5, 25, 50, and 75 m downwind from potted milkweed and blooming forbs in 4-32 ha sod fields and quantified the extent of directed flights toward the resources. We also observed flight patterns of 49 non-experimental monarchs in our study areas.

A unimodal pattern of movement was observed when monarchs were released 5, 25, or 50 m downwind of the resources; however, the majority of flights were with the wind rather than toward the resources. For those monarchs that flew into the wind, direct flights to the resources 10 ranged from 3-125 m for eight experimental and 49 non-experimental monarchs. Based on these observations, we conclude that the perceptual range was functionally ≤ 50 m. We classified the dispersal behavior of the monarchs in this resource-poor habitat as direct displacement, based on the similarity of effective and total distances traveled and shallow turn angles. These results provide improved input estimates for spatially explicit agent-based modeling and conservation plans.

In Chapter 6, we undertook a study that combined the handheld radio telemetry methods with the automated radio telemetry system to track monarch butterflies released at field sites with a more diverse suite of land cover classes. This chapter will be submitted to Applied Ecology in

2021. The goal was to advance understanding of how female monarch movement behavior and space use are influenced by different configurations of contiguous habitat classes with varying density of milkweed and nectar resources. Monarch butterflies were released in maize fields, restored or remnant prairies, grass-dominated fields and roadside rights of ways, and habitat edges. Landscape settings included an 800 m2 mosaic configuration of habitat classes and secondary gravel roadsides with and without access to resource-rich areas. Monarchs were visually tracked for up to one hour. Visual observations were aided using automated and handheld radio telemetry systems. This study provided an opportunity to compare landscape- scale empirical evidence to simulated model predictions of perceptual range, flight directionality, and flight step lengths needed to produce biologically-relevant recommendations for conservation practices.

To date, this study provides the most robust dataset of directly tracked movement of individual breeding-season monarch butterflies to assess utilization of agricultural landscapes.

Monarchs were observed moving among habitat classes and, although not always, were observed 11 performing natural up-wind search behavior. Step lengths and turn angles were mostly below 50 m and directional, regardless of habitat class or spatial configuration. This observation was likely a response to floral and milkweed density rather than the presence or absence of resources. As reported by Fisher, Adelman, et al. (2020) and Fisher and Bradbury (2021), we observed large exiting steps from study areas of up to ≥ 1,000 m. These dispersal steps occurred most frequently from resource-rich habitat and could indicate initiation of a long-range search for a new resource-rich habitat, which in turn would increase egg distribution across the landscape. In summary, monarchs successfully located resource-rich habitat; 78.9% of monarchs not released in a resource-rich habitat were observed in resource-rich habitat, with some monarchs flying over

500 m to find resource-rich areas. These results suggest that model assumptions are reasonable.

Chapter 7: Prediction of mitochondrial genome-wide variation through sequencing of mitochondrion-enriched extracts

Chapter 7 addresses the development of a method to improve the means to quantify mitochondrial DNA (mtDNA) variation. Evaluating spatial-temporal mtDNA variation and natal origins of monarchs across the breeding range in comparison to the overwintering population in

Mexico will significantly advance fundamental knowledge of dispersal ecology and genetic fluctuation in monarch populations, which is essential for evaluating potential resiliency and developing effective conservation strategies. The impact of a genetic shift on monarch health and capacity to adapt to changing landscapes and environmental conditions is unknown. Recent population declines suggest an increased probability that genetic drift could drastically shift mtDNA genetic frequencies. Thus, accumulated genetic diversity among back-migrating monarchs from northcentral and eastern states could, in turn, have an increased influence on the genetic diversity within the admixed overwintering subpopulation in Mexico. From 2016-2020, wild monarch butterflies were collected and preserved (stored at -20 °C) across North America 12 and the Pacific basin. In 2016, adult monarchs were collected from six northcentral states (IA,

OH), eastern (PA, DE), and western states (NV, ID), Hawaii, and Australia. Each year from

2016-2020, adult monarchs were collected from May-October in Iowa, including spring migrants, summer residents, and fall migrants. These samples provide an opportunity to explore population structure, genetic drift, and migration.

Prior studies reported a lack of significant variation, including at microsatellite loci in the nuclear genome (Lyons et al. 2012, Pierce et al. 2014, Zhan et al. 2014, Pfeiler et al. 2017, Talla et al. 2020) and within mtDNA fragments (Brower and Jeansonne 2004, Servín-Garcidueñas and

Martínez-Romero 2014, Pfeiler et al. 2017), between eastern and western migratory North

American monarchs collected in their overwintering habitats. However, when these samples were compared to non-migratory monarchs in Mexico, Central America, Florida, Hawaii,

Australia, and New Zealand, genetic differences were detected. The reported lack of differentiation between overwintering subpopulations based on analyses of mtDNA data was likely premature because of study design limitations. Specifically, these prior studies compared the full mtDNA genome of 3 individuals (Servín-Garcidueñas and Martínez-Romero 2014) or short 1,555 bp fragments of a single mtDNA gene from 38 individuals (Brower and Jeansonne

2004). In addition, these studies assessed genetic variations in monarch butterflies primarily sampled from the overwintering sites, where genetic admixture occurs annually. Although overwintering monarchs showed the greatest overall genetic diversity, a study by Eanes and

Koehn (1978) relied upon data from six isoenzymes to suggest random genetic drift may lead to transient genetic differentiation among non-migratory subpopulations across the northcentral and northeastern states during the summer breeding season. 13

Although mtDNA haplotype variation is often used to estimate population dynamics and phylogenetic relationships, economic and generalized methods for entire mtDNA genome enrichment prior to high-throughput sequencing are not readily available. Colony reared Ostrinia nubilalis were used as a model organism to demonstrate the utility of differential centrifugation to enrich for mitochondrion within cell extracts prior to DNA extraction, short-read sequencing, and assembly using exemplars from eight maternal lineages. Compared to controls, enriched extracts showed a significant mean increase of 48.2- and 86.1-fold in mtDNA based on quantitative PCR and proportion of subsequent short sequence reads that aligned to the O. nubilalis reference mitochondrial genome, respectively. Compared to the reference genome, our de novo assembled O. nubilalis mitochondrial genomes contained 82 intraspecific substitution and insertion/deletion mutations and provided evidence for correction of mis-annotated 28 C- terminal residues within the NADH dehydrogenase subunit 4. Comparison to a more recent O. nubilalis mtDNA assembly from unenriched short read data analogously showed 77 variant sites.

Twenty-eight variant positions and a triplet ATT codon (Ile) insertion within ATP synthase subunit 8 were unique within our assemblies. This study provides a generalizable pipeline for whole mitochondrial genome sequence acquisition adaptable to applications across a range of taxa. This chapter was published in Scientific Reports in August of 2020 (Fisher, Bradbury, et al.

2020) and is the foundation of a USDA-NIFA Education and Workforce Development

Postdoctoral Fellowship Program grant application proposing to study monarch butterfly migration using the wild monarch butterflies that were collected and preserved (stored at -20 °C) across North America and the Pacific basin from 2016-2020. The grant was submitted in August

2020 under the guidance of Dr. Alan Wanamaker, Dr. Brad Coates, and Dr. Steven Bradbury.

Chapter 8: General Conclusion

The final chapter of this dissertation summarizes the overall research effort presented in chapters 2-7. Additionally, this chapter provides suggestions for future studies and applications of research findings in insect conservation biology.

Contributions of this Dissertation

References

Barton, K. A., B. L. Phillips, J. M. Morales, and J. M. J. Travis. 2009. The evolution of an ‘intelligent’ dispersal strategy: biased, correlated random walks in patchy landscapes. Oikos. 118: 309–319.

Bergin, T. M., L. B. Best, K. E. Freemark, and K. J. Koehler. 2000. Effects of landscape structure on nest predation in roadsides of a midwestern agroecosystem: a multiscale analysis. Landscape Ecology. 15: 131–143.

Blader, T. R. 2018. Milkweed patch size effects on monarch butterfly oviposition within Iowa prairies and roadsides (Master of Science).

Borkin, S. S. 1982. Notes on Shifting Distribution Patterns and Survival of Immature Danaus Plexippus (Lepidoptera: Danaidae) on the Food Plant Asclepias Syriaca. The Great Lakes Entomologist. 15: 9. 15

Brower, A. V. Z., and M. M. Jeansonne. 2004. Geographical populations and “subspecies” of new world monarch butterflies (Nymphalidae) share a recent origin and are not phylogenetically distinct. Ann. Entomol. Soc. of Am. 97: 519–523.

Brower, L. P., O. R. Taylor, E. H. Williams, D. A. Slayback, R. R. Zubieta, and M. I. Ramírez. 2012. Decline of monarch butterflies overwintering in Mexico: is the migratory phenomenon at risk?: Decline of monarch butterflies in Mexico. Insect Conservation and Diversity. 5: 95–100.

Cochran, W. W., D. W. Warner, J. R. Tester, and V. B. Kuechle. 1965. Automatic Radio- Tracking System for Monitoring Animal Movements. BioScience. 15: 98–100.

Crewe, T. L., J. E. Deakin, A. T. Beauchamp, and Y. E. Morbey. 2019. Detection range of songbirds using a stopover site by automated radio-telemetry. Journal of Field Ornithology. 90: 176–189.

Eanes, W. F., and R. K. Koehn. 1978. An analysis of genetic structure in the monarch butterfly, Danaus plexippus L. Evolution. 32: 784–797.

Fisher, K. E., J. S. Adelman, and S. P. Bradbury. 2020. Employing Very High Frequency (VHF) Radio Telemetry to Recreate Monarch Butterfly Flight Paths. Environmental Entomology. 49: 312–323.

Fisher, K. E., and S. P. Bradbury. 2021. Estimating perceptual range of female monarch butterflies (Danaus plexippus) to potted vegetative common milkweed (Asclepias syriaca) and nectar resources. Environ Entomol. In review, Submitted December 2020.

Fisher, K. E., S. P. Bradbury, and B. S. Coates. 2020. Prediction of mitochondrial genome-wide variation through sequencing of mitochondrion-enriched extracts. Scientific Reports. 10: 19123.

Fisher, K. E., P. M. Dixon, G. Han, J. S. Adelman, and S. P. Bradbury. 2020. Locating large insects using automated VHF radio telemetry with a multi-antennae array. Methods Ecol Evol. DOI: 10.1111/2041-210X.13529.

Fisher, K. E., R. L. Hellmich, and S. P. Bradbury. 2020. Estimates of common milkweed (Asclepias syriaca) utilization by monarch larvae (Danaus plexippus) and the significance of larval movement. J Insect Conserv. 24: 297–307.

Flockhart, D. T. T., L. P. Brower, M. I. Ramirez, K. A. Hobson, L. I. Wassenaar, S. Altizer, and D. R. Norris. 2017. Regional climate on the breeding grounds predicts variation in the natal origin of monarch butterflies overwintering in Mexico over 38 years. Global Change Biology. 23: 2565–2576.

Flockhart, D. T. T., J.-B. Pichancourt, D. R. Norris, and T. G. Martin. 2015. Unravelling the annual cycle in a migratory animal: breeding-season habitat loss drives population declines of monarch butterflies. Journal of Animal Ecology. 84: 155–165. 16

Geest, E. A., L. L. Wolfenbarger, and J. P. McCarty. 2019. Recruitment, survival, and parasitism of monarch butterflies (Danaus plexippus) in milkweed gardens and conservation areas. J Insect Conserv. 23: 211–224.

Grant, T. J., and S. P. Bradbury. 2019. The Role of Modeling in Monarch Butterfly Research and Conservation. Front. Ecol. Evol. 7: 197.

Grant, T. J., H. R. Parry, M. P. Zalucki, and S. P. Bradbury. 2018. Predicting monarch butterfly (Danaus plexippus) movement and egg-laying with a spatially-explicit agent-based model: The role of monarch perceptual range and spatial memory. Ecological Modelling. 374: 37–50.

Hanski, I. 1998. Metapopulation dynamics. Nature. 396: 41–49.

Jepsen, S., D. F. Schweitzer, B. Young, N. Sears, M. Ormes, and S. H. Black. 2015. Conservation status and ecology of the monarch butterfly in the United States. 36.

Kays, R., S. Tilak, M. Crofoot, T. Fountain, D. Obando, A. Ortega, F. Kuemmeth, J. Mandel, G. Swenson, T. Lambert, B. Hirsch, and M. Wikelski. 2011. Tracking Animal Location and Activity with an Automated Radio Telemetry System in a Tropical Rainforest. The Computer Journal. 54: 1931–1948.

Kissling, W. D., D. E. Pattemore, and M. Hagen. 2014. Challenges and prospects in the telemetry of insects. Biological Reviews. 89: 511–530.

Knight, S. M., G. M. Pitman, D. T. T. Flockhart, and D. R. Norris. 2019. Radio-tracking reveals how wind and temperature influence the pace of daytime insect migration. Biol. Lett. 15: 20190327.

Larkin, R. P., A. Raim, and R. H. Diehl. 1996. Performance of a Non-Rotating Direction-Finder for Automatic Radio Tracking (Desempeño de un Localizador No-Rotativo Para Radiotelemetría Automática). Journal of Field Ornithology. 67: 59–71.

Lenske, A. K., and J. J. Nocera. 2018. Field test of an automated radio-telemetry system: tracking local space use of aerial insectivores. Journal of Field Ornithology. 89: 173–187.

Lima, S. L., and P. A. Zollner. 1996. Towards a behavioral ecology of ecological landscapes. Trends in Ecology & Evolution. 11: 131–135.

Lyons, J. I., A. A. Pierce, S. M. Barribeau, E. D. Sternberg, A. J. Mongue, and J. C. De Roode. 2012. Lack of genetic differentiation between monarch butterflies with divergent migration destinations. Mol. Ecol. 21: 3433–3444.

MacDonald, Z. G., J. H. Acorn, J. Zhang, and S. E. Nielsen. 2019. Perceptual Range, Targeting Ability, and Visual Habitat Detection by Greater Fritillary Butterflies Speyeria cybele (Lepidoptera: Nymphalidae) and Speyeria atlantis. Journal of Insect Science. 19: 10. 17

Malcolm, S., B. J. Cockrell, and L. Brower. 1993. Spring recolonization of eastern North America by the monarch butterfly: successive brood or single sweep migration?, pp. 253– 267. In Malcolm, S., Zalucki, M.P. (eds.), Biology and Conservation of the Monarch Butterfly. Natural History Museum of Los Angeles County, Los Angeles, CA.

Martínez-Meza, F., M. A. Mendoza-Pérez, M. Cruz-Piña, and G. Mondragón-Contreras. 2020. Superficie forestal ocupada por las colonias de mariposas monarca en Mexico durante la hibernacion de 2019-2020. (https://monarchjointventure.org/images/uploads/documents/2019_Monitoreo_Mariposa_ Monarca_en_M%C3%A9xico_2019-2020.pdf).

Midwest Association of Fish and Wildlife Agencies. 2018. Mid-America Monarch Conservation Strategy. (http://www.mafwa.org/wp- content/uploads/2018/07/MAMCS_June2018_Final.pdf).

Misenhelter, M., and J. Rotenberry. 2000. Choices and Consequences of Habitat Occupancy and Nest Site Selection in Sage Sparrows. Ecology. 81: 2892–2901.

Nail, K. R., C. Stenoien, and K. S. Oberhauser. 2015. Immature Monarch Survival: Effects of Site Characteristics, Density, and Time. Ann Entomol Soc Am. 108: 680–690.

Oberhauser, K., R. Wiederholt, J. E. Diffendorfer, D. Semmens, L. Ries, W. E. Thogmartin, L. Lopez‐Hoffman, and B. Semmens. 2017. A trans-national monarch butterfly population model and implications for regional conservation priorities. Ecological Entomology. 42: 51–60.

Pauls, S. U., C. Nowak, M. Bálint, and M. Pfenninger. 2013. The impact of global climate change on genetic diversity within populations and species. Molecular Ecology. 22: 925– 946.

Pelton, E. M., C. B. Schultz, S. J. Jepsen, S. H. Black, and E. E. Crone. 2019. Western monarch population plummets: Status, probable causes, and recommended conservation actions. Front. Ecol. Evol. 7.

Pfeiler, E., N. O. Nazario-Yepiz, F. Pérez-Gálvez, C. A. Chávez-Mora, M. R. L. Laclette, E. Rendón-Salinas, and T. A. Markow. 2017. Population genetics of overwintering monarch butterflies, Danaus plexippus (Linnaeus), from Central Mexico inferred from mitochondrial DNA and microsatellite markers. J Hered. 108: 163–175.

Pierce, A. A., M. P. Zalucki, M. Bangura, M. Udawatta, M. R. Kronforst, S. Altizer, J. F. Haeger, and J. C. de Roode. 2014. Serial founder effects and genetic differentiation during worldwide range expansion of monarch butterflies. Proceedings of the Royal Society B: Biological Sciences. 281: 20142230.

Pitman, G. M., D. T. T. Flockhart, and D. R. Norris. 2018. Patterns and causes of oviposition in monarch butterflies: Implications for milkweed restoration. Biological Conservation. 217: 54–65. 18

Pleasants, J. M., and K. S. Oberhauser. 2013. Milkweed loss in agricultural fields because of herbicide use: effect on the monarch butterfly population. Insect Conservation and Diversity. 6: 135–144.

Pyle, R. M. 2015. Monarchs in the mist: new perspectives on monarch distribution in the Pacific Northwest, pp. 236–246. In Monarchs in a Changing World: Biology and Conservation of an Iconic Butterfly. Cornell University Press, Ithaca, NY.

Rawlins, J. E., and R. C. Lederhouse. 1981. Developmental Influences of Thermal Behavior on Monarch Caterpillars (Danaus plexippus): An Adaptation for Migration (Lepidoptera: Nymphalidae: Danainae). Journal of the Kansas Entomological Society. 54: 387–408.

Schierenbeck, K. A. 2017. Population-level genetic variation and climate change in a biodiversity hotspot. Ann Bot. 119: 215–228.

Semmens, B. X., D. J. Semmens, W. E. Thogmartin, R. Wiederholt, L. López-Hoffman, J. E. Diffendorfer, J. M. Pleasants, K. S. Oberhauser, and O. R. Taylor. 2016. Quasi-extinction risk and population targets for the Eastern, migratory population of monarch butterflies (Danaus plexippus). Scientific reports. 6: 23265.

Servín-Garcidueñas, L. E., and E. Martínez-Romero. 2014. Complete mitochondrial genome recovered from the gut metagenome of overwintering monarch butterflies, Danaus plexippus (L.) (Lepidoptera: Nymphalidae, Danainae). Mitochondrial DNA. 25: 427– 428.

Stevens, V. M., A. Trochet, H. V. Dyck, J. Clobert, and M. Baguette. 2012. How is dispersal integrated in life histories: a quantitative analysis using butterflies. Ecology Letters. 15: 74–86.

Talla, V., A. A. Pierce, K. L. Adams, T. J. B. de Man, S. Nallu, F. X. Villablanca, M. R. Kronforst, and J. C. de Roode. 2020. Genomic evidence for gene flow between monarchs with divergent migratory phenotypes and flight performance. Molecular Ecology. 29: 2567–2582.

Thogmartin, W. E., L. López-Hoffman, J. Rohweder, J. Diffendorfer, R. Drum, D. Semmens, S. Black, I. Caldwell, D. Cotter, P. Drobney, L. L. Jackson, M. Gale, D. Helmers, S. Hilburger, E. Howard, K. Oberhauser, J. Pleasants, B. Semmens, O. Taylor, P. Ward, J. F. Weltzin, and R. Wiederholt. 2017. Restoring monarch butterfly habitat in the Midwestern US: ‘all hands on deck.’ Environ. Res. Lett. 12: 074005.

Urquhart, F. 1960. The Monarch Butterfly. University of Toronto Press, Toronto.

USFWS. 2020. Monarch (Danaus plexippus) Species Status Assessment Report, version 2.1 September 2020. (https://www.fws.gov/savethemonarch/pdfs/Monarch-SSA-report.pdf).

Ward, M. P., J. H. Sperry, and P. J. Weatherhead. 2013. Evaluation of Automated Radio Telemetry for Quantifying Movements and Home Ranges of Snakes. J. of Herpetology. 47: 337–345. 19

Wassenaar, L. I., and K. A. Hobson. 1998. Natal origins of migratory monarch butterflies at wintering colonies in Mexico: New isotopic evidence. PNAS. 95: 15436–15439.

Wiens, J. A. 1989. Spatial Scaling in Ecology. Functional Ecology. 3: 385–397.

Zalucki, M. P., and R. L. Kitching. 1982a. Dynamics of oviposition in Danaus plexippus (Insecta: Lepidoptera) on milkweed, Asclepias spp. Journal of Zoology. 198: 103–116.

Zalucki, M. P., and R. L. Kitching. 1982b. Temporal and spatial variation of mortality in field populations of Danaus plexippus L. and D. chrysippus L. Larvae (Lepidoptera: Nymphalidae). Oecologia. 53: 201–207.

Zalucki, M. P., and R. L. Kitching. 1984. The dynamics of adult Danaus plexippus L. (Danaidae) within patches of its food plant, Asclepias spp. J. Lepid. Soc. 38: 209–2219.

Zalucki, M. P., and J. H. Lammers. 2010. Dispersal and egg shortfall in Monarch butterflies: what happens when the matrix is cleaned up? Ecological Entomology. 35: 84–91.

Zalucki, M. P., H. R. Parry, and J. M. Zalucki. 2016. Movement and egg laying in Monarchs: To move or not to move, that is the equation. Austral Ecology. 41: 154–167.

Zhan, S., W. Zhang, K. Niitepõld, J. Hsu, J. F. Haeger, M. P. Zalucki, S. Altizer, J. C. de Roode, S. M. Reppert, and M. R. Kronforst. 2014. The genetics of monarch butterfly migration and warning colouration. Nature. 514: 317–321.

CHAPTER 2. ESTIMATES OF COMMON MILKWEED (Asclepias syriaca) UTILIZATION BY MONARCH LARVAE (Danaus plexippus) AND THE SIGNIFICANCE OF LARVAL MOVEMENT

Kelsey E. Fisher1, Richard L. Hellmich1,2, and Steven P. Bradbury1,3

1Department of Entomology, Iowa State University, Ames, IA, 50011

2United States Department of Agriculture, Agriculture Research Service, Corn Insects and Crop

Genetics Research Unit, Ames, IA 50011

3Department of Natural Resource Ecology and Management, Iowa State University, Ames, IA

50011

Modified from a manuscript published in Journal of Insect Conservation.

Fisher, KE, RL Hellmich & SP Bradbury. 2020. Estimates of common milkweed (Asclepias syriaca) utilization by monarch larvae (Danaus plexippus) and the significance of larval movement. Journal of Insect Conservation. doi: 10.1007/s10841-019-00213-2

Abstract

The population of monarch butterflies east of the Rocky Mountains has noticeably declined over the past two decades. The decline is due, in part, to loss of breeding and forage habitat in the Southern and Midwestern US. To support a resilient overwintering population of six hectares of occupied forest canopy, approximately 1.6-1.8 billion additional ramets of milkweed are needed in the summer breeding range. Milkweed establishment that facilitates natural behavior of monarchs is necessary for effective conservation restoration. This study explored the effect of milkweed ramet density on larval search behavior, milkweed utilization, and survival without predation, parasitism, or competition. 21

Under our experimental greenhouse conditions, monarch larvae abandoned their natal ramet, and subsequent ramets, prior to the pre-pupal wandering stage and before all available leaf biomass on a ramet was consumed. This is consistent with previous field observations. Larvae consumed biomass from three or four milkweed ramets that totaled the approximate biomass of single 10-35 cm ramet. Movement behavior suggests that isolated ramets may not support development through pupation, even though an isolated ramet could provide enough biomass.

Our results suggest milkweed patches containing at least two to four ramets of closely-spaced common milkweed would provide sufficient biomass for development and increase the likelihood that larvae moving in random directions would encounter non-natal ramets to support development. Larval movement behavior and biomass requirements are critical aspects of monarch larval biology that should be considered in habitat restoration and maintenance plans, monitoring survey designs and protocols, and population modeling.

Introduction

Monarch butterflies (Danaus plexippus) east of the Rocky Mountains are an iconic species known for their annual, tri-country migration. The hectares (ha) of forest surface area occupied by congregations of butterflies is an indirect measure of monarch overwintering population (Garcia-Serrano et al. 2004). From the winter of 2003-2004 through 2018-2019, the average overwintering population was 3.5 ha, which is below an estimated resilient population size of 6 ha (Brower et al. 2012; Oberhauser et al. 2017). Current conditions pose a quasi- extinction risk to the annual migration phenomenon (Semmens et al. 2016). Causes of the decline include loss of overwintering habitat in Mexico, loss of breeding and forage habitat in the

Southern and Midwestern US, and extreme weather conditions (Brower et al. 2012; Inamine et al. 2016; Thogmartin et al. 2017). 22

Breeding success, and therefore the size of the overwintering population, is dependent on the status of milkweed species (Asclepias sp.), in particular common milkweed (Asclepias syriaca) (Malcolm et al. 1993; Geest et al. 2019), which are obligate host plants for oviposition and larval development. Because of urbanization and conversion of grasslands to row crop agriculture with more efficient weed management technologies, common milkweed abundance has declined in the Midwest U.S. (Hartzler 2010; Pleasants and Oberhauser 2013; Pleasants

2017). Approximately 1.6-1.8 billion additional milkweed ramets are needed to support an overwintering population goal of 6 ha of occupied forest canopy (Thogmartin et al. 2017). The spatial arrangement and density of new milkweed habitat patches that align with monarch behavior are necessary for effective conservation. The adult female monarch is vagile and simulation studies indicate higher egg densities in the landscape are expected with uniform establishment of small habitat patches, as compared to larger aggregates of habitat widely dispersed (Zalucki & Lammers 2010; Zalucki et al. 2016; Grant et al. 2018). The extent to which increased realized fecundity will result in increased population size depends on larval movement and survival within habitat patches with varying milkweed ramet densities (Zalucki et al 2016;

Zalucki and Kitching 1982c).

Current understanding of larval Lepidoptera dispersal and movement behavior comes largely from observations of species that lay eggs in clusters and whose larvae are polyphagous or oligophagous. With these species, neonates can undergo undirected, long distance dispersals to move away from their siblings by “ballooning” (a behavior where the neonate hangs from the host plant by a strand of silk and is blown downwind; Zalucki, Clarke, & Malcolm 2002;

Goldstein et al. 2010; Razze et al 2011; Razze & Mason 2012). Movement by larger larvae occurs by walking, because their weight negates their ability to balloon (Zalucki, Clarke, & 23

Malcolm 2002). For example, both 4-day and 7-day old tobacco budworm (Heliothis virescens) were observed to abandon their natal transgenic cotton plant and move to adjacent plants (Parker

& Luttrell 1999). Over 15 days, western bean cutworm (Striacosta albicosta) larvae traveled up to 3.54 m to a new corn plant (Pannuti et al. 2016). In general, dispersal behavior is observed with species capable of rejecting their natal host plant in search of more suitable or higher quality plant species (Zalucki, Clarke, & Malcolm 2002); e.g., the bay checkerspot butterfly

(Euphydryas editha bayensis) switches host species based on weather and plant conditions

(Hellman 2002).

The extent to which observations of larval movement and natal plant abandonment with poly/oligophagous species, which lay eggs in clusters, can be extrapolated to the monarch, a monophagous species that lays individual eggs on its host plant, is unclear. Long-range monarch neonate dispersal has not been reported. Numerous field studies suggest third, fourth, and early fifth instars commonly abandon their natal common milkweed ramet and move to co-located milkweed (Urquhart 1960; Rawlins and Lederhouse 1981; Borkin 1982; Zalucki and Rochester

2004; DeAnda and Oberhauser 2015). The extent to which movement observed in the field is due to top-down or bottom-up drivers is unknown. In addition, the interplay of ramet density, larval movement behavior, leaf consumption rates, developmental rates, and survival rates have not been examined experimentally.

The present study was designed to improve understanding of monarch larvae utilization of common milkweed and movement between milkweed ramets in a greenhouse setting without predation, parasitism, and intra- or interspecies competition. In a series of experiments, we quantified biomass consumed, number of leaves with feeding, and plants visited through development when larvae were provided two, three, or four ramets of common milkweed. We 24 hypothesized that prior to natal ramet abandonment, there would be no differences in larval behavior, consumption of biomass, and number of leaves with feeding injury, despite the number of available ramets. Assuming that larval movement is a correlated random walk, we further hypothesized that larvae would find new host plants more successfully, have a higher cumulative survival rate from neonate to pupation, and shorter developmental times when larvae had access to an increasing number of milkweed ramets. During these experiments, we also collected ancillary data to inform potential motivation for ramet abandonment.

Methods

This study assessed the relationship between milkweed ramet availability and monarch larval behavior and survival. In a preliminary study conducted in 2016, 33 neonates were placed in cages with single stems of common milkweed and observed until pupation. When larvae were observed off their milkweed ramet, they were placed on a new ramet to simulate finding a new plant. All 33 larvae abandoned milkweed ramets 2-3 times during development. Given these consistent observations with a single ramet, the present study employed an experimental design with 2, 3, or 4 caged milkweed ramets to assess if there is an effect of milkweed density on larval survival, movement, and milkweed biomass consumption.

Plants

Common milkweed seed was collected from plants around Ames, Iowa USA in 2016 and

2017. Seeds were stratified by placing a mix of cleaned seed with sand, vermiculite, and water in a cold room (0:24 L:D; 4°C) for at least 6 weeks. Seeds were planted in 128-cell plug trays

(Landmark Plastics, Akron, OH) containing soil (Sungro Professional Growing Mix SS#1-F1P,

Agawam, MA) and time-release fertilizer (Osmocote Pro 19-5-8, Dublin, OH). When seedlings had 4-6 leaves, they were transplanted into 8.9 cm square deep perennial pots (Kord, Ontario,

Canada). Seedlings and plants were maintained in a greenhouse (sodium light-augmented 16:8 25

L:D cycle, approximately 56% RH, windows opened when temperatures exceed 21°C). Plants were watered twice per day. Experiments were initiated approximately 8 weeks after planting when ramet height was between 10 and 35 cm, consistent with monarch oviposition preferences

(Urquhart 1960; Bergstrom et al. 1995; Fischer et al. 2015).

Insects

The United States Department of Agriculture (USDA), Agriculture Research Service

(ARS), Corn Insects and Crop Genetics Research Unit (CICGRU) in Ames, IA maintains multiple colonies of monarch butterflies. Each year a new colony is established from field collected eggs. Neonates from the colony established the year prior to experimentation were used for each trial to reduce effects of colony inbreeding (i.e., 2016 or 2017 colonies were using in 2017 and 208 experiments). Experiments were initiated with neonates acquired within five hours of hatching.

Experimental Design

Two, three, or four individually-potted common milkweed ramets were placed at least 16 cm from each other in cages created from mesh pop-up laundry baskets (57 x 37 x 55 cm; Honey-

Can-Do HMP-03891 Mesh Hamper with Handles, Walmart, Rogers, AK) and “no-see-em” netting

(Arrowhead Fabric Outlet, Duluth, MN) under greenhouse conditions (Fig 1). Ramets were labeled to document larval movement among plants. Milkweed ramets were measured from soil surface to new leaf growth and fully extended leaves were counted (median = 14 leaves per ramet). Leaves with signs of injury or disease were removed prior to initiation of an experiment.

Pots were buried in potting soil to provide a uniform surface at the base of the ramets.

Three treatments consisted of cages with two, three or four milkweed ramets. Cages were distributed in a random complete block design, where each block (rows on greenhouse benches) contained three replicates of each treatment. In 2017, four trials with three blocks were 26 conducted (108 total cages). In 2018, three trials were conducted. Because of plant availability, the first trial contained three blocks, the second trial contained nine blocks, and the third trial contained four blocks (144 total cages).

Monarch neonates were added to cages in random order. One neonate was randomly placed on the top portion of one ramet in each cage; this ramet was termed the “natal ramet”.

Twice daily (at approximately 0600 h and 1600 h), cages were observed for larvae. If a larva was found dead, the cage was terminated. If a larva could not be found, the cage was monitored during the next two observation periods. A larva missing a total of three consecutive observation periods, or 1.5 days, was assumed dead and the cage terminated.

Trials in 2017 focused on larval ramet abandonment behavior. Ramet occupied, instar, and molting status were recorded until pupation. Twenty-four hours after pupation, pupae were collected and weighed. Day of ramet abandonment was defined as the day when the larva was first observed off the ramet.

Trials in 2018 quantified monarch plant utilization and biomass consumed. Twice daily

(at approximately 0600 h and 1600 h), ramet occupied, portion of the ramet occupied (top, middle, or bottom third), surface of the leaf occupied (new leaf growth, top side of leaf, underside of leaf), instar, and molting status were recorded until larvae moved from their natal ramet. When larvae were first observed off the natal ramet, the condition of each plant in the cage was ranked based on the percent of leaves consumed (see Table 1) and the height and number of full leaves were recorded. All leaves with signs of feeding were removed from the ramet, photographed with a Kodak Easyshare camera (model # C1550), and stored in a drying oven (30°C, 0% RH). If 90% of the leaf or more was consumed, a leaf from a non-experimental milkweed ramet of comparable size to the missing leaf was used to estimate surface area 27 consumed and mass of the consumed leaf (base of the leaf with feeding was matched to the base of a non-experimental full leaf without feeding). Based on results from the 2017 study, there was typically insufficient biomass remaining in cages with 2 or 3 plants after leaves with evidence of feeding were removed. Therefore, after abandonment of a natal ramet, cages containing two or three ramets were terminated. Larvae in cages with four ramets were monitored until pupation. Every time a larva abandoned a ramet, the height of plants in the cage were measured and number of full leaves were recorded. Leaves with feeding were removed, photographed, and dried. Some fifth instars consumed entire leaves and it was not possible to determine the size of missing leaves. These leaves were reported as consumed, but biomass was not estimated. Twenty-four hours after pupation, pupae were collected and weighed and the height of milkweed ramets measured and fully extended leaves counted. Leaves collected with evidence of feeding were dried at room temperature for at least one week. Leaf material was subsequently weighed to estimate dry leaf mass (mg) not consumed per leaf.

Biomass consumed was calculated using ImageJ Software (Rasband 2018; Schindelin et al. 2012) to estimate leaf area consumed (cm2). Monarch larvae feed on milkweed leaves in predictable patterns (leaf tips, notches out of the side of a leaf, holes in the center of a leaf) and all milkweed leaves have generally the same shape, therefore the area consumed was estimated by measuring the area to fill in the leaf. Dry mass per cm2 was calculated for each leaf based on the dry leaf mass and the computed leaf area not consumed. An estimate of dry biomass consumed per leaf was determined by multiplying the mass per cm2 by the total area consumed calculated with ImageJ.

Statistical Analysis

Characteristics of natal ramet abandonment for individuals that survived to natal abandonment in 2017 and 2018 were analyzed with generalized linear models in RStudio version 28

1.0.153 (RStudio Team 2016) using the package emmeans. Variables to test our additional hypotheses (time observed on plant material, survival rates, pupal weights, and developmental times) were analyzed similarly with data collected in 2017. Generalized linear models accounted for trial (with year embedded), block, and treatment; trial, block, and their interactions showed no effect at a 0.05 level of significance. In all analyses, data residuals were normally distributed and appropriately dispersed. Differences in instar survival rates were evaluated with a chi-square analysis in RStudio version 1.0.153 (RStudio Team 2016).

Ramet abandonment was explored with all individuals that survived to pupation in 2017.

Number of plants visited through complete development and total number of ramets abandoned, including the natal abandonment, were summarized. Generalized linear models were used to understand the effect of trial, block, and ramet treatments on these characteristics. Timing of movements (day vs. night) were analyzed with movement data from 2017 and 2018 using a

Wilcoxan sign rank test in RStudio 1.0.153 (RStudio Team 2016).

With individuals from 2018, estimates of ramet utilization were quantified. Portion of ramet utilized (top, middle, or bottom third), portion of leaf utilized (top, underside, new growth, or stem), biomass consumed prior to natal ramet abandonment, number of leaves with feeding prior to natal abandonment, and plant condition rank (Table 1) at abandonment were quantified.

Biomass consumption was analyzed by instar at time of abandonment using generalized linear models. For larvae provided four plants in 2018, number of plants with signs of feeding and number of leaves with feeding were summed within a cage. Biomass consumption estimates through complete development were made for nine larvae. Two sample t-tests assuming equal variance were conducted in RStudio 1.0.153 (RStudio Team 2016) to determine if larvae consumed more leaves and biomass from their natal ramet or from subsequently visited ramets. 29

Results

Survival and Developmental Endpoints

Number of ramets had no significant effect on the number of individuals surviving to pupation (F = 2.494; df = 2, 155; P = 0.0826). Across both years, 67% (105/156) of individuals survived to pupation (74% in 2017; 52% in 2018 with four ramet cages; see Table 2). Of the 51 individuals that did not survive to pupation, 41 were observed as dead bodies and 10 were assumed dead because they were missing for three consecutive observation periods. In both cases, mortality was assigned to the instar stage when last observed alive. Mortality rates generally declined with developmental stage, but were not significantly different (X-squared =

0.88198; df = 4; P = 0.9271); 10% died in the first instar, 8.3% died in the second instar, 4.5% died in the third instar, 6.4% died in the fourth instar, and 3.2% died in the fifth instar. Similar trends were noted when data were examined by year and number of ramets per cage (see Table

2).

Based on the experiments conducted in 2017, number of ramets present in the cage had no effect on neonate to pupation development time (F = 0.096; df = 2, 76; P = 0.9080), pupal weight (F = 0.031; df = 2, 75; P = 0.9693), length of time in the pupal stage (F = 0.053; df = 2,

74; P = 0.9485), or the developmental time from neonate to adult eclosion (F = 0.113; df = 2, 74;

P = 0.8929). Development from neonate to pupation took 11.6 ± 1.60 (sd) days. Pupae weighed

1.31 ± 0.188 g and pupal duration lasted 9.65 ± 1.21 days. Time from neonate to eclosion was

21.3 ± 2.44 days.

Natal Ramet Abandonment

As noted previously, in the 2016 preliminary study all 33 instars abandoned their natal ramets. Across 2017 and 2018 trials, all larvae that did not die (n=162 out of 252) abandoned their natal ramet. There was no significant effect of number of ramets on number of days 30 elapsed prior to natal ramet abandonment (F = 0.201; df = 2, 161; P = 0.8179) or on instar at ramet abandonment (F = 0.008; df = 2, 161; P = 0.9921). Larvae abandoned the natal ramet after an average of 6.73 ± 2.04 (sd) days, 6.33 ± 2.06 days, and 6.62 ± 2.53 days for cages with two, three, or four ramets, respectively. Abandonment occurred most frequently in the fourth instar

(mode = 4; mean = 4; sd = 1; max = 5, min = 1). At natal ramet abandonment in 2018, the average ramet condition rank was 2.35 ± 1.36 (min=0, max=4); i.e., on average approximately

25-50% of the leaf material remained at natal ramet abandonment.

Larval Movement Behavior

Larvae were observed more often on the top portion of the ramet than on the middle or the bottom (Fig 2a; Z < -3.013; df = 2, 254; P < 0.0073). Larvae also were observed more often on the underside of a leaf than on the top side of the leaf, on the stem, or on new growth (Fig 2b;

Z < -5.037; df = 3, 339; P < 0.0001). After larvae abandoned their natal ramet and successfully found another, they continued to abandon ramets until pupation. The number of ramets had no effect on total ramet abandonments (F = 0.004; df = 2, 76; P = 0.9961); on average larvae abandoned ramets 3.00 ± 1.26 times (min=1, max=6, mode=2). Most post-natal ramet abandonment occurred during the fifth instar (Fig 3; Z < -6.473; df = 4, 384; P < 0.0001), and more of these movements occurred during the day than during the night (mean = 2.37 during day, 1.23 during night; P = 4.275e-07).

Plant Utilization

Number of ramets in a cage had no effect on biomass consumption prior to natal ramet abandonment (F = 0.086; df = 2, 73; P = 0.9179) or on the number of leaves with feeding injury prior to ramet abandonment (F = 1.766; df = 2, 78; P = 0.1710). Though larvae abandoned their natal ramet most frequently in the 4th instar, some abandoned and 2nd, 3rd, or 5th instars. The amount of biomass consumed prior to ramet abandonment was dependent on instar stage at 31 abandonment (F = 15.213; df = 3, 73; P < 0.0001). On average, second or third instars that abandoned their natal ramet consumed less than those that abandoned their natal ramet in the fourth or fifth instar (Fig 4a; Z < 3.338; df = 3, 73; P < 0.0047). Those that abandoned in the fifth instar consumed significantly more than any other instar (Fig 4a; Z > 3.971; df = 3,73; P <

0.0004). Those that abandoned their natal ramet in the second, third, fourth or fifth instar consumed 9.03 ± 4.34 mg (sd), 21.7 ± 12.5 mg, 99.6 ± 71.9 mg, and 732 ± 524 mg of dry ramet material, respectively. Similarly, number of leaves on the natal ramet with feeding injury depended on instar stage at abandonment (F = 12.899; df = 3, 78; P < 0.0001). More leaves on the natal ramet had feeding injury when larvae remained on their natal ramet until fifth instar

(Fig 4b; Z > 3.882; df = 3,78; P < 0.0006). Larvae that abandoned the natal ramet in the second, third, fourth, and fifth instar fed from 3.75 ± 1.49 leaves (sd), 3.71 ± 1.86 leaves, 5.87 ± 2.14 leaves, and 10.7 ± 5.02 leaves on the natal ramet, respectively.

In 2018, neonates placed in cages with four ramets that survived to pupation consumed

1209.61 ± 412.30 mg of total dry plant material (n=9; note 21 larvae survived to pupation, but consumed entire leaves, which precluded the means of estimating consumed plant material).

Biomass was consumed from 17.11 ± 5.09 leaves. Most total biomass was consumed through development (t = -6.7208; df = 10.626; P = 3.891e-05) and leaves with feeding through development (t = -3.4293; df = 14.942; P = 0.003786) were from non-natal ramets. Larvae consumed 119 ± 165 mg of biomass from 5.78 ± 2.94 leaves of their natal ramet and 1090 ± 401 mg from 11.33 ± 3.87 leaves on ramets visited after they abandoned their natal ramet.

In 2017, larvae provided 2, 3, or 4 ramets visited an average or 1.96, 2.33, and 2.93 ramets, respectively; larvae visited more ramets when more ramets were provided, but did not always visit all ramets provided. On average, larvae returned to an already abandoned ramet 32

0.86 ± 0.90 times. Those that were provided four ramets in 2018 fed on an average of 3.22 ± 0.44 ramets (max=4, min=3). When larvae were presented with four ramets, they were more frequently observed on plant material in comparison to “non-plant material” (side of cage, soil, plant labels) than when larvae were provided two ramets (Fig 5; Z = -3.417; df = 2, 155; p =

0.0018).

Discussion

The present study explored the effect of milkweed ramet availability on monarch larval survival and behavior. We observed no effect of milkweed ramet availability on survival rates;

67% of individuals survived to pupation. Monarch larval survival has been estimated in several field studies. Nail et al. (2015) estimated 7-10% survival from egg to fifth instar while Geest et al. (2019) estimated 14-30% survival from egg to adult. The higher survival rate in our study is likely attributed to the exclusion of predators and parasitoids. Our instar specific survival was similar to that observed by Malcolm (1994) in outdoor natural enemy exclusion experiments.

Larval milkweed abandonment is documented in several field studies (Urquhart 1960;

Rawlins and Lederhouse 1981; Borkin 1982; Zalucki and Rochester 2004; De Anda and

Oberhouser 2015). In our 2016 preliminary greenhouse study, we observed natal ramet abandonment through larval development with each of 33 larvae individually placed on isolated milkweed ramets. We hypothesized that larvae would abandon their natal ramet regardless of the number of co-located ramets. Results of the 2017 and 2018 experiments, with 2, 3, or 4 ramets in a cage indicated all larvae abandoned their natal ramet prior to completion of larval development. On average, larvae abandoned their natal ramet during the fourth instar, i.e, on the sixth or seventh day of development.

Natal ramet abandonment did not occur because all available biomass was consumed; approximately 25-50% of the leaf material remained at the time of natal ramet abandonment. 33

Natal ramet biomass consumption rates were similar regardless of the number of ramets in the cage; however, those instars that remained on their natal ramet later in development consumed more biomass from more leaves than larvae that abandoned the natal ramet earlier in development. Early instars fed from four leaves while later instars fed from as many as 11 leaves.

Larvae that abandoned a ramet were likely in search of a new milkweed ramet; however, they probably are not able to detect the presence of surrounding milkweed. Larval perceptual range is estimated to be less than 5 cm (Urquhart 1960). Previous reports suggest that monarch larvae move without orientation (Urquhart 1960; Rawlins and Lederhouse 1981; Borkin 1982).

Our study further supports a random walk behavior. If movement was directed, larvae would detect the presence of, and move directly to, a new ramet after abandonment. In contrast, we observed an association between number of ramets present in the cage and number of times monarch larvae were observed on a ramet. Larvae were observed more frequently on plant material when more ramets were provided, presumably because there were more opportunities to randomly find a host plant. Likewise, larvae returned to milkweed ramets that were previously abandoned, suggesting that the movement was not directed.

Consequently, we hypothesized that larvae with access to more ramets would have higher survival rates and pupal weights with shorter developmental times because they would spend more time on plant material. There was, however, no significant effect of ramet number, which. suggests even with a random walk, larvae were likely not off a ramet for a sufficient time to show any effect of reduced feeding. This observation may be due to the distance between ramets

(16 cm) in our experimental design. Distances monarch larvae are capable of moving when searching for a new ramet remains uncertain with only Borkin’s (1982) observation of a larva 34 that traveled 2.25 m before finding a new ramet. If our experimental design was repeated using larger cages with a greater distance between ramets, we hypothesize that survival rates and pupal weights would increase and/or developmental times would decrease as the number of ramets increased.

After larvae successfully found a new ramet, they continued to abandon milkweed ramets up to five more times. Larvae typically returned to an already abandoned ramet once during development. These movements were not an artifact of the wandering stage prior to pupal formation (Truman and Riddiford 1974) because larvae continued feeding when they arrived on new ramets. Most of the total biomass and leaves consumed through development were from ramets visited after natal ramet abandonment. Therefore, subsequent ramets may be more important biomass for development. Larvae visited more ramets when more were provided, but did not always visit all that were available. We estimated that a larva consumed approximately the same mass as one 11.0-34.1 cm milkweed ramet (1726 mg dried weight ± 829 mg), but they acquired this mass from three or four separate ramets.

Field observations suggest monarch females can ‘dump’ eggs on isolated ramets of milkweed. Isolated milkweed ramets can contain up to ten monarch eggs at one time (Zalucki and Kitching 1982b). Larvae on isolated stems are also thought to have greater survival rates because of reduced predation (Zalucki and Kitching 1982a; Nail et al. 2015). Based on our observations of natal ramet abandonment, it is unlikely that multiple eggs on one ramet would develop to multiple adult butterflies unless additional ramets were within close proximity. Even if larvae did not abandon their natal ramet, there would likely be insufficient biomass with spring plants to support more than one larva, assuming no inter-species competition or predation, and the size and age of milkweed are similar to those used in the present study. Interactions among 35 multiple larvae on an isolated ramet have yet to be explored. Studies in progress are assessing the effect of intra-species competition on characteristics of larval movement, survival, growth, and development.

While the causes of observed larval movement in field studies may be due to intra- or inter-specific competition, avoidance of predators and/or parasitoids, and/or insufficient forage, these environmental factors were not present in our experiments. We consistently observed natal ramet abandonment over three years of experiments with 1, 2, 3, or 4 ramets and three different monarch cohorts. Our findings suggest abandonment of the natal ramet is an innate behavior; it occurs consistently, predictably and without prior larval experience. What environmental cue(s) could be responsible for triggering this behavior?

Reduced plant quality due to feeding could cause a negative taxis response. Leaf quality changes as leaves mature. Younger leaves may be more readily digested than older leaves and have higher concentrations of nitrogen and water. Herbivorous insects that feed on young leaves have higher survival, faster development, and larger mass than those that feed on older leaves

(Stamp & Bowers 1990; Jordano & Gomariz 1994 and references therein). Reports also suggest monarch females prefer to lay eggs on young milkweed plants with new leaves (Urquhart 1987;

Bergstrom et al. 1995; Fischer et al. 2015). Our findings suggest that monarch larvae may prefer new growth over older leaf material; larvae may have abandoned their natal ramet in search of plants with new growth. Larvae were observed significantly more times on the top third of the milkweed ramet, where new growth is present, than on the middle or bottom third of the ramet.

Often, leaves from the top third of a ramet were consumed at larval abandonment, with the older leaves remaining on the bottom of the plant (Fig 6). 36

Monarch larvae perform best on milkweed species that contain intermediate cardenolide concentrations (Agrawal et al. 2014; Agrawal 2017). Some cardenolides are constitutive and others can be induced within five days of monarch feeding (Malcolm 1994; Malcolm & Zalucki

1996; Van Zandt and Agrawal 2004; Rasmann et al. 2009; Agrawal et al. 2012, 2014). In our study, monarch larvae abandoned their natal ramet after six to seven days, aligning with the time to induce cardenolide production. Though increased milkweed leaf cardenolide concentrations due to herbivory will not cause increased mortality rates (Agrawal et al. 2014), feeding on these leaves could have a fitness cost. Milkweed species with high cardenolide concentrations produce smaller pupae than species low in cardenolides (Agrawal 2017). In our studies, it is possible induced cardenolides reached levels that invoked negative chemotaxis. Future studies assessing larval movement on and between milkweed plants, with and without prior herbivory and including measurements of cardenolide concentrations, would help determine if there is a threshold concentration that elicits natal ramet abandonment.

Alternatively, the consistent observations of natal ramet abandonment could indicate negative chemotaxis to “call for help” compounds released by the milkweed in response to herbivory. While we are not aware of any studies documenting release of compounds by milkweed to attract predators and parasites in response to feeding damage, this response has been reported with other plants (e.g., see Turlings et al 1995; Pare and Tumlinson 1999; Gershenzon

2007; War et al. 2012; Aljbory and Chen 2018). Monarch larvae may also be capable of detecting these secondary plant compounds, which in turn elicits natal ramet abandonment.

Consistent natal ramet abandonment could also indicate negative tactile taxis due to reduced leaf cover with increased herbivory. Also noted previously, 100% of the larvae in the current study abandoned their natal ramet, which ranged between 10-35 cm in height. 37

Abandonment of the natal ramet occurred after 25 to 50% of the leaves had been consumed.

Larvae also were observed most frequently on the underside of the leaf. Similar observations have been reported by Rawlins and Lederhouse (1981). Loss of leaf material results in less cover ‘to hide’ from predators and parasitoids. Consequently, with less foliage for camouflage, larvae could be more easily detected by predators/parasitoids because these species often search for frass or feeding injury (Price et al 1980; De Mores et al. 1998; Oberhauser et al. 2015). In response to predation/parasitism, natural selection may have favored natal ramet abandonment.

Leaving a milkweed ramet is risky. However, abandoning a milkweed and successfully finding another, may be associated with lower rates of predation/parasitism and a greater probability of development to a reproductively viable adult, as compared to rates of mortality for larvae remaining on the natal ramet. Future studies with larger plants and more leaves could assess the extent to which abandonment is correlated with a decreased amount of leaf cover.

Regardless of the mechanism(s) that invoke this seemingly innate movement behavior, it has implications for larval monitoring projects. Larvae in our studies abandoned plants an average of three times through development and this movement occurred most often during the day. Most monitoring protocols are based on observing milkweed leaves and ramets for eggs and larvae during the day (Prysby and Oberhauser 2004; De Anda and Oberhauser 2015; Nail et al. 2015; Geest et al. 2019). In these protocols, when larvae are expected in a subsequent sampling period and not found, they are presumed to be dead (Prysby and Oberhauser 2004; Nail et al. 2015; Geest et al. 2019). When a larva is observed but there was no detection in the previous sampling period, data records are adjusted under the assumption of missed detection in the prior sampling period (De Anda and Oberhauser 2015; Nail et al. 2015; Geest et al. 2019).

Although these studies account for failed detection when new larvae are observed, they did not 38 account for larvae moving out of the patch. Survey design should account for imperfect larval detection, which is often disregarded when studying invertebrates because of their general abundance in comparison to vertebrates (Kellner and Swihart 2014). For larval monarch surveys, our results suggest current monitoring protocols may over estimate field mortality and do not account for larval movement into and out of habitat patches. Monitoring the ground and vegetation surrounding milkweed ramets as well as searching ramets for standardized durations of time during observation periods may reduce bias in survey results.

Larval movement behavior and biomass requirements are critical aspects of monarch larval biology that should be incorporated in habitat restoration and maintenance practices.

Because the eastern North American monarch population grows exponentially over the summer prior to fall migration, it is important that the first generation in the Upper Midwest is productive

(Flockhart et al. 2015; Oberhauser et al. 2017). Under our greenhouse conditions and consistent with previous field studies, monarch larvae unfailingly abandon their natal ramet of common milkweed, and subsequent ramets, before all of the available leaf biomass on a ramet is consumed and prior to the pre-pupal wandering stage. In the spring, when monarchs first arrive in the Upper Midwest, milkweed are 10-35 cm in height and likely provide sufficient milkweed biomass to support development of a single egg to pupation. However, because larvae abandon their natal ramets, if the natal ramet is isolated from other milkweed plants the larvae would likely die before finding another plant during a random walk.

In conclusion, while previous modeling studies indicate that isolated ramets in the matrix can increase the number of eggs laid in the landscape (Zalucki et al. 2016), our findings suggest that eggs or larvae observed on isolated ramets will have low survival. Patches containing at least two to four ramets of closely-spaced common milkweed provide sufficient biomass for 39 development and would increase the likelihood that larvae moving in random directions would encounter non-natal milkweed ramets to support development through pupation.

Data Accessibility

Data, metadata, and R code pertaining to this manuscript are publicly available through

GitHub: https://github.com/kelseyefisher/larvalmonarchmovementandcompetition

Acknowledgements

This work was supported, in part, by USDA-NRCS/CIG Agreement 69-3A75-16-006 and the Iowa State University, College of Agriculture and Life Sciences. We would like to thank

Keith Bidne (USDA, ARS-CICGRU) for maintaining monarch butterfly colonies and providing materials and larvae used in these experiments. We thank Julia Pfeiffer, Signey Hilby, Riley

Nylin, Cody Acevedo, Jenna Nixt, and Kara Weber for their help with experimental set up and data collection; without their help twice daily data collection would not have been possible.

Additional thanks to ISU Statistics Consulting, Kathleen Rey, Audrey McCombs, and Dr. Philip

Dixon. Mention of a proprietary product does not constitute an endorsement or a recommendation for its use by Iowa State University or USDA.

References

Agrawal AA (2017) Hatching and defending. In: Monarchs and milkweed: a migrating butterfly, a poisonous plant, and their remarkable story of coevolution. Princeton University Press, New Jersey, pp 90-118

Agrawal AA, Hastings AP, Patrick ET, Knight AC (2014) Specificity of herbivore-induced hormonal signaling and defensive traits of five closely related milkweeds (Asclepias spp.). J Chem Ecol. http://doi.org/10.1007/s10886-014-0449-6

Agrawal AA, Petschenka G, Bingham RA, Weber MG, Rasmann S (2012) Toxic cardenolides: chemical ecology and coevolution of specialized plant-herbivore interactions. New Phytol. https://doi.org/10.1111/j.1469-8137.2011.04049.x

Alijbory Z, Chen MS (2018) Indirect plant defense against insect herbivores: a review. Insect Sci. https://doi.org/10.1111/1744-7917.12436

Bergstrom G, Rothschild M, Groth I, Crighton C (1994/1995) Oviposition by butterflies on young leaves: investigation of leaf volatiles Chemoecology 5/6: 147-158

Borkin SS (1982) Notes on shifting distribution patterns and survival of immature Danaus plexippus (Lepidoptera: Danaidae) on the food plant Asclepias syriaca. Great Lakes Entomol 15(3): 199-206

Brower LP, Taylor OR, Williams EH, Slayback DA, Zubieta RR, Ramirez MI (2012) Decline of monarch butterflies overwintering in Mexico: in the migratory phenomenon at risk? Insect Conserv Diver. https://doi.org/ 10.1111/j.1752-4598.2011.00142.x

De Anda A, Oberhauser KS (2015) Invertebrate natural enemies and stage-specific mortality rrates of monarch eggs and larvae. In: Oberhauser KS, Nail KR, Altizer S (eds) Monarchs in a changing world. Cornell University Press, New York, pp 60-70

De Moraes CM, Lewis WJ, Pare PW, Alborn HT, Tumlinson JH (1998) Herbivore-infested plants selectively attract parasitoids. Nature 393: 570-573

Fischer SJ, Williams EH, Brower LP (2015) Enhancing monarch butterfly reproduction by mowing fields of common milkweed. Am Midl Nat. https://doi.org/10.1674/amid-173- 02-229-240.1

Flockhart DT, Pichancourt JB, Norris DR, Martin TG (2015) Unravelling the annual cycle in a migratory animal: breeding-season habitat loss drives population declines of monarch butterflies. J Anim Ecol http://doi.org/10.1111/1365-2656.12253

Garcia-Serrano E, Reyes JL, Alvarez BXM (2004) Locations and area occupied by monarch butterflies overwintering in Mexico from 1993 to 2002. In: Oberhauser KS, Solensky MJ (eds) The monarch butterfly biology and conservation, 1 st edn. Cornell University Press, New York, pp 129-133

Geest EA, Wolfenbarger LL, McCarty JP (2019) Recruitment, survival, and parasitism of monarch butterflies (Danaus plexippus) in milkweed gardens and conservation areas. J Insect Conserv. https://doi.org/10.1007/s10841-018-0102-8

Gershenzon J (2007) Plant volatiles carry both public and private messages. PNAS. https://www.pnas.org/content/104/13/5257

Goldstein JA, Mason CE, Pesek J (2010) Dispersal and movement behavior of neonate European corn borer (Lepidoptera: Crambidae) on non-bt and transgenic bt corn. J Econ Entomol. http://doi.org/10.1603/ec09304

Grant TJ, Parry HR, Zalucki MP, Bradbury SP (2018) Predicting monarch butterfly (Danaus plexippus) movement and egg-laying with a spatially-explicit agent-based model: The role of monarch perceptual range and spatial memory. Ecol Modell. https://doi.org/10.1016/j.ecolmodel.2018.02.011 41

Hartzler RG (2010) Reduction in common milkweed (Asclepias syriaca) occurrence in Iowa cropland from 1999 to 2009. Crop Prot. https://doi.org/10.1016/j.cropro.2010.07.018

Hellman JJ (2002) The effect of an environmental change on mobile butterfly larvae and the nutritional quality of their hosts. J Anim Ecol. https://doi.org/10.1046/j.1365- 2656.2002.00658.x

Inamine H, Ellner SP, Springer JP, Agrawal AA (2016) Linking the continental migratory cycle of the monarch butterfly to understand its population decline. Oikos. https://doi.org/10.1111/oik.03196

Jordano D, Gomariz Gines (1994) Variation in phenology and nutritional quality between host plants and its effect on larval performance in a specialist butterfly, Zerynthis rumina. Entomol Exp Appl. https://doi.org/10.1111/j.1570-7458.1994.tb01794.x

Kellner KF, Swihart RK (2014) Accounting for imperfect detection in ecology: a quantitative review. PLoS One. http://doi.org/10.1371/journal.pone.0111436

Malcolm SB (1994) Milkweeds, monarch butterflies and the ecological significance of cardenolides. Chemoecology 5:101-117

Malcolm SB, Cockrell BJ, Brower LP (1993) Spring recolonization of the eastern North America by the monarch butterfly: successive brood or single sweep migration? In: Malcolm SB, Zalucki MP (eds) Biology and conservation of the monarch butterfly. Natural History Museum of Los Angeles County, California pp 253-267

Malcolm SB, Zalucki MP (1996) Milkweed latex and cardenolide induction may resolve the lethal plant defence paradox. Entomol Exp Appl 80: 193-196

Nail KR, Stenoien C, Oberhauser KS (2015) Immature monarch survival: effects of site characteristics, density, and time. Ann Entomol Soc Am. http://doi.org/10.1093/aesa/sav047

Oberhauser KS, Anderson M, Anderson S, Caldwell W, De Anda AP, Hunter MD, Kaiser M, Solensky MJ (2015) Lacewings, wasps and flies – oh my: insect enemies take a bite out of monarchs. In: Oberhauser KS, Nail KR, Altizer S (eds) Monarchs in a changing world. Cornell University Press, New York, pp 71-82

Oberhauser KS, Wiederholt R, Diffendorfer JE, Semmens D, Ries L, Thogmartin WE, Lopex- Hoffman L, Semmens B (2017) A trans-national monarch butterfly population model and implications for regional conservation priorities. Ecol Entomol. http://doi.org/10.1111/een.12351

Pannuti LER, Paula-Moraes SV, Hunt TE, Baldin ELL, Dana L, Malaquias JV (2016) Plant-to- plant movement of Striacosta albicosta (Lepidoptera: Noctuidae) and Spodoptera frugiperda (Lepidoptea: Noctuidae) in maize (Zea mays). J Econ Entomol. https://doi.org/10.1093/jee/tow042

Pare PW, Tumlinson JH (1999) Plant volatiles as a defense against insect herbivores. Plant Physiol 121:325-331

Parker CD, Luttrell RG (1999) Interplant movement of Heliothis virescens (Lepidoptera: Noctuidae) larvae in pure and mixed plantings of cotton with and without expression of the Cry1Ac δ-endotoxin protein of Bacillus thuringiensis berliner. J Econ Entomol. https://doi.org/10.1093/jee/92.4.837

Pleasants JM (2017) Milkweed restoration in the Midwest for the monarch butterfly recovery: estimates of milkweeds lost, milkweeds remaining and milkweeds that must be added to increase the monarch population. Insect Conserv Diver. https://doi.org/10.1111/icad.12198

Pleasants JM, Oberhauser KS (2013) Milkweed loss in agriculture fields because of herbicide use: Effects on the monarch butterfly population. Insect Conserv Diver. https://doi.org/10.1111/j.1752-4598.2012.00196.x

Price PW, Bouton CE, Gross P, McPheron BA, Thompson JN, Weis AE (1980) Interactions among three trophic levels: Influence of plants on interactions between insect herbivores and natural enemies. Ann Rev Ecol Syst. https://doi.org/10.1146/annurev.es.11.110180.000353

Prysby MD, Oberhauser KS (2004) Temporal and geographic variation in monarch densities: Citizen scientists document monarch population patterns. In: Oberhauser KS, Solensky MJ (eds) The monarch butterfly biology and conservation, 1 st edn. Cornell University Press, New York, pp 9-20

Rasband WS (2018) ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, https://imagej.nih.gov/ij/

Rasmann S, Agrawal AA, Cook SC, Erwin AC (2009) Cardenolides, induced responses, and interactions between above- and belowground herbivores of milkweed (Asclepias spp.). Ecology 90:2393-2404

Rawlins JE, Lederhouse RC (1981) Developmental influences of thermal behavior on monarch caterpillars (Danaus plexippus): An adaptation for migration (Lepidoptera: Nymphalidae: Danainae). J Kans Entomol Soc 54:387-408

Razze JM, Mason CE, Pizzolato TD (2011) Feeding behavior of neonate Ostrinia nubilalis (Lepidoptera: Crambidae) on Cry1Ab bt corn: Implications for resistance management. J Econ Entomol. http://doi.org/10.1603/ec10287 43

Razze JM, Mason CE (2012) Dispersal behavior of neonate European corn borer (Lepidoptera: crambidae) on bt corn. J Econ Entomol. http://doi.org/10.1603/ec11288

RStudio Team (2016) RStudio: Integrated Development for R. RStudio, Inc., Boston, MA URL https://www.rstudio.com/

Schindelin J, Arganda-Carreras I, Frise E, et al (2012) Fiji: an open-source platform for biological-image analysis. Nat Methods 9:676-682

Semmens BX, Semmens DJ, Thogmartin WE, et al (2016) Quasi-extinction risk and population targets for the Eastern, migratory population of monarch butterflies (Danaus plexippus). Sci Rep. https://doi.org/10.1038/srep23265

Stamp NE, Bowers MD (1990) Phenology of nutritional differences between new and mature leaves and its effect on caterpillar growth. Ecol Entomol. https://doi.org/10.1111/j.1365- 2311.1990.tb00827.x

Thogmartin WE, Lopez-Hoffman L, Rohweder J, et al (2017) Restoring monarch butterfly habitat in the Midwestern US: ‘all hands on deck’. Environ Res Lett. https://doi.org/10.1088/1748-9326/aa7637

Truman JW, Riddiford LM (1974) Physiology of insect rhythms III. The temporal organization of the endocrine events underlying pupation of the tobacco hornworm. J Exp Biol. 60:371-382

Turlings TCJ, Loughrin JH, McCall PJ, Rose US, Lewis WJ, Tumlinson JH (1995) How caterpillar-damaged plants protect themselves by attracting parasitic wasps. Proc Natl Acad Sci. 92: 4169-4174

Urquhart FA (1960) The monarch butterfly. University of Toronto Press, Canada.

Urquhart FA (1987) The monarch butterfly: International traveler. Nelson-Hall Inc., Chicago, IL.

Van Zandt PA, Agrawal AA (2004) Specificity of induced plant responses to specialist herbivores of the common milkweed Asclepias syriaca. Oikos. https://doi.org/10.1111/j.0030-1299.2004.12964.x

War AR, Paulraj MG, Ahmad T, Buhroo AA, et al (2012) Mechanisms of plant defense against insect herbivores. Plant Signal Behav. https://doi.org/10.4161/psb.21663

Zalucki MP, Clarke AR, Malcolm SB (2002) Ecology and behavior of first instar larval Lepidoptera. Annu. Rev. Entomol. https://doi.org/10.1146/annurev.ento.47.091201.145220

Zalucki MP, Kitching RL (1982a) Temporal and spatial variation of mortality in field populations of Danaus plexippus L. and D. chrysippus L. larvae (Lepidoptera: Nymphalidae) Oecologia 53:201-207

Zalucki MP, Kitching RL (1982b) Dynamics of oviposition in Danaus plexippus (Insecta: Lepidoptera) on milkweed Asclepias spp. J Zool. https://doi.org/10.1111/j.1469- 7998.1982.tb02063.x

Zalucki MP, Kitching RL (1982c) The analysis and description of movement in adult Danaus plexippus L. (Lepidoptera: Danainae). Behaviour 80:174-198

Zalucki MP, Lammers JH (2010) Dispersal and egg shortfall in monarch butterflies: what happens when the matrix is cleaned up? Ecol Entomol. https://doi.org/10.1111/j.1365- 2311.2009.01160.x

Zalucki MP, Parry HR, Zalucki JM (2016) Movement and egg laying in monarchs: to move or not to move, that is the equation. Austral Ecol. https://doi.org/10.1111/aec.12285

Zalucki MP, Rochester WA (2004) Spatial and temporal population dynamics of monarchs down-under: lessons for North America In: Oberhauser KS, Solensky MJ (eds) The monarch butterfly biology and conservation, 1st edn. Cornell University Press, New York, pp 219-228 Figures and Tables

Figure 1. Experimental cages with common milkweed ramets. a. Two ramets of common milkweed. b. Three ramets of common milkweed. c. Four ramets of common milkweed. Potted ramets were placed approximately 16 cm apart and covered with potting soil to provide a single surface

Figure 2. Relative time monarch larvae were observed on portions of milkweed ramets. Relative time is the number of observations associated with a specific portion or a ramet or leaf surface divided by the total number of observations. a. Observed location of larva on ramet (top, middle, or bottom third). b. Surface of leaf where larvae were observed (underside, top, new growth). Bars represent the mean ± one standard deviation. Different number of asterisks (*) above bars represent a significance value less than 0.05

Figure 3. Average frequency of instar specific ramet abandonments. Bars represent the mean ± one standard deviation. Different number of asterisks (*) above bars represent a significance value less than 0.05

Figure 4. Larval milkweed consumption at natal ramet abandonment. a. Estimated natal milkweed ramet consumed by instar. b. Number of natal ramet leaves with feeding injury by instar. Bars represent the mean ± one standard deviation. Different number of asterisks (*) above bars represent a significance value less than 0.05

Figure 5. Proportion of larval observations on milkweed ramet (number of observations on a ramet divided by number of observations on ramets, soil, and cage surface). Bars represent the mean ± one standard deviation. Different number of asterisks (*) above bars represent a significance value less than 0.05

Figure 6. Representative common milkweed ramet after a 4th instar abandonment. Leaves on the top of the ramet are totally consumed, while leaves on the middle and bottom have no evidence of feeding

Table 1. Description of milkweed ramet condition ranking

Rank Ramet Condition 5 No feeding 4 Between 0% and 25% of leaf material consumed 3 Between 25% and 50% of leaf material consumed 2 Between 50% and 75% of leaf material consumed 1 Between 75% and 100% of leaf material consumed 0 All leaves consumed

Table 2. Number (and percent) of individuals that survived to developmental stages by year, number of ramets, and pooled across years and number of ramets

Number of Individuals Developmental 2017 2017 2017 2018 Stage 2 Ramets 3 Ramets 4 Ramets 4 Ramets Pooled 1st Instar 35 (97%) 32 (89%) 33 (92%) 40 (83%) 140 (90%) 2nd Instar 34 (94%) 30 (83%) 31 (86%) 32 (67%) 127 (81%) 3rd Instar 34 (94%) 29 (80%) 29 (80%) 28 (58%) 120 (77%) 4th Instar 32 (89%) 24 (67%) 28 (78%) 26 (54%) 110 (70%) 5th Instar 30 (83%) 22 (61%) 28 (78%) 25 (52%) 105 (67%) Total 36 36 36 48 156

Appendix. Impact of Intraspecific Competition on Larval Monarch (Danaus plexippus) Movement Behavior and Fitness

Introduction

All monarch larvae reared individually on 2, 3, or 4 ramets of common milkweed under greenhouse conditions in the absence of predation abandoned their natal milkweed ramet during the fourth instar (after 6 or 7 days of development), prior to consumption of the entire milkweed ramet and prior to the completion of larval development. For the duration of development, larvae continued to find, feed on, and abandon new milkweed ramets between 2-6 times. This behavior appears to be an innate behavior because it occurs consistently, predictability, and without prior larval experience (Chapter 2 and Fisher et al. 2020). Motivations for this seemingly innate behavior have yet to be explored; however, because monarchs were reared individually, it is not in response to the presence of intraspecific competition.

Field observations suggest monarch females can ‘dump’ eggs on isolated ramets of milkweed, with isolated milkweed ramets containing up to ten monarch eggs at one time

(Zalucki and Kitching 1982a). Larvae on isolated stems are also thought to have greater survival rates because of reduced predation (Zalucki and Kitching 1982b, Nail et al. 2015). The influence of intraspecific competition on ramet abandonment behavior and larval fitness have yet to be explored. Here, we aimed to assess the effect of intraspecific competition on larval movement and fitness under controlled greenhouse conditions in the absence of predation.

Methods

Plants

Common milkweed seed was collected from plants around Ames, IA USA in 2018. Seeds were stratified, planted, and maintained as in Fisher et al. (2020). Experiments were initiated approximately 8 weeks after planting when ramet height was between 10 and 35 cm, consistent 49 with monarch oviposition preferences (Urquhart 1960, Bergström et al. 1994, Fischer et al.

2015).

Insects

The United States Department of Agriculture (USDA), Agriculture Research Service

(ARS), Corn Insects and Crop Genetics Research Unit (CICGRU) in Ames, IA maintains multiple colonies of monarch butterflies (Krishnan et al. 2020). Experiments were initiated with neonates from the colony established in 2017 within five hours of hatching.

Experimental Design

Three trials, each with three treatments and 18 replicates (n=162 neonates), were conducted in a random complete block design on greenhouse benches during the summer of

2019. Four common milkweed (Asclepias syriaca) ramets were measured and placed in the corners of cages created from mesh pop-up laundry baskets (57×37×55 cm; Honey-Can-Do

HMP-03891 Mesh Hamper with Handles, Walmart, Rogers, AK) and “no-see-em” netting

(Arrowhead Fabric Outlet, Duluth, MN) under greenhouse conditions, as in Fisher et al. (2020).

Treatments included: 1) one neonate placed on one of the four milkweed ramets, 2) two neonates placed on diagonally opposite ramets, or 3) two neonates placed on the same ramet. Neonates were assigned to cages in random order. The initial plant to receive the first neonate was selected randomly. Neonates were always placed on individual leaves at the top of the milkweed ramet.

Twice daily (at approximately 0600 h and 1600 h), the ramet occupied, portion of the ramet occupied (top, middle, or bottom third), surface of the leaf occupied (new leaf growth, top side of leaf, underside of leaf), instar, and molting status were recorded for all larvae in a cage until pupation. It was noted when both larvae within a cage were sharing space (i.e., same ramet, same portion of the ramet, same leaf). Twenty-four hours after pupation of the last larva in a cage, pupae were collected, weighed, and pinned in plastic cups. Plant height was measured and 50 number of leaves with and without feeding were counted. Twenty-four hours after eclosion, adults were sexed and weighed. All data can be found on the public github repository: https://github.com/kelseyefisher/larvalmonarchmovementandcompetition

Statistical Analyses

The number of individuals to survive to pupation were analyzed with chi-square test of independence in RStudio (chisq.test) version 4.0.3 (R Core Team 2020, RStudio Team 2020).

Elapsed days until pupation and pupal mass were analyzed with generalized linear models using the ‘emmeans’ package in RStudio. Additional general linearized model analyses will be conducted including the frequency of sharing space (same ramet, same portion of ramet, same leaf), elapsed days until eclosion, adult sex, adult mass, elapsed days until natal ramet abandonment, and frequency of ramet abandonment.

Preliminary Results and Discussion

Survival rates were not influenced by the number of larvae in a cage (χ2 = 6, df = 4, p =

0.1991). At least 72% of larvae survived to pupation in every treatment. This finding suggests that multiple eggs oviposited on the same stem of milkweed would not affect survival of larvae if there are additional milkweed stems near the natal plant. All larvae pupated after approximately

28 observations (14 days). When two larvae were present, independent of treatment, one would weigh less than and one would weigh more than the control (F=17.676, df=4, 170, p < 0.0001;

Fig 7), which suggests that intraspecific competition has a potential influence on larval fitness.

Figure 7. Mean pupal mass (±sd) of the smaller of two larvae initially placed on the same plant, the smaller of the two larvae initially placed on diagonally -opposite plants, the control larvae with only one larva per cage, the larger of the two larvae initially placed on the same plant and the larger of the two larvae initially placed on diagonally-opposite plants. When two larvae were in a cage, independent of where they started initially, one larva would be smaller than the control and one would be larger than the control. Different letters above the bars indicate significant difference at a p = 0.0001.

CHAPTER 3. EMPLOYING VERY HIGH FREQUENCY (VHF) RADIO TELEMETRY TO RECREATE MONARCH BUTTERFLY (Danaus plexippus) FLIGHT PATHS

Kelsey E. Fisher1, James Adelman2,3, and Steven P. Bradbury1,3

1Department of Entomology, Iowa State University, Ames, IA, 50011

2Department of Biological Sciences, The University of Memphis, Memphis, TN 38117

3Department of Natural Resource Ecology and Management, Iowa State University, Ames, IA,

50011

Modified from a manuscript published in Environmental Entomology.

Fisher, KE, J Adelman & SP Bradbury. 2020. Employing very high frequency (VHF) radio telemetry to recreate monarch butterfly (Danaus plexippus) flight paths. Journal of Environmental Entomology. doi: 10.1093/ee/nvaa019

Abstract

The overwintering population of eastern North American monarch butterflies (Danaus plexippus) has declined significantly. Loss of milkweed (Asclepias sp.), the monarch’s obligate host plant in the Midwest U.S., is considered to be a major cause of the decline. Restoring breeding habitat is an actionable step towards population recovery. Monarch butterflies are highly vagile; therefore, the spatial arrangement of milkweed in the landscape influences movement patterns, habitat utilization, and reproductive output. Empirical studies of female movement patterns within and between habitat patches in representative agricultural landscapes support recommendations for habitat restoration.

To track monarch movement at distances beyond human visual range, we employed VHF radio telemetry with handheld antennae to collect movement bearings on a biologically-relevant 53 time scale. Attachment of 220-300 mg transmitters did not significantly affect behavior and flight capability. Thirteen radio-tagged monarchs were released in a restored prairie, and locations were estimated every minute for up to 39 minutes by simultaneous triangulation from four operators. Monarchs that left the prairie were tracked and relocated at distances up to 250 m.

Assuming straight flights between locations, the majority of steps within the prairie were below

50 m. Steps associated with exiting the prairie exceeded 50 m with high directionality. Because butterflies do not fly in straight lines between stationary points, we also illustrate how occurrence models can use location data obtained through radio telemetry to estimate movement within a prairie and over multiple land cover types.

Introduction

Since the early 1990s, the overwintering population of eastern North American monarch butterflies (Danaus plexippus) has declined significantly (Brower et al. 2012; Oberhauser et al.

2017). Many factors contribute to this decline; yet, loss of milkweed (Asclepias sp.), the monarch’s obligate host plant, in the breeding range is generally considered to be a significant cause (Brower et al. 2012; Inamine et al. 2016; Thogmartin et al. 2017). Restoring breeding habitat in the Midwest U.S., including milkweed and nectar sources, is an actionable step towards population recovery (Thogmartin et al. 2017; Pitman et al. 2018). To maximize the efficacy of conservation efforts, milkweed should be established in the landscape in spatial patterns that optimize female reproductive output. Improved understanding of how monarch females move through agricultural landscapes and distribute eggs is needed.

The monarch butterfly is a vagile species (Zalucki and Lammers 2010; Zalucki and

Kitching 1984) that can traverse up to 15 kilometers/day during the breeding season (Zalucki et al. 2016). During their lifetime, females produce 300-400 eggs (Urquhart 1960; Oberhauser

2004) that are typically laid singly per milkweed ramet (Zalucki and Kitching 1982a). Given 54 these behaviors, the spatial arrangement of milkweed in the landscape is expected to influence movement patterns, habitat utilization, and reproductive output (Bergin et al. 2000; Misenhelter and Rotenberry 2000; Pitman et al. 2018; Grant and Bradbury 2019). Higher realized fecundity is predicted when habitat is distributed in many small patches as compared to a smaller number of large patches (Zalucki and Lammers 2010; Zalucki et al. 2016).

Typically, animal movement models are based on correlated random walk algorithms without incorporating behavioral factors (Siniff and Jessen 1969; Smouse et al. 2010; Zhao et al.

2015). These algorithms can be problematic in heterogeneous landscapes that require flight direction decisions at habitat edges in the mosaic of cropland, pastures, conservation land, and roadside rights of way in the monarch’s summer breeding range (Grant et al. 2018). Recently, individual-based models were developed to explore the inclusion of female monarch behavioral factors in simulating landscape-scale movement and egg distribution (Zalucki et al. 2016; Grant et al. 2018). Behavioral factors were derived from expert opinion and limited empirical evidence

(Zalucki 1983; McIntire et al. 2007; Zalucki et al. 2016; Grant et al. 2018). An uncertainty analysis indicated model predictions in a spatially-explicit Iowa U.S. landscape are sensitive to assumptions concerning flight step lengths and directionality, in addition to perceptual range and spatial memory (Grant et al. 2018). Empirical studies of female monarch movement patterns within and between habitat patches in representative agricultural landscapes would reduce uncertainty in predicting landscape-scale egg densities for alternate habitat restoration plans.

Several approaches could be employed to quantify monarch movement patterns across agricultural landscapes. To date, insect flight behavior has been studied through visual observation or the use of technologies such as harmonic radar (Drake and Reynolds 2012), very high frequency (VHF) radio telemetry (Wikelski et al. 2006; Hagen et al. 2011; Svensson et al. 55

2011; Levett and Walls 2011; Fornoff et al. 2012; Liegeois et al. 2016; Martin W; Kissling et al.

2014; Wang et al. 2019), and radio frequency identification (RFID; Schneider et al. 2012). The strengths and limitations of each approach to quantify flight movement depend on the behaviors of interest within relevant spatial and temporal domains.

Visual observation, aided by binoculars (Merckx et al. 2003) or recording theodolites

(Zalucki and Kitching 1982b), in tandem with Geospatial Positioning System (GPS) units

(Schultz and Crone 2001; Schultz et al. 2012; Kallioniemi et al. 2014; Fernandez et al. 2016;

Schultz et al. 2017) have been used to record locations of over 25 butterfly species up to 50 m, with one report of tracking a sliver-studded blue butterfly (Plebejus argus) 259 m. Flagging locations between flight steps and measuring distances and compass bearings between locations have been used to track the Fender’s blue butterfly (Icaricia icarioides fender), alpine butterfly

(Parnassius smintheus) and the endangered scarce large blue butterfly (Phengaris teleius) up to

37.5 m (Turchin et al. 1991; Schultz 1998; Fownes and Roland 2002; Skorka 2013). These approaches provide the means to recreate flight paths within a habitat patch, assuming the time between successive two- or three-dimensional locations can be quantified. Given limitations of the human visual range, these approaches have limited means to quantify movement patterns within and between habitat patches at landscape scales.

Studies that implement VHF radio telemetry have relocated migrating dragonflies (Anax junius) over 1,500,000,000 hectares (ha) (Wikelski et al. 2006) and bumblebees (Bombus sp.) over 44 ha (Hagen et al. 2011), both with the aid of aircraft. Increased frequency of data collection can theoretically provide greater certainty of estimated flight paths; however, to date insects have been relocated once or twice per day (Wikelski et al. 2006; Hagen et al. 2011;

Svensson et al. 2011; Levett and Walls 2011; Fornoff et al. 2012; Liegeois et al. 2016; Martin W; 56

Kissling et al. 2014). Recently, Wang et al. (2019) employed VHF radio telemetry to collect data from migratory golden birdwing butterflies (Troides aeacus) every 30 minutes; however, successful location estimates were achieved in 38% of the observation periods. Thirty-minute to

24- or 48-hour data collection intervals are insufficient to reconstruct monarch flight patterns with sufficient resolution to enhance individual-based model predictions of monarch movement at habitat edges in landscapes. New approaches with tracking technology need to be implemented to provide higher frequency estimates of locations and flight paths.

Based on Kissling et al.’s (2014) review of the advantages and limitations of a variety of technologies, we are applying VHF radio telemetry methods to reduce uncertainty in estimated flight paths of female monarch butterflies at distances beyond human visual limits. Here we report: (1) a 220 mg transmitter does not significantly affect monarch butterfly behavior and flight capability, (2) a technique using handheld antenna to collect location data every minute, which is a time-scale relevant for monarch flight behavior, and (3) preliminary insights on monarch butterfly movement with a straight-line flight path analysis over distances larger than previously reported with visual data collection. Because monarch flight paths are not straight lines between stationary locations, we also illustrate how continuous-time movement models can use estimated locations to estimate flight patterns better. Employing the VHF radio telemetry methods described here, combined with continuous-time movement models, will provide the means to estimate flight paths in landscapes with multiple types of land cover.

Methods

Field Sites

Studies were undertaken in restored prairie areas of at least four ha containing a diversity of grasses, blooming forbs, and Asclepias sp. in Boone, IA (UTM zone 15N, E: 427169.32, N:

4656779.47), Ames, IA (UTM zone 15N, E: 447691.83, N: 4653386.83), and Rochester, MN 57

(UTM zone 15N, E: 541666.13, N: 4873350.54), U.S.A. All work was conducted with non- migratory monarchs in late June to early August in 2016, 2017, and 2018. All studies occurred on clear or mostly clear days between 1000 h and 1500 h with temperatures between 21-30C and wind speeds below 16 kph.

Insects

We collected male and female monarch butterflies in and around field site locations.

Butterflies were held in a soft-sided, mesh cage for 1-5 days before use in an experiment, in accordance with Zalucki and Kitching (1982b). Gatorade (Chicago, IL) was provided ad libitum as a sugar source.

Transmitter Attachment

Transmitters (LB-2X, 220 mg, Holohil Systems Ltd, Ontario, Canada) were attached to the ventral side of the monarch’s abdomen with superglue by Loctite (Henkel Canada,

Brampton, Ontario, Canada; Boiteau et al. 2009; Fig 1a). Monarchs with a transmitter attached are referred to as “radio-tagged.” Monarchs with attached watch batteries (300 mg; Energizer

AZ10DP, St. Louis, MO, U.S.A.) are referred to as “sham-tagged” (Fig. 1b). Sham-tagged monarchs were used to assess the extent to which added weight altered behavior and flight ability. After transmitter or watch battery attachment, butterflies were anesthetized in a

Styrofoam cooler containing dry ice wrapped in a cloth for one minute to prevent fright behavior

(Schultz and Crone 2001; Schultz et al. 2012; Kallioniemi et al. 2014; Schultz et al. 2017;

Fernandez et al. 2016). Handling and attachment were completed in less than three minutes.

Effects of Transmitter Attachment

To determine the extent to which added weight affected behavior time allocation, we observed male and female butterflies with (“sham-tagged”) and without (“untagged”) an attached watch battery (sham-transmitter). Sham-tagged and untagged butterflies were released 58 individually on either a blooming common milkweed ramet or a blooming forb in a restored prairie at one of the field sites. Monarchs were observed for twenty minutes or until they flew from the observation area. Using stopwatches, three observers stood approximately 5 m from the butterfly (Skorka 2013) and recorded the continuous-time intervals of various behaviors, including resting, feeding, and flying (oviposition was not observed). In 2016 and 2017, 16 and

28 individual monarchs were observed (Table 1). Behaviors were subsequently classified as “in- air” (flying) or “on-plant” (resting and feeding). In 2018, observations of 49 individuals (Table

1) were classified as “on-plant” or “in-air.”

For statistical analyses, observations from all three years were combined into on-plant or in-air groups. RStudio version 1.2.1335 (RStudio Team 2016; R Core Team 2019) with the

Estimated Marginal Means (emmeans) package (Lenth et al. 2020) was used to fit the proportion of time observed in-air to a binomial generalized linear model that accounted for tag attachment, monarch sex, and year. An analysis of the proportion of time observed resting while on-plant in

2016 and 2017 was conducted in a similar manner. There was no effect of monarch sex or year in either analysis.

Tracking Female Monarch Butterflies with VHF Radio Telemetry

Over five days in August 2016, radio-tagged female monarchs were released one at a time on vegetation in the center of the 4-ha Boone, IA site (Fig 2) to assess the means to track female-specific movements and behaviors; e.g., oviposition which has a direct impact on population growth (Grant et al. 2018). One observer stood approximately 5 m from the butterfly to observe behavior, as described previously. To recreate a flight path, compass bearings in the direction of the transmitter were collected from multiple locations simultaneously and repeatedly to triangulate sequential estimated locations of the transmitter. Radio telemetry operators were stationed in the four corners of the prairie (Fig 2; 50-135 m from the monarch release location). 59

Each operator held a compass, and a 3-element directional Yagi antenna (directional Yagi 3 element antenna for frequencies ranging from 164.000-166.000 MHz, Johnson's Telemetry, El

Dorado Springs, MO) connected with a 0.9 m coax cable (Johnson's Telemetry, El Dorado

Springs, MO) to a VHF radio telemetry receiver (Alinco DJ-X11; Toyama, Japan). Every minute after release, operators noted the compass bearing (Fig 2) for the loudest signal with an iPhone tape recorder application. Signals were noted every minute until the butterfly crossed the edge of the prairie or was observed within the prairie a maximum of 39 minutes. At this point, butterfly capture was attempted. When capture was unsuccessful (n=4), tracking with telemetry continued by foot or in vehicles until the butterfly was relocated. When monarchs were captured, their location was georeferenced (Trimble Geo7x; Sunnyvale, CA). Transmitters were removed from the radio-tagged monarch and re-used to track other monarchs. No transmitter was used more than 48 total hours (or more than 25% of their estimated battery life).

Space-Use Analysis

Bearings from all four corners of the prairie with associated date and time were transcribed for 13 butterflies. Using Location of a Signal software (LOAS; Ecological Software

Solutions LLC, www.ecostats.com), locations and error ellipses were estimated each minute based on triangulation of at least 3 of the four bearings with a maximum likelihood estimator.

Locations collected with the Trimble Geo7x (i.e., release sites and recapture sites) were considered true locations without error ellipses (error was < 30 cm). Step lengths (distance traveled in one minute) between two consecutive locations and turn angle created from three consecutive locations were calculated using the movement.pathmetrics package in Geospatial

Modeling Environment (GME; Spatial Ecology LLC, www.spatialecology.com/gme/). All step lengths for an individual were summed to estimate total distance traveled; Euclidian distances from start to end were measured in ArcMap 10.3 to estimate true displaced distances. 60

Straight-line analyses are typically used to infer movement; however, this approach leads to overconfidence in movement inferences because 1) movement is likely not a straight line between two data collection points, and 2) no error is assumed in location estimates (Calabrese et al. 2016). Occurrence models created using the continuous-time movement model (ctmm) package (Flemming et al. 2019) in RStudio (RStudio Team 2016; R Core Team 2019) can estimate an animal’s location during the tracking period, including times between data collection

(Calabrese et al. 2016). As sampling frequency increases, the occurrence model output more closely resembles the true movement path (Fleming et al. 2016). In the present study, occurrence models were selected based on the most representative semi-variogram estimates with the smallest AIC value. Output was a raster-image scaled with the proportion of time spent in each raster cell. Since each butterfly was observed for a short time period (i.e., up to 39 minutes), occurrence models for nine butterflies were summed to create one output. Four of the 13 monarchs were removed from the occurrence analysis because too few location estimates were made or error ellipses were too large to employ the model properly. A GeoTIFF of the field site

(Iowa Geographic Map Server 2015) was digitized by land use type and converted into a raster image in ArcMap 10.3. Land use types included: restored prairie, grass-dominated habitat, forest, agriculture, road and roadside, and residential. Area covered by intermediate to high occurrence probability within the prairie was estimated. Percent of time in each land use type was calculated with raster calculation in ArcMap 10.3. Because our study was primarily designed to assess the feasibility of employing VHF radio telemetry to track high-resolution monarch flight behavior, individuals were typically captured before they left the prairie. Consequently, while our findings are relevant for assessing behavior within the prairie, the results of the occurrence models over multiple land-use types should not be considered representative of flight patterns across larger, 61 more complex landscapes. This study does, however, illustrates a workflow that can be applied in future studies to analyze movement data within and across different land cover types without assuming straight-line flights.

Results

Effects of Transmitter Attachment

Of the monarchs observed with (n=45) or without (n=48) sham-transmitters, 88.9% of sham-tagged and 91.7% of sham-untagged monarchs flew at least once during their observation period. There was no significant effect of sham-transmitter attachment on time spent in-air (Fig

3a; F = 0.097; df = 1, 96; P = 0.7551). Sham-tagged monarchs were observed in-air for 27.5 ±

33.0% (sd) of the observation period, while untagged flew for 32.1 ± 36.4% of their observation period. Twenty-three sham-tagged (51.1%) and 25 untagged (52.1%) monarchs left the prairie before the 20-minute observation period was completed and were not recaptured.

For those individuals with behavior observations on-plant (22 sham-tagged and 26 untagged), there was no significant difference in the percent of time spent resting with or without a sham-transmitter (Fig 3b; F = 3.034; df = 1, 47; P = 0.0816). While not significant at the

P=0.05 level, there was an indication that sham-tagged monarchs rest more than untagged monarchs. Monarchs with sham-transmitters rested for 66.2 ± 35.6% of the observation period, while those without sham-transmitters rested for 36.9 ± 35.0% of the period.

Space-Use Analysis

Thirteen female monarch butterflies were tracked with radio telemetry for up to 39 minutes. Attempts were made to track additional monarchs for up to 90 minutes; however, bearing estimates were unreliable and occurred at inconsistent time intervals due to operator fatigue. Four radio-tagged monarchs were not captured at the edge of the prairie, but were successfully tracked and recaptured in surrounding habitat up to 250 m away from the prairie. 62

Radio-tagged monarchs were recaptured on a variety of plants, including trees, maize, tall grass, and forbs.

Straight-line flight paths were constructed for the 13 butterflies with 4-39 one-minute separated sequential estimated locations (example flight path in Fig 4; Supp. Fig 6 for all 13 straight-line flight paths). Within these flight paths, the majority of step-lengths were below 50 m

(Fig 5a) and generally associated with foraging, as noted by an observer 5 m from the butterfly.

However, after 2-30 short steps below 50 m, four individuals took steps ranging from 75-213 m and exited the prairie (Fig 5b). Summing the distance covered in the one-minute time-steps for each butterfly indicates individuals traversed 84.6-497.8 m during their observation periods.

However, Euclidian distance between the start and end locations ranged from 1.6-238.9 m, suggesting that true displacement was much smaller than the total distance traversed. Consistent with these findings, although turn angles spanned 360 degrees, the majority of turn angles were approximately 180 degrees, which indicates butterflies turned completely around and moved in the opposite direction of where it was originally headed (Fig 6).

The calculated error ellipse size was variable and ranged from 0.05-6560.22 m2. Most of the error ellipses were over 100 m2 (101 of 145 total). When error is large relative to movement distances, it should be included in space-use analyses to provide a more robust analysis of movement (Calabrese et al. 2016). Occurrence models were calculated for nine of the 13 monarch butterflies, as four individuals showed too few points or too large of error ellipses to employ this technique. The nine occurrence models were combined to better estimate the utilization of the multiple habitat types (Fig 7a). Areas of high probability of occurrence are associated with many estimated locations (Fig 7b). Intermediate or high occurrence was observed over 1.5-ha (15267.59 m2) of the 4-ha prairie (37.5%). Though our dataset was biased 63 for locations within the prairie, because error ellipses of the estimated locations spanned multiple land cover types, the occurrence model estimated that of the time observed, monarchs spent

98.95% of their time in the prairie, 1.00% in crop fields, 0.03% over the road, and 0.02% in the forest (Table 2).

Discussion

To date, insect movement studies within the range of visual observations provide reasonable estimates of movement at small scales (e.g., within a habitat patch; Zalucki and

Kitching 1982b). Tracking technology adapted for flying insects creates an opportunity to quantify movement at larger spatial scales. Application of this technology can advance understanding of how butterflies move at habitat patch edges within landscape scales. Here, we report a new application of VHF radio telemetry to improve the means of quantifying monarch butterfly flight movement at distances beyond the line of sight in a landscape containing a diversity of land cover classes.

Typically, VHF radio telemetry tracking is used to study the movement of large, slow- moving animals. In these instances, a researcher can walk around the radio-tagged animal and calculate an estimated location (Garton et al. 2001); however, flying insects move too quickly for this approach. Recently, Wang et al. (2019) described a novel application of insect radio telemetry to track the migration of golden birdwing butterflies (Troides aeacus) over distances up to 4314 m. Data were collected at 30-minute intervals by a single operator with a 38% success rate for estimating locations. Errors associated with estimated locations were not explicitly reported. The approach of Wang et al. (2019) shows promise for tracking flights of a large- bodied butterflies over 4000 m for up to 4 days; however, the 30-minute collection interval precludes the means to quantify flight patterns of butterflies encountering habitat edges within a landscape over shorter periods of time; e.g., over 30 to 90 minutes. Our VHF radio telemetry 64 approach, using the smaller-bodied, non-migratory monarch, with four operators collecting simultaneous bearings at one-minute intervals, provided the means to quantify errors associated with estimated locations every minute for up to 39 minutes.

Attaching a transmitter to an insect may have an effect on its behavior or flight ability

(Kissling et al. 2014). In studies of butterflies and moths with attached weight equaling up to

15% of their body weight, there was no effect on flight ability (Srygley & Kingsolver 2000;

Liegeois et al. 2016). However, when the transmitter was 66-100% of bumblebees’ (Bombus sp.) body weight, transmitter attachment caused increased resting, fewer flower visitations, and more time spent on a single flower (Hagen et al. 2011). To assess the extent to which our 220 mg transmitter could perturb the flight behavior of monarchs with a mass between 300 and 700 mg

(transmitter weight 31-73% of body weight), we conducted experiments with sham-transmitters

(300 mg watch batteries) to observe any effects on behavior and flying ability. Approximately ninety percent of our sham-tagged butterflies flew at least once during the observation periods.

On average, monarchs flew for 30% of their observed time and approximately half of the monarchs flew from the prairie release point in excess of 100 m. In one trial, we attached a battery to a 300 mg female monarch, and it robustly flew away from the release point, and we were unable to recapture her. Consistent with Hagen et al. (2011), our findings suggest monarch butterflies with transmitters may rest more than those without transmitters. This observation, however, did not preclude the ability to use radio telemetry transmitters to track their flight behavior. Based on our observations, sham-tagged monarchs will likely respond to environmental stimuli and make movement decisions in a manner similar to untagged monarchs.

Inputs of female monarch movement behavior within individual-based models (Zalucki

1983; Zalucki et al. 2016; Grant et al. 2018) are based primarily on an observational study of 65 monarch butterflies using recording theodolites in a grass-dominated, 0.03-ha experimental plot with planted milkweed (Zalucki and Kitching 1982b). In that study, monarchs concentrated most of their activity and flights in areas of the plot that contained milkweed. Within the plot, monarch locations and step lengths were estimated every 1.25 minutes. In the 0.03-ha plot, all reported steps were under 5 m, and butterflies flew generally straight, with turn angles mostly around 0 degrees. While Zalucki and Kitching (1982b) could estimate the exit angle of monarchs flying out of their 0.03-ha plot, their methodology precluded the means to track flight patterns further.

Using radio telemetry, we expanded understanding of flight steps and directionality by tracking locations of radio-tagged monarchs every minute in a 4-ha restored prairie containing milkweed, nectar plants, and grasses at natural densities and distributions. Within the prairie, we quantified small, undirected steps, while steps associated with exiting the prairie were large with high directionality. These movement patterns are qualitatively consistent with simulation modeling (Grant et al. 2018). We found that the majority of steps within the prairie were below

50 m, and most were under 5 m, consistent with Zalucki & Kitching (1982b). In contrast to

Zalucki & Kitching’s (1982b) observation of turn angles centering on 0 degrees, we recorded wide turn angles that centered around 180 degrees suggesting that monarchs were pivoting.

Within the prairie, these 180-degree turn angles were associated with foraging behavior and monarchs flying small distances from flower to flower in a pattern with low directionality.

Consequently, the total distance traveled was much greater than the Euclidian distance displacement.

With the use of radio telemetry, we attained more information about monarch butterfly movement than was possible by visual observation only. Simple mark-recapture or mark-resight studies with flying insects have an estimated recovery rate of 52% (Ide 2002). When our 66 operators lost sight of a butterfly, they were rarely able to relocate the individual with confidence, consistent with observations of sham-tagged monarchs. Four of our radio-tagged monarchs left the prairie, and although we could not detect them visually, each was relocated up to 250 m from the prairie using radio telemetry. When monarchs left the prairie, they took directional steps exceeding 50 m. To the best of our knowledge, quantitative measurements of flight steps of this length with non-migratory monarchs have not previously been reported. Our findings indicate that by using radio telemetry with monarch butterflies, we can attain movement information at scales relevant for landscape analyses.

Our error ellipses were large, suggesting it is not appropriate to base space utilization on estimated locations between steps (Calabrese et al. 2016; Fleming et al. 2016). Therefore, we incorporated estimated locations and error into occurrence models for nine of the 13 monarchs.

Based on the continuous-time movement model, we estimated monarchs spent 98.95% of their time within the prairie, and 1.00, 0.03, and 0.01% in agriculture, road and roadside, and forest, respectively because error ellipses of the estimated locations included in the model spanned these land cover types. In this time, monarchs had intermediate to high probability of occurring in

37.5% of the 4-ha prairie. Since we captured the monarchs as they crossed the edge of the prairie, these results are not surprising and should not be considered representative of habitat utilization estimates based on data collected for longer time periods and/or across more complex settings. However, these results are encouraging and suggest with sufficient frequency of data collection, continuous-time movement models can be used to estimate habitat utilization when monarchs are tracked over multiple land cover types. These analyses are especially useful to understand how animals move as they encounter different landscape features and to identify corridors between resources (Calabrese et al. 2016; Fleming et al. 2016). 67

When employing a continuous-time movement model, biologically-relevant data collection intervals for the study species is critical. Model estimates of locations and error ellipses are dependent on the elapsed time between data collection periods (Calabrese et al. 2016;

Fleming et al. 2016). Improved precision in utilization estimates could be gained with increased frequency of location estimates (Calabrese et al. 2016). While wide time intervals for collecting location data may be appropriate for large, slow-moving animals, for small, quick-moving, vagile species, like the monarch butterfly, data collection intervals in the order of one minute or less are required. Prior to our study, radio telemetry data sets for flying insects with the shortest collection interval was 30 minutes (Wang et al., 2019), which is too coarse to quantify adult monarch habitat utilization. By using simultaneous radio telemetry with a one-minute data collection interval, we could successfully track monarch butterflies and recreate flight paths.

Although our method helps increase the temporal resolution of radio telemetry for small, flying insects, there are limitations. Estimated locations were calculated with triangulation of simultaneously estimated bearings. Identifying the true bearing with this method can be difficult.

Both errors in bearing estimates and timing of data collection correlated to an increased error in location estimation. Bearings were based on human auditory perception. There was inter- operator variation in perceived signal strength when individuals were listening to the same signal. In addition, radio signals can reflect and occasionally produce loud signals in the wrong direction. This method also required four operators to collect bearing information at the same time consistently. To reach this operational objective, bearings were collected at the top of the minute. However, maintaining a one-minute collection interval rushed personnel in making bearing decisions, and occasionally collection times were missed. In these instances, we used three bearings instead of 4 to calculate locations. To improve bearing estimates, a receiver that 68 reports signal strength would provide a more objective means to identify the bearing with the strongest signal (Wang et al. 2019).

Through the use of radio telemetry, we could track monarch movement in areas larger than reported in previously published studies based on human observation (up to 50 m; Turchin et al. 1991; Schultz 1998; Schultz and Crone 2001; Fownes and Roland 2002; Schultz et al.

2012; Skorka 2013; Kallioniemi et al. 2014; Fernandez et al. 2016; Schultz et al. 2017). In our prairie, setting monarchs fly distances well beyond human visual range, and some of the radio- tagged monarchs flew greater than 250 m in a few minutes. At these distances estimating locations and flight paths by sight is difficult, if not impossible. While advancing the means to track monarchs at greater distances, our approach requires a team of five field technicians (1 to release the monarch and 4 to collect simultaneous bearings). To ensure data collection occurred simultaneously, and signal would be detected by the four operators at the same time, individuals needed to be stationed within 50 – 135 m of the release point.

To increase the area covered using this telemetry method, additional operators could be stationed further from the release site to create a grid. Alternatively, the use of automated systems has been suggested to increase scalability and data collection intervals while potentially reducing errors in location estimates (Kissling et al. 2014). A customized, automated system deployed in Panama successfully tracked 38 avian and mammalian species (Kays et al. 2011).

Currently, our group is developing and employing an automated radio telemetry system from commercially available equipment to increase the frequency of data collection for flying insects, using the monarch as a model species. This system can complement operator-based, simultaneous radio telemetry reported here to track flying insects and quantify movement patterns and habitat utilization. 69

Data Accessibility

Data, metadata, and R code pertaining to this manuscript are publicly available through

GitHub: https://github.com/kelseyefisher/employing-vhf-radio-telemetry-to-recreate-monarch- flight-paths

Acknowledgements

This work was supported by the Agriculture and Food Research Initiative Pollinator

Health Program (Grant No. 2018-67013-27541) from the USDA National Institute of Food and

Agriculture, Garden Club of America Board of Associates Centennial Pollinator Fellowship received in 2017, Holohil Grant Program received in 2018, ISU College of Agriculture and Land

Stewardship, and the Iowa Monarch Conservation Consortium.

We thank Signey Hilby and Chloe Weingarten for collecting behavior allocation data in

2017 and 2018, respectively. Additional thanks to Jacqueline Appelhans, Alec Euken, Cory

Haggard, Signey Hilby, Anthony Kerker, Riley Nylin, and Julia Pfeiffer for aid in monarch tracking. Dr. Philip Dixon and Audrey McCombs, ISU Statistics Department, provided statistical consulting. Appreciation to the property landowners, especially Barb Licklider, for allowing us to chase butterflies in their restored prairies, wooded areas, and agriculture fields. Special thanks to Justin Calabrese, Christen Fleming, Michael Noonan, and Joe Kolowski for leading the

February 2019 AniMove workshop and use of the ctmm package in RStudio for analyzing tracking data. Finally, thank you to Myron Zalucki and Chip Taylor for incredibly valuable insight and guidance when developing and planning this project.

References

Bergin, T. M., L. B. Best, K. E. Freemark, K. J. Koehler. 2000. Effects of landscape structure on nest predation in roadsides of midwestern agroecosystem: a multiscale analysis. Landsc Ecol. 15:131-143. 70

Boiteau G., F. Meloche, C. Vincent, and T. C. Leskey. 2009. Effectiveness of glues used for harmonic radar tag attachment and impact on survival and behavior of three insect pests. Environ Entomol. 38:168-175.

Brower, L. P., O. R. Taylor, E. H. Williams, D. A. Slayback, R. R. Zubieta, and M. I. Ramirez. 2012. Decline of monarch butterflies overwintering in Mexico: is the migratory phenomenon at risk? Insect Conserv Diver. https://doi.org/ 10.1111/j.1752- 4598.2011.00142.x

Calabrese, J. M., C. H. Fleming, and E. Gurarie. 2016. ctmm: an R package for analyzing animal relocation data as a continuous-time stochastic process. Methods Ecol Evol. 7:1124-1132.

Drake, V. A. and D. R. Reynolds. 2012. Radar entomology: observing insect flight and migration. CABI, Wallingford, UK.

Fernandez, P., A. Rodrigues, R. Obregon, S. de Haro, D. Jordano, J. Fernandez-Haeger. 2016. Fine scale movement of the butterfly Plebejus argus in a heterogeneous natural landscape as revealed by GPS tracking. J Insect Behav. 29:80-98.

Fleming, C. H., J. M. Calabrese, X. Dong, K. Winner, G. Peron, M. J. Noonan, B. Kranstauber, E. Gurarie, K. Safi, P. C. Cross, T. Mueller, R. C. de Paula, T. Akre, K. Drescher- Lehman, A. Harrison. 2019. ctmm: Continuous-Time Movement Modeling. R Package Version 0.5.8.

Fleming, C. H., W. F. Fagan, T. Mueller, K. A. Olson, P. Leimgruber, and J. M. Calabrese. 2016. Estimating where and how animals travel: an optimal framework for path reconstruction from autocorrelated tracking data. Ecology. https://doi.org/10.1890/15-1607.1

Fornoff, F., D. Dechmann, M. Wikelski. 2012. Observation of movement and activity via radio- telemetry reveals diurnal behavior of the neotropical katydid Philophyllia ingens (Orthoptera: Tettigoniidae). Ecotropica. 18:27-34.

Fownes, S. and J. Roland. 2002. Effects of meadow suitability on female behaviour in the alpine butterfly Parnassius smintheus. Ecol Entomol. 27:457-466.

Garton, E. O., M. J. Wisdom, R. A. Leban, and B. K. Johnson. 2001. Experimental design for radiotelemetry studies, pp. 16-44. In J. J. Millspaugh and J. M. Marzluff (eds.), Radio tracking and animal populations. Academic Press, San Diego, CA, USA.

Grant, T. J., H. R. Parry, M. P. Zalucki, M. P., S. P. Bradbury. 2018. Predicting monarch butterfly (Danaus plexippus) movement and egg-laying with a spatially-explicit agent- based model: The role of monarch perceptual range and spatial memory. Ecol Modell. https://doi.org/10.1016/j.ecolmodel.2018.02.011

Grant, T. J. and S. P. Bradbury. 2019. The role of modeling in monarch butterfly research and conservation. Front Ecol Evol. https://doi.org/10.3389/fevo.2019.00197 71

Hagen, M., M. Wikelski, and W. D. Kissling. 2011. Space use of bumblebees (Bombus spp.) revealed by radio-tracking. PLoS ONE. https://doi.org/10.1371/journal.pone.0019997

Ide, J. 2002. Mating behavior and light conditions cause seasonal changes in the dispersal pattern of the satyrine butterfly Lethe diana. Ecol Entomol. 27:33-40.

Inamine, H., S. P. Ellner, J. P. Springer, A. A. Agrawal. 2016. Linking the continental migratory cycle of the monarch butterfly to understand its population decline. Oikos. https://doi.org/10.1111/oik.03196

Iowa Geographic Map Server. 2015. Summer 2015 Orthophoto. Iowa State University Geographic Information Systems Support and Research Facility. https://ortho.gis.iastate.edu/

Kays, R., S. Tilak, M. Crofoot, T. Fountain, D. Obando, A. Ortega, F. Kuemmeth, J. Mandel, G. Swenson, T. Lambert, B. Hirsch, and M. Wikelski. Tracking animal location and activity with an automated radio telemetry system in a tropical rainforest. Comput J. 5(12)1931- 1948.

Kallioniemi, E., A. Zannese, J. E. Tinker, and A. M. A. Franco. Inter- and intra-specific differences in butterfly behavior at boundaries. Insect Conserv Diver. 7:232-240.

Kissling, W. D., D. E. Pattemore, and M. Hagen. 2014. Challenges and prospects in the telemetry of insects. Biol Rev. 89:511-530.

Lenth, R., H. Singmann, J. Love, P. Buerkner, and M. Herve. 2020. emmeans: Estimated Marginal Means. R Package Version 1.4.4.

Levett, S., S. Walls. 2011 Tracking the elusive life of the Emperor Dragonfly Anax imperatpr Leach. J Br Dragonfly Soc. 27:59-68.

Liegeois, M., P. Tixier, L. Beaudoin-Ollivier. 2016. Use of radio telemetry for studying flight movements of Paysandisia archon (Lepidoptera: Castniidae). J Insect Behav. 29(2):199- 213.

McIntire, E. J. B., C. B. Schultz, and E. E. Crone. 2007. Designing a network for butterfly habitat restoration: where individuals, populations, and landscapes interact. J Appl Ecol. 44:725-736.

Merckx, T., H. Van Dyck, B. Karlsson, and O. Leimar. 2003. The evolution of movements and behaviour at boundaries in different landscapes: a common arena experiment with butterflies. Proc R Soc Lond. 270:1815-1821.

Misenhelter, M. D. and J. T. Rotenberry. 2000. Choices and consequences of habitat occupancy and nest site selection in sage sparrows. Ecology. 8:2892-2901. 72

Oberhauser, K. S. 2004. Overview of monarch breeding biology, pp. 3-7. In K. S. Oberhauser and M. J. Solensky (eds), The monarch butterfly biology and conservation. Cornell University Press, New York, NY.

Oberhauser, K. S., R. Wiederholt, J. E. Diffendorfer, D. Semmens, L. Ries, W. E. Thogmartin, L. Lopez-Hoffman, B. Semmens. 2017. A trans-national monarch butterfly population model and implications for regional conservation priorities. Ecol Entomol. http://doi.org/10.1111/een.12351

Pitman, G. M., D. T. T. Flockhart, D. R. Norris. 2018. Patterns and causes of oviposition in monarch butterflies: Implications for milkweed restoration. Biol Conserv. https://doi.org/10.1016/j.biocon.2017.10.019

R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/

RStudio Team. 2016. RStudio: Integrated Development for R. RStudio, Inc., Boston, MA URL https://www.rstudio.com/

Schneider, C. W., J. Tautz, B. Grunewald, and S. Fuchs. 2012. RFID tracking of sublethal effects of two neonicotinoid insecticides on the foraging behavior of Apis mellifera. PLoS ONE. https://doi.org/10.1371/journal.pone.0030023

Schultz, C. B, B. G. Pe'er, C. Damiani, L. Brown, and E. E. Crone. 2017. Does movement behaviour predict population densities? A test with 25 butterfly species. J Anim Ecol. 86:384-393.

Schultz, C. B. 1998. Dispersal behavior and its implications for reserve design in a rare Oregon butterfly. Conserv Biol. 12(2):284-292.

Schultz, C. B. and E. E. Crone. 2001. Edge-mediated dispersal behavior in a prairie butterfly. Ecology. 82(7):1879-1892.

Schultz, C. B., A. M. A. Franco, E. E. Crone. 2012. Response of butterflies to structural and resource boundaries. J Anim Ecol. 81:724-734.

Siniff, D. B. and C. R. Jessen. 1969. A simulation model of animal movement patterns. Adv Ecol Res. 6:185-219.

Skorka, P., P. Nowicki, M. Lenda, M. Witek, E. B. Sliwinska, J. Settele, and M. Woyciechowski. 2013. Different flight behaviour of the endangered scarce large blue butterfly Phengaris teleius (Lepidoptera: Lycaenidae) within and outside its habitat patches. Landsc Ecol. 28:535-546.

Smouse, P. E., S. Focardi, P. R. Moorcroft, J. G. Kie, J. D. Forester, J. M. Morales. 2010. Stochastic modeling of animal movement. Philos Trans R Soc Lond. 365:2201-2211. 73

Srygley, R. B. and J. G. Kingsolver. 2000. Effects of weight loading on flight performance and survival of palatable Neotropical Anartia fatima butterflies. Biol J Linn Soc. 70:707-725.

Svensson, G. P., U. Sahlin, B. Brage, M. C. Larsson. 2011. Should I stay or should I go? Modeling dispersal strategies in saproxylic insects based on pheromone capture and radio telemetry a case study on the threatened hermit beetle Osmoderma eremita. Biodivers Conserv. 20:2993-2902.

Thogmartin, W. E., L. Lopez-Hoffman, J. Rohweder, et al. 2017. Restoring monarch butterfly habitat in the Midwestern US: ‘all hands on deck’. Environ Res Lett. https://doi.org/10.1088/1748-9326/aa7637

Turchin, P., F. J. Odendaal, and M. D. Rauscher. 1991. Quantifying insect movement in the field. Environ Entomol. 20:955-963.

Urquhart, F. A. 1960. The monarch butterfly. University of Toronto Press, Canada.

Wang Z., Y. Huang, and N. E. Pierce. 2019. Radio telemetry helps record the dispersal patterns of birdwing butterflies in mountainous habitats: Golden Birdwing (Troides aeacus) as an example. J Insect Conserv. 23:729-738.

Wikelski, M., D. Moskowitz, J. S. Adelman, J. Cochran, D. S. Wilcove, and M. L. May. 2006. Simple rules guide dragonfly migration. Biol Lett. 2:325-329.

Zalucki, M. P. 1983. Simulation of movement and egg-laying in Danaus plexippus (Lepidoptera: Nymphalidae). Res Popul Ecol. 25:353-365.

Zalucki, M. P. and J. H. Lammers. 2010. Dispersal and egg shortfall in monarch butterflies: what happens when the matrix is cleaned up? Ecol Entomol. https://doi.org/10.1111/j.1365- 2311.2009.01160.x

Zalucki, M. P. and R. L. Kitching. 1982a. Dynamics of oviposition in Danaus plexippus (Insecta: Lepidoptera) on milkweed, Asclepias spp.

Zalucki, M. P. and R. L. Kitching. 1982b. The analysis and description of movement in adult Danaus plexippus L. (Lepidoptera: Danainae). Behaviour 80:174-198.

Zalucki, M. P. and R. L. Kitching. 1984. The dynamics of adult Danaus plexippus L. around patches of its host plant Asclepias spp.

Zalucki, M. P., H. R. Parry, J. M. Zalucki. 2016. Movement and egg-laying in monarchs: to move or not to move, that is the equation. Austral Ecol. https://doi.org/10.1111/aec.12285

Zhao, K., R. Jurdak, J. Liu, D. Westcott, B. Kusy, H. Parry, P. Sommer, and A. McKeown. 2015. Optimal Levy-flight foraging in a finite landscape. J R Soc. https://doi.org/10.1098/rsif.2014.1158

Figures and Tables

Figure 1. Monarch butterflies with a 220 mg LB-2X radio telemetry transmitter (a) or sham- transmitter (b; 300 mg watch battery) attached to the ventral side of the abdomen with superglue

Figure 2. Field site in Boone, IA, U.S.A., where flight paths of monarch butterflies were tracked with VHF radio telemetry. Colors indicate land cover types at the site. The red circle is the release site for transmitter-tagged female monarch butterflies, and yellow stars indicate the locations of four operators equipped with a compass, a 3-element Yagi antenna, and a VHF radio telemetry receiver. Dashed lines are example compass bearings collected by operators in the direction of the transmitter at the release site 75

Figure 3. Proportion of time sham-tagged (300 mg watch battery) and sham-untagged monarchs were observed engaging in various behavior classifications. There was no significant effect of sham-transmitter attachment on time spent in-air; sham-tagged monarchs were observed in-air for 27.5 ± 33.0% (sd) of the observation period, while those without sham-transmitters flew for 32.1 ± 36.4% of their observation period (a). For those individuals observed on-plant, there was no significant difference in the percent of time spent resting with or without a sham-transmitter (b). While not significant at the P=0.05 level, there was an indication that monarchs with sham- transmitters rest more than sham-untagged monarchs. Monarchs with sham-transmitters rested for 66.2 ± 35.6% of the observation period, while those without sham-transmitters rested for 36.9 ± 35.0% of the period

Figure 4. Example recreated flight path of one monarch butterfly. Teal circles indicate estimated locations, label next to the point represents the minute of observation. Colored lines connecting points represent the straight-line steps between the points. Butterfly moved in order from yellow to red to blue. Grey circles represent the error ellipse calculated for each estimated location using Location of a Signal software (LOAS). Land cover types are represented by different colors

Figure 5. Characterization of 13 female monarch butterfly flight steps (distances between two estimated locations) tracked with radio telemetry in a restored prairie. The majority of step lengths were less than 50 m (a). Four monarchs with steps larger than 75 m flew out of the prairie after 2 to 19 steps (b)

Figure 6. Frequency of turn angles (angle created from 3 consecutive, estimated locations) for 13 female monarch butterflies tracking with radio telemetry in a restored prairie. Turn angles near 0 degrees indicate directional flight, while turn angles of approximately 180 degrees indicate pivoting in a more tortuous flight path

Figure 7. Occurrence model created using the ctmm package in RStudio for nine radio-tagged monarch butterflies. Black and dark grey shading represent areas with a low probability of occurrence. Light grey and white shading represent areas with a high occurrence probability. Landcover types are represented by different colors (a). Occurrence model of nine radio-tagged monarchs (see a) overlaid with estimated locations of the monarchs based on simultaneous triangulation (b)

Table 1. Number of female and male wild-caught individuals observed with and without sham- transmitters (300 mg watch batteries) in 2016, 2017, and 2018 to determine if there was an effect of weight attachment on behavior or flight ability

Sham-Tagged Untagged Year Female Male Female Male 2016 5 1 9 1 2017 4 12 2 10 2018 10 13 16 10

Table 2. Percent of time observed monarchs occurred in various landcover types, as estimates with ctmm occurrence models

Land Cover Type Percent Restored Prairie 98.95 Agriculture 1.00 Road and Roadside 0.03 Forest 0.01 Grass Dominated 0.00 Residential 0.00

Appendix. Supplemental Information

Figure 8. Individual recreated straight-line flight paths for all 13 VHF radio-tagged monarch butterfly females tracked in July 2016 at the field site in Boone, IA, U.S.A. (a-m). Teal circles indicate estimated locations, label next to the point represents the minute of observation. Colored lines connecting points represent the straight-line steps between the points. Butterfly moved in order from yellow to red to blue. Grey circles represent the error ellipse calculated for each estimated location using Location of a Signal software (LOAS). Land cover types are represented by different colors 79

CHAPTER 4. LOCATING LARGE INSECTS USING AUTOMATED VHF RADIO TELEMETRY WITH A MULTI-ANTENNAE ARRAY

Fisher, K.E.1,5, Dixon, P.M.2,5, Han, G.2, Adelman, J.S.3,4, and Bradbury, S.P.1,3

1Department of Entomology, Iowa State University, Ames, IA, 50011

2Department of Statistics, Iowa State University, Ames, IA, 50011

3Department of Natural Resource Ecology and Management, Iowa State University, Ames, IA,

50011

4Department of Biological Sciences, The University of Memphis, Memphis, TN 38117

5Equal contributions from the first two authors

Modified from a manuscript published in Methods in Ecology and Evolution.

Fisher, KE, PM Dixon, G Han, JS Adelman, & SP Bradbury. Locating large insects using automated VHF radio telemetry with multi-antennae array. Methods in Ecology and Evolution. (Accepted October 23, 2020).

Abstract

We describe an automated radio telemetry system (ARTS) designed for estimating the location of 0.50g butterflies that was constructed with commercially available materials.

Previously described systems were not designed to estimate fine-scale locations of large insects within approximately 200m2 study areas. The ARTS consists of four receiving stations. Each receiving station has four 3-element, directional Yagi antennae (separated by 60) connected to an automated receiver that records detected power sequentially from each antenna. To develop and evaluate the ARTS performance, four receiving stations were installed in the corners of 4-ha and 6.25-ha square fields with varying heights of vegetative cover. The location of a 0.22g 80 transmitter was estimated with a statistical method implementing both distance- and angle-power relationships. Calibrated model parameters were based on power detected from transmitters at known locations. Using independently collected data, model performance was evaluated based on estimated locations of a georeferenced stationary transmitter, a moving transmitter with a known georeferenced path, and a transmitter attached to a monarch butterfly (Danaus plexippus).

Estimated locations were calculated as frequently as every 5 seconds, which is at least 12 times greater than the sampling frequency previously reported for tracking insects. When sufficient power data was received, the median estimated locations of a transmitter attached to an investigator's hat were <16m from the true location. The median effective radius of the 95% confidence ellipse was 18.3m for stationary targets and 15.9m for a moving transmitter. Greater error in location estimation was expected when the transmitter was attached to a monarch butterfly due to interference from vegetation and variability in antenna orientation and transmitter height. As such, the median distance between the estimated and true locations was

72m. After applying a correction for the effect of vegetation, median location error was reduced by 12m. While our ARTS has likely reached the limit of current technology, the system is still a substantial methodological advancement for locating butterflies. Our efforts should provide a benchmark as technology improves.

Introduction

Understanding animal movement across varying spatial and temporal scales is an active area of fundamental ecological research, with practical applications in the fields of conservation biology and natural resource management (Kissling, Pattemore, & Hagen, 2014; Kays, Crofoot,

Jetz, & Wikelski, 2015; Chan, Tibbitts, Lok, Hassell, Pend, Ma, Zhang, & Piersma, 2019;

Nandintsetseg et al. 2019; Tobias & Pigot, 2019; Wittemyer, Northrup, & Bastille-Rousseau,

2019). Such knowledge can support elucidation of migration patterns, home ranges, and 81 responses to global and regional fluctuations in environmental conditions. Technology advances provide extensive potential for tracking vertebrates (Hallworth & Marra, 2016; McMahon,

Rachlow, Shipley, Forbey, Johnson, & Olsoy, 2017; Telemetry Solutions, 2017; ARGOS

System, 2020; Lotek, 2020); however, options to study the movement ecology of insects are limited.

Historically, and most commonly, insect movement has been studied through visual observation. Observations, with aid of binoculars, recording theodolites, Geospatial Positioning

System (GPS) units, and/or compasses, allow researchers to flag and georeference locations and estimate exit angles from habitat sites when the insect can no longer be followed on foot.

Researchers have employed these methods to track at least 30 butterfly species up to 50m

(Zalucki & Kitching, 1982; Turchin, Odendaal, & Rauscher, 1991; Schultz, 1998; Schultz &

Crone, 2001; Fownes & Roland, 2002; Merckx, Van Dyck, Karlsson, & Leimar, 2003; Schultz,

Franco, & Crone, 2012; Skorka, Nowicki, Lenda, Witek, Sliwinska, Settele, & Woyciechowski,

2013; Kallioniemi, Zannese, Tinker, & Franco, 2014; Fernandez, Rodrigues, Obregon, de Haro,

Jordano, & Fernandez-Haeger, 2016; Schultz, Pe’er, Damiani, Brown, & Crone, 2017). Given limitations of the human visual range and associated error, potential interference of the investigator on natural behavior of the insect, and intensive effort by the researcher, visual observation provides limited means to quantify movement patterns within and between habitat patches at landscape scales. Less frequently, movement ecology of insects has been studied with the aid of tracking technologies including harmonic radar (Drake & Reynolds, 2012), very high frequency (VHF) radio telemetry (Wikelski, Moskowitz, Adelman, Cochran, Wilcove, & May,

2006; Hagen, Wikelski, & Kissling, 2011; Levett & Walls, 2011; Svensson, Sahlin, Brage, &

Larsson, 2011; Fornoff, Dechmann, & Wikelski, 2012; Kissling, Pattermore, & Hagen, 2014; 82

Liegeois, Tixier, & Beaudoin-Ollivier, 2016; Wang, Haung, & Pierce, 2019; Fisher, Adelman, &

Bradbury, 2020), and radio frequency identification (RFID; Schneider, Tautz, Grünewald, &

Fuchs, 2012). The strengths and limitations of each approach depend on the behaviors of interest within relevant spatial and temporal domains (Fisher, Adelman, & Bradbury, 2020). The most promising technology for exploring movement, habitat use, behavior, and migration of insects is

VHF radio telemetry (Kissling, Pattermore, & Hagen, 2014); however, applying this technology is constrained to the size of the animal. Transmitters exceeding 3% of a vertebrate animal's body weight have been shown to alter movement behavior (Murray & Fuller, 2000; Barron, Brawn, &

Weatherhead, 2010; Hallworth & Marra, 2016). Although insects can carry transmitters up to approximately 50% of their mass (Srygley & Kingsolver 2000; Hagen, Wikelski, & Kissling,

2011; Liegeois, Tixier, & Beaudoin-Ollivier 2016; Fisher, Adelman, & Bradbury, 2020), because of the relatively small size of insects, they require small, low-powered transmitters. For example, monarch butterflies (Danaus plexippus) weighing approximately 0.50g exhibit normal flight behavior while carrying 0.30 to 0.25g transmitters, which are among the smallest commercially available (Fisher, Adelman, & Bradbury, 2020).

Very high frequency radio telemetry requires a receiver that detects the active transmitter signal, typically via a handheld or vehicle-mounted antenna (Kissling, Pattemore, & Hagen,

2014). The transmitted signal can be stored in the receiver and/or converted to an audible sound scaled to the received power. Directional antennae can be used to estimate the location of an animal without visual confirmation. If the animal is not moving, or moving slowly, the researcher can encircle the animal and determine its location based on triangulation of multiple estimated compass bearings in the direction of the most intense received signals (Lenth, 1981;

White & Garrott, 1990). However, specifically when tracking insects, it is difficult, if not 83 impossible, for one researcher to collect multiple bearings to estimate locations and reconstruct flight patterns. Alternatively, directional antennae can be used to approach an animal and facilitate visual confirmation of its true location. In the case of rapidly moving insects, researchers have followed tagged individuals on foot or in vehicles using handheld directional antennae (Hagen, Wikelski, & Kissling, 2011). When that approach failed, some research teams attempted to relocate transmitter signals from small aircraft, which required significant training and financial investment (Wikelski, Moskowitz, Adelman, Cochran, Wilcove, & May, 2006;

Hagen, Wikelski, & Kissling, 2011).

Recent advancements have improved the means of using VHF radio telemetry to track insects. For example, the Motus Wildlife Tracking System is a collaborative, global automated radio-telemetry network of over 900 receiving stations designed to facilitate landscape-scale research of migratory animals (Taylor et al., 2017; Motus Wildlife Tracking System, 2020), and was used by Knight, Pitman, Flockhart, and Norris (2019) to track the North American fall monarch butterfly migration. These researchers estimated that migratory monarchs traversed an average of 61 +/- 42 km/day. On a landscape scale, a single researcher tracked migratory golden birdwing butterflies (Troides aeacus) across 4km over a 4-day period using a directional, handheld antenna and a ''directional-strength" analytical technique (Wang, Haung, & Pierce,

2019). Attempts were made to detect tagged butterflies every 30 minutes; approximately 38% of these attempts resulted in estimated locations. These techniques provide temporally coarse estimates of insect movement and are not applicable for quantifying movement at finer scales that are still beyond human visual acuity. To fill this methodological gap, Fisher, Adelman, and

Bradbury (2020) stationed four researchers with directional antenna-receiver units around the release site of radio-tagged monarch butterflies and collected simultaneous bearings at one- 84 minute intervals. These bearings were subsequently triangulated to estimate locations within and immediately adjacent to a 4-ha restored prairie. This method provided the means to quantify estimated locations with associated error on a biologically relevant timescale to recreate movement paths and estimate habitat utilization. While successful, this method has practical limitations, including operator fatigue during long data collection sessions, inter-operator differences in audible detection of signals, inability to estimate locations in less than one-minute intervals, and the need for additional operators to increase spatial coverage.

Automated radio telemetry systems (ARTS) have been proposed as a means to increase scalability, frequency in data collection intervals, and potentially minimize errors in location estimates for insects (Kissling, Pattemore, & Hagen, 2014). Several research teams have described ARTS for tracking mammals and birds (Cochran, Warner, Tester, & Kuechle, 1965;

Larkin, Raim, & Diehl, 1996; Cochran, Swenson, & Pater, 2001; Kays et al., 2011; Ward,

Sperry, & Weatherhead, 2013; Lenske & Nocera, 2018; Crewe, Deakin, Beauchamp, & Morbey,

2019; Knight, Pitman, Flockhart, & Norris, 2019; Motus Wildlife Tracking System, 2020). Most notably Kays et al. (2011) designed and utilized an ARTS to track rainforest birds and mammals on Barro Colorado Island, Panama. This system incorporated seven custom, automated receiving units on 40m receiving stations to cover an area of roughly 1500ha. Each receiving station housed six, six-element directional Yagi antennae pointing in known bearing directions. The advances of Kays et al. (2011) and others to track animals ranging from 21g to 10,000g in size, provide the framework we used to develop an ARTS designed for tracking 0.5g insects, carrying low powered transmitters, over spatial scales of 4-6ha.

Here, we describe an ARTS that was originally designed to study movement of non- migratory monarch butterflies beyond human visual range (approximately 200m2) on a fine 85 temporal scale. The monarch butterfly is at risk of quasi-extinction (Semmens et al., 2016), in part, from the loss of breeding habitat in agricultural landscapes of the Midwest United States

(Thogmartin et al., 2017). Because adult females are not patch-resident, egg abundance and distribution across the landscape is a function of their perceptual range, flight directionality, and flight step lengths (Zalucki, Parry, & Zalucki, 2016; Grant, Parry, Zalucki, & Bradbury, 2018).

Increased understanding of female monarch movement through empirical studies in landscapes with different spatial patterns of restored breeding habitat is needed to improve predictions of realized fecundity (Fisher, Adelman, & Bradbury, 2020). Our ARTS was built modularly with commercially-available materials to reduce costs and simplify installation at different research sites. We demonstrate successful location estimates of stationary and moving transmitters and address the influence of transmitter orientation and vegetative cover on the detection of radio- tagged monarchs and the precision of their estimated locations. We conclude with a discussion of the ARTS strengths, limitations, and future research directions.

Materials and Methods

Overview

This ARTS, designed for estimating locations of insects beyond human visual ability on a fine temporal scale, was built with commercially available materials and adapted from the system described by Kays et al. (2011). We describe our ARTS design and statistical model, including:

1. transmitter, antennae, receivers, and design to collect multiple measurements of signal

strength

2. statistical method development and evaluation to estimate location from the strength of

the received signal. 86

Our ARTS consisted of four receiving stations installed and georeferenced (Trimble

Geo7x; Sunnyvale, CA, USA) in the corners of a square grid. Each receiving station had an automated receiver and four directional antennae pointing in different directions. We used 3- element Yagi antennae, rather than 6-element (Kays et al. 2011), because they are lighter, which increased receiving station stability. Because we used different materials and antennae arrays, the mathematical model used by Kays et al. (2011), which incorporated a relationship between received power and angle, was not appropriate to estimate transmitter locations with our system.

Our ARTS utilized a model that incorporated both power-distance and power-angle relationships. Performance of the ARTS was evaluated using a low-powered, 0.22g transmitter, taking into account factors including height and orientation of the transmitter, vegetation between the transmitter and the receiving station, and rate and pattern of transmitter movement

(see Kays et al., 2011; Ward, Sperry, & Weatherhead, 2013; Lenske & Nocera, 2018; Crewe,

Deakin, Beauchamp, & Morbey, 2019).

System Design

Receiving stations were created from 7m tall aluminum telescoping banner stands (Figure

1a; 23 foot Mondo Outdoor Flag, San Diego Sign Company, Carlsbad, CA), which were stabilized with a guy wire system (EZ-GW-KIT, Solid Signal, Novi, MI, USA) that was augured

(auger anchor kit, Grainger, Lake Forest, IL, USA) approximately 0.5m into the ground. The top

1.25m of the pole was wrapped in electrical tape to reduce potential signal interference. Four 3- element, directional Yagi antennae (calibrated for frequencies from 164.000-166.000 MHz;

Johnson's Telemetry, El Dorado Springs, MO, USA) were attached to the top of the pole with hose clamps and a series of RAM mounts (RAP-B-400U, RAM-B-201U, RAM-B-231Z-

2NUBU; RAM Mounts, Seattle, WA, USA). Mounts were separated by 0.3m, with the top mount 0.9m higher than the bottom mount. The claw clamp of the mount (RAP-B-400U) was 87 attached between the middle and bottom parallel elements of a Yagi antenna, which held the antenna in the vertical orientation (Figure 1b). A vertical orientation was used to maximize detection of transmitter signals with increasing distance from a receiving station, rather than a horizontal orientation that has greater ability to discern directionality, but with a lower limit of detection with distance (Cochran, Warner, Tester, & Kuechle, 1965; Larkin, Raim, & Diehl,

1996; Cochran, Swenson, & Pater, 2001). Yagi antennae were positioned 60 apart, so the top and bottom antennae were pointed 180 from each other. The space between the middle two

Yagi antennae pointed toward the center of the field. To obtain the true direction of the Yagi antennae, compass bearings were taken while standing behind each antenna.

Figure 1. Each receiving station in our automated telemetry system consisted of four 3-element Yagi antennae attached to the top of a 7m pole, supported by a guy wire system that was augured 0.5m into the ground (a). Yagi antennae were placed in vertical polarization (b) separated by 60 around the pole's axis (c).

Yagi antennae were connected with 7.62m coax cables (Lotek Wireless, Newmarket, ON,

Canada) to an automated receiving unit (SRX800-D, Lotek Wireless, Newmarket, ON, Canada) at the base of the pole. Receivers were programmed to scan for a single, specific transmitter 88 frequency and collect data from each Yagi antenna for six seconds; one full scan through all four antennae occurred every 24 seconds (Supplemental Table ST1 lists the receiver programming parameters). The power of a detected transmitter was reported in Received Signal Strength

Indicator (RSSI) units that are logarithmically scaled, analogous to decibels, and stored in the receiver. Based on results from a preliminary evaluation of the antenna's signal detection sensitivity and directionality (Figure 6), receiver gain was set to 65, which minimized detection of signal behind the antenna.

We used LB-2X transmitters (Holohil Systems, Ltd., Ontario, Canada), which are among the smallest available commercially (0.22g), have approximately 8 days of battery life, and have been used successfully on monarch butterflies (Fisher, Adelman, and Bradbury, 2020). Data that aided in development and evaluation of the statistical model to predict transmitter location were collected with the transmitter attached to a stationary investigator, a mobile investigator, and a radio-tagged butterfly. After noting that LB-2X transmitters emit signal with a directional pattern when held in horizontal orientation (Figure 7), we ensured the transmitter antenna was maintained in a vertical orientation, which removed the directional pattern. When worn by an investigator, the transmitter was placed in 15ml conical tube (Fisher Scientific, Waltham, MA,

USA) and vertically affixed to a baseball cap. When attached to the ventral side of the butterfly's abdomen with superglue (Fisher, Adelman, and Bradbury, 2020), the transmitter antenna was bent at a 90 angle approximately 5mm after the end of the monarch's abdomen to maintain vertical orientation. Transmitters with a 90 bend of the antenna to the vertical orientation behaved similar to transmitters attached to the cap in vertical orientation (Figure 7). Locations of hat and stationary monarch-affixed transmitters were georeferenced (Trimble Geo7x; Sunnyvale,

CA) and adjusted with the closest base station; after correction, error ranged from 0-50mm. 89

Conceptual Model to Estimate Transmitter Location

Telemetry systems can estimate location through two, non-exclusive, basic principles: 1) the strength of a received signal decreases as distance from the source increases and 2) the strength of a received signal decreases as the direction to the source deviates from the antenna's boresight angle. Traditional direction-based triangulation requires finding only the antenna angle resulting in the greatest signal strength. We devised a method to estimate location implementing both angle and distance.

Antenna physics implies that distance effects and angle effects are additive when power is measured on a logarithmic scale, like decibels or RSSI units (Balanis, 2016). The model for power, P, at an arbitrary distance, d, and angle from the antenna to the transmitter, , is then:

The intercept, distance function, and angle function can be estimated by recording power at a range of distance and angles. Figure 8 shows the locations to calibrate equation (1) for one of our data collection periods.

An unknown transmitter location, (x, y), can be estimated from contemporaneous measurements of power at multiple antennae on the same receiving station. Implementing an estimator requires specifying the form of fdistance (d) and fangle (), estimating any unknown parameters in those functions, and devising an estimator of the location. Details of each of these steps are provided in the following subsections.

Determining fdistance (d)

According to Friis's law (1946), the power of a received signal, measured in decibels, decreases linearly with log transformed distance to the source. In practice, there is random variation in the received power, resulting in the model 90

where P* is the power, d is the distance between the antenna and the transmitter, and  is the random variation.

Automated receivers have a limit of detection for radio signals. Only radio signals with power above the detection limit are recorded; signals below that limit are ignored. Hence, the recorded power has a distribution that is truncated at the detection limit. When the random variation in power follows a normal distribution, the recorded power follows a truncated normal distribution. The expected recorded power is then

where P is the mean recorded power, P* is the mean power given by equation (2), c is the receiver detection limit,  is the standard deviation of power values, () is the standard normal probability density function and () is the standard normal cumulative probability density function. The second term in equation (3) is the shift in mean recorded power due to receiver failure to record below the detection limit. That shift is generally small unless the power predicted by Friis's law is close to or below the detection limit.

To confirm the theoretical relationship between distance and power given by fdistance, equations (2) and (3), we collected data at a range of distances from a receiving station. We installed two receiving stations 400m apart in a 4-ha, flat, open, harvested maize field with low to no vegetation to prevent interactions between the radio signals and the environment (Goldman

& Swensen, 1999). Each receiving station had an antenna pointing directly towards the other receiving station. An investigator, wearing the transmitter, walked along a straight line between the two receiving stations four times, stopping for two minutes every 25m (Figure 8). Because 91 the receiver sampled the antenna every 24 seconds, there were up to four instantaneous power readings per receiving station/antenna at each location. Preliminary investigation indicated that a small fraction of power readings were abnormally low over a two-minute sampling period; therefore, the median power of the two-minute interval was used to summarize the power for each of the four visits to each location. Distance from the transmitter to each receiving station was calculated assuming all antennae were 7m above ground. The difference between a transmitter's distance to the receiving station base and its distance to the antennae is < 0.5m at

50m or more from the receiving station and < 1m at 25m.

The relationship between power and log distance is linear between 25m and approximately 140m (Figure 2, left panel). Closer than 25m, signals saturated the receiver. In this field setting, the receiver detection limit was about 58 RSSI units. The truncated power model, equation (3), reasonably describes the observed power from 25m to 200m. That fitted model is essentially identical to the linear model up to distances of ca. 125m, then asymptotes at the receiver detection limit.

Determining fangle ()

When the receiving antenna is directional, the power of a received signal also depends on the angle between the transmitter and the antenna boresight. The specific relationship depends on the antenna design. For commonly used Yagi antennas, the relationship between angle and power depends on orientation of the antenna and the number, size, and spacing of parallel elements (Kuechle & Kuechle, 2012). The specific relationship requires specialized software to determine. Some antenna manufacturers provide this information, but ours did not. We digitized the radiation pattern (Figure 9) for a three-element Yagi antenna with dimensions identical to ours. For angles up to 80 (~1.4 radians) from the antenna boresight, the power is well

0.95 approximated by P = 0 + 1 (1 – cos ) . The pattern of the ``back lobe'', for angles between

90 and 270, is more complex. At least part of that theoretical pattern can be approximated by P

0.95 = 2 + 3 (1 + cos ) . The front- and back-lobe curves can be described by a single model by forcing the two parts to be continuous at 90 and -90. That model for data at a constant distance is

To confirm this relationship in a field setting, we installed four receiving stations in the corners of a harvested maize field with four antennae pointed at -60, 0, 60, and 120. An investigator walked in a semi-circle 25m from each receiving station base and stopped at 13 evenly distributed locations for two-minute intervals (Figure 8). This provided power readings at a variety of angles from the four antennae boresights, but at a constant distance. The theoretical pattern was fit by a segmented regression joined at 90 and -90 to account for the different 93 curves in the front and back lobes (Figure 2, right panel). The fitted angle model for these antennae is

The large random variation in power means that the bearing with the strongest signal may be quite different from the true bearing (Figure 2, right panel). Therefore, estimating location from only bearing angles does not work as well in our system as it has in others (see Kays et al. 2011).

Random Effects of Receiving Stations and Antennae

Evaluation of the residuals from the fitted regression supported the assumption that the random variation in received power could be approximated by a normal distribution with constant variance. With the details of fdistance and fangle equation (1) becomes

In the above equations, * is the mean power before adjusting for truncation or angle effects,  is the mean power at distance d and angle , and P is the observed power.

Inspection of the power readings from up to four antennae on each of up to four receiving stations suggested repeatable effects of individual antennae. In particular, we noted antenna- specific variation in 0, 2, and 3. If these random effects are ignored, the errors associated with each observation are not independent. We assessed these effects by fitting a mixed model generalization of equation (12) that allowed 0, 2, and 3 to randomly vary by antenna. 94

When this model (equations 12 and 13) was fit to distance and angle data from a harvested maize field, the estimated variance components for 0i, 0i, and 0i were 51, 145, and

57, while the estimated residual variance was 159. Only the data from distances between 25m and 140m were used in the fit to essentially eliminate the consequences of truncation. Evaluation of other possible random effects indicated no evidence of an antenna effect on, 1, the slope for log(distance), or a receiving station effect on 0, the overall intercept.

Estimating Location

Consider a transmitter at an unknown location (x,y) and an antenna j pointing in direction

ij at height h on receiving station i located at (ei,ni), the distance, dij, from the transmitter to the antenna is

and the angle between the antenna direction and the transmitter location, ij, is

where the arc tangent function is one that correctly identifies the appropriate quadrant, e.g. atan2() in R (R Core Team, 2019). The expected power observed at antenna j on receiving station i from a transmitter at location (x,y) is given by equation (12) evaluated using d = dij, ij, 95

estimated regression coefficients, 0, 1, 2, and 3 and the best linear unbiased predictors

(BLUPs) of the random effects, 0i, 0i, and 0i.

The regression coefficients and random effects in equation (12) are predicted by fitting a calibration data set with known locations. We found it necessary to collect calibration data for each new field and set of receiving station locations. Results reported here are based on calibrations using 50-75 locations per study field. The calibration locations included many collected in 25m radius circles around each receiving station (angle calibrations), some at varying distance from one or two receiving stations (distance calibrations) and some at random locations.

When the errors, ij associated with each observation satisfy the assumptions of independence, a normal distribution, and equal variance, the unknown location can be estimated by non-linear least squares, or equivalently maximum likelihood. This was implemented using the optim() function in R (R Core Team, 2019). For most data sets, we found minimization of the error sum-of-squares to be more numerically robust than maximization of the log likelihood.

The estimated variance-covariance matrix ̂ for the estimated location, l̂ = (x̂, ŷ), is given by the negative inverse of the Hessian matrix of the log-likelihood function evaluated at the estimated location. Equivalently, when the location is estimated by minimizing the error sum-of-squares,

where H is the Hessian matrix of second derivatives of the sum-of-squares with respect to the two location coordinates, x and y. 96

A bivariate 1 -  confidence region can be computed using a normal approximation

(Johnson & Wichern, 2002, p. 221, with S/n replaced by ̂). For two parameters, x and y, the confidence region boundary is the set of locations l = (x, y) that satisfy

Alternatively, the confidence region could be computed using profile sum-of-squares (Bates and

Watts, 2007).

The overall uncertainty in the location can be summarized by the effective radius of the

95% confidence interval:

When the confidence interval for the location is circular, this value is the radius of that confidence interval circle. When the confidence interval is elliptical, this value is radius of a circle with the same area as the confidence ellipse.

Estimating location of stationary transmitters at random locations

Locations and 95% confidence ellipses were estimated for all ten test locations (Figure

3). Six of the transmitter's actual locations were located inside the 95% confidence ellipse for the estimated location. The other four estimated locations were outside of the 95% confidence ellipse, but were within 49.2m of the actual location. The median distance between the estimated and actual locations was 14.7m; the median effective radius of the confidence ellipses was

18.3m. 97

Figure 3. The system accurately estimated locations of six of the ten test locations of stationary transmitters on a sod field at Blue Grass Enterprises. Receiving station locations are shown by green cross; short black lines away from the green cross indicate the direction of each antenna. True locations are shown in red and estimated locations are shown in black. The 95% confidence ellipses are shown in black when they include the true location and red if the estimated location does not fall within the ellipse.

When a transmitter was within 25m of a single receiving station, the received power saturated the receiver, which limited the precision of location estimates in these situations.

Similarly, if a transmitter was located outside the upper limit of the linear distance-power relationship (ca. 150m for the system in this study), recorded power readings were less reliable for estimating location because the power vs. distance curve asymptotes.

Evaluating the Method

Estimating location of stationary transmitter at random locations

At Blue Grass Enterprises sod farm in Linn County, IA, four receiving stations were installed 250m apart in the four corners of an approximately 6.25ha, flat, open area. An 98 investigator, with the transmitter attached to the hat, stood at 62 randomly or deliberately chosen, georeferenced locations for two-minute intervals. Data from 52 strategically chosen locations were used to calibrate the model, which was then used to estimate the locations of the remaining

10 points.

Estimating locations of a moving transmitter

To imitate monarch butterfly flight, while having control over speed and path, an investigator walked a straight line through the 250m2 detection area at a flat, open, harvested maize field in Story County, IA, at a pace of approximately 50m/min for approximately seven minutes. Known locations along the straight line path were georeferenced. Because the receiver only records one antenna at a time, synchronous power readings for the four antennae at a receiving station were obtained by smoothing the power vs time curve for each antenna (Figure

10). The power vs time curves were smoothed by fitting a generalized additive model with a thin-plate regression spline using gam() in the R mgcv library (Wood 2003, Wood 2011). Default parameters were used except for setting gamma=2 to obtain smoother fits. Predicted power for all antennae was calculated every five seconds, except when there was a gap of over 60 seconds, which corresponded to two consecutive cycles without a recorded power.

Estimating locations of a mobile butterfly

Four receiving stations were installed to create a 250m2 detection area in a restored prairie with habitat in Floyd County, IA, comprised of native grasses and forbs of approximately

1.0m in height. A radio-tagged monarch butterfly was released on a non-blooming forb and observed for forty minutes by an investigator stationed 5m away to record the nature and elapsed time of behaviors. When the butterfly was stationary, the location was georeferenced and the true time spent at that location was recorded. Likewise, the true time spent flying between stationary locations was recorded. Periods of low-level variation in received power and periods of erratic 99 changes in received power were associated with periods when the monarch was stationary or flying, respectively. Time sequences of power received during the observation period were smoothed using a gam() model, as described in the section above. Because the vegetation present in the prairie restoration was expected to lower confidence in estimated location in comparison to the sod and harvested maize fields (Larkin, Raim, & Diehl, 1996; Goldman & Swenson, 1999;

Kays et al., 2011; Crewe, Deakin, Beauchamp, & Morbey, 2019), we conducted a calibration test to quantify the distance-power relationship with the transmitter placed in a potted plant

(Chrysanthemum sp.; Figure 11). The resulting relationship paralleled that observed in the open field, but was dampened by 22 RSSI. Locations in the restored prairie were estimated every 30 seconds with and without employing the 22 RSSI adjustment to evaluate the extent to which the

'vegetation correction factor' improved estimates.

Results

Estimating locations of a moving transmitter

Using the smoothing calculation, sequential locations were estimated using the predicted power (Figure 4) of a transmitter that moved at a rate of approximately 50m/min. Although not shown, 93% of the estimated locations occurred within the 9% confidence ellipse and half of the estimated locations were within 15.9m of the actual location. All estimated locations of a moving transmitter were between 5.0m and 47.9m from the true location. When the transmitter was within 25m of a receiving station, estimated locations were not always consistent with actual, sequential locations. With increasing distance from receiving stations, the sequence of estimated and actual locations were concordant. Inconsistency between actual and estimated locations near receiving stations likely reflects the received power saturating the receivers.

Although location could be estimated from as few as 3 power readings, there was uncertainty in such estimates. When the confidence interval, equation (17), was estimated from 100

the 3 power values, the F2,1(0.95) quantile was 200. With four readings, the F quantile was much smaller: F2,2(0.95) = 19.0$. For example, the median sample size for the locations collected with the moving transmitter was 8, with a range from 3 to 11. Our experience suggests aiming to have at least 6 power readings from 3 receiving stations.

Estimating location of a mobile butterfly

Within the 40-minute observation period, the radio-tagged monarch flew for a maximum of 20 seconds among four stationary locations a maximum of 44m apart within the 250m 2 detection area of the restored prairie. The power received by the receiving stations was stable and consistent when the monarch was stationary. Variability in power was observed during times of known flight (Figure 12). All the raw data detected by the receiving stations was used to estimate

79 locations every 30 seconds. Consequently, we assume estimated locations are associated with the monarch stationary in the vegetation, since it is extremely unlikely a location could be estimated during the short time the monarch was in flight. Estimated locations were a median of

72m from the true stationary locations. Confidence ellipses could not be calculated for all estimated locations because of numerical difficulties estimating the variance-covariance matrix.

Two typical confidence ellipses, shown in Figure 5, had effective radii of 209m and 310m. Error estimates were large because each location was estimated with only 5 or 6 recorded power readings and those readings were variable. To account for dampened signals due to the presence of vegetation, we added 22 RSSI to each of the raw values (see section Estimating locations of a mobile butterfly), which improved our location estimates. The 79 estimated locations based on the 22 RSSI adjustment were a median of 46m from their associated true locations. The effective radii of the confidence intervals shown in Figure 5 were 120m and 240m. Although the difference between actual and estimated locations ranged from 38 to 93m, the pattern of estimated locations were concordant with the true locations between flights. 101

Figure 4. The system accurately captured the path of an investigator walking from the southeast to northwest receiving station at a pace of 50m/min at a harvested maize field. We estimated locations every 5s, depicted here as numerals in chronological order. Blue dots mark the investigator's georeferenced locations along the path. Receiving station locations are shown by green crosses; short black lines away from the green cross indicate the direction of each antenna.

102

Figure 5. We used the ARTS to estimate locations of a radio-tagged monarch butterfly that flew for a maximum of 20 seconds between known stationary points associated with distinct patches of vegetation that were a maximum of 44m apart (red vs. blue open circles; left panel). Raw RSSI signals detected by the receiving stations yielded clusters of estimated locations (filled circles) that corresponded to the known stationary points (open circles). These estimated locations were a median distance of 58m from the butterfly’s associated known locations. To offset the effect of vegetation on received signal strength, we added 22 RSSI to each of the raw values (see section 2.4.3 Estimating locations of a mobile butterfly), which yielded estimated points that were a median of 46m from the true locations (right panel). Representative confidence ellipses are shown for one estimated point within each cluster. Green crosses represent locations of receiving stations.

Discussion

Because of their relatively small size, techniques for tracking and quantifying insect movement are limited. To date, most methods are labor-intensive, rely on visual observation, or collect data on coarse timescales. We devised an automated VHF radio telemetry system to estimate locations of insects with low-powered transmitters at a fine temporal scale beyond human visual acuity. Although additional research and future technological advances could improve error estimates, our system provides a substantial advancement in comparison to previous methods to locate butterflies. Implementation of the ARTS described here increases 103 detection range to approximately 150m, which exceeds a 50m tracking limit based on visual observations (Zalucki & Kitching, 1982; Turchin, Odendaal, & Rauscher, 1991; Schultz, 1998;

Schultz & Crone, 2001; Fownes & Roland, 2002; Merckx, Van Dyck, Karlsson, & Leimar, 2003;

Schultz, Franco, & Crone, 2012; Skorka, Nowicki, Lenda, Witek, Sliwinska, Settele, &

Woyciechowski, 2013; Kallioniemi, Zannese, Tinker, & Franco, 2014; Fernandez, Rodrigues,

Obregon, de Haro, Jordano, & Fernandez-Haeger, 2016; Schultz, Pe’er, Damiani, Brown, &

Crone, 2017), and provides the means to estimate location errors. Our system is not intended to replace traditional handheld or vehicle-mounted radio telemetry methods, but rather augment methods and alleviate personnel requirements as compared to following insects by foot or vehicle; therefore, the ARTS also provides practical advantages as compared to personnel following insects on foot with handheld radio telemetry (Wikelski, Moskowitz, Adelman,

Cochran, Wilcove, & May, 2006; Hagen, Wikelski, & Kissling, 2011; Levett & Walls, 2011;

Svensson, Sahlin, Brage, & Larsson, 2011; Fornoff, Dechmann, & Wikelski, 2012; Kissling,

Pattermore, & Hagen, 2014; Liegeois, Tixier, & Beaudoin-Ollivier, 2016; Wang, Haung, &

Pierce, 2019; Fisher, Adelman, & Bradbury, 2020).

Our statistical method to estimate location incorporated both the effect of distance and angle on the received power by each of four Yagi antennae on a receiving station, which was not explored previously (Cochran, Warner, Tester, & Kuechle, 1965; Larkin, Raim, & Diehl, 1996;

Cochran, Swenson, & Pater, 2001; Kays et al., 2011). With this method, we were able to successfully estimate locations of stationary transmitters at distances beyond visual detection with a median accuracy of 14.7m. This accuracy from a distance of up to 150m from a receiving station is appropriate for the spatial scale of our movement questions and is a significant improvement compared to visual observations at distances up to, but not exceeding, 50m with no 104 error estimates. In addition, our method produced comparable accuracy when locating a mobile transmitter attached to an investigator's hat at a temporal resolution that is 12 times more frequent than the shortest telemetry data collection interval previously reported with insects

(Fisher, Adelman, & Bradbury, 2020; Hagen, Wikelski, & Kissling, 2011; Kissling, Pattemore,

Hagen, 2014; Knight, Pitman, Flockhart, & Norris, 2019; Wang, Haung, & Pierce, 2019; Fisher,

Adelman, & Bradbury, 2020). These results suggest under optimized conditions, with the transmitter antenna consistently vertical and 1.6m above the ground, our system and method of estimating location can be used to reasonably quantify locations and error estimates of stationary and moving low-mass transmitters at distances beyond human vision.

When the transmitter was attached to a mobile monarch butterfly, increased error in location estimates were expected due to interference of vegetation, variability in transmitter antenna orientation when in flight, and height of the transmitter (Kays et al., 2011; Ward, Sperry,

& Weatherhead, 2013; Crewe, Deakin, Beauchamp, & Morbey, 2019). Estimated locations were a median of 58m from their associated stationary location. To compensate for decreased received power, we added an empirically-derived constant 22 RSSI factor to each receiving station, which improved our location estimates and decreased the associated error. Estimated locations with the correction ranged in accuracy from 38 to 93m and the pattern of estimated and actual locations were concordant. This approach assumes a constant vegetative effect; however, the impact of vegetation likely varies with plant height and density (Kays et al., 2011; Ward, Sperry, &

Weatherhead, 2013; Crewe, Deakin, Beauchamp, & Morbey, 2019). Further investigation could ascertain the need to use site-specific vegetation to implement appropriate corrections for interference. Representative confidence ellipses had effective radii of 120m and 240m (see

Figure 5 right panel) because locations were estimated from 5 or 6 detected power values, which 105 impacts the multiplier that scales the variance-covariance matrix into a bivariate confidence ellipse. For example, for N=5, the multiplier is 5.04 and for N=6, the multipler is 4.17. With larger sample sizes, the multiplier is considerably smaller, e.g. 3.17 for N=10. Although location can be estimated from relatively few detected power values, large confidence ellipses are expected.

We confirmed that periods of stable received power occurred when the monarch was stationary and periods of variability in signal strength detection were associated with flight.

While we can detect when a monarch is in flight, it may prove difficult to predict location during flight because of the thin-plate regression spline smoothing function applied in our estimation method (Figure 12). The thin-plate regression spline smoothing function controlled for differences in the precise time of estimates from different receiving stations, and reasonably estimated received power in our field trials. When transmitters attached to the cap were moving slowly, fluctuations in received power over short time intervals likely represented background noise, rather than true differences in location. However, if a radio-transmitter was moving rapidly or erratically, fluctuations in received power could represent true differences in the location. A smoothing function applied to such data over time would produce estimated power values that reduce true differences over time, thus reducing the precision of location estimates.

Modifications to components of our ARTS could enhance its performance; however, these modifications are associated with additional financial cost, reduced transmitter battery life, and/or altered insect flight behavior. Detection range of our ARTS is limited by the distance- power relationship. With our current receiver settings, locations can be estimated if the transmitter is between 25 to ca. 150m from a receiving station. Changing the receivers' gain could increase the distance at which signals are consistently detected or reduce the distance at 106 which saturation occurs; additional calibration data would be required to describe the distance- power and angle-power relationships. Likewise, future transmitters that emit stronger signals, with low mass batteries, could also extend detection range and improve error estimates by increasing detection of power readings by receiving stations. Alternatively, additional receiving stations could be installed to extend the detectable range of the system; e.g., rather than deploying four receiving stations to create a 250m2 detection area, nine receiving stations would create a 500m2 detection area. Additional receiving stations, antennae on receiving stations, and/or antennae with more elements could improve tracking insects in flight, as suggested by

Lenske and Nocera (2018) for tracking birds. Using a transmitter with a greater pulse rate could also aid in tracking insects in flight (e.g. see Lenske & Nocera 2018) as well as decrease the error ellipse estimates.

To the best of our knowledge, this is the first ARTS designed to track insects (see

Kissling, Pattemore, Hagen, 2014 for a recent review of insect telemetry techniques). Using current, commercially-available equipment, our findings likely represent the limit of existing

ARTS technology to track non-migratory movement of insects. Although error estimates are large compared to ARTS using heavier, higher power transmitters for small mammals and birds, our system can estimate fine scale locations within a 250m$^2$ area at a frequency appropriate for our research questions. Estimating locations based on visual observations is not possible at these spatial and temporal scales. When combined with handheld radio telemetry to verify locations of tagged insects outside an ARTS’ grid, this system provides a cost-effective means to estimate locations of insects at distances beyond human visual range. Our method with 0.22g transmitters has potential to be adapted for use with other large insect species, small mammals, and small birds. 107

Data Accessibility

Data, metadata and R codes for this manuscript are publicly available through Github,

Zenodo and Dryad: https://github.com/kelseyefisher/locating-large-insects-using-automated-vhf- radio-telemetry-with-multi-antennae-array; https://zenodo.org/badge/latestdoi/26021 4221

(Fisher, 2020). Locating large insects using automated VHF radio telemetry with a multi- antennae array, Dryad Digital Repository https://doi.org/10.5061/dryad.9kd51c5fn (Fisher &

Dixon, 2020).

Acknowledgements

Conversations with Anuj Sharma at Iowa State University's Institute for Transportation provided critical design ideas. Materials and logistical support were provided by Richard L.

Hellmich, Keith G. Bidne, and Craig A. Abel with the USDA-ARS CICGRU in Ames, IA.

Thanks to our field site owners and managers for allowing access to their property: Mike Loan and Sarah Nolte at Blue Grass Enterprises, Adam Thomes and Ben Pease at Iowa State

University's Turf Research Station at the Horticulture Farm, Justin Stensland at Stensland Sod,

Warren Pierson at Iowa State University's Field Extension Education Laboratory (FEEL) Farm,

Kent Burns at the Iowa State University Dairy Farm, and David Bruene at Iowa State

University's Beef Teaching Farm. Field site recruitment was possible with guidance from Mark

Honeyman and Adam Thomes. Special thanks to our technicians; data collection and troubleshooting this system could not have been accomplished without their dedication and positive attitudes: Cody Acevedo, Kevin Anderson, Cory Haggard, Signey Hilby, Jenna Nixt,

Riley Nylin, Michael Roth, Samantha Shimota, Brooklyn Snyder, Kara Weber, and Elke

Windschitl. Kathleen Ray and Iowa State University's Statistic Consulting Group provided coding support. 108

This work was supported, in part, by the Agriculture and Food Research Initiative

Pollinator Health Program (Grant No. 2018-67013-27541) from the USDA National Institute of

Food and the Agriculture and Iowa Agriculture and Home Economics Experiment Station.

Project No. IOW03617 is supported by USDA/NIFA and State of Iowa funds. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of the U.S. Department of Agriculture.

References

ARGOS System. (2020). https://www.argos-system.org/ accessed 27 March 2020.

Balanis, C. A. (2016). Antenna Theory: Analysis and Design (4th ed). Hoboken, NJ: Wiley.

Barron, D. G., J. D. Brawn, and P. J. Weatherhead. (2010). Meta-analysis of transmitter effects on avian behavior and ecology. Methods in Ecology and Evolution. 1, 180-197. doi: 10.1111/j.2041-210X.2010.00013.x

Bates, D.M. and Watts, D.G., 2007. Nonlinear regression analysis and its applications. New York: Wiley.

Chan, Y., T. L. Tibbitts, T. Lok, C. J. Hassell, H. Peng, Z. Ma, Z. Zhang, and T. Piersma. (2019). Filling knowledge gaps in a threatened shorebird flyway through satellite tracking. J. Appl. Ecol. 56: 2305–2315. doi: 10.1111/1365-2664.13474

Cochran, W. W., D. W. Warner, J. R. Tester, and V. B. Kuechle. (1965). Automatic radio- tracking system for monitoring animal movements. BioScience. 15, 98-100. doi: 10.2307/1293346

Cochran, W., G. Swenson, and L. Pater. (2001). Radio finding for wildlife research. Antenna Applications Symposium. University of Illinois at Urbana-Champaign, IL, USA.

Crewe, T. L., J. E. Deakin, A. T. Beauchamp, and Y. E. Morbey. (2019). Detection range of songbirds using a stopover site by automated radio‐telemetry. J. Field Ornithol., 90: 176- 189. doi:10.1111/jofo.12291

Drake, V. A. and D. R. Reynolds. (2012). Radar entomology: observing insect flight and migration. CABI, Wallingford, United Kingdom.

Fernandez, P., A. Rodrigues, R. Obregon, S. de Haro, D. Jordano, and J. Fernandez-Haeger. (2016). Fine scale movement of the butterfly Plebejus argus in a heterogeneous natural landscape as revealed by GPS tracking. J. Insect Behav. 29: 80–98.

109

Fisher, K. E. (2020), Locating large insects using automated VHF radio telemetry with a multi- antennae array, Zenodo, https://zenodo.org/badge/latestdoi/260214221

Fisher, K. E., P. Dixon (2020), Locating large insects using automated VHF radio telemetry with a multi-antennae array, Dryad, Dataset, https://doi.org/10.5061/dryad.9kd51c5fn

Fisher, K. E., J. S. Adelman, S. P. Bradbury. (2020). Employing very high frequency (VHF) radio telemetry to recreate monarch butterfly flight paths. Environmental Entomology. doi: 10.1093/ee/nvaa019

Fornoff, F., D. Dechmann, and M. Wikelski. (2012). Observation of movement and activity via radio-telemetry reveals diurnal behavior of the neotropical katydid Philophyllia ingens (Orthoptera: Tettigoniidae). Ecotropica. 18: 27–34.

Fownes, S., and J. Roland. (2002). Effects of meadow suitability on female behaviour in the alpine butterfly Parnassius smintheus. Ecol Entomol. 27: 457–466.

Friis, H.T. (1946). A note on a simple transmission formula. Proceedings of the IRE. 34(5), 254- 256. doi: 10.1109/JRPROC.1946.234568

Goldman J. and G. W. Swenson. (1999). Radio wave propagation through woods. IEEE Antennas and Propagation Magazine. 41(5), 34-36. doi: 10.1109/74.801512

Grant, T. J., H. R. Parry, M. P. Zalucki, and S. P. Bradbury. (2018). Predicting monarch butterfly ( Danaus plexippus) movement and egg-laying with a spatially-explicit agent-based model: the role of monarch perceptual range and spatial memory. Ecological Modelling. 374, 37–50. doi:10.1016/j.ecolmodel.2018.02.011.

Hagen, M., M. Wikelski, and W. D. Kissling. (2011). Space use of bumblebees ( Bombus spp.) revealed by radio tracking. PLOS One. 6(5), e19997. doi: 10.1371/journal.pone.0019997

Hallworth, M. T. and P. P. Marra. 2016. Miniaturized GPS tags identify non-breeding territories of small breeding migratory songbird. Scientific Reports. 5, 11069. doi: 10.1038/srep11069

Johnson, R. A. and D. W. Wichern. (2002). Applied Multivariate Statistical Analysis, 5th ed., Prentice Hall, NJ.

Kallioniemi, E., A. Zannese, J. E. Tinker, and A. M. A. Franco. (2014). Inter- and intra-specific differences in butterfly behavior at boundaries. Insect Conserv Diver. 7: 232–240.

Kays, R., M. C. Crofoot, W. Jetz, and M. Wikelski. (2015). Terrestrial animal tracking as an eye on life and planet. Science. 348(6240), aaa2478. doi: 10.1126/science.aaa2478

110

Kays, R., S. Tilak, M. Crofoot, T. Fountain, D. Obando, A. Ortega, F. Kuemmeth, J. Mandel, G. Swenson, T. Lambert, B. Hirsch, and M. Wikelski. (2011). Tracking animal location and activity with an automated radio telemetry system in a tropical rainforest. The Computer Journal. 5(12), 1931–1948. doi: 10.1093/comjnl/bxr072

Kissling, W. D., D. E. Pattemore, and M. Hagen. (2014). Challenges and prospects in the telemetry of insects. Biological Reviews. 89, 511–530. doi: 10.1111/brv.12065

Knight S. M., G. M. Pitman, D. T. T. Flockhart, and D. R. Norris. (2019). Radio-tracking reveals how wind and temperature influence the pace of daytime insect migration. Biology Letters. 15, 20190327. doi: 10.1098/rsbl.2019.0327

Kuechle, V. B. and P. J. Kuechle. (2012). Radio telemetry in fresh water: the basics. pp 91-137 In Adams, N. S., J. W. Beeman, & J. H. Eiler (eds.) 2012. Telemetry techniques: a user guide for fisheries research. American Fisheries Society, Bethesda, MD.

Larkin, R. P., A. Raim, and R. H. Diehl. (1996). Performance of non-rotating direction-finder for automatic radio telemetry. Journal of Field Ornithology. 67, 59-71.

Lenske, A. K. and J. J. Nocera. (2018). Field test of an automated radio‐telemetry system: tracking local space use of aerial insectivores. Journal of Field Ornithology 89: 173-187. doi:10.1111/jofo.12254

Lenth, R.V. (1981). On finding the source of a signal. Technometrics. 23, 149-154. doi: 10.2307/1268030

Levett, S., S. Walls. (2011). Tracking the elusive life of the Emperor Dragonfly Anax imperatpr Leach. J. Br. Dragonfly Soc. 27: 59–68.

Liegeois, M., P. Tixier, L. Beaudoin-Ollivier. (2016). Use of radio telemetry for studying flight movements of Paysandisia archon (Lepidoptera: Castniidae). J. Insect Behav. 29(2): 199–213.

Lotek Wireless Inc. (2020). https://www.lotek.com/products/pinpoint-gps/ accessed 27 March 2020.

McMahon, L. A., J. L. Rachlow, L. A. Shipley, J. S. Forbey, T. R. Johnson, and P. J. Olsoy. (2017). Evaluation of micro-GPS receivers for tracking small-bodied mammals. PLOS One. 12(3), e0173185. doi: 10.1371/journal.pone.0173185

Merckx, T., H. Van Dyck, B. Karlsson, and O. Leimar. (2003). The evolution of movements and behaviour at boundaries in different landscapes: a common arena experiment with butterflies. Proc. Biol. Sci. 270: 1815–1821.

Motus Wildlife Tracking System. (2020). https://motus.org accessed 27 March 2020.

111

Murray, D. L. and M. R. Fuller. (2000). A critical review of the effects of marking on the biology of vertebrates. In Research Techniques in Animal Ecology (ed. F. T. Boitani L), pp. 15-64. New York: Columbia University Press.

Nandintsetseg, D., C. Bracis, K. A. Olson, K. Bohning-Gaese, J. M. Calabrese, B. Chimeddorj, W. F. Fagan, C. H. Fleming, M. Heiner, P. Kaczensky, P. Leimgruber, D. Munkhnast, T. Strratmann, and T. Mueller. (2019). Challenges in the conservation of wide-ranging nomadic species. J. Appl. Ecol. 56: 1916-1926. doi: 10.1111/1365-2664.13380

R Core Team (2019). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Schneider, C. W., J. Tautz, B. Grünewald, and S. Fuchs. (2012). RFID tracking of sublethal effects of two neonicotinoid insecticides on the foraging behavior of Apis mellifera. PLoS One. 7: e30023.

Schultz, C. B. (1998). Dispersal behavior and its implications for reserve design in a rare Oregon butterfly. Conserv. Biol. 12(2): 284–292.

Schultz, C. B. and E. E. Crone. (2001). Edge-mediated dispersal behavior in a prairie butterfly. Ecology. 82(7): 1879–1892.

Schultz, C. B., A. M. Franco, and E. E. Crone. (2012). Response of butterflies to structural and resource boundaries. J. Anim. Ecol. 81: 724–734.

Schultz, C. B., B. G. Pe’er, C. Damiani, L. Brown, and E. E. Crone. (2017). Does movement behaviour predict population densities? A test with 25 butterfly species. J. Anim. Ecol. 86: 384–393.

Semmens B. X., D. J. Semmens, W. E. Thogmartin, R. Wiederholt, L. Lopez-Hoffman, J. E. Diffendorfer, J. M. Pleasants, K. S. Oberhauser, and O. R. Taylor. (2016). Quasi- extinction risk and population targets for the Eastern, migratory population of monarch butterflies ( Danaus plexippus). Scientific Reports. 6, 23265. doi: 10.1038/srep23265

Skorka, P., P. Nowicki, M. Lenda, M. Witek, E. B. Sliwinska, J. Settele, and M. Woyciechowski. (2013). Different flight behaviour of the endangered scarce large blue butterfly Phengaris teleius (Lepidoptera: Lycaenidae) within and outside its habitat patches. Landsc. Ecol. 28: 535–546.

Srygley, R. B., and J. G. Kingsolver. (2000). Effects of weight loading on flight performance and survival of palatable Neotropical Anartia fatima butterflies. Biol. J. Linn. Soc. 70: 707– 725.

112

Svensson, G. P., U. Sahlin, B. Brage, M. C. Larsson. (2011). Should I stay or should I go? Modeling dispersal strategies in saproxylic insects based on pheromone capture and radio telemetry a case study on the threatened hermit beetle Osmoderma eremita. Biodivers. Conserv. 20: 2993–2902.

Taylor P. D., T. L. Crewe, S. A. Mackenzie, D. Lepage, Y. Aubry, Z. Crysler, G. Finney, C. M. Francis, C. G. Guglielmo, D. J. Hamilton, R. L. Holberton, P. H. Loring, G. W. Mitchell, D. R. Norris, J. Paquet, R. A. Ronconi, J. R. Smetzer. P. A. Smith, L. J. Welch, and B. K. Woodworth. (2017). The MOTUS wildlife tracking system: a collaborative research network. Avian Conservation and Ecology. 12, 8. doi: 10.5751/ACE-00953-120108

Telemetry Solutions. (2017). https://www.telemetrysolutions.com/wildlife-tracking-devices/gps- collars/micro-collars/ accessed 27 March 2020.

Thogmartin, W. E., L. Lopez-Hoffman, J. Rohweder, J. Diffendorfer, R. Drum, S. Semmens, S. Black, I. Caldwell. D. Cotter, P. Drobney, L. L. Jackson, M. Gale, D. Helmers, S. Hilburger, E. Howard, K. Oberhauser, J. Pleasants, B. Semmens, O. Taylor, P. Ward, J. F. Weltzin, and R. Wiederholt. (2017). Restoring monarch butterfly habitat in the Midwestern US: ‘all hands on deck’. Environmental Research Letters. 12, 074005. doi:10.1088/1748–9326/aa7637.

Tobias, J. A. and A. L. Pigot. (2019). Integrating behaviour and ecology into global biodiversity conservation strategies. Phil. Trans. R. Soc. B. 374: 20190012. doi: 10.1098/rstb.2019.0012

Turchin, P., F. J. Odendaal, and M. D. Rauscher. (1991). Quantifying insect movement in the field. Environ. Entomol. 20: 955–963.

Wang Z., Y. Haung, and N. E. Pierce. (2019). Radio telemetry helps record the dispersal patterns of birdwing butterflies in mountainous habitats: Golden Birdwing ( Troides aeacus) as an example. Journal of Insect Conservation. 23, 729-738. doi: 10.1007/s10841-019-00167-5

Ward, M., J. Sperry, and P. Weatherhead. (2013) Evaluation of automated radio telemetry for quantifying movements and home ranges of snakes. Journal of Herpetology 47: 337-345. doi: 10.1670/12-018.

White, G. C. and R. A. Garrott. (1990). Analysis of Wildlife Radio-Tracking Data. Academic Press, San Diego CA.

Wikelski, M., D. Moskowitz, J. S. Adelman, J. Cochran, D. S. Wilcove, and M. L. May. (2006). Simple rules guide dragonfly migration. Biology Letters 2, 325-329. doi: 10.1098/rsbl.2006.0487

Wittemyer, G., J. M. Northrup, and G. Bastille-Rousseau. (2019). Behavioural valuation of landscapes using movement data. Phil. Trans. R. Soc. B. 374: 20180046. doi: 10.1098/rstb.2018.0046 113

Wood, S.N. (2003). Thin-plate regression splines. Journal of the Royal Statistical Society (B). 65(1), 95-114. doi: 10.1111/1467-9868.00374

Wood, S.N. (2011). Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society (B). 73(1), 3-36. doi: 10.1111/j.1467-9868.2010.00749.x

Yagi-Uda.com (2008).verb@http://yagi-uda.com/three_elements_yagi-uda.php@ Web page accessed 9 April 2020.

Zalucki, M. P., and R. L. Kitching. (1982). The analysis and description of movement in adult Danaus plexippus L. (Lepidoptera: Danainae). Behaviour 80: 174–198.

Zalucki, M. P., H. R. Parry, and J. M. Zalucki. (2016). Movement and egg-laying in monarchs: to move or not to move, that is the equation. Austral Ecology. 41, 154–167. doi:10.1111/aec.12285

Appendix. Supplemental Information

To employ a similar system:

To employ a similar system, investigators will need to calculate their own constants for each of the formulae provided in this paper. Similar methods to those described should be utilized to understand the relationship between angle and power, the relationship between distance and power, as well as, any random effects of the towers, antennae, receivers, and field site locations.

The following is included to provide a glimpse into the effort required to install and calibrate the receiving towers. The first-time towers were augered into the ground, a full day of work was required; a gas-powered auger (BT 24 Earth Auger; Stihl; Waiblingen, Germany) proved useful. If holes and augers were already installed, it took approximately 45 minutes to erect one tower; four towers were on and functioning within four hours. When towers were left fully installed, but turned off overnight, it took approximately one hour to turn on all of the towers and ensure they were functioning. Distance, angle, and random location data collection could be completed within 2-3 weeks. 114

Table 1. Receiver settings. SRX800-D receiving units have the ability to collect power readings from more than one transmitter. However, it would decrease the temporal data resolution to include more than one transmitter at a time because the receiver would cycle through all Yagi antennae for one transmitter and then cycle through all Yagi antennae for the second transmitter; while it is recording one transmitter, it is not able to detect power from another transmitter. Category Criteria Setting Scan Settings GPS clock enabled Enabled Scan Time 6 seconds

Frequency and Channels Tag Types Beeper BPM Window 36 ± 20

Antenna Configuration SRX800D Antenna Switch Enabled Antenna Order 1, 2, 3, 4 Gain 65 Radio Master Options Not using a master

Filter Settings Filter Type No Filter

Sensors Code Set No Sensors

Figure 6. Power was recorded at 13 locations along a 100 m circle surrounding one receiving station at three gain settings: 50, 65, and 75. Zero values indicated when no power was detected. Based on these power-angle relationships, all gain settings provided similar information around the front of the antenna, but the gain setting of 65 exhibited limited potential "back-lobe" interference. This information provided justification to set the receiver gain to 65 for this study.

115

Figure 7. Variability in power for three configurations of transmitting antenna: parallel to the ground (horizontal orientation), straight up (vertical orientation), and bent so the electronics were horizontal but the transmitting antenna was vertical. For each antenna type and distance, power was recorded when the transmitter was oriented 8 directions, spaced 45 degrees around the circle. When the transmitter was held in horizontal orientation, power readings were more variable than when the transmitter was held in vertical orientation or when the transmitting antenna was bent.

Figure 8. Locations used to calibrate the model for power as a function of distance and angle from the transmitter to the receiving antenna.

116

Figure 9. Theoretical power vs angle from a 3 element Yagi antenna pointing in the indicated direction. Redrawn from data shown in 3el Yagi-Uda.pdf at yagi-uda.com (last accessed 9 April 2020, domain name expired 11 April 2020)

Figure 10. Using thin-plate regression smoothing provided an effective alternative to calculating the median power recorded over time. Dots and dashed lines show observed power readings for three antennae on one tower over time, colored lines show smoothed values. The effective degrees of freedom are 1.8, 2.0, and 4.6.

117

Figure 11. Received power as a function of distance for a transmitter worn by an investigator (in hat) and one mounted on a butterfly in a plant. 118

Figure 12. Received power over time at four antennae when a monarch was allowed to move freely through a field. Green segments at the bottom indicate when the buttery was visually observed to be flying. The system could detect when the buttery was stationary and when it was flying. At approximately 55000 seconds past midnight, the buttery was observed to y out of sight. The system indicates that it appears to be been stationary for approximately 12 minutes after that. 119

CHAPTER 5. ESTIMATING PERCEPTUAL RANGE OF FEMALE MONARCH BUTTERFLIES (Danaus plexippus) TO VEGETATIVE COMMON MILKWEED (Asclepias syriaca) AND NECTAR RESOURCES

Kelsey E. Fisher1 and Steven P. Bradbury1,2

1Department of Entomology, Iowa State University, Ames, IA, 50011

2Department of Natural Resource Ecology and Management, Iowa State University, Ames, IA,

50011

Modified from a manuscript submitted to Environmental Entomology in December 2020.

Abstract

Habitat loss in the summer breeding range contributes to eastern North American monarch (Danaus plexippus) population decline. Habitat restoration efforts include increasing native prairie plants for adult forage and milkweed (Asclepias sp.) for oviposition and larval development. Previous modeling efforts indicate simulated increases of monarch egg density in spatially-explicit or theoretical landscapes depend on spatial patterns of new habitat patches and assumptions of female movement behavior, including their perceptual range to resources. The lack of empirical data for monarch perceptual range introduces significant uncertainty in simulations. To refine estimates of perceptual range, we released 188 experimental monarch females 5, 25, 50, and 75 m downwind from potted milkweed and blooming forbs in 4-32 ha sod fields and quantified the extent of direct flights toward the resources. We also opportunistically observed flight patterns of 49 non-experimental, wild monarchs. 120

A unimodal pattern of movement was observed when monarchs were released 5, 25, or

50 m downwind of the resources; however, the majority of flights were with the wind rather than toward the resources. For those monarchs that flew into the wind, direct flights to the resources ranged from 3-125 m for eight experimental and 49 non-experimental monarchs. Based on these observations, we conclude that the perceptual range was functionally ≤ 50 m. We classified the dispersal behavior of the monarchs in this resource-poor habitat as direct displacement, based on the similarity of effective and total distances traveled and shallow turn angles. These results provide improved input estimates for spatially-explicit agent-based modeling and conservation plans.

Introduction

Discerning animal movement patterns in fragmented landscapes is foundational to understanding population dynamics, population persistence, and population distribution (Hanski

1998, Barton et al. 2009, Stevens et al. 2012, MacDonald et al. 2019). As animals disperse across a landscape, they are guided by olfactory and visual cues. Dispersal and movement behaviors are directly impacted by an animal’s perceptual range: the maximum distance individuals are able to detect stimuli with their sensory organs (Wiens 1989). For insects, directed movement behavior is performed when attractive stimuli are detected. Undirected or random walk behavior occurs in the absence of stimuli (Bell 1991, Carde and Willis 2008). Olfaction and vison likely work synergistically at a fine-scale (e.g., within meters of a resource), while olfactory cues alone attract insects from distances 30 meters or greater (Rutowski 2003, MacDonald et al. 2019).

Insect search behavior is well documented. Insects fly upwind, in and out of an odor plume, until they reach the source of an attractant chemical signal (Bell 1991, Carde and Willis

2008). Researchers take advantage of this behavior to study the role of olfactory cues by employing wind tunnels to study mate-finding or host-finding behaviors (Baker and Linn 1984, 121

Finch 1986). In the wind tunnel experimental design, the source of a volatile chemical is positioned at one end of the tunnel, and an insect is released at the opposite end. A fan blows a slow and steady air current from the stimulus toward the insect, and the insect’s response is observed.

At a landscape-scale, a mixture of olfactory cues complicates perceptual range studies for insects. Consequently, observational studies of perceptual range often are simplified by focusing on surrogates for the attractive stimuli, such as suitable habitat (Lima and Zollner 1996, Zollner and Lima 1997). Butterflies are exceptional models for these studies because individuals are easily released at varying distances from habitat edges and are easily observed within the limits of visual detection (MacDonald et al. 2019). Results indicate that butterfly perceptual range is highly variable, ranging from ≤ 1 m for cabbage white butterflies (Pieris rapae) (Fahrig and

Paloheimo 1987) to up to 100 m for the speckled wood (Pararge aegeria) (Merckx and Van

Dyck 2007). Although empirical evidence is lacking, expert opinion suggests that the perceptual range of monarchs (Danaus plexippus) is as large as 400 m (Grant et al. 2018).

Given its influence on movement, and therefore population dynamics, perceptual range is a key input parameter in many landscape-explicit population models (Lima and Zollner 1996).

One application of these models is for conservation planning. The decline of the overwintering eastern North American population of monarchs is associated with the loss of their obligate host plant, milkweed (Asclepias sp.), in part due to land conversion to row crop agriculture in their breeding range (Hartzler 2010, Pleasants and Oberhauser 2013, Pleasants 2017). As a result, plans for additional habitat establishment are underway (Midwest Association of Fish and

Wildlife Agencies 2018). Because the monarch is a vagile species, the spatial configuration of restored habitat can impact realized fecundity and subsequent adult recruitment (Zalucki et al. 122

2016, Grant et al. 2018). Agent-based modeling suggests establishing small habitat patches at distances within the monarch’s perceptual range supports increased egg densities in the landscape (Zalucki et al. 2016, Grant et al. 2018). Simulations of female movement in a spatially-explicit Iowa, USA landscape are, however, highly dependent on an assumed perceptual range between 50 and 400 m (Grant et al. 2018).

Here, we aim to advance understanding of a non-migratory female monarch perceptual range to suitable habitat and provide improved input estimates for spatially-explicit agent-based modeling and conservation plans. Suitable habitat for monarchs is often isolated milkweed stems or small milkweed/forage patches surrounded by fields of maize, soybean, or cool-season, non- native glasses (Zalucki and Kitching 1982a, Pleasants 2015, Baker and Potter 2019).

Consequently, we designed a field-scale ‘wind-tunnel’ study to quantify perceptual range by releasing monarch females 5 to 75 m downwind from potted common milkweed (Asclepias syriaca) and blooming forbs in 4-32 ha bluegrass fields.

Methods

Insects

As described in Fisher et al. (2020), non-migratory adult monarchs were caught in restored prairies and roadsides in Story, Dallas, and Linn County, Iowa, USA. Females were collectively housed in a soft-sided, mesh cage for 1-5 days before use in an experiment. Because collected monarchs starved overnight were lethargic upon release in an experimental trial the following day (unpublished observations), Gatorade (Chicago, IL) was provided ad libitum as a sugar source. On the morning of a trial, monarchs were transferred to individual glassine envelopes and transported to the field site in a cooler with an ice pack wrapped in a cloth.

123

Perceptual Range of Experimental Monarchs to Potted Resources

To estimate the perceptual range of monarchs toward isolated stimuli at distances beyond those possible under controlled laboratory conditions, we mimicked a wind-tunnel design under field conditions. The goal was to determine perceptual range by observing the extent of direct flight toward milkweed and nectar resources using radio-tagged, sham tagged, and untagged monarchs. Radio telemetry was employed to aid in relocating monarchs that exited the boundaries of experimental sod fields. Telemetry equipment included directional 3-element Yagi antennae (frequencies ranging from 164.000–166.000 MHz) connected to VHF receivers (Alinco

DJ-X11; Toyama, Japan) with 0.9-m coax cables (Johnson’s Telemetry, El Dorado Springs, MO)

As described in Fisher et al. (2020), monarchs were fitted with a 220 mg LB-2X transmitter

(Holohil Systems Ltd, Ontario, Canada) using superglue. Additional monarchs released without radio-tags (untagged) or with watch batteries attached (sham-tagged; 300 mg; Energizer

AZ10DP, St. Louis, MO) were used to evaluate potential effects of the transmitters on flight behavior (also see Fisher et al. 2020).

Experimental sites were 4-32 ha sod fields in Story, Dallas, and Linn County, Iowa,

USA, that were mowed weekly to 5 cm in height. These experimental sites, which were surrounded by corn and soybean fields, were selected to minimize the presence of confounding olfactory stimuli. Two milk crates containing five greenhouse-grown vegetative-stage Asclepias syriaca plants and one milk crate containing five greenhouse-grown blooming Liatris sp. were placed near the center of the sod fields-site and served as oviposition and feeding stimuli. A potted, vegetative-stage Monarda fistulosa or Liatris sp. was positioned either 5, 25, 50, or 75 m, because of field size constraints, downwind of the potted resources to serve as a release plant that provided no ovipositing or forage resources (Fig 1). Trials occurred on clear or mostly clear days from 700-1500 h when the temperature was between 20 and 33C and wind speed was between 0 124 and 16 kph. Prior to the release of each monarch, temperature and wind speed were measured with an anemometer (Extech Instruments; Nashua, NH; mini thermo-anemometer 45118), and wind direction was recorded. Release plants and potted resources remained at the same positions for all experiments on a given day even if the wind direction shifted slightly; consequently, resources were not always directly upwind of the release plant.

In 2017 and 2018, 112 radio-tagged, 60 sham-tagged, and 16 untagged female monarchs

(N = 188 total) were anesthetized with dry-ice (Fisher et al. 2020) and released individually.

Behavior was observed for 0.2-52 minutes. Monarchs were considered unresponsive if they did not move from the release plant within 10 minutes, and the observation period was terminated, similar to Zalucki and Kitching (1982). For monarchs that moved from the release plant, sequential stationary locations between flight steps were georeferenced (Trimble Geo7x;

Sunnyvale, CA) when feasible. Monarchs were caught by technicians as they approached the boundary of the sod field or when they were relocated with handheld radio telemetry in the surrounding agricultural fields. Technicians were positioned at least 25 m from the release plant.

Perceptual Range of Non-experimental Monarchs to Potted Resources

The potted milkweed and blooming Liatris sp. in the sod field acted as isolated stimuli in an otherwise resource-deprived landcover and attracted monarchs flying in the area. We opportunistically observed monarchs directly flying upwind to the potted resources. To estimate the perceptual range of these monarchs, investigators laid a flag at the monarch’s location where direct flight was first observed, and the distance from the flag to the resources was measured.

These observations were made under the same environmental conditions as described previously.

Statistical Analyses

Consistent with previous observations that a sham-transmitter does not impact monarch behavior (Fisher et al. 2020), we observed no significant difference in the frequency of 125 unresponsive individuals in the tagged (24/112), sham-tagged (13/60), and untagged (3/13) groups (Table 1; χ2=0.067975, df=2, p=0.9666), based on the number of days captive (Table 2;

χ2=3.6211, df=4, p=0.4597), or based on the release distance from potted resources (Table 3;

χ2=7.1975, df=3, p=0.06586). Radio-tagged, sham tagged, and untagged individuals, regardless of days in captivity, across 2017 and 2018, were combined for statistical analyses. All statistical analyses were performed in R version 4.0.3 (R Core Team 2020, RStudio Team 2020).

Directions of movement from release to capture, relative to the bearing in the direction of the resources (effective bearing), and the effective bearing of the first step (bearing from release point to first stationary location) were calculated using the movement.pathmetrics package in

Geospatial Modeling Environment (GME; Spatial Ecology LLC, www.spatialecology.com/gme/). Effective bearings were analyzed at each release distance with the R package ‘circular’ (Agostinelli 2017). Effective bearings for all individuals at a release distance were analyzed with the Rayleigh Test of Uniformity (rayleigh.test) to determine if effective movements exhibited a unimodal pattern. Release distances that were statistically unimodal were analyzed in comparison to the bearing in the direction of the resources and the wind direction with the Watson-Wheeler Test for Homogeneity of Means (watson.wheeler.test) to determine which stimulus guided movement.

We assumed monarchs were flying towards the resources when their effective bearings were ± 45 of the potted milkweed and blooming Liatris sp because this covers 90 or ¼ of the total directionality possibility. The frequency of flight steps ± 45 of the resources along movement paths, the number of individuals with effective bearings ± 45 of the resources, and the number of individuals that landed on the resources was analyzed by release distance with chi- square tests of independence. Wind speed and temperature during observations of individuals 126 whose effective bearings were ± 45 of the resources and individuals whose effective bearings were < -45 or > 45 of the resources were analyzed with two-sample t-tests.

To gain further insights on monarch movement patterns in the sod fields, additional movement parameters were calculated with the movement.pathmetrics package in GME, including effective distance from release to capture, step lengths between two consecutive stationary locations until the investigators lost sight or telemetry signal of a monarch that was not recovered. Turn angles were calculated from three consecutive stationary locations. Total distance traveled was calculated by summing all of the step lengths for an individual. Flight step lengths associated with landing on the resources, as well as steps within the sod field and associated with exiting the sod field were recorded.

Results

Perceptual Range of Experimental Monarchs to Potted Resources

When monarchs were released 5 or 25 m downwind of the potted resources, effective bearings relative to the resources at the point of capture displayed a highly significant unimodal pattern (Fig 2a; 5 m: n = 94, test statistic = 0.3242, p = 1e-04; 25 m: n = 32, test statistic =

0.6212, p = 0). The pattern of effective bearings from release to capture for monarchs released 50 m downwind from the resources was marginally unimodal (Fig 2a; n = 11, test statistic = 0.5309, p = 0.0412). Effective bearings from release to capture when monarchs were released 75 m downwind of the resources did not show a directional pattern (n = 11, test statistic = 0.4391, p =

0.1193). The first steps of movement paths followed a similar trend. Unimodal movement was observed when resources were placed 5, 25, and 50 m from the release site; however, a directional pattern was not observed at 75 m (Fig 2b; 5 m: test statistic = 0.2449, p = 0.0054; 25 m: test statistic = 0.6368, p = 0; 50 m: test statistic = 0.6442, p = 0.0074; 75 m: test statistic = 127

0.301, p = 0.3778). Interestingly, the unimodal flight patterns following releases at 5, 25, and 50 m were not oriented toward the resources. Rather, the monarchs predominately flew with the wind (Table 4). There was no relationship between wind speed (t = 1.778, df = 18, 120, p =

0.08679) or temperature (t = -0.72856, df = 18, 120, p = 0.4733) and the direction of movement.

Although the direction of monarch movement was predominately with the wind, 19 individuals flew upwind and within ± 45 of the resources (see Fig. 2). Eight of these monarchs landed on the Liatris sp. There was no significant effect of release distance on the frequency of steps towards the resource (χ2 = 12, df = 9, p = 0.2133), effective bearings to orient toward the resources (χ2 = 12, df = 9, p = 0.2133), or number of landings on the resources (χ 2 = 8, df = 6, p =

0.2381). Of the 94 monarchs that flew from the 5 m release point, 16 flew upwind toward the resource, and seven of them landed on the potted Liatris sp. Across all the monarchs that flew from the 5 m release point, 12.6% of their total flight steps (41/325) were oriented within ± 45 of the resources. Two of the 39 monarchs that flew from the 25 m release point had effective movement upwind, but neither of them landed on the resources. Eight percent of the total flight steps performed by all individuals that flew from the 25 m release point (9/111) were oriented within ± 45 of the resources. None of the monarchs released at 50 m (0/11) had effective movements in the direction of the resources. One of the 11 individuals that flew from the 75 m release point moved toward the resources and landed on the potted Liatris sp. Of the monarchs that flew from the 50 and 75 m release points, 3% (1/39) and 17% (6/35) of the total flight steps, respectively, were oriented toward the resources. Across all the release points, the length of the flight step associated with landing on the Liatris sp. were 3.5, 3.8, 4.9, 5.0, 5.0, 5.0, 5.0, and 59.7 m. 128

To further characterize movement, regardless of orientation with the wind or towards the resources, ancillary data for all 148 individuals that flew from the 5, 25, 50, and 75 m release points were collected. The mean effective distance (± SD) was 114.0 ± 182.5 m, with a maximum effective distance of 1,479.6 m (Fig 3). The mean total distance was 114.8 ± 182.0 m, with a maximum of 1491.8 m (Fig 3). The essentially identical effective and total distances indicate highly directional flight. Consistent with the similar effective and total distances, the majority of turn angles were close to zero (Fig 4), which also indicates the monarchs moved in nearly straight lines. Step lengths that were summed to calculate the total distance traveled were a mean of 53.9 ± 117.2 m, with a maximum step length of 800.6 m (Fig 5a). Flight steps within the sod field were a mean of 18.4 ± 24.1 m (Fig 5b). The 46 steps associated with exiting the sod field ranged from 13.8 to 800.6 m with a mean of 292.8 ± 192.5 m (Fig 5c).

Perceptual Range of Non-experimental Monarchs to Potted Resources

In 2017 and 2018, 86 non-experimental monarchs (sex unknown) were attracted to, landed on, and fed from the potted blooming Liatris sp. The behavior of 37 monarchs was not observed prior to their discovery on the plants. Forty-nine of these monarchs were observed flying directly upwind for 3-125 m to the resources (Fig 6). Thirty-five monarchs had direct flights between < 5 and 25 m to the resources, while 13 had direct flights of 50 to 100 m. One monarch had a direct flight of 125 m.

Discussion

To support conservation plans for the declining population of eastern North American monarchs, agent-based population models that include female movement algorithms suggest establishing small habitat patches at distances within the monarch’s perceptual range will more effectively increase egg densities within landscapes, as opposed to establishing patches in random locations (Zalucki et al. 2016, Grant et al. 2018). However, uncertainty in monarch 129 perceptual range, between 50 to 400 m based on expert elicitation, results in significantly different simulated movement patterns in a spatially-explicit Iowa, USA landscape (see Grant et al. 2018; Figure 5). We designed a field-scale ‘wind-tunnel’ study to quantify female monarch perceptual range and further inform input estimates for spatially-explicit, agent-based modeling and conservation planning. Radio-tagged, sham-tagged, and untagged monarch females were released 5, 25, 50, or 75 m downwind of potted Asclepias syriaca and Liatris sp. at several sod fields surrounded by corn, soybean, or non-native cool-season grasses. The goal was to determine the extent of directed flight into the wind toward the potted resources.

Overall, a unimodal pattern of movement was observed when monarchs were released 5,

25, or 50 m downwind of the resources; however, this movement was with the wind rather than toward the resources. No directional pattern was observed for individuals released 75 m downwind of the resources, likely because of a reduced sample size. Even in our largest field site, monarchs released 75 m downwind of the potted resources were within 80 m of the sod field boundary, and 5 of the 11 monarchs (45%) landed in the surrounding corn or soybean fields but could be relocated. A larger field site would provide the means to effectively assess flight patterns with a greater number of individuals released at larger distances from the resources.

For monarchs released 5, 25, and 50 m from the resource plants, the direction of movement was not influenced by wind speed, temperature, or presence of oviposition and forage resources within the sod fields. Rather, in this setting, monarchs flew with the wind for an average total distance of approximately 115 m over 0.2-52-minute observation periods. Most of the individuals that flew 19.6 to 1479.6 m with the wind landed in weedy fence rows, roadsides, trees, or crop fields and then remained stationary. The ability to relocate radio-tagged monarchs that exited a sod field further establishes the applicability of radio telemetry in larger landscapes, 130 as in Fisher et al. 2020. However, even when using radio telemetry, we were unable to locate the end of some flights that were greater than several hundred to over 1000 meters due to loss of signal because of interference from vegetation or because the transmitter was beyond the limit of detection. In 2019, we released 11 monarchs in an early-season soybean field 40, 60, or 80 m downwind of a restored prairie with the expectation that in a more natural setting, the monarchs would fly to the prairie (data not shown). Consistent with the results for monarchs released in the sod fields, only two individuals flew to the prairie (18%). Our observation that flying with the wind was a dominant behavior may be a response of the monarchs being released in an exposed, resource-deprived sod and early-season soybean fields. This prevalent movement behavior is often reported for migratory insects that use the wind to facilitate large distance dispersal

(Chapman et al. 2015); however, our experiments were conducted with females from non- migratory populations.

Despite the dominant trend of the monarchs moving with the wind, we observed 19 of the

148 experimental monarchs (12.8%) exhibiting effective movement in the direction of the resources, most from the 5m release distance. Eight of these monarchs landed on the Laitris sp.

Flights towards the resources may indicate the perception of an olfactory and/or visual stimulus; however, for the 11 monarchs that did not land on the resource, we can only speculate on the environmental stimuli that influenced their behavior. Of the eight experimental monarchs that landed on the resources, seven performed direct flight from ≤ 5 m, and one monarch’s direct flight to the resources was 59.7 m.

Eighty-six non-experimental monarchs were observed on the potted Laitris sp. Because our experiments were conducted in a resource deprived area, these monarchs were likely attracted during dispersal flights across the landscape. Distances of direct flight upwind to the 131 resources for 49 non-experimental monarchs where flight behavior was observed ranged from 3-

125 m, with 13 observations of direct flight between 50 to 100 m. Based on observations from experimental and non-experimental monarchs, we conclude that the monarch perceptual range to resources under our field settings was at least 50 m; however, a perceptual range up to 125 m is possible.

Non-migratory monarchs are a vagile species that traverse the landscape and lay eggs across embedded habitat patches (Zalucki and Lammers 2010, Zalucki et al. 2016, Grant et al.

2018). Zalucki et al. (2016) assumed monarch perceptual range of 5 or 25 m in an agent-based population model that was used to simulate female movement in an artificial landscape with varying spatial arrangements of habitat patches. In an agent-based model to explore monarch movement and egg distribution in a spatially-explicit Iowa USA landscape (Grant et al. 2018), perceptual range assumptions ranged from 50-400 m, based on expert elicitation. When model inputs included a perceptual range of 400 m, monarchs remained in high-quality habitat, and eggs accumulated in high-density clusters. Alternatively, when movement was modeled with a perceptual range of 50 m, movement patterns were considered more realistic, and eggs were distributed more evenly across high-quality habitat locations in the landscape. Our empirical evidence indicates a perceptual range assumption of 50 m is reasonable.

We designed a study to identify perceptual range by attracting monarchs to an isolated resource in a setting that would limit confounding chemical stimuli. While we likely limited exposure to confounding stimuli, the artificial experimental environment seemed to stimulate dispersal behavior. Because of the tortuousness of movements dispersal movement can be classified as either direct displacement or foray searches (Delattre et al. 2010). Direct displacement with shallow turn angles, also known as a correlated random walk, is presumed to 132 maximize survival probabilities and time and energy efficiencies in heterogeneous environments

(Heinz and Strand 2006, Schtickzelle et al. 2007, Barton et al. 2009). Tortuous foray searches maximize the chance of finding resources (Bell 1991, Delattre et al. 2010). As evidenced by the similarity between effective and total distance traversed, as well as 42% of the calculated turn angles between + 25, experimental monarchs observed in our study performed direct displacement dispersal. We observed 46 directional flight steps existing the sod fields ranging from 13.8 to 800.6 m with a mean of 292.8 ± 192.5 m. Similar observations of large, directional flight steps by monarchs exiting a prairie were reported by Fisher et al. (2020), and direct flights by monarchs moving through a grass matrix among small patches of milkweed were described by Zalucki and Kitching (1982b).

Observations in this study also identify opportunities to refine current agent-based model inputs with regard to assumptions concerning directionality and flight step lengths of non- migratory female monarchs. As noted above, directionality can be influenced by land cover classification with a direct displacement over unfavorable land cover and torturous movement over favorable habitats (Zalucki and Kitching 1982b, Zalucki 1983, Evans et al. 2020, Fisher et al. 2020). Consistent with this literature, Grant et al. (2018) assumed directionality increases as habitat quality decreases; thus, directionality for a monarch’s next step is dependent on the quality of habitat it currently occupies. Using an example from Grant et al. (2018), if directionality was set to a range of 0.5-0.75, directionality would be 0.5 in the best habitat (a habitat score = 0.09) and increase linearly to 0.75 in the poorest habitat (habitat score = 0). Grant et al. (2018) simulated realistic flight patterns when employing directionality ranging from 0.5 in high-quality habitat to 0.75 in low-quality land cover and assuming a 50 m perceptual range.

Results from this study and those of Fisher et al. (2020) suggest a non-linear relationship 133 between habitat quality and directionality could be explored across a broader range of directionality values. Results from a landscape-scale study in progress are quantifying movement patterns across crop fields, grass-dominated areas, road rights-of-way, and restored prairies.

Results from this study will provide additional information to assess relationships between habitat quality and flight directionality.

Likewise, results from this study and those in progress can provide information to refine input assumptions of step lengths over a diversity of land cover classes in agent-based models.

Grant et al. (2018) selected a step length of 30 m, in part due to limited empirical data, but noted variable step lengths could be drawn from a distribution derived from empirical observations of distances traveled in set time-periods. In the current study, we have established an initial step length distribution for a low-quality land cover with the majority of flight steps ≤ 50 m; however, with the aid of handheld radio telemetry, flight steps of up to 800.5 m were observed. Fisher et al. (2020) also reported occasional flight steps of several hundred meters when monarchs exited a restored prairie.

Olfactory and visual cues guide insect movement as they disperse across a landscape in search of resources (Bell 1991, Carde and Willis 2008), with directed movement in response to attractive stimuli and undirected movement when no stimuli are detected (Baker and Linn 1984,

Finch 1986, Bell 1991). Consequently, perceptual range is a key input parameter in landscape- explicit population models (Lima and Zollner 1996), and habitat restoration plans for species of interest should be created to facilitate behavioral attributes of movement (Lima and Zollner

1996, Zollner and Lima 1997). Our empirical evidence and model output estimates suggest an assumption of female monarch perceptual range to host plant and nectar forage resources of approximately 50 m (Grant et al., 2018) is reasonable. Establishing habitat patches 50 m apart 134 will likely create agroecosystems with spatial arrangements of resources that are appropriate for monarch movement behavior.

Data Accessibility

Data, metadata, and R code pertaining to this manuscript are publicly available through

GitHub: https://github.com/kelseyefisher/estimating-monarch-butterfly-perception-of-potted- resources

Acknowledgements

This work was supported by the Agriculture and Food Research Initiative Pollinator

Health Program (Grant 2018-67013-27541) from the USDA National Institute of Food and

Agriculture, Garden Club of America Board of Associates Centennial Pollinator Fellowship received in 2017, Holohil Grant Program received in 2018, ISU College of Agriculture and Life

Sciences, and the Iowa Monarch Conservation Consortium. We have a special appreciation for the property managers and landowners that allowed us to conduct experiments and catch monarchs on their property: Mike Loan and Sarah Nolte at Blue Grass Enterprises, Justin

Stensland at Stensland Sod, Adam Thomes and Ben Pease at the ISU Turf Research Station,

David Bruene at the ISU Beef Teaching Farm, Kent Berns at the ISU Dairy Farm, Daniel

Gudahl, Rob Davis, Penny Perkins, and Liz Garst at Whiterock Conservancy, Jesse Randell, and

Brab Licklider. We thank our tireless field technicians for their enthusiasm, support, and assistance: Cody Acevedo, Matthew Dollenbacher, Alec Euken, Signey Hilby, Jenna Nixt, Riley

Nylin, Michael Roth, and Kara Weber. Additional thanks to Dr. James Adelman for his help on experimental design and input prior to undertaking these experiments and Dr. Jarad Niemi for his guidance on statistical analyses. Finally, thank you to Myron Zalucki, Chip Taylor, Rick

Hellmich, and USDA-ARS CICGRU for incredibly valuable insight and guidance when developing and planning this project. 135

References

Agostinelli, C. 2017. Circular Statistics. R Package Version 0.4-93. https://cran.r- project.org/web/packages/circular/circular.pdf. Accessed 22 December 2020.

Baker, A. M., and D. A. Potter. 2019. Configuration and Location of Small Urban Gardens Affect Colonization by Monarch Butterflies. Front. Ecol. Evol. 7.

Baker, T. C., and C. E. Linn. 1984. Wind tunnels in pheromone research, pp. 75–110. In Hummel, H.E., Miller, T.A. (eds.), Techniques in Pheromone Research, Springer Series in Experimental Entomology. Springer, New York, NY.

Bell, W. J. 1991. Searching Behavior: The behavioural ecology of finding resources, Chapman and Hall Animal Behaviour Series. Springer Netherlands.

Carde, R. T., and M. A. Willis. 2008. Navigational strategies used by insects to find distant, wind-borne sources of odor. J Chem Ecol. 34: 854–866.

Chapman, J. W., D. R. Reynolds, and K. Wilson. 2015. Long-range seasonal migration in insects: mechanisms, evolutionary drivers and ecological consequences. Ecology Letters. 18: 287–302.

Delattre, T., F. Burel, A. Humeau, V. M. Stevens, P. Vernon, and M. Baguette. 2010. Dispersal mood revealed by shifts from routine to direct flights in the meadow brown butterfly Maniola jurtina. Oikos. 119: 1900–1908.

Evans, L. C., R. M. Sibly, P. Thorbek, I. Sims, T. H. Oliver, and R. J. Walters. 2020. The importance of including habitat-specific behaviour in models of butterfly movement. Oecologia. 193: 249–259.

Fahrig, L., and J. E. Paloheimo. 1987. Interpatch dispersal of the cabbage butterfly. Can. J. Zool. 65: 616–622.

Finch, S. 1986. Assessing host-plant finding by insects. In Miller, J.R., Miller, T.A. (eds.), Insect-Plant Interactions, Springer Series in Experimental Entomology. Springer, New York, NY.

Fisher, K. E., J. S. Adelman, and S. P. Bradbury. 2020. Employing very high frequency (VHF) radio telemetry to recreate monarch butterfly flight paths. Environ Entomol. 49: 312–323.

Hanski, I. 1998. Metapopulation dynamics. Nature. 396: 41–49.

Hartzler, R. G. 2010. Reduction in common milkweed (Asclepias syriaca) occurrence in Iowa cropland from 1999 to 2009. Crop Protection. 29: 1542–1544.

Heinz, S. K., and E. Strand. 2006. Adaptive patch searching strategies in fragmented landscapes. Evol Ecol. 20: 113–130.

Lima, S. L., and P. A. Zollner. 1996. Towards a behavioral ecology of ecological landscapes. Trends in Ecology & Evolution. 11: 131–135.

Merckx, T., and H. Van Dyck. 2007. Habitat fragmentation affects habitat-finding Ability of the speckled wood butterfly, Pararge aegeria L. Animal Behaviour. 74: 1029–1037.

Midwest Association of Fish and Wildlife Agencies. 2018. Mid-America Monarch Conservation Strategy. (http://www.mafwa.org/wp- content/uploads/2018/07/MAMCS_June2018_Final.pdf).

Pleasants, J. 2017. Milkweed restoration in the Midwest for monarch butterfly recovery: estimates of milkweeds lost, milkweeds remaining and milkweeds that must be added to increase the monarch population. Insect Conservation and Diversity. 10: 42–53.

Pleasants, J. M. 2015. Monarch butterflies and agriculture, pp. 169–178. In Oberhauser, K.S., Nail, K.R., Altizer, S. (eds.), Monarchs in a Changing World. Cornell University Press, Ithaca, NY.

R Core Team. 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

RStudio Team. 2020. RStudio: Integrated Development for R. RStudio, PBC, Boston, MA.

Rutowski, R. L. 2003. Visual ecology of adult butterflies, pp. 9–25. In Boggs, C.L., Watt, W.B., Ehrlich, P.R. (eds.), Butterflies: Ecology and Evolution Taking Flight. University of Chicago Press, Chicago, IL.

Schtickzelle, N., A. Joiris, H. Van Dyck, and M. Baguette. 2007. Quantitative analysis of changes in movement behaviour within and outside habitat in a specialist butterfly. BMC Evolutionary Biology. 7: 4. 137

Stevens, V. M., A. Trochet, H. V. Dyck, J. Clobert, and M. Baguette. 2012. How is dispersal integrated in life histories: a quantitative analysis using butterflies. Ecology Letters. 15: 74–86.

Wiens, J. A. 1989. Spatial Scaling in Ecology. Functional Ecology. 3: 385–397.

Zalucki, M. P. 1983. Simulation of movement and egglaying in Danaus plexippus (Lepidoptera: Nymphalidae). Res. Popul. Ecol. 23: 353–365.

Zalucki, M. P., and R. L. Kitching. 1982a. Dynamics of oviposition in Danaus plexippus (Insecta: Lepidoptera) on milkweed, Asclepias spp. Journal of Zoology. 198: 103–116.

Zalucki, M. P., and R. L. Kitching. 1982b. The Analysis and Description of Movement in Adult Danaus plexippus L. (Lepidoptera: Danainae). Behaviour. 80: 174–198.

Zalucki, M. P., and J. H. Lammers. 2010. Dispersal and egg shortfall in Monarch butterflies: what happens when the matrix is cleaned up? Ecological Entomology. 35: 84–91.

Zalucki, M. P., H. R. Parry, and J. M. Zalucki. 2016. Movement and egg laying in Monarchs: To move or not to move, that is the equation. Austral Ecology. 41: 154–167.

Zollner, P. A., and S. L. Lima. 1997. Landscape-Level Perceptual Abilities in White-Footed Mice: Perceptual Range and the Detection of Forested Habitat. Oikos. 80: 51–60.

Figures and Tables

138

Figure 2. Bearings of the effective distance from release point to capture (a) and of the first flight step from release point to first stationary location (b) of 148 responsive, experimental monarchs released at distances that displayed a unimodal movement pattern (5, 25, and 50 m from the potted resources). All effective bearings are represented as dots colored to correspond with their release distance. The mean bearing of monarch movement at each release distance is displayed as colored, solid arrows. As a frame of reference, the dot-dash green arrow is pointing from the release site to the resources; all directions of movement were normalized to the resources to be positioned at 0. Experimental monarchs were released on vegetative stage Monarda fistulosa 5, 25, 50, or 75 m from potted vegetative stage Asclepias syriaca and blooming Liatris sp in sod fields of 4-32 ha. 139

Figure 4. Histogram of turn angles estimated from three consecutive stationary locations along flight paths of 99 experimental monarchs released on vegetative stage Monarda fistulosa 5, 25, 50, or 75 m from potted vegetative stage Asclepias syriaca and blooming Liatris sp in sod fields of 4-32 ha. Forty-two percent of the calculated turn angles (105/250) were between + 25 degrees, which indicates monarch flight patterns were highly directional. 140

Figure 6. Histogram of opportunistically observed direct flights upwind to potted Asclepias syriaca and Liatris sp. by 49 non-experimental monarchs in 4-32 ha sod fields in 2017 and 2018.

Table 1. The number of individuals that flew or were unresponsive (UNR; no movement within 10 minutes of release) based on attachment treatment (χ2=0.067975, df=2, p=0.9666).

Treatment Flew UNR Total % UNR Radio- tagged 88 24 112 21.4 Sham-tag 47 13 60 21.7 Untagged 13 3 16 18.8 Overall 148 40 188 21.3

Days Captive Flew UNR Total % UNR 1 71 15 86 17.4 2 53 20 73 27.4 3 17 3 20 15.0 4 2 0 2 0.0 5 5 2 7 28.6 Overall 148 40 188 21.3

142

Table 3. The number of individuals that flew or were unresponsive (UNR; no movement within 10 minutes of release) based on the release distance to the potted resources (χ2=7.1975, df=3, p=0.06586).

Release Distance Flew UNR Total % UNR 5 m 94 30 124 24.2 25 m 32 3 35 8.6 50 m 11 6 17 35.3 75 m 11 1 12 8.3 Overall 148 40 188 21.3

Table 4. The unimodal movement patterns displayed at 5, 25, and 50 m releases were not oriented toward the resources.

Mean Effective Release Bearing Distance Relative to Resources F df p 5 m 172.8 249.24 1, 174 < 2.2e-16 25 m 146.5 323.13 1, 74 < 2.2e-16 50 m 203.5 68.62 1, 20 6.77E-08 75 m No unimodal movement pattern observed

Table 5. The unimodal movement patterns displayed at 5, 25, and 50 m releases were in the same direction as the wind.

Mean Effective Release Bearing Distance Relative to Wind F df p 5 m 2.5 0.13216 1, 172 0.7167 25 m 13.2 3.0201 1, 74 0.0864 50 m 26.0 0.61499 1, 20 0.4421 75 m No unimodal movement pattern observed

143

CHAPTER 6. INFLUENCE OF RESOURCE ABUNDANCE AND HABITAT CONFIGURATION ON FEMALE MONARCH BUTTERFLY (Danaus plexippus) MOVEMENT AND HABITAT UTILIZATION PATTERNS AT A LANDSCAPE-SCALE

Kelsey E. Fisher1 and Steven P. Bradbury1,2

1Department of Entomology, Iowa State University, Ames, IA, 50011

2Department of Natural Resource Ecology and Management, Iowa State University, Ames, IA,

50011

Modified from a manuscript to be submitted to Journal of Applied Ecology.

Abstract

To this end, we employed VHF radio telemetry to track monarchs. We aimed to understand how female monarch movement behavior and space use are influenced by different, complex spatial configurations of contiguous habitat classes with varying densities of milkweed and nectar resources. Radio-tagged monarch butterflies were released in maize fields, restored or remnant prairies, grass-dominated fields and roadside rights of ways, and habitat edges in an 800 m2 144 mosaic configuration of habitat classes and secondary gravel roadsides with and without access to resource-rich areas.

To date, this study provides the most robust dataset of directly tracked movement of individual breeding-season monarch butterflies to assess utilization of agricultural landscapes.

Monarch movements among habitat classes were tracked and, in some cases, were observed flying up-wind, consistent with search behavior. Likely in response to floral and milkweed density, step lengths and turn angles were mostly below 50 m and directional, regardless of habitat class or spatial configuration. Consistent with results of previous reports, we observed large exiting steps from the research area of up to ≥ 1,000 m, which suggests dispersal in search of a new research-rich habitat and in turn results in eggs being distributed across the landscape.

In support of this hypothesis, 78.9% of monarch not released in a resource-rich habitat flew to resource rich habitat, with some flying over 500 m before locating forage or oviposition resources. This research advances understanding of movement behavior at a landscape scale and would not have been possible without the aid of radio telemetry to relocate individuals beyond human visual range.

Introduction

Understanding animal movement and space use ecology is fundamental to advancing conservation biology and natural resource management programs (Kissling et al. 2014, Kays et al. 2015, Chan et al. 2019, Nandintsetseg et al. 2019, Tobias and Pigot 2019, Wittemyer et al.

2019). Population dynamics, persistence, and distribution are emergent properties of animal movement behavior and habitat utilization (Hanski 1998, Barton et al. 2009, Stevens et al. 2012,

Moorter et al. 2016). Landscape-scale movements of insects across fragmented habitats are often inferred with indirect methods such as trapping, molecular markers, and mass mark-recapture

(Reynolds et al. 1997, Southwood and Henderson 2000, Osborne et al. 2002). Direct methods to 145 study movement are more precise than indirect methods. However, tracking individual insects over large distances is very difficult (Osborne et al. 2002); successful examples have been reported for visual observations within 50 m of an observer (Zalucki and Kitching 1982, Schultz and Crone 2001, Merckx and Van Dyck 2007, Delattre et al. 2010, Schultz et al. 2012, 2017,

Kallioniemi et al. 2014, Fernández et al. 2016). A limited number of successful individually labeled mark-recapture studies have been reported (Urquhart 1987, Showers 1997, Southwood and Henderson 2000, Ide 2002), while the application of tracking technologies are rare (Osborne et al. 2002, Kissling et al. 2014, Wang et al. 2019, Fisher, Adelman, et al. 2020).

The eastern North American monarch butterfly is at risk of quasi-extinction (Semmens et al. 2016) and determined to be a candidate species for listing under the Endangered Species Act

(USFWS 2020). The population decline is due, in part, to the loss of breeding habitat in agricultural landscapes of the Midwest United States (Thogmartin et al. 2017). Because adult females are not patch-residents, egg abundance and distribution across the landscape is a function of their perceptual range, flight directionality, and flight step lengths (Zalucki et al.

2016, Grant et al. 2018). Agent-based modeling with parameters derived from limited empirical evidence suggests establishing small habitat patches at distances within the monarch’s perceptual range will support increased egg densities and subsequent adult recruitment in the landscape

(Zalucki et al. 2016, Grant et al. 2018). To refine model simulations, and recommendations to conservation practices, field observations of perceptual range, flight directionality, and flight step lengths at a landscape scale are necessary (Grant et al. 2018, Grant and Bradbury 2019).

Monarch butterflies have been successfully located with handheld VHF radio telemetry in a 4-ha restored prairie (Fisher, Adelman, et al. 2020) and 6.25-ha sod field (Fisher and

Bradbury 2021) and with an automated VHF radio telemetry system in a 6.25-ha sod field 146

(Fisher, Dixon, et al. 2020). Data from these studies provide the means to recreate flight paths and analyze movement patterns in areas larger than reported in previously published studies.

These data also generated empirical information of perceptual range, step length, and directionality for model parameterization. However to increase understanding of movement and space use across larger landscapes, tracking studies in areas with more complex spatial patterns of restored breeding habitat within agricultural settings are needed (Fisher, Adelman, et al.

2020).

Here, we hypothesize that female monarch movement behavior and space use are influenced by different spatial configurations of contiguous habitat classes with varying density of milkweed and nectar resources. Monarch butterflies were released in maize fields, restored or remnant prairies, grass-dominated fields, roadside rights of ways, and habitat edges. Landscape settings included an 800 m2 mosaic configuration of habitat classes and secondary gravel roadsides with and without access to resource-rich areas. Monarchs were visually observed for up to one hour with the aid of handheld and automated radio telemetry. We predict that monarchs released in these configurations would move among habitat classes and orient into the wind, consistent with insect search behavior (Bell 1991, Carde and Willis 2008). Based on model simulations and reports from previous studies (Zalucki and Kitching 1982, 1984, Zalucki et al.

2016, Grant et al. 2018, Fisher, Adelman, et al. 2020, Fisher and Bradbury 2021), we predict that monarchs would perform directed movement with large steps in resource-poor areas and undirected movement with small steps when forage and oviposition resources are abundant.

Additionally, we predict that independent of release location, monarchs would locate and preferentially use areas with abundant milkweed and blooming forbs (nectar resources) as compared to areas without these resources. To date, this study provides the most robust dataset 147 of directly tracked movement of individual breeding-season monarch butterflies to assess utilization of agricultural landscapes.

Methods

Field Sites

Three habitat configurations were selected: 1) an 800 m2 focal research area in Floyd

County, Iowa consisting of a mosaic of CP42 restored prairie established in 2015, grass- dominated fields, and crop fields (Fig 1a); 2) a 536 m linear focal research area of a secondary gravel roadside corridor boarded by maize to the west and the Anderson-Dyas remnant prairie to the east in Story County, Iowa (Fig 1b green box); and 3) a 536 m linear focal research area of a secondary roadside corridor bordered by maize and soy fields to the east and west that was > 500 m from prairie sites in Story County, Iowa (Fig 1b blue box). These field sites will be referred to as “mosaic,” “roadside corridor with resource-rich habitat,” and “roadside corridor without resource-rich habitat,” respectively (Table 1). Field studies were undertaken in July – August

2019 at the mosaic configuration and June – August 2020 at the roadside corridor configurations.

Vegetation surveys were conducted along 100 m transects in crop fields, grass-dominated areas, and remnant/restored prairies in mid-July of each year to estimate milkweed and blooming floral resource abundance. The percent of grasses, forbs, milkweed, woody vegetation, bare ground, and litter were estimated within a 0.5 m2 Daubenmire frame, with associated bins of

“zero”, “trace”, “3%”, “16%”, “38%”, “63%”, “86%”, and “98%”, every 10 m along the 100 m transect. At the same locations along the transect, estimates of vegetation height to the nearest 10 cm were made with a Robel pole. Additionally, blooming inflorescences (e.g., solitary, spike, raceme, panicle, corymb, cyme, and umbel) and milkweed ramets were counted within a 1 m band parallel to the 100 m transect (Table 2). Survey transect locations and data can be found in the digital repository associated with this project. 148

Based on results from these surveys, areas were classified as “crop field,” “resource-poor habitat,” or “resource-rich habitat” (Table 1; Table 3; Fig 1). Crop fields were either mid-season maize or soybean fields. Based on Daubenmire estimates, resource-poor habitat sites (N = 6) had a mean of 69.6% grasses and 3.2% forbs with 30.5% bare ground and 69% litter that averaged

4.1 cm in depth. Blooming species found in the resource-poor habitats included Cirsium arvense

(non-native; noxious weed), Daucus carota (native), Erigeron strigosus (native), Lactuca serriola (non-native), and Pastinaca sativa (non-native). Resource-rich habitats (N = 7) were a mean of 48.7% grasses and 35.2% forbs with 26.5% bare ground and 71.5% litter that averaged

3.25 cm in depth. Monarda fistulosa (native) and Ratibida pinnata (native) were abundant with >

90 blooms/100 m in all the resource-rich habitats. Additional species that were abundant in at least one of the seven resource-rich habitats, included Erigeron stigosus (native), Heliopsis helianthoides (native), Melilotus albus (non-native), Melilotus officinalis (non-native),

Rudbeckia hirta (native), and Rudbeckia triloba (native). Additional, less abundant blooming species present in resource-rich habitats included Cirsium arvense (non-native; noxious weed),

Hypericum perforatum (non-native), Oenothera biennis (native), Pastinaca sativa (non-native),

Pycnanthemum virginianum (native), Trifolium pratense (non-native), and Verbena stricta

(native).

Monarchs

Adult female monarch butterflies were caught in multiple restored prairies and roadsides, including our focal research areas in Floyd, Story, and Boone counties in central Iowa. For 1-4 days post-capture, monarchs were housed in a soft-sided, mesh cage (pop-up laundry basket; 57

× 37 × 55 cm; Honey-Can-Do HMP-03891 Mesh Hamper with Handles, Walmart, Rogers, AK with “no-see-em” netting; Arrowhead Fabric Outlet, Duluth, MN). Gatorade (Chicago, IL) was provided ad libitum as a sugar source (Fisher, Adelman, et al. 2020). On the day of an 149 experimental trial, monarchs were transferred to individual glassine envelopes and transported to a field site in a cooler with an ice pack wrapped in a cloth. Experimental trials occurred on days with a temperature of 18.6 - 35.4C and wind speed below 15.0 kph. Trials at the mosaic field site were undertaken from July 2 - August 9, 2019 and from June 24 - August 4, 2020, at the roadside corridor sites.

Tracking Female Monarch Butterflies

Monarchs were radio-tagged with a LB-2X transmitter (220 mg, Holohil Systems Ltd,

Ontario, Canada) attached to their ventral abdomen with super glue (Fisher, Adelman, et al.

2020). Additional monarchs without transmitters (untagged) served as controls. After being anesthetized with CO2 for one minute (Fisher, Adelman, et al. 2020), monarchs were released individually on smooth bromegrass (Bromus inermis) or a maize plant (mosaic site only). To assess the influence of release habitat on behavior, at the mosaic site monarchs were released once in a maize field (n=22), resource-rich habitat (n=25), resource-poor habitat (n=22), or in the habitat boundary between a crop field and resource-rich habitat (n=33) (Table 1). Monarchs were released once in the resource-poor, grassy roadside on the east side of the roadside corridors with

(n=55) or without (n=38) access to a resource-rich habitat (Table 1) to explore if the presence of a resource-rich area within 5 m of the release site influences monarch movement patterns.

Monarchs that did not move from their release location within 15 minutes were considered unresponsive and the observation period was terminated. At the roadside corridor sites, the definition of an unresponsive monarch was modified to include individuals that moved from the release location but did not move a second time within 30 minutes to ensure movement patterns were observed. For monarch butterflies that moved from the release plant, stationary locations between flight steps were visually confirmed by technicians stationed within the focal 150 research area, marked with flagging tape, and georeferenced (Trimble Geo7x; Sunnyvale, CA).

In cases where monarchs flew beyond the limit of visual detection (ca. 50 m), monarchs were relocated with the aid of handheld radio telemetry equipment, including directional 3-element

Yagi antennae (frequencies ranging from 164.000–166.000 MHz) connected to VHF receivers

(Alinco DJ-X11; Toyama, Japan) with 0.9-m coax cables (Johnson’s Telemetry, El Dorado

Springs, MO). Monarch capture was attempted after a one-hour observation period within the focal research areas, or until they were relocated with the aid of handheld radio telemetry outside of the focal research areas, or when unsuccessful relocation efforts elapsed three hours.

Additionally, at the roadside corridor site, capture was attempted when a monarch crossed 20 rows into a bordering maize or soy fields.

Visual observation and handheld radio telemetry were augmented with an automated radio telemetry system (Fisher, Dixon, et al. 2020) at the mosaic field site. A square grid of nine towers, each separated by approximately 250 m, was employed to create 500 m2 detection area within over the 800 m2 focal research area (Fig 1a). Power readings collected with this system were stored and analyzed post-hoc in the lab; however, because of vegetation interference and limited power emitted by our transmitters, location estimates were not possible (see supplemental). Because of concern for investigator safety, limited personnel support due to

SARS-CoV-2 restrictions, and potential signal interference from utility lines, the tower system was not employed at the roadside corridor sites.

Statistical Analyses

Responsive individuals in all tagging treatments were combined for statistical analyses.

We observed no significant effect of transmitter attachment (Table 4; χ 2 = 0.023007, df = 1, p =

0.8794) or time in captivity post-capture (Table 5; χ2 = 6.0703, df = 4, p = 0.194) on monarch butterfly responsiveness at the 15 min threshold when analyzed with chi-square tests of 151 independence (chisq.test in RStudio version 4.0.3 (R Core Team 2020, RStudio Team 2020), similar to previous reports (Fisher, Adelman, et al. 2020, Fisher and Bradbury 2021). Release habitat had a marginal influence on the rate of responsiveness at the 15 min threshold (Table 6;

χ2=8.536, df=3, p=0.03614), suggesting that release site did not create a consistent bias in responsive monarchs across all sites. However, monarchs released in the mosaic configuration at the habitat edge between the maize field and a resource-rich field (n=14) were more unresponsive than monarchs released in resource-poor fields (χ2=7.11, df=1, p=0.007665).

Monarch habitat preference was analyzed with several approaches. The number of monarchs that remained in their release habitat or crossed a habitat boundary during their observation period were summed by release habitat and habitat configuration and analyzed with chi-square tests of independence (chisq.test) in RStudio version 4.0.3. Of the individuals that crossed a boundary at least once, the total number of crossings was analyzed by release habitat with a generalized linear model using the ‘emmeans’ package in RStudio (Lenth et al. 2020).

The total time individuals spent in each habitat classification were analyzed by habitat configuration, release habitat, and habitat classification with generalized linear models using the

‘emmeans’ package in RStudio. Habitat visitations by monarchs from each release habitat were summarized.

Locations of monarchs were unknown while in flight or when observers lost visual detection. Assuming direct flight between stationary locations, straight-line analyses were used to connect georeferenced locations. Step lengths (the distance between two consecutive stationary points) and turn angles (angles created from three consecutive points) were calculated using the ‘movement.pathmetrics’ package in Geospatial Modeling Environment (GME; Spatial

Ecology LLC, www.spatialecology.com/gme/) in combination with ArcMap 10.3. Step lengths 152 and the time elapsed prior to steps greater than 50 m were analyzed by habitat configuration, release habitat, and habitat classification with generalized linear models using the ‘emmeans’ package in RStudio. Distributions of turn angles within habitat classifications were analyzed with the Watson-Wheeler Test for Homogeneity of Angles with the ‘circular’ package in

RStudio (Agostinelli 2017). The orientation of effective movement (bearing of movement from start to end of the observation period) and effective movement relative to wind direction (wind direction minus orientation of effective movement) were analyzed with the Rayleigh Test of

Uniformity with the ‘circular’ package in RStudio.

Because movement is likely not a straight line between two stationary points, locations between the known stationary locations were estimated with occurrence models created with the

‘continuous-time movement model’ (CTMM) package in RStudio (Calabrese et al. 2016,

Fleming et al. 2016, Fleming and Calabrese 2020). CTMM is employed to estimate an animal’s location between sampling location fixes. Model estimates resemble the animal’s movement patterns more accurately with increased sampling frequency. Occurrence models were selected based on semivariogram estimates with the smallest AIC value for individual monarchs. The output of the models produced raster-images scaled with the estimated probability of occurrence in each cell, which ranged from zero to one. Raster cells with non-zero values that ranged from

0.10 to 0.33 were classified as “low probability,” from 0.34 to 0.66 were classified as “moderate probability,” and from 0.67 to 1.00 were classified as “high probability.”

Raster-images were resampled to create images with 0.05 m2 cells in ERDAS Imagine

2020 (Hexagon Geospatial, Madison, AL). Geotiff images of the field sites were digitized and converted to raster images with 0.05 m2 cells in ArcMap 10.7. Raster cells were scaled by habitat class, including “crop,” “resource-poor,” “resource-rich,” and “edge”; edge raster cells were 153 those within 2.5 m of any boundary line. Occurrence models for each butterfly were clipped to each habitat class to identify the number of raster cells with an occurrence probability in each habitat class. The proportion of cells with an occurrence probability was calculated by dividing the number of cells with a non-zero occurrence probability within a habitat class by the total number of cells within the same habitat class. To identify areas with moderate to high probability, the number of cells with probabilities greater than 0.33 were summed by habitat class and divided by the total number of cells in the class. Proportions were logit transformed

(Warton and Hui 2011) and analyzed by release habitat and habitat class with generalized linear models using the ‘emmeans’ package in RStudio.

Results

One hundred and ninety-five female monarchs were released at georeferenced locations across the three habitat configurations: habitat mosaic, roadside without resource-rich habitat, and roadside with resource-rich habitat. One hundred and fourteen monarchs were considered responsive at the 15-min threshold at the mosaic configuration and 30-min threshold at the roadside configurations and were observed for 75 – 7,500 s (1.25 – 125 min). Twenty-one of the responsive monarchs remained in their release habitat for the entirety of their observation period

(3,600 s). The other 93 monarchs crossed a habitat boundary at least once. There was no difference in the number of individual monarchs that crossed habitat boundaries based on release habitat (chi.test: χ2 = 30, df =25 p = 0.2243) or habitat configuration (chi.test: χ2 = 12, df = 10, p- value = 0.2851).

Seventy-five responsive monarchs were released in the mosaic habitat configuration. Of the 18 responsive monarchs that were released in the maize field, two remained in the crop field for the entirety of their observation period (3,600 s). The remaining 16 monarchs left the crop field after 38 to 1,136 s. Of these, 12 spent a mean of 793.2 ± 933.8 s in a resource-rich habitat, 3 154 spent 381 to 1,483 s in a resource-poor habitat, and two returned to a crop field after 227 and

1,306 s. Of the 18 responsive monarchs released in a resource-rich habitat, nine remained in the habitat for the entire observation period; six exited the focal survey area after spending 45 to

2,319 s in the resource-rich habitat. One monarch crossed into a crop field after 362 s, one crossed into a resource-poor habitat after 478 s, and one monarch crossed to a crop field after

1,154 s and returned to a resource-rich habitat after an additional 102 s. Fifteen of 20 responsive monarchs released in a resource-poor habitat exited their release habitat after 49 to 2,325 s, while five individuals remained in the resource-poor habitat for 3,600 s. Eight of the 15 monarchs that left resource-poor habitat spent 1,882.7 ± 726.7 s in a resource-rich habitat, one visited a crop field for 420 s, and one returned to the resource-poor habitat after 3,512 s. Twelve of the 19 responsive monarchs released on the habitat boundary between a resource-rich habitat and a maize field entered the resource-rich habitat. Five initially entered the crop field, with three of these individuals subsequently flying to the resource-rich habitat. Two of the 19 individuals immediately flew from the focal research area and did not land in the resource-rich habitat or the maize field and were not subsequently captured.

Thirty-nine responsive monarchs were released in grassy roadsides at the two roadside corridor habitat configurations. Nineteen responsive monarchs were released on a roadside without resource-rich habitat. Fourteen of these monarchs exited the grassy roadside and entered bordering maize or soybean fields. Two of these individuals subsequently returned to the roadside. One individual flew 500 m across a maize field and was relocated in a restored prairie area. In the roadside with a resource-rich prairie remnant present within 5 m of the release location, nine of 20 responsive monarchs flew to the resource-rich habitat. At various times during the 3,600 s observation periods, 11 monarchs entered a maize field; however, seven 155 returned to the roadside, and two subsequently flew to the remnant prairie. Monarchs released in the roadside with access to a resource-rich habitat crossed habitat boundaries more often than monarchs released in the other two configurations (Fig 2. joint_tests: df = 5, 108, F = 8.994, p <

0.0001; pairs: z < -3.785, p < 0.0021).

The orientation of effective movement was analyzed based on the release location rather than the release habitat because geographic characteristics, such as proximity to an edge or view of a high point on the horizon, may influence movement behavior (O. Taylor, personal communication). Responsive monarchs were released at 13 locations across the release habitats and three habitat configurations. Nine locations had greater than four responsive monarchs released, while the other four locations only had one responsive monarch. These four locations were removed from the analysis. Movement of monarchs at eight of the nine locations with greater than four responsive monarchs displayed no directional pattern (0.2276 > test statistic <

0.4838, p > 0.1209). Monarchs released in the mosaic configuration in the habitat edge between the east side of the maize field and the west side of the restored prairie (see Fig 1a) had a directional pattern toward the prairie (Rayleigh test: test statistic = 0.5512, p=0.0031; mean direction of movement = 137.0).

Although monarchs flew with the wind when released in a sod field (Fisher and Bradbury

2021), it was hypothesized that releases in diverse land cover would elicit flights into the wind, assuming flight behavior is influenced by olfactory cues for monarchs seeking nectar forage or milkweed for ovipositing. Monarchs at seven of the nine release locations with greater than four released monarchs showed no pattern of directional movement relative to the wind direction

(Rayleigh test: 0.1042 > test statistic < 0.4898, p > 0.1142). Directional movement into the wind was observed with monarchs released on the roadside without access to a resource-rich area (see 156

Fig 1b blue box; Rayleigh test: test statistic = 0.4085, p=0.0474; Watson-Williams: mean of wind = 180, mean of movement = 1.33, F = 75.639, df1 = 1, 35, p-value = 2.864e-10).

Directional movement into the wind was also noted with monarchs released in the mosaic configuration on the habitat edge between the east side of the maize field and the west side of the prairie (see Fig 1a; Rayleigh: test statistic = 0.6132, p=6e-04; Watson-Williams: mean of wind =

180, mean of movement = 342, F = 182.6, df1 = 1, 34, p-value = 3.139e-15).

Across all the monarchs, 748 step lengths based on the distance between two georeferenced stationary points were calculated. An additional 42 step lengths were estimated based or the distance between a georeferenced stationary point and an estimated location based on last visual observation or last radio signal from a monarch not recovered. Step lengths ranged from 0.16 m to 1919.63 m; 90.5% (715/790) of the steps were below 50 m (Fig 3). Steps greater than 50 m were classified as “large steps”; most large steps were initiated in resource-rich habitats (Table 7). There was no difference in elapsed time prior to a large step based on the length of the step (50-99m, 100-249m, 250-499m, 500-999m, or >1000m; F=1.739, df=4, 70, p=0.1383) or the habitat of step initiation (F = 0.629, df = 2, 72, p = 0.5331). However, monarchs in the mosaic configuration initiated large steps earlier in the observation period

(863.90 ± 795.23 s) than those in both the roadside without resource-rich habitat (1806.00 ±

970.77 s) and the roadside with resource-rich habitat (1858.67 ± 878.80 s) (Fig 4; joint_test: F =

12.016, df = 2, 72, p < 0.0001; pairs: z < -3.667, p < 0.0001).

Steps between two points within the same habitat classification were categorized “within habitat” (n = 631) and those that spanned two habitat classifications were categorized “cross habitat boundary” (n = 159). Step lengths within a habitat (17.23 ± 50.43 m) were shorter than steps when monarchs crossed a habitat boundary (82.92 ± 243.25 m) (Fig 5a. F = 45.531, df = 1, 157

788, p < 0.0001). There was no influence of habitat configuration (F = 0.729, df = 2, 788, p =

0.4531) on step length. There was also no effect of habitat class on step lengths (Fig 5b. F =

2.285, df = 2, 628, p = 0.1017); however, habitat configuration had an impact on step lengths within habitat classes (joint_tests: F = 4.599, df = 2, 628, p = 0.0101), with longer step lengths in crop fields at the roadside without resources (57.8 ± 108.2 m) than in resource-poor habitat (14.1

± 34.9 m) or resource-rich habitat (15.6 ± 47.5 m) at the mosaic configuration in the mixed land use classification (pairs: z < -3.433, p < 0.0174).

There was no influence of habitat configuration on step lengths exiting (F = 0.841, df = 2,

156, p = 0.4314) or entering (F = 0.857, df = 3, 155, p = 0.4244) habitat classes. Monarchs took larger steps when exiting resource-rich habitat (188.71 ± 418.29 m) than resource-poor habitat

(Fig 5c; 33.08 ± 48.46 m; joint_tests: F = 6.614, df = 2, 156, p = 0.0013; pairs: z=-3.403, p=0.0019) or crop fields (48.08 ± 109.69 m; pairs: z = -2.942, p = 0.0091). Larger steps were taken when monarchs exited the survey areas (218.90 ± 447.98 m) than when they entered resource rich (Fig. 5d; 27.91 ± 24.93 m; joint_tests: F=6.09, df = 3, 155, p = 0.0004; pairs: z =

3.796, p = 0.0008), resource poor (67.86 ± 144.34 m; pairs: z = 2.597, p = 0.0464), or crop fields

(32.66 ± 60.87 m; pairs: z = -3.684, p = 0.0013).

Steps crossing into a resource-rich habitat provided the means to estimate monarch butterfly perceptual range. Forty monarchs crossed into resource-rich habitat 47 times. These steps ranged from 0.36 m to 126.79 m (Fig 6a). Fourteen monarchs took 14 steps into a resource- rich habitat while flying ±45 into the wind (Fig 6b), which provided the opportunity to draw insights about the perceptual range of odor plumes and olfactory stimuli. Steps into resource-rich habitat while flying into the wind ranged from 0.53 m to 126.79 m. Eight-six percent (12/14) of 158 the steps flying into the wind that resulted in monarchs landing in resource-rich habitats were below 50 m.

Turn angles created from three consecutive stationary locations were analyzed by habitat class with the hypothesis that turn angles would be narrow (high directional flight) in crop fields and resource-poor habitats because of the lack of milkweed and floral diversity. Conversely, turn angles were expected to be wide (low directional flight) in resource-rich areas because of the presence of forage and oviposition resources. In all habitat classifications, turn angles followed a normal distribution with narrow turn angles occurring most abundantly, which suggests highly directional movement in general (Fig 7). Within each habitat configuration, there was no difference in the distribution of turn angles based on habitat class (3.3707 > W < 5.5162, df = 4, p > 0.1854). Likewise, there was no difference in the distributions of turn angles within resource- rich habitats (W = 1.3377, df = 2, p-value = 0.5123) or crop-fields (W = 6.7254, df = 4, p-value

= 0.1511) across habitat configurations. However, the distributions of turn angles within resource-poor habitats depended on habitat configuration with a wider distribution of turn angles in the mosaic configuration than in both roadside corridors (W = 22.845, df = 2, p-value =

1.095e-05), suggesting that flights along roadsides are more directional.

Because the location of monarchs in flight was unknown and movement is likely not a straight line between two points, the occurrence probability for each monarch in each habitat classification was estimated with continuous-time movement models. Models were generated for

15 of 18 monarchs released in maize, 12 of 19 released on the edge of maize and prairie, 18 of 20 released in resource-poor habitat, 15 of 18 released in resource-rich habitat, 18 of 19 released on the roadside with access to resource-rich habitat, and 18 of 20 released on the roadside with 159 access to resource-rich habitat. Models were not successful for 18 monarchs because the observation period did not provide enough data to assess estimate uncertainty.

The habitat class where a monarch was released had no effect on the proportion of raster cells in each habitat classification with a non-zero probability of occurrence (Fig 8a; F = 0.060, df = 5, 378, p = 0.9976) or on the amount of time monarchs spent in each habitat class (F=0.632, df=5, 108, p=0.6753). When monarchs were released in resource-rich habitat, resource-poor habitat, or a crop field, the highest percent of raster cells with non-zero occurrence probabilities were within their release habitat, 73.4%, 63.8%, and 36.7%, respectively. The second-highest percentage of occurrence probability for monarchs released in resource-poor habitats and crop fields was within a resource-rich habitat, 29.7%, and 36.5%, respectively. Similarly, when monarchs were released on the edge of a resource-rich habitat and a crop field, 58.0% of their occurrence was within resource-rich habitats, and only 3.2% of their occurrence was in crop fields. Monarchs released in the roadside corridors had the highest probability of occurrence in resource-poor habitats, which were grassy roadsides. Those released on the roadside with a resource-rich habitat had 13.4% of their non-zero occurrence probabilities in resource-rich habitat. Although resource-rich habitat was over 500 m from the release site on the roadside without resource-rich, 4.4% of raster cells with occurrence probabilities were within a resource- rich habitat, showing that monarchs were successful at locating these areas.

There was a higher proportion of cells with a non-zero probability of occurrence in resource-rich habitats and resource-poor habitats than in crop fields (Fig 8b; joint_tests: F =

6.368, df = 3, 380, p = 0.003; pairs: z < -3.260, p < 0.0061) and no difference when resource-rich habitats, resource-poor habitats, and crop fields were compared to edge habitat (pairs: z > -2.253, p > 0.1094). Likewise, based on visual observations, monarchs spent less time in crop fields than 160 in resource-poor or resource-rich habitats (joint_tests: F = 6.604, df = 2, 111, p = 0.0014; pairs: z

< -2.820, p < 0.0133). There was a higher proportion of raster cells with a moderate (0.34 to

0.66) or high (0.67 to 1.00) probability of occurrence in resource-poor habitats than in crop fields

(Fig 8c; joint_tests: F = 3.108, df = 3, 380, p = 0.0253; pairs: z = -3.021, p = 0.0134); however, there was no difference in comparison to resource rich or edge habitats (pairs: -1.268 < z >

1.312, p > 0.2962). For reference, despite similarities in the proportion of cells with non-zero occurrence probabilities, the number of raster cells with non-zero occurrence probabilities in resource-rich habitat was approximately 4 times greater than that of crop fields, resource-poor, or edge habitats (Fig 8d).

Discussion

Direct methods to study animal movement, such as visual observations, individual mark- recapture, and tracking technologies, are more precise than indirect methods; however, it is very difficult to directly track individual insects over large distances (Osborne et al. 2002). Monarch butterflies have been successfully located with handheld VHF radio telemetry in a 4-ha restored prairie (Fisher, Adelman, et al. 2020) and 6.25-ha sod field (Fisher and Bradbury 2021) and with an automated VHF radio telemetry system in a 6.25-ha sod field (Fisher, Dixon, et al. 2020). To increase understanding of movement and space use across larger landscapes, tracking studies in areas with more complex spatial patterns of restored breeding habitat within agricultural settings are needed (Fisher, Adelman, et al. 2020).

We aimed to employ VHF radio telemetry to understand if female monarch movement behavior and space use are influenced by different spatial configurations of contiguous habitat classes with varying densities of milkweed and nectar resources. Radio-tagged monarch butterflies were released in maize fields, restored or remnant prairies, grass-dominated fields and roadside rights of ways, and habitat edges in an 800 m2 mosaic configuration of habitat classes 161 and secondary gravel roadsides with and without access to resource-rich areas. We predicted that:1) monarchs released in these configurations would move among habitat classes; 2) monarchs would orient into the wind, consistent with insect search behavior (Bell 1991, Carde and Willis 2008); 3) monarchs would perform directed movement with large steps in resource- poor areas and undirected with small steps when forage and oviposition resources are abundant

(Zalucki and Kitching 1982, 1984, Zalucki et al. 2016, Grant et al. 2018, Fisher, Adelman, et al.

2020, Fisher and Bradbury 2021); and 4), independent of release location, monarchs would locate and preferentially use areas with abundant milkweed and blooming forbs (nectar resources) as compared to areas without these resources.

Movement among Habitat Classes

The monarch butterfly is described as a vagile species, that is assumed to traverse up to

15 km/d during the breeding season (Zalucki et al. 2016). Consequently, we hypothesized that monarch butterflies would move among habitat classes in our research areas. We observed

81.5% of monarchs cross a habitat boundary at least once with no influence of habitat release class or habitat configuration. Flight steps that crossed a habitat boundary averaged 82.92 ±

243.25 m and were significantly larger than steps within a habitat type (17.23 ± 50.43 m).

Interestingly, 9.5% of the steps we observed ranged from 50 to > 1,000 m, were directed, and were associated with monarchs dispersing from our focal research area. Most large dispersal steps were initiated in resource-rich habitats (45.3%) with approximately a quarter of the dispersal steps initiated in crop fields (25.3%) or resource-poor habitats (29.3%). Similar observations of exiting flight steps of up to 250 m and approximately 1500 m from a restored prairie and turf field were also reported by Fisher, Adelman, et al. (2020) and Fisher and

Bradbury (2021), respectively.

162

Orientation of Movement

Because monarchs oriented with the wind when released in a sod field (Fisher and

Bradbury 2021), likely as a response to the lack of milkweed or blooming forbs, we hypothesized that releases in diverse land cover would elicit flights into the wind, assuming flight behavior is influenced by olfactory cues from nectar forage or milkweed for ovipositing.

Generally, the effective movements of monarchs did not display directional orientation relative to the wind or in reference to geographic features at any of the release sites. However, monarchs that were released on the roadside without access to a resource-rich area showed directional movement into the wind, suggesting that when no resources were in close proximity, monarchs exhibited stereotypical up-wind search behavior. At the mosaic configuration directional movement into the wind and toward a resource-rich area also was observed when monarchs were released on the edge of a maize field and a resource-rich habitat. These observations suggest monarchs may have responded to visual and/or olfactory cues in these release settings.

Perception likely is synergistically detected by olfaction and vision at a fine-scale (e.g., within meters of a resource) and olfactory cues alone attract insects from distances 30 m or greater (Rutowski 2003, MacDonald et al. 2019). Specifically, monarch butterflies are estimated to rely on visual detection within 5 m to locate milkweed plants (Garlick 2007). In a sod field with potted milkweed and blooming forbs, the monarch’s olfactory perceptual range was estimated to be approximately 50 m (Fisher and Bradbury 2021). In the current study, flight steps that entered resource-rich areas, irrelevant of wind direction, provided an opportunity to further quantify visual functional perceptual range, while steps associated with flights into the wind that entered a resource-rich habitat provided the opportunity to draw inferences of olfactory perceptual range. Like observations reported by Fisher and Bradbury (2021), monarch butterfly olfactory perceptual range extended up to 125 m, but most frequently was estimated to be ≤ 50 163 m, which aligns with an agent-based model assumptions of a 50 m perceptual range (Grant et al.

2018).

Step Lengths and Directionality

We hypothesized that step lengths and turn angles would be influenced by milkweed and nectar resource availability, with short steps and undirected turn angles of foraging monarchs in areas with abundant resources as reported by Fisher, Adelman et al. (2020) and by Zalucki and

Kitching (1982, 1984). We observed no difference in step lengths based on habitat classification, with the majority of steps ≤ 50 m, which implies model assumptions with 30 m step lengths

(Grant et al. 2018) are reasonable. Similarly, there was no difference in turn angles across habitat classifications. Typically turn angles with habitat classes displayed normal distributions with directed movement occurring most frequently. While in resource rich habitat, monarchs were observed exhibiting foraging behavior. Different patterns of step lengths and turn angles observed in our 800 m2 mosaic configuration as compared to the undirected turn angles observed by Fisher, Adelman et al. (2020) in a 4-acre restored prairie likely reflects differences in habitat spatial scale and density of resources. Qualitatively, the 4-ha restored prairie had a higher density of blooming forbs than the resource-rich habitats studied here.

In relation to site configuration, step lengths and turn angles were larger and more direct at the roadside corridor configurations than at the mosaic configuration. In addition, large exiting steps were initiated earlier in the observation period in the mosaic configuration than in the roadside corridor configurations, suggesting that monarchs were less motivated to exhibit dispersal behavior outside roadside corridors when presented with linear, resource-poor habitat.

We observed monarchs released in the roadside within 5 m of a prairie disperse into the resource-rich habitat. In total, these observations are consistent with a spatially-explicit agent- based model of female monarch butterfly movement across Story county Iowa (Grant et al. 164

2018). Assuming a perceptual range of 50 m and step lengths of 30 m, the model simulates most monarchs moving along roadsides bordered by crop fields, with excursions outside roadsides when bordered by resource rich habitat.

Space Use

Despite the relation to wind direction and proximity of the release location to a resource- rich area, we hypothesized the monarchs would locate resource-rich habitats. When monarchs were released on the habitat edge between resource-rich habitat and a maize field, 78.9% (15/19) ultimately ended their movement in resource-rich habitat with 12 individuals immediately entering the resource rich habitat and three additional monarchs entering the resource rich habitat during their 3,600 s observation periods. Of the 57 monarchs released in areas other than resource-rich habitats, 45 (78.9%) were observed in a resource-rich habitat at some point during their observation period. Of note, a monarch that was released on the roadside ≥ 500 m from the nearest resource-rich habitat ended their observation period in a resource-rich area.

Because we hypothesized that monarchs would find resource-rich habitats, we also hypothesized that monarchs would occur in these areas more that in habitats without milkweed and nectar resources. Of the 21 monarchs that remained in their release habitat for the duration of their observation period, nine were released in a resource-rich habitat with abundant milkweed and nectar resources and five were released in a resource-poor habitat that had less abundant milkweed and nectar resources, suggesting that monarchs utilize resources when they are co- located. Although there is no difference in the proportion of raster cells with a probability of occurrence between resource-rich and resource-poor habitats, monarchs had non-zero occurrence probabilities four times greater in resource-rich raster cells than resource-poor raster cells.

Additionally, monarchs spent more time in resource-rich and resource-poor areas than in crop fields. Results from these analyses are concordant with movement patterns simulated by an 165 agent-based model with 30 m step lengths, 50 m perceptual range, and high directionality (Fig.

5a and 5c in Grant et al. 2018).

In conclusion, to date, this study provides the most robust dataset of directly tracked movement of individual breeding-season monarch butterflies to assess utilization of agricultural landscapes. Monarchs were observed moving among habitat classes and, although not always, were observed performing natural up-wind search behavior. Step lengths and turn angles were mostly below 50 m and directional, regardless of habitat class or spatial configuration. This observation was likely a response to floral and milkweed density. As reported by Fisher,

Adelman, et al. (2020) and Fisher and Bradbury (2021), we observed large exiting steps from study areas of up to ≥ 1,000 m. These dispersal steps occurred most frequently from resource- rich habitat and could indicate initiation of long-range search for a new research-rich habitat, which in turn would increase egg distribution across the landscape. In summary despite proximity to resource rich habitat, 78.9% of monarch not released in a resource-rich habitat were observed in resource rich habitat, with some monarch flying over 500 m to find resource-rich areas.

Results of this study provide empirical landscape-scale estimates of perceptual range, flight directionality, flight step lengths, and habitat utilization that are needed to simulate monarch population dynamics (Zalucki et al. 2016, Grant et al. 2018, Grant and Bradbury 2019) and produce biologically-relevant recommendations for conservation practices. This work advances knowledge at a landscape scale and would not have been possible without the aid of radio telemetry to relocate individuals beyond human visual range.

Data Accessibility

Data, metadata, and R code pertaining to this manuscript are publicly available through

GitHub: https://github.com/kelseyefisher/monarchoccurrence 166

Acknowledgements

This work was supported by the Agriculture and Food Research Initiative Pollinator

Health Program (Grant 2018-67013-27541) from the USDA National Institute of Food and

Agriculture, Garden Club of America Board of Associates Centennial Pollinator Fellowship received in 2017, Holohil Grant Program received in 2018, ISU College of Agriculture and Life

Sciences, the Iowa State University Foundation, and the Iowa Monarch Conservation

Consortium. We have a special appreciation for the property managers and landowners that allowed us to install equipment, conduct experiments, and catch monarch butterflies on their property: Kent Berns at the ISU Dairy Farm, the Mull and Gambaiani families, Jesse Randell, and Brab Licklider, as well as Brian and Cheryl Fuller for providing field housing, bonfires, and s’mores. We thank our tireless field technicians for their enthusiasm, support, and assistance:

Kevin Anderson, Jenna Nixt, Michael Roth, Samantha Shimota, and Brooklyn Snyder.

Additional thanks to Daniel Borchardt with Pheasants Forever and Quail Forever for aid in identifying potential field sites, Dr. James Adelman for his help on experimental design and input prior to undertaking these experiments, Dr. Philip Dixon for his assistance interpreting data collected from the tower system, and Dr. Jarad Niemi for his guidance on statistical analyses.

Finally, thank you to Myron Zalucki, Chip Taylor, Rick Hellmich, and USDA-ARS CICGRU for incredibly valuable insight and guidance when developing and planning this project.

References

Agostinelli, C. 2017. Circular Statistics. R Package Version 0.4-93. https://cran.r- project.org/web/packages/circular/circular.pdf. Accessed 22 December 2020.

Bell, W. J. 1991. Searching Behavior: The behavioural ecology of finding resources, Chapman and Hall Animal Behaviour Series. Springer Netherlands. 167

Calabrese, J. M., C. H. Fleming, and E. Gurarie. 2016. ctmm: an r package for analyzing animal relocation data as a continuous-time stochastic process. Methods in Ecology and Evolution. 7: 1124–1132.

Carde, R. T., and M. A. Willis. 2008. Navigational strategies used by insects to find distant, wind- borne sources of odor. J Chem Ecol. 34: 854–866.

Chan, Y.-C., T. L. Tibbitts, T. Lok, C. J. Hassell, H.-B. Peng, Z. Ma, Z. Zhang, and T. Piersma. 2019. Filling knowledge gaps in a threatened shorebird flyway through satellite tracking. Journal of Applied Ecology. 56: 2305–2315.

Fernández, P., A. Rodríguez, R. Obregón, S. de Haro, D. Jordano, and J. Fernández-Haeger. 2016. Fine Scale Movements of the Butterfly Plebejus argus in a Heterogeneous Natural Landscape as Revealed by GPS Tracking. J Insect Behav. 29: 80–98.

Fisher, K. E., J. S. Adelman, and S. P. Bradbury. 2020. Employing very high frequency (VHF) radio telemetry to recreate monarch butterfly flight paths. Environ Entomol. 49: 312–323.

Fisher, K. E., P. M. Dixon, G. Han, J. S. Adelman, and S. P. Bradbury. 2020. Locating large insects using automated VHF radio telemetry with a multi-antennae array. Methods in Ecology and Evolution. DOI: 10.1111/2041-210X.13529.

Fleming, C. H., and J. M. Calabrese. 2020. ctmm: Continuous-Time Movement Modeling. R Package Version 0.5.11. https://cran.r-project.org/web/packages/ctmm/index.html. Accessed 1 January 2021.

Garlick, K. M. 2007. Visual and olfactory sensory systems employed by monarch butterflies (Danaus plexippus) to locate their milkweed host plants. (M.Sc. Thesis).

Grant, T. J., and S. P. Bradbury. 2019. The Role of Modeling in Monarch Butterfly Research and Conservation. Front. Ecol. Evol. 7: 197.

Hanski, I. 1998. Metapopulation dynamics. Nature. 396: 41–49.

Ide, J.-Y. 2002. Mating behaviour and light conditions cause seasonal changes in the dispersal pattern of the satyrine butterfly Lethe diana. Ecological Entomology. 27: 33–40.

Kallioniemi, E., A. Zannese, J. E. Tinker, and A. M. A. Franco. 2014. Inter- and intra-specific differences in butterfly behaviour at boundaries. Insect Conservation and Diversity. 7: 232–240.

Kays, R., M. C. Crofoot, W. Jetz, and M. Wikelski. 2015. Terrestrial animal tracking as an eye on life and planet. Science. 348: 11.

Kissling, W. D., D. E. Pattemore, and M. Hagen. 2014. Challenges and prospects in the telemetry of insects. Biological Reviews. 89: 511–530.

Lenth, R., H. Singmann, J. Love, P. Buerkner, and M. Herve. 2020. emmeans: Estimated Marginal Means, aka Least-Squares Means. R Package Version 1.4.4. https://cran.r- project.org/web/packages/emmeans/index.html. Accessed 1 January 2021.

Merckx, T., and H. Van Dyck. 2007. Habitat fragmentation affects habitat-finding ability of the speckled wood butterfly, Pararge aegeria L. Animal Behaviour. 74: 1029–1037.

Moorter, B. V., C. M. Rolandsen, M. Basille, and J.-M. Gaillard. 2016. Movement is the glue connecting home ranges and habitat selection. Journal of Animal Ecology. 85: 21–31.

Nandintsetseg, D., C. Bracis, K. A. Olson, K. Böhning‐Gaese, J. M. Calabrese, B. Chimeddorj, W. F. Fagan, C. H. Fleming, M. Heiner, P. Kaczensky, P. Leimgruber, D. Munkhnast, T. Stratmann, and T. Mueller. 2019. Challenges in the conservation of wide-ranging nomadic species. Journal of Applied Ecology. 56: 1916–1926.

Osborne, J., H. Loxdale, and I. P. Woiwod. 2002. Monitoring insect dispersal: methods and approaches, pp. 24–49. In Bullock, J.M, Kenward, R.E., Hails, R.S. (eds.) Dispersal Ecology. Blackwell Publishing, Oxford.

R Core Team. 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Reynolds, D., N. Armes, R. Cooter, M. Tucker, and J. Colvin. 1997. Techniques for quantifying insect migration, pp. 111–145. In Dent, D., Walton, M. (eds.), Methods in Ecological and Agricultural Entomology. CAB International, Wallingford, UK.

RStudio Team. 2020. RStudio: Integrated Development for R. RStudio, PBC, Boston, MA. 169

Schultz, C. B., and E. E. Crone. 2001. Edge-Mediated Dispersal Behavior in a Prairie Butterfly. Ecology. 82: 1879–1892.

Schultz, C. B., A. M. A. Franco, and E. E. Crone. 2012. Response of butterflies to structural and resource boundaries. Journal of Animal Ecology. 81: 724–734.

Schultz, C. B., B. G. Pe’er, C. Damiani, L. Brown, and E. E. Crone. 2017. Does movement behaviour predict population densities? A test with 25 butterfly species. Journal of Animal Ecology. 86: 384–393.

Showers, W. B. 1997. Migratory Ecology of the Black Cutworm. Annual Review of Entomology. 42: 393–425.

Southwood, T. R. E., and P. Henderson. 2000. Ecological Methods, 3rd ed. Blackwell Science, Oxford.

Stevens, V. M., A. Trochet, H. V. Dyck, J. Clobert, and M. Baguette. 2012. How is dispersal integrated in life histories: a quantitative analysis using butterflies. Ecology Letters. 15: 74–86.

Tobias, J., and A. Pigot. 2019. Linking behaviour to dynamics of populations and communities: application of novel approaches in behavioural ecology to conservation. Phil. Trans. R. Soc. B. 374: 20190008.

Urquhart, F. 1987. The Monarch Butterfly: International Traveller. Nelson-Hall, Chicago, IL.

USFWS. 2020. Monarch (Danaus plexippus) Species Status Assessment Report, version 2.1 September 2020. (https://www.fws.gov/savethemonarch/pdfs/Monarch-SSA-report.pdf).

Wang, Z., Y. Huang, and N. E. Pierce. 2019. Radio telemetry helps record the dispersal patterns of birdwing butterflies in mountainous habitats: Golden Birdwing (Troides aeacus) as an example. J Insect Conserv. 23: 729–738. 170

Warton, D. I., and F. K. C. Hui. 2011. The arcsine is asinine: the analysis of proportions in ecology. Ecology. 92: 3–10.

Wittemyer, G., J. M. Northrup, and G. Bastille-Rousseau. 2019. Behavioural valuation of landscapes using movement data. Phil. Trans. R. Soc. B. 374: 20180046.

Zalucki, M. P., and R. L. Kitching. 1982. The Analysis and Description of Movement in Adult Danaus plexippus L. (Lepidoptera: Danainae). Behaviour. 80: 174–198.

Zalucki, M. P., and R. L. Kitching. 1984. The dynamics of adult Danaus plexippus L. (Danaidae) within patches of its food plant, Asclepias spp. J. Lepid. Soc. 38: 209–2219.

Zalucki, M. P., H. R. Parry, and J. M. Zalucki. 2016. Movement and egg laying in Monarchs: To move or not to move, that is the equation. Austral Ecology. 41: 154–167.

Figures and Tables

Figure 1. The spatial configuration of landcover types at the mosaic (a, yellow box) and the roadside corridor habitats (b) with (green box) and without (blue box) access to resource-rich habitat. Inset maps display focal locations and the surrounding 2.6 km2 landscape. Example release sites are depicted as yellow stars. Automated radio telemetry system towers are represented as purple circles in the mosaic field site. The automated radio telemetry system was not employed at the roadside corridor sites. 171

Figure 2. The mean number of times female monarchs (n=88) left release habitat and crossed a habitat boundary during the observation period of 75 to 7,500 s. Crop, resource-rich, resource- poor, and edge (between a maize field and restored prairie) release habitat were in the mosaic configuration site (See Figure 1a and Table 6). Roadsides with and without resource-rich habitat were at the roadside configuration sites (see Figure 1b and Table 6). Monarchs released on the roadside with access to the resource-rich habitat crossed habitat boundaries more than monarchs released in the other habitat classifications. Different letters above bars represent a significant difference (p < 0.05). 172

Figure 3. Histograms of step lengths for 114 female monarchs released in all habitat classes; 0- 100 m (a) and 100-2000 m (b). Bars in both histograms represent 10 m bins; note the different y- axis scales for (a) and (b). 173

Figure 4. The time elapsed prior to initiating a large step (> 50 m) by habitat class and habitat configuration. Less time elapsed prior to initiating a > 50 m step in the mosaic configuration habitat than in either roadside corridor, as indicated by the “a” or “b” over the bars of each habitat configuration (p < 0.05). There was no difference in duration before a large step based on the length of the step (step length data assigned to 50-99, 100-249, 250-499, 500-999, >1000 m categories, see Table 7 and Figure 4) or the habitat in which the step was initiated. 174

Figure 5. All steps lengths for female monarchs within habitats and crossing habitat boundaries (a), step lengths within each habitat classification (b), step lengths that crossed a boundary based on the exiting habitat (c), and step lengths that crossed a boundary based on the entering habitat (d). Steps that ended with the monarch outside of our focal research area are displayed as “exit survey area” in d. Different letters above groups of bars denote a significant difference in the step category (p < 0.05). Note different y-axis scales in each graph. Habitat configuration had no effect on step lengths associated with for exiting or entering a habitat class. Steps were shorter when monarchs were moving within a habitat classification than when they crossed a habitat boundary. There was no effect of habitat class on step lengths within habitats; however, habitat configuration had a marginal impact on step lengths within habitat classifications, with longer step lengths in the roadside without resources available than in the mosaic configuration. Monarchs took larger steps when leaving a resource-rich habitat class than resource-poor habitat class or crop fields. Larger steps were observed when monarchs exited the focal research area than when they entered resource-rich habitat, resource-poor habitat, or crop fields. 175

Figure 6. Histograms of step lengths for 40 female monarchs that entered a resource-rich habitat (n = 47 steps) (a) and step lengths of 14 monarchs that entered resource-rich habitat while flying ±45 into the wind (n = 14 steps) (b). All steps entering resource-rich habitat were 0.36 m to 126.79 m. Step lengths for monarchs that flew into the wind ranged from 0.53 m to 126.79 m; 86% (12/14) of the steps were below 50 m. 176

Figure 7. Histograms of turn angles (bins = 10) created from three sequential stationary locations for female monarchs within each habitat classification (columns) and each habitat configuration (rows). Generally, a normal distribution around a mean of 0 degrees was observed, suggesting directional movement. Within each habitat configuration, there was no difference in the distribution of turn angles in each habitat type. Likewise, there was no difference in the distributions of turn angles within resource-rich habitats or crop-fields across habitat configurations. Distributions of turn angles within resource-poor habitats differed by spatial configuration, with a narrow distribution in the roadside corridor configurations than the mosaic configuration. 177

Figure 8. Summaries of raster cells with a non-zero probability of occurrence by habitat class. There was no difference (p < 0.05) in the percentages of raster cells with non-zero occurrence probability in the habitat classes based on release habitat, although there was a trend of higher occurrence probability in resource rich habitat classes when monarchs were released in, or in close proximity of resource rich habitat (a). A higher proportion of raster cells had a non-zero probability of occurrence within resource-poor and resource-rich habitats than in crop habitats (b). The proportion of raster cells within 2.5 m of a habitat boundary (edge) was no different in the probability of occurrence than resource-poor, resource-rich, or crop habitats. There was no difference in the proportion of raster cells with an occurrence probability based on release habitat. There was a higher proportion of raster cells with a moderate to a high probability of occurrence (> 0.33 probability) in resource-poor habitats than crop (c). The total number of raster cells with occurrence probability was highest in resource-rich habitats (d). Different letters above bars or individual bars denote a significant difference (p < 0.05).

178

Table 1. Terms and definitions for field sites, release sites, and habitat types. Terms are described further in subsequent tables.

Term Definition Habitat Configuration Mosaic Roadside without Resource-Rich Roadside with Resource-Rich

Release Habitat Crop Field Resource-Poor Resource-Rich Roadside without Resource-Rich Roadside with Resource-Rich

Habitat Classification Crop Field Resource-Poor Resource-Rich

Table 2. Categorization of milkweed ramet and blooming species abundance estimated within 100 m transect surveys to classify resource plant density as zero, low, moderate, or high.

Classification # Milkweed Ramets # Blooms Zero 0 0 Low ≤ 15 ≤ 100 Moderate 16-100 101-1000 High ≥ 101 ≥ 1001

Table 3. Combination of milkweed and bloom classifications (see Table 2) to identify habitat type as “crop field,” “resource-poor,” or “resource-rich.” Crop fields had zero blooming forbs and zero milkweed ramets. Resource-poor habitats were identified when either the bloom or milkweed classification was zero, low, or moderate. When both the bloom and milkweed classifications were moderate or when either the bloom or milkweed classification was high, the habitat was identified as resource-rich.

Bloom Classification Zero Low Moderate High Zero Crop Poor Poor Rich Milkweed Low Poor Poor Poor Rich Classification Moderate Poor Poor Rich Rich High Rich Rich Rich Rich

179

Treatment Flew UNR Total % UNR Radio- tagged 105 33 138 23.9 Untagged 44 13 57 22.8 Overall 149 46 195 23.6

Days Captive Flew UNR Total % UNR 1 56 14 70 20.0 2 62 22 84 26.2 3 27 7 34 20.6 4 4 3 7 42.9 Overall 149 46 195 23.6

Table 6. The number of female monarchs released in the mosaic configuration and the roadside corridor configuration that flew or were unresponsive (UNR; no movement within 15 minutes of release).

Spatial Release Configuration Habitat Flew UNR Total % UNR Resource-Poor 20 2 22 9.1 Resource-Rich 18 7 25 28.0 Mosaic Crop Field 18 4 22 18.2 Edge 19 14 33 42.4 Total 75 27 102 26.5 Resource-Poor 30 8 38 21.1 Roadside Corridor Resource-Rich 44 11 55 20.0 Total 74 19 93 20.4

180

Table 7. The number of large steps (≥ 50 m) by step length categories and habitat of step initiation combined across all habitat configurations.

Number of Large Steps Initiated in Habitat Class Crop Field Resource-Poor Resource-Rich Total 50-99 m 7 14 19 40 100-149 m 6 5 8 19 Step Length 250-499 m 5 3 3 11 500-999 m 1 0 1 2 ≥ 1000 m 0 0 3 3 Total 19 22 34 75

Appendix. Supplemental Information: Automated Radio Telemetry System (ARTS)

Data and R code associated with the ARTS can be found on the digital repository for this project: https://github.com/kelseyefisher/monarchoccurrence

System and Field Design

The tower ARTS designed by Fisher, Dixon, et al. 2020 was employed at the mosaic field configuration in 2019. Instead of using 4 towers each 250 m apart, we designed a 9-tower grid with each tower 250 m apart which created a 500 m2 area of detection (Fig 1). To calibrate the

ARTS in this field setting, data was collected from July 6-8, 2019 to generate power-distance and power-angle relationships. With all towers’ automated receivers functioning, technicians stood for one minute while wearing a transmitter attached to a baseball cap at locations every 25 m on the diagonals from tower 2 to 4, 4 to 8, 8 to 6, and 6 to 2 for a total of 50 locations (Fig 1).

Additionally, power data was collected at an additional 11-16 locations per tower arranged in semi-circles with a 25 m radius from each tower (Fig 1). Data associated with these tests can be found in the digital repository for with this project. 181

Figure 1. Location of towers and data collection for calibration at the mosaic field configuration in 2019. Towers and data collection locations were located in resource-rich habitats, resource- poor habitats and crop fields.

182

Monarchs

In 2019 there were 13 days that radio tagged monarchs were released and observed while the 9-tower ARTS was powered to collect data. Across these days, 42 radio-tagged monarchs were observed for 300 – 5,400 s (5 – 90 min). The beats per minute (bpm) window of the transmitters was 36.8, hence, a power detection could theoretically be detected by one antenna on each tower every two seconds; i.e., 150 – 2,700 power detections for each tower. Actual power detections were much lower than these theoretical maximum ‘hits.’ The largest percent of total successful power detections of the anticipated detections was 22%, followed by 12% and

7%. Power readings either came from two towers over a long period of time or scattered in time across up to 6 towers. Of the 42 butterflies, six (14.3%) had zero power detections from any of the towers. Two towers (tower 1 and 4) never detected power during times when monarch butterfly trials were conducted.

Because of the limited power readings that were detected were either from two or fewer towers or did not overlap in time in this field configuration, we were unable to estimate any locations with this system. Similarly, few location estimates were possible when a transmitter was attached to a butterfly under ideal environmental conditions with limited surrounding vegetation. As reported in Fisher, Dixon et al. (2020), power can be dampened by 22 RSSI when vegetation surrounded the transmitter. Our findings in the mosaic field conditions further demonstrate the power emitted by the 0.22 g transmitter is not sufficiently strong enough to overcome signal dampening due to vegetation and/or orientation of the transmitter antenna.

Looking Forward

Likewise, future transmitters that emit stronger signals, with low mass batteries, could also extend detection range and improve error estimates by increasing detection of power readings by receiving stations. Alternatively, additional receiving stations, antennae on receiving stations, and/or antennae with more elements could improve tracking insects in flight, as suggested by

CHAPTER 7. PREDICTION OF MITOCHONDRIAL GENOME-WIDE VARIATION THROUGH SEQUENCING OF MITOCHONDRION ENRICHED EXTRACTS

Kelsey E. Fisher1, Steven P. Bradbury1,2, and Brad S. Coates3

1Department of Entomology, Iowa State University, Ames, IA, 50011

2Department of Natural Resource Ecology and Management, Iowa State University, Ames, IA 50011

3United States Department of Agriculture, Agriculture Research Station, Corn Insects and Crop Genetics Research Unit, Ames, IA 50011

Modified from a manuscript published in Scientific Reports.

Fisher, KE, BS Coates, & SP Bradbury. 2020. Prediction of mitochondrial genome-wide variation using mitochondrion enrichment and next-generation sequencing methods. Scientific Reports. doi:10.1038/s41598-020-76088-0

Abstract

Although mitochondrial DNA (mtDNA) haplotype variation is often applied for estimating population dynamics and phylogenetic relationships, economical and generalized methods for entire mtDNA genome enrichment prior to high-throughput sequencing are not readily available. This study demonstrates the utility of differential centrifugation to enrich for mitochondrion within cell extracts prior to DNA extraction, short-read sequencing, and assembly using exemplars from eight maternal lineages of the insect species, Ostrinia nubilalis. Compared to controls, enriched extracts showed a significant mean increase of 48.2- and 86.1-fold in mtDNA based on quantitative PCR and proportion of subsequent short sequence reads that aligned to the O. nubilalis reference mitochondrial genome, respectively. Compared to the reference genome, our de novo assembled O. nubilalis mitochondrial genomes contained 82 intraspecific substitution and insertion/deletion mutations, and provided evidence for correction 185 of mis-annotated 28 C-terminal residues within the NADH dehydrogenase subunit 4.

Comparison to a more recent O. nubilalis mtDNA assembly from unenriched short-read data analogously showed 77 variant sites. Twenty-eight variant positions, and a triplet ATT codon

(Ile) insertion within ATP synthase subunit 8, were unique within our assemblies. This study provides a generalizable pipeline for whole mitochondrial genome sequence acquisition adaptable to applications across a range of taxa.

Introduction

Insect mitochondrial genomes are relatively small (range: 14 to 40 kbp) and encode 13 protein-coding genes, 22 transfer RNAs, and two ribosomal RNAs [1] in a highly conserved order and orientation [2]. Due to higher rates of sequence evolution [3,4], low recombination rates [5], and lower effective population size of the materially-inherited molecule compared to biparental nuclear loci, mitochondrial DNA (mtDNA) is often used to predict evolutionary relationships and estimate species divergence times [6–11], albeit for shallow time scales due to accumulating effects of homoplasy [12]. Additionally, intraspecies variation can shed light on population genetics, demographics, and female dispersal [13]. Although the majority of studies are based on haplotype variation estimated from a relatively small number of mitochondrial genes (typically one or a few), whole mtDNA genome analyses offer advantages for phylogenetic, population genetic, and functional genomic studies [14–16].

Ostrinia nubilalis is a lepidopteran pest that feeds on and causes yield loss to cultivated maize and other crops across its native range of Europe and western Asia, as well as within an introduced and invasive areas of North America [17]. Feeding damage is reduced in the United

States following the widespread adoption of transgenic maize encoding insecticidal Bacillus thuringiensis (Bt) toxins [18], yet O. nubilalis remains a model for the study of sympatric population divergence and incipient species formation [19]. Although low-levels of mtDNA 186 cytochrome c oxidase subunit I (coxI) variation were significant between sympatric and allopatric O. nubilalis ecotypes differing in the number of annual reproductive generations [20], mtDNA variation is generally considered to be low and uninformative within this species

[21,22]. This is in stark contrast to the greater inter-individual (haplotype) variation reported in the related species, O. furnacalis [23,24], and other species of Lepidoptera [25,26]. The evolutionary rate of nucleotide change varies across mitochondrial genome regions for animals

[27], including insects [7], indicating studies based on short mtDNA gene fragments may be skewed by genome sampling ascertainment bias [14].

The number of sequenced whole mitochondrial genomes has greatly increased in recent years [28]. This increase has facilitated mitochondrial genome assemblies directly from high- throughput short-read sequencing of shotgun total genomic libraries [8], or libraries enriched for mtDNA using commercial miniprep- [29], rolling circle- [30], or probe-based methods [31].

Regardless, many of these methods co-enrich nuclear-fragments with integrated mitochondrial

DNA fragments (NUMTs) [32–36], or arguably require expensive commercial reagents. The current study establishes a relatively rapid, less expensive, and portable enrichment method, which, when applied to samples prior to high-throughput DNA sequence library construction and sequencing, allows for downstream assessment of full mtDNA genome variation. Differential centrifugation, adapted from prior methods [37], is used to obtain mitochondrial- and nuclear- enriched fractions from Ostrinia nubilalis thorax homogenates. This enrichment method, which minimized the number of reads derived from NUMTs, may be especially useful for assembly of novel reference mitochondrial genomes. The de novo mitochondrial genome assembly and annotation are described, and we report two major differences compared to the reference assemblies (AF442957.1 and MN793322.1). This study demonstrates the utility of 187 mitochondrion enrichments within a full mtDNA genome sequencing protocol, and how detected variation can be used in population and phylogenetic studies.

Results

Optimization of Differential Centrifugation and DNA Extraction

Following initial centrifugation at 1,000 relative centrifugation force (rcf) to remove cell debris, second centrifugation at 6,000 rcf optimally separated intact nuclei from mitochondria within thorax cellular homogenates. This combination resulted in the greatest sedimentation of nuclei (chromosomal DNA) and retention of mitochondria (mtDNA) in the supernatant.

Subsequent centrifugation of the supernatant at 13,000 rcf provided a precipitate rich in mitochondria, as estimated by Janus Green B stained organelles (Fig. 1a). Enrichment was also demonstrated by the ratio of amplified fragment intensities by semi-quantitative PCR of mitochondrial (coxI) compared to the apn1 nuclear target gene (Fig. 1b, S1). By comparison,

DNA extracts centrifuged at 2,000 and 4,000 rcf resulted in either no or inconsistent nuclear amplification, respectively, indicating ineffective separation of nuclei and mitochondria (data not shown). Mitochondrial coxI amplification was also inconsistent or non-existent within DNA extracts obtained from fractions pelleted at 2,000 and 4,000 rcf. Nuclear and mitochondrial target genes were both PCR amplified from fractions collected at 8,000 rcf, suggesting co-precipitation of both organelles (data not shown).

Quantification of Mitochondrion Enrichment at Optimized Parameters

Using mitochondrial coxI primers to amplify mitochondrial fractions as template with real-time quantitative PCR (qPCR) resulted in an estimated mean CT (14.83±0.09) that was significantly lower compared to that estimated from nuclear fractions (17.43±0.17) or unenriched controls (17.54±0.41) (Fig. 2a; -5.069 < Z > 3.205; df = 2, 34; p < 0.0039; Table S1). By comparison, when mitochondrial fractions, nuclear fractions, and unenriched control samples 188

were amplified with nuclear apn1 primers, the mean CT values were 30.25±0.21, 24.56±0.29, and 26.58±0.21, respectively, and were significantly different from each other (-6.905 < Z >

3.763; df = 2, 34; p < 0.0005; Table S1).

A mean of 3,047,025 (±162,751) reads was obtained across eight individual mitochondrion-enriched Illumina MiSeq libraries (BioProject: PRJNA604593; accessions

SRR5182721-SRR5182724; Table S1). Among these reads, a mean of 781,200 (±77,773) aligned to the O. nubilalis reference mitochondrial genome sequence (AF442957.1) with a Q- score > 30 (mean ~26.7±2.0 % alignment rate). The unenriched controls had a mean of 183,349

(±34,965) mapped reads with Q-score > 30 from among a mean of 35,319,000 (±4,800,041) total reads (~0.6±0.1 % mean alignment rate). The increase in the proportions of reads that aligned to the mitochondrial reference from libraries generated from mitochondrion enriched (~26.7%) compared to unenriched libraries (~0.6%) was significant (Fig. 2b; t = 13.065; df = 1, 11; p <

0.0001; Table S1).

The resulting ratio of CT values estimated from real-time qPCR amplification of mitochondrial-enriched extracts compared to the mean CT of unenriched extracts showed an estimated mean 86.1±8.25-fold enrichment (Fig. 2c). Based on the ratio of aligned Illumina

MiSeq reads from enriched to unenriched libraries, a mean enrichment of 48.2±3.6-fold was estimated. Although there was a significant difference in the estimated fold-enrichments based on real-time qPCR compared to Illumina MiSeq read alignment (Fig. 2c, t = -4.0958; df = 1, 7; p

= 0.0046), a ≥ 29.8-fold mitochondrial DNA enrichment was predicted among all samples with both estimation methods (Table S1).

De novo Mitochondrial Genome Assembly, Annotation and Variant prediction

Mitochondrion-enriched short read library data were de novo assembled, annotated from query results of O. nubilalis mitochondrial genome RefSeq models using the BLASTn algorithm, 189 and submitted to NCBI GenBank (MT492030.1-MT492037.1; Table 1). Each assembly was annotated with 13 protein-coding genes (PCGs), 22 tRNAs, and two rRNAs typical of animal mtDNA genomes. Based on the invertebrate mitochondrial code, all PCG translations were initiated with Met start codon (ATA or ATG), with the exception of coxI that had an atypical Arg

(CGA). Stop codons, TAA or TAG, were predicted for all PCGs, with the exception of coxII that ends in a single T nucleotide that putatively forms a functional TAA stop codon following polyadenylation during transcript maturation.

Our assemblies were compared to the Sanger sequence-based reference O. nubilalis mitochondrial genome assembly, AF442957.1 [38] (RefSeq NC_003367.1), and a recent

Illumina short shotgun read-based assembly [39]. For the former comparison, a total of 82 mutations were predicted among our de novo assemblies (MT492030.1 – MT492037.1; Table 1) and the reference AF442957.1 (RefSeq NC_003367.1; Table S2). Because each of our de novo mitochondrial genomes had a different start position due to starts at random seeds, homologous nucleotide sites with respect to the reference were defined within each accession (Table S3).

Among these predicted variants, 58 (69.8%) were within protein-coding sequences (CDS; coxI, coxII, atp8, atp6, coxIII, nd3, nd5, nd4, nd4L, cytb, or nd1; Table S2), for which 35 generated nonsynonymous (amino acid) changes (60.3% of CDS mutations; 42.2% of all mutations).

Twenty-two of the remaining CDS mutations were synonymous (silent; 26.8%). In addition, an in-frame insertion of an intact triplet ATT (Ile) codon was predicted within atp8 from

F1_Family_ID_04 (accession MT492032.1; Fig. 3a). This inserted ATT codon was not present in any other previously sequenced Ostrinia mitochondrial genome but was found in the atp8 CDS of the butterfly species. Ochlodes venata (Lepidoptera: Hesperiidae). Moreover, the two adjacent

Ile residues within this region of ATP8 showed variable presence among species of Lepidoptera, 190 with both omitted from Parnassius epaphus (Lepidoptera: Papilionidae) (Fig. 3a). Our analyses also predicted a frameshift impacting three consecutive codons in coxI among 4 of our assemblies (MT492030.1, MT492035.1 to MT492037.1) compared to the reference, and shared with other Ostrinia including the recent Illumina sequence-base O. nubilalis assembly

(MN793322.1; [39]) (Fig. 3b). This frameshift is predicted to result from upstream insertion and downstream deletion mutations that function in tandem to maintain the frame within the downstream CDS beyond the three affected codons (Fig. 3b). The remaining 25 predicted mutations compared to the reference were within rRNALSU (8 SNVs and three indels), rRNASSU

(6 indels), and among tRNAs (5 SNVs and two indels; Table S2).

The frequency of each variant site among the eight F1 family-specific assemblies compared to the reference ranged from 0.13 to 1.00, of which 20 of the 82 variable positions

(22.4%; coxI frameshift not counted as variable position) were conserved across all eight F1 families (Freq = 1.00 in Table S2). All of the 20 fixed differences with respect to the reference were SNVs, with the exception of a deletion in nd4. For this nd4 deletion, all assemblies from this study lacked a C nucleotide reported for position 8200 in reference AF442957.1, which caused a discrepancy in predicted C-terminal amino acids starting at position 418 of nd4 (Fig. 4).

Subsequent multiple nd4 protein sequence alignment and phylogenetic comparison showed ≥

83.33% amino acid identity between the terminal 60 residues of nd4 from our O. nubilalis assemblies and other lepidopteran species (Fig. 4), whereas nd4 from AF442957.1 showed complete mismatch in the terminal 29 residues. Inspection of bam alignments of short reads from each F1 family-specific library confirmed a lack of this C compared to the O. nubilalis reference

(Fig. S2). Seven substitutions predicted in coxI and coxII genes among our assemblies were previously observed in a prior population genetics study, of which mutations at reference 191 positions 2,554 and 3,050 were validated using HaeIII and Sau3AI PCR-restriction fragment length polymorphism (PCR-RFLP) assays within that study (Table S2) [20].

Analogously, there were a total of 77 variant sites, plus an Ile (ATT) triplet codon insertion within atp8 and a frameshift mutation in coxI identified within the multiple sequence alignment among our eight assemblies and a previous unenriched Illumina HiSeq read data- assembled mitochondrial genome sequence [39] (MN792233.1; Fig. S2). Among these 77 variant sites, 43 (55.1%) were within CDS, of which 24 (55.8%) were nonsynonymous and predicted to cause an amino acid change (Table S4; Fig S3). Eight mutations were fixed differently between MT492031.1 to MT492034.1 compared to MN792233.1. Identical to comparison with AF442957.1, the Ile codon insertion within our F1 Family_ID_04

(MT492032.1) was not present with the atp8 CDS of MN792233.1 (Fig. 3a; Fig S3).

Additionally, compared to reference AF442957.1 (RefSeq NC_003367.1), the cox1 frameshift identified within four of our eight assemblies was also present in MT492032.1 (Fig. 3b; Fig S3).

The remaining 35 mutations were within the rRNALSU (n = 14), rRNASSU (n = 6), tRNAs (n =

14), and non-coding sequence (n = 1). Three of the substitutions within coxI and coxII genes were previously described by Coates et al. [20], including two validated by HaeIII and Sau3AI

PCR-RFLP assays (Table S4).

Comparison between enriched and unenriched reads for de novo mitochondrial genome assembly

Our use of reads from mitochondrion-enriched and unenriched libraries as input for

SPAdes and SOAPdenovo2, respectively, resulted in assemblies that showed different levels of contiguity and computational efficiency. Specifically, library read data from enriched samples were assembled in 52±8 seconds compared to 1380±180 seconds among the unenriched libraries.

Additionally, the total number of contigs within resulting assembles were lower among enriched 192

(98±41) compared to unenriched libraries (337,151±154,655), wherein mean contig lengths were also greater for enriched (1,110±469 bp) compared to unenriched libraries (117±992 bp)

(remaining data not shown). Among the assembled contigs a mean of 31% and 0.0006% from enriched and unenriched libraries, respectively, showed homology to MtDNA based on BLASTn query with the mitochondrial reference genome sequence (Table S6). Moreover, among these

MtDNA contigs, normalized read depth (RPKM) the estimated mean was ~65 times higher and significantly greater (t = 14.435; df = 1, 667; p < 2.2x10 -16) among enriched samples compared to unenriched samples (Fig. S4). Furthermore, there was a distribution of contigs with both high

RPKM estimate and length across all assemblies generated from enriched libraries that were of predicted to have mtDNA origin (Fig S5), and deemed to be of putative mitochondrial genome origin. In addition, the observed distribution of contigs of MtDNA origin with correspondingly low RPKM estimates and comparative bias toward short contig length were categorized at putative NUMTs. In contrast, contigs with high RPKM estimates in assemblies from each of the unenriched libraries were mostly of non-MtDNA origin (Fig. S5).

Discussion

Variation in mtDNA sequence data is frequently used within interspecies phylogenetic and intraspecies population genetic analyses. However, a vast majority of studies assess variation based on a few genes or gene fragments, which potentially impact estimates of genetic variation and bias results, leading to the possibility of arriving at specious conclusions. This premise resides in evidence that the mutation rates vary among regions of the mitochondrial genome within plant species [40] and humans [41], as well as between species [42]. Methods have recently been developed to more efficiently obtain sequence data from entire mtDNA genomes, capitalizing on advancements in DNA sequencing technologies. These include generation of low-pass short shotgun sequencing reads from total genomic DNA, from which success of 193 mtDNA genome assemblies rely on the proportionally high copy number of mtDNA compared to nuclear DNA [8,10], as recently reported for several species of Ostrinia [39]. Targeted mtDNA sequencing employs various enrichments prior to library construction, including those based on commercial reagents [43], isolation of highly intact circular mtDNA [30], the requirement to develop specific antibodies for organelle pull-down [44], or necessitating prior knowledge of the mtDNA genome sequence for probe-based nucleotide capture [31,32].

Although differential centrifugation was analogously used in methods to isolate human mitochondrion prior to mtDNA genome sequencing, the commercial kit protocols used are tailored to specific species [45] and not directly transferable to other organisms. Here, we developed and validated a relatively rapid and low-cost differential centrifugation method to prepare mitochondrion-rich fractions from Ostrinia nubilalis thorax homogenates prior to DNA extraction and subsequent high-throughput DNA sequence library construction and sequencing.

Although not tested here, this enrichment method is likely transferrable to other non-model organisms where no prior genome assemblies are available, due to our evidence that enrichments lead to significant reductions in raw data input and computational time requirements, as well as increase contiguity of resulting MtDNA assemblies. Furthermore, the additive effect of these efficiencies likely may enable individual-wise assessment of mitochondrial genome-wide haplotype variation on a population scale.

Our protocol development empirically determined that a second centrifugation step of

6,000 rcf maximized the separation of mitochondria and nuclei in thorax homogenates based on qualitative Janus-Green visualization and semi-quantitative PCR comparisons between enriched and unenriched extracts (Fig. 1). Subsequent RT-qPCR and read alignment analyses indicated that while our differential centrifugation did no provide pure organelle f ractions (Fig. 2a), it did 194 provide for subsequent extracts enriched with mtDNA (Fig. 2b). Although these two methods lead to significantly different estimates of enrichment, a mean fold-enrichment of ≥ 48.2 was estimated among the libraries (Fig. 2c). These results suggest that secondary enrichment steps, such as the use of equilibrium density-gradient centrifugation [46], could likely be incorporated into our protocol in cases where higher purity isolations are desired. While the level of enrichment and purity of our mitochondrion fractions were lower compared to methods employing mitochondrion surface protein-specific antibody capture [44,47] or probe-assisted mtDNA isolation [31,32], our method has the advantage of not requiring upfront development of custom biological reagents or reliance upon tailored commercial kits. Our differential centrifugation achieved 26% of reads aligning to a reference mitochondrial genome, which is comparable to a 22% alignment rate obtained following a two-step protocol that used a commercial circular DNA miniprep followed by antibody tagged paramagnetic bead isolation

[43]. Regardless, due to variation in the mass of organelles, it may be anticipated that rcf values will need to be optimized for application to different species. Since mitochondrion enrichment prior to sequencing produced at least a 29.8-fold greater read alignment rate to the mitochondrial reference compared to unenriched controls (Table S1), showing that our optimized protocol facilitates a significant reduction in short sequence read input required for downstream assemblies.

Comparison of processes involved in the assembly of enriched compared to unenriched libraries, showed is a significant (~27-fold) reduction in computation time and lower required

CPU node count and random access memory. Moreover, although it was possible to de novo assemble the mitochondrial genome from unenriched samples, the resulting assemblies were less contiguous compared to those from enriched libraries (Table S6). Specifically, the mean size of 195 mitochondrial-derived contigs among unenriched assemblies were nearly 10-fold shorter and nearly more numerous compared to contigs assembled from enriched libraries (Fig. S4, S5).

Interestingly, these increases were achieved despite a 11.7- to 65.7-fold reduction in input reads among enriched libraries (Table S1) compared to that within unenriched SRA data files

(SRX2498822-SRX2498825). Furthermore, due to lower sequencing costs and computational time, as well as increased contiguity, we suggest that enrichment provides greater accessibility to the potential for higher throughput assessment of haplotype variation among haplotypes of a species or among species. This may provide for downstream large-scale population and phylogenetic analyses that arguably remain out of reach using other methods.

In this study, we applied the full mtDNA genome assemblies derived from our enrichment method to assess variation across O. nubilalis maternal lineages. The genus Ostrinia is a model for the study of sympatric population divergence and incipient species formation

[19,48] and comprises two corn borer species, O. nubilalis, and O. furnacalis, which are major pest insects that feed on cultivated maize in North America, Europe and/or Asia. Compared to prior O. nubilalis mtDNA genome assemblies AF442957.1 [20] and MN792233.1 [39], our eight

F1 families showed 82 and 77 SNVs, respectively (Table S2, S4). These comparisons also predicted a single insertion and frameshift mutation. Among the mutations we predicted across all eight F1 family-based assemblies, 20 and 8 were fixed differently compared to the reference

AF442957.1 [20] and MN792233.1 [39], respectively. Fixed differences could represent natural intraspecies haplotype differences, but given the small sample size of the eight F1 families, some of the fixed differences could be a consequence of sampling bias from a laboratory colony that has undergone a genetic bottleneck and influenced by random genetic drift for >12 generations.

Alternatively, some of these SNVs could have resulted from Sanger sequencing errors 196 incorporated into the reference assembly AF442957.1 [20]. The latter might be concluded, for a

C nucleotide inserted within nd4 at position 8200 from the Sanger reference as compared to all

Illumina-based assemblies (Fig. S2; MN792032.1), which caused a fixed discrepancy in the C- terminal 29 residues of nd4 compared to assemblies from other Ostrinia and lepidopteran species generated from short-read data (Fig. 4). This putative correction suggests that the high read depths obtained from short read sequencing of enriched mtDNA libraries provides superior error correction capabilities compared to dual-pass Sanger read data.

The structural gene annotation of our eight assembles showed two major differences compared to the reference. Firstly, our results also predicted a novel ATT insertion within F1 family 4 (MT492032.1) residing at reference position 3,953 within the atp8, causing a putative in-frame Ile codon duplication at atp8 amino acid position 32 (Fig. 3a). This insertion is novel compared to all other exemplar sequences within the genus Ostrinia and other species of

Lepidoptera, with the exception of a Leu codon (TTA) found at the orthologous position of the

Ochlodes venata mitochondrial genome (Fig. 3b). Moreover, the two adjacent Ile residues are deleted at orthologous sites from the butterfly Parnassius epaphus. These comparisons suggest that sites orthologous to Ostrinia atp8 amino acid positions 52 and 53 may show relaxed functional constraint, and thus indels involving these amino acids might putatively be impacted by purifying selection to a lesser degree. Although intriguing, testing this hypothesis is beyond the scope of the current study. Secondly, out of eight assemblies, four show a pair of indels that cause a frameshift over three amino acids of coxI (Fig. 3b). Specifically, compensation for the upstream deletion by a 10bp downstream insertion retains the predicted downstream coding frame. Typically, mitochondrial frameshifts are highly deleterious [49], but can result in apparently non-deleterious changes depending upon position and context [50,51]. In other 197 instances, suppressor tRNA mutations or “leaky” ribosomal frameshift compensation mechanisms have been proposed [52,53]. Regardless, the frameshift described here, resulting in the putative alteration of 3 amino acids of coxI, has an unknown functional consequence and was not tested further. Due to presence within reads at this position, the indel is likely valid, but may be unique among individuals within a relatively inbred laboratory colony and not representative of wild individuals under the influence of selective forces in field conditions.

The variation we detected cannot be interpreted as representing population frequencies, but would require large-scale screening and comparison of random samples from wild populations. Regardless, prior population screening of 1,414 O. nubilalis individuals from North

American populations determined that 89.6% were comprised of a single haplotype based on

PCR-RFLP, wherein the sequencing of a 2,156 bp fragment comprised of coxI and coxII from 14 individuals predicted 26 polymorphic positions [20] (≤ 0.8% variation). Interestingly, of the 82 variant sites predicted in the current study, seven were identical to the set of 26 previously predicted by Coates et al. [20]. Furthermore, substitutions validated by HaeIII and Sau3AI PCR-

RFLP were predicted within comparisons of our assemblies to the Sanger sequence assembled reference [38] and the more recent short read-based assembly [39], suggesting variants are real as opposed to sequencing or assembly artifacts.

Prior phylogenetic reconstructions differ between Ostrinia species when using variation in sequence data from coxII [54] coxI [55] gene fragment data, as compared to full mitochondrial genome alignments [39]. Specifically, full genome analysis by Zhou et al. [39] predicted that corn borer sister species, O. nubilalis, and O. furnacalis, were within a clade along with O. scapulalis. In comparison, an O. nubilalis and O. scapulalis clade were separated from

O. furnacalis by O. zealis based on coxII gene [54] or coxI sequence data [55]. Although not 198 investigated further for Ostrinia, a disparity in the evolutionary rate between mitochondrial genome regions (genes or gene fragments) is described across taxa [7,27], suggesting that phylogenetic relationships could be biased when applying a subset of mitochondrial haplotype variation within short fragments. Therefore, phylogenetic relationships predicted based on the full mtDNA genome sequence might arguably be more reflective of evolutionary divergence patters, at least over appropriate timespans [14].

Prior to our analysis, Ostrinia nubilalis mtDNA variation was generally estimated to be low and uninformative in a population context [21,22]. By implementing a mitochondrial enrichment method and predicting variation across the full mitochondrial genome sequences, we identified a greater number of variant sites within O. nubilalis than previously reported. The future application of analogous full mtDNA sequencing to population or phylogenetic studies will contribute to unbiased estimates of inter- and intra-species variation. This study demonstrated that a differential centrifugation method could be a component of this goal due to demonstrated enrichment of cell homogenates for mitochondrion and correspondingly removing nuclei prior to DNA extraction. This method provides a low-cost option for organelle enrichment, resulting in samples with a high proportion of mtDNA for library construction and sequencing. Undoubtedly further cost efficiencies could be implemented through indexing of a greater number of sample libraries within higher capacity flow cells compared to the MiSeq that was used here. Furthermore, we demonstrate increased downstream computational efficiencies and resulting assembly contiguities achieved during assembly of reads from mitochondrion- enriched libraries compared to those obtained from unenriched libraries. Although we applied these methods to O. nubilalis, the protocol could be adapted and optimized for other species, thereby facilitating a higher throughput of mtDNA genome sequencing for application in 199 unbiased population and phylogenetic studies, where additive effects of time and cost saving may facilitate performance on larger scales.

Methods

Specimens and sampling

Z-pheromone race Ostrinia nubilalis pupae were obtained from the United States

Department of Agriculture, Agriculture Research Service, Corn Insects and Crop Genetics Unit

(USDA-ARS, CICGRU) in Ames, IA. Pupae were individually maintained in a Percival incubator (Percival, Boone, IA) (16:8 [L:D] h; 26° C; 40 to 60% RH). Upon eclosion, single mate pairs were initiated to create F1 families (maternal-specific haplotype lineages). Pairs were maintained in wire mesh cages with wax paper for oviposition substrate, as previously described

[56]. Mated pairs were monitored, and eggs were collected daily for seven days. Wax papers with egg masses were stored in the incubator (16:8 [L:D] h; 26° C; 40-60% RH) until near hatch.

Egg masses from each F1 family were separately placed into individual 10-cm diameter plastic containers with approximately 50-ml of standard O. nubilalis meridic diet [57]. Pupae were collected from each family and placed into individual paper cups until eclosion. Adults were either live-dissected or frozen at -20 °C.

Optimization of Differential Centrifugation and DNA Extraction

Thorax tissue containing flight muscle was dissected from three live O. nubilalis adults per F1 family. Since mitochondria are maternally inherited without recombination, tissues were pooled by family in 2.0-mL glass Dounce homogenizers (Corning Inc., Corning, NY) and homogenized on ice in 1.0-mL of a homogenization medium (0.32 M sucrose, 1.0 mM EDTA,

10.0 mM Tris-HCl, 4 °C, pH 7.8). Each F1 family homogenate was transferred to a separate 1.5 mL microcentrifuge tube. 200

Differential centrifugation was used to fractionate extracts into nuclear and non-nuclear fractions based on molecular mass. The insoluble exoskeleton, cell debris, and other particles were removed from samples at 1,000 relative centrifugal force (rcf) for 10 minutes at 4 °C in an

Eppendorf 5417R centrifuge (Eppendorf, Hauppauge, NY). Aqueous supernatant was transferred to new 1.5-mL microcentrifuge tubes. Replicate samples within each F1 family were centrifuged at 2,000, 4,000, 6,000, and 8,000 rcf at 4 °C for 10 minutes to optimize the force that most effectively pelleted nuclei (referred to as nuclear fraction) while retaining mitochondria in the supernatant. The remaining supernatants of each replicate (F1 family pool) were transferred to new 1.5-mL microcentrifuge tubes and centrifuged at 4 °C for 10 minutes at 13,000 rcf to pellet the remaining organelles (referred to as mitochondrial fraction).

The optimal centrifugal force was determined using microscopy and the ratio of amplified DNA fragment intensities as determined by semi-quantitative/qualitative PCR of mitochondrial (coxI), and nuclear (apn1) indicator genes. Subsamples of the 13,000 rcf pellet were treated with Janus Green B cell normalization stain (Abcam Inc., Cambridge, MA); stained mitochondria were observed by light microscopy (Olympus Microscopy, model BH-2, Tokyo,

58 Japan) [ ]. DNA was then extracted from each of the fractionated replicates within F1 family pools using the DNEasy Blood and Tissue Extraction Kit according to manufacturer directions

(Qiagen, Hilden, Germany), except DNA was eluted from silica columns using 50.0 µL of

Elution Buffer. Samples were serially diluted using nuclease-free water, and 1.0 µL from each dilution and was PCR amplified in duplicate using mitochondrial primers (HCO/LCO) [59] and nuclear primers (OnFLWA-01) in separate reactions as described earlier [60]. Entire products were separated by 2% agarose gel electrophoresis with the Lambda HindIII and EcoRI ladder.

Based on staining and PCR results, the following method was employed: 1) initial pelleting of 201 cell debris at 1,000 rcf; 2) resultant supernatant centrifuged at 6,000 rcf to obtain nuclear fraction; 3) resultant supernatant centrifuged at 13,000 rcf to obtain the mitochondrial fraction.

All centrifugation steps were conducted for 10 minutes at 4 °C.

The empirically-determined optimal rcf was employed to pellet nuclei with three frozen/preserved adult individuals of each F1 family. Fractions were subsequently extracted with

DNEasy extraction kits, as described above. Unenriched controls were prepared by identical

DNA extraction from dissected thorax tissue without nuclear or mitochondrion fractionation from individual Z-strain O. nubilalis field samples collected at South Shore, SD (SDSS ♀1 and

SDSS ♀2; contributed by Dr. Micheal Catangui, South Dakota State University).

Quantification of Mitochondrion Enrichment at Optimized Parameters

Using the optimized centrifugation forces (6,000 and 13,000 rcf), the fold-enrichment of mitochondrion compared to nuclei was estimated by two different methods: 1) analysis of cycle threshold (CT) obtained through real-time quantitative PCR (qPCR) with nuclear and mitochondrial primers, and 2) high-throughput read alignment rates to a mitochondrial reference genome of enriched and unenriched samples.

For real-time qPCR, DNA was quantified from the nuclear and mitochondrial fractions of the eight F1 families and the two unenriched controls on a NanoDrop 2000 spectrophotometer

(Thermo Scientific, Wilmington, DE, USA). Concentrations were adjusted to ~1.0-ng/μl with nuclease-free water. All samples were PCR amplified in duplicate on an MX3000P real-time qPCR system (Stratagene, La Jolla, CA) with 2.0-ng of the template along with 10.0-nM of each forward and reverse primer in iQsupermix reactions per to manufacturer instructions (BioRad,

Hercules, CA). Mitochondrial DNA-specific primers OnCOXIrtS, 5’-CCT GAT ATA GCA TTC

CCA CGA ATA-3’, and OnCOXIrtA, 5’-AAC CAG TTC CTG CTC CAT TT-3’, were designed using the RealTime qPCR Assay tool [61]. Single locus nuclear primers, APN1RT-1F and 202

APN1RT-1R, amplifying the aminopeptidase N1 gene (apn1), were designed and applied as previously described [56]. Statistical significance of differences in CT values from mitochondrial and nuclear primer amplification reactions of the mitochondrial and nuclear fractions, and unenriched controls, were analyzed with a general linearized model, one-way ANOVA in

RStudio version 1.0.153 [62,63] with the Estimated Marginal Means (emmeans) package [64].

Summaries of paired comparisons were made with Tukey’s HSD method using the

Visualizations of Paired Comparisons (multcompView) package [65].

DNA extracts from each mitochondrial-enriched fraction for each F1 family were submitted to the Iowa State University (ISU) DNA Facility (≤ 100-ng each). Individual indexed short insert libraries were generated using AmpliSeq Library Plus methods according to manufacturer instructions (Illumina, San Diego, CA). All indexed libraries were sequenced in a single lane of a MiSeq system (Illumina, San Diego, CA). Data were received in fastq format and submitted to the National Center for Biotechnology Information (NCBI) GenBank short read archive (SRA). Additionally, four previously generated Illumina HiSeq reads from the whole genomic sequence of pooled O. nubilalis population samples were downloaded from the NCBI

SRA database (Accessions SRX2498822-SRX2498825 [66]).

Reads from each of our mitochondrion-enriched libraries (n = 8) and whole genomic libraries (n = 4) were aligned to the 14,535 nucleotides O. nubilalis mitochondrial genome reference sequence (NCBI GenBank accession AF442957.1; RefSeq NC_003367.1) assembled previously from Sanger sequencing data from overlapping PCR amplicons [38]. Nucleotide quality was initially assessed and visualized with fastqc [67] (fastqc/0.11.7-d5mgqc7). Reads were trimmed for those with a Phred quality score (q) ≥ 20 over a 4-nucleotide sliding window: the first 15 nucleotides (HEADCROP:15) and the last ten nucleotides (TAILCROP:10) were 203 removed using trimmomatic [68] (trimmomatic/0.36-lkktrba). Paired reads were aligned to the reference sequence using bowtie2 [69] (bowtie2/2.3.4.1-py2-jl5zqym) in unpaired mode (-U).

Resulting sequence alignment and map (.sam) files were converted to binary alignment and map

(.bam) format, and inclusive reads filtered for those that mapped to the reference sequence (view

–bhF 2) and had a mapping quality score (Q) ≥ 30 (view –bhq 30) using SAMtools [70]

(SAMtools/1.9-k6deoga). All bioinformatic operations were performed on the ISU Pronto server

(See Table S5 for an example of Linux shell commands). The proportion of reads to meet filtering criteria in the mitochondrial-enriched and unenriched whole genomic libraries were calculated, and the statistical significance of the difference in the proportion of aligned reads was estimated using a two-sample t-test at a threshold α = 0.05 [62,63].

Mitochondrial fold-enrichment was calculated for real-time qPCR and MiSeq read alignment methods. For qPCR, the relative difference in copy number between CT values from enriched and unenriched fractions was used to calculate relative fold enrichments. Delta CT

(ΔCT) values were calculated by subtracting the CT value produced with mitochondrial primers from the CT value produced from the same sample with nuclear primers. For each family, mitochondrial fold-enrichment was calculated with the formula: 2(mitochondrial fraction ΔCT – unenriched control ΔCT). Likewise, nuclear fold-enrichment was calculated with 2(nuclear fraction ΔCT – unenriched control

ΔCT). In the read alignment method, mitochondrial fold-enrichment was calculated as the proportion of reads from each enriched sample that aligned to the reference divided by the mean proportion of reads that analogously aligned from the unenriched samples. Statistical significance in fold enrichment estimations was calculated using paired t-tests at a threshold α =

0.05 [62,63].

204

De novo Mitochondrial Genome Assembly, Annotation, and Variant prediction

Alignments for mitochondrion-enriched reads to the O. nubilalis mitochondrial genome reference (AF442957.1; RefSeq NC_003367.1) were sorted (SAMtools -sort), and the wrapper script plasmidspades.py was used to de novo assemble reads with SPAdes [71] (spades/3.11.1- py2-arfy7sc; default parameters) to increase the coverage of the mitochondrial genome in comparison to the reference genome (See Table S5 for an example of Linux shell commands).

Resulting Kmer contigs of the mitochondrion-enriched de novo assembled reads were used as subjects against the mitochondrial reference genome query [38] (AF442957.1; RefSeq

NC_003367.1) with the BLASTn algorithm [72] (ncbi-rmblastn/2.6.0-2kyyml7; default parameters) because the de novo assembly increased the coverage of the mitochondrial genome in comparison to the reference genome. Kmer contigs from each F1 family that generated

BLASTn hits with E-values ≥ 10-90 were assembled separately with the Sequence Assembly

Program, CAP3 [73] (cap3/2015-02-11-2jwa5sb; default parameters). Gene features in each assembled CAP3 contig (mitochondrial genome) were annotated by queries with protein and

RNA coding sequences from the reference assembly (AF442957.1; RefSeq NC_003367.1) using the BLASTn algorithm. Any discrepancies were corrected based on evidence from corresponding bowtie2-generated bam alignments visualized in Integrative Genome Viewer

(IGV) [74]. Final annotated assemblies were submitted to the NCBI GenBank nr database.

Variable nucleotide positions among the reference and F1 family-specific mitochondrial genomes were identified from corresponding filtered .bam files with bcftools [75] (bcftools/1.9-womp5gh) using the call and varFilter options (See Table S5 for an example of Linux shell commands; minimum read depth ≥ 2,000). Results were output in VCF format v4.2 [76]. Bedtools v 2.29.2

[77] was implemented to retrieve genome sequence features (CDS, rRNA, and tRNAs). 205

Variations between our submitted annotated genomes and the reference genome were manually verified in Microsoft Word.

Additionally, a multiple sequence alignment was generated between our assembled mitochondrial genomes and the 15,248 bp O. nubilalis mitochondrial genome assembled from unenriched Illumina HiSeq read data [39] (MN793322.1). This O. nubilalis sample was collected from the Yili area of the Xinjiang Autonomous region of western China. Nucleotide sequences for our eight assembles and MN793322.1 were loaded into the MEGA8.0 alignment utility [78], and aligned using the ClustalW algorithm [79] (default parameters) with adjustments to codon frame made manually. Wrapping to a width of 100 bp and demarcation of variant sites was performed with the Multiple Alignment Viewer [80]; alignment was manually decorated with corresponding gene intervals. The position of variant sites within each corresponding assembly and the consensus were retrieved from the multiple sequence alignment NCBI MSA viewer [81].

The location and impact of variants on codon use within protein CDS were identified manually and verified by the alignment of corresponding nucleotide and protein sequences retrieved from

GenBank accessions.

A discrepancy in the putative translation of ND4 was predicted between the reference

(AF442957.1; RefSeq NC_003367.1) and all subsequent Illumina short read-based sequence assemblies (MT793322.1 and this study). To investigate this discrepancy further, a multiple protein sequence alignment was generated among ND4 orthologs from Ostrinia species (O. nubilalis, O. furnacalis, O. scapulalis, O. zealis, O. penitalis, and O. palustralis) and a subset of related lepidopteran species using the ClustalW algorithm [79] within the MEGA8.0 alignment utility [78] (default parameters; accessions provided in Fig. 4). The general reversible mitochondrial model of protein sequence evolution [82] (mtREV24) with among site frequency 206 variation (F) and empirically-derived gamma distribution (G) maximized the Bayesian

Information Criterion (BIC) and was chosen as the most appropriate model. The mtREV24 + G

+ F model was subsequently applied within a Maximum-Likelihood (ML) approach to reconstruct an unrooted phylogeny with a consensus tree constructed from 1,000 iterative bootstrap pseudo-replicate sampling steps.

Comparison between enriched and unenriched reads for de novo mitochondrial genome assembly

Short Illumina reads from four unenriched libraries Accessions SRX2498822-SRX2498825

[66], that contained from 44.6 to 157.7 million reads, were de novo assembled with

SOAPdenovo2 [83] (soapdenovo2/240-bg2qxy6; default parameters). Time to complete assemblies and total contigs assembled were quantified, and compared to those for our eight libraries constructed from mitochondrion-enriched extracts. Read depth among contigs within each resulting SOAPdenovo2 assembly were determined by realigning component reads using bowtie2 [69] (bowtie2/2.3.4.1-py2-jl5zqym) in unpaired mode (-U) as described above, and subsequent .bam file conversion, indexing, and retrieval of read counts performed using

Samtools [70] -view, -index, and -idxstats commands respectively. Normalized read counts for contig length was performed by calculation of reads per kilo base per million mapped reads

(RPKM; number of reads / (contig length/1000 * total number of reads/1,000,000). RPKM was analogously estimated from .bam output from SPAdes assemblies performed above for enriched library read data, and significant of comparative difference in assembled contig length and read depth estimates using paired t-tests within Rstudio [62,63] and evaluated at a threshold α = 0.05.

The SOAPdenovo2 assembled contigs assembled from each of the unenriched library reads were loaded into separate local BLAST databases. Each database was then queried with the entire reference O. nubilalis mitochondrial genome sequence (AF442957.1) using the BLASTn 207 algorithm [72], with results filtered by an E-value cutoff of ≤ 10-40 and sorted by query start position. The number and percent identity of MtD fragments within each assembly were compared to that obtained from separate assemblies derived from enriched libraries (see above).

Data Accessibility

Illumina HiSeq3000 reads can be found on National Center for Biotechnology

Information (NCBI) BioProject PRJNA604593 under Short Read Archive (SRA) accession numbers SRR11007774-SRR11007781. Annotated mitochondrial genome sequence assemblies are submitted to the NCBI non-redundant (nr) database under accessions MT492030.1-

MT492037.1. The Ostrinia nubilalis mitochondrial genome used as a reference in this study can be found under NCBI GenBank accession AF442957.1. O. nubilalis population Pool-seq reads that were considered unenriched controls are available in the SRA database (Accessions

SRX249882-SRX2498825 [62]). Additional supplemental information can be found associated with the publication: https://static-content.springer.com/esm/art%3A10.1038%2Fs41598-020-

76088-0/MediaObjects/41598_2020_76088_MOESM1_ESM.pdf

Acknowledgements

This work is supported by the United States Department of Agriculture (USDA),

Agricultural Research Service (ARS) (CRIS Project 5030-22000-018-00D), and the Iowa

Agriculture and Home Economics Experiment Station, Ames, IA (Project 3543). Any mention of commercial products in this work does not represent an endorsement or recommendation by

USDA for its use. We thank Dr. Andrew Severin and Dr. Maryam Sayadi at the Iowa State

University Genomic Informatics Facility for guidance in computational methods. We also thank

Keith Binde from the USDA-ARS Corn Insects and Crop Genetics Research Unit, Ames, IA, for maintaining and providing insects used in this study. USDA is an equal opportunity employer and provider. 208

References

1. Crozier, R. H. & Crozier, Y. C. The mitochondrial genome of the honeybee Apis mellifera: complete sequence and genome organization. Genetics 133, 97–117 (1993).

2. Wolstenholme, D. R. Genetic novelties in mitochondrial genomes of multicellular animals. Current Opinion in Genetics & Development 2, 918–925 (1992).

3. Brown, W. M., George, M. & Wilson, A. C. Rapid evolution of animal mitochondrial DNA. Proc Natl Acad Sci U S A 76, 1967–1971 (1979).

4. Moritz, C., Dowling, T. E. & Brown, W. M. Evolution of Animal Mitochondrial Dna: Relevance for Population Biology and Systematics. Annual Review of Ecology and Systematics 18, 269–292 (1987).

5. Piganeau, G., Gardner, M. & Eyre-Walker, A. A broad survey of recombination in animal mitochondria. Mol. Biol. Evol. 21, 2319–2325 (2004).

6. Avise, J. C. Gene Trees and Organismal Histories: A Phylogenetic Approach to Population Biology. Evolution 43, 1192–1208 (1989).

7. Cameron, S. L. Insect Mitochondrial Genomics: Implications for Evolution and Phylogeny. Annual Review of Entomology 59, 95–117 (2014).

8. Coates, B. S. Assembly and annotation of full mitochondrial genomes for the corn rootworm species, Diabrotica virgifera virgifera and Diabrotica barberi (Insecta: Coleoptera: Chrysomelidae), using Next Generation Sequence data. Gene 542, 190–197 (2014).

9. Douglas, D. A. & Gower, D. J. Snake mitochondrial genomes: phylogenetic relationships and implications of extended taxon sampling for interpretations of mitogenomic evolution. BMC Genomics 11, 14 (2010).

10. Sun, W. et al. Comparison of complete mitochondrial DNA sequences between old and new world strains of the cowpea aphid, Aphis craccivora (Hemiptera: Aphididae). Agri Gene 4, 23–29 (2017).

11. Yu, L., Li, Y.-W., Ryder, O. A. & Zhang, Y.-P. Analysis of complete mitochondrial genome sequences increases phylogenetic resolution of bears (Ursidae), a mammalian family that experienced rapid speciation. BMC Evolutionary Biology 7, 198 (2007).

12. Archie, J. W. Homoplasy excess ratios: New indices for measuring levels of homoplasy in phylogenetic systematics and a critique of the consistency index. Syst Biol 38, 253–269 (1989).

13. Harrison, R. G. Animal mitochondrial DNA as a genetic marker in population and evolutionary biology. Trends Ecol. Evol. (Amst.) 4, 6–11 (1989). 209

14. Boore, J. L., Macey, J. R. & Medina, M. Sequencing and comparing whole mitochondrial genomes of animals. Meth. Enzymol. 395, 311–348 (2005).

15. Bevers, R. P. J. et al. Extensive mitochondrial population structure and haplotype-specific phenotypic variation in the Drosophila Genetic Reference Panel. bioRxiv 466771 (2018) doi:10.1101/466771.

16. Eimanifar, A., Kimball, R. T., Braun, E. L. & Ellis, J. D. Mitochondrial genome diversity and population structure of two western honey bee subspecies in the Republic of South Africa. Scientific Reports 8, 1333 (2018).

17. Mason, C. E. et al. European Corn Borer - Ecology and Management and Association with other Corn Pests. (Iowa State University Extension and Outreach, 2018).

18. Hutchison, W. D. et al. Areawide Suppression of European Corn Borer with Bt Maize Reaps Savings to Non-Bt Maize Growers. Science 330, 222–225 (2010).

19. Coates, B. S., Dopman, E. B., Wanner, K. W. & Sappington, T. W. Genomic mechanisms of sympatric ecological and sexual divergence in a model agricultural pest, the European corn borer. Curr Opin Insect Sci 26, 50–56 (2018).

20. Coates, B. S., Sumerford, D. V. & Hellmich, R. L. Geographic and voltinism differentiation among North American Ostrinia nubilalis (European corn borer) mitochondrial cytochrome c oxidase haplotypes. Journal of Insect Science 4, (2004).

21. Hoshizaki, S. et al. Limited variation in mitochondrial DNA of maize-associated Ostrinia nubilalis (Lepidoptera: Crambidae) in Russia, Turkey and Slovenia. EJE 105, 545–552 (2013).

22. Marçon, P. C., Taylor, D. B., Mason, C. E., Hellmich, R. L. & Siegfried, B. D. Genetic similarity among pheromone and voltinism races of Ostrinia nubilalis (Hübner) (Lepidoptera: Crambidae). Insect Mol. Biol. 8, 213–221 (1999).

23. Li, J. et al. The genetic structure of Asian corn borer, Ostrinia furnacalis, populations in China: haplotype variance in northern populations and potential impact on management of resistance to transgenic maize. J. Hered. 105, 642–655 (2014).

24. Wang, Y. et al. Introgression between divergent corn borer species in a region of sympatry: Implications on the evolution and adaptation of pest arthropods. Mol. Ecol. 26, 6892–6907 (2017).

25. Sperling, F. A. & Hickey, D. A. Mitochondrial DNA sequence variation in the spruce budworm species complex (Choristoneura: Lepidoptera). Mol Biol Evol 11, 656–665 (1994).

26. Ciminera, M. et al. Genetic Variation and Differentiation of Hylesia metabus (Lepidoptera: Saturniidae): Moths of Public Health Importance in French Guiana and in Venezuela. J Med Entomol 56, 137–148 (2019). 210

27. Eo, S. H. & DeWoody, J. A. Evolutionary rates of mitochondrial genomes correspond to diversification rates and to contemporary species richness in birds and reptiles. Proceedings of the Royal Society B: Biological Sciences 277, 3587–3592 (2010).

28. Smith, D. R. The past, present and future of mitochondrial genomics: have we sequenced enough mtDNAs? Brief Funct Genomics 15, 47–54 (2016).

29. Mascolo, C. et al. Comparison of mitochondrial DNA enrichment and sequencing methods from fish tissue. Food Chemistry 294, 333–338 (2019).

30. Marquis, J. et al. MitoRS, a method for high throughput, sensitive, and accurate detection of mitochondrial DNA heteroplasmy. BMC Genomics 18, 326 (2017).

31. Richards, S. M. et al. Correction: Low-cost cross-taxon enrichment of mitochondrial DNA using in-house synthesised RNA probes. PLOS ONE 14, e0213296 (2019).

32. Li, M., Schroeder, R., Ko, A. & Stoneking, M. Fidelity of capture-enrichment for mtDNA genome sequencing: influence of NUMTs. Nucleic Acids Res. 40, e137 (2012).

33. Grau, E. T. et al. Survey of mitochondrial sequences integrated into the bovine nuclear genome. Scientific Reports 10, 2077 (2020).

34. Lenglez, S., Hermand, D. & Decottignies, A. Genome-wide mapping of nuclear mitochondrial DNA sequences links DNA replication origins to chromosomal double-strand break formation in Schizosaccharomyces pombe. Genome Res 20, 1250–1261 (2010).

35. Srinivasainagendra, V. et al. Migration of mitochondrial DNA in the nuclear genome of colorectal adenocarcinoma. Genome Medicine 9, 31 (2017).

36. Hazkani-Covo, E., Zeller, R. M. & Martin, W. Molecular Poltergeists: Mitochondrial DNA Copies (numts) in Sequenced Nuclear Genomes. PLOS Genetics 6, e1000834 (2010).

37. Graham, J. M. Isolation of Mitochondria from Tissues and Cells by Differential Centrifugation. Current Protocols in Cell Biology 4, 3.3.1-3.3.15 (1999).

38. Coates, B. S., Sumerford, D. V., Hellmich, R. L. & Lewis, L. C. Partial mitochondrial genome sequences of Ostrinia nubilalis and Ostrinia furnicalis. Int J Biol Sci 1, 13–18 (2005).

39. Zhou, N., Dong, Y., Qiao, P. & Yang, Z. Complete Mitogenomic Structure and Phylogenetic Implications of the Genus Ostrinia (Lepidoptera: Crambidae). Insects 11, (2020).

40. Barr, C. M., Keller, S. R., Ingvarsson, P. K., Sloan, D. B. & Taylor, D. R. Variation in Mutation Rate and Polymorphism Among Mitochondrial Genes of Silene vulgaris. Molecular Biology and Evolution 24, 1783–1791 (2007).

41. Hagelberg, E. Recombination or mutation rate heterogeneity? Implications for Mitochondrial Eve. Trends Genet. 19, 84–90 (2003). 211

42. Martin, A. P. & Palumbi, S. R. Body size, metabolic rate, generation time, and the molecular clock. Proc Natl Acad Sci U S A 90, 4087–4091 (1993).

43. Quispe-Tintaya, W., White, R. R., Popov, V. N., Vijg, J. & Maslov, A. Y. Rapid mitochondrial DNA isolation method for direct sequencing. Methods Mol. Biol. 1264, 89–95 (2015).

44. Franco, L. M. et al. Integrative genomic analysis of the human immune response to influenza vaccination. eLife 2, e00299 (2013).

45. Gould, M. P. et al. PCR-free enrichment of mitochondrial DNA from human Blood and cell lines for high quality next-generation DNA sequencing. PLoS ONE 10, e0139253 (2015).

46. Lodish, H. et al. Molecular Cell Biology. (W. H. Freeman, 2000).

47. Hornig-Do, H.-T. et al. Isolation of functional pure mitochondria by superparamagnetic microbeads. Anal. Biochem. 389, 1–5 (2009).

48. Dopman, E. B., Robbins, P. S. & Seaman, A. Components of reproductive isolation between North American pheromone strains of the European corn borer. Evolution 64, 881–902 (2010).

49. Kytövuori, L. et al. Case report: a novel frameshift mutation in the mitochondrial cytochrome c oxidase II gene causing mitochondrial disorder. BMC Neurol 17, (2017).

50. Beckenbach, A. T., Robson, S. K. A. & Crozier, R. H. Single Nucleotide +1 Frameshifts in an Apparently FunctionalMitochondrial Cytochrome b Gene in Ants of the Genus Polyrhachis. J Mol Evol 60, 141–152 (2005).

51. Russell, R. D. & Beckenbach, A. T. Recoding of Translation in Turtle Mitochondrial Genomes: Programmed Frameshift Mutations and Evidence of a Modified Genetic Code. J Mol Evol 67, 682–695 (2008).

52. Atkins, J. F. & Björk, G. R. A Gripping Tale of Ribosomal Frameshifting: Extragenic Suppressors of Frameshift Mutations Spotlight P-Site Realignment. Microbiol. Mol. Biol. Rev. 73, 178–210 (2009).

53. Fox, T. D. & Weiss-Brummer, B. Leaky +1 and −1 frameshift mutations at the same site in a yeast mitochondrial gene. Nature 288, 60–63 (1980).

54. Kim, C., Hoshizaki, S., Huang, Y., Tatsuki, S. & Ishikawa, Y. Usefulness of mitochondrial COII gene sequences in examining phylogenetic relationships in the Asian corn borer, Ostrinia furnacalis, and allied species (Lepidoptera: Pyralidae). Applied Entomology and Zoology 34, 405–412 (1999).

55. Ruisheng, Y., W. Zhenying, & H. Kanglai. Genetic diversity and phylogeny of the genus Ostrinia (Lepidoptera: Crambidae) inhabiting China inferred from mitochondrial COI gene. Journal of Nanjing Agricultural University (2011). 212

56. Coates, B. S., Sumerford, D. V., Siegfried, B. D., Hellmich, R. L. & Abel, C. A. Unlinked genetic loci control the reduced transcription of aminopeptidase N 1 and 3 in the European corn borer and determine tolerance to Bacillus thuringiensis Cry1Ab toxin. Insect Biochemistry and Molecular Biology 43, 1152–1160 (2013).

57. Lewis, L. C. & Lynch, R. E. Rearing the European corn borer, Ostrinia nubilalis (Hubner), on diets containing corn leaf and wheat germ. Iowa State journal of science 44, 9–14 (1969).

58. Lazarow, A. & Cooperstein, S. J. Studies on the mechanism of Janus green B staining of mitochondria. I. Review of the literature. Exp. Cell Res. 5, 56–69 (1953).

59. Folmer, O., Black, M., Hoeh, W., Lutz, R. & Vrijenhoek, R. DNA primers for amplification of mitochondrial cytochrome c oxidase subunit I from diverse metazoan invertebrates. Mol Mar Biol Biotechnol 3, 294–299 (1994).

60. Coates, B. S. et al. Frequency of hybridization between Ostrinia nubilalis E-and Z- pheromone races in regions of sympatry within the United States. Ecol Evol 3, 2459–2470 (2013).

61. Integrated DNA Technologies. https://www.idtdna.com/scitools/Applications/RealTimePCR/.

62. RStudio Team. RStudio: Integrated Development for R. (RStudio, PBC, 2020).

63. R Core Team. R: A language and environment for statistical computing. (R Foundation for Statistical Computing, 2019).

64. Lenth, R., Singmann, H., Love, J., Buerkner, P. & Herve, M. emmeans: Estimated Marginal Means, aka Least-Squares Means. (2020).

65. Graves, S. & Dorai-Raj, H.-P. P. and L. S. with help from S. multcompView: Visualizations of Paired Comparisons. (2019).

66. Kozak, G. M. et al. A combination of sexual and ecological divergence contributes to rearrangement spread during initial stages of speciation. Mol. Ecol. 26, 2331–2347 (2017).

67. Andrews, S. Babraham Bioinformatics - FastQC A Quality Control tool for High Throughput Sequence Data. https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ (2010).

68. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120 (2014).

69. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nature Methods 9, 357–359 (2012).

70. Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078– 2079 (2009). 213

71. Bankevich, A. et al. SPAdes: A New Genome Assembly Algorithm and Its Applications to Single-Cell Sequencing. J Comput Biol 19, 455–477 (2012).

72. Altschul, S. F., Gish, W., Miller, W., Myers, E. W. & Lipman, D. J. Basic local alignment search tool. Journal of Molecular Biology 215, 403–410 (1990).

73. Huang, X. & Madan, A. CAP3: A DNA Sequence Assembly Program. Genome Res 9, 868– 877 (1999).

74. Robinson, J. T. et al. Integrative Genomics Viewer. Nat Biotechnol 29, 24–26 (2011).

75. Narasimhan, V. et al. BCFtools/RoH: a hidden Markov model approach for detecting autozygosity from next-generation sequencing data. Bioinformatics 32, 1749–1751 (2016).

76. Danecek, P. et al. The variant call format and VCFtools. Bioinformatics 27, 2156–2158 (2011).

77. Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010).

78. Kumar, S., Nei, M., Dudley, J. & Tamura, K. MEGA: a biologist-centric software for evolutionary analysis of DNA and protein sequences. Brief. Bioinformatics 9, 299–306 (2008).

79. Thompson, J. D., Gibson, T. J. & Higgins, D. G. Multiple sequence alignment using ClustalW and ClustalX. Curr Protoc Bioinformatics Chapter 2, Unit 2.3 (2002).

80. MView < Multiple Sequence Alignment < EMBL-EBI. https://www.ebi.ac.uk/Tools/msa/mview/.

81. NCBI Multiple Sequence Alignment Viewer 1.15.0. https://www.ncbi.nlm.nih.gov/projects/msaviewer/.

82. Adachi, J. & Hasegawa, M. Model of amino acid substitution in proteins encoded by mitochondrial DNA. J. Mol. Evol. 42, 459–468 (1996).

83. Luo, R. et al. SOAPdenovo2: an empirically improved memory-efficient short-read de novo assembler. Gigascience 1, 18 (2012).

214

Figures and Tables

Figure 1. Representative example of differentially fractionated samples of Ostrinia nubilalis homogenized thorax tissue to acquire nuclear and mitochondrial components at 6000 and 13,000 rcf, respectively. Janus Green B staining of the mitochondrial fraction indicated abundant mitochondria (a). Gel electrophoresis of serial diluted DNA extracted from the mitochondrial and nuclear fractions amplified in duplicate with mitochondrial primers (HCO/LCO) and nuclear primers (Onapn1-F and -R) (b). Mitochondrial fraction (13,000 rcf) was PCR amplified with mitochondrial primers (i) and nuclear primers (ii). Nuclear fraction (6000 rcf) was PCR amplified with mitochondrial primers (iii) and nuclear primers (iv). Although nuclear DNA was present in both the mitochondrial and the nuclear fractions, mtDNA was only present in the mitochondrial fraction. 215

Figure 2. Estimates of mitochondrion enrichment of Ostrinia nubilalis from homogenized thorax tissue through differential fractionation prior to DNA extraction. (a) Significantly lower CT values were estimated by real-time PCR among mitochondrial fractions (14.83 ± 0.37) amplified with mitochondrial cytochrome c oxidase subunit I (coxI) primers (OnCOXIrtS and OnCOXIrtA) compared to the nuclear fractions (17.43 ± 0.67) or unenriched controls (20.0 ± 4.34) (ANOVA, p < 0.001). Different letters above bars represent a significant difference. (b) Significant increase in the ratio of reads from enriched mitochondrion samples (0.255 ± 0.054) that aligned to the O. nubilalis mitochondrial genome reference (AF442957.1) compared to unenriched controls (0.0005 ± 0.0017; two-sample t- test, p < 0.0001) (c) Difference in fold-enrichment estimates for mitochondrion fractions estimated by with mitochondrial target amplicons estimate enrichment by real-time quantitative PCR (mean 86.1 ± 22.3) and read alignment (mean 48.2 ± 10.2). Although significant (paired t-test, p = 0.0046), both estimation methods indicate consistent trends in enrichment. 216

Figure 3. In-frame and frameshift mutations predicted among assembled Ostrinia mitochondrial genomes. (a) Multiple sequence alignment of an ATP synthase subunit 8 (atp8) gene region showing a putative in-frameATT nucleotide (Isoleucine, I) duplication within F1_Family_ID_07 (MT492034.1) compared to other Ostrinia. (b) Frameshift impacting three codons of the COI protein-coding sequence caused by flanking compensatory insertion/deletion mutations that return sequences to the same downstream frame. Genus species (Gsp)abbreviations: Onub, O. nubilalis; Ofur, O. furnacalis; Osca, O. scapulalis; Ozea, O. zealis; Opal, O. palustralis. Open, O. penitalis; Oven, Ochlodes venata (Lepidoptera: Hesperiidae); Pepa, Parnassius epaphus (Lepidoptera: Papilionidae). Conserved sites highlighted in grey. 217

Figure 4. Alignment of 60 C-terminal residues from representative NADH Dehydrogenase subunit 4 (ND4) protein sequence accessions. Conserved residues (highlighted) showed ≥ 83.33% identity between species, with the exception of the 29 terminal residues from O. nubilalis AAL66246.1 predicted in the reference mitochondrial genome (AF442957.1). Frameshift in ND4 accession AAL66246.1 predicted from the reference AF442957.1 putatively caused by insertion of a cytosine (C) nucleotide at position 1254 of the ND4 CDS (genome position 8013; Table S2). Companion Maximum-Likelihood (ML) constructed based on the mtREV + F + G model of sequence evolution that maximized the Bayesian Information Criterion (BIC) score at 4929.63, and used an empirically-derived discrete gamma distribution (shape parameter; G) of 0.3988 that minimized the loglikelihood score (-2330.81). Total branch length of 0.56 amino acid changes per site. Species abbreviations: Sfru, Spodoptera frugiperda; Tni; Trichoplusia ni; Epos, Epiphyas postvittana; Mvit, Maruca vitrata; Ostrinia species abbreviations as in Fig. 3).

Table 1. Ostrinia nubilalis mitochondrial genome sequence assembly accessions

Accession Description Assembly Len Citation Read type/platform AF442957.1 RefSeq NC_003367.1 14,535 bp Coates et al. 2004 [20] Sanger/ABI3700 MN792233.1 Yili, Xinjiang Region 15,248 bp Zhou et al. 2020 [35] Illumina/HiSeq MT492030.1 F1_Family_ID_02 14,838 bp This study Illumina/MiSeq MT492031.1 F1_Family_ID_03 14,817 bp This study Illumina/MiSeq MT492032.1 F1_Family_ID_04 14,691 bp This study Illumina/MiSeq MT492033.1 F1_Family_ID_06 14,621 bp This study Illumina/MiSeq MT492034.1 F1_Family_ID_07 14,622 bp This study Illumina/MiSeq MT492035.1 F1_Family_ID_11 15,082 bp This study Illumina/MiSeq MT492036.1 F1_Family_ID_12 14,809 bp This study Illumina/MiSeq MT492037.1 F1_Family_ID_18 15,064 bp This study Illumina/MiSeq

218

CHAPTER 8. GENERAL CONCLUSION

This dissertation successfully addressed gaps in the knowledge of monarch butterfly biology that are instrumental to creating responsive conservation strategies. Habitat restoration plans for species of interest should be created to facilitate behavioral attributes of movement

(Lima and Zollner 1996, Zollner and Lima 1997). Using a variety of experimental techniques, including observation, VHF radio telemetry, and next-generation sequencing, we investigated larval and adult monarch movement behavior at the host plant, patch, and landscape scales.

Results of the research will help inform conservation plans with recommendations for the spatial arrangement of restored habitat patches to create a connected landscape for monarch butterflies in their summer breeding range. In addition to providing valuable insights, advancements were made by developing new methods, employing technology with novel approaches, and applying current analyses in spatial statistics. This dissertation also opens doors for further research.

At the milkweed patch scale explored in chapter 2, we determine that larval movement behavior and biomass requirements are critical aspects of monarch larval biology that should be incorporated in habitat restoration and maintenance practices. Because the eastern North

American monarch population grows exponentially over the summer prior to fall migration, it is important that the first generation in the Upper Midwest is productive (Flockhart et al. 2015,

Oberhauser et al. 2017). Under our greenhouse conditions and consistent with previous field studies, monarch larvae unfailingly abandon their natal ramet of common milkweed and subsequent ramets before all of the available leaf biomass on a ramet is consumed and prior to the pre-pupal wandering stage. In the spring, when monarchs first arrive in the Upper Midwest, milkweed are 10–35 cm in height and likely provide sufficient milkweed biomass to support the development of a single egg to pupation. However, because larvae abandon their natal ramets, if 219 the natal ramet is isolated from other milkweed plants, the larvae will likely die before finding another plant during a random walk. In conclusion, while previous modeling studies indicate that isolated ramets in the matrix can increase the number of eggs laid in the landscape (Zalucki et al.

2016), our findings suggest that eggs or larvae observed on isolated ramets will have low survival. Patches containing at least two to four ramets of closely-spaced common milkweed provide sufficient biomass for development and would increase the likelihood that larvae moving in random directions would encounter non-natal milkweed ramets to support development through pupation.

We conclude that regardless of the mechanism(s) that invoke the seemingly innate movement behavior of milkweed ramet abandonment by monarch larvae, there are implications for larval monitoring projects. Larvae in our studies abandoned plants an average of three times through development, and this movement occurred most often during the day. Most monitoring protocols are based on observing milkweed leaves and ramets for eggs and larvae during the day

(Prysby and Obserhauser 2004, De Anda and Oberhauser 2015, Nail et al. 2015, Geest et al.

2019). In these protocols, when larvae are expected in a subsequent sampling period and not found, they are presumed to be dead (Prysby and Obserhauser 2004, Nail et al. 2015, Geest et al.

2019). When a larva is observed, but there was no detection in the previous sampling period, data records are adjusted under the assumption of missed detection in the prior sampling period

(De Anda and Oberhauser 2015, Nail et al. 2015, Geest et al. 2019). Although these studies account for failed detection when new larvae are observed, they do not account for larvae moving out of the patch. Survey design should account for imperfect larval detection, which is often disregarded when studying invertebrates because of their general abundance in comparison to vertebrates (Kellner and Swihart 2014). For larval monarch surveys, our results suggest 220 current monitoring protocols may overestimate field mortality and do not account for larval movement into and out of habitat patches. Monitoring the ground and vegetation surrounding milkweed ramets as well as searching ramets for standardized durations of time during observation periods may reduce bias in survey results.

From this work, several new questions have been identified. Specifically, why do monarch larvae abandon their natal ramet of milkweed prior to consuming all the available leaf biomass and prior to the pre-pupal larval wandering phase? Hypotheses include finding cover to avoid predation, finding young leaves that have higher nitrogen concentrations, are easier to digest, and lower concentrations of plant-induced defensive chemicals. These hypotheses could be tested under controlled greenhouse of laboratory conditions with quantification of leaf texture, nitrogen content, and defensive compounds. Additionally, there are further analyses possible with the already conducted experiments presented in the Appendix for Chapter 2, including the association of ramet abandonment with molting status and the impacts of intraspecific competition on timing and frequency of natal ramet abandonment behavior.

Through five years of landscape-scale movement experiments (chapters 3-6), we released

393 radio-tagged, sham-tagged, and untagged monarch butterflies in various land cover types.

With the use of radio telemetry, we attained more information about monarch butterfly movement than was possible by visual observation only. We developed a new data collection system capable of estimating the location of a monarch butterfly in a sod field. We analyzed movement characteristics, including step length, directionality, and perceptual range. We employed novel approaches, specifically continuous-time movement models, to analyze monarch butterfly occurrence. Finally, we used the results to inform parameter selection in a spatially explicit agent-based model reported by Grant et al. (2018). 221

Using radio telemetry, in chapter 3, we expanded understanding of flight steps and directionality by tracking locations of radio-tagged monarchs every minute in a 4-ha restored prairie containing milkweed, nectar plants, and grasses at natural densities and distributions.

Within the prairie, we quantified small, undirected steps, while steps associated with exiting the prairie were large with high directionality. These movement patterns are qualitatively consistent with simulation modeling (Grant et al. 2018). We found that the majority of steps within the prairie were below 50 m, and most were under 5 m, consistent with Zalucki and Kitching (1982).

In contrast to Zalucki and Kitching’s (1982) observation of turn angles centering on 0 degrees, we recorded wide turn angles that centered around 180 degrees suggesting that monarchs were pivoting. Within the prairie, these 180-degree turn angles were associated with foraging behavior and monarchs flying small distances from flower to flower in a pattern with low directionality.

Consequently, the total distance traveled was much greater than the Euclidian distance displacement.

To the best of our knowledge, the automated radio telemetry system (ARTS) described in chapter 4 is the first ARTS designed to track insects (see Kissling et al. 2014 for a recent review of insect telemetry techniques). Our statistical method to estimate location incorporated both the effect of distance and angle on the received power by each of four Yagi antennae on a receiving station, which was not explored previously (Cochran et al. 1965, Larkin et al. 1996, Kays et al.

2011). With this method, we were able to successfully estimate locations of stationary transmitters at distances beyond visual detection with a median accuracy of 14.7 m. This accuracy from a distance of up to 150 m from a receiving station is appropriate for the spatial scale of our movement questions and is a significant improvement compared to visual observations at distances up to, but not exceeding, 50 m with no error estimates. In addition, our 222 method produced comparable accuracy when locating a mobile transmitter attached to an investigator's hat at a temporal resolution that is 12 times more frequent than the shortest telemetry data collection interval previously reported with insects (Kissling et al. 2014, Knight et al. 2019, Wang et al. 2019, Fisher et al. 2020). These results suggest under optimized conditions, with the transmitter antenna consistently vertical and 1.6 m above the ground, our system, and method of estimating location can be used to reasonably quantify locations and error estimates of stationary and moving low-mass transmitters at distances beyond human vision.

Using current, commercially available equipment, our findings likely represent the limit of existing ARTS technology to track non-migratory movement of insects. Although error estimates were large compared to ARTS using heavier, higher power transmitters for small mammals and birds, our system can estimate fine-scale locations within a 250 m2 area at a frequency appropriate for our research questions. Estimating locations based on visual observations is not possible at these spatial and temporal scales. When combined with handheld radio telemetry to verify locations of tagged insects outside an ARTS grid, this system provides a cost-effective means to estimate locations of insects at distances beyond human visual range. Our method with 0.22 g transmitters has the potential to be adapted for use with other large insect species, small mammals, and small birds.

As reported in chapter 5, olfactory and visual cues guide insect movement as they disperse across a landscape in search of resources (Bell 1991, Carde and Willis 2008), with directed movement in response to attractive stimuli (within perceptual range) and undirected movement when no stimuli are detected (Baker and Linn 1984, Finch 1986, Bell 1991). To support conservation plans for the declining population of eastern North American monarchs, agent-based population models that include female movement algorithms suggest establishing 223 small habitat patches at distances within the monarch’s perceptual range will more effectively increase egg densities within landscapes, as opposed to establishing patches in random locations

(Zalucki et al. 2016, Grant et al. 2018). However, uncertainty in monarch perceptual range, between 50 to 400 m based on expert elicitation, results in significantly different simulated movement patterns in a spatially-explicit Iowa, USA landscape (see Grant et al. 2018; Figure 5).

We designed a field-scale study to quantify female monarch perceptual range and further inform input estimates for spatially-explicit, agent-based modeling and conservation planning.

Radio-tagged, sham-tagged, and untagged monarch females were released 5, 25, 50, or 75 m downwind of potted Asclepias syriaca and Liatris sp. at several sod fields surrounded by corn, soybean, or non-native cool-season grasses. The goal was to determine the extent of directed flight into the wind toward the potted resources. Our empirical evidence and model output estimates suggest an assumption of female monarch perceptual range to host plant and nectar forage resources of approximately 50 m (Grant et al. 2018) is reasonable. Establishing habitat patches 50 m apart will likely create agroecosystems with spatial arrangements of resources that are appropriate for monarch movement behavior.

In chapter 6, we report the most robust dataset of directly tracked movement of individual breeding-season monarch butterflies to assess utilization of agricultural landscapes to date.

Monarchs were observed moving among habitat classes and, although not always, were observed performing natural up-wind search behavior. Step lengths and turn angles were mostly below 50 m and directional, regardless of habitat class or spatial configuration. This observation was likely a response to floral and milkweed density rather than the presence or absence of resources. As reported by Fisher et al. (2020) and Fisher and Bradbury (2021), we observed large exiting steps from study areas of up to ≥ 1,000 m. These dispersal steps occurred most frequently from 224 resource-rich habitat and could indicate initiation of a long-range search for a new research-rich habitat, which in turn would increase egg distribution across the landscape. In summary, despite proximity to resource-rich habitat, 78.9% of monarchs not released in a resource-rich habitat were observed in resource-rich habitat, with some monarchs flying over 500 m to find resource- rich areas.

Results from chapters 3-6 provide empirical landscape-scale estimates of perceptual range, flight directionality, flight step lengths, and habitat utilization that are needed to simulate monarch population dynamics (Zalucki et al. 2016, Grant et al. 2018, Grant and Bradbury 2019).

This work advances knowledge at a landscape scale and would not have been possible without the aid of radio telemetry to relocate individuals beyond human visual range. Although many reports have been made, there is a great deal that has yet to be explored, specifically related to the order of movement behaviors. The sequence of behaviors could be analyzed to provide the foundation for Hidden-Markov Models to explore observed and hidden states. Additional exploration of the order of step lengths and turn angles would determine if any auto-correlation is present. Finally, the order of visitations and observations of time spent feeding on various floral resources could be analyzed to determine feeding preferences, which would be invaluable in providing seed mix recommendations for habitat establishment in Iowa.

As reported in chapter 7, we successfully developed a method to investigate variation in mitochondrial DNA. This chapter lays the foundation to further investigate monarch butterfly haplotype at the continental scale to aid in estimating the effective population size (Ne) of the overwintering population and assessing the potential resiliency of the eastern North American monarch to ecological change, which was proposed as a USDA-NIFA Education and Workforce 225

Development Postdoctoral Fellowship Program application in August 2020 under the guidance of Dr. Alan Wanamaker, Dr. Brad Coates, and Dr. Steven Bradbury.

Migratory monarchs have experienced population declines for a variety of reasons. It is unknown if the effects of the monarch butterfly population decline and associated potential loss of genetic diversity reduce the ability of monarchs to adapt to a changing climate. Using the workflow I established with O. nubilalis in this dissertation, I will be able to objectively analyze variation within full mtDNA genome sequences across breeding, overwintering, and non- migratory locations and estimate spatial and temporal variance, population structure, and Ne.

Identifying stable isotopic signatures of the same individuals will verify natal origins and define putative migratory paths with greater precision compared to previous studies. Combining genetic variation with stable isotope-based estimates of natal origin will provide more complete datasets that allow estimation of population variability, gene flow, and Ne, which in turn will provide insights on the resiliency of the eastern North American subpopulation monarch butterfly migration. The objectives of this proposal are:

1. Quantify mtDNA genetic variation frequencies and natal origins of non-migratory

summer resident monarchs across six northcentral, eastern, and western states.

Compare variation frequencies to those observed in migratory monarchs

overwintering in Mexico and known outgroup populations from Hawaii and

Australia to estimate subpopulation diversities, migration rates, Ne, and the

degree of inbreeding.

2. Estimate spatial-temporal effects by sampling monarchs in Iowa from May-

October. Compare genetic drift observed among years to variation observed in the

overwintering population, spring breeding generation in Texas, and fall migrants 226

intercepted in Texas. Complement genetic analysis with stable isotope analyses to

verify the natal origins of each monarch for geospatial analyses.

In conclusion, research conducted in this dissertation provided an opportunity to work with diverse, interdisciplinary teams to plan, develop, and execute novel methods and approaches that directly support conservation and management of our nation’s natural resources.

Research plans and findings were actively and efficiently communicated within the scientific community, with decision-makers, and with stakeholders to help inform the implementation of science-based conservation recommendations. Work here helped inform a listing decision for the monarch butterfly under the Endangered Species Act.

References

Bell, W. J. 1991. Searching Behavior: The behavioural ecology of finding resources, Chapman and Hall Animal Behaviour Series. Springer Netherlands.

Carde, R. T., and M. A. Willis. 2008. Navigational strategies used by insects to find distant, wind-borne sources of odor. J Chem Ecol. 34: 854–866.

Cochran, W. W., D. W. Warner, J. R. Tester, and V. B. Kuechle. 1965. Automatic Radio- Tracking System for Monitoring Animal Movements. BioScience. 15: 98–100.

De Anda, A., and K. S. Oberhauser. 2015. Invertebrate natural enemies and stage-specific mortality rrates of monarch eggs and larvae, pp. 60–70. In Oberhauser, K.S., Nail, K.R., Altizer, S. (eds.), Monarchs in a Changing World. Cornell University Press, New York.

Finch, S. 1986. Assessing host-plant finding by insects. In Miller, J.R., Miller, T.A. (eds.), Insect-Plant Interactions, Springer Series in Experimental Entomology. Springer, New York, NY.

Fisher, K. E., J. S. Adelman, and S. P. Bradbury. 2020. Employing very high frequency (VHF) radio telemetry to recreate monarch butterfly flight paths. Environ Entomol. 49: 312– 323.

Grant, T. J., and S. P. Bradbury. 2019. The Role of Modeling in Monarch Butterfly Research and Conservation. Front. Ecol. Evol. 7: 197.

Kellner, K. F., and R. K. Swihart. 2014. Accounting for Imperfect Detection in Ecology: A Quantitative Review. PLoS ONE. 9: e111436.

Kissling, W. D., D. E. Pattemore, and M. Hagen. 2014. Challenges and prospects in the telemetry of insects. Biological Reviews. 89: 511–530.

Knight, S. M., G. M. Pitman, D. T. T. Flockhart, and D. R. Norris. 2019. Radio-tracking reveals how wind and temperature influence the pace of daytime insect migration. Biol. Lett. 15: 20190327.

Lima, S. L., and P. A. Zollner. 1996. Towards a behavioral ecology of ecological landscapes. Trends in Ecology & Evolution. 11: 131–135.

Nail, K. R., C. Stenoien, and K. S. Oberhauser. 2015. Immature Monarch Survival: Effects of Site Characteristics, Density, and Time. Ann Entomol Soc Am. 108: 680–690.

228

Prysby, M. D., and K. S. Obserhauser. 2004. Temporal and geographic variation in monarch densities: Citizen scientists document monarch population patterns, pp. 9–20. In Oberhauser, K.S., Solensky, M.J. (eds.), The Monarch Butterfly Biology and Conservation. Cornell University Press, New York.

Zalucki, M. P., and R. L. Kitching. 1982. The Analysis and Description of Movement in Adult Danaus plexippus L. (Lepidoptera: Danainae). Behaviour. 80: 174–198.

Zalucki, M. P., H. R. Parry, and J. M. Zalucki. 2016. Movement and egg laying in Monarchs: To move or not to move, that is the equation. Austral Ecology. 41: 154–167.

Zollner, P. A., and S. L. Lima. 1997. Landscape-Level Perceptual Abilities in White-Footed Mice: Perceptual Range and the Detection of Forested Habitat. Oikos. 80: 51–60.