1

INFORMING SURVEILLANCE FOR THE LOWLAND PLAGUE FOCUS IN AZERBAIJAN USING A HISTORIC DATASET

By

LILLIAN REBECCA MORRIS

A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT

OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE

UNIVERSITY OF FLORIDA

2013

2

© 2013 Lillian Rebecca Morris

3

To my mom and dad

4

ACKNOWLEDGMENTS

This project was funded by the United States Defense Threat Reduction Agency

(DTRA) through the Cooperative Biological Engagement Program under the

Cooperative Biological Research Project AJ-3. Funding was administered by the joint

University Partnership managed by the University of New Mexico.

I thank my family for their constant support, guidance and motivation for all of my

endeavors inside and outside of academia. I would also like to thank the members of

the SEER lab for their help, encouragement and patience throughout this process. I

would like to thank my advisor, Jason Blackburn, for his dedication and enthusiasm. He

has continuously challenged me to learn and grow as a scientist over the past two

years. I would also like to thank the members of my committee, Liang Mao and Mike

Binford, for their participation and guidance in writing this thesis.

Finally, I would like to thank FUEL for providing a much needed and appreciated

outlet throughout my graduate school career.

5

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS .................................................................................................. 4

LIST OF TABLES ............................................................................................................ 7

LIST OF FIGURES .......................................................................................................... 8

LIST OF ABBREVIATIONS ............................................................................................. 9

ABSTRACT ................................................................................................................... 10

CHAPTER

1 INTRODUCTION .................................................................................................... 12

Plague Background ................................................................................................ 12 Primary Host and Focus ......................................................................................... 15 Plague in Azerbaijan ............................................................................................... 18 STAMP ................................................................................................................... 20 Plague and Climate ................................................................................................ 21 Human Plague Risk ................................................................................................ 23 Objectives ............................................................................................................... 25

2 METHODS .............................................................................................................. 27

Study Area .............................................................................................................. 27 Meriones libycus Data ............................................................................................. 28 Database Development .......................................................................................... 29 STAMP ................................................................................................................... 30 Stable Regions ....................................................................................................... 31 Environmental Data ................................................................................................ 32 Environmental Analysis ........................................................................................... 33 Human Population Analysis .................................................................................... 34

3 RESULTS ............................................................................................................... 44

STAMP ................................................................................................................... 44 Stable Regions ....................................................................................................... 44 Environmental Data ................................................................................................ 45 Environmental Analysis ........................................................................................... 45 Human Population .................................................................................................. 46

4 DISCUSSION ......................................................................................................... 59

6

Future Directions .................................................................................................... 66 APPENDIX

A STAMP OUTPUTS ................................................................................................. 67

B POPULATION DISTRIBUTION .............................................................................. 93

LIST OF REFERENCES ............................................................................................... 94

BIOGRAPHICAL SKETCH .......................................................................................... 101

7

LIST OF TABLES

Table page 2-1 List of host and vector species found in annual yearbooks ................................ 37

2-2 GlobCover classifications. .................................................................................. 38

3-1 Changes in the area from STAMP analysis ........................................................ 48

3-2 Mann Whitney U test, 5 categories. .................................................................... 49

3-3 Descriptive statistics for environmental variables. .............................................. 50

3-4 Mann Whitney U test, 3 categories ..................................................................... 51

3-5 The percentages of each land cover type ........................................................... 52

8

LIST OF FIGURES

Figure page 2-1 Azerbaijan with land cover. ................................................................................. 39

2-2 Plague focus and protected areas in Azerbaijan ................................................ 40

2-3 Extracting data from annual yearbooks .............................................................. 41

2-4 Hand drawn maps .............................................................................................. 42

2-5 Event definitions for STAMP ............................................................................... 43

3-1 STAMP outputs from 1976 to 1977. ................................................................... 53

3-2 Stable regions for each abundance category ..................................................... 54

3-3 The distribution of environmental variables across Azerbaijan ........................... 55

3-4 Regions of stability for each abundance level .................................................... 56

3-5 The percentage of each land cover type ............................................................ 57

3-6 The percentage population change between 1970 and 2010. ............................ 58

A-1 STAMP outputs for very low Meriones libycus abundance ................................. 68

A-2 STAMP outputs for low Meriones libycus abundance ......................................... 71

A-3 STAMP outputs for average Meriones libycus abundance ................................. 77

A-4 STAMP outputs for high Meriones libycus abundance ....................................... 83

A-5 STAMP outputs for very high Meriones libycus abundance ............................... 89

B-1 HYDE population grids for 1970 and 2010. ........................................................ 93

9

LIST OF ABBREVIATIONS

APS Republican Anti-Plague Station

APD Anti-Plague Division

CDC Center for Disease Control

CEPF Critical Ecosystems Partnership Fund

GIS Geographic Information System

GlobCover Global composites and land cover maps

HYDE History Database of the Global Environment

RMS root-mean-square error

STAMP Space Time Analysis of Moving Polygons

WHO World Health Organization

WorldClim Global Climate Data

10

Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science

INFORMING SURVEILLANCE FOR THE LOWLAND PLAGUE FOCUS IN

AZERBAIJAN USING A HISTORIC DATASET

By

Lillian Rebecca Morris

May 2013

Chair: Jason Blackburn Major: Geography

Yersinia pestis is a gram-negative, zoonotic pathogen that causes plague.

Plague is maintained in nature through a transmission cycle between partially resistant

rodent hosts and fleas. There are natural reservoirs on almost every continent, and the

number of human plague cases has increased in recent years. Azerbaijan is a country

at the crossroads of Eastern Europe and western Asia that has a history of

environmental plague foci. Informing plague surveillance in this region is imperative

due to the deteriorating public health system that resulted from the collapse of the

Soviet Union. This study aims to prioritize regions for plague surveillance in Azerbaijan.

A 14 year historic data set was employed to analyze the spatio-temporal pattern of the

primary plague host in the country, Meriones libycus, using the Space Time Analysis of

Moving Polygons method (STAMP). This method has the utility to identify areas that

remained stable over time, which is meaningful when analyzing historic patterns. The

relationship between stable M. libycus abundance and environmental variables

including temperature, altitude, land cover type and annual precipitation was explored.

Changes in human population density over the historic period to modern times were

11

also analyzed. We were particularly interested in identifying increasing population

trends in the area surrounding regions characterized by historically high M. libycus

abundance, as the risk of human plague increases as humans come into close

proximity with hosts and vectors. We found that there was variation in M. libycus

abundance over the historic period, but regions of stability were identified. There are

significantly different climatic and land cover conditions associated with different levels

of abundance. The population in Azerbaijan has steadily increased over the past 30

years, including regions bordering plague foci. Surveillance should be prioritized for

regions with historically stable high host abundance, regions with climatic conditions

associated with high abundance, and regions with increasing populations surrounding

plague foci.

12

CHAPTER 1 INTRODUCTION

Plague Background

Plague is a flea borne zoonosis caused by the gram negative bacterium Yersinia

pestis (Pollitzer 1954, Perry and Fetherston 1997, Gage and Kosoy 2005). Plague is a

notorious disease that has been associated with three massive pandemics throughout

history (Gage and Kosoy 2005). The first plague pandemic spread across the

Mediterranean in the 6th century and has been linked to the weakening of the Byzantine

empire (Perry and Fetherston 1997). The Black Death, which killed almost one third of

Europe’s population in the fourteenth century, has been attributed to plague. There is

evidence that the black death originated from an environmental plague focus, which is a

region on the landscape that maintains the pathogen, in central Asia and spread to the

south and east (Pollitzer 1951). The third pandemic started in China in the 19th century

(Perry and Fetherston 1997, Stenseth et al. 2008). The spread of this pandemic was

facilitated by rat infested steamships travelling to major ports (Dennis 1998). An

estimated 26 million plague cases resulted, with approximately 12 million fatalities

(Dennis 1998). The geographic range of plague has greatly expanded since the onset

of the most recent pandemic (Gage and Kosoy 2005). This expansion includes the

Western United States, portions of India, and Southeast Asia (Dennis 1998, Gage and

Kosoy 2005, Pollitzer 1954). There has also been an increase in human plague cases.

From 1981 to 1995 approximately 20,000 human plague cases from 25 countries were

reported to the World Health Organization (WHO).

The Center for Disease Control (CDC) classifies Y. pestis as a Category A

biological agent (Rotz et al. 2002). Category A agents have the greatest potential for

13

adverse public health impacts through mass casualties, which presents a bioterrorism

concern. These agents require preparedness efforts including improved surveillance

(Rotz et al. 2002). The greatest challenge related to plague control and surveillance is

its presence and maintenance in rural, natural cycles (Dennis 1998). Plague has a

history of appearing in new populations or reemerging in a region that has not seen

cases for extended time periods , which is difficult to predict (Dennis 1998). The 1994

pneumonic human outbreak in India exemplifies the necessity of monitoring this

important pathogen because it reiterates the idea that there is a constant risk of human

plague surrounding natural foci (Dennis 1994).

Yersinia pestis is maintained in nature through a transmission cycle between

partially resistant rodent hosts and adult hematophagous fleas (Gage and Kosoy 2005,

Meyer 1942). Yersinia pestis is a successful vector borne pathogen because it is able to

survive in its host and also spreads from the site of a flea bite, which serves as a source

of infection for feeding fleas (Anisimov 2002). The infected flea spreads the bacteria to

other rodent hosts. It is believed that the bacteria can be maintained indefinitely in these

enzootic or maintenance cycles (Gage and Kosoy 2005, Beran 1994, Gage, Ostfeld and

Olson 1995). The enzootic or primary host is able to maintain plague foci without the

involvement of other hosts as long as sufficient numbers of flea vectors are present

(Gage and Kosoy 2005). When Y. pestis spreads from enzootic hosts to more highly

susceptible or secondary hosts then rapid die offs take place (Gage and Kosoy 2005).

Classically it is hypothesized that plague transmission relies on the flea vector.

High numbers of fleas tend to be correlated with a higher prevalence of plague (Conrad,

LeCocq and Krain 1968). There are more than 200 potential species of plague vectors

14

that are specific to rodent hosts (Gage and Kosoy 2005). Fleas are efficient vectors of

Y. pestis as a result of a blocking process that takes place during the transmission

cycles (Bacot and Martin 1914). Once a flea has fed on an infected host the bacteria

rapidly replicates in the flea’s gut (Bacot and Martin 1914, Bibikova 1977). The bacterial

colonies grow large enough to block the movement of blood from the midgut to the

foregut, which essentially starves the flea. The starving flea feeds rapidly by using

pharyngeal muscles to draw blood, but eventually has to relax those muscles, which

leads to contaminated blood being flushed back to the feeding site infecting the host

(Webb et al. 2006, Bacot and Martin 1914).

There is ongoing research on the role of partially blocked and unblocked fleas in

plague transmission (Gage and Kosoy 2005). It has been suggested that partially

blocked fleas could play an important, short term role in transmission that helps sustain

outbreaks (Webb et al. 2006). Webb et al. (2006) found that blocked fleas cannot

exclusively drive epizootics in prairie dogs. Short term reservoirs including other small

mammals, infectious carcasses, and transmission from unblocked fleas likely drive

disease spread during an outbreak.

Over 200 species of rodents have been identified as playing a role in plague

transmission cycles (Gratz 1999). A successful host must have enough resistance to Y.

pestis to allow bacteria to cycle through the blood and infect fleas (Meyer 1942). The

amount of resistance in a host is related to variables including species, age, breeding

status, immune and physiological status, and season (Meyer 1942, Hubbert and

Goldenberg 1970). A host population tends to contain a combination of susceptible

individuals that die as a result of infection and more resistant individuals that survive

15

infection (Pollitzer 1954, Gage and Kosoy 2005). Recovered hosts likely become

chronic carriers, which allows populations to maintain the bacteria between seasons

(Pollitzer 1954). It has been suggested that the impact of host populations on disease

transmission depends on the ratio of resistant to susceptible individuals within the

population (Gage and Kosoy 2005). An effective host population tends to be heavily

infected with multiple flea vectors, and live in burrows with high densities of rodents and

fleas (Pavlovsky, Plous JR and Levine 1966).

Studies have found that there is also a relationship between host abundance and

plague prevalence in host communities. Davis et al. (2004) found that plague

persistence in Kazakhstan’s desert region was related to the total abundance of the

Great Gerbil, Rhombomys opimus, host population. They suggest that the colonization

and extinction dynamic affecting gerbil abundance determines the meta-population

context of plague (Davis et al. 2004). In order to successfully monitor and control

plague on the landscape, it is necessary to understand the distribution and abundance

of hosts (Stenseth et al. 2008, Gage and Kosoy 2005).

Primary Host and Focus

Large, natural foci of plague are active in Asia, part of the Russia Federation, and

Asian republics (Gratz 1999). Over the past twenty years, human plague cases have

reemerged in this region (Pollitzer 1954). Bertherat et al. (2007) reported a human

plague outbreak in Algeria in 2003, which was the first reported instance of plague in

the country for 50 years. Human plague cases have also been reported in Saudi Arabia

(Saeed, Al-Hamdan and Fontaine 2005), Jordan (Arbaji et al. 2005), Afghanistan (Leslie

et al. 2011), and a limited outbreak in Libya (Tarantola et al. 2009) over the past

decade. The reemergence of human plague in the region suggests that there are still

16

active environmental reservoirs. The Republic of Azerbaijan is a country at the

crossroads of western Asia and Eastern Europe that has three well documented

historical plague foci (Gurbanov and Akhmedova 2010).

Meriones libycus is a major plague host in this part of the world (Gratz 1999,

Gage and Kosoy 2005, Gurbanov and Akhmedova 2010, Bakanidze et al. 2010).

Meriones libycus has been documented as the primary plague host in the Trans-

continental lowland foothills plague focus in Azerbaijan (Gurbanov and Akhmedova

2010). As early as 1953, M. libycus prevalence was linked to a plague epizootic in the

Apshnon peninsula of Azerbaijan (Bakanidze et al. 2010).

In addition to its established role as a plague host, this ground dwelling gerbil

plays a critical role in the maintenance of leishmaniasis in the greater Mediterranean

area and Western Asia. Leishmanisis is a vector born disease caused by a protozoa

transmitted by the bite of a sand-fly, and presents a critical public health concern in

several countries (González et al. 2009, Desjeux 2001). Meriones libycus is one

species that maintains the zoonotic form of the disease in the environment (Nadim and

Faghih 1968). Human cases can result from humans coming into contact with sandflies

inhabiting M. libycus burrows. In the Republic of Iran, M. libycus is the primary reservoir

for zoonotic cutaneous leishmaniasis (Rassi et al. 2006, Yaghoobi-Ershadi, Akhavan

and Mohebali 1996). Meriones libycus has also been linked to the maintenance of

leishmaniasis in Saudi Arabia (Ibrahim et al. 1994). It is clear that monitoring and

controlling M. libycus populations has important implications for multiple public health

initiatives.

17

The natural range of M. libycus spans from the Western Sahara to Western

Xinjiang, China (Nowak and Paradiso 1999). They are found in a range of habitats

including clay and sandy deserts, bush country, arid steppe, low plains, cultivated fields,

grasslands, and mountain valleys (Nowak and Paradiso 1999, Daly and Daly 2009).

Meriones libycus has a rat like appearance characterized by narrow well developed

ears, a tail the length of their head and body, elongated hind legs for leaping, and strong

claws (Nowak and Paradiso 1999). They eat green vegetation, roots, bulbs, seeds,

cereal, fruits, and insects (Nowak and Paradiso 1999). Meriones libycus live in complex

multi entranced burrows, which are approximately 1.5 meters deep, and expand

outward approximately 3 to 4 meters (Nowak and Paradiso 1999). Their burrows

generally have food storage chambers near the top, and nesting chambers deeper in

the ground (Naumov et al. 1973). Individuals store as much as 10 kilograms of seeds at

a time in the northern part of its range (Naumov et al. 1973). Their preferred burrow

location is under bush cover (Daly and Daly 2009), which puts them at close proximity

to flea vectors (Gage and Kosoy 2005). Many individuals burrow in the same vicinity

(Naumov et al. 1973). They spend the majority of their time underground, and are prone

to nocturnal activity (Daly and Daly 2009). Males traverse large overlapping home

ranges, while females prefer smaller more exclusive ranges (Daly and Daly 2009,

Nowak and Paradiso 1999). Females will remain in the same vegetative depression

surrounded by desert for several consecutive months. Home ranges are between 50

and 120 meters in diameter (Nowak and Paradiso 1999). Reproduction occurs year

round, but usually takes place in late winter to early autumn. Each female produces 2

18

to 3 litters, and litter sizes range from 1 to 12. The gestation period is 20 to 31 days

(Nowak and Paradiso 1999).

Plague in Azerbaijan

The concern about plague and its hosts has a long history in Azerbaijan. During

the Soviet era, an Anti–plague system (APS) was created to respond to outbreaks of

plague and other bacterial and viral diseases (Ouagrham-Gormley 2006, Zilinskas

2006). These stations addressed disease through surveillance, research production

and training efforts; however, their primary objective was disease control (Ouagrham-

Gormley 2006, Zilinskas 2006). During the Soviet era, the APS grew to include over 100

facilities throughout the region, which was the largest system of its kind the world had

ever seen. Today, The APS in Baku is responsible for surveillance, identification,

documentation and preventive measures against very dangerous infections (10

conditions including plague, cholera and anthrax). It has five regional branches known

as Anti-Plague Divisions (APD) that obtain laboratory samples and perform an initial

diagnosis before referring samples to the APS for confirmation testing. However,

contemporary field sampling for plague has been limited since independence from the

Soviet Union.

The first APD in Azerbaijan was a result of a plague outbreak in the Hadrut

district of Azerbaijan in 1930. This outbreak prompted the establishment of a plague

department within the Institute of microbiology in Baku, the capital city of Azerbaijan

(Ouagrham-Gormley 2006). In 1934 the department became the central Anti – plague

station in Azerbaijan (Ouagrham-Gormley 2006). After several plague epizootic events

in Azerbaijan the source was identified as a natural plague reservoir in Azerbaijan by

Soviet scientists in 1953. Three new field stations were established as a result

19

(Ouagrham-Gormley 2006). As part of surveillance efforts in Azerbaijan, a detailed

annual yearbook was published outlining every aspect of the APS effort for that year.

Topics included preventative measures, prophylaxis of plague and other diseases,

epidemiological observations, and epizootic observations. The yearbook also outlined

lab research and field related work from the year. Other details such as the structure,

financial status of the institute, and training efforts that took place at the institute were

also discussed. In Azerbaijan, annual yearbooks were published for more than 40

years. Today, these yearbooks are maintained in the archival library at the Republican

APS in Baku.

APS yearbooks provide a unique and important opportunity to understand the

history of plague in Azerbaijan. This is particularly relevant because surveillance efforts

have been greatly limited since the collapse of the Soviet Union in 1991. The Republic

of Azerbaijan gained its independence at this time; however, this collapse was

characterized by diminished funding and maintenance for the APS system. With fewer

resources available for surveillance, this thorough and detailed historic dataset can be

used to inform modern surveillance efforts.

One of the most thorough and repeated datasets in the yearbook collection

outlined seasonal M. libycus abundance across the country. These hand drawn maps

depicted M. libycus abundance using polygons, and cross hatched patterns to represent

different levels of abundance. Each level of abundance is represented with a polygon of

a particular pattern or color that differentiates it from the other abundance levels and

host species. This is an important dataset to explore because of the continued

significance of M. libycus as a plague host in Azerbaijan. The yearbooks spanned

20

several years and represented the entire plague focus, allowing for the analysis of M.

libycus abundance across space and over time.

STAMP

Exploring historical changes in M. libycus spatial patterns and abundance across

the landscape can provide important insight into distribution and ecology of this

important plague host. Identifying explicit regions with extended periods of high or low

abundance can be particularly informative. Specifically, such analyses could provide a

historical baseline of spatio-temporal patterns of host abundance that might inform

contemporary efforts to direct enzootic sampling in Azerbaijan. The Space Time

Analysis of Moving Polygons (STAMP) approach is one method for addressing these

objectives with data such as those available in the APS yearbooks. STAMP uses

polygons representing a phenomenon from two consecutive time periods and describes

the type of change that is taking place between polygons. STAMP overlays polygons

from consecutive time periods, and each incremental time change is categorized based

on the associated spatial change (Robertson et al. 2007). STAMP’s change definitions,

also called event types, are derived from the spatial proximity or overlap between the

two time periods. The analysis includes the identification of regions that do not change,

or that remain stable (Robertson et al. 2007).

STAMP has been applied to a variety of diverse research questions. This

method was initially employed to determine how a wildfire spread over time (Robertson

et al. 2007). Robertson et al. (2007) were able to determine the rate and direction of fire

spread over a three week period using STAMP. This method was also used to

determine differences in the area used by solitary female grizzly bears verses females

with cubs over a two year time period (Smulders et al. 2012). They were particularly

21

interested in identifying differences in site fidelity between the two groups, which can be

derived from areas classified as stable during a STAMP analysis (Smulders et al. 2012).

Nelson et al. (2011) employed STAMP to explore changing caribou home ranges. A

home range describes the space an animal uses over a specified time period and is

commonly defined by polygons. A STAMP analysis of changing home ranges revealed

the type of home range drift that occurred over the study interval (Nelson 2011). As

these applications reveal, STAMP can be used to compare two time periods, or can

create a sequence to describe how a polygon changes over multiple time periods

(Robertson et al. 2007).

Our analysis has many similarities to previous applications of STAMP. Our data

consists of polygons to describe M. libycus abundance over a multiyear period. STAMP

has the utility to analyze this type of data, and identify the regions of stability. The

stable areas across years are of particular interest because human plague epidemics,

as well as epizootic events, have been correlated to M. libycus natural foci in the past

(Bakanidze et al. 2010, Daly and Daly 2009). Historical trends might be able to suggest

current regions more likely to support high levels of plague host abundance.

Plague and Climate

Throughout history, links have been made between climatic conditions and

plague prevalence. Precipitation, temperature, and relative humidity have been found

to have the highest impact on the plague cycle (Ben Ari et al. 2011). Seasonal plague

events in India have been related to variations in temperature and humidity (Rogers

1928). A pattern of fewer outbreaks in extreme hot or cold temperatures has been

described for several African countries (Davis 1953). Stenseth et al. (2006) found that

warmer springs and wetter summers increase occurrence and levels of plague

22

prevalence. Environmental foci tend to be found in semi-arid to arid regions, and low

humidity forests (Perry and Fetherston 1997). Plague foci do not typically persist for

long periods in humid, tropical lowlands, or the highest, driest deserts (Perry and

Fetherston 1997). Elevation is another environmental variable that has been linked to

plague persistence on the landscape. One study identified elevation as one of the most

important variables in a model for predicting plague risk across the Lushoto district of

Tanzania (Neerinckx et al. 2010). A model of human plague risk in the southwest United

States also identified elevation as a significant predictor (Eisen et al. 2007b).

Important relationships have been described between the plague host and

climatic conditions. Cavanaugh and Marshall (1972) suggest that high intensity rainfall

is detrimental to rodent survival, as burrows become flooded. There appears to be an

association between annual rainfall, rodent densities and seasonal distribution of hosts

(Ben Ari et al. 2011). A recent study in New Mexico presented a trophic cascade

hypothesis where precipitation resulted in increased plant production and rodent food

sources (Parmenter et al. 1999). The subsequent increase in host populations

increases the likelihood of epizootics and human cases (Parmenter et al. 1999). There

is a less apparent direct relationship between hosts and temperature. Though in

temperate regions, low winter temperatures can impact food availability, which

influences rodent densities and distribution (Korslund and Steen 2005).

Climate plays an important role in dictating the presence of plague on the

landscape; therefore, climatic variables should be included when exploring questions

relating to plague and its hosts. This analysis included an exploration of the climatic

conditions associated with rodent abundance over time. We focused on precipitation,

23

temperature and elevation because they are known to influence the presence of plague

and plague hosts on the landscape in other regions. Understanding relationships

between environmental factors and host abundance will help us to identify regions that

might be characterized by higher levels of host abundance. For example, if a region

was historically characterized by high levels of rodent abundance we can hypothesize

that stable climate conditions will maintain similar levels of plague and rodents in the

region. Similarly, if there are climatic shifts this analysis can help detect new regions at

risk of increasing host abundance and guide surveillance efforts.

Human Plague Risk

In order to thoroughly explore regions at the highest risk of plague transmission,

it is important to consider human population dynamics, as human plague cases

frequently result from humans coming into contact with flea vectors (Mann et al. 1979,

Saeed et al. 2005). Plague epidemics still occur in developing countries and are

commonly related to large epizootic events. When there is a large die off of host

species, the flea vectors are forced to find another source of food, which causes the

bacteria to spread outside of the enzootic, maintenance cycle (Eskey and Haas 1940,

Kartman 1970, Gage and Kosoy 2005). In California, studies have suggested that

increased serological prevalence of plague in coyotes, Canis latrans, could be an

indicator of a plague epizootic event in the region (Thomas and Hughes 1992).

Carnivores also provide a vehicle for infected fleas to travel to other populations (Perry

and Fetherston 1997). Flea vectors on dogs and cats is one method of facilitating

human infection (Mann et al. 1979). Human outbreaks are sometimes correlated with

unsanitary or rat infested environments, as the hosts, vectors and human populations

are in a close spatial proximity (Gage and Kosoy 2005). A human plague epidemic in

24

Vietnam in the winter of 1967 was partially attributed to a rapidly increasing rat

population and a high flea index (Conrad et al. 1968). In the southwest United States,

distance to host habitat is also a significant predictor of high risk human plague regions

(Eisen et al. 2007a). Handling or ingesting wild animal carcasses is another

documented infection pathway. There was one reported human plague outbreak in

Saudi Arabia linked to eating raw camel liver (Saeed et al. 2005). More recently, a large

outbreak with 83 cases, of which 17 were fatal, was attributed to eating camel meat in

Afghanistan (Leslie et al. 2011).

Human cases generally start as the bubonic form of the disease, particularly if

the bacterium was contracted from a flea bite. Symptoms include fever, headache,

chills, and swollen lymph nodes, called buboes (Perry and Fetherston 1997). Symptoms

develop within 2 to 6 days of coming into contact with the bacteria. Septicemic plague

develops as the infection spreads into the bloodstream (Hull, Montes and Mann 1987).

Pneumonic plague results from the lungs becoming infected. This is the most fatal form

of plague. Pneumonic plague can be transmitted from person to person via respiratory

droplets through close contact (Perry and Fetherston 1997).

Protected lands, such as national parks, create a dynamic that can bring humans

and vectors into close proximity. This dynamic has the potential to effect Azerbaijan

due to its inclusion in the Caucasus Biodiversity Hotspot, which is one of 34 biodiversity

hotspots in the world (Zazanashvili et al. 2009). This designation is a result of the

presence of diverse ecosystems, endangered species, and endangered vegetation

(CEPF 2003). The Caucasus region also has the greatest biodiversity out of any

temperate forest in the world (Zazanashvili et al. 2009). The bezoar goat, goitered

25

gazelle, mouflon tar, and Caspian seal are all critically endangered species found in

Azerbaijan (Zazanashvili et al. 2009). Today approximately 10% of the country is

protected lands, which includes national parks, strict nature preserves, and wildlife

sanctuaries. The number of protected lands in Azerbaijan has doubled since the

collapse of the Soviet Union, and there are significant efforts to continue expanding the

system in order to preserve the biodiversity in the country (Zazanashvili et al. 2009).

The presence of non-governmental organizations concerned with conservation, and

organizations such as the World Wildlife Fund, The Critical Ecosystem Partnership

Fund, and the Eco Region Conservation Fund are all contributing to expanding the

protected lands system (Zazanashvili et al. 2009, CEPF 2003). Part of this effort

includes the diversification of protected lands. Traditionally, Azerbaijan has created

strict nature reserves that have minimal access and recreational opportunities for

people. In recent years, the number of national parks with recreational opportunities

and public access has greatly increased (Zazanashvili et al. 2009).

The spatial proximity of human populations to natural reservoirs is clearly a risk

factor for human cases. It is particularly important to monitor changing population

densities in regions neighboring natural plague foci in order to gauge changing risks of

human infection. Prioritizing limited resources for control efforts should incorporate an

evaluation of regions at the greatest risk of human infection.

Objectives

The objective of this study was to inform and identify priorities for plague

surveillance and control in Azerbaijan, which is critically important due to the limited

resources available for public health efforts after the collapse of the Soviet Union. To

achieve this objective, we explored changes in M. libycus abundance over space and

26

time, and identified regions of historically stable host abundance. A second objective

was to identify specific climatic factors associated with stable regions of mammal

abundance by category. The relationship between stable host abundance and

changing population levels were also analyzed in order to identify regions characterized

by an increased risk of human infection.

27

CHAPTER 2 METHODS

Study Area

Azerbaijan is a small country located in the Caucasus region at the intersection

of Western Asia and Eastern Europe (Figure 2-1). There are 59 administrative districts,

or rayons, in the country. Azerbaijan is bordered by the Caspian Sea to the east,

Armenia and Georgia to the west and Russia to the north. The capital city, Baku, is on

the east coast of Azerbaijan. The Greater and Lesser Caucasus mountain ranges run

through the country. The Greater Caucasus range extends from Sochi Russia on the

northeastern shore of the black sea and ends adjacent to Baku. The lesser Caucasus

range runs parallel to the Greater range, about 100km to the south. In Azerbaijan, the

Greater Caucasus mountain range parallels the northern border. The Lesser Caucasus

range spans the south western border between Armenia and Azerbaijan.

This study focused on one of three plague foci in the country. APS scientists

define the plague focus used for this study as the natural habitat of M. libycus in the

region (Figure 2-2). This area is characterized by constant presence of the Libyan jird,

which was determined through year round perennial investigations. Meriones libycus

natural habitat was employed to define this boundary because it is the primary carrier of

plague in the Transcaucasian Plain-sub Mountain Natural focus.

There are currently 44 protected areas in Azerbaijan, which encompasses

National parks, strict nature reserves, wildlife refuges and wildlife sanctuaries

(Zazanashvili et al. 2009) (Figure 2-2). Spatial data on protected lands was

downloaded from the World Database on Protected Lands (http://www.wdpa.org/). This

database was derived from information from national governments, non-government

28

organizations, academic institutions, and many other sources to provide a

comprehensive global spatial dataset on marine and terrestrial protected areas.

Meriones libycus Data

Annual APS yearbooks thoroughly described every aspect of the stations

activities for the year including lab activities, field work, and control efforts, which were

reported using tables, charts, and maps. Historic yearbooks contained information on

multiple plague hosts and vectors (Table 2-1). Host and vector data was displayed in

hand drawn maps using polygons, colors, and patterns to represent different species, or

abundance levels for the year. These datasets were found in the archival library at the

Republican APS in Baku. Our research team extracted data from yearbooks in order to

convert them to an electronic form more conducive to analysis in a Geographic

Information System (GIS) (Figure 2-3). For this analysis, we use a subset of APS

yearbook data from 1972 to 1985. We limited this analysis to data on M. libycus

because of its history as an important plague host in the country. Additionally, M.

libycus abundance maps were consistently published throughout the fourteen year

period. The historic Azerbaijan dataset included M. libycus data collected during the fall

season across the country.

Meriones libycus themed maps defined abundance using 5 categories from 1972

to 1980 in the historic yearbooks: very low, low, average, high, and very high (Figure 2-

4). Very low was defined as 0 – 1 specimen per hectare. Low was 2-5 specimens per

hectare. Average was defined as 6 – 10 specimens per hectare. 11 – 20 specimens

per hectare was high abundance. Greater than 20 specimens per hectare defined very

high abundance. Each definition was based on APS zoological sampling surveys

conducted in the fall of each year. After 1980, very low and low were collapsed into one

29

category defined as less than or equal to 5 specimens per hectare. Very low was no

longer used due to the inexpediency in identifying areas with close to absent rodent

populations. Instead, the regions of very low abundance were identified by zoologists

based on visual inspection of the landscape for burrows or signs of recent rodent

activity, but were not included in annual abundance maps. This analysis included an

exploration of changes in M. libycus abundance over the historic time period.

Database Development

In order to analyze data available in annual yearbooks we photographed all maps

from all yearbooks between 1972 and 1985; minus 1979 and 1984, as we could not

locate these maps in the collection of yearbooks found in the Baku APS facility. All

photographs were captured as digital *.jpg images and organized by year. A small

Olympus digital camera was used to capture high resolution photos. Each page for

each yearbook was photographed two times to allow for quality comparisons. The

digital camera was held near perpendicular to the map image laid flat on a table. All

maps were cataloged according to translated titles and legends.

Each map image was georeferenced using the georeferencing tools in ArcGIS v

9.3.1 (ESRI, Redlands CA). All maps were referenced back to a 1-m polygon outline of

Azerbaijan from the Geo-Community Geoportal. Each image was georeferenced to the

WGS 1984 geographic projection of the world provided by ESRI using a minimum of

eight control points. Control points are coordinate pairs assigned by the user that match

the map position to real world coordinates. We employed an n+1 control point

experiment to determine the most suitable number of control points from 5 to 25 to

accurately georeferenced the images. There was minimal improvement after eight

control points. We use the root-mean-square error (RMS) and high spatial overlap

30

between known Azerbaijani features (such as political boundaries) from existing GIS

data and features on the images to minimize georeferencing error. Once images were

georeferenced (with quality control), we constructed new GIS data layers to match each

item in the map legend (points, line, and polygons) and then heads-up digitized these

shapes into spatially-referenced shapefiles (Curtis, Blackburn and Sansyzbayev 2006);

the shapefile is an ESRI proprietary data format that is used worldwide. In this way, a

single usable GIS layer was created for each map item.

STAMP

A STAMP analysis was used to explore the spatial and temporal variability of M.

libycus abundance across the landscape from 1972 to 1985. A separate STAMP

analysis was conducted on each of the five abundance categories. Each STAMP

analysis included polygons from every year that data were available for each

abundance category. The very low abundance analysis incorporated 1972 to 1978, as

very low and low abundance were collapsed after this time period. If a year was

unavailable the next available year was used in its place. For example, 1979 was

unavailable for all abundance categories, so polygons representing 1978 were

compared to 1980. STAMP is a freely available toolbar extension for ArcGIS v 9.3

(http://www.geog.uvic.ca/spar/stamp/help/index.html).

STAMP works by overlaying polygons from two consecutive time periods. The

change between intervals is defined based on the type of spatial overlap (Figure 2-5). If

polygons from two consecutive time periods are overlapping that area is classified as

contraction, or expansion. Contraction indicates that the area was only present in the

first time interval, while expansion indicates that a new area was created in the second

time interval. Disappearance and generation events are spatially isolated, which means

31

that they are not connected to any region delineating the opposing time period.

Disappearance is only present in the first time period, while generation is only found in

the second time period. Stable is an area common to both time periods.

STAMP requires spatially unique polygons. To prepare digitized data, the erase

tool was used to create unique polygons for each abundance level for each year.

Georeferenced images of hand drawn maps were consulted to ensure the accurate

categorization of polygons. All polygon layers were then projected

(WGS_1984_UTM_zone_38N). Finally, multipart features were separated into single

part features, which allowed STAMP to evaluate each polygon independently.

The STAMP process is comprised of four steps (Robertson et al. 2007). The first

step is to ensure that polygon’s input layers are spatially exclusive. The second step

creates change layers and assigns them to one of the five event categories: stable,

generation, expansion, disappearance or contraction (Robertson et al. 2007). The third

step uses the change layers and computes the directional relationship between

polygons by creating voronoi polygons (Robertson et al. 2007). The final step, polygon

metrics, calculates the metrics and ratios from the change layers (Robertson et al.

2007). The area of each abundance category for each year was reported in square

kilometers.

Stable Regions

The STAMP analyses identified regions that remained stable between

consecutive years. Those outputs were used to evaluate inter-annual variability for

each category across all consecutive time periods. We were also interested in

identifying areas characterized by long-term persistence. To identify stable regions

spanning more than two years, stable event polygons were extracted from all STAMP

32

outputs for each abundance level. Overlapping stable layers from consecutive years

within each abundance category were identified using the intersect tool in ArcGIS v 10.

The overlapping portion of the polygon was used to create regions of stability. These

stable regions were then used to evaluate environmental conditions associated with

each abundance category.

Environmental Data

STAMP allowed us to identify regions where a particular abundance category

persisted for an extended time interval, and we were interested in determining if there

were climatic conditions associated with different levels of stable M. libycus abundance.

Interpolated world climate surfaces were used to explore potential relationships

between the environment and stable regions. WorldClim climatic data are freely

accessible on the WorldClim website (www.WorldClim.org) (Hijmans et al. 2005).

These data are available at a 1-km spatial resolution. WorldClim data were derived

from a network of ground based weather stations, and reflect data collected between

1950 and 2000 (Hijmans et al. 2005). A smoothing spline algorithm was used for

interpolation using latitude, longitude and elevation as independent variables (Hijmans

et al. 2005). WorldClim includes over 20 variables describing climate in a variety of

ways. Some of these variables include precipitation in the wettest month, precipitation

in the driest month, minimum temperature of coldest month, and mean diurnal range.

Annual mean temperature (BIO1), annual precipitation (BIO12), and altitude were

downloaded for analyses.

Similarly to our climate analysis, we attempted to identify unique land cover types

associated with different levels of stable rodent abundance. We obtained land cover

data from GlobCover v 2.2 (ESA GlobCover project). GlobCover is a land cover dataset

33

at a 300 meter resolution with a thematic resolution of 22 classes

(http://www.ionia1.esrin.esa.int). The data was created from satellite images obtained

from 2005 to 2006. GlobCover has been employed for a variety of recent scientific

analyses. Linard et al. (2011) found that using GlobCover resulted in the most accurate

human population modelling across Africa compared to other available land cover

datasets. Falcucci et al. (2013) incorporated GlobCover data into their model of wolf re-

colonization of the alpine range. The GlobCover dataset identified 16 land cover

classifications found throughout Azerbaijan. We collapsed the land cover categories

comprising a small percentage of Azerbaijan, which resulted in 8 categories used for

analyses: rain-fed cropland, mosaic cropland, deciduous forest, evergreen forest, mixed

forest, shrubland/grassland, sparse/barren, and water (Figure 2-1, Table 2-2).

Environmental Analysis

Climatic data were paired with stable regions of abundance for each category to

explore relationships between climatic conditions and M. libycus abundance. Identifying

unique characteristics related to different levels of abundance could help to predict

abundance levels for modern surveillance efforts. To make WorldClim data compatible

with stable regions, the area of all stable regions from each abundance category were

combined using the merge and dissolve tools in ArcGIS v 10. Temperature,

precipitation, and altitude values were extracted from polygons representing the stable

area for each abundance level using the raster clip tool in ArcGIS v 10. The resulting

attribute tables were imported to the R statistical package for analyses (http://www.r-

project.org/). The maximum value, minimum value, mean, median and range were

calculated for each variable at each abundance level. The Mann Whitney U test was

34

used to determine if climatic conditions were significantly different for each abundance

level.

The Mann Whitney U test is a nonparametric test that compares the median

differences between two distributions (Crichton 2000). This is an appropriate test when

data being compared are naturally linked and do not follow a normal distribution

(Crichton 2000). The null hypothesis of the Mann Whitney U test is that the median

difference between paired values is equal to zero (Wilcoxon 1945, Mann and Whitney

1947). This test, run in R, was used to compare each abundance category with the

other abundance categories for all three environmental variables. The level of

confidence was set at 95%.

To identify distinctive land cover descriptions for each abundance level, the land

cover data set was clipped to each of the polygons representing stability for each level

of abundance to determine if there were unique land cover conditions for different

abundance levels. The total number of raster cells within each polygon was calculated.

The proportion of each land cover type for each abundance category was computed by

dividing the number of cells representing each land cover type by the total number of

raster cells within the polygon.

Human Population Analysis

The risk of human infection should be considered when prioritizing regions on the

landscape for control and preventative measures, as human plague risk is inherently

linked to the proximity of human populations to the pathogen. Toward this, the objective

of this analysis was to inform surveillance of natural plague foci throughout Azerbaijan.

History Database of the Global Environment v 3.1 (HYDE) gridded time series

population data were downloaded (http://themasites.pbl.nl/tridion/en/themasites/hyde/)

35

in an effort to explore the relationship between M. libycus abundance and human

population levels (Klein Goldewijk et al. 2011). HYDE includes georeferenced historical

gridded population data for the past 12,000 years at a 5’ (9.56 km) resolution (Goldewijk

2005). The data were derived from a combination of population statistics and satellite

information with specific allocation algorithms that change over time (Klein Goldewijk et

al. 2011). Available population totals for each country were crosschecked with other

sources, which were primarily local country studies (Goldewijk 2005). The Populstat

database (Lahmeyer 2000), and Gazetteer (2004) were the two primary sources

employed for deriving HYDE population maps from 1700 to 2000 (Goldewijk 2005).

Gaps in data were overcome by interpolating data points (Goldewijk 2005). HYDE data

has a variety of applications including environmental assessments, and biological

diversity analyses. The Global environmental outlook for the United Nations

Environmental Program (UNEP, 1997), have used HYDE data. Gaston et al. (2003)

used HYDE data in their analysis of global avian biodiversity loss.

In an effort to understand how population levels have changed from our historic

study interval to modern times, the population grids for 1970 and 2005 (the most recent

available surface) were downloaded. The 2005 surface was projected to 2010 using the

United Nations, medium variant, inter-censal growth rate by country (UNPD world

population prospects), which is in line with the methods of Hay et al.(2009). The 2010

population surface was derived using

Where is the population within each pixel, represents the population at year x in

the same pixel, r is average growth rate, and t is the number of years between 2010 and

36

x (Hay et al. 2009). The resulting 2010 population grid was used to create a surface

representing the percentage change in population from 1970 to 2010. All calculations

were computed using the raster calculator in ArcGIS v 10.

37

Table 2-1. List of host and vector species found in annual yearbooks

Common Name Genus Species

Hosts

Libyan jird Meriones M. libycus

Persian jird Meriones M. persicus

Tristram's jird, Malaysian jird Meriones M. tristami

Vinogradovi's jird Meriones M. vinogradovi

Common Vole Microtus M. arvalis

Social Vole Microtus M. socialis

Vectors

Fleas Callopsylla C. caspius

Ctenophthalmus C. wladimiri

Nosopsyllus N. consilimis

Nosopsyllus N. iranus

Nosopsyllus N. laericeps

Xenopsylla X. conformis

38

Table 2-2. GlobCover classifications for land cover types found throughout Azerbaijan and definitions of land cover classifications. The definitions report the combined definitions of categories that were collapsed.

Name Definition

Mixed Forest Closed to open (>15%) mixed broadleaved and needleleaved forest (>5m)

Deciduous Forest Closed or open Broadleaved Deciduous Forest (>5m)

Evergreen Forest Closed or open Needleleaved Deciduous or Evergreen Forest (>5m)

Shrubland / Grassland Mosaic grassland, shrubland (broadleaved or needleleaved, evergreen or deciduous, <5m), herbaceous vegetation (grassland, savannas or lichens/mosses)

Rainfed Cropland Rainfed Croplands

Mosaic Cropland and Vegetation 20 - 70% Cropland, 20 - 70% vegetation (grassland/shrubland/forest)

Sparse / Barren Sparse (<15%) vegetation or bare areas

Water Water Bodies

39

Figure 2-1. Azerbaijan with land cover, bordering countries and important rayons identified. The land cover categories were derived from GlobCover v 2.2, and reflect the final eight categories resulting from collapsing the 16 categories identified in Azerbaijan by the GlobCover surface.

40

Figure 2-2. Plague focus and protected areas in Azerbaijan. The focus is defined by M. libycus natural habitat in the area.

The protected areas are depicted based on the year of their designation.

41

Note: Photos courtesy of Jason Blackburn

Figure 2-3. Extracting data from annual yearbooks in the APS archival library in Baku, Azerbaijan

42

bottom map is from 1981.

Note: Photos courtesy of Jason Blackburn Figure 2-4. Hand drawn maps delineating Meriones libycus abundance with translated

legends from annual yearbooks. A) 1972 Meriones libycus abundance. B) 1980 Meriones libycus abundance.

A

Libyan Gerbil

0 – 1 Specimens per ha 1 – 5 Specimens per ha 6 – 10 Specimens per ha 11 – 20 Specimens per ha More than 20 Specimens per ha Settlement type – pesthole

Very High

Low

Average

High

Common Vole

Vinograd’s Gerbil

B

43

Figure 2-5. Event definitions for Space Time Analysis of Moving Polygons (STAMP).

Each circle represents an area in one of two consecutive time periods. Concentration, stable, and expansion are defined by overlapping areas, but contraction is only in time period one while expansion is only in time period two. Disappearance and generation are events that are spatially isolated from other areas.

Close Spatial Proximity

Time 1 Contraction

Time 2 Expansion

Time 1 + Time 2 Stable

Time 1 Disappearance

Time 2 Generation

Spatially Independent

44

CHAPTER 3 RESULTS

STAMP

There were no data available on M. libycus abundance for 1979, or 1984 in any

abundance category. Every year from 1972 to 1985 included M. libycus abundance for

low, average, and high abundance categories. Very low abundance data was available

from 1972 to 1978. Very high abundance was available for every year except 1973,

1976, and 1982 (Table 3-1).

The boundaries for each abundance category changed every year (Table1).

There were no years when the area of any abundance category was completely stable.

Low rodent abundance had the largest average area over the study interval. Very high

abundance had the smallest area over the study area. Very high abundance was

frequently found close to Samukh. Low abundance spanned the entire central portion

of the country. The area calculations for all years and all abundance categories are

shown in Table 2. All inter-annual STAMP outputs are shown in Appendix A.

Stable Regions

Stable regions spanning more than two years were identified for every

abundance category (Figure 3-2). Regions of low stable abundance were the most

widespread across Azerbaijan, and tended to persist longer than other abundance

categories. Stable regions of low abundance were concentrated directly east of Baku,

and bordered the lesser Caucasus region. The longest period of stability was a low

abundance region that persisted from 1976 to 1985. Stable areas associated with

average and high abundance did not consistently persist for more than five years.

Areas of average abundance were more widespread, and tended to be smaller than low

45

and very low abundance stable regions. Average abundance stable regions were the

most prominent in the Samukh region, and the western border of Baku. All stable

regions defining high and very high abundance were in Samukh. There were four

regions associated with stable high abundance, and only one region of stability was

identified for very high abundance.

Environmental Data

Precipitation, temperature, and altitude ranges across Azerbaijan are shown in

Figure 3-3. The two Caucasus mountain ranges in Azerbaijan are characterized by

unique environmental conditions compared to the rest of the country. The highest

altitudes are found in these regions in addition to the highest precipitation levels and

lowest temperatures. The central portion of Azerbaijan, from the Samukh rayon to

south of Baku, showed relatively stable levels of precipitation, temperature, and altitude.

Environmental Analysis

Altitude, temperature, and precipitation variables were extracted and analyzed for

each of the five abundance categories. The Mann Whitney U test revealed that there

was not a significant difference between very high and high for any of the environmental

variables (Table 3-2). Data on very low abundance was not available after 1978. As a

result, we were compelled to collapse very low and low into one category and high and

very high into another category for environmental analyses. This resulted in three

categories: low, average, and high abundance (Figure 3-4). These three abundance

categories were used for all subsequent analyses.

Each abundance category was characterized by significantly different climatic

and land cover conditions. These distinctive characteristics can help identify regions

that should be prioritized for surveillance and control. Stable areas of high abundance

46

had the highest average annual temperature, altitude, and annual precipitation (Table 3-

3). Stable low regions had the lowest average annual temperature and altitude. Stable

average abundance had the highest average precipitation, as a result of its overlap with

the Lesser Caucasus mountain range along the western border of the country. Low

abundance stretched up the northeast border of Azerbaijan and into the Greater

Caucasus region, across a wider range of temperatures, altitudes and precipitation

levels than the other two abundance categories. Stable high abundance was limited to

Samukh, and was characterized by the smallest ranges for all three variables. All

abundance levels were significantly different for each environmental variable except

temperature in high abundance regions compared to temperature in low abundance

regions (Table 3-4).

The region associated with each abundance category had a different land cover

composition. The landscape delineating low rodent abundance was dominated by

mosaic cropland based on GlobCover characterizations (Figure 3-5, Table 3-5). High

abundance regions were dominated by shrubland/grassland and sparse/barren land

cover. The landscape associated with average abundance was approximately 30%

mosaic cropland and 22% grassland / shrubland. Mixed forest, evergreen, and

deciduous all comprised less than 1% of total land cover for each abundance category,

but are the most prominent in low abundance.

Human Population

The highest percentage increase in human populations was along the southern

border of the plague focus in the lesser Caucus region (Figure 3-6, Appendix B). The

area surrounding Samukh, which included the stable region of high abundance, was

characterized by approximately 50% increases in population levels since 1970. The

47

central portion of Azerbaijan also experienced approximately 50% increase in

population levels over the 40 year period. The Greater Caucus region, which borders

stable low abundance, was another region that experienced up to a 100% increase in

human population between 1970 and 2010.

48

Table 3-1. Changes in the area representing each abundance category over time. The total area associated with each

abundance category for each year is reported.

Note: ND = No Data

Year Very Low Abundance Area (Sq KM)

Low Abundance Area (Sq Km)

Average Abundance Area (Sq Km)

High Abundance Area (Sq Km)

Very High Abundance Area (sq KM)

1972 4859.67 15333.38 5419.4 717.91 305.05

1973 8738.92 13904.51 8478.61 6354.78 0

1974 4050.69 25070.44 5659.06 2639.57 290.63

1975 22564.16 9440.97 4178.55 1128.08 235.61

1976 10604.22 6824.64 2370.56 580.41 0

1977 15417.32 4303.45 2072.98 501.02 77.37

1978 7025.66 7687.52 2205.41 1026.39 165.53

1980 ND 27657.74 2853.18 626.12 241.45

1981 ND 21291.7 5033.66 544.52 948.64

1982 ND 28038.79 6114.36 56.77 0

1983 ND 30504.17 3815.76 6.64 160.36

1985 ND 29033.75 2108.9 305.42 172.22

49

Table 3-2. Wilcoxon signed rank test results for temperature, precipitation, and altitude comparing five abundance

categories. The null hypothesis being tested is that there is no significant difference in environmental variables between abundance categories.

* Reject the null hypothesis with 95% confidence ** Reject the null hypothesis with 99% confidence

High v. Very high

High v. Average

High v. Low

High v. Very Low

Very High v.

Average

Very High v.

Low

Very High v. Very Low

Average v. Low

Average v. Very Low

Low v. Very Low

Temperature

Statistic 667 394278 2780926 846997 43117 313012 98752 12856601

9 38356594 196215207

P.value 0.44 0.07 <0.001** <0.001** 0.57 0.04* <0.001** <0.001** <0.001** <0.001**

Precipitation

Statistic 958 2775322 3028050 752339 353292 392333 95538 55356675

6 146705200 173157853

P.value 0.14 <0.001** <0.001** <0.001**

<0.001** <0.001** <0.001** <0.001** <0.001** <0.001**

Altitude

Statistic 840 461061 3234434 712025 50585 360680 79071 13316060

6 29822902 149264722

P.value 0.58 <0.001** <0.001** <0.001** 0.08 <0.001** 0.06 <0.001** <0.001** <0.001**

50

Table 3-3. Descriptive statistics for the distribution of temperature, precipitation, and altitude associated with each of the three abundance levels.

Low Average High

Temperature (Degrees Celsius) Min. 10.1 10.1 13.7 1st Qu. 13.5 14 14.3 Median 14.2 14.6 14.5 Mean 14.03 14.31 14.55 3rd Qu. 14.7 14.8 14.8 Max. 15.4 15.3 15.2 Precipitation (Millimeters) Min. 243 243 337 1st Qu. 348 343 396 Median 373 371 416 Mean 377.9 369.7 406.9 3rd Qu. 398 395 429.5 Max. 672 601 473 Altitude (Meters) Min. -42 -32 70 1st Qu. -7 15 154 Median 41 122 209 Mean 126.2 161.8 202.8 3rd Qu. 204 259 257.5 Max. 1023 993 402

51

Table 3-4. Mann Whitney U test results for temperature, precipitation, and altitude

comparing three abundance categories. The null hypothesis being tested is that there is no significant difference in environmental variables between abundance categories.

* Reject the null hypothesis with 95% confidence ** Reject the null hypothesis with 99% confidence.

High v. Average High v. Low Average v. Low

Temperature (Degrees Celsius)

Statistic 437395.00 4039687.00 167000000.00

P.value 0.05 >0.001** >0.001**

Precipitation (Millimeters)

Statistic 3128614.00 4268260.00 700271956.00

P.value >0.001** >0.001** >0.001**

Altitude (Meters)

Statistic 511645.00 4386209.00 162983507.50

P.value >0.001** >0.001** >0.001**

52

Table 3-5. The percentages of each land cover type that comprises the landscape for

high, medium, and low M. libycus abundance.

Land Cover Type Low Average High

Rainfed Cropland 11.13% 7.49% 7.92%

Mosaic Cropland 55.90% 30.10% 9.43%

Deciduous 0.06% 0.03% 0.00%

Evergreen 0.33% 0.20% 0.00%

Mixed Forest 0.20% 0.03% 0.00%

Grassland / Shrubland 14.60% 22.19% 29.51%

Sparse / Barren 16.30% 36.04% 36.38%

Water 1.48% 3.85% 16.75%

53

Figure 3-1. STAMP outputs from 1976 to 1977. A) is the change in very low abundance from 1976 to 1977; B) is low

abundance from 1976 to 1976. C) is average abundance from 1976 to 1977. D) is high abundance from 1976 to 1977. E) is very high abundance from 1976 to 1977.

54

Figure 3-2. Stable regions for each abundance category representing regions that remained stable for at least 2

consecutive years. A) shows the region of Azerbaijan that each of the subsequent maps focuses on. B) is very low abundance. C) is low regions of stability that persisted for at least 5 years. D) is average abundance stable regions. E) is high abundance stable regions. F) is very high abundance stable regions.

55

Figure 3-3. The distribution of environmental variables across Azerbaijan. A) shows temperature. B) is precipitation. C) is altitude across the country. All climatic data was downloaded from WorldClim.

56

Figure 3-4. Regions of stability for each abundance level. Each polygon represents the

area that was stable for at least 2 consecutive years anytime between 1972 and 1985. A) is low abundance. B) is average abundance. C) is high abundance.

57

Figure 3-5. The percentage of each land cover type associated with regions of high, average, and low rodent abundance

58

Figure 3-6. The percentage population change between 1970 and 2010. Dark red indicates an increase in population. Stable regions for high, average, and low abundance area are also shown.

59

CHAPTER 4 DISCUSSION

Plague presents a pressing public health concern in the developing world.

Plague has a history of devastating pandemics, and there has been an increase in

human cases in recent years. There is also evidence that climate change could further

increase human disease risk. This threat is particularly relevant in the former Soviet

Union, and its bordering countries. The collapse of the Soviet Union was accompanied

by a deteriorating public health system that was left with limited resources for disease

control and surveillance efforts. There is also evidence of active historical foci in

countries surrounding Azerbaijan. In Iran, a recent study found high numbers of host

and vectors in a historically active focus. Yersinia pestis was detected in rodent hosts,

which suggests that plague is being maintained in the environment (Esmaeili 2013).

Continued plague surveillance in these regions is essential.

The goal of this study was to inform plague surveillance in one environmental

focus in Azerbaijan. The first objective was to use a historic data set to evaluate

patterns of persistence in M. libycus abundance in Azerbaijan from 1972 – 1985. It has

been established in the literature that plague prevalence is related to the distribution

and density of hosts on the landscape, and Meriones libycus is an important plague

host in Azerbaijan (Gage and Kosoy 2005, Davis et al. 2004). We were particularly

interested in identifying geographic regions of long-term stability in M. libycus

abundance, as these regions of historical M. libycus persistence should be prioritized.

A second objective was to associate environmental and climatic characteristics with

different levels of M. libycus abundance. Environmental characteristics correlated to

different levels of abundance would provide a way to predict M. libycus across the

60

country. Similarly, stable environmental conditions have the potential to be associated

with stable levels of M. libycus abundance. Finally, we explored changes in human

population densities. Effectively prioritizing regions for surveillance must consider the

risk of human infection.

Overall, this study identified variability in the spatial distribution of M. libycus over

the 13 year study period in all abundance categories. Regions characterized by low M.

libycus abundance were the most stable over time with several regions persisting for

more than five years. For the other abundance categories, regions of stability did not

tend to persist for more than five years, but every abundance category had a stable

region that was maintained for at least two years. Stable regions of low rodent

abundance were the most widespread and continuous across Azerbaijan. Areas of

stable average rodent abundance were found on eastern and western portion of the

country, but were not as continuous. Landscapes consistently characterized by high

abundance were limited to Samukh in the northwest portion of Azerbaijan.

It is important to consider the variability in abundance that the STAMP analysis

established. There were no abundance levels that were spatially or temporally stable

over the entire study period. Abundance levels fluctuated within the boundaries

representing each category; however, those boundaries represent regions that have

been characterized by the persistence of one abundance level for multiple years at

some point between 1972 and 1985. From a surveillance perspective, these historically

stable regions can help inform surveillance by prioritizing regions with a history of higher

abundance levels. If a region was historically characterized by high abundance, modern

surveillance should prioritize that region for control efforts.

61

There were significant differences in environmental characteristics associated

with different rodent abundance categories. Higher counts of M. libycus were correlated

with the greatest amount of annual precipitation, higher mean temperatures, and higher

altitudes. There is little variability in temperature, precipitation, and altitude across the

region of Azerbaijan delineated by M. libycus data employed in this analysis, but these

fine scale differences are important when considered within the context of the entire

plague system. The distinguishing characteristics correlated with each abundance level

can help predict changes or continued stability in M. libycus abundance. If climate

remains stable, the levels of rodent abundance will likely reflect historical levels of M.

libycus abundance. If conditions become warmer and wetter then we might hypothesize

M. libycus abundance is more likely to increase.

It is also important to consider the relationship between plague prevalence and

host abundance. Ideally, control efforts would be prioritized for M. libycus populations

with active Y. pestis. Here we have shown a relationship between relatively warm, wet

climatic conditions and M. libycus host abundance in Azerbaijan. An association

between warm, moist climate and higher levels of flea vectors has also been discussed

in the literature (Davis 1953, Cavanaugh and Marshall 1972). Plague prevalence in

several places worldwide have also been correlated to warm springs and wet summers.

Plague is a complex system with numerous factors affecting prevalence and spread;

however, in general terms, high host abundance, high vector abundance and pathogen

presence have been related to similar, broad scale environmental conditions. It is also

important to consider that plague is more likely to persist in the environment with a

higher abundance of rodent hosts (Gage and Kosoy 2005, Stenseth et al. 2008). These

62

findings suggest that regions associated with stable high rodent abundance are likely at

a higher risk of plague prevalence. Our findings suggest that areas historically

categorized as high abundance should be prioritized for surveillance efforts because

they are the most likely to support plague prevalence.

The landscapes delineating each abundance category had different land cover

compositions. The land cover in Samukh, which contains stable high abundance, is

characterized by higher proportions of sparse/barren landscape, shrubland, and

grassland than the remainder of the country. In contrast, northwestern Azerbaijan has

more forested area and mosaic cropland than Samukh where high abundance stability

occurred. These land cover compositions can be used to prioritize landscapes more

likely to support high levels of plague hosts.

Several studies have suggested that human proximity to plague hosts is a risk

factor for human infection (Schotthoefer et al. 2012, Eisen et al. 2007b, Kartman 1970).

Schotthoefer et al. (2012) found that living close to environmental plague reservoirs in

New Mexico was a significant risk factor for human plague. They also found that

changing socioeconomic demographics were correlated to changes in the location of

plague cases (Schotthoefer et al. 2012). Essentially, changes in human population

dynamics are important to consider when prioritizing surveillance. In Azerbaijan, human

populations have increased in and around the plague focus. This is particularly

concerning in the area surrounding regions of historically high M. libycus abundance.

Increasing human populations, high levels of host abundance, and stable climate could

create an intensified human plague risk in this region. Locations with growing human

63

populations in close proximity to natural plague foci should be prioritized for surveillance

and control efforts.

The increasing number of protected lands, coupled with growing populations has

potential implications for plague risk in Azerbaijan. In other parts of the world, national

parks have been correlated with increased human disease risk. In Kyrgyzstan, the

highest risk region of tick borne encephalitis is in Ala-Archa National Park, which is

close to the capital city (Briggs et al. 2011). The park is a popular attraction for hikers,

climbers, and tourists who come into close proximity with the rodent hosts and tick

vectors that inhabit the park (Briggs et al. 2011). In California, human plague cases

have also been linked to host and vector populations in national parks. One model

predicting plague host distribution identified national parks in California as a probable

location for vector distribution (Adjemian et al. 2006). These predicted distributions also

reflected sites of previous human cases (Adjemian et al. 2006). In the 1970s, California

experienced an increase in human plague cases, and many of those cases were

attributed to host populations in national parks (Nelson 1980). Wilderness areas, and

national monuments were also locations marked as high risk for plague maintenance in

the state (Nelson 1980). The increase in National Parks in Azerbaijan has the potential

to provide an opportunity for plague hosts, such as M. libycus, and flea vectors to come

into contact with people visiting the park. National parks, particularly those in close

proximity to large population centers, should be prioritized for surveillance.

Working with historical data is inherently associated with challenges. Our original

data were in the form of hand drawn maps that were photographed, georeferenced and

hand digitized. It is likely that there were errors associated with this process. As a

64

result of the nature of the original data, the areas of polygons were not exact, and this

analysis was based on comparing the relative size difference in polygons. The historic

M. libycus dataset also contains some sampling bias. It is unclear if the same regions

were sampled consistently over the fourteen year period. We know that sampling efforts

were largely driven by expert opinion of known rodent habitat, which creates a biased

sampling design. It is possible that some of the spatial variation identified in M. libycus

abundance is due to this sampling bias. While it is important to consider the challenges

related to historic datasets, the benefits associated with analyzing historical data are

particularly relevant to the study of disease.

Similarly to the challenges we encountered with our data, the literature has

established sampling bias, or non-uniform reporting as the primary hurdle associated

with employing historic data sets (Cliff et al. 1997). Despite these complications,

historic data still has the potential to provide context for current outbreaks, and should

be used (Cliff, Haggett and Smallman-Raynor 2008). Cliff et al. (1997) liken the use of

historic public health records to the use of long term climate data. In order to effectively

improve short term forecasts, it is necessary to understand the long term patterns (Cliff

et al. 1997). The reemergence and resilience of disease across the globe is another

reason studying past patterns is important (Cliff et al. 1997, Cliff et al. 2008). Using

geographic information systems (GIS) is an effective method for analyzing historic data

because of its capacity to integrate data from different sources and dates (Gregory and

Healey 2007). GIS also provides a framework for exploring change over a historic time

interval (Gregory and Healey 2007). Cunfer (2005) used annual agricultural data and

environmental data sets to explore the cause of the great plains dust storm. Skinner et

65

al. (2000) also used GIS to analyze trends in fertility rates in China’s from the 1960s to

the 1990s.

Some additional limitations associated with this study include the STAMP

toolbar, which is not compatible with ArcGIS v 10, which is the most recent version. The

applicability of STAMP to a wide range of questions calls for an updated version of the

software. The GlobCover dataset employed was derived from satellite images taken in

2005 and 2006, which might not reflect conditions at the time this M. libycus data was

collected. Our research group is currently investigating historical land cover change in

Azerbaijan to determine the degree of land cover stability over time using a remote

sensing approach.

Despite these limitations, valuable information was obtained from this analysis.

Our findings are important and relevant for three primary reasons. Plague still affects

human populations worldwide, and needs to be carefully monitored (Stenseth et al.

2008, Perry and Fetherston 1997). Impending climate change has the capacity to

increase prevalence of gerbil hosts in the future, which can increase the frequency at

which plague can occur (Stenseth et al. 2006). Finally, changes in the public health

system that accompanied the collapse of the Soviet Union have limited resources

available for plague surveillance.

Our results can be used to prioritize limited resources for the surveillance and

control of environmental plague foci in Azerbaijan. Regions of historically high M.

libycus abundance should be prioritized. Regions of the country with climatic conditions

associated with high abundance should also be prioritized. Employing predictive

climate change models is one way to identify regions likely to experience a climatic shift

66

towards favorable plague conditions. Similarly, stable climate is likely to be associated

with stable rodent abundance. Regions with increasing human populations bordering

plague foci should be monitored, as these present a high risk of human infection.

Finally, protected lands with public access should be monitored because of the close

proximity of humans and vectors.

Future Directions

An important future direction will be to analyze M. libycus abundance at a finer

temporal scale. In order to thoroughly understand the impact of climate on rodent

abundance we need to study seasonal variations in M. libycus abundance and the

correlated climatic variation. Ideally, future sampling efforts could take place on a

seasonal basis throughout Azerbaijan. This effort would also require a climatic dataset

at a finer temporal scale to analyze seasonal variations. A modern analysis compared

to historical data would allow us to determine if a shift in abundance levels has taken

place since 1985. A more thorough understanding of stable regions could be obtained.

A modern data set describing rodent abundance throughout Azerbaijan would be a

necessary first step to meet these objectives.

The annual historic yearbooks contain data on multiple other hosts and vectors.

The exploration of these additional data sets could provide a more thorough picture of

how plague hosts and vectors are spread across the landscape. Incorporating

yearbook data that includes lab analysis and spatial information on plague isolates

would provide an informative element to this analysis. In order to evaluate the plague

risk in Azerbaijan, we need to understand the roles of all hosts, vectors, and the

prevalence of the pathogen in the region

67

APPENDIX A STAMP OUTPUTS

All STAMP outputs for all years and all five categories (very low, low, average,

high, and very high) of M. libycus abundance are shown. A-1 includes Very low

abundance STAMP outputs from 1972 to 1978. Each STAMP output reflects the

change between two consecutive years. For example, the change in area representing

very low M. libycus abundance from 1972 to 1973 is one map, and the change in area

from 1973 to 1974 is shown on a separate map. A-2 includes the STAMP outputs for

low abundance from 1972 to 1985, A-3 is the STAMP outputs for average abundance

from 1972 to 1985, A-4 includes STAMP outputs for high abundance from 1972 to 1985,

and A-5 is very high abundance STAMP outputs from 1972 to 1985.

68

Figure A-1. STAMP outputs for very low Meriones libycus abundance. A) through F) shows STAMP outputs from 1973 to 1978.

1972 1973

1973 1974

Contraction

Disappearance Expansion Generation Stable

A

B

69

1974 1975

1975 1976

Contraction

Disappearance Expansion Generation

Stable

C

D

Figure A-1 Continued

70

Figure A-1 Continued

1976 1977

E

1977 1978

Contraction Disappearance Expansion Generation Stable

F

71

Figure A-2. STAMP outputs for low Meriones libycus abundance. A) to K) shows STAMP outputs from 1973 to 1985.

Contraction

Disappearance Expansion Generation

Stable

A

B

1972 1973

1973 1974

72

Contraction

Disappearance Expansion Generation

Stable Figure A-2 Continued

C

D

1974 1975

1975 1976

73

Contraction

Disappearance Expansion Generation

Stable Figure A-2 Continued

E

F

1976 1977

1977 1978

74

Contraction

Disappearance Expansion Generation

Stable Figure A-2 Continued

G

H

1978 1980

1980 1981

75

Contraction

Disappearance Expansion Generation

Stable Figure A-2 Continued

I

J

1981 1982

1982 1983

76

Contraction

Disappearance Expansion Generation

Stable

Figure A-2 Continued

K

1983 1985

77

Figure A-3. STAMP outputs for average Meriones libycus abundance. A) to K) shows STAMP outputs from 1973 to 1985.

Contraction

Disappearance Expansion Generation

Stable

A

B

1972 1973

1973 1974

78

Contraction

Disappearance Expansion Generation

Stable Figure A-3 Continued

C

D

1974 1975

1975 1976

79

Contraction

Disappearance Expansion Generation

Stable Figure A-3 Continued

E

F

1977 1978

1976 1977

80

Contraction

Disappearance Expansion Generation

Stable Figure A-3 Continued

G

H

1978 1980

1980 1981

81

Contraction

Disappearance Expansion Generation

Stable Figure A-3 Continued

I

J

1981 1982

1982 1983

82

Contraction

Disappearance Expansion Generation

Stable

Figure A-3 Continued

K

1983 1985

83

Figure A-4. STAMP outputs for high Meriones libycus abundance. A) to K) shows STAMP outputs from 1973 to 1985.

Contraction

Disappearance Expansion Generation

Stable

A

B

1972 1973

1973 1974

84

Contraction

Disappearance Expansion Generation

Stable Figure A-4 Continued

C

D

1974 1975

1975 1976

85

Contraction

Disappearance Expansion Generation

Stable Figure A-.4 Continued

E

F

1976 1977

1977 1978

86

Contraction

Disappearance Expansion Generation

Stable Figure A-4 Continued

G

H

1978 1980

1980 1981

87

Contraction

Disappearance Expansion Generation

Stable Figure A-4 Continued

I

J

1981 1982

1982 1983

88

Contraction

Disappearance Expansion Generation

Stable

Figure A-4 Continued

K

1983 1985

89

Figure A-5. STAMP outputs for very high Meriones libycus abundance. A) to H) shows STAMP outputs from 1973 to 1985.

Contraction

Disappearance Expansion Generation

Stable

A

B

1972 1974

1974 1975

90

Contraction

Disappearance Expansion Generation

Stable Figure A-5 Continued

C

D

1975 1977

1977 1978

91

Contraction

Disappearance Expansion Generation

Stable Figure A-5 Continued

E

F

1978 1980

1980 1981

92

Contraction

Disappearance Expansion Generation

Stable Figure A-5 Continued

G

H

1981 1983

1983 1985

93

APPENDIX B POPULATION DISTRIBUTION

Figure B-1. HYDE population grids showing the population distribution in Azerbaijan.

A) for 1970. B) 2010.

94

LIST OF REFERENCES

Adjemian, J. C. Z., E. H. Girvetz, L. Beckett & J. E. Foley (2006) Analysis of Genetic Algorithm for Rule-Set Production (GARP) Modeling Approach for Predicting Distributions of Fleas Implicated as Vectors of Plague, Yersinia pestis, in California. Journal of medical entomology, 43, 93-103.

Anisimov, A. (2002) Factors of Yersinia pestis providing circulation and persistence of

plague pathogen in ecosystems of natural foci. Communication 2. MOLECULAR GENETICS MICROBIOLOGY AND VIROLOGY C/C OF MOLEKULIARNAIA GENETIKA MIKROBIOLOGIIA I VIRUSOLOGIIA, 1-14.

Arbaji, A., S. Kharabsheh, S. Al-Azab, M. Al-Kayed, Z. Amr, M. Abu Baker & M. Chu

(2005) A 12-case outbreak of pharyngeal plague following the consumption of camel meat, in north–eastern Jordan. Annals of tropical medicine and parasitology, 99, 789-793.

Bacot, A. W. & C. Martin (1914) LXVII. Observations on the mechanism of the

transmission of plague by fleas. The Journal of hygiene, 13, 423. Bakanidze, L., P. Imnadze, S. Chubinidze, N. Tsertsvadze, G. Mgeladze, I.

Shalutashvili, S. Tsanava, M. Shavishvili, J. Manvelyan & N. Ninashvili (2010) Surveillance on Plague in Natural Foci in Georgia. Emerging and Endemic Pathogens, 21-28.

Ben Ari, T., S. Neerinckx, K. L. Gage, K. Kreppel, A. Laudisoit, H. Leirs & N. C.

Stenseth (2011) Plague and Climate: Scales Matter. PLoS Pathog, 7, e1002160. Beran, G. W. 1994. Handbook of Zoonoses, Section B: Viral Zoonoses. CRC. Bertherat, E., S. Bekhoucha, S. Chougrani, F. Razik, J. B. Duchemin, L. Houti, L.

Deharib, C. Fayolle, B. Makrerougrass & R. Dali-Yahia (2007) Plague reappearance in Algeria after 50 years, 2003. Emerging Infectious Diseases, 13, 1459.

Bibikova, V. (1977) Contemporary views on the interrelationships between fleas and the

pathogens of human and animal diseases. Annual review of entomology, 22, 23-32.

Briggs, B. J., B. Atkinson, D. M. Czechowski, P. A. Larsen, H. N. Meeks, J. P. Carrera,

R. M. Duplechin, R. Hewson, A. T. Junushov & O. N. Gavrilova (2011) Tick-Borne Encephalitis Virus, Kyrgyzstan. Emerging Infectious Diseases, 17, 876.

Cavanaugh, D. C. & J. D. Marshall (1972) The Influence of Climate On the Seasonal

Prevalence of Plague in the Republic of Vietnam. Journal of wildlife diseases, 8, 85-94.

95

CEPF. 2003. Ecosystem Profile for the Caucasus Biodiversity Hotspot. Cliff, A. D., P. Haggett & M. Smallman-Raynor (2008) An exploratory method for

estimating the changing speed of epidemic waves from historical data. International journal of epidemiology, 37, 106-112.

Cliff, A. D., P. Haggett, M. Smallman-Raynor, D. F. Stroup & G. D. Williamson (1997)

The importance of long-term records in publichealth surveillance: the US weekly sanitaryreports, 1888–1912, revisited. Journal of Public Health, 19, 76-84.

Conrad, F. G., F. R. LeCocq & R. Krain (1968) A recent epidemic of plague in Vietnam.

Archives of internal Medicine, 122, 193-198. Crichton, N. (2000) Wilcoxon signed rank test. Journal of Clinical Nursing, Blackwell

Science Ltd. Cunfer, G. 2005. On the Great Plains: agriculture and environment. TAMU Press. Curtis, A., J. Blackburn & Y. Sansyzbayev (2006) Using a geographic information

system to spatially investigate infectious disease. Encyclopedia of Infectious Diseases: Modern Methodologies, 405-424.

Daly, M. & S. Daly (2009) Socio-ecology of Saharan gerbils, especially Meriones

libycus. Mammalia, 39, 289-312. Davis, D. (1953) Plague in Africa from 1935 to 1949: a survey of wild rodents in African

territories. Bulletin of the World Health Organization, 9, 665. Davis, S., M. Begon, L. De Bruyn, V. S. Ageyev, N. L. Klassovskiy, S. B. Pole, H.

Viljugrein, N. C. Stenseth & H. Leirs (2004) Predictive thresholds for plague in Kazakhstan. Science, 304, 736-738.

Dennis, D. (1994) Plague in India. BMJ, 309, 893-894. Dennis, D. T. (1998) Plague as an emerging disease. Emerging infections, 169-184. Desjeux, P. (2001) The increase in risk factors for leishmaniasis worldwide.

Transactions of the Royal Society of Tropical Medicine and Hygiene, 95, 239-243.

Eisen, R. J., R. E. Enscore, B. J. Biggerstaff, P. J. Reynolds, P. Ettestad, T. Brown, J.

Pape, D. Tanda, C. E. Levy & D. M. Engelthaler (2007a) Human plague in the southwestern United States, 1957-2004: spatial models of elevated risk of human exposure to Yersinia pestis. Journal of medical entomology, 44, 530-537.

96

Eisen, R. J., P. J. Reynolds, P. Ettestad, T. Brown, R. E. Enscore, B. J. Biggerstaff, J. Cheek, R. Bueno, J. Targhetta & J. A. Montenieri (2007b) Residence-linked human plague in New Mexico: a habitat-suitability model. The American journal of tropical medicine and hygiene, 77, 121-125.

Eskey, C. R. & V. Haas (1940) Plague in the western part of the United States. Public

Health Bulletin. Esmaeili, S. A., Kayhan; Naddaf, Saied Reza; Rajerison, Minoariosa; Carniel, Elisabeth;

Mostafavi, Ethsan (2013) A serological Survey of Plague in Animals in Western Iran. Emerging Infectious Diseases.

Falcucci, A., L. Maiorano, G. Tempio, L. Boitani & P. Ciucci (2013) Modeling the

potential distribution for a range-expanding species: Wolf recolonization of the Alpine range. Biological Conservation, 158, 63-72.

Gage, K. L. & M. Y. Kosoy (2005) NATURAL HISTORY OF PLAGUE: Perspectives

from More than a Century of Research*. Annu. Rev. Entomol., 50, 505-528. Gage, K. L., R. S. Ostfeld & J. G. Olson (1995) Nonviral vector-borne zoonoses

associated with mammals in the United States. Journal of Mammalogy, 695-715. Gaston, K. J., T. M. Blackburn & K. K. Goldewijk (2003) Habitat conversion and global

avian biodiversity loss. Proceedings of the Royal Society of London. Series B: Biological Sciences, 270, 1293-1300.

Goldewijk, K. K. (2005) Three centuries of global population growth: A spatial

referenced population (density) database for 1700–2000. Population & Environment, 26, 343-367.

González, U., M. Pinart, M. Rengifo-Pardo, A. Macaya, J. Alvar & J. Tweed (2009)

Interventions for American cutaneous and mucocutaneous leishmaniasis (Review). Cochrane Database Syst Rev, 2.

Gratz, N. (1999) Rodent reservoirs and flea vectors of natural foci of plague. Plague

manual: epidemiology, distribution, surveillance and control, 63-96. Gregory, I. N. & R. G. Healey (2007) Historical GIS: structuring, mapping and analysing

geographies of the past. Progress in Human Geography, 31, 638-653. Gurbanov, S. & S. Akhmedova (2010) Especially Dangerous Infections in Azerbaijan.

Emerging and Endemic Pathogens, 39-43. Hay, S. I., C. A. Guerra, P. W. Gething, A. P. Patil, A. J. Tatem, A. M. Noor, C. W.

Kabaria, B. H. Manh, I. R. F. Elyazar & S. Brooker (2009) A world malaria map: Plasmodium falciparum endemicity in 2007. PLoS Medicine, 6, e1000048.

97

Hijmans, R. J., S. E. Cameron, J. L. Parra, P. G. Jones & A. Jarvis (2005) Very high resolution interpolated climate surfaces for global land areas. International journal of climatology, 25, 1965-1978.

Hubbert, W. T. & M. I. Goldenberg (1970) Natural resistance to plague: genetic basis in

the vole (Microtus californicus). American Journal of Tropical Medicine and Hygiene, 19, 1015-19.

Hull, H. F., J. M. Montes & J. M. Mann (1987) Septicemic plague in New Mexico.

Journal of Infectious Diseases, 155, 113-118. Ibrahim, E., M. Mustafa, S. Al Amri, S. AlSeghayer, S. Hussein & L. Gradoni (1994)

Meriones libycus (Rodentia: Gerbillidae), a possible reservoir host of zoonotic cutaneous leishmaniasis in Riyadh province, Saudi Arabia. Transactions of the Royal Society of Tropical Medicine and Hygiene, 88, 39-39.

Kartman, L. (1970) Historical and oecological observations on plague in the United

States. Tropical and geographical medicine, 22, 257. Klein Goldewijk, K., A. Beusen, G. Van Drecht & M. De Vos (2011) The HYDE 3.1

spatially explicit database of human‐induced global land‐use change over the past 12,000 years. Global Ecology and Biogeography, 20, 73-86.

Korslund, L. & H. Steen (2005) Small rodent winter survival: snow conditions limit

access to food resources. Journal of Animal Ecology, 75, 156-166. Lahmeyer, J. 2000. Population Statistics: Growth of the population per country in a his

torical perspective, including their administrative divisions and principal towns. Leslie, T., C. Whitehouse, S. Yingst, C. Baldwin, F. Kakar, J. Mofleh, A. Hami, L.

Mustafa, F. Omar & E. Ayazi (2011) Outbreak of gastroenteritis caused by Yersinia pestis in Afghanistan. Epidemiology and Infection, 139, 728-735.

Linard, C., M. Gilbert & A. J. Tatem (2011) Assessing the use of global land cover data

for guiding large area population distribution modelling. GeoJournal, 76, 525-538. Mann, H. B. & D. R. Whitney (1947) On a test of whether one of two random variables is

stochastically larger than the other. The annals of mathematical statistics, 18, 50-60.

Mann, J. M., W. J. Martone, J. M. Boyce, A. F. Kaufmann, A. M. Barnes & N. S. Weber

(1979) Endemic human plague in New Mexico: risk factors associated with infection. Journal of Infectious Diseases, 140, 397-401.

Meyer, K. (1942) The ecology of plague. Medicine, 21, 143-174.

98

Nadim, A. & M. Faghih (1968) The epidemiology of cutaneous leishmaniasis in the Isfahan province of Iran: I. The reservoir II. The human disease. Transactions of the Royal Society of Tropical Medicine and Hygiene, 62, 534-542.

Naumov, N. P., V. S. Lobachev, P. P. Dmitriev, V. Kanatov Iu & V. M. Smirin (1973)

[Experience in studying the rate of spread and paths of transmission of plague epizootics in the Northern Desert]. Zh Mikrobiol Epidemiol Immunobiol, 50, 78-85.

Neerinckx, S., A. T. Peterson, H. Gulinck, J. Deckers, D. Kimaro & H. Leirs (2010)

Predicting potential risk areas of human plague for the Western Usambara Mountains, Lushoto District, Tanzania. Am J Trop Med Hyg, 82, 492-500.

Nelson, B. C. (1980) Plague studies in California—the roles of various species of

sylvatic rodents in plague ecology in California. Nelson, T. A. (2011) Quantifying Wildlife Home Range Changes. Modern Telemetry. Nowak, R. M. & J. L. Paradiso. 1999. Walker's mammals of the world. Cambridge Univ

Press. Ouagrham-Gormley, S. B. (2006) Growth of the Anti-plague System during the Soviet

Period. Critical reviews in microbiology, 32, 33-46. Parmenter, R. R., E. P. Yadav, C. A. Parmenter, P. Ettestad & K. L. Gage (1999)

Incidence of plague associated with increased winter-spring precipitation in New Mexico. The American journal of tropical medicine and hygiene, 61, 814-821.

Pavlovsky, E. N., F. K. Plous JR & N. D. Levine (1966) Natural nidality of transmissible

diseases. The American Journal of the Medical Sciences, 252, 161. Perry, R. D. & J. D. Fetherston (1997) Yersinia pestis--etiologic agent of plague. Clinical

microbiology reviews, 10, 35-66. Pollitzer, R. (1951) Plague studies: I. A summary of the history and a survey of the

present distribution of the disease. Bulletin of the World Health Organization, 4, 475.

Pollitzer, R. (1954) Plague. Geneva: WHO. Rassi, Y., E. Javadian, M. Amin, S. Rafizadeh, H. Vatandoost & H. Motazedian (2006)

Meriones libycus is the main reservoir of zoonotic cutaneous leishmaniasis in south Islamic Republic of Iran. Eastern Mediterranean Health Journal, 12, 474.

Robertson, C., T. A. Nelson, B. Boots & M. A. Wulder (2007) STAMP: spatial–temporal

analysis of moving polygons. Journal of Geographical Systems, 9, 207-227.

99

Rogers, L. (1928) The yearly variations in plague in India in relation to climate: forecasting epidemics. Proceedings of the Royal Society of London. Series B, Containing Papers of a Biological Character, 103, 42-72.

Rotz, L. D., A. S. Khan, S. R. Lillibridge, S. M. Ostroff, J. M. Hughes, L. Rotz, A. Khan,

S. Lillibridge, S. Ostroff & J. Hughes (2002) Public health assessment of potential biological terrorism agents. Emerging Infectious Diseases, 8, 225.

Saeed, A. A. B., N. A. Al-Hamdan & R. E. Fontaine (2005) Plague from eating raw

camel liver. Emerging Infectious Diseases, 11, 1456. Schotthoefer, A. M., R. J. Eisen, K. J. Kugeler, P. Ettestad, P. J. Reynolds, T. Brown, R.

E. Enscore, J. Cheek, R. Bueno Jr & J. Targhetta (2012) Changing Socioeconomic Indicators of Human Plague, New Mexico, USA. Emerging Infectious Diseases, 18, 1151.

Skinner, G. W., M. Henderson & J. Yuan (2000) China's fertility transition through

regional space: using GIS and census data for a spatial analysis of historical demography. Social Science History, 24, 613-648.

Smulders, M., T. A. Nelson, D. E. Jelinski, S. E. Nielsen, G. B. Stenhouse & K. Laberee

(2012) Quantifying spatial–temporal patterns in wildlife ranges using STAMP: A grizzly bear example. Applied Geography, 35, 124-131.

Stenseth, N. C., B. B. Atshabar, M. Begon, S. R. Belmain, E. Bertherat, E. Carniel, K. L.

Gage, H. Leirs & L. Rahalison (2008) Plague: past, present, and future. PLoS Medicine, 5, e3.

Stenseth, N. C., N. I. Samia, H. Viljugrein, K. L. Kausrud, M. Begon, S. Davis, H. Leirs,

V. Dubyanskiy, J. Esper & V. S. Ageyev (2006) Plague dynamics are driven by climate variation. Proceedings of the National Academy of Sciences, 103, 13110-13115.

Tarantola, A., T. Mollet, J. Gueguen, P. Barboza & E. Bertherat (2009) Plague outbreak

in the Libyan Arab Jamahiriya. Euro surveillance: bulletin europeen sur les maladies transmissibles= European communicable disease bulletin, 14.

Thomas, C. U. & P. E. Hughes (1992) Plague surveillance by serological testing of

coyotes (Canis latrans) in Los Angeles County, California. Journal of wildlife diseases, 28, 610-613.

Webb, C. T., C. P. Brooks, K. L. Gage & M. F. Antolin (2006) Classic flea-borne

transmission does not drive plague epizootics in prairie dogs. Proceedings of the National Academy of Sciences, 103, 6236-6241.

100

Wilcoxon, F. (1945) Individual comparisons by ranking methods. Biometrics Bulletin, 1, 80-83.

Yaghoobi-Ershadi, M., A. Akhavan & M. Mohebali (1996) < i> Meriones Iibycus</i>

and< i> Rhombomys opimus</i>(Rodentia: Gerbillidae) are the main reservoir hosts in a new focus of zoonotic cutaneous leishmaniasis in Iran. Transactions of the Royal Society of Tropical Medicine and Hygiene, 90, 503-504.

Zazanashvili, N., K. Manvelyan, E. Askerov, V. Krever, S. Kalem, B. Avcıoğlu, S.

Galstyan, R. Mnatsekanov & M. Bitsadze (2009) Territorial protection of globally threatened species in the Caucasus. Zazanashvuli, N. and Mallon, D.(Eds.) Status and Protection of Globally Threatened Species in the Caucasus, 222-231.

Zilinskas, R. A. (2006) The anti-plague system and the Soviet biological warfare

program. Critical reviews in microbiology, 32, 47-64.

101

BIOGRAPHICAL SKETCH

Lillian Morris was born in 1989 in Baltimore, Maryland. She grew up in Baltimore

and graduated from the Bryn Mawr School in 2007. She attended the University of

Colorado in Boulder, and graduated with a Bachelor of Arts in geography in 2011. She

was accepted to the UF Department of Geography master’s program, and was offered a

research assistantship in the Spatial Epidemiology and Ecology Research lab in

conjunction with the Emerging Pathogens Institute in 2011. She received her Master of

Science from the University of Florida in spring 2013.