The Pennsylvania State University

The Graduate School

BIOLOGICAL TIMING ACROSS MULTIPLE ECOLOGICAL SCALES AND SYSTEMS

A Dissertation in

Biology

Damie Pak

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

December 2020

The dissertation of Damie Pak was reviewed and approved by the following:

Ottar Bjørnstad Distinguished Professor of Entomology and Biology Dissertation Advisor

Jim Marden Distinguished Professor of Biology Chair of Committee

Jesse Lasky Assistant Professor of Biology and Ecology

Jessica Conway Assistant Professor of Mathematics

David Biddinger Associate Professor of Entomology

Timothy Jegla Professor of Biology Head of Biology

iii

ABSTRACT

Phenology, the timing of biological events, is crucial for understanding many of the ecological processes that occur at multiple organismal scales. At the individual-level, organisms respond to abiotic and biotic cues to appropriately time life-cycle events to the seasonal environment. The aggregated individual responses then make up the population’s phenology which is represented through peaks and variance in the time series. Finally, a community of coexisting species provides insights into how abiotic and biotic interactions influence phenology and possibly offer a mechanism of coexistence. By investigating the phenological mechanisms linking these scales across multiple ecological systems, one can then find commonalities useful for the prediction of the ecological consequences of climate change. As many human industries rely on monitoring phenology for decision making (ex: fish harvesting, pest, management, etc.), managers must adapt their policies to reflect ecological changes. In this dissertation, I explore different aspects of phenology in a guild of tortricid moth pests, the Pennsylvanian tick community, and the plant species in hyper-diverse neotropical rainforests. With the tortricid moth pests, I show how climate variability influences individual vital rates and used this information to formulate phenological models. These models can be valuable for understanding how phenology can shift under different climate scenarios. Finally, long-term time series of the tick and plant communities can be useful for exploring how seasonal activities between coexisting species is influenced by biotic interactions. I further demonstrate that these communities can shift over in time. In conclusion, the ubiquity of biological timing across ecological scales and diverse species provide a special opportunity to understand the common phenological mechanisms between different systems and further study of phenology will help foster a broad framework for ecological theory and management strategies in the face of climate change.

TABLE OF CONTENTS

LIST OF FIGURES ...... vii

LIST OF TABLES ...... xii

ACKNOWLEDGEMENTS ...... xiii

Chapter 1 Introduction ...... 1

Individual-level: understanding how environmental variables influence individuals ...... 1 Development ...... 1 Diapause ...... 2 Population-level: Phenological peaks and variance ...... 3 Timing ...... 3 Variance...... 3 Management ...... 4 Community-level: coexistence and multi-species management ...... 5 Varying responses ...... 5 Biotic interactions ...... 5 Community management ...... 6 Objectives ...... 7

Local and regional climate variables driving spring phenology of tortricid pests: a 36- year study ...... 12

Abstract ...... 12 Introduction ...... 13 Materials and Methods ...... 16 Study Site ...... 16 Climate data: local temperatures and regional indices ...... 17 Tortricid species ...... 18 Growing degree‐day model to calculate the spring flight index ...... 21 Partial least‐squares regression ...... 23 Results ...... 25 Meteorological data ...... 25 Annual spring flight index of the tortricid species ...... 27 Partial least‐squares regression ...... 30 Discussion ...... 35 Acknowledgements ...... 38

Incorporating diapause to predict the interannual dynamics of an important agricultural pest ...... 39

Abstract ...... 39 Introduction ...... 40

Methods ...... 42 Study species ...... 42 Time-series...... 42 The modeling framework...... 43 Vital rate functions ...... 45 Model calibration and sensitivity analysis ...... 52 Calculating developmental synchrony ...... 52 Results ...... 54 Discussion ...... 61 Acknowledgment ...... 64 Data Availability ...... 64

Community phenology and multi-species management ...... 65

Abstract ...... 65 Introduction ...... 66 Method ...... 69 The mathematical model...... 69 Single species case ...... 72 Two species with the same economic thresholds ...... 73 Two species with different economic threshold ...... 75 Results ...... 76 Single species management is influenced by developmental variability ...... 76 Control for two species is affected by coefficient of variation and the shift between species ...... 77 The difference between the species’ economic thresholds influence management..80 Discussion ...... 82

A 117-year retrospective analysis of Pennsylvania tick community dynamics ...... 87

Introduction ...... 88 Methods ...... 91 Study locations ...... 91 Submissions ...... 91 Identification ...... 93 Spatial distribution ...... 93 Temporal analysis ...... 94 Host associations ...... 94 Results ...... 95 Spatial analysis ...... 97 Temporal shifts in species abundance ...... 99 Seasonality ...... 100 Host association ...... 103 Discussion ...... 105 Shifts in tick community composition ...... 105 Host Association ...... 107 Multi-faceted approach to tick surveillance ...... 107 Conclusions ...... 109

Multi-scale phenological niches in hyperdiverse Amazonian plant communities ..112

Abstract ...... 112 Introduction ...... 113 Methods ...... 116 Study sites ...... 116 Seed rain data ...... 117 Seed dispersal mechanisms ...... 118 Weather data ...... 119 Statistical analysis ...... 120 Taxonomic and seed dispersal groups ...... 121 Climatic drivers of synchrony vs anti-synchrony ...... 122 Results ...... 123 Community-wide phenology...... 123 Phenology among confamilials ...... 125 Phenology among species sharing dispersal modes ...... 127 The role of temperature and precipitation on community-wide synchrony versus compensatory phenology...... 130 Discussion ...... 132 Do communities exhibit synchronous or compensatory reproduction? ...... 133 Is evidence for phenological niche partitioning strongest among ecologically similar species? ...... 134 Synchrony among related species or species sharing dispersal mode ...... 136 Community phenology and abiotic fluctuations ...... 137 Conclusion ...... 138 Acknowledgements ...... 138

Appendix A Supplementary Information for Chapter 2 ...... 139

Appendix B Supplementary Information for Chapter 3 ...... 143

Appendix C Supplementary Material for Chapter 4 ...... 172

Appendix D Supplementary Material for Chapter 5...... 187

Appendix E Supplementary Material for Chapter 6 ...... 192

References...... 193

vii

LIST OF FIGURES

Figure 2-1: The average monthly temperatures (minimum and maximum) between September and April from 1980 to 2016 at the Fruit Research and Extension Center FREC (Biglerville, Pennsylvania). Linear regressions were performed between the average monthly temperatures and years ...... 26

Figure 2-2: The annual spring flight indices for five tortricid species across 1981–2016...... 28

Figure 2-3: A plot of the Pearson correlation between a pair of species' annual departures from their average date of emergence. A positive coefficient indicates that the two species had similar annual departures with either earlier or later emergences in their spring phenology. All correlations were positive and significant...... 29

Figure 2-4: Scaled and centered regression coefficients from the partial least‐squares regression (PLS) analysis of the bivoltine species' spring flights. We ran three separate PLS regressions for the smoothed daily minimum temperatures, and North Atlantic Oscillation (NAO) and Arctic Oscillation (AO) indices. The colored bars represent predictor variables that were significant with variable importance in projection (VIP) scores > 0.8. Negative regression coefficients (blue) mean that low values of the predictor variable were associated with later emergence. The positive regressions coefficients indicate that an increase in the predictor values led to later emergences...... 31

Figure 2-5: Scaled and centered regression coefficients from the PLS analysis of the multivoltine species' spring flights We ran three separate PLS regressions for the smoothed daily minimum temperatures, and NAO and AO indices. The colored bars represent predictor variables that were significant with VIP scores > 0.8. Negative regression coefficients (blue) mean that low values of the predictor variable was associated with later emergence. The positive regressions coefficients indicate that an increase in the predictor values led to later emergences...... 32

Figure 2-6: A plot of the Pearson correlation coefficients between the monthly temperatures (minimum, maximum, and average) and the spring flight indices for the five tortricid species in the period 1981–2016. The colors were based on the absolute values of the Pearson correlation coefficients. Negative coefficients indicate that decreasing temperatures were linked to later spring emergences. The positive coefficients indicate that increasing temperatures were linked to later spring emergences...... 34

Figure 3-1: The stage-structured physiological model with individuals flowing into and out of stages through development, mortality, and diapause processes. All fifth instar larvae, regardless of age, can be induced into diapause. Diapausing larvae go through the two stages of diapause before entering the pupal stage. To incorporate variable delays due to development or diapause, we used the ‘linear chain trick’ where individuals move through a series of subcompartments within a single stage (Bjørnstad et al. 2016) ...... 43

Figure 3-2: Temperature-dependent responses for the per capita birth rate. The line shows the fit of a gaussian function to experimental data. See Appendix B for parameter estimates...46

viii

Figure 3-3:Temperature responses for development rates across the different life stages. The lines show the fit of the sigmoid function to the experimental data. See Appendix B for parameter estimates...... 47

Figure 3-4: Temperature responses for mortality rates across the different life stages. We assumed that the larval instars share the same temperature-dependent mortality function. The lines represent the fit of the modified Wang function to the experimental data. See Appendix B for parameter estimates...... 48

Figure 3-5:The diapause termination rate is described as a sigmoid function which is dependent on the day of year (left). The diapause termination rate of stage 1 and stage 2 is described as a sigmoidal function and is dependent on the change in daily temperature (right). See Appendix B for parameter estimates...... 50

Figure 3-6: Population dynamics of Cydia pomonella predicted by the model in black. In red is the time series of the pheromone trap data from 1984 to 2016. The y-axis is the log- transformed reproductive adult abundance and the x-axis is the day of year. The period of pheromone capture varies somewhat between years...... 54

Figure 3-7: The sensitivity indices (absolute value) of the timing in adult emergence to changes in the model parameters. Lighter colors represent an increase in sensitivity and indicates that the parameter of interest is important in influencing the model output...... 55

Figure 3-8: The sensitivity indices (absolute value) of accumulated egg, diapausing larvae, and reproductive adult abundance to changes in model parameters. Lighter colors represent the parameter having significant influence on the model output...... 56

Figure 3-9: The population dynamics of C. pomonella predicted by the model in blue with unchanged average temperatures (T). In red is the population dynamics predicted by the model with average temperatures increased by 3 ℃ (T + 3℃) and diapause induction being delayed. All scenarios were simulated from 1984 to 2016 with the y-axis as the log- transformed reproductive adult abundance and the x-axis is the day of year...... 57

Figure 3-10: Top: The circular variance averaged across the day of year with the historical temperature from the FREC (1984-2016) in blue and with daily temperature increased by 3°C in red. The ribbons represent the 95% confidence interval. An increase in circular variance is associated with more overlapping life stages at a given time. Bottom: The averaged circular variance across the day of year with the blue line representing the circular variance calculated with the historical temperature data (1984 to 2016) and an unchanged 푛퐷퐿 and 푛푃. In red, is the circular variance calculated with the daily temperature increased by 3°C and an unchanged 푛퐷퐿 and 푛푃. The orange and yellow represent the averaged circular variance of model run under high-emission scenario with 푛퐷퐿 풐풓 푛푃 varied at 10 and 5 respectively...... 59

Figure 4-1: The schematic of the pesticide induced mortality function described with the Hill Equation (Left) and the decay of pesticide concentration over time (Right). Parameters in the Supplementary Material...... 70

Figure 4-2: A diagram representing the model with the linear chain trick. Individuals flow through the subcompartments of stage A until they are in stage B. If the dwelling time in the entirety of stage A is 1/nα than this can be equivalent to 푛 number of subcompartments ‘chained’ together. As 푛 increases, the CV decreases and there is less developmental variability...... 72

Figure 4-3: A schematic on the two-species management where intervention for the earlier emerging species in red affects the later emerging species in green. The two solid lines represent the species where no intervention occurred, and the dashed lines represents the population after the pesticide spray (vertical line). The horizontal line represents the action threshold which is the pest density when pesticide is applied...... 74

Figure 4-4: The number of sprays needed for species sprayed at action thresholds 40 and 80. On the x-axis is the coefficient of variation while the y-axis is the decay rate of the pesticide. 76

Figure 4-5: The inflow rate of individuals into the last stage at the time of the first pesticide spray. On the x-axis is the coefficient of variation, on the y-axis is the decay rate. The different colors represent the different action thresholds...... 77

Figure 4-6: The number of sprays needed to control for two species when their action thresholds are the same (in this case 40). On the x-axis is the coefficient of variation while the y-axis is the decay rate and each panel represent the shift between the two species...... 78

Figure 4-7: The difference in the number of sprays when interventions influence all species versus when intervention does not affect the other species. More negative value suggests that more sprays are needed for the individual case. On the x-axis is the coefficient of variation while the y-axis is the decay rate. The panel are the shift is the number of days that are between the species...... 78

Figure 4-8: The damage of the second emerging species when only spraying for the first species. On the x-axis is the coefficient of variation, the y-axis is the decay rate, and the panels represent the shift between the species...... 79

Figure 4-9: The number of sprays needed when the earlier emerging species has the action threshold at 40 and the later emerging species has a higher threshold that is +10, +20, and + 30 than the first species action threshold. The x-axis is the coefficient of variation and the y- axis is the decay rate...... 81

Figure 4-10: The total damage of the later emerging species based on the intervention of the earlier emerging species. On the left represents the scenario in which the earlier emerging species has a lower action threshold than the later emerging species...... 82

Figure 5-1: Annual reported cases of Lyme disease. By state from 2006–2017 (left) and by counties in Pennsylvania from 2006–2017 (right). Public data from the Centers for Disease Control and Prevention ...... 89

Figure 5-2: Distribution of the five most abundant tick species across Pennsylvania over time. Prevalence rates (tick counts per 100,000 population, left) represent tick abundance adjusted

by county population for time periods 1960–1969, 1990–1999, 2000–2009 and 2010–2018. Cumulative counts of ticks by species shown on the right ...... 98

Figure 5-3: Annual submissions of tick specimens by year. On the left is the annual sum of all tick counts (log-transformed) from 1900 to 2017. On the right are the proportional contributions of the five major tick species to the total tick counts (1900–2017). The grey shaded area represents periods where there were few or no tick submissions from the top five most abundant taxa...... 100

Figure 5-5:The seasonal distribution of D. variabilis, I. cookei and I. scapularis specimens by life stages from 1900 to 2017. The proportion was calculated by comparing the monthly abundance of each life stage (larvae, nymphs and adults) to the cumulative sum of all stages by species...... 102

Figure 5-6: Chord diagram representing associations between tick species and vertebrate hosts parasitized. Submissions (not counts) were used to quantify host association. We chose submissions over counts to avoid skews in abundance by hosts. The wider the chord, the more submissions exist for any given tick species-to-host...... 104

Figure 6-1: The wavelet modulus ratio (WMR) of the aggregate Yasuní (1067 species) and Cocha Cashu (654 species) community phenology (top panels) and time-series of the total species in traps (middle panels) and total estimated seeds in traps (bottom panels, natural log) in each period. In top panels, red indicates synchronous dynamics (high WMR) while blue indicates compensatory dynamics (low WMR). The black contour lines bound the points in time and scale (years) when the WMR was significant through phase-randomized bootstrap (n=1000). Here, nearly all significant regions are high WMR (yellow to red), with only two very small regions of significant low WMR (blue) in 2005 in Cocha Cashu at 0.1- 0.3 month scales. The cone of influence (white shading in the top panels) detonates the regions where the wavelet transforms are affected by the boundaries of the sampling period.124

Figure 6-2:The averaged wavelet modulus ratio of different plant families in Yasuní at the sub- annual (left) and interannual (right) scales. The number in parenthesis represents the number of species analyzed within the family. Colored points represent either significant synchronous (red) or compensatory (blue) dynamics at the time scale...... 126

Figure 6-3: The averaged wavelet modulus ratio of the families in the Cocha Cashu community at the sub-annual (left) and interannual (right) scales. The number in parenthesis represents the number of species analyzed within the family. Colored points represent either significant synchronous (red) or compensatory (blue) dynamics at the time scale...... 126

Figure 6-4: The averaged wavelet modulus ratio for the two dispersal groups that include all growth form in Yasuní. The number in parenthesis represents the number of species with the dispersal group. The ribbon represents the null distribution generated through bootstrapping

(n=1000). Any points that lie above the ribbon was considered significant and synchronous while any points below the ribbon indicated significant, compensatory dynamics ...... 127

Figure 6-5: The averaged wavelet modulus ratio for the different dispersal groups of tree species in Yasuní. The number in parenthesis represents the number of species with the dispersal group. The ribbon represents the null distribution generated through bootstrapping (n=1000). Any points that lie above the ribbon was considered significant and synchronous while any points below the ribbon indicated significant, compensatory dynamics...... 128

Figure 6-6: The averaged wavelet modulus ratio for all growth forms (animal, wind, ballistic) in Cocha Cashu. The number in parenthesis represents the number of species with the dispersal group. The ribbon represents the null distribution generated through bootstrapping (n=1000). Any points that lie above the ribbon was considered significant and synchronous while any points below the ribbon indicated significant, compensatory dynamics...... 129

Figure 6-7:The averaged wavelet modulus ratio for the five dispersal groups in tree species for Cocha Cashu. The number in parenthesis represents the number of species with the dispersal group. The ribbon represents the null-distribution generated through bootstrapping (n=1000). Any points that lie above the ribbon was considered significant and synchronous while any points below the ribbon indicated significant, compensatory dynamics...... 130

Figure 6-8: The Pearson correlation coefficient at varying scales (months) between the WMR and minimum temperature/precipitation for Yasuní. The grey lines represent the null distribution through phase-randomization permutation (n=1000). Any points outside the distribution is considered significant (colored in red)...... 131

Figure 6-9: The Pearson correlation coefficient at varying scales (months) between the WMR and minimum temperature/precipitation for Cocha Cashu. The grey lines represent the null distribution through phase-randomization permutation (n=1000). Any points outside the distribution is considered significant (colored in red)...... 132

xii

LIST OF TABLES

Table 2-1: Life‐history traits and base‐threshold temperatures for the five tortricid species..19

Table 2-2: The average day of year (DOY) when the accumulated catch of the annual spring flights reached the fifth, 25th, 50th, and 75th percentiles. The closest calendar dates corresponding to the day of the year are also listed...... 22

Table 5-1: The total submissions to the PSU Department of Entomology/Frost Entomological Museum from 1900 to 2017. Generic names that have been changed since the submission date are shown in parentheses...... 96

xiii

ACKNOWLEDGEMENTS

I would like to thank my adviser, Ottar Bjørnstad, for his support during my PhD. I will always be thankful for his mentorship and providing me the intellectual freedom to pursue whatever I wish. I would also like to thank my committee members: Jim Marden for giving me insight into what a good scientist is,

David Biddinger for sharing all the information he has on the fascinating biology of the tortricid moths,

Jesse Lasky for letting me dive into a wonderous time-series and providing insights into an extremely cool system, and Jessica Conway for her immense support and inspiring me to work harder in mathematical biology.

Other mentors I would like to thank is Kat Shea for providing me all her helpful advice and scientific insights. I am also grateful to Joyce Sakamoto for allowing me to collaborate on a wonderful dataset and always advising me on the importance of science communication. I would also like to thank William

Nelson who was always a great host when I visited in Queens University. Finally, I like to thank Aaron

Iverson who was my undergraduate mentor and I am still thankful for his advice.

I would like to thank my best friend, Rebecca Johnson for her friendship, food, and doing random, nonsensical things with me. I like to thank David Stupski for always being ready to talk science while getting Sheetz hotdogs. I would like to thank Spencer Carran for always being supportive with his advice and I appreciate our discussions on musicals and fermentation.

I like to thank my girlfriend Caitlin Lienkaemper, a brilliant mathematician and all amazing person. Her company made me barely notice the 2020 pandemic.

I would like to thank my friends that work with me in CIDD, Entomology, and Biology: Renuka, Allyson,

xiv

Nikki, Emily, Fhallon, Tyler, Thu, Edelio, Sarah, Makaylee, Mario, Lizz, Natalie, Zeinab, Lauren Q,

Hide, and way too many to list. I would also like to thank all the visiting scholars that came and let me take them to my favorite parts of State College (this again usually involved Sheetz hotdogs): Barbara,

Benno, Walter, and Fabienne.

Finally, I like to thank my family. I like to thank my father who says he’s proud of me even though he keeps saying that I could publish a lot more. I like to thank my mother for being incredibly supportive in everything I do without any judgement. I would like to thank my older sister, Somie, who is always ready to send me her love and support and packages of food.

This research has been graciously funded through the NSF Graduate Research Fellowship. Findings and conclusions do not necessarily reflect the view of the funding agency.

Chapter 1

Introduction

Phenology is the study of recurring biological events that occur in an organism’s lifecycle.

Events like the spring emergence of insects and flowering in plants are important to observe the passing of seasons and inform management decisions notably in agriculture and horticulture. There has been a call to investigate the mechanisms that influence biological timing and understand phenology’s role across multiple ecological scales (Forrest and Miller-Rushing 2010). For example, to explain the phenological patterns at the population-level, one must understand the underlying physiological mechanism that relate changes in the environment to individual responses. At higher ecological scales like at the community- level, the phenology of multiple, coinhabiting species provides insight on how timing is influenced by biotic interactions and how coexistence is possible through temporal niche partitioning. Finally, with growing concerns of climate change, understanding phenology across multiple scales will be crucial to predict ecological consequences and inform for effective management decisions as phenology shifts.

Herein, I discuss the phenological processes at the individual, population, and community level and show how the interactions between these scales is crucial for understanding biological timing.

Individual-level: understanding how environmental variables influence individuals

Development

Individual-level phenology focuses on how abiotic variables influence individual physiology like growth and development. For plants and ectothermic organisms, temperature plays a direct role in controlling individual’s metabolism and other critical functions (Régnière et al. 2012). As insects are critical in agriculture and can either be beneficial or damaging species, there is a rich body of empirical

2 work investigating the relationship between physiological rates and ambient temperature (Ratte 1985,

Gordo and Sanz 2005). Increasing temperatures have generally been linked to faster growth, increased fecundity, and higher survivorship among insects (Aghdam et al. 2009a, Zhou et al. 2010, Ma et al.

2017). Most importantly, development rate, the most important factor in determining the ‘speed’ of the individual’s life-cycle, was found to increase with rising temperatures (Jarosik et al. 2004, Logan et al.

2006). This direct relationship between ambient temperature and individual development rate serves as a fundamental building block to predict population-level phenology. By tracking accumulated thermal units or degree days, growers have been able to predict important life-cycle events with remarkable accuracy

(Yamanaka et al. 2012, Nelson et al. 2013, Johnson et al. 2016) .

Diapause

In many insect species, individuals must also time when they go into programmed dormancy or diapause where metabolic activities like development are paused. Only certain life stages go into diapause and in temperate latitudes, diapause is an adaptation to persist through freezing, winter temperatures. To enter and break dormancy at appropriate times, individuals mainly rely on changes in the day-length with temperatures being able to modify the response (Ashby and Singh 1990, Pumpuni et al. 1992, Steinberg et al. 1992, Powell et al. 2000). In most phenological models, diapause is excluded as there are many physiological aspects of this process that still need investigation and is not required for short term predictions (Hodek 1996) . However, the inclusion of diapause may be necessary for long-term predictions especially with warming winters. Like development rate, there can be changes to diapause responses due to warming temperatures such as delays in the diapause induction and the synchrony of how individuals terminate diapause (Neven et al. 2000, Stålhandske et al. 2015). Consequently, this could make species response to climate change more complicated.

Population-level: Phenological peaks and variance

Timing

Population-level phenology reflects the collective response of the individuals to abiotic and biotic influences. At this scale, much of the research is dedicated toward describing the position of the phenological distributions across time through quantifying the date of first, mean and peak abundance

(Forrest 2016). Because phenological peaks are easy to identify in time-series, both their timing and frequency are useful measures for understanding how variability in environmental variables influence phenology. For example, warmer temperatures can quicken the rate at which plants or insects become active in spring leading to earlier peaks while unfavorable condition can delay activity (Stefanescu et al.

2003). Consequently, an earlier start and prolongment of the growing season can then allow for additional generations which in turn can have important economic impact on pest management (Altermatt 2010,

Bjørnstad et al. 2016).

Variance

Another important population-level pattern is the phenological variance which reflects the degree of synchrony in the individual responses. This synchrony can reflect an evolutionary adaptation with strong selections pressure. For example, more asynchronous emergence may be a “bet-hedging” against unforeseen hazards such as early freezing (Calabrese and Fagan 2004, Kivelä et al. 2016, ten Brink et al.

2020). Additionally, pest genetic variability and subtle differences in microclimates can contribute to a distribution of different phenological responses. In the case of development rate, some individuals may develop faster than others and how much variation there is in development time is known as

‘developmental asynchrony’ (Gilbert and Raworth 1996, Bjørnstad et al. 2016). At the population-level, a

4 higher developmental asynchrony among individuals can contribute to ‘generation overlapping’ in multivoltine insects.

Developmental variability can vary across the season with some populations losing their synchrony with each additional generation. Ultimately, winter temperatures force the population to be resynchronized either by gating individuals from developing or through diapause induction (Gurney et al.

1992, Bjørnstad et al. 2016). Diapause termination also influences phenological variability as how synchronously individuals terminate diapause determines the initial synchrony of the population which would then have influence on the rest of the growing season. For many insect pests, successfully breaking diapause requires a minimum number of ‘chill-days’ or days when temperatures fall below a certain threshold. If the chill-day requirement is not met, there is a scattered emergence of individuals which suggests that climate change could lead to more complicated phenological responses than simply earlier emergence (J. R. Forrest, 2016; Neven, Ferguson, & Knight, 2000; Stålhandske, Lehmann, Pruisscher, &

Leimar, 2015).

Management

Of all the ecological-scales, population-level phenology is the most important for decision making in the management of natural resources most notably in agricultural processes. For example, pest- management in agriculture relies on monitoring the pest population to determine when vulnerable life- stages are present. With warming temperatures, there has been a heavy focus on predicting the additional number of generations per growing season as this can increase the number of interventions needed.

However, phenological variance can also influence the frequency of control efforts as species with more developmental asynchrony will have a wider distribution across time while a more synchronous population will have a narrower window of time for management (Bjørnstad et al. 2016). Therefore, a full

5 understanding of population-level process will investigate both the timing and variance of critical life stages will help inform better management.

Community-level: coexistence and multi-species management

Varying responses

Finally, community-level phenology provides important insight into how shared environmental variables influence community interactions and composition. The diversity in biological timing among coinhabiting species suggests reliance on different shared abiotic cues. (Forrest and Miller-Rushing

2010). For example, in the case of diapause termination, certain species may be more sensitive to increasing day-length than to rising temperatures due to their unique life-histories (ex: diapausing stage).

Therefore, coinhabiting species may show different phenological responses making prediction of ecological consequences under climate change challenging. Previous research across many systems have already shown the remarkable diversity in how species have shifted in their phenology in response to climate change (Navarro-Cano et al. 2015) . Changes in species’ phenology then affect ecological interactions (ex: herbivores and plants, parasites and hosts, etc.) and great enough shifts can possibly lead to phenological mismatch (Nakazawa and Doi 2012, Ovaskainen et al. 2013a). Therefore, being able to predict community-level phenology requires understanding the mechanism at the lower ecological scales.

Biotic interactions

Community-level phenology is also crucial for understanding how niche partitioning through time can promote the coexistence of certain species. Most of the work has been focused on plant communities and the timing of life-cycle events such as leafing bud bursts, flowering, and fruiting (Lasky

6 et al. 2016). Community phenology can show synchronous or asynchronous dynamics depending on the seasonality of both abiotic variables and biotic interactions. In the neotropical rainforests, seasonality is defined more by precipitation patterns than by temperature and rainy seasons were found to synchronize reproduction across multiple species though some species may show asynchronous dynamics due to unique adaptations that allow them to reproduce in unfavorable conditions.

Most interestingly, biotic interactions can shape phenology such as through competition. Shared pollinators and seed dispersers can select for more asynchronous phenology to prevent overlapping

(Wheelwright 1985, Elzinga et al. 2007). However, it is possible that synchronous phenology can be selected for if there is positive-density dependence with synchronous reproduction being helpful to satiate seed predators (ex: seed masting) (Kelly et al., 2000; Silvertown, 1980). Finally, it is important to note that that all species have an evolutionary history and through phylogenetic niche conservatism, closely related species may show synchrony through shared ecological traits. Disentangling the different mechanisms influencing community phenology requires understanding a species role in the context of other species.

Community management

Finally, an understanding of community-level phenology can inform better management decisions. In infectious disease cycles, there are diverse communities of ectoparasites such as ticks, that can share hosts and pathogens. Tick phenology can be unique due to the main hosts that each species relies upon, but all are faced with similar abiotic restrictions that come with changing seasons. Exploring phenology at the community-level is informative for public health as the presence of certain tick species can fluctuate throughout the season. Knowing the likely species present at various time during the year

7 would allow for quicker identification of the species and for better diagnosis of the possible pathogens being transmitted.

Community-level phenology can also provide us a framework for a multi-species approach in the management of natural resource. For example, in pest management, species phenology is carefully monitored to target the vulnerable life-stages. If one considers that control efforts like pesticide may inadvertently affect other non-targeted pest species, then it may be possible to use management approaches that target multiple species by accounting for multiple timing. Thus, it may be possible to manage multiple species with less control effort or interventions.

Objectives

The ultimate research philosophy explored in this dissertation is that phenology is a process that unifies ecological scales. As all organisms must exist in fluctuating environments, biological timing is ubiquitous and affects all biotic interactions. Even among diverse species, there are commonalities in that the environment influences individual physiology which in turn has implications across higher ecological levels. By exploring multiple species in phenological research, I demonstrate that each species’ unique mechanisms are a testament to how all species must synchronize their life cycles to the changing environment. Therefore, the following studies focus on a variety of species including a guild of tortricid moths that are pests of stone and pome fruits, the tick species of Pennsylvania, and the neotropical plant communities of Yasuní and Cocha Cashu. Though closer examination of these species I show the importance for phenological research across ecological scales.

The first three chapters in the dissertation investigate the biological timing of tortricid moths and how phenology can guide management. These five coinhabiting tortricid moth species are important pests

8 of apple orchards and show considerable diversity in life-histories. Because of their differences in phenology, I was interested in the relationship between climate variables that influence spring emergence.

In Chapter 2, I investigate the relationship between phenology and the local temperatures as well as regional climate through the North Atlantic Oscillation and the Arctic Oscillation. These large-scale oscillations control most of the climate variability in the northern hemisphere and provides a holistic measure of climate. By using partial-least squares regression, we found the specific time periods when both the local and regional climate were influential for each species’ phenology. Thus, the unique phenology that we see among the coinhabiting tortricid moths was due to each species relying on climate variables at different periods of the year. Our most important finding was that autumn and winter conditions have a significant impact on spring phenology.

In Chapter 3, I then focused on one of the tortricid species, Cydia pomonella or the codling moth, to create a population-level phenology model. Due to its important pest status, predictive models are necessary to schedule interventions. Because we demonstrated that in Chapter 2 that autumn and winter processes are important for determining spring phenology in the previous chapter, I included diapause induction and termination in my C. pomonella model. To capture the field dynamics of a 33-year time series data, I collected previously published empirical data showing the relationship between temperature and individual vital rates (see Appendix B). The physiological stage-structured model incorporated distributed delay differential equations to show how developmental asynchrony is crucial for capturing the population-level processes which can have important management consequences. Finally, we showcase the benefits of the model as being able to predict long-term dynamics under changing climate.

In Chapter 4, I then investigated how different species phenology can influence the feasibility of community-level management. Specifically, I used pest management to explore how similar the species’

9 phenology must be for a single application of pesticide to control both pest species. As in Chapter 3, I use distributed delay equations with a focus on developmental variability due to its importance in management. Specifically, a wider distribution of individuals over time would require more sprays than a pest species with a narrower distribution. By varying the different timing of sprays and the overlap between the species, I found that future approaches to multispecies management should focus on developmental variability, the shift between the species timing, and the differences in economic importance.

Chapter 5 and Chapter 6 are collaborative projects that show the importance of long-term time series for phenological research at the community level. As with Chapter 2, I focused on the seasonality of the different species and their varying phenologies. Chapter 5 explores geographical range and seasonality of Pennsylvanian tick communities across 117 years of passive surveillance data. Chapter 6 investigates the community-level dynamics of the plant species in the neotropical rainforests of Yasuní and Coshu Cashu. In Chapter 5, we found that tick species and saw that their differences in phenology between tick species may be due to varying sensitivity to abiotic cues but also may be due to shifts in host differences availability. The long-term data sets also show signs of a shift in tick species dominance over time which then can have important public health implications. While the changes in tick species composition is due to increased suburbanization, our findings suggest that changes to the environment like climate change can also alter the structure of a community. Due to ticks being important vectors of pathogens for both humans and livestock, this has serious implications for public health and econoics.

Finally, in Chapter 6, we investigated the reproductive phenology of a hyper-diverse ecological community with hundreds of different plant species. Using multivariate wavelet-analysis, we investigated if there was synchronous/asynchronous timing among groups that are related or share similar dispersers.

Synchronous phenology among species suggests the influence of shared abiotic cues such as precipitation.

However, species can show anti-synchronous phenology especially among closely related species way to temporally partition their niche and allow for coexistence. The chapter cements the importance of community-level phenology in understanding the role of abiotic and biotic interactions in shaping phenology.

With this dissertation I hope to leave readers with a new insight into biological timing through exploring phenology across many different ecological scales and systems. By investigating the unique mechanisms in phenology, one can make generalities about other species and systems. For example, the community-level phenology of plants in a neotropical rainforest show how evolutionary legacies, shared abiotic cues, and biotic interactions can shape timing. These findings can easily be applied to other systems like the tortricid pest species where they share similar seasonal hazards and reliance on similar host plants. An important role of an ecologist is to appreciate the immense diversity of living organisms but at the same time be able to find the commonalities among ecological processes.

For future work, I would like to better understand the consequences of climate change on phenology. While there are some general predictions made of insect populations under warmer temperatures, species will show their own phenological responses. I am specifically interested in the role of dormancy and diapause with climate change as winter temperatures are crucial for resynchronizing the population. While I explored the possible consequences of warming temperatures on diapause termination in Chapter 3, I would like to further explore how different diapauses (ex: species requiring chill-days versus those that do not) might influence overall phenology. A warmer spring might make it harder for species to accumulate the required chill-days to terminate diapause such that their phenology is not earlier as expected. Therefore, it is possible that there are counterintuitive phenological responses to climate change.

Additionally, I would also like to further work on bridging phenological processes between the population and the community level. Specifically, as phenology changes with climate change, biotic interactions may change and lead to a phenological mismatch. I have shown how a pest species may change under a warmer temperature in Chapter 3 but without investigating how the phenology of the host plant would shift as well, we do not have a full understanding of possible management strategies.

Therefore, my future work would be dedicated to thinking of phenological changes of a species in relations to its resources, predators, etc.

I would like to conclude that the importance of phenology lies not only in the field of ecology but as a crucial field of science that may help to bridge the gap between scientific community and the general public. Humans, like any biological organism, have an inherent understanding of seasonal changes and changes to climate have been observed and documented across generational memory. With the threat of climate change, I believe that phenological research would help the public better understand and quantify the extent to which our world is changing.

Local and regional climate variables driving spring phenology of tortricid pests: a 36-year study

Abstract

Insect phenology is driven by local climate variables, most notably temperature. Increased warming has been linked to advancements in critical phenophases such as the spring flight of reproductive adults in the mid‐Atlantic region of the U.S.A. Local climate is governed by the fluctuations of large‐scale climate oscillations. In the northern hemisphere, both the North Atlantic Oscillation (NAO) and the Arctic Oscillation (AO) control the local autumn and winter severity. Low NAO and AO indices are associated with colder autumns and winters, which can delay spring phenology. In this study, 36 years of data from experimental fruit orchards in Biglerville, Pennsylvania, were used to run partial least‐ squares regressions in order to determine the climate variables related to the spring phenology of five tortricid pest species. The phenology of the tortricid pests did not advance, even though there was evidence of warming at the research site. Spring temperatures were found to be the most significant climate variables in determining the timing of the spring flights. However, autumn–winter temperatures were also important. For the NAO and the AO, it was found that these oscillations affected the tortricid moths by influencing autumn–winter conditions. The oscillations of the NAO and AO can obscure long‐ term changes in phenology. These findings suggest that the inclusion of large‐scale climate oscillations can provide important insights into how climate conditions can influence insect phenology and presents an opportunity for improving the ability to forecast spring emergence.

Note: This chapter was published in the Ecological Entomology (2019)*- Damie Pak, David Biddinger, and Ottar Bjørnstad. DP conceived the idea, performed the statistical analyses, created the figures/tables,

13 and wrote the manuscript with support from the co-authors. DB provided the original data, wrote the

‘study site’ subsection, and helped with the critical revisions. OB assisted with the project conception and helped with critical revision.

*Pak, Damie, David Biddinger, and Ottar N. Bjørnstad. "Local and regional climate variables driving spring phenology of tortricid pests: a 36 year study." Ecological Entomology 44.3 (2019): 367-379.

Introduction

Phenology, the timing of recurring biological events, has garnered more attention with growing concerns of climate change. A large proportion of research has been dedicated to insects due to their reliance on abiotic factors, most importantly temperature, to synchronize their life cycles with the seasonal environment (Danks 1978a, Powell and Logan 2005). Insects develop faster with rising temperatures, and warmer springs have accelerated spring emergence across multiple insect groups (Roy and Sparks 2000, Forister and Shapiro 2003, Gordo and Sanz 2005). Less studied are the effects of autumn and winter conditions on the spring phenology of insects. Of the few studies to examine these effects, some have reported a significant relationship between milder winters and delayed emergence

(Bale and Hayward 2010a, Stålhandske et al. 2015, Forrest 2016, Lehmann et al. 2017). Local climate conditions change considerably from year to year and much of their variability is due to the changes in the large‐scale climate oscillations. Therefore, investigating how large‐scale climate cycles interact with local weather conditions to influence spring emergence is necessary to predict future phenological changes.

In the northern hemisphere, the North Atlantic Oscillation (NAO) and the Arctic Oscillation (AO) are the dominant cycles governing much of the local weather phenomena. The NAO is described as the oscillation of atmospheric mass between the Azores Islands and Iceland, which determines the strength of

14 the westerlies and the position of storm tracks (Hurrell 1995). The AO is associated with changes in the pressure systems between the Arctic and mid‐latitudes, known as the polar vortex, where the cold Arctic air is either confined to the Arctic or allowed to move to the equator (Kryzhov and Gorelits 2015).

Lacking any dominant periodicities, both the NAO and the AO fluctuate in magnitude and direction as they oscillate between negative and positive phases at the monthly to the decadal scale. The effect of the

NAO and AO on local climates is strongest in the months between November and April (Stenseth et al.

2002). When the NAO and AO are in their negative phases, the eastern U.S.A. faces a colder and drier winter due to the weakened westerlies that allow for the cold arctic air to reach more southerly latitudes.

In the positive phase, however, stronger westerlies confine the Arctic air to the more northerly latitudes, which then leads to a wetter and warmer winter (Hurrell 1995). The NAO and AO indices are holistic measures that influence not only local temperatures but also precipitation, snow cover, wind speed, and other meteorological variables. In some cases, large‐scale climate indices were found to be better predictors of ecological processes than a single climate measure and they have been useful in identifying periods when climatic conditions greatly influence phenological events (Ottersen et al. 2001, Stenseth et al. 2002)

The NAO and the AO have been shown to influence the spring phenology of various insects through their effects on the local weather. Positive winter NAO and AO indices have been linked to milder and warmer conditions, leading to an earlier start of the growing season and accelerated growth and development in insects (Ottersen et al. 2001, Briers et al. 2004, Schaefer et al. 2005, Cook et al.

2005). In the U.K., an increase in the NAO winter indices was linked to earlier spring flight dates for multiple butterfly species and earlier migrations in the green spruce aphids (Westgarth-Smith et al. 2007,

2012). Although warm winters can speed up development, the opposite can also be true. To avoid adverse periods like winter, some insect species go into diapause, a state of reduced metabolic activities and suspended development. The duration of diapause can be enhanced by winter severity, which varies year‐

15 to‐year due to the NAO and AO (Stålhandske et al. 2015, Forrest 2016). As diapause must be terminated before development can resume, diapause length could then have subsequent effects on the timing of spring flights. Some diapausing insect species have a chilling requirement where a minimum number of cold‐weather days are required to terminate diapause. Previous work on the orange tip butterfly has found a negative correlation between cold‐weather days and post‐diapause development, suggesting that warmer winters can sometimes negate the effects of milder spring temperatures on insect phenology (Stålhandske et al. 2015)

In our study, we investigated the spring phenology of five moth species (Family: Tortricidae):

Cydia pomonella/codling moth (L.), Choristoneura rosaceana/oblique banded leafroller (Harris),

Platynota idaeusalis/tufted apple‐bud moth (Walker), Grapholita molesta/oriental fruit moth (Busck), and

Argyrotaenia velutinana/red banded leafroller (Walker). These are important pests of pome (apples, pears, etc.) and stone (peach, plum, etc.) fruits across the U.S.A. Observing spring phenology is critical for the effective management of these pests, as their spring emergence signals to growers that they should begin counting thermal accumulation or ‘degree days’ to predict the appearance of vulnerable life stages and to time interventions properly. Therefore, our goal was to investigate the periods during the year when climatic variables strongly influence the spring emergence of the tortricid pests. We used NAO and

AO indices to investigate how autumn and winter severity can also be related to spring phenology. We obtained long‐term trapping data spanning more than three decades (1981–2016) of monitoring from research orchards at Pennsylvania State University's Fruit Research and Extension Center (FREC). We first investigated whether there were trends in the spring flight across 36 years from 1981 to 2016 and then aimed to find periods during the year when the climatic variables strongly influence the spring emergence.

Materials and Methods

Study Site

We obtained time series data of adult male moths captured in pheromone traps from 1981 to 2016 from FREC. The FREC consists of a 140‐acre research centre located in south central Pennsylvania near the city of Biglerville (latitude 39.936°N, longitude −77.25°W) and the research orchards from

Arendtsville, located 4.8 km away (latitude 39.923°N, longitude −77.29°W). These study sites exhibit typical temperate climate with four distinct seasons, including hot, humid summers and winters with average annual snowfall of 68.58 cm.

The five totricid moth species were monitored in multiple mature blocks of apples and peaches of fruit‐bearing age, ranging from 2 to 4 ha in size. Trapping occurred in multiple blocks and with different pheromone lures and traps as the technology evolved, but a minimum of six to eight traps of each species were monitored weekly over the 36‐year study. Until about 2006, Pherocon 1C Wing traps (Adair,

Oklahoma) were used for monitoring the various tortricid pests with specific sex pheromone rubber lures that were placed within. After that time, the use of longer‐lasting and more weather‐proof Pherocon 6 traps made of corrugated plastic shelters were used. Over the 36 years, the load rates and types of pheromone dispensers/lures varied to some extent, but the specific pheromone components for each tortricid species was discovered at least a decade before this study began and thus did not change significantly. As pest control phenology models use first sustained adult flight as a biofix to begin the

‘clock’ for predicting pest control spray timing, traps were placed in the orchards several weeks in advance of the first anticipated moth flight for each species. In general, one trap for each individual species was used for every hectare of orchard.

Although insecticides were used at the FREC, we assumed that all analyses on adult tortricid pests in this study were not confounded by agroecosystem inputs. For research evaluations of the insecticides at the research site, single tree plots with unsprayed buffer trees were maintained around the replicated treatments, as were unsprayed control trees, so that up to 50% of all the trees in each block were not sprayed with insectides at all. It is important to note that some experimental insecticides were not very effective and tortricid pest pressure was much higher than normally found in commercial orchards. Most of the insecticide applications that would affect tortricid pests are applied after bloom, by which time initial adult flight for three of the five species had already started. Thus, excluding abandoned orchards, the experimental study site at FREC might be the closest to an unmanaged orchard.

The types of pesticides evolved greatly over this long trapping period, with the earlier years consisting mostly of organophosphate and carbamate insecticides in the 1980s and 1990s, giving way to various types of insect growth regulators and neonicotinoids in the late 1990s. Synthetic pyrethroid insecticides were never recommended in Pennsylvania apple orchards due to the disruption of biological mite, scale and aphid control in that crop, but were and are still used extensively in the less integrated pest management‐intensive peach crop. Most organophosphate and carbamate insecticides were lost for use in fruit orchards around 2003–2005 due to restrictions by the Food Quality and Protection Act and replaced with newer neonicotinoid insect growth regulators, diamide and macrocylic lactone insecticides, which are currently in use.

Climate data: local temperatures and regional indices

We retrieved the daily and monthly temperature data (minimum, maximum, and average) from the local weather station at the FREC (Biglerville). To investigate if there was any significant warming at the research site, we ran a series of linear regressions of the monthly temperatures between September and

April against years. For the NAO and AO data, we obtained the monthly and daily indices from the National

Weather Service Climate Prediction. The NAO index was calculated as the normalized anomaly of the sea‐ level pressure between the Azores and Iceland, whereas the AO index was based on any anomalies poleward of 20°N. To investigate if there was a relationship between the large‐scale climate oscillations and local temperatures, we ran a Pearson correlation between the monthly temperatures (minimum, maximum, and average) and the corresponding monthly NAO and AO indices between September and April.

Tortricid species

For the tortricid moths in Mid‐Atlantic fruit farms, the primary hosts are apples and peaches

(although C. pomonella does not normally attack peach), with larvae feeding on the shoots and fruits.

From past literature and observations made at the FREC, we listed the possible life‐history traits that could influence the timing of spring flight in Table 1.

† The upper‐base threshold for A. velutinana was unavailable so we assumed that it has a similar upper‐base threshold as the other leafrollers (C. rosaceana and P. idaeusalis)

Table 2-1: Life‐history traits and base‐threshold temperatures for the five tortricid species

Of the five tortricid species, C. pomonella, C. rosaceana, and P. idaeusalis are usually bivoltine, though a favorable growing season allows for a partial third generation. The multivoltine species include

G. molesta which has three to five generations annually and A. velutinana which generally has three to four generations in Pennsylvania. Previous work done at FREC found that G. molesta's survivorship, development rate, and spring phenology were affected by the type of host plant (Myers et al. 2007).

Therefore, we split the G. molesta data based on the two host types that the species primarily consumed: peach (1999–2016) and apples (1981–2016).

In terms of overwintering strategies, all five species are induced into diapause by shortening day lengths and decreasing temperatures. At the study site, the tortricid species were found to diapause as immature larvae (second to fourth instars), mature larvae (fourth to fifth instars), or pupae. Cydia pomonella, C. rosaceana, and G. molesta, are known to require a chilling period to break diapause

(Neven et al. 2000, Omeg 2001). It is not known if A. velutinana has a chill‐day requirement and P. idaeusalis is the only species known not to require chilling (Rock and Shaffer 1983). In terms of spring phenology, A. velutinana overwinters as a pupa and initiates its spring flight just as green tissues are showing on the apples. Both C. pomonella and G. molesta overwinter as mature larvae, with G. molesta emerging at the start of apple bloom and C. pomonella emerging during apple bloom. C.rosaceana and P. idaeusalis diapause as immature larvae and overwinter in the trees and in the orchard ground cover, respectively. Both C. rosaceana and P. idaeusalis then emerge 3–4 weeks after bloom.

For our spring flight analyses, we excluded years in which the pheromone traps were not set out early enough to capture the peak emergence of the tortricid moths. For C. rosaceana we excluded data from 1981, 1991, 2005, and 2016. For P. idaeusalis, we excluded 1981, 2002, 2007, and 2008. Finally, we excluded 1985, 1986, 1987, 1989, 1990, and 2016 for A. velutinana.

Growing degree‐day model to calculate the spring flight index

As insect development is driven by temperature, we used a growing degree‐day model that incorporates the effects of daily temperature on the insect development rate. The simplest degree‐day model was calculated by subtracting the species' lower‐base threshold, the minimum temperature at which development can occur, from the average daily temperature (Table 1). The upper‐base threshold is the point at which the development rate does not increase due to physiological stress. If the average daily temperature exceeded the upper limit, we substituted it with the species' upper‐base threshold. We then accumulated the resulting degree days from the start of each year (1 January). All base thresholds were taken from degree‐day models used at the FREC that were based on the PETE models developed in the mid‐1970s (Pennsylvania Tree Fruit Production Guide, 2016).

Due to the bias that could arise from observational uncertainty with regard to the absolute first appearance of the tortricid moths, we calculated the cumulative degree days up to when the accumulated catch of the annual spring flights reached the 25th percentile and defined it as the spring flight index. In calculating the annual spring flight index, we partitioned each species' data to capture only the first‐ generation spring flight, which would be visualized as the distinct first peak in an annual time series. For example, we only looked at the catch data within the first 1000-degree days of each year for C. rosaceana. We then fitted an exponential model (Equation 1) with the proportion of moths captured (푃푚) against the accumulated degree days to estimate the accumulated degree day corresponding to the 25th percentile (Brown and Mayer 1988). The parameters r and b were the rate of increase and lag, respectively, and were estimated using weighted non‐linear least‐squares estimates.

푃푚 = exp (− exp(−푟퐷퐷 + 푏) (Equation 1)

Here, DD refers to degree days. For ease of interpretation, we subsequently converted the accumulated degree days into day of year for all further analyses. Although only the 25th percentile was used for this study, we also calculated the average time (as day of year) at which the fifth, 50th and 75th percentiles of the accumulated catch were reached (Table 2).

Table 2-2: The average day of year (DOY) when the accumulated catch of the annual spring

flights reached the fifth, 25th, 50th, and 75th percentiles. The closest calendar dates

corresponding to the day of the year are also listed.

To see if there were any significant shifts in the spring flight indices from 1981 to 2016, we calculated the annual departures for each species. These were calculated by subtracting the total average of the annual spring flight indices from each annual spring flight index. A positive annual departure indicates a later emergence compared with the total average, whereas a negative departure indicates an earlier emergence. We then ran a linear regression with the species' annual departures against years to look for any significant trends. In addition to the linear regression analysis, we used the non‐parametric

Mann–Kendall test to detect if there was a monotonic upward or downward trend in the annual departures.

Finally, we investigated whether the five species showed similar phenological responses across

1981 to 2016. We investigated the different pairwise correlations between the species' annual departures.

The correlations would indicate if the two species showed similar patterns of delays and advancements in their spring phenology. For this analysis, we only looked at G. molesta (apple), due to the lower number of years in the G. molesta (peach) time series data.

Partial least‐squares regression

Partial least‐squares regression (PLS) was used to investigate the ‘critical periods’ during which the local and large‐scale climate variables have a significant influence on the spring phenology. Similar to a principal component analysis, the PLS uses a linear combination of predictor variables to maximize the variance explained in the dependent variable (Wold 1985) Partial least‐squares regression has previously been used in other phenological studies, as it is useful in situations where there are more predictor variables than phenological observations (Luedeling and Gassner 2012, Luedeling et al. 2013). Partial least‐squares regression also accounts for the strong collinearity and autocorrelation that would occur in the predictors, such as in climate data (Carrascal et al. 2009). It is important, however, to consider the statistical importance of each predictor in the context of insect biology. For example, there may be short time periods when the predictors are statistically significant, but these may not be biologically relevant.

Therefore, we recommend a broad interpretation of the PLS results.

We ran three separate PLS regressions for the minimum daily temperatures, NAO, and AO indices between 1 September and 30 April. The predictors were smoothed using a running average of 11 days. Because G. molesta (apple and peach) and A. velutinana emerge in April and we were only interested in the predictors preceding their spring flights, we excluded April predictors for these species.

To account for leap years, we removed the daily temperatures and the large‐scale climate indices of 29

February. The important outputs of the PLS include the variable importance in projection (VIP) scores, which estimate the importance of each predictor in explaining the variance in the dependent variable.

Predictors with larger VIP scores are more relevant to the model and we selected for predictor variables with a VIP > 0.8 (Wold 1985). After variable selection with the VIP scores, the regression coefficients then quantify the strength and the direction of each predictor's effect on the full model. Predictors with positive regression coefficients indicate that, at this period, higher minimum temperatures or climate indices would delay the spring flight. Predictors with negative regression coefficients indicate an opposite effect, with lower minimum temperature or indices linked to later emergence. Finally, we wanted to investigate whether our PLS results are valid using a simpler method. We ran the Pearson correlation between the local monthly temperatures (minimum, maximum, and average) and the spring flight indices of each species to compare these with the PLS outputs.

All statistical analyses were done with R, version 3.3.2 (R Core Team, 2016) with the Kendall package for the Mann–Kendall test (McLeod 2011) and the ‘pls’ package for the partial‐least squares regression (Mevik and Wehrens 2007).

Results

Meteorological data

From 1980 to 2016, we found a significant increase (P < 0.05) in the minimum temperatures between September and December (September, 퐹1,34 = 36.05, 푅2 = 0.50, P < 0.001; October, 퐹1,34 =

14.60, 푅2 = 0.28, P < 0.001; November, 퐹1,34 = 6.48, 푅2 = 0.14, P = 0.02; December, 퐹1,34= 9.4, 푅2 =

0.19, P = 0.004) and in April (퐹1,34 = 10.82, 푅2 = 0.22, P < 0.01) (Figure 1). However, there was considerable interannual variation in both the minimum and maximum monthly temperatures across 36 years. For the autumn and winter months, this variability could be related to the fluctuations of the NAO and AO. The Pearson correlations between the climate indices and the monthly temperatures were significant and negative for months between October and January) (Appendix A: Tables S1, S2). The strongest association was with the climate indices and December maximum temperatures (r = 0.58, P <

0.001 and r = 0.54, P < 0.001 for NAO and AO, respectively). The negative coefficients indicate that decreases in the NAO and AO indices were linked to lower temperatures.

Annual spring flight index of the tortricid species

Across all species, we found no significant upward or downward trend in the timing of spring flight [C. pomonella, F1,31 = 1.975, R2 = 0.03, P = 0.17; C. rosaceana, F1,30 = 1.657, R2 = 0.02, P = 0.21; P. idaeusalis, F1,30 = 0.717, R2 = −0.009, P = 0.40; G. molesta (Apple), F1,34 = 0.892, R2 = −0.003, P = 0.35;

G. molesta (peach), F1,16 = 0.010, R2 = −0.06, P = 0.92; A. velutinana, F1,28 = 0.88, R2 = −0.003, P = 0.35]

(Figure 2). Instead, there was an oscillatory pattern of varying delays and advancements in the annual emergence relative to the average for the whole period. The annual departures in the spring flight indices varied in both strength and direction among the five species. The more negative the departure, the earlier the spring flight was compared with the average date of emergence, and vice versa.

Figure 2-2: The annual spring flight indices for five tortricid species across 1981–2016.

For Cydia pomonella and Grapholita molesta (peach), the trapping started in 1984 and 1999, respectively.

Missing points represent years which were excluded from the analyses. Linear regressions were performed between the spring flight indices and years.

The Pearson correlations in the annual departures between pairs of species were significant and positive, showing that all the tortricid species had similar phenological responses across 36 years (Figure

3). The strongest correlation was between the annual departures of C. rosaceana and P. idaeusalis (r =

0.78, P < 0.001) and between those of C. rosaceana and G. molesta (r = 0.78, P < 0.001).

Partial least‐squares regression

All five tortricid species showed negative regression coefficients for autumn and spring temperatures, suggesting that colder conditions delayed spring emergence (Figure 4 and Figure 5).

Specifically, the minimum temperatures between March and May and between November and December were important periods influencing their phenology (Appendix A: Figure S1, S2). For G. molesta (apple and peach) and A. velutinana, we also found significant negative regression coefficients for the minimum temperatures between January and March. For C. pomonella and G. molesta (peach), their PLS regression showed positive regression coefficients for minimum temperatures between September and October and also for January temperatures, meaning that warmer temperatures in these periods were associated with delayed spring phenology.

The colored bars represent predictor variables that were significant with variable importance in projection

(VIP) scores > 0.8. Negative regression coefficients (blue) mean that low values of the predictor variable were associated with later emergence. The positive regressions coefficients indicate that an increase in the predictor values led to later emergences.

NAO and AO indices. The colored bars represent predictor variables that were significant with VIP scores

> 0.8. Negative regression coefficients (blue) mean that low values of the predictor variable was associated with later emergence. The positive regressions coefficients indicate that an increase in the predictor values led to later emergences.

The Pearson correlations between the monthly temperatures and the spring flight indices corroborated the PLS results, with colder monthly temperatures delaying spring flight (Figure 6). Except for C. pomonella, the spring phenology of the tortricid species was again negatively associated with both

March and November temperatures. Like the results of the PLS regression, we also found significant negative correlations between the spring flight indices and January–February temperatures for G. molesta

(apple) (January, r = −0.47, P = 0.003; February, r = −0.47, P = 0.003), G. molesta (peach) (January, r =

−0.48, P = 0.043; February, r = −0.59, P = 0.008) and A. velutinana (January, r = −0.48, P = 0.006;

February, r = −0.47, P = 0.008). The only contradictions between the PLS regression and the Pearson correlations were the influence of September temperatures on the spring flight of G. molesta (peach), and of January temperatures on the spring flight of C. pomonella. The positive PLS regression coefficients indicated that higher September temperatures delayed the flight of G. molesta (peach), whereas the negative Pearson correlation coefficient indicated the opposite effect (r = −0.40, P = 0.09). For C. pomonella, we found January temperatures to be significant in the PLS regression, but they were not significant in the Pearson correlation (r = −0.009, P > 0.10). These contradictions were most probably due to differences in statistical analyses, as the PLS considers other predictors when trying to best explain the variance in the dependent variable, unlike a simple Pearson correlation.

2016. The colors were based on the absolute values of the Pearson correlation coefficients. Negative coefficients indicate that decreasing temperatures were linked to later spring emergences. The positive coefficients indicate that increasing temperatures were linked to later spring emergences.

In the mid‐Atlantic, the PLS regression revealed that the association of NAO and AO with local conditions was strongest between September and February (Figure 4; Figure 5). Both climate oscillations were significant at similar periods and shared the same regression coefficient signs (positive or negative).

In the autumn, between September and December, the NAO and AO predictors generally had negative regression coefficients. The negative coefficients suggested an association between lower climate indices and delayed spring phenology. However, between January and March, we found the opposite effect, with the winter NAO and AO indices having positive regression coefficients in the model. This indicated that, in winter, higher NAO and AO climate indices were related to later emergence.

Discussion

From 1981 to 2016, we found no significant advancement in the spring phenology of the tortricid species, even though we found warming at FREC. Our findings are not uncommon, as previous works found multiple insect species that exhibited no long‐term changes in their emergence even with significant warming at the study site (Roy & Sparks, 2000; Marcel et al., 2003; Both et al., 2004; Forrest,

2016). Instead, there was considerable year‐to‐year variability in the timing of spring flight, with all species showing similar delays and advancements in their phenology. This suggests that the tortricid species are influenced by climatic variables at similar times to capitalize on plentiful resources and to avoid dangers like freezing. We found that autumn and winter severity was an important predictor of spring phenology and that large‐scale climate oscillations such as the NAO and AO can potentially obscure the effects of long‐term climate change on phenology.

Spring temperatures, most notably in March, were the most significant variables when it came to predicting spring flight, with colder temperatures correlating with a delay in emergence. In Pennsylvania,

March is generally when daily temperatures first rise above the minimum threshold required for insect

36 development to occur. Warmer spring temperature can then shorten the time for juveniles to complete their development before the adult flight (Ratte 1985, Roy and Sparks 2000, Stefanescu et al. 2003). In addition to colder spring conditions, we found that colder autumn and winter temperatures were also generally correlated with delayed emergence (Figure 4, Figure 5). Specifically, the autumn temperatures between September and December contributed significantly to the year‐to‐year variability in spring flight for all species. The influence of winter temperatures between January and February was strongest for the spring phenology of G. molesta (apple and peach) and A. velutinana, with colder temperatures delaying their flight. Interestingly, these species are both multivoltine and emerge earlier than the other species in the study, all of which are bivoltine. Therefore, it is possible that the multivoltine species are sensitive to early‐season temperature cues to prevent mistiming their emergence (Pau et al. 2011). This suggests that life‐history variation plays a crucial role in determining the significant periods during the year when climate variables influence phenology.

The negative PLS regression coefficients in both the autumn and winter periods suggest that harsher conditions can delay spring emergence by influencing diapause and post‐diapause processes. At the FREC, there have been anecdotal sightings of active tortricid larvae during uncharacteristically warm winters, indicating that during warm autumns and winters, development can continue and potentially lead to earlier spring flight. From previous work, warmer temperatures during autumn and winter were found to delay the onset of dormancy in some insect species. This resulted in less intensive diapause, with fewer thermal accumulations needed for termination (Denlinger and Iv 1981, McWatters and Saunders 1998,

Fantinou and Kagkou 2000, Bale and Hayward 2010b). Alternatively, colder temperatures during the onset of diapause can enhance the diapause response and possibly extend its length (Collier and Finch

1983, Tauber et al. 1986, Bale and Hayward 2010b, Chen et al. 2014a). For insects in diapause, warmer temperatures can increase the insect's metabolic rates and lead to depletion of nutrient reserves and early termination of diapause (Han and Bauce 1998).

Chilling requirement further complicates our understanding of how autumn and winter conditions influence spring phenology. In our study, three of the five tortricid species are known to require chilling in order to terminate diapause and, therefore, we predicted that warmer winters would delay the onset of spring flight (Danks 1978b, Han and Bauce 1998) Although we were not able to conclusively support our prediction, from the PLS regression, we found that warmer September temperatures delayed emergence for two species with known chilling requirements: C. pomonella and G. molesta (peach) (Figure 4a,b).

For C. pomonella, we also found January temperatures with positive regression coefficients, suggesting that warmer temperatures in this period delay spring emergence. However, it is unclear whether these findings were related to chilling requirements. It is also possible that a lack of chill days could affect the synchrony or distribution of adult emergence and not the actual timing (Forrest 2016) Although we do not know the full physiological mechanism, our analyses highlight the consequences of autumn and winter severity on the timing of spring flights and should thus help to guide future studies of diapause triggers.

For the NAO and AO, we observed two distinct periods in which they exerted the strongest influence on local weather. Between September and December, lower NAO and AO indices delayed the onset of spring flight. As lower NAO and AO indices were linked to colder local conditions in autumn

(Appendix A: Tables S1, S2), the delay in spring flight are most likely due to temperature‐related effects described earlier (Appendix A: Tables S1, S2). However, we found that the NAO and AO indices between January and March generally had positive regression coefficients, which suggested an opposite pattern: higher indices in winter were linked to later flights. Although higher NAO and AO indices are generally linked to warmer conditions in the eastern U.S.A., we did not find a significant relation between monthly temperatures and the climate indices between January and March (Appendix A: Tables S1, S2).

Most probably, the winter NAO and AO are influencing spring phenology through other non‐temperature meteorological variables that can affect diapause and post‐diapause processes. Although the mechanism is

38 still unclear, the inclusion of NAO and AO indices in phenological studies highlights their usefulness in discovering windows during which non‐temperature variables influence spring phenology. Recent models were able to forecast the winter NAO indices a year ahead, which can be useful for timing management interventions (Dunstone et al. 2016, Wang et al. 2017).

Finally, we propose that there was no significant shift in the timing of the spring flights, because although autumn temperatures have increased, spring temperatures have not. Due to the importance of spring temperatures in the emergence of the tortricid species, it is possible that spring conditions have not shown a large enough change to significantly affect phenology (Figure 1). It is likely that the phenological responses of each tortricid species will eventually change with a warming future, although these changes may differ based on their unique biology and life history. Species‐specific changes in phenology could have important consequences, especially with trophic interactions (Harrington et al.

1999). Therefore, it is imperative that future work on phenology take a community‐level approach to predict possible ecological consequences.

Acknowledgements

We would like to thank the faculty and staff of the Penn State Fruit Research Extension Center for their immense help and insight into the system, and especially Dr Larry Hull (retired) who provided much of the trap data from his almost 40 years as an entomologist at the FREC. This work was funded by the

National Science Graduate Research Fellowship and the National Science Foundation, grant DEB‐

1354819. We would also like to thank the reviewers and Rebecca M. Johnson for their helpful suggestions. The authors declare they have no conflicts of interest.

Incorporating diapause to predict the interannual dynamics of an important agricultural pest

Abstract

We develop a new population-scale model incorporating diapause induction and termination that allows multi-year predictions of pest dynamics. In addition to predicting phenology and voltinism, the model also allows us to study the degree of overlapping among the life-stages across time; a quantity not generally predicted by previous models yet a key determinant of how frequently management must be done to maintain control. The model is a physiological, stage-structured population model that includes temperature-dependent vital rates, diapause processes, and plasticity in development. The model is statistically fitted with a 33-year long weekly term time series of C. pomonella adults captured in pheromone traps from a research orchard in southern Pennsylvania. The multiannual model allows investigation of both within season control strategies, as well as the likely consequences of climate change for this important agricultural pest. The model predicts that warming temperatures will cause earlier spring emergences, additional generations, and increased overall abundance. Most importantly, by calculating the circular variance, we find that warmer temperatures are associated with an increase in overlap among life-stages especially at the beginning of the growing season. Our findings highlight the importance of modeling diapause to fully understand C. pomonella lifecycle and to better inform management for effectively controlling this pest in a warmer future.

Note: DP conceived the idea, performed the statistical analyses, created the figures/tables, and wrote the manuscript with support from the co-authors. Spencer Carran contributed to the model validation and contributed to the manuscript. DB provided the original data, wrote the ‘study site’ subsection, and helped

40 with the critical revisions. OB and William Nelson assisted with the project conception and helped with critical revision.

Introduction

Cydia pomonella L. (Tortricidae), commonly known as the codling moth, is a destructive insect pest of pome fruits across the world. The larvae tunnel into the host fruits to consume their seeds leading to premature ripening and early fruit drops. An unmanaged population can infest whole orchards leading to significant fruit damages as high as 80% loss (Myburgh 1980, Juszczak et al. 2013). Since control must be applied before newly hatched larvae bore into the fruits to be effective, precise timing of chemical applications is necessary (Khan et al. 2018). Growers currently rely on established degree-day models that forecast pest activity using the predictable relationship between insect development and ambient temperatures. By calculating the amount of accumulated heat units or degree-days, growers can predict the timing of emergence and the total number of generations per year in order to determine the optimal times to intervene (Ratte 1985, Jarosik et al. 2004, Logan et al. 2006, Rebaudo and Rabhi 2018).

One of the limitations of degree-day models, however, is that an annual start date to initiate degree day accumulation (called the “biofix”) must be empirically determined each season to initiate degree-day accumulation. Biofixes are set by growers who monitor for the start of pest activity with the aid of pheromone traps. However, establishing the biofix can be a source of confusion as some programs recommend setting the biofix with the first adult moth catch while others recommend the first consistent adult moth catch (Knight and Light 2012). The biofix is necessary as current models do not track the species through winter when individuals are in diapause, a physiological state of suspended development.

In C. pomonella, mature larvae (4th-5th instar) are cued into diapause by shortening daylengths and decreasing temperatures, with dormancy being terminated when enough chill days are accumulated

(Neven et al. 2000). By explicitly modeling both the induction and termination of diapause, we should be able to follow the individuals through the winter thus forgoing the biofix requirement.

In addition to being an important survival strategy to bide through unfavorable times, diapause also serves a crucial role in resynchronizing the population to a single life-stage at the end of each year

(Tauber and Tauber 1976). C. pomonella exhibits developmental asynchrony or overlapping life-stages which increases through the growing season due to developmental variability and environmental heterogeneity (Bjørnstad et al. 2016). Diapause then homogenizes the population structure ensuring of the synchronized adult emergence in the spring. Many insect species like C. pomonella need a minimum chilling period for diapause termination and warming winters can potentially reduce population synchrony (Stålhandske et al. 2015, Forrest 2016). It has been observed in C. pomonella that when the diapausing larvae fail to meet the chilling requirement, there is a less synchronized adult emergence which can influence the degree of overlapping of suspectable stages in the next generation (Neven et al.

2000, Dyck 2005). Thus, diapause has important implications for management under climate change.

Here, we propose a physiological stage-structured model that runs across multiple growing seasons, eliminating the need to set a biofix and enabling us to carry over the effect of past year’s conditions on future pest activity. Three important biological features are incorporated into the model: (1) temperature-dependent vital rates such as development rate parameterized from previous experimental work, (2) diapause induction and termination, and (3) individual variability in development. First, we run our model over multiple years with average daily temperatures collected in a research orchard in southern

Pennsylvania and validate the model using a 33-year time series of weekly adult males captured in pheromone traps at the same site. After validating the model with historical data, we study predictions with daily temperatures increased by 3°C which is the current climate scenario projected by 2100 (Beits et al. 2011). We quantify the degree of developmental overlap as this will influence the frequency of

42 required management using circular variance. Finally, we explore whether the variability in diapause termination will significantly increase the amount of overlapping life-stages through the growing season.

Methods

Study species

C. pomonella diapauses as mature larvae (4th to 5th instar larvae). After terminating diapause and completing its development, the adults emerge in early May to mate and oviposit. The newly hatched first instar larvae then burrow into the blossoms or fruits within twenty-four hours (Lacey and Unruh 1998).

After fully developing within their hosts, the mature larvae emerge from the fruits to pupate. In

Pennsylvania, C. pomonella complete two generations, though a favorable growing season allows for a partial third generation.

Time-series

We study time series data spanning 33 years (1984-2016) of pheromone trap data from the

Pennsylvania State University’s Fruit Research and Extension Center (FREC). The FREC consists of a

140-acre research center located in south central Pennsylvania near the city of Biglerville (latitude: 39.93

º N, longitude: -77.25º W) and the research orchards of Arendtsville, located three miles away (latitude:

39.92 ºN, longitude: -77.29 ºW). Trapping of adult males occurred in 21 blocks of multiple mature, fruit bearing apple blocks that are of 2-4 hectares in size. Because there was no single block that had continuous trapping across thirty-three years, we averaged the multiple time-series (In addition, we also used Multiple Autoregressive State Space modeling to combine multiple time-series as an alternative method, see Appendix B).

The modeling framework

We created a physiological stage-structured model that incorporated the six life stages of C. pomonella: egg, the five larval instars, the diapausing fifth instar larvae, pupae, reproductive adults, and senescent adults (Figure 1). We incorporated stage-specific vital rates that are dependent on ambient temperature as determined from previous laboratory experiments. The full model and its derivation are provided in

Appendix B.

The basic form of the stage-structure model is as follows:

푑푁 푖 = 푅 − 푅 − 훿 푁 (푡) (Equation 1) 푑푡 푁푖−1 푁푖 푁푖 푖

Each term describes the flow of individuals in and out of the stage 푁푖 with 푖 representing the different

life-stages . 푅푁푖−1 is the number of individuals recruited from the previous stage 푁푖−1 with 푅푁푖 being the

individuals maturing into the next stage 푁푖+1. 훿푁푖 represent the stage-specific instantaneous per capita mortality rate at time 푡. The recruitment of individuals both in and out of a stage is dependent on the development rate 휇푖(푡, 푥) which is the instantaneous rate that an individual in stage 푁푖 and of age 푥 will mature on to the next stage (Equation 2).

푓푖(푡, 푥) 휇푖(푡, 푥) = (Equation 2) 1 − 퐹푖(푡, 푥)

Here, 푓푖(푡, 푥) is the probability distribution of the through-stage development time for stage 푁푖, and

퐹푖(푡, 푥) is the cumulative probability of 푓푖(푎, 푡). Generally, ordinary differential equations assume an exponentially distributed developmental period with individuals sharing the same probability of maturing out of a stage regardless of age. It is more biologically realistic then to assume that there must be some developmental delay and individual variability. Therefore, we assumed the probability distribution to follow an Erlang distribution (a case of the gamma distribution with a rate parameter λ and an integer shape parameter 푘) (Manetsch 1976). A special property of the Erlang distribution is that it can be constructed as a chain of 푘 independent exponential distributions, which allows the underlying integrodifferential equations to be solved with a system of ODEs through the ‘linear chain trickery’ (Metz and Diekmann 2017).

Looking at the egg (퐸) and first instar larvae (퐿1) as an example, individuals will flow through

푛퐸 and 푛퐿1 sub-compartments respectively according to:

푑퐸 (푡) 푏(푡) 퐴(푡) − 푛 휇 (푡)퐸 (푡) − 훿 (푡)퐸 (푡) 푖푓 푖 = 1 (Equation 3) 푖 = { 퐸 퐸 1 퐸 1 푑푡 푛퐸휇퐸(푡)퐸푖−1(푡) − 푛퐸휇퐸(푡)퐸푖(푡) − 훿퐸(푡)퐸푖 (푡) 푖푓 푖 > 1

푑퐿1 (푡) 푛 휇 (푡)퐸 (푡) − 푛 휇 퐿1 (푡) − 훿 (푡)퐿1 (푡) 푖푓 푖 = 1 (Equation 4) 푖 { 퐸 퐸 푛퐸 퐿1 퐿1 1 퐿 1 = 푖푓 푖 > 1 푑푡 푛퐿1휇퐿1(푡)퐿1푖−1(푡) − 푛퐿1휇퐿1(푡)퐿1푖 (푡) − 훿퐿(푡)퐿푖(푡)

Here 푏(푡) is the per capita birth rate and 훿(푡) is the per capita mortality rate. 휇퐸(푡) and 휇퐿1(푡) represent the development rate at time t for the egg and first instar larvae respectively. 휇푖(푡) is a function of time as the instantaneous development rate can change due to the ambient temperature. The mean duration for

1 1 individuals in the egg and first instar stage will be and because the average rate of movement 휇퐸 휇퐿1

1 1 through each compartment are 푛퐸휇퐸 and 푛퐿1휇퐿1. The coefficient of variation is and √푛퐸 √푛퐿1 respectively.

Vital rate functions

We characterize the functional forms of the vital rates including birth rate, development, mortality rates, and finally diapause induction and termination from previously published laboratory data

(see Appendix B).

Birth rate

Figure 3-2: Temperature-dependent responses for the per capita birth rate. The line shows the fit of a gaussian function to experimental data. See Appendix B for parameter estimates.

We fit a Gaussian function to describe how the per capita birth rate 푏(푇) depend on temperature.

2 −(푇 − 푇 ) 표푝푡 (Equation 5) 푏(푇) = 퐵퐸 exp 2 ∗ exp(−퐶퐴(푡)) 2휎퐸

Here, 퐵퐸 is the maximum number of eggs laid at the optimal temperature 푇표푝푡, 휎퐸 describes the reduction away from the optimum. We also included an Allee effect to the birth rate, 퐶, due to the importance of synchronous emergence for finding mates. Here, the birth rate is influenced by the adult density 퐴(푡) (Friedenberg et al. 2007, Yamanaka and Liebhold 2009).

Development rate

The stage-specific development rates were described as a sigmoid function as it best fit the data.

Here, 푀푗 representing the maximum developmental rate, 푇표푝푡푗 describing the temperature at which there is an inflection point, and 훽푗 being a fitted parameter for stage 푗 according to:

푀푗 휇푗= (Equation 6) ( ) 1 + exp −훽푗 ∗ (푇 − 푇표푝푡푗)

Mortality rate

For the instantaneous mortality rate, we fit a modified Wang model for each life stage (Wang et al. 1982). We assumed that all larval instars share the same temperature dependent mortality functions according to:

1 훿푗(푇) = 1 − exp(−푇 − 푇푑푗) exp(푇퐷퐿 − 푇) (Equation 7) ex p ((1 + )(1 + )) ∗ 푏푗 푑푗 푑퐷퐿

Here, the 푇푗 is the optimal temperature for survival with 푑푗 and 푏푗 being the fitted parameters.

Diapausing Induction and Termination

Environmental variables such as temperature and photoperiod influence the induction and termination of diapause in C. pomonella (Ashby and Singh 1990). The species is induced into diapause at a critical daylengths of 13.5 to 15 hours, though warm temperatures are known to delay dormancy

(Garcia-Salazar et al. 1988). We assumed that all fifth instar larvae, regardless of age, can enter dormancy and that the induction rate is only dependent on the day of the year modulo 365 (푡365) .

As photoperiod is consistent year to year, the induction rate remains the same across the years such that We described the rate as a sigmoid function with 퐼푎 as the maximum rate of diapause,

퐼푏describes as the steepness of the function, while 퐼푐 is the day of year where there is an inflection point

(Equation 8).

퐼퐴 퐼(푡) = 휇푗= (Equation 8) 1 + exp(−퐼퐵 ∗ (푡365 − 퐼퐶))

To terminate diapause, C. pomonella has a chilling requirement of at least 60 days below 10 ℃

(Neven 2013). We assumed that, analogous to accumulating enough degree days to mature into the next stage, diapausing larvae must accumulate enough chill days to terminate dormancy. Therefore, to model the stage of diapause, we had two compartments which we call 퐷1 and 퐷2. In the first compartment, the individuals move through the stage depending on how cool the temperature is (Equation 9). For 퐷2, we assume that the individuals are in the state of quiescence or continuing their development into the pupal stage. Like development rate in the other life-stages, the development out of stage 퐷2 increases with warmer temperature (Equation 10). For both 퐷1 and 퐷2, we use a sigmoid function with 푀푑 being the maximum termination rate, 푇퐷 describing the temperature at which there is an inflection point, and 퐵푑 as the parameter determining the steepness of the slope.

푀퐷1 휇퐷1(푇) = (Equation 9) (1 + exp(훽퐷1(푇 + 푇퐷1 ))

푀퐷2 휇퐷2(푇) = (Equation 10) (1 − exp(−훽퐷2(푇 − 푇퐷2 ))

All vital rate functions were parameterized with previous laboratory data when available. We ran the model with average daily temperature collected from the FREC for the 33 years.

Model calibration and sensitivity analysis

The model was fit to the C. pomonella pheromone trap data to estimate unknown parameters using trajectory matching and maximum likelihood, assuming a normal distribution for the counts. For each year, we calculated the root mean square error and the Pearson correlation between the simulated and observed data to quantify fit. We also ran a local sensitivity analysis (using the method of differences) to determine which model parameters were most influential in affecting the timing of the first adult emergence and the accumulated abundance of the egg, diapausing larvae, and adult. Specifically, sensitivity was calculated as:

푂푗(1.05푝푖) − 푂푗(0.95푝푖) 푆푒푛푠푖푡푖푣푖푡푦 = (Equation 11) 0.10 푂푗(푝푖)

Here, 푂푗 represents the model output with 푗 either being the timing of emergence or the overall abundance and the 푝푖 represents the parameter of interest. The sensitivity is described as the percent change in the model output in response to the percent change in each parameter. To quantify the timing of the first adult emergence, we first ran the model and then used a smoothing spline to determine the day of year when there is the first peak in reproductive adults. To investigate changes in the overall abundance of the egg, diapausing larvae, and reproductive adults, we accumulated the total number of each life stage at the end of each year.

Calculating developmental synchrony

We used circular variance (V) that have been modified for stage frequency data to quantify the degree of overlap in life stages at a given time (Fischer 1983, Bjørnstad et al. 2016). By distributing the developmental indices around a circle with 0 representing the newly laid eggs and 2휋 representing the

53 senescent adults, we can calculate the order parameter, R, to get the circular variance 푉 = 1 − 푅 according to:

2 2 푘 푘 1 푅 = √(∑ 푦 cos 휃 ) + (∑ 푦 sin 휃 ) (Equation 12) ∑푘 푦 푗 푗 푗 푗 푗=1 푗 푗=1 푗=1

Here, the 푦푗 is the population density of stage j and 휃푗 is the mid-point (in radians) of the developmental index of stage 푗 and 푘 is the total number of stages (see Appendix B). We calculated the circular variance across 33 years of simulation and averaged the quantity for each day of the calendar year.

To investigate how warmer temperatures may influence pest dynamics, we simulated the model with average daily temperatures from 1984-2016 increased by 3℃. As there is evidence that warmer temperatures would delay diapause induction in Cydia pomonella, we modified the parameter (퐼푐) so that diapause was induced later (Stoeckli et al. 2012). This was to show how flexible our framework can be in investigating how different diapause responses could influence pest dynamics under warmer temperatures. We derived key phenology parameters and circular variance to compare the high temperature scenario to the current ambient temperatures. We used this to test the hypothesis that more variability in diapause termination will increase the overlap between the life stages across the growing season. For this, we ran the model under the higher temperature scenario with 푛퐷퐿 being decreased to 10 and 5 Smaller number of subcompartments corresponds to an increase in termination variability. We calculated the average circular variance per day of year across the years.

All analyses were done in R (Version 3.53) and the full code and maximum likelihood was done in the ‘pomp’ package (King).

Results

The model successfully characterized 33 years of data without the need of a biofix at the beginning of each year. This allowed us to track diapausing fifth instar larvae through the winter until reemergence in the spring. Individuals could be tracked through winter with the fifth instar larvae being induced into diapause and terminating diapause in the coming spring. There is general agreement between our model and the field trap data, especially when it comes to the first peak or spring emergence (Figure 6; see

Appendix B). However, the model tends to overestimate the total pest abundance especially in the period

2008- 2012.

Figure 3-6: Population dynamics of Cydia pomonella predicted by the model in black. In red is the time series of the pheromone trap data from 1984 to 2016. The y-axis is the log-transformed reproductive adult abundance and the x-axis is the day of year. The period of pheromone capture varies somewhat between years.

The sensitivity analysis shows that the timing in adult emergence was significantly influenced by the development rate of the pupae (Figure 7). Specifically, the most sensitive parameter was the one that controlled the temperature at which inflection happens in pupal development (푇푃). Additionally, we found the parameters controlling the maximum development rate (푀푃) and the steepness of the slope (훽푃) to be influential as well.

For the abundance of egg, diapausing larvae, and adult, the most sensitive parameter was the day of year when there is the inflection point in the diapause induction (퐼푐 ) (Figure 8). Changes to the optimal temperature of reproduction (푇표푝푡 ) was also influential in affecting the overall abundance among the three life-stages. We also see that the parameters related to the maximum rate of diapause termination

(푀퐷1 and 푀퐷2) was significant for the diapausing larvae. Generally, the sensitivity analysis indicates the importance of diapause induction on the overall abundance of C. pomonella.

When the average daily temperatures were increased by 3ºC, the model predicted earlier spring flights with adults emerging in April instead of early May. In addition to an earlier spring phenology, we also see longer adult emergences under the high-emission scenario. The warmer temperatures also allow for prolonged pest activity with adults being present through the late fall. The overall pest abundance increased under the high-emission scenario due to decreased mortality especially for the larvae and the diapausing larvae. Most importantly, increased winter survivorship of diapausing larvae can lead to more reproductive adults emerging in the spring (Figure 9).

Figure 3-9: The population dynamics of C. pomonella predicted by the model in blue with unchanged average temperatures (T). In red is the population dynamics predicted by the model with average temperatures increased by 3 ℃ (T + 3℃) and diapause induction being delayed. All scenarios were simulated from 1984 to 2016 with the y-axis as the log-transformed reproductive adult abundance and the x-axis is the day of year.

60 represent the 95% confidence interval. An increase in circular variance is associated with more overlapping life stages at a given time. Bottom: The averaged circular variance across the day of year with the blue line representing the circular variance calculated with the historical temperature data (1984 to 2016) and an unchanged 푛퐷퐿 and 푛푃. In red, is the circular variance calculated with the daily temperature increased by 3°C and an unchanged 푛퐷퐿 and 푛푃. The orange and yellow represent the averaged circular variance of model run under high-emission scenario with 푛퐷퐿 풐풓 푛푃 varied at 10 and 5 respectively.

For both the current and high-emission scenarios, we found the circular variance to decrease in the winter as only the cold-tolerant diapausing larvae can survive the lethal temperatures. The circular variance then increases after day 100 (early April) as the diapausing larvae terminate dormancy and resume their development to emerge as adults. Under warmer temperatures, however, the developmental synchrony erodes more quickly with a sharp increase in the circular variance due to the earlier emergence of adults (Figure 10: Top).

To investigate if the circular variance changes significantly if the variability in diapause termination increases, we increased the coefficient of variation (decreased the number of subcompartments 푛퐷퐿) under the high-emission scenario. We found no notable difference with the average circular variance predicted by the model under the current scenario and the high-emission scenarios (Figure 11). As we decreased the number of subcompartments (푛퐷퐿), we found an earlier shift for when the circular variance increases though the maximum value did not differ between the different number of subcompartments. When we increased the variability in the development rate of the pupal stage, however, we instead found a significant change in circular variance. As we increase the CV, there

61 was an earlier and larger increase in circular variance. Since C. pomonella is a boring fruit pest, increased circular variance means that each generation may need more than one pesticide application.

Discussion

A modeling framework that can track pest populations through multiple years to predict short and long-term dynamics is useful for improving pest management. By incorporating diapause, our stage- structured model robustly predicted 33 years of weekly codling moth trap data without the need for any annual biofix; a deficiency of previous phenology models. Our model captured the seasonal patterns found in the field which further supports the crucial role that temperatures have in driving pest dynamics.

Furthermore, by calculating the circular variance, the model allows us to understand and predict the dynamics of developmental desynchronization through the growing season; another challenge for previous models. Winter serves a crucial role in resynchronizing the life-stages as cold temperatures halt development and increase mortality of the non-diapausing life-stages. As only the diapausing stages can survive through winter, the emergence of the overwintering generations will be synchronized due to similar developmental indices though that synchrony erodes through the growing season until winter homogenizes the stage-structure again. Therefore, our study highlights how including diapause processes is essential for understanding how C. pomonella’s life-cycle tracks the seasonal fluctuations in its habitat

(Tauber and Tauber 1976, Powell and Logan 2005, Friedenberg et al. 2007).

When simulating the model under a high-emission climate scenario, the model predicts earlier emergences in adults, an additional generation per year, and an increase in overall abundance (Figure 6).

Mechanistically, it is likely that after diapause is terminated, warmer spring temperatures will hasten C. pomonella’s development allowing the adults to emerge earlier in the spring (Stefanescu et al. 2003,

Forister and Shapiro 2003, Pak et al. 2018). Warmer temperatures will also prolong the growing season and decrease the development time allowing for an additional generation as predicted by other phenological models of C. pomonella (Luedeling et al. 2011, Stoeckli et al. 2012). Finally, pest abundance would increase with warmer temperatures due to increased fecundity and survivorship specifically with lower winter mortality allowing for more diapausing larvae to survive and develop into adults. Our findings suggest that managing C. pomonella will have to change in a warmer future.

A complete understanding of C. pomonella’s management also requires quantifying the degree of developmental synchrony as it would influence the window for when vulnerable life-stages are present.

Overlapping life-stages would make it harder for growers to find a suitable time for pesticide applications.

Under the high-emission climate scenario, the model predicts an earlier overlapping of life-stages with the circular variance increasing as the growing season progressed. This indicates that interventions will need to be more frequent not only because of an additional generation but due to the increase in developmental asynchrony. Mechanistically, insect life stages can vary in the minimum temperatures required for development and this can generate developmental synchrony (Powell et al. 2000). For example, both C. pomonella’s fifth instar larvae and pupae have lower developmental thresholds to hasten development in the spring while the second instar larvae have the highest developmental threshold to assist with diapause initiation (Setyobudi 1989). We propose at warmer temperature the stage specific developmental thresholds may be less effective in synching the life-cycles to the seasonal environment thus leading to earlier and stronger overlap among the life-stages.

We hypothesize that the earlier decorrelation in development may also be amplified when there is increased variability in diapause termination. Warmer winters would make it harder for C. pomonella to accumulate the necessary chill days and failure to meet the chilling requirement have been linked to asynchronous adult emergence. (Neven et al. 2000, Stålhandske et al. 2015). However, when we

63 increased the variability in the diapause termination rate by decreasing the number of subcompartment

푛퐷퐿 we did not see a significant change in developmental synchrony (Figure 11). Instead, we saw that increasing the variability in the pupal development rate lead to an earlier and higher increase in circular variance. While our findings do not fully support our hypothesis, we do show that pre-emergence processes, specifically post-diapause development, may serve an important role in determining the initial population synchrony.

Future work may refine these findings as our approach uses a relatively simplistic model for the complicated diapause processes. Empirical work should further be directed towards more clearly defining different phases of diapause induction and termination. Previous experiments that have measured diapause termination in C. pomonella generally quantify the rate based on the fraction of diapausing larvae that successfully pupate in the spring (Ashby and Singh 1990). It is important to note that this is not true diapause termination, as larvae may have terminated diapause but remain in a state of quiescence until temperatures rise above a minimum threshold (Tauber and Tauber 1976). Additionally, it would be useful to understand the role of warm temperatures on delaying insects from entering dormancy

(Steinberg et al. 1992). This can prove difficult as depending on the latitude, C. pomonella differ in their overwintering behavior as diapause becomes increasingly obligate with latitude (Setyobudi 1989).

Finally, another future direction is incorporating not only environmental seasonality but resource availability. C. pomonella’s survival is dependent on timing its life-cycle with the phenology of their host plants. As fruits are harvested in commercial orchards, it is conceivable the pest dynamics observed in an unmanaged orchard like the field data considered here can differ from what is observed in a commercial one.

Our model can be easily modified for other insect pests especially any species that diapauses with a chilling requirement. For example, the FREC hosts multiple tortricid pests that overwinter through

64 diapause with growers being primarily concerned with C. pomonella and the oriental fruit moth

(Grapholita molesta). Therefore, it could be beneficial then to use our modeling framework to determine the timing of interventions by considering the phenology of multiple pest species integrated pest management. If a single application of pesticide can target multiple species due to vulnerable life-stages overlapping in time, this would lower the economic and environmental cost of pesticides. Here, we provide a useful, adaptable tool for multiple pest species which can help inform for effective management strategies in the warming future.

Acknowledgment

We would like to thank faculty and staff of the Penn State Fruit Research Extension Center for their immense help. The work was funded by the National Science Graduate Research Fellowship and the

National Science Foundation grant DEB – 1354819. The findings and conclusions do not necessarily reflect the view of the funding agency

Data Availability

All codes are uploaded to https://github.com/pakdamie/codlingmothdiapause

Community phenology and multi-species management

Abstract

The timing of biological event, or phenology, is crucial for informing for management decisions in many human activities most notably in agriculture. However, many management policies generally take a single species approach which may not consider how management actions targeted for one species can influence other coexisting species. There has been a call to consider the phenology of all species for a multi-species approach. Using pest-management as an example, pesticide spray applied to one species may affect other pest species depending on the similarity of their biological timing. This can be quite advantageous for managers as it can lower the cost of management actions. However, it is crucial to explore the characteristics of the species’ phenology that could allow a multispecies approach to management.

Therefore, we formulated a simple mathematical model with distributed delays and pesticide interventions to investigate how difference in phenology can influence the benefits of management. We found that there are three characteristics that must be considered which is the (1) developmental variability in the species, (2) the phenological shifts and (3) the difference in economic importance. These aspects can guide future work into multi-species management.

Note: This was written with Jessica Conway, S. David Stupski, Ottar Bjornstad, and Caitlin Lienkaemper.

O.B helped with the project conception. DP and S.D.S conceived the idea, formulated the model, and contributed to the writing of the manuscript. C.L and J.M helped with the modeling, providing coding, assisted with the writing of the manuscript.

66 Introduction

The study of biological timing, or phenology, is a critical foundation for many biological and ecological processes. All organisms must appropriately time their life cycles with the environment as resource availability and seasonal hazards fluctuate. In the management of natural resources such as in forestry, fishery, and agriculture, phenology is monitored to schedule timely management actions. For example, in farms and fisheries, managers record the organism's schedule like the fruiting of plant crops and the migration of fishes as it would guide the appropriate times for harvesting (Chmielewski 2013,

Staudinger et al. 2019). Likewise, in pest management, control efforts must be timed to the pest phenology as interventions are only effective when the pest is at its vulnerable life-stages (Prokopy and

Kogan 2009). In fields such as public health, phenology is also relied upon for management as each infectious disease has its own seasonality such as vector-borne diseases (Moore et al. 2014, Touré et al.

2016). Across the scope of human activities, the ubiquity of biological timing indicates that phenology is crucial for decision-making.

Historically, there has been a single-species approach to management with careful monitoring for the date of first appearance, the peak abundance, and, to a lesser-extent, the developmental variability in pest phenology. The timing of peaks varies across the years, and for ectotherms, this is mostly due to variability in local temperatures (Rebaudo and Rabhi 2018). Less explored is the developmental variance which describes the degree of synchrony in individual development. This is a great interest to managers because the developmental variance determines the window for actions (Bjørnstad et al. 2016). In pest management, a pest species that develop and emerge synchronously over a short period of time may give growers a shorter window to apply chemical control. A species that develops more asynchronously may have a wider distribution across time, which may require more interventions. Therefore, it is not only phenological peaks that are useful but also the developmental variance as it informs for the frequency of management actions.

67 The limitation to the single-species management approach, however, does not account for the multiple species that coexist within a system. There have been calls from ecologists, specifically in fishery research, to incorporate a multi-species management approach (Quassi et al. 2013). For decision- making models in fisheries, researchers incorporated species interactions with complex food webs and community-wide disturbances (Sanchirico and Wilen 2001, Mangel and Levin 2005). It is likely that any management efforts targeted at one species such as mass harvesting in fisheries and non-specific pesticide sprays in agriculture may inadvertently influence other co-inhabiting species.

While all species have their own unique phenology, co-inhabiting species are influenced by similar abiotic cues like temperature, face the same seasonal hazards (example: winter), and rely on similar resources. This indicates that, at the community level, there are some general phenological patterns with some degree of temporal overlap (Ovaskainen et al. 2013b). Therefore, it is possible that, depending on the similarity in the species' phenology, any management efforts for one species can potentially influence the other. Such efforts could prove beneficial if applied to controlling multiple pest species as it could minimize the total cost of control.

A multi-species approach to pest-management has been documented informally by some growers as they found that, in certain years, the emergence of two pest species may be synchronized. Growers found that spraying for one pest species would affect the other thus decreasing the total sprays needed.

The minimization of chemical control would be advantageous for growers as overuse of pesticide has been linked to detrimental effects such as increasing the risk of insecticide resistance and negatively affecting beneficial insect species (Whalon et al. 2008). Many agricultural systems use integrated pest management (IPM), a program that advocates for ecologically based decisions, to inform for control efforts as it limits unnecessary sprays by setting species specific thresholds that growers can tolerate.

68 Accounting for the economic importance of each pest, one can then explore how different timings determine when multi-species management is possible.

We propose a simple mathematical model as a framework to investigate the role of multi-species management. Using distributed-delay differential equations, we can control the variability in development to show how this could affect the frequency of interventions. We then show how in a two-species scenario, how difference in phenology and economic importance may influence management strategies.

Most importantly, our framework can be easily adapted to any phenological process in any ecological system.

Our questions are:

1) How does developmental variability affect the total interventions needed? How does this change with the decay rate of the pesticide?

2) For two species, how does the phenological similarity in the species affect the number of sprays needed for the two species? How does this change with the decay rate of the pesticide?

3) If one species is more economically important than the other (one has a lower threshold of tolerance), how does this affect community management? Does the order matter as well?

Though we use pest management as a case study, our methods and findings are more general and can be applied to any other systems where phenology is important for scheduling management actions.

69 Method

The mathematical model

We use a simple two-stage model to represent a multi-stage organism where only one stage is vulnerable to pesticide. We assume that individuals develop through the first stage, 퐴, until they mature into the final stage, 퐵, which is affected by pesticides. The general form of the model is:

푑퐴 (Equation 1) = −훼 퐴(푡) 푑푡

푑퐵 (Equation 2) = 훼퐴(푡) − 휇퐵(푡) − 퐻(푡) 푚(푃(푡))퐵(푡) 푑푡

푛 (Equation 3) 푃(푡) = 푝 ∑ 푒푥푝(−푏(푡 − 푡푗)) 퐻(푡 − 푡푗) Type푗=1 equation here.

where α represents a constant transition rate, and we assume that there is no mortality in the first stage, 퐴.

In the final stage, 퐵, the individuals are removed with a constant natural mortality rate μ. 푃(푡) denotes the total pesticide in the environment at time 푡 and is the summation of all sprays made at 푡1 to 푡푛 (Equation

3) . In a single spray case, an initial pesticide concentration 푝 is introduced into the environment at time 푡푗 through the Heaviside Step function, 퐻(푡). The pest concentration degrades through a declining exponential function with the decay rate 푏 (Figure 1). The pesticide induced mortality rate, 푚(푡) is dependent on the pesticide concentration in the environment which we have modeled as a Hill Function.

푘 (푃/푚50) (Equation 4) 푚(푃) = 푚푚푎푥 푘 1 + (푃/푚50)

70 Here, 푚푚푎푥 is the maximum rate of pesticide induced mortality, 푚50 is the pesticide concentration where mortality rate is half of its maximum, and 푘 is the Hill coefficient that determines the steepness of the slope.

Figure 4-1: The schematic of the pesticide induced mortality function described with the Hill Equation

(Left) and the decay of pesticide concentration over time (Right). Parameters in the Supplementary

Material.

We incorporate variable delay into stage 퐴 by assuming the dwelling time to have an Erlang distribution

(Equation 6). The Erlang distribution is a special case of the gamma distribution with an integer shape parameter 푘 and the rate parameter α. For 푛 = 1, the Erlang distribution decomposes into the exponential distribution and as 푛 approaches infinity, the distribution approaches the Dirac function. Therefore, one can control the phenological variability seen in stage 퐵 by manipulating 푛 of stage 퐴. Therefore, we can modify Equations 1 and 2 to produce Equation 6 with 푔(τ) being the Erlang distribution:

푑퐴 0 (Equation 5) = ∫ −푔(τ)퐴(푡 + τ)푑 τ 푑푡 −∞

71 푑퐵 0 (Equation 6) = ∫ 푔(τ)퐴(푡 + τ)푑τ − μ퐵(푡) − 퐻(푡)푚(푃)퐵(푡) 푑푡 −∞

As integrodifferential equations such as Equation 6 and 7 can be difficult to solve, we use the useful property of the Erlang distribution where it can be described as a sum of 푛 i.i.d exponential distributions.

In this ‘linear chain trickery’ if the mean dwelling time in stage 퐴 is 1/α then it is equivalent to 푛

1 identical subcompartments with exponential distributions that have the mean dwelling time of (Figure 푛α

1 2). The coefficient of variation (CV) describing the development rate is then and we use this value to √(푛) determine developmental variability. If the developmental variability is high in stage A, then there will be a more asynchronous emergence of individuals in stage B.

For example, through this ‘linear chain trickery’, Equations 6 and 7 can then be rewritten equivalently as:

푑퐴1 (Equation 7) = −푛α퐴 푑푡 1

푑퐴2 (Equation 8) = 푛α퐴 − 푛α퐴 푑푡 1 2

⋮

푑퐴푘 (Equation 9) = 푛α퐴 − 푛α퐴 푑푡 푘−1 푘

푑퐵 (Equation 10) = 푛훼퐴 − 휇퐵(푡) − 퐻(푡)푚(푡)퐵(푡) 푑푡 푛

1/nα than this can be equivalent to 푛 number of subcompartments ‘chained’ together. As 푛 increases, the

CV decreases and there is less developmental variability.

Single species case

We first explored the relationship between developmental variability and the number of sprays needed in a single-species case. In IPM, growers are advised to spray every time the pest population reaches a pre-determined pest density which is called the action threshold. This is a preemptive intervention to ensure that the pest numbers do not grow to economically damaging numbers. For all analysis, a lower action threshold suggests that the grower has less tolerance for the pest species, and therefore that the species is more economically important.

73 To quantify the difference in developmental variability, we used the CV of 0.10, 0.20, 0.30, 0.40, and 0.50 which correspond to the integer shape parameters 4, 6, 11, 25, and 100 respectively. We also varied the decay rate of the pesticide spray to compare frequency of sprays between a slow and fast decaying pesticide. To ensure that the frequency of control can be compared across different developmental variability, the initial number of individuals was changed to ensure the total area above the action threshold was the same in the absence of pesticide.

Two species with the same economic thresholds

To model the phenology of the second species, we shifted forward the timing of the first species by 5, 30, and 80 days. Here, we define multi-species management as a scenario in which a species is sprayed at its action threshold with the pesticide affecting the other species (Figure 3). For example, spraying the earlier emerging species at its action threshold can control the later-emerging species such that no additional intervention is needed. The benefits of multispecies management are dependent on the overlap between the two species which is dependent on both the developmental variability and shift between species. Additionally, how long the pesticide last in the environment could the determine if the pesticide last long enough to affect the other species. We assume that all species rely on different resources or on the different parts of the same resources (ex: shoots versus fruits) such that the threshold cannot be summed.

To investigate the number of sprays needed to manage the two species with the same action- threshold, we varied the CV in development rate, the shift, and the pesticide decay rate. We then quantified the benefit of multispecies management by calculating the number of sprays needed when the model assumes that interventions affect both species versus a hypothetical situation where the pesticide only affects the targeted species but not the other. To do this we subtracted the number of sprays needed for first species case multiplied by two from the number of sprays needed for the multispecies management. We also quantified the benefit of multispecies management by investigating how only spraying for the earlier emerging species influences the total damage the second species would have.

Specifically, we sprayed only the first species at its threshold and then calculated the second species damage which we calculated as the area above the second species’ action threshold.

75 Two species with different economic threshold

Finally, we then explored scenarios in which the action thresholds of the two species differed.

Specifically, we were interested if the order in which the more economically important species appeared mattered for management. For example, we were interested in the outcomes depending on if the more economically important species emerges earlier than later. We ran the model with the first case having the action thresholds of the earlier emerging species be lower than the later emerging species. In the second case, the earlier emerging species had the action threshold set higher than the later emerging species. To better quantify the effects of the order, we again sprayed only the earlier emerging species and the damage of the second species.

All analyses were done in R (4.0.2) with the desolve package for solving the differential equations

(Soetaert et al. 2010) and the pracma package (Hans et al. 2017) for calculating the area under the curve.

76 Results

Single species management is influenced by developmental variability

Figure 4-4: The number of sprays needed for species sprayed at action thresholds 40 and 80. On the x- axis is the coefficient of variation while the y-axis is the decay rate of the pesticide.

The one-species scenario reveals that the developmental variability has a significant influence on the number of sprays that were needed to keep the pest density below the action threshold (Figure 4;

Appendix C). Generally, with species that have a wider distribution of emerging individuals there more sprays were needed. However, this pattern was reversed when the decay rate of the pesticide was low. To further investigate this, we calculated the rate of individuals entering the last stage at the time of the first pesticide spray. We saw that there was a positively correlation between the coefficient of variation and the rate of inflowing individuals (Appendix C; see Discussion).

Control for two species is affected by coefficient of variation and the shift between species

When two species have the same action thresholds, the shift between the two species as well as the developmental variability influenced the overlapping between species (Figure 5: Appendix C). For example, when the shift between species is 5 days, there is a large degree of overlap such that when the pesticide is sprayed for the earlier emerging species, it will likely affect the later emerging species. When the shift increased, however, we saw that more sprays were needed as the pesticide spray of the first species had less effect on the second species. This was also further true if the pesticide decays rapidly from the environment.

When quantifying the benefits of multispecies management, we saw that a more significant difference in the number of sprays when the shift in the two species was low and with larger CV. The patched appearance of Figure 7 is due to the nature of small integers/step-functions,

79 for the individual case. On the x-axis is the coefficient of variation while the y-axis is the decay rate. The panel are the shift is the number of days that are between the species.

Additionally, when looking at the total damage that the later emerging species would have if only the first species was sprayed, we found similar results. Generally, the intervention of the earlier emerging species can ensure that the damage of the second species is lower though the effects wane when there is a further shift between the species.

The difference between the species’ economic thresholds influence management

Here, we found that the order of the economically important pest species affected the benefits of multispecies management. In Figure 9, when the first emerging species has a lower action threshold than the later emerging species, the number of sprays decreases as the action threshold of the later emerging species increased. In the second scenario when the earlier emerging species, we found that there was a similar number of sprays to the first scenario (Appendix C)

82 We found that the benefits of multispecies management differed between these scenarios (Figure 10).

Specifically, when the economically important species emerged first, there little to no damage of the second species when the shift was at 5. However, when the economically important species emerged later, there was considerable damage which increased with difference in action threshold. Damage of the second species for both scenarios increased with shift.

Discussion

We formulated a mathematical model with distributed delays to provide a framework for understanding how phenology can inform for multi-species management. We found that to have a have a complete theory of management, it is necessary to understand both the species timing and developmental

83 variability. Generally, we found that when there was less synchrony among individual development rate, more sprays were needed to keep the pest density below the action threshold (Figure 4). Additionally, our model suggested that the phenological shift between species was an important factor in determining to what extent intervention on one species affected the other (Figure 6). Finally, we found that the order in which the most economically important species emerged influenced the benefits that multispecies management can provide (Figure 9). This suggests that phenological similarity between species and their difference in economic importance must be both considered for management decisions. In conclusion, our model suggests that multispecies management should focus on these three features (1) the developmental variability, (2) the phenological shift between the species, and finally (3) the difference in economic importance.

First, developmental variability reflects the degree to which individuals develop synchronously.

In our model, the shape integer controls for the developmental variability and has an important role in determining how frequently the pesticide sprays are required. Generally, we found a higher CV in development was linked to increase in the number of interventions required though this pattern was reversed when the pesticide decay rate was low (Figure 4). This is most likely due to a more asynchronous development being associated with a lower rate of individuals entering the last stage

(Figure 5). This means that when the pesticide decays slowly from the environment, residual toxicity can control for species with more developmental variability. This finding suggests that managers must consider the species’ growth rate at the time when management action is taken.

A crucial aspect of developmental variability that must be considered is that it can vary drastically among species. This in turn influences the degree of phenological overlap between species which influences if intervention on one species is likely to affect the other. It is also important to note that developmental variability in species can change over time especially when considering species with

84 multiple generations. Specifically, many species tend to lose their developmental synchrony over the growing season as they accumulate desynchronizing factors due to genetic variability and (Gilbert et al.

2004, Bjørnstad et al. 2016). Because each species can show varying developmental variability over the growing season, it is possible that multispecies management in the later growing season might be drastically different than the earlier part of the season. This highlights an advantage of our model in that developmental variability is easily modifiable through a single parameter. Therefore, managers can partition the growing season based on the emergence of each generation and adjust the parameter accordingly.

Secondly, the phenological shift between the two species heavily influenced if the pesticide spray for one species would affect the other. Generally, we found that as the distance between two species increased, the pesticide spray of the first species had less influence on the second species. This effect was further amplified when using a rapidly decaying pesticide as it might have degraded before the later species emerged. Therefore, the closer the two species are in their timing, the less sprays are needed to control for both species (Figure 7). For decision-making, it is important for growers to that note the difference in shift would vary year to year due to environmental fluctuations especially with temperature.

For example, warmer conditions can lead to earlier emergences while cooler temperatures could delay emergence. However, the phenological responses are unique to each species such that the shift between the two species timing can vary season to season (Pak et al. 2018). Ultimately, it is possible that as temperatures continue to rise with climate change, the species can shift from each other considerably.

Finally, we found that multispecies management must not only consider the phenological similarity between species but the difference in their economic importance. Specifically, we found that the order at which the most economically important pest species emerged influenced the management (Figure 10).

Our model showed that if the species with the lower action threshold emerged first then there is less

85 damage done by the later emerging species. Conversely, we found that if the least economically important species emerged first, then the interventions for the first species may not be enough to control for the later emerging species. These differences are further exacerbated as the difference in the economic importance between species increases. This suggests that even if species are phenologically similar, the difference in economic importance or action thresholds in this can affect the total benefits of multispecies management.

The advantage of this model is that it is modular for different species and distributed delay equations used in our model have already been widely used to model the phenology of species with medical or agricultural importance (Manetsch 1976, Gilbert et al. 2004, Nelson et al. 2013, Bjørnstad et al. 2016). By incorporating the life-stages with temperature dependent vital rates into the model, managers can then predict how phenological variance and shifts can change with changes in the environment. This in turn gives some information on the climatic conditions that managers can capitalize on for multispecies management. The model can also be more informative for actual decision making if managers implement the cost of management versus the cost of not taking into action. Finally, though our study focuses on two species, more species can be added to the model but it likely that the analysis can become cumbersome and unhelpful. To balance between complexity and applicability, future work should approach how multispecies management can work if the most economically significant species were targeted. Then seeing if interventions for that species can influence the less economically important species.

In conclusion, phenological research is crucial for the management of natural resources. By understanding each species’ phenology in relation to other coinhabiting species, we can begin research into incorporating multispecies management. Simple mathematical models provide crucial insight into generalizing common phenomenon across multiple species and systems. As biological timing is common in many human industries, we provided a way to explore multispecies management.

86 Acknowledgement

We would like to thank Spencer Carran for providing feedback. The work was funded by the National

Science Graduate Research Fellowship and the National Science Foundation grant DEB – 1354819. The findings and conclusions do not necessarily reflect the view of the funding agency.

A 117-year retrospective analysis of Pennsylvania tick community dynamics

Abstract

Tick-borne diseases have been increasing at the local, national, and global levels. Researchers studying ticks and tick-borne diseases need a thorough knowledge of the pathogens, vectors, and epidemiology of disease spread. Both active and passive surveillance approaches are typically used to estimate tick population size and risk of tick encounter. Our data consists of a composite of active and long-term passive surveillance, which has provided insight into spatial variability and temporal dynamics of ectoparasite communities and identified rarer tick species. We present a retrospective analysis on compiled data of ticks from Pennsylvania over the last 117 years. We compiled data from ticks collected during tick surveillance research, and from citizen-based submissions. Most of the specimens were submitted by citizens. However, a subset of the data was collected through active methods (flagging or dragging, or removal of ticks from wildlife). We analyzed all data from 1900–2017 for tick community composition, host associations, and spatio-temporal dynamics. In total there were 4491 submission lots consisting of 7132 tick specimens. Twenty-four different species were identified, with the large proportion of submissions represented by five tick species. We observed a shift in tick community composition in which the dominant species of tick (Ixodes cookei) was overtaken in abundance by

Dermacentor variabilis in the early 1990s and then replaced in abundance by I. scapularis. We analyzed host data and identified overlaps in host range amongst tick species. In conclusion, we highlight the importance of long-term passive tick surveillance in investigating the ecology of both common and rare tick species. Information on the geographical distribution, host-association, and seasonality of the tick community can help researchers and health-officials to identify high-risk areas.

88 Note: This chapter was published in Parasite and Vector (2019)*- Damie Pak, Steven B. Jacobs & Joyce

M. Sakamoto. DP did all the analyses, data-visualizations, and helped write the manuscript. SJ provided guidance on the matter of the database and contributed to writing the manuscript. JS directed the project and contributed to writing the manuscript.

*Pak, Damie, Steven B. Jacobs, and Joyce M. Sakamoto. "A 117-year retrospective analysis of

Pennsylvania tick community dynamics." Parasites & vectors 12.1 (2019): 189.

Introduction

The Centers for Disease Control and Prevention reported a 3.5× increase in vector-borne diseases in the USA between 2004–2016, with 76.5% of cases caused by tick-borne pathogens (Rosenberg et al.

2018). The increase in tick-borne disease is attributed to climate change, land use changes, and expanding geographical ranges for several important endemic tick species, posing novel risks to local communities

(Simon et al. 2014, Sonenshine 2018) . Although there are many tick-borne pathogens, the vast majority of tick-borne disease cases are caused by Borrelia burgdorferi (Eisen et al. 2017, Rosenberg et al. 2018), the main etiological agent of Lyme disease in the USA. Pennsylvania has had the highest number of total

Lyme disease cases since 2000, with increasing numbers of annual cases across several counties (Figure

1).

Figure 5-1: Annual reported cases of Lyme disease. By state from 2006–2017 (left) and by counties in Pennsylvania from 2006–2017 (right). Public data from the Centers for Disease

Control and Prevention

90 Surveillance data collected over multiple decades may reveal spatio-temporal changes in ectoparasite communities (Rand et al. 2007) . Data such as spatial distribution and occurrence of both abundant and rare species of ticks can be correlated with land use (e.g. habitat loss, fragmentation, management), fluctuating environmental conditions, or changes in human or animal behavior (e.g. encroachment may bring reservoir hosts in closer proximity) (Brownstein et al. 2005, Simon et al. 2014).

Long-term surveillance data can also reveal shifts in the temporal dynamics of tick populations and communities (Sonenshine 2018). Although the seasonality of known tick species has been described, year-to-year distribution of tick species may be influenced by inter-annual variability (e.g. local climate) and biotic factors (e.g. local reservoir species abundance). These data can be useful for developing predictive models that accurately measure the risks of tick-borne zoonotic agents.

We present a retrospective analysis of tick collection data in Pennsylvania from the early 1900s to

June of 2017. Some of the data prior to 1968 had been published in a progress report on ticks from

Pennsylvania but were presented in a format that included anecdotes and overall percentages rather than raw number breakdowns by species (Snetsinger 1968). We revisit these specimens and utilize the raw data from both these time periods (1900–1967 and 1968–June 2017) to identify shifts in tick community composition, phenological patterns, and host associations. We used our database to map the distribution of major tick species at the county level, investigate tick community spatiotemporal dynamics, and explore host associations by tick species.

Methods

Study locations

The state of Pennsylvania (PA) is located in the Northeast/Mid-Atlantic region of the USA (State

Center: 40.9699889, −77.7278831(“Node: Pennsylvania (316987717)” n.d.). Climate in Pennsylvania is variable by location, but broadly classified as a continental type with warm, humid summers (mean temperature ranges of 17.7–23.33°C, (“Pennsylvania State Climatologist” n.d.). The majority of

Pennsylvania’s land-use is dedicated to agriculture (both croplands and pastures), forestland, with some dense urban areas (“Pennsylvania State Climatologist” n.d., Pennsylvania. n.d.)There have been significant changes in the human populations of PA from 1960 to 2010, but a large proportion of the PA population has remained heavily clustered around Philadelphia and Pittsburgh, which are located in southeastern and in south-western PA respectively (Figure S1).

Submissions

The PSU Frost Entomological Museum (‘Frost Museum’) houses arthropod samples collected by researchers, teaching collections, and samples submitted by the public for identification. We present our analysis of the tick specimens from 1900 to June 2017, although some Frost Museum collections date as far back as the late 1800s. Because tick samples were submitted to the Department of Entomology or the

Frost Museum over a period of 117 years, they represent multiple collection/submission periods (early

1900–1959; 1960–1969; 1970–1988; Tick Research Lab (TRL) submissions from 1990 to 1993 and

1995–present). Two public campaigns account for the majority of the citizen-submitted specimens. The first campaign (between 1963–1967) was conducted by Dr Robert Snetsinger. He enlisted the help of the

92 public through advertisements in radio, television, and newspapers to obtain 700 specimens(Snetsinger

1968). Additionally, he utilized active surveillance methods to collect approximately 500 ticks using a combination of dragging, sweeping, live animal trapping, and roadkill examinations of mammals and birds to assess tick abundance in localized areas (Snetsinger 1968) A second funded campaign dedicated to estimating tick abundance and species diversity was launched by Steven B. Jacobs (second author) from 1990 to 1993 (TRL), in which he cataloged, identified, and labeled each specimen. Specimens were accompanied by additional data: date of tick discovery, location and vegetation type of tick encounter, and host species.

Data from both campaigns and subsequent submissions were combined into a single dataset for our analyses. For analysis on the distribution of tick species over time, we used total tick counts. For quantifying host association, however, we used “submission” number which we defined as a vial or lot containing one or more ticks. We chose to use this more conservative measure rather than total specimen count to avoid skews in abundance by hosts. For example, a submission lot of 1 tick versus 50 ticks from a host were both classified as “one submission”.

While most specimens were collected within state boundaries, a few were declared from people either visiting or returning from visiting other states. Tick specimens identified as species that are not commonly found in Pennsylvania were later discovered to have been imported from other states/countries or found on exotic animals. Non-PA data were excluded from state-wide analyses, but were included in the supplements Tables S1 and Table S2.

93 Identification

Ticks were morphologically identified to species and life stage using taxonomic keys for

Argasidae, Ixodidae east of the Mississippi, Dermacentor, nymphal Ixodes, and nymphs of Amblyomma

(Cooley and Kohls 1944, 2011, Keirans and Clifford 1978, Yunker et al. 1986, Keirans and Litwak 1989,

Keirans and Durden 1998). Species-level identification is crucial since at least 3 Dermacentor species, 3 species of Amblyomma, and 9 different Ixodes species have been reported in Pennsylvania. If diagnostic characters were missing due to damage to the specimen, the next level of taxonomic identification was used (e.g. samples with missing mouthparts that were clearly Prostriata were identified as “Ixodes spp.”).

In a few cases, samples were not identified beyond “tick” and were designated “Ixodidae” for hard ticks or “Argasidae” for soft ticks. Unusual specimens or those that were difficult to identify were sent to the

National Tick Collection, Georgia Southern University for confirmation (by Dr James Oliver at the time of confirmation).

Spatial distribution

We focused on the geographical distribution of the five most abundant species of significant public health and veterinary importance: Amblyomma americanum (Linnaeus), Dermacentor variabilis

(Say), Ixodes cookei (Packard), Ixodes scapularis (Say), and Rhipicephalus sanguineus (Latreille).

Working on the assumption that counties with higher populations would submit more specimens than less populated counties, we estimated the prevalence rate (the total numbers of individual ticks per 100,000 people). This was done by adjusting the total tick count of each species by the county’s total population.

We looked at relevant time periods during the surveillance programme: 1960–1969; 1990–1999; 2000–

2009; and 2010–2018. For each time period, we used the United States Census data for 1960, 1990, 2000 and 2010, respectively, to calculate the tick prevalence rate (Figure S1) (“U.S. Census data from 1960 to

1990.” n.d., “US Census data: U.S. Census Bureau” n.d.)

94 Temporal analysis

We used the annual sum of all individual tick specimens to investigate how annual submission rates changed over time. We then analyzed the temporal dynamics of the five most abundant taxa (A. americanum, D. variabilis, I. cookei, I. scapularis and R. sanguineus). We did not evaluate total counts by year as these varied drastically due to the active campaigning for citizen submissions or the introduction of identification fees. Therefore, we looked at the proportional contribution of each species to the annual summed counts of all the five major species. To detect if there have been any monotonic trends (i.e. gradual shifts in abundance), we ran a non-parametric, two-sided Mann-Kendall trend test on the yearly proportion of each of the species between 1900–2017.

We investigated the seasonal distribution of the tick community by analyzing the monthly frequency of submissions. Citizen submissions were to include the date of discovery, but for specimens that lacked these, we used the date that a given submission was received. We analyzed the seasonal distribution of motile life stages (larvae, nymphs and adults) for the five most abundant taxa. The proportions were calculated by comparing the monthly abundance of each life stage (larvae, nymphs and adults) to the cumulative sum of all stages.

Host associations

Host information (combined by family, except for dog, cat, human and groundhog) was available for many of the tick submissions. Host data were classified as either domestic or wildlife. We analyzed the host-tick data by summing the total submissions by both the tick species and the host group. We constructed a circular network map to visualize the relationships between tick species and hosts. All host association analyses were done with R (Version 3.4.13) with the packages Kendall for the Mann-Kendall

95 test and the circlize package for chord diagrams of host association mapping (Gu et al. 2014, Mcleod and

Mcleod 2015)

Results

A total of 4491 submission lots consisting of 7132 tick specimens across 23 species were identified (Table 1). Five species of ticks accounted for the majority (91%) of the total number of tick specimens: Dermacentor variabilis (n = 3172); Ixodes scapularis (n = 1899); Ixodes cookei (n = 897);

Rhipicephalus sanguineus (n = 332); and Amblyomma americanum (n = 196). Other tick species that had at least 100 specimens were Dermacentor albipictus (Packard) (n = 107), Ixodes dentatus (Marx)

(n = 120), and Ixodes texanus (Banks) (n = 111). The remaining ticks had less than 100 specimens/species and included both hard and soft tick species.

97 Spatial analysis

Ticks were submitted from all 67 counties in Pennsylvania (Figure S2). We hypothesized that more tick submissions would come from areas with higher human populations. As expected, tick submissions were heavily clustered around Allegheny and Philadelphia Counties, where Pittsburgh and

Philadelphia are located, respectively. When we adjusted the total tick count by county population- decade, we found higher prevalence rates in less populated counties. For example, in 1990–2000, the highest prevalence rates of I. scapularis submissions were from Elk County (870 individuals per 100,000 population). Neighboring Forest and Cameron counties also had high submissions of I. scapularis with

116.64 and 589.97 individuals per 100,000 respectively (Figure 2).

Figure 5-2: Distribution of the five most abundant tick species across Pennsylvania over time.

Prevalence rates (tick counts per 100,000 population, left) represent tick abundance adjusted by county population for time periods 1960–1969, 1990–1999, 2000–2009 and 2010–2018.

Cumulative counts of ticks by species shown on the right

Dermacentor variabilis distribution was largely localized to the southern parts of the state. From

1990 to 2000, the highest proportion of D. variabilis submissions came from Greene County, the most

southeastern county of Pennsylvania (865.45 submissions per 100,000). Other southern counties with

high tick loads per capita included Fulton County (350.60 per 100,000) and Franklin County (117.26 per

100,000). Ixodes cookei was more evenly distributed throughout Pennsylvania, although like I.

scapularis, it was more highly abundant in the northern counties. In 1990–2000, Forest County had the

highest prevalence rates of Ixodes cookei with 80.87 per 100,000. R. sanguineus and A. americanum had

very few submissions and their distribution was mostly scattered across Pennsylvania (Figure 2).

For the tick species with less than 150 submissions across 1900 to 2017, we aggregated the submissions by genus. Multiple species within the genera Ixodes and Dermacentor were widely distributed across Pennsylvania (I. scapularis, I. cookei, D. andersoni and D. albipictus) (Figure S2).

Other species in the genera Amblyomma, Argas, Carios (Ornithodoros) and Haemaphysalis were not as widely distributed, possibly because these species are not commonly encountered or because the specimens were introduced from their native geographical ranges (Figure S2, Figure S3).

Temporal shifts in species abundance

Prior to the 1990s, the majority of the tick submissions were identified as I. cookei and R. sanguineus (Figure 3). The spike in the number of submissions in 1990 was largely due to D. variabilis, but gradually, I. scapularis became the dominant taxon submitted. Results from the Mann-Kendall test supports these observations with an upward trend in the I. scapularis counts (tau = 0.288, P = 0.02) and a significant downward trend in D. variabilis (tau = −0.408, P = 0.002). The Mann-Kendall also indicate that the proportional contributions of I. cookei (tau = −0.607, P < 0.001) and R. sanguineus (tau = −0.377,

P = 0.005) to the total count have also significantly shifted over a century.

100

Seasonality

Overall, we find that the majority of tick specimens were received in the months between April and July with the highest proportion of tick submissions in May (Figure 4). Submissions of D. variabilis, A. americanum, I. cookei and R. sanguineus were most abundant during the period between May and July.

Dermacentor variabilis and A. americanum were most abundant from March to October. Ixodes cookei and R. sanguineus samples were submitted throughout the year, with peak abundance in June. Samples of

I. scapularis were also submitted year-round, but the peak abundances were bimodally distributed with a large peak between May-June and a second peak between October-November.

101

Figure 5-4: Seasonal distribution of tick submissions over time. On the left is the total proportion of tick specimens received at different months of the years from 1900 to 2017. We On the right are the proportional seasonal abundances of each of the five major tick species compared (1900–2017) seasonal distribution by lifestage for 6233 of the 7132 total of tick specimens (Figure 5). Four percent of total submissions were larvae (n = 237), 20% were nymphs (n = 1271), and 75% of the submissions were adults (n = 4 725). Of the D. variabilis submissions, there was a total of 32 larvae, 33 nymphs and 3059 adults from 1960 to 2017 (Figure 5), with a unimodal distribution peaking around June.

102

(larvae, nymphs and adults) to the cumulative sum of all stages by species.

Prior to 1990 I. scapularis were very rare so we have only shown their lifestage-specific seasonal abundance since 1990. Overall, the nymphal and larval submissions showed a unimodal pattern with the

103 highest proportion of submissions received in June. For the adult submissions, there were prominent bimodal peaks in May and October with similar proportions of submissions received in both seasons.

The majority of I. cookei submissions were nymphs (n = 521), followed by adults (n = 182) and larvae (n = 88). The submission patterns indicate that I. cookei specimens can be encountered year-round, but that nymphs were the most commonly encountered lifestage. Across all lifestages, we see that the distributions are unimodal with peaks in early summer between May and June. There were too few lifestage submissions for A. americanum and R. sanguineus to make sufficient comparisons with the seasonality of these species.

Host association

One of the assumptions about passive surveillance is that there is an inherent bias toward humans as the hosts, particularly since most specimens submitted by the humans from themselves, their pets, or other domestic animals. By far the majority of submissions were associated with humans and their domestic animals and this reflects the fact that many of the specimens in our collection were submitted by people on themselves or their pets (Fig. 6). Of 4491 submissions from PA, there were 2662 attached to humans, 666 associated with cats or dogs, 20 from other domestic animals, and 168 submissions pooled from multiple hosts (mixed). There were 11 additional submissions found on various exotic animals.

There were 689 submissions for which there was no host record or the ticks were not attached to a host.

The remaining 275 were found on various wildlife.

104

105 Discussion

Our data is unique in that it contains details about tick community composition and spatio- temporal dynamics from Pennsylvania over a 117-year period. Subsets of our data had been reported as percentages or combined with data from other museums and literature reviews to estimate the distribution of one or more tick species across the state of Pennsylvania (Snetsinger 1968, Snetsinger et al. 1993). To our knowledge, this is the first time that these data have been compiled in their entirety and analyzed in this format. We were able to detect seasonality, shifts in tick community composition, and host associations that have not been well-documented in a quantitative manner. The seasonality data for the five most abundant tick species inferred by our passive surveillance data is consistent with previous records of seasonality described by other researchers (Kollars et al. 2000, Burg 2001, Kollars and Oliver

2003, Simmons et al. 2015) demonstrating that these types of passive data contain biologically meaningful signal.

Shifts in tick community composition

In the 1960s, PA tick communities consisted predominantly of three species: I. cookei, D. variabilis, and R. sanguineus (Figure 2). The most abundant species at that time, I. cookei, often referred to as a groundhog tick, is actually a broad- host tick feeding principally on medium mammals, although humans and dogs will also be parasitized (Bishopp and Trembley 1945). The second most abundant species, D. variabilis, was widely distributed and eventually became the dominant species submitted over

I. cookei in the 1990s. By 1991 D. variabilis had been identified from all but 4 counties (Figure 2)

(Snetsinger et al. 1993). After 1995, D. variabilis annual submission rates declined as I. scapularis submission rates increased (Figure 2). Because there were gaps in submission numbers for certain years, we cannot say for certain why there were shifts in the abundance of these species.

106 Although we cannot directly infer a causal negative relationship between D. variabilis and I. cookei with I. scapularis, this pattern was also observed in neighboring Ohio. The passive surveillance programme run by Ohio Department of Health (started in 1978) did not detect I. scapularis (=formerly I. dammini) until 1989 (Pretzman, Daugherty, Poetter, & Ralph, 1990). At that time, the dominant species were D. variabilis (~97% of submissions) and I. cookei (1.2%) (Kollars and Oliver 2003). Between 1989 until 2008, I. scapularis accounted for less than 1% of the total submissions, but after 2009, the abundance began to increase. By 2012 they accounted for 24.8% of ticks submitted to the Ohio Department of Health

(Wang et al. 2014).

Ixodes cookei abundance was highest prior to 1990 but has since become very rare in our dataset.

Yet, we know that the abundance of I. cookei in the Maine passive surveillance programme has been constant, even as I. scapularis submissions have increased (Rand et al. 2007). It is possible that I. cookei abundance in PA has also remained relatively stable, but that we lack sufficient power to detect I. cookei.

This inconsistency in submission rates may be due to the shift from free to per-submission charges for tick identifications that occurred in the mid-late 1990s.

The fourth and fifth most abundant tick species in our database were R. sanguineus, the brown dog tick (287 submission lots consisting of 332 specimens) and A. americanum (183 submission lots consisting of 196 specimens) (Table 1). Although R. sanguineus originated in Africa, it is a cosmopolitan urban pest species found worldwide in association with humans and their canine companions (Brites-Neto et al. 2015). Snetsinger (Snetsinger 1968) suggested in 1968 that R. sanguineus had established breeding populations in Pennsylvania, but the abundance tapered off after 1968 and none exist in our database since 2002. In contrast, the submission rates of A. americanum increased and then leveled off from 1960 to 2000. Although not commonly encountered, we have specimens from as recent as 2016.

107 Host Association Vector-host associations (including host specificity) are important for predicting the risk of pathogen transmission and identifying key players in a sylvatic disease cycle. The specificity to host varies with the tick species. In our dataset generalist tick species (e.g. D. variabilis, I. scapularis, I. cookei) were found parasitizing a wide range of vertebrate hosts, while specialist species (e.g. I. dentatus,

I. marxi, I. muris) were associated with a single host or limited to host size (e.g. small mammals or birds).

This was consistent with other data in the literature.

Less commonly encountered tick species can sometimes lead to incorrect assumptions about host preferences and subsequent risk of pathogen transmission. While some tick species may be presumed to hold strict host preferences, they may bite humans if given the opportunity. For instance, I. dentatus bit humans in cabins that had been inhabited by their squirrel hosts in Maine and Vermont, and I. cookei was found on humans in West Virginia (Hall et al. 1991, Lubelczyk et al. 2010). Ixodes texanus was only found from raccoons in our dataset, but this species is known to feed on several mammalian hosts

(Brillhart et al. 1994, Kollars and Oliver 2003, Cohen et al. 2010). Commonly held misconceptions about host associations of certain tick species (e.g. I. cookei as “groundhog ticks”) based on lack of encounter may result in ignoring a potentially epidemiologically important vector (e.g. I. cookei is a vector of

Powassan encephalitis virus and may potentially be another vector of Ehrlichia muris (Xu et al. 2018)).

Multi-faceted approach to tick surveillance

Surveillance can be a powerful tool for the detection of introduced species (transient or established), emergent arthropod-borne pathogens, and disease risks due to increases or changes in vector community composition. Both passive and active surveillance strategies have their strengths and weaknesses but combined, they provide a more complete picture of tick community dynamics. Active tick surveillance approaches such as dragging, flagging, CO2-trapping, or live animal capture, can be very

108 effective for assessing tick load by habitat (Snetsinger 1968, Brillhart et al. 1994, Kollars and Oliver

2003, Bouchard et al. 2013). It can, however, be labor-intensive, costly, and difficult to implement over a wide geographical area. Passive surveillance is more cost-effective and less labor-intensive and can provide insight into ectoparasite abundance, host associations, or habitat associations across a wider geographical area (Oliver et al. 2017). Passive surveillance (particularly based on submissions by citizens) may run the risk of under-representing certain taxa or reflect a bias toward certain host associations. However, citizen-submitted tick collections can provide valuable baseline data on prevalence and likelihood of tick encounters and may be more strongly correlated with reported human cases of tick-borne diseases than active surveillance alone (Cortinas and Spomer 2014, Barrett et al. 2015,

Xu et al. 2016, Ripoche et al. 2018). A community engagement programme that actively recruits ticks submitted by citizens should be coupled with support for a rigorously curated database of tick submissions.

Utilizing complementary strategies can help fill in knowledge gaps about tick prevalence. In a study using a combination of retrospective literature review, data compilation of specimens from archival collections, and active collection (dry ice, dragging and flagging) in counties presumed to be free of A. americanum, 68 of 77 counties of Oklahoma were determined to be colonized (Barrett et al. 2015). The metadata associated with a multi-pronged approach to tick surveillance (assuming proper data curation and management) can provide insight into tick-host associations, vegetation, seasonality, and shifts in population structure that can be used for modeling disease risk. Archival tick samples (or their DNA) can be useful for retrospective mining for research on the population genetics of ticks to detect gene flow, host shifts, or on their microbial inhabitants.

Implicit in any tick surveillance strategy is having trained tick biologists who can readily distinguish species by morphological and molecular characteristics. In the last 20 years, I. scapularis has

109 become the most abundant tick species in Pennsylvania. While distinguishing Ixodes from other genera of ticks is fairly simple, species-level identification requires more detailed morphological examination, since there are six endemic species of Ixodes and three exotic species that could potentially be misidentified as

I. scapularis. More generally, although many tick species are incompetent vectors of B. burgdorferi, they may be vectors and/or reservoirs of other pathogens/parasites, or acquire pathogens during co-feeding

(Courtney et al. 2003, Kollars and Oliver 2003, Baer-Lehman et al. 2012, Brown et al. 2015, Campagnolo et al. 2018). It is therefore important to correctly identify tick species, not only for the determination of disease risk but also because the treatments for the pathogens they transmit may differ significantly.

Conclusions

An ideal tick surveillance programme would not only utilize multiple approaches and have a dedicated tick biologist proficient at species identification on staff, but it would also take a proactive stance that is not limited strictly to immediate threats. Since 1993 (~25 years) there have been 28 publications on ticks from Pennsylvania, and 22 of them were focused on I. scapularis and/or the microbiota (mostly on pathogens) (Anderson et al. 1990, Serfass et al. 1992, Snetsinger et al. 1993,

Daniels et al. 1993, Lord et al. 1994, Magnarelli et al. 1995, yeh et al. 1995, Courtney et al. 2003, Dick et al. 2003, Lo Re et al. 2004, Schoelkopf et al. 2005, Steiner et al. 2008, Devevey and Brisson 2012, Chen et al. 2014b, Rogers et al. 2014, Shock et al. 2014, Stromdahl et al. 2014, Crowder et al. 2014, Han et al.

2014, Hutchinson et al. 2015, Simmons et al. 2015, Springer et al. 2015, Edwards et al. 2015, Miller et al.

2016, Sakamoto et al. 2016, Chinuki et al. 2016, Waits et al. 2018). While I. scapularis is an important vector that warrants this attention, other tick species are being ignored. The consequences of neglecting other potential ticks of epidemiological significance include missing shifts in tick biodiversity, not identifying the potential causes of said shifts, not monitoring changes in range expansions of vectors, and not detecting the presence of introduced or established species. Recently, the presence of H. longicornis

110 was reported in Pennsylvania, but we have no data on whether it has been introduced previously, or whether it has established populations. Given the potential risks that this parthenogenic tick poses (wide host range, possible vector of multiple pathogens, may induce meat allergies), this example highlights the potential dangers of focusing on only one vector-pathogen system(Chen et al. 2014b, Chinuki et al. 2016,

Rainey et al. 2018, Tufts et al. 2019). Hybrid datasets from tick collections derived from multiple sources represent a powerful tool for mining past ecological and epidemiological events. Many states maintain county records on passive tick submissions to veterinary or medical health officials, but there may be other cryptic collections (and associated data) housed in museums, universities, government institutions, or with private individuals. Combining these data with other ectoparasite databases and currently unexplored collections will provide ectoparasite researchers the robust dataset needed for a massive meta- analysis. These cryptic ectoparasite collections will provide the basis for exploring hypotheses such as: (i) are shifts in tick populations correlated with increasing human encroachment on natural habitats; (ii) what are some phenological reasons for the increase in I. scapularis abundance; or (iii) if displacement of a dominant tick community species occurs, what are the implications for tick-borne disease risk? We anticipate that participating in such a study will fill in the gaps of knowledge about less-studied tick species as well as highlight the intrinsic value of museum collections of ectoparasites.

Acknowledgments

The authors would like to thank Dr István Mikó for critical manuscript review, citizens and researchers who have contributed ticks to PSU, The Frost Entomological Museum for access to the collection, and critical suggestions from reviewers. This work was supported by the USDA National

Institute of Food and Agriculture and Hatch Appropriations under Project #PEN04691 and Accession

#1018545. Additional support came from NSF GRFP DGE1255832 (to DP), and start-up funds from the

Huck Institutes of Life Sciences, and the Penn State College of Agriculture to JMS. The findings and conclusions do not necessarily reflect the view of the funding agency.

111

Data availability

Data supporting the conclusions of this article are provided within the article and its additional files. The datasets supporting the conclusions of this article are available at GitHub in the https://github.com/pakdamie/passive_surveillance_tick_2018/tree/master/MAIN_DAT, and at Zenodo

(https://zenodo.org/record/1476091).

112

Multi-scale phenological niches in hyperdiverse Amazonian plant communities

Abstract

Phenology has long been hypothesized as an axis of niche partitioning which promotes species coexistence. Tropical plant communities exhibit seasonal phenology but show immense diversity in reproductive timing. Here, we investigated (1) whether this phenological diversity is non-random, (3) at what temporal scales community phenological patterns are structured, and (3) what are factors that drive the phenology in different phylogenetic or dispersal groups. We applied multivariate wavelet analyses on the long-term time series of seed rain of hyperdiverse plant communities in the western Amazon. We then characterized the community-level phenology as being either synchronous or compensatory (i.e. anti- synchronous) across temporal scales. We found significant synchronous whole-community phenology at a wide range of time scales, consistent with shared environmental responses or positive interactions among species. We also observed compensatory phenology within groups of species likely to share traits

(confamilials) and especially those who share seed dispersal mechanisms. Our results show that community phenology is shaped by shared environmental responses but that the diversity tropical of plant phenology is partly a product of non-random temporal niche partitioning. The scale-specificity and time- localized nature of synchronous and compensatory dynamics indicates the importance of multiple and changing drivers of phenology.

Note: This was a collaborative work with Damie Pak who ran all the statistical analyses/ contributing to the writing with Jesse Lasky as the lead primary investigator.

113 Introduction

Ecological communities often exhibit interspecific diversity in phenology or biological timing.

This diversity may represent an axis of niche partitioning among species that reflects the action of community assembly and evolutionary processes (Ashton et al. 1988, Gonzalez and Loreau 2009,

Wolkovich and Cleland 2011, Bernard-Verdier et al. 2012, Godoy and Levine 2014). Species differences in phenology may limit interspecific competition and cause complementarity through time in resource- use, apparent competition mediated by natural enemies, or interactions with mutualists, potentially promoting species coexistence (Robertson 1895, Rathcke and Lacey 1985). Alternatively, periodically harsh environmental conditions may limit the possible phenological traits capable of persisting, and pulses in resource supplies may promote phenological synchrony (Gentry 1974, Rathcke and Lacey 1985,

Vasseur et al. 2014, Usinowicz et al. 2017, Detto et al. 2018). However, phenology remains a relatively poorly characterized dimension of functional diversity in many communities, especially for communities with many long-lived organisms, owing to the multi-scale complexity of phenological patterns and lack of long-term monitoring required to characterize phenology (Wolkovich et al. 2014).

Co-occurring species with similar reproductive phenology might be more likely to compete for mutualist frugivores (Saracco et al. 2005) or other resources, given that reproduction is a resource demanding activity (Karlsson and Méndez 2005). As a result, those species capable of coexisting might partition phenological space and thus resource use through time. For example, coexisting species with shared mutualists might exhibit phenological differences generating temporally based niche partitioning

(Robertson 1895, Botes et al. 2008). Researchers have studied evidence for this axis of niche variation in tropical forests e.g. (Gentry 1974, Stiles 1977, Wheelwright 1985, Ashton et al. 1988, Poulin et al. 1999,

Jones and Comita 2010) and other plant communities (Elzinga et al. 2007, Botes et al. 2008, Albrecht et al. 2015). Within a diverse community, phenological niche partitioning might be strongest among species with shared mutualists and/or resource requirements (Encinas‐Viso et al. 2012) as is often the case among

114 species that are phylogenetically related (Robertson 1895, Prinzing et al. 2001, Donoghue 2008, Davies et al. 2013). However, past studies have found little evidence for the importance of temporal niche partitioning in mutualist interactions across diverse communities. Part of the challenge of detecting such partitioning is that there are simultaneous and opposing processes acting on phenology, leading patterns indicative partitioning to only emerge at certain time scales or over certain periods of time (Vasseur et al.

2005, Keitt 2008, Lasky et al. 2016).

Tropical plant communities are remarkable because there are often multiple species reproducing at any given time of the year (Frankie et al. 1974, Gentry 1974, van Schaik et al. 1993). This phenological diversity may be made possible by favorable temperature and in rainforests, abundant moisture throughout the year (Gentry 1974, Usinowicz et al. 2017). One explanation for the great phenological diversity in tropical plant communities is that without abiotic constraints, phenology is free to evolve neutrally across the year. Alternatively, different species may be limited by different conditions fluctuating across the year (e.g. light, moisture, heat), thus phenological diversity may be a consequence of distinct strategies or sensitivities to seasonality in resources. Furthermore, despite the year-round presence of at least some reproductive species, tropical tree communities often exhibit some synchrony among a subset of the community, perhaps due to shared responses to abiotic seasonality and role of environmental filtering in community assembly (van Schaik et al. 1993, Lasky et al. 2016, Detto et al.

2018) or seasonality in effective frugivory (Poulin et al. 1999). Alternatively, positive density dependent interactions among species may also result in synchronous reproduction, for example when greater reproductive output among species decreases rates of seed predation (Ashton et al. 1988, Jones and

Comita 2010) or when reproduction by one species facilitates frugivory on individuals of a neighboring species (Carlo 2005). Community patterns of synchronous versus compensatory reproduction might occur during periods with specific abiotic conditions, suggesting the specific abiotic constraints that drive species interactions (Vasseur et al. 2005).

115

Here, we study two long-term and high frequency records of community-wide phenology in hyperdiverse communities. We use multivariate wavelet transformation, a type of spectral analysis for nonstationary data, to test for evidence of non-random phenology to address the following questions:

1. Do communities exhibit compensatory (anti-synchronous) patterns of reproduction through time, suggesting phenological niche partitioning could promote coexistence? Alternatively, do species primarily show synchrony of reproduction, suggesting shared constraints or exploitation of shared resources shapes phenology?

2. Is evidence for phenological niche partitioning strongest among functionally similar species, potentially those with the greatest likelihood of interspecific competition? Specifically, do species with similar fruit morphology, suggesting shared dispersers, exhibit stronger compensatory dynamics? Do related species, potentially sharing abiotic constraints on phenology or seed dispersers, exhibit stronger compensatory dynamics?

3. Is evidence for compensatory and synchronous dynamics scale-specific, with changing patterns at different scales? Such a finding would suggest that failure to conduct multi-scale analysis might miss evidence for scale-specific niche partitioning.

4. Are phenological niche patterns mediated by abiotic fluctuations? Greater phenological diversity might be more possible when abiotic conditions are permissive which may allow many species to be reproductive.

116 Methods

Study sites

We studied two forest plots in the western Amazon basin, in Cocha Cashu, Peru and Yasuní,

Ecuador (Appendix E: Figure S1). These plots were monitored continuously for different intervals, from

February 2000-February 2017 in Yasuní and September 2002-January 2011 in Cocha Cashu.

The study plot in Ecuador was located in Yasuní National Park at the Estación Científica Yasuní

(0° 41’ S, 76° 24’ W), a research station maintained by Pontificia Universidad Católica del Ecuador. The

Yasuní lowland rainforest is in the wettest and least seasonal region of the Amazon (Xiao et al. 2006,

Silman 2011). Mean annual rainfall is 2826 mm, with no months having <100 mm rainfall on average

(Valencia et al. 2004a, 2004b). Seed traps were placed within the 50-ha Yasuní Forest Dynamics Plot

(YFDP, established in 1995), where elevation ranged from 216 to 248 m. This is a hyperdiverse forest, with 1104 tree species recorded on the first 25a of the YFDP (Valencia et al. 2004a, 2004b)

The study plot in Peru, Cocha Cashu Biological Station, is located within Manu National Park

(11°54'S, 71°22'W), at the western margin of the Madre de Dios river basin in southeastern Peru. We established a 4-ha (200 × 200 m) long-term forest dynamics plot situated in mature floodplain forest habitat. Less than 10 cm of rain falls during the dry season from June to October, although this may vary annually (Terborgh 1990). Community-wide fruiting begins at the onset of the rainy season, peaking once in November and again between January and March (Terborgh 1990). During this period, roughly between September and April, an excess of fruit resources is available for frugivorous vertebrate consumers (Terborgh 1986) suggesting the potential for plant competition to attract frugivores. Fruit resources are scarce between May and August, during the transition from the rainy to the early dry season

117 (Terborgh 1990). During this period, frugivores adjust their behavior to cope with limited resources

(Terborgh 1983, 1986a; Van Schaik et al. 1993).

Seed rain data

In each plot, an array of seed traps was established. At Yasuní we followed the methods of

(Wright and Calderon 1995). In February 2000, 200 seed traps were placed in the 50 ha YFDP along trails but >50 m from the plot border. Every 13.5 m along the trails, a trap was placed a random distance between 4 and 10 m perpendicular from the trail, alternating left and right. Traps were constructed of

PVC tubes and 1 mm fiberglass mesh, positioned 0.75 m above ground, with an area of 0.57 m2. Twice monthly from February 2000 to February 2017 all reproductive parts in each trap were counted and identified to species or morphospecies using a reference collection of seeds and fruits maintained on site.

Seeds and whole mature fruits were counted individually; fruit segments (such as capsule values) were aggregated and counted as the equivalent number of a whole fruit. The number of seeds per fruit was counted directly from fresh specimens, our reference collection or photographs, or estimated from generic or familial data.

At Yasuní, some species were not well-separated in earlier years and thus species were excluded from analyses. We censored the time series of species belonging to Clusiaciae (due to issues with Clusia identification) and Solanaceae before 1/1/2007. For species belonging to the family Moraceae, their time series were censored before 1/1/2008 (due to issues with Ficus identification). We excluded all censored species from the analyses.

At Cocha Cashu, year-round quantitative data on fruit and seed fall were collected between 2002 and 2011 within the 4-ha plot. A 17 × 17 array of 289 evenly spaced seed-fall traps was installed within the central 1.44 ha (120 × 120 m) of the plot at the beginning of the study. Seed traps consisted of 0.49

118 m2 (70 × 70 cm) open bags made of 1-mm nylon mesh sewn to wire frames with 0.5-mm monofilament line. Corners of the traps were attached to nearby trees with 1-mm monofilament line so that the traps were suspended approximately 1 m above the ground. The contents of the traps were collected every 2 weeks, and all seeds, fruit and fruit parts (capsules, valves, pods, etc.) were identified to species and recorded. For fruits collected, fruit counts were converted to seed counts by multiplying by the average number of seeds per fruit for that species. Data on seeds per fruit were obtained from the literature

(Alvarez-Buylla and Martinez-Ramos 1992, Gentry 1996, Kalko and Condon 1998, Stevenson et al.

2002, Russo 2003, Cornejo and Janovec 2010).

Seed dispersal mechanisms

For Yasuní, we grouped species into different dispersal syndromes. We conducted two separate classification efforts, one for all species, and another focused-on tree species (excluding lianas, herbaceous, and woody shrub species). For all plants, we focused analysis on classifications as animal

(650 species) or wind (134 species) dispersed. For trees with animal-dispersed seeds, we followed

Harrison et al. 2013 and further classified them as terrestrial animal-dispersed (25 species), species with small (< 2cm, 209 species), medium (2-5 cm long, 62 species), and large (> 5 cm, 10 species) seeds dispersed by canopy animals. For trees with seeds dispersed abiotically, we included ballistically dispersed seeds (16 species) and wind dispersed seeds (30 species).

Likewise for Cocha Cashu, dispersal mechanism was assigned in a prior study using information from published studies conducted in the Madre de Dios basin and other long-term Neotropical rainforest sites (Bagchi et al. 2018). Tree species were classified based on their proportional dispersal by members of seven dispersal groups: (1) large and medium- bodied vertebrates (e.g. tapirs, spider monkeys, capuchins, guans, toucans, trumpeters), (2) small bodied non-volant arboreal mammals (e.g. tamarins,

119 night monkeys, kinkajous), (3) small birds (e.g. manakins, contingas and tanagers), (4) bats of the genus

Artibeus , (5) terrestrial rodents (e.g. agoutis and squirrels), (6) ballistic, (7) wind dispersal), and (8) those with unknown dispersal mechanism (two species of Calatola, Icacinaceae) (Bagchi et al. 2018). We took these published estimates and performed k-means clustering to produce six mutually exclusive groups of species with similar dispersal modes. These approximately correspond (based on cluster means) to groups dispersed mostly by large vertebrates (52 species), small vertebrates (20 species), wind (8 species), and bats (5 species), ordered by decreasing number of plant species in each group.

For all analysis on the taxonomic and dispersal groups, we insured that each group have at least 5 species. We did not use a lower threshold on number of records for inclusion of a species, as species contributions to group-wide phenological dynamics are weighted by number of seeds in the analyses below.

Weather data

We estimated monthly precipitation and minimum temperature at the plot level for each study site. Because local weather station data contained many missing observations, we used remotely sensed data. We used a ten-day precipitation time series estimated on a 0.05º grid by (Funk et al. 2014) using both remote and locally-sensed data. We used ECMWF/ERA-Interim reanalysis 4-hr temperature data at

2 m height estimated on a N128 Gaussian (~2º) grid (European Centre for Medium-Range Weather

Forecasts 2009) and calculated daily minimum temperatures and then monthly values.

We used weather station data from a site to estimate rough seasonality in wind speed in the

Peruvian Amazon in the region of Cocha Cashu. Specifically, we calculated average monthly wind speed

120 from a station in 150 km away, within 100 m elevation of Cocha Cashu, for the years 2004-2009

(http://atrium.andesamazon.org/meteo_station_display_info.php?id=12).

Statistical analysis

We used wavelet analysis to describe community-wide phenology in a time-localized manner across different temporal scales. These analyses were used to determine whether the whole community or subsets of species exhibited synchronous or compensatory (anti-synchronous) seed rain or whether they exhibited compensatory (anti-synchronous) seed rain. By analyzing seed rain dynamics in a way that allows us to resolve non-stationary and scale-specific patterns, we may improve our ability to detect multiple opposing processes affecting seed rain dynamics at different temporal scales or points in time.

Like other spectral analyses, wavelet transformations decompose signals at different scales but with the added advantage of being able to characterize time-localized and nonstationary patterns (Terrence and

Compo 1998, Keitt 2008). In the transformation, the base wavelet is translated across the time-series at varying dilations to identify the important timescales that contribute to the variability in the signal

(Cazelles et al. 2008). For each species’ time series, we first used continuous wavelet transformations:

∞ − 푡 − 휏 푤푘(푡, 푠) = 푠 1 ∫ 휓 ( ) 푥푘(휏)푑휏 −∞ 푠 (Equation 1)

Here, s is the scales of interest , 휓 is the basis function which we chose the Morlet wavelet, and 푥푘 is the seed abundance at scale 푠 at time 휏.

To then compare each species’ phenology in relation to each other, we calculated the wavelet modulus ratio (WMR) which quantifies the variance in the aggregate community-wide reproduction to the species-level reproduction (Keitt 2008, Keitt 2014, Lasky et al. 2016)

121 Λ (|∑ 푤 (휏, 푠)|) 푊푀푅(푡, 푠) = t,s 푘 푘 Λt,s ∑푘 |푤푘 (휏, 푠)|

1 푡−휏 2 ∞ − ( ) ( ) 2 푠 ( ) | | Here, Λ푡,푠 . = ∫−∞ 푒 ∙ 푑휏 and ∙ represents the complex norm. The numerator represents the aggregated variation in the seed abundance at time 푡 and scale 푠, whereas the denominator is the individual species variation. At unity, the WMR at the time period signifies complete phenological synchrony among the species, as species-level phenological dynamics are completely reflected at the community level. At zero, the WMR indicates complete compensation or anti-synchrony: all species-level dynamics are compensated so that community level reproduction is constant.

We calculated the total community WMR for all species in Yasuní (0.10 to 8.5 year periods) and

Cocha Cashu (0.08 to 4.2 year periods) respectively, with the periods differing between sites because of the frequency of trap collection or the duration of the study. The minimum scales were calculated as twice the median distance between successive dates and the maximum scales were calculated as half the maximum distance between the dates. We tested statistical significance of whole community WMR using a phase-randomized bootstrap (n=1000). All analyses were run in R (v 3.3.2). WMR was calculated using the package ‘mvcwt’ (Keitt 2014).

Taxonomic and seed dispersal groups

To investigate if species that are closely related share similar phenological niches and thus synchronous seed rain, or partition phenological niches and thus compensatory seed rain dynamics, we focused on taxonomic groups of species. Our analyses were done at the family-level to ensure sufficient sample size as confamilials still often share characteristics making them likely to exhibit evolutionary niche conservation or character displacement (Wiens and Graham 2005). Similarly, groups with shared seed dispersal mechanisms might be more likely to exhibit non-random phenology for reasons outlined

122 above. Thus, we separately grouped species based on their dispersal syndromes, as determined with the methods described above.

For these grouped analyses (family or dispersal syndrome), we first calculated the WMR on the wavelet-transformed phenology for each taxonomic or dispersal group. To test the hypothesis that species within a group exhibited non-random phenology, we generated a null distribution of each group’s WMR using permutations. We permuted species labels while maintaining the number of each species in each group, calculated WMR, and then repeated this permutation 1000 times. If the observed WMR of a group averaged across time points was above the 97.5th percentile of the permutation-based null distribution, we considered it as significant synchrony, if the observed WMR of a group averaged across time points was below the 2.5th percentile of the null distribution, we considered it as significant compensatory dynamics.

Climatic drivers of synchrony vs anti-synchrony

To determine whether climatic fluctuations might influence phenology among community members, we investigated the role of local temperature and precipitation on whole community WMR.

Because the time scale of the climate data was in months, we aggregated seed rain data to monthly average counts and then calculated the wavelet modulus ratio for Yasuní (2-70 months scales) and Cocha

Cashu (2-50 months scales). We then used wavelet reconstruction to filter the climate variables at the specific scales so that we could calculate their relationship with community WMR. Specifically, at each scale, we calculated the Pearson correlation coefficients between the community WMR and the wavelet- transformed minimum temperature or precipitation. The significance of the relationship was verified by comparing it to a null distribution of the Pearson correlation coefficient. Null distributions were generated by permutating the starting point of the climate time series while maintaining periodic boundaries and

123 then calculating the Pearson correlation between the randomized time series and the WMR (n=1000). If the actual correlation coefficients lie outside the null distribution, we considered the climate variable significant at that time scale. The wavelet reconstruction of the time series was done in the R package

“WaveletComp” (Rösch and Schmidbauer 2016).

Results

Community-wide phenology

At Yasuní (1067 species) we found a general trend of strong community synchrony in seed rain at scales of less than 50 days, while larger sub-annual periods were typically non-significant (Figure 1). At the annual scale, we found a shift back toward significant synchrony for most of the study. Additionally, we found significant synchrony at scales greater than 2 years with period of 3.85 years being the strongest. These patterns were largely stationary across the time period of the study, especially the pattern of strong synchrony occurring at subannual scale. However, there was a weakening to non-significant of annual-scale synchrony from 2012-2015.

By contrast, we did not find much significant synchrony at the sub-annual time scales at Cocha

Cashu (654 species) until 2006. From 2006 into early 2008 we found the emergence of community-wide synchrony across a wide range of temporal scales. Additionally, there was consistent significant synchrony at the 1-2 year scales across the duration of the study, indicating some shared annual and biannual dynamics among species.

124

Figure 6-1: The wavelet modulus ratio (WMR) of the aggregate Yasuní (1067 species) and Cocha Cashu

(654 species) community phenology (top panels) and time-series of the total species in traps (middle panels) and total estimated seeds in traps (bottom panels, natural log) in each period. In top panels, red indicates synchronous dynamics (high WMR) while blue indicates compensatory dynamics (low WMR).

The black contour lines bound the points in time and scale (years) when the WMR was significant through phase-randomized bootstrap (n=1000). Here, nearly all significant regions are high WMR

(yellow to red), with only two very small regions of significant low WMR (blue) in 2005 in Cocha Cashu at 0.1-0.3 month scales. The cone of influence (white shading in the top panels) detonates the regions where the wavelet transforms are affected by the boundaries of the sampling period.

125 Phenology among confamilials

We found distinct patterns of family-level phenology at Yasuní versus Cocha Cashu. At Yasuní among the 29 families analyzed, we found that species within the same family often tended to exhibit significant compensatory dynamics at sub-annual timescales (Figure 2). In particular, species in the

Cecropiaceae, Malpighiaceae, and Myristicaceae families exhibited significant compensatory dynamics at the 2-5 month timescales. Additionally, we found evidence of strong compensatory dynamics at the longer timescales such as in the 5-8 years period for Nyctaginaceae and Olacaceae. By contrast, at Cocha

Cashu among the 8 studied families, members of the same family exhibited a mix of significant synchrony or anti-synchrony, depending on the family, especially at the sub-annual time scales (Figure 3).

We found some consistency of family-level phenology shared across sites. Annonaceae at Cocha

Cashu exhibited significant compensatory dynamics at sub annual timescales (1-4 months) like at Yasuní.

By contrast, Bignoniaceae and Fabaceae species showed significant synchrony for sub-annual time scales at both sites.

126

Figure 6-2:The averaged wavelet modulus ratio of different plant families in Yasuní at the sub-annual

(left) and interannual (right) scales. The number in parenthesis represents the number of species analyzed within the family. Colored points represent either significant synchronous (red) or compensatory (blue) dynamics at the time scale.

127 Phenology among species sharing dispersal modes

Among the species sharing dispersal mechanisms, we found significant synchronous and compensatory phenology at multiple scales for few dispersal modes at Yasuní (Figure 4; Figure 5). When considering all growth forms, we found animal dispersed species exhibit synchrony compensatory dynamics at the 2-4-year scales and synchronous dynamics at the 3 month time scales. For the wind- dispersed species, we found synchronous dynamics at the 6 month scale (Figure 4).

The ribbon represents the null distribution generated through bootstrapping (n=1000). Any points that lie above the ribbon was considered significant and synchronous while any points below the ribbon indicated significant, compensatory dynamics

128

When considering canopy trees dispersed by animals, we found that trees with seeds that are less than 2cm to have synchronous dynamics at the 6-year scale. Additionally, for the ballistically dispersed tree species, we see compensatory dynamics at around 2 years with the wind dispersed tree species having compensatory dynamics at periods greater than 8 years (Figure 4).

Figure 6-5: The averaged wavelet modulus ratio for the different dispersal groups of tree species in

Yasuní. The number in parenthesis represents the number of species with the dispersal group. The ribbon represents the null distribution generated through bootstrapping (n=1000). Any points that lie above the ribbon was considered significant and synchronous while any points below the ribbon indicated significant, compensatory dynamics.

At Cocha Cashu, we found significantly non-random phenology within groups of similar dispersal mechanisms again primarily at sub-annual timescales (Figure 6). Among all growth forms, wind-dispersed species exhibited strong synchrony at the 6- and 12-month timescales, while animal-

129 dispersed species showed compensatory dynamics at 6 month and 2-year timescales. Tree species dispersed by specific groups of vertebrates (small birds, small vertebrates, or large vertebrates) showed periods of reproductive synchrony at various sub-annual scales (Figure 7).

Figure 6-6: The averaged wavelet modulus ratio for all growth forms (animal, wind, ballistic) in Cocha

Cashu. The number in parenthesis represents the number of species with the dispersal group. The ribbon represents the null distribution generated through bootstrapping (n=1000). Any points that lie above the ribbon was considered significant and synchronous while any points below the ribbon indicated significant, compensatory dynamics.

130

Figure 6-7:The averaged wavelet modulus ratio for the five dispersal groups in tree species for Cocha

Cashu. The number in parenthesis represents the number of species with the dispersal group. The ribbon represents the null-distribution generated through bootstrapping (n=1000). Any points that lie above the ribbon was considered significant and synchronous while any points below the ribbon indicated significant, compensatory dynamics.

The role of temperature and precipitation on community-wide synchrony versus compensatory phenology

There was a significant relationship between the climate variables and community-wide dynamics

(Figure 6, Figure 7). Yasuní had a strongly significant positive temperature-WMR relationships at the 6-7 month scales, indicating twice yearly increases in among-species seed rain synchrony with warming

131 temperatures. At Yasuní, there was a significant positive correlation between WMR and minimum temperature at the 4- 5.8 year scale. The positive coefficients indicate that warmer periods of were correlated with increases in the WMR, i.e. community synchrony. Additionally, we found negative correlation coefficients between the minimum temperature and the WMR at the 3-year periods. At Cocha

Cashu, minimum temperature and precipitation had a significant relationship with community WMR. At the 1.5 year scale, there was a significant negative correlation between the WMR and the minimum temperature, indicating that increasing temperatures were associated with a lower WMR, or more compensatory seed rain dynamics.

(colored in red).

132

Discussion

Communities harbor extensive phenological diversity among member species, especially in tropical wet forests (Frankie et al. 1974). This phenological diversity might help explain species coexistence within communities if phenology is a key axis of niche partitioning among species. That is, if phenological differences reduce competition among species for resources and mutualists involved in reproduction (e.g. frugivores), then these species might be more likely to coexist (Godoy and Levine

2013). However, aspects of phenology remain lesser known dimensions of diversity in ecological communities. Particularly, in hyperdiverse communities, most pairs of species are likely have weak interactions which make it challenging to infer evidence for species interactions in structuring communities. Here, we used flexible spectral analyses, the multivariate wavelet transformation, to show

133 evidence that tree species frequently exhibit significant reproductive synchrony at the community-wide level, but that species sharing dispersal mechanisms or related groups of species often exhibited much stronger compensatory (anti-synchronous) reproductive dynamics.

Do communities exhibit synchronous or compensatory reproduction?

Overall, we found many cases of strong synchronous dynamics in seed rain at the whole community level, and almost no significant whole-community compensatory dynamics. This synchrony suggests whole-community dynamics are largely driven by shared environmental (abiotic or biotic) responses or positive density dependent interactions (van Schaik et al. 1993). By contrast, we previously found whole-community compensatory dynamics of seed rain in a much less diverse tropical dry forest

(Lasky et al. 2016).

At both Yasuní and Cocha Cashu, synchronous seed fall was strongest and most consistent across sites at scales of approximately the 1 month, 1 year, and greater than 3 year periods. The fastest scale of synchrony at Yasuní may represent shared and rapid community responses to brief weather events, perhaps bursts of rainy periods alternating with brief dry periods, like what was observed in a much more seasonally dry forest (Lasky et al. 2016). However, we did not find a correlation between temperature or rainfall and WMR at this scale at Yasuní. Additionally, the synchrony at 1 and >3 year periods at both sites might reflect shared responses to annual or longer time scale oscillations of environmental conditions, given that we found WMR associated with temperature or precipitation fluctuations at these scales. At both sites, the yearly and super-annual patterns of synchrony were consistent throughout the study. Clear yearly fluctuations were visible even in the raw species or seed counts over time, highlighting the strength of synchrony at this scale (Figure 1).

134

At Cocha Cashu, we observed strong non-stationarity in synchrony. We observed a shift from essentially random among-species phenology to synchronous dynamics, across all time scales <1 year in

2007. The ecological explanation for this shift is unclear, but many species decreased reproduction in this period and then subsequently increased to a high level of reproduction (Figure 1). The deep decrease followed a high peak in reproduction might indicate multiple species were accumulating resources synchronously (reducing reproduction) in order to subsequently invest a large amount in reproduction, akin to, but not as dramatic as, bursts of reproduction in masting species (Janzen 1974, Ashton et al.

1988).

Is evidence for phenological niche partitioning strongest among ecologically similar species?

In contrast to our findings at the whole community level, we found widespread significant compensatory dynamics within focal groups of species. Specifically, among confamilials and species with similar animal seed dispersers we often found significant compensatory dynamics. Our findings indicate that community-wide patterns in highly diverse communities may obscure phenological niche partitioning that occurs within groups of closely interacting species. Previous studies that have shown evidence for such partitioning have been largely focused on groups of closely related species hypothesized to be closely interacting due to shared pollinators or mutualists (Gentry 1974, Ashton et al. 1988, Botes et al.

2008). Future efforts might use our analytical approach to identify these groups of species based on non- random phenological patterns instead of relying on substantial prior knowledge of these groups.

Furthermore, two families showed similar patterns of compensatory reproduction for within-year timescales at both sites, Annonaceae and Myristicaceae, perhaps indicating consistency across sites in phenological niche partitioning among these groups.

135 We found opposing patterns of synchrony and compensatory across groups sharing similar dispersers at Cocha Cashu vs Yasuní. At Cocha Cashu, all animal dispersed species showed significant compensatory dynamics within years while subsets of tree species with similar vertebrate dispersers tended to show synchrony within years (Figure 6; Figure 7). At Yasuní, all animal-dispersed plant species showed some synchrony the 3 months scales, and of the animal-dispersed tree species, there was some evidence of compensatory dynamics at the 2-4 year scale (Figure 4; Figure 6).

The ultimate mechanisms leading to compensatory reproduction among species may be due to varying responses to environmental fluctuations, or due to community assembly or coevolutionary processes favoring cooccurrence of species with distinct phenology due to niche partitioning (Robertson

1895, Rathcke and Lacey 1985). When groups of species exhibit compensatory dynamics, it is challenging to determine conclusively from our approach whether it is due to phenological niche partitioning with respect to abiotic or biotic (e.g. frugivores) components of environment. An example of a potentially misleading scenario might be where a group of species share dispersal mechanisms but also other traits, the latter of which cause niche overlap that is ameliorated by phenological niche partitioning.

To show a role conclusively of abiotic or biotic niche partitioning would require additional evidence. For example, Botes and colleagues (2008), showed how compensatory flowering dynamics occurred for Aloe species that deposit pollen on the same location on pollinators’ bodies (suggesting competition or interference in pollination) but not for species depositing pollen on different locations. Wheelwright

(1985) studied fruit removal directly to test whether overlapping phenology of Lauraceae species indicated competition for frugivores and found that it did not. If compensatory phenology of seed rain does reflect biotic niche partitioning, it will still require additional study to determine if compensation is due to seed dispersal processes or processes affecting other stages (e.g. flowering, or seed/seedling survival). For example, Jones and Comita (2010) found negative within-species effects of fruiting density

136 on pre-dispersal seed predation, which could favor compensatory reproduction for species sharing seed predators.

Synchrony among related species or species sharing dispersal mode

We found that among all growth forms, wind-dispersed species exhibit strong synchrony of seed rain, particularly at time scales indicative of shared abiotic niches. At both Yasuni and Cocha Cashu, we found significant 6-month synchrony among wind-dispersed species. This is consistent with the twice- yearly peaks in wind speed observed at a weather station near the site in Cocha Cashu (Figure 4 and

Figure 6). This site has a much stronger peak in wind speed in the September with a small peak in March, also consistent with the synchrony at this site observed among wind-dispersed species at the 1-year periods. Relatedly, we observed synchrony observed among members of Bignoniaceae many of which are wind dispersed. This synchrony among wind-dispersed species during windy seasons has been well- supported by previous studies (Frankie et al. 1974, Janzen 1974, Detto et al. 2018), but here it may be viewed as a partial “positive control” for our approach to detect non-random phenology among species sharing dispersal mechanisms.

We additionally found significant synchrony among members of the same families as well as among specific vertebrate dispersed groups at Cocha Cashu. Several factors may explain this synchrony.

Synchrony may indicate that shared responses to environmental fluctuations drive phenology of trees dispersed by different groups of animals species, or that seasonal fluctuations in biotic interactions select for synchrony among species sharing partners (e.g. frugivores), or alternatively that related species tend to share dispersers and seasonal phenology simply because of niche conservatism (Davies et al. 2013) and a lack of selection for phenological divergence.

137 Community phenology and abiotic fluctuations

We found evidence that community-wide phenology was driven by climate fluctuations, with

WMR being significantly associated with temperature at both sites at multiple scales. Gentry (1974) hypothesized that the phenological (flowering specifically) diversity of tropical plant communities was enabled by the potential year-round reproductive period compared to temperate communities with a narrower range of potential phenologies. We do not see evidence for this based on fluctuations in climate within our two tropical sites. The direction of the WMR-temperature associations and the lack of whole- community WMR contradicts this hypothesis. At both sites, sub-annual WMR was positively associated with temperature, indicating greater synchrony (as opposed to phenological diversity) at warmer times of the year. Our findings may signify that the community-wide trend is for a degree of synchrony to exploit favorable conditions during warmer parts of the year, potentially due to greater light resources (Detto et al. 2018).

Super-annual WMR showed strong associations with temperature at some scales (but not rainfall). For Yasuní, at the > 5 year scales we found higher synchrony in warmer periods, perhaps corresponding to enhanced community synchrony driven by longer range climatic oscillations like the El

Niño Southern Oscillation. At Barro Colorado Island in Panama, (Detto et al. 2018) found community- wide peaks in seed rain during ENSO events, presumably as trees had shared responses to increased light during these dry and warm periods.

138 Conclusion

Here, we showed how whole-community phenology in hyperdiverse plant communities is largely characterized by synchronous reproduction, and to a certain degree in association with warmer temperatures. However, we also uncovered evidence that groups of related or ecologically similar species often show compensatory patterns of seed rain, indicating potential phenological axes of niche partitioning that might promote species coexistence. Our results highlight the scale-specific and sometimes non-stationary characteristics of community phenology. Flexible multi-scale analyses may help researchers characterize evidence of scale-specific niche partitioning and environmental filtering.

Acknowledgements

Work at Yasuní was supported by funding to NCG and collaborators from the Andrew W. Mellon

Foundation, Natural Environment Research Council (GR9/04037), British Airways, Department of

Botany, Natural History Museum, and the National Science Foundation (DEB-0614525, DEB-1122634,

DEB-1754632, DEB- 1754668). We thank the Ecuadorian Ministerio del Ambiente for permission to work in Yasuní National Park (under No 012-2018-IC-PNY-DPAO/AVS, No 008-2017-IC-PNY-

DPAO/AVS, No. 012–2016-IC-FAU-FLO-DPAO-PNY, No. 014-2015-FLO-MAE-DPAO-PNY, and earlier permits). We very gratefully thank Milton Zambrano for collecting most of the trap data from

2002-2017. We also thank Viveca Persson for help initiating the censuses in 2000-2002, with assistance from Zornitza Aguilar, Paola Barriga and Matt Priest, and Gorky Villa, Alvaro Perez and Pablo Aliva for help identifying species. The findings and conclusions do not necessarily reflect the view of the funding agency.

139

Appendix A

Supplementary Information for Chapter 2

Min Temp Max Temp Avg. Temp

Month Correlation p Correlation p Correlation p

September 0.12 0.45 -0.11 0.50 0.07 0.66

October 0.12 0.45 0.49 0.001** 0.32 0.04*

November 0.19 0.25 0.23 0.16 0.23 0.17

December 0.48 0.002** 0.58 0.0001*** 0.55 0.0004***

January 0.25 0.12 0.34 0.03* 0.30 0.06.

February 0.01 0.01 -0.04 0.78 0.14 0.38

March 0.004 0.97 0.03 0.81 0.18 0.27

April 0.25 0.13 0.015 0.92 0.18 0.28

Table S1: The Pearson correlation between the monthly NAO indices and temperatures (minimum, maximum, average) across 1980 to 2016.

140

Min Temp Max Temp Avg Temp

Month Correlation p Correlation p Correlation p

September 0.26 0.10. -0.001 0.99 0.149 0.37

October -0.06 0.71 0.33 0.04* 0.13 0.13

November 0.33 0.04* 0.35 0.02* 0.37 0.02*

December 0.38 0.01* 0.54 0.00049*** 0.48 0.002***

January 0.25 0.12 0.34 0.03* 0.30 0.07.

February 0.01 0.91 0.28 0.09 0.16 0.33

March 0.01 0.95 0.24 0.14 0.14 0.38

April 0.29 0.08. 0.21 0.20 0.27 0.10.

Table S2: The Pearson correlation between the monthly AO indices and temperatures (minimum, maximum, and average) across 1980 to 2016.

141

Figure S1: Variable Importance of the Projection scores of the daily minimum temperatures, NAO, and

AO indices for the three bivoltine species.

142

Figure S2: Variable Importance of the Projection scores of the daily minimum temperatures, NAO, and

AO indices for the two multivoltine species.

143

Appendix B

Supplementary Information for Chapter 3

1. The full mathematical model of Cydia pomonella

The mathematical formulation for Chapter 3 use renewal equations to describe the physiological, age-structured model (see Breda et al. 2012 and Bjornstad et al. 2016). The model tracks the individuals at these life-stages at time t: egg 퐸(푡) , the larval instars 퐿1(푡), 퐿2(푡), 퐿3(푡), 퐿4(푡), 퐿5(푡), the diapause stage 1 퐷1(푡) , the diapause stage 2 퐷2(t), the reproductive adults 퐴푟(t), and finally the senescent adults

퐴푠(t).

Individuals flow in and out of the each stage through development and mortality rates that are dependent on temperature T(t) at time 푡. Time-dependent per-capita adult rate is denotated by B(t) and the per capita mortality rate of stage 푖 at time 푡 is given denoted by δ푖(푡). The only rate not dependent on temperature is the diapause induction rate 퐼(푡) in stage 퐿5 which is assumed to depend on photoperiod. Finally, μ푖(푡) is the thorough-stage flux into the next stage (senescence for the adults) and is calculated from the probability density of the individual at some τ days ago ℎ푖(τ) which is the probability density of the individual at some 휏 days ago have matured to the next stage (Equation 1). and the the cumulative distribution function 퐻푖(τ) according to

ℎ푖(휏) 휇푖(휏) = 1 − 퐻푖(휏)

The total flux of individuals that enter stage 푖 at time 푡 depends on the number of individuals that entered the previous stage 푖 − 1 at time 푡 − 휏 and survived to time 푡 to mature out according to the developmental probability distribution. Likewise, the total individuals that leave stage 푖 to be recruited into the next stage 푖 + 1 are those who entered stage 푖 at time t- 휏 and survived to 푡 as weighted by the on-

144 ward development probability distribution. At any one time the flux of individuals leaving and entering a stage is based on integrating across 휏.

The general form of the renewal equation with 푁푖(푡) being the population density of stage 푖 at time 푡 is:

∞ ∞ 푑푁푖(푡) ℎ푖−1(휏) ℎ푖(휏) = ∫ 푁푖−1(푡 − 휏)푒푥푝(−훿푖−1휏) 푑휏 − ∫ 푁푖(푡 − 휏)푒푥푝(−훿푖휏) 푑휏 − 훿푖푁푖(푡) 푑푡 ⏟0 1 − 퐻 푖− 1 ( 휏 ) ⏟0 1 − 퐻 푖 (휏 ) 퐼푛푓푙표푤 푓푟표푚 푡ℎ푒 푛푒푥푡 푠푡푎푔푒 푂푢푡푓푙표푤 푡표 푡ℎ푒 푛푒푥푡 푠푡푎푔푒

With this general form in mind, we can describe the flow of individuals in and out of the insect model as below:

∞ 푑퐸(푡) ℎ퐸(휏) (2) = 푏(푡)퐴(푡) − ∫ 퐸(푡 − 휏)푒푥푝(−훿퐸휏) 푑휏 − 훿퐸퐸(푡) 푑푡 0 1 − 퐻퐸(휏)

푑퐿1(푡) ∞ ℎ퐸(휏) ∞ ℎ퐿1(휏) (3) = ∫ 퐸(푡 − 휏)푒푥푝(−훿퐸휏) 푑휏 − ∫ 퐿1(푡 − 휏) 푒푥푝(−훿퐿휏) 푑휏 − 훿퐿(푡)퐿1(푡) 푑푡 0 1−퐻퐸(휏) 0 1−퐻퐿1(휏)

ℎ (휏) ℎ (휏) 푑퐿푛(푡) ∞ 퐿푛−1 ( ) ∞ 퐿푛 ( ) ( ) (4) = ∫0 퐿푛−1(푡 − 휏) 푒푥푝 −훿퐿휏 푑휏 − ∫0 퐿푛(푡 − 휏)푒푥푝 −훿퐿휏 푑휏 − 훿퐿(푡)퐿푛 푡 푑푡 1−퐻퐿푛−1(휏) 1−퐻퐿푛(휏)

푑퐿 (푡) ∞ ℎ (휏) ∞ ℎ (휏) (5) 5 = 퐿4 퐿 (푡 − 휏)푒푥푝(−푢 휏) 푑휏 − 퐿5 퐿 (푡 − 휏)푒푥푝(−푢 휏) 푑휏 − 푢 (푡)퐿 (푡) − 퐼(t)L (t) ∫0 ( ) 4 퐿 ∫0 ( ) 5 퐿 퐿 5 5 푑푡 1−퐻퐿4 휏 1−퐻퐿5 휏

푑퐷1(푡) ∞ ℎ퐷1(휏) (6) = 퐼(t)L5(t) − ∫ 퐷1(푡 − 휏)푒푥푝(−푢퐷휏) 푑휏 − 푢퐷(푡)퐷1(푡) 푑푡 0 1−퐻퐷1(휏)

푑퐷 (푡) ∞ ℎ (휏) ∞ ℎ (휏) (7) 2(푡) = 퐷1 퐷 (푡 − 휏)푒푥푝(−푢 휏) 푑휏 − 퐷2 퐷 (푡 − 휏)푒푥푝(−푢 휏) 푑휏 − 푢 (푡)퐷 (푡) ∫0 ( ) 1 퐷 ∫0 ( ) 2 퐷 퐷 2 푑푡 1−퐻퐷1 휏 1−퐻퐷2 휏

푑푃(푡) ∞ ℎ (휏) ∞ ℎ (휏) (8) = 퐿5 퐿 (푡 − 휏)푒푥푝(−푢 휏) 푑휏 − 푃 푃(푡 − 휏)푒푥푝(−푢 휏) 푑휏 − 푢 푃(푡) ∫0 ( ) 5 푃 ∫0 ( ) 푃 푃 푑푡 1−퐻퐿5 휏 1−퐻푃 휏

( ) ( ) ℎ (휏) 푑퐴푟 푡 ∞ ℎ푃 휏 ( ) ( ) ∞ 퐴푟 ( ) ( ) ( ) ( ) (9) = ∫0 푃 푡 − 휏 푒푥푝 −푢푃휏 푑휏 − ∫0 퐴푟 푡 − 휏 푒푥푝 −푢퐴휏 푑휏 − 푢퐴 푡 퐴 푡 푑푡 1−퐻푃(휏) 1−퐻퐴푟(휏)

( ) ℎ (휏) 푑퐴푠 푡 ∞ 퐴푟 ( ) ( ) ( ) ( ) (10) = ∫0 퐴푟 푡 − 휏 푒푥푝 −푢퐴휏 푑휏 − 푢퐴 푡 퐴푠 푡 푑푡 1−퐻퐴푟(휏)

145

The development rate and linear chain trickery

Generally, mathematical models using ordinary differential equations assume that the dwelling time in any stage to be exponentially distributed such that individuals can mature out of their stage at a constant rate. To incorporate biological realism of through-stage developmental delays with individual variability, we can gain computational tractability by assuming an Erlang probability distribution, a special case of the gamma distribution with an integer shape parameter 푘 and the rate parameter 휆. When

푘 = 1, the Erlang distribution reduces to an exponential form and as 푘 goes to infinity, the Erlang distribution becomes a Dirac function which then turns the distributed delay integrodifferential equation into a DDE delay differential equation. The most biologically realistic form lies between the exponential

(ODE) and the Dirac (DDE) model (Figure S1).

Figure S1: The probability distribution of the developmental times assuming an Erlang distribution with the mean age of developing out of a stage to be age 50. When 푘 = 1, the probability distribution becomes exponential (in purple). As 푘 → ∞ approaches infinity, the Erlang distribution approaches the Dirac function (in black).

146

Under the assumption of Erlang distributed development times, equations 2-10 can be solved using a system of ordinary differential equations through the so-called `gamma-chain’ or ‘linear-chain trickery’ (Figure S2) due to a special property of the Erlang distribution. Specifically, an Erlang distribution can be described as the summation of 푘 i.i.d exponential distributions. From the example below, we first assume that 퐴, the mean duration time in state 푋 to be Erlang distributed. Individuals must go through a series of subcompartments (푋1, 푋2, … 푋푘) with each subcompartment having an exponential

퐴 distribution with the average sojourn time as . The individuals flow through each subcompartments at a 푘

푘 rate of ( ) to ensure the desired mean and variance transitions are assured. 퐴

Figure S2: An illustration describing the ‘linear chain trick’

147

The full Cydia pomonella model with parameter estimates

This is the full gamma-chain model used in the model. The full model has all five larval instars. Here

퐵(푡) represents the birth rate, 휇푘(푡)represents the stage-specific instantaneous development rate

(senescent rate for reproductive adults and diapause ‘development rate’ for diapausing larvae). 훿푘(푡) is the stage-specific instantaneous mortality rate. 푛푘represents the sub-stages for each of the life-stage.

1 푑퐸 (푡) 퐵(푡) 퐴(푡) − 푛 휇 (푡)퐸 (푡) − 훿 (푡)퐸 (푡) 푖푓 푖 = 1 (11) 푖 = {2 퐸 퐸 1 퐸 1 푑푡 푛퐸휇퐸(푡)퐸푖−1(푡) − 푛퐸휇퐸(푡)퐸푖(푡) − 훿퐸(푡)퐸푖(푡) 푖푓 푖 > 1

푑퐿1 (푡) 푛 휇 (푡)퐸 (푡) − 푛 휇 (푡)퐿1 (푡) − 훿 (푡)퐿1 (푡) 푖푓 푖 = 1 (12) 푖 = { 퐸 퐸 푛퐸 퐿1 퐿1 1 퐿 1 푑푡 푛퐿1휇퐿1(푡)퐿1푖−1(푡) − 푛퐿1휇퐿1(푡)퐿1푖(푡) − 훿퐿(푡)퐿1푖 (푡) 푖푓 푖 > 1

푑퐿2 (푡) 푛 휇 (푡)퐿1 (푡) − 푛 휇 (푡)퐿2 (푡) − 훿 (푡)퐿2 (푡) 푖푓 푖 = 1 (13) 푖 = { 퐿1 퐿1 푛퐿1 퐿2 퐿2 1 퐿 1 푑푡 푛퐿2휇퐿2(푡)퐿2푖−1(푡) − 푛퐿2휇퐿2(푡)퐿2푖(푡) − 훿퐿(푡)퐿2푖 (푡) 푖푓 푖 > 1

푑퐿3 (푡) 푛 휇 (푡)퐿2 (푡) − 푛 휇 (푡) 퐿3 (푡) − 훿 (푡)퐿3 (푡) 푖푓 푖 = 1 (14) 푖 = { 퐿2 퐿2 푛퐿2 퐿3 퐿3 1 퐿 1 푑푡 푛퐿3휇퐿3(푡)퐿3푖−1(푡) − 푛퐿3휇퐿3(푡)퐿3푖(푡) − 훿퐿(푡)퐿3푖 (푡) 푖푓 푖 > 1

푑퐿4 (푡) 푛 휇 (푡)퐿3 (푡) − 푛 휇 (푡)퐿4 (푡) − 훿 (푡)퐿4 (푡) 푖푓 푖 = 1 (15) 푖 = { 퐿3 퐿3 푛퐿3 퐿4 퐿4 1 퐿 1 푑푡 푛퐿4휇퐿4(푡)퐿4푖−1(푡) − 푛퐿4휇퐿4(푡)퐿4푖(푡) − 훿퐿(푡)퐿4푖(푡) 푖푓 푖 > 1

푑퐿5 (푡) 푛 휇 (푡)퐿4 (푡) + 푛 휇 퐷퐿 (푡) − 푛 휇 (푡)퐿5 (푡) − 퐼(푡) 퐿5 (푡) − 훿 (푡)퐿5 (푡) 푖푓 푖 = 1 (16) 푖 = { 퐿4 퐿4 푛퐿4 퐷퐿 퐷퐿 푛퐷퐿 퐿5 퐿5 1 1 퐿 1 푑푡 푛퐿5휇퐿5(푡)퐿5푖−1(푡) − 푛퐿5휇퐿5(푡)퐿5푖(푡) − 퐼(푡) 퐿5푖(푡) − 훿퐿(푡)퐿5푖 (푡) 푖푓 푖 > 1

푑퐷퐿 (푡) 퐼(푡)퐿5(푡) − 푛 휇 (푡)퐷퐿 (푡) − 훿 (푡)퐷퐿 (푡) 푖푓 푖 = 1 (17) 푖 = { 퐷퐿 퐷퐿 1 퐷퐿 1 푑푡 푛퐷퐿휇퐷퐿(푡)퐷퐿푖−1(푡) − 푛퐷퐿휇퐷퐿(푡)퐷퐿푖 (푡) − 훿퐷퐿(푡)퐷퐿푖 (푡) 푖푓 푖 > 1

푑푃 (푡) 푛 휇 (푡)퐿5 (푡) − 푛 휇 (푡)푃 (푡) − 훿 (푡)푃 (푡) 푖푓 푖 = 1 (18) 푖 = { 퐿5 퐿5 푛퐿5 푃 푃 1 푃 1 푑푡 푛푃휇푃(푡)푃푖−1(푡) − 푛푃휇푃(푡)푃푖(푡) − 훿푃(푡)푃푖 (푡) 푖푓 푖 > 1

푑퐴푟 (푡) 푛 휇 (푡)푃 (푡) − 푛 휇 (푡)퐴푟 (푡) − 훿 (푡)퐴푟 (푡) 푖푓 푖 = 1 (19) 푖 = { 푃 푃 푛푃 퐴푟 퐴푟 1 퐴 1 푑푡 푛퐴푟휇퐴푟 (푡)퐴푟푖−1 − 푛퐴푟휇퐴푟 (푡)퐴푟푖(푡) − 훿퐴(푡)퐴푟푖 (푡) 푖푓 푖 > 1

푑퐴푠(푡) (20) = 푛 휇 (푡)퐴푟 (푡) − 훿 (푡)퐴푠(푡) 푑푡 퐴푟 퐴푟 푛퐴푟 퐴

148

2. Parameter value for the model

Function Param Value Description eters 퐵퐸 10.52 Maximum birth rate at the optimum temperature 푇 24.62 Optimum temperature (°C) for the maximum birth rate Birth Rate 표푝푡 휎퐸 3.42 Variability about the optimum temperature C* 6.12 ∗ 10−3 Allele constant 푀퐸 0.19 Maximum development rate for Egg 훽퐸 0.30 Scalar determining the steepness for Egg 푇퐸 18.59 The temperature (°C) at which inflection occurs for Egg 푀퐿1 0.31 Maximum development rate for Instar 1 훽퐿1 0.33 Scalar determining the steepness for Instar 1 푇퐿1 17.60 The temperature (°C) at which inflection occurs for Instar 1 푀퐿2 0.45 Maximum development rate for Instar 2 훽퐿2 0.23 Scalar determining the steepness for Instar 2 푇퐿2 21.24 The temperature (°C) at which inflection occurs for Instar 2 푀퐿3 0.33 Maximum development rate for Instar 3 훽퐿3 0.35 Scalar determining the steepness for Instar 3 푇 18.46 The temperature (°C) at which inflection occurs for Instar 3 Dev. Rate 퐿3 푀퐿4 0.30 Maximum development rate for Instar 4 훽퐿4 0.43 Scalar determining the steepness for Instar 4 푇퐿4 21.54 The temperature (°C) at which inflection occurs for Instar 4 푀퐿5 0.33 Maximum development rate for Instar 5 훽퐿5 0.21 Scalar determining the steepness for Instar 5 푇퐿5 20.95 The temperature (°C) at which inflection occurs for Instar 5 푀푃 0.09 Maximum development rate for Pupae 훽푃 0.28 Scalar determining the steepness for Pupae 푇푃 18.48 The temperature (°C) at which inflection occurs for Pupae 푀퐴.푟 0.13 Maximum senescent rate for R. Adults 훽퐴.푟 0.18 Scalar determining the steepness for R. Adults 푇퐴.푟 20.22 The temperature (°C) at which inflection occurs for R. Adults 푑퐸 2.06 Fitted scalar for Egg −4 푏퐸 1.31 ∗ 10 Fitted scalar for Egg 푇푑퐸 13.57 The optimum temperature (°C) of survival for Egg 푑퐿* 5.00 Fitted scalar for Larvae −4 푏퐿* 4.02 ∗ 10 Fitted scalar for Larvae 푇푑 16.24 The optimum temperature (°C) of survival for Larvae Mortality Rate 퐿 푑푃 2.86 Fitted scalar for Pupae −5 푏푃 1.04∗ 10 Fitted scalar for Pupae 푇푑푝 5.46 The optimum temperature (°C) of survival for Pupae 푑퐴푟 2.86 Fitted scalar for R. Adults −5 푏퐴푟 1.04∗ 10 Fitted scalar for R. Adults 푇푑퐴푟 20.00 The optimum temperature (°C) of survival for R. Adults 푑퐷퐿* 3.09 Fitted scalar for Diapausing Larvae −4 푏퐷퐿 * 3.21 ∗ 10 Fitted scalar for Diapausing Larvae 푇푑퐷퐿* 3.72 The optimum temperature (°C) of survival for Diapausing Larvae 푛퐸 50 Number of sub-compartment for Egg Compartments 푛퐿1 30 Number of sub-compartment for Instar 1 푛퐿2 30 Number of sub-compartment for Instar 2

149

푛퐿3 30 Number of sub-compartment for Instar 3 푛퐿4 30 Number of sub-compartment for Instar 4 푛퐿5 30 Number of sub-compartment for Instar 5 푛푃 15 Number of sub-compartment for Pupae 푛퐴.푟 25 Number of sub-compartment for R. Adults 푛퐷1 25 Number of sub-compartment for the Diapausing Larvae 1 푛퐷2 5 Number of sub-compartment for the Diapausing Larvae 2 퐼푎 4.6 Maximum diapause induction rate Diapause 퐼푏 0.50 Scalar determining the steepness for diapause induction Induction 퐼푐 230 The day of year at which inflection occurs for diapause induction Diapause 푀퐷1 0.04 Maximum diapause termination rate for Diapausing Larvae 1 Termination 훽퐷1 0.30 Scalar determining the steepness for Diapausing Larvae 1 푇퐷1 -2.0 The temperature (°C) at which inflection occurs for Diapausing Larvae 1 푀퐷2* 0.03 Maximum diapause termination rate Diapausing Larvae 2 훽퐷2 0.5 Scalar determining the steepness for Diapausing Larvae 2 푇퐷2* 0.04 The temperature (°C) at which inflection occurs for Diapausing Larvae 2

Table S1: The full parameters for the Cydia pomonella model. Stars represent parameters estimated through maximum-likelihood.

3. Multiple autoregressive state systems (MARSS)

To combine the time series of multiple pheromone traps, we fitted a Multivariate Autoregressive State-

Space Modeling (MARSS) to the time series (specifically, see Chapter 15 of Holmes et al. 2012). The sampled blocks varied at the FREC across the years and no single block was sampled continuously from

1984 to 2016. Due to the gaps in data due to winter, we fitted the MARSS separately for each year. We only fitted the MARSS model for years when there were three or more blocks. For years with only two traps, we found that one trap had more consistent catches while the other trap had very few moths captured (see: 1985-1998 excluding 1986 in Figure S3).

The general MARSS model is shown below with a hidden state process (풙) and an observation (풚) model:

(1) 푥푡 = 퐵푡푥푡−1 + 푢푡 + 푤푡 푤ℎ푒푟푒 푤푡~ 푀푉푁(0, 푄)

(2) 푦푡 = 푍푥푡 + 푎 + 푣푡 푤ℎ푒푟푒 푣푡~푀푉푁(0, 푅)

150

Here 퐵, 푍, 푢, 푎, 푄, 푅 푎푛푑 푢 are the parameters that are estimated through maximum likelihood. The true population size is represented by the state 푥푡 with B describing the interactions between the different processes and 푢 describing the mean trend. Assuming there are 푚 traps per year, we first test if the time series can be combined in that there is one state process model to describe the population. We first we modify the matrix 푍 which is a 푚 푥 1 matrix with each element describing which individual block time series is associated with what trajectory. For example, if there are three blocks (A, B, C), we can test multiple hypotheses: the three blocks can be represented by one state-process model, the three blocks have to be represented by three state process models, A and B are more similar to each other and can be represented by one state process model while C is different and must be described with another model, etc.

We assume that all the blocks have the same observation variance and that the errors are i.i.d

( 풗 ~ 푀푉푁 (0, 푹)). Therefore, we assume that 푹 is a 푚 푥 푚 diagonal matrix with equal variance 푟.

푦1 1 0 푣1 푟 … 0 (3) [ ⋮ ] = [⋮] 푥푡 + [ ⋮ ] + [ ⋮ ] 푣푡~푀푉푁 (0, [⋮ ⋱ ⋮] ) 푦푚 1 푎푚 푣3 0 … 푟

Finally, the 푎 are the scaling parameters and represent the bias between the block and the total population.

The first element in the 푚 푥 1 푚푎푡푟푖푥 will be fixed at 0 and the multiple time series are scaled against each other. After creating different hypotheses, we chose the model with the lowest AICc. We are then able to get the estimated underlying states and if there are more than two states, we chose the one with more sites across 33 years.

151

Figure S3: Log-transformed time series of adult Cydia pomonella captured in pheromone traps across

152

4. Circular variance

Circular variance is a useful measure of dispersion in periodical time series. The important summary statistics consist of angular observations 휃1 … 휃푛 where n represents the total number of individuals. The mean resultant length 푅̅ is the average vector length when individuals are distributed on a circle.

1 (4) 푅̅ = √( ∑푛 cos 휃 )2 + ( ∑푛 sin 휃 )2 푛 푖=1 푖 푖=1 푖

The circular variance is then calculated as

(5) 푉 = 1 − 푅̅

Calculating the circular variance assumes that we know where the individuals are in the state of development, yet in our simulation we only know what stage and subcompartment they are in. We assume then that the number of individuals is evenly distributed in each subcompartment. We modify equation 4 then by integrating over the integral of each subcompartment

̅ 1 ∑푘 휃푖 2 ∑푘 휃푖 2 (6) 푅 = 푘 √( 푖=1 ℎ푖 ∫휃 −1 cos 푥 푑푥) + ( 푖=1 ℎ푖 ∫휃 −1 sin 푥 푑푥) ∑푖=1 푦푖 푖 푖

ℎ푖is the mean density at each subcompartment (푖) with 푦푖 being the total number of individuals at the subcompartments

푦푖 (7) ℎ푖 = 휃푖−휃푖−1

By taking the integral, the final form of this equation is

̅ 1 푘 2 푘 2 (8) 푅 = 푘 √( ∑푖=1 ℎ푖 (cos (휃푖 − 휃푖−1 )) + ( ∑푖=1 ℎ푖(sin(휃푖 − 휃푖−1 )) ∑푖=1 푦푖

153

5. Model validation for Cydia pomonella

Figure S4: The Real Mean Squared Error and the Pearson correlation between model output and the field-trap data.

154

6. Data on the vital rates of Cydia pomonella

Temperature Oviposition Period Mean eggs Daily egg rate Source (°C) (Days) (lifetime) (1/day) 10.00 0.00 0.0 0.0 (Aghdam et al. 2009a) 14.00 25.38 0.0 0.0 (Aghdam et al. 2009a) 20.00 11.23 48.85 4.35 (Aghdam et al. 2009a) 25.00 8.75 89.25 10.29 (Aghdam et al. 2009a) 27 7.69 66.32 8.62 (Aghdam et al. 2009a) 30 6.64 20.18 3.04 (Aghdam et al. 2009a) 33 6.50 0.0 0.0 (Aghdam et al. 2009a) 35 0.00 0.0 0.0 (Aghdam et al. 2009a) Table S2: The average per capita birth rate of Cydia pomonella female adults at different constant temperatures

155

Dev. Dev. Temperature duration rate Source (°C) (Days) (1/Days) 10.00 NA 0.00 Penn State Extension 13.90 22.70 0.044 (Howell and Schmidt 2002) 14.00 18.67 0.050 (Aghdam et al. 2009b) 14.00 18.67 0.050 (Aghdam et al. 2011) 14.80 17.70 0.056 (Howell and Schmidt 2002) 15.00 19.39 0.052 (Blomefield and Giliomee 2009) 16.44 14.00 0.071 (Glenn 1922) 17.28 12.67 0.079 (Glenn 1922) 18.30 10.66 0.094 (Glenn 1922) 19.71 9.35 0.107 (Glenn 1922) 20.00 9.19 0.109 (Blomefield and Giliomee 2009) 20.00 9.34 0.107 (Aghdam et al. 2009b) 20.00 10.00 0.100 (Aghdam et al. 2011) 20.10 8.70 0.115 (Howell and Schmidt 2002) 20.50 8.78 0.115 (Glenn 1922) 21.85 7.72 0.130 (Glenn 1922) 22.84 7.00 0.143 (Glenn 1922) 23.81 6.60 0.152 (Glenn 1922) 25.00 5.75 0.174 (Glenn 1922) 25.00 4.80 0.208 (Aghdam et al. 2009b) 25.23 6.12 0.163 (Glenn 1922) 25.5 5.30 0.189 (Howell and Schmidt 2002) 25.95 5.95 0.168 (Glenn 1922) 26.74 5.71 0.175 (Glenn 1922) 28.25 5.52 0.181 (Glenn 1922) 28.88 5.53 0.181 (Glenn 1922) 29.6 4.40 0.227 (Howell and Schmidt 2002) 30 4.04 0.248 (Aghdam et al. 2009b) 30 4.23 0.236 (Blomefield and Giliomee 2009) 33 4.19 0.239 (Aghdam et al. 2009b) 34 NA 0.000 (Aghdam et al. 2009b)

Table S3: Development duration and daily development rate of eggs at different constant temperatures.

156

Development Temperature Development Rate Instar Duration Source (°C) (1/Days) (Days 10.00 NA 0.000 (Setyobudi 1989) 20.00 4.20 0.238 (Williams and McDonald 1982) 1st Instar 25.00 4.20 0.238 (Williams and McDonald 1982) 25.00 4.80 0.208 (Williams and McDonald 1982) 30.00 2.70 0.370 (Williams and McDonald 1982) 35.00 3.50 0.286 (Williams and McDonald 1982) 9.60 NA 0.000 (Setyobudi 1989) 20.00 4.20 0.238 (Williams and McDonald 1982) 25.00 4.40 0.227 (Williams and McDonald 1982) 2nd Instar 25.00 3.00 0.333 (Williams and McDonald 1982) 30.00 2.10 0.476 (Williams and McDonald 1982) 35.00 2.50 0.400 (Williams and McDonald 1982) 12.00 NA 0.000 (Setyobudi 1989) 20.00 4.10 0.244 (Williams and McDonald 1982) 3rd Instar 25.00 3.40 0.294 (Williams and McDonald 1982) 25.00 3.70 0.270 (Williams and McDonald 1982) 30.00 3.20 0.313 (Williams and McDonald 1982) 35.00 2.60 0.385 (Williams and McDonald 1982) 14.23 NA 0.000 (Setyobudi 1989) 20.00 8.00 0.125 (Williams and McDonald 1982) 4th Instar 25.00 4.50 0.222 (Williams and McDonald 1982) 30.00 2.10 0.476 (Williams and McDonald 1982) 35.00 5.90 0.169 (Williams and McDonald 1982) 11.14 NA 0.000 (Setyobudi 1989) 20.00 5.30 0.189 (Williams and McDonald 1982) 5th Instar 25.00 5.90 0.169 (Williams and McDonald 1982) 30.00 3.60 0.278 (Williams and McDonald 1982) 35.00 3.00 0.333 (Aghdam et al. 2009b)

Table S4: Average development duration and development rate of Cydia pomonella larval instars at constant temperatures.

157

Temperature Average development Development rate (°C) duration Source (1/Day) (Day) 9.82 NA 0 (Setyobudi 1989) 11.44 45.50 0.022 (Glenn 1922) 13.16 35.20 0.028 (Glenn 1922) 13.41 34.01 0.029 (Glenn 1922) 14.72 29.77 0.034 (Glenn 1922) 14.80 53.70 0.019 (Glenn 1922) 15.00 56.25 0.018 (Blomefield and Giliomee 2009) 17.00 37.90 0.026 (Blomefield and Giliomee 2009) 20.00 21.90 0.046 (Williams and McDonald 1982) 20.00 27.48 0.036 (Williams and McDonald 1982) 20.10 23.30 0.043 (Howell and Schmidt 2002) 20.71 13.80 0.072 (Glenn 1922) 21.55 12.70 0.079 (Glenn 1922) 22.83 11.50 0.087 (Glenn 1922) 23.84 10.73 0.093 (Glenn 1922) 24.87 10.02 0.100 (Glenn 1922) 25.00 14.90 0.067 (Williams and McDonald 1982) 25.00 18.80 0.053 (Williams and McDonald 1982) 25.50 15.30 0.065 (Howell and Schmidt 2002) 26.18 9.44 0.106 (Glenn 1922) 27.15 9.43 0.106 (Glenn 1922) 28.15 9.24 0.108 (Glenn 1922) 29.60 13.50 0.074 (Howell and Schmidt 2002) 30.00 10.70 0.093 (Setyobudi 1989) 30.00 13.78 0.073 (Williams and McDonald 1982) 35.00 12.40 0.081 (Williams and McDonald 1982)

Table S5: Average development duration and development rate of Cydia pomonella pupae at constant temperatures from different sources.

158

Temperature Dev. Duration Dev. Rate Source (°C) (Day) (1/Day) 14 25.38 0.039 (Aghdam et al. 2009b) 20 11.23 0.089 (Aghdam et al. 2009b) 25 8.75 0.114 (Aghdam et al. 2009b) 30 6.64 0.151 (Aghdam et al. 2009b) 33 6.50 0.154 (Aghdam et al. 2009b)

Table S6: Average development duration and development rate of Cydia pomonella reproductive adults at constant temperatures.

159

Life- Temperature Duration Per Capita Survivorship Source stage (°C) (Days) Mortality -1.00 23 0.035 0.063 Other 0.00 7.00 0.134 0.287 (Yokoyama et al. 1987) 0.10 8.75 0.114 0.004 (Moffitt and Burditt 1989) 13.90 22.70 0.591 0.023 (Howell and Neven 2000) Egg 14.80 17.70 0.843 0.009 (Howell and Neven 2000) 20.10 8.70 0.743 0.034 (Howell and Neven 2000) 25.50 5.30 0.495 0.130 (Howell and Neven 2000) 29.60 4.40 0.370 0.220 (Howell and Neven 2000) 34.40 1.00 0.50 1.000 (Chidawanyika and Terblanche 2011) -25.00 1.00 0.020 1.000 (Khani and Moharramipour 2010) -20.00 1.00 0.001 6.900 Other -15.00 1.00 0.080 2.520 Other -10.00 1.00 0.980 0.020 (Khani and Moharramipour 2010) -1.00 21.00 0.113 0.103 Other 0.00 28.00 0.920 0.029 (Neven 2013a) Larvae 5.00 28.00 0.780 0.008 (Neven 2013b) 14.80 53.30 0.620 0.008 (Howell and Neven 2000) 20.10 26.30 0.750 0.010 (Howell and Neven 2000) 25.50 16.30 0.780 0.015 (Howell and Neven 2000) 29.60 15.30 0.590 0.034 (Howell and Neven 2000) 35.00 15.70 0.180 0.109 (Howell and Neven 2000) 14.80 53.70 0.747 0.005 (Howell and Neven 2000) 20.10 23.30 0.853 0.006 (Howell and Neven 2000) Pupae 25.5 15.30 0.947 0.003 (Howell and Neven 2000) 29.60 13.50 1.000 0.000 (Howell and Neven 2000) 35.00 0.00 0.000 1.000 (Howell and Neven 2000) 19.00 14.62 0.996 0.0002 (Chen et al. 2019) 22.00 9.57 0.994 0.0005 (Chen et al. 2019) 25.00 7.28 0.954 0.006 (Chen et al. 2019) Pupae 28.00 7.52 0.795 0.030 (Chen et al. 2019) (OFM) 31.00 6.96 0.596 0.074 (Chen et al. 2019) 12.70 129.8 0.760 0.002 (Chen et al. 2019) 15.10 80.7 0.825 0.002 (Graf et al. 2018) D. 17.60 60.90 0.818 0.003 (Graf et al. 2018) Larvae 19.90 46.50 0.802 0.004 (Graf et al. 2018) 22.40 37.00 0.793 0.006 (Graf et al. 2018) 25.10 29.10 0.858 0.005 (Graf et al. 2018)

Table S7: Instantaneous mortality rate (per day) for each of the life stages

160

6. Full model code of Cydia pomonella

“Pomp” is a R package that is useful for writing models in C using Csnippets. It is also easy to implement trajectory matching (King et al. 2016). Here is the underlying code and can be easily retrieved from

Github: https://github.com/pakdamie/codlingmothdiapause

1. library(pomp2) #ENSURE THAT YOU ARE USING POMP2 NOT POMP 2. #library(here) 3. library(ggplot2) #This is ensuring for the plotting 4. library(dplyr) #This is to match up the original data and the simulation 5. library(viridis) 6. 7. ###THE COVARIATES THAT INCLUDE THE DAILY TEMPERATURE (TT), and 8. ###THE DIFFERENCE IN PHOTOPERIOD (PP) AND THE REAL TRAP DATA (ACT) 9. 10. 11. ###NOTE: the original covar has the averaged trap data when it's more 12. ### appropriate to use the MARSS time series 13. 14. ###Just load this RData file that has the data.frame- 15. ###originally, the covar's ARC is the averaged abundance and NOT 16. ###the MARSS Data, which must be loaded separately 17. 18. load(here("Data","covariates","covar.RData")) 19. 20. ###This is the MARSS abundance- must load! 21. load(here("Data","covariates","NEW_CM_ABUNDANCE.RData")) 22. 23. ###New pomp packages require you to use the covariate-table- so run this 24. 25. ###We replace the old ARC wih the new ARC 26. covartable$ARC <- NEW_CM_ABUNDANCE$Abund 27. COVAR <- covariate_table(covartable,times = 'time') 28. 29. ########################################################################## 30. ###MORE FOR FITTING PURPOSES############################################## 31. ########################################################################## 32. 33. 34. ###covartable_fit <- covartable[1:4018,] 35. ###COVAR_FIT <- covariate_table(covartable_fit,times = 'time') 36. 37. 38. ############################################################# 39. #################THE POMP MODEL BELOW ####################### 40. ############################################################# 41. 42. 43. ###THE NUMBER OF SUBCOMPARMENTS FOR EACH LIFE-STAGE## 44. 45. ###Remember as the number of compartment increases- the variability 46. ###in development rate decreases and vice versa.

161

47. 48. 49. nE <- 50 #Egg 50. nL1 <-30 #First Instar Larvae 51. nL2 <- 30 #Second Instar Larvae 52. nL3 <-30 #Third Instar Larvae 53. nL4 <-30 #Fourth Instar Larvae 54. nL5 <-30 #Fifth Instar Larvae 55. nP <-25 #Pupae 56. nAr <-25 #Reproductive adult 57. nDL5 <-25 #Diapausing fifth instar larvae – Stage 1 58. nD <- 5 #Diapausing fifth instar larvae – Stage 2 59. 60. ###This is what you put in the pomp so this just says 61. ###Hey C, I need these integers that will set the number 62. ### of subcompartments 63. 64. globs <- c(paste0("static int nE = ",nE,";"), 65. paste0("static int nL1 = ",nL1,";"), 66. paste0("static int nL2 = ",nL2,";"), 67. paste0("static int nL3 = ",nL3,";"), 68. paste0("static int nL4 = ",nL4,";"), 69. paste0("static int nL5 = ",nL5,";"), 70. paste0("static int nDL5 = ",nDL5,";"), 71. paste0("static int nP = ",nP,";"), 72. paste0("static int nAr = ",nAr,";"), 73. paste0("static int nAr = ",nD,";"),) 74. 75. ################################################################################ 76. #############DETERMINISTIC SKELETON ############################################ 77. ################################################################################ 78. 79. 80. DIA_SKEL <- Csnippet(" 81. 82. //I need this function for the birth rate and diapause rate 83. // Example:the number of eggs is dependent on all the reproductive adults 84. // in the subcompartments so I have to sum it up 85. // C doesn't have a function for that so I have to create it and I name it //// 86. // sum 87. 88. const double sum(const double arr[], int n) 89. { 90. double sum = 0.0; 91. for (int i = 0; i < n; i++) 92. sum += arr[i]; 93. return sum; 94. } 95. 96. ////////////////////////////// 97. //Subcompartments in stages// 98. ///////////////////////////// 99. 100. 101. 102. // Here are the states that will require pointers to form 103. // an array of vectors. For all stages except for senescent adult/ 104. // check King's pomp documentation and there's more information there/ 105.

162

106. 107. double *E = &E1; //EGG 108. double *DE = &DE1; 109. 110. double *L1 = &L11; //LARVAE1 111. double *DL1 = &DL11; 112. 113. double *L2 = &L21; //LARVAE2 114. double *DL2 = &DL21; 115. 116. double *L3 = &L31; //LARVAE3 117. double *DL3 = &DL31; 118. 119. double *L4= &L41; //LARVAE4 120. double *DL4 = &DL41; 121. 122. double *L5= &L51; //LARVAE5 123. double *DL5 = &DL51; 124. 125. double *DIAL5= &DIAL51; //DIAPAUSING LARVAE 5-1 126. double *DDIAL5 = &DDIAL51; 127. 128. double *DW= &DW1; //DIAPAUSING LARVAE 5-2 129. double *DW = &DW1; 130. 131. double *P= &P1; //PUPAE 132. double *DP = &DP1; 133. 134. double *Ar= &Ar1; //REPRODUCTIVE ADULT 135. double *DAr = &DAr1; 136. 137. int i; 138. 139. 140. ///////////// 141. //Birth Rate/ 142. ////////////// 143. long double B; 144. 145. 146. // the birth rate is a Gaussian Function 147. B = birth_a * exp(0.50* powl ((TT- birth_t), 2.0) 148. / powl (birth_b, 2.0)); 149. 150. /////////////////////////////////////// 151. //Development Rate of all the stages// 152. /////////////////////////////////////// 153. 154. double alphaE, alphaL1, alphaL2, alphaL3, 155. alphaL4, alphaL5, alphaP, alphaA, alphaDL; 156. 157. 158. //Sigmoidal functions- the warmer it is, the faster one develops 159. alphaE = alphaE_a/ (1+exp(-alphaE_b*(TT - alphaE_c))); 160. alphaL1 = alphaL1_a/(1+exp(-alphaL1_b*(TT - alphaL1_c))); 161. alphaL2 = alphaL2_a/(1+exp(- alphaL2_b * (TT - alphaL2_c))); 162. alphaL3 = alphaL3_a/(1+exp(- alphaL3_b*(TT - alphaL3_c))); 163. alphaL4 = alphaL4_a/(1+exp(- alphaL4_b*(TT - alphaL4_c))); 164. alphaL5 = alphaL5_a/(1+exp(- alphaL5_b*(TT - alphaL5_c)));

163

165. alphaP = alphaP_a/(1+ exp(-alphaP_b*(TT - alphaP_c))); 166. alphaA = alphaA_a/(1+exp(-alphaA_b*(TT - alphaA_c))); 167. //Diapause development rate 168. alphaDL = alphaDL_a/(1+exp(alphaDL_b *(TT + alphaDL_c))); 169. 170. 171. ///////////////////////////////////// 172. ///Mortality Rate of all the stages// 173. ///////////////////////////////////// 174. double muE, muL, muP, muA, muDL; 175. //Wang mortality functions 176. muE = 1- (1/(exp((1+exp(-(TT-T_E_OPT)/MORT_E_A))*(1+exp(- 177. (T_E_OPT- TT)/MORT_E_A ))*MORT_E_B))); 178. 179. muL = 1- (1/(exp((1+exp(-(TT-T_L_OPT)/MORT_L_A))*(1+exp(- 180. (T_L_OPT- TT)/MORT_L_A))*MORT_L_B))); 181. 182. muP = 1- (1/(exp((1+exp(-(TT-T_P_OPT)/MORT_P_A))*(1+exp(- 183. (T_P_OPT- TT)/MORT_P_A))*MORT_P_B ))); 184. 185. muA = 1- (1/(exp((1+exp(-(TT-T_A_OPT)/MORT_DL_A))*(1+exp(- 186. (T_A_OPT- TT)/MORT_DL_A))*MORT_A_B ))); 187. 188. muDL = 1- (1/(exp((1+exp(-(TT-T_DL_OPT)/MORT_DL_A))*(1+exp(- 189. (T_DL_OPT- TT)/MORT_DL_A))*MORT_DL_B ))); 190. 191. 192. ////////////////////// 193. //Diapause induction// 194. ////////////////////// 195. 196. long double DIA_INDUC_L5; 197. 198. //Sigmoidal function 199. DIA_INDUC_L5 =DIA_A/(DIA_B+exp(DIA_C *PP_DIFF)); 200. 201. /////////////////////// 202. //BALANCE THE EQUATION/ 203. //////////////////////// 204. 205. 206. //General structure is shown below : 207. 208. //dS/dt = (recruitment in due to birth, 209. // maturation or diapause termination) - (recruitment out due to matura 210. tion or diapause) - background mortality 211. 212. /////// 213. //EGG// 214. /////// 215. 216. // Incorporating Allee effect on the birth rate and the use of the sum 217. //function 218. 219. DE[0] = 0.5*B*exp(C*sum(Ar,nAr))*sum(Ar,nAr) - 220. (nE * alphaE* E[0]) - (muE* E[0]); 221. for ( i = 1; i < nE; i ++){ 222. DE[i] = nE * alphaE * (E[i-1]- E[i]) - muE * E[i]; 223. }

164

224. 225. ///////////// 226. // LARVAE 1// 227. ///////////// 228. DL1[0] = nE * alphaE * E[nE1] - (nL1 * alphaL1 * L1[0])- 229. (muL*L1[0]); 230. for ( i = 1; i < nL1; i++){ 231. DL1[i] = nL1 * alphaL1 * (L1[i-1]- L1[i])- muL * L1[i];} 232. 233. //////////// 234. //LARVAE 2// 235. //////////// 236. DL2[0] = nL1 * alphaL1 * L1[nL11] - (nL2 * alphaL2 * L2[0])- 237. (muL*L2[0]); 238. for (i = 1; i < nL2 ; i++){ 239. DL2[i] = nL2 * alphaL2 * (L2[i-1]- L2[i])- muL*L2[i]; 240. } 241. 242. //////////// 243. //LARVAE 3// 244. //////////// 245. DL3[0] = nL2 * alphaL2 * L2[nL21] - (nL3 * alphaL3 * L3[0])- 246. (muL*L3[0]); 247. for (i = 1; i < nL3; i++){ 248. DL3[i]= nL3 * alphaL3 *(L3[i-1]- L3[i])- muL* L3[i]; 249. } 250. 251. //////////// 252. //LARVAE 4// 253. //////////// 254. DL4[0] = nL3 * alphaL3 * L3[nL3-1] - (nL4 * alphaL4 * L4[0])- 255. (muL*L4[0]); 256. for (i = 1; i < nL4; i++){ 257. DL4[i]= nL4 * alphaL4 *(L4[i-1]- L4[i])- muL* L4[i]; 258. } 259. 260. //////////// 261. //LARVAE 5// 262. //////////// 263. 264. 265. DL5[0] = (nL4 * alphaL4 * L4[nL4-1]) -(nL5 * alphaL5 *L5[0]) - 266. (DIA_INDUC_L5 * L5[0]) - 267. (muL*L5[0]); 268. 269. for (i = 1; i < nL5; i++){ 270. DL5[i]= nL5 * alphaL5 * (L5[i-1] -L5[i]) - 271. (DIA_INDUC_L5 * L5[i]) - 272. (muL *L5[i]); 273. } 274. 275. ////////////////////////////// 276. //DIAPAUSING LARVAL STAGE 5 // 277. ////////////////////////////// 278. 279. DDIAL5[0] =DIA_INDUC_L5*(sum(L5,nL5)) - (nDL5 * alphaDL*DIAL5[0])- 280. muDL * DIAL5[0]; 281. 282. for (i = 1; i < nDL5; i++){

165

283. DDIAL5[i]= nDL5* alphaDL*(DIAL5[i-1] - DIAL5[i]) - 284. muDL * DIAL5[i]; 285. } 286. ////////////////////// 287. //Diapausing Stage 2// 288. ////////////////////// 289. DW[0] = nDL5 * alphaDL* DIAL5[nDL5-1] - 290. nD * alphaW* W[0]- 291. (muDL *W[0]); 292. for (i = 1; i < nD; i++){ 293. DW[i]= nD*alphaW*(W[i-1] - W[i])- 294. (muDL *W[i]); 295. } 296. 297. ///////// 298. //PUPAE// 299. ///////// 300. 301. DP[0] = nD*alphaW*W[nD] + nL5 * alphaL5 * (L5[nL5-1]) 302. - nP * alphaP * P[0] - muP * P[0] ; 303. for (i = 1; i

166

341. W=c(100,rep(0,nD-1)),#Starting out with 30 diapausing larvae in each nDL5 compartment 342. P = rep(0,nP), 343. Ar = rep(0,nAr), 344. As = 0)} 345. ################################### 346. ###THIS IS THE FULL POMP MACHINE### 347. ################################### 348. 349. 350. ###THE DMEASURE IS IN A DIFFERENT SCRIPT# 351. ###FOR RUNNING SIMULATION YOU DON'T NEED IT SO I PUT IT AS NULL######### 352. 353. POMP<- pomp(covartable, 354. times="time", 355. t0=1, 356. globals=globs, 357. dmeasure = NULL, 358. covar =COVAR, 359. covarnames = c("TT", "s"), 360. obsnames = c("ARC"), 361. rinit = init , 362. skeleton = vectorfield(DIA_SKEL), 363. paramnames= 364. c("birth_a","birth_t","birth_b", 365. "alphaE_a", 366. "alphaE_b", 367. "alphaE_c", 368. "alphaL1_a", 369. "alphaL1_b", 370. "alphaL1_c", 371. "alphaL2_a", 372. "alphaL2_b", 373. "alphaL2_c", 374. "alphaL3_a", 375. "alphaL3_b", 376. "alphaL3_c", 377. "alphaL4_a", 378. "alphaL4_b", 379. "alphaL4_c", 380. "alphaL5_a", 381. "alphaL5_b", 382. "alphaL5_c", 383. "alphaP_a", 384. "alphaP_b", 385. "alphaP_c", 386. "alphaDL_a", 387. "alphaDL_b", 388. "alphaDL_c", 389. "alphaA_a", 390. "alphaA_b", 391. "alphaA_c", 392. "alphaW_a", 393. "alphaW_b", 394. "alphaW_c", 395. "MORT_E_A","MORT_E_B",'T_E_OPT', 396. "MORT_L_A","MORT_L_B",'T_L_OPT', 397. "MORT_P_A","MORT_P_B","T_P_OPT", 398. "MORT_A_A","MORT_A_B","T_A_OPT",

167

399. "MORT_DL_A","MORT_DL_B",'T_DL_OPT', 400. "DIA_A","DIA_B","DIA_C","C",'COMP'), 401. statenames=c(sprintf("E%1d",seq_len(nE)), 402. sprintf("L1%1d",seq_len(nL1)), 403. sprintf("L2%1d",seq_len(nL2)), 404. sprintf("L3%1d",seq_len(nL3)), 405. sprintf("L4%1d",seq_len(nL4)), 406. sprintf("L5%1d",seq_len(nL5)), 407. sprintf("DIAL5%1d",seq_len(nDL5)), 408. sprintf("W%1d",seq_len(nD)), 409. sprintf("P%1d",seq_len(nP)), 410. sprintf("Ar%1d",seq_len(nAr)), 411. "As")) 412. 413. #################################### 414. #THE PARAMETERS AND THEIR VALUES ### 415. #################################### 416. 417. PARAMETERS<- c(birth_a =10.52 , #Egg Fecundity 418. birth_t=24.629, #... 419. birth_b = 3.427, #... 420. 421. alphaE_a=0.1927, #Egg Development rate 422. alphaE_b=0.3039, #... 423. alphaE_c= 18.5929, #... 424. 425. alphaL1_a=0.30, #Instar 1 Development rate 426. alphaL1_b=0.327,#... 427. alphaL1_c= 17.60, #... 428. 429. alphaL2_a=0.457, #Instar 2 Development rate 430. alphaL2_b=0.2301,#... 431. alphaL2_c=21.23, #... 432. 433. alphaL3_a= 0.338, #Instar 3 Development rate 434. alphaL3_b=0.350, #... 435. alphaL3_c=18.45, #... 436. 437. alphaL4_a=0.305, #Instar 4 Development rate 438. alphaL4_b=0.429, #... 439. alphaL4_c= 21.54, #... 440. 441. alphaL5_a= 0.33, #Instar 5 Development rate 442. alphaL5_b=0.2144, #... 443. alphaL5_c= 20.94, #... 444. 445. alphaP_a=0.09287, #Pupae Development rate 446. alphaP_b=0.28966, #... 447. alphaP_c=18.48736,#... 448. 449. alphaA_a=0.13, #Adult development rate 450. alphaA_b=0.18, #... 451. alphaA_c= 20.2206,#... 452. 453. alphaDL_a=0.0716149819 , #Diapause development 454. alphaDL_b=.0899064656, 455. alphaDL_c= 0.53 , 456. 457. MORT_E_A =2.063e+00, #Egg Mortality rate

168

458. MORT_E_B=1.306e-04 , #... 459. T_E_OPT =1.357e+01 , #... 460. 461. MORT_L_A= 5.0000000000 , #Larval Mortality rate 462. MORT_L_B= 0.0004022344 , #... 463. T_L_OPT= 16.24,#... 464. 465. MORT_P_A=2.862e+00 , #Pupal mortality rate 466. MORT_P_B=1.036e-05 , #... 467. T_P_OPT=5.461e+00 ,#... 468. 469. MORT_A_A= 2.862e+00, #Adult mortality rate 470. MORT_A_B=1.306e-04, #... 471. T_A_OPT= 20, #... 472. 473. MORT_DL_A= 6.491816, #Diapause larvae mortality rate 474. MORT_DL_B= 0.000321, #... 475. T_DL_OPT= 3.725070 ,#... 476. 477. DIA_A=0.500 , #Diapause induction rate 478. DIA_B=0.5,#... 479. DIA_C= 220, #... 480. 481. C = 0.0061202543, #Allee constant , 482. 483. 484. 485. #################################### 486. #THE PARAMETERS AND THEIR VALUES ### 487. #################################### 488. 489. PARAMETERS<- c(birth_a =10.52 , #Egg Fecundity 490. birth_t=24.629, #... 491. birth_b = 3.427, #... 492. 493. alphaE_a=0.1927, #Egg Development rate 494. alphaE_b=0.3039, #... 495. alphaE_c= 18.5929, #... 496. 497. alphaL1_a=0.30, #Instar 1 Development rate 498. alphaL1_b=0.327,#... 499. alphaL1_c= 17.60, #... 500. 501. alphaL2_a=0.457, #Instar 2 Development rate 502. alphaL2_b=0.2301,#... 503. alphaL2_c=21.23, #... 504. 505. alphaL3_a= 0.338, #Instar 3 Development rate 506. alphaL3_b=0.350, #... 507. alphaL3_c=18.45, #... 508. 509. alphaL4_a=0.305, #Instar 4 Development rate 510. alphaL4_b=0.429, #... 511. alphaL4_c= 21.54, #... 512. 513. alphaL5_a= 0.33, #Instar 5 Development rate 514. alphaL5_b=0.2144, #... 515. alphaL5_c= 20.94, #... 516.

169

517. alphaP_a=0.09287, #Pupae Development rate 518. alphaP_b=0.28966, #... 519. alphaP_c=18.48736,#... 520. 521. alphaA_a=0.13, #Adult development rate 522. alphaA_b=0.18, #... 523. alphaA_c= 20.2206,#... 524. 525. alphaDL_a= 0.04 , #Diapause development 526. alphaDL_b=.3, 527. alphaDL_c=-2, 528. 529. alphaW_a=0.02679687, #Diapause development 530. alphaW_b=.5, 531. alphaW_c= 0.0390625, 532. 533. MORT_E_A =2.063e+00, #Egg Mortality rate 534. MORT_E_B=1.306e-04 , #... 535. T_E_OPT =1.357e+01 , #... 536. 537. MORT_L_A= 2.0000000000 , #Larval Mortality rate 538. MORT_L_B= 0.0004022344 , #... 539. T_L_OPT= 16.24,#... 540. 541. MORT_P_A=2.862e+00 , #Pupal mortality rate 542. MORT_P_B=1.036e-05 , #... 543. T_P_OPT=5.461e+00 ,#... 544. 545. MORT_A_A= 2.862e+00, #Adult mortality rate 546. MORT_A_B=1.306e-04, #... 547. T_A_OPT= 20, #... 548. 549. MORT_DL_A=3.091816, #Diapause larvae mortality rate 550. MORT_DL_B= 0.000321, #... 551. T_DL_OPT= 3.725070 ,#... 552. 553. DIA_A=4.6, #Diapause induction rate 554. DIA_B=1,#... 555. DIA_C= 230, #... 556. 557. C = 0.0061202543, #Allee constant , 558. COMP =0) #Density dependent ) 559. 560. 561. 562. ################################################################### 563. ###This runs the desolver and we integrate- TRAJ_MODEL IS WHAT I USE # 564. ### FOR ALL ANALYSIS################################################## 565. TRAJ_MODEL<- trajectory(POMP, 566. PARAMETERS 567. ,times=seq(1,12054), 568. format = 'data.frame') 569. 570. ###################################################################### 571. ###WE NEED TO GIVE THE DATE SO THAT WE CAN COMBINE THE REAL LIFE DATA# 572. ###SIMULATION # 573. ###################################################################### 574. 575. DATE<- seq.Date(as.Date('1/1/1984', format = '%m/%d/%Y'),

170

576. as.Date('12/31/2016', format = '%m/%d/%Y'), 577. 'days') 578. rr############################################################## 579. ###To get the total number of individuals within a life-stage# 580. ###We have to sum up the whole subcompartments ############### 581. ############################################################## 582. ########################################################################## 583. ###To get the total abundance of egg we have to sum up all the indviduals# 584. ###in the compartments together E = nE1 + nE2 and so on # 585. ###Isn't there a more efficient way of doing this? Yes ################### 586. ########################################################################## 587. ###EGG 588. EGG<-data.frame(DATE, Abund= rowSums(TRAJ_MODEL [,1:(nE)]), id = 'Egg') 589. 590. ###ALL LARVAL STAGES 591. LARVAE1<- data.frame(DATE, Abund = rowSums(TRAJ_MODEL [,(nE+1):(nE+nL1)]), id= 'L1') 592. 593. LARVAE2<- data.frame(DATE, Abund = rowSums(TRAJ_MODEL[,(nE+nL1+1):(nE+nL1+nL2)]), id = 'L2') 594. 595. LARVAE3<- data.frame(DATE, Abund = rowSums(TRAJ_MODEL[,(nE+nL1+nL2+1):(nE+nL1+nL2+nL3)]), id = 'L3') 596. 597. LARVAE4<- data.frame(DATE, Abund = rowSums(TRAJ_MODEL[,(nE+nL1+nL2+nL3+1):(nE+nL1+nL2+nL3+nL4)]), id = 'L4') 598. 599. LARVAE5<- data.frame(DATE, Abund = rowSums(TRAJ_MODEL[,(nE+nL1+nL2+nL3+nL4+1):(nE+nL1+nL2+nL3+nL4+nL5)]), id = 'L5') 600. 601. ###The Diapausing Larval Stage 602. 603. DIAL5<- data.frame(DATE, Abund = rowSums(TRAJ_MODEL[,(nE+nL1+nL2+nL3+nL4+nL5+1):(nE+nL1+nL2+nL3+nL4+nL5+nDL5)]), 604. id = 'DL5') 605. 606. DIAL52<- data.frame(DATE, Abund = rowSums(TRAJ_MODEL[,(nE+nL1+nL2+nL3+nL4+nL5+nDL5+1):(nE+nL1+nL2+nL3+nL4+nL5+nDL5+nD)]), 607. id = 'DL52') 608. ###Pupal Stage 609. Pupae<-data.frame(DATE, Abund = 610. rowSums(TRAJ_MODEL[,(nE+nL1+nL2+nL3+nL4+nL5+nDL5+nD+1):(nE+nL1+nL2+nL3+nL4+nL5+nDL5+nD+ nP)]), 611. id = 'Pupae') 612. ###Reproductive Adult 613. R.adult<- data.frame(DATE, Abund = rowSums(TRAJ_MODEL[,(nE+nL1+nL2+nL3+nL4+nL5+nDL5+nD+nP+1): 614. (nE+nL1+nL2+nL3+nL4+nL5+nDL5+nD+nP+nAr)]), id = 'R.adult') 615. 616. ###This is for matching up with the actual pheromone trap data 617. R.adult_2 = rowSums(TRAJ_MODEL[,(nE+nL1+nL2+nL3+nL4+nL5+nDL5+nD+nP+1): 618. (nE+nL1+nL2+nL3+nL4+nL5+nDL5+nD+nP+nAr)]) 619. ### Senescent Adult 620. S.adult<- data.frame(DATE, Abund= TRAJ_MODEL[,(nE+nL1+nL2+nL3+nL4+nL5+nDL5+nP+nAr+1)], id = 'S.Adult') 621. 622. #################################################################

171

623. ####################################################################### 624. ###THE FULL MODEL IS NOW ALL THE TOTAL INDIVIDUALS IN A SINGLE DF ##### 625. ####################################################################### 626. FULL_Model <- rbind.data.frame(EGG, LARVAE1, LARVAE2, 627. LARVAE3, LARVAE4, LARVAE5, 628. DIAL5,DIAL52, 629. Pupae, R.adult, S.adult) 630. 631. ###Putting in the year and the day of year 632. FULL_Model$Year <- as.numeric(format(FULL_Model$DATE, '%Y')) 633. FULL_Model$Julian <- as.numeric(format(FULL_Model$DATE, '%j')) 634. 635. 636. 637. ###If you want to see a plot of all the life-stages together 638. ###Just for the sake of making sure nothing is negative 639. ggplot(subset(FULL_Model,FULL_Model$id %in% c("L5","DL5","DL52","Pupae")), 640. aes (x = Julian, y = log(Abund+1),color = id))+ 641. geom_line(size =1.2)+ 642. facet_wrap(~Year)+ 643. scale_color_viridis(discrete=TRUE)+ 644. theme_bw() 645. 646. ggplot(FULL_Model, 647. aes (x = Julian, y = log(Abund+1),color = id))+ 648. geom_line(size =1.2)+ 649. facet_wrap(~Year)+ 650. scale_color_viridis(discrete=TRUE)+ 651. theme_bw() 652. 653. ###CREATING A DATAFRAME TO MERGE BOTH THE ACTUAL DATA (ARC) AND 654. ###THE SIMULATED DATA 655. 656. 657. SIM_DF<- cbind.data.frame(DATE ,SIM =(R.adult_2)) 658. ABUND_DF <- cbind.data.frame(DATE, ACT = subset(covartable, covartable$time <12055)$ARC ) 659. 660. SIM_ABUND_DF <- cbind(ACT=CM_F_3[,2], SIM_DF) 661. SIM_ABUND_DF$YEAR <- as.numeric(format(SIM_ABUND_DF$DATE,'%Y')) 662. SIM_ABUND_DF$julian <- as.numeric(format(SIM_ABUND_DF$DATE,'%j')) 663. 664. 665. ggplot(na.omit(SIM_ABUND_DF), aes(x = julian, y =(ACT)))+ 666. geom_line(size = 1,col='#fa1e3f',alpha = 1)+ 667. geom_point( aes(x =julian, y = (ACT)),col='red', size=1)+ 668. geom_line(data = SIM_ABUND_DF, aes(x= julian, y =log( 0.5*SIM+1)),size =1)+ 669. facet_wrap(~YEAR)+ 670. xlab("Day of Year")+ 671. ylab("Abundance of adult moths (log)")+ 672. theme_bw()+ 673. theme(strip.background = element_blank(), 674. strip.text.y = element_blank(), 675. panel.grid.major = element_blank(), 676. panel.grid.minor = element_blank(), 677. text = element_text(size = 16))

172

Appendix C

Supplementary Material for Chapter 4

Parameters Description Value

n The number of subcompartments for the first stage A 4, 6, 11, 25, and 100

훼 The development rate of the individuals going through stage A 0.02 (푑푎푦−1)

μ The instantaneous mortality rate of individuals in stage B 0.35 (푑푎푦−1)

−1 푀푚푎푥 The maximum mortality rate of the Hill’s equation 8 (푑푎푦 )

푀50 The concentration at which mortality is half of M1 0.5

푀푘 The Hill coefficient that determines the steepness 5

푏 The decay rate of the pesticide 0.01-0.25 (푑푎푦−1)

Table S1: The parameter values used for all analyses

173

A. The number of sprays needed for one species

Figure S1: The number of species required for the single-species case depending on the different action thresholds (20,40,60, and 80). On the x-axis is the coefficient of variation and on the y- axis is the decay rate of the pesticide.

174

B. The number of sprays needed for two species when the action thresholds are the same

Figure S2: The number of sprays needed for two species with the action thresholds of 20. The panels represent the different shifts in days between the two species. The x-axis represents the coefficient of variation while the y-axis represents the pesticide decay rate.

175

Figure S3: The number of sprays needed for two species with the action threshold of 40. The panels represent the different shifts in days between the two species. The x-axis represents the coefficient of variation while the y-axis represents the pesticide decay rate.

176

Figure S4: The number of sprays needed for two species with the action threshold of 60. The panels represent the different shifts in days between the two species. The x-axis represents the coefficient of variation while the y-axis represents the pesticide decay rate.

177

Figure S5: The number of sprays needed for two species with the action threshold of 80. The panels represent the different shifts in days between the two species. The x-axis represents the coefficient of variation while the y-axis represents the pesticide decay rate.

178

B. The difference in sprays

Figure S6: The difference in the number of sprays when interventions influence all species versus when intervention does not affect the other species (Action threshold: 20). More negative value suggests that more sprays are needed for the individual case. On the x-axis is the coefficient of variation while the y- axis is the decay rate. The panels are the shifts between the species.

179

Figure S7: The difference in the number of sprays when interventions influence all species versus when intervention does not affect the other species (Action threshold: 40). More negative value suggests that more sprays are needed for the individual case. On the x-axis is the coefficient of variation while the y- axis is the decay rate. The panels are the shifts between the species.

180

Figure S8: The difference in the number of sprays when interventions influence all species versus when intervention does not affect the other species (Action threshold: 60). More negative value suggests that more sprays are needed for the individual case. On the x-axis is the coefficient of variation while the y- axis is the decay rate. The panels are the shifts between the species.

181

Figure S9: The difference in the number of sprays when interventions influence both species versus when intervention does not affect the other species (Action threshold: 80). More negative value suggests that more sprays are needed for the individual case. On the x-axis is the coefficient of variation while the y- axis is the decay rate. The panels are the shifts between the species.

182

C. Damage of the second species

Figure S10: The damage of the second emerging species when only spraying for the first species (Action threshold is 20). On the x-axis is the coefficient of variation, the y-axis is the decay rate, and the panels represent the shift between the species.

183

Figure S11: The damage of the second emerging species when only spraying for the first species (Action threshold is 40). On the x-axis is the coefficient of variation, the y-axis is the decay rate, and the panels represent the shift between the species.

184

Figure S12: The damage of the second emerging species when only spraying for the first species (Action threshold is 60). On the x-axis is the coefficient of variation, the y-axis is the decay rate, and the panels represent the shift between the species.

185

Figure S13: The damage of the second emerging species when only spraying for the first species (Action threshold is 80). On the x-axis is the coefficient of variation, the y-axis is the decay rate, and the panels represent the shift between the species.

186

Figure S14: The number of sprays needed when the later emerging species has the action threshold at 40 and the earlier emerging species has a higher threshold that is +10, +20, and + 30 than the later emerging species. The x-axis is the coefficient of variation and the y-axis is the decay rate. The panel labels on the top represent shift, while the panel label shows the difference in action threshold.

187

Appendix D

Supplementary Material for Chapter 5

Table S1: Specimens submitted from outside of the state of Pennsylvania.

188

Table S2: Exotic tick specimens. Specimens that were not native to Pennsylvania, to northeastern United

States, or to North America were likely, either introductions on exotic animals or other unknown sources

(Keirans and Litwak 1989). Presence of exotic specimens may be indicative of introduction, but we do not have enough data to determine whether these species ever establish themselves as breeding populations. However, given the right conditions, an introduced exotic species can successfully establish a breeding population (for example, the parthenogenic H. longicornis) and become epidemiologically important to veterinary and human health

189

Figure S1: Total population of Pennsylvania counties over time. Data were taken from the US Census for

1960, 1990, 2000, and 2010.

190

Figure S2: Dot-density map of all individual tick specimens across Pennsylvania from 1900 to 2017.

Each point represents an individual tick specimen with its placement randomized within the county and each colored dot represents a different tick species.

191

Figure S3: Presence or absence map of species with less than 150. Counties with zero submissions are dark, while the light areas represent the presence of one or more specimens for each genus.

192

Appendix E

Supplementary Material for Chapter 6

Figure S1: Map showing locations of the Yasuní (YAS) and Cocha Cashu (CC) plots.

193

References

Aghdam, H. R., Y. Fathipour, and D. C. Kontodimas. 2011. Evaluation of non-linear models to describe

development and fertility of codling moth at constant temperatures. Entomologia Hellenica 20:3–16.

Aghdam, H. R., Y. Fathipour, D. C. Kontodimas, G. Radjabi, and M. Rezapanah. 2009a. Age-Specific

Life Table Parameters and Survivorship of an Iranian Population of the Codling Moth (Lepidoptera:

Tortricidae) at Different Constant Temperatures. Annals of the Entomological Society of America

102:233–240.

Aghdam, H. R., Y. Fathipour, G. Radjabi, and M. Rezapanah. 2009b. Temperature-Dependent

Development and Temperature Thresholds of Codling Moth (Lepidoptera: Tortricidae) in Iran.

Environmental Entomology 38:885–895.

Altermatt, F. 2010. Climatic warming increases voltinism in european butterflies and moths. Proceedings

of the Royal Society B: Biological Sciences.

Anderson, J. F., P. H. Duray, and L. A. Magnarelli. 1990. Borrelia burgdorferi and Ixodes dammini

prevalent in the greater Philadelphia area.

Ashby, M. D., and P. Singh. 1990. Control of diapause in codling moth larvae.

Ashton, P. S., T. J. Givnish, and S. Appanah. 1988. Staggered flowering in the Dipterocarpaceae: new

insights into floral induction and the evolution of mast fruiting in the aseasonal tropics. American

Naturalist.

Baer-Lehman, M. L., T. Light, N. W. Fuller, K. D. Barry-Landis, C. M. Kindlin, and R. L. Stewart. 2012.

Evidence for competition between Ixodes scapularis and Dermacentor albipictus feeding

concurrently on white-tailed deer. Experimental and Applied Acarology.

Bagchi, R., V. Swamy, J. P. Latorre Farfan, J. Terborgh, C. I. A. Vela, N. C. A. Pitman, and W. G.

Sanchez. 2018. Defaunation increases the spatial clustering of lowland Western Amazonian tree

communities. Journal of Ecology.

194

Bale, J. S., and S. A. L. Hayward. 2010a. Insect overwintering in a changing climate. Journal of

Experimental Biology 213.

Bale, J. S., and S. A. L. Hayward. 2010b. Insect overwintering in a changing climate. Journal of

Experimental Biology 213:980–994.

Barrett, A. W., B. H. Noden, J. M. Gruntmeir, T. Holland, J. R. Mitcham, J. E. Martin, E. M. Johnson,

and S. E. Little. 2015. County scale distribution of amblyomma americanum (Ixodida: Ixodidae) in

oklahoma: Addressing local deficits in tick maps based on passive reporting. Journal of Medical

Entomology.

Beits, R. A., M. Collins, D. L. Hemming, C. D. Jones, J. A. Lowe, and M. G. Sanderson. 2011. When

could global warming reach 4°C? Pages 67–84 Philosophical Transactions of the Royal Society A:

Mathematical, Physical and Engineering Sciences.

Bernard-Verdier, M., M. L. Navas, M. Vellend, C. Violle, A. Fayolle, and E. Garnier. 2012. Community

assembly along a soil depth gradient: Contrasting patterns of plant trait convergence and divergence

in a Mediterranean rangeland. Journal of Ecology.

Bishopp, F. C., and H. L. Trembley. 1945. Distribution and Hosts of Certain North American Ticks. The

Journal of Parasitology.

Bjørnstad, O. N., W. A. Nelson, and P. C. Tobin. 2016. Developmental synchrony in multivoltine insects:

generation separation versus smearing. Population Ecology 58:479–491.

Blomefield, T. L., and J. H. Giliomee. 2009. Development Rates of the Embryonic and Immature Stages

of Codling Moth, Cydia pomonella (L.) (Lepidoptera: Tortricidae), at Constant and Fluctuating

Temperatures. African Entomology 17:183–191.

Botes, C., S. D. Johnson, and R. M. Cowling. 2008. Coexistence of succulent tree aloes: Partitioning of

bird pollinators by floral traits and flowering phenology. Oikos.

Bouchard, C., P. A. Leighton, G. Beauchamp, S. Nguon, L. Trudel, F. Milord, L. R. Lindsay, D.

Bélanger, and N. H. Ogden. 2013. Harvested white-tailed deer as sentinel hosts for early

195

establishing ixodes scapularis populations and risk from vector-borne zoonoses in Southeastern

Canada. Journal of Medical Entomology.

Briers, R. A., J. H. R. Gee, and R. Geoghegan. 2004. Effects of the North Atlantic Oscillation on growth

and phenology of stream insects. Ecography 27:811–817.

Brillhart, D. B., L. B. Fox, and S. J. Upton. 1994. Ticks (Acari: Ixodidae) collected from small and

medium-sized Kansas mammals. Journal of Medical Entomology. ten Brink, H., J. R. Gremer, and H. Kokko. 2020. Optimal germination timing in unpredictable

environments: the importance of dormancy for both among- and within-season variation.

Brites-Neto, J., K. M. R. Duarte, and T. F. Martins. 2015. Tick-borne infections in human and animal

population worldwide. Veterinary World.

Brown, R. F., and D. G. Mayer. 1988. Representing Cumulative Germination.: 2. The Use of the Weibull

Function and Other Empirically Derived Curves. Annals of Botany 61:127–138.

Brown, S. M., P. M. Lehman, R. A. Kern, and J. D. Henning. 2015. Detection of Borrelia burgdorferi and

Anaplasma phagocytophilum in the black-legged tick, Ixodes scapularis, within southwestern

Pennsylvania. Journal of Vector Ecology.

Brownstein, J. S., D. K. Skelly, T. R. Holford, and D. Fish. 2005. Forest fragmentation predicts local

scale heterogeneity of Lyme disease risk. Oecologia.

Burg, J. G. 2001. Seasonal activity and spatial distribution of host-seeking adults of the tick Dermacentor

variabilis. Medical and Veterinary Entomology.

Calabrese, J. M., and W. F. Fagan. 2004. Lost in time, lonely, and single: Reproductive asynchrony and

the allee effect. American Naturalist.

Campagnolo, E. R., D. Tewari, T. S. Farone, J. L. Livengood, and K. L. Mason. 2018. Evidence of

Powassan/deer tick virus in adult black-legged ticks (Ixodes scapularis) recovered from hunter-

harvested white-tailed deer (Odocoileus virginianus) in Pennsylvania: A public health perspective.

Zoonoses and Public Health.

196

Carrascal, L. M., I. Galván, and O. Gordo. 2009. Partial least squares regression as an alternative to

current regression methods used in ecology. Oikos 118:681–690.

Chen, C., Q.-W. Xia, S. Fu, X.-F. Wu, and F.-S. Xue. 2014a. Effect of photoperiod and temperature on

the intensity of pupal diapause in the cotton bollworm, Helicoverpa armigera (Lepidoptera:

Noctuidae). Bulletin of Entomological Research 104:12–18.

Chen, X., S. Xu, Z. Yu, L. Guo, S. Yang, L. Liu, X. Yang, and J. Liu. 2014b. Multiple lines of evidence

on the genetic relatedness of the parthenogenetic and bisexual haemaphysalis longicornis (Acari:

Ixodidae). Infection, Genetics and Evolution.

Chen, Z. Z., L. X. Xu, L. L. Li, H. B. Wu, and Y. Y. Xu. 2019. Effects of constant and fluctuating

temperature on the development of the oriental fruit moth, Grapholita molesta (Lepidoptera:

Tortricidae). Bulletin of Entomological Research.

Chidawanyika, F., and J. S. Terblanche. 2011. Rapid thermal responses and thermal tolerance in adult

codling moth Cydia pomonella (Lepidoptera: Tortricidae). Journal of Insect Physiology 57:108–117.

Chinuki, Y., K. Ishiwata, K. Yamaji, H. Takahashi, and E. Morita. 2016. Haemaphysalis longicornis tick

bites are a possible cause of red meat allergy in Japan. Allergy: European Journal of Allergy and

Clinical Immunology.

Chmielewski, F. M. 2013. Phenology in agriculture and horticulture. Page Phenology: An Integrative

Environmental Science.

Cohen, S. B., J. D. Freye, B. G. Dunlap, J. R. Dunn, T. F. Jones, and A. C. Moncayo. 2010. Host

associations of dermacentor, amblyomma, and ixodes (Acari: Ixodidae) ticks in Tennessee. Journal

of Medical Entomology.

Collier, H., and S. Finch. 1983. Effects of Intensity and Duration of Low Temperatures in Regulating

Development of the Cabbage Root Fly ( Dzxza Radzcum ) 34:193–200.

Cook, B. I., T. M. Smith, and M. E. Mann. 2005. The North Atlantic Oscillation and regional phenology

prediction over Europe. Global Change Biology 11:919–926.

197

Cooley, R. A., and G. M. Kohls. 1944. The Genus Amblyomma (Ixodidae) in the United States. The

Journal of Parasitology.

Cooley, R. A., and G. M. Kohls. 2011. The Argasidae of North America, Central America and Cuba, by

R.A. Cooley ... and Glen M. Kohls ... Contribution from the Rocky Mountain Laboratory, Hamilton,

Montana. Division of Infectious Diseases, National Institute of Health, United States Public Heal.

Page The Argasidae of North America, Central America and Cuba, by R.A. Cooley ... and Glen M.

Kohls ... Contribution from the Rocky Mountain Laboratory, Hamilton, Montana. Division of

Infectious Diseases, National Institute of Health, United States Public Heal.

Cortinas, R., and S. M. Spomer. 2014. Occurrence and county-level distribution of ticks (Acari:

Ixodoidea) in nebraska using passive surveillance. Journal of Medical Entomology.

Courtney, J. W., R. L. Dryden, J. Montgomery, B. S. Schneider, G. Smith, and R. F. Massung. 2003.

Molecular characterization of Anaplasma phagocytophilum and Borrelia burgdorferi in Ixodes

scapularis ticks from Pennsylvania. Journal of Clinical Microbiology.

Crowder, C. D., H. E. Carolan, M. A. Rounds, V. Honig, B. Mothes, H. Haag, O. Nolte, B. J. Luft, L.

Grubhoffer, D. J. Ecker, S. E. Schutzer, and M. W. Eshoo. 2014. Prevalence of Borrelia miyamotoi

in Ixodes ticks in Europe and the United States. Emerging Infectious Diseases.

Daniels, T. J., D. Fish, J. F. Levine, M. A. Greco, A. T. Eaton, P. J. Padgett, and D. A. LaPointe. 1993.

Canine exposure to Borrelia burgdorferi and prevalence of Ixodes dammini (Acari: Ixodidae) on

deer as a measure of Lyme disease risk in the northeastern United States. Journal of medical

entomology.

Danks, H. V. 1978a. Some effects of photoperiod, temperature, and food on emergence in three species of

chironomidae (diptera). The Canadian Entomologist.

Danks, H. V. 1978b. Modes of seasonal adaptation in the insects. The Canadian Entomologist 110:1167–

1205.

Denlinger, B. Y. D. L., and J. Y. B. Iv. 1981. Duration of Pupal Diapause in the Tobacco Hornworm Is

198

Determined By Number of Short Days Received By the Larva. Experimental Animals:331–337.

Detto, M., S. J. Wright, O. Calderón, and H. C. Muller-Landau. 2018. Resource acquisition and

reproductive strategies of tropical forest in response to the El Niño-Southern Oscillation. Nature

Communications.

Devevey, G., and D. Brisson. 2012. The effect of spatial heterogenity on the aggregation of ticks on

white-footed mice. Parasitology.

Dick, C. W., M. R. Gannon, W. E. Little, and M. J. Patrick. 2003. Ectoparasite Associations of Bats from

Central Pennsylvania. Journal of Medical Entomology.

Dunstone, N., D. Smith, A. Scaife, L. Hermanson, R. Eade, N. Robinson, M. Andrews, and J. Knight.

2016. Skilful predictions of the winter North Atlantic Oscillation one year ahead. Nature Geoscience

9:809–814.

Dyck, V. A. 2005. Rearing Codling Moth for the Sterile Insect Technique. Page Springer, Dordrecht, The

Netherlands.

Edwards, M. J., L. A. Barbalato, A. Makkapati, K. D. Pham, and L. M. Bugbee. 2015. Relatively low

prevalence of Babesia microti and Anaplasma phagocytophilum in Ixodes scapularis ticks collected

in the Lehigh Valley region of eastern Pennsylvania. Ticks and Tick-borne Diseases.

Eisen, R. J., K. J. Kugeler, L. Eisen, C. B. Beard, and C. D. Paddock. 2017. Tick-borne zoonoses in the

United States: Persistent and emerging threats to human health. ILAR Journal.

Elzinga, J. A., A. Atlan, A. Biere, L. Gigord, A. E. Weis, and G. Bernasconi. 2007. Time after time:

flowering phenology and biotic interactions.

Fantinou, A. A., and E. A. Kagkou. 2000. Effect of thermoperiod on diapause induction of Sesamia

nonagrioides (Lepidoptera-Noctuidae). Environ Entomol 29:489–494.

Forister, M. L., and A. M. Shapiro. 2003. Climatic trends and advancing spring flight of butterflies in

lowland California. Global Change Biology 9:1130–1135.

Forrest, J., and A. J. Miller-Rushing. 2010. Toward a synthetic understanding of the role of phenology in

199

ecology and evolution. Philosophical Transactions of the Royal Society of London B: Biological

Sciences 365.

Forrest, J. R. 2016. Complex responses of insect phenology to climate change.

Friedenberg, N. A., J. A. Powell, and M. P. Ayres. 2007. Synchrony’s double edge: Transient dynamics

and the Allee effect in stage structured populations. Ecology Letters 10:564–573.

Garcia-Salazar, C., H. Podoler, and M. E. Whalon. 1988. Effects of temperature on diapause induction in

the codling moth, Cydia pomonella (L.) (Lepidoptera: Olethreutidae). Environmental Entomology

17:626–628.

Gentry, A. H. 1974. Flowering Phenology and Diversity in Tropical Bignoniaceae. Biotropica.

Gilbert, E., J. A. Powell, J. A. Logan, and B. J. Bentz. 2004. Comparison of three models predicting

developmental milestones given environmental and individual variation. Bulletin of Mathematical

Biology 66:1821–1850.

Gilbert, N., and D. a. Raworth. 1996. FORUM: INSECTS AND TEMPERATURE—A GENERAL

THEORY. The Canadian Entomologist 128:1–13.

Glenn, P. A. 1922. Relation of temperature to development of codling moth. Journal of Economic

Entomology 15:193–198.

Godoy, O., and J. M. Levine. 2014. Phenology effects on invasion success: Insights from coupling field

experiments to coexistence theory. Ecology.

Gonzalez, A., and M. Loreau. 2009. The causes and consequences of compensatory dynamics in

ecological communities. Annual Review of Ecology, Evolution, and Systematics.

Gordo, O., and J. J. Sanz. 2005. Phenology and climate change: a long-term study in a Mediterranean

locality. Oecologia 146:484–495.

Graf, B., H. Höhn, H. U. Höpli, and S. Kuske. 2018. Predicting the phenology of codling moth, Cydia

pomonella, for sustainable pest management in Swiss apple orchards. Entomologia Experimentalis

et Applicata 166:618–627.

200

Gu, Z., L. Gu, R. Eils, M. Schlesner, and B. Brors. 2014. Circlize implements and enhances circular

visualization in R. Bioinformatics.

Gurney, W. S. C., P. H. Crowley, and R. M. Nisbet. 1992. Locking life-cycles onto seasons: Circle-map

models of population dynamics and local adaptation. Journal of Mathematical Biology 30:251–279.

Hall, J. E., J. W. Amrine, R. D. Gais, V. P. Kolanko, B. E. Hagenbuch, V. F. Gerencser, and S. M. Clark.

1991. Parasitization of humans in West Virginia by Ixodes cookei (Acari: Ixodidae), a potential

vector of Lyme borreliosis. Journal of medical entomology.

Han, E. R. N., and E. Bauce. 1998. Timing of diapause initiation, metabolic changes and overwintering

survival of the spruce budworm, Choristoneura fumiferana. Ecological Entomology 23:160–167.

Han, G. S., E. Y. Stromdahl, D. Wong, and A. C. Weltman. 2014. Exposure to borrelia burgdorferi and

other tick-borne pathogens in gettysburg national military park, south-central pennsylvania, 2009.

Vector-Borne and Zoonotic Diseases.

Hans, A., W. Borchers, and M. H. W. Borchers. 2017. Package ‘ pracma .’ R package version.

Harrington, R., I. Woiwod, and T. Sparks. 1999. Climate change and trophic interactions. Trends in

Ecology & Evolution 14:146–150.

Harrison, R. D., S. Tan, J. B. Plotkin, F. Slik, M. Detto, T. Brenes, A. Itoh, and S. J. Davies. 2013.

Consequences of defaunation for a tropical tree community. Ecology Letters.

Hodek, I. 1996. Diapause development, diapause termination and the end of diapause. European Journal

of Entomology 93:475–487.

Holmes, E. E., E. J. Ward, and K. Wills. 2012. MARSS: Multivariate autoregressive state-space models

for analyzing time-series data. R Journal 4:11–19.

Howell, J. F., and L. G. Neven. 2000. Physiological Development Time and Zero Development

Temperature of the Codling Moth (Lepidoptera: Tortricidae). Environmental Entomology 29:766–

772.

Howell, J. F., and R. S. Schmidt. 2002. Codling moth (Lepidoptera: Tortricidae): development at constant

201

and at orchard temperatures. Journal of Agricultural and Urban Entomology 19:15–23.

Hurrell, J. W. 1995. Decadal trends in the North Atlantic oscillation: Regional temperatures and

precipitation. Science.

Hutchinson, M. L., M. D. Strohecker, T. W. Simmons, A. D. Kyle, and M. W. Helwig. 2015. Prevalence

Rates of Borrelia burgdorferi (Spirochaetales: Spirochaetaceae), Anaplasma phagocytophilum

(Rickettsiales: Anaplasmataceae), and Babesia microti (Piroplasmida: Babesiidae) in Host-Seeking

Ixodes scapularis (Acari: Ixodidae) from Pennsylvania. Journal of Medical Entomology.

Jarosik, V., L. Kratochvil, A. Honek, and A. F. G. Dixon. 2004. A general rule for the dependence of

developmental rate on temperature in ectothermic animals. Proceedings of the Royal Society B:

Biological Sciences 271:S219–S221.

Johnson, C. A., R. M. Coutinho, E. Berlin, K. E. Dolphin, J. Heyer, B. Kim, A. Leung, J. Lou Sabellon,

and P. Amarasekare. 2016. Effects of temperature and resource variation on insect population

dynamics: the bordered plant bug as a case study. Functional Ecology 30:1122–1131.

Jones, F. A., and L. S. Comita. 2010. Density-dependent pre-dispersal seed predation and fruit set in a

tropical tree. Oikos.

Juszczak, R., L. Kuchar, J. Leśny, and J. Olejnik. 2013. Climate change impact on development rates of

the codling moth (Cydia pomonella L.) in the Wielkopolska region, Poland. International Journal of

Biometeorology.

Karlsson, P. S., and M. Méndez. 2005. The Resource Economy of Plant Reproduction. Page Reproductive

Allocation in Plants.

Keirans, J. E., and C. M. Clifford. 1978. The genus Ixodes in the United States: a scanning electron

microscope study and key to the adults. Journal of medical entomology. Supplement.

Keirans, J. E., and L. A. Durden. 1998. Illustrated Key to Nymphs of the Tick Genus Amblyomma

(Acari: Ixodidae) Found in the United States. Journal of Medical Entomology.

Keirans, J. E., and T. R. Litwak. 1989. Pictorial key to the adults of hard ticks, family Ixodidae (Ixodida:

202

Ixodoidea), east of the Mississippi River. Journal of medical entomology.

Kelly, D., A. L. Harrison, W. G. Lee, I. J. Payton, P. R. Wilson, and E. M. Schauber. 2000. Predator

satiation and extreme mast seeding in 11 species of Chionochloa (Poaceae). Oikos.

Khan, A. A., A. R. Wani, F. A. Zaki, R. K. Nehru, and S. S. Pathania. 2018. Pests of Apple. Pages 457–

490 in Omkar, editor. Pests and Their Management. Springer Singapore, Singapore.

Khani, A., and S. Moharramipour. 2010. Cold Hardiness and Supercooling Capacity in the

Overwintering Larvae of the Codling Moth, Cydia pomonella . Journal of Insect Science 10:1–12.

King, A. A., D. Nguyen, and E. L. Ionides. 2016. Statistical inference for partially observed markov

processes via the R package pomp. Journal of Statistical Software.

Kivelä, S. M., P. Välimäki, and K. Gotthard. 2016. Evolution of alternative insect life histories in

stochastic seasonal environments. Ecology and Evolution.

Knight, A. L., and D. M. Light. 2012. Monitoring Codling Moth (Lepidoptera: Tortricidae) in Sex

Pheromone-Treated Orchards With (E)-4,8-Dimethyl-1,3,7-Nonatriene or Pear Ester in

Combination With Codlemone and Acetic Acid. Environmental Entomology.

Kollars, J., J. Oliver, L. A. Durden, and P. G. Kollars. 2000. Host associations and seasonal activity of

Amblyomma americanum (Acari: Ixodidae) in Missouri. Journal of Parasitology.

Kollars, T. M., and J. H. Oliver. 2003. Host Associations and Seasonal Occurrence of Haemaphysalis

leporispalustris, Ixodes brunneus, I. cookei, I. dentatus, and I. texanus (Acari: Ixodidae) in

Southeastern Missouri. Journal of Medical Entomology.

Kryzhov, V. N., and O. V. Gorelits. 2015. The Arctic Oscillation and its impact on temperature and

precipitation in Northern Eurasia in the 20th Century. Russian Meteorology and Hydrology 40:711–

721.

Lacey, L. A., and T. R. Unruh. 1998. Entomopathogenic nematodes for control of codling moth, Cydia

pomonella (Lepidoptera: Tortricidae): Effect of nematode species, concentration, temperature, and

humidity. Biological Control.

203

Lasky, J. R., M. Uriarte, and R. Muscarella. 2016. Synchrony, compensatory dynamics, and the functional

trait basis of phenological diversity in a tropical dry forest tree community: Effects of rainfall

seasonality. Environmental Research Letters 11.

Lehmann, P., W. Van Der Bijl, S. Nylin, C. W. Wheat, and K. Gotthard. 2017. Timing of diapause

termination in relation to variation in winter climate. Physiological Entomology 42:232–238.

Logan, J. D., W. Wolesensky, and A. Joern. 2006. Temperature-dependent phenology and predation in

arthropod systems. Ecological Modelling.

Lord, R. D., V. R. Lord, J. G. Humphreys, and R. G. McLean. 1994. Distribution of Borrelia burgdorferi

in host mice in Pennsylvania. Journal of Clinical Microbiology.

Lubelczyk, C., B. K. Cahill, T. Hanson, J. Turmel, E. Lacombe, P. W. Rand, S. P. Elias, and R. P. Smith.

2010. Tick (Acari: Ixodidae) infestation at two rural, seasonal camps in Maine and Vermont. Journal

of Parasitology.

Luedeling, E., and A. Gassner. 2012. Partial Least Squares Regression for analyzing walnut phenology in

California. Agricultural and Forest Meteorology 158–159:43–52.

Luedeling, E., A. Kunz, and M. M. Blanke. 2013. Identification of chilling and heat requirements of

cherry trees-a statistical approach. International Journal of Biometeorology 57:679–689.

Luedeling, E., K. P. Steinmann, M. Zhang, P. H. Brown, J. Grant, and E. H. Girvetz. 2011. Climate

change effects on walnut pests in California. Global Change Biology 17:228–238.

Ma, L., X. Wang, Y. Liu, M. Z. Su, and G. H. Huang. 2017. Temperature effects on development and

fecundity of Brachmia macroscopa (Lepidoptera: Gelechiidae). PLoS ONE.

Magnarelli, L. A., K. C. Stafford, T. N. Mather, M. T. Yeh, K. D. Horn, and J. S. Dumler. 1995.

Hemocytic rickettsia-like organisms in ticks: Serologic reactivity with antisera to ehrlichiae and

detection of DNA of agent of human granulocytic ehrlichiosis by PCR. Journal of Clinical

Microbiology.

Manetsch, T. J. 1976. Time-Varying Distributed Delays and Their Use in Aggregative Models of Large

204

Systems. IEEE Transactions on Systems, Man and Cybernetics 6:547–553.

Mangel, M., and P. S. Levin. 2005. Regime, phase and paradigm shifts: Making community ecology the

basic science for fisheries.

Mcleod, A. A. I., and M. A. I. Mcleod. 2015. Kendall rank correlation and Mann-Kendall trend test.

McLeod, A. I. 2011. Package “Kendall.”

McWatters, H. G., and D. S. Saunders. 1998. Maternal temperature has different effects on the

photoperiodic response and duration of larval diapause in blow fly (Calliphora vicina) strains

collected at two latitudes. Physiological Entomology 23:369–375.

Metz, J., and O. Diekmann. 2017. Exact Finite Dimensional Representations of Models for

Physiologically Structured Populations.I. Pages 269–290 Differential Equations with Applications in

Biology, Physics, and Enqineering.

Mevik, B. H., and R. Wehrens. 2007. The pls package: Principal component and partial least squares

regression in R. Journal of Statistical Software.

Miller, M. J., H. J. Esser, J. R. Loaiza, E. A. Herre, C. Aguilar, D. Quintero, E. Alvarez, and E.

Bermingham. 2016. Molecular ecological insights into neotropical bird-tick interactions. PLoS

ONE.

Moffitt, H. R., and A. K. Burditt. 1989. Effects of Low Temperatures on Three Embryonic Stages of the

Codling Moth (Lepidoptera: Tortricidae). Journal of Economic Entomology 82:1379–1381.

Moore, S. M., R. J. Eisen, A. Monaghan, and P. Mead. 2014. Meteorological influences on the

seasonality of lyme disease in the United States. American Journal of Tropical Medicine and

Hygiene.

Myburgh, A. . 1980. Infestation potential of the codling moth. Deciduous Fruit Grower 10:368–377.

Myers, C. T., L. A. Hull, and G. Krawczyk. 2007. Effects of Orchard Host Plants (Apple and Peach) on

Development of Oriental Fruit Moth (Lepidoptera: Tortricidae). Journal of Economic Entomology.

Nakazawa, T., and H. Doi. 2012. A perspective on match/mismatch of phenology in community contexts.

205

Oikos.

Navarro-Cano, J. A., B. Karlsson, D. Posledovich, T. Toftegaard, C. Wiklund, J. Ehrlén, and K. Gotthard.

2015. Climate change, phenology, and butterfly host plant utilization. Ambio.

Nelson, W. a, O. N. Bjørnstad, and T. Yamanaka. 2013. Recurrent insect outbreaks caused by

temperature-driven changes in system stability. Science (New York, N.Y.) 341:796–9.

Neven, L. G. 2013. Effects of Short Photoperiod on Codling Moth Diapause and Survival. Journal of

Economic Entomology.

Neven, L. G., H. J. Ferguson, and A. Knight. 2000. Sub-zero cooling synchronizes post-diapause

development of codling moth, Cydia pomonella. Cryo-Letters 21:203–214.

Node: Pennsylvania (316987717). (n.d.). .

Oliver, J. D., S. W. Bennett, L. Beati, and L. C. Bartholomay. 2017. Range Expansion and Increasing

Borrelia burgdorferi Infection of the Tick Ixodes scapularis (Acari: Ixodidae) in Iowa, 1990-2013.

Journal of medical entomology.

Omeg. 2001. Biology and Management of the Obliquebanded Leafroller, Choristoneura rosaceana

(Harris) (Lepidoptera: Tortricidae), in Sweet Cherry. Thesis 0:1–103.

Ottersen, G., B. Planque, A. Belgrano, E. Post, P. Reid, and N. Stenseth. 2001. Ecological effects of the

North Atlantic Oscillation. Oecologia 128:1–14.

Ovaskainen, O., S. Skorokhodova, M. Yakovleva, A. Sukhov, A. Kutenkov, N. Kutenkova, A.

Shcherbakov, E. Meyke, and M. del M. Delgado. 2013a. Community-level phenological response to

climate change. Proceedings of the National Academy of Sciences 110:13434–13439.

Ovaskainen, O., S. Skorokhodova, M. Yakovleva, A. Sukhov, A. Kutenkov, N. Kutenkova, A.

Shcherbakov, E. Meyke, and M. Del Mar Delgado. 2013b. Community-level phenological response

to climate change. Proceedings of the National Academy of Sciences of the United States of

America.

Pak, D., D. Biddinger, and O. N. Bjørnstad. 2018. Local and regional climate variables driving spring

206

phenology of tortricid pests: a 36 year study. Ecological Entomology.

Pau, S., E. M. Wolkovich, B. I. Cook, T. J. Davies, N. J. B. Kraft, K. Bolmgren, J. L. Betancourt, and E.

E. Cleland. 2011. Predicting phenology by integrating ecology, evolution and climate science.

Pennsylvania., L. U. | N. (n.d.). No Title.

Pennsylvania State Climatologist. (n.d.). .

Poulin, B., S. J. Wright, G. Lefebvre, and O. Calderón. 1999. Interspecific synchrony and asynchrony in

the fruiting phenologies of congeneric bird-dispersed plants in Panama. Journal of Tropical Ecology.

Powell, J. A., J. L. Jenkins, J. A. Logan, and B. J. Bentz. 2000. Seasonal temperature alone can

synchronize life cycles. Bulletin of Mathematical Biology 62:977–998.

Powell, J. A., and J. A. Logan. 2005. Insect seasonality: circle map analysis of temperature-driven life

cycles. Theoretical Population Biology 67:161–179.

PRETZMAN, C., N. DAUGHERTY, K. POETTER, and D. RALPH. 1990. The Distribution and

Dynamics of Rickettsia in the Tick Population of Ohio. Annals of the New York Academy of

Sciences.

Prokopy, R., and M. Kogan. 2009. Integrated Pest Management. Page Encyclopedia of Insects.

Pumpuni, C. B., J. Knepler, and G. B. Craig. 1992. Influence of temperature and larval nutrition on the

diapause inducing photoperiod of Aedes albopictus. Journal of the American Mosquito Control

Association.

Rainey, T., J. L. Occi, R. G. Robbins, and A. Egizi. 2018. Discovery of Haemaphysalis longicornis

(Ixodida: Ixodidae) Parasitizing a Sheep in New Jersey, United States. Journal of medical

entomology.

Rand, P. W., E. H. Lacombe, R. Dearborn, B. Cahill, S. Elias, C. B. Lubelczyk, G. A. Beckett, and R. P.

Smith. 2007. Passive surveillance in Maine, an area emergent for tick-borne diseases. Journal of

Medical Entomology.

Rathcke, B., and E. P. Lacey. 1985. Phenological patterns of terrestrial plants. Annual review of ecology

207

and systematics. Vol. 16.

Ratte, H. T. 1985. Temperature and Insect Development. Pages 33–66 in K. H. Hoffmann, editor.

Environmental Physiology and Biochemistry of Insects. Springer Berlin Heidelberg, Berlin,

Heidelberg.

Lo Re, V., J. L. Occi, and R. R. MacGregor. 2004. Identifying the Vector Lyme Disease. American

Family Physician.

Rebaudo, F., and V. B. Rabhi. 2018. Modeling temperature-dependent development rate and phenology in

insects: review of major developments, challenges, and future directions.

Régnière, J., J. Powell, B. Bentz, and V. Nealis. 2012. Effects of temperature on development, survival

and reproduction of insects: Experimental design, data analysis and modeling. Journal of Insect

Physiology.

Ripoche, M., S. Gasmi, A. Adam-Poupart, J. K. Koffi, L. R. Lindsay, A. Ludwig, F. Milord, N. H.

Ogden, K. Thivierge, and P. A. Leighton. 2018. Passive tick surveillance provides an accurate early

signal of emerging lyme disease risk and human cases in Southern Canada. Journal of Medical

Entomology.

Robertson, C. 1895. The Philosophy of Flower Seasons, and the Phaenological Relations of the

Entomophilous Flora and the Anthophilous Insect Fauna. The American Naturalist.

Rock, G. C., and P. L. Shaffer. 1983. Tufted Apple Budmoth ( Lepidoptera : Tortricidae ): Effects of

Constant Daylengths and Temperatures on Larval Diapause Development l. Environmental

Entomology 12:71–75.

Rogers, M. B., L. Cui, A. Fitch, V. Popov, A. P. A. Travassos Da Rosa, N. Vasilakis, R. B. Tesh, and E.

Ghedin. 2014. Short report: Whole genome analysis of Sierra Nevada virus, a novel mononegavirus

in the family Nyamiviridae. American Journal of Tropical Medicine and Hygiene.

Rosenberg, R., N. P. Lindsey, M. Fischer, C. J. Gregory, A. F. Hinckley, P. S. Mead, G. Paz-Bailey, S. H.

Waterman, N. A. Drexler, G. J. Kersh, H. Hooks, S. K. Partridge, S. N. Visser, C. B. Beard, and L.

208

R. Petersen. 2018. Vital Signs : Trends in Reported Vectorborne Disease Cases —

United States and Territories, 2004–2016 . MMWR. Morbidity and Mortality Weekly

Report.

Roy, D. B., and T. H. Sparks. 2000. Phenology of British butterflies and climate change. Global Change

Biology 6:407–416.

Sakamoto, J. M., T. F. F. Ng, Y. Suzuki, H. Tsujimoto, X. Deng, E. Delwart, and J. L. Rasgon. 2016.

Bunyaviruses are common in male and female Ixodes scapularis ticks in central Pennsylvania. PeerJ.

Sanchirico, J. N., and J. E. Wilen. 2001. A bioeconomic model of marine reserve creation. Journal of

Environmental Economics and Management.

Saracco, J. F., J. A. Collazo, M. J. Groom, and T. A. Carlo. 2005. Crop size and fruit neighborhood

effects on bird visitation to fruiting Schefflera morototoni trees in Puerto Rico. Biotropica.

Schaefer, K., A. S. Denning, and O. Leonard. 2005. The winter Arctic Oscillation, the timing of spring,

and carbon fluxes in the Northern Hemisphere. Global Biogeochemical Cycles 19:1–17.

Schoelkopf, L., C. E. Hutchinson, K. G. Bendele, W. L. Goff, M. Willette, J. M. Rasmussen, and P. J.

Holman. 2005. New ruminant hosts and wider geographic range identified for Babesia odocoilei

(Emerson and Wright 1970). Journal of Wildlife Diseases.

Serfass, T. L., L. M. Rymon, and R. P. Brooks. 1992. Ectoparasites from river otters in Pennsylvania.

Journal of wildlife diseases.

Setyobudi, L. 1989. Seasonality of codling moth, Cydia pomonella L. (Lepidoptera: Oleuthreutidae) in

the Willamette Valley of Oregon: role of photoperiod and temperature. Oregon State University-

PhD dissertation.

Shock, B. C., A. Moncayo, S. Cohen, E. A. Mitchell, P. C. Williamson, G. Lopez, L. E. Garrison, and M.

J. Yabsley. 2014. Diversity of piroplasms detected in blood-fed and questing ticks from several

states in the United States. Ticks and Tick-borne Diseases.

Silman, M. R. 2011. Plant species diversity in Amazonian forests. Page Tropical Rainforest Responses to

209

Climatic Change.

SILVERTOWN, J. W. 1980. The evolutionary ecology of mast seeding in trees. Biological Journal of the

Linnean Society.

Simmons, T. W., J. Shea, M. A. Myers-Claypole, R. Kruise, and M. L. Hutchinson. 2015. Seasonal

Activity, Density, and Collection Efficiency of the Blacklegged Tick (Ixodes scapularis) (Acari:

Ixodidae) in Mid-Western Pennsylvania. Journal of Medical Entomology.

Simon, J. A., R. R. Marrotte, N. Desrosiers, J. Fiset, J. Gaitan, A. Gonzalez, J. K. Koffi, F. J. Lapointe, P.

A. Leighton, L. R. Lindsay, T. Logan, F. Milord, N. H. Ogden, A. Rogic, E. Roy-Dufresne, D.

Suter, N. Tessier, and V. Millien. 2014. Climate change and habitat fragmentation drive the

occurrence of Borrelia burgdorferi, the agent of Lyme disease, at the northeastern limit of its

distribution. Evolutionary Applications.

Snetsinger, R. 1968. Distribution of ticks and tick-borne diseases in Pennsylvania. University Park, PA.

Snetsinger, R., S. B. Jacobs, K. C. Kim, and D. Tavris. 1993. Extension of the range of Dermacentor

variabilis (Acari: Ixodidae) in Pennsylvania. Journal of medical entomology.

Soetaert, K., T. Petzoldt, and R. W. Setzer. 2010. Solving differential equations in R: Package deSolve.

Journal of Statistical Software.

Sonenshine, D. E. 2018. Range expansion of tick disease vectors in north america: Implications for spread

of tick-borne disease.

Springer, Y. P., C. S. Jarnevich, D. T. Barnett, A. J. Monaghan, and R. J. Eisen. 2015. Modeling the

present and future geographic distribution of the lone star tick, amblyomma americanum (ixodida:

Ixodidae), in the continental United States. American Journal of Tropical Medicine and Hygiene.

Stålhandske, S., P. Lehmann, P. Pruisscher, and O. Leimar. 2015. Effect of winter cold duration on spring

phenology of the orange tip butterfly, Anthocharis cardamines. Ecology and evolution 5:5509–20.

Staudinger, M. D., K. E. Mills, K. Stamieszkin, N. R. Record, C. A. Hudak, A. Allyn, A. Diamond, K. D.

Friedland, W. Golet, M. E. Henderson, C. M. Hernandez, T. G. Huntington, R. Ji, C. L. Johnson, D.

210

S. Johnson, A. Jordaan, J. Kocik, Y. Li, M. Liebman, O. C. Nichols, D. Pendleton, R. A. Richards,

T. Robben, A. C. Thomas, H. J. Walsh, and K. Yakola. 2019. It’s about time: A synthesis of

changing phenology in the Gulf of Maine ecosystem.

Stefanescu, C., J. Penuelas, and I. Filella. 2003. Effects of climatic change on the phenology of butterflies

in the northwest Mediterranean Basin. Global Change Biology 9:1494–1506.

Steinberg, S., H. Podoler, and S. W. Applebaum. 1992. Diapause induction in the codling moth, Cydia

pomonella: effect of larval diet. Entomologia Experimentalis et Applicata 62:269–275.

Steiner, F. E., R. R. Pinger, C. N. Vann, N. Grindle, D. Civitello, K. Clay, and C. Fuqua. 2008. Infection

and co-infection rates of Anaplasma phagocytophilum variants, Babesia spp., Borrelia burgdorferi,

and the rickettsial endosymbiont in Ixodes scapularis (Acari: Ixodidae) from sites in Indiana, Maine,

Pennsylvania, and Wisconsin. Journal of Medical Entomology.

Stenseth, N. C., A. Mysterud, G. Ottersen, J. W. Hurrell, K. S. Chan, and M. Lima. 2002. Ecological

effects of climate fluctuations.

Stiles, F. G. 1977. Coadapted competitors: The flowering seasons of hummingbird-pollinated plants in a

tropical forest. Science.

Stoeckli, S., M. Hirschi, C. Spirig, P. Calanca, M. W. Rotach, and J. Samietz. 2012. Impact of climate

change on voltinism and prospective diapause induction of a global pest insect - cydia pomonella

(l.). PLoS ONE 7.

Stromdahl, E., S. Hamer, S. Jenkins, L. Sloan, P. Williamson, E. Foster, R. Nadolny, C. Elkins, M. Vince,

and B. Pritt. 2014. Comparison of phenology and pathogen prevalence, including infection with the

Ehrlichia muris-like (EML) agent, of Ixodes scapularis removed from soldiers in the midwestern

and the northeastern United States over a 15 year period (1997-2012). Parasites and Vectors.

Tauber, M. J., and C. A. Tauber. 1976. Insect Seasonality: Diapause Maintenance, Termination, and

Postdiapause Development. Annual Review of Entomology 21:81–107.

Tauber, M. J., C. A. Tauber, and S. Masaki. 1986. Seasonal adaptations of insects. Page Ecology.

211

Terborgh, J. 1986. Community aspects of frugivory in tropical forests.

Terborgh, J. 1990. An overview of research at Cocha Cashu Biological Station. Pages 48–59 in A. H.

Gentry, editor. Four Neotropical rainforests. Yale University Press, New Haven, Connecticut,USA.

Touré, M., D. Sanogo, S. Dembele, S. I. Diawara, K. Oppfeldt, K. L. Schiøler, D. Ben Haidara, S. F.

Traoré, M. Alifrangis, F. Konradsen, and S. Doumbia. 2016. Seasonality and shift in age-specific

malaria prevalence and incidence in Binko and Carrière villages close to the lake in Selingué, Mali.

Malaria Journal.

Tufts, D. M., M. C. Vanacker, M. P. Fernandez, A. Denicola, A. Egizi, and M. A. Diuk-Wasser. 2019.

Distribution, host-seeking phenology, and host and habitat associations of haemaphysalis

longicornis ticks, Staten Island, New York, USA. Emerging Infectious Diseases.

U.S. Census data from 1960 to 1990. (n.d.). .

US Census data: U.S. Census Bureau. (n.d.). .

Usinowicz, J., C. H. Chang-Yang, Y. Y. Chen, J. S. Clark, C. Fletcher, N. C. Garwood, Z. Hao, J.

Johnstone, Y. Lin, M. R. Metz, T. Masaki, T. Nakashizuka, I. F. Sun, R. Valencia, Y. Wang, J. K.

Zimmerman, A. R. Ives, and S. J. Wright. 2017. Temporal coexistence mechanisms contribute to the

latitudinal gradient in forest diversity. Nature.

Valencia, R., R. Condit, R. B. Foster, K. Romoleroux, G. Villa-Muñoz, J. C. Svenning, E. Magård, M.

Bass, E. C. Losos, and H. Balslev. 2004a. Tropical forest diversity and dynamism: Findings from a

large-scale plot network. Page Tropical forest diversity and dynamism: Findings from a large-scale

plot network.

Valencia, R., R. B. Foster, G. Villa, R. Condit, J. C. Svenning, C. Hernández, K. Romoleroux, E. Losos,

E. Magård, and H. Balslev. 2004b. Tree species distributions and local habitat variation in the

Amazon: Large forest plot in eastern Ecuador. Journal of Ecology.

Vasseur, D. A., J. W. Fox, A. Gonzalez, R. Adrian, B. E. Beisner, M. R. Helmus, C. Johnson, P. Kratina,

C. Kremer, C. de Mazancourt, E. Miller, W. A. Nelson, M. Paterson, J. A. Rusak, J. B. Shurin, and

212

C. F. Steiner. 2014. Synchronous dynamics of zooplankton competitors prevail in temperate lake

ecosystems. Proceedings of the Royal Society B: Biological Sciences.

Waits, K., M. J. Edwards, I. N. Cobb, R. S. Fontenele, and A. Varsani. 2018. Identification of an

anellovirus and genomoviruses in ixodid ticks. Virus Genes.

Wang, L., M. Ting, and P. J. Kushner. 2017. A robust empirical seasonal prediction of winter NAO and

surface climate. Scientific Reports 7.

Wang, P., M. N. Glowacki, A. E. Hoet, G. R. Needham, K. A. Smith, R. E. Gary, and X. Li. 2014.

Emergence of Ixodes scapularis and Borrelia burgdorferi, the Lyme disease vector and agent, in

Ohio. Frontiers in Cellular and Infection Microbiology.

Westgarth-Smith, A. R., S. A. G. Leroy, P. E. F. Collins, and R. Harrington. 2007. Temporal variations in

English populations of a forest insect pest, the green spruce aphid (Elatobium abietinum), associated

with the North Atlantic Oscillation and global warming. Quaternary International 173–174:153–160.

Westgarth-Smith, A. R., D. B. Roy, M. Scholze, A. Tucker, and J. P. Sumpter. 2012. The role of the

North Atlantic Oscillation in controlling U.K. butterfly population size and phenology. Ecological

entomology 37:221–232.

Whalon, M. E., D. Mota-Sanchez, and R. M. Hollingworth. 2008. Global pesticide resistance in

arthropods. Page Global Pesticide Resistance in Arthropods.

Wheelwright, N. T. 1985. Competition for Dispersers, and the Timing of Flowering and Fruiting in a

Guild of Tropical Trees. Oikos.

Wiens, J. J., and C. H. Graham. 2005. Niche conservatism: Integrating evolution, ecology, and

conservation biology.

Williams, D. G., and G. McDonald. 1982. THE DURATION AND NUMBER OF THE IMMATURE

STAGES OF CODLING MOTH CYDIA POMONELLA (L.) (TORTRICIDAE: LEPIDOPTERA).

Australian Journal of Entomology 21:1–4.

Wold, H. 1985. Partial Least Squares. Pages 581–591 Encyclopedia of Statistical Sciences.

213

Wolkovich, E. M., and E. E. Cleland. 2011. The phenology of plant invasions: A community ecology

perspective.

Wolkovich, E. M., B. I. Cook, and T. J. Davies. 2014. Progress towards an interdisciplinary science of

plant phenology: Building predictions across space, time and species diversity.

Xiao, X., S. Hagen, Q. Zhang, M. Keller, and B. Moore. 2006. Detecting leaf phenology of seasonally

moist tropical forests in South America with multi-temporal MODIS images. Remote Sensing of

Environment.

Xu, G., T. N. Mather, C. S. Hollingsworth, and S. M. Rich. 2016. Passive Surveillance of Ixodes

scapularis (Say), Their Biting Activity, and Associated Pathogens in Massachusetts. Vector-Borne

and Zoonotic Diseases.

Xu, G., P. Pearson, and S. M. Rich. 2018. Ehrlichia muris in ixodes cookei ticks, northeastern United

States, 2016–2017.

Yamanaka, T., and A. M. Liebhold. 2009. Mate-location failure, the Allee effect, and the establishment of

invading populations. Population Ecology 51:337–340.

Yamanaka, T., W. a. Nelson, K. Uchimura, and O. N. Bjørnstad. 2012. Generation Separation in Simple

Structured Life Cycles: Models and 48 Years of Field Data on a Tea Tortrix Moth. The American

Naturalist 179:95–109. yeh, M. T., J. M. Bak, R. Hu, M. C. Nicholson, C. Kelly, and T. N. Mather. 1995. Determining the

duration of Ixodes scapularis (Acari: Ixodidae) attachment to tick-bite victims. Journal of medical

entomology.

Yokoyama, V. Y., G. T. Miller, and A. M. Harvey. 1987. Development of Oriental Fruit Moth

(Lepidoptera: Tortricidae) on a Laboratory Dietl. J. Econ. Entomol 80:272–276.

Yunker, C. E., J. E. Keirans, C. M. Clifford, and E. R. Easton. 1986. Dermacentor ticks (Acari: Ixodidea:

Ixodidae) of the New World: A scanning electron microscope atlas.

Zhou, Z. S., J. Y. Guo, H. S. Chen, and F. H. Wan. 2010. Effects of temperature on survival,

214 development, longevity, and fecundity of ophraella communa (Coleoptera: Chrysomelidae), a potential biological control agent against ambrosia artemisiifolia (Asterales: Asteraceae).

Environmental Entomology.

VITA

Damie Pak EDUCATION Pennsylvania State University (University Park, Pennsylvania) • PhD Candidate in Biology (2015 - Current) University of Michigan (Ann Arbor, Michigan) • Bachelor of Science in Ecology and Evolutionary Biology (2011-2014) • University Honors (4 semesters)

PUBLICATIONS Canto, M.N, Hall, M.D., Pak, D., Young, P.R., Holmes, E.C., McGraw, E.A. (2019). Intra-host growth kinetics of dengue virus in the mosquito Aedes aegypti. PlosPathogens. Iverson, A, Gonthier, D, Pak, D., Ennis, K., Burnham, R, Perfecto, I, Vandermeer, J. (2019). Resolving a key dilemma in biodiversity conservation and farmer livelihoods. Biological Conservation, 238. Pak, D., Jacobs, S. B., & Sakamoto, J. M. (2019). A 117-year retrospective analysis of Pennsylvania tick community dynamics. Parasites & vectors, 12(1), 189. Pak, D., Biddinger, D., & Bjørnstad, O. N. (2019). Local and regional climate variables driving spring phenology of tortricid pests: a 36 year study. Ecological Entomology, 44(3), 367-379. Nelson, W. A., Joncour, B., Pak, D., & Bjørnstad, O. N. (2019). Asymmetric interactions and their consequences for vital rates and dynamics: the smaller tea tortrix as a model system. Ecology, 100(2). Karp, Daniel S., et al [including Pak, Damie]. Crop pests and predators exhibit inconsistent responses to surrounding landscape composition. Proceedings of the National Academy of Sciences 115.33 (2018. Pak, D*., Iverson, A*. L., Ennis, K. K., Gonthier, D. J., & Vandermeer, J. H. (2015). Parasitoid wasps benefit from shade tree size and landscape complexity in Mexican coffee agroecosystems. Agriculture, Ecosystems & Environment, 206, 21-32.

* Co- First Authors