The Pennsylvania State University

The Graduate School

Department of Entomology

PROJECTING THE INFLUENCE OF CLIMATE CHANGE AND

DAILY TEMPERATURE RANGE ON INSECT PHENOLOGY AND

RISK PERIOD OF VECTOR BORNE DISEASES

A Dissertation in

Entomology and Operations Research

by

Shi Chen

 2011 Shi Chen

Submitted in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

August 2011 ii

The dissertation of Shi Chen was reviewed and approved* by the following:

Shelby J. Fleischer Professor of Entomology Dissertation Co-Advisor Co-Chair of Committee

Michael C. Saunders Professor of Entomology Dissertation Co-Advisor Co-Chair of Committee

Matthew B. Thomas Professor of Entomology

Paul Heinemann Professor of Agricultural and Biological Engineering

John Fricks Assistant Professor of Statistics

Gary Felton Professor of Entomology Head of the Department of Entomology

*Signatures are on file in the Graduate School

iii

ABSTRACT

Understanding how climate change would influence the life history and population dynamics of insects is a key component needed to evaluate the impact of climate change on ecological and epidemiological systems, since insects are relatively sensitive to their surrounding environment, especially ambient temperature. Previous research has demonstrated that rising mean temperatures can influence voltinism, geographic distribution, and vectoring capacity. However, the distributional change among life stages for specific climate scenarios, and the importance of the daily temperature range (DTR), or daily temperature fluctuations, has not been emphasized. In this thesis we employed an agroecosystem and an epidemiological system and quantitatively study how changes in daily temperature range, as well as in mean temperature, would shift pest/vector insect phenology.

In the first chapter (chapter II) following the introduction chapter (chapter I), I developed an individual based Monte Carlo simulation method to project grape berry moth (Paralobesia viteana) life history (e.g. number of generations per year, mean and distribution of emergence time for each generation) under two different climate change scenarios, A1fi and B1, which represent upper and lower global warming conditions, respectively. The results showed that under A1fi condition the number of generations per year would increase nonlinearly after an initial lag of approximately 35 years (from

2009), and by the end of this century we might expect almost one more generations per year on average and a 5th generation would be more common. I then further investigated

iv how changes in the daily temperature range would influence degree day accumulation and consequently P. viteana phenology in chapter III. I discovered that mean temperature alone could not sufficiently and accurately compute degree day accumulation, and various daily temperature range conditions substantially influences degree day accumulation. Large changes in daily temperature fluctuations around current climatic conditions were sufficient to allow the voltinism of P. viteana to approach that simulated to occur in at the end of this century. The simulations based on different daily temperature ranges and mean temperature conditions revealed that, even when mean temperature decreased, larger daily temperature ranges could compensate the effect of decreasing mean temperature, resulting in a similar number of generations per year.

These results confirmed that daily temperature range should be an important factor in modeling insect phenology and population dynamics.

In chapter IV I considered to an epidemiological system and investigated if degree day models could predict West Nile virus (WNV) incidence dates in Pennsylvania. The original degree days model did not consider a temperature-dependent extrinsic incubation period or longevity changes in vector mosquitoes (Culex spp.), and did not capture virus transmission period well in four locations (Harrisburg, Philadelphia, Pittsburgh, and

Williamsport) in Pennsylvania from 2002 to 2008. By treating decreasing the extrinsic incubation period (EIP) at the beginning of the season, and incorporating adult mosquito longevity a variable as well, model performance was increased significantly. I also demonstrated that mosquito species composition should be an important factor for predicting WNV transmission period. The calibrated models performed well to predict

v

WNV emergence period in Pennsylvania, with a success rate of more than 70% compared to less than 30% in the uncalibrated model. In the successive research in chapter V I demonstrated that mean temperature alone did not provide enough information to compute degree day accumulation and predict disease transmission period well. I quantified WNV transmission period by comprehensively exploring combinations of changing daily temperature range and mean temperature in four locations: Harrisburg,

Philadelphia, Pittsburgh, and Williamsport in Pennsylvania. I investigated vector mosquito life history change according to these different climate change conditions. The results again showed that even when the mean temperatures remained the same, a large daily temperature variation results in more degree day accumulation and consequently a longer transmission period. Increasing mean temperature with increasing or decreasing the daily temperature range would yield earlier first incidences and later last incidences. I also demonstrated that increasing daily temperature range would impact disease transmission more dramatically in relatively cooler areas (Williamsport, Harrisburg, and

Pittsburgh) than relatively warm areas (Philadelphia). These results showed how daily temperature range alters WNV epidemics quantitatively and confirmed that I need to consider daily temperature range in our research in addition to mean temperature.

vi

TABLE OF CONTENTS

LIST OF FIGURES ...... viii

LIST OF TABLES ...... xi

ACKNOWLEDGEMENTS ...... xiii

Chapter 1 Introduction ...... 1

References Cited ...... 19

Chapter 2 Projecting Insect Voltinism under High and Low Greenhouse Gas Emission Conditions ...... 32

Introduction ...... 33

Materials and Methods ...... 35

Results ...... 41

Discussion ...... 45

Reference Cited ...... 50

Chapter 3 Investigating the Relationship between Daily Temperature Range and Insect Phenology under Climate Change ...... 65

Introduction ...... 66

Materials and Methods ...... 68

Results ...... 71

Discussion ...... 78

vii

Reference Cited ...... 83

Chapter 4 Capturing West Nile Virus Risk Using Degree Day Model in Pennsylvania ...... 111

Introduction ...... 112

Materials and Methods ...... 114

Results ...... 118

Discussion ...... 121

Reference Cited ...... 125

Chapter 5 Changing Daily Temperature Range, Degree Day Accumulation, and Disease Transmission Period ...... 146

Introduction ...... 148

Materials and Methods ...... 150

Results ...... 155

Discussion ...... 162

Reference Cited ...... 167

Chapter 6 Conclusion ...... 215

Appendix Calculate Daily Temperature Profile and Degree Day Accumulation ...... 219

viii

LIST OF FIGURES

Figure 2-1: Conceptual model for a single simulation run to collect number of generations per year, initialized by a diapausing overwintering life stage...... 59

Figure 2-2: Degree Day requirement of first generation emergence of adult grape berry moth from overwintering diapausing pupae. PDF is the density function, and CDF the cumulative density function. For both functions, modeled values are from Tobin et al. 2002, and observation are from field data collected in 2002...... 60

Figure 2-3: Historical and projected number of mean generations of grape berry moth in North East, PA, under low and high emission conditions (B1 and A1fi emission conditions, respectively) described by the Intergovernmental Panel on Climate Change (IPCC 2007b)...... 61

Figure 2-4: Density functions expressing the emergence date for all generations of grape berry moth in 2007 (top panel, for observed data collected from malaise traps, and for simulated data), and for 2099 under low (middle panel) and high (lower panel) emission conditions (B1 and A1fi emission conditions, respectively) described by the Intergovernmental Panel on Climate Change (IPCC 2007b)...... 62

Figure 2-5: Projected mean emergence date for all generations of grape berry moth under low (B1) emission conditions described by the Intergovernmental Panel on Climate Change (IPCC 2007b)...... 63

Figure 2-6: Projected mean emergence date for all generations of grape berry moth under high (A1fi) emission conditions described by the Intergovernmental Panel on Climate Change (IPCC 2007b)...... 64

Figure 3-1: Daily Temperature Profiles From a Sine Negative Exponential Model in Different Days Bracketing the Seasonal Development of Paralobesia viteana ...... 93

Figure 3-1: Daily Temperature Profiles From a Sine Negative Exponential Model in Different Days Bracketing the Seasonal Development of Paralobesia viteana ...... 94

Figure 3-2-1: Degree Day Accumulation in Day 100 under Various DTR Conditions ...... 94

Figure 3-2-2: Degree Day Accumulation in Day 150 under Various DTR Conditions ...... 95

Figure 3-2-3: Degree Day Accumulation in Day 200 under Various DTR Conditions ...... 96

Figure 3-2-4: Degree Day Accumulation in Day 250 under Various DTR Conditions ...... 97

Figure 3-3: Predicted Number of Generation per Year under Different DTR Conditions ..... 98

Figure 3-4: Predicted Change of Number of Generation per Year under Different DTR Conditions ...... 99

ix

Figure 3-5: Predicted Mean Emergence Date of First Generation ...... 100

Figure 3-6: Predicted Mean Emergence Date of Second Generation ...... 101

Figure 3-7: Predicted Mean Emergence Date of Third Generation ...... 102

Figure 3-8: Predicted Mean Emergence Date of Fourth Generation...... 103

Figure 3-9: Predicted Mean Emergence Date of Fifth Generation ...... 104

Figure 3-10: Predicted Number of Mean Generations Per Year in Category I: no change in mean and symmetric change in the DTR ...... 105

Figure 3-11: Number of Generation Change in Category II (k2=0.4) ...... 106

Figure 3-12: Predicted Mean Emergence Day in Category II (k2=0.4) ...... 107

Figure 3-13: Predicted Number of Generation Change in Category III (k1=0.4) ...... 108

Figure 3-14: Predicted Number of Generation Change in Category IV (k1=-0.4) ...... 109

Figure 3-15: Predicted Mean Emergence Day in Category IV (k1=-0.4) ...... 110

Figure 4-1: Topography of Pennsylvania and the Locations of the Four Sites (red triangles) Used to Model WNV ...... 135

Figure 4-2: Difference of Observation and Model Prediction in Original Model ...... 136

Figure 4-3: Difference of Observation and Model Prediction in Adjusted Model ...... 137

Figure 4-4: WNV Positive Species Composition in Four Locations from 2002-2008 ...... 138

Figure 4-5: Difference of Observation and Model Prediction in Adjusted Model with Calibration for Philadelphia ...... 139

Figure 5-1: Degree Day Accumulation under Different DTR Conditions in Harrisburg in Day 100 ...... 185

Figure 5-2: Degree Day Accumulation under Different DTR Conditions in Harrisburg in Day 200 ...... 186

Figure 5-3: Degree Day Accumulation under Different DTR Conditions in Philadelphia in Day 100...... 187

Figure 5-4: Degree Day Accumulation under Different DTR Conditions in Philadelphia in Day 200...... 188

Figure 5-5: Change of First, Last Incidence, Transmission Period, and Area in Category I .. 189

x

Figure 5-6: Change of First, Last Incidence, Transmission Period, and Area in Category II ...... 191

Figure 5-7: Change of First, Last Incidence, Transmission Period, and Area in Category III ...... 193

Figure 5-8: Change of First, Last Incidence, Transmission Period, and Area in Category IV...... 195

Figure 5-9: Change of First, Last Incidence, Transmission Period, and Area in Category V ...... 197

Figure 5-10: Accumulated Degree Days under Same Mean with Various DTR Conditions in Harrisburg ...... 200

Figure 5-11: Accumulated Degree Days under Same Mean with Various DTR Conditions in Philadelphia ...... 202

xi

LIST OF TABLES

Table 2-1: Regression of mean emergence dates (y) on years (x) under low (B1) and high (A1fi) emission conditions...... 57

Table 2-2: Influence of low or high emission condition (modeled as a categorical variable) and year (modeled as a covariate) on mean emergence dates...... 58

Table 3-1: Six Major Categories of Changing Daily Temperature Range...... 86

Table 3-2: Degree Day Accumulation for specific days under Different Mean Temperature and DTR Conditions...... 87

Table 3-3: P. viteana phenology under different DTR conditions...... 89

Table 3-4: Mean Generation Duration (in Days) of 2nd-5th Generation in Category II (k2 = 0.4)...... 91

Table 3-5: Mean Generation Duration (in days) of 2nd-5th Generation in Category IV (k1=-0.4)...... 92

Table 4-1: Annual Mean Temperature Summary from 2002-2008 in 4 Locations...... 129

Table 4-2: Predicted, Observed Emergence Dates and Difference (in days) in Original Model...... 130

Table 4-3: Predicted, Observed Emergence Dates and Difference in Adjusted Model ...... 132

Table 4-4: Predicted, Observed Emergence Dates and Difference For Philadelphia...... 134

Table 5-1: Six Major Categories of Changing Daily Temperature Range ...... 172

Table 5-2: Degree Day Accumulation in Harrisburg...... 173

Table 5-3: Degree Day Accumulation in Philadelphia...... 174

Table 5-4: Degree Day Accumulation in Pittsburgh ...... 175

Table 5-5: Degree Day Accumulation in Williamsport...... 176

Table 5-6: Incidence Time and Transmission Period Change in Harrisburg ...... 177

Table 5-7: Incidence Time and Transmission Period Change in Philadelphia...... 179

Table 5-8: Incidence Time and Transmission Period Change in Pittsburgh...... 181

xii

Table 5-9: Incidence Time and Transmission Period Change in Williamsport ...... 183

xiii

ACKNOWLEDGEMENTS

I would sincerely thank first and foremost all my current and previous graduate committee members, Shelby J. Fleischer, Michael C. Saunders, Matthew B. Thomas,

Paul Heinemann, John Fricks, Ottar Bjonstad, and Scott Isard, for their advice, guidance, and encouragement during my graduate program at Penn State. Particular thanks go to my advisors, Shelby and Mike for their support and instruction on my graduate study in all these five years.

I would also pay my tributes to many excellent faculty instructors in Department of Entomology, to name a few, Consuelo De Moraes, Diana Cox-Foster, Gary Felton,

Kechung Kim, Kelly Hoover, Mark Mescher, Michael Saunders, Shelby Fleischer, and

Tom Baker. Their dedication to the courses and professional development guidance has shaped me the next generation of entomologist.

As a student majoring in operations research I owe my thanks to those responsible instructors, John Fricks, Naomi Altman and Steven Arnold from Department of Statistics, Jose Ventura and M. Jeya Chandra from Department of Industrial

Engineering, Asok Ray from Department of Mechanical Engineering, Christopher B.

Byrne from Department of Mathematics, and many more, for their efforts on teaching and exposing me to the knowledge of various computational skills.

xiv

During my research at Penn State I am grateful for the help, advising, and collaboration from Matt Ferrari, Jim Marden, and Howard Fescemyer, Department of

Biology; Justine Blanford, Department of Geosciencies; and Ray Najjar, Department of

Meteorology. I also appreciate many outside researchers, for instance, Patrick Tobin from

USDA-Forest Service, and Melanie Fitzerpatrick from the Union of Concerned Scientists, for their help on developing the grape berry moth manuscripts. I would like to thank my former undergraduate advisors, Jianxiu Chen and Cheng Huang, who established sound entomology and computational background for me. I would thank many fellow students at and outside Penn State for the discussion on many research ideas and projects as well.

I would like to thank Penn State entomology staff, Ellen Johnson, LuAnn

Weatherholtz, Roxie Smith, and Karen Dreibelbis, for their kindness and help over all these years, and Dave Love and Scott Smiles, for their help on the farm work. I cannot achieve my goals without all their help.

I would like to acknowledge my research and salary support from Penn State

Department of Entomology for all these five years, Penn State native bee research project during summer 2008, WHO measles burden research project during summer 2009, and

Pennsylvania Department of Agriculture invasive grape berry moth project during summer 2010. They have graciously supported my research and my life during my study at Penn State.

Last but not least I would like to thank my family and friends for their tremendously selfless material and spiritual support throughout my life and academic

xv career, especially my charming and supportive girlfriend, Qian Xu. Without them my study would not have been possible.

1 Chapter 1

Introduction

What is climate change?

Climate change is one of the most important and controversial topics in scientific research and people’s daily life. As of 2011, a Google search returned more than 50 million websites and more than 1.5 million papers related to climate change

(scholar.google.com). Today the term climate change (as well as global warming, and relatively recently global cooling) is mentioned repeatedly in the press, and people are more or less informed of it. However, there are opinions about climate change that are incorrect or misleading, and which only emphasize the negative aspects of recent climate change (Shackley and Wynne 1996, Schneider 1997, Sonnet 2010, Russill 2010).

Far from being an evil, climate change is a natural process and happens all the time.

There would be no life without climate change— the newborn earth, at around 4.6 billion years ago, was a cauldron of hot melting lava and toxic gases, very similar to current conditions on Venus (Wilde et al. 2001). The paleoclimate, with an average temperature of 50-80ºC, was too harsh for any form of life to evolve and flourish. The brutal climate slowly alleviated to around 20-30 ºC 1.5 billion years ago (Robert and Chaussidon 2006), when we saw the first self-replicating RNA, the first primitive cell, and the dawn of life.

However climate change is not always bliss to life – at least not all life forms when the change is dramatic. The famous Cretaceous- Tertiary (K/T) mass extinction, which wiped all dinosaurs out, was due to the (in the geological time scale) rapid climate change after an asteroid hitting the earth (Courtilotte 1999). But the K/T mass extinction is by far not the most destructive one. The earlier Permian-Triassic (P/T) mass extinction terminated 2 the dominance of trilobes, and killed more than 80% of all insect genera at that time

(Lanbadeira and Sepkoski 1997, Jin et al. 2000).

Nevertheless, after every dramatic climate change period (and the successive mass extinction event), survivors struggled to thrive again. Were there no such climate changes, we might still live in a trilobe- or dinosaur- dominated world – in the other world, mammals might never have evolved to Homo sapiens (Dawkins 2004). Thanks to global cooling during the Ice Age, major continents were interconnected by land bridges, and our ancestors could cross the Bering Strait and enter the Americas (Goebel et al.

2008).

Climate change, being a natural process, has its own internal and external propulsions. The earth is one of the eight planets in the solar system and a tiny fraction of all solar power could warm up our planet and drive all atmospheric, oceanic, and biological activities directly or indirectly. More than 95% of the solar radiation is in the form of short wave radiation (0.29-3um), and more than 99% of terrestrial radiation is long wave (3-100 um). Solar radiation is not constant from year to year and we have dimming and brightening events (Pinker et al. 2005). These events, even though imperceptible to our senses, could steadily affect evaporation (Roderick and Farquhar

2002), soil moisture (Robock et al. 2005), hydrological cycles (Rosefeld 2000), daily temperature range (Travis et al. 2002), and wind systems (Rotstayn and Lohmann 2002).

Earth’s climate is also influenced by space weather and cosmic ray effects, which significantly influence the earth’s cloud coverage and temperature (Svensmark 2000). In addition, variations in the earth’s orbital characteristics are strongly correlated with climate change. The orbital induced climate change caused the Ice Age (Hays et al. 1976). 3 Besides these external factors, earth’s own volcanic activity helps shift the climate,

especially atmospheric circulation (Kodera 1994). All these factors together influence

and shape earth’s climate.

Evidence and effect of climate change: Climate Change is not only Global Warming

Evidence of recent global climate changes are undeniable. Global warming and an increased greenhouse effect could be the most evident clues of climate change (IPCC

2007a, IPCC 2007b). The cause of global warming is the excessive amount of greenhouse

gases in the atmosphere. The earth would be icy cold (-30 ºC) if there were no

greenhouse effect. The greenhouse gases are primarily in the form of carbon dioxide

(CO2) and methane (CH4), which currently contribute about 81% of the total radiative

forcing. Although CO2 has a much higher concentration in the atmosphere than CH4, the

latter has more than 25 times more radiative ability than CO2. These greenhouse gases

heat the atmosphere in an indirect way. They reabsorb the long wave radiation emitted

from the earth, preventing this energy from leaking to the space. This is very similar to

how greenhouses function and is where the term greenhouse effect comes from

(Bursinger and Fleagle 1980). Nevertheless, global warming is not the only aspect of

global climate change. We have already observed many other indicators: changes in

atmospheric circulation, especially long term trends in expansion of the lower

stratosphere (Rosenlof 2002); weather changes in tropic and mid latitudes – the well

known El Nino Southern Oscillation effect (ENSO), which leads to increased

precipitation on the central and eastern tropical Pacific coast, while decreasing

precipitation in the western Pacific (Philander 1989); rising sea levels due to the melting 4 ice from glaciers and ice caps, which already threatens many coastal regions and islands

(Munk 2002); ocean acidification, which is probably due to increased CO2 concentration dissolved in ocean water and impacts directly on pH-sensitive plankton (Royal Society

2005); changes in ice sheets , most notably the ice cover shrinkage near the poles (IPCC

2007b); and coastline degradation (Valiela 2006), etc. All these observations confirm that our current climate is changing at an abnormally high rate.

Through the industrialization starting in the 1800s, Homo sapiens has significantly altered the atmosphere’s composition and thus the climate. The concentration of CO2 has been increasing from 280ppmv (parts per million by volume) in

1748 AD to 379 ppmv currently – an approximately 40% increase. The increase of CH4 concentration is even more dramatic: from 0.72ppmv to 1.77ppmv, a relative increase of more than 120% (IPCC 2007b, Letcher 2009). Meanwhile the mean temperature of the global surface has increased by 1.5 ºC (NOAA 2009). Although we cannot conclude that there is a simple linear relationship between global warming and increased concentration of greenhouse gases (for instance, the mean temperature between 1750 AD and 1840 AD was decreasing while greenhouse gas concentration was increasing), there is undoubtedly a significant and strong correlation between them, especially in the past 50 years (Hansen and Sato 2004).

Ecological responses to climate change, or ants on the hot pot: why insects are especially sensitive to climate change

5 The Intergovernmental Panel on Climate Change (IPCC 2007a) projects increases in global mean surface temperatures of 1.1-6.4 ºC by the end of the 21st century if greenhouse gas emissions continue to increase at current rates. Such changes in climate will likely impact many ecosystem functions, especially many natural processes that are temperature dependent (Wing et al. 2005, Deutsch et al. 2008, Inouye 2008). Plants are a fundamental keystone of terrestrial ecosystems. Increased temperatures will influence plant growth, soil nutrient cycling, and water evaporation and transportation, (Midgley et al. 2007). Studies show substantial evidence that the phenology of some plants has been altered because the spring warming is becoming earlier by approximately 2-5 days (IPCC

2007b, Myneini et al. 1997). Both the tempo and spatial distribution of temperature is altered, with a poleward and altitude shift (Kelly and Goulden 2008). One of the most important ecosystems to humans, the agroecosystem, is also impacted by rapid climate change. Changes might be beneficial because of increased concentration of CO2 and temperature, at least in the short time scale (Hulme 2005). C3 plants would grow better than C4 plants with increased CO2 level, and Crassulacean acid metabolisms (CAM) plants should adapt to increased temperature. But warmer environments will also influence population dynamics and damage from arthropods and pathogens , making it is difficult to predict their impact in a warmer climate(Dixon 2006). Sea life (pelagic and planktonic ecosystem), which is usually overlooked by the public, plays a fundamental role in regulating global environment and producing biomass. More than 99% of pelagic and planktonic organisms are poikilotherms and they are extremely sensitive to changing temperature (Atkinson and Sibly 1997). Many studies have highlighted that sea life are responding to changes such as sea surface warming, precipitation, and wind patterns. 6 Both community composition and species abundance are influenced by climate changes

(Edwards and Richardson 2004). Coral reefs are especially negatively impacted by climate change. Many are bleaching in response to rising sea temperature. Today coral reefs are considered as one of the most vulnerable systems under climate change

(Peterson et al. 2008).

Unlike plants, animals are much more mobile and their responses and adaptations to climate change are easier to observe. Researchers have confirmed range shifts in many birds. For instance, in North America birds are moving northward, with an average shift of about 2.4 km/yr (Hitch and Leberg 2007). For migratory species, the mean advancing time of blackbirds is about 17 days based on a 47 year study (Huppop and Huppop 2003).

Avian reproduction timing, is also advancing (Crick et al. 1997). Mammals also show temporal and spatial responses to recent climate change. Some long-term studies have confirmed Bergmann’s rule that a warmer climate would cause smaller body size in mammals. For instance, body size of woodrats in New Mexico has decreased by 15% on average in the last 10 years, coinciding with an increase of 3 ºC of summer temperature

(Smith et al. 1998). Mammal distribution and abundance are also influenced by climate change. Currently, warm area expansion results in increased mammalian diversity

(Gaston 2000), but imposes significant threats to polar mammals which often rely on the presence of frozen seas.

Smaller animals, such as insects, are more appropriate model candidates on which to study the effects of climate change. First, insects are poikilotherms and hence their internal temperature is highly dependent on ambient temperature. Consequently, insect development is driven primarily by temperature (Stinner et al. 1975, Logan et al. 1976, 7 Pruess 1983, Tauber et al. 1986, Lowry and Lowry 1989, Wagner et al. 1984). Second,

many insects have relatively short life spans, which is conducive to the development of

laboratory and field-based research designed to measure the impact of climate change. In

multivoltine taxa, the number of generations per year under current and projected

climatic regimes can be evaluated to quantify the influence of changing climates

(Yamamura and Kiritani 1998, van Asch et al. 2007, Post et al. 2008). However, many

other factors, both biotic and abiotic, could also influence insect seasonality. For

example, many insects diapause and often the main induction factor is photoperiod

(Tauber et al. 1986, Denlinger 2002). Furthermore, the nonlinear interaction of

temperature and photoperiod makes it challenging to investigate and understand changes

in insect voltinism in response to climate change (Tobin et al. 2008). Although universal

predictions of how climate change would influence insect phenology is not feasible,

doing so for representative insect species could be valuable in efforts to understand the

consequences of climate change in a general context.

Understanding key drivers of insect voltinism, such as development and diapause,

are essential in efforts to quantify the consequences of climate change on insect

seasonality (Logan et al. 2003). Such drivers are known for the grape berry moth,

Paralobesia viteana (Clemens) (Lepidoptera: Tortricidae) (Nagarkatti et al. 2001, Tobin

et al. 2001, 2002, 2003). P. viteana is native to North America and feeds on native wild and cultivated (not necessarily native) Vitis host spp., and can be an economically important pest in the latter (Riedl and Taschenberg 1984). It is also non-migratory, and movement patterns are spatially limited; thus, life stages are not subjected to different temperature profiles, and projections of P. viteana at a given location are reasonable 8 predictors of the effects of climate change. Adults emerge in spring from diapausing

pupae, mate, and females oviposit on Vitis spp. flowers or fruit. Upon hatching, larvae burrow into the fruit, exiting to pupate in leaves or bark. Pupae emerge as adults that initiate subsequent generations. As day length decreases following the summer solstice, there is a critical photoperiod independent of temperature that initiates diapause, and eggs deposited at these decreasing photoperiods will develop into diapausing pupae

(Nagarkatti et al. 2001). Hence, as in many temperate insect species, voltinism is influenced by the degree-day accumulation prior to the arrival of photoperiodic conditions that initiate diapause. There are generally 2-3 generations per year along the grape production belt along Lake Erie (Tobin et al. 2003). Tobin et al. (2008) previously highlighted the consequence of interactions between temperature (subject to climate change) and photoperiod (not subject to climate change) in driving P. viteana voltinism.

Disease dynamics among pathogens, vectors and hosts under climate change has been

studied intensively as well. Rogers and Packer (1993) first systematically introduced the

relationship between vector born diseases (VBD) and global climate change. Daszak's

group (2000) introduced three different types of emergence patterns of infectious disease:

from domestic to wild life; related to humans via host/parasite translocation; and with no

human or domestic animals. Lemon's workshop summary (Lemon et al. 2008) is a good

synthesis of relationships between climate change and VBD. While Lemon focuses on

biological aspects of infectious disease and introduces the famous VBD triad (pathogen-

host-vector), other researchers (Mayer and Pizer 2005) give more attention to the

sociological aspect of infectious diseases. 9 There is still much debate on the causal relationship between climate change and increasing VBD (Semenza and Menne 2009). It is unsafe to simply infer that because temperature is rising and vectors are poikilotherms, we will see more VBD cases because of more rapid population growth of vectors. This reasoning overlooks the complex, nonlinear dynamics between different populations (vector, host, pathogen, predator, etc.) in the ecosystem. Lafferty ignites a heated debate (Lafferty, 2009a) by suggesting there is not necessarily a causal relationship between climate change and infectious diseases.

Lafferty (2009a) argued that there are too many intertwined socioeconomic factors to posit a direct relationship between climate and disease dynamics. I believe more powerful mathematical models using better and more reliable data sources could improve our understanding of future disease dynamics. After Lafferty's paper, five other researchers gave their comments, either pro or con, to his idea. Harvell's group (2009) believes that not only the pathogen but also the host life history is influenced by climate change.

Besides the external environmental factors such as temperature, humidity, other intrinsic factors like host physiology, especially immune responses, play a key role in the host- pathogen dynamics. Hence I should not focus only on the abiotic factors. Ostfeld casts his suspicion on Lafferty's reasoning that disease might shift geographically rather than expand (Ostfeld 2009). He argues we should consider non-climate factors. Also we should carefully consider if transmission rates really exceed 1, which will yield totally different ecological outcomes. While Pascual agrees VBD might shift rather than expand

(Pascual et al. 2009), she notes that human distribution on the earth is not homogeneous so that even if the infectious disease shifts its range rather than expands, it might cause significant health problem to those areas where people have never experienced those 10 diseases (and thus carry no immunity to them). Consequently, this might lead to epidemics of some diseases. Randolph (Randolph 2009) says we should have a better understanding of how extrinsic and intrinsic forces influence disease dynamics. Dobson's team has expanded the single pathogen single host system to multiple hosts, and multiple pathogens dynamics (Dobson et al. 2009). They propose a fine-tuned mathematical model (TSEIR, time series exposed-infectious-recover) to actually see the disease dynamics and expand it to multi-species scenario. As a summary of this forum, Lafferty

(Lafferty 2009b) concludes that climate change may actually reduce the spread of some infectious diseases such as malaria.

Among many VBDs, West Nile Virus (WNV) is relatively new to North America.

It was introduced accidentally via shipment from Africa to New York in 1999. WNV belongs to the Flaviviridae, which is a group 4 positive single stranded RNA (+ss RNA) virus. Its RNA is usually 10-12kb nucleotides long. WNV primarily infect birds especially the family Corvidae (crows and jays). Mammals, including humans, horses, and rodents, are other potential hosts. The vectors are mosquitoes especially those in

Culex genus (Family Culicidae) such as Cx pipiens. Ever since its arrival in the US,

WNV spread quickly westward and now all of the continental US has reported human and bird cases. The rapid spread of WNV is largely due to the high mobility of its primary host, migratory (Careful…I don’t think crows are considered migratory) birds.

However, WNV did not draw much attention until 3-4 years later when people realized its threat not only to wild life, especially bird populations, but also to our health as well.

A total of 149 human infections were reported between 1999 and 2001, including 18 deaths. Pennsylvania first saw a WNV case in 2001, 2 years after its invasion from New 11 York, with 3 cases. However, things have been getting worse since then. During the

2003 WNV outbreak, almost 10,000 cases were reported with 264 fatalities. In 2007 the

infection dropped to 3,623 and in 2008 only 236 cases were reported.

Since then researchers have paid attention to the relationship between WNV, its

vector mosquitoes, and environmental factors. Bolling's group in Texas (Bolling et al.

2005) identified four primary vector mosquitoes in their region: Aedes vexans, Culex

erraticus, Culex salinarius, and Psorophora columbiae. They found that environmental

factors such as temperature, precipitation, dew point, and canopy coverage are important

to determine dominant species; hence we could utilize different strategies to deal with

different dominant species. Cruz-Pacheco's team has focused on mathematical modeling

(Cruz-Pacheco et al., 2005) and they used R0, the intrinsic growth rate related to the vector species, as an indicator of WNV disease dynamics to test whether the disease dies out or become epidemic. They also showed that R0 is related to temperature. Shaman and

other researchers have studied the relationship between drought/wet weather and WNV

epidemics (Shaman et al., 2005). They investigated how spring drought influences

occurrence of WNV in birds (as well as humans). Ward (2005) applied a degree-day

concept into a model to predict occurrence of WNV in horse populations in Indiana. He

also showed that the distribution of WNV is heterogeneous across the state. Kilpatrick

and his fellow researchers have been studying WNV dynamics for quite a few years and

have published a series of papers. They used the Galapagos islands as an ideal

environment to simulate WNV invasion (Kilpatrick et al., 2006a) because the Galapagos

are surrounded by WNV threats and may have different paths of WNV introduction:

airplane, wind, boat, human-related bird, migratory birds, and humans. They utilized a 12 simulation model to assess the possibility of WNV infection in Galapagos. The model shows that introductions by airplane carry the most mosquitoes with pathogens into the

Galapagos. In another paper (Kilpatrick et al., 2006b) they challenged the previous conclusion that crows are most responsible for WNV spread. Instead, a single relatively uncommon robin species could cause mosquitoes to feed and transfer the pathogen to other hosts. In another paper, they focused on host change (Kilpatrick et al. 2006c). Culex pipiens is the primary enzootic and bridge vector of WNV. During late summer and early fall, it shifts its preference from bird to human as much as 7 times, which might cause

WNV epidemics in humans. Cx. tarsalis is another vector in Colorado and California.

These host shift studies are very useful for predicting starting time and intensity of human

WNV epidemics. Later in 2007, LaDeau et al. (2007) summarized almost 30 years of population dynamics of North American bird populations. WNV significantly impacts

North America bird populations, especially American crows, which lost about 45% of their population in 26 years.. In 2008, they combined temperature, viral genetics and transmission of WNV by Cx. pipiens to study WNV disease dynamics, which links extrinsic and intrinsic factors, and their ideas are very helpful for future research.

Lampman traced Cx. dynamics and WNV transmission in Illinois (Lampman et al., 2006).

They focused on Cx. pipiens and Cx. restuans. Their dates of crossover (when the proportion of egg rafts from both species is equal) was said to be linearly dependent with date of last spring frost.

Outside the US, European researchers are also concerned with WNV. Poncon et al.

(2007) studied the population dynamics of mosquitoes in France: Aedes caspius (Pallas),

Culex modestus (Ficalbi), Culex pipiens L. and Anopheles hyrcanus (Pallas) which are 13 vectors of WNV and malaria. Their life history differs significantly with ambient

temperature, rainfall, and seasonality. While researchers and the public are worried about

WNV epidemics, Snappin et al. 2007 showed the growth rate of WNV in North America

was declining. There are two strains of WNV, WN02 and NY99. The WN02 strain had

almost 3 times more epidemiological growth rate than the NY99 strain; however even the

WN02 strain had a decreased growth rate, so WNV may have already reached its peak

prevalence in North America. Their findings coincide with declining human and birds

WNV cases in North America. Bouden et al. (2008) developed a temporal-spatial

dynamics model of mosquitoes and birds populations. DeGroote argued different

landscapes, demographics, and climates will cause different WNV disease patterns. Drier

conditions tend to be associated with WNV diseases (DeGroote et al. 2008). Trawinski

and Mackay (2008) consider these meteorological factors in a SARIMA model (seasonal

Auto Regressive Moving Average) and predict population dynamics of Cx. pipiens and

restuans. The factors include: Tmin, Tmax, Tavg, RH (relative humidity), and

P(precipitation). Gale et al. (2009) assessed the impact of climate change and different disease dynamics: African swine fever (ASFV), WNV, RVFV, Crimean Congo haemorrhagic fever (CCHFV), and African Horse sickness (AHSV). They were assessed by routes of entry: vectors, livestock, meat products, wildlife, pets and human. Climate change was reported to increase all these diseases, and AHSV, CCHFV, and WNV most significantly.

Degree day models have been applied to determine WNV emergence in various locations (Ward et al. 2005, Bohm et al. 2006, Zou et al. 2007, Kilpatrick et al. 2008,

Konrad et al. 2009). While more complex and sensitive models have been developed 14 (Bolling et al. 2005, Tachiiri et al. 2006, Gong et al. 2010), there is always a trade-off between model accuracy and complexity, and I still think degree day models are useful because of their simplicity. .

Importance of variability in the daily temperature range (DTR): too trivial to consider?

So far, I have argued that the rapidly changing climate would impact our planet earth, especially (almost) all ecosystems. Of all potential candidates to model the impact of this change, insects are an appropriate group. While most of the current references focus on long-term (annual, multi-annual, and some periodical) trends due to climate change , very little effort is given to daily temperature range variability under climate change. This is, from my perspective, ignoring some basic knowledge of insect life history because insects do experience temperature variation in a day. The constant temperature provided in a lab’s incubators never occurs in nature. Hence I might need to consider how daily temperature range (DTR) influences insect life history.

In short, DTR is closely related to global radiation (Eg), which is the total solar radiation absorbed on earth’s horizontal surface (the bottom of atmosphere, Bristow and

Campbell 1984). There are still controversial arguments whether DTR will increase or decrease under current climate change conditions. A 40 year study shows DTR increased in Yom Kippur, Israel by 0.31 ºC, and other places had an 18% increase (Stanhill and

Moreshet 1992, Stanhill and Cohen 1997). But others argue the increase in daily minimum is faster than that of maximum temperature, hence DTR is actually decreasing

(Roderick and Farquhar 2002). Some researchers are using DTR change to study the 15 influence of Eg on global temperatures (Wild et al. 2007). In our research I focus on increased DTR and its impact on insects.

Currently, a few research projects have been done and showed how DTR would influence vector mosquito dynamics in Africa (Paajimans et al. 2009, 2010). However I need to understand the impact of changing DTR more explicitly and quantitatively.

Research Goals : predicting insect phenology, and vector borne disease dynamics, under changing mean and daily temperature range conditions

In this thesis, I focus on understanding how changes in both mean and DTR influence agricultural pests in agroecosystems and vectors in disease dynamics. I first use degree day models to estimate voltinism and phenology of a certain pest and vector. I then add different DTR conditions to see how they respond to climate change in a much finer temporal scale.

I extend the work on how changing mean climate would influence voltinism

(Patrick et al. 2008) by examining the interaction of temperature and photoperiod based upon scenarios of the rate of greenhouse gas emissions and their effect on temperature

(Hayhoe et al. 2007). In Chapter II I develop individual-based Monte Carlo methods to explore how climate change can alter insect voltinism under varying greenhouse gas emissions scenarios using input distributions of diapause termination or spring emergence, development rate, and diapause initiation, linked to daily temperature and photoperiod. I show concurrence of these projections with a field dataset, and then explore changes in grape berry moth, Paralobesia viteana (Clemens), voltinism that may occur with climate projections developed from the average of three climate models using 16 two different future emissions scenarios from the International Panel of Climate Change

(IPCC). Based on historical climate data from 1960-2008, and projected downscaled climate data until 2099 under both high (A1fi) and low (B1) greenhouse gas emission scenarios, I used concepts of P. viteana biology to estimate distributions of individuals entering successive generations per year. Changes in voltinism in this and other species in response to climate change will likely cause significant economic and ecological impacts, and the methods presented here can be readily adapted to other species for which the input distributions are reasonably approximated. Thus, I present an individual-based

Monte Carlo approach to quantify changes in voltinism and highlight not only generalized trends but the variability among individuals within a population, and provide open-source computer code to enable these methods to be extended to other species.

In the next step (Chapter III) I emphasize the effects of DTR and quantify how various DTR conditions influence P. viteana phenology. As stated before, the life history of P. viteana has been well studied, and the phenology of P. viteana is determined primarily by the interaction of ambient temperature, which is subject to climate change, and photoperiod, which is considered invariable (Nagarkatti et al. 2001, Tobin et al.

2002, Tobin et al. 2003, Tobin et al. 2008, Chen et al. 2011 in press). I first demonstrate why and how different degree day accumulation models yield different results (mean temperature model is compared to a sine curve model), then add various DTR conditions to current climate data to see its effect on degree day accumulation; finally I examine how these DTR conditions affect P. viteana phenology. 17 In the following chapter (Chapter IV), I shift to a vector-disease system and study

the relationship between temperature and Culex mosquito phenology. Environmental

factors, especially temperature, play an important role in WNV dynamics because the life

history of its major vectors, mosquitoes, depend on ambient temperature. Mosquito

longevity, development rate, and fecundity can all be modeled as functions of

temperature (Dohm et al. 2002a, Bellan et al. 2010). From an epidemiological perspective,

the extrinsic incubation period (EIP) is critical for determining whether the vectors

carrying pathogen can be infectious, and hence have the capacity to transmit the pathogen

effectively. Reisen et al. (2006) showed that the inverse of EIP (or, equivalently, the

extrinsic incubation rate) of WNV is linearly proportional to temperature and that the

median EIP of WNV is 109 degree days based on experiments for Culex tarsalis.

One of the most commonly used method to determine potential WNV emergence period

is proposed by Zou et al. (2007). This method works well in Wyoming. However it might

not be directly applicable to Pennsylvania. First, it treats EIP as a constant across the

season. Konrad et. al (2007) argues this method does not perform well in Santa Babara,

California unless the EIP is adjusted to about 76 degree days. Secondly, it assumes

mosquito longevity is the same across the season, while in fact mosquito longevity is

very sensitive to ambient temperature. Furthermore the specific parameters work with

Culex tarsalis in the Western US. Only C. pipiens and C. resturans are major WNV vectors in Pennsylvania and I have no idea if these parameters work for C.pipiens and

C.resturans.

In Chapter IV I investigate whether degree day models could: capture minimal

WNV transmission criterion; capture first detection of WNV in the year; and capture last 18 detection of WNV in the year.To answer these questions I first test the simple degree days model (Zou et al. 2007) against field observations collected from four locations in

Pennsylvania (Harrisburg, Philadelphia, Pittsburgh, and Williamsport) from 2002 to 2008 to see if this method also works in Pennsylvania. I demonstrate that a simple degree day model does not capture virus emergence period in these four locations during these 6 years. Then I propose adjustments by considering variable EIP and longevity in different times of the year. I suggest Extrinsic Incubation Period (EIP) variable at the beginning and end of the season and adult mosquito longevity a variable as well. The calibrated models improve the prediction of the WNV emergence period in Pennsylvania. These models also provide useful tools to assess potential WNV risk in the future, and could be easily adapted to other regions with different vectors when proper parameters are provided.

In the next research step (Chapter V), I still use Zou’s and Konrad’s approaches as our basic modeling framework (Zou et al. 2007, Konrad et al. 2007) but with our own calibrations (see Chapter IV for details). I quantify the effect of various DTR conditions on the WNV transmission period. I first demonstrate why and how different degree day accumulation models yield different results (mean temperature versus sine curve model), then add various DTR conditions to current climate data to see the effect of DTR on degree day accumulation; finally I examine how these DTR conditions affect WNV transmission in different locations in Pennsylvania. 19

References Cited Atkinson, D. and R. M. Sibly. 1997. Why are organisms usually bigger in colder environments? Making sense of a life history puzzle. Trends Ecol. Evol. 12: 235- 239. Bale, J.S., G. J. Masters, I. D. Hodkinson, C. Awmack, T. M. Bezemer, V. K. Brown,

J. Butterfield, A. Buse, J. C. Coulson, J. Farrar, J. E. G. Good, R.

Harrington, S. Hartley, T. H. Jones, R. L. Lindroth, M. C. Press , I.

Symrnioudis, A. D. Watt and J. B. Whittaker. 2002. Herbivory in global

climate change research: direct effects of rising temperature on insect herbivores.

Glob. Ch. Biol. 8: 1-16.

Bellan, S.E. 2010. The importance of age dependent mortality and the extrinsic

incubation period in models of mosquito-borne disease transmission and control.

PLoS Pathogen. 5: e10165.

Bolling, B. G., J. H. Kennedy, and E. G. Zimmerman. 2005. Seasonal dynamics of four potential West Nile vector species in north-central Texas. J Vector Ecol. 30: 186–194. Bouden, M., B. Moulin, and P. Gosselin. 2008. The geosimulation of West Nile virus propagation: a multi-agent and climate sensitive tool for risk management in public health. Int J Health Geogr. 7: 7–35. Bristow, R. L. and G. S. Campbell. 1984. On the relationship between incoming solar radiation and daily maximum and minimum temperature. Agriculture and Forestry Meteorology 31:159--166. Businger, J. A., and R. G. Fleagle 1980. An introduction to atmospheric physics.

International geophysics series (2nd ed.). Cambridge University Press. 20 Courtillot, V. 1999. Evolutionary Catastrophes: The Science of Mass Extinction.

Cambridge University Press.

Couzin, J. 2004. Hybrid Mosquitoes Suspected in West Nile Virus Spread. Science 303:

5663.

Crick, H.Q.P., C. Dudley, D. E. Glue, and D. L. Thompson. 1997. UK birds are laying

eggs earlier. Nature 388: 526.

Cruz-Pacheco, G., L. Esteva, A. Montao-Hirose, and C. Vargas. 2005. Modelling the dynamics of West Nile Virus. Bull Math Biol. 67(6): 1157–1172. Curreano, E. D., P. Wilf, S. L. Wing, et al. 2008. Sharply increased insect herbivory

during the Paleocene-Eocene Thermal Maximum. Proc. Natl. Acad. Sci. USA 105

(6): 1960-1964.

Dawkins, R. 2004. The ancestor’s tale: a pilgrimage to the dawn of evolution, 3rd ed

.Houghton Miffin Harcourt Press.

Daszak, P., A. A. Cunningham, and A. D. Hyatt. 2000. Emerging infectious diseases of wildlife – threats to biodiveristy and human health. Science 287: 443–449. DeGroote, J. P., Sugumaran, R., Brend, S. M., Tucker, B. J., and Bartholomay, L. C. 2008. Landscape, demographic, entomological, and climatic associations with human disease incidence of West Nile virus in the state of Iowa, USA. Int J Health Geogr. 1: 7–19. Denlinger, D.L. 2002. Regulation of diapause. Ann. Rev. Entomol. 47: 93-122. Deutsch, C.A., J. J. Tewksbury, R. B. Huey, et al. 2008. Impacts of climate warming on terrestrial ectotherms across latitude. Proc. Natl. Acad. Sci. USA 105 (18): 6668-6672. Dixon, G. R. 2006. The biology of Plasmodiophora brassicae wor. – a review of recent

advances. Acta Hortic 706: 271-282. 21 Dobson, A. 2009. Climate variability, global change, immunity, and the dynamics of

infectious diseases. Ecology 90: 920–927.

Dohm, D. J., M. L. O’Guinn and M. J. Turell. 2006. Effect of environmental

temperaeture on the ability of Culex pipiens (Diptera: Culicidae) to transmit West

Nile virus. J. Med. Entomol. 39: 221-225.

Dohm, D. J., M. R. Sardelis and M. J. Turell. 2006. Experimental vertical trnamission

of West Nile virus by Culex pipiens (Diptera: Culicidae). J. Med. Entomol. 39:

640-644.

Edwards, M. & A. Richardson. 2004. Impact of climate change on marine pelagic

phenology and trophic mismatch. Nature 430: 881-884.

Gale, P., A. Brouwer, V. Ramnial, L. Kelly, R. Kosmider, A. Fooks, and E. Snary.

2009. Assessing the impact of climate change on vector borne viruses in the EU

through the elicitation of expert opinion, Epidemiol. Infect. 1: 1–12.

Gaston, K. J. 2000. Global patterns in biodiversity. Nature 405:220-227.

Gingrich, J.B., and L. Casillas. 2004. Selected mosquito vectors of West Nile virus:

comparison of their ecological dynamics in four woodland and marsh habitats in

Delaware, J Am Mosq Control Assoc. 20: 138–145.

Goebel, T., M. R. Waters, and D. H. O'Rourke. 2008. The Late Pleistocene Dispersal

of Modern Humans in the Americas. Science 319: 1497–1502.

Gong, H, A. T. DeGaetano, and L. C. Harrington. 2010. Climate-based models for

West Nile Culex mosquito vectors in the Northeastern US. Int. J. of Biometeorol.

DOI: 10.1007/s00484-010-0354-9. 22 Hansen, J. and M. Sato. 2004. Greenhouse gas growth rates. Proc Natl Acad Sci USA 101: 16109-16114. Hartmann, D. L. 1994. Global Pysical Climatology. Academic Press Harvell, Drew, S. Altizer, I. M. Cattadori, L. Harrington, and E. Weil. 2009. Climate change and wildlife diseases: when does the host matter the most?. Ecology 90(4): 912–920. Hayhoe, K., C.P. Wake, T.G. Huntington, L. Luo, M. Schwartz, J. Sheffield, E.

Wood, B. Anderson, J. Bradbury, A. DeGaetano, T. Troy, and D. Wolfe.

2007. Past and future changes in climate and hydrological indicators in the U.S.

Northeast. Climate Dynamics 28:381-407.

Hays, J. D., J. Imbre, and N. J. Shackleton. 1976. Variations in the Earth's orbit:

Pacemaker of the Ice Ages. Science 194: 1121–1131.

Hulme, P. E. 2006. Adapting to climate change: is there scope for ecological

management in the face of global threat?. J. Applied Ecology 42: 784–794.

Huppop, O. and K. Huppop. 2003. North Atlantic oscillation and timing of spring

migration in birds. Proc. R. Soc. Lond. B 270: 233-240.

IPCC. 2007a. Summary for policymakers: 4th report. http://www.ipcc.ch/ipccreports/ar4-

wg1.htm.

IPCC. 2007b. Climate Change 2007: The Physical Science Basis. Contribution of

Working Group I to the Fourth Assessment Report of the Intergovernmental Panel

on Climate Change (AR4). Solomon, S., D. Qin, M. Manning, Z. Chen, M.

Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.), Cambridge University

Press, Cambridge, UK and NY, USA. 23 Jin, Y.G., Y. Wang, W. Wang, Q. H. Shang, C. Q. Cao, and D. H. Erwin. 2000.

Pattern of Marine Mass Extinction Near the Permian–Triassic Boundary in South

China. Science 289: 432–436.

Karl, T. R. and K. E. Trenberth. 2003. Modern Global Climate Change. Science

302:1719-1723.

Kelly, A. E. and M. L. Goulden. 2008. Rapid shifts in plant distribution with recent

climate change. Proc. Natl. Acad. Sci. USA 105:11591-11592.

Kent, R., L. Juliusson, M. Weissmann, S. Evans, and N. Komar. 2009. Seasonal blood feeding behavior of Culex tarsalis in Weld County, Colorado, 2007. J. Med Entomol. 46(2): 380–390, 2009. Kilpatrick, A. M., L. D. Kramer, M. J. Jones, P. P. Marra, and P. Daszak. 2006. West Nile virus epidemics in North America are driven by shifts in mosquito feeding behavior, PLoS Biol. 4(4), 82 Kilpatrick, A. M., P. Daszak, S. Goodman, H. Rogg, L. D. Kramer, V. Cedeno, and A. Cunningham. 2006. Predicting pathogen introduction: West Nile Virus spread to Galapagos. Conservation Biology 20(4): 1224–1231. Kilpatrick, A. M., P. Daszak, M. J. Jones, P. P. Marra, and L. D. Kramer. 2006. Host heterogeneity dominates West Nile virus transmission. Proc. R. Soc. B 273: 2327–2333. Kilpatrick, A. M, M. A.Meola, Moudy, M. Robin, and L. D. Kramer. 2008. Temperature, Viral Genetics, and the Transmission of West Nile Virus by Culex pipiens Mosquitoes. PLoS Pathog. 4: e1000092. Kodera. K. 1994. Influence of volcanic eruptions on the troposphere through

stratospheric dynamical processes in the northern hemisphere winter. J. of

Geophy Res. 99: 1273-1282. 24 Konrad, S. K., S. N. Miller, W. K. Reeves, and N. S. Tietze. 2009. Spatially explicit

West Nile virus risk modeling in Santa Clara County, California. Vector Borne

Zoonotic Diseases 9: 267-274.

Labandeira, C. C. and J. J. Sepkoski. 1993. Insect diversity in the fossil record.

Science 261: 310–315.

LaDeau, S. L., A. M. Kilpatrick, and P. P. Marra. 2007. West Nile virus emergence and large-scale declines of North American bird populations. Nature 447: 710– 713. Lafferty, K. D. 2009. The ecology of climate change and infectious diseases. Ecology 90(4): 888–900. Lafferty, K. D. 2009. Calling for an ecological approach to studying climate change and infectious diseases. Ecology 90(4): 932–933. Lampman, R., M. Slamecka, N. Krasavin, K. Kunkel, and R. Novak. 2006. Culex

population dynamics and West Nile virus transmission in east-central Illinois. J

Am Mosq Control Assoc. 22(3): 390–400.

Lemon, S. M., P. Sparling, H. Frederick, A. Margaret, D. A. Relman, E. R. Choffnes, and A. Mack. 2008. Vector- Borne diseaes: understanding environmental, human healt hand ecological connections. the National Academies Press. Letcher, T. M. 2009. Climate change: observed impacts on planet earth. Elesevier.

Logan, J. A., D. J. Wollkind, S. C. Hoyt, and L. K. Tanigoshi. 1976. An Analytic

Model for Description of Temperature Dependent Rate Phenomena in

Arthropods. Environ. Entomol. 5: 1133-1140.

Logan, J. A., J. Regniere, and J. A. Powerll. 2003. Assessing the impacts of global

warming on forest pest dynamics. Front Ecol. Environ. 1: 130-137. 25 Lovejoy, T. E., and L. Hannah. 2005. Climate Change and Biodiversity. Yale

University Press.

Lowry, W. P. and P. P. Lowry II. 1989. Fundamentals of Biometeorology (volume 1),

Peavine Publications, McMinnville.

Mayer, K H., and H. F. Pizer. 2008. The social ecology of infectious diseases. Academic Press. Meehl, G. A. and C. Tebaldi. 2004. More intense, more frequent and longer lasting heat

waves in the 21st century. Science 305: 994-997.

Midgley, G.F., S. L. Chown, and B. S. Kgope. 2007. Monitoring effects of

anthropogenic climate change on ecosystems: A role for systematic ecological

observation?. South African Journal of Science 103:282-286.

Munk, W. 2002. Twenty Century Sea Level: an Enigma. Proc Natl Acad Sci USA 99:

6550–6555.

Myneni, R. B., C. D. Keeling, C. J. Tucker, G. Asrar, and R. R. Nemani. 2002.

Increased plant growth in the northern high latitudes from 1981 to 1991. Nature

186: 698–702.

Nagarkatti, S., P. C. Tobin and M. C. Saunders. 2001. Diapause induction in the

Grape Berry Moth (Lepidoptera: Tortricidae). Environ. Entomol. 30: 540-544.

NOAA. 2010.

ftp://ftp.ncdc.noaa.gov/pub/data/anomalies/annual_land_and_ocean.ts.

Nobel, P. S. 1996. High productivity of certain agronomic CAM species. Biochemistry, ecophysiology and evolution. Springer, Berlin. Ostfeld, R. S. 2009. Climate change and the distribution and intensity of infectious diseases. Ecology 90: 903–905. 26 Paaijmans, K. P., A. F. Read, and M. B. Thomas. 2009.Understanding the link

between malaria risk and climate. Proc. Natl. Acad. Sci. USA 106: 13844-13849.

Paaijmans, K. P., S. Blanford, A. S. Bell, J. I. Blanford, A. F. Read, and M. B.

Thomas. 2010.Influence of climate on malaria transmission depends on daily

temperature variation. Proc. Natl. Acad. Sci. USA 107: 15135-15139.

Pascual, M, and M. J. Bouma. 2009. Do rising temperatures matter?. Ecology 90(4): 906–912. Peterson, E. L, M. Beger, and Z. T. Richards. 2008. Thinking outside reef. Science 319: 1759. Pinker, R. T., B. Zhang, and E. G. Dutton. 2005. Do satellites detect trends in surface

solar radiation. Science 368: 850-854.

Poncon, N., C. Toty, G. L'ambert, G. le Goff, C. Brengues, F. Schaffner, and D. Fontenille. 2007. Population dynamics of pest mosquitoes and potential malaria and West Nile virus vectors in relation to climatic factors and human activities in the Camargue, France. Med Vet Entomol. 21(4): 350–357. Post, E. S., C. Pedersen, C. C. Wilmers, and Mads C. Forchhammer. 2008.

Phenological sequences reveal aggregate life history response to climatic

warming. Ecology 89 (2): 363-370.

Pruess, K. P. 1983. Day-Degree Methods for Pest Management. Environ. Entomol. 12:

613-619.

Randolph, S. E. 2009. Perspectives on climate change impacts on infectious diseases, Ecology 90(4): 927–931. Reisen, W. K., Y. Fang, and V. M. Martinez. 2006. Effects of temperature on the

transmission of West Nile virus by Culex tarsalis (Diptera: Culicidae). J. Med.

Entomol. 43: 309-317. 27 Riedl, H. and E.F. Taschenberg. 1984. Grape Berry Moth. Grape IPM Insect

Identification Sheet No.1, New York State Agricultural Experiment Station,

Geneva.

Roberick, M. L. and G. D. Faquhar. 2002. The cause of decreased pan evaporation

over the past 50 years. Science 298: 1410–1411.

Roberit, F. and M. Chaussidon. 2006. A palaeotemperature curve for the Precambrian

oceans based on silicon isotopes in cherts. Nature 443: 969–972.

Robock, A., M. Mu, K. Vinnikov, I. V. Trofimova, and T. I. Adamenko. 2005. Forty-

five years of observed soil moisture in the Ukraine: no summer desiccation yet.

Geo. Res. Letter 32: doi:10.1029/2004GL021914.

Rogers, D. J., and M. J. Packer. 1993. Vector-borne diseases, models, and global change. Lancet 342: 1282–1284. Rosenfeld, D. 2000. Suppression of rain and snow by urban and industrial air pollution.

Science 487: 1793-1769.

Rosenlof, K. H. 2002. Transport changes inferred from HALOE water and methane

measurements. J. Meteorol. Soc. Japan 80:831– 848.

Rotstayn, L. D., and U. Lohmann. 2002. Tropical rainfall trends and the indirect

aerosol effect. J. of Climate 15: 2103-2116.

Royal Society. 2005. Ocean acidification due to increasing atmospheric carbon dioxide.

Policy Document. Royal Society, London.

Russill, C. 2010. Stephen Schneider and the “double ethical bind” of climate change

coomunication. Buletin of Science Technology Society 30: 60-69. 28 Philander, S. G. 1989. El Nino, La Nina, and the Southern Oscillation. Elsevier, New

York.

Schneider, S. 1997. Integrated assessment modeling of global climate change:

transparent rational tool for policy making or opqque screen hiding value-laden

assumptions?. Environmental modeling and assessment 2: 229-249.

Semenza, J. C., and B. Menne. 2009. Climate change and infectious diseases in Europe.

Lancet Infect. Dis. 9: 365–375

Shackley, S. and B. Wynne. 1996. Representing uncertainty in global climate change

science and policy: boundary-ordering devices an authority. Science Technology

Human Values 21: 275-302.

Shaman, J., J. F. Day, and M. Stieglitz. 2005. Drought-induced amplification and epidemic transmission of West Nile virus in southern Florida. J Med Entomol. 42: 134–141. Smith, F. A., H. Browning, and U. L. Shepherd. 1998. The influence of climate change on the body mass of woodrats. Ecography 21: 140–148. Snapinn, K. W., E. C. Holmes, D. S. Young, K. A. Bernard, L. D. Kramer, and G. D. Ebel. 2007. Declining growth rate of West Nile virus in North America. J. Virol. 81(5): 2531–2534. Sonnett, J. 2010. Climates of risk: afield analysis of global climate change in US media

discourse, 1997-2004. Public Understanding of Science 19: 698-716.

Stanhill, G. and S. Moreshet 1992. Global radiation climate change in Israel. Climate

Change 22:121-138.

Stanhill, G. and S. Cohen. 1997. Recent changes in solar irradiance in Antarctica. J.

Climat. 10: 2078-2086. 29 Stinner, R.E., G. D. Butler, J. S. Bacheler, and C. Tuttle. 1975. Simulation of

Temperature-dependent Development in Population Dynamics Model. Can. Ent.

107: 1167-1174.

Tachiiri, K., B. Klinkenberg, S. Mak and J. Kazmi. 2006. Predicting outbreaks: a

spatial risk assessment of West Nile virus in British Columbia. International

Journal of Health Geographics 5: 1-21.

Tauber, M. J., C. A. Tauaber, and S. Masaki. 1986. Seasonal Adaptations of Insects,

Oxford University Press, New York.

Svensmark, H. 2000. Cosmic rays and Earth's climate. Space Sci. Rev. 93: 175-185.

Thomas R. Karl and Kevin E. Trenberth. 2003. Modern Global Climate Change.

Science 302 (5651): 1719 – 1723.

Tobin, P. C., S. Nagarkatti, and M. C. Saunders. 2001. Modeling Development in

Grape Berry Moth (Lepidoptera: Tortricidae). Environ. Entomol. 30: 692-699.

Tobin, P. C., S. Nagarkatti, and M. C. Saunders. 2002. Diapause Maintenance and

Termination in Grape Berry Moth (Lepidoptera: Tortricidae). Environ. Entomol.

31 (4): 708-713.

Tobin, P. C., S. Nagarkatti, and M. C. Saunders. 2003. Phenology of Grape Beery

Moth (Lepidoptera: Tortricidae) in Cultivated Grape at Selected Geographic

Locations. Environ. Entomol. 32: 340-346.

Travis, D.J., A. M. Carleton, and R. Lauritsen. 2002. Contrails Reduce Daily

Temperature Range, Nature 418: 601. 30 Trawinski, P. R., and D. S. Mackay. 2008. Meteorologically conditioned time-series predictions of West Nile virus vector mosquitoes. Vector Borne Zoonotic Dis. 8(4): 505–521. Valiela, I. 2006. Global coastal change. Blackwell, Malden, MA.

Van Asch M, P.H. van Tienderen, L. J. M. Holleman, M. E. Visser. 2007. Predicting

adaptation of phenology in response to climate change, an insect herbivore

example. Global Change Biology 13: 1596-1604.

Wagner, T. L., H-I. Wu, Peter J. H. Sharp, Robert M. Schoolfield, Robert M.

Coulson. 1984. Modeling Insects Development Rates: A Literature Review and

Application of Biophysical Model. Ann. Entomol. Soc. Am. 77: 208-225.

Ward, M.P. 2005. Epidemic West Nile virus encephalomyelitis: a temperature- dependent, spatial model of disease dynamics. Prev Vet Med. 71(3): 253–264. Wild, M., A. Ohmura, and K. Makowski 2007. Impact of global dimming and brightening on global warming. Geophys. Res. Lett. 34: L04702 doi:10.1029/2006GL028031. Wilde, S. A.. J. W. Valley, W. H. Peck, and C. M. Graham. 2001. Evidence from

detrital zircons for the existence of continental crust and oceans on the earth

4.4Gyr ago. Nature 409: 175-178.

Wing, S. L., G. J. Harington, Francesca A. Smith, Jonathan I. Bloch, Douglas M.

Boyer, Katherine H. Freeman. 2005. Transient Floral Change and Rapid Global

Warming at the Paleocene-Eocene Boundary. Science 310(5750): 993-996.

Yamamura K, Kiritani K. 1998. A simple method to estimate the potential increase in

the number of generations under global warming in temperate zones. Applied

Entomology and Zoology 33: 289-298. 31 Zwiers, F and G. Hegerl. 2008. Climate Change: attributing cause and effect. Nature

453 (7193): 296-297.

Zou, L., S. N. Miller and E. T. Schmidtmann. 2007. A GIS tool to estimate West Nile

virus risk based on a degree-day model. Environ. Monit. Assess 129: 413-420.

32

Chapter 2

Projecting Insect Voltinism under High and Low Greenhouse Gas Emission Conditions

ABSTRACT

I develop individual-based Monte Carlo methods to explore how climate change can alter insect voltinism under varying greenhouse gas emissions scenarios using input distributions of diapause termination or spring emergence, development rate, and diapause initiation, linked to daily temperature and photoperiod. I show concurrence of these projections with a field dataset, and then explore changes in grape berry moth,

Paralobesia viteana (Clemens), voltinism that may occur with climate projections developed from the average of three climate models using two different future emissions scenarios from the International Panel of Climate Change (IPCC). Based on historical climate data from 1960-2008, and projected downscaled climate data until 2099 under both high (A1fi) and low (B1) greenhouse gas emission scenarios, I used concepts of P. viteana biology to estimate distributions of individuals entering successive generations per year. Under the low emissions scenario, I observed an earlier emergence from diapause and a shift in mean voltinism from 2.8 to 3.1 generations per year, with a fraction of the population achieving a 4th generation. Under the high emissions scenario, up to 3.6 mean generations per year were projected by the end of this century, with a very small fraction of the population achieving a 5th generation. Changes in voltinism in this and other species in response to climate change will likely cause significant economic 33 and ecological impacts, and the methods presented here can be readily adapted to other species for which the input distributions are reasonably approximated.

Introduction

There is much interest in understanding the ecological effects of increasing concentrations of atmospheric greenhouse gases forcing climate change (Bale et al. 2002,

Karl and Trenberth 2003, Thomas and Trenberth 2003, Meehl and Tebaldi 2004, Mills

2005, Zwiers and Hegerl 2008, Curreano et al. 2008). The Intergovernmental Panel on

Climate Change (IPCC 2007a) projects increases in global mean surface temperatures of

1.1-6.4 ºC by the end of the 21st century if greenhouse gas emissions continue to increase at current rates. Such changes in climate will likely impact many ecosystem functions, especially many natural processes that are temperature dependent (Wing et al. 2005,

Deutsch et al. 2008, Inouye 2008).

Insects are appropriate model candidates on which to study the effects of climate change. First, insects are poikilotherms and hence their internal temperature is highly dependent on ambient temperature. Consequently, insect development is driven primarily by temperature (Stinner et al. 1975, Logan et al. 1976, Pruess 1983, Tauber et al. 1986,

Lowry and Lowry 1989, Wagner et al. 1984). Second, many insects have relatively short life spans, which is conducive to the development of laboratory and field-based research designed to measure the impact of climate change. In multivoltine taxa, the number of generations per year under current and projected climatic regimes can be evaluated to quantify the influence of changing climates (Yamamura and Kiritani 1998, van Asch et 34 al. 2007, Post et al. 2008). However, many other factors, both biotic and abiotic, could

also influence insect seasonality. For example, many insects diapause and often the main

induction factor is photoperiod (Tauber et al. 1986, Denlinger 2002). Furthermore, the

nonlinear interaction of temperature and photoperiod makes it challenging to investigate

and understand changes in insect voltinism in response to climate change (Tobin et al.

2008). Although universal predictions of how climate change would influence insect

phenology is not feasible, doing so for representative insect species could be valuable in

efforts to understand the consequences of climate change in a general context.

Understanding key drivers of insect voltinism, such as development and diapause,

are essential in efforts to quantify the consequences of climate change on insect

seasonality (Logan et al. 2003). Such drivers are known for the grape berry moth,

Paralobesia viteana (Clemens) (Lepidoptera: Tortricidae) (Nagarkatti et al. 2001, Tobin

et al. 2001, 2002, 2003). P. viteana is native to North America and feeds on native wild and cultivated (not necessarily native) Vitis host spp., and can be an economically important pest in the latter (Riedl and Taschenberg 1984). It is also non-migratory, and movement patterns are spatially limited; thus, life stages are not subjected to different temperature profiles, and projections of P. viteana at a given location are reasonable predictors of the effects of climate change. Adults emerge in spring from diapausing pupae, mate, and females oviposit on Vitis spp. flowers or fruit. Upon hatching, larvae burrow into the fruit, exiting to pupate in leaves or bark. Pupae emerge as adults that initiate subsequent generations. As day length decreases following the summer solstice, there is a critical photoperiod independent of temperature that initiates diapause, and eggs deposited at these decreasing photoperiods will develop into diapausing pupae 35 (Nagarkatti et al. 2001). Hence, as in many temperate insect species, voltinism is

influenced by the degree-day accumulation prior to the arrival of photoperiodic

conditions that initiate diapause. There are generally 2-3 generations per year along the

grape production belt along Lake Erie (Tobin et al. 2003). Tobin et al. (2008) previously

highlighted the consequence of interactions between temperature (subject to climate

change) and photoperiod (not subject to climate change) in driving P. viteana voltinism.

In this paper, I extend this work by examining the interaction of temperature and

photoperiod based upon scenarios of the rate of greenhouse gas emissions and their effect

on temperature (Hayhoe et al. 2007). I also present an individual-based Monte Carlo

approach to quantify changes in voltinism and highlight not only generalized trends but

the variability among individuals within a population, and provide open-source computer

code to enable these methods to be extended to other species.

Materials and Methods

I developed our model based upon prior studies of P. viteana biology conducted at the Lake Erie Regional Grape Laboratory in North East, PA (42.2 ºN, 79.9 ºW) that describe development and diapause as functions of temperature and photoperiod

(Nagarkatti et al. 2001, Tobin et al. 2001, 2002). Photoperiods for North East were obtained from the Naval Oceanography Portal (2008). Projected monthly temperature data for this location, estimated to the end of this century under both high (A1fi) and low

(B1) greenhouse gas emissions scenarios of the IPCC Special Report on Emissions

Scenario (Nakicenovic et al. 2000), are available from the Northeast Climate Impacts 36 Assessment (2006). The A1fi scenario simulates climate under current economic development conditions without reducing greenhouse gas emissions, and the B1 scenario assumes significant global reductions in greenhouse gas emissions; these scenarios have been used previously to contrast climate-driven processes under different greenhouse-gas emissions scenarios (Hayhoe et al. 2007, Kunkel et al. 2008).

I obtained higher resolution downscaled daily weather data in the location of interest from 1960 to 2099 (M. Fitzerpatrick, personal communication). A more detailed description of climate data downscaling techniques and applications is provided by

Hewitson and Crane (1996). The data are the average of three general circulation models

– the Geophysical Fluid Dynamics Laboratory Model (GFDL), the Hadley Centre for

Climate Prediction and Research Model (HadCM3) and the National Center for

Atmospheric Research Parallel Climate Model (PCM) – each with a different sensitivity to changes in greenhouse gas amounts. Using these models with the A1fi and B1 emission scenarios, daily mean temperatures were derived.

Our population model is individual based and divided into three sub-models: diapause termination (Tobin et al. 2002), degree-day development (Tobin et al. 2001) and diapause induction (Nagarkatti et al. 2001). I focused our attention on modeling the progression of individuals among life stages independent of population density. These simulation processes are presented in figure 1-1 as a conceptual model.

Diapause termination: Diapause termination in P. viteana is primarily driven by temperature, and initial and 50% adult emergence generally occurs at 148 and 210 degree-days accumulated after January 1, respectively (base threshold = 8.4 °C, Tobin et al. 2002). Because P. viteana overwinters as a pupa, the total degree-day requirement for 37 adult emergence from diapausing pupae is different from that of the subsequent

generations, which develop from egg to adult. Also, since the distribution of degree-day

accumulation required for adults to emerge from diapausing pupae is often highly

skewed, I used a negative binomial distribution with k=2, which I estimated from field

observations of emerging adults reported in Tobin et al. (2001), to simulate a distribution

of the required degree days at which emergence would occur. Because the minimum

degree-day requirement is 148 instead of 0, which is typical of a binomial distribution, I

shifted the distribution to the right, accordingly. The calendar day of emergence of each

adult from a population of 10,000 diapausing pupae was then recorded. I defined this

population of emerging adults as the first generation.

Development: After adult emergence, another 75 degree-days were added for female sexual maturation (Luciani 1987, Tobin et al. 2003). If I assume that females oviposit all eggs in a single day and only one offspring survives to the adult stage (e.g., the population size doesn’t increase), I can treat the second generation offspring as another generation of its parent. For the second (and subsequent) generation(s), an average of 424 degree-days is required for development from egg to adult (Tobin et al.

2001). This developmental rate, however, also follows a negative binomial distribution, with a mean of 424 and k=2, as determined from initial investigation of the field data

(Tobin et al. 2001). I shifted this negative binomial distribution to the right by 250 because the minimum degree-day requirement for egg-to-adult development is 250

(Tobin et al. 2001). Each first generation adult was advanced independently with respect to the distribution of egg-to-adult development. This yielded the required calendar days for the development of the population of second generation adults. 38 Diapause induction: The egg is the sensitive stage for diapause induction, and

depending on the photoperiod at which an egg is laid, it will either eventually develop

into an adult or into a diapausing pupa (Nagarkatti et al. 2001). In field studies, eggs laid

before June 25 at the latitude of North East, PA never entered diapause, and after August

11 all eggs entered diapause (Nagarkatti et al 2001). The relationship between entering

diapause, Pr, is a function of the change in photoperiod (in hours) between a given date and its precedent date following the summer solstice, Pc, when the egg is laid,

Pr=100(1-e-3.957Pc) . (1)

For any given individual, I used a uniform distribution to run a Monte Carlo simulation to

determine whether the individual of the second generation or later enters diapause. For

instance, if the probability (Pr) of entering diapause is 0.7, I chose a random number

generated from a uniform distribution [0,1]. If that random number was smaller than 0.7,

then the individual egg developed into a diapausing pupa; otherwise, that egg matured

into an adult. In the latter case, an additional 424 degree-days (determined from the

development-rate distribution described above) would be accumulated by that individual

to reach the adult stage, and another 75 degree-days would be accumulated to allow for

oviposition by that adult. This iterative process was continued until all individuals were

either in diapause or dead (i.e., degree-day accumulation was insufficient to complete the

life cycle to an ovipositing adult).

Field validation: Independent of these modeling efforts and the published

literature, I conducted a field study to determine the distribution of adult P. viteana

emergence dates at North East, PA, in 2007. Using a combination of malaise traps and

light traps, I monitored female P.viteana at 4 locations in the area of North East, PA. 39 Unlike sex pheromone trap-catch data of male moths, which is useful to determine first generation emergence (Tobin et al. 2003) but not subsequent generations (M.C.S., unpublished data), female trap catch data provided us with discrete generational peaks that could be compared to the results from our simulations.

Simulation models and analyses: I simulated a population of 10,000 individuals, initialized as overwintering diapausing pupae, which progressed through diapause termination, development, and diapause initiation (Fig. 1). I first tested whether the negative binomial distribution was appropriate to simulate the degree day requirement by conducting a two-sided Kolmogorov-Smirnov test. To model the influence of climate on the voltinism of this population, I determined two outcomes for each individual (1) the number of generations each individual and its progeny could complete until it either entered diapause or died, and (2) the calendar date of adult emergence for each generation it experienced. Each individual could only complete one generation and have exactly one offspring, regardless of regulation factors such as predation and disease.

Hence, each offspring represents a parent individual in the future generation, and I tracked the outcome for that individual. I quantified the number of generations from each iteration, and then used 10,000 iterations to calculate the distribution of the number of generations per year. In each generation, I also determined the distribution of emergence dates and the number of completed generations for the field-collected data from 2007, and two projected emission scenarios extending to the end of the century. I also compared the predicted and observed distribution of emergence dates using a two-sided

Kolmogorov- Smirnov test for that year. The simulation code, presented in the Appendix, was written in MATLAB 2007 (The MathWorks Inc., Natick, MA). I summarized these 40 outcomes as time-series plots, graphing the mean number of completed generations, and the mean calendar day of adult emergence extending to the end of this century, under the low and high emission scenarios developed from the average of GFDL, HadCM3, and

PCM climate models.

I used linear regression, using the mean number of generations as a dependent variable and year as the independent variable, to test whether there was a significant increasing trend of mean generations per year under the historical and low emission condition projections. For the projected high emission condition, after initial scrutiny, I discovered a significant nonlinear trend; thus, I used nonlinear regression by fitting a

Gompertz function because a Gompertz curve has asymptotic properties that are more realistic than other unrestricted nonlinear functions (Laird 1964).

To explore the seasonal dynamics of different generations under our climate change scenarios, I used the default kernel estimation function in R (R Core Development

Team 2010) to estimate the kernel density of each generation in 2007 (from the field data), and in the low and high conditions in 2099. These estimated kernel functions express the proportion of the population that is present for every date that each generation completes its life history. The area under the curve for each generation sums to one.

Under both the low and high emission scenarios, I investigated the shift in emergence date for all four generations throughout the simulation period, from 2009 to

2099. For each generation in both scenarios, I used linear regression of emergence date by year to test if there was a significant trend. Furthermore, I used Analysis of

Covariance to test for differences in the emergence date over year (as a covariate), using 41 the different emissions conditions (low and high emission) as a categorical main effect.

All statistical analyses were conducted in R 2.10 (R Core Development Team 2010).

Results

The projected probability density function (PDF), projected cumulative density function (CDF), and observed CDF (e.g., Tobin et al. 2003) of adult emergence from overwintering, diapausing pupae are shown in figure 1-2 for 2001. The simulated distribution had predicted mean, minimum, and maximum degree-days of 210, 148, and

530, respectively, which coincides well with the observed data (mean, minimum, and maximum degree-days of 210, 148, and 512, respectively; Tobin et al. 2003). I also observed congruence between model predictions and empirical observations in the emergence profile by conducting a two sided Kolmogorov-Smirnov test of the first generation (figure 2-2). The results show no significant difference between predicted and observed distributions (D=0.05, p=0.22), suggesting that the negative binomial distribution was appropriate in our efforts to simulate degree-day requirements for diapause termination and emergence of the first generation, which were then progressed through the development rate and diapauses initiation subroutines

Using this modeling framework (figure 2-1), the estimated number of mean generations increased during the recent historical past (from 1960 to 2008), and is projected to continue to increase throughout the century in both greenhouse gas emission scenarios (figure 2-3). From 1960 to 2008, there was a significant increase in the number of generations (F=5.32; df=1, 47; P<0.05). In both the low (F=29.49; df=1, 89: P<0.001) 42 and high (F=32.57; df=1, 89; P<0.001) emission conditions, the number of generations was projected to increase (figure 2-3). To iterate, the projected number of generations in a specific year is the mean based upon 10,000 individual simulations.

The seasonal dynamics of different generations under different climate change scenarios can be illustrated by the distribution of adult emergence for each generation.

The comparison between the simulated density function of emergence date for each generation in 2007 and observations from the field in 2007 is presented in figure 2-4 (top panel). Note that the amplitude of the first generation does not imply that its population size is larger than other generations; rather, the curve is a density function, capturing the total area under the curve such that the area sums to 1 for each generation. The emergence timespan of the first generation is more narrowly distributed than later generations, which leads to a higher amplitude. The observation curve is scaled so that the maximum number of observations in each generation is the same as the maximum value of the projected density function in that generation. Also, note that these curves reflect the distribution, but not the number, of individuals contributing to each generation.

For example, in 2007, there were 13, 25, 38 and 11 individuals measured in the field for generations 1, 2, 3, and 4, respectively. The higher amplitude and shorter range associated with the first generation reflects the more discrete timing over which the first generation presents itself, in contrast to the wider ranges that result in overlapping generations in subsequent generations. I also ran a two sided Kolmogorov-Smirnov test and found no significant differences between predicted and observed distributions of emergence dates in all four generations (D = 0.38, 0.87, 0.49, and 0.21, and p=0.35, 0.08,

0.19, and 0.56, for generations 1 through 4, respectively). From these comparisons, I 43 verified that our modeling predictions almost match the observed field data; generally

with a deviation of ± 3 d (figure 2-4, top panel). The exception was that our projected

emergence date for the second generation was approximately 5 d later than the observed

date.

The estimated density functions of emergence date for each generation in 2099

under both low and high emission scenarios are also presented in figure 2-4 (middle and

bottom panels). By the end of the century and under the low emission scenario, the initial

emergence dates of the first two generations have advanced ≈ 15-20 d relative to 2007.

For example, in the 1st and 2nd generation, the timing of their respective initial emergence in 2007 is approximately the time at which 50% emergence is predicted to occur by 2099, while the 3rd and 4th generations are predicted to be advanced ≈ 20-30 d relative to the

2007 data under the low emission scenario.

Under the high emission scenario, this tendency for advanced emergence is much

more pronounced. All four generations advanced almost 30 d compared to 2007, and

under this scenario, adult emergence from overwintering pupae is projected to be almost

completed before observed emergence from 2007 would have begun. Moreover, about

30% of the 2nd generation would be completed before a 2nd generation in 2007 would

have begun. This advance of a full generation follows throughout the time series, so by

the time a rare 4th generation could have conceivably occurred in 2007, the model not

only predicted four full generations in 2099, but also sufficient time for a partial 5th

generation. I suggest this is due to a combination of earlier first generation emergence

and faster developmental times in subsequent generations under climate warming.

The projected mean emergence dates from 2009 to 2099 under both emission 44 scenarios are presented in figures 2-5 and 2-6. Under both the low and high emissions

scenario, a very significant (P < 0.001) and negative trend toward earlier emergence dates

was observed in all four generations as time progressed through the 21st century (Table 2-

1). The slope estimates express the rate of decrease in number of days needed to complete each generation, and range from −0.120 to −0.278 under low emission conditions, and −0.316 to −0.548 in the high emission conditions. In both emissions scenarios, there is a trend towards a faster decline (i.e., steeper negative slopes) as generations increase. Also, there is a consistent trend towards a more predictive relationship (a higher R2) as generations increase within each emission condition (Table

1). Relationships of emergence date with year were consistently more predictive from

simulations conducted under the high-emission condition (R2 of 0.74 to 0.91) than the low-emission condition (R2 of 0.31 to 0.58) (Table 2-1). When comparing the effect of

emissions conditions given the time-dependent decrease for each generation using an

ANCOVA, both the categorical variable (emission conditions) and covariate (year) had a

highly significant influence (P<0.001) on emergence dates (Table 2-2).

As noted by the density kernels (Fig. 4), there is an opportunity for a 5th generation to

develop under the high emission scenario by the end of the century. In our

summarization of mean emergence dates through time, I saw a very small proportion

(≈50 out of 10,000 individuals) of the population entering a 5th generation under the high

emission conditions, beginning in 2050 (figure 2-6). 45 Discussion

There is a need to estimate the influence of climate change on insect populations to provide guidance in policy decisions, and to enable adaptations in agricultural practices and public health efforts. Climate models under varying emissions scenarios are becoming increasingly available and statistically downscaled, both temporally and spatially, enabling abiotic drivers (temperature and photoperiod) of insect phenology models to connect to projections of climate change. Among the variables that define climate (i.e., temperature, precipitation, and wind), temperature is often the more tractable to predict as advances in statistical downscaling enable more reasonable predictions of local daily temperatures (Marshall et al. 2007), while photoperiod remains effectively constant. Thus, abiotic drivers of insect voltinism can be estimated. The degree-day requirements for spring emergence and in-season development have been estimated for many insect species (e.g., Taylor 1981, Nietschke et al. 2007); however, methods to connect these degree-day based insect phenology models with models of climate and climate projections are still needed. Here, I provide an individual-based approach to explore how climate change can alter insect voltinism under varying emissions scenarios using input distributions of diapause termination or spring emergence, development rate, and diapause initiation, linked to daily temperature and photoperiod. These individual-based methods enable projection of both means and distributions of emergence dates. I showed concurrence of these projections with a field dataset, and explored projections in our model system. 46 Our model system assumes that development of individuals is driven solely by air temperature, and diapause solely by photoperiod using a population initialized at 10,000 individuals, which can only be reduced due to individuals entering diapause or due to mortality when individuals fail to complete development to a diapause-capable stage. I assumed no immigration, emigration, or variation among life stages in how they are exposed to air temperature. These simplifying assumptions enabled us to develop a modeling framework for examining the influence of future climates on voltinism at a single spatial location. Future work should consider climatic influences on populations over wider spatial scales, incorporating variation in the drivers of insect seasonality such as the effects of host quality (Hunter and McNeil 1997), and geographic clines in diapause initiation probabilities (Ruberson et al. 2001).

New York, Michigan, and Pennsylvania are respectively the second, fourth, and fifth largest grape producing states in the US, and the vast majority of the grapes in these states are grown along the shores of Lake Michigan, Lake Erie, and in the Finger Lakes region of New York. All of these grape growing regions are close in latitude to North

East, PA, from where the field study was based. In recent years, late season infestations of P.viteana have surprised many growers in this area, resulting in an increased amount of fruit being rejected for consumption due to insect damage (M.C.S., unpublished data).

These late season infestations have caused considerable consternation to both growers and the industry. The timing of insecticide sprays against P. viteana populations have remained unchanged since 1991 with the introduction of the Grape Berry Moth Risk

Assessment Program (Martinson et al. 1991). For a high risk vineyard, this protocol calls for an insecticide application at 10 d post bloom (targeting 1st generation P. viteana), 47 early August (targeting the 2nd generation), and if necessary, late August (targeting the 3rd

generation). Although the recommended first spray is tied to bloom time, which in turn,

is driven by temperature, the timing of subsequent treatments are currently based solely

on calendar date irrespective of temperature. During a growing season with average

temperatures, these management guidelines could work fairly well, but are inadequate

when temperatures do not follow an “average” year.

The reality of climate change will result in a need to adjust overall management

guidelines for many insect species, including P. viteana, as well as dynamic strategies

within the growing season to account for developmental and climatic variability. Recent

failures in P. viteana management programs suggest that this shift in voltinism could

already be happening. The economic damages associated with future additional

generations of insect pests of agriculture, forestry, animal health, and other sectors are

likely to be severe (Walther et al. 2002, Kiritani 2006, Reisen et al. 2006, Kilpatrick et al.

2008, Lafferty 2009). I note that in P. viteana, the inherent lag associated with climate

suggests that I will see increasing numbers of individuals reaching a third generation in

North East, PA, regardless of policy or other factors that might influence emissions, until

approximately mid-century (Fig. 3). It is not unreasonable to assume similar shifts in

volintism in other systems. After 2050, however, there is likely to be a dramatic influence

of emission scenarios on P. viteana voltinism, particularly under the high emission scenario. It is important to note that I am defining the “high” emission scenario as a continuation of the status quo in emission rates, which is fossil-fuel intensive and is the highest future emission trajectory considered by the IPCC (Nakicenovic et al. 2000). 48 However, recent observations show global emissions have been higher than this “high” emission scenario since 2004 (Canadell et al. 2007, Raupach et al. 2007).

In addition to the mean temperature, daily temperature variability could also significantly influence insect life history. Past work has demonstrated that changes in the range of daily temperatures can affect the longevity, mortality, and other life history variables in mosquitoes, which could in turn affect mosquito-borne disease dynamics

(Paaijmans 2009, 2010). Since increasing climate variability is an important aspect of global climate change (IPCC 2007b), future studies should also focus on the influence of climate variability on insect voltinism in this and other systems.

Increasing levels of greenhouse gases will likely result in increases in annual global average temperature. This in turn will affect many aspects of ecosystem function, and will alter the dynamics and distribution of individual species and the communities in which they interact (IPCC 2007a, IPCC 2007b). Here, I focused our attention to two emission scenarios: A1fi and B1. Although insect voltinism is strongly determined by temperature and thus would also be affected by increasing temperatures, other factors such as photoperiod, which is not influenced by global climate change, also play an important role in insect seasonality. I show here how the nonlinear interaction between temperature and photoperiod is critical in understanding the link between climate change and voltinism. I also provide projected patterns of voltinism for a representative multivoltine insect species under these two climate change scenarios. Moreover, I also propose a conceptual framework, through the incorporation of diapause termination, degree-day development, and diapause induction for addressing climate change and 49 voltinism in other insect systems. The conceptual framework, and the source code provided in Appendix 1, could be readily modified for use in other systems.

Acknowledgements

I thank Dr. Melanie Fitzpatrick of the Union for Concerned Scientists and Dr. K.

Hayhoe from Texas Tech University who provided us with climate data from 1960 to

2099. Funding was provided through a University Graduate Student Fellowship from

Pennsylvania State University. 50

References Cited

Bale, J.S., G. J. Masters, I. D. Hodkinson, C. Awmack, T. M. Bezemer, V. K. Brown,

J. Butterfield, A. Buse, J. C. Coulson, J. Farrar, J. E. G. Good, R.

Harrington, S. Hartley, T. H. Jones, R. L. Lindroth, M. C. Press , I.

Symrnioudis, A. D. Watt and J. B. Whittaker. 2002. Herbivory in global

climate change research: direct effects of rising temperature on insect herbivores.

Glob. Ch. Biol. 8: 1-16.

Canadell, J. G., C. Le Quéré, M. R. Raupach, C. B. Field, E. T. Buitenhuis, P. Ciais,

T. J. Conway, N. P. Gillett, R. A. Houghton and G. Marland. 2007.

Contributions to accelerating atmospheric CO2 growth from economic activity,

carbon intensity, and efficiency of natural sinks. Proc. Natl. Acad. Sci. USA

104(47): 18866-18870.

Curreano, E. D., P. Wilf, S. L. Wing, et al. 2008. Sharply increased insect herbivory

during the Paleocene-Eocene Thermal Maximum. Proc. Natl. Acad. Sci. USA 105

(6): 1960-1964.

Denlinger, D.L. 2002. Regulation of diapause. Ann. Rev. Entomol. 47: 93-122.

Deutsch, C.A., J. J. Tewksbury, R. B. Huey, et al. 2008. Impacts of climate warming

on terrestrial ectotherms across latitude. Proc. Natl. Acad. Sci. USA 105 (18):

6668-6672.

Hayhoe, K., C.P. Wake, T.G. Huntington, L. Luo, M. Schwartz, J. Sheffield, E.

Wood, B. Anderson, J. Bradbury, A. DeGaetano, T. Troy, and D. Wolfe. 51 2007. Past and future changes in climate and hydrological indicators in the U.S.

Northeast. Climate Dynamics 28:381-407.

Hewitson, B. C. and R.G. Crane. 1996. Climate downscaling: technique and application.

Climate Research 7: 97-110.

Hunter, M. D. and J. N. McNeil. 1997. Host-plant quality influences diapause and

voltinism in a polyphagous insect herbivore. Ecology 78: 977-986.

Inouye DW. 2008. Effects of climate change on phenology, frost damage and floral

abundance of montane wildflowers. Ecology 89: 353-362.

IPCC. 2007a. Summary for policymakers: 4th report. http://www.ipcc.ch/ipccreports/ar4-

wg1.htm.

IPCC. 2007b. Climate Change 2007: The Physical Science Basis. Contribution of

Working Group I to the Fourth Assessment Report of the Intergovernmental Panel

on Climate Change (AR4). Solomon, S., D. Qin, M. Manning, Z. Chen, M.

Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.), Cambridge University

Press, Cambridge, UK and NY, USA.

Karl, T. R. and K. E. Trenberth. 2003. Modern Global Climate Change. Science

302:1719-1723.

Kilpatrick, A. M., M. A. Meola, R. M. Moudy and L. D. Kramer. 2008. Temperature,

viral genetics, and the transmission of West Nile virus by Culex pipiens

mosquitoes. PLoS Pathogen 4: e1000092.

Kiritani, K. 2006. Predicting impacts of global warming on population dynamics and

distribution of arthropods in Japan. Popul. Ecol. 48: 5-12. 52 Kunkel, K.E., H.-C. Huang, X.-Z. Liang, J.-T. Lin, D. Wuebbles, Z. Tao, A.

Williams, M. C., J. Zhu, and K. Hayhoe. 2008. Sensitivity of future ozone

concentrations in the northeast USA to regional climate change. Mitigation and

Adaptation Strategies for Global Change. 13:5-6.

Lafferty, K. D. 2009. The ecology of climate change and infectious diseases. Ecology 90:

888-900.

Laird, A. K. 1964. Dynamics of tumor growth. British Journal of Caner 18: 490-502.

Logan, J. A., D. J. Wollkind, S. C. Hoyt, and L. K. Tanigoshi. 1976. An Analytic

Model for Description of Temperature Dependent Rate Phenomena in

Arthropods. Environ. Entomol. 5: 1133-1140.

Logan, J. A., J. Regniere, and J. A. Powerll. 2003. Assessing the impacts of global

warming on forest pest dynamics. Front Ecol. Environ. 1: 130-137.

Lowry, W. P. and P. P. Lowry II. 1989. Fundamentals of Biometeorology (volume 1),

Peavine Publications, McMinnville.

Luciani, M. A. 1987. The biology of the grape berry moth, Endopiza viteana (Clemens)

(Lepidoptera: Tortricidae) in southern Ontario. M.S. Thesis, University of

Guelph, Guelph.

Marshall, S. J. , M. J. Sharp, D. O. Burges and F. W. Anslow. 2007. Near-surface-

temperature lapse rates on the Prince of Wales Icefield, Ellesmere Island, Canada:

implications for regional downscaling of temperature. Int. J. Climatol. 27:385-398.

Martinson, T. E., C. J. Hoffman, T. J. Dennehy, S. J. Kamas and T. Weigle. 1991.

Risk Assessment of Grape Berry Moth and Guidelines for Management of the 53 Eastern Grape Leafhopper. New York's Food and Life Sciences Bulletin Number

138.

Meehl, G. A. and C. Tebaldi. 2004. More intense, more frequent and longer lasting heat

waves in the 21st century. Science 305: 994-997.

Mills, E. 2005. Insurance in a Climate of Change. Science 309: 1040-1044.

Nagarkatti, S., P. C. Tobin and M. C. Saunders. 2001. Diapause induction in the

Grape Berry Moth (Lepidoptera: Tortricidae). Environ. Entomol. 30: 540-544.

Nakicenovic, N., J. Alcamo, G. Davis, B. de Vries, J. Fenhann, S. Gaffin, K. Gregory,

A. Grübler, T.Y. Jung, T. Kram, E.L. La Rovere, L. Michaelis, S. Mori, T.

Morita, W. Pepper, H. Pitcher, L. Price, K. Riahi, A. Roehrl, H. Rogner, A.

Sankovski, M. Schlesinger, P. Shukla, S. Smith, R. Swart, S. van Rooijen, N.

Victor, and Z. Dadi. 2000. Special report on emissions scenarios: An

Intergovernmental Panel on Climate Change (IPCC) report. Cambridge, UK:

Cambridge University Press.

Naval Oceanography Portal. 2008. http://www.usno.navy.mil/.

Nietschke, B., R. D. Magarey, D. M. Borchert, D. D. Calvin, and E. Jones. 2007. A

developmental database to support insect phenology models. Crop Protection 26:

1444-1448.

Northeast Climate Impacts Assessment (NECIA). 2006. Climate Change in the US

Northeast, Cambridge, MA: Union of Concerned Scientists.

http://www.northeastclimateimpacts.org/.

Paaijmans, K. P., A. F. Read, and M. B. Thomas. 2009.Understanding the link

between malaria risk and climate. PNAS 106: 13844-13849. 54 Paaijmans, K. P., S. Blanford, A. S. Bell, J. I. Blanford, A. F. Read, and M. B.

Thomas. 2010.Influence of climate on malaria transmission depends on daily

temperature variation. PNAS 107: 15135-15139.

Post, E. S., C. Pedersen, C. C. Wilmers, and Mads C. Forchhammer. 2008.

Phenological sequences reveal aggregate life history response to climatic

warming. Ecology 89 (2): 363-370.

Pruess, K. P. 1983. Day-Degree Methods for Pest Management. Environ. Entomol. 12:

613-619.

Reisen, W. K., Y. Fang, and V. M. Martinez. 2006. Effects of temperature on the transmission of West Nile virus by Culex tarsalis (Diptera: Culicidae). J. Med. Entomol.

43: 309-317.

Riedl, H. and E.F. Taschenberg. 1984. Grape Berry Moth. Grape IPM Insect

Identification Sheet No.1, New York State Agricultural Experiment Station,

Geneva.

Ruberson, J. R., K. V. Yeargan, and B. L. Newton. 2001. Variation in diapause

responses between geographic populations of the predator Geocoris punctipes

(Heteroptera: Geocoridae). Ann. Entomol. Soc. Am. 94: 116-122.

Raupach, M. R., G. Marland, P. Ciais, C. Le Quere, J. G. Canadell, G. Klepper and

C. B. Field. 2007. Global and regional drivers of accelerating CO2 emissions.

Proc. Natl. Acad. Sci. USA 104(24): 10288-10293.

Stinner, R.E., G. D. Butler, J. S. Bacheler, and C. Tuttle. 1975. Simulation of

Temperature-dependent Development in Population Dynamics Model. Can. Ent.

107: 1167-1174. 55 Tauber, M. J., C. A. Tauaber, and S. Masaki. 1986. Seasonal Adaptations of Insects,

Oxford University Press, New York.

Taylor, F. 1981. Ecology and the evolution of physiological time in insects. The

American Naturalist 117: 1-23.

Thomas R. Karl and Kevin E. Trenberth. 2003. Modern Global Climate Change.

Science 302 (5651): 1719 – 1723.

Tobin, P. C., S. Nagarkatti, and M. C. Saunders. 2001. Modeling Development in

Grape Berry Moth (Lepidoptera: Tortricidae). Environ. Entomol. 30: 692-699.

Tobin, P. C., S. Nagarkatti, and M. C. Saunders. 2002. Diapause Maintenance and

Termination in Grape Berry Moth (Lepidoptera: Tortricidae). Environ. Entomol.

31 (4): 708-713.

Tobin, P. C., S. Nagarkatti, and M. C. Saunders. 2003. Phenology of Grape Beery

Moth (Lepidoptera: Tortricidae) in Cultivated Grape at Selected Geographic

Locations. Environ. Entomol. 32: 340-346.

Tobin, P. C., S. Nagarkatti, G. Loeb, and M. C. Saunders. 2008. Historical and

projected interactions between climate change and voltinism in a multivoltine

insect species. Global Change Biology. 14 (5): 951-957.

Van Asch M, P.H. van Tienderen, L. J. M. Holleman, M. E. Visser. 2007. Predicting

adaptation of phenology in response to climate change, an insect herbivore

example. Global Change Biology 13: 1596-1604.

Wagner, T. L., H-I. Wu, Peter J. H. Sharp, Robert M. Schoolfield, Robert M.

Coulson. 1984. Modeling Insects Development Rates: A Literature Review and

Application of Biophysical Model. Ann. Entomol. Soc. Am. 77: 208-225. 56 Walther, G-R, E. Post , P. Convey, A. Menzel, C. Parmesan, Trevor J. C. Beebee, J-

M. Fromentin, O. Hoegh-Guldberg, F. Bairlein. 2002. Ecological responses to

recent climate change. Nature 416: 389-395.

Wing, S. L., G. J. Harington, Francesca A. Smith, Jonathan I. Bloch, Douglas M.

Boyer, Katherine H. Freeman. 2005. Transient Floral Change and Rapid Global

Warming at the Paleocene-Eocene Boundary. Science 310(5750): 993-996.

Yamamura K, Kiritani K. 1998. A simple method to estimate the potential increase in

the number of generations under global warming in temperate zones. Applied

Entomology and Zoology 33: 289-298.

Zwiers, F and G. Hegerl. 2008. Climate Change: attributing cause and effect. Nature

453 (7193): 296-297. 57

Table 2-1. Regression of mean emergence dates (y) on years (x) under low (B1) and high (A1fi) emission conditions

Emission Generation a b R2 F p Condition

Low 1 -0.12 390 0.31 41 <0.001 2 -0.137 468 0.5 91 <0.001 3 -0.166 560 0.6 131 <0.001 4 -0.278 817 0.58 123 <0.001 High 1 -0.316 786 0.74 247 <0.001 2 -0.361 920 0.86 557 <0.001 3 -0.426 560 0.91 907 <0.001 4 -0.548 817 0.91 851 <0.001

* the regression has the form y (emergence date) = ax (year) + b 58 Table 2-2. Influence of low or high emission condition (modeled as a categorical variable) and year (modeled as a covariate) on mean emergence dates

Generation R2 F p(emmission p condition) (year)

1 0.667 119 <0.001 <0.001 2 0.819 269 <0.001 <0.001 3 0.877 421 <0.001 <0.001 4 0.841 312 <0.001 <0.001

59

Fig. 2-1. Conceptual model for a single simulation run to collect number of generations per year, initialized by a diapausing overwintering life stage. 60

Fig. 2-2. Degree Day requirement of first generation emergence of adult grape berry moth from overwintering diapausing pupae. PDF is the density function, and CDF the cumulative density function. For both functions, modeled values are from Tobin et al. 2002, and observation are from field data collected in 2002. 61

Fig. 2-3. Historical and projected number of mean generations of grape berry moth in North East, PA, under low and high emission conditions (B1 and A1fi emission conditions, respectively) described by the Intergovernmental Panel on Climate Change (IPCC 2007b). 62

Fig. 2-4. Density functions expressing the emergence date for all generations of grape berry moth in 2007 (top panel, for observed data collected from malaise traps, and for simulated data), and for 2099 under low (middle panel) and high (lower panel) emission conditions (B1 and A1fi emission conditions, respectively) described by the Intergovernmental Panel on Climate Change (IPCC 2007b). 63

Fig. 2-5. Projected mean emergence date for all generations of grape berry moth under low (B1) emission conditions described by the Intergovernmental Panel on Climate Change (IPCC 2007b). 64

Fig. 2-6. Projected mean emergence date for all generations of grape berry moth under high (A1fi) emission conditions described by the Intergovernmental Panel on Climate Change (IPCC 2007b).

65 Chapter 3

Investigating the Relationship between Daily Temperature Range and Insect

Phenology under Climate Change

ABSTRACT

Increased climate variability, such as changes in the daily temperature range (DTR) is

one projected outcome of global climate change. Although insects experience fluctuating

temperatures in the field, there is little in the literature addressing the importance of daily

temperature range (DTR) and its influence on insect phenology and population dynamics.

I demonstrate how both symmetrical and asymmetrical changes in DTR alter daily

temperature profiles and hence degree day accumulation by the grape berry moth

(Paralobesia viteana). I then use an individual-based model to simulate the seasonal life

history of P. viteana (number of generations per year, average emergence date of each generation) using a comprehensive set of DTRs (81 different conditions) in North East,

Pennsylvania. When the mean temperature remains the same, larger DTR conditions result in more generations per year and earlier emergence time for each generation. The results also show that larger DTR variations could compensate for the effect of decreased mean temperature. These simulated results using P. viteana demonstrate explicitly and quantitatively how shifts in DTR could alter insect life histories beyond those expected from changes in mean temperature alone. 66

Introduction

Many environmental factors, especially ambient temperature, influence insect life histories because insects are poikilotherms (Lowry and Lowry 1989), whose development is driven primarily by temperature (Logan et al. 1976, Wagner et al. 1984). Many species have relatively short life spans and multiple generations per year, which is conducive to the development of laboratory and field-based research designed to measure the impact of climate change. While most of the research on biological effects due to climate change focuses on the influence of increasing or decreasing mean daily temperature, the daily temperature range (DTR) is also expected to change (Letcher 2009). However, unlike the strong consensus that mean temperature is rising, scientists are still debating whether

DTR will be larger or smaller, as well as highly variable among sites. Furthermore, I am still not sure to what extent the DTR would change, and whether changes will be symmetric about the mean (Bristow and Campbell 1984, Stanhill and Moreshet 1992,

Stanhill and Cohen 1997, Roderick and Farquhar 2002, Travis et al. 2002, Wild et al.

2007).

Insects in the field experience fluctuating temperatures rather than constant temperatures, and the daily temperature range can influence insect phenology. Pest managers, who often utilize phenology models to optimize timing of management tactics to specific life stages, typically recognize that a small amount of development accrues even when the daily mean is below threshold and adjust recommendations accordingly.

Researchers have also demonstrated how varying the DTR could alter mosquito life history and disease transmission patterns (Paaijmans et al. 2009, 2010, Lambrechts et al. 67 2011). Because I do not know how DTR will change in the future, it is critical to develop a method that allows us to consider all possible changes in the DTR, which, in turn, will also influence the mean temperature. This is important if I would like to explore how changing DTR might influence pest phenology systematically and quantitatively, and provide useful tools to assess potential pest risk in the future.

In this chapter, I seek to quantify how various DTR conditions influence phenology of herbivorous multivoltine species with relatively short life spans, using the grape berry moth (Paralobesia viteana) as a model species. P. viteana is a native North

American pest which primarily consumes wild and cultivated (not necessarily native)

Vitis host species. It is one of the most economically important pests in eastern US vineyards (Riedl and Taschenberg 1984). It is also non-migratory, and movement patterns are spatially limited; thus, life stages are not subjected to different temperature profiles, and projections of P. viteana at a given location are reasonable predictors of the effects of climate change. The life history of P. viteana has been well studied. The phenology of P. viteana is determined primarily by the interaction of ambient temperature, which is subject to climate change, and photoperiod, which is considered invariable (Nagarkatti et al. 2001, Tobin et al. 2002, Tobin et al. 2003, Tobin et al. 2008).

These life history traits and previous knowledge enabled the projections of how changing greenhouse gas conditions would influence voltinism, and time and distribution of occurrence of specific life stages, of P. viteana (Chen et al. 2011). These projections were developed out to the end of this century, using an individual-based modeling framework, and downscaled mean temperatures at a single location. In this study, I 68 explore the effects of both symmetric and asymmetric changes in the DTR on the daily

temperature profile, and on degree day accumulation by P. viteana, at a single site (North

East, PA) and a single year (2009). I focus on the impact of changing DTR conditions on

P. viteana phenology by studying the change in number of generations per year, and average emergence date of all generations.

Materials and Methods

1 Effect of various DTR on degree day accumulation

I use the modified sine curve method (also called sine negative exponential curve method, Parton and Logan 1981) to construct daily temperature profiles based on four inputs: daily maximum temperature, daily minimum temperature, number of the day in a year, and latitude. The latter two variables are used to generate day length, which is also essential to reconstruct the daily temperature curve and determine diapause initialization.

A detailed description of the procedure for modeling the daily temperature profile is provided in Appendix 1.

I modify the daily maximum and minimum temperature in order to simulate the

TT effect of DTR change. I use min max to approximate daily mean temperature, which is 2

used to adjust daily maximum and minimum temperature. I define two parameters, k1 and

k2, for the adjustment:

TT T'() T  k min max min min 1 2

TTmin max Tmax'() T max  k2 2 (Equation 1) 69

The adjusted daily minimum temperature (Tmin’) is the original daily minimum

temperature (Tmin) minus a fraction (k1) of daily mean temperature of that day, and

adjusted daily maximum (Tmax’) is the original maximum (Tmax) plus another fraction

(k2) of the daily mean. Note that this modeling framework allows either symmetric or

asymmetric changes in the pattern of change in the DTR.

Whenever k1 = k2, the magnitude of change is the same for both the minimum and maximum temperature, and when k1 does not equal k2, and there is an asymmetrical

change in the DTR. This enables simulation for the full range of possibilities that may

occur under climate change. For each simulation, after this DTR adjustment is made, the

computation of degree day accumulation is based the on sine curve methods as described

above and in Appendix 1.

Currently the annual surface mean temperature in Pennsylvania is approximately

10 ºC and, according to the 4th annual report of the Intergovernmental Panel on Climate

Change (IPCC 2007b), the annual surface mean temperature at the end of this century

will be about 1.1-4.0 ºC higher. It is important to note that changing the DTR can

influence the mean: in other words, the change of mean temperature can be derived from

the change in DTR. To demonstrate, I assume daily mean temperature is the arithmetic

mean of daily maximum and minimum temperature. According to equation 1, I could

approximate the new daily mean temperature based on the adjusted daily maximum and

minimum temperature as follows: 70

TTTTmin max min  max 24 Tmin k1()()  T max  k 2 1 TTmin'' max 2 2 TTmean'' h   24h1 2 2

(Equation 2)

where Th’ is the hourly temperature, and Tmean’ is the new daily mean temperature. Note that this also shows how symmetry, or the lack of symmetry, in DTR changes influence the mean. When the DTR change is symmetric (e.g., k1 = k2), there is no change in the

mean, but only a change in the DTR, for that day. However, when is there is an

asymmetrical change in the DTR (k1 does not equal k2), the mean for that day is shifted in

the direction of the larger absolute k value.

In order to limit the change in mean temperature to that projected by the IPCC

2007b report, I set the maximum of both adjustment parameters, k1 and k2, to 0.4; and I

also set the minimum as -0.4 so that it is symmetrical, and both parameters have a 0.1

step size. This will make the largest mean temperature change approximately 4 ºC. Note

that this is only for annual mean temperature, in each day the maximum/minimum/mean

temperature varies. These parameter settings and step sizes results in a total of 81

different climate change conditions (in fact, 80 novel conditions because k1 = k2 =0

represents current the climate condition, with no change in either mean temperature or

DTR).

Mathematically, these 81 combinations can be divided into six major categories (I

to VI, see table 1): same mean temperature with variable DTR (k1 =k2); increased mean

temperature with increased DTR (k2>0 and abs(k1)

with increased DTR (k1>0 and abs(k1)>abs(k2)); increased mean temperature with 71

decreased DTR (k1<0 and abs(k1)>abs(k2)); decreased mean temperature with decreased

DTR (k2<0 and abs(k1)

k2). Although not all categories are equally likely, given the uncertainty in how DTR will change, I simulate all possible combinations. To illustrate variation in degree day accumulation (degree days above the developmental threshold of 8.4 ºC) under these six categories, I examine days 100, 150, 200, and 250 (approximately April 10, May 30, July

19, and September 7). These dates allow us to step through the calendar year, and they are important to the life history of P. viteana because they bracket the time frame of

emergence from diapause, seasonal development, and re-entering diapause at the end of

the season (Tobin et al. 2008, Chen et al. 2011).

2 Effect of various DTR on P. viteana phenology

Under each of these 81 conditions, I use the temperature data for 2009 (obtained

from NOAA) from the town of North East, PA (42°12’N, 79°50’W) and modified in

equation 1 to simulate P. viteana population dynamics, using an individual-based

phenology model (Chen et al. 2011, described in detail in chapter II) and determine the

mean emergence dates for each generation, and the mean number of generations, for this

one calendar year. All the simulation scripts are written in R 2.10.1.

Results

1 Effect of various DTR on degree day accumulation

I present current daily temperature profiles for day 100, 150, 200, and 250 in

figure 3-1, when the mean temperature (dashed line) is below the developmental 72 threshold (solid line), the actual temperature profile has a small portion crossing that boundary. Hence using the sine curve model I still accumulate heat units even when the mean temperature is below the developmental threshold. In table 3-2 and figure 3-2 I present the simulated degree day accumulations under the 81 DTR conditions. In general, the degree day accumulation increases as k1 decreases and k2 increases, and this pattern is very consistent across the season in all four dates. While the effect of k1 and k2 on degree day accumulation is subtle at the beginning of the season (a range of <1 degree-day on day 100), there are substantial differences later (day 150, 200, and 250), where some conditions with either larger or smaller DTR conditions have more than 20 degree days accumulated in a day, while some decreased temperature conditions only accumulate 5 degree days. As noted above, these conditions bracket those in which there is no change in either the seasonal mean or the DTR (i.e., where k1 = k2 = 0), and are based on observed daily temperatures for a recent year (2009).

2 Effect of various DTR on P. viteana phenology

General Results

I use population based life history simulations (Chen et al. 2011) to demonstrate how various DTR conditions influence mean number of generations per year, as shown in figure 3-3 and table 3-3. As described in Chen et al. 2011, under current climate conditions (represented here as 2009 data) there are on average approximately 3 generations per year, and increased mean temperatures, regardless of fluctuations in DTR

(category II and category IV), will yield more generations per year comparing to the 73 current 3 generations. Some simulated populations even reach a 5th generation, if either -

k1 and k2 is large (or both).

Decreased mean temperatures with an increased DTR, in general, (category III,

k1>k2) result in fewer generations per year, but not as few as in the decreased mean

temperature with decreased DTR (category V, k1

category III (k1>>k2 I observe almost as many generations per year as in the current

condition, which indicates increasing the DTR could compensate for the effect of

decreased mean temperature. The projected change in mean number of generations

(relative to current climate condition, k1 = k2 = 0) is shown in figure 3-4.

Spring emergence dates (Figures 3-5 through 3-9, Table 2-3) follow a similar pattern. As expected, the mean emergence dates advance when mean temperatures increase

(categories II and IV) regardless of changes in the DTR. Also, later generations consistently develop faster than the first generation, resulting in shorter generation durations. Decreased mean temperature conditions (categories III and V), on the other hand, have an initially similar first generation emergence date but much later emergence dates for successive generations. However, in category III I still observe that under some large DTR conditions the generation duration and emergence dates are similar very similar to current condition. These are additional indicators that DTR influences

P.viteana phenology beyond that easily predicted from mean temperature alone.

Category I: Same Mean Temperature with Different and Symmetric DTR Conditions

Under current climate conditions (k1 = k2 = 0) I observe approximately 3

generations (on average) of P. viteana per year. If the mean temperature remains the 74

same but DTR increases (k1 = k2 > 0) I see an increase in mean generations per year. The

projected number of generations (y) is a simple linear function of the absolute magnitude

of DTR (figure 3-10, y=3+0.3*x, F1,5 = 457, p<0.001). However the mean time of

emergence of earlier generations is not advanced as much as later generations (Table 3-3).

st For instance, under a large DTR variation condition (k1 = k2 =0.4), the mean time of 1

generation emergence advances about 5 days while the 3rd and 4th generation emergence

date is about 7-8 days earlier, relative to current climatic conditions (k1 = k2 =0.4) . These results confirms that even when the mean temperature is unchanged, increasing DTR would have a positive impact on number of generations per year (and decreasing DTR results in decreasing number of mean generations). Although this change in voltinism is relatively small (~0.25 generations per year when k2=0.4), it might still represent a large

number of potentially damaging insects late in the growing season.

These results reveal why using mean temperature alone as a model input could

lead to inaccurate projections of population dynamics. I could expect higher or lower

numbers of generations per year if DTR is larger or smaller, even when the mean

temperature remains the same.

Category II: Increased Mean Temperature and Increased DTR

In this category, where the daily maximum is increasing faster than the minimum

is decreasing, or while the minimum is decreasing, the mean is also increasing. Thus, I

consistently see an increase in number of generations and advanced time of each

generation. In order to quantitatively examine how DTR variability, which is driven

heavily by increasing maximum temperatures in this category, would impact P. viteana 75

phenology I need to fix either k1 or k2 and make the other one changeable, so that these

conditions are comparable. Here I chose to fix k2 at 0.4 to investigate how a daily

minimum temperature (e.g, changing k1), affects P. viteana phenology when the daily

maximum is increasing the most within our tested range.. Again, I observe a strong linear

relationship, in this case between the magnitude of change in the DTR (reflected by the

varying k1) and generations per year (figure 3-11, F1,5 =5685, p<0.001). Although the

DTR is always larger than the current condition, and increasing in all of these conditions when k1 varies from -0.3 to 0.3 with k2 fixed at 0.4, a smaller magnitude of DTR change

results in more generations per year. Although counter-intuitive, this can be seen by first

realizing that as k1 decreases, and become more negative, daily minimum temperature increases, and thus the absolute magnitude in DTR decreases. Figure 3-11 shows that as k1 decreases, and become more negative, there is an increase in the number of generations per year.

Again I confirm that change in DTR would also change emergence dates.

Regressions of emergence dates against DTR (k1) are consistent and linear (figure 3-12,

F=1581, p<0.001; F1,5 =1062, p<0.001; F1,5 =1141, p<0.001; F1,5 =1286, p<0.001; F1,5

=96, p<0.001, for 1st through 5th generations, respectively). Also, the mean duration of

each generation is compressed as DTR decreases (Table 3-4). I believe such a large

change in P. viteana phenology would undoubtedly impose serious problems to agricultural and ecological applications.

Category III: Decreased Mean Temperature and Increased DTR 76 Although maximum temperatures are increasing in this category, the minimum is

decreasing faster, shifting the annual mean downward. Thus, simple projections based on

means might show fewer generations per year. However, I show that large DTR

variations can compensate for the effect of decreased annual mean temperatures. For

instance, if the mean temperature is not decreasing significantly (k1=0.3, k2=0.2; k1=0.4,

k2=0.3), the number of mean generations is 2.95 and 3.00, respectively. This is very close to current observations (3.00) and the mean emergence time of each generation is very similar as well, with differences of less than 3 days.

For illustration, I examine k1=0.4 because it has the most k2 possibilities (k2=-

0.3:0.3). I run a simple linear regression again to demonstrate how larger DTR (larger k2 values) could result in increased number of generations, even when mean temperature is decreasing (figure 3-13, F1,5 =118.6, p<0.001). I do not show further results in this

category because I believe it is the most unlikely condition under future greenhouse gas

emission scenarios.

Category IV: Increased Mean Temperature and Decreased DTR

There is no doubt that the global mean temperature is increasing. But some

researchers suggest that the DTR is decreasing because the increasing rate of minimum

temperature is greater than that of maximum temperature (Sanhill and Cohen 1997).

Though the DTR is decreasing relative to current climate condition, I still observe more

generations and earlier emergence times when the reduction in DTR is relatively small.

To better illustrate these conditions, I fix k1 at -0.4 (reflecting the strongest increase in

minimum temperatures in the range I simulated) and see how P. viteana phenology 77

changes according to the change of k2. Qualitatively speaking, I are having higher daily

maximums when k2 is large, which means DTR is large. As in the other categories, I see a

strong linear relationship between the number of generations per year and larger DTR

(larger k2, because k1 is fixed, in figure 3-14, F1,5 =3038, p<0.001).

For the emergence dates of all generations, a larger DTR would result in an earlier emergence date. I run linear regression of emergence dates against DTR (k2) and show

consistent significant relationship (F1,5 =6327, p<0.001; F1,5 =298, p<0.001; F1,5 =440,

p<0.001; F1,5 =233, p<0.001; F1,5 =27, p<0.005, for 1st to 5th generation, respectively.

See figure 3-15). In addition to earlier emergence dates, the mean duration of each generation is shortened as DTR enlarges (Table 3-5).

Category V: Decreased Mean Temperature and Decreased DTR

According to the IPCC report (IPCC 2007), there is little evidence that I will experience global cooling with less DTR variation in the future, so I only present results in Table 3-3 and do not discuss in details.

Category VI: Same DTR and Increased/Decreased Mean Temperature

It is relatively easy and straightforward to interpret the results for this condition. Since the DTR remains the same, degree day accumulation completely relies on the minimum temperature (the maximum temperature could be calculated without any modification because DTR does not shift in this condition, I could imagine it’s just a translation of current daily temperature profile to a lower/higher level). 78 I would expect the smallest number of generations (2.0 generations/year) when the minimum temperature is lowest (k1 = 0.4, k2 = -0.4), and largest number of generations

(4.4 generations/year) when minimum temperature is highest (k1=-0.4, k2=0.4) (Table 3-

3).

Discussion

These results demonstrate that the advanced onset differs for each generation.

Generally speaking, under current climatic conditions for this species at this location, for the first generation the advanced time is small. This is due to the cold weather at the beginning of the season: by day 100 the temperature is still low, and the increased maximum temperature or minimum temperature rarely surpasses the development threshold. Hence, the accumulated heat unit is zero at the beginning of the season regardless of what the DTR situation is. When it gets warmer (more specifically, the daily maximum temperature exceeds the basal development threshold, while the mean temperature is still below that threshold), increased DTR begins to result in a larger number of degree day accumulation. Nevertheless the amount of heat units accumulated is relatively small, so even under a large DTR variation condition, there is little effect on the emergence time of first generation that much.

Similarly I could understand that later in the season, especially during the hot midsummer, when the lowest temperature consistently exceeds the threshold, larger DTR variation could give much more heat unit accumulation than smaller DTR variations.

Both relative and absolute value of excessive degree day accumulation finally result in much earlier emergence dates in the later generations (3, 4, and 5). Furthermore, the time 79 to complete a full generation is shorter later in the season, because the degree day requirement is a constant. Those two factors (advanced emergence time and shortened life span) together account for the increased number of generations per year in larger

DTR variation conditions. I have predicted that under some very large DTR variation conditions there could be one more complete generation comparing to our current estimation, which has severe implications for management of this pest..

I have demonstrated how different DTR conditions influence pest phenology in the Lake Erie area. According to our research in the disease dynamics system (chapter V in the thesis), larger DTR has more impact in the relatively cooler areas than warmer areas. I have shown that the transmission period of the vector mosquitoes under large

DTR variation has extended to almost twice as long as in current condition in cool areas, whereas in relatively warmer areas (2C warmer) it is only about 20% longer in transmission period. Thus, increased DTR imposes a heavier and more significant burden for cooler areas. Unlike the mosquito system, the P. viteana system is more complicated because temperature is not the only factor that determines its life history. In the previous chapter (chapter II) I have described how the nonlinear interaction between temperature and photoperiod influence P. viteana phenology. It would be an interesting exercise to compare the phenological change of P. viteana in different locations. However, currently

I have no idea how DTR variation would influence P. viteana phenology other than the

Lake Erie area because of data availability and model complexity. I would like to expand this research in the future.

Although I use a population model to simulate P. viteana phenology under climate change conditions, some life history parameters, for instance, basal development 80 threshold, are set to be the same for the entire population. This is a simplified assumption while in reality different individuals should have different developmental thresholds.

Generally speaking, the individuals having relatively lower basal development thresholds should develop more rapidly than other members in that population, and emerge earlier in the field. However it’s not safe to draw the conclusion that they undoubtedly gain superiority over the other moths. For instance, their life history might desynchronize with their host plants. I am not sure if this will actually happen, because this will depend on the phenological change of their hosts, which is rarely observed and documented.

Nevertheless I could still use simulation techniques to estimate the impact of changing climate on a heterogeneous population.

Besides the relatively complicated interaction between host plant phenology and

P. viteana phenology, I am also interested in investigating how the natural enemies of P. viteana respond to daily temperature range change. It is reported that predators might respond more promptly than the herbivores. If this is true then climate change, especially global warming, might not cause as much damage to agroecosystems as I expected. I could apply differential equation systems (predator-prey interaction, in this case) and assign temperature-dependent coefficients to the system to see how increased DTR influence population dynamics quantitatively. All these possible research paths could be carried out in future studies.

Nevertheless I admit our modeling approach is not perfect. First, I am using one- year climate data (2009 daily data) and projecting into the future. The modified data in the remaining 80 combinations are more or less determined by this specific dataset.

However I do not know exactly what climate condition would be in the future. One 81 possible improvement could be using multiple-year-averaged climate data as the input.

But this approach bears the criticism that I have no field data to validate it. Since our

main objective of this research is to demonstrate the importance of DTR, I do not think

this is a serious issue.

Another concern in our modeling approach lies in the assumption that the DTR

rate of change is the same across the year. Again I do not know explicitly how this rate

changes in different part of the year and this assumption simplifies our modeling

procedure. But in reality it might change from day to day, or season to season. Currently I

believe this assumption is generally reasonable and, based on our research objectives, it is

adequate. I have already demonstrated the importance of finer scale climate change by

our examination of the DTR. Hence I am still relying on finer scale prediction of future

climate and weather conditions and looking for collaborations with meteorologists to deal

with this situation.

In this study I have shown that even under decreased mean temperature conditions,

if DTR is large enough, the number of generations per year and mean emergence dates

are not affected very much. On the other hand (and much more likely to happen in the

future), increasing DTR even with the same mean temperature would result in

approximately 0.3 generations more than current estimation, and under increased mean

temperature situations it is much more severe. I could observe almost 1 more generation

and approximately 1 month advancement in the 2nd and successive generations.

In summary, research on DTR variation and its impact on insect phenology and potential impacts on agroecosystem management is still a young and emerging field which requires further and careful research. It is our hope that this examination of DTR on the 82 population dynamics of P. viteana will serve as a useful template for further explorations into this potentially important arena of study. 83

Reference Cited Bonan, G. 2008. Ecological climatology: concepts and applications. Cambridge Press.

Bristow, R. L. and G. S. Campbell. 1984. On the relationship between incoming solar

radiation and daily maximum and minimum temperature. Agriculture and

Forestry Meteorology 31:159--166.

Chen, S., S, J. Fleischer, P. C. Tobin and M. C. Saunders. 2011. Projecting insect

voltinism under high and low greenhouse gas emission conditions.

Environmental Entomology (in press).

Forsythe, W. C., E. J. Rykiel, R. S. Stahl, H. Wu, and R. M. Schoolfield. 1995. A

model comparison for daylength as a function of latitude and day of a year.

Ecological Modelling 80: 87-95.

Lambrechts, L., K. P. Paaijmans, T. Fansiri, L. B. Carrington, L. D. Kramer, M. B.

Thomas, and T. W. Scott. 2011. Impact of daily temperature fluctuations on

dengue virus transmission by Aedes aegypti. PNAS (in print).

Letcher, T. M. 2009. Ecological Climate change: observed impacts on planet earth.

Elsevier.

Logan, J. A., D. J. Wollkind, S. C. Hoyt, and L. K. Tanigoshi. 1976. An Analytic

Model for Description of Temperature Dependent Rate Phenomena in

Arthropods. Environ. Entomol. 5: 1133-1140.

Logan, J. A., J. Regniere, and J. A. Powerll. 2003. Assessing the impacts of global

warming on forest pest dynamics. Front Ecol. Environ. 1: 130-137.

Lowry, W. P. and P. P. Lowry II. 1989. Fundamentals of Biometeorology (volume 1),

Peavine Publications, McMinnville. 84 IPCC. 2007a. Summary for policymakers: 4th report. http://www.ipcc.ch/ipccreports/ar4-

wg1.htm.

IPCC. 2007b. Climate Change 2007: The Physical Science Basis. Contribution of

Working Group I to the Fourth Assessment Report of the Intergovernmental Panel

on Climate Change (AR4). Solomon, S., D. Qin, M. Manning, Z. Chen, M.

Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.), Cambridge University

Press, Cambridge, UK and NY, USA.

Nagarkatti, S., P. C. Tobin and M. C. Saunders. 2001. Diapause induction in the

Grape Berry Moth (Lepidoptera: Tortricidae). Environ. Entomol. 30: 540-544.

Paaijmans, K. P., A. F. Read, and M. B. Thomas. 2009.Understanding the link

between malaria risk and climate. PNAS 106: 13844-13849.

Paaijmans, K. P., S. Blanford, A. S. Bell, J. I. Blanford, A. F. Read, and M. B.

Thomas. 2010. Influence of climate on malaria transmission depends on daily

temperature variation. PNAS 107: 15135-15139.

Parton, W. J. and J. A. Logan. 1981. A model for diurnal variation in soil and air

temperature. Agricultural Meteorology 23: 205-216.

Riedl, H. and E.F. Taschenberg. 1984. Grape Berry Moth. Grape IPM Insect

Identification Sheet No.1, New York State Agricultural Experiment Station,

Geneva.

Roberick, M. L. and G. D. Faquhar. 2002. The cause of decreased pan evaporation

over the past 50 years. Science 298: 1410–1411.

Stanhill, G. and S. Moreshet 1992. Global radiation climate change in Israel. Climate

Change 22:121-138. 85 Stanhill, G. and S. Cohen. 1997. Recent changes in solar irradiance in Antarctica. J.

Climat. 10: 2078-2086.

Stinner, R.E., G. D. Butler, J. S. Bacheler, and C. Tuttle. 1975. Simulation of

Temperature-dependent Development in Population Dynamics Model. Can. Ent.

107: 1167-1174.

Tobin, P. C., S. Nagarkatti, and M. C. Saunders. 2001. Modeling Development in

Grape Berry Moth (Lepidoptera: Tortricidae). Environ. Entomol. 30: 692-699.

Tobin, P. C., S. Nagarkatti, and M. C. Saunders. 2002. Diapause Maintenance and

Termination in Grape Berry Moth (Lepidoptera: Tortricidae). Environ. Entomol.

31 (4): 708-713.

Tobin, P. C., S. Nagarkatti, and M. C. Saunders. 2003. Phenology of Grape Beery

Moth (Lepidoptera: Tortricidae) in Cultivated Grape at Selected Geographic

Locations. Environ. Entomol. 32: 340-346.

Travis, D.J., A. M. Carleton, and R. Lauritsen. 2002. Contrails Reduce Daily

Temperature Range, Nature 418: 601.

Wagner, T. L., H-I. Wu, Peter J. H. Sharp, Robert M. Schoolfield, Robert M.

Coulson. 1984. Modeling Insects Development Rates: A Literature Review and

Application of Biophysical Model. Ann. Entomol. Soc. Am. 77: 208-225.

Wann, M., D. Yen and H. J. Gold. 1985. Evaluation and calibration of three models for

daily cycle of air temperature. Agricultural and Forest Meteorology 34: 121-128.

Wild, M., A. Ohmura, and K. Makowski 2007. Impact of global dimming and brightening on global warming. Geophys. Res. Lett. 34: L04702 doi:10.1029/2006GL02803 86

Table 3-1. Six Major Categories of Changing Daily Temperature Range

Category Description Mean DTR k1 k2 abs(k1)-abs(k2) Max increasing (decreasing) at same rate as min

I is decreasing (increasing) No Change Variable k1 = k2 Max increasing- faster than min is increasing, or II min is decreasing Increasing Increasing Positive Negative Min decreasing-faster than max is increasing, or III max is decreasing Decreasing Increasing Positive Positive Min increasing- faster than max is increasing, or IV max is decreasing Increasing Decreasing Negative Positive Max decreasing-faster than min is decreasing, or V min is increasing Decreasing Decreasing Negative Negative Max increasing (decreasing) at same rate as min

VI is increasing (decreasing) Variable No Change k1 = -k2 87

Table 3-2. Degree Day Accumulation for specific days under Different Mean Temperature and DTR Conditions Difference in Degree Day on each Calendar Day Day relative to k1=k2=0 k1 k2 100 150 200 250 100 150 200 250 -0.4 -0.4 0.03 12.05 12.31 13.83 -0.29 -0.57 -0.56 -0.14 -0.4 -0.3 0.09 13.15 13.43 14.96 -0.23 0.53 0.56 0.99 -0.4 -0.2 0.16 14.25 14.55 16.09 -0.16 1.63 1.67 2.13 -0.4 -0.1 0.25 15.35 15.66 17.22 -0.07 2.73 2.79 3.26 -0.4 0 0.34 16.45 16.78 18.35 0.02 3.83 3.91 4.39 -0.4 0.1 0.44 17.55 17.9 19.48 0.12 4.94 5.03 5.52 -0.4 0.2 0.55 18.66 19.02 20.61 0.23 6.04 6.15 6.65 -0.4 0.3 0.67 19.76 20.14 21.74 0.35 7.14 7.27 7.78 -0.4 0.4 0.79 20.86 21.26 22.87 0.46 8.24 8.39 8.91 -0.3 -0.4 0.03 11.09 11.33 12.73 -0.29 -1.53 -1.54 -1.23 -0.3 -0.3 0.09 12.19 12.45 13.86 -0.23 -0.43 -0.42 -0.1 -0.3 -0.2 0.15 13.29 13.57 14.99 -0.17 0.67 0.7 1.03 -0.3 -0.1 0.24 14.39 14.69 16.12 -0.08 1.77 1.82 2.16 -0.3 0 0.34 15.49 15.81 17.25 0.01 2.88 2.93 3.29 -0.3 0.1 0.43 16.6 16.92 18.38 0.11 3.98 4.05 4.42 -0.3 0.2 0.54 17.7 18.04 19.52 0.21 5.08 5.17 5.55 -0.3 0.3 0.65 18.8 19.16 20.65 0.33 6.18 6.29 6.68 -0.3 0.4 0.77 19.9 20.28 21.78 0.45 7.28 7.41 7.82 -0.2 -0.4 0.03 10.13 10.35 11.63 -0.3 -2.49 -2.52 -2.33 -0.2 -0.3 0.09 11.23 11.47 12.76 -0.24 -1.39 -1.4 -1.2 -0.2 -0.2 0.15 12.33 12.59 13.89 -0.17 -0.29 -0.28 -0.07 -0.2 -0.1 0.24 13.43 13.71 15.02 -0.08 0.82 0.84 1.06 -0.2 0 0.33 14.54 14.83 16.16 0.01 1.92 1.96 2.19 -0.2 0.1 0.42 15.64 15.95 17.29 0.1 3.02 3.07 3.32 -0.2 0.2 0.53 16.74 17.06 18.42 0.21 4.12 4.19 4.46 -0.2 0.3 0.64 17.84 18.18 19.55 0.32 5.22 5.31 5.59 -0.2 0.4 0.76 18.94 19.3 20.68 0.44 6.32 6.43 6.72 -0.1 -0.4 0.03 9.17 9.37 10.53 -0.3 -3.45 -3.5 -3.43 -0.1 -0.3 0.08 10.27 10.49 11.67 -0.24 -2.35 -2.38 -2.3 -0.1 -0.2 0.15 11.37 11.61 12.8 -0.17 -1.24 -1.26 -1.17 -0.1 -0.1 0.23 12.48 12.73 13.93 -0.09 -0.14 -0.14 -0.03 -0.1 0 0.33 13.58 13.85 15.06 0 0.96 0.98 1.1 -0.1 0.1 0.42 14.68 14.97 16.19 0.1 2.06 2.1 2.23 -0.1 0.2 0.52 15.78 16.09 17.32 0.2 3.16 3.22 3.36 -0.1 0.3 0.63 16.88 17.2 18.45 0.31 4.26 4.33 4.49 -0.1 0.4 0.75 17.98 18.32 19.58 0.43 5.36 5.45 5.62 0 -0.4 0.03 8.21 8.4 9.44 -0.3 -4.41 -4.47 -4.52 0 -0.3 0.08 9.31 9.52 10.57 -0.24 -3.3 -3.36 -3.39

88

0 -0.2 0.15 10.42 10.63 11.7 -0.17 -2.2 -2.24 -2.26 0 -0.1 0.23 11.52 11.75 12.83 -0.09 -1.1 -1.12 -1.13 0 0 0.32 12.62 12.87 13.96 0 0 0 0 0 0.1 0.41 13.72 13.99 15.09 0.09 1.1 1.12 1.13 0 0.2 0.51 14.82 15.11 16.22 0.19 2.2 2.24 2.26 0 0.3 0.62 15.92 16.23 17.36 0.3 3.3 3.36 3.39 0 0.4 0.74 17.02 17.35 18.49 0.42 4.41 4.47 4.52 0.1 -0.4 0.03 7.25 7.42 8.34 -0.3 -5.36 -5.45 -5.62 0.1 -0.3 0.08 8.36 8.54 9.47 -0.24 -4.26 -4.33 -4.49 0.1 -0.2 0.15 9.46 9.66 10.6 -0.18 -3.16 -3.22 -3.36 0.1 -0.1 0.22 10.56 10.77 11.73 -0.1 -2.06 -2.1 -2.23 0.1 0 0.32 11.66 11.89 12.87 0 -0.96 -0.98 -1.1 0.1 0.1 0.41 12.76 13.01 14 0.09 0.14 0.14 0.03 0.1 0.2 0.5 13.86 14.13 15.13 0.18 1.24 1.26 1.17 0.1 0.3 0.61 14.96 15.25 16.26 0.29 2.35 2.38 2.3 0.1 0.4 0.73 16.07 16.37 17.39 0.41 3.45 3.5 3.43 0.2 -0.4 0.03 6.3 6.44 7.24 -0.3 -6.32 -6.43 -6.72 0.2 -0.3 0.08 7.4 7.56 8.38 -0.24 -5.22 -5.31 -5.59 0.2 -0.2 0.14 8.5 8.68 9.51 -0.18 -4.12 -4.19 -4.46 0.2 -0.1 0.22 9.6 9.8 10.64 -0.1 -3.02 -3.07 -3.32 0.2 0 0.31 10.7 10.92 11.77 -0.01 -1.92 -1.96 -2.19 0.2 0.1 0.4 11.8 12.03 12.9 0.08 -0.82 -0.84 -1.06 0.2 0.2 0.5 12.9 13.15 14.03 0.18 0.29 0.28 0.07 0.2 0.3 0.6 14.01 14.27 15.16 0.28 1.39 1.4 1.2 0.2 0.4 0.72 15.11 15.39 16.29 0.39 2.49 2.52 2.33 0.3 -0.4 0.03 5.34 5.46 6.15 -0.3 -7.28 -7.41 -7.82 0.3 -0.3 0.08 6.44 6.58 7.28 -0.24 -6.18 -6.29 -6.68 0.3 -0.2 0.14 7.54 7.7 8.41 -0.18 -5.08 -5.17 -5.55 0.3 -0.1 0.21 8.64 8.82 9.54 -0.11 -3.98 -4.05 -4.42 0.3 0 0.31 9.74 9.94 10.67 -0.01 -2.88 -2.93 -3.29 0.3 0.1 0.4 10.84 11.06 11.8 0.08 -1.77 -1.82 -2.16 0.3 0.2 0.49 11.95 12.17 12.93 0.17 -0.67 -0.7 -1.03 0.3 0.3 0.59 13.05 13.29 14.07 0.27 0.43 0.42 0.1 0.3 0.4 0.7 14.15 14.41 15.2 0.38 1.53 1.54 1.23 0.4 -0.4 0.03 4.57 4.5 5.05 -0.3 -8.05 -8.37 -8.91 0.4 -0.3 0.08 5.64 5.62 6.18 -0.24 -6.98 -7.25 -7.78 0.4 -0.2 0.14 6.72 6.74 7.31 -0.18 -5.9 -6.13 -6.65 0.4 -0.1 0.21 7.8 7.86 8.44 -0.11 -4.82 -5.01 -5.52 0.4 0 0.3 8.89 8.98 9.58 -0.02 -3.73 -3.9 -4.39 0.4 0.1 0.39 9.98 10.09 10.71 0.07 -2.64 -2.78 -3.26 0.4 0.2 0.49 11.08 11.21 11.84 0.17 -1.54 -1.66 -2.13 0.4 0.3 0.58 12.17 12.33 12.97 0.26 -0.44 -0.54 -0.99 0.4 0.4 0.69 13.28 13.45 14.1 0.37 0.66 0.58 0.14

89

Table 3-3. P. viteana phenology under different DTR conditions Mean Emergence Date for Generation Mean number of k1 k2 generations One Two Three Four Five -0.4 -0.4 2.9 142 187 225 253 NA -0.4 -0.3 3.04 140 182 217 244 251 -0.4 -0.2 3.22 137 176 210 235 259 -0.4 -0.1 3.43 135 172 204 228 255 -0.4 0 3.65 133 168 198 223 245 -0.4 0.1 3.83 131 165 194 218 237 -0.4 0.2 4.01 129 162 190 214 232 -0.4 0.3 4.18 127 159 186 209 227 -0.4 0.4 4.38 125 157 182 205 222 -0.3 -0.4 2.75 145 193 232 278 NA -0.3 -0.3 2.92 142 187 224 253 NA -0.3 -0.2 3.06 139 181 216 244 262 -0.3 -0.1 3.25 137 176 209 234 261 -0.3 0 3.47 135 171 203 227 251 -0.3 0.1 3.67 133 168 198 222 243 -0.3 0.2 3.86 131 164 193 217 236 -0.3 0.3 4.04 129 162 189 213 231 -0.3 0.4 4.21 127 159 185 209 226 -0.2 -0.4 2.56 148 199 241 267 NA -0.2 -0.3 2.79 144 191 231 272 NA -0.2 -0.2 2.95 141 185 222 253 NA -0.2 -0.1 3.08 138 180 215 241 NA -0.2 0 3.28 136 175 208 233 261 -0.2 0.1 3.5 134 171 202 227 247 -0.2 0.2 3.7 132 167 197 222 241 -0.2 0.3 3.9 130 164 192 217 235 -0.2 0.4 4.07 128 161 188 212 230 -0.1 -0.4 2.31 152 206 255 NA NA -0.1 -0.3 2.6 147 198 240 277 NA -0.1 -0.2 2.82 143 190 229 265 NA -0.1 -0.1 2.97 140 184 221 249 NA -0.1 0 3.12 138 179 214 239 258 -0.1 0.1 3.32 135 174 207 231 264 -0.1 0.2 3.54 133 170 201 226 249 -0.1 0.3 3.74 131 166 196 221 241 -0.1 0.4 3.92 129 163 192 216 235 0 -0.4 2.11 155 216 280 NA NA 0 -0.3 2.36 150 205 252 309 NA 0 -0.2 2.65 146 196 238 297 NA 0 -0.1 2.86 142 189 228 269 NA

90

0 0 3 139 183 220 247 NA 0 0.1 3.15 137 178 212 238 261 0 0.2 3.36 135 173 206 231 257 0 0.3 3.59 133 169 200 225 247 0 0.4 3.78 131 165 195 220 240 0.1 -0.4 2.03 159 229 328 NA NA 0.1 -0.3 2.15 154 213 275 NA NA 0.1 -0.2 2.43 149 203 249 291 NA 0.1 -0.1 2.7 145 194 236 287 NA 0.1 0 2.89 141 187 226 260 NA 0.1 0.1 3.03 138 181 218 246 NA 0.1 0.2 3.19 136 176 211 236 267 0.1 0.3 3.41 134 172 205 230 255 0.1 0.4 3.62 132 168 199 224 246 0.2 -0.4 2 164 248 NA NA NA 0.2 -0.3 2.04 157 225 314 NA NA 0.2 -0.2 2.2 152 211 269 NA NA 0.2 -0.1 2.49 147 201 246 NA NA 0.2 0 2.75 143 193 234 281 NA 0.2 0.1 2.92 140 186 224 258 NA 0.2 0.2 3.06 137 180 217 245 256 0.2 0.3 3.24 135 175 209 235 269 0.2 0.4 3.46 133 171 204 228 252 0.3 -0.4 2 169 281 NA NA NA 0.3 -0.3 2.01 161 241 NA NA NA 0.3 -0.2 2.07 155 221 300 NA NA 0.3 -0.1 2.29 150 207 262 NA NA 0.3 0 2.57 145 198 243 NA NA 0.3 0.1 2.79 141 191 232 278 NA 0.3 0.2 2.95 138 184 223 258 NA 0.3 0.3 3.1 136 179 215 243 NA 0.3 0.4 3.29 134 174 208 234 267 0.4 -0.4 2 174 323 NA NA NA 0.4 -0.3 2 164 264 NA NA NA 0.4 -0.2 2.02 158 233 337 NA NA 0.4 -0.1 2.12 152 216 286 NA NA 0.4 0 2.38 147 204 255 NA NA 0.4 0.1 2.66 143 195 240 293 NA 0.4 0.2 2.86 140 188 229 269 NA 0.4 0.3 3 137 182 220 252 NA 0.4 0.4 3.15 134 177 213 240 258

91

Table 3-4. Mean Generation Duration (in Days) of 2nd-5th Generation in Category II (k2 = 0.4)

k1 Generation -0.3 -0.2 -0.1 0 0.1 0.2 0.3 2nd 32.36 32.99 33.85 34.94 36.34 38.09 39.98 3rd 26.25 27.44 28.63 29.72 31.19 32.69 34.18 4th 23.34 23.93 24.45 24.78 24.67 24.92 25.84 5th 17.58 17.56 18.32 20.28 22.28 23.99 33.33

92

Table 3-5. Mean Generation Duration (in days) of 2nd-5th Generation in Category IV (k1=-0.4)

k2 Generation -0.3 -0.2 -0.1 0 0.1 0.2 0.3 2nd Gen 42.28 39.26 37.07 34.84 33.75 32.91 32.18 3rd Gen 35.23 33.59 32.08 30.44 29.07 27.76 26.5 4th Gen 27.22 24.76 24.06 24.43 24.17 23.9 23.43 5th Gen 6.75 24.59 26.56 22.24 18.69 17.92 17.4

93

Fig 3-1. Daily Temperature Profiles From a Sine Negative Exponential Model in Different Days Bracketing the Seasonal Development of Paralobesia viteana. (Solid Curve: Interpolated Daily Temperature Fluctuation; Solid Line: Development Threshold; Dotted Line: Mean Temperature)

94

DD

k1 k2

Fig 3-2-1. Degree Day Accumulation in Day 100 under Various DTR Conditions (x, y axes represent k1 and k2, respectively; z axis represents DD accumulation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values. Notice when k2 is small the degree day accumulations are almost zero for any k1 values, such as the yellow, red, light green, and green balls. The degree day accumulation in day 100 is relatively small.)

95

DD

k2

k1

Fig 3-2-2. Degree Day Accumulation in Day 150 under Various DTR Conditions (x, y axes represent k1 and k2, respectively; z axis represents DD accumulation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values.)

96

DD

k2 k1

Fig 3-2-3. Degree Day Accumulation in Day 200 under Various DTR Conditions (x, y axes represent k1 and k2, respectively; z axis represents DD accumulation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values.)

97

DD

k2

k1

Fig 3-2-4. Degree Day Accumulation in Day 250 under Various DTR Conditions (Note difference in scale of day 100 relative to the other three days) (x, y axes represent k1 and k2, respectively; z axis represents DD accumulation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values.)

98

Number of Generations

k2

k1

Fig 3-3. Predicted Number of Generation per Year under Different DTR Conditions (x, y axes represent k1 and k2, respectively; z axis represents predicted number of generations per year. Balls with same color represent degree day accumulation of same k2 values but varying k1 values.)

99

Change of Generations

k2 k1

Fig 3-4. Predicted Change of Number of Generation per Year under Different DTR Conditions (x, y axes represent k1 and k2, respectively; z axis represents change of number of generations per year, that is, predicted number of generations per year minus number of generations per year at current status where k1=k2=0. Balls with same color represent degree day accumulation of same k2 values but varying k1 values.)

100

k2

k1

Fig 3-5. Predicted Mean Emergence Date of First Generation (x, y axes represent k1 and k2, respectively; z axis represents mean Julian date of 1st generation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values.)

101

Emergence Date of 2nd Generation

k2 k1

Fig 3-6. Predicted Mean Emergence Date of Second Generation (x, y axes represent k1 and k2, respectively; z axis represents mean Julian date of 2nd generation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values.)

102

Emergence Date of 3rd Generation

k2

k1

Fig 3-7. Predicted Mean Emergence Date of Third Generation (x, y axes represent k1 and k2, respectively; z axis represents mean Julian date of 3rd generation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values.)

103

Emergence Date of 4th Generation

k2 k1

Fig 3-8. Predicted Mean Emergence Date of Fourth Generation (x, y axes represent k1 and k2, respectively; z axis represents mean Julian date of 4th generation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values. Notice not all combinations could have a potential 5th generation, hence not all of them have emergence dates of 4th generation.)

104

k2

k1 Fig 3-9. Predicted Mean Emergence Date of Fifth Generation (x, y axes represent k1 and k2, respectively; z axis represents mean Julian date of 5th generation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values. Notice not all combinations could have a potential 5th generation, hence not all of them have emergence dates of 5th generation.)

105

Fig 3-10. Predicted Number of Mean Generations per Year in Category I: no change in mean and symmetric change in the DTR (x-axis represents DTR change, which is both k1 and k2 because in this category k1=k2. Notice the significant linear increasing trend of number of generations against DTR change, though the magnitude of change is relatively small.)

106

Fig 3-11. Predicted Number of Generation Change in Category II (k2=0.4) (x-axis represents k1 values when k2 is fixed at 0.4. Notice the significant linear decreasing trend of number of generations against DTR change, which means here smaller DTR correlate with more number of generations per year. However in this category DTR is still larger than current condition where k1=k2=0.)

107

Fig 3-12. Predicted Mean Emergence Day in Category II (k2=0.4) (x-axis represents k1 values when k2 is fixed at 0.4. Notice the significant linear increasing trend of emergence dates against DTR change, which means here smaller DTR (smaller k1 value) correlate with earlier emergence dates. However in this category DTR is still larger than current condition where k1=k2=0.)

108

Fig 3-13. Predicted Number of Generation Change in Category III (k1=0.4) (x-axis represents k2 values when k1 is fixed at 0.4. Notice the significant linear increasing trend of number of generations against DTR change, which means here larger DTR correlate with more number of generations per year. Also notice here when DTR is large (k2=0.3), the number of generations per year is almost the same as in the current condition (~3.0 gen/year), despite the fact that the mean temperature in this category is decreasing.)

109

Fig 3-14. Predicted Number of Generation Change in Category IV (k1=-0.4) (x-axis represents k2 values when k1 is fixed at -0.4. Notice the significant linear increasing trend of number of generations against DTR change, which means here larger DTR correlate with more number of generations per year. )

110

Fig 3-15. Predicted Mean Emergence Day in Category IV (k1=-0.4) (x-axis represents k2 values when k1 is fixed at -0.4. Notice the significant linear decreasing trend of emergence dates against DTR change, which means here larger DTR (larger k2 value) correlate with earlier emergence dates. )

111

Chapter 4

Capturing West Nile Virus Risk Using Degree Day Model in Pennsylvania

ABSTRCT

I provide calibrated degree day models to predict West Nile virus (WNV) transmission periods in

Pennsylvania. Previous degree day models treat the extrinsic incubation period (EIP) and mosquito longevity as constants, and do not capture virus transmission period well in four locations

(Harrisburg, Philadelphia, Pittsburgh, and Williamsport) in Pennsylvania from 2002 to 2008. I propose adjustments based on literature and model simulations to better understand how environmental factors (temperature and photoperiod) determine WNV emergence period. I suggest that the EIP is smaller at the beginning of the season, and that adult mosquito longevity be incorporated as a variable. I also consider species composition as an important factor for the WNV transmission period. The calibrated models increase the ability to successfully predict WNV emergence period in Pennsylvania to 70% compared to less than 30% in the uncalibrated model.

These models provide useful tools to assess potential WNV risk and could be easily adapted to other regions with different vectors when appropriate parameters are provided.

112

Introduction

A brief introduction of West Nile Virus

WNV belongs to the Flaviviridae family, defined as group 4 positive single stranded RNA

(+ss RNA) viruses. WNV primarily infects avians, especially crows (Corvidae). Other mammals,

including humans, horses and rodents, serve as potential hosts and reservoirs of WNV. The vectors

of WNV are mosquitoes, especially those in the Culex genus (Family Culicidae) such as Cx pipiens,

Cx restuans, and Cx tarsalis (Campbell et al. 2002).

Since the arrival of WNV in New York City in 1999 (Lanciotti et al. 1999) there have been

more than 30,600 people infected, resulting in approximately 1,200 fatalities to date (CDC 2010).

Within a few years the virus spread across the continental U.S. (Peterson 2001, Nash et al. 2001,

Sugumaran et al. 2009). Each year between 60 to 9,000 human cases of WNV have been reported

(CDC 2010), with mild symptoms ranging from coughs and headaches, to life threatening symptoms

and death.

WNV can also bring about significant ecological impacts by killing bird populations. A

severe outbreak killed nearly half of North American crows (Corvidae) in 2007 , and American

robin (Turdus migratorius) populations have also been significantly impacted (Komar et al. 2001,

LaDeau et al. 2007, Rahbek 2007). Thus, WNV is a severe threat to both public health and ecological health.

Importance of Temperature

Temperature plays an important role in both the life history traits (e.g. longevity, rate of development and fecundity) of mosquitoes as well as the dynamics of WNV (Rueda et al. 1990,

Dohm et al. 2002a, Bellan et al. 2010). For example, between 10º C and 30º C, transmission of

WNV by Cx. tarsalis occurred much faster at higher temperatures (Reisen et al. 2006). At around

113

10º C it takes more than 30 days on average for the mosquito to transmit the virus whereas near 30º

C it takes less than 1 week. Kilpatrick et al. (2008) showed a similar effect for other Culex species

(Cx. pipiens) and pointed out the relationship of temperature and virus maturity inside the mosquito body is quartic (power of 4, after square root arc-sine transformation) rather than linear; hence I might underestimate WNV transmission rates under high temperature conditions (Kilpatrick et al.

2008).

Degree Day Modeling of WNV

Degree day modeling techniques have been introduced by Allen (1976) and applied widely in ecological and epidemiological approaches. Degree day models have been used in several studies to estimate potential WNV emergence (Zou et al. 2007, Konrad et al. 2009). For example, Zou et al.

(2007) used this method to predict WNV occurrence in Wyoming between 2003 and 2005 in mid summer for Cx. tarsalis, where it was able to achieve 80% accuracy in predictions. However this method did not perform well in Santa Barbara, California unless the extrinsic incubation period (EIP) for Cx. tarsalis was adjusted to 76 degree days (DD, Konrad et al. 2009)) from theoriginal 109 DD used by Zou et al. 2007.

Research Objectives

Here, I investigate whether this type of degree day modeling method may be applicable to

Pennsylvania, where the WNV vectors differ from those in the western USA. In Pennsylvania, the major vectors are Cx. pipiens and Cx. restuans, but not Cx tarsalis. Thus, to assess whether this model can adequately capture WNV dynamics in Pennsylvania, I investigated whether the model was able to successfully capture both first detection of WNV emergence in a year, and last detection of WNV in a year. To answer these questions I used the original degree day model by Zou et

114 al.( 2007) and examined how well it predicts WNV occurrence, as described above, at four locations in Pennsylvania (Philadelphia, Harrisburg, Williamsport and Pittsburgh across seven years (2002 to

2008). Then I changed some parameters (e.g. EIP and longevity)based on optimization modeling to improve the model performance in Pennsylvania.

Materials and Methods

Climate, Geographic, and Epidemiological Background

Daily minimum and maximum temperature data for Pennsylvania were obtained from

NOAA (www.noaa.gov). For the purpose of this study, I selected four sites across Pennsylvania and used meteorological station data for Harrisburg, Pittsburgh, Philadelphia, and Williamsport, as seen in Figure 4-1. For the purpose of this study, I selected data points that were within 15km of the meteorological stations.

Mosquito surveillance data were provided by the Pennsylvania Department of Environment

Protection. These data had been collected between 2002 and 2008. Each data record was georeferenced and contains additional information such as date the mosquitoes were captured, trap type, habitat type, species and poolsize. Only adult mosquitoes were captured and each mosquito was identified to species level. They were pooled in the same species and the same trap type and each pooled sample was tested for WNV using RT-PCR. A total of 145,536 mosquitoes were tested as WNV positive and were collected in 789 pools during the 7 year study period. The pool size varied from 3 individuals to 100 individuals per date and trap type. Although dead birds can be used and have been extensively used in a variety of other studies (LaDeau et al. 2007, Konrad et al. 2009) it is not collected in the State of Pennsylvania therefore for this study only WNV-positive mosquitoes and the collection dates will be used.

115

Modeling WNV Transmission Using Original Degree Days Model Approach

To examine WNV emergence for each of the seven years (2002 to 2008) at four locations

(Harrisburg, Philadelphia, Pittsburgh, and Williamsport) I first applied the degree day model as proposed by Zou et al. (2007). This approach estimated potential WNV risk on each trapping day by computing degree day accumulation over a previous time frame defined by the vector mosquito longevity, and determining if that accumulation exceeded a threshold defined by the extrinsic incubation period of the virus. The extrinsic incubation period (EIP) is set as a threshold of potential

WNV emergence. Initially the EIP was set as 109 DD. Model simulations were run for the duration of the WNV field surveillance season starting early April (day 150 of year) until the end of October

(day 300). The model parameters were those used by Zou et al. (2007) and include vector longevity

(12 days as the average adult Culex longevity, Dohm et al. 2006), base development threshold of Cx tarsalis is 14.3C (Reisen et al. 2006), and the number of degree days it takes for the virus to become transmittable as 109 DD (Reisen et al. 2006). A detailed description of this modeling approach was provided in appendix B. In this study I considered a prediction that was within 10 days of the observed date as a successful prediction. Model accuracy was determined by comparing first and last prediction dates for WNV-positive vectors with those observed based on the rtPCR measures of the pooled samples.

To more accurately and precisely model degree days accumulation and WNV emergence I also considered daily temperature fluctuation since it plays an important role in vector insect phenology and disease dynamics (Paaijmans et al. 2009, 2010). Here I applied a more realistic degree days accumulation routine by taking daily maximum temperature, minimum temperature, and photoperiod into consideration (Zou et al. 2007). I incorporated the daily maximum and minimum temperatures to reconstruct the hourly temperature profile in a day, along with photoperiod which determined when the maximum and minimum temperatures occur (Forstythe et al. 1995, Martinez

116

1991, Parton and Logan 1981). It should be reasonable that if the maximum temperature in a day is

higher than the base development threshold (but with the mean temperature lower than the threshold),

the mosquito could still accumulate a proportion of degree days for that day. The following equation

showed how I reconstructed a daily temperature profile:

 m Ti( Tmax  T min )sin( )  T min ( t n  t  t x ) D 2 a bn  Z Ti Tmin ( T s  T min ) e ( t x  t  t n ' ) where D is day length (determined by latitude and day in a year), Z is night length, Ts is the

temperature at the moment of sunset, m is number of hours after minimum temperature until sunset, n is number of hours after sunset until the time of minimum temperature of the next day, a is the lag coefficient for the maximum temperature, and b is the nighttime temperature coefficient (Parton and

Logan 1981). A detailed description and derivation of the degree day accumulation model that I am using is presented in appendix A.

Because I did not exactly know the life history parameters (EIP, longevity, etc) for Cx. pipiens and Cx. restuans in Pennsylvania, I explored the optimal parameter values (which minimize the total error of model prediction) within a large parameter space to determine whether 109DD and

12 days longevity were appropriate for Pennsylvania. For the EIP, I set the lowest value as 89 degree days (similar to Konrad et al. 2008) and the upper bound as 129 degree days (same interval from the median 109 degree day), with a 10 degree days step. For longevity, I tested values from 10 to 14 days (which is plus/ minus 2 days from the mean longevity of 12 days of Cx. mosquitoes). There were a total of 25 different parameter combinations (5 for EIP and 5 for longevity). I ran the original degree day model on each of these combinations and selected the combination that yielded the smallest error from observation. The results were provided as Table 1 in appendix C.

Improving model fit.

117

As the simple degree days model failed to capture the WNV emergence period in

Pennsylvania (table 2), I considered several adjustments. Reisen et al. (2006) argued the EIP is not a constant across the season and different strains of WNV (original NY99 and later WN02) have significantly different EIPs (Moudy et al. 2007). Dohm et al. (2002a) and Reisen et al. (2006) reported that it takes longer for Culex mosquito to successfully transmit WNV in the lab under low temperatures (20ºC). More explicitly, Kilpatrick (et al. 2008) argued that the transmission rate of

WNV increased much faster at higher temperatures and a new WNV strain, WN02, had replaced the original NY99 strain. However many previous DD models were based on the life history parameters of the NY99 strain (Reisen et al. 2006, Zou et al. 2007). Konrad et al. 2009 had already shown that the EIP should be lowered in order to successfully predict WNV transmission period in Santa

Barbara, California, which coincided well with Kilpatrick’s argument. To include this information in our model, I suggested that EIP threshold to enable WNV transmission should be relatively higher at the beginning of the season and lower later on when temperature rose.

In Zou’s original approach they set mosquito longevity as a constant of 12 days across the season. But mosquito survivorship and longevity are temperature dependent as well. Juricic et al.

(1974) reported optimal temperature for C. tarsalis development was around 25ºC and Tachiiri et al.

(2006) further demonstrated the relationship between daily mortality rate and temperature. Thus I allowed longevity to vary to improve our model fit as well.

I ran the simulation model again to find optimal parameter combination (EIP and longevity) that minimizes the error between predicted WNV emergence dates and field observed dates. In our modeling approach I divide the entire season (day 150 through day 300) into two intervals: day 150 to day 210 (May 5 to July 29) as early season and day 211 (July 30) afterwards as late season. Each interval is assigned with a specific value of EIP and longevity. Again I allow the EIP to vary from

89DD to 129DD, with a 10DD step, and longevity to vary from 10 to 14 days. This creates 625

118 different combinations and I compare predictions from these simulations to the observed data.

Results differed for Philadelphia compared to the other sites (see below. Note: The dataset of this approach is simply too large to show in an appendix. It only employs a basic summing operation and choosing the smallest error value, which was accomplished using scripts to directly pick up the optimal combination, which I show in the results.)

Results

Climate and Mosquito Data

Of all four locations (Figure 4-1), Philadelphia had the hottest temperature in all years. Its mean temperature, on average, is about 3ºC more than that of Harrisburg, 3.5ºC more than Pittsburgh and 3.8ºC more than Williamsport. So I would expect Philadelphia to have the longest duration of

WNV emergence (Table 1) There is no significant increasing or decreasing trend of mean temperature (p>0.05) in all four locations but I do see that 2002 and 2006 are relatively hot years for

Philadelphia and 2003 a relatively cool year for Pittsburgh and Williamsport (table 1).

In Pennsylvania the distribution of WNV-positive species differed between Philadelphia and the other sites across our experiment period, whether viewed from the perspective of positive batchs, or numbers of individuals in the positive batches (Fig. 4-4, top and bottom panels, respectively). I did not see appreciable variation in the species composition among the 7 years. Cx. pipiens and Cx. restuans were dominant species (>10% in species composition in WNV-positive population) in all locations, but Cx. salinurius was the most dominant in Philadelphia., Cx. salinurius concentrated in Philadelphia exclusively. They were rare to absent in in the other three locations.

Original degree days Model Results

119

I record first and last WNV emergence dates in each year based on the model prediction and compare them with field observation in all four locations (Table 1 in appendix C). I confirm that the current parameter combination (109 DD as EIP and 12 days as longevity) is indeed the best of all 25 combinations. But this model does not perform well in Pennsylvania even with this parameter combination. This model consistently overestimates WNV emergence period (table 4-2, figure 4-2): the predicted first WNV emergence date tends to be earlier and last emergence date later than observed emergence dates. For 18 observed first emergence data points, the model only successfully predicts 6 cases within 10 days. For the remaining 13 unsuccessful predictions, only 1 of them is underestimated, which means the predicted first emergence date is later than observed. All other 13 unsuccessful predictions consistently overestimate emergence dates. Similarly, for 25observed last incidence data points, the model successfully predicts 4 cases within the applicable range. Four of the 19 unsuccessful predictions are underestimations and the remaining 16 are overestimations (table

4-2, figure 4-5). Hence this model has a tendency to overestimate the transmission period in

Pennsylvania.

Moreover, in 2004, the original model fails to predict a reasonable first emergence date for

Williamsport (see footnote of table 4-2). In addition, I know Philadelphia is the hottest location

(table 4-1) hence I expect Philadelphia should have longest duration of WNV incidence.

Nevertheless the observation data show Philadelphia has a much shorter WNV period than predicted, which suggest there might be a different virus emergence pattern in Philadelphia. All these results indicate the necessity of alternative model designs. As discussed above, I consider dividing the season, and assigning different EIP and longevity to the different parts of the season.

Variable EIP and Longevity Adjustment

120

I know that one of the original model assumptions, that the EIP and longevity remain

constant across the season, is not completely valid (Tachiiri et al. 2006, Gong et al. 2010). So in our

calibrated model I maintain the EIP of 109 degree days at the beginning of the season (day 150), but

reduce it to 89 degree days elsewhere, and set mosquito longevity as 10 days at the beginning of the

season, then 12 days afterward. This specific combination is selected because it minimized total

prediction error in all four locations during the 7 years surveillance period.

These adjustments improve model fitting substantially (table 4-3, figure 4-3) and there is a

substantial difference between the adjusted model and the original model. For all 19 first emergence

data points the adjusted model predicts 13 of them within 10 days of the observed values. The

remaining 6 unsuccessful predictions are all overestimations. For 25 last emergence dates the

adjusted model correctly reports 15 incidences. Of 10 unsuccessful cases, 7 are underestimations and

3 are overestimations. Philadelphia, however, fails to yield a better fit after model adjustment, and it

has the largest prediction error.

So far I was able to enhance model prediction but Philadelphia remains problematic. As

noted above, the mosquito species composition of Philadelphia is quite different from the other

locations (figure 4-4, showing species composition in all four locations). Cx. salinarius became the dominant infectious species during the late season in Philadelphia, while all other locations did not have many records of Cx. salinarius. Thus I suspected species composition difference might result in the failure of our model prediction because different vector species might have quite different life history parameters. Unfortunately, there was no record of WNV specific parameters for Cx. salinarius. Nevertheless I can infer them from our observation data by testing the EIP threshold and longevity parameters from a potential range and see which values best approximates the observed results. I chose 159 DD as threshold for early season detection and 89 DD for later on. The predicted and observed emergence dates are presented in table 4 and provide a substantial improvement of the

121 model fit than the original model (see figure 4-5). Of all 19 first emergence dates, the adjusted model with calibration for Philadelphia successfully predicted 14 incidences within 10 days, a 74% success rate. Of all 25 last incidence data the model successfully predicted 21 of them, a success rate 84%.

Comparing these percentages with the less than 30% success rate using simple degree day model, our model adjustment was necessary and effective.

Discussion

Our final model could provide robust and generally accurate (+/- 10-day) predictions of

WNV emergence time in multiple locations across different years. The model performs consistently among the 3 locations that shared a similar community composition. Generally speaking the model shows minor performance variability across years as well. But in 2005 and 2006 the model performance is relatively poor (Table 3). In 2005, the model fails to capture first emergence in

Philadelphia and Williamsport; and last emergence in Philadelphia as well. In 2006, the model fails to predict last emergence of Harrisburg and Pittsburgh. Based on the weather data, 2005 and 2006 are both average years, with moderate temperature and precipitation which excludes the influence of extreme climate conditions. Indeed, the errors in these two years are so large that it is possible to be due to observation error. Because I am using field data, observation errors are important to determine potential (optimal) degree day requirement (EIP) for Culex mosquitoes to become infectious. However I believe I still need better basic knowledge of Culex mosquito life history to improve the model. Researchers need better monitoring and sampling systems to enhance data quality as well, but I also know these systems require much labor and financial inputs and might not be easy to implement.

There is a possible flaw in our model. Based on simulation I suggest the optimal longevity should be 10 days at the beginning of the season (Julian day 150 to Julian day 210) and 12 days later

122 on. However some research based on lab experiments under constant temperature (Reisen et al. 2006,

Tachiiri et al. 2006) suggest that adult mosquitoes have longer longevity (hence smaller mortality rate) towards lower temperatures. So our model assumption seems contradictory to their findings.

Nevertheless from an operational perspective I think that treating longevity as shorter in the early season is satisfactory, because I am using PA-DEP collected field data and I do not know adult mosquito longevity in the field. Although I do not think the difference between 10 and 12 days longevity makes a big difference ecologically, these differences do have a large impact on model performance. I suggest more detailed knowledge and field observations for the mosquitoes are definitely needed and important for better modeling mosquito life history.

Ideally, site-specific and year-specific models should be the most accurate (as seen from the

Philadelphia case, where the species composition differed). But site- and year-specific models are more computationally expensive and not generic enough to serve as predictive models. Our goal is to provide not only a retrospective model which determines the best parameter combinations, but also a predictive model. There is always a trade-off between model accuracy and model simplicity. Our suggestion is that I might divide different regions in Pennsylvania based on vector mosquito species composition (not necessarily by region) and use different model parameters based on the species composition. I believe it will deliver better model accuracy and be more useful for public health and ecological responses.

In our revised model I allowed parameters (EIP and longevity) to change during the season, using a cut-off point dividing two intervals in the season. This use of a categorical cut-off point, and the timing of that changewas, of course, arbitrary, especially in the presence of global climate change. However it is relatively difficult to determine when actually the cut-off point would be.

Again since I would like to present a simple and generic model, I use simple values, but would be interested in exploring the temporal variation in EIP and longevity for future modeling efforts.

123

Besides the cut-off point criterion, determining whether the model prediction is good is also arbitrary.

Zou et al. (2007) set it as 7 days. I have, however, a much larger dataset (7 vs 2 years), and I set the acceptable interval as 10 days. Ideally this interval should be flexible in disease control. For example, in outbreak years this interval needs to be small, otherwise the model is useless for early warning of emerging incidences, and help in estimating when to safely decrease management efforts.

Temperature is one of the most important environmental drivers for disease dynamics.

Kilpatrick et al. (2008) further argues that WNV transmission increases sharply with increasing temperature. They suggest that the transmission of WNV is best fit with the power of 4 of temperature. I do not have daily infectious probability data for each mosquito, and could not test the influence of increasing temperature using this model. Nevertheless, climate change, especially global warming, imposes serious threats to public health (Lafertty 2009, Randolph 2009). I would like to use our model to predict WNV emergence under climate change conditions and assess potential risk in Pennsylvania. In our modeling approach I use simulations based on different parameter combinations to determine potential degree day requirement and other life history parameters for

Culex to become infectious. However I still need better basic knowledge of Culex mosquito life history to improve the model. I need better monitoring and sampling system to enhance data quality as well.

Not only is the adult stage influenced by temperature, other life stages, especially larva, are temperature dependent as well. Some researchers have built complicated and sensitive life history models to predict WNV outbreaks (Tachiiri et al. 2006, Gong et al. 2010). While these models could track mosquito life history more explicitly and provide improved accuracy, they are usually computationally intensive and require much more and accurate model inputs (e.g. degree day requirement for larva development, survivorship, etc.), and might be more sensitive to error in parameter estimates as many relevant functions tend to be nonlinear. Here, I focused on basic degree

124 day models and balance model accuracy and simplicity. I would definitely explore the multi-stage models if I am provided with more detailed data. However based on our current datasets I think our model still provides a feasible tool for predicting WNV risk in Pennsylvania.

Besides environmental factors, I also suggest that sampling efficacy would significantly influence field incidence detection sensitivity. In our model I simply consider the entire cohort of mosquitoes is WNV positive at the beginning of simulation. However it is very likely that only a small fraction of the mosquito population has virus particles early in the season, which imposes the further burden to detect the incidences in the field. From an operational perspective I suggest increasing sampling frequency and efficacy at the beginning of the season so that the first emergence could be better recognized. I also suspect that different mosquito collecting methods (i.e. different traps) could influence our monitoring efforts. Gravid traps are designed to attract virus-positive mosquitoes while light traps have no preference on vector status (containing virus particles or not).

Currently gravid trap is the most widely applied in Pennsylvania followed by light trap and a few others (magnet, aspiration, faye print, etc). Of all the WNV-positive batches, more than 77% were captured in gravid trap, and light traps got about 19%.

Moreover, other than sampling efficacy alone, many anthropogenic factors might also influence WNV dynamics and alter the emergence dates. For instance, consistent and regular pesticide sprays, especially during the early time in the season, may significantly reduce mosquito populations to a very low level.

In summary, I have proposed a successful modeling framework to assess potential WNV risk.

My modeling approach could be easily adapted to other locations and could be utilized to model future disease risk under climate change conditions.

125

References Cited

Allen, J. C. 1976. A modified sine wave method for calculating degree-days. nviron. Entomol.

5:388-396.

Bellan, S.E. 2010. The importance of age dependent mortality and the extrinsic incubation period in

models of mosquito-borne disease transmission and control. PLoS Pathogen. 5: e10165.

Bolling, B. G., C. M. Barker, C. G. Moore, W. J. Pape, and L. Eisen. 2009. Modeling/GIS, risk

assessment, economic impact: seasonal patterns for entomological measures of risk for

exposure to Culex vectors and West Nile virus in relation to human disease cases in

Northeastern Colorado. J. Med. Entomol. 46: 1519-1531.

Campbell, G.L., A.A. Marfin, R.S. Lanciotti, and D.J. Gubler. 2002. West Nile virus. Lancet

Infect. Dis. 2: 519-529.

CDC website of West Nile virus in the U.S.. 2011. http://www.cdc.gov/NCIDOD/dvbid/westnile/.

Dohm, D. J., M. L. O’Guinn, and M. J. Turell. 2006a. Effect of environmental temperature on the

ability of Culex pipiens (Diptera: Culicidae) to transmit West Nile virus. J. Med. Entomol.

39: 221-225.

Dohm, D. J., M. R. Sardelis, and M. J. Turell. 2006b. Experimental vertical transmission of West

Nile virus by Culex pipiens (Diptera: Culicidae). J. Med. Entomol. 39: 640-644.

Forsythe, W. C., E. J. Rykiel, R. S. Stahl, H. Wu, and R. M. Schoolfield. 1995. A model

comparison for daylength as a function of latitude and day of a year. Ecological Modelling

80: 87-95.

126

Gong, H, A. T. DeGaetano, and L. C. Harrington. 2010. Climate-based models for West Nile

Culex mosquito vectors in the Northeastern US. Int. J. of Biometeorol. DOI:

10.1007/s00484-010-0354-9.

Juricic, D., B. E. Eno, and G. Parikh. 1974. Mathematical modeling of a virus vector: Culex

tarsalis. Biomed. Sci. Instrum. 10: 23-28.

Kilpatrick, A. M., M. A. Meola, R. M. Moudy, and L. D. Kramer. 2008. Temperature, viral

genetics, and the transmission of West Nile virus by Culex pipiens mosquitoes. PLoS

Pathogen 4: e1000092.

Komar, N., N. A. Panella, J. E. Burns, S. W. Dusza, T. M. Mascarenhas, and T. O. Talbot.

2001. Serologic evidence for West Nile virus infection in birds in the New York City vicinity

during an outbreak in 1999. Emerging Infectious Diseases 7: 621-625.

Konrad, S. K., S. N. Miller, W. K. Reeves, and N. S. Tietze. 2009. Spatially explicit West Nile

virus risk modeling in Santa Clara County, California. Vector Borne Zoonotic Diseases 9:

267-274.

LaDeau, S. L, A. M. Kilpatrick, and P. P. Marra. 2007. West Nile virus emergence and large-

scale declines of North American bird populations. Nature 447: 710-713.

Lafferty, K. D. 2009. The ecology of climate change and infectious diseases. Ecology 90: 888-900.

Lanciotti, R. S., J. T. Roehrig, V. Deubel, J. Smith, M. Parker, K. Steele, B. Crise, K. E. Volpe,

M. B. Crabtree, J. H. Scherret, R. A. Hall, J. S. MacKenzie, C. B. Cropp, B.

Panigraphy, E. Ostlund, B. Schmitt, M. Malkinson, C. Banet, J. Weissman, N. Komar,

H. M. Savage, W. Stone, T. McNamara, and D. J. Gubler. 1999. Origin of the West Nile

virus responsible for an outbreak of encephalitis in the northeastern United States. Science

286:2333-2337.

127

Moudy, R. M., M. A. Meola, L. L. Morin, G. D. Ebel, and L. D. Kramer. 2007. A newly

emergent genotype of West Nile virus is transmitted earlier and more efficiently by Culex

mosquitoes. Am. J. Trop. Med. Hyg. 77: 365-370.

Nash, D., F. Mostashari, J. Miller, D. O’Leary, K. Murray, A. Huang, A. Rosenberg, A.

Greenberg, M. Sherman, S. Zaki, P. Smith, and M. Layton. 2001. The outbreak of West

Nile virus infection in the New York City area in 1999. The New England Journal of

Medicine 344: 1807-1814.

Oda, T., Y. Eshita, K. Uchida, M. Mine, K. Kurokawa, Y. Ogawa, K. Kato, and H. Tahara.

2002. Reproductive activity and survival of Culex pipiens pallens and Culex

quinquefasciatus (Diptera: Culicidae) in Japan at high temperature. J. Med. Entomol. 39:

185-190.

Parton, W. J. and J. A. Logan. 1981. A model for diurnal variation in soil and air temperature.

Agricultural Meteorology 23: 205-216.

Peterson, L.R. 2001. West Nile virus: a reemerging global pathogen. Rev. Biomed. 12: 208-216.

Rahbek, C. 2007. Disease ecology: the silence of the robins. Nature 447: 652-653.

Randolph, S. E. 2009. Perspectives on climate change impacts on infectious diseases. Ecology 90:

927-931.

Reisen, W. K., Y. Fang, and V. M. Martinez. 2006. Effects of temperature on the transmission of

West Nile virus by Culex tarsalis (Diptera: Culicidae). J. Med. Entomol. 43: 309-317.

Rueda, L. M., K. J. Patel, R. C. Axtell, and R. E. Stinner. 1990. Temperature dependent

development and survival rates of Culex Quinquefasciatus and Aedes aegypti. J. Med.

Entomol. 27: 892-898.

128

Sugumaran, R., S. R. Larson, and J. P. DeGroote. 2009. Spatio-temporal cluster analysis of

county-based human West Nile virus incidence in the continental United States. International

Journal of Health Geographics 8:43.

Tachiiri, K., B. Klinkenberg, S. Mak, and J. Kazmi. 2006. Predicting outbreaks: a spatial risk

assessment of West Nile virus in British Columbia. International Journal of Health

Geographics 5: 1-21.

Wann, M., D. Yen and H. J. Gold. 1985. Evaluation and calibration of three models for daily cycle

of air temperature. Agricultural and Forest Meteorology 34: 121-128.

Zou, L., S. N. Miller, and E. T. Schmidtmann. 2007. A GIS tool to estimate West Nile virus risk

based on a degree-day model. Environ. Monit. Assess 129: 413-420.

129

Table 4-1. Annual Mean Temperature Summary from 2002-2008 in 4 Locations

2002 2003 2004 2005 2006 2007 2008 Average Harrisburg Mean 12.28 10.71 11.22 11.47 12.1 11.55 11.29 11.58 Min -7.22 -11.67 -12.78 -11.11 -6.67 -14.44 -10 -6.46 Max 30 28.33 26.67 29.44 30.56 28.33 29.44 29.3 Philadelphia Mean 15.72 14.32 14.45 15.12 15.73 13.85 14.57 14.67 Min -2.78 -7.78 -8.33 -9.44 -4.44 -10 -5.56 -7.15 Max 33.33 31.11 30.56 33.33 33.89 30.56 32.78 32.36 Pittsburgh Mean 11.61 10.35 11.04 11.03 11.48 11.38 10.69 11.116 Min -8.89 -15 -13.89 -13.33 -11.11 -16.11 -12.78 -12.56 Max 27.78 26.11 25.56 28.89 28.33 28.33 28.33 27.63 Williamsport Mean 11.21 9.85 10.5 10.68 11.59 10.96 10.6 10.8 Min -10.56 -12.78 -14.44 -12.78 -7.78 -12.78 -11.11 -11.32 Max 28.89 27.78 26.11 28.33 29.44 28.89 28.33 28.47

130

Table 4-2. Predicted, Observed Emergence Dates and Difference (in days) in Original Model

Year/location Harrisburg Philadelphia Pittsburgh Williamsport 2002 First Last First Last First Last First Last Predicted 3-Jul 28-Aug 28-Jun 14-Sep 15-Jul 22-Aug 5-Jul 24-Aug Observed 17-Jul 2-Oct 9-Jul 8-Oct NA 24-Sep 25-Jun 26-Sep Difference 14 36 12 24 NA 33 -9 33

2003 First Last First Last First Last First Last Predicted 8-Jul 15-Aug 30-Jun 3-Sep 4-Aug 27-Oct 10-Apr 27-Oct Observed 11-Jul 18-Sep 26-Jun 17-Oct NA 16-Sep 15-Jul 15-Sep Difference 3 34 -4 45 NA -41 96 -42

2004 First Last First Last First Last First Last Predicted 13-Jul 27-Oct 22-Jun 6-Sep 10-Apr 27-Oct NA* 25-Oct Observed NA 8-Sep 9-Jul 21-Sep 13-Jul 13-Sep NA 10-Aug Difference NA -49 17 15 94 -44 NA -78

2005 First Last First Last First Last First Last Predicted 15-Jun 17-Oct 13-Jun 26-Sep 20-Jul 20-Aug 6-Jul 14-Aug Observed NA NA 25-Jul 13-Sep 11-Jul 29-Sep 21-Jul 14-Sep Difference NA NA 42 -13 -9 40 15 31

2006 First Last First Last First Last First Last Predicted 21-Jul 10-Aug 7-Jun 29-Aug 12-Jun 8-Aug 3-Jul 10-Aug Observed 21-Jun 11-Oct NA 12-Sep NA 11-Oct NA 11-Sep Difference -30 61 NA 15 NA 64 NA 33

2007 First Last First Last First Last First Last Predicted 6-Aug 10-Aug 14-Jul 18-Aug 7-Jul 12-Aug 5-Aug 3-Oct Observed 25-Jul 4-Oct 24-Jul 18-Oct NA 26-Sep 7-Aug NA Difference -12 51 10 61 NA 45 2 NA

2008 First Last First Last First Last First Last Predicted 18-Jun 23-Jul 13-Jun 11-Sep NA* 27-Oct 20-Jun 25-Sep Observed 22-Jul 23-Sep 24-Jun 25-Sep 31-Jul 30-Sep 28-Jul NA Difference 35 61 11 14 NA -27 38 NA

131

Note: NAs usually indicate no record in the field data. NA*s indicate the model fails to predict a (reasonable) emergence date. For instance in 2004 for Williamsport the model prediction of first emergence was in October, which is already later than observed last emergence date (10-August) so we believe this prediction is unreasonable hence use NA* to indicate that.

132

Table 4-3. Predicted, Observed Emergence Dates and Difference in Adjusted Model

Year/location Harrisburg Philadelphia Pittsburgh Williamsport 2002 First Last First Last First Last First Last Predicted 29-Jun 26-Sep 27-Jun 7-Sep 30-Jun 24-Sep 30-Jun 24-Sep Observed 17-Jul 2-Oct 9-Jul 8-Oct NA 24-Sep 25-Jun 26-Sep Difference 18 6 12 35 NA 0 -5 2

2003 First Last First Last First Last First Last Predicted 9-Jul 9-Sep 1-Jul 6-Sep 19-Aug 9-Sep 7-Jul 9-Sep Observed 11-Jul 18-Sep 26-Jun 17-Oct NA 16-Sep 15-Jul 15-Sep Difference 2 9 -5 41 NA 7 -8 6

2004 First Last First Last First Last First Last Predicted 13-Jul 18-Sep 24-Jun 6-Sep 7-Jul 19-Sep 8-Jul 20-Sep Observed NA 8-Sep 9-Jul 21-Sep 13-Jul 13-Sep NA 10-Aug Difference NA -10 15 15 6 -6 NA -41

2005 First Last First Last First Last First Last Predicted 15-Jun 28-Sep 14-Jun 25-Sep 8-Jul 21-Sep 14-Jun 21-Sep Observed NA NA 25-Jul 13-Sep 11-Jul 29-Sep 21-Jul 14-Sep Difference NA NA 41 -12 3 8 37 -7

2006 First Last First Last First Last First Last Predicted 19-Jun 5-Sep 23-Jun 30-Aug 19-Jul 31-Aug 11-Jul 5-Sep Observed 21-Jun 11-Oct NA 12-Sep NA 11-Oct NA 11-Sep Difference 2 36 NA 13 NA 41 NA 6

2007 First Last First Last First Last First Last Predicted 18-Jul 10-Oct 16-Jul 6-Sep 9-Jul 30-Sep 12-Jun 9-Oct Observed 25-Jul 4-Oct 24-Jul 18-Oct NA 26-Sep 7-Aug NA Difference 7 -6 8 42 NA -4 56 NA

2008 First Last First Last First Last First Last Predicted 25-Jul 15-Sep 13-Jun 10-Sep 29-Jul 24-Sep 28-Jul 17-Sep Observed 22-Jul 23-Sep 24-Jun 25-Sep 31-Jul 30-Sep 28-Jul NA Difference -3 8 11 15 2 6 0 NA

Note: NAs usually indicate no record in the field data. NA*s indicate the model fail to predict a (reasonable) emergence date. For instance in 2004 for Williamsport the model prediction of first emergence was in October,

133 which is already later than observed last emergence date (10-August) so we believe this prediction is unreasonable hence use NA* to indicate that.

134

Table 4-4. Predicted, Observed Emergence Dates and Difference For Philadelphia

Year 2002 2003 2004 2005 2006 2007 2008 First Obs. 27-Jun 1-Jul 24-Jun 24-Jul NA 24-Jul 24-Jun First Pred. 19-Jun 2-Jul 28-Jun 16-Jun 28-Jun 6-Aug 19-Jun Difference 8 -1 -4 39 NA -13 5

Last Obs. 8-Oct 17-Oct 21-Sep 13-Sep 12-Sep 18-Oct 25-Sep Last Pred. 8-Oct 9-Oct 26-Sep 7-Oct 18-Sep 11-Oct 25-Sep Difference 0 8 -5 -24 -6 7 0

135

Fig. 4-1. Topography of Pennsylvania and the Locations of the Four Sites (red triangles) Used to Model WNV

136

Fig. 4-2. Difference of Observation and Model Prediction in Original Model (Upper panel: first emergence dates; lower panel: last emergence dates. Two dashed lines: +/-10 days interval of success prediction, The last emergence incidence dates are always poorly predicted in the original model. Missing observation data: Pittsburgh first in 2002, 2003; Williamsport first in 2004; Harrisburg first and last in 2005; Philadelphia, Pittsburgh, and Williamsport first in 2006; Pittsburgh and Williamsport first in 2007; Williamsport last in 2008. )

137

Fig. 4-3. Difference of Observation and Model Prediction in Adjusted Model (Upper panel: first emergence dates; lower panel: last emergence dates. Two dashed lines: +/-10 days interval of success prediction. Note substantial improve over the original model. Missing observation data: Pittsburgh first in 2002, 2003; Williamsport first in 2004; Harrisburg first and last in 2005; Philadelphia, Pittsburgh, and Williamsport first in 2006; Pittsburgh and Williamsport first in 2007; Williamsport last in 2008.)

138

Fig 4-4. WNV Positive Species Composition in Four Locations from 2002-2008 (Upper panel: number of positive batches, regardless of pool size in each batch. Cx pipiens is the most abundant species in all locations, followed by Cx restuans except Philadelphia. In Philadelphia the second most abundant species is Cx salinurius. Lower panel: number of total mosquitoes in positive batches. Cx salinurius is the most abundant in Philadelphia, but not present in Pittsburgh at all, and has a low quantity in Harrisburg and Williamsport as well. This figure summarizes the species composition over entire seven years period. I do not observe substantial species composition change among different years.)

139

Fig. 4-5. Difference of Observation and Model Prediction in Adjusted Model with Calibration for Philadelphia (Upper panel: first emergence dates; lower panel: last emergence dates. Two dashed lines: +/-10 days interval of

success prediction. Missing observation data: Pittsburgh first in 2002, 2003; Williamsport first in 2004;

Harrisburg first and last in 2005; Philadelphia, Pittsburgh, and Williamsport first in 2006; Pittsburgh and

Williamsport first in 2007; Williamsport last in 2008.)

140

Appendix

Appendix A. Degree Day Modeling Procedure Incorporating Daily Temperature Range

Mean Temperature Model

This is the simplest model which treats daily mean temperature as the arithmetic mean of daily maximum and minimum temperature:

TT T  max min mean 2

The disadvantage of this model is obvious. Although Tmean is derived from daily maximum and minimum temperature, it does not contain any information about them, that is, I am ignoring DTR information completely.

Sine-negative exponential Model

To overcome the problem inherited in the mean temperature model, biometerologists have proposed sine curve model (sinusoidal model) to simulate DTR profile, which has the basic form as:

Ti  asin bi  c

Four inputs are required in order to reconstruct DTR profile: daily maximum temperature, minimum temperature to determine coefficient a and c. Coefficient b is determined by photoperiod, which itself is further considered as a function of latitude and number of the day in the year. Thus I will need daily maximum, minimum temperature, latitude, and number of day in the year as inputs. The accumulated degree day is basically the integration over basic development threshold. To reduce computational effort I interpolate hourly temperatures and use these temperatures to approximate degree day accumulation above threshold in a day. This is a direct analog of Riemann Integral.

Forsythe et al. (1994) have proposed the following equations to calculate daylength:

141

 0.2163108  2arctan(0.9671396 tan(0.00860  (J  186)))  arcsin(0.39795cos  ) p L  sin sin sin 24 D 24  cos1 180 180 L  cos cos 180

where  is the revolution angle predicted by number of day in the year (J),  is the declination angle, L is

latitude, and D is the daylength. p is the sun position constant and equals 0.8333 according to US government

definition (Forsythe et al. 1994).

While I have calculated the daylength I could further determine temperature profile in a day:

()t t   T( T  T )sin(r )  T ( t  t  t ) imax minD2( a   ) min n x 1 (())t t  t   1 1 r s T( T  T ')  ( T  T ')cos(2 ) ( t  t  t ) imax min max minD x n ' 2 2 24  a   2

where t is current time, D is day length, tr and ts are the sunrise and sunset time derived from daylength, Tmin’ is

minimal temperature of the next day, and a and  are predetermined coefficients (Wann et al. 1985).

The sine-negative exponential model is a fine tuning version of the original sine curve model. Field observation

data suggest that DTR is not necessarily symmetric in a day and the temperature curve after sunset is better

fitted with an negative exponential curve. Thus the final model is formulated as:

 m Ti( Tmax  T min )sin( )  T min ( t n  t  t x ) D 2 a bn  Z Ti Tmin ( T s  T min ) e ( t x  t  t n ' )

where D is day length (determined by latitude and day in a year), Z is night length, Ts is the temperature at the

moment of sunset, m is number of hours after minimum temperature until sunset, n is number of hours after sunset until the time of minimum temperature of the next day, a is the lag coefficient for the maximum temperature, and b is the nighttime temperature coefficient (Parton and Logan 1981). Because sine-negative

142

exponential model is derived from sine curve model, I will call it sine curve model in the following text for

simplicity.

Appendix B: Original degree day modeling for predicting WNV emergence dates

In this model (also described in Zou et al. 2007, Konrad et al. 2009) I use the median extrinsic

incubation period (EIP50) as the criterion of potential WNV transmission. More specifically, in a given day if

the EIP50 could be completed within the mean longevity of Culex mosquitoes’ life span (12 days before and including that day) then I consider that day is possible for WNV transmission. For example, consider a day in the very early season, say, Feb. 1st (day 32), and I accumulate the degree day (DD) above development

threshold for Culex mosquito (14.3 °C, Reisen et al. 2006) from Feb. 1st and 11 days prior. That particular

timing is too early, temperatures are too low, and so the DD accumulation during the longevity period is zero,

which simply means the mosquito cannot develop at all. When it gets warmer, for instance in day 150 (late

May), the total DD accumulated during mosquito’s longevity is about 50 in Harrisburg, Pittsburgh, and

Williamsport; or 75 in Philadelphia (see figure 1 in appendix A, the starting point). However these accumulated

DD are not high enough to allow the virus particles in the mosquito to be completely matured and able to be

transmitted, because they are substantially below the EIP50 value, which is determined as 109 DD by Reisen et

al. 2006.

I computed the DD accumulation for each day across the season (from Julian day 150 to Julian day 300,

other than that, the climate is too harsh and temperature is below the basic development threshold for

mosquitoes, even for Philadelphia). I record the first and last day in the season that the DD accumulation is

above the EIP50 threshold (the first and last intersection of the DD curve (in red) and the EIP50 threshold line

(dotted), see figure 1 in appendix B). I then compare these predicted dates with field observation data and

compute the error (difference between the observed and predicted days) and use this error as the criterion to

evaluate model performance (basically parameter performance).

143

Figure 4-1 (Appendix). DD Accumulation in Mosquito Longevity and Potential WNV Transmission Period in 2002 (Solid red line: DD accumulation during mosquito longevity; Dotted black line: EIP50 threshold=109DD. The first and last intersections of these two lines indicate first and last potential WNV emergence dates, respectively.

144 Appendix C: Selecting the optimal parameter combination in the original model

Table 4-1 (Appendix). Error (in Days, comparing to field observation) of Original Model with Different Parameter Combinations

Longevity 10 11 12 13 14 EIP 89 99 109 119 129 89 99 109 119 129 89 99 109 119 129 89 99 109 119 129 89 99 109 119 129 H.F.2002 20 19 16 13 -20 39 19 18 15 13 40 33 14 17 18 40 36 33 18 16 41 40 35 18 17 H.L.2002 35 36 37 38 40 33 34 35 36 38 32 33 36 35 34 10 32 32 34 34 9 10 31 32 33 H.F.2003 9 7 1 92 92 9 7 6 1 92 10 8 3 5 6 10 9 6 5 5 16 10 8 5 5 H.L.2003 26 31 33 -39 -39 18 26 30 33 -39 16 19 34 30 25 13 15 22 25 27 10 14 15 22 25 H.F.2004 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA H.L.2004 32 58 -49 -49 -49 2 33 -49 -49 -49 -1 3 -49 -49 55 -3 -1 32 35 -49 -5 -3 0 33 34 H.F.2005 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA H.L.2005 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA H.F.2006 -4 -15 -29 -32 -45 -3 -7 -15 -30 -32 -3 -4 -30 -14 -7 13 -4 -6 -8 -15 13 -3 -5 -7 -9 H.L.2006 46 61 62 63 65 42 47 60 61 62 41 42 61 60 47 40 41 42 57 59 38 40 41 42 57 H.F.2007 52 9 -11 -16 106 52 51 8 -11 -16 52 51 -12 8 50 53 51 50 25 7 53 52 51 49 25 H.L.2007 48 49 52 55 -23 46 47 48 52 54 20 46 51 48 47 19 23 46 47 48 18 20 23 45 46 H.F.2008 40 39 36 35 103 41 40 39 36 34 41 40 35 37 39 42 41 39 39 37 42 41 40 39 37 H.L.2008 44 45 61 98 -34 42 43 45 61 62 29 41 61 45 43 8 35 41 43 46 6 16 35 40 42 Ph.F.2002 37 36 32 12 11 79 36 35 32 12 80 36 12 33 35 80 36 36 34 33 80 78 36 34 33 Ph.L.2002 14 15 27 42 43 11 13 15 40 41 3 11 24 15 13 1 4 11 13 17 -1 2 4 11 13 Ph.F.2003 10 -1 -2 -4 -6 10 9 -2 -3 -5 11 9 -4 -2 6 11 10 8 1 -1 12 10 9 6 1 Ph.L.2003 25 43 46 48 49 20 25 43 46 47 19 21 45 42 25 18 19 21 39 41 17 18 20 21 39 Ph.F.2004 51 48 19 11 -1 51 49 43 18 11 51 50 17 44 48 51 50 48 45 44 53 50 49 47 44 Ph.L.2004 6 14 15 30 45 2 6 12 15 21 0 2 15 11 6 -1 0 3 6 10 -10 -2 1 3 8 Ph.F.2005 45 44 43 41 40 47 45 44 42 41 47 46 42 43 45 47 46 45 45 43 47 46 45 45 44 Ph.L.2005 -15 -14 -12 -9 22 -17 -16 -14 -13 -10 -20 -17 -13 -15 -16 -24 -20 -18 -16 -15 -25 -24 -20 -18 -16 Ph.F.2006 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA Ph.L.2006 13 14 15 20 33 11 13 14 15 19 9 11 15 13 12 -7 9 10 12 13 -8 -6 8 10 12 Ph.F.2007 52 51 12 9 -11 53 52 27 11 8 53 52 10 26 51 53 52 51 49 25 53 52 51 50 46 Ph.L.2007 34 35 62 63 67 18 34 35 62 63 9 33 61 35 34 8 10 33 34 37 7 8 9 32 34 Ph.F.2008 14 13 12 11 9 17 13 12 12 11 17 15 11 12 14 18 16 14 13 12 18 16 15 14 13 Ph.L.2008 6 8 15 46 48 4 6 9 14 46 1 5 14 8 6 0 2 5 6 8 -2 0 3 5 6 Pi.F.2002 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA

145 Pi.L.2002 30 32 36 50 -33 28 30 32 33 50 14 27 33 31 29 10 22 27 29 31 3 10 22 27 28 Pi.F.2003 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA Pi.L.2003 25 -41 -41 -41 -41 17 25 -41 -41 -41 13 18 -41 -41 25 11 14 19 25 -41 9 13 14 19 25 Pi.F.2004 -3 94 94 94 94 22 -4 94 94 94 22 21 94 94 94 49 21 -3 94 94 51 48 -3 -4 94 Pi.L.2004 60 -44 -44 -44 -44 57 60 -44 -44 -44 3 57 -44 -44 -44 1 3 57 -44 -44 -5 1 4 57 -44 Pi.F.2005 28 27 -9 92 92 29 27 26 -9 92 29 28 -9 25 26 29 28 27 25 4 29 28 27 26 6 Pi.L.2005 39 39 47 -28 -28 37 38 39 41 -28 33 36 40 39 38 10 33 36 37 39 4 30 34 35 37 Pi.F.2006 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA Pi.L.2006 61 63 65 -16 -16 44 61 62 64 -16 42 44 64 62 60 41 42 44 60 61 40 41 42 57 59 Pi.F.2007 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA Pi.L.2007 26 28 47 -31 -31 16 26 44 46 -31 14 17 45 44 26 13 16 17 40 43 11 13 16 20 41 Pi.F.2008 48 44 112 112 112 48 47 44 112 112 49 48 112 42 44 49 48 47 44 112 49 49 47 45 44 Pi.L.2008 53 105 -27 -27 -27 50 62 104 -27 -27 49 51 -27 104 61 22 49 51 61 -27 14 22 49 51 59 W.F.2002 -3 -4 -8 -42 76 -2 -3 -6 -8 76 -2 -3 -9 -7 -5 0 -3 -4 -6 -7 13 0 -3 -5 -7 W.L.2002 31 33 35 53 -31 30 30 32 34 -31 27 29 33 32 30 11 27 29 30 31 5 23 27 28 30 W.F.2003 9 6 96 96 96 12 9 -33 96 96 12 11 96 96 9 13 11 10 8 -33 18 12 10 9 8 W.L.2003 24 30 -42 -42 -42 22 24 30 -42 -42 17 21 -42 -42 24 15 18 20 24 31 13 16 18 20 24 W.F.2004 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA W.L.2004 -78 -78 -78 -78 -78 -28 -78 -78 -78 -78 -32 25 -78 -78 -78 -34 -32 -78 -78 -78 -36 -34 -32 -78 -78 W.F.2005 38 37 36 102 102 39 37 36 35 -2 39 38 15 35 37 39 38 37 36 34 40 39 38 37 35 W.L.2005 24 25 29 -43 -43 22 23 24 29 55 20 22 31 24 23 -7 20 21 22 24 -11 13 19 21 22 W.F.2006 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA W.L.2006 32 33 34 35 -46 14 31 32 33 35 12 29 33 32 31 11 12 29 30 31 10 11 12 28 29 W.F.2007 40 23 3 119 119 65 40 22 2 119 65 64 2 21 39 66 64 40 39 20 66 65 63 39 27 W.L.2007 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA W.F.2008 44 42 39 109 109 44 43 41 39 109 44 44 38 40 43 45 44 43 41 39 46 45 44 42 40 W.L.2008 NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA NA

Note: for the first column names, H.F. represents Harrisburg First emergence, H.L. Harrisburg Last emergence, Ph.F, Ph.L, Pi.F, Pi.L, W.F, W.L represent Philadelphia First, Philadelphia Last, Pittsburgh First, Pittsburgh Last, Williamsport First, and Williamsport Last, respectively. NA indicates no records in the field observation data. I choose the optimal parameter combination that minimizes the total absolute value across all columns and the result shows that EIP=109 and Longevity = 12 days combination has the minimum total error. Note the performance of this optimal combination is still very bad for Pennsylvania.

146

Chapter 5

Changing Daily Temperature Range, Degree Day Accumulation, and Disease

Transmission Period

ABSTRCT

I demonstrate that mean temperature does not provide enough information to compute degree day accumulations for predicting disease transmission periods, and show how a changing daily temperature range (DTR), which would change vector life history, influences disease dynamics. I quantify West Nile Virus transmission period by comprehensively exploring combinations of changing DTR and mean temperature in four locations in Pennsylvania: Harrisburg, Philadelphia, Pittsburgh, and Williamsport. I first examine how different DTR conditions influence degree day accumulation, and then investigate vector mosquito life history changes according to these different climate change conditions. The results show that even when the mean temperatures remain the same, a larger DTR results in more degree days and consequently a longer transmission period. Increasing mean temperature with increasing or decreasing DTR yields earlier first incidences and later last incidences, however the magnitude of these two are not the same. I also demonstrate that increasing DTR would impact disease transmission more dramatically in relatively cooler areas (Williamsport, Harrisburg, and Pittsburgh) than already warm areas (Philadelphia). Also total accumulated degree days during the entire 147 transmission period is nonlinearly related to DTR and highly regional specific. These results show how DTR significantly changes West Nile virus epidemics and confirms that

I need to consider the DTR in addition to mean temperature for modeling transmission periods.

148 Introduction

Temperature is a key factor in the rate of development of a variety of insects

(Logan et al. 1976, Wagner et al. 1984, Lowry and Lowry 1989). Recently, Lafferty ignited a heated debate (Lafferty, 2009a) about whether there is a simple causal relationship between climate change and insect life history and infectious diseases transmission. Lafferty (2009a) argued too many other socio-economic factors are intertwined. After Lafferty's paper, five other researchers gave their comments, either pro or con, to his idea (Harvell et al. 2009, Ostfeld et al. 2009, Pascual and Bouna 2009,

Randolph et al. 2009, Dobson et al. 2009). They used either empirical or theoretical methods to demonstrate how disease dynamics could be influenced under climate change conditions. As a summary of this debate, Lafferty (2009b) concludes that climate change may change disease transmission in some vector borne diseases.

However, unlike the consensus that mean temperature is rising under global climate change, it is still unclear how daily temperature range (DTR) will be affected – that is will DTR increase, remain unchanged or decrease (Letcher 2009) in the future.

Furthermore, it is not clear the degree to which DTR changes would be site-specific

(Bristow and Campbell 1984, Stanhill and Moreshet 1992, Stanhill and Cohen 1997,

Roderick and Farquhar 2002, Travis et al. 2002, Wild et al. 2007). Although it is unclear how DTR may change in the future, it is important to explore how changes in DTR may affect the outcome of disease dynamics and therefore assess potential disease burdens in the future.

Mosquitoes, because of their ability to transmit various infectious diseases, have been studied intensively to understand their life history and temperature interactions

149 (Rueda et al. 1990, Dohm et al. 2006). While most of the research focuses on the increasing and decreasing of mean temperature and long-term trend of climate change

(annual, multi-annual, and some periodical trends, Lafferty 2009a), daily temperature range (DTR) is also expected to change (Letcher 2009). Some recent papers (e.g. Gu and

Novak 2006, Paaijmans et al., 2009, 2010) have demonstrated that with changes in DTR, transmission patterns of disease can be altered significantly.

West Nile virus, a mosquito borne disease, (WNV, Flaviridae) is the main cause of arboviral encephalitis in the US and has become endemic throughout North America

(Gyure, 2009). WNV was accidentally introduced into America in 1999 in New York

City (CDC 2010) and within 4 years has subsequently spread across the continental U.S

(Peterson 2001, Nash et al. 2001, Guyre 2009). During 2002 and 2003, fatalities by WNV in humans were reaching close to 300 cases annually and infecting between 4,000 and close to 10,000 people. Since 2008 the annual number of infected cases ranges between

700 – 1000/year with between 32 to 45 deaths per year (CDC, 2010). Not only does

WNV affect humans but has also had important ecological impacts on avian populations such as, crows (Corvidae) and American robins (Turdus migratorius) populations (Komar et al. 2001, LaDeau et al. 2007, Rahbek 2007). Thus, viruses such as WNV are considered a threat to both public and ecological health.

The purpose of this study is to examine the effect of various changing DTR’s and mean temperature conditions on the transmission of WNV. To achieve this I will (1) first impose different DTR conditions to current climate data to see their effects on daily degree day accumulations relevant to predicting WNV transmission period (using the adjusted model in Chapter IV) and (2) examine different DTR conditions under different

150 climate change scenarios. These scenarios are divided into six major DTR/mean temperature changing categories with 81 subsequent cases in these six major categories. I will analyze these differences by examining how these changes will affect WNV transmission in four locations in Pennsylvania.

Material and Methods

Modeling Degree Day Accumulation in a Single Day

Mean temperature alone does not sufficiently capture the daily temperature profile in the field. Biometeorologists (Allen 1976) have proposed a sine curve model

(sinusoidal model) to simulate DTR profile, which has the basic form of:

T asin bi  c i

Four inputs are required in order to reconstruct a daily temperature profile: daily maximum temperature, and minimum temperature to determine coefficients a and c.

Coefficient b is determined by photoperiod, which itself is further considered as a function of latitude and number of the day in the year. Thus I will need daily maximum, minimum temperature, latitude, and number of day in the year as inputs. The accumulated degree day is basically the integration over a basic development threshold.

To reduce computational effort I interpolate hourly temperatures and use these temperatures to approximate degree day accumulation above threshold in a day. This is a direct analog of the Riemann Integral.

Forsythe et al. (1994) proposed the following equations to calculate daylength:

151  0.2163108  2arctan(0.9671396 tan(0.00860  (J  186)))  arcsin(0.39795cos  ) p L  sin sin sin 24 D 24  cos1 180 180 L  cos cos 180 where  is the revolution angle predicted by number of day in the year (J),  is the

declination angle, L is latitude, and D is the day length p is the sun position constant and

equals 0.8333 according to US government definition (Forsythe et al. 1994).

After I have calculated the day length, I could further determine temperature profile in a

day:

()t t   T( T  T )sin(r )  T ( t  t  t ) imax minD2( a   ) min n x 1 (())t t  t   1 1 r s T( T  T ')  ( T  T ')cos(2 ) ( t  t  t ) imax min max minD x n ' 2 2 24  a   2

where t is current time, D is day length, tr and ts are the sunrise and sunset time derived

from daylength, Tmin’ is minimal temperature of the next day, and a and  are

predetermined coefficients (Wann et al. 1985).

The sine-negative exponential model is a fine-tuning version of the original sine curve

model (reference?). Field observation data suggest that temperature profile is not

necessarily symmetric in a day and the temperature curve after sunset is better fitted with

a negative exponential curve. Thus the final model is formulated as:

 m Ti( Tmax  T min )sin( )  T min ( t n  t  t x ) D 2 a bn  Z Ti Tmin ( T s  T min ) e ( t x  t  t n ' )

152

where D is day length (determined by latitude and day in a year), Z is night length, Ts is

the temperature at the moment of sunset, m is number of hours after minimum temperature until sunset, n is number of hours after sunset until the time of minimum temperature of the next day, a is the lag coefficient for the maximum temperature, and b is the nighttime temperature coefficient (Parton and Logan 1981). Because the sine- negative exponential model is derived from a sine curve model, I will call it sine curve model in the following text for simplicity.

Modeling Effect of various DTR on Temperature Profile

Daily maximum and minimum temperatures were obtained from NOAA for

Harrisburg, Philadelphia, Pittsburgh, and Williamsport for 2008. This year was selected because it was the most recent complete field observation data on WNV incidence in mosquitoes provided to us by the Department of Environmental Protection for

Pennsylvania (PA-DEP). Latitudes of these locations were determined in order to reconstruct daily temperature profile and estimate degree day accumulation.

In Pennsylvania mosquitoes emerge during spring and the WNV transmission period is between June and September approximately (depending on locations), therefore for the purpose of this study I selected a range of dates that occurred within the WNV transmission season. Simulations were run using the sine curve model for days 100, 150,

200, and 250, from mid-April to mid-September, which bracket some important events for WNV transmission.

I use two parameters, k1 and k2, for the adjustment of current daily climate data:

153 TT T'() T  k min max min min 1 2

TTmin max Tmax'() T max  k2 2 Equation 1

TTn x (Note: I use 2 to approximate daily mean temperature, which is only applied to adjust daily maximum and minimum temperature, and computation of degree day accumulation is still based on sine curve methods.)

Using this method, the adjusted daily minimum temperature is the original daily minimum temperature minus a fraction of the daily mean temperature of that day, and adjusted daily maximum is the original maximum plus another fraction of the daily mean.

Currently the mean temperature in Pennsylvania is approximately 10 ºC and according to the 4th annual report of IPCC (IPCC 2007b) the mean temperature will be about 1.1-4 ºC

higher by the end of this century. While I do not know how DTR would change in the

future, I need to consider these conditions comprehensively by using simulation to

investigate their impacts. Because the change of mean temperature could be derived from

the change of DTR (but NOT vice versa!), I set the two adjustment parameters, k1 and k2, to change from -0.4 to 0.4, with 0.1 step size, so the largest mean temperature change could be around 4 ºC (Note: this is only for annual mean temperature, in each day the maximum/minimum/mean temperature varies significantly). This results in a total of 81 different climate change conditions (in fact, 80 new conditions because k1 = k2 = 0

represents the current climate condition, with no change in either the mean or DTR).

Basically these 81 combinations could be divided into six major categories:

decreased mean temperature with decreased DTR (k2<0 and abs(k1)

154 mean temperature with decreased DTR (k1<0 and abs(k1)>abs(k2)); decreased mean temperature with increased DTR (k1>0 and abs(k1)>abs(k2)); increased mean temperature with increased DTR (k2>0 and abs(k1)

DTR (k1 =k2); same DTR with variable mean temperature (k1 = -k2). Note: the terms increased and decreased are relative to current climate condition. All these categories are summarized in table 5-1. I also present an illustration of daily temperature profiles of these six major categories in appendix 2.

I emphasize how variable DTR conditions influence degree day accumulation in the same mean temperature conditions (k1 = k2) and summarize the results in table 5-2 through 5-5. I also present degree day accumulation of all 81 different combinations in 3-

D plots, with the x-axis and y-axis representing k values and the z-axis degree days, for

Philadelphia and Pittsburgh in day 100 and 200 in a year. These two locations have different climate conditions (Philadelphia is about 2°C higher than other three locations), and I choose day 100 and 200 to emphasize different degree day accumulation on different daily temperature profiles.

WNV Transmission Period under Climate Change Conditions

Based on these modifications of current climate data I run a degree-day based simulation (the modified model described in chapter IV in this thesis, also see Zou et al.

2007, and Konrad et al. 2009,) to determine dates of first and last incidence of positive

West Nile virus in vector mosquitoes. I report both first and last incidence dates, transmission period, change of these dates and period relative to current climate condition,

155 and total accumulated degree days above threshold during the entire transmission period, which are presented in table 5-6 through 5-9.

I also demonstrate how different DTR conditions could influence transmission period and emergence dates when the mean temperature remains the same in category I. I plot first and last incidence dates, transmission period, and area above threshold against various k values in figure 5. For category II to V I choose the k1 (or k2 ) values that have the most k2 (or k1 ) combinations. For instance in category II I choose k2 = 0.4 so that k1 varies from -0.3 to 0.3 in this category and compare the first incidence dates, transmission period, and area above threshold. These results are shown in figure 5-10 to

5-12.

I show how various DTR and mean temperature conditions (81 k1, k2 combinations) influence the change of first and last incidence dates and transmission period relative to current climate condition in appendix figure 1 to12. To explore temporal and spatial differences I use the same mean with various DTR conditions as examples to illustrate why advancing time of first incidence is always smaller than that of last incidence and how different locations respond to climate change. I select Harrisburg and Philadelphia and show our findings in figure 10 and 11. An additional two locations are shown in appendix figure 5-13 and 5-14.

Results

Effect of various DTR on Temperature Profile

I report the daily temperature profile in day 100, 150, 200, and 250 under the same mean temperature conditions (k1 = k2) for the four locations (table 5-2 through 5-5

156 for Harrisburg, Philadelphia, Pittsburgh, and Williamsport, respectively) to demonstrate

how DTR influences degree day accumulation when the mean temperature is the same.

All the 81 combinations are reported in figure 5-1 to figure 5-4.

From the following tables it is clearly seen that across all four locations, larger

DTRs result in more degree day accumulations on any given dates. These results provide

sound evidence that using mean temperature alone to calculate degree day (or heat unit)

accumulation is insufficient, and sometimes very unreliable. For example, on day 150 in

Harrisburg, the largest DTR condition (k = 0.4) yields twice as many degree days as in

the smallest condition (k = -0.4). Considering our current WNV modeling approach that

computes the degree day accumulation during the adult mosquito’s life span (10-12 days),

the daily errors will be inevitably accumulated and magnified. Consequently, it is

dangerous to use mean temperature alone to model insect phenology and vector-borne

diseases dynamics in research.

In addition to the same mean temperature conditions, I also present other

conditions in figure 1to16 in the appendix. In each major category (category II- VI), I still

see a consistent pattern that larger DTR would yield more degree day accumulation. For

example, in category III when k1 = 0.4 and k2 varies from -0.3 to 0.3, they are all in the

decreased mean with increased DTR category (relative to current condition). From the

figure I could see larger DTR (more positive k2) yields more degree days than smaller

DTR (negative k2). However this conclusion is not as persuasive as in the same mean temperature category, because the mean temperatures in the larger DTR conditions are higher than those in smaller DTR conditions. I could observe this general pattern from

157 figure 5-1 through figure 5-16. (Note: again the term decreased mean temperature is

relative to the current climate condition where k = 0.)

Effect of various DTR on disease transmission

I summarize first and last incidence time, transmission period, and their relative changes to current condition in tables 5-6 to table 5-9 for the four locations. I first demonstrate how DTR influence WNV dynamics in the same mean condition (category I).

I plot change of first incidence, last incidence, transmission period, and area above the threshold in figure 5-5. The figure clearly shows that larger DTR (larger k values) will result in earlier first incidence and later last incidence. Nevertheless the magnitudes of change in first and last incidences are not identical. In Philadelphia and Pittsburgh, when

DTR is large, the change of the last incidence exceeds the change of first incidence. In

Williamsport and Harrisburg the difference is not that substantial. The other finding is the area above the threshold grows nonlinearly (perhaps exponentially) when DTR increases.

In category II (increased daily mean temperature with increased DTR) I observe the trend that instead of larger DTR, (relatively) smaller DTR variability results in longer transmission periods. For instance, if k2 = 0.4 then k1 could vary from -0.3 to 0.3 and

satisfy the requirement (abs(k1)

Harrisburg as an example, the transmission period changes most dramatically when k1 = -

0.3 (which is the smallest DTR in this category, approximately 4°C increase in daily

mean temperature), with 30 days longer in transmission period, which represents a

substantial 60% increase in that period compared to current conditions. All other

locations show similar pattern, with 53, 43, and 45 days longer in transmission period, in

158

Philadelphia, Pittsburgh, and Williamsport, respectively. When k2 is fixed, smaller (more negative) k1 value relates to higher daily minimum temperatures (because the adjusted

daily minimum temperature is the original daily minimum temperature minus a

proportion, represented by k1, of daily mean temperature. See equation 1 in material and

method) and hence higher daily mean temperature (because k2 is fixed, daily maximum

temperature does not change. Daily mean temperature rises because the minimum

temperature rises.), but with smaller DTR variability. Nevertheless, note that all

conditions in this category have an increased DTR relative to the current condition, where

k1 = k2 = 0.

I observe that even in category III (the decreased mean with increased DTR

category) the transmission period is still longer than the current condition in some large

DTR conditions. For instance when k1 = 0.4, k2 = 0.3 (about -4°C decrease in daily minimum and 3°C increase in daily maximum, which results in an approximately 1°C decrease in daily mean temperature), Harrisburg, Pittsburgh, and Williamsport experience

3 days , 2 days, and 11 days longer transmission periods, respectively (figure 5-7). Again from the degree day accumulation result I know larger DTR would compensate for the decreased mean temperature by having more effective heat unit accumulation above the development threshold. Though it’s not always true, considering I are accumulating degree days in a 10 or 12 days period, it is almost certain that the total accumulated degree days in the decreased mean with large DTR conditions could exceed original temperature condition, though the difference is relatively small.

159

In category IV (increased mean with decreased DTR) the trend still remains. If k1

is fixed at -0.4 and k2 varies from -0.3 to 0.3, I see longer transmission periods with larger

DTR variation (larger k2, see figure 5-8 When k2 = -0.2 (approximately 1°C increase in

daily mean temperature), Philadelphia experiences less than 10 days increase in

transmission period, whereas Williamsport has a significant increase of about 40 days.

The other two locations have about 20 days increase of transmission period. When k2 = 0

(approximately 2°C increase in daily mean temperature), Philadelphia experiences less

than 20 days increase in transmission period, whereas other three locations have more

than 40 days increase of transmission period. Under the extreme condition, where k2 = 0.3

(approximately 3.5-4°C increase in daily mean temperature) all of these locations would experience more than 40 days longer transmission period.

For the other two major categories, category V (decreased mean with decreased

DTR) is very unlikely to happen in the future; and category VI is irrelevant to DTR change. Nevertheless in category V I still observe the trend that larger DTR would result in earlier incidence dates of first emergence and later last emergence, but only for

Philadelphia because in this category the other locations do not have enough degree day accumulation to enable WNV transmission (figure 5-9).

Sometimes different conditions in different categories would result in similar increased or decreased mean temperature. For instance, if the mean temperature is about to increase by 2°C there are at least five k1 and k2 combinations to achieve this

increment: k1 varies from -0.4 to 0 and k2 always 0.4 larger than k1, which will all have the

same mean with 2°C increase from the original mean temperature. I don’t observe much

160 difference in terms of first and last incidence dates and transmission period in these five conditions in all four locations. For instance, the changes of transmission period in

Harrisburg are 54, 54, 55, 55, and 55 days under these five conditions. Nevertheless this does not mean DTR is not important, but rather the effect of increasing minimum and maximum should be equivalent. This indicates I should pay the same attention to rising minimum temperatures in addition to rising maximum temperatures.

In appendix figure 17 to 28 I plot the relative change of first incidence, last incidence, and transmission period in all 81 combinations. I observe very strong asymmetry: most of the points are above the zero-plane, which indicates that the duration of the WNV transmission period becomes longer in most of the cases. However I am not too surprised with this outcome. From the previous discussion I have already seen that larger DTR would result in larger degree day accumulation, and since I am accumulating over 10 or 12 days (the longevity time for adult mosquitoes), even subtle differences in

DTR cause a noticeable difference in the transmission period. Also in these figures I observe consistent and substantial differences in the advancement of first incidence compared to delay of last incidence time of WNV in all four locations. For instance, in many cases the advancing time in Philadelphia is around 5-10 days, whereas the delay time is around 20-40 days. This coincides well with our previous results in figure 5 that change in the last incidence is more dramatic than the change in the first incidence.

Besides temporal differences at the beginning and end of transmission period, I also observe regional differences. In relatively cooler areas, for example, Williamsport, the predicted transmission period is 51 days currently. Under large DTR variation and large increased mean temperature conditions, the transmission period is 96 days, a 45

161 days increase, which is almost a 90% increase in the transmission period. This pattern is also true for Harrisburg and Pittsburgh where temperature is significantly lower than

Philadelphia (about 2°C in annual mean temperature). On the other hand, though

Philadelphia experiences a similar 50 days transmission period increase, it is a relatively smaller increase when compared to its prediction of a 90 days transmission period under current climate condition, a 55% increase of the transmission period. Hence I suggest that if I focus on the transmission period as an indicator of WNV disease, in the future increased mean with increased/decreased DTR conditions, relatively cooler areas should prepare more for the elongated transmission period.

162

Discussion

From the results I discover that increasing DTR would influence disease transmission significantly. I have also noticed asymmetric changes in first compared to last emergence dates in all locations. The first emergence dates, even under the most dramatic climate variability condition (k1 = -0.4, k2 = 0.4), only advance at most 7-10 days.

On the other hand, last emergence dates differ significantly across different DTR conditions. Under the extreme condition I simulated (k1 = -0.4, k2 = 0.4) the last emergence dates could be almost 2 months later in all locations. This asymmetry could be due two factors: first, the degree day accumulation curve is steeper at the beginning of the season than at the end of the season (figure 10 is the most significant one). Therefore, it’s relatively easier for the curve to cross the threshold line at the end of the season than in the beginning. However Harrisburg is somehow different and it seems the steepness is similar in both ends. Besides the steepness, I also argue that the relatively higher threshold at the beginning of the season (129 DD vs 109 DD, or 159 DD vs 109 DD) further prevent the first incidence change to be too dramatic. It is much easier for the curve to hit the threshold line of 109 DD than both 159 DD in Philadelphia and 129 DD in other locations at the end of the season.

The model results reveal an important outcome of changing DTR. I have already known that the disease transmission period is going to be larger under climate change conditions. But most of the previous research has focused on the advancing of first incidence of disease. Our finding, on the other hand, suggests the delay of last incidence is more substantial than the advance of first incidence. I have also discovered a similar

163 pattern in other insect phenology system (see chapter III). Hence I need to pay attention to the delay of the last disease incidence as much as, if not more important than, the advance of first incidence.

Different locations also response to changing DTR and changing mean temperature conditions differently. I have already demonstrated that in the warmer areas such as Philadelphia, the relative change of transmission period is not as large as in those cooler areas. In fact, Philadelphia has regular and frequent pesticide spray applications which would suppress mosquito population and hence decrease WNV risk. So management efforts may be better prepared to minimize effects from the predicted longer transmission period on Philadelphia. But on the other side, the cooler areas might experience relatively much longer transmission periods, and if the ecosystem or the human health department are not forewarned and prepared, the virus might cause more severe disruptions.

Although in this research I focus on the phenology of the vector mosquitoes, I also notice that another variable, the area above the threshold, is highly related to climate conditions. That area, in a more accurate sense, represents total degree days accumulation in the transmission period above the transmission threshold. Even when the transmission period is very similar, the area might differ a lot. For instance, in Harrisburg when k1 = 0 and k2 = 0.1, the transmission period is only 4 days longer, however the area is 434 degree days, much larger than that in the original condition where the area is merely 16. I could also see very clearly that in Philadelphia this area is much larger than in other locations (table 5-5 to 5-8, figure 5-10 and 5-11). Currently Philadelphia has 89 days transmission period and approximately 2100 degree day accumulated in this period. But

164 in the extreme conditions (k1 = -0.4 and k2 = 0.4) it could accumulate more than 14,000 degree days in 144 days transmission period. That is, a 7 fold increase in the total degree day accumulation but only 56% increases in transmission period. It is even more dramatic in Harrisburg, where in the extreme conditions the area is almost 400 times more. This nonlinear relationship between degree-day accumulation and transmission period could also be observed in figure 5-17. When the mean temperature remains the same, large

DTR could result in exponential growth of the degree day accumulation in the transmission period. I believe such a big change must be related to disease dynamics as well. Though not explicitly and quantitatively explored, I suggest the total degree day accumulation in the transmission period might impact population growth and population size, since in our model I do not consider population dynamics and only focus on the transmission period. I also suggest that the large area above threshold could result in an explosion of the mosquito population, hence impose more serious burden to disease control. I would like to further study this problem in future research and link it to population dynamics.

Our research not only comprehensively and quantitatively explores how different

DTR conditions would influence WNV transmission period in Pennsylvania, but also sets up a framework to demonstrate the importance of DTR from the very basic application of degree days accumulation, which is frequently overlooked. Our WNV model is a relatively simple system where the transmission period is said to be related to degree days accumulated in the adult mosquito lifespan and EIP. Nevertheless our modeling procedure could be easily transferred to other disease or ecological systems which are more complicated yet share similar research question. Since global climate change is a

165 hot spot in scientific research now, I strongly argue that I need to consider the influence of DTR and not to use mean temperature alone. Using mean temperature alone might yield unreliable and skewed results.

Nevertheless I admit our modeling approach is not perfect. First, I am using one-year climate data (2008 daily data) and project into the future. The modified data in the remaining 80 combinations are more or less determined by this specific dataset. I choose year 2008 because in this year the original WNV incidence prediction model yields the best results among 7 years of field observations (thesis chapter IV). But I do not know exactly what climate conditions will exist in the future and using one year data to do the extrapolation seems to be risky. One possible improvement could be using multiple-year- averaged climate data as the basis. But this approach bears the criticism that I do not know if the averaged data would generate reliable incidence dates in the basic condition

(k = 0). As seen in chapter IV, the WNV incidence date prediction model performance does vary in different years. However our main objective of this research is to demonstrate the importance of DTR, this should not be a serious issue from our perspective.

Another flaw in our modeling approach lies in the assumption that the DTR changing rate is the same across the year. I do not know exactly how this rate changes in the year and this assumption simplifies our modeling procedure. But in reality it might change from day to day, or season to season. Currently I believe this assumption is generally reasonable and based on our research objective it is adequate for now. I have already demonstrated the importance of finer scale climate change, the DTR. Hence I are

166 still relying on finer scale prediction of future climate and weather conditions and looking for collaboration with meteorologists to deal with this situation.

In our research I assume all mosquitoes have identical longevity (at the same time)

and minimal development threshold. It is a simplification from a population perspective

and I ignore individual difference in our model. But heterogeneity and mutation occur

all the time in the population. All these factors would eventually lead to evolution of

both the virus and the vector. Scientists are facing more variability than we have

expected. Besides, not only is the adult mosquito stage influenced by ambient climate,

other life stages such as larva, egg, and pupa are also temperature dependent (Gong et al.

2010). Moreover, large DTR might influence vector mortality as well (Paaijmans et al.

2009). Extremely cold or hot weather, if beyond the insect’s tolerance threshold, would

inevitably be unfavorable for the insects. I have already known that for most insects,

mortality rate is correlated to temperature. Thus a more comprehensive multi-stage

population based model is more accurate to predict disease transmission.

In summary, our research investigates some fundamental aspects of insect

phenology and disease transmission and demonstrates that DTR substantially influences

relevant degree day accumulation, and consequently alters disease emergence dates and

transmission period. Our research proposes a finer scale and systematic framework for

future research of climate change and ecological/epidemiological responses.

167

References Cited

Allen, J. C. 1976. A modified sine wave method for calculating degree days. Environ.

Entomol. 5: 38~396.

Bolling, B. G., C. M. Barker, C. G. Moore, W. J. Pape and L. Eisen. 2009.

Modeling/GIS, risk assessment, economic impact: seasonal patterns for

entomological measures of risk for exposure to Culex vectors and West Nile virus

in relation to human disease cases in Northeastern Colorado. J. Med. Entomol. 46:

1519-1531.

CDC website of West Nile virus in the U.S.. 2010.

http://www.cdc.gov/NCIDOD/dvbid/westnile/.

Dobson, A. 2009. Climate variability, global change, immunity, and the dynamics of

infectious diseases. Ecology 90: 920–927.

Dohm, D. J., M. L. O’Guinn and M. J. Turell. 2006. Effect of environmental

temperaeture on the ability of Culex pipiens (Diptera: Culicidae) to transmit West

Nile virus. J. Med. Entomol. 39: 221-225.

Dohm, D. J., M. R. Sardelis and M. J. Turell. 2006. Experimental vertical trnamission

of West Nile virus by Culex pipiens (Diptera: Culicidae). J. Med. Entomol. 39:

640-644.

168 Forsythe, W. C., E. J. Rykiel, R. S. Stahl, H. Wu, and R. M. Schoolfield. 1995. A

model comparison for daylength as a function of latitude and day of a year.

Ecological Modelling 80: 87-95.

Gong, H, A. T. DeGaetano, and L. C. Harrington. 2010. Climate-based models for

West Nile Culex mosquito vectors in the Northeastern US. Int. J. of Biometeorol.

DOI: 10.1007/s00484-010-0354-9.

Gyure, K. A.. 2009. West Nile virus infections. J. Neuropatho. Exp. Neuro. 68: 1053-

1060.

Gu, W. D., and R. J. Novak. 2007. Stochastic estimation of degree days of mosquito

development under fluctuating temperatures in the field. Journal of Vector

Ecology 31: 107-112.

Harvell, Drew, S. Altizer, I. M. Cattadori, L. Harrington, and E. Weil. 2009. Climate

change and wildlife diseases: when does the host matter the most? Ecology 90:

912–920.

Juricic, D., B. E. Eno, and G. Parikh. 1974. Mathematical modeling of a virus vector:

Culex tarsalis. Biomed. Sci. Instrum. 10: 23-28.

Kilpatrick, A. M., M. A. Meola, R. M. Moudy and L. D. Kramer. 2008. Temperature,

viral genetics, and the transmission of West Nile virus by Culex pipiens

mosquitoes. PLoS Pathogen 4: e1000092.

169 Komar, N., N. A. Panella, J. E. Burns, S. W. Dusza, T. M. Mascarenhas, and T. O.

Talbot. 2001. Serologic evidence for West Nile virus infection in birds in the

New York City vicinity during an outbreak in 1999. Emerging Infectious Diseases

7: 621-625.

Konrad, S. K., S. N. Miller, W. K. Reeves, and N. S. Tietze. 2009. Spatially explicit

West Nile virus risk modeling in Santa Clara County, California. Vector Borne

Zoonotic Diseases 9: 267-274.

LaDeau, S. L, A. M. Kilpatrick, and P. P. Marra. 2007. West Nile virus emergence

and large-scale declines of North American bird populations. Nature 447: 710-

713.

Lafferty, K. D. 2009. The ecology of climate change and infectious diseases. Ecology

90: 888-900.

Lafferty, K. D. 2009. Calling for an ecological approach to studying climate change and infectious diseases. Ecology 90(4): 932–933.

Moudy, R. M., M. A. Meola, L. L. Morin, G. D. Ebel, and L. D. Kramer. 2007. A

newly emergent genotype of West Nile virus is transmitted earlier and more

efficiently by Culex mosquitoes. Am. J. Trop. Med. Hyg. 77: 365-370.

Nash, D., F. Mostashari, J. Miller, D. O’Leary, K. Murray, A. Huang, A. Rosenberg,

A. Greenberg, M. Sherman, S. Zaki, P. Smith, and M. Layton. 2001. The

outbreak of West Nile virus infection in the New York City area in 1999. The

New England Journal of Medicine 344: 1807-1814.

170 Oda, T., Y. Eshita, K. Uchida, M. Mine, K. Kurokawa, Y. Ogawa, K. Kato, and H.

Tahara. 2002. Reproductive activity and survival of Culex pipiens pallens and

Culex quinquefasciatus (Diptera: Culicidae) in Japan at high temperature. J. Med.

Entomol. 39: 185-190.

Ostfeld, R. S. 2009. Climate change and the distribution and intensity of infectious

diseases. Ecology 90: 903–905.

Paaijmans, K. P., A. F. Read, and M. B. Thomas. 2009.Understanding the link

between malaria risk and climate. PNAS 106: 13844-13849.

Paaijmans, K. P., S. Blanford, A. S. Bell, J. I. Blanford, A. F. Read, and M. B.

Thomas. 2010.Influence of climate on malaria transmission depends on daily

temperature variation. PNAS 107: 15135-15139.

Pascual, M, and M. J. Bouma. 2009. Do rising temperatures matter?. Ecology 90(4):

906–912.

Parton, W. J. and J. A. Logan. 1981. A model for diurnal variation in soil and air

temperature. Agricultural Meteorology 23: 205-216.

Peterson, L.R. 2001. West Nile virus: a reemerging global pathogen. Rev. Biomed. 12:

208-216.

Rahbek, C. 2007. Disease ecology: the silence of the robins. Nature 447: 652-653.

171 Randolph, S. E. 2009. Perspectives on climate change impacts on infectious diseases,

Ecology 90(4): 927–931.

Reisen, W. K., Y. Fang, and V. M. Martinez. 2006. Effects of temperature on the

transmission of West Nile virus by Culex tarsalis (Diptera: Culicidae). J. Med.

Entomol. 43: 309-317.

Rueda, L. M., K. J. Patel, R. C. Axtell, and R. E. Stinner. 1990. Temperature

dependent development and survival rates of Culex Quinquefasciatus and Aedes

aegypti. J. Med. Entomol. 27: 892-898.

Tachiiri, K., B. Klinkenberg, S. Mak and J. Kazmi. 2006. Predicting outbreaks: a

spatial risk assessment of West Nile virus in British Columbia. International

Journal of Health Geographics 5: 1-21.

Wann, M., D. Yen and H. J. Gold. 1985. Evaluation and calibration of three models for

daily cycle of air temperature. Agricultural and Forest Meteorology 34: 121-128.

Zou, L., S. N. Miller and E. T. Schmidtmann. 2007. A GIS tool to estimate West Nile

virus risk based on a degree-day model. Environ. Monit. Assess 129: 413-420.

172

Table 5-1. Six Major Categories of Changing Daily Temperature Range

Category Description Mean DTR k1 k2 abs(k1)-abs(k2) Max increasing (decreasing) at same rate as min

I is decreasing (increasing) No Change Variable k1 = k2 Max increasing- faster than min is increasing, or II min is decreasing Increasing Increasing Positive Negative Min decreasing-faster than max is increasing, or III max is decreasing Decreasing Increasing Positive Positive Min increasing- faster than max is increasing, or IV max is decreasing Increasing Decreasing Negative Positive Max decreasing-faster than min is decreasing, or V min is increasing Decreasing Decreasing Negative Negative Max increasing (decreasing) at same rate as min

VI is increasing (decreasing) Variable No Change k1 = -k2 173

Table 5-2. Degree Day Accumulation in Harrisburg

Day Day Day Day k1,k2 100 150 150 200 -0.4 0.08 3.2 8.36 8.19 -0.3 0.41 3.31 8.5 8.22 -0.2 0.78 3.43 8.65 8.25 -0.1 1.18 3.75 8.79 8.28 0 1.57 4.24 8.93 8.31 0.1 1.96 4.76 9.12 8.38 0.2 2.35 5.31 9.45 8.71 0.3 2.74 5.88 9.98 9.2 0.4 3.14 6.46 10.6 9.76

174 Table 5-3. Degree Day Accumulation in Philadelphia

Day Day Day Day k1,k2 100 150 150 200 -0.4 0 7.05 15.56 13.94 -0.3 0 7.19 15.75 13.98 -0.2 0 7.33 15.93 14.02 -0.1 0.01 7.46 16.12 14.06 0 0.19 7.64 16.31 14.09 0.1 0.42 7.98 16.49 14.13 0.2 0.7 8.51 16.68 14.17 0.3 0.98 9.11 16.87 14.2 0.4 1.27 9.74 17.11 14.24

175 Table 5-4. Degree Day Accumulation in Pittsburgh

Day Day Day Day k1,k2 100 150 150 200 -0.4 0 0.1 11.07 11.47 -0.3 0 0.46 11.23 11.5 -0.2 0 0.86 11.39 11.54 -0.1 0 1.27 11.55 11.57 0 0 1.69 11.71 11.61 0.1 0 2.12 11.87 11.64 0.2 0.02 2.54 12.07 11.68 0.3 0.16 2.96 12.41 11.71 0.4 0.35 3.39 12.94 11.93

176 Table 5-5. Degree Day Accumulation in Williamsport

Day Day Day Day k1,k2 100 150 150 200 -0.4 0 4.25 7.56 8.21 -0.3 0 4.42 7.7 8.24 -0.2 0.04 4.83 7.84 8.28 -0.1 0.2 5.35 7.98 8.31 0 0.42 5.91 8.13 8.35 0.1 0.65 6.47 8.37 8.62 0.2 0.91 7.07 8.82 9.08 0.3 1.17 7.68 9.41 9.63 0.4 1.44 8.29 10.03 10.2

177 Table 5-6. Incidence Time and Transmission Period Change in Harrisburg

k1 k2 First Last T.Period F. Change L. Change P. Change Area -0.4 -0.4 NA NA NA NA NA NA NA -0.4 -0.3 205 261 56 2 2 4 209 -0.4 -0.2 204 261 57 3 2 5 771 -0.4 -0.1 202 262 60 5 3 8 1460 -0.4 0 199 270 71 8 11 19 2236 -0.4 0.1 198 276 78 9 17 26 3178 -0.4 0.2 197 278 81 10 19 29 4240 -0.4 0.3 197 279 82 10 20 30 5363 -0.4 0.4 196 280 84 11 21 32 6545 -0.3 -0.4 NA NA NA NA NA NA NA -0.3 -0.3 NA NA NA NA NA NA NA -0.3 -0.2 205 261 56 2 2 4 245 -0.3 -0.1 203 261 58 4 2 6 816 -0.3 0 202 262 60 5 3 8 1515 -0.3 0.1 199 270 71 8 11 19 2293 -0.3 0.2 198 276 78 9 17 26 3251 -0.3 0.3 197 277 80 10 18 28 4317 -0.3 0.4 197 279 82 10 20 30 5446 -0.2 -0.4 NA NA NA NA NA NA NA -0.2 -0.3 NA NA NA NA NA NA NA -0.2 -0.2 225 225 0 -18 -34 -52 2 -0.2 -0.1 205 261 56 2 2 4 282 -0.2 0 203 261 58 4 2 6 862 -0.2 0.1 202 262 60 5 3 8 1570 -0.2 0.2 198 270 72 9 11 20 2356 -0.2 0.3 198 276 78 9 17 26 3328 -0.2 0.4 197 278 81 10 19 29 4399 -0.1 -0.4 NA NA NA NA NA NA NA -0.1 -0.3 NA NA NA NA NA NA NA -0.1 -0.2 NA NA NA NA NA NA NA -0.1 -0.1 225 225 0 -18 -34 -52 3 -0.1 0 205 261 56 2 2 4 325 -0.1 0.1 203 261 58 4 2 6 914 -0.1 0.2 201 262 61 6 3 9 1629 -0.1 0.3 198 270 72 9 11 20 2433 -0.1 0.4 198 276 78 9 17 26 3416 0 -0.4 NA NA NA NA NA NA NA 0 -0.3 NA NA NA NA NA NA NA 0 -0.2 NA NA NA NA NA NA NA 0 -0.1 NA NA NA NA NA NA NA 0 0 207 259 52 0 0 0 16 0 0.1 205 261 56 2 2 4 434 0 0.2 203 261 58 4 2 6 1057 0 0.3 199 269 70 8 10 18 1788

178 0 0.4 198 270 72 9 11 20 2644 0.1 -0.4 NA NA NA NA NA NA NA 0.1 -0.3 NA NA NA NA NA NA NA 0.1 -0.2 NA NA NA NA NA NA NA 0.1 -0.1 NA NA NA NA NA NA NA 0.1 0 NA NA NA NA NA NA NA 0.1 0.1 206 260 54 1 1 2 41 0.1 0.2 205 261 56 2 2 4 524 0.1 0.3 203 261 58 4 2 6 1171 0.1 0.4 199 269 70 8 10 18 1909 0.2 -0.4 NA NA NA NA NA NA NA 0.2 -0.3 NA NA NA NA NA NA NA 0.2 -0.2 NA NA NA NA NA NA NA 0.2 -0.1 NA NA NA NA NA NA NA 0.2 0 NA NA NA NA NA NA NA 0.2 0.1 NA NA NA NA NA NA NA 0.2 0.2 205 261 56 2 2 4 135 0.2 0.3 203 261 58 4 2 6 676 0.2 0.4 202 262 60 5 3 8 1351 0.3 -0.4 NA NA NA NA NA NA NA 0.3 -0.3 NA NA NA NA NA NA NA 0.3 -0.2 NA NA NA NA NA NA NA 0.3 -0.1 NA NA NA NA NA NA NA 0.3 0 NA NA NA NA NA NA NA 0.3 0.1 NA NA NA NA NA NA NA 0.3 0.2 225 259 34 -18 0 -18 7 0.3 0.3 205 261 56 2 2 4 322 0.3 0.4 183 261 78 24 2 26 911 0.4 -0.4 NA NA NA NA NA NA NA 0.4 -0.3 NA NA NA NA NA NA NA 0.4 -0.2 NA NA NA NA NA NA NA 0.4 -0.1 NA NA NA NA NA NA NA 0.4 0 NA NA NA NA NA NA NA 0.4 0.1 NA NA NA NA NA NA NA 0.4 0.2 NA NA NA NA NA NA NA 0.4 0.3 206 261 55 1 2 3 97 0.4 0.4 205 261 56 2 2 4 596

Note: T. Period=Transmission Period (in days). F.Change, L.Change, P.Change represent first incidence date change (relative to current condition k1=k2=0, in days), last incidence date change, and transmission period change, respectively. Area indicates total degree day accumulation during transmission period. This note also applies for table 7-10.

179 Table 5-7. Incidence Time and Transmission Period Change in Philadelphia k1 k2 First Last T.Period F. Change L. Change P. Change Area -0.4 -0.4 168 254 86 -3 0 -3 1783 -0.4 -0.3 165 262 97 0 8 8 2749 -0.4 -0.2 162 270 108 3 16 19 3870 -0.4 -0.1 161 276 115 4 22 26 5138 -0.4 0 160 292 132 5 38 43 6808 -0.4 0.1 156 295 139 9 41 50 8622 -0.4 0.2 155 296 141 10 42 52 10534 -0.4 0.3 154 296 142 11 42 53 12505 -0.4 0.4 153 297 144 12 43 55 14504 -0.3 -0.4 161 253 92 4 -1 3 1123 -0.3 -0.3 168 254 86 -3 0 -3 1857 -0.3 -0.2 164 262 98 1 8 9 2841 -0.3 -0.1 162 270 108 3 16 19 3970 -0.3 0 161 276 115 4 22 26 5268 -0.3 0.1 160 291 131 5 37 42 6953 -0.3 0.2 156 295 139 9 41 50 8774 -0.3 0.3 155 296 141 10 42 52 10686 -0.3 0.4 154 296 142 11 42 53 12658 -0.2 -0.4 162 249 87 3 -5 -2 552 -0.2 -0.3 160 253 93 5 -1 4 1178 -0.2 -0.2 167 254 87 -2 0 -2 1935 -0.2 -0.1 164 262 98 1 8 9 2935 -0.2 0 162 270 108 3 16 19 4072 -0.2 0.1 161 276 115 4 22 26 5407 -0.2 0.2 160 291 131 5 37 42 7099 -0.2 0.3 156 295 139 9 41 50 8926 -0.2 0.4 154 296 142 11 42 53 10841 -0.1 -0.4 163 224 61 2 -30 -28 230 -0.1 -0.3 161 249 88 4 -5 -1 596 -0.1 -0.2 160 253 93 5 -1 4 1233 -0.1 -0.1 167 254 87 -2 0 -2 2015 -0.1 0 163 262 99 2 8 10 3031 -0.1 0.1 162 270 108 3 16 19 4173 -0.1 0.2 161 276 115 4 22 26 5552 -0.1 0.3 160 292 132 5 38 43 7248 -0.1 0.4 156 295 139 9 41 50 9084 0 -0.4 165 220 55 0 -34 -34 32 0 -0.3 163 224 61 2 -30 -28 257 0 -0.2 161 249 88 4 -5 -1 642 0 -0.1 160 253 93 5 -1 4 1287 0 0 165 254 89 0 0 0 2099 0 0.1 163 262 99 2 8 10 3128 0 0.2 162 270 108 3 16 19 4275 0 0.3 161 276 115 4 22 26 5702 0 0.4 160 292 132 5 38 43 7405

180 0.1 -0.4 NA NA NA NA NA NA NA 0.1 -0.3 165 221 56 0 -33 -33 46 0.1 -0.2 162 224 62 3 -30 -27 283 0.1 -0.1 161 249 88 4 -5 -1 689 0.1 0 157 253 96 8 -1 7 1342 0.1 0.1 165 254 89 0 0 0 2187 0.1 0.2 163 270 107 2 16 18 3226 0.1 0.3 162 271 109 3 17 20 4380 0.1 0.4 161 276 115 4 22 26 5866 0.2 -0.4 NA NA NA NA NA NA NA 0.2 -0.3 NA NA NA NA NA NA NA 0.2 -0.2 165 222 57 0 -32 -32 65 0.2 -0.1 162 225 63 3 -29 -26 310 0.2 0 161 249 88 4 -5 -1 747 0.2 0.1 169 253 84 -4 -1 -5 1405 0.2 0.2 165 254 89 0 0 0 2281 0.2 0.3 163 270 107 2 16 18 3332 0.2 0.4 161 271 110 4 17 21 4502 0.3 -0.4 NA NA NA NA NA NA NA 0.3 -0.3 NA NA NA NA NA NA NA 0.3 -0.2 NA NA NA NA NA NA NA 0.3 -0.1 164 222 58 1 -32 -31 93 0.3 0 162 240 78 3 -14 -11 349 0.3 0.1 160 249 89 5 -5 0 841 0.3 0.2 168 253 85 -3 -1 -4 1508 0.3 0.3 165 254 89 0 0 0 2411 0.3 0.4 162 270 108 3 16 19 3477 0.4 -0.4 NA NA NA NA NA NA NA 0.4 -0.3 NA NA NA NA NA NA NA 0.4 -0.2 NA NA NA NA NA NA NA 0.4 -0.1 167 211 44 -2 -43 -45 2 0.4 0 163 223 60 2 -31 -29 148 0.4 0.1 161 243 82 4 -11 -7 453 0.4 0.2 156 250 94 9 -4 5 1006 0.4 0.3 168 253 85 -3 -1 -4 1698 0.4 0.4 164 270 106 1 16 17 2632

181 Table 5-8. Incidence Time and Transmission Period Change in Pittsburgh k1 k2 First Last T.Period F. Change L. Change P. Change Area -0.4 -0.4 215 262 47 -4 -6 -10 185 -0.4 -0.3 210 268 58 1 0 1 811 -0.4 -0.2 201 270 69 10 2 12 1680 -0.4 -0.1 201 278 77 10 10 20 2534 -0.4 0 198 279 81 13 11 24 3453 -0.4 0.1 196 281 85 15 13 28 4449 -0.4 0.2 195 282 87 16 14 30 5491 -0.4 0.3 194 294 100 17 26 43 6594 -0.4 0.4 194 295 101 17 27 44 7816 -0.3 -0.4 225 260 35 -14 -8 -22 18 -0.3 -0.3 215 263 48 -4 -5 -9 235 -0.3 -0.2 205 268 63 6 0 6 934 -0.3 -0.1 200 270 70 11 2 13 1736 -0.3 0 200 278 78 11 10 21 2591 -0.3 0.1 197 279 82 14 11 25 3517 -0.3 0.2 196 281 85 15 13 28 4520 -0.3 0.3 195 283 88 16 15 31 5566 -0.3 0.4 194 294 100 17 26 43 6680 -0.2 -0.4 NA NA NA NA NA NA NA -0.2 -0.3 225 261 36 -14 -7 -21 31 -0.2 -0.2 211 263 52 0 -5 -5 277 -0.2 -0.1 205 268 63 6 0 6 992 -0.2 0 200 276 76 11 8 19 1797 -0.2 0.1 200 278 78 11 10 21 2660 -0.2 0.2 197 280 83 14 12 26 3594 -0.2 0.3 196 282 86 15 14 29 4601 -0.2 0.4 195 291 96 16 23 39 5652 -0.1 -0.4 NA NA NA NA NA NA NA -0.1 -0.3 NA NA NA NA NA NA NA -0.1 -0.2 220 261 41 -9 -7 -16 36 -0.1 -0.1 211 264 53 0 -4 -4 319 -0.1 0 204 269 65 7 1 8 1057 -0.1 0.1 200 276 76 11 8 19 1872 -0.1 0.2 200 278 78 11 10 21 2743 -0.1 0.3 197 280 83 14 12 26 3688 -0.1 0.4 196 282 86 15 14 29 4699 0 -0.4 NA NA NA NA NA NA NA 0 -0.3 NA NA NA NA NA NA NA 0 -0.2 NA NA NA NA NA NA NA 0 -0.1 215 261 46 -4 -7 -11 47 0 0 211 268 57 0 0 0 383 0 0.1 204 270 66 7 2 9 1147 0 0.2 203 277 74 8 9 17 1974 0 0.3 199 278 79 12 10 22 2851 0 0.4 197 281 84 14 13 27 3806

182 0.1 -0.4 NA NA NA NA NA NA NA 0.1 -0.3 NA NA NA NA NA NA NA 0.1 -0.2 NA NA NA NA NA NA NA 0.1 -0.1 225 260 35 -14 -8 -22 3 0.1 0 215 262 47 -4 -6 -10 68 0.1 0.1 211 268 57 0 0 0 487 0.1 0.2 202 275 73 9 7 16 1268 0.1 0.3 201 277 76 10 9 19 2110 0.1 0.4 199 280 81 12 12 24 2990 0.2 -0.4 NA NA NA NA NA NA NA 0.2 -0.3 NA NA NA NA NA NA NA 0.2 -0.2 NA NA NA NA NA NA NA 0.2 -0.1 NA NA NA NA NA NA NA 0.2 0 225 261 36 -14 -7 -21 10 0.2 0.1 215 268 53 -4 0 -4 118 0.2 0.2 211 270 59 0 2 2 686 0.2 0.3 202 276 74 9 8 17 1481 0.2 0.4 201 278 77 10 10 20 2324 0.3 -0.4 NA NA NA NA NA NA NA 0.3 -0.3 NA NA NA NA NA NA NA 0.3 -0.2 NA NA NA NA NA NA NA 0.3 -0.1 NA NA NA NA NA NA NA 0.3 0 NA NA NA NA NA NA NA 0.3 0.1 225 261 36 -14 -7 -21 26 0.3 0.2 215 268 53 -4 0 -4 274 0.3 0.3 210 276 66 1 8 9 984 0.3 0.4 201 278 77 10 10 20 1795 0.4 -0.4 NA NA NA NA NA NA NA 0.4 -0.3 NA NA NA NA NA NA NA 0.4 -0.2 NA NA NA NA NA NA NA 0.4 -0.1 NA NA NA NA NA NA NA 0.4 0 NA NA NA NA NA NA NA 0.4 0.1 225 260 35 -14 -8 -22 7 0.4 0.2 215 268 53 -4 0 -4 91 0.4 0.3 211 270 59 0 2 2 617 0.4 0.4 204 277 73 7 9 16 1383

183 Table 5-9. Incidence Time and Transmission Period Change in Williamsport k1 k2 First Last T.Period F. Change L. Change P. Change Area -0.4 -0.4 222 261 39 -12 0 -12 27 -0.4 -0.3 205 265 60 5 4 9 297 -0.4 -0.2 204 271 67 6 10 16 905 -0.4 -0.1 201 275 74 9 14 23 1686 -0.4 0 199 279 80 11 18 29 2599 -0.4 0.1 198 280 82 12 19 31 3591 -0.4 0.2 198 283 85 12 22 34 4636 -0.4 0.3 197 293 96 13 32 45 5744 -0.4 0.4 196 297 101 14 36 50 6984 -0.3 -0.4 NA NA NA NA NA NA NA -0.3 -0.3 215 261 46 -5 0 -5 57 -0.3 -0.2 204 266 62 6 5 11 337 -0.3 -0.1 203 271 68 7 10 17 955 -0.3 0 200 276 76 10 15 25 1747 -0.3 0.1 199 279 80 11 18 29 2670 -0.3 0.2 198 281 83 12 20 32 3666 -0.3 0.3 197 283 86 13 22 35 4717 -0.3 0.4 197 293 96 13 32 45 5842 -0.2 -0.4 NA NA NA NA NA NA NA -0.2 -0.3 260 260 0 -50 -1 -51 6 -0.2 -0.2 215 261 46 -5 0 -5 69 -0.2 -0.1 203 266 63 7 5 12 373 -0.2 0 203 271 68 7 10 17 1011 -0.2 0.1 200 276 76 10 15 25 1817 -0.2 0.2 199 279 80 11 18 29 2752 -0.2 0.3 198 281 83 12 20 32 3752 -0.2 0.4 197 283 86 13 22 35 4810 -0.1 -0.4 NA NA NA NA NA NA NA -0.1 -0.3 NA NA NA NA NA NA NA -0.1 -0.2 260 260 0 -50 -1 -51 6 -0.1 -0.1 210 261 51 0 0 0 88 -0.1 0 205 271 66 5 10 15 417 -0.1 0.1 202 272 70 8 11 19 1087 -0.1 0.2 200 278 78 10 17 27 1906 -0.1 0.3 199 280 81 11 19 30 2857 -0.1 0.4 198 282 84 12 21 33 3864 0 -0.4 NA NA NA NA NA NA NA 0 -0.3 NA NA NA NA NA NA NA 0 -0.2 NA NA NA NA NA NA NA 0 -0.1 260 261 1 -50 0 -50 9 0 0 210 261 51 0 0 0 114 0 0.1 204 271 67 6 10 16 492 0 0.2 202 272 70 8 11 19 1193 0 0.3 199 279 80 11 18 29 2039 0 0.4 198 280 82 12 19 31 2997

184 0.1 -0.4 NA NA NA NA NA NA NA 0.1 -0.3 NA NA NA NA NA NA NA 0.1 -0.2 NA NA NA NA NA NA NA 0.1 -0.1 NA NA NA NA NA NA NA 0.1 0 260 261 1 -50 0 -50 14 0.1 0.1 207 270 63 3 9 12 156 0.1 0.2 204 272 68 6 11 17 626 0.1 0.3 201 276 75 9 15 24 1359 0.1 0.4 199 279 80 11 18 29 2232 0.2 -0.4 NA NA NA NA NA NA NA 0.2 -0.3 NA NA NA NA NA NA NA 0.2 -0.2 NA NA NA NA NA NA NA 0.2 -0.1 NA NA NA NA NA NA NA 0.2 0 260 260 0 -50 -1 -51 3 0.2 0.1 215 261 46 -5 0 -5 27 0.2 0.2 206 271 65 4 10 14 242 0.2 0.3 204 272 68 6 11 17 847 0.2 0.4 200 278 78 10 17 27 1616 0.3 -0.4 NA NA NA NA NA NA NA 0.3 -0.3 NA NA NA NA NA NA NA 0.3 -0.2 NA NA NA NA NA NA NA 0.3 -0.1 NA NA NA NA NA NA NA 0.3 0 NA NA NA NA NA NA NA 0.3 0.1 260 260 0 -50 -1 -51 8 0.3 0.2 215 265 50 -5 4 -1 71 0.3 0.3 204 272 68 6 11 17 464 0.3 0.4 203 277 74 7 16 23 1163 0.4 -0.4 NA NA NA NA NA NA NA 0.4 -0.3 NA NA NA NA NA NA NA 0.4 -0.2 NA NA NA NA NA NA NA 0.4 -0.1 NA NA NA NA NA NA NA 0.4 0 NA NA NA NA NA NA NA 0.4 0.1 NA NA NA NA NA NA NA 0.4 0.2 215 261 46 -5 0 -5 21 0.4 0.3 210 272 62 0 11 11 214 0.4 0.4 204 276 72 6 15 21 819

185

DD

k2 k1

Fig. 5-1. Degree Day Accumulation under Different DTR Conditions in Harrisburg in Day 100 (x, y axes represent k1 and k2, respectively; z axis represents DD accumulation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values. Notice when k2 is small the degree day accumulations are almost zero for any k1 values, such as the yellow and red balls. The degree day accumulation in day 100 is relatively small.)

186

DD

k2 k1

Fig. 5-2. Degree Day Accumulation under Different DTR Conditions in Harrisburg in Day 200 (x, y axes represent k1 and k2, respectively; z axis represents DD accumulation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values. Notice the pattern that for the balls with same colors, smaller k1 values (larger minimum temperature) will lead to sharper increase of degree day accumulation. Also the scale of degree day accumulation is different from day 100 (see figure 1).)

187

DD

k2K2 k1k1

Fig. 5-3. Degree Day Accumulation under Different DTR Conditions in Philadelphia in Day 100 (x, y axes represent k1 and k2, respectively; z axis represents DD accumulation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values. Notice when k2 is small the degree day accumulations are almost zero for any k1 values, such as the yellow, red, light green, and green balls. The degree day accumulation in day 100 is relatively small.)

188

DD

k2 k1

Fig. 5-4. Degree Day Accumulation under Different DTR Conditions in Philadelphia in Day 200 (x, y axes represent k1 and k2, respectively; z axis represents DD accumulation. Balls with same color represent degree day accumulation of same k2 values but varying k1 values. Notice the pattern that for the balls with same colors, smaller k1 values (larger minimum temperature) will lead to almost linear increase of degree day accumulation. Also the scale of degree day accumulation is very different from day 100 (see figure 3). Also the scale is different from that of Harrisburg, too (see figure 2).)

189

Fig. 5-5. Change of First, Last Incidence, Transmission Period, and Area in Category I

190 (Topleft panel: change of first emergence incidence relative to current condition where k1=k2=0. Topright panel: change of last emergence incidence relative to current condition. Bottomleft panel: change of total transmission period relative to current condition. Bottomright panel: area above threshold. The area represents total degree day accumulation over entire transmission period. I believe this should also be an important measurement of mosquito phenology and population dynamics. The legends showing in bottomright panel is also valid for all other three panels. For instance under current condition the first and last incidence Julian dates in Harrisburg are 207 and 259, respectively. And when k1=k2=0.1, the first and last incidence Julian dates are 206 and 260, respectively. Hence I compute the change of first incidence date as 207-206= +1, which means under k1=k2=0.1 condition, the first incidence date advances 1 day. And the change of last incidence date is 260-259= +1, which indicates the last incidence date postpones for 1 day. Therefore the total change of transmission period is 1+1=2 days. Notice the different calculation method for first and last incidence dates could ensure correct computation of change of transmission period. Also note significantly larger area value of Philadelphia.)

191

Fig. 5-6. Change of First, Last Incidence, Transmission Period, and Area in Category II

192 (Topleft panel: change of first emergence incidence relative to current condition where k1=k2=0. Topright panel: change of last emergence incidence relative to current condition. Bottomleft panel: change of total transmission period relative to current condition. Bottomright panel: area above threshold. In this category k2 is fixed at 0.4 and k1 varies from -0.3 to 0.3. More negative k1 values represent higher daily minimum temperature hence smaller DTR (because k2 is fixed, rising daily minimum would result in smaller DTR but higher daily mean temperature). There are strong decreasing trends of all the four changes: first incidence, last incidence, transmission period, and area above threshold, against larger k1 values (smaller DTR). Nevertheless note in any k1 condition the DTR is still larger than current condition, and I could see in any condition the value of changes in first, last, and transmission period are all positive (relative to current condition). Also note significantly larger area value of Philadelphia but shorter change of transmission period.)

193

Fig. 5-7. Change of First, Last Incidence, Transmission Period, and Area in Category III

194 (Topleft panel: change of first emergence incidence relative to current condition where k1=k2=0. Topright panel: change of last emergence incidence relative to current condition. Bottomleft panel: change of total transmission period relative to current condition. Bottomright panel: area above threshold. In this category k1 is fixed at 0.4 and k2 varies from -0.3 to 0.3. More negative k2 values represent lower daily maximum temperature hence smaller DTR (because k1 is fixed, lowering daily maximum would result in smaller DTR). There are increasing trends of all the four changes: first incidence, last incidence, transmission period, and area above threshold. Also I could see when k2 value is large (e.g. k2=0.3), the changes of first, last incidence, and transmission period are almost zero (or even a little larger), which means even when mean temperature is decreasing, large DTR would result in similar disease transmission timing as current condition. This again emphasizes the importance of DTR besides mean temperature. Note any change in this category does not result in any potential transmission of WNV for Harrisburg. )

195

Fig. 5-8. Change of First, Last Incidence, Transmission Period, and Area in Category IV

196 (Topleft panel: change of first emergence incidence relative to current condition where k1=k2=0. Topright panel: change of last emergence incidence relative to current condition. Bottomleft panel: change of total transmission period relative to current condition. Bottomright panel: area above threshold. In this category k1 is fixed at -0.4 and k2 varies from -0.3 to 0.3. More negative k2 values represent lower daily maximum temperature hence smaller DTR (because k1 is fixed, lowering daily maximum would result in smaller DTR). There are strong increasing trends of all the four changes: first incidence, last incidence, transmission period, and area above threshold, against larger k2 values (larger DTR). Note in any condition the value of changes in first, last, and transmission period are all positive (relative to current condition).Also note significantly larger area value of Philadelphia but shorter change of transmission period.)

197

Fig. 5-9. Change of First, Last Incidence, Transmission Period, and Area in Category V

198 (Topleft panel: change of first emergence incidence relative to current condition where k1=k2=0. Topright panel: change of last emergence incidence relative to current condition. Bottomleft panel: change of total transmission period relative to current condition. Bottomright panel: area above threshold. Note any change in this category does not result in any potential transmission of WNV for Harrisburg, Pittsburgh and Williamsport. Only Philadelphia could result in WNV transmission in some conditions (k1>0).)

199

200

Fig. 5-10. Accumulated Degree Days under Same Mean with Various DTR Conditions in Harrisburg (Two Dashed Horizontal Lines: EIP threshold of 109 and 89 DD at the beginning and rest of the season, respectively. Left panel: k<0; Right panel: k>0. )

201

202

Fig. 5-11. Accumulated Degree Days under Same Mean with Various DTR Conditions in Philadelphia (Two Dashed Horizontal Lines: EIP threshold of 159 and 89 DD at the beginning and rest of the season, respectively. Left panel: k<0; Right panel: k>0. )

203

Appendix 2 Figure 1 to 12: Change of first incidence, last incidence, and transmission period relative to current condition (k1=k2=0). x, y axes represent k1 and k2, respectively; z axis represents days. Balls with same color represent the same k2 values but varying k1 values.

204

Fig. 1. Change of First Incidence in Harrisburg

Fig. 2. Change of Last Incidence in Harrisburg

205

Fig. 3. Change of Transmission Period in Harrisburg

Fig. 4. Change of First Incidence in Philadelphia

206

Fig. 5. Change of Last Incidence in Philadelphia

Fig. 6. Change of Transmission Period in Philadelphia

207

Fig. 7. Change of First Incidence in Pittsburgh

Fig. 8. Change of Last Incidence in Pittsburgh

208

Fig. 9. Change of Transmission Period in Pittsburgh

Fig. 10. Change of First Incidence in Williamsport

209

Fig. 11. Change of Last Incidence in Williamsport

Fig. 12. Change of Transmission Period in Williamsport

210

211

Fig. 13. Accumulated Degree Days under Same Mean with Various DTR Conditions in Pittsburgh (Two Dashed Horizontal Lines: EIP threshold of 109 and 89 DD at the beginning and rest of the season, respectively. Left panel: k<0; Right panel: k>0. )

212

213

Fig. 14. Accumulated Degree Days under Same Mean with Various DTR Conditions in Williamsport

(Two Dashed Horizontal Lines: EIP threshold of 109 and 89 DD at the beginning and rest of the season, respectively. Left panel: k<0; Right panel: k>0.

214 Appendix 2. Illustration of Daily Temperature Profile in Six Major Categories

215

Chapter 6

Conclusion

I am always aware of our environment. I have always missed the painting-like beautiful mountains, skies, creeks, and waterfalls in my home town, which are almost all gone now, due to rapid industrialization and urbanization. About 12 years ago when I was in high school I participated in a UN program aimed at monitoring and assessing water and air quality. For the first time I was able to become familiar with and use modern equipment to conduct chemical experiments and quantitatively assess human impacts on ecosystems. As an undergraduate my research interest shifted from chemistry-based to biology-based. I was interested in biodiversity and ecological integrity, especially for arthropods, threatened by of human activities. In June 2008, after being accepted at Penn

State for a while, I attended a Northeast Climate Impact Assessment meeting. I realized my training in both entomology and operations research is very appropriate for studying the relationship between environmental change and insect responses using mathematical models. The spark from that meeting inspired me and shaped my graduate thesis study.

We are undoubtedly facing global climate change now and ecological systems are especially sensitive to climate change. As a Ph.D. candidate of both entomology and operations research, I am very interested in investigating how change in the daily temperature range (DTR) might influence degree day accumulation, insect phenology, and vector-borne disease transmission. I feel that daily temperature range and its relative change have somehow been overlooked by researchers. I suggest that mean temperature 216 alone cannot capture temperature fluctuation, which is experienced by insects in the field rather than in incubators. Finer scale daily temperature fluctuation models bring more accurate degree day accumulation estimation, and I suggest DTR has substantial influence on degree day accumulation. I show that DTR and mean temperature are not independent. Consequently, insect phenology and vector-borne disease transmission should also be related to DTR. So a primary objective of this research is to investigate how various DTR conditions influence insect phenology and periods of vector competency.

In chapter II I have explored how different environmental factors, primarily temperature and photoperiod, interact and influence grape berry moth (Paralobesia viteana) life history under two major climate change conditions, which I conduct out to the end of this century. These two conditions, A1fi and B1, represent different social- economic development scenarios in the future. I first develop individual-based methods for studying this process in a manner that are readily adaptable to multiple species, and verify the ability of this approach to model the distribution of each life stage by comparing simulations to field data. I have found that both conditions would increase number of generations of P. viteana, and I quantify the degree to which A1fi would have a much more significant impact on P. viteana voltinism and time-of-occurrence of each life stage.

Successively in the next chapter (chapter III) I use the individual-based methods developed in Chapter II, and further explore how different combination of DTR and mean temperature would alter P. viteana life history, contrasting results to current climatic conditions. Again, I develop methods that are applicable to other species, and enable us

217 to study a full range of DTR conditions, which I simulate a range that is bounded by the potential limits to the mean suggested by climate models. The results show that advancing time of emergence in different generations are not the same. Generally speaking, for the first generation the advancing time is not as much as later generations in most of the increased DTR conditions, and all increased mean temperature conditions. I also show that strong changes in DTR, even with current climatic conditions, has the potential to influence insect phenology similar to that which can be projected from long- term increases in mean temperature.

Then in chapter IV I switch to a vector-borne disease system (West Nile Virus,

WNV) and model Culex mosquito life history in four locations in Pennsylvania. I have proposed a calibrated degree day model to predict first and last WNV incidence dates.

The calibrated model assumes changing extrinsic incubation period (EIP) and mosquito longevity at different time of the season, because temperature has a nonlinear impact on both parameters. I also demonstrate species composition is an important factor for WNV transmission period. The calibrated models perform well to predict the WNV emergence period in Pennsylvania based on seven years of field observation data.

The last chapter recaptures previous research ideas and assesses the impact of changing DTR on WNV transmission. I have demonstrated quantitatively why mean temperature alone is insufficient to capture WNV transmission period. I have discovered asymmetric changes in first and last emergence dates in all locations. The first emergence dates, even under the most dramatic climate variability condition, only advance at most

7-10 days. On the other hand, last emergence dates differ significantly across different

218 DTR conditions. Also different locations response to changing DTR and changing mean temperature conditions differently.

The work presented in my thesis is a summary of my graduate research at Penn

State University. It is an accumulation of my research ideas and thoughts over the past years. It prepares me for my future research career, reflects my original intention of protecting our environment, and always reignites my passion to make the public aware of how we are destroying our world, how we might suffer from it, and how we could prevent it.

219

Appendix

Calculate Daily Temperature Profile and Degree Day Accumulation

Mean Temperature Model

This is the simplest model which treats daily mean temperature as the arithmetic mean of

daily maximum and minimum temperature:

TT T  max min mean 2

The disadvantage of this model is obvious. Although Tmean is derived from daily

maximum and minimum temperature, it does not contain any information about them,

that is, we are ignoring DTR information completely. For instance, consider two days

having exactly the same Tmean, when the first has 10 degrees DTR (0°C to 10°C) while the other has 20 degrees (-5°C to 15°C), for an insect the latter condition should be more favorable provided that in both conditions the minimal temperatures are above lethal threshold.

Sine-negative exponential Model

To overcome the problem inherited in the mean temperature model, biometerologists have proposed sine curve model (sinusoidal model, Allen 1976) to simulate DTR profile, which has the basic form as:

Ti  asin bi  c 220 Four inputs are required in order to reconstruct DTR profile: daily maximum temperature,

minimum temperature to determine coefficient a and c. Coefficient b is determined by photoperiod, which itself is further considered as a function of latitude and number of the day in the year. Thus I will need daily maximum, minimum temperature, latitude, and number of day in the year as inputs. The accumulated degree day is basically the integration over basic development threshold. To reduce computational effort I interpolate hourly temperatures and use these temperatures to approximate degree day accumulation above threshold in a day. This is a direct analog of Riemann Integral.

Forsythe et al. (1994) have proposed the following equations to calculate daylength:

 0.2163108  2arctan(0.9671396 tan(0.00860  (J  186)))  arcsin(0.39795cos  ) p L  sin sin sin 24 D 24  cos1 180 180 L  cos cos 180

where  is the revolution angle predicted by number of day in the year (J),  is the declination angle, L is latitude, and D is the daylength. p is the sun position constant and equals 0.8333 according to US government definition (Forsythe et al. 1994).

While I have calculated the daylength I could further determine temperature profile in a day:

()t t   T( T  T )sin(r )  T ( t  t  t ) imax minD2( a   ) min n x 1 (())t t  t   1 1 r s T( T  T ')  ( T  T ')cos(2 ) ( t  t  t ) imax min max minD x n ' 2 2 24  a   2

221

where t is current time, D is day length, tr and ts are the sunrise and sunset time derived

from daylength, Tmin’ is minimal temperature of the next day, and a and  are

predetermined coefficients (Wann et al. 1985).

The sine-negative exponential model is a fine tuning version of the original sine curve

model. Field observation data suggest that DTR is not necessarily symmetric in a day and

the temperature curve after sunset is better fitted with an negative exponential curve.

Thus the final model is formulated as:

 m Ti( Tmax  T min )sin( )  T min ( t n  t  t x ) D 2 a bn  Z Ti Tmin ( T s  T min ) e ( t x  t  t n ' ) where D is day length (determined by latitude and day in a year), Z is night length, Ts is

the temperature at the moment of sunset, m is number of hours after minimum temperature until sunset, n is number of hours after sunset until the time of minimum temperature of the next day, a is the lag coefficient for the maximum temperature, and b is the nighttime temperature coefficient (Parton and Logan 1981). Because sine-negative exponential model is derived from sine curve model, I will call it sine curve model in the following text for simplicity.

VITA

Shi Chen Department of Entomology School of Agricultural Sciences and Industries PENN STATE UNIVERSITY University Park, PA 16802 Office Phone: (814) 863-7657 Cellphone: (814) 880-0738 e-mail: [email protected]

EDUCATION

8/2006-08/2011 The Pennsylvania State University Ph.D. in Entomology and Ph.D. in Operations Research with graduate minor in statistics

9/2002-7/2006 Nanjing University, China B.S. in Biological Sciences

PUBLICATIONS

Chen, S., (2010). A Dynamic Programming Implemented Resource Competition Game Theoretic Model. Journal of Ecological Modelling, 221(16): 1847-1851.

Chen, S., Tobin, P. C., Fleischer, S. J., and Saunders, M. C. (2011). Projecting Insect Voltinism under High and Low Greenhouse Gas Emission Conditions. Environmental Entomology, in press

Chen, S., Fricks, J., and Ferrari, M. (2011). Tracking measles infection through nonlinear state space models, Journal of Royal Society of Statistics Series C, in press

Articles in preparation

Chen, S., Bao, S., and Byrne, C. C. A Game Theoretic Model for Predation and Antipredation behavior.

Chen, S., Blanford, J., and Thomas, M. B. Modeling Westnile virus emergence in Pennsylvania.