1 ANALYZING SEASONAL RISK INDICATORS OF MEDITERRANEAN FRUIT FLY CERATITIS CAPITATA (MEDFLY) IMPORTATION INTO FLORIDA VIA COMMODITY IMPORTS AND PASSENGER TRAFFIC By ANNA MARIA SZYNISZEWSKA A DISSERTATION PRESENTED TO THE GRADU ATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2013
2 2013 Anna Maria Szyniszewska
3 To my Dad wish you wer e here
4 ACKNOWLEDGMENTS This dissertation would not be possible without the immense support of my extremely talented and supportive advisor Dr. Andy Tatem, who has always been ready to give advice, comment and provide very valuable feedback on my work. I wish to convey my sincere gratitude to D r. Peter Waylen, who has been a tremendous academic and moral support to me and my family throughout the graduate program at the University of Florida. Dr. Michael Binford devoted much of his time to supervise the project from which my dissertation was funded and I am very thankful to him for that. I would also like to express my deepest appreciation to Dr. Norman Leppla for his invaluable input to the development of my research proposal. This dissertation would no t have this shape and form without his vast expertise, knowledge and willingness to help. I would also like to thank my collaborators, Dr. Roger Magarey, Dr. Manuel Colunga Garcia Dr. Dan Borchert and John Stewart, for their generous insight and assistanc e in developing my research proposal, which in result was successfully financed by the USDA APHIS PPQ CPHST Cooperative Agreement No. 12 8130 0158 Disciplinary Tools to Assess Seasonal Risk of Mediterranean Fruit Fly Cerati tis C apitata I have had an amazing time being a student at the Department of Geography. All the memories of friends I made in the department and the inspirational people I had a chance to spend time with, w ill stay with me forever. There are too many to mention but I would like to thank Dr. Risa Patarasuk, Dr. Deepa Pindolia, Andrea Wolf, Dr. Jaclyn Hall, Nicholas Campiz, Dr. Sanchayeetta Adhikari, Erin Bunting, Dr. Carlos Canas, Dr. Betty Lininger, Claudia Monzon, Dr. HuiPing Tsai, Dr. Zhuojie Huang, Amy Panikowski,
5 Renee Bullock, Caroline Staub, Jessica Steele, Matthew Graham, Dr. Tracy Van Holt, Ryan Kremser, Sam Schramsky, Ying Yang, Michala Jones, Johanna Engstrm David Keelings, Dr. Tracy Van Holt, Dr. Corene Matyas, Dr. Keith Yearwood and many others for their amazing support and friendship over so many years. Finally, I would like to express my deepest gratitude to my family my extremely supportive husband for his understanding, patience and continu ous encouragement, as well as my daughter for her everyday smile, energy and for being the biggest comfort of my life. I would also like to thank my amazing Mom, for her patience and bright, positive spirit. My proud Father has been anxiously waiting for m e to complete this degree and I wish I could share and celebrate this m ilestone with him.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............. 8 LIST OF FIGURES ................................ ................................ ................................ .......... 9 LIST OF ABBREVIATIONS ................................ ................................ ........................... 11 ABSTRACT ................................ ................................ ................................ .................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ ..... 14 2 GLOBAL SEASONAL POTENTIAL DISTRIBUTION OF MEDITERRANEAN FRUIT FLY CERATITIS CAPITATA ................................ ................................ ........ 18 Background ................................ ................................ ................................ ............. 18 Data and Methods ................................ ................................ ................................ ... 20 Occurrence data ................................ ................................ ............................... 21 Covariates used in modeling ................................ ................................ ............. 22 Choice of seasons ................................ ................................ ............................ 23 Maximum entropy modeling ................................ ................................ .............. 24 Model testing ................................ ................................ ................................ ..... 27 Results ................................ ................................ ................................ .................... 28 Occurrence data ................................ ................................ ............................... 28 Environmental niche modeling ................................ ................................ .......... 29 Relative importance of predictor variables ................................ ........................ 30 Discussion ................................ ................................ ................................ ............... 30 Conclusions ................................ ................................ ................................ ............. 34 3 ANALYZING MEDITERRANEAN FRUIT FLY SEASONAL IMPORTATION RISK INDICATORS THROUGH COMMODITY IMPORT TRAFFIC TO FLORIDA ................................ ................................ ................................ ................. 42 Background ................................ ................................ ................................ ............. 42 Data and methods ................................ ................................ ................................ ... 46 Results ................................ ................................ ................................ .................... 50 Discussion ................................ ................................ ................................ ............... 53 Conclusions ................................ ................................ ................................ ............. 55 4 ANALYZING SEASONAL IMPORTATION RISK INDICATORS FOR MEDITERRANEAN FRUIT FLY VIA AIR PASSENGER TRAFFIC TO FLORIDA .. 65
7 Background ................................ ................................ ................................ ............. 65 Methods ................................ ................................ ................................ ................... 67 Passenger flow data and airport locations ................................ ........................ 67 Medfly distribution ................................ ................................ ............................. 69 Risk indicator calculations ................................ ................................ ................. 70 Interception data ................................ ................................ ............................... 71 Results ................................ ................................ ................................ .................... 7 2 Seat capacity on direct flight connections to Florida ................................ ......... 72 Pass enger flows ................................ ................................ ................................ 72 Seasonal risk indicators and passenger flow ................................ ............. 73 Interception data ................................ ................................ ............................... 76 Discussion ................................ ................................ ................................ ............... 77 5 CONCLUSIONS ................................ ................................ ................................ ...... 90 LIST OF REFERENCES ................................ ................................ ................................ 93 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 102
8 LIST OF TABLES Table page 2 1 Data sets used in the development of ecological niche models, including source and sp atial resolution. ................................ ................................ ............. 36 2 2 The list of the data sets used in the development of seasonal ecological niche models. ................................ ................................ ................................ ...... 36 2 3 Model acc uracy scores. ................................ ................................ ...................... 36 2 4 Relative influence of the contribution of the variables to the model. ................... 37 3 1 Seasonal risk indicators for c ommodities entering ports of entry in Florida (PRI) from Medfly infested countries according to EPPO (2001). ....................... 57 3 2 R isk indicator for commodities arriving from international origin locations (ORI). ................................ ................................ ................................ .................. 58 3 3 Highest total tonnage origin destination pairs and their seasonal port of entry origin risk indicators (PORI). ................................ ................................ ............... 59 4 1 Main origin airports by region from the Medfly infested countries arriving to Florida according to the largest total passenger numbers. ................................ 81 4 2 Main destination airports in Florida accordin g to the largest total passenger numbers.. ................................ ................................ ................................ ............ 82 4 3 Pairs of selected origin destination airports.. ................................ ...................... 83 4 4 Type of host and the nu mber of Medfly interception at Miami airport (2003 2008). ................................ ................................ ................................ .................. 84
9 LIST OF FIGURES Figure page 2 1 Occurrence data for Ceratitis capitata used in the stu dy.. ................................ .. 38 2 2 Global environmental suitability for Medfly as predicted by MaxEnt model.. ...... 39 2 3 3 panel seasonal maps showing the environmental suitability for Ceratitis capitata occurrence annually according to the MaxEnt model. ........................... 40 2 4 ROC curves representing mean AUC ................................ ................................ 41 3 1 Top maritime importers of fruits and vegetables (FV) to Florida (2004 2008) according to the total volume in tonnes. ................................ ............................. 60 3 2 The total tonnage of fruits and vegetab le (FV) imported to Florida between 2004 and 2008 from worldwide origin countries via maritime pathway. .............. 60 3 3 Total volume of fruit and vegetable maritime imports in each Florida ports betw een 2004 2008. ................................ ................................ ........................... 61 3 4 Share of maritime fruits and vegetable imports that carry main, occasional or no host for Medfly classified by total tonnage for 2004 2008. ............................. 62 3 5 Intraannual and interannual variability in the incoming volume of FV imports to Florida via maritime pathway between 2004 and 2008. ................................ .. 63 3 6 Intraa nnual variability in the total volume of incoming maritime fruit and vegetable (FV) imports to Florida per port of entry (2004 2008). ........................ 63 3 7 Seasonal risk factors for Medfly importation. ................................ ...................... 64 4 1 Direct international connections to Florida and the annual seat capacity total on those routes (OAG2010). ................................ ................................ ............... 85 4 2 The sea sonal seat capacity total on incoming flights to Florida from the 15 th highest volume countries of origin. ................................ ................................ ...... 85 4 3 Countries with the highest number of passenger arrivals to Florida (up to 2 connections). ................................ ................................ ................................ ....... 86 4 4 Countries that have Medfly officially present with the highest number of arriving passengers to Florida (up to 2 connections) by destination airport. ....... 86 4 5 Medfly suitability adjusted numbers of passengers arriving from Medfly infested countries compared to the international arrivals total. ........................... 87 4 6 Ratio of passengers arriving from Medfly infested countries .............................. 87
10 4 7 Three panel map illustrating the Medly suitability adjusted volume of passenger flows to Florida per season. ................................ .............................. 89 4 8 Number of Medfly interceptions in passenger luggage at Miami International Airport (MIA) between 2003 and 2008 according to PestID. ............................... 89
11 LIST OF ABBREVI ATIONS AORI Airport Origin Risk Indicator APHIS Animal and Plant Health Inspection Services ARI Airport Risk Indicator EPPO European and Mediterranean Plant Protection Organization FLL Fort Lauderdale International Airport GNV Gainesville Regional Airport M AX E NT Maximum Entropy Algorithm M EDFLY Mediterranean fruit fly Ceratitis capitata (Wiedemann) MCO Orlando International Airport MIA Miami International Airport NDVI Normalized Diffe rence Vegetation Index ORI Origin Risk Indicator PBI Palm Beach Internatio nal Airport PIE St. Petersburg Clearwater International Airport PORI Port Origin Risk Indicator PPP Purchasing Power Parity PPQ Plant and Pest Quarantine PRA Pest Risk Assessment PRI Port Risk Indicator SFB Orlando Sanford International Airport TLH Tallaha ssee Regional Airport TPA Tampa International Airport USDA United States Department of Agriculture
12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the D egree of Doctor of Philosophy ANALYZING SEASONAL RISK INDICATORS OF MEDITERRANEAN FRUIT FLY CERATITIS CAPITATA (MEDFLY) IMPORTATION INTO FLORIDA VIA COMMODITY IMPORTS AND PASSENGER TRAFFIC By Anna Szyniszewska December 2013 Chair: Andrew Tatem Major: Ge ography Mediterranean fruit fly (Medfly) is considered to be one of the most destructive invasive pests. It is highly polyphagus and widely spread across the world. It has invaded Florida multiple times and each time required costly eradication. Should it agriculture. Extensive transportation links through both passenger and cargo, and the of new pests. The surveillance resources at its borders however are limited. Therefore, making use of existing information to guide the tailoring and targeting of inspection efforts a t ports of entry is vital. A nalyses that can aid the identification of bo th routes and times of year when the long distance movement of the pest is more likely to occur are presented herein The assessment is based on a combination of seasonal Medfly suitability maps with statistics on the seasonal patterns of cargo and passeng er transportation characteristics. I identify seasonally changing risks of Medfly importation from various origin points depending on the frequency, volume and infestation potential of transported commodities, modeled air passenger flow volumes, as well as the
13 destination port of entry in Florida. Presented study identifies the times of year, ports of entry and origin countries with the highest relative risks of Medfly arrival, based on available information. I demonstrate how risk indicators for Medfly arr ival change seasonally and therefore underscore the need for information that can guide allocation of resources at the ports of entry
14 CHAPTER 1 INTRODUCTION Biological invasions have always been part of natural processes. Species were colonizing new h abitats following major disturbances and changes in climate via natural means of spread wind, terrestrial or aquatic migration. Today however, distance and travel time pose much lower constraints to species migration than they did several centuries ago. Continued globalization, trade liberalization and increasing passenger traffic play a facilitating role in the movement of various biological organisms between even distant geographic regions at unprecedented pace. Transport network expansion provides fast connections and gateways into new regions for the spread of exotic species, and therefore results in inevitable increases in pest arrival rates for many countries (Mack et al. 2000; Drake and Lodge 2004; Tatem and Hay 2007; Hulme et al. 2008; Varela et al 2008; Floerl et al. 2009; Perry and Vice 2009). How often a species is transported and how many individuals survive are recognized as important correlates of establishment success (Lockwood et al. 2005; Hulme et al. 2008; Colunga Garcia et al. 2009; Loc numbers of individuals involved in the colonization event (propagule size) and the number of discrete events of colonization (propagule number) (Lockwood et al. 2005). Climate has been identified as another important factor that not only contributes to chances of exotic pest survival upon arriving to new destination, but also drives its population dynamics in its established range (Escudero Coloma r et al. 2008; Lockwood et al. 2009). It influences host plant availability, potential generation time, as well as seasonality and overwintering potential determined by the intensity and duration of low
15 temperatures. Despite the importance of these two fac tors, analyses on the potential role of far reaching transport networks combined with climatic similarity in biological invasions remain few (Drake and Lodge 2004; Tatem et al. 2006b; Tatem et al. 2006c; Tatem et al. 2006d; Tatem et al. 2006a; Tatem and Ha y 2007). Among a variety of pests, insect pest species are of especially high concern for the health of agriculture and the environment. It is estimated that worldwide, insect pests destroy approximately 14% of all potential food production, despite the ye arly application of more than 3000 million kilograms of pesticides (Pimentel et al. 2005). The Mediterranean fruit fly, Ceratitis capitata (Medfly) (Wiedemann) is considered one of the s a country commodities internationally (Papadopoulos et al. 2001b; Papadopoulos et al. 2001a; Papadopoulos et al. 2002). It originated from sub Saharan Africa and subsequently invaded parts of southern Europe, Central and South America, the South P acific Islands and Australia (Gasperi et al. 2002). The Medfly poses a serious economic threat to fresh fruit and vegetable production in many states of the US. Some of the main host plants include avocado, banana, bittermelon, carambola, coffee, guava, ma ngo, papaya, peppers and persimmon and citrus (Weems Jr 1981; Liquido et al. 1990; Liebhold et al. 2006). The pest has been established for about a century in Hawaii and despite continuous and costly eradication efforts, repeatedly reintroduced into two st ates: California and Florida (Carey 1996; Jang 2007). It has the potential for tremendous economic damage and eradication costs are hugely expensive. It is estimated that the cost of each of its previous multiple incursions into the US (eradication and ind ustry loss) ranged from $300,000 to $200 million and the cost of
16 potential establishment is estimated at $821 million or more per year (APHIS 1992). Medfly outbreaks in California during the past 25 years have cost taxpayers nearly $500 million, while the Medfly outbreak in Florida's Tampa Bay region in 1997 resulted in $25 million spent on eradication (Cross 2004). Frequent Medfly invasions in the US have been attributed to propagule pressure (Liebhold et al. 2006). Invasions of Medfly are most likely due to multiple introductions, which allow the maintenance and enhancement of genetic variability in the adventive populations, resulting in improved potential invasiveness (Malacrida et al. 2007). Except for a small Medfly infestation in Texas, all Medfly det ections in the continental United States have occurred in Florida and California. The insect was first found in Florida in 1929, and first detected in California in 1975. Since that time, multiple infestations have occurred, and while some researchers beli ev e it is continuously reintroduc ed into the state of California (Liebhold et al. 2006), some claim that Medfly might be already established there (Carey 1991; Carey 1996). Without control measures, Florida could sustain high Medfly populations because of both favorable climate and availability of preferred hosts throughout the year. The seasonal population occurrence of Medfly is well documented in regional studies from several areas of the world (Harris and Olalquiaga 1991; Harris et al. 1993; Israely et al. 1997; Katsoyannos et al. 1998; Mavrikakis et al. 2000; Papadopoulos et al. 2001a; Maelzer et al. 2004; Israely et al. 2005; Escudero Colomar et al. 2008; Martnez Ferrer et al. 2010). Some of these studies relate seasonality to temperature and the abi lity of the Medfly to overwinter in the areas studied. Others link seasonality also to the rainy or dry seasons, or particular host availability (Sciarretta and Trematerra 2011).
17 Medfly development has been shown to have a linear relationship with temperat ure (Duyck and Quilici 2002) and development thresholds have been recorded for all four life stages of the insect (Grout and Stoltz 2007). However, it is known that the pest is capable of adapting to a wide range of climates and the larvae can overwinter i nside of certain fruit hosts (Papadopoulos et al. 1996). Escudero Colomar et al. (2008) surveyed annual Medfly population development in Northern Spanish orchards, illustrating the strong seasonal pattern in its occurrence. Evidence exists that not only pe st population dynamics, but also exotic pest arrival rates and commodity movements show seasonal patterns (Dobbs and Brodel 2004; Caton et al. 2006; McCullough et al. 2006; Tatem and Hay 2007; Tatem 2009). An understanding of the processes that govern inse ct pest species movements is of central importance in building evidence based border surveillance and control strategies to minimize their adverse impacts (Floerl and Inglis 2005; Venette et al. 2010). Pest Risk Assessment (PRA) guidelines outlined by the Secretariat of the International Plant Protection Convention (IPPC) in the International Standards for Phytosanitary Measures (ISPM) consider multiple risk factors, however are not concerned about how these factors are changing seasonally in a calendar yea r and what the interannual variation in these factors is. This thesis aims to explore the relationship between seasonal variations of these factors that are likely to interact and produce variability in seasonal Medfly interception risk at the points of en try to Florida. With limited funds available for surveillance and control, multidisciplinary approaches can aid the identification of both routes and times of year when the long distance movement of pest species are most likely to occur, facilitating the t argeting of resources
18 CHAPTER 2 GLOBAL SEASONAL POTENTIAL DISTRIBUTION OF MEDITERRANEAN FRUIT FLY CERATITIS CAPITATA Background Invasive alien species are directly associated with biodiversity loss, ecosystem service changes, and negative impacts on hum an health, agriculture, forestry and fisheries. In Europe alone, these losses and impacts are estimated to cost at least EUR 12 billion per year (Riccardo Scalera et al. 2012). The Mediterranean fruit fly, Ceratitis capitata (Wiedemann), commonly referred most destructive pests. It is a highly polyphagus species, able to feed on over 300 hosts and known to be capable of adapting to a wide range of climates (Liquido et al. 1990; Papadopoulos et al. 2001b; Papado poulos et al. 2001a). It causes significant damage to fruits and vegetables, and its economic impacts are substantial. It is a quarantine pest and countries with established Medfly populations have significant trade barriers imposed on their exports, as in the case of Argentina, where Medfly was first detected in the orchards of Buenos Aires at the beginning of 19th century and is currently well established in most regions of the country (Vera et al. 2002; Segura et al. 2004; Ovruski et al. 2010). Some co un tries, including New Zealand Chile and Belize were successful in Medfly eradication (EPPO 2009). Comprehensive global information on Medfly occurrence, both in terms of spatial and temporal presence, is crucial for understanding not only the current and historical extent of its presence, but also the conditions where it is able to survive and areas susceptible to potential invasion. For similar reasons, it is also essential to track A version of this chapter is under review for publication: Szyniszewska AM and Tatem AJ: Global seasonal potential distribution of Mediterranean fruit fly Ceratitis cap itata PL O S ONE Under Review
19 historical spread routes and the history of invasion. Occurrence records with temporal reference are important for understanding the drivers of Medfly seasonal population dynamics, which can be valuable for guiding eradication and control strategies. CABI and EPPO maintain the official global records on where Medfly is establis hed. Both sources define this at the country scale and in some cases, on the provincial scale (CABI 2000; EPPO 2001 2009 ). There are no widely available current expert opinion maps defining the environmental range of known Medfly occurrence. Recent years have seen a handful of studies aiming to define the potential distribution of Medfly. In a study by De Meyer et al. (De Meyer et al. 2008), a genetic algorithm for rule set prediction (GARP) and principal component analysis (PCA) were used to estimate the potential geographical range of Medfly using native range distributional data derived from a database maintained by the Royal Museum for Central Africa. This data was complemented by non native range information gathered from the literature and electronic resources. Outputs showed areas of high and low suitability for Medfly presence globally without providing information on what constituted the threshold for such categories, or any seasonal changes. CLIMEX ( http://www.csiro.au/solutions/ps1h3 ) was used in a different study to assess the seasonal and year to year variation in climatic suitability for Medfly worldwide with emphasis on Argentina and Australia (Vera et al. 2002). No occurrence data was used for the modeling, but rather parameters of its population dynamics were used, specifically a CLIMEX growth index derived from a study of Medfly populations in Thessaloniki. Gutierrez and Ponti (2011) also assessed the invasive potential of Medfly
20 in Califo rnia and Arizona using GRASS GIS, based on age structured dynamics of Medfly life stages and temperature variability in the region. It is well documented in regional studies from several areas of the world that Medfly has a highly seasonal pattern to its population dynamics (Harris and Olalquiaga 1991; Harris et al. 1993; Israely et al. 1997; Katsoyannos et al. 1998; Mavrikakis et al. 2000; Papadopoulos et al. 2001a; Maelzer et al. 2004; Escudero Colomar et al. 2008; Martnez Ferrer et al. 2010). However, spatiotemporal datasets to quantify these patterns on a global scale have yet to be assembled, while previous global mapping of the suitability for Medfly presence has not accounted for seasonal shifts. Here I present the results of a study that has focused on constructing the most comprehensive database on confirmed Medfly occurrence records, the timing of these records and their locations. Moreover, information on hosts, life stages and capture method were also recorded. Finally, seasonally varying gridded environmental variables were linked to these presence records in a niche modeling framework to produce predictions of the annual and seasonal distributions of suitability for Medfly presence on a global scale, with an aim of identifying regions that can be potential risk areas for Medfly invasions depending on the season Data and Methods There is significant information available in the literature regarding the biology, environmental preference s and Medfly occurrence which is valuable in terms of building a comprehensive, multidisciplinary understanding of the various biotic and abiotic variables that govern its seasonal population dynamics (Harris et al. 1993; Vera et al. 2002; Israely et al. 2005; De Meyer et al. 2008; Carey 2011). The seasonal dynamics of Medfly populations are well documented in regional studies from several
21 areas of the world (Harris and Olalquiaga 1991; Harris et al. 1993; Israely et al. 1997; Katsoyannos et al. 1998; Mavr ikakis et al. 2000; Papadopoulos et al. 2001a; Maelzer et al. 2004; Escudero Colomar et al. 2008; Martnez Ferrer et al. 2010). Occurrence data Occurrence data for Ceratitis capitata were searched for in online open a ccess museum collections data, published articles, reports and conference proceedings. The literature search result ed in 158 publications and reports containing potential data to be reviewed. Of these publications, 101 contained information about Medfly occurrence that could be ge o located, and 64 contained information about the month when Medfly was observed. A databas e was constructed to store these data and it contains current temporal and historical data pertaining to Medfly presence. For each entry, information about the author, year and type of publication, country, two administrative levels and locality, georefer enced location and source of coordinates, the quality of information about the location, year and month of the occurrence record, sampling technique, developmental stage of Medfly and host plant were recorded. This data record protocol was built upon that pioneered for recent studies of malaria vectors worldwide (Sinka et al. 2010a; Sinka et al. 2010b; Hay et al. 2010; Sinka et al.). Locations of Medfly observations were georeferenced, either based on coordinates included in the source material or dependent on the name of the location found in the source. Additional supporting sources of information on the recorded presence of Medfly were obtained from the Global Biodiversity Information Facility (GBIF http://data.gbif.or g ), Belgian Biodiversity Platform (BeBIF www.biodiversity.be ) and the Royal Museum of Central Africa (De Meyer et al. 2008)
22 Covariates used in modeling A suite of environmental variables was constructed in preparation f or use in niche modeling. Medfly is known to be sensitive to climate, and its main limitation in development is low temperature that may hinder its ability to overwinter, plus high precipitation which may have an adverse impact on the pupae development in the soil (Israely et al. 1997; Papadopoulos et al. 2001a). It is unable to overwinter at high altitudes on the fringes of suitable areas, and its dynamics are linked to the availability of hosts (Israely et al. 2004; Escudero Colomar et al. 2008). Given these factors, I obtained land surface temperature (LST) and normalized difference vegetation index (NDVI) images from the Advanced Very High Resolution Radiometer (AVHRR) satellite sensor. The products are available at 8x8km s patial resolution for over 20 years time ( http://daac.gsfc.nasa.gov/ ). Digital elevation model data (DEM) were obtained from the Shuttle Radar Topo graphy Mission (SRTM) ( http://www2.jpl.nasa.gov/srtm/ ) Finally, annual and quarterly average, minimum and maximum precipitation data were derived from the Worldclim database ( http://www.worldclim.org/ ). The data represent interpolated rainfall measures derived from the world wide network of weather stations for the time period of 1950 2000 (Hijmans et al. 2005). All of the gridded datasets underwent a number of proc essing steps prior to being used in modeling (Hay et al. 2006; Sinka et al. 2010b; Hay et al. 2010). Each gridded dataset was processed to ensure that the size, location and extent matched for every layer. For datasets where the remotely sensed information was multi temporal, Fourier analysis was used to ordinate the data by decomposing the temporal signal into an additive series of harmonics of different seasonal frequencies (Hay et al. 2006; Sinka
23 et al. 2010a; Sinka et al. 2010b; Hay et al. 2010; Sinka e t al.). All of the gridded datasets were resampled to produce matching extents and a grid cell size of 5km x 5km. The layers used in the modeling is likely well beyond the northern limit of Medfly suitability anyway. All of the data layers were tested for pairwise Pearson correlation prior to building and running the model. Although MaxEnt is known to be a stable model in the face of correlated variables (Elith et al. 2011), those with the high correlation coefficients (r >= 0.85) were excluded from the analysis. These were average precipitation and LST in the general model, and average precipitation i n the seasonal model. Average LST was removed from the first season, average LST of the coldest month was removed from the second season and the average LST of the warmest month remained as a covariate only in the first season model Choice of seasons The year was divided into three Medfly relevant seasons for seasonal mapping to strike a balance between ensuring that the seasonal variation in Medfly population dynamics was captured, and having sufficient data points to produce reliable maps for each time p eriod. Dividing year according to calendar seasons did not correspond well with the activity of Medflies, especially in the northern hemisphere, where the seasonal differences tend to be most pronounced due to a larger proportion of land located in higher latitudes. On the northern limits of the Medfly distribution, the pest tends to be inactive between January and April (pupates in the soil), and the onset of activity starts anywhere between May and August, depending on location and condition in a particul ar year. The peak of Medfly activity is observed in the fall usually between September and November, and a sharp decline is observed between November and December.
24 This led us to divide the year into three seasons (January April, May August and September December), which is both significant from the phenological point of view for Medfly, and at the same time preserves some common environmental characteristics of the seasons Maximum entropy modeling The Maximum Entropy Modeling tool (MaxEnt version 3.3.3k) was used to map by finding the maximum entropy distribution, in other words, distribution closest to uniform (Elith et al. 2006; Phillips et al. 2006). In the model the environmental values found at the occurren ce localities impose certain constraints on the output distribution. The constraints are expressed as simple functions of the environmental variables called features, and each feature in the model should have a mean close to the empirical average. The model looks for a set of probability distributions that satisfy the constraints and chooses the most unconstrained one (Phillips et al. 2006). In the logistic output of the model, every grid square has an assign ed value between 0 and 100, which represents the relative suitability of species occurrence. There were many reasons that dictated the choice of this model. Most importantly, in a review of 16 species modeling methods, MaxEnt was among the best performing methods when evaluated using the AUC and correlation statistics (Elith et al. 2006). Secondly, the method holds a strict mathematical definition and can accommodate diverse types of predictor variables both categorical and continuous. Moreover, it does n ot require absence data, can handle a relatively low sample size and gives a simple to interpret continuous output. Finally, the method is well documented and available for free download ( http:// www.cs.princeton.edu/~schapire/maxent ).
25 The default MaxEnt model parameters have been calibrated on a wide range of data (a convergence threshold of 10 5 a maximum iteration value of 500 and the maximum number of background points as 10000). These setti ngs are recommended to achieve good model performance for species at ecological equilibrium (Phillips and Dudk 2008). Because Medfly is an invasive organism, I modified these settings to better predict the nature of the potential niche of an invader. The convergence threshold was left at the default of 10 5 and the number of iterations was increased to 5000 to allow the model adequate time for convergence and avoid under or over prediction of the relationships. I explored the choice of features and increa sed model regularization. MaxEnt allows various feature types to be used by default (if there are enough sample points on species presence available), which results in complex functions. Using less or only one feature is recommended for simpler models and I chose hinge features which allow the model to fit nonlinear functions of varying complexity, but without the sudden steps of threshold features (Elith et al. 2011). Increased regularization parameters increase the degree of level smoothing, however the A UC score of the model was consequently declining with increased regularization. Therefore, I left the regularization setting at the default of 1. The output maps illustrate the mean results of the replicated runs. There are no precise scientific guidelines that dictate the choice of settings, so my choice was based on a visual assessment of their influence on the partial dependence plots, AUC scores and the prediction maps. The logistic habitat suitability output values were used for the model output which i s simple to interpret (probability range of occurrence between 0 and 1). The model estimates the relative influence of each variable used in the prediction. It is scaled so
26 the sum of the relative influence of each variable adds to 100, with higher numbers indicating stronger contribution on the outcome (Elith et al. 2008). In the case where multiple Medfly presence points in my constructed database were registered at a single location, only one record was used in the MaxEnt modeling Only points of occurre nce and sources from 1980 onwards were taken into consideration to build the models. By default, MaxEnt selects its own background samples from the entire study region, which implies that this entire space is available to species and surveillance (Elith et al. 2010). Another option is to include a mask that will allow MaxEnt to choose a background sample only from pre selected areas. In the case of Medfly these could be areas that are accessible for Medfly where the species currently is present and no era dication efforts are currently ongoing, and no quarantine measures against Medfly are in place. It could also include areas that were accessible to Medfly over decades. Finally, the mask could help represent the sampling bias of Medfly presence records, ho wever this was not feasible here as the data comes from various sources over long temporal scales. Since the goal of this study was to represent Medfly potential and not the actual niche, and also how environmental variables favorable for its presence are changing according to the season, plus I only have political boundaries of Medfly current presence and not expert drawn distribution maps, I did not use a mask for the general potential niche model, but I applied a mask for the three seasonal models. The m ask was obtained by using a probability threshold of Medfly occurrence equal to or above 0.1 from the general model It helped to avoid overprediction of the model and filter out areas that can only be suitable for Medfly occurrence during a few months of the year and therefore cannot sustain populations over climatically unfavorable months
27 Model testing The accuracy of the distribution models was evaluated by partitioning the data within MaxEnt into training (75%) and testing (25%) subsets and performing validation statistical analyses on each of the partitions. Each of the settings were run on 30 replicates using a subsample run type with a random seed, so that Max E nt could average the results from all of the models created. Firstly, the area under the Re ceiver Operating Characteristic (ROC) was used to measure model performance. The plot of the ROC curve is illustrative of the ratio of correctly classified positives to the total number of positive cases (sensitivity) versus the false positive rate (specif icity) at all thresholds of presence absence classification. In this case, I do not have actual absence data in the study. Therefore, tests show whether the model classifies presence more accurately than a random prediction. The ROC plot for a model whose predictive ability is the equivalent of random assignment will lie near the diagonal, where the true positive rate equals the false positive rate for all thresholds. The area under the curve (AUC) is therefore a good measure of the overall model performan ce and has a possible range of 0 1, where 0 indicates that prediction is equal to a random assignment while an AUC score of 1 indicates a perfect presence absence prediction. Secondly, a threshold dependent binomial test of omission was performed. If in a specific cell I observe a value of 0.10 or above, that cell is classified as suitable for Medfly. This approach transforms the prediction output from continuous into binary. The number of Medfly suitable cells was compared to the number of cells known to h ave had Medfly presence. A one tailed binomial test was used to find out whether the model outperformed a random model predicting Medfly to be present in the same number of cells. MaxEnt provides test statistics for binomial tests for 10 different threshol d values.
28 The extrinsic omission rate represented the fraction of the test localities that were assigned into pixels which are not predicted as suitable for Medfly. Low omission rate is highly advisable for a good model (Anderson et al. 2003) Results Occu rrence data The search for data on Medfly presence and absence resulted in records from 43 countries and nearly 500 unique localities. The oldest records come from 1898 and the most recent from 2011. 171 locations contained information about the year of Me dfly occurrence, and 125 about a specific month where Medfly was observed. The majority of the records identified Medfly occurrence at the adult stage of development, with some in the pupae and larvae stage. Some of the most common hosts included apricots, guava, peach, various types of citrus (mainly varieties of oranges and mandarines), apples, fig, peach, loquat and coffee. The dominant method of recording occurrence data was through food or trimed lure baited traps (McPhail and Jackson, but also Nadel, M axitrap, Steiner and Lynfield). For the purpose of niche modeling points with uncertain locations, duplicate points and those collected before 1980 were removed for the analysis. Additionally, the data were supplemented with occurrence points derived from the GBIF, BeBIF and MCA datasets, which further increased the total number of sample points available for niche modeling In the annual model (produced using all ge o located points available), 463 occurrence points were used. For the seasonal models, 139 points were used for the January through April model, 158 for May through August and 157 for September through December. Figure 2 1 shows the locations of Medfly occurrence worldwide obtained from all of the data sources, starting from 1980 onwards. The po ints where
29 information about Medfly occurrence after 1980 were available are marked in red (463 unique locations), and those marked with crosses are locations that had the month of record information on file (270 unique locations) Environmental niche mo deling The annual Medfly niche suitability model, produced using all geolocated occurrence records since 1980, is presented in Figure 2 2. The largest suitable areas for Medfly presence are located in South America, east ern and south ern Africa and eastern Asia. Other suitable areas appear across a variety of climate zones, including warm temperate and semi tropical and tropical, mostly in coastal areas. This inc l udes the Mediterranean basin, Gulf of Mexico, western coast of South America and coastal areas o f India and Australia. The model prediction performs significantly better than random with a binomial test result of p < 3.9 40 The AUC score for the training and testing datasets is 0.882 and 0.878 respectively, representing strong predictive performance given the fractional predicted area of 0.307 (Table 2 3). In the January April and September December models the lowest fraction of the study area is predicted as a suitable (0.376 for Jan Apr, 0.335 for May Aug, and 0.336 for Sept Dec) (Fig ure 2 3). In the Jan Apr season, the least amount of land area in the Mediterranean basin is shown as suitable for Medfly. The area surrounding the Gulf of Mexico and Caribbean Basin, as well as the Pampa in Argentina and eastern Brazil are predicted to be highly suita ble. In Africa, the highest suitability is observed in the Sahel belt, some parts of Abyssinia and the southern part of the continent, including Madagascar. High suitability is also apparent in southeast ern Asia, where Medfly is not yet known to be establi shed. The May Aug season largely corresponds with summer in the northern hemisphere. Consequently, the largest proportions of areas in Europe,
30 North America and Asia appear as suitable, compared to the other seasons. The expansion in predicted suitable ran ge is also apparent in central Africa and northern Aust ralia. In the Sept Dec season, the suitability is largely contained in the Mediterranean basin in Europe, Southeast United States, but is expanded in the southern hemisphere. The AUC scores for the tra ining data for all the seasons are shown in Figure 2 4 and Table 3, consistently exceeding 0.86, while for the test data it remains above 0.84 (Figure 2 4, Table 2 3). As in the case of annual suitability model, the seasonal predictions return extremely lo w p values, indicating that the models perform significantly better than random (Table 2 3) Relative importance of predictor variables Table 2 4 represents the relative influence of the variables on the model. The most significant environmental contributo r appears to be temperature (minimum for the general model and Jan Apr season and average for the Sep Dec season).In May Aug maximum precipitation appears as the most important predictor. These are followed by NDVI, DEM and minimum precipitation Discussio n Medfly ecological niche suitability models were constructed here using the most comprehensive dataset on its occurrence assembled to date. The distribution model outputs represent the first global assessment of the seasonally changing potential distribut ion of Medfly, illustrating the significant shifts in environmental suitability that occurs throughout a typical year. Previous mapping has rarely addressed these seasonal variations, and the output maps provide a basis for global assessments of shifting i nvasion risks.
31 The database assembled here includes the greatest number of geolocated Medfly occurrence data points compared with any other dataset available. It also includes a new level of detail regarding Medfly presence including the sampling method, h ost type and relative abundance of Medfly presence in different months. In terms of data coverage, Medfly presence is relatively well documented in Mediterranean Europe, but most data from the Middle East comes from Israel, with little information availabl e from other countries in the region. While there does exist comprehensive data on the native range of Medfly across most of Africa, few records are available from the northern part of the continent. Central and South America generally have sparse coverage with Argentina being an exception, where many comprehensive studies were performed and a large number of occurrence data is available. Only around 25% of records gathered include information about the month of Medfly occurrence. Given the environmental s ensitivity of the species and the resultant significant seasonality in distributions and abundance, future studies should ideally prioritize the collection and assembly of such valuable temporal information. The mapping outputs presented here identify area s of similar environmental suitability to those where Medfly has been found previously, both across a typical average year and by season. Some areas identified as highly suitable of course do not have Medfly population established at present and this may b e a result of either lack of introduction, eradication efforts or presence of another dominant species, like in the case of eastern Australia, where the Queensland fruit fly has displaced Medfly (Dominiak and Daniels 2012). The seasonal prediction maps ref lect changes in the environmental suitability for Medfly, and it should be noted that while the insect may be
32 able survive in the regions shown to be suitable for one or two of the three seasons mapped, it will likely not be able to become established, due to unsuitable conditions for the remainder of the year. The global annual suitability model presented here can be compared with two previous published studies on Medfly range. The most recent was performed using two approaches: a genetic algorithm for rul e set prediction (GARP) and principal component analysis (PCA) (De Meyer et al. 2008). A comprehensive native and non native dataset on Medfly occurrence was compiled and used in modeling, together with eight environmental covariates consisting of temperat ure and precipitation parameters. In the were presented, but thresholds for the division between them were not specified. The GARP model was judged to perform better by the authors; therefore I compare the results to the GARP output. It is noticeable that the MaxEnt annual model presented here tends to be more conservative and return a narrower range of Medfly suitability. It is particularly apparent in Africa and Austral ia. The models show less agreement in terms of suitability in Americas and good agreement on the suitability in Europe and al. 2002). This model inferred the climatic cond itions it can tolerate, based on the CLIMEX Growth Index. Additional modeling was then performed incorporating the effect of irrigation on Medfly abundance. Suitability was illustrated by 3 different suitability indices represented by various sized dots. I n this case, the MaxEnt annual model presented here shows substantially closer agreement in terms of the most suitable
33 areas in both the Americas and Europe, while it shows a more constrained suitability range in Africa, Australia and Asia. Significant unc ertainties in the outputs presented here remain. The models have been built on the most comprehensive dataset of Medfly presence points yet assembled, but this still has a limited amount of data in many parts of the world. Moreover, the data are often lack ing consistency in sampling methodology and possibly subject to errors in spatial and temporal referencing that are difficult to track. Further, while a detailed set of global seasonal environmental covariate datasets has been assembled and utilized in the modeling here, there are many factors that influence Medfly presence and abundance for which global spatial data do not exist these include, for example, the distribution of competitor species, the distribution of host plants, control method coverage an d produce movement patterns. Spatial data on these would likely improve modeling output fidelity and tackle some of the unexplained variance seen. Despite these caveats, the output maps represent one of the most comprehensive attempts to model the potentia l distribution and the first attempt to damaging pest species. Due to continuous efforts towards its elimination in many countries and trade and custom regulations that ai m at reducing the risk of its importation, the presence of Medfly is not necessarily continuous across a region, but fragmented. To get a better picture of the seasonal aspects of Medfly activity, the outputs presented here need to be matched and adjusted to known Medfly suitability areas, where the pest could overwinter and become established. Future work will aim to tackle this and link the
34 seasonal distribution maps presented here with seasonally changing commodity movement and human travel data to work towards building predictive models of Medfly importation risk and how it likely changes seasonally. The analyses presented here have shown how the suitability for Medfly changes throughout a typical year, but the riskiest movements of people and commoditie s for Medfly importation to suitable areas also change seasonally, and assessing reliably the risk of both Medfly importation and establishment should account for both. Such approaches can likely aid surveillance planning in prioritizing limited resources Conclusions Very few studies exist on the seasonal modeling of species distributions, and these studies are usually performed on local or regional scales. For species that are highly sensitive to environmental conditions that display strong seasonal patte rns in distributions and abundances, seasonal modeling of environmental suitability can be crucial in terms of understanding and predicting when and where a pest is most likely to be at the peak of its population activity, potentially informing targeted su rveillance at borders. Continued effort in gathering information about Medfly occurrence locations not only in spatial terms, but recording activity on seasonal scales, can serve as a tool to understand the spatio temporal population dynamics of the speci es. Increasing numbers of tools are available to model species potential distributions, and with finer resolutions of global spatial environmental datasets as well as ever increasing computing power to handle such large datasets more accurate prediction of species potential distributions can likely be performed. Even the most robust methods however, are limited in their performance where occurrence data is incomplete or scarce. Continued efforts to document Medfly and other pest species presence and make su ch
35 records available are therefore essential for improvement of our understanding and prediction of their distributions
36 Table 2 1. Data sets used in the development of ecological niche models, including source and spatial resolution. Variables in the an nual model Source Spatial R esolution Maximum land surface temperature (LST) AVHRR 5 km Minimum LST AVHRR 5 km Maximum LST AVHRR 5 km Elevation SRTM 5 km Minimum monthly precipitation total WorldClim 5 km Maximum monthly precipitation total WorldClim 5 km Mean NDVI AVHRR 5 km Table 2 2. The list of the data sets used in the development of seasonal ecological niche models. Variables in the seasonal model Source Spatial R esolution Seasonal M odel Average LST in the season AVHRR 5 km 2, 3 Average LST of the coldest month in the season AVHRR 5 km 1, 3 Average LST of the warmest month in the season AVHRR 5 km 1 Elevation SRTM 5 km 1, 2, 3 Precipitation total of the wettest month WorldClim 5 km 1, 2, 3 Precipitation total of the driest month Avera ge quarterly NDVI in the season WorldClim AVHRR 5 km 5 km 1, 2, 3 1, 2, 3 Table 2 3. Model accuracy scores Model General Jan Apr May Aug Sep Dec No. of points 463 139 158 157 Mean training AUC 0. 882 0.891 0.866 0.881 Mean test AUC 0.878 0.855 0.84 8 0.853 Test AUC standard deviation 0.012 0.025 0.025 0.022 Mean f ractional predicted area 0. 307 0.376 0.335 0.336 Training omission rate 0.0 97 0.098 0.098 0.098 Test omission rate 0.118 0.165 0.130 0.161 p value 3 948 40 2.495 9 3 795 10 1 154 8
37 Table 2 4. Relative influence of the contribution of the variables to the model. Variable General Jan Apr May Aug Sep Dec DEM 7 .4 7.9 10.2 6.1 LST min 63.7 32.5 3.4 LST max 23.3 5.8 LST avg 4.2 17.4 Prec min 0.8 9.1 3.5 1.1 Prec max 2.8 5. 9 24.7 14.5 NDVI 2 8.4 18.1 8.8 Mask 30.4 39.3 48.6
38 Figure 2 1 Occurrence data for Ceratitis capitata used in the study. Countries/regions where Medfly is present are colored with yellow and where it is currently u nder eradication are marked with green (EPPO 200 9 ).
39 Fig ure 2 2 Global environmental suitability for Medfly as predicted by MaxEnt model. Blue, purple and red colors show high confidence in predicted suitability, while yellow represents low confid ence and predicted absence.
40 Fig ure 2 3 3 panel seasonal maps showing the environmental suitability for Ceratitis capitata occurrence annually according to the MaxEnt model.
41 Figure 2 4. ROC curves representing mean AUC (red line) and mean +/ one standard deviation (blue field) Black line represents random prediction.
42 CHAPTER 3 ANALY Z ING MEDITERRANEAN FRUIT FLY SEASONAL IMPORTATION RISK INDICATORS THROUGH COMMODITY IMPORT TRAFFIC TO FLORIDA Background According to the Internation al Maritime Organization, approximately 90% of world trade is carried by sea (IMO 2011). About 75% by value of non North American trade to and from the US is shipped via cargo container (Willis and Ortiz 2004). Previous studies have shown that shipping can serve as a prominent human mediated pathway for Colizza et al. 2006; Tatem et al. 2006a; Meyerson and Mooney 2007; Westphal et al. 2008; Floerl et al. 2009; Tatem 2009; T oy and Newfield 2010; Seebens et al. 2013). Manufactured and agricultural goods, associated packaging material, cargo containers, airline baggage, mail and ship ballast water are examples of exotic pest pathways facilitated by air and ship transport (Dobbs and Brodel 2004; Work et al. 2005; Caton et al. 2006; McCullough et al. 2006; Drake and Lodge 2007; Colunga Garcia et al. 2009). Historical examples of invasions include Aedes aegypti introduction from West Africa enabling yellow fever epidemics in some o f the port locations of North America in the two last centuries (Tatem et al. 2006c). Another prominent event was the accidental introduction of the main African malaria vector, Anopheles gambiae to Brazil in 1930, which resulted in extensive malaria epid emics, costing 16,000 lives and around 3 billion USD (modern day estimate) in healthcare, drugs and the vector eradication program (Killeen et al. 2002). Modern container ships facilitated numerous alien mosquito species introductions including Aedes albop ictus a vector of multiple arboviruses including dengue, yellow and West Nile fever (Hawley 1988; Reiter 1998; Lounibos 2002; Tatem et al. 2006a), as well as chikungunya, for which it caused a recent
43 outbreak in northern Italy (Rezza et al. 2007; Tatem et al. 2012). It is believed that the Dutch elm disease arrived to the UK from Canada via the shipment of Rock Elm (Brasier and Gibbs 1973). Moreover, the invasive pathogen pine wilt disease has most likely spread across the world via contaminated wood in sh ips (Webster and Mota 2008). The risk and rates of non indigenous organism introductions is apparently increasing each year as global trade and travel continue to increase (Levine and Among a variety of pests, insect pest species are of especially high concern for the health of agriculture and the environment. It is estimated that worldwide, insect pests destroy approximately 14% of all potential food production, despite the yearly application of more than 3000 million kilo grams of pesticides (Pimentel 2007). Mediterranean fruit fly, Ceratitis capitata (Medfly) (Wiedemann) is considered one of the Americas, Australia and the Pacific region, inc luding Hawaii (Liquido et al. 1990; Papadopoulos et al. 2001b; Papadopoulos et al. 2001a; Papadopoulos et al. 2002). It has potential for tremendous economic damage and eradication costs are expensive (Cross 2004). Evidence exists that exotic pest arrival rates, commodity movements and climatic similarity between origin and destination transport network hubs show seasonal patterns (Dobbs and Brodel 2004, Caton et al. 2006, Tatem 2007, 2009, Escudero Colomar et al. 2008). How often a species is transported a nd how many individuals survive are recognized as important correlates of establishment success (Lockwood et al. 2005; Lockwood et al. 2009; Colunga Garcia et al. 2010). The seasonal population occurrence
44 of Medfly is well documented in regional studies fr om several areas of the world (Harris and Olalquiaga 1991, Harris et al. 1993, Israely et al. 1997, Katsoyannos et al. 1998, Mavrikakis et al. 2000, Papadopoulos et al. 2001, Maelzer et al. 2004, Escudero Colomar et al. 2008, Martnez Ferrer et al. 2010). Some of these studies relate seasonality to temperature and the ability of the Medfly to overwinter in the areas studies. Others link seasonality also to particular host availability (Sciarretta and Trematerra 2011). Medfly development has shown a linear r elationship with temperature (Duyck and Quilici 2002) and development thresholds have been recorded for all four life stages of the insect (Grout and Stoltz 2007). However, it is known that the pest is capable of adapting to a wide range of climates and th e larvae can overwinter inside of certain fruit hosts (Papadopoulos et al. 1996). Ports in Florida are one of the primary arrival sites for entry of maritime cargo to the USA, transporting large quantities of foreign grown fruit and vegetable products. Bo th state and federal agencies have a number of regulations that safeguard US agriculture by preventing the entry of invasive plants, animal pests and diseases. These include quarantines that prohibit entry of certain plant species, depending in some cases on origin. Some regulations require certain plants arriving from specific origins to be inspected upon departure at the port of embarkation. In the case of Florida, this may also apply to shipments arriving not only from abroad, but also different states. Because the standard of the certificate of inspections of arriving goods can vary according to the procedures of certifying agencies, inspection of imported plants is the next level of security. USDA APHIS has inspected cargo and baggage arriving at US en try points since 1972 and its predecessor conducted inspections since the early 1900s Since
45 2003 these activities are conducted by the Department of Homeland and Security Customs and Border Protection department (DHS CBP). Cargo shipments are supposed to be targeted for efficient inspections based on manifest descriptions of containers. This assessment occurs upon entry or more frequently at the departure point, which allows inspectors more time to target those products (Wasem et al. 2004). Resources for invasive species detection at the border tend to be limited. Currently, only about 2% of all commodities arriving to Florida are inspected at the border due to increasing numbers of incoming goods and limited staffing (NRC 2002; Work et al. 2005; Magarey et al. 2009). In order to minimize the risk of entry of invasive species, surveillance prioritization and gap identification are of crucial importance (Kumschick et al. 2012; Bacon et al. 2012; Bradie et al. 2013). Here I propose an approach that can aid t he identification of both routes and times of year when the long distance movement of the pest is more likely to occur. I examine seasonally changing risk factors, including that from fruit and vegetable commodity import data, evaluate their risk of Medfly infestation depending on host suitability, and combine them with the information about seasonal population dynamics at the origin points. Medfly is a very climatically sensitive organism, therefore has seasonal dynamics in its populations. These dynamics vary by origin location, as well as suitability for survival and establishment within Florida. Further, commodity demands and origins vary seasonally. Quantifying, understanding and accounting for these seasonal interactions has the potential to provide v aluable information for better guiding the tailoring of inspection and surveillance efforts
46 Data and methods A comprehensive search for Medfly occurrence records globally was undertaken, and these were assembled into a database (Szyniszewska and Tatem, i n review). The database contained 2328 unique geolocated entries on Medfly presence from 43 countries and nearly 500 unique localities. Of these, 270 localities had information on the month when Medfly was recorded. I combined these records with the two re maining largest resources of Medfly presence worldwide the Global Biodiversity Information Facility (GBIF) and Museum of Central Africa (MCA). Records from 1980 until present day were used to model seasonal environmental suitability for Medfly using Maxi mum Entropy Algorithm (MaxEnt) and a set of seasonally varying environmental variables, including minimum, mean and maximum temperature, minimum, maximum and sum of the rainfall, normalized difference vegetation index (NDVI) and digital elevation model (DE M) (Szyniszewska and Tatem, in review). Three seasons were chosen in the study: January April, May August and September December. These particular seasons were chosen, as they depicted well the seasonality in Medfly activity on both hemispheres (Szyniszews ka and Tatem, in review). The output of this work was used here to classify countries to according to its infestation risk (Figure 2 3) and EPPO records from 2001 were use d to classify countries as to whethe r Medfly is officially present Descriptive stati stics across all grid cells per country where Medfly is known to be present were calculated (mean x standard deviation s and maximum max ). According to these results I classified countries as being at low risk (weight = 0.2) if x was lower or equal to 0.1 max was lower or equal to 0.8 and s was lower or equal to 0.15. Countries were categorized as medium risk and assigned weight 0.6 if x value was less or equal to 0.3
47 and max was less or equal to 0.9. In all other instances, the country was classified as high risk and assigned risk weight was 1 (Equation 3 1 Figure 3 7 ). (3 1) Information about the volume and type of goods imported to ports of entry in Florida were obtained from the PPQ280 database maintained by APHIS. It provides detailed information about land, sea and air regulated cargo arriving to the U.S., including the total volume of goods, ports by type, origin and date of arrival. Access to the database for the years of 2003 2008 was provided by APHIS. Initially I consider ed examining the volume and frequency of both the maritime and air cargo data, however no hosts associated with Medfly were shown to be transported via air, and therefore this pathway is not included in the analysis. Only the maritime commodity arrival rec ords to ports in Florida were used in the analysis shipments with further destinations outside of Florida were excluded. Moreover, only fruit and vegetable imports were considered, as this group of goods may serve as Medfly host and therefore is most lik ely to be associated with its arrival. The majority of commodities had kilogram listed as their unit of weight; however there were a few instances where the amount of shipped cargo was listed in boxes, or individual units. I applied a correction to these u nits to represent as closely as possible to the likely weight in kilograms, so the units were uniform across all the commodities used in the analysis. I assigned each commodity a category for the potential to serve as a Medfly host (k). Two sources to det ermine the species host potential were used: EPPO and the Featured Creatures website maintained by the University of Florida (Thomas et al.
48 2001; EPPO 2001, 2009). Commodities with heavy or main infestation risk were classified as Main, and all other poten tial hosts (either occasional, unknown, laboratory) were categorized as Occasional/Rare. Plant species not listed in any of these two sources were classified as posing no risk of Medfly infestation (None). For each of these respective categories, I assigne d a weight to represent the potential of the commodity being infested (None = 0, Rare/Occasional = 0.5, Main = 1). These arbitrarily chosen weights do not represent a statistical likelihood of infestation with Medfly. There have been few studies that have estimated the approach rates of nonindigenous species on specific pathways associated with cargo. These can also be very dynamic and come with a high degree of uncertainty. Due to a vast number of host species, and multiple factors that can influence Medfl y feeding preferences in diverse landscapes in a global setting, and additionally the lack of reliable estimate of infestation rates from the incoming cargo, assigning such weights was not feasible. The chosen weights facilitate a simple calculation of the arriving cargo risk indicators, depending on the origin points, season and port of entry. For each port in Florida ( i ) and each origin point ( l ) I calculated the total cargo volume m il I further broke down this volume by the three seasons of interest ( j ) to receive m ijl The risk indicators of commodities arriving to port i in season j were defined as PR I ij The risk indicators for commodities arriving from an origin location l in season j were defined as OR I j l The risk indicator for a port i in season j from origin l are defined as POR I ij l h k stands for a commodity host suitability, o lj stands for the seasonal suitability at the origin country, m stands for the commodity tonnage, where m ijlk is the total of commodity tonnage for a port of entry I seas on j and origin l (Equation 3 2).
49 (3 2) In order to compare the highest volumes and risk routes with the survey results at the border, I examined the Medfly interception data at the US ports of entry from the Port Information Network (PIN) database which tracks interceptions in agricultural commodities in cargo, air passenger baggage and other conveyances. The database dates back to 1984 and was maintained by USDA APHIS PPQ by the year 2003. Starting in March 2003 DHS took over the in spection activities, and PIN was renamed as PestID. I received access to the records from the 2003 2008 time period, however I excluded the data from 2003 due to the risk of erroneous data inputs due to major procedural changes. The database provides very detailed information about the ports of entry and origins and types of commodities where the pest has been found, species name, taxonomic order, date, name of pest, the part of the plant where the pest was found and various other information. It is limited to information about cases where an actionable pest has been found. It is important to note, that due to lack of systematic sampling scheme, and varying detection priorities due to changing commodities and pests of concern and additionally lack of records of negative inspection results, these data are not statistically valid for estimating approach rates of nonindigenous spe cies into US (Work et al. 2005)
50 Results The main exporters of fruits and vegetables (FV) to Florida via maritime pathways include mai nly Central and South American countries, with Costa Rica, Honduras and Guatemala being the leaders, followed by Belize, Peru, Ecuador and Dominican Republic (Figure 3 1). The main exporter outside this region is China. Lower volumes of FV imports arrived from European countries and the Middle East, with the least from the African continent (Figure 3 2). The largest share of imported FV constitutes species classified as occasional or rare Medfly hosts. The largest volume of imported FV classified as the mai n Medfly hosts arrived from Peru, Bahamas, Guatemala, Ecuador and Haiti. The largest share of commodities classified as most popular (main) Medfly hosts per total FV exported tonnage arrived from Bahamas, Haiti and Peru, while relatively high absolute numb ers were shipped from Guatemala. Between all the ports of entry in Florida, Port Everglades received the highest share of all the fruit and vegetable maritime imports to the state in the study period by ton nage. It was followed by Miami Seaport Tampa and Port Manatee, while the remaining ports received significantly less tonnage of FV imports (Figure 3 3). The east coast ports including Ft. Pierce, Miami, Fort Lauderdale/Port Everglades, West Palm Beach and Jacksonville received the highest ratio of commod ities that are the main or occasional Medfly hosts, with Fort Pierce receiving the vast majority of imports at high risk of Medfly infestation, due to host suitability (Figure 3 4). There was significant variability in the tonnage of imported FV arriving t o Florida on intra annual and interannual scales (Figure 3 5).The highest inflow of goods is observed between January and April (with a peak in March). Consistently, the second half of the year, corresponding to summer and fall in Florida, receives the low est share
51 of FV annually. On interannual scales I observed that the majority of years have low interannual variability of import volume in the second half of the year. The biggest variability was observed between December and April. This trend was persiste nt in various years across all the major ports of entry in Florida (Figure 3 6). A significant seasonality in the changing risk of Medfly potential infestation at the origin points, mainly in Europe, but also in Australia, and some locations of South and C entral America was notable. Three seasons (January April, May August and September December) are identified in the panel of maps in Figure 3 7. In each country, where Medfly wa s present according to EPPO (2001) the seasonal presence risk is classified as low, medium and high according to presented earlier classification. The pie charts represent the volume and type of FV exported to Florida from various countries. The size of the charts represents relative difference between larger and smaller volume of th e incoming commodities total. As the differences were too large to proportionately represent them on this map, the size of circles helps to rank the exporters but is not directly proportional to the volume of exports. The ratio in commodities at risk that are exported from South America is subject to substantial seasonal variability, especially in the cases of Peru, Ecuador and Colombia, when the highest share can be observed in the first four months of the year. Risk indicators for commodities arriving to each of the ports of entry in Florida were calculated according to season (Table 3 1). Obtained metrics show the relative ratio of the total number of commodities in each season that are likely to have an increased risk of Medfly infestation. I see that t he largest share of annual imports arrives into Florida in the first season, and for Port Everglades and Miami Seaport this is also
52 the month with the highest ratio of higher risk of Medfly entry. In some instances, the calculated risk in the first trimest er is extremely low, or is zero, as in the case of West Palm Beach, Fort Pierce, Fort Lauderdale and Fernandina Beach. The port in Tampa has very high risk values for the second and third trimester, and none for the first trimester, where the risk is null. In several ports the risk remains very low all year round including West Palm Beach and Fort Pierce. The overall highest risk indicator for Medfly importation in the study period can be seen for Panama City. Three major entry ports sustain a relatively high risk annually, with high variability e.g. in Tampa the risk in the first trimester is null, while in the second is 0.48 and in the latter 0.28. Significant variability in the risk indicators was observed for various countries of origin (Table 3 2). For the main exporters of FV to Florida (Costa Rica, Honduras and Guatemala), the risk is relatively high throughout the year and decreases in the last trimester. Peru has an extremely high RI of importation of goods carrying Medfly in the first trimester of the year, while Mexico in the second. Panama displays the lowest risk in the first trimester of the year, with even lower risk in the latter two. Colombia presents no risk in the third trimester of the year. Pairs of ports of entry and origin countries with the highest tonnage travelled on routes include Port Everglades Honduras, where the risk indicators remain in the range of 0.45 0.46 during the first two trimesters, and slightly drops in the third one (0.26) (Table 3 3). The second route with the h eaviest tonnage is Costa Rica Tampa with risk indicators around 0.50, decreasing in the third trimester. Several other routes with large volume of exported FV show a higher variability in seasonal risk indicator
53 values. I can see elevated increased risk from Peru to Miami Seaport in the first trimester, compared to the two following trimesters of the year Discussion Because of multiple transportation links and a mild climate, the major industry of Florida agriculture is highly vulnerable to invasions The risk of arrival of one of the Medfly is persistent, the cost of eradication is enormous and the risk of establishment is real. Currently pest risk assessments (PRA) rely on general information on which origin locations and comm odities carry high risks of pest infestation (Campbell 2001). This helps the border authorities to target shipments from specific origins and specific goods on them. However, the seasonal nature of the population dynamics of various species remains often u naccounted for, and a deeper understanding of how various seasonal factors can interact, may enhance surveillance prioritization and efficiency by guiding a strategic allocation of limited resources. I have shown here that for some pairs of origin countrie s and ports of entry, key risk indicators can change sharply depending on the season, whereas for others, this risk remains proportionally equal, even for seasons with highly variable seasonal volumes of incoming FV. Some of the major exporters from Latin America sustain a high climatic suitability for Medfly annually, but I can observe a high variability in the volume and type of shipped goods depending on the season. As in the case of Peru, the ratio of goods serving as main Medfly hosts is the highest in the first trimester, and consequently there is a sharply increased risk of Medfly importation in this trimester to various ports of entry in Florida. In Nicaragua, the lower risk of Medfly importation in the second and third trimester can be attributed to a lower ratio of Medfly suitable goods, and lower environmental suitability in the third trimester compared to the first trimester
54 of the year, when the risk is overall significantly higher. I can also observe a high seasonal variation in interaction betw een environmental suitability and the types of items being shipped from Ecuador. The first trimester in this country has a lower environmental suitability compared to the latter two, but relatively high amount of items suitable as Medfly host are being exp orted during this time of year. The second trimester has a higher environmental suitability, but the majority of shipped items are classified as not suitable or rarely suitable as Medfly hosts. Consequently, decreased RI values are observed in the second t rimester, medium in the first and highest in the last. It is important to acknowledge several sources of uncertainty in the modeling undertaken here. Firstly, while I have made use of some of the most comprehensive datasets available, the cargo movement da ta to Florida was limited, with the most recent records being from 2008. The dynamics of trade, including changing trade agreements, differences arising from changing commodity prices and supply and demand, mean that quantities have likely changed in more recent years. A more recent picture of the commodity movement could enhance the analysis presented here. Additional sources of uncertainly include the lack of knowledge about the actual infestation rates of commodities and error that is inherent to the Med fly seasonal potential presence model. Due to the lack of a systematic sampling scheme at the border I was unable to validate my model against existing interception data. This type of approach however, offers potential for identifying inspection gaps. The models have been built on the most comprehensive dataset of Medfly occurren ce points yet assembled (Szyniszewska and Tatem, in review), but this still has a limited amount of data in many parts of the world. I do not have information about the spatial loca tion of
55 fields where almost 300 plants known as Medfly hosts are grown. Therefore, I could only rely on the political boundaries of regions where Medfly is known to be present. It is clear that great seasonal variations in Medfly presence around the world and commodity imports to Florida occur and there remains significant potential to utilize such information in allocation of inspection and surveillance resources. Various origin points, commodity types and seasons can be ranked according to their seasonal risk, which can result in recommendations on proportionately higher or lower inspection focus at different times of year. Commodities at higher risk in a certain season could be subjected to more intensive surveillance rates, while the pathways ranked as l ow could be subjected to minimal surveillance and the remaining resources used for higher risk pathways. This methodology could be applied for any other invasive organism of concern, which is subject to seasonal population dynamics that result in significa nt variability in population densities, and has seasonal variations in host commodity import rates, and also to any other state in the country, or country in the World Conclusions Quantifying and accounting spatially for various seasonal aspects of commod ity flow to ports of entry can serve as a useful tool in order to prioritize and optimize the border surveillance. For species that are highly sensitive to environmental conditions that display strong seasonal patterns in distributions and abundances, and in markets subjected to seasonally varying demand, the seasonal modeling of import risk can be crucial in terms of predicting not only the highest risk pathways and origin points but also the time of year when the risk is the highest. Such an approach has the potential to be applied to any invasive insect species, or any organism subject to seasonal
56 population dynamics and density in any country concerned with invasive organism introduction
57 Table 3 1. Seasonal risk indicators for commodities entering p orts of entry in Florida (PRI) from Medfly infested countries according to EPPO (200 1 ). The overall ratio is an annual risk indicator for certain port. Port of Entry Jan Apr May Aug Sep Dec Total Tonnage Total Ratio Port Everglades 0.42 0.35 0.22 4074459. 2 0.35 Tampa 0.00 0.48 0.28 1625260.1 0.42 Miami Seaport 0.34 0.27 0.24 1426654.1 0.31 Port Manatee 0.48 0.50 0.30 659801.5 0.42 Cape Canaveral 0.34 0.35 0.25 125924.5 0.30 West Palm Beach 0.00 0.05 0.03 105993.8 0.20 Ft. Pierce 0.00 0.00 0.00 56081. 3 0.00 Panama City 0.50 0.53 0.50 46686.5 0.50 Jacksonville 0.20 0.37 0.18 1994.4 0.22 Ft. Lauderdale 0.00 0.50 0.00 117.4 0.50 Fernandina Beach 0.00 0.30 0.25 44.9 0.29
58 Table 3 2. Risk indicator for commodities arriving from international origin locations (ORI). The overall ratio is an annual risk indicator for certain country. Origin Jan Apr May Aug Sep Dec Total T onnage Overall R atio Costa Rica 0.44 0.47 0.27 2399155 0.41 Honduras 0.44 0.45 0.26 2239555 0.41 Guatemala 0.41 0.40 0.23 1942865 0 .35 Peru 0.86 0.30 0.22 192101 0.46 Ecuador 0.55 0.42 0.58 191366 0.53 Panama 0.24 0.48 0.43 125303 0.32 Colombia 0.51 0.28 0.00 68963 0.35 Mexico 0.51 0.60 0.56 53702 0.53 Nicaragua 0.44 0.28 0.15 40824 0.35 Brazil 0.11 0.30 0.64 36836 0.38 Venezu ela 0.49 0.50 0.45 22027 0.48 Argentina 0.04 0.23 0.30 11128 0.11 El Salvador 0.43 0.15 0.19 10175 0.26 Spain 0.30 0.00 0.00 331 0.02 Guinea 1.00 0.00 1.00 111 0.44 Cote D`Ivoire 0.00 0.50 0.00 18 0.50
59 Table 3 3. Highest total tonnage origin destin ation pairs and their seasonal port of entry origin risk indicators (PORI). Port of Entry Origin Country Jan Apr May Aug Sep Dec Total Tonnage Overall Ratio Port Everglades Honduras 0.45 0.46 0.26 1,787,196 0.41 Tampa Costa Rica 0.47 0.50 0.30 1,115,58 4 0.44 Port Everglades Guatemala 0.44 0.44 0.24 927,991 0.38 Port Everglades Costa Rica 0.38 0.39 0.22 603,018 0.34 Port Manatee Costa Rica 0.48 0.50 0.30 532,246 0.43 Tampa Guatemala 0.41 0.37 0.24 484,500 0.36 Miami Seaport Honduras 0.41 0.37 0 .26 405,460 0.40 Miami Seaport Guatemala 0.33 0.27 0.17 317,797 0.28 Miami Seaport Costa Rica 0.34 0.40 0.23 144,049 0.33 Cape Canaveral Guatemala 0.34 0.35 0.22 115,570 0.30 Miami Seaport Peru 0.81 0.32 0.28 114,825 0.42 Port Everglades Ecuador 0.56 0.38 0.53 99,602 0.51 Port Manatee Guatemala 0.47 0.50 0.28 96,985 0.39 Port Everglades Peru 0.89 0.14 0.09 75,658 0.54 Miami Seaport Panama 0.24 0.47 0.42 75,257 0.31 Miami Seaport Ecuador 0.57 0.40 0.67 60,388 0.58 Port Everglades Panama 0 .23 0.49 0.44 47,104 0.33 Panama City Mexico 0.50 0.53 0.50 46,686 0.50 West Palm Beach Honduras 0.48 0.43 0.30 44,853 0.47 Port Everglades Colombia 0.50 0.29 0.27 36,149 0.35 Miami Seaport Nicaragua 0.40 0.20 0.06 26,180 0.31 Port Manatee Ecuado r 0.50 0.50 0.50 21,707 0.50 Miami Seaport Brazil 0.38 0.35 0.71 19,707 0.57 Port Everglades Brazil 0.05 0.13 0.44 16,959 0.14 Port Everglades Nicaragua 0.54 0.46 0.21 13,733 0.42 Port Everglades Venezuela 0.50 0.50 0.44 12,467 0.48 Tampa Colombi a 0.50 0.30 0.30 12,296 0.34 Miami Seaport Colombia 0.55 0.17 0.18 11,113 0.35 Tampa Ecuador 0.50 0.50 0.58 9,669 0.51 Miami Seaport Venezuela 0.47 0.49 0.46 9,495 0.47
60 Figure 3 1. Top maritime importers of fruits and vegetables (FV) to Florida (200 4 2008) according to the total volume in tonnes. Commodities are classified according to their Medfly host status. Fig ure 3 2. The total tonnage of fruits and vegetable (FV) imported to Florida between 200 4 and 200 8 from worldwide origin countries via maritime pathway.
61 Fig ure 3 3. Total volume of fruit and vegetable maritime imports in each Florida ports between 200 4 2008.
62 Fig ure 3 4. Share of maritime fruits and vegetable imports that carry main, occa sional or no host for Medfly classified by total tonnage for 200 4 2008.
63 Fig ure 3 5. Intraannual and interannual variability in the incoming volume of FV imports to Florida via maritime pathway between 200 4 and 2008. Fig ure 3 6. Intraannual var iability in the total volume of incoming maritime fruit and vegetable (FV) imports to Florida per port of entry (200 4 2008).
64 Figure 3 7. Seasonal risk factors for Medfly importation.
65 CHAPTER 4 ANALYZING SEASONAL IMPORTATION RISK INDICATORS FOR MEDITERRANEAN FRUIT FLY VIA AIR PASSENGER TRAFFIC TO FLORIDA Background Traffic connections greatly enhance both intended and unintended spread of organisms, including invasive pest species (Klassen et al. 2002; Drake and Lodge 2004; Tatem et al. 2006a; W estphal et al. 2008; Hulme 2009). Air travel has seen the most rapid rise in recent decades with a steady growth at about 5.9% per annum in 2011 (IATA 2012). There are about 30 million scheduled flights per year with the number of passengers transported an nually exceeding 2.8 billion (IATA 2012). The increasing reach of the air traffic network and volumes carried on it are being mirrored by increasing rates of migration and dispersal of species. Disease carrying mosquitos have been shown to have survived lo ng haul flights in aircraft cabins (Lounibos 2002; Tatem et al. 2006b; Benedict et al. 2007) and many invasive pest species are being found in both cargo and passenger luggage (Work et al. 2005; Liebhold et al. 2006; McCullough et al. 2006). While air trav el means that distance is posing a rapidly diminishing obstacle to the spread of invasive pests, the number of individuals travelling on a route (a conditions an invasive species encounters are stil l fundamental constraints to establishment (Levine and Lockwood et al. 2009; Tatem 2009). Mediterranean fruit fly, Ceratitis capitata (Medfly) (Wiedemann) is considered one of Americas, Australia and the Pacific region, including Hawaii (Liquido et al. 1990; Papadopoulos et al. 2001b; Papadopoulos et al. 2001a; Papadopoulos et al. 2002). It
66 has the potential for tremendous economic damage and eradication costs are hugely expensive. It is estimated that the cost of each of its previous multiple incursions into the US (eradication and industry loss) ranged from $300,000 to $200 million and the cos t of potential establishment is estimated at $821 million or more per year (APHIS 1992). Medfly outbreaks in California during the past 25 years have cost taxpayers nearly $500 million, while the Medfly outbreak in Florida's Tampa Bay region in 1997 result ed in $25 million spent on eradication (Cross 2004).. While the international air travel network is constantly expanding and traffic on it increasing also, the resources for surveillance and in consequence the rates of organism interceptions at borders ha ve not kept up this the steady rise (Klassen et al. 2002; McCullough et al. 2006). US border agencies maintain a record of pest interceptions at the border, known as PestID database (formerly PIN). Between 1984 and 2000, 725,000 pest entries were recorded by this database. Among those as much as 73% occurred at airports and only 9% at marine ports (McCullough et al. 2006). More than half of those interceptions were associated with small parcels and baggage carried by travelers In total, insects represented 73.5 to 84.6% interceptions each year. Miami, JFK and Los Angeles airports accounted for 43% of all interceptions. In Florida, 69% of interceptions occurred on shipments that arrived from South and Central America and 22% from the Caribbean. Roughly 62% w ere associated with baggage, 30% associated with cargo and 7% with plant propagative material. Dobbs and Brodel (2004) undertook a study that surveyed randomly selected cargo aircrafts arriving to Miami International Airport between September 1998 and Augu st 1999 in order to look for hitchhiking
67 passenger luggage is known to be a major pathway of Medfly entry into United States, most commonly through transported fruits (Liebhold et al. 2006). The numbers of interceptions are positively related to the volume of traffic from that country and negatively associated with the GDP of that country. It has sustained a steady level of arrival to US that can explain the repeated d etection of this insect in southern California even after its repeated eradication (Liebhold et al. 2006). Evidence exists that the arrival rates of many invasive organisms and population dynamics at their origin locations show strong seasonal patterns (C aton et al. 2006; Liebhold et al. 2006; Escudero Colomar et al. 2008). Accounting for seasonality in tailoring passenger luggage screening and surveillance strategies, may enhance efficiency in utilization of limited resources at the border. This research aims to quantify the seasonally changing risks of importation of one of the most economically significant pest species to Florida, a state that is highly vulnerable to economically damaging biological invasions due to its mild climate, strong international transport links and its main industry being agriculture. Here I utilize global spatial datasets of relevance to Medfly historical presence, lifecycle, environmental preferences and a global passenger air travel flow model, to estimate how various risk fac tors interplay in Medfly transportation risk. I account for seasonally changing Medfly environmental suitability at the airport of origin and the rates of international arrivals from various origins to the major airports in the state of Florida. Methods Pa ssenger flow data and airport locations Airport coordinates, city name and IATA airport code were obtained for 3,416 airports from the Flightstats ( www.flightstats.com ) website. Information on flight routes
68 and scheduled seat capacity on each flight by month for 2010 were purchased from OAG International ( www.oag.com ). Directly connected airports pairs were utilized to construct a graph of international air travel connections to Florida in 2010, consisting of 103 nodes and 133 edges. This dataset is however limited in estimating the actual passenger flow, as flights often do not operate at full capacity and therefore it tends to overestimate the actual flow. Moreover, it provides information only on point to point connections and thus itineraries requiring transfers are not captured (Johansson et al. 2011; Huang and Tatem 2013; Huang et al. 2013). Actual passenger flow data exist but remain costly and difficult to obtain for research purposes due t o prohibitive costs and significant confidentiality and legal restrictions. Therefore, for the passenger flow analysis I utilized an open access modeled passenger flow matrix for the global air network in 2010 (Huang et al. 2013).In brief, travel volumes o n routes were modeled based on primarily publicly available datasets under a generalized linear model network. To reflect the complex hierarchical network in which cities are situated, G Econ data for local area Purchasing Power Parity (PPP) per capita wer e utilized ( http://gecon.yale.edu/ ) as covariate to reflect the economic links of the airport region. Due to computational power limitations, only airports serving cities of populations of 100,000 or more were includ ed in the model. Predicted flows for routes were constructed based on the adjacency matrix defined by the OAG dataset, and actual travel volumes for training and validation were extracted and assembled from various transportation organizations in the USA, Canada and the European Union. (Huang et al. 2013).
69 A subset of 813 airports which connect to one of the airpor ts in Florida via modeled air traffic flow was extracted from the passenger flow matrix (Huang et al. 2013). This included the origin, the number of stops (between 0 to 2), the final travel destination and the estimated number of passengers taking this route annually. Medfly distribution Medfly seasonal environmental suitability maps were used in order to estimate s surrounding each airport of origin (Szyniszewska and Tatem, in review). Records from 1980 until present day from the most up to data collection of data on Medfly presence were used to model seasonal environmental suitability for Medfly using the Maximum Entropy species distribution modeling algorithm (MaxEnt) combined with a set of seasonally varying environmental variables, including minimum, mean and maximum temperature, minimum, maximum and sum of the rainfall, normalized difference vegetation index ( NDVI) and digital elevation model (DEM) (Szyniszewska and Tatem, in review). Three seasons were chosen in the study: January April, May August and September December. These particular seasons depict well the seasonality in Medfly activity on both hemispher es. The output of this work was utilized here to classify areas around each airport in the countries where Medfly is officially present (EPPO 2001) according to its infestation risk. In order to establish a risk weighting, a 200km circular buffer was draw n around each airport location, and descriptive statistics across all grid cells per buffer were calculated (mean x standard deviation s and maximum max around airports) were classified as being at low risk for Medfly presence (weight = 0.2) if x was lower or equal to 0.1, max was lower or equal to 0.8 and s was lower or equal to 0.15. Airport regions were categorized as medium risk for Medfly presence and
70 assigned weight 0.6 if x value was less or equa l to 0.3 and max was less or equal to 0.9. In all other instances, the airport region was classified as high risk for Medfly presence and the assigned risk weight was 1 (Equation 4 1). (4 1) Risk indicator calculations Three risk ind icator values were calculated for an airport of origin, destination and a pair of origin destination locations in order to assess the relative ratio of passengers at risk and compare it between three seasons. The values were obtained by weighting the actua l passenger flow by the Medfly presence risk at the origin, and dividing it by the passenger flow number on the route. For each airport in Florida ( i ) and each passenger itinerary start location ( l ) I calculated the passenger flow p ij I further divided th is number by the three seasons of interest ( j ) to receive seasonally adjusted flow p ijl The risk indicators for passengers arriving from all international origin locations to a Florida airport i in season j were defined as AR I ij The risk indicators for c ommodities arriving from an origin airport l in season j were defined as OR I jl The risk indicator for an airport i in season j from origin l are defined as AOR I ijl o jl stands for the seasonal presence risk at the airport of origin, p stands for the passe nger number, where p ijl is the total passenger numbers at a port of entry I in season j and for origin l (Equation 4 2).
71 (4 2) Interception data In order to compare the highest volumes and risk routes with the border inspection surv eillance, I examined Medfly interception data at US ports of entry from the Port Information Network (PIN) database. The database tracks interceptions in agricultural commodities in cargo, air passenger baggage and other conveyances. The database dates bac k to 1984 and was maintained by USDA APHIS PPQ up until 2003. Starting in March 2003 the Department of Homeland Security (DHS) took over the inspection activities, and PIN was renamed as PestID. I received access to the records from the 2003 2008 time peri od. The database provides detailed information on the ports of entry and origins and types of commodities where the pest has been found, species name, taxonomic order, date, name of pest, the part of the plant where the pest was found and various other inf ormation. It is limited to information about cases where an actionable pest has been found. It is important to note that due to the lack of a systematic sampling scheme, and varying detection priorities due to changing commodities and pests of concern and additionally lack of records of negative inspection results, these data are not statistically valid for estimating approach rates of
72 statistics was used to compare the int erception numbers with the Medfly presence risk weight adjusted passenger volume from specific countries for each season. Results Seat capacity on direct flight connections to Florida Miami airport contained the highest number of international direct conne ctions (77), followed by Orlando International Airport (29) and Fort Lauderdale (12). This order is also reflected in seat capacity on international connections, with Miami (MIA) having about 11.7 million seats on incoming flights in 2010, including over 6 million seats annually from countries where Medfly is officially present. Three additional airports maintain direct connections with countries where Medfly is officially present Orlando (over 400,000 seats annually, representing 17% of seat capacity on international connections) and Fort Lauderdale (63,000, representing only 6% of total seat capacity on international routes). The traffic in terms of seat capacity on individual international direct connections to Florida in 2010 is shown in Figure 4 1. Th e highest number of seats on scheduled flights comes from Canada, United Kingdom and Puerto Rico. There are multiple connections with the Caribbean region, Central and South America and Western Europe. The highest seat capacities on direct connections from Medfly infested locations come from Mexico, Brazil and Colombia. I plotted the cumulative seat capacity on the routes with the highest seat capacity volumes by trimesters and concluded that no significant seasonal trend can be observed (Figure 4 2). Passe nger flows The passenger flow model provides a much broader picture of the actual flow of passengers to Florida from international locations. The model includes connections to Florida airports from 666 origin locations worldwide, including 350 locations f rom Medfly
73 infested countries. The network contains a total of 674 nodes (i.e. points) and 2797 edges (i.e. links), including 1453 edges between Florida airports and departure points located within countries or regions where Medfly is established according to EPPO (200 1 ). The countries with the highest number of passengers arriving to Florida each year include Colombia, Canada, Haiti, United Kingdom and Mexico. Notably, there is a high flow of passengers from countries where Medfly is established, such as Brazil, Ecuador, Costa Rica, Guatemala, Honduras, Bolivia and Panama (Figure 4 3 and 4 4). The airports with the highest number of arriving passengers (over 500,000 annually) include Miami, Orlando and Fort Lauderdale. Tampa, Sanford, St. Petersburg, Talla hassee and Gainesville receive considerably less international passenger arrivals (below 250,000 each). Figure 4 5 shows the estimated annual international passenger flows to selected airports as estimated by the flow model and considering the seat capacit y on direct connections only. Miami receives the highest share of arrivals, followed by Orlando and Fort Lauderdale. The lowest number of arrivals who started their journey outside of the US is received by smaller airports, including St. Petersburg, Tallah assee and Gainesville. Seasonal risk indicators and passenger flow Considering the passenger volume flow adjusted by the seasonal risk of Medfly presence at the origin from international origin points, as a share of total international arrivals, there is a high variability in the ratio of passengers at risk vs. passengers at no risk between various ports (Figure 4 6). The major arrival airport of Miami receives Gainesville an d Tallahassee. Fort Lauderdale has a ratio of less than 40%, with the lowest number in the first trimester of the year, while Orlando and St. Petersburg show
74 values of 29% and 24% respectively, with similar ratios year round. The lowest numbers of passenge rs at risk arrive into Tampa and Orlando Sanford airport. Figure 4 7 shows a 3 panel map illustrating the risk adjusted volume of passenger flows to Florida per trimester. The modeled seasonal Medfly environmental suitability at the origin countries or re gions where Medfly is officially present according to EPPO (200 1 ) is shown. Origin airports classified according to the risk level of seasonal Medfly presence and their seasonally weighted flow of passengers at risk is also shown. The model identifies 812 origin locations from which passengers arrive to Florida, and I plot only those in Medfly infested countries. Seasonal changes in the Medfly environmental suitability at the port of origin are combined with the changing flow of origin weight risk adjusted passenger volumes on the pathways. I observed a decreased volume of passengers from Medfly suitable regions in Europe between January and April, due to less suitable climatic conditions for Medfly development during those months. This risk increases to med ium and high in the following two trimesters of the year. Most of the Central and South American origin regions maintain high or medium environmental suitability for Medfly all year round. Most of the origin regions located in Central America and eastern B razil have environmental suitability classified as medium between September and December, and therefore the volume of passengers at risk is lower compared to the rest of the year. The majority of African origin regions also sustain a high level of risk for most of the year, with some variability in Western Africa, when the risk is at medium level for most of the year in northern locations. It is clear that all regions are relatively well connected with Florida, both in
75 terms of the number of itineraries and the volume of passengers travelling between those locations. Summary tables present the risk indicator scores for origin locations (ORI), Florida destination airport (ARI) and pair of origin destination ports (AORI) respectively (Table 4 1, 4 2 and 4 3). In Table 4 1 the origin risk indicators (ORI) for the main international airports in Medfly infested countries to Florida according to the total annual passenger numbers are presented. The airports are divided according to several world regions. The highes t volume of passengers to Florida comes from Central and South America. Relatively lower passenger numbers arrive from Europe, Africa and the Middle East. As a corollary to results presented in Figure 4 7, the locations in South America tend to maintain hi gh suitability at the origin locations all year round, while many origin regions in Central America have a lower suitability between September and December. Europe, Africa and the Middle East display the highest variability in seasonal suitability for Medf ly. Table 4 2 presents the airport risk indicators (ARI), which represents the risk adjusted origin Medfly suitability volume of passengers arriving to each airport. The first set of risk indicators applies to the seasonal origin risk adjusted volume of pa ssengers compared to the total volume of passengers from locations where Medfly is present, and the second set of risk indicators compares the risk adjusted volume to the total volume of international arrivals. Miami (MIA) not only has the highest volume o f international arrivals (over 2 million) but also the highest ratio of arrivals from Medfly present origin locations (0.67). Gainesville airport has the lowest number of international arrivals, but does have a similarly high ratio (0.63). Orlando (MCO) an d Fort Lauderdale
76 (FLL) follow Miami in terms of the number of international passenger arrivals, however the ratios of arrivals from Medfly present countries are much lower (0.31 and 0.42 respectively). While FLL has ARI are close to 1 between January and August, this risk becomes lower in the last four months of the year. Miami sustains a relatively stable risk all year round, with a slight decrease in the fall. The lowest risk in the first four months of the year is at Orlando Sanford Airport (SFB) (0.71) with even lower risks shown when the volume of passengers at risk is compared to all international arrivals (0.09). Table 4 3 presents AORI scores for chosen pairs of destination airports and some of the main origin regions for each. Most of the major or igins include locations in Central and South America (mostly Guatemala, Mexico, Costa Rica and Colombia), but also locations in Europe (Paris, Zurich, Tel Aviv) and Africa (Kinshasa and Bamako). Interception data The total number of interception records fo r Medfly in Florida for 2003 2008 displays high seasonal variability with two peaks in May and July (Figure 4 8). Medfly sustains a relatively high arrival rate to the US, including Florida, mostly via the means of passenger luggage (Liebhold et al. 2006). The most common hosts include peach ( Prunus persica ), citruses and Annona muricata (Table 4 4). The most common origin of intercepted luggage with Medfly infested items is Europe (59 interceptions), South America (26), Central America (15) and North Afric a and the Middle East (4). The highest number of interceptions from Europe occurred in July (42), with some in August (1), November (1), March (3) and May (12). Unsurprisingly, weak correlations (r<0.15) between the number of insect interceptions and the M edfly presence risk adjusted volume by season were found.
77 Discussion Increasing the effectiveness of border inspection targeting efforts under limited resources and increasing global travel will be reliant upon an improved evidence base and understanding o f pest importation dynamics. This study has focused on one of the most economically significant pests, that is sustaining a high level of arrivals to Florida, where it can potentially become established and cause substantial economic damages as a result. M edfly has been known to arrive into Florida via shipped commodities and air passenger luggage. Previous analyses have examined commodity trends (Chapter 3) Here I have utilized a global air passenger flow matrix to estimate the arrivals from various world destinations to the airports in Florida. Further, because Medfly is very climate sensitive and its population dynamics rely heavily on weather conditions, I have examined the seasonal environmental suitability at origin airport regions according to their risk of Medfly presence and population density at a particular time of year. Through combining incoming passenger volumes with climatic suitabilities, I can shed light on the varying proportions in terms of volume and risk of airline baggage that may arriv e to destinations in Florida from Medfly present origins for three different seasons of the year. The results highlight how well connected Florida is to the majority of Medfly he number of passengers travelling) comes from locations in Central and South America, as well as the Caribbean. The Caribbean region is officially Medfly free at present, and therefore excluded from the analyses. Locations in Central and South America hav e the highest volume of passenger flow to Florida and also display highest levels of environmental suitability all year round compared to other regions, and a relatively high
78 number of Medfly have been found on fruit in passenger luggage from these regions Relatively smaller numbers of passengers and significantly lower environmental suitability is observed for Europe. However, the highest number of pest interceptions according to PestID database comes from this continent and the majority of interceptions occurred in the summer, which is expected. Three interceptions that occurred in early spring month (March) arrived from the UK and could possibly be not correctly classified the goods carried most likely arrived from another location. My results rely on two modeling approaches, and each are subject to a range of uncertainties and limitations. The first set of uncertainties stems from the seasonal Medfly environmental suitability model, which is built on the most comprehensive dataset of Medfly presence p oints yet assembled, but is likely still missing information from many locations, is built on data that may be prone to errors in the reported spatial and temporal references, as well as lacking consistency in sampling methodology. Further, while a detaile d set of global seasonal environmental covariate datasets has been assembled and utilized in the modeling here, there are many factors that influence Medfly presence and abundance for which global spatial data do not exist, i.e. the distribution of competi tor species and host. Finally, the seasonal variation in environmental suitability is subject to interannual variation and the presented results represent merely the mean values. A second set of uncertainties stems from limitations in the global passenger flow matrix model. Firstly, the model represents a 2010 snapshot and therefore does not account for recent changes in routes and increases in passenger volumes. Secondly, the model assumes that the passengers will utilize the
79 nearest airport, but this is n ot always the case. Moreover, longer haul international flights with 3 or more connections are not represented A lack of information and deep understanding on the mechanisms governing Medfly dispersal also represents a limitation. Firstly, it's likely th at the number of passengers arriving from certain locations does not directly translate to the amount of carried luggage, or carried fruits and vegetables within this luggage. Regional variations travel and trip length are likely to exist (Liebhold et al. 2006) While the border interception data provides valuable information on possible pathways, due to the lack of systematic sampling scheme it is unfortunately subject to a range of biases. First ly, passengers from some regions may tend to be screened more by border authorities. Also, inspectors may focus their attention on inspecting certain types of fruits from Central and South America commonly infested with other types of pests of concern (e.g Anastrepha ) and thus be less likely to inspect citruses for Medfly presence (APHIS 1992). Moreover, the fact that few fruit fly interceptions occur in cargo compared to baggage, does not necessarily indicate that more fruit fly individuals arrive via the latter. Due to the much larger volume of fruits and vegetables arriving by cargo ship, and more complex inspection procedures, fruit cutting is not routinely performed at most locations during ship boarding and clearance. Finally, knowledge is lacking on how many individuals may be sufficient to establish a successful new Medfly colony at the arrival location. Improved knowledge about propagule pressures on these pathways would help immensely in more precise identification of pathways at risk.
80 As inspect ion agencies attempt to prioritize their inspection efforts, this work sheds light on potential inspection gaps and proposes a new approach to enable the consideration of seasonal factors of Medfly interception risk. It may help in identification of ports of entries at relatively increased or decreased infestation risk and highlights how these risks change in time between seasons. All of the interceptions in Florida between 2003 and 2008 occurred at Miami airport, but there are multiple air and seaports of entry at risk. The proposed methodology could also be employed in the modeling of other high priority pests, providing an evidence base for guiding how surveillance efforts could be redistributed accordingly.
81 Table 4 1. Main origin airports by region from the Medfly infested countries arriving to Florida according to the largest total passenger numbers. For each season origin risk indicator (ORI) is given. Region Country Airport IATA code Jan Apr May Aug Sep Dec Total passenger No Central Mexico MEX 1 1 1 174085 America Guatemala GUA 1 1 0.6 129062 Honduras SAP 1 1 0.6 92398 Mexico CUN 1 1 0.6 86221 Panama PTY 1 1 1 82470 Costa Rica LIR 1 1 0.6 69761 Costa Rica SJO 1 1 0.6 69720 South Colombia MDE 1 1 0.6 125494 America Colombia BOG 1 1 1 12 2783 Colombia CLO 1 1 1 93781 Ecuador UIO 1 1 1 92657 Bolivia VVI 1 1 1 91175 Colombia CTG 1 1 1 84527 Ecuador GYE 1 1 1 80702 Brazil REC 1 0.6 1 71480 Europe France CDG 0.2 1 0.6 48541 Italy FCO 0.6 1 1 39035 Italy MXP 0.2 0.6 0.6 20513 Austria VIE 0.2 0.6 0.2 1368 Bulgaria VAR 0.2 0.6 1 668 Africa and Congo FIH 1 1 1 16062 Middle East Nigeria ABV 1 1 1 2860 Burkina Faso OUA 0.2 0.6 0.6 1476 Cape Verde RAI 1 0.6 1 1468 Cameroon DLA 1 1 1 977 Israel TLV 0.6 0.6 0.6 954 Egypt CAI 0.6 0.6 1 1025 Morocco CMN 0.6 0.6 1 975 Benin COO 1 1 1 929
82 Table 4 2. Main destination airports in Florida according to the largest total passenger numbers. For each season, an airport risk indicator (ARI) is given. ARI: Medfly countries on ly ARI all international arrivals Destination Jan Apr May Aug Sep Dec Jan Apr May Aug Sep Dec Annual passenger volume Arrivals ratio from Medfly countries MIA 0.93 0.96 0.86 0.63 0.64 0.58 2080149 0.67 MCO 0.97 0.98 0.89 0.30 0.30 0.27 937268 0.31 FL L 0.99 1.00 0.75 0.42 0.42 0.32 615017 0.42 TPA 0.93 0.93 0.89 0.17 0.17 0.16 225657 0.18 SFB 0.71 0.84 0.87 0.09 0.10 0.11 91943 0.12 PIE 0.92 0.95 0.88 0.25 0.26 0.24 9282 0.27 TLH 0.95 0.96 0.83 0.54 0.55 0.48 5610 0.57 GNV 0.90 0.94 0.82 0.57 0.59 0.52 1190 0.63
83 Table 4 3 Pairs of selected origin destination airports. For each season, an airport origin risk indicator (AORI) is given. Destination Origin Jan Apr May Aug Sep Dec Total Passenger No FLL BOG 1 1 1 32958 FLL CUN 1 1 0.6 23588 FLL C DG 0.2 1 0.6 470 FLL MAD 0.6 0.6 1 177 GNV MEX 1 1 1 74 GNV CUN 1 1 0.6 20 GNV TLV 0.6 0.6 0.6 16 TLH GUA 1 1 0.6 281 TLH MEX 1 1 1 266 TLH SJO 1 1 0.6 108 TPA GUA 1 1 0.6 2348 TPA REC 1 0.6 1 1417 TPA FOR 1 0.6 0.6 673 TPA RAI 1 0.6 1 640 MIA UIO 1 1 1 89949 MIA MDE 1 1 0.6 75350 MIA REC 1 0.6 1 69575 MIA CDG 0.2 1 0.6 45546 PIE GUA 1 1 0.6 288 PIE OUA 0.2 0.6 0.6 72 PIE SJO 1 1 0.6 26 MCO MEX 1 1 1 78878 MCO BOG 1 1 1 45600 MCO PTY 1 1 1 34092 MCO ZRH 0.2 0.6 0.6 718 SFB HRE 1 1 1 1 349 SFB BKO 0.2 0.6 0.6 876 SFB FIH 1 1 1 795 SFB VAR 0.2 0.6 1 288
84 Table 4 4 Type of host and the number of Medfly interception at Miami airport (2003 2008). Host No of interceptions Annona muricata 12 Annona squamosa 3 Capsicum annuum 1 Capsic um sp. 6 Citrus sp. 13 Malus sp. 1 Mangifera indica 9 Manilkara zapota 3 Pouteria sapota 2 Prunus avium 1 Prunus persica 33 Prunus sp. 6 Punica granatum 4 Pyrus communis 1 Terminalia catappa 5 Unknown Fruit 4
85 Figure 4 1. Direct internati onal connections to Florida and the annual seat capacity total on those routes (OAG2010). Figure 4 2. The seasonal seat capacity total on incoming flights to Florida from the 15 th highest volume countries of origin.
86 Figure 4 3. Countries with t he highest number of passenger arrivals to Florida (up to 2 connections). Countries marked with an asterisk(*) indicate countries where Medfly is known to be present (EPPO 200 1 ) Figure 4 4. Countries that have Medfly officially present with the highest number of arriving passengers to Florida (up to 2 connections) by destination airport. TPA Tampa, TLH Tallahassee, SFB Orlando Sanford, PIE St. Petersburg Clearwater MIA Miami, MCO Orlando, GNV Gainesville, FLL Fort Lauderdale.
87 Figure 4 5. Medfly suitability adjusted numbers of passengers arriving from Medfly infested countries compar ed to the international arrivals total. TPA Tampa, TLH Tallahassee, SFB Orlando Sanford, PIE St. Petersburg Clearwater MIA Miami, MCO Orlando, GNV Gainesville, FLL Fort Lauderdale. Figure 4 6. Ratio of passengers arriving f rom Medfly infested countries (total number for each trimester is adjusted by suitability weight value).
89 Figure 4 7. Three panel map illustrating the Medly suitability adjusted volume of passenger flows to Florida per season. Origin airports a re marked according to their environmental suitability category for Medfly. Figure 4 8. Number of Medfly interceptions in passenger luggage at Miami International Airport (MIA) between 2003 and 2008 according to PestID.
90 CHAPTER 5 CONCLUSIONS Geogra phy bridges various sciences by offering tools and approaches to handle both human and environmental datasets using quantitative methods. These characteristics make geography attractive to other disciplines, policy makers and governmental agencies. Global data is becoming more accessible, less expensive and increasingly comprehensive every year. This massive flow of multi disciplinary information, combined with constantly increasing computer processing power, offers an exciting promise for further growth of geography as an overarching and synthesizing discipline able to handle problems on regional, global and multi temporal scale. This study is an example of geography research that combines methodologies and information from a broad set of sciences in order to tackle an interdisciplinary problem. The main purpose of this work was to quantify and combine various seasonal factors that contribute to changing risks of arrival of one of the most economically significant pest species, Medfly, to Florida. Seasonal p opulation dynamics at the origin locations driven by environmental factors the properties of commodity imports and the characteristics of the air passenger movement have all been accounted for. Although it has been shown in the literature that both climat e and the number of individuals transported contribute to invasion success, few studies exist that combine the two. Very few studies attempt to model species distributions seasonally, and these studies are usually performed on local or regional scales. Fo r species highly sensitive to environmental conditions that display strong seasonal patterns in distributions and abundances, seasonal modeling of environmental suitability can be crucial in terms of understanding and predicting when and where a pest is mo st likely to be at the peak or
91 trough of its population activity. In Chapter 1 Medfly ecological niche suitability models are presented that employ the most comprehensive dataset on its occurrence assembled to date, as a result of extensive literature sear ch for Medfly presence records. The distribution model outputs represent the first global assessment of the seasonally changing potential distribution of Medfly, illustrating the significant shifts in environmental suitability that occur throughout a typic al year. Quantifying and accounting spatially for various seasonal aspects of commodity flow to ports of entry is another factor analyzed here that can serve as an evidence base for guiding border surveillance strategies. Chapter 3 analyses the properties of maritime cargo shipments to Florida, with an emphasis on fruits and vegetable imports and the particular commodities that can serve as Medfly hosts. In markets subjected to seasonally varying demand, the seasonal modeling of import risk can be crucial in terms of predicting not only the highest risk pathways and origin regions, but also the time of year when the risk is highest. As Medfly is most commonly intercepted on passenger luggage, in Chapter 4 contemporary world air passenger flow data are utili zed to assess passenger flows to Florida from various international origin locations. Information about changing seasonal environmental suitability for Medfly at the origin locations is compared with the volume of passengers arriving from world wide destin ations to ports of entry in the state. A new approach has been presented here for assessing how Medfly importation risk changes according to origin, destination and season, which can aid various inspection and control activities. First of all, it can faci litate prioritization and optimization of the border surveillance efforts struggling with limited resources and staffing. It can
92 also be used to target interventions, enable or deny seasonal trade, track commodity areas of countries, define data acquisitio n at ports of entry, and be incorporated into risk assessments for commodity importation. This methodology, in principle, has potential to be applied to any invasive insect species, or any organism subject to seasonal population dynamics and density in a c ountry/state concerned with invasive organism introduction via air passenger luggage and can be widely adoptable both in scienti fic and policy making community
93 LIST OF REFERENCES Anderson RP, Lew D, Peterson AT (2003) Evaluating predictive models of speci distributions: criteria for selecting optimal models. Ecol Model 162:211 232. doi: 10.1016/S0304 3800(02)00349 6 APHIS (1992) Risk assessment, Mediterranean fruit fly. Planning and Risk Analysis Systems. Policy and Program Development. Animal and Plant Health Inspection Service USDA, Washington D.C. APHIS (1999) The Mediterranean Fruit Fly Factsheet PPQ. USDA APHIS Bacon SJ, Bacher S, Aebi A (2012) Gaps in Border Controls Are Related to Quarantine Alien Insect Invasions in Europe. PLoS ONE 7:e47689. doi: 10.1371/journal.pone.0047689 Benedict MQ, Levine RS, Hawley WA, Lounibos LP (2007) Spread of the tiger: global risk of invasion by the mosquito Aedes albopictus. Vector Borne Zoonotic Dis 7:76 85. Bradie J, Chivers C, Leung B (2013) Importing risk: q uantifying the propagule pressure establishment relationship at the pathway level. Divers Distrib 19:1020 1030. doi: 10.1111/ddi.12081 Brasier CM, Gibbs JN (1973) Origin of the Dutch elm disease epidemic in Britain. Nat UK 242:607 609. CABI (2000) Ceratiti s capitata. Distribution Maps of Plant Pests. http://www.cabi.org/isc/?compid=5&dsid=12367&loadmodule=datasheet&page=4 81&site=144 Campbell FT (2001) The science of risk assessment for phytosanitary regulation and the impact of changing trade regulations. BioScience 51:148 153. Carey (1991) Establishment of the Mediterranean fruit fly in California. Science 253:1369 1373. doi: 10.1126/science.1896848 Carey JR (1996) The Incipient Mediterranean Fruit Fly Population in California: Implications for Invasion Biology. Ecology 77:1690 1697. doi: 10.2307/2265775 Carey JR (2011) Biodemography of the Mediterranean fruit fly: Aging, longevity and adaptation in the wi ld. Exp Gerontol 46:404 411. doi: 10.1016/j.exger.2010.09.009 Caton BP, Dobbs TT, Brodel CF (2006) Arrivals of hitchhiking insect pests on international cargo aircraft at Miami International Airport. Biol Invasions 8:765 785.
94 CDFA (2003) Preventing biologi cal pollution: the Mediterranean fruit fly exclusion program. Colizza V, Barrat A, Barthlemy M, Vespignani A (2006) The role of the airline transportation network in the prediction and predictability of global epidemics. Proc Natl Acad Sci U S A 103:2015 Colunga Garcia M, Haack RA, Adelaja AO (2009) Freight transportation and the potential for invasions of exotic insects in urban and periurban forests of the United States. J Econ Entomol 102:237 246. Colunga Garcia M, Haack RA, Magarey RA, Margosian ML ( 2010) Modeling Spatial Establishment Patterns of Exotic Forest Insects in Urban Areas in Relation to Tree Cover and Propagule Pressure. J Econ Entomol 103:108 118. doi: 10.1603/EC09203 Committee on the Scientific Basis for Predicting the Invasive Potential of Nonindigenous Plants and Plant Pests in the United States, National Research Council (2002) Predicting Invasions of Nonindigenous Plants and Plant Pests. The National Academies Press, Washington, D.C. Cross E (2004) Mediterranean fruit flies attempt to sneak in again. Cust. Bord. Prot. TODAY Dobbs TT, Brodel CF (2004) Cargo Aircraft as a Pathway for the Entry of Nonindigenous Pests into South Florida. Fla Entomol 87:65 78. Dominiak BC, Daniels D (2012) Review of the past and present distribution of Med iterranean fruit fly (Ceratitis capitata Wiedemann) and Queensland fruit fly (Bactrocera tryoni Froggatt) in Australia. Aust J Entomol 51:104 115. doi: 10.1111/j.1440 6055.2011.00842.x Drake JM, Lodge DM (2004) Global hot spots of biological invasions: Eva luating options for ballast water management. Proc R Soc Lond B Biol Sci 271:575. Drake JM, Lodge DM (2007) Hull fouling is a risk factor for intercontinental species exchange in aquatic ecosystems. Aquat. Invasions Duyck P f., Quilici S (2002) Survival a nd Development of Different Life Stages of Three Ceratitis Spp. (Diptera: Tephritidae) Reared at Five Constant Temperatures. Bull Entomol Res 92:461 469. doi: 10.1079/BER2002188 Elith J, Graham CH, Anderson RP, et al. (2006) Novel methods improve predictio n of 151. Elith J, Kearney M, Phillips S (2010) The art of modelling range shifting species. Methods Ecol Evol 1:330 342. doi: 10.1111/j.2041 210X.2010.00036.x
95 Elith J, Leathwick JR, Hastie T (2 008) A working guide to boosted regression trees. J Anim Ecol 77:802 813. doi: 10.1111/j.1365 2656.2008.01390.x Elith J, Phillips SJ, Hastie T, et al. (2011) A statistical explanation of MaxEnt for ecologists. Divers Distrib 17:43 57. doi: 10.1111/j.1472 4 642.2010.00725.x Enkerlin W, Mumford J (1997) Economic Evaluation of Three Alternative Methods for Control of the Mediterranean Fruit Fly (Diptera: Tephritidae) in Israel, Palestinian Territories, and Jorda. J Econ Entomol 90:1066 1072. EPPO (2009) Data Sh eets on Quaran tine Pests: Ceratitis capitata. EPPO (2001) PQR EPPO database on quarantine pests. http://www.eppo.int Escudero Colomar LA, Vilajeliu M, Batllori L (2008) Seasonality in the occurrence of the Mediterr anean fruit fly [Ceratitis capitata (Wied.)] in the north east of Spain. J Appl Entomol 132:714 721. doi: 10.1111/j.1439 0418.2008.01372.x Floerl O, Inglis GJ (2005) Starting the invasion pathway: the interaction between source populations and human transp ort vectors. Biol Invasions 7:589 606. Floerl O, Inglis GJ, Dey K, Smith A (2009) The importance of transport hubs in stepping stone invasions. J Appl Ecol 46:37 45. Gasperi G, Bonizzoni M, Gomulski L, et al. (2002) Genetic differentiation, gene flow and t he origin of infestations of the medfly, Ceratitis capitata. Genetica 116:125 135. Grout TG, Stoltz KC (2007) Developmental Rates at Constant Temperatures of Three Economically Important Ceratitis spp. (Diptera: Tephritidae) From Southern Africa. Environ E ntomol 36:1310 1317. doi: 10.1603/0046 225X(2007)36[1310:DRACTO]2.0.CO;2 Gutierrez AP, Ponti L (2011) Assessing the invasive potential of the Mediterranean fruit fly in California and Italy. Biol Invasions. doi: 10.1007/s10530 011 9937 6 Harris EJ, Olalqui aga G (1991) Occurrence and Distribution Patterns of Mediterranean Fruit Fly (Diptera: Tephritidae) Desert Areas in Chile and Peru. Environ Entomol 20:174 178. Harris EJ, Vargas RI, Gilmore JE (1993) Seasonality in Occurrence and Distribution of Mediterran ean Fruit Fly (Diptera: Tephritidae) in Upland and Lowland Areas on Kauai, Hawaii. Environ Entomol 22:404 410. Hawley W (1988) The biology of Aedes albopictus. J Am Mosq Control Assoc Suppl 1:1 39.
96 Hay SI, Sinka ME, Okara RM, et al. (2010) Developing Globa l Maps of the Dominant Anopheles Vectors of Human Malaria. PLoS Med 7:e1000209. doi: 10.1371/journal.pmed.1000209 Hay SI, Tatem AJ, Graham AJ, et al. (2006) Global Environmental Data for Mapping Infectious Disease Distribution. In: Simon I. Hay AG and DJR (ed) Adv. Parasitol. Academic Press, pp 37 77 Hijmans RJ, Cameron SE, Parra JL, et al. (2005) Very high resolution interpolated climate surfaces for global land areas. Int J Clim 25:1965 1978. Huang Z, Tatem AJ (2013) Global malaria connectivity through ai r travel. Malar J 12:269. Huang Z, Wu X, Garcia AJ, et al. (2013) An Open Access Modeled Passenger Flow Matrix for the Global Air Network in 2010. PLoS ONE 8:e64317. doi: 10.1371/journal.pone.0064317 Hulme PE (2009) Trade, transport and trouble: managing i nvasive species pathways in an era of globalization. J Appl Ecol 46:10 18. Hulme PE, Bacher S, Kenis M, et al. (2008) Grasping at the routes of biological invasions: a framework for integrating pathways into policy. J Appl Ecol 45:403 414. IMO (2011) Inter national shipping facts and figures information resources on trade, safety, security, environment. International Maritime Organization Israely N, Ritte U, Oman SD (2004) Inability of Ceratitis capitata (Diptera: Tephritidae) to Overwinter in the Judean H ills. J Econ Entomol 97:33 42. doi: 10.1603/0022 0493 97.1.33 Israely N, Yuval B, Kitron U, Nestel D (1997) Population Fluctuations of Adult Mediterranean Fruit Flies (Diptera: Tephritidae) in a Mediterranean Heterogeneous Agricultural Region. Environ Ento mol 26:1263 1269. Israely N, Ziv Y, Oman SD (2005) Spatiotemporal Distribution Patterns of Mediterranean Fruit Fly (Diptera: Tephritidae) in the Central Region of Israel. Ann Entomol Soc Am 98:77 84. doi: 10.1603/0013 8746(2005)098[0077:SDPOMF]2.0.CO;2 Jan g EB (2007) Fruit flies and their impact on agriculture in Hawaii. http://scholarspace.manoa.hawaii.edu/handle/10125/1297 Accessed 21 Nov 2011 Johansson MA, Arana Vizcarrondo N, Bigge rstaff BJ, et al. (2011) On the Treatment of Airline Travelers in Mathematical Models. PLoS ONE 6:e22151. doi: 10.1371/journal.pone.0022151
97 Katsoyannos BI, Kouloussis NA, Carey JR (1998) Seasonal and Annual Occurrence of Mediterranean Fruit Flies (Diptera: Tephritidae) on Chios Island, Greece: Differences Between Two Neighboring Citrus Orchards. Ann Entomol Soc Am 91:43 51. Killeen GF, Fillinger U, Kiche I, et al. (2002) Eradication of Anopheles gambiae from Brazil: lessons for malaria control in Africa? La ncet Infect Dis 2:618 627. Klassen W, Brodel CF, Fieselmann DA (2002) Exotic pests of plants: current and future threats to horticultural production and trade in Florida and the Caribbean Basin. MICRONESICA AGANA 35:5 27. Kumschick S, Bacher S, Dawson W, et al. (2012) A conceptual framework for prioritization of invasive alien species for management according to their impact. NeoBiota 15:69 100. International Trade. Conserv Bio l 17:322 326. doi: 10.1046/j.1523 1739.2003.02038.x Liebhold AM, Work TT, McCullough DG, Cavey JF (2006) Airline baggage as a pathway for alien insect species invading the United States. Am Entomol 52:48 54. Liquido NJ, Cunningham RT, Nakagawa S (1990) Hos t Plants of Mediterranean Fruit Fly (Diptera: Tephritidae) on the Island of Hawaii (1949 1985 Survey). J Econ Entomol 83:1863 1878. Lockwood JL, Cassey P, Blackburn T (2005) The role of propagule pressure in explaining species invasions. Trends Ecol Evol 2 0:223 228. doi: 10.1016/j.tree.2005.02.004 Lockwood JL, Cassey P, Blackburn TM (2009) The more you introduce the more you get: the role of colonization pressure and propagule pressure in invasion ecology. Divers Distrib 15:904 910. Lounibos LP (2002a) Inva sions by insect vectors of human disease. Annu Rev Entomol 47:233 266. Lounibos LP (2002b) Invasions by insect vectors of human disease. Annu Rev Entomol 47:233 266. doi: 10.1146/annurev.ento.47.091201.145206 Mack RN, Simberloff D, Mark Lonsdale W, et al. (2000) Biotic invasions: causes, epidemiology, global consequences, and control. Ecol Appl 10:689 710. Maelzer DA, Bailey PT, Perepelicia N (2004) Factors supporting the non persistence of fruit fly populations in South Australia. Aust J Exp Agric 44:109. doi: 10.1071/EA01128
98 Magarey RD, Colunga Garcia M, Fieselmann DA (2009) Plant biosecurity in the United States: roles, responsibilities, and information needs. BioScience 59:875 884. Martnez Ferrer MT, Navarro C, Campos JM, et al. (2010) Seasonal and annu al trends in field populations of Mediterranean fruit fly, Ceratitis capitata, in Mediterranean citrus groves: comparison of two geographic areas in eastern Spain. Span J Agric Res 8:757 765. Mavrikakis PG, Economopoulos AP, Carey JR (2000) Continuous Wint er Reproduction and Growth of the Mediterranean Fruit Fly (Diptera: Tephritidae) in Heraklion, Crete, Southern Greece. Environ Entomol 29:1180 1187. doi: 10.1603/0046 225X 29.6.1180 McCullough DG, Work TT, Cavey JF, et al. (2006) Interceptions of nonindige nous plant pests at US ports of entry and border crossings over a 17 year period. Biol Invasions 8:611 630. De Meyer M, Robertson MP, Peterson AT, Mansell MW (2008) Ecological niches and potential geographical distributions of Mediterranean fruit fly (Cera titis capitata) and Natal fruit fly (Ceratitis rosa). J Biogeogr 35:270 281. doi: 10.1111/j.1365 2699.2007.01769.x Meyerson LA, Mooney HA (2007) Invasive alien species in an era of globalization. Front Ecol Environ 5:199 208. Ovruski SM, Schliserman P, Van Nieuwenhove GA, et al. (2010) Occurrence of Ceratitis capitata and Anastrepha fraterculus (Diptera: Tephritidae) on Cultivated, Exotic Fruit Species in the Highland Valleys of Tucuman in Northwest Argentina. Fla Entomol 93:277 282. doi: 10.1653/024.093.02 19 Papadopoulos NT, Carey JR, Katsoyannos BI, Kouloussis NA (1996) Overwintering of the Mediterranean Fruit Fly (Diptera: Tephritidae) in Northern Greece. Ann Entomol Soc Am 89:526 534. Papadopoulos NT, Katsoyannos BI, Carey JR (2002) Demographic Parameter s of the Mediterranean Fruit Fly (Diptera: Tephritidae) Reared in Apples. Ann Entomol Soc Am 95:564 569. doi: 10.1603/0013 8746(2002)095[0564:DPOTMF]2.0.CO;2 Papadopoulos NT, Katsoyannos BI, Carey JR, Kouloussis NA (2001a) Seasonal and Annual Occurrence of the Mediterranean Fruit Fly (Diptera: Tephritidae) in Northern Greece. Ann Entomol Soc Am 94:41 50. doi: 10.1603/0013 8746(2001)094[0041:SAAOOT]2.0.CO;2 Papadopoulos NT, Katsoyannos BI, Kouloussis NA, et al. (2001b) Early Detection and Population Monitori ng of Ceratitis capitata (Diptera: Tephritidae) in a Mixed Fruit Orchard in Northern Greece. J Econ Entomol 94:971 978.
99 Perry G, Vice D (2009) Forecasting the Risk of Brown Tree Snake Dispersal from Guam: a Mixed Transport Establishment Model. Conserv Biol 23:992 1000. doi: 10.1111/j.1523 1739.2009.01169.x Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecol Model 190:231 259. Phillips SJ, Dudk M (2008) Modeling of species distributions with Maxent: new extensions and a comprehensive evaluation. Ecography 31:161 175. doi: 10.1111/j.0906 7590.2008.5203.x Pimentel D (2007) Area wide pest management: environmental, economic and food issues. Area Wide Control Insect Pes ts 35 47. Pimentel D, Zuniga R, Morrison D (2005) Update on the environmental and economic costs associated with alien invasive species in the United States. Ecol Econ 52:273 288. doi: 10.1016/j.ecolecon.2004.10.002 Reiter P (1998) Aedes albopictus and the world trade in used tires, 1988 1995: the shape of things to come? J Am Mosq Control Assoc 14:83 94. Rezza G, Nicoletti L, Angelini R, et al. (2007) Infection with chikungunya virus in Italy: an outbreak in a temperate region. Lancet 370:1840 1846. doi: 1 0.1016/S0140 6736(07)61779 6 Riccardo Scalera, Genovesi P, Franz Essl, Rabitsch W (2012) The impacts of invasive alien species in Europe. European Environmental Agency, Luxembourg Sciarretta A, Trematerra P (2011) Spatio temporal distribution of Ceratitis capitata population in a heterogeneous landscape in Central Italy. J Appl Entomol 135:241 251. doi: 10.1111/j.1439 0418.2010.01515.x Seebens H, Gastner MT, Blasius B (2013) The risk of marine bioinvasion caused by global shipping. Ecol Lett n/a n/a. doi: 10.1111/ele.12111 Segura DF, Vera MT, Cladera JL, Barnes BN (2004) Host utilization by the Mediterranean fruit fly, Ceratitis capitata (Diptera: Tephritidae). Isteg Scientific Publications, pp 83 90 Simberloff D (2005) Th e politics of assessing risk for biological invasions: the USA as a case study. Trends Ecol Evol 20:216 222. Sinka ME, Bangs MJ, Manguin S, et al. (2010a) The dominant Anopheles vectors of human malaria in Africa, Europe and the Middle East: occurrence dat a, distribution maps and bionomic prcis. Parasit Vectors 3:117. doi: 10.1186/1756 3305 3 117
100 Sinka ME, Bangs MJ, Manguin S, et al. The dominant Anopheles vectors of human malaria in the Asia Pacific region: occurrence data, distribution maps and bionomic prcis. Parasit Vectors 4:89 89. doi: 10.1186/1756 3305 4 89 Sinka ME, Rubio Palis Y, Manguin S, et al. (2010b) The dominant Anopheles vectors of human malaria in the Americas: occurrence data, distribution maps and bionomic prcis. Parasit Vectors 3:72. d oi: 10.1186/1756 3305 3 72 Szyniszewska AM and Tatem AJ (In review) Global seasonal potential distribution of Mediterranean fruit fly Ceratitis capitata PLoS ONE Tatem AJ (2009) The worldwide airline network and the dispersal of exotic species: 2007 2010. Ecography 32:94 102. Tatem AJ, Hay SI (2007) Climatic similarity and biological exchange in the worldwide airline transportation network. Proc R Soc B Biol Sci 274:1489 1496. doi: 10.1098/rspb.2007.0148 Tatem AJ, Hay SI, Rogers DJ (2006a) Global traffic and disease vector dispersal. Proc Natl Acad Sci 103:6242 6247. doi: 10.1073/pnas.0508391103 Tatem AJ, Huang Z, Das A, et al. (2012) Air travel and vector borne disease movement. Parasitol Camb 139:1816. Tatem AJ, Rogers DJ, Hay SI (2006b) Estimating the malaria risk of African mosquito movement by air travel. Malar J 5:57. Tatem AJ, Rogers DJ, Hay SI (2006c) Global transport networks and infectious disease spread. Adv Parasitol 62:293 343. Tatem AJ, Rogers DJ, Hay SI (2006d) T8. 5: Traffic in disease vect ors: 2005, 2015 and Infectious Diseases: preparing for the future. Lond. UK Off. Sci. Innov. Thomas MC, Heppner JB, Woodruff RE, et al. (2001) Featured Creatures: Mediterranean Fruit Fly Florida Department of Agriculture and Consumer Services, Division of Plant Industry. University of Florida. http://entnemdept.ufl.edu/Creatures/fruit /mediterranean_fruit_fly.htm Toy SJ, Newfield MJ (2010) The accidental introduction of invasive animals as hitchhikers through inanimate pathways: a New Zealand perspective. Rev Sci Tech Int Epiz 29:123 133. Varela LG, Johnson MW, Strand L, et al. (2008 California worries commodity groups. Calif Agric 62:57 61. doi: 10.3733/ca.v062n02p57
101 Venette RC, Kriticos DJ, Magarey RD, et al. (2010) Pest Risk Maps for Invasive Alien Species: A Roadmap for Improvement. Bioscience 60:349 362. doi: 10.1525/bio.2010.60.5.5 Vera MT, Rodriguez R, Segura DF, et al. (2002) Potential Geographical Distribution of the Mediterranean Fruit Fly, Ceratitis capitata (Diptera: Tephritidae), with Emphasis on Argentina and Australia. Environ Entomol 31:1009 1022. doi: 10.1603/0046 225X 31.6.1009 Wasem RE, Lake J, Seghetti L, et al. (2004) Border Security: Inspections Practices, Policies, and Issues. Webster J, Mota M (2008) Pine wilt disease: global issues, trade and economic impact. http://dspace.uevora.pt/rdpc/handle/10174/6797 Accessed 12 Aug 2013 Weems Jr H (1981) Mediterranean fruit fly, Ceratitis capitata (Wiedemann)(Diptera: Tephritidae). Gainesv. Div. Plant Ind. 12pEntomology Ci rc. 230Links Westphal MI, Browne M, MacKinnon K, Noble I (2008) The link between international trade and the global distribution of invasive alien species. Biol Invasions 10:391 398. Willis HH, Ortiz DS (2004) Evaluating the security of the global contain erized supply chain. Rand Corporation Work TT, McCullough DG, Cavey JF, Komsa R (2005) Arrival rate of nonindigenous insect species into the United States through foreign trade. Biol Invasions 7:323 332.
102 BIOGRAPHICAL SKETCH Anna Szyniszewska graduated w ith PhD degree in Geography at the University of Florida (UF), working in cooperation with US Department of Agriculture, Animal Plant and Health Inspection Services (USDA APHIS) on a seasonal population modeling of invasive insect pest species and assessin g the risk of its spread over the global transportation networks. Her primary academic interest lies in the application of quantitative and geospatial techniques in solving various agricultural problems. Her obability of extreme rainfall events in Thailand with significance to the agriculture. Before joining UF Anna worked as a research fellow in the Japanese research group at the Kanazawa University on a project researching long term climate variability in Ce ntral Asia. She obtained BA at the Adam Mickiewicz University at the Dept. of Geographical and Geological Sciences in Poznan, Poland. Anna has recently joined the Computational Biology Department at the Rothamsted Research in the UK to work as a postdoctor al scientist on the early detection strategies of emerging cassava viruses in Africa.
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EMPS0JPAW_XJD0UU INGEST_TIME 2014-05-13T17:06:41Z PACKAGE UFE0046114_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC