DETERMINANTS OF MALARIA VECTOR HABITAT USE, SPATIAL
DISTRIBUTION, AND COMMUNITY COMPOSITION, WITH A FOCUS ON
ENVIRONMENTAL FACTORS AND INSECTICIDE-TREATED BED NETS
By
ROBERT SEAN MCCANN

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Entomology - Doctor of Philosophy
Ecology, Evolutionary Biology, and Behavior - Dual Major
2013

ABSTRACT
DETERMINANTS OF MALARIA VECTOR HABITAT USE, SPATIAL
DISTRIBUTION, AND COMMUNITY COMPOSITION, WITH A FOCUS ON
ENVIRONMENTAL FACTORS AND INSECTICIDE-TREATED BED NETS
By
ROBERT SEAN MCCANN
Malaria is one of the deadliest vector-borne diseases known. Controlling malaria
requires an understanding of the ecology of local malaria vectors. The potential for
microdams to provide larval habitats for malaria vectors is poorly understood, despite the
importance of microdams for water management in rural areas of Africa. The perimeters
of microdam reservoirs in western Kenya were sampled for Anopheles larvae in the dry
and rainy seasons. In the dry season, both malaria vector and non-vector species were
found in microdam-associated habitats, suggesting that microdams may contribute to
population persistence through the dry season. In the rainy season, microdams may
provide important habitat for Anopheles gambiae s.l., as this species dominated
Anopheles communities in microdams. Thus, microdams may represent a conflict
between public health concerns about malaria and people’s need for stable and reliable
water sources.
Anopheles gambiae s.l. larvae also use a range of smaller, temporary bodies of
stagnant water as habitat. Predictive models of the locations of these habitats may provide
a basis for understanding the spatial determinants of malaria transmission. Four landscape
variables and accumulated precipitation were used to model larval habitat locations
through two methods: logistic regression and random forest. The random forest models
were more accurate than the logistic regression models, especially when accumulated

precipitation was included to account for seasonal differences in precipitation. Larval
habitats were more likely to be present in locations with a lower slope to contributing
area ratio, closer to streams, and in agricultural land use. Differences among soil types
were also found, and the probability of larval habitat presence increased with increasing
accumulated precipitation. This model was used to assess the contribution of larval
habitat proximity to houses to the number of adult malaria vectors in those houses.
Houses were sampled for adult Anopheles females, and variation in household-level
insecticide-treated net (ITN) use was assessed. The number of adult female An. gambiae
s.l. per house increased with increasing mean probability of larval habitat presence within
500 m, which was a better predictor than distance to the nearest larval habitat. Thus, the
configuration of larval habitats within a given landscape may affect the relationship
between larval habitat location and adult malaria vector spatial distribution. Also, houses
in which all residents slept under ITNs the previous night had fewer adult female An.
gambiae s.l. than other houses. There was no difference in the number of An. gambiae s.l.
females collected in houses without ITNs and houses where only some of the residents
slept under ITNs, highlighting an important difference between the effects of ITN
ownership at the household-level and individual ITN use.
Finally, the reemergence of Anopheles funestus in an area of long-term ITN use is
described here. ITN use in Asembo, a community in western Kenya, has been high since
1998. The abundance of An. funestus in Asembo was relatively low from 1998-2008.
However, the majority of the Anopheles collected here in 2010 and 2011 were An.
funestus. This has important implications for malaria transmission in the region, given the
current reliance on ITNs as a long-term, stand-alone method of vector control.

Copyright by
ROBERT SEAN MCCANN
2013

ACKNOWLEDGEMENTS

I would like to thank Ned Walker for his advice, mentoring, support and sense of
humor. Ned, it has truly been a joy to work with you. I would also like to thank the
members of my graduate committee, Dave MacFarlane, Joe Messina and Jim Miller for
their guidance throughout my program. Jim, you are an inspiration for your creativity and
your love of science. I would like to thank my collaborators from the Centers for Disease
Control and Prevention and the Kenya Medical Research Institute, especially John
Gimnig, Nabie Bayoh, Maurice Ombok, George Olang and John Vulule. Thank you to all
of the current and past members of the Walker/Kaufman/Tsao Lab. Especially to Gabe
Hamer and Megan Fritz for showing me the ropes my first two years. Also to Betsy
Brouhard, Becky Morningstar, Lora Anderson and Kelvin Chen for helping me find
things in the lab, where I felt perpetually lost. And to all of the amazing undergraduate
lab assistants who helped me process samples in the lab at MSU: Milicia Dimic, Erica
Brown, Monica MacDonald, Garrett Berry, Cherie Bryant, Geoff Grzesiak, Ingrid
Folland, Simanjit Mand and Susan Boehl.
The collection of data for all of these projects, including thousands of mosquitoes
and thousands of GPS readings, involved help from more field assistants than I can count.
Thank you to everyone who worked on the bed net trap sampling done with Bill Hawley
in the 1990s; to whoever set and collected all of the CDC light traps in the 2000s; and to
everyone involved with the HLC sampling in 2011. Those with whom I worked directly
in the field included Jared Sudhe, Evans Owino, Peter Owera, Michael Nyonga, James
Omolo, Edwin Milando, James Odongo, Baraza Lucas, Peter Kudha, Phillip Owera and
Richard Owera; I sincerely appreciate your efforts. I am grateful to the staff at the

v

KEMRI/CDC field station in Kisian: Samson Otieno for his lessons on identifying
mosquitoes; Ben Oloo, Laban, Bishop, and George Musula for their constant helpfulness
and good nature; Gideon Nyasakera and Richard Amimo for their help rearing
mosquitoes; Professor Joseph Nduati, Elizabeth Nyawera and Eric Ochomo for guiding
me through DNA extractions, PCRs and ELISAs and for providing entertainment; and
Nduta for her help and advice on managing both data and people. Of course, the friendly
and understanding people of Asembo are greatly appreciated for allowing us access to
their land and their homes. Also, thank you to Heather Lenartson-Kluge, Carolyn
Devereaux, Linda Gallagher, Carrie Mulks, Christine VanDeuren, Coreena Spitzley and
the rest of the office staff in both the Department of Entomology and the Department of
Microbiology and Molecular Genetics for helping me navigate the labyrinth of MSU.
And to Jan Eschbach for always printing posters at the last minute for me. My graduate
degree and research were supported by the National Science Foundation (Grant EF0723770), a Dissertation Completion Fellowship from the Graduate School and the
College of Agriculture and Natural Sciences, and with additional support from the
Rhodes Thompson Memorial Fellowship Fund.
I wouldn’t be here today without the endless love and support of my family.
Thank you so much, Mom, Dad, and Michelle! I also have to thank all of my wonderful
friends for making the last four years so enjoyable. To Eric, Ben, Jackie, Nduta and Liz,
thank you for inviting me into Lunch Club and making me feel like less of a Dullard by
comparison. To all of my other friends in Kisumu: Denno, Caro, Joy, Krystin, Amy, Tom
and Edna, thank you for kuku choma nights, watching football, and ridiculous
adventures. To Dan for keeping me up to date on the latest funny YouTube videos and

vi

being willing to help me with anything at the drop of a hat; to Pike, Liu, Ange, Amanda,
Nate, Emily, Pete, Mitch and Brendan for good times at the Green Door, Paul Revere’s,
and other fine establishments talking about music, bugs, sports and how to change the
world; and to Julia for simply being awesome! Thank you to Dylan, Lando, Kevin,
Keiffer, and Tunseth for making every trip home feel like I’d never left, and to King and
Sison for always being there when it counts. And thank you to Megan. First and foremost
for reading every word of this dissertation and providing so many helpful suggestions,
comments and edits. But also for adding color to my life, even if it sometimes comes with
glitter.

vii

TABLE OF CONTENTS

LIST OF TABLES

x

LIST OF FIGURES

xiii

INTRODUCTION
Malaria as a public health concern
Spatial distribution of malaria
Agriculture, water management, and malaria
The physical landscape, precipitation variation, and malaria
Vector control interventions
Biology of African malaria vectors
Objectives
REFERENCES

1
1
1
3
3
4
6
7
8

CHAPTER I: Anopheline larvae in microdam reservoirs.
Abstract
Introduction
Study site
Methods
Sampling larvae
Statistical analyses
Results
Discussion
APPENDIX
REFERENCES

16
16
16
18
20
20
21
22
24
28
39

CHAPTER II: Precipitation data improve predictions of larval malaria vector habitat
locations because of seasonal differences in precipitation.
Abstract
Introduction
Study site
Methods
Larval habitat ground surveys
Environmental data
Statistical methods
Results
Discussion
APPENDIX
REFERENCES

44
44
44
47
48
48
50
51
53
55
60
76

CHAPTER III: Proximity of larval Anopheles gambiae habitats to houses and
insecticide-treated bed net use both explain variation in adult An. gambiae indoor
resting densities.

81

viii

Abstract
Introduction
Methods
Study site
Larval habitats
Adult mosquito sampling
Statistical analysis
Results
Discussion
APPENDIX
REFERENCES

81
82
84
84
85
85
87
89
90
94
106

CHAPTER IV. Reemergence of Anopheles funestus as a vector of
Plasmodium falciparum in western Kenya after long-term implementation of
insecticide-treated bed nets
Abstract
Introduction
Methods
Study site
Sampling the indoor resting populations
Sampling the human landing populations
Historical comparison
Mosquito identification, sporozoite rates and blood meal analysis
Insecticide bioassays
Ethical approval
Results
Indoor resting samples
Human landing samples
Historical comparison
Insecticide bioassays
Discussion
APPENDIX
REFERENCES

112
112
112
115
115
116
116
117
118
119
119
119
119
120
121
122
122
126
139

CHAPTER V: Conclusions
REFERENCES

144
147

ix

LIST OF TABLES

Table 1.1: Record of deposition of voucher specimens for Chapter I. The specimens
listed below have been deposited in the Laboratory of Entomology at the Center for
Global Health Research, Kenya Medical Research Institute/Centers for Disease
Control and Prevention in Kisian, Kenya, as samples of those species or other taxa,
which were used in this research. Voucher recognition labels bearing the voucher
number 2013-14 have been attached or included in fluid-preserved specimens.
30
Table 1.2: Number of each species of Anopheles collected from the perimeters of
microdam reservoirs in Asembo in the dry (February) and rainy (May) seasons of
2012. Percent of total Anopheles identified in each season shown in parentheses.

31

Table 1.3: Estimated changes in the density of larvae collected per microdam
according to soil types and percent vegetation along the perimeter. Soil type 1, friable
clay/sandy clay loam; soil type 2, friable clay; soil type 3, firm, silty clay/clay. Bold
indicates p < 0.10. Effect on Anopheles funestus was not analyzed for rainy season
sampling because only three An. funestus larvae were collected.
32
Table 2.1: Record of deposition of voucher specimens for Chapter II. The specimens
listed below have been deposited in the Laboratory of Entomology at the Center for
Global Health Research, Kenya Medical Research Institute/Centers for Disease
Control and Prevention in Kisian, Kenya, as samples of those species or other taxa,
which were used in this research. Voucher recognition labels bearing the voucher
number 2013-14 have been attached or included in fluid-preserved specimens.
62
Table 2.2: Pearson’s correlation coefficient (r) matrix of four cumulative precipitation
measures for (A) 22 days of larval habitat ground surveys in the 10 by 10 km area
from 17 May to 4 July 2011, and (B) the last day of larval habitat ground surveys
each month in Aduoyo-Miyare and Nguka from April 2011 to June 2012. 0-day refers
to precipitation total for the day of ground surveys, while 6-day, 14-day, 21-day, and
30-day refer to the cumulative precipitation total for the day of the ground surveys
plus the previous 6 days, 14 days, 21 days, or 30 days, respectively. Note that similar
correlations were found among other cumulative n-day precipitation totals at 1-day
increments from 1-day to 30 day (data not shown).
63
Table 2.3: The top four logistic regression candidate models for the 15 monthly
ground surveys in Aduoyo-Miyare and Nguka based on Bayesian information
criterion (BIC), in order of increasing difference in BIC from the top model (ΔBIC)
and decreasing BIC weight, w. TWI, topographic wetness index; LULC, land
use-land cover; DS, distance to the nearest stream; Precip., cumulative 30-day
precipitation total.

64

Table 2.4: Parameter estimates as odds ratios with 95% CI for the top logistic
regression model for the 15 monthly ground surveys in Aduoyo-Miyare and Nguka.

65

x

Table 2.5: The top five logistic regression candidate models for the 10 by 10 km area
based on Bayesian information criterion (BIC), in order of increasing difference in
BIC from the top model (ΔBIC) and decreasing BIC weight, w. TWI, topographic
wetness index; LULC, land use-land cover; DS, distance to the nearest stream;
Precip., cumulative 30-day precipitation total.
66
Table 2.6: Parameter estimates as odds ratios with 95% CI for the top logistic
regression model for the 10 by 10 km area.

67

Table 2.7: Comparison of models predicting the presence of larval Anopheles
gambiae s.l. habitats in the 10 by 10 km area. Two random forest (RF) models are
shown with and without Precip. (the cumulative 14-day precipitation total). The best
logistic regression (LR) model is also show. TWI, topographic wetness index; LULC,
land use-land cover; DS, distance to the nearest stream.
68
Table 2.8: Comparison of models predicting the presence of larval Anopheles
gambiae s.l. habitats in Aduoyo-Miyare and Nguka. Two random forest (RF) models
are shown with and without Precip. (the cumulative 14-day precipitation total). The
best logistic regression (LR) model is also shown. TWI, topographic wetness index;
LULC, land use-land cover; DS, distance to the nearest stream.
69
Table 3.1: Household-level variables recorded for houses where sampling was done
for female Anopheles. For factors, all levels of the factor are listed, with the number
of houses observed at each level shown in parentheses.
96
Table 3.2: Record of deposition of voucher specimens for Chapter III. The specimens
listed below have been deposited in the Albert J. Cook Arthropod Research
Collection, Michigan State University (MSU) museum as samples of those species or
other taxa, which were used in this research. Voucher recognition labels bearing the
voucher number 2013-14 have been attached.
97
Table 3.3: Indices used to quantify the proximity of larval habitats to houses where
sampling was done for female Anopheles.

98

Table 3.4: Leading candidate models (sum of wi > 0.90) for comparison of larval
habitat indices contribution to the number of An. gambiae s.l. in houses. Results for
the distance to the nearest habitat are also shown for comparison. Bayesian
Information Criteria (BIC), difference in BIC from the lowest BIC (ΔBIC), model
weight (wi), correlation of habitat index with mean probability of larval habitat within
500 m (r). Remaining candidate models had ΔBIC > 7.0 and wi < 0.020.
Table 3.5: Leading candidate models (sum of wi > 0.90) for quantifying the variation
in number of An. gambiae s.l. in houses using household-level variables. Bayesian
Information Criteria (BIC), difference in BIC from the lowest BIC (ΔBIC), model

xi

99

weight (wi), degrees of freedom (df). Remaining candidate models had ΔBIC > 6.5
and wi < 0.025.

100

Table 3.6: Relative rate ratios of the number of An. gambiae s.l. per house, according
to the variables in the leading negative binomial regression model. Note that all
models were analyzed with larval habitat indices as continuous variables; mean
P(habitat) within 500 m is presented here as a factor for a more intuitive
interpretation.
101
Table 4.1: Record of deposition of voucher specimens for Chapter IV. The specimens
listed below have been deposited in the Laboratory of Entomology at the Center for
Global Health Research, Kenya Medical Research Institute/Centers for Disease
Control and Prevention in Kisian, Kenya, as samples of those species, which were
used in this research. Voucher recognition labels bearing the voucher number
2013-14 have been attached.
128
Table 4.2: Number of Anopheles collected per house during pyrethrum spray catch
sampling shown by species, sex, and year. Percentages of the total for each column
are shown in parentheses. P-values are from non-parametric Mann-Whitney tests to
determine differences between years for each species within each sex.

129

Table 4.3: Plasmodium falciparum sporozoite-negative and positive female
mosquitoes, according to enzyme-linked immunosorbent assay (ELISA), from
pyrethrum spray catch (PSC) sampling, shown by species with percent positive in
parentheses.

130

Table 4.4: Comparison of human landing catch (HLC) and bed net trap (BNT)
sampling methods for Anopheles funestus and Anopheles gambiae s.l., Nov 1992 to
May 1993. Relative rates of collection and associated 95% CI and p-value were
calculated using negative binomial regression. P-values indicate the probability that
the relative collection rate of BNT sampling was different from that of HLC.

131

xii

LIST OF FIGURES

Figure 1.1: Pictures of six of the sampled microdams in Asembo.

33

Figure 1.2: Daily precipitation totals for 1 December 2011 (12/01/11) through 31
May 2012 (05/31/12). Horizontal black bar indicates dry season period for Anopheles
sampling from microdams, while the red bar indicates the rainy season sampling
period.
34
Figure 1.3: Total number of samples taken for Anopheles larvae by habitat type in the
dry and rainy seasons of 2012.
35
Figure 1.4: Mean number of Anopheles larvae per sample, collected from three
habitat types on the perimeters of microdams in the dry season. Error bars are 95%
confidence intervals. Results shown for An. pharoensis are for larvae identified as
An. pharoensis/squamosus.

36

Figure 1.5: Mean number of Anopheles gambiae s.l. larvae per sample, collected
from three habitat types on the perimeters of microdams in the rainy season. Error
bars are 95% confidence intervals.

37

Figure 1.6: Dry and rainy season species richness of Anopheles larvae collected from
microdam reservoir perimeters, by the percent of the perimeter that was vegetated.
Points show observed data (n = 31 microdams per season). Solid lines show estimates
from separate linear regressions for each season, with broken lines showing 95%
confidence intervals.
38
Figure 2.1: (A) Map of Kenya with red square indicating location of the study region
in western Kenya. (B) Map showing the boundaries of Asembo and the streams
within the community, with black dots representing all households. (C) 10 by 10 km
study site shown as dashed black line with thirty-one 500 by 500 m quadrats
surveyed for larval habitats shown as red boxes. (D) Location of the neighboring
villages of Aduoyo-Miyare and Nguka, sites of the 15 monthly ground surveys for
larval habitats, shown in red.
70
Figure 2.2: Scatter plot of the number of habitats recorded in Aduoyo-Miyare and
Nguka each month (n = 15) by the cumulative 30-day precipitation for the last day of
ground surveys for that month. Line shows prediction of linear regression,
2
R = 0.1931, p = 0.1012. The red and blue boxes highlight variation in the residual
error discussed further in the text.
72
Figure 2.3: Precipitation data from the Kisumu airport (about 40 km east of Asembo)
for March 2011 through July 2012. Bars show daily precipitation total. Solid line
shows the cumulative 30-day precipitation total used in the logistic regression model
for the 15-month dataset, while the dashed line shows the cumulative 21-day total

xiii

used in the random forest model of the same data. Circles and squares show the last
day of ground surveys for each month in Aduoyo-Miyare and Nguka.

73

Figure 2.4: Precipitation data from the Kisumu airport (about 40 km east of Asembo)
for 2 May 2011 to 4 July 2012. Bars show daily precipitation total. Solid line shows
the cumulative 6-day total used in the logistic regression model of the 10 by 10 km
dataset, while the dashed line shows the cumulative 14-day precipitation total used in
the random forest model of the same data. Squares and circles show the days of the
ground surveys in the 31 quadrats.
74
Figure 2.5: Probability of larval habitat presence across the 10 by 10 km study site as
predicted by (A) the logistic regression model using TWI, distance to stream and soil
type; and (B) the random forest model using TWI, distance to stream, soil type and
LULC.
75
Figure 3.1: (A) Map of Kenya with red square indicating location of the study region
in western Kenya. (B) Map showing the boundaries of Asembo and the streams
within the community, with black dots representing all compounds in the community.
(C) Probability of larval habitat presence as predicted by a random forest model
across a 10 by 10 km area in Asembo. (D) 8 by 8 km border for pyrethrum spray
catch sampling shown as dashed black line, with red dots showing the 525 houses
sampled for adult Anopheles and solid red line representing the extent of the larval
habitat model.
102
Figure 3.2: Estimated number of adult female An. gambiae s.l. per house according to
the leading model. Shading bounded by dashed lines indicates 95% confidence
intervals. Levels of LLIN use refer to whether all, some, or none of the residents used
LLINs in the house the previous night. Estimates are shown for houses where cattle
are present in the compound (i.e. cattle = yes), but residents did not cook in the house
(cooked = no).
104
Figure 3.3: Scatterplot of the mean probability of larval habitat presence within
500 m of a house and the distance to the nearest larval habitat to the house (r = 0.22)
for the 525 houses sampled in this study.
105
Figure 4.1: (A) Map of Kenya with box indicating location of the study region in
Nyanza Province. (B) Map showing the location of Asembo on the shores of Lake
Victoria, about 50 km west of the city of Kisumu.

132

Figure 4.2: Proportion of blood meals taken from humans and cattle by Anopheles
species collected during pyrethrum spray catch in Asembo. Non-amplifiers to
polymerase chain reaction for blood meal identification are not shown (29%).

133

Figure 4.3: Number of Anopheles females collected per collector per night during
human landing catch sampling shown by species and location (indoors/outdoors).
Error bars are 95% confidence intervals. Plasmodium falciparum sporozoite rates as

xiv

determined by enzyme-linked immunosorbent assays are shown as percentages above
each species. Anopheles gambiae s.l. specimens were identified as members of the
An. gambiae species complex according to morphological methods but could not be
further differentiated by polymerase chain reaction.
134
Figure 4.4: (A) Mean number of Anopheles females collected per sample in June
through July from 1993 through 2011 using bed net traps (BNT; 1993--1997), light
traps (LT; 2002--2008), and human landing catch (HLC; 2011). (B) Mean number of
Anopheles females collected, adjusted for comparison to HLC sampling using the
relative rates of BNT to HLC (1993--1997) and LT to HLC (2002--2008).
(C) Proportion of total Anopheles females collected identified as either Anopheles
funestus or Anopheles gambiae s.l. Error bars indicate 95% confidence intervals.
Data were not available from 1998-2001, 2009, and 2010.
135
Figure 4.5: Percent mortality after 24 hours of wild collected Anopheles funestus and
Anopheles gambiae s.l. adults from Asembo when exposed to permethrin,
deltamethrin, or bendiocarb in one-hour World Health Organization tube bioassays. 138

xv

INTRODUCTION
Malaria as a public health concern
Malaria continues to be one of the most significant infectious diseases in the
world. Although significant reductions in malaria-related morbidity and mortality have
occurred globally since 2001, there were still an estimated 219 million cases of malaria
worldwide in 2010, resulting in 660,000 deaths (World Health Organization 2012).
Furthermore, the burden of malaria is heaviest for children living in poverty (World
Health Organization 2012). Poor families are less likely to be able to prevent malaria
through measures like window screens and insecticide-treated bed nets (ITNs), which
reduce malaria vector to human contact (Noor et al. 2009). Families living in poverty are
also less likely to be able to treat a child with malaria medication (Mbachu et al. 2012).
At the same time, malaria contributes to keeping people in poverty, producing a ‘malaria
trap’ (Berthélemy et al. 2013). Individuals with malaria are more likely to be absent from
school or work, and families may deplete their savings for treatment (Sachs and Malaney
2002). However, when poverty is reduced through community-wide economic
development, malaria prevalence decreases as people have more economic flexibility
(Pronyk et al. 2012).
Spatial distribution of malaria
The fact that malaria displays marked variation across space has long been
recognized, such that Italian law in 1809 prevented housing in the vicinity of irrigated
meadows (Watson 1949). As the etiology and transmission of malaria became clear
following the work of Laveran, Ross, and Grassi, the connection between malaria and
water was made explicit (Boyd 1949). Given that transmission of malaria requires the

1

bite of an infected Anopheles mosquito, the disease is limited to locations where malaria
vector populations can persist. Additional factors have been shown to contribute to the
spatial heterogeneity of malaria at the local scale. Even relatively small socioeconomic
differences among households in a community can lead to differential risk of infection
with malaria (Greenwood 1989, Gamage-Mendis et al. 1991). Immunological differences
among people due to genetics and previous exposure to malaria may also contribute
significantly to variation in malaria risk within a community (Greenwood 1989,
Mackinnon et al. 2000). Nevertheless, the locations of people’s homes relative to that of
the malaria vectors’ larval habitats has been shown to be an important determinant of
malaria risk in a wide range of ecological and epidemiological settings (Trape et al. 1992,
Ribeiro et al. 1996, Thompson et al. 1997, Ghebreyesus et al. 1999). Clearly, the spatial
distribution of malaria vectors contributes to the spatial distribution of malaria. Thus,
identifying determinants of the spatial distribution of malaria vectors facilitates a better
understanding of variation in malaria risk.
The association between water and malaria is driven by the requirements of the
aquatic life stages of the malaria vector mosquitoes. The specific aquatic habitats used by
local malaria vector species are unique to the ecology of each species (Silver 2008), but
the immature life stages of every species are restricted spatially to the locations of the
suitable aquatic habitat. In this sense, the aquatic habitat of malaria vector immature
stages may be thought of as a focus of malaria transmission (Carter et al. 2000).
Importantly, the spatial extent, or dimensions, of the these foci should be largely
determined by: 1) the relative densities of the aquatic habitats and people’s homes; and 2)
the dispersal behavior of the adult female malaria vectors (Carter et al. 2000). Therefore,

2

it is vital to identify the specific aquatic habitat types used locally by malaria vector
immature stages, and to understand where these habitats are located, to better understand
a primary driver of the spatial distribution of malaria vectors (Killeen et al. 2004).
Agriculture, water management, and malaria
Agriculture has been linked with malaria transmission, especially malaria caused
by Plasmodium falciparum (Welsh), probably since the very emergence of the disease in
humans (Webb 2009). This relationship is driven by the effects of agriculture on water
management, as irrigation and poor drainage can result in the creation of suitable aquatic
habitat for the immature stages of malaria vectors (Packard 2007). Of course, agriculture
alone does not cause malaria. In cases where modest capital investments in agriculture
have resulted in community-wide economic improvements, malaria prevalence actually
decrease because of residents increased ability to prevent and treat malaria (Ijumba and
Lindsay 2001, Ijumba et al. 2002). However, larval Anopheles habitats have been
associated with agricultural land use in a wide range of settings (Ramasamy et al. 1992,
Afrane et al. 2004, Pope et al. 2005, Mutuku et al. 2009). The construction of dams for
agricultural water management or hydroelectric power has also been long recognized as
potentially leading to increased larval Anopheles habitat (Keiser et al. 2005). In this case,
thoughtful design and management of dams to make the reservoir less suitable for local
malaria vector species can limit this risk (Kitron and Spielman 1989, Yohannes et al.
2005).
The physical landscape, precipitation variation, and malaria
In addition to the influence of human land use in determining the locations of
malaria vector larval habitats, the physical landscape also has considerable influence on

3

these locations (Smith et al. 2013). For instance, in landscapes with high topographic
relief, larval Anopheles habitats are generally restricted to valley bottoms (Balls et al.
2004, Zhou et al. 2007, Himeidan et al. 2009). Even in landscapes with low topographic
relief, the larval habitats of malaria vectors may be concentrated along a river or stream
(van der Hoek et al. 2003, Bogh et al. 2003). One method for quantifying the hydrologic
potential of a landscape is to use a topographic wetness index (Beven and Kirkby 1979),
which has been used to show that larval Anopheles habitats may be found in locations
having a combination of greater upslope area contributing to drainage and less slope
(Clennon et al. 2010, Li et al. 2011, Nmor et al. 2013).
The spatial distribution of larval Anopheles habitats is not necessarily static over
time. Indeed, seasonal differences in rainfall may cause variation in larval habitat
abundance and distribution across the landscape (Gimnig et al. 2001, Mutuku et al. 2009,
Munga et al. 2009, Clennon et al. 2010, Li et al. 2011). These changes have important
consequences for malaria transmission. An increased availability of larval habitats
contributes to higher malaria vector population growth (Killeen et al. 2004), while
increased spatial range of larval habitats across the landscape results in an increased
range of malaria transmission (Carter et al. 2000). At longer temporal scales and broader
spatial scales, the size and distribution of malaria vector populations may expand or
contract regionally due to climate change (Thomas et al. 2004, Ermert et al. 2013).
Vector control interventions
Public health interventions may also have a significant impact on the distribution
of malaria vectors. One of the primary malaria vector control measures used since the
1990s is mass distribution of insecticide-treated bed nets (ITNs), which must be retreated

4

every six to nine months. More recently, ITNs have been largely replaced by long-lasting
impregnated nets (LLINs), with effective insecticidal activity lasting up to three years
(World Health Organization 2012). On a broad scale, ITNs impact malaria vectors by
reducing their population sizes (Gimnig et al. 2003a, 2003b), which reduces communitywide morbidity and mortality due to malaria (Hawley et al. 2003b). At the household
scale, variation in ITN ownership and use may lead to variation in the number of adult
Anopheles found indoors. The presence of ITNs in houses may reduce the rate of entry
(Mathenge et al. 2001), or increase the rate of exiting (Mathenge et al. 2001, Malima et
al. 2008), by malaria vectors.
While the effectiveness of ITNs is well documented, many potential limitations
exist, especially when ITNs are used as the only means of vector control (Hawley et al.
2003a, Beier et al. 2008). First, ITNs target only indoor-biting malaria vectors, leaving
exophagic populations unaffected (Bayoh et al. 2010). Second, despite the substantial
increase in the number of households owning ITNs in many malarious countries over the
last decade (World Health Organization 2012), variation in ITN use by individuals within
households leaves some people unprotected (Korenromp et al. 2003, Eisele et al. 2009).
This is true even in countries implementing universal coverage with LLINs (Macintyre et
al. 2011). Finally, there is considerable potential for the development of insecticide
resistance in malaria vector populations (Ranson et al. 2011, Trape et al. 2011). Only one
class of insecticides, the pyrethroids, is approved for use with ITNs (Zaim et al. 2000),
and pyrethroid resistance has been reported in malaria vectors across Africa (Ranson et
al. 2011).

5

Biology of African malaria vectors
The vast majority of malaria cases (80%) and deaths from malaria (91%) in 2010
occurred in Africa (World Health Organization 2012). The most widely distributed
malaria vectors in Africa are Anopheles funestus Giles, Anopheles gambiae s.s. Giles and
Anopheles arabiensis Patton (Sinka et al. 2010). The latter two are members of a species
complex of eight closely related, morphologically indistinguishable species known
collectively as Anopheles gambiae s.l. (Coetzee et al. 2013). Anopheles funestus and An.
gambiae s.s. are two of the most important malaria vectors in the world due to their
preference for feeding almost exclusively on human blood (Garrett-Jones 1964).
Anopheles arabiensis demonstrates a broader variation in host preference for blood
feeding, opportunistically taking blood meals from cattle in addition to humans (Takken
and Verhulst 2013). However, An. arabiensis is the primary malaria vector in many
regions (Massebo et al. 2013) and can become a more important vector when control
measures reduce the size of other malaria vector populations (Bayoh et al. 2010).
These three species utilize a range of aquatic habitat types across their
distributions, but they are generally limited to stagnant fresh water bodies (Gillies and De
Meillon 1968). In many regions the larval habitats of An. gambiae s.s. and An. arabiensis
are smaller, temporary bodies of stagnant water (Gimnig et al. 2001, Mutuku et al. 2006),
and the two species are often sympatric within the same larval habitats (Charlwood and
Edoh 1996, Minakawa et al. 1999, Gimnig et al. 2001). Anopheles arabiensis will
additionally exploit larger, permanent habitats such as rice fields when they are available
(Githeko et al. 1996). Anopheles funestus larvae are found more often in larger, more
permanent aquatic habitats with emergent vegetation (Gimnig et al. 2001).

6

Objectives
In 2007, Bill and Melinda Gates called for the eradication of malaria from the
globe within their lifetime (Roberts and Enserink 2007). If we are to succeed at this goal
without experiencing the widespread resurgences of malaria that followed the Global
Malaria Eradication Program of the 1950-60s (Feachem et al. 2010, Nájera et al. 2011),
we must take into account the ecology of the local malaria vector species (Ferguson et al.
2010). Along these lines, this dissertation addresses four aspects of malaria vector
biology in order to strengthen our understanding of the contribution of Anopheles
mosquito malaria vectors to transmission dynamics in rural landscapes heavily altered by
human activity. The objectives of this dissertation are to: 1) investigate the contribution
of microdam-associated habitat to the production of malaria vectors in the dry and rainy
seasons; 2) create a model for predicting larval An. gambiae s.l. habitat locations using
landscape variables that predict the likelihood of stagnant water bodies and account for
seasonal changes in habitat probability based on accumulated precipitation; 3) quantify
the relative contributions of larval habitat proximity and household characteristics to the
spatial distribution of adult An. gambiae s.l.; and 4) describe the reemergence of
Anopheles funestus in a region with long-term ITN coverage.

7

REFERENCES

8

REFERENCES

Afrane, Y. A., E. Klinkenberg, P. Drechsel, K. Owusu-Daaku, R. Garms, and T. Kruppa.
2004. Does irrigated urban agriculture influence the transmission of malaria in the
city of Kumasi, Ghana? Acta Tropica 89:125–134.
Balls, M. J., R. Bødker, C. J. Thomas, W. Kisinza, H. A. Msangeni, and S. W. Lindsay.
2004. Effect of topography on the risk of malaria infection in the Usambara
Mountains, Tanzania. Transactions of the Royal Society of Tropical Medicine and
Hygiene 98:400–408.
Bayoh, M. N., D. K. Mathias, M. R. Odiere, F. M. Mutuku, L. Kamau, J. E. Gimnig, J.
M. Vulule, W. A. Hawley, M. J. Hamel, and E. D. Walker. 2010. Anopheles
gambiae: historical population decline associated with regional distribution of
insecticide-treated bed nets in western Nyanza Province, Kenya. Malaria Journal
9:62.
Beier, J. C., J. Keating, J. I. Githure, M. B. Macdonald, D. E. Impoinvil, and R. J. Novak.
2008. Integrated vector management for malaria control. Malaria Journal 7 Suppl
1:S4–S4.
Berthélemy, J.-C., J. Thuilliez, O. Doumbo, and J. Gaudart. 2013. Malaria and protective
behaviours: is there a malaria trap? Malaria Journal 12:200.
Beven, K. J., and M. J. Kirkby. 1979. A physically based, variable contributing area
model of basin hydrology. Hydrological Sciences Bulletin 24:43–69.
Bogh, C., S. E. Clarke, M. Jawara, C. J. Thomas, and S. W. Lindsay. 2003. Localized
breeding of the Anopheles gambiae complex (Diptera: Culicidae) along the River
Gambia, West Africa. Bulletin of Entomological Research 93:279–287.
Boyd, M. F. 1949. Historical overview. Pages 3–25 in M. F. Boyd, editor. Malariology:
A comprehensive survey of all aspects of this group of diseases from a global
standpoint. W.B. Saunders Co., Philadelphia.
Carter, R., K. N. Mendis, and D. Roberts. 2000. Spatial targeting of interventions against
malaria. Bulletin of the World Health Organization 78:1401–1411.
Charlwood, J. D., and D. Edoh. 1996. Polymerase chain reaction used to describe larval
habitat use by Anopheles gambiae complex (Diptera: Culicidae) in the environs of
Ifakara, Tanzania. Journal of Medical Entomology 33:202–204.
Clennon, J. A., A. Kamanga, M. Musapa, C. Shiff, and G. E. Glass. 2010. Identifying
malaria vector breeding habitats withremote sensing data and terrain-based
landscapeindices in Zambia. International Journal of Health Geographics 9:58.

9

Coetzee, M., R. H. Hunt, R. C. Wilkerson, A. della Torre, M. B. Coulibaly, and N. J.
Besansky. 2013. Anopheles coluzzii and Anopheles amharicus, new members of the
Anopheles gambiae complex. Zootaxa 3619:246–274.
Eisele, T. P., J. Keating, M. Littrell, D. Larsen, and K. Macintyre. 2009. Assessment of
insecticide-treated bednet use among children and pregnant women across 15
countries using standardized national surveys. American Journal of Tropical
Medicine and Hygiene 80:209–214.
Ermert, V., A. H. Fink, and H. Paeth. 2013. The potential effects of climate change on
malaria transmission in Africa using bias-corrected regionalised climate projections
and a simple malaria seasonality model. Climatic Change 120:741–754.
Feachem, R. G., A. A. Phillips, J. Hwang, C. Cotter, B. Wielgosz, B. M. Greenwood, O.
Sabot, M. H. Rodríguez, R. R. Abeyasinghe, T. A. Ghebreyesus, and R. W. Snow.
2010. Shrinking the malaria map: progress and prospects. Lancet 376:1566–1578.
Ferguson, H. M., A. Dornhaus, A. Beeche, C. Borgemeister, M. Gottlieb, M. S. Mulla, J.
E. Gimnig, D. Fish, and G. F. Killeen. 2010. Ecology: a prerequisite for malaria
elimination and eradication. PLoS Medicine 7:e1000303.
Gamage-Mendis, A. C., R. Carter, C. Mendis, A. P. De Zoysa, P. R. Herath, and K. N.
Mendis. 1991. Clustering of malaria infections within an endemic population: risk of
malaria associated with the type of housing construction. American Journal of
Tropical Medicine and Hygiene 45:77–85.
Garrett-Jones, C. 1964. The human blood index of malaria vectors in relation to
epidemiological assessment. Bulletin of the World Health Organization 30:241–261.
Ghebreyesus, T. A., M. Haile, K. H. Witten, A. Getachew, A. M. Yohannes, M.
Yohannes, H. D. Teklehaimanot, S. W. Lindsay, and P. Byass. 1999. Incidence of
malaria among children living near dams in northern Ethiopia: community based
incidence survey. BMJ 319:663–666.
Gillies, M. T., and B. De Meillon. 1968. The Anophelinae of Africa south of the Sahara.
Pages 1–343. South African Institute for Medical Research, Johannesburg, South
Africa.
Gimnig, J. E., J. M. Vulule, T. Q. Lo, L. Kamau, M. S. Kolczak, P. A. Phillips-Howard,
E. M. Mathenge, F. O. ter Kuile, B. L. Nahlen, A. W. Hightower, and W. A. Hawley.
2003a. Impact of permethrin-treated bed nets on entomologic indices in an area of
intense year-round malaria transmission. American Journal of Tropical Medicine and
Hygiene 68:16–22.
Gimnig, J. E., M. Ombok, L. Kamau, and W. A. Hawley. 2001. Characteristics of larval

10

anopheline (Diptera: Culicidae) habitats in western Kenya. Journal of Medical
Entomology 38:282–288.
Gimnig, J. E., M. S. Kolczak, A. W. Hightower, J. M. Vulule, E. Schoute, L. Kamau, P.
A. Phillips-Howard, F. O. ter Kuile, B. L. Nahlen, and W. A. Hawley. 2003b. Effect
of permethrin-treated bed nets on the spatial distribution of malaria vectors in
western Kenya. American Journal of Tropical Medicine and Hygiene 68:115–120.
Githeko, A. K., M. W. Service, C. M. Mbogo, and F. K. Atieli. 1996. Resting behaviour,
ecology and genetics of malaria vectors in large scale agricultural areas of western
Kenya. Parassitologia 38:481–489.
Greenwood, B. M. 1989. The microepidemiology of malaria and its importance to
malaria control. Transactions of the Royal Society of Tropical Medicine and Hygiene
83 Suppl:25–29.
Hawley, W. A., F. O. ter Kuile, R. S. Steketee, B. L. Nahlen, D. J. Terlouw, J. E. Gimnig,
Y. P. Shi, J. M. Vulule, J. A. Alaii, A. W. Hightower, M. S. Kolczak, S. K. Kariuki,
and P. A. Phillips-Howard. 2003a. Implications of the western Kenya permethrintreated bed net study for policy, program implementation, and future research.
American Journal of Tropical Medicine and Hygiene 68:168–173.
Hawley, W. A., P. A. Phillips-Howard, F. O. ter Kuile, D. J. Terlouw, J. M. Vulule, M.
Ombok, B. L. Nahlen, J. E. Gimnig, S. K. Kariuki, M. S. Kolczak, and A. W.
Hightower. 2003b. Community-wide effects of permethrin-treated bed nets on child
mortality and malaria morbidity in western Kenya. American Journal of Tropical
Medicine and Hygiene 68:121–127.
Himeidan, Y. E., G. Zhou, L. Yakob, Y. A. Afrane, S. Munga, H. Atieli, E.-A. El-Rayah,
A. K. Githeko, and G. Yan. 2009. Habitat stability and occurrences of malaria vector
larvae in western Kenya highlands. Malaria Journal 8:234.
Ijumba, J. N., and S. W. Lindsay. 2001. Impact of irrigation on malaria in Africa: paddies
paradox. Medical and Veterinary Entomology 15:1–11.
Ijumba, J. N., F. W. Mosha, and S. W. Lindsay. 2002. Malaria transmission risk
variations derived from different agricultural practices in an irrigated area of northern
Tanzania. Medical and Veterinary Entomology 16:28–38.
Keiser, J., M. C. de Castro, M. F. Maltese, R. Bos, M. Tanner, B. H. Singer, and J.
Utzinger. 2005. Effect of irrigation and large dams on the burden of malaria on a
global and regional scale. American Journal of Tropical Medicine and Hygiene
72:392–406.
Killeen, G. F., A. A. Seyoum, and B. G. J. Knols. 2004. Rationalizing historical successes
of malaria control in Africa in terms of mosquito resource availability management.

11

American Journal of Tropical Medicine and Hygiene 71:87–93.
Kitron, U., and A. Spielman. 1989. Suppression of transmission of malaria through
source reduction: antianopheline measures applied in Israel, the United States, and
Italy. Reviews of infectious diseases 11:391–406.
Korenromp, E. L., J. Miller, R. E. Cibulskis, M. K. Cham, D. Alnwick, and C. Dye. 2003.
Monitoring mosquito net coverage for malaria control in Africa: possession vs. use
by children under 5 years. Tropical Medicine and International Health 8:693–703.
Li, L., L. Bian, L. Yakob, G. Zhou, and G. Yan. 2011. Analysing the generality of
spatially predictive mosquito habitat models. Acta Tropica 119:30–37.
Macintyre, K., M. Littrell, J. KEATING, B. Hamainza, J. Miller, and T. P. Eisele. 2011.
Determinants of hanging and use of ITNs in the context of near universal coverage in
Zambia. Health Policy and Planning.
Mackinnon, M. J., D. M. Gunawardena, J. Rajakaruna, S. Weerasingha, K. N. Mendis,
and R. Carter. 2000. Quantifying genetic and nongenetic contributions to malarial
infection in a Sri Lankan population. Proceedings of the National Academy of
Sciences of the United States of America 97:12661–12666.
Malima, R. C., S. M. Magesa, P. K. Tungu, V. Mwingira, F. S. Magogo, W. Sudi, F. W.
Mosha, C. F. Curtis, C. Maxwell, and M. Rowland. 2008. An experimental hut
evaluation of Olyset nets against anopheline mosquitoes after seven years use in
Tanzanian villages. Malaria Journal 7:38.
Massebo, F., M. Balkew, T. Gebre-Michael, and B. Lindtjorn. 2013. Entomologic
Inoculation Rates of Anopheles arabiensis in Southwestern Ethiopia. American
Journal of Tropical Medicine and Hygiene.
Mathenge, E. M., J. E. Gimnig, M. Kolczak, M. Ombok, L. W. Irungu, and W. A.
Hawley. 2001. Effect of permethrin-impregnated nets on exiting behavior, blood
feeding success, and time of feeding of malaria mosquitoes (Diptera: Culicidae) in
western Kenya. Journal of Medical Entomology 38:531–536.
Mbachu, C. O., O. E. Onwujekwe, B. S. Uzochukwu, E. Uchegbu, J. Oranuba, and A. L.
Ilika. 2012. Examining equity in access to long-lasting insecticide nets and
artemisinin-based combination therapy in Anambra state, Nigeria. BMC Public
Health 12:315.
Minakawa, N., C. M. Mutero, J. I. Githure, J. C. Beier, and G. Yan. 1999. Spatial
distribution and habitat characterization of anopheline mosquito larvae in western
Kenya. American Journal of Tropical Medicine and Hygiene 61:1010–1016.
Munga, S., L. Yakob, E. Mushinzimana, G. Zhou, T. Ouna, N. Minakawa, A. K. Githeko,

12

and G. Yan. 2009. Land use and land cover changes and spatiotemporal dynamics of
anopheline larval habitats during a four-year period in a highland community of
Africa. American Journal of Tropical Medicine and Hygiene 81:1079–1084.
Mutuku, F. M., J. A. Alaii, M. N. Bayoh, J. E. Gimnig, J. M. Vulule, E. D. Walker, E.
Kabiru, and W. A. Hawley. 2006. Distribution, description, and local knowledge of
larval habitats of Anopheles gambiae s.l. in a village in western Kenya. American
Journal of Tropical Medicine and Hygiene 74:44–53.
Mutuku, F., M. N. Bayoh, A. Hightower, J. M. Vulule, J. E. Gimnig, J. Mueke, F.
Amimo, and E. D. Walker. 2009. A supervised land cover classification of a western
Kenya lowland endemic for human malaria: associations of land cover with larval
Anopheles habitats. International Journal of Health Geographics 8:19.
Nájera, J. A., M. González-Silva, and P. L. Alonso. 2011. Some lessons for the future
from the Global Malaria Eradication Programme (1955–1969). PLoS Medicine
8:e1000412.
Nmor, J. C., T. Sunahara, K. Goto, K. Futami, G. Sonye, P. Akweywa, G. Dida, and N.
Minakawa. 2013. Topographic models for predicting malaria vector breeding
habitats: potential tools for vector control managers. Parasites & Vectors 6:14.
Noor, A. M., J. J. Mutheu, A. J. Tatem, S. I. Hay, and R. W. Snow. 2009. Insecticidetreated net coverage in Africa: mapping progress in 2000-07. Lancet 373:58–67.
Packard, R. M. 2007. The making of a tropical disease: a short history of malaria. Johns
Hopkins University Press, Baltimore.
Pope, K., P. Masuoka, E. Rejmankova, J. Grieco, S. Johnson, and D. Roberts. 2005.
Mosquito habitats, land use, and malaria risk in Belize from satellite imagery.
Ecological applications : a publication of the Ecological Society of America
15:1223–1232.
Pronyk, P. M., M. Muniz, B. Nemser, M.-A. Somers, L. McClellan, C. A. Palm, U. K.
Huynh, Y. Ben Amor, B. Begashaw, J. W. McArthur, A. Niang, S. E. Sachs, P.
Singh, A. Teklehaimanot, and J. D. Sachs. 2012. The effect of an integrated
multisector model for achieving the Millennium Development Goals and improving
child survival in rural sub-Saharan Africa: a non-randomised controlled assessment.
Lancet 379:2179–2188.
Ramasamy, R., R. De Alwis, A. Wijesundere, and M. S. Ramasamy. 1992. Malaria
transmission at a new irrigation project in Sri Lanka: the emergence of Anopheles
annularis as a major vector. American Journal of Tropical Medicine and Hygiene
47:547–553.
Ranson, H., R. N’Guessan, J. Lines, N. Moiroux, Z. Nkuni, and V. Corbel. 2011.

13

Pyrethroid resistance in African anopheline mosquitoes: what are the implications for
malaria control? Trends in Parasitology 27:91–98.
Ribeiro, J. M. C., F. Seulu, T. Abose, G. Kidane, and A. Teklehaimanot. 1996. Temporal
and spatial distribution of anopheline mosquitos in an Ethiopian village: Implications
for malaria control strategies. Bulletin of the World Health Organization 74:299–305.
Roberts, L., and M. Enserink. 2007. Did They Really Say... Eradication? Science, New
Series 318:1544–1545.
Sachs, J., and P. Malaney. 2002. The economic and social burden of malaria. Nature
415:680–685.
Silver, J. B. 2008. Sampling the larval population. Pages 137–338 in Mosquito ecology:
field sampling methods, 3rd edition. Springer, New York.
Sinka, M. E., M. J. Bangs, S. Manguin, M. Coetzee, C. M. Mbogo, J. Hemingway, A. P.
Patil, W. H. Temperley, P. W. Gething, C. W. Kabaria, R. M. Okara, T. Van
Boeckel, H. C. J. Godfray, R. E. Harbach, and S. I. Hay. 2010. The dominant
Anopheles vectors of human malaria in Africa, Europe and the Middle East:
occurrence data, distribution maps and bionomic précis. Parasites & Vectors 3:117.
Smith, M. W., M. G. Macklin, and C. J. Thomas. 2013. Earth-Science Reviews. Earth
Science Reviews 116:109–127.
Takken, W., and N. O. Verhulst. 2013. Host Preferences of Blood-Feeding Mosquitoes.
Annual review of entomology 58:433–453.
Thomas, C. J., G. Davies, and C. E. Dunn. 2004. Mixed picture for changes in stable
malaria distribution with future climate in Africa. Trends in Parasitology 20:216–
220.
Thompson, R., K. Begtrup, N. Cuamba, M. Dgedge, C. Mendis, A. Gamage-Mendis, S.
M. Enosse, J. Barreto, R. E. Sinden, and B. Hogh. 1997. The Matola malaria project:
a temporal and spatial study of malaria transmission and disease in a suburban area of
Maputo, Mozambique. American Journal of Tropical Medicine and Hygiene 57:550–
559.
Trape, J.-F., A. Tall, N. Diagne, O. Ndiath, A. B. Ly, J. Faye, F. Dieye-Ba, C. Roucher,
C. Bouganali, A. Badiane, F. D. Sarr, C. Mazenot, A. Touré-Baldé, D. Raoult, P.
Druilhe, O. Mercereau-Puijalon, C. Rogier, and C. Sokhna. 2011. Malaria morbidity
and pyrethroid resistance after the introduction of insecticide-treated bednets and
artemisinin-based combination therapies: a longitudinal study. The Lancet Infectious
Diseases 11:925–932.
Trape, J.-F., E. Lefebvre-Zante, F. Legros, G. Ndiaye, H. Bouganali, P. Druilhe, and G.

14

Salem. 1992. Vector density gradients and the epidemiology of urban malaria in
Dakar, Senegal. American Journal of Tropical Medicine and Hygiene 47:181–189.
van der Hoek, W., F. Konradsen, P. H. Amerasinghe, D. Perera, M. K. Piyaratne, and F.
P. Amerasinghe. 2003. Towards a risk map of malaria for Sri Lanka: the importance
of house location relative to vector breeding sites. International Journal of
Epidemiology 32:280–285.
Watson, R. B. 1949. Location and mosquito proofing of dwellings. Pages 1184–1202 in
M. F. Boyd, editor. Malariology: A comprehensive survey of all aspects of this group
of diseases from a global standpoint. W.B. Saunders Co., Philadelphia.
Webb, J. L. A. 2009. Humanity's burden: a global history of malaria. Pages 18–41.
Cambridge University Press, New York.
World Health Organization. 2012. World malaria report 2012. Pages 1–276. World
Health Organization, Geneva.
Yohannes, M., M. Haile, T. A. Ghebreyesus, K. H. Witten, A. Getachew, P. Byass, and
S. W. Lindsay. 2005. Can source reduction of mosquito larval habitat reduce malaria
transmission in Tigray, Ethiopia? Tropical Medicine and International Health
10:1274–1285.
Zaim, M., A. Aitio, and N. Nakashima. 2000. Safety of pyrethroid-treated mosquito nets.
Medical and Veterinary Entomology 14:1–5.
Zhou, G., S. Munga, N. Minakawa, A. K. Githeko, and G. Yan. 2007. Spatial relationship
between adult malaria vector abundance and environmental factors in western Kenya
highlands. American Journal of Tropical Medicine and Hygiene 77:29–35.

15

CHAPTER I: Anopheline larvae in microdam reservoirs.
Abstract
Microdams are important for water management in rural areas of Africa.
Microdams retain water used by people for domestic and agricultural use, but the
potential for malaria vector larvae to use aquatic habitats associated with microdams is
poorly understood. The perimeters of microdam reservoirs in western Kenya were
sampled for Anopheles larvae in the dry and rainy seasons. In the dry season, malaria
vector species and non-vector species were found in microdam-associated habitats. The
primary malaria vector Anopheles funestus was more common in vegetated habitats
relative to open water and hoof print aggregations. In the rainy season, Anopheles
gambiae s.l., another primary malaria vector, dominated Anopheles communities in
microdam reservoirs. Anopheles gambiae s.l. larvae were more common in hoof print
aggregations relative to other habitat types along the microdam reservoir perimeters.
Microdams may provide a dry season refuge habitat for malaria vectors, contributing to
population persistence through the dry season. Microdams also appear to provide
important habitat for An. gambiae s.l. larvae in the rainy season. This suggests a potential
conflict between public health concerns about malaria and people’s need for stable and
reliable sources of water. To alleviate this conflict, variation in habitat suitability within
and among microdams could potentially be exploited to reduce conditions favorable to
malaria vector larvae.
Introduction
Malaria transmission is intricately linked with water due to the obligate aquatic
life stage of all malaria vector mosquito species, though the specific ecology of local

16

malaria vectors determines which types of water bodies contribute to malaria
transmission in a region (Smith et al. 2013, McKeon et al. 2013). While factors such as
adult stage dispersal (Carter et al. 2000), host preference (Garrett-Jones 1964) and
survival (Macdonald 1957, Smith et al. 2007), determine a malaria vector population’s
capacity to transmit malaria, identification of the specific aquatic habitat types utilized by
malaria vector immature stages is critical for understanding malaria vector population
dynamics across space and time (Killeen et al. 2004). The three most widely distributed
primary malaria vector species in Africa, Anopheles arabiensis, Anopheles gambiae s.s.
and Anopheles funestus, utilize a range of aquatic habitat types across their distributions,
but they are generally limited to stagnant fresh water bodies (Gillies and De Meillon
1968). Anopheles gambiae s.s. and An. arabiensis are two members of the An. gambiae
s.l. species complex, a group of at least eight closely related species (Coetzee et al. 2013),
which are morphologically indistinguishable but exhibit important ecological and
behavioral differences in the larval and adult stages. Larvae of An. gambiae s.s. and An.
arabiensis both utilize smaller, temporary aquatic habitats without vegetation and are
often found in the same habitats (Charlwood and Edoh 1996, Minakawa et al. 1999,
Gimnig et al. 2001), but An. arabiensis will additionally exploit larger, permanent
habitats such as rice fields when they are available (Githeko et al. 1996). Anopheles
funestus larvae are found more often in larger, more permanent aquatic habitats with
emergent vegetation (Gimnig et al. 2001).
Infrastructure designed to manipulate the flow and retention of water, such as an
irrigation scheme or dam, is often designed to meet societal needs for reliable sources of
water for domestic or agricultural use. Effects on the production of malaria vectors and

17

thus effects on malaria transmission are not generally considered in the design of such
infrastructure, and therefore the potential for increased malaria transmission exists
(Keiser et al. 2005). Notably, water-related infrastructure projects can potentially reduce
malaria transmission when the projects improve the socioeconomic status of local
residents (Ijumba and Lindsay 2001, Ijumba et al. 2002) or the ecology of the local
malaria vector can be exploited by management techniques (Kitron and Spielman 1989).
An estimated 800,000 small dams, defined by Keiser and colleagues (2005) as
3

impoundments less than 15 m high or storing less than 3 million m of water, had been
built worldwide as of 2001, yet the effects of these dams on malaria in Africa have not
been widely studied. In Ethiopia, children living near microdams had an increased risk of
malaria incidence compared to children living 8-10 km from microdams (Ghebreyesus et
al. 1999), but environmental management of larval habitats has shown potential for
reducing this risk (Yohannes et al. 2005).
The use of aquatic habitats along the perimeters of microdam-created reservoirs
by malaria vectors in western Kenya has not been previously investigated. Therefore, the
objective of this study was to quantify the abundance of malaria vector species along
microdam reservoir perimeters. Additionally, we quantified the abundance of non-vectors
and characterized the overall anopheline community. Furthermore, the contributions of
environmental conditions and seasonal differences in rainfall to variation in the
Anopheles communities within microdam reservoirs were assessed.
Study site
This study took place in Asembo, a rural community of 79 villages in western
Kenya (Nyanza Province). The region sits in the lowlands along the shores of Lake

18

Victoria, with elevations ranging from 1,100 m to 1,400 m above sea level and low
topographic relief. Networks of streams run across the region and drain into Lake
Victoria. Microdams in Asembo are constructed along these drainage basins. The region
2

is relatively densely populated, with about 60,000 residents living in an area of 200 km .
The majority of residents are subsistence farmers, primarily growing maize, sorghum,
cassava, millet, or vegetables, and raising cattle, goats, or chickens (Phillips-Howard et
al. 2003). Cattle are an important resource in the community, being used primarily for
plowing fields and as an economic investment in addition to sources of dairy and meat
products. Thus, a source of water for cattle in Asembo is vital. Reservoirs created by
microdams are a primary source of water for livestock and for domestic use, including
drinking water (Crump et al. 2005).
Rainfall in Asembo is seasonally bimodal with peaks usually occurring in
October-November and March-May. However, rainfall may occur year round, with
monthly precipitation totals ranging from 7 to 49 mm and yearly totals ranging from
1,100 to 1,800 mm from 2003 through 2012. Malaria is holoendemic in the region, with
parasitemia rates in children under 5 being around 50% in 2009 (Hamel et al. 2011). The
predominant species of malaria is Plasmodium falciparum (Welch). The primary malaria
vectors in the region are An. funestus, An. gambiae s.s. and An. arabiensis. Other species
of Anopheles known to occur in the region include Anopheles coustani, Anopheles
rufipes, Anopheles pharoensis and Anopheles squamosus. None of these species are
known to known to transmit malaria in western Kenya, although some have been reported
as locally important malaria vectors in other regions of Africa (Carrara et al. 1990,
Mwangangi et al. 2013).

19

According to Jewsbury and Imevbore (1988), around 50,000 microdams were
constructed in Nyanza Province over three years in the 1950s. Seventy-two functional
microdams were identified within Asembo by local residents from each village for the
current study, twenty-four of which had been built or renovated since 2000. Three of
these were excluded from our study because they were located in a village where
larvicide was being applied to larval Anopheles habitats, including microdam reservoirs.
Eighteen microdams were excluded because they were within 1 km of villages where
insecticide-treated wall lining was applied to every house as part of a separate, controlled
study. Of the remaining 51 microdams in Asembo, 31 still held water in February 2012
and were included in the current study (Figure 1.1).
Methods
Sampling larvae. Microdams were sampled for Anopheles larvae in the first dry
season and the first rainy season of 2012. Each microdam was sampled once from 3-22
February 2012 (dry season) and once from 30 April - 23 May (rainy season).
Standardized samples were taken along the perimeter of each microdam reservoir at 20 m
intervals. Each sample was taken from a quadrat measuring 2 m along the perimeter and
1 m into the reservoir, and consisted of twenty 300 ml dips taken at 20 s intervals. If
sampling locations were primarily aggregations of hoof prints, dipping was considered
impractical and all larvae within the 2 by 1 m quadrat were collected using plastic
pipettes. All collected larvae were kept in separate containers for each instar and
sampling location. The larvae were reared in the lab to the forth instar and identified to
species according to Gillies and Coetzee (1987). In some cases the larvae were reared to
adults, which were also identified to species according to Gillies and Coetzee (1987).

20

Specimens identified morphologically as part of the An. gambiae species complex were
further differentiated to the species level by PCR (Scott et al. 1993). It was not possible to
differentiate An. pharoensis larvae from An. squamosus larvae. Mosquito voucher
specimens were deposited in the Laboratory of Entomology at the Center for Global
Health Research, Kenya Medical Research Institute/Centers for Disease Control and
Prevention in Kisian, Kenya (Table 1.1).
The habitat type for each sample was classified as either open water, vegetated, or
aggregations of hoof prints. The percent vegetated habitat for each microdam was
determined by calculating the proportion of samples classified as vegetated. The soil type
for each microdam was determined using the 1:1,000,000 exploratory soil map of Kenya,
compiled by the Kenya Soil Survey in 1980 (Sombroek et al. 1982). The three soil types
were 1) friable clay/sandy clay loam, 2) friable clay, and 3) firm, silty clay/clay. Daily
precipitation totals for December 2011 through May 2012, as measured by the weather
station at the Kisumu Airport (about 40 km east of Asembo), were downloaded from the
National Climatic Data Center’s Global Summary of Day (GSoD) database.
Statistical analyses. All analyses were performed separately for each season. To
assess differences in anopheline abundance among habitats, the number of larvae of a
given species per sample was related to habitat type using general linear mixed models.
We used a compound symmetrical correlation structure to account for repeated measures
within microdams. Separate analyses were performed for each malaria vector species and
for the most abundant non-vector species.
We were also interested in determining how microdam characteristics influenced
overall anopheline abundance within microdams. To account for variation in reservoir

21

size, the number of larvae collected per meter of reservoir perimeter was calculated for
each microdam, separately for each species. This metric of larval density was related to
percent vegetated habitat and soil type using general linear models. Finally, Anopheles
species richness in each microdam was also related to percent vegetated habitat and soil
type using general linear models.
Results
As expected, there was a dramatic difference in rainfall between the dry and rainy
season sampling periods, both in the weeks preceding the sampling periods and during
the sampling periods (Figure 1.2). Corresponding differences were found in the
Anopheles larvae collected from microdams (Table 1.2). In the dry season, a total of 419
Anopheles larvae were collected, 85% of which were identified to species. The primary
reason for not being able to identify a specimen was the larva dying as an early instar or
pupa. A mixture of malaria vector species and non-vector species occurred in the dry
season. The most common species was Anopheles pharoensis/squamosus. Species
richness per microdam ranged from 0 (n = 10) to 6 (n = 1). In the rainy season, a total of
2,750 Anopheles larvae were collected (75% identified to species), most of which were
An. gambiae s.l. (Table 1.2). A total of 980 An. gambiae s.l. were identified to species by
PCR, 96% of which were An. arabiensis. Species richness in the rainy season ranged
from 0 (n = 2) to 4 (n = 2) species per microdam.
The perimeters of the microdam reservoirs also changed between seasons. While
the water level in many of the reservoirs remained relatively unchanged, the water level
in some of the reservoirs changed dramatically. The mean perimeter length increased
from 168 m (range: 60-540 m) to 183 m (60-580 m). Changes in reservoir water level

22

were associated with changes in the types of habitat occurring along the reservoir
perimeter. Open water was the most common habitat type in the dry season, but vegetated
habitat was more common in the rainy season (Figure 1.3).
Larval abundances of the most common Anopheles species differed among habitat
types in both seasons. In the dry season, Anopheles pharoensis/squamosus larvae were
more common in samples taken from hoof print aggregations and vegetated locations
relative to open water (Figure 1.4). Anopheles funestus larvae were most common in
vegetated locations. Anopheles gambiae s.l. larvae were found in all three habitat types at
similar abundances in the dry season. In the rainy season, An. gambiae s.l. larvae were
most common in hoof print aggregations (Figure 1.5).
Anopheles communities in the microdam reservoirs were associated with the
percent vegetation of the reservoir perimeter. Species richness of Anopheles larvae
increased with increasing vegetation in both seasons, though the effect was stronger in
the dry season (Figure 1.6). Soil type did not have a significant effect on species richness.
Table 1.3 shows the effects of percent vegetated perimeter and soil types on the relative
density of larvae in the microdams. Higher densities of An. funestus larvae were found
with higher percent vegetation in the dry season, and they were lowest in the firm, silty
clay/clay soil type. In both seasons, the relative density of An. pharoensis/squamosus
larvae was associated with the percent vegetation and soil type of the microdams. Finally,
densities of An. gambiae s.l. larvae in the rainy season were higher in microdams with
lower percent vegetation.

23

Discussion
Aquatic habitats along the edges of microdam reservoirs in western Kenya were
utilized in this study by Anopheles larvae, including malaria vectors, indicating a
potential conflict between public health concerns about malaria and people’s need for
stable and reliable sources of water. However, Anopheles communities in the microdam
reservoirs varied considerably between seasons. This variation between seasons was
likely driven by a combination of factors, including the population dynamics of
Anopheles species, the availability of other aquatic habitats, and catchment-scale
hydrology. The seasonal variation of An. funestus and An. gambiae s.l. population sizes
are well characterized in this region, generally correlating positively with lagged
precipitation (Beier et al. 1990, Taylor et al. 1990, Odiere et al. 2007). If microdam
habitats were used equally, relative to other larval habitats on the landscape, higher
collections of these two species would have been expected in the rainy season compared
to the dry season.
The lower relative abundance of An. funestus larvae in the rainy season sampling
suggests that other larval habitats are more important than microdam habitats for An.
funestus during its yearly peak in population abundance. However, microdam habitats are
potentially important for the reproduction of An. funestus and other anophelines in the dry
season. Microdam reservoirs are some of the few water bodies available to anopheline
larvae in the dry season in Asembo. Dry season precipitation accumulation in this region
varies from year to year, but January-February 2012 was a particularly dry period for
Asembo. Water bodies in these drier seasons are limited to a few small springs fed by
ground water, Lake Victoria, and reservoirs at microdams. Thus, habitats along the

24

perimeters of microdam reservoirs potentially represent dry season refuge, which An.
funestus and other anopheline populations use to persist when other aquatic habitats dry
up.
Anopheles gambiae s.l. larvae were found in microdam habitats in both seasons.
Contrary to the findings of Webb (1961) in Tanganyika (Tanzania), An. gambiae s.l.
larvae were found along the perimeters of reservoirs that persisted throughout the dry
season. The relative abundance of An. gambiae s.l. larvae was considerably higher in the
rainy season, agreeing with previous findings regarding the population dynamics of An.
gambiae s.l. However, the exact mechanisms by which An. gambiae s.l. larvae enter
microdam habitats and the importance of these habitats for An. gambiae s.l. population
dynamics remain unclear. Potentially, female anophelines may oviposit in microdam
reservoirs. Additionally, larvae may be aggregated in microdams during the rainy season
when they are flushed from more typical aquatic habitats (Gimnig et al. 2001, Paaijmans
et al. 2007). Heavy rainfall during the rainy season in this region flows across the
landscape, and microdams are designed to collect water flowing along these catchments.
Therefore, even if An. gambiae s.l. females do not oviposit in microdam habitats, these
habitats may contribute to An. gambiae s.l. population dynamics through larvae
completing at least part of their lifetime in them. Of course, additional work must be done
to determine whether An. gambiae s.l. adults emerge from microdam habitats, given the
low correlation between larval abundance and adult emergence within larval habitats
(Mutuku et al. 2006).
Variations in the use of microdam habitats within species and within a season
were associated with the types of available habitat along microdam reservoir perimeters.

25

These associations corresponded with previous findings about habitat preferences of
anopheline larvae. Many Anopheles species, including An. funestus, An. squamosus, and
An. pharoensis, are associated with emergent vegetation in larval habitats (Gillies and De
Meillon 1968) because of the protection it provides from predators. Accordingly, these
species were found at higher abundances in vegetated habitats. Additionally, species
richness was higher in microdam reservoir perimeters with higher percent vegetation,
suggesting that vegetated perimeters provide more ecological niches for Anopheles
larvae.
In contrast, An. gambiae s.l. larvae are often found in habitats without vegetation
(Gimnig et al. 2001). The absence of predators in the small, temporary aquatic habitats
generally associated with An. gambiae s.l. larvae suggests a strategy of avoiding
predators altogether rather than being adapted to find protection in emergent vegetation.
While all of the reservoirs sampled in this study were permanent, and therefore likely
harbored predators, biological communities varied among habitats within the reservoirs.
Thus, An. gambiae s.l. larvae may have been able to utilize specific habitats within the
reservoir perimeters, such as hoof print aggregations, that did not contain predators.
In addition to the effects of vegetation, differences in soil type were associated
with differences among microdams in the number of larvae found for some species. Soil
type potentially influences the abundance of larvae in a microdam reservoir indirectly
through its influence on microbial communities (Bossio et al. 1998). Anopheles larvae
feed primarily on bacteria (Merritt et al. 1992) or algae (Kaufman et al. 2006), and
variation in nutrient availability among soil types may lead to differences in microbial
biomass, diversity, and community composition. Additionally, the microbial community

26

may produce semiochemicals that provide oviposition cues for Anopheles females (Lindh
et al. 2008). However, the roles of soil type and microbial community ecology in the
habitat use of Anopheles mosquitoes are not yet fully understood.

27

APPENDIX

28

Chapter I Tables and Figures

29

Table 1.1: Record of deposition of voucher specimens for Chapter I. The specimens listed
below have been deposited in the Laboratory of Entomology at the Center for Global
Health Research, Kenya Medical Research Institute/Centers for Disease Control and
Prevention in Kisian, Kenya, as samples of those species or other taxa, which were used
in this research. Voucher recognition labels bearing the voucher number 2013-14 have
been attached or included in fluid-preserved specimens.
Species
Life Stage Quantity Preservation
Anopheles gambiae s.l.
Larvae
5 70% EtOH
Anopheles pharoensis/squamosus
Larvae
5 70% EtOH
Anopheles rufipes
Larvae
5 70% EtOH
Anopheles coustani
Larvae
5 70% EtOH
Anopheles funestus
Larvae
5 70% EtOH
Anopheles ardensis
Larvae
5 70% EtOH

30

Table 1.2: Number of each species of Anopheles collected from the perimeters of
microdam reservoirs in Asembo in the dry (February) and rainy (May) seasons of 2012.
Percent of total Anopheles identified in each season shown in parentheses.

Species
Anopheles gambiae s.l.
Anopheles pharoensis/squamosus
Anopheles rufipes
Anopheles coustani
Anopheles funestus
Anopheles ardensis

No. Collected
Dry (%)
Rainy (%)
45 (12.6%)
1,963 (95.9%)
185 (52.0%)
55 (2.7%)
14 (3.9%)
17 (0.8%)
53 (14.9%)
7 (0.3%)
33 (9.3%)
3 (0.1%)
26 (7.3%)
2 (0.1%)

31

Table 1.3: Estimated changes in the density of larvae collected per microdam according to soil types and percent vegetation along the
perimeter. Soil type 1, friable clay/sandy clay loam; soil type 2, friable clay; soil type 3, firm, silty clay/clay. Bold indicates p < 0.10.
Effect on Anopheles funestus was not analyzed for rainy season sampling because only three An. funestus larvae were collected.
Dry season
Species
Habitat type
Estimated change ± SE
Anopheles funestus
Percent vegetation
0.26 ± 0.11
Soil type 1-2
-0.10 ± 0.09
Soil type 1-3
-0.26 ± 0.11
Anopheles gambiae
Percent vegetation
-0.06 ± 0.10
Soil type 1-2
-0.09 ± 0.08
Soil type 1-3
-0.08 ± 0.10
Anopheles pharoensis/squamosus Percent vegetation
0.45 ± 0.25
Soil type 1-2
-0.40 ± 0.22
Soil type 1-3
-0.57 ± 0.27

32

p
0.027
0.310
0.035
0.526
0.289
0.457
0.083
0.073
0.040

Rainy season
Estimated change ± SE
NA
NA
NA
-4.08 ± 2.09
1.24 ± 1.79
1.77 ± 2.21
0.19 ± 0.08
-0.03 ± 0.07
-0.15 ± 0.08

p
NA
NA
NA
0.061
0.497
0.431
0.027
0.643
0.081

Figure 1.1: Pictures of six of the sampled microdams in Asembo. For interpretation of the
references to color in this and all other figures, the reader is referred to the electronic
version of this dissertation.

33

Figure 1.2: Daily precipitation totals for 1 December 2011 (12/01/11) through 31 May 2012. Horizontal black bar indicates dry season
period for Anopheles sampling from microdams, while the red bar indicates the rainy season sampling period.

34

Figure 1.3: Total number of samples taken for Anopheles larvae by habitat type in the dry
and rainy seasons of 2012. Hoof, hoof print aggregations; open, open water; veg,
vegetated habitats.

35

Figure 1.4: Mean number of Anopheles larvae per sample, collected from three habitat
types on the perimeters of microdams in the dry season. Error bars are 95% confidence
intervals. Results shown for An. pharoensis are for larvae identified as An.
pharoensis/squamosus.

Mean number larvae / sample

2.5

hoof

open

vegetated

2.0
1.5
1.0
0.5
0.0
An. funestus

An. gambiae s.l.

36

An. pharoensis

Figure 1.5: Mean number of Anopheles gambiae s.l. larvae per sample, collected from
three habitat types on the perimeters of microdams in the rainy season. Error bars are
95% confidence intervals.

Mean no. larvae / sample

30
25
20
15
10
5
0
hoof

open

37

vegetated

Figure 1.6: Dry and rainy season species richness of Anopheles larvae collected from
microdam reservoir perimeters, by the percent of the perimeter that was vegetated. Points
show observed data (n = 31 microdams per season). Solid lines show estimates from
separate linear regressions for each season, with broken lines showing 95% confidence
intervals.

38

REFERENCES

39

REFERENCES

Beier, J. C., P. V. Perkins, F. K. Onyango, T. P. Gargan, C. N. Oster, R. E. Whitmire, D.
K. Koech, and C. R. Roberts. 1990. Characterization of malaria transmission by
Anopheles (Diptera: Culicidae) in western Kenya in preparation for malaria vaccine
trials. Journal of Medical Entomology 27:570–577.
Bossio, D., K. Scow, N. Gunapala, and K. Graham. 1998. Determinants of Soil Microbial
Communities: Effects of Agricultural Management, Season, and Soil Type on
Phospholipid Fatty Acid Profiles. Microbial ecology 36:1–12.
Carrara, G. C., V. Petrarca, M. Niang, and M. Coluzzi. 1990. Anopheles pharoensis and
transmission of Plasmodium falciparum in the Senegal River delta, West Africa.
Medical and Veterinary Entomology 4:421–424.
Carter, R. R., K. N. K. Mendis, and D. D. Roberts. 2000. Spatial targeting of
interventions against malaria. Bulletin of the World Health Organization 78:1401–
1411.
Charlwood, J. D., and D. Edoh. 1996. Polymerase chain reaction used to describe larval
habitat use by Anopheles gambiae complex (Diptera: Culicidae) in the environs of
Ifakara, Tanzania. Journal of Medical Entomology 33:202–204.
Coetzee, M., R. H. Hunt, R. C. Wilkerson, A. della Torre, M. B. Coulibaly, and N. J.
Besansky. 2013. Anopheles coluzzii and Anopheles amharicus, new members of the
Anopheles gambiae complex. Zootaxa 3619:246–274.
Crump, J. A., P. O. Otieno, L. Slutsker, B. H. Keswick, D. H. Rosen, R. M. Hoekstra, J.
M. Vulule, and S. P. Luby. 2005. Household based treatment of drinking water with
flocculant-disinfectant for preventing diarrhoea in areas with turbid source water in
rural western Kenya: cluster randomised controlled trial. BMJ 331:478.
Garrett-Jones, C. 1964. The human blood index of malaria vectors in relation to
epidemiological assessment. Bulletin of the World Health Organization 30:241–261.
Ghebreyesus, T. A., M. Haile, K. H. Witten, A. Getachew, A. M. Yohannes, M.
Yohannes, H. D. Teklehaimanot, S. W. Lindsay, and P. Byass. 1999. Incidence of
malaria among children living near dams in northern Ethiopia: community based
incidence survey. BMJ 319:663–666.
Gillies, M. T., and B. De Meillon. 1968. The Anophelinae of Africa south of the Sahara.
Pages 1–343. South African Institute for Medical Research, Johannesburg, South
Africa.
Gillies, M. T., and M. Coetzee. 1987. A supplement to the Anophelinae of Africa south

40

of the Sahara (Afrotropical Region). Pages 1–143. South African Institute for
Medical Research, Johannesburg, South Africa.
Gimnig, J. E., M. Ombok, L. Kamau, and W. A. Hawley. 2001. Characteristics of larval
anopheline (Diptera: Culicidae) habitats in western Kenya. Journal of Medical
Entomology 38:282–288.
Githeko, A. K., M. W. Service, C. M. Mbogo, and F. K. Atieli. 1996. Resting behaviour,
ecology and genetics of malaria vectors in large scale agricultural areas of western
Kenya. Parassitologia 38:481–489.
Hamel, M. J., K. Adazu, D. Obor, M. Sewe, J. M. Vulule, J. M. Williamson, L. Slutsker,
D. R. Feikin, and K. F. Laserson. 2011. A reversal in reductions of child mortality in
western Kenya, 2003-2009. American Journal of Tropical Medicine and Hygiene
85:597–605.
Ijumba, J. N., and S. W. Lindsay. 2001. Impact of irrigation on malaria in Africa: paddies
paradox. Medical and Veterinary Entomology 15:1–11.
Ijumba, J. N., F. W. Mosha, and S. W. Lindsay. 2002. Malaria transmission risk
variations derived from different agricultural practices in an irrigated area of northern
Tanzania. Medical and Veterinary Entomology 16:28–38.
Jewsbury, J. M., and A. M. Imevbore. 1988. Small dam health studies. Parasitology
Today 4:57–59.
Kaufman, M. G., E. Wanja, S. Maknojia, M. N. Bayoh, J. M. Vulule, and E. D. Walker.
2006. Importance of Algal Biomass to Growth and Development of Anopheles
gambiaeLarvae. Journal of Medical Entomology 43:669–676.
Keiser, J., M. C. de Castro, M. F. Maltese, R. Bos, M. Tanner, B. H. Singer, and J.
Utzinger. 2005. Effect of irrigation and large dams on the burden of malaria on a
global and regional scale. American Journal of Tropical Medicine and Hygiene
72:392–406.
Killeen, G. F., A. A. Seyoum, and B. G. J. Knols. 2004. Rationalizing historical successes
of malaria control in Africa in terms of mosquito resource availability management.
American Journal of Tropical Medicine and Hygiene 71:87–93.
Kitron, U., and A. Spielman. 1989. Suppression of transmission of malaria through
source reduction: antianopheline measures applied in Israel, the United States, and
Italy. Reviews of infectious diseases 11:391–406.
Lindh, J. M. J., A. A. Kännaste, B. G. J. Knols, I. I. Faye, and A. K. A. Borg-Karlson.
2008. Oviposition responses of Anopheles gambiae s.s. (Diptera: Culicidae) and
identification of volatiles from bacteria-containing solutions. Journal of Medical

41

Entomology 45:1039–1049.
Macdonald, G. 1957. The epidemiology and control of malaria. Oxford University Press,
Oxford.
McKeon, S. N., C. D. Schlichting, M. M. Povoa, and J. E. Conn. 2013. Ecological
Suitability and Spatial Distribution of Five Anopheles Species in Amazonian Brazil.
American Journal of Tropical Medicine and Hygiene 88:1079–1086.
Merritt, R. W., R. H. Dadd, and E. D. Walker. 1992. Feeding behavior, natural food, and
nutritional relationships of larval mosquitoes. Annual review of entomology 37:349–
376.
Minakawa, N., C. M. Mutero, J. I. Githure, J. C. Beier, and G. Yan. 1999. Spatial
distribution and habitat characterization of anopheline mosquito larvae in western
Kenya. American Journal of Tropical Medicine and Hygiene 61:1010–1016.
Mutuku, F. M., M. N. Bayoh, J. E. Gimnig, J. M. Vulule, L. Kamau, E. D. Walker, E.
Kabiru, and W. A. Hawley. 2006. Pupal habitat productivity of Anopheles gambiae
complex mosquitoes in a rural village in western Kenya. American Journal of
Tropical Medicine and Hygiene 74:54–61.
Mwangangi, J. M., E. J. Muturi, S. M. Muriu, J. Nzovu, J. T. Midega, and C. M. Mbogo.
2013. The role of Anopheles arabiensis and Anopheles coustani in indoor and
outdoor malaria transmission in Taveta District, Kenya. Parasites & Vectors 6:114.
Odiere, M., M. N. Bayoh, J. E. Gimnig, J. M. Vulule, L. Irungu, and E. D. Walker. 2007.
Sampling outdoor, resting Anopheles gambiae and other mosquitoes (Diptera:
Culicidae) in western Kenya with clay pots. Journal of Medical Entomology 44:14–
22.
Paaijmans, K. P., M. O. Wandago, A. K. Githeko, and W. Takken. 2007. Unexpected
High Losses of Anopheles gambiae Larvae Due to Rainfall. PLoS ONE 2:e1146.
Phillips-Howard, P. A., B. L. Nahlen, J. A. Alaii, F. O. ter Kuile, J. E. Gimnig, D. J.
Terlouw, S. P. Kachur, A. W. Hightower, A. A. Lal, E. Schoute, A. J. Oloo, and W.
A. Hawley. 2003. The efficacy of permethrin-treated bed nets on child mortality and
morbidity in western Kenya I: development of infrastructure and description of study
site. American Journal of Tropical Medicine and Hygiene 68:3–9.
Scott, J. A., W. G. Brogdon, and F. H. Collins. 1993. Identification of single specimens
of the Anopheles gambiae complex by the polymerase chain reaction. American
Journal of Tropical Medicine and Hygiene 49:520–529.
Smith, D. L., F. E. McKenzie, R. W. Snow, and S. I. Hay. 2007. Revisiting the Basic
Reproductive Number for Malaria and Its Implications for Malaria Control. PLoS

42

Biology 5:e42.
Smith, M. W., M. G. Macklin, and C. J. Thomas. 2013. Earth-Science Reviews. Earth
Science Reviews 116:109–127.
Sombroek, W. G., H. M. H. Braun, and B. J. A. van der Pouw. 1982. Exploratory soil
map and agro-climatic zone map of Kenya, 1980. Scale 1: 1,000,000. Kenya Soil
Survey, Nairobi, Kenya.
Taylor, K. A., J. K. Koros, J. Nduati, R. S. Copeland, F. H. Collins, and A. D. BrandlingBennett. 1990. Plasmodium falciparum infection rates in Anopheles gambiae, An.
arabiensis, and An. funestus in western Kenya. American Journal of Tropical
Medicine and Hygiene 43:124–129.
Webbe, G. G. 1961. Breeding sites of malaria vectors in relation to water management
and agricultural practice in the Sukumaland area of Tanganyika. East African
Medical Journal 38:239–245.
Yohannes, M., M. Haile, T. A. Ghebreyesus, K. H. Witten, A. Getachew, P. Byass, and
S. W. Lindsay. 2005. Can source reduction of mosquito larval habitat reduce malaria
transmission in Tigray, Ethiopia? Tropical Medicine and International Health
10:1274–1285.

43

CHAPTER II: Precipitation data improve predictions of larval malaria vector habitat
locations because of seasonal differences in precipitation.
Abstract
Predictive models of malaria vector larval habitat locations may provide a basis
for understanding the spatial determinants of malaria transmission. Four landscape
variables (topographic wetness index, soil type, land use-land cover, and distance to
stream) and accumulated precipitation were used to model larval habitat locations
through two methods: logistic regression and random forest. The random forest models
were more accurate than the logistic regression models, especially when accumulated
precipitation was included to account for seasonal differences in precipitation. Larval
habitats were more likely to be present in locations with a lower slope to contributing
area ratio, closer to streams, and in agricultural land use. Differences among soil types
were also found, and the probability of larval habitat presence increased with increasing
accumulated precipitation. Estimated larval habitat probabilities from the random forest
model produced here can be used to study the influence of larval habitat proximity to
houses on the number of malaria vectors resting indoors in this landscape. Finally, the
sampling strategy employed here for model parameterization could serve as a framework
for creating predictive larval habitat models to assist in larval control efforts.
Introduction
Malaria remains as one of the most significant infectious diseases affecting people
in poverty, with an estimated 219 million cases of malaria worldwide in 2010 killing
660,000 people (World Health Organization 2012). In many settings, the spatial
distribution of malaria is heterogeneous, with malaria prevalence differing among
households within a community (Greenwood 1989, Gamage-Mendis et al. 1991, Trape et

44

al. 1992, Clark et al. 2008). Of course, socioeconomic and immunological differences
contribute to heterogeneity of malaria prevalence in people (Greenwood 1989, GamageMendis et al. 1991). Additionally, the spatial distribution of malaria vectors may coincide
with the spatial distribution of parasitemia in people, suggesting a causative relationship
(Trape et al. 1992). Furthermore, the spatial distribution of the adult malaria vectors is
determined, in part, by the spatial distribution of the aquatic habitats of the vector
mosquito larvae. In many landscapes, houses closer to larval habitats have more adult
malaria vectors resting indoors relative to houses farther from larval habitats (Trape et al.
1992, Ribeiro et al. 1996, Zhou et al. 2004, Bogh et al. 2007). Therefore, understanding
the factors that determine the distribution of the larval habitats facilitates our
understanding of the spatial determinants of malaria transmission.
The vast majority of deaths from malaria (91%) now occur in Africa (World
Health Organization 2012), where the primary vector mosquitoes are perhaps the most
efficient vectors of malaria in the world. Two of the most widely distributed vectors in
Africa are Anopheles gambiae s.s. Giles and Anopheles arabiensis Patton, which are both
members of a species complex of eight closely related, morphologically indistinguishable
species known collectively as Anopheles gambiae s.l. (Coetzee et al. 2013). In many
regions the larval habitats of An. gambiae s.s. and An. arabiensis are similar, and in fact,
the two species are often found within the same larval habitats (Charlwood and Edoh
1996, Minakawa et al. 1999, Gimnig et al. 2001). These larval habitats are generally
smaller, temporary bodies of stagnant water (Gimnig et al. 2001, Mutuku et al. 2006)
with rain being the main source of the water.

45

The locations of larval An. gambiae s.l. habitats are associated with environmental
features in many landscapes. Previous studies have found more larval habitats closer to
streams (Mutuku et al. 2009) and in locations with agricultural land uses (Mushinzimana
et al. 2006, Mutuku et al. 2009). Others have used a topographic wetness index (TWI;
Beven and Kirkby 1979) to predict the locations of larval habitats, finding more larval
habitats in locations having a combination of greater upslope area contributing to
drainage and less slope (Clennon et al. 2010, Li et al. 2011, Nmor et al. 2013). The
influence of soil types on the presence of larval habitats has largely been ignored,
although Bogh and colleagues (2003) found larval habitats exclusively in alluvial soils in
The Gambia. Finally, seasonal differences in rainfall likely influence the number of larval
habitats on the landscape (Gimnig et al. 2001, Mutuku et al. 2009, Munga et al. 2009,
Clennon et al. 2010, Li et al. 2011).
The objectives of this study were to create a model for predicting larval An.
gambiae s.l. habitat locations using landscape variables that predict the likelihood of
stagnant water bodies, and to account for seasonal changes in habitat probability based on
accumulated precipitation. A model predicting accurately the locations of malaria vector
larval habitats would allow researchers to investigate the links between larval habitat
distribution and adult malaria vector distribution across large landscapes where manually
mapping the larval habitats is infeasible (Bogh et al. 2007). Additionally, such a model
could be useful for malaria control programs, allowing program managers to focus their
efforts to areas where larval habitats are most likely to occur.

46

Study site
The Asembo region of Rarieda District in western Kenya (Figure 2.1A) is a rural
2

community of about 60,000 people covering about 200 km . Most of the residents are
subsistence farmers, and the landscape is largely dominated by small-scale agriculture.
Small plots of land generally surround family-based groups of houses, or compounds,
further arranged into villages. While the compounds are highly dispersed within villages,
the boundaries between villages are often discernable only by residents (Phillips-Howard
et al. 2003) (Figure 2.1B). Asembo sits in the lowlands along the shores of Lake Victoria,
with elevations ranging from 1,100 m to 1,400 m above sea level and low topographic
relief. Networks of streams run across the region and drain into Lake Victoria. Farmland
is common in these low-lying drainage basins, as well as throughout the region. Houses
are mostly absent within 100 m of the streams. Rainfall is seasonally bimodal but may
occur year round, with monthly precipitation totals ranging from 7 to 49 mm and yearly
totals ranging from 1,100 to 1,800 mm from 2003 through 2012.
Malaria is holoendemic in Asembo, with parasitemia rates in children under 5
being around 50% in 2009 (Hamel et al. 2011). Similar to rainfall patterns, malaria
transmission occurs year round, with seasonal peaks in May-July and October-November.
The predominant species of malaria is Plasmodium falciparum (Welch). Two of the
primary malaria vectors in the region are An. gambiae s.s. and An. arabiensis, the only
two members of the An. gambiae s.l. species complex found here. Larval An. gambiae s.l.
habitats are numerous and widespread in Asembo yet heterogeneously distributed. This
makes it difficult to establish a relationship between larval habitats and the spatial
distribution of the adult vectors or malaria prevalence in people. A 10 by 10 km study site

47

was defined within Asembo to capture variation in the determinants of larval habitat
location across a relatively large area. Because we wanted to include the lakeshore in the
2

study site, the southern border fell largely within the lake, leaving 96.43 km of actual
landmass in the 10 by 10 km site.
Methods
Larval habitat ground surveys. The study site was divided into 500 by 500 m
quadrats for larval habitat ground surveys. After excluding the quadrats that fell
completely in the lake, from the remaining 393 quadrats, we selected 31 for larval habitat
ground surveys using spatially stratified random sampling (Figure 2.1C). Thirty-one
quadrats were the maximum we could sample during the targeted time frame, which was
the end of the long rainy season to coincide with the peak An. gambiae s.l. population
level (Beier et al. 1990, Taylor et al. 1990, Odiere et al. 2007). For spatial stratification
the study area was divided into 25, 2 by 2 km quadrats, containing 16 of the 500 by 500
m quadrats each. The smaller quadrats were randomly selected from groups defined by
the larger quadrats. Spatial stratification was implemented to avoid the problem of
sampling a cluster of quadrats in a certain area of the grid and missing variation in the
predictor variables (Hurlbert 1984).
The quadrats were each surveyed exhaustively on 1 of 22 days between 17 May
2011 and 4 July 2011, during the end of the long rainy season. All potential An. gambiae
s.l. habitats found in the quadrats were georeferenced with GPS units. Six field workers
spaced 20 m from each other walked from one end of a quadrat to the other, using
ArcPad on a GPS unit for navigating the borders of the quadrat. This was repeated until
the entire quadrat was covered, usually in four to five passes. This approach allowed us to

48

say, with certainty, where habitats were absent during the survey. Where habitats were
present, we recorded the location and presence or absence of Anopheles larvae. Larval
An. gambiae s.l. habitats were defined as any stagnant body of water, regardless of
whether Anopheles were present on the day of the ground survey, and falling under the
following categories: drainage channel, burrow pit, rain pool, runoff, cluster of hoof
prints, stream bed pool, pond/reservoir, wet meadow, well and tire track (Mutuku et al.
2006). For a subset of habitats (the first five Anopheles-positive habitats for each of the
four to five passes across each quadrat), Anopheles larvae and pupae were collected and
transported to the lab for species identification to confirm that the habitats were being
used by An. gambiae s.l. Larvae were raised to fourth-stage instars for identification,
while pupae were allowed to eclose as adults before identification. All identifications
were done according to Gillies and Coetzee (1987). Mosquito voucher specimens were
deposited in the Laboratory of Entomology at the Center for Global Health Research,
Kenya Medical Research Institute/Centers for Disease Control and Prevention in Kisian,
Kenya (Table 2.1).
To capture variation in habitat location across time due to seasonal rainfall
patterns, monthly ground surveys were done in two neighboring villages, Aduoyo-Miyare
2

and Nguka, covering an area of 6.22 km within the 10 by 10 km study site (Figure
2.1D). Two local field workers with extensive knowledge of the villages walked
throughout the whole of each village over the course of one to three days, depending on
the number of habitats encountered, each month from April 2011 through June 2012.
Potential An. gambiae s.l. habitats were defined and recorded as above, and we had the
ability to say where habitats were absent within the two villages each month.

49

Environmental data. Spatial data for soils, land use-land cover (LULC), distance
to the nearest stream, and TWI were created at a 20 m resolution across the study site. All
data were assembled in ArcGIS (ESRI, Redlands, CA) in raster data structures. Soil data
were taken from the 1:1,000,000 exploratory soil map of Kenya, compiled by the Kenya
Soil Survey in 1980 (Sombroek et al. 1982). The three soil types in Asembo were 1)
friable clay, 2) friable clay/sandy clay loam, and 3) firm, silty clay/clay. A satellite image
from the IKONOS-2 sensor was used to create the LULC classification. Briefly,
unsupervised classification was done using the K-means method. Classes were combined
into a binary data layer of agricultural or non-agricultural land use. All streams in
Asembo were mapped using GPS units, and the Euclidean distance to the nearest stream
was calculated.
The TWI data were derived from a digital elevation model (DEM) of the study
site. The DEM was created using local universal kriging to interpolate 11,130 GPS
elevation records previously taken within Asembo (Hightower et al. 1998, Ombok et al.
2010). The ArcGIS extension TauDEM 5.0 (Tarboton, Utah State University) was used
to calculate a TWI. First, flow direction and slope were calculated. Contributing area
catchments were calculated using the flow direction and slope data. Finally, TWI was
calculated as the ratio of the slope to the contributing area, and the values rescaled to the
range, from 0-100, by taking:
(TWIi – min(TWI)) × 100 / (max(TWI) – min(TWI)) for each i value of TWI.
Because the TWI was calculated as the ratio of slope to contributing area, the lowest
values (near 0) represent the wettest locations, while the highest TWI values represent the
driest areas.

50

Daily precipitation totals for March 2011 to July 2012, as measured by the
weather station at the Kisumu Airport (about 40 km east of Asembo), were downloaded
from the National Climatic Data Center’s Global Summary of Day (GSoD) database
(Figures 2 and 3). For missing daily data at the Kisumu weather station (n = 19), the
inverse distance weighted mean of surrounding GSoD weather stations (within 250 km)
was used. We summed cumulative precipitation totals from 0 to 30 days prior to each day
of larval habitat ground surveys, and selected the best cumulative n-day total based on the
criteria outlined below.
Statistical methods. We used two approaches for modeling the distribution of
larval habitats across the landscape, logistic regression and random forest, and each
approach was used separately on the two datasets (one from the 10 by 10 km area and the
other from the 15 monthly ground surveys in Aduoyo-Miyare and Nguka). While logistic
regression has been used in species distribution modeling more frequently, ecologists
have recently started using the random forest method as well, because it does not require
any assumptions about the distribution of the data (Bisrat et al. 2011, Hernández et al.
2013).
We built a series of candidate logistic regression models to select the most useful
predictor variables. Univariate models were built for each of the cumulative precipitation
totals from 0 to 30 days, and the model with the lowest BIC was selected as the best
precipitation measure to use in further logistic regression models. All possible
combinations of our five predictor variables (TWI, soil, LULC, distance to stream, and
precipitation) were used to build 31 candidate logistic regression models, with the top
models again being selected according to the lowest BIC. These analyses were

51

implemented in the statistical software R 2.14.2 (R Development Core Team, Vienna,
Austria).
Random forest is a machine learning classification method extending on the
classification and regression tree approach (CART), which works by recursive binary
partitioning of the data space into increasingly homogenous regions (Breiman et al. 1984,
Bisrat et al. 2011). Random forest works by fitting and combining many CARTs to create
a more accurate prediction (Breiman 2001, Bisrat et al. 2011). We implemented the
random forest approach using the R package ‘randomForest’ (Liaw and Wiener 2002).
The best cumulative precipitation measure for use in the random forest models was
determined by the mean decrease in the Gini impurity criterion when removing the
variable from the full model (TWI, soil, LULC, distance to stream, and precipitation).
The cumulative precipitation measure with the highest mean decrease in the Gini
impurity criterion was used in the final random forest models.
The top models from both approaches within each dataset were evaluated by
determining their accuracy at predicting larval habitat presence and absence for holdout
data. Fifty percent of each dataset was randomly selected as a holdout dataset before
model building. Evaluation of model accuracy required the selection of a threshold at
which to convert predicted probabilities into larval habitat presence or absence. Because
threshold specific accuracy statistics can be sensitive to the threshold used for
conversion, we generated an optimal threshold value by minimizing the absolute value of
the difference between sensitivity and specificity (Liu et al. 2005). This approach was
chosen because both sensitivity and specificity were equally important for the intended
application of the predictive models. To assess the performance of each model, we

52

calculated the sensitivity, specificity, percent correctly classified (PCC), and kappa of
each approach at each of the thresholds from the methods above. We also calculated the
threshold-independent area under the ROC curve (AUC) statistic from ROC plots using
the R package ‘SDMTools’.
Finally, we assessed correlation among the cumulative precipitation measures
using Pearson’s correlation coefficient. The contribution of cumulative 30-day
precipitation to variation in the number of habitats found each month in Aduoyo-Miyare
and Nguka was quantified using simple linear regression.
Results
In the 31 sampling quadrats selected from the 10 by 10 km study site, we recorded
the locations of 1,673 larval An. gambiae s.l. habitats. Six of the quadrats did not have
any larval habitats, while the mean number of habitats per 500 by 500 m quadrat was 54.
Anopheles larvae were present in 921 of the 1,673 habitats on the day each habitat was
recorded. Of the Anopheles larvae and pupae collected from 141 of the habitats, 79%
were identified as An. gambiae s.l. In 15 monthly ground surveys in Aduoyo-Miyare and
Nguka, a total of 6,770 larval An. gambiae s.l. habitats were recorded. The number of
larval habitats in this area varied by month, ranging from 104 to 953 with a mean of 451.
The number of larval habitats recorded in the two villages each month increased with
increasing cumulative 30-day precipitation on the final day of ground surveys each month
2

(R = 0.1931, p = 0.1012; Figure 2.2).
The best cumulative precipitation total to use in the models differed between the
datasets and between the modeling approaches, although all cumulative precipitation
totals were highly correlated with each other (Table 2.2). For the 15 monthly ground

53

surveys in Aduoyo-Miyare and Nguka, the univariate logistic regression model with the
lowest BIC within the precipitation candidate models used the cumulative 30-day
precipitation (Figure 2.3), whereas the random forest model using the cumulative 21-day
precipitation (Figure 2.3) had the highest mean decrease in the Gini impurity criterion.
For the 10 by 10 km data, the univariate logistic regression model with the lowest BIC
used the cumulative 6-day precipitation (Figure 2.4), while the random forest model
using the cumulative 14-day precipitation (Figure 2.4) had the highest mean decrease in
the Gini impurity criterion. Each of these precipitation measures was used within its
respective dataset and modeling approach moving forward.
The environmental variables used in the best logistic regression models for
predicting the locations of larval An. gambiae s.l. habitats differed slightly between the
datasets. For the 15-month dataset from Aduoyo-Miyare and Nguka, the logistic
regression model with the lowest BIC included all five of the variables (Table 2.3). No
other model had a ΔBIC less than 20. Larval habitats were more likely to be found in
locations with a lower TWI (i.e. wetter because of a lower slope to contributing area
ratio), closer to streams, in agricultural land use, and in the friable clay/sandy clay loam
soil type (Table 2.4). The probability of a larval habitat increased with increasing
cumulative 30-day precipitation (Table 2.4).
For the 10 by 10 km dataset, the logistic regression model with the lowest BIC
included four of the variables (TWI, distance to stream, soil type, and cumulative 6-day
precipitation). No other model had a ΔBIC less than 9 (Table 2.5). Larval habitats were
again more likely to be found in locations with a lower TWI, closer to streams, and in the
friable clay/sandy clay loam soil type (Table 2.6). Counterintuitively, the probability of a

54

larval habitat according to this model decreased with increasing cumulative 6-day
precipitation.
The most accurate model for predicting larval An. gambiae s.l. habitat locations in
the 10 by 10 km area was built using the random forest method and all five variables
(AUC = 0.864). Removing the cumulative 14-day precipitation from this model reduced
the accuracy of the model (Table 2.7). The best logistic regression model for the 10 by 10
km area was less accurate than the random forest model when evaluated against the
holdout data (Table 2.7). When the probabilities of larval habitat location across the
entire 10 by 10 km study site were estimated using the most accurate models from each
method, the random forest model clearly produced a more heterogeneous landscape at a
fine scale than that produced by the logistic regression method (Figure 2.5).
Similarly, the most accurate model for predicting larval An. gambiae s.l. habitat
locations over the 15 monthly surveys in Aduoyo-Miyare and Nguka was built using the
random forest method and all five variables (Table 2.8). Again, removing the cumulative
21-day precipitation from this model reduced its accuracy, and the best logistic regression
model for the 15 monthly surveys in Aduoyo-Miyare and Nguka was less accurate than
the random forest model when evaluated against the holdout data (Table 2.8).
Discussion
The use of models to predict the distribution of species is common in ecology
(Guisan and Zimmermann 2000), and novel approaches to building these models such as
random forest have become more widely available in recent years. We used two methods
to predict the probability of larval An. gambiae s.l. habitat across the landscape and over
time, and the random forest method created more accurate models than the logistic

55

regression method. This may be due to the differences between the two methods in
heterogeneity of larval habitat probability on a fine scale, which can be seen in the
estimates across the entire 10 by 10 km study site (Figure 2.5). The estimates from the
most accurate random forest model are more fragmented, in that high-probability
locations are nearer to low-probability locations relative to the estimates from the logistic
regression model. At a broad scale, a general pattern for higher probability locations is
still apparent with the random forest estimates. However, the fine scale heterogeneity in
the random forest estimates more closely reflects the nature of actual larval habitat
distribution on the ground, where larval An. gambiae s.l. habitats are distributed as many
small patches rather than one continuous, large patch.
The creation of larval An. gambiae s.l. habitats (which are temporary, small
bodies of stagnant water) depends on rainfall, which varies seasonally across the range of
the species complex. Therefore, an important question in the application of predictive
larval habitat models is whether models parameterized with data for habitat locations in
one season are applicable to another season (Li et al. 2011). One strategy to deal with
differences between seasons is to account for variation in precipitation. We found more
larval habitats in months with more precipitation compared to the same area in months
with less precipitation. Thus, including accumulated precipitation in our models
improved the accuracy of larval habitat location predictions.
While this confirms that variation in precipitation influences larval An. gambiae
s.l. habitats, the relationship may be more complex than it first appears. For example, it
may not be linear. Rather, the number of larval habitats may increase monotonically with
accumulated precipitation up to a threshold, after which more of the water on the

56

landscape flows as surface sheet or channeled water, which is unsuitable aquatic habitat
for An. gambiae s.l. larvae. Additionally, different habitat types may respond differently
to increasing accumulated precipitation. Stagnant water forming in drainage channels and
stream bed pools may be described better by a threshold relationship than the water
filling burrow pits, hoof prints and tire tracks, because the former develops from channel
and sheet water made stationary by diminished water flows, whereas the latter forms
from water accumulating into various catchments not associated with channels. These
additional factors may explain some of the uneven residual error seen in Figure 2.2,
where the 4 months in the red box falling above the fitted regression line have more
larval habitats with a lower accumulated precipitation relative to the 3 months in the blue
box falling below the fitted regression line.
An unexpected result was found for the best logistic regression model of the 10 by
10 km data, as the odds of larval habitat presence decreased with increasing cumulative
6-day precipitation. Most likely this reflects a limitation of the 10 by 10 km data
collection rather than the true influence of precipitation on larval habitat presence, given
the range of cumulative 6-day precipitation over the 49-day period (1.5 mm – 51.1 mm;
Figure 2.4). The sampling strategy for those data was designed to capture variation in
landscape variables over space. While precipitation varied among the days of the ground
surveys, we were not able to capture that variation over the full range of values for the
landscape variables. Instead, the effect of accumulated precipitation in this particular
model may be an indication of some other property differing between the quadrats
sampled on days of higher and lower accumulated precipitation. Furthermore, the
temporal scale over which larval habitats respond to variation in accumulated

57

precipitation is probably closer to monthly than daily. That is, ground surveys conducted
at monthly intervals in the same area are more likely to be different than daily samples
within a month in the same area.
The sampling designs of these two datasets allowed us to address two
complementary goals. The monthly surveys in Aduoyo-Miyare and Nguka captured
variation in precipitation across both dry and rainy seasons in the same landscape. This
provided a stronger logical basis for inferences about the relationship between seasonal
variation in precipitation and variation in the location and number of larval habitats.
However, the small spatial extent of that landscape limited the applicability of the model
results across a larger area. Conversely, limiting the ground surveys of the 31 quadrats
from the 10 by 10 km study site to be within one season likely impeded our ability to
infer much about the affect of precipitation on these data. On the other hand, replicating
our sampling effort across space in the 31 quadrats captured more variation in landscape
variables, allowing us to apply the results of models based on these data to a larger area.
Thus, we are currently using estimated larval habitat probabilities from the random forest
model in the 10 by 10 km study site to study the influence of larval habitat proximity to
houses on the number of malaria vectors resting indoors in this landscape where mapping
2

all of the larval habitats across 100 km is infeasible.
Additionally, the sampling strategy used in the 10 by 10 km site could serve as a
framework for creating predictive larval habitat models for larval control. Targeted larval
control is often cited as a useful application of predictive larval habitat models (Li et al.
2011, Nmor et al. 2013), and we agree that there is potential for this application. For
example, malaria control programs could identify areas suited to environmental

58

management such as filling in burrow pits and engineering drainage channels to drain
more completely. Additionally, allowing application crews to focus on areas with a
higher probability of larval habitat presence would reduce the time, and therefore the
cost, of larviciding. However, models fitted to data from a single geographic location
may have limited generalizability (Strauss and Biedermann 2007). Spatially stratified
random samples, repeated across a variable landscape, may allow malaria control
programs to build models that are useful over larger areas.
While the models developed here exclusively used physical and environmental
factors as predictor variables, the formation of larval An. gambiae s.l. habitats also
depends on human behavior. For example, landowners in Asembo create small drainage
channels around fields. Stagnant water left behind in the channels creates habitats for An.
gambiae s.l. larvae (Mutuku et al. 2006). The locations of these drainage channels are
often in low-lying agricultural areas, and therefore our models were able to predict the
locations of most of the drainage channels. However, drainage channels are not found in
100% of low-lying agricultural areas, probably in part because of individual variation in
landowner decision-making. Larval habitats formed from burrow pits and aggregations of
hoof prints are also subject to variation in human behavior. While our models were able
to correctly predict the locations of most of these habitats, interactions between the
physical landscape and human behavior likely account for some of the locations
identified incorrectly by the models.

59

APPENDIX

60

Chapter II Tables and Figures

61

Table 2.1: Record of deposition of voucher specimens for Chapter II. The specimens
listed below have been deposited in the Laboratory of Entomology at the Center for
Global Health Research, Kenya Medical Research Institute/Centers for Disease Control
and Prevention in Kisian, Kenya, as samples of those species or other taxa, which were
used in this research. Voucher recognition labels bearing the voucher number 2013-14
have been attached or included in fluid-preserved specimens.
Species
Anopheles gambiae s.l.

Life Stage Quantity Preservation
Larvae
5 70% EtOH

62

Table 2.2: Pearson’s correlation coefficient (r) matrix of four cumulative precipitation
measures for (A) 22 days of larval habitat ground surveys in the 10 by 10 km area from
17 May to 4 July 2011, and (B) the last day of larval habitat ground surveys each month
in Aduoyo-Miyare and Nguka from April 2011 to June 2012. 0-day refers to precipitation
total for the day of ground surveys, while 6-day, 14-day, 21-day, and 30-day refer to the
cumulative precipitation total for the day of the ground surveys plus the previous 6 days,
14 days, 21 days, or 30 days, respectively. Note that similar correlations were found
among other cumulative n-day precipitation totals at 1-day increments from 1-day to 30
day (data not shown).
0-Day 6-Day
A) 17 May - 4 July 2011
0-Day 1.000
6-Day 0.411 1.000
14-Day 0.332 0.721
21-Day 0.405 0.477
30-Day 0.392 0.611

14-Day 21-Day 30-Day

1.000
0.794
0.578

1.000
0.517

1.000

B) April 2011 - June 2012
0-Day 1.000
6-Day 0.812 1.000
14-Day 0.665 0.826 1.000
21-Day 0.408 0.555 0.841
30-Day 0.336 0.537 0.819

1.000
0.979

1.000

63

Table 2.3: The top four logistic regression candidate models for the 15 monthly ground
surveys in Aduoyo-Miyare and Nguka based on Bayesian information criterion (BIC), in
order of increasing difference in BIC from the top model (ΔBIC) and decreasing BIC
weight, w. TWI, topographic wetness index; LULC, land use-land cover; DS, distance to
the nearest stream; Precip., cumulative 30-day precipitation total; NA, not applicable.
Model
TWI + LULC + DS + Soil + Precip.
TWI + DS + Soil + Precip.
TWI + LULC + DS + Precip.
TWI + DS + Precip.

BIC
45933.1
45955.6
46045.5
46073.7

64

ΔBIC
NA
22.5
112.3
140.5

w
0.9999
<0.001
<0.001
<0.001

Table 2.4: Parameter estimates as odds ratios with 95% CI for the top logistic regression
model for the 15 monthly ground surveys in Aduoyo-Miyare and Nguka. TWI,
topographic wetness index; LULC, land use-land cover; DS, distance to the nearest
stream; Precip., cumulative 30-day precipitation total; Ag:NonAg, odds ratio of
agricultural to nonagricultural LULC; Type2:Type3, odds ratio of the friable clay/sandy
clay loam soil type to the firm, silty clay/clay soil type.

(Intercept)
TWI
LCLU, Ag:NonAg
DS
Soil, Type2:Type3
Precip.

Odds ratio
0.0378
0.9365
1.3371
0.9980
0.7127
1.0033

95% lower
0.0334
0.9324
1.2096
0.9978
0.6715
1.0029

65

95% upper
0.0426
0.9405
1.4780
0.9981
0.7564
1.0036

p-value
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001

Table 2.5: The top five logistic regression candidate models for the 10 by 10 km area
based on Bayesian information criterion (BIC), in order of increasing difference in BIC
from the top model (ΔBIC) and decreasing BIC weight, w. TWI, topographic wetness
index; LULC, land use-land cover; DS, distance to the nearest stream; Precip.,
cumulative 30-day precipitation total.
Model
TWI + DS + Soil + Precip.
TWI + LULC + DS + Soil + Precip.
TWI + DS + Precip.
TWI + LULC + DS + Precip.
TWI + DS + Soil

BIC
6482.6
6491.8
6492.2
6502.0
6658.0

66

ΔBIC
0
9.3
9.6
19.4
175.4

w
0.9822
0.0096
0.0082
< 0.001
< 0.001

Table 2.6: Parameter estimates as odds ratios with 95% CI for the top logistic regression
model for the 10 by 10 km area. TWI, topographic wetness index; DS, distance to the
nearest stream; Precip., cumulative 6-day precipitation total; Type1:Type3, odds ratio of
the friable clay soil type to the friable clay/sandy clay loam soil type; Type2:Type3, odds
ratio of the friable clay soil type to the firm, silty clay/clay soil type.

(Intercept)
TWI
DS
Soil, Type1:Type2
Soil, Type1:Type3
Precip.

Odds ratio
0.3156
0.9117
0.9973
1.9105
1.4970
0.9700

95% lower
0.2472
0.8988
0.9969
1.4994
1.2024
0.9656

67

95% upper
0.4030
0.9248
0.9977
2.4343
1.8639
0.9745

p-value
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001
< 0.001

Table 2.7: Comparison of models predicting the presence of larval Anopheles gambiae s.l. habitats in the 10 by 10 km area. Two
random forest (RF) models are shown with and without Precip. (the cumulative 14-day precipitation total). The best logistic regression
(LR) model is also show. TWI, topographic wetness index; LULC, land use-land cover; DS, distance to the nearest stream; AUC, area
under the receiver operating curve; PCC, percent correctly classified.
Model
AUC Sensitivity Specificity PCC Kappa
RF: TWI + DS + Soil + LULC + Precip. 0.864
0.806
0.789 0.790 0.216
RF: TWI + DS + Soil + LULC
0.808
0.750
0.725 0.726 0.145
LR: TWI + DS + Soil + Precip.
0.799
0.748
0.709 0.711 0.133

68

Table 2.8: Comparison of models predicting the presence of larval Anopheles gambiae s.l. habitats in Aduoyo-Miyare and Nguka.
Two random forest (RF) models are shown with and without Precip. (the cumulative 14-day precipitation total). The best logistic
regression (LR) model is also shown. TWI, topographic wetness index; LULC, land use-land cover; DS, distance to the nearest
stream; AUC, area under the receiver operating curve; PCC, percent correctly classified.
Method of optimizing
AUC Sensitivity Specificity PCC Kappa
RF: TWI + DS + Soil + LULC + Precip. 0.871
0.820
0.773 0.774 0.102
RF: TWI + DS + Soil + LULC
0.827
0.659
0.936 0.930 0.268
LR: TWI + DS + Soil + Precip.
0.733
0.621
0.704 0.703 0.045

69

Figure 2.1: (A) Map of Kenya with red square indicating location of the study region in western Kenya. (B) Map showing the
boundaries of Asembo and the streams within the community, with black dots representing all households. (C) 10 by 10 km study site
shown as dashed black line with thirty-one 500 by 500 m quadrats surveyed for larval habitats shown as red boxes. (D) Location of the
neighboring villages of Aduoyo-Miyare and Nguka, sites of the 15 monthly ground surveys for larval habitats, shown in red.

70

Figure 2.1 (cont’d)

71

Figure 2.2: Scatter plot of the number of habitats recorded in Aduoyo-Miyare and Nguka each month (n = 15) by the cumulative
2
30-day precipitation for the last day of ground surveys for that month. Line shows prediction of linear regression, R = 0.1931,
p = 0.1012. The red and blue boxes highlight variation in the residual error discussed further in the text.

72

Figure 2.3: Precipitation data from the Kisumu airport (about 40 km east of Asembo) for March 2011 through July 2012. Bars show
daily precipitation total. Solid line shows the cumulative 30-day precipitation total used in the logistic regression model for the 15month dataset, while the dashed line shows the cumulative 21-day total used in the random forest model of the same data. Circles and
squares show the last day of ground surveys for each month in Aduoyo-Miyare and Nguka.

73

Figure 2.4: Precipitation data from the Kisumu airport (about 40 km east of Asembo) for 2 May 2011 to 4 July 2012. Bars show daily
precipitation total. Solid line shows the cumulative 6-day total used in the logistic regression model of the 10 by 10 km dataset, while
the dashed line shows the cumulative 14-day precipitation total used in the random forest model of the same data. Squares and circles
show the days of the ground surveys in the 31 quadrats.

74

Figure 2.5: Probability of larval habitat presence across the 10 by 10 km study site as predicted by (A) the logistic regression model
using TWI, distance to stream and soil type; and (B) the random forest model using TWI, distance to stream, soil type and LULC.

75

REFERENCES

76

REFERENCES

Beier, J. C., P. V. Perkins, F. K. Onyango, T. P. Gargan, C. N. Oster, R. E. Whitmire, D.
K. Koech, and C. R. Roberts. 1990. Characterization of malaria transmission by
Anopheles (Diptera: Culicidae) in western Kenya in preparation for malaria vaccine
trials. Journal of Medical Entomology 27:570–577.
Beven, K. J., and M. J. Kirkby. 1979. A physically based, variable contributing area
model of basin hydrology. Hydrological Sciences Bulletin 24:43–69.
Bisrat, S. A., M. A. White, K. H. Beard, and D. Richard Cutler. 2011. Predicting the
distribution potential of an invasive frog using remotely sensed data in Hawaii.
Diversity and Distributions 18:648–660.
Bogh, C., C. Bogh, S. E. Clarke, M. Jawara, C. J. Thomas, and S. W. Lindsay. 2003.
Localized breeding of the Anopheles gambiae complex (Diptera: Culicidae) along the
River Gambia, West Africa. Bulletin of Entomological Research 93:279–287.
Bogh, C., S. W. Lindsay, S. E. Clarke, A. Dean, M. Jawara, M. Pinder, and C. J. Thomas.
2007. High spatial resolution mapping of malaria transmission risk in the Gambia,
west Africa, using LANDSAT TM satellite imagery. American Journal of Tropical
Medicine and Hygiene 76:875–881.
Breiman, L. 2001. Random forests. Machine learning 45:5–32.
Breiman, L., J. H. Friedman, R. A. Olshen, and C. J. Stone. 1984. Classification and
regression trees (CART). Wadsworth International Group, Belmont, CA, USA.
Charlwood, J. D., and D. Edoh. 1996. Polymerase chain reaction used to describe larval
habitat use by Anopheles gambiae complex (Diptera: Culicidae) in the environs of
Ifakara, Tanzania. Journal of Medical Entomology 33:202–204.
Clark, T. D., B. Greenhouse, D. Njama Meya, B. Nzarubara, C. Maiteki Sebuguzi, S. G.
Staedke, E. Seto, M. R. Kamya, P. J. Rosenthal, and G. Dorsey. 2008. Factors
determining the heterogeneity of malaria incidence in children in Kampala, Uganda.
The Journal of Infectious Diseases 198:393–400.
Clennon, J. A., A. Kamanga, M. Musapa, C. Shiff, and G. E. Glass. 2010. Identifying
malaria vector breeding habitats withremote sensing data and terrain-based
landscapeindices in Zambia. International Journal of Health Geographics 9:58.
Coetzee, M., R. H. Hunt, R. C. Wilkerson, A. della Torre, M. B. Coulibaly, and N. J.
Besansky. 2013. Anopheles coluzzii and Anopheles amharicus, new members of the
Anopheles gambiae complex. Zootaxa 3619:246–274.

77

Gamage-Mendis, A. C., R. Carter, C. Mendis, A. P. De Zoysa, P. R. Herath, and K. N.
Mendis. 1991. Clustering of malaria infections within an endemic population: risk of
malaria associated with the type of housing construction. American Journal of
Tropical Medicine and Hygiene 45:77–85.
Gillies, M. T., and M. Coetzee. 1987. A supplement to the Anophelinae of Africa south
of the Sahara (Afrotropical Region). Pages 1–143. South African Institute for
Medical Research, Johannesburg, South Africa.
Gimnig, J. E., M. Ombok, L. Kamau, and W. A. Hawley. 2001. Characteristics of larval
anopheline (Diptera: Culicidae) habitats in western Kenya. Journal of Medical
Entomology 38:282–288.
Greenwood, B. M. 1989. The microepidemiology of malaria and its importance to
malaria control. Transactions of the Royal Society of Tropical Medicine and Hygiene
83 Suppl:25–29.
Guisan, A., and N. E. Zimmermann. 2000. Predictive habitat distribution models in
ecology. Ecological Modelling 135:147–186.
Hamel, M. J., K. Adazu, D. Obor, M. Sewe, J. M. Vulule, J. M. Williamson, L. Slutsker,
D. R. Feikin, and K. F. Laserson. 2011. A reversal in reductions of child mortality in
western Kenya, 2003-2009. American Journal of Tropical Medicine and Hygiene
85:597–605.
Hernández, J., I. Núñez, A. Bacigalupo, and P. E. Cattan. 2013. Modeling the spatial
distribution of Chagas disease vectors using environmental variables and people s
knowledge. International Journal of Health Geographics 12:29.
Hightower, A. W., M. Ombok, R. Otieno, R. Odhiambo, A. J. Oloo, A. A. Lal, B. L.
Nahlen, and W. A. Hawley. 1998. A geographic information system applied to a
malaria field study in western Kenya. American Journal of Tropical Medicine and
Hygiene 58:266–272.
Hurlbert, S. H. 1984. Pseudoreplication and the design of ecological field experiments.
Ecological Monographs 54:187–211.
Li, L., L. Bian, L. Yakob, G. Zhou, and G. Yan. 2011. Analysing the generality of
spatially predictive mosquito habitat models. Acta Tropica 119:30–37.
Liaw, A., and M. Wiener. 2002. Classification and Regression by randomForest. R news
2:18–22.
Liu, C., P. M. Berry, T. P. Dawson, and R. G. Pearson. 2005. Selecting thresholds of
occurrence in the prediction of species distributions. Ecography 28:385–393.

78

Minakawa, N., C. M. Mutero, J. I. Githure, J. C. Beier, and G. Yan. 1999. Spatial
distribution and habitat characterization of anopheline mosquito larvae in western
Kenya. American Journal of Tropical Medicine and Hygiene 61:1010–1016.
Munga, S., L. Yakob, E. Mushinzimana, G. Zhou, T. Ouna, N. Minakawa, A. K. Githeko,
and G. Yan. 2009. Land use and land cover changes and spatiotemporal dynamics of
anopheline larval habitats during a four-year period in a highland community of
Africa. American Journal of Tropical Medicine and Hygiene 81:1079–1084.
Mushinzimana, E., S. Munga, N. Minakawa, L. Li, C.-C. Feng, L. Bian, U. Kitron, C.
Schmidt, L. Beck, G. Zhou, A. K. Githeko, and G. Yan. 2006. Landscape
determinants and remote sensing of anopheline mosquito larval habitats in the
western Kenya highlands. Malaria Journal 5:13.
Mutuku, F. M., J. A. Alaii, M. N. Bayoh, J. E. Gimnig, J. M. Vulule, E. D. Walker, E.
Kabiru, and W. A. Hawley. 2006. Distribution, description, and local knowledge of
larval habitats of Anopheles gambiae s.l. in a village in western Kenya. American
Journal of Tropical Medicine and Hygiene 74:44–53.
Mutuku, F., M. N. Bayoh, A. Hightower, J. M. Vulule, J. E. Gimnig, J. Mueke, F.
Amimo, and E. D. Walker. 2009. A supervised land cover classification of a western
Kenya lowland endemic for human malaria: associations of land cover with larval
Anopheles habitats. International Journal of Health Geographics 8:19.
Nmor, J. C., T. Sunahara, K. Goto, K. Futami, G. Sonye, P. Akweywa, G. Dida, and N.
Minakawa. 2013. Topographic models for predicting malaria vector breeding
habitats: potential tools for vector control managers. Parasites & Vectors 6:14.
Odiere, M., M. N. Bayoh, J. E. Gimnig, J. M. Vulule, L. Irungu, and E. D. Walker. 2007.
Sampling outdoor, resting Anopheles gambiae and other mosquitoes (Diptera:
Culicidae) in western Kenya with clay pots. Journal of Medical Entomology 44:14–
22.
Ombok, M., K. Adazu, F. Odhiambo, M. N. Bayoh, R. Kiriinya, L. Slutsker, M. J.
Hamel, J. Williamson, A. Hightower, K. F. Laserson, and D. R. Feikin. 2010.
Geospatial distribution and determinants of child mortality in rural western Kenya
2002-2005. Tropical Medicine and International Health 15:423–433.
Phillips-Howard, P. A., B. L. Nahlen, J. A. Alaii, F. O. ter Kuile, J. E. Gimnig, D. J.
Terlouw, S. P. Kachur, A. W. Hightower, A. A. Lal, E. Schoute, A. J. Oloo, and W.
A. Hawley. 2003. The efficacy of permethrin-treated bed nets on child mortality and
morbidity in western Kenya I: development of infrastructure and description of study
site. American Journal of Tropical Medicine and Hygiene 68:3–9.
Ribeiro, J. M. C., F. Seulu, T. Abose, G. Kidane, and A. Teklehaimanot. 1996. Temporal
and spatial distribution of anopheline mosquitos in an Ethiopian village: Implications

79

for malaria control strategies. Bulletin of the World Health Organization 74:299–305.
Sombroek, W. G., H. M. H. Braun, and B. J. A. van der Pouw. 1982. Exploratory soil
map and agro-climatic zone map of Kenya, 1980. Scale 1: 1,000,000. Kenya Soil
Survey, Nairobi, Kenya.
Strauss, B., and R. Biedermann. 2007. Evaluating temporal and spatial generality: How
valid are species–habitat relationship models? Ecological Modelling 204:104–114.
Taylor, K. A., J. K. Koros, J. Nduati, R. S. Copeland, F. H. Collins, and A. D. BrandlingBennett. 1990. Plasmodium falciparum infection rates in Anopheles gambiae, An.
arabiensis, and An. funestus in western Kenya. American Journal of Tropical
Medicine and Hygiene 43:124–129.
Trape, J.-F., E. Lefebvre-Zante, F. Legros, G. Ndiaye, H. Bouganali, P. Druilhe, and G.
Salem. 1992. Vector density gradients and the epidemiology of urban malaria in
Dakar, Senegal. American Journal of Tropical Medicine and Hygiene 47:181–189.
World Health Organization. 2012. World malaria report 2012. Pages 1–276. World
Health Organization, Geneva.
Zhou, G., N. Minakawa, A. K. Githeko, and G. Yan. 2004. Spatial distribution patterns of
malaria vectors and sample size determination in spatially heterogeneous
environments: a case study in the West Kenyan highland. Journal of Medical
Entomology 41:1001–1009.

80

CHAPTER III: Proximity of larval Anopheles gambiae habitats to houses and insecticidetreated bed net use both explain variation in adult An. gambiae indoor resting densities.
Abstract
The spatial heterogeneity of malaria has long been recognized, but the specific
determinants of spatial variation in malaria risk may vary according to the landscape.
Additionally, public health interventions may influence the spatial distribution of malaria.
Houses in western Kenya were sampled for adult Anopheles females at the end of the
rainy season in 2011. The proximity of the sampled houses to larval habitats was assessed
using the predictive larval habitat model described in Chapter III. Variation in householdlevel factors, including use of long-lasting impregnated nets (LLINs), was also assessed.
The number of adult female An. gambiae s.l. per house increased with increasing mean
probability of larval habitat presence within 500 m. This index of habitat proximity was a
better predictor of An. gambiae s.l. female abundance than simply measuring the distance
to the nearest larval habitat. Also, houses in which all residents slept under LLINs the
previous night had fewer adult female An. gambiae s.l. than houses in which only some,
or none, of the residents used LLINs the previous night. There was no difference in the
number of An. gambiae s.l. females collected in houses without LLINs and houses where
only some of the residents slept under LLINs. Our results describing the relationship
between larval habitat locations and the number of adult malaria vectors in houses differ
from previous findings, suggesting that this relationship may vary depending on the
configuration of larval habitats within a given landscape. The difference between
individual LLIN use and LLIN ownership at the household-level highlights an important
consideration for malaria control program managers, as well as those modeling the
potential effects of LLINs as an intervention.

81

Introduction
The spatial distribution of malaria prevalence within communities is
heterogeneous in a wide range of ecological and epidemiological settings, especially
when community-wide transmission intensity is low to moderate (Carter et al. 2000).
While the recognition of spatial variation in malaria risk predates even the etiology of the
disease (Boyd 1949), the specific determinants of malaria’s spatial distribution vary
locally. Even within communities, socioeconomic and immunological differences affect
people’s risk of malaria prevalence (Greenwood 1989, Gamage-Mendis et al. 1991,
Mackinnon et al. 2000). Additionally, the ecology of the local malaria vectors determines
the spatial distribution of malaria, often through the relative locations of the vectors’
larval habitats and people’s homes (Trape et al. 1992, Ribeiro et al. 1996, Thompson et
al. 1997, Ghebreyesus et al. 1999). While closer proximity to the larval habitats of
malaria vectors may lead to increased (Clark et al. 2008) or decreased (Clarke et al. 2002)
incidence of clinical malaria, depending on the epidemiological context, it is nevertheless
clear that the spatial distribution of malaria vectors has important implications for malaria
transmission.
The spatial distribution of adult malaria vectors is determined, in part, by the
landscape through its effects on hydrology and the locations of aquatic habitat for
Anopheles larvae (Smith et al. 2013). In landscapes where larval habitats are restricted to
a linear feature such as along a river or swamp, adult Anopheles densities per house
increase with decreasing distance to the nearest larval habitat (Trape et al. 1992, van der
Hoek et al. 2003, Bogh et al. 2007, Zhou et al. 2007). A similar relationship exists for
some Anopheles species in landscapes where relatively few larval habitats are dispersed

82

among people’s homes (Minakawa et al. 2002, Walker et al. 2013). If larval Anopheles
habitats are considered as foci of vector production or malaria transmission (Carter et al.
2000), this relationship is intuitive. However, in regions where numerous larval habitats
are spread across the landscape, the relationship between larval habitat locations and the
spatial distribution of adult Anopheles is less clear, possibly because the foci of vector
production overlap.
Public health interventions aimed at reducing malaria transmission also influence
the spatial distribution of malaria. As one of the primary vector control measures used
since the 1990s, insecticide-treated bed nets (ITNs) have a significant impact on malaria
vectors, reducing their population sizes on a broad scale (Gimnig et al. 2003a, 2003b). At
the household scale, variation in ITN ownership and use may lead to variation in the
number of adult Anopheles found indoors. The presence of ITNs in houses may reduce
the rate of entry (Mathenge et al. 2001), or increase the rate of exiting (Mathenge et al.
2001, Malima et al. 2008), by malaria vectors. Furthermore, despite the substantial
increase in the number of households owning ITNs in many malarious countries over the
last decade (World Health Organization 2012), variation in ITN use by individuals within
households leaves some people unprotected (Korenromp et al. 2003, Eisele et al. 2009),
even in countries implementing universal coverage with long-lasting impregnated nets
(LLINs) (Macintyre et al. 2011). Quantifying the contribution of these interventions to
the heterogeneity of malaria, relative to that of landscape factors such as the distribution
of larval habitats, is critical for the effective implementation and evaluation of
interventions.

83

The primary objective of this study was to quantify the contribution of larval
habitat proximity to variation in the number of Anopheles gambiae s.l. resting in houses,
within a landscape where larval habitats are numerous yet heterogeneously distributed.
As an observational study, we also accounted for household-level variables that
potentially influenced the number of An. gambiae s.l. in houses. In particular, LLIN use
was of interest because of the relatively high proportion of households owning LLINs in
the study site. We included additional variables in our analysis based on previous studies’
findings of variation in the number of Anopheles found indoors associated with the
following factors (Table 3.1): presence of cattle in the compound (Minakawa et al. 2002);
whether the inhabitants had cooked in the house (Bockarie et al. 1994, Hiscox et al.
2013); different wall types (Zhou et al. 2007, Kirby et al. 2008); different roof types
(Zhou et al. 2007); and the number of residents in the house (Kirby et al. 2008, Walker et
al. 2013).
Methods
Study site. The Asembo region of Rarieda District in western Kenya (Figure
2

3.1A) is a rural community of about 60,000 people covering about 200 km . Most of the
residents are subsistence farmers, and the landscape is largely dominated by small-scale
agriculture. Small plots of land generally surround family-based groups of houses, or
compounds, further arranged into villages. While the compounds are highly dispersed
within villages, the boundaries between the 79 villages in Asembo are often discernable
only by residents (Phillips-Howard et al. 2003). This is apparent in Figure 3.1B, which
shows the roughly 10,500 compounds georeferenced within Asembo as of 2009
(Hightower et al. 1998, Ombok et al. 2010). Asembo sits in the lowlands along the shores

84

of Lake Victoria, with elevations ranging from 1,100 m to 1,400 m above sea level and
low topographic relief. Rainfall is seasonally bimodal but may occur year round.
Similar to rainfall patterns, malaria transmission occurs year round, with seasonal
peaks in May-July and October-November. Parasitemia rates in children under 5 were
around 50% in 2009 (Hamel et al. 2011), although entomological inoculation rates have
been estimated at less than 15 infectious bites per person per year since 2003 (MN
Bayoh, unpublished data). The predominant species of malaria is Plasmodium
falciparum. Two of the primary malaria vectors in the region are An. gambiae s.s. and
Anopheles arabiensis (Beier et al. 1990, Taylor et al. 1990, Bayoh et al. 2010), the only
two members of the An. gambiae s.l. species complex found here.
Larval habitats. Larval An. gambiae s.l. habitats are numerous and widespread in
Asembo, yet heterogeneously distributed. The locations of 1,673 larval An. gambiae s.l.
habitats were recorded in ground surveys conducted in thirty-one 500 by 500 m quadrats
from 17 May to 4 July 2011 (Chapter III). The surveyed quadrats were randomly
selected, after spatial stratification, from a 10 by 10 km area within Asembo. Using a
topographic wetness index, land use-land cover, soil type, and distance to stream, we
applied the random forest method (Breiman 2001) to produce a landscape model of larval
habitats, predicting the probability of larval habitat presence at a 20 m resolution across
the 10 by 10 km area (Figure 3.1C).
Adult mosquito sampling. Houses were sampled for indoor-resting adult
Anopheles between 16 May and 24 June 2011 using the pyrethrum spray catch method
(PSC) (Gimnig et al. 2003a, Silver 2008). This period was chosen to coincide with
seasonal peaks of Anopheles populations, which occur after the March-May rainy season

85

(Beier et al. 1990, Taylor et al. 1990, Odiere et al. 2007). All collected Anopheles were
morphologically identified to species according to Gillies and Coetzee (1987). Mosquito
voucher specimens were deposited in the Albert J. Cook Arthropod Research Collection
at Michigan State University (Table 3.2). To avoid edge effects in the assessment of
larval habitat proximity on the number of adult Anopheles collected in houses, no houses
were sampled within 1 km of the larval habitat model border (Figure 3.1D). Houses
within the 8 by 8 km area were selected for sampling using two-stage cluster sampling.
First, the 8 by 8 km area was divided into 1 by 1 km quadrats. Forty of these quadrats
were randomly selected, and one house in each quadrat was randomly selected as the
starting point for cluster sampling. The order in which the clusters were sampled was also
randomly selected. Nine to nineteen nearest neighbor houses were sampled in each
cluster, as time permitted, resulting in 40 clusters of 10 to 20 houses being sampled over
the six-week period.
Seven household-level variables that potentially influenced the number of
Anopheles in the sampled houses (Table 3.1) were assessed at the time of mosquito
sampling through a visual inspection of the house and via a standardized survey. While
specific numbers of LLIN users were counted in the surveys, houses were later
categorized into three groups of LLIN use. The first group was houses where everyone
who slept in the house the previous night used an LLIN. The second group of houses
included those where some residents had slept under an LLIN the previous night while
other residents had not. The third group of houses was those where no one had slept
under an LLIN the previous night.

86

The locations of the sampled houses were recorded using GPS units to quantify
the proximity of the houses to larval An. gambiae s.l. habitats based on the model
described above. The best method for measuring the proximity of the houses to larval
habitats was not clear a priori. In contrast to previous studies using only the distance to
the nearest larval habitat, larval habitats in Asembo are abundant and may surround a
house from multiple directions. Thus, the number of adult mosquitoes found in the house
may depend on the density of larval habitats surrounding the house. Accordingly, several
indices of larval habitat proximity were calculated for each house sampled during PSC
(Table 3.3), and each index was evaluated as a univariate candidate model during model
selection (below).
To calculate the distance to the nearest larval habitat and the number of larval
habitats surrounding a house, the probability output from the predictive habitat model
was converted to a binary surface of presence and absence using a threshold value
(Chapter III). Alternatively, variation in the probability of habitat presence within
locations above the threshold could contribute additional information about how the
number of adult An. gambiae s.l. in a house is influenced by the surrounding landscape.
Therefore, we assessed the probability values surrounding each house by calculating their
maximum, minimum, and mean. Finally, the scale at which to measure these statistics
was not obvious, though the dispersal range of An. gambiae s.l. is thought to be less than
1 km, generally (Carter et al. 2000). Thus, the statistics were calculated across a range of
distances from 50 to 1,000 m (Table 3.3).
Statistical analysis. We used multiple regression with generalized linear models
to quantify the relative contribution of the measured variables to the number of An.

87

gambiae s.l. in houses. These generalized linear models used a log link function and
negative binomial error distribution to account for the over-dispersion and high number
of zeros in the observed count data. To control for potential dependence among houses
within the clusters of sampled houses described above, we used generalized estimating
equations (GEE) within the generalized linear models. Additionally, we compared
models with and without GEE to determine the potential effect of sampling the houses in
clusters.
We evaluated a series of candidate models with the measured variables, using an
information-theoretical approach for selecting models that best explained the observed
variation in the number of An. gambiae s.l. in houses (Burnham and Anderson 2002). We
considered eight variables (seven listed in Table 3.1 plus habitat proximity) to be
potentially useful as predictors of adult female An. gambiae s.l. in houses, and we fitted
models with all possible subsets of these parameters, calculating Bayesian Information
Criterion (BIC) and model weights (wi) for the candidate models. BIC quantifies the fit of
the model to the observed data based on maximum likelihood while accounting for a bias
of increased likelihood with an increasing number of parameters in the model. While
models the lower BIC values were considered the best-fitting models to the data, model
weights were used to reflect the degree of uncertainty about differences among
competing models (Burnham and Anderson 2004). Habitat proximity indices were not
combined into the same candidate model. Rather, univariate models with each of the
habitat proximity indices were compared first. For the purposes of comparing the habitat
proximity indices to each other, we compared the model weights of those univariate
models. To quantify the contribution of larval habitat proximity to adult An. gambiae s.l.

88

abundances relative to the seven other household-level variables, we used the habitat
proximity index from the univariate model with the lowest BIC.

Results
A total of 227 female An. gambiae s.l. were collected in the 525 houses sampled
(mean = 0.432, range: 0-8). Houses varied considerably in their proximity to larval An.
gambiae s.l. habitats. Distance from the sampled houses to the nearest predicted larval
habitat ranged from 1 to 539 m (mean = 48 m). The number of predicted larval habitats
within 500 m of the sampled houses ranged from 0 to 977 (mean = 462), while the mean
probability of larval habitat presence within the same area ranged from 0.0 to 11.3%
(mean = 3.5%). The larval habitat proximity index that best fit the observed variation in
female An. gambiae s.l. collected per house was the mean probability of larval habitat
presence within 500 m of a house (Table 3.4). However, there was support for
considering three additional indices of larval habitat proximity to houses, though all three
were highly and positively correlated with the mean probability of larval habitat presence
within 500 m of a house. The number of habitats within 500 m and 300 m, as well as the
mean probability of larval habitat presence within 300 m of a house, combined to make
up the top 90% of model weights for the univariate larval habitat proximity candidate
models (Table 3.4).
Overall, the model that best fit the observed variation in female An. gambiae s.l.
collected per house included four variables: larval habitat proximity (mean probability
within 500 m), LLIN use, whether residents cooked in the house, and whether cattle were
present in the compound. Three additional models could not be ruled out based on the top

89

90% of model weights, but all three models included the four variables from the leading
model (Table 3.5). Furthermore, in each of those models, the 95% confidence intervals
for the regression estimates of the fifth variable overlapped with an estimate of no effect.
In the leading model, after accounting for LLIN use, cooking in the house, and cattle in
the compound, the number of adult female An. gambiae s.l. per house increased with
increasing mean probability of larval habitat presence within 500 m (Figure 3.2). Houses
in which all residents slept under an LLIN the previous night had fewer adult female An.
gambiae s.l. than the other two groups of houses, in which only some, or none, of the
residents used an LLIN the previous night (Table 3.6). We also collected fewer An.
gambiae s.l. adult females from houses in which residents had cooked the previous night,
and from houses without cattle in the compound (Table 3.6). Controlling for the effect of
sampling in clusters resulted in a change of the standard errors for model coefficients of
less than one percent.
Discussion
We have shown that the proximity of houses to larval An. gambiae s.l. habitats
contributes significantly to the spatial distribution of adult An. gambiae s.l. in a region
where larval habitats are numerous and spread widely across the landscape. This agrees
broadly with findings in other landscapes (Trape et al. 1992, Bogh et al. 2007, Zhou et al.
2007), suggesting that An. gambiae s.l. females preferentially disperse relatively short
distances between aquatic larval habitats, which are also oviposition sites, and adult-stage
blood feeding sites. Presumably, this relates to the energetic costs of flight (Nayar and
Van Handel 1971, Foster 1995).

90

We found variation in the predictive ability of different indices of larval habitat
proximity. Indices related to the density of larval habitats within 300 to 500 m were better
predictor variables for our data than simply measuring the distance to the nearest larval
habitat (Table 3.4). This is likely due to a combination of the relatively large number of
habitats present during the rainy season in this region and the distribution of the habitats
around people’s houses in this landscape. For houses less than about 100 m from the
nearest larval habitat, we found considerable variation in the mean probability of larval
habitat presence within 500 m (Figure 3.3). In contrast, in landscapes where the distance
to the nearest larval habitat and the density of larval habitats surrounding a house are
more closely related, their usefulness for explaining the number of adult Anopheles in
houses will be similar. Thus, although larval habitats are related to the distribution of
adult Anopheles in general, the specific predictive relationship between the two may vary
depending on the configuration of larval habitats within a given landscape.
Variation in LLIN use also contributed to variation among houses in the number
of An. gambiae s.l. in this study. In the houses we sampled, only households that did not
own LLINs made up the group of houses in which none of the residents used LLINs. As
expected, these houses had more An. gambiae s.l. than houses in which all of the
residents used LLINs the previous night. Notably, houses in which only some of the
residents used LLINs the previous night also had more An. gambiae s.l. than houses in
which all of the residents used LLINs the previous night. In fact, there was no difference
between the number of An. gambiae s.l. females collected in houses without LLINs and
houses where only some of the residents slept under LLINs. A likely explanation is that
people not sleeping under LLINs, even in houses with LLINs, represent potential hosts

91

for An. gambiae s.l. females. This is an important consideration for malaria control
program managers, as well as those modeling the potential effects of LLINs as an
intervention.
In addition to larval habitat proximity and LLIN use, the presence of cattle in the
compound also contributed to variation in the number of An. gambiae s.l. among sampled
houses. Anopheles arabiensis females are relatively opportunistic blood feeders, readily
utilizing cattle as hosts in addition to humans (Takken and Verhulst 2013). Anopheles
gambiae s.s. females are highly anthropophilic and respond preferentially to human odors
at short distances (Pates et al. 2001). However, An. gambiae s.s. females respond to
carbon dioxide (CO2) alone (i.e. in the absence of other host cues) at distances up to 50 m
(Lorenz et al. 2013), potentially using the CO2 emitted from cattle to orient toward their
eventual human host when cattle are kept near people’s homes. Together, these responses
to cattle potentially account for the higher number of An. gambiae s.l. in houses with
cattle present in the compound.
Finally, cooking in the house the previous night was associated with collecting
fewer An. gambiae s.l. females. Wood and charcoal are the predominant fuels for cooking
in Asembo, and the smoke from firewood may reduce the number of An. gambiae s.l.
indoors (Bockarie et al. 1994). However, Biran and colleagues (2007) found little
evidence for a protective effect of smoke from domestic fires against mosquitoes in a
recent systematic review. They note that three observational studies from the early 20

th

century in South and East Africa (De Meillon 1930, Symes 1930, Gibbins 1933) found no
difference in the numbers of Anopheles between homes with and without smoke from

92

domestic fires. Furthermore, the increased risk of respiratory diseases linked to smoke
from biofuel sources (Po et al. 2011) likely outweighs any potential decrease in the
number of Anopheles indoors.
This study was aimed at quantifying the determinants of variation in An. gambiae
s.l. females in houses during the yearly peak in population size. Clearly, Anopheles
populations in the region vary seasonally (Beier et al. 1990, Taylor et al. 1990, Gimnig et
al. 2003a, Odiere et al. 2007), and the relationships found here may change in magnitude
or even direction with the seasons. Additionally, variation among years in precipitation
patterns could influence the relationships found here. The advantage to this crosssectional approach was our ability to cover a relatively large area, capturing greater
variation in the landscape and potentially making our results more generalizable.

93

APPENDIX

94

Chapter III Tables and Figures

95

Table 3.1: Household-level variables recorded for houses where sampling was done for
female Anopheles. For factors, all levels of the factor are listed, with the number of
houses observed at each level shown in parentheses. LLIN, long-lasting impregnated net.
Variable
Number of total residents
LLIN use
Wall type
Roof type
Cattle present
Goats present
Cooked in house

Summary of observed values
mean = 3
range: 1 - 9
all (314)
some (81)
none (130)
mud (280) plastered (109)
other (136)
thatch (161)
iron (364)
no (166)
yes (359)
no (209)
yes (316)
no (342)
yes (183)

96

Table 3.2: Record of deposition of voucher specimens for Chapter III. The specimens
listed below have been deposited in the Albert J. Cook Arthropod Research Collection,
Michigan State University (MSU) museum as samples of those species or other taxa,
which were used in this research. Voucher recognition labels bearing the voucher number
2013-14 have been attached.
Species
Anopheles gambiae s.l.
Anopheles gambiae s.l.

Life Stage Quantity Preservation
Adult female
5
Pinned
Adult male
5
Pinned

97

Table 3.3: Indices used to quantify the proximity of larval habitats to houses where sampling was done for female Anopheles.
NA, not applicable.
*
larval habitats predicted using a random forest model, converting probabilities to presence and absence with a threshold
†

probability of larval habitat presence predicted using a random forest model

Larval habitat proximity index
Distance to nearest habitat

Measured at x =

*

*

NA

Number of habitats within x meters

50, 100, 300, 500, 1000 m

Minimum probability of a habitat within x meters

50, 100, 300, 500, 1000 m

Maximum probability of a habitat within x meters

50, 100, 300, 500, 1000 m

Mean probability of a habitat within x meters

50, 100, 300, 500, 1000 m

†

†

†

98

Table 3.4: Leading candidate models (sum of wi > 0.90) for comparison of larval habitat
indices contribution to the number of An. gambiae s.l. in houses. Results for the distance
to the nearest habitat are also shown for comparison. BIC, Bayesian Information Criteria;
ΔBIC, difference in BIC from the lowest BIC; wi, model weight; r, correlation of habitat
index with mean probability of larval habitat within 500 m; P(habitat), probability of
habitat presence. Remaining candidate models had ΔBIC > 7.0 and wi < 0.020.
Habitat index
Mean P(habitat) within 500 m
Number of habitats within 500 m
Number of habitats within 300 m
Mean P(habitat) within 300 m
Distance to nearest habitat

BIC
897.0
899.8
900.6
900.9
910.6

99

ΔBIC
0.0
2.8
3.7
4.0
13.6

wi
0.616
0.152
0.098
0.084
< 0.001

r
1.00
0.86
0.69
0.77
-0.22

Table 3.5: Leading candidate models (sum of wi > 0.90) for quantifying the variation in number of An. gambiae s.l. in houses using
household-level variables. BIC, Bayesian Information Criteria; ΔBIC, difference in BIC from the lowest BIC; wi, model weight; df,
degrees of freedom; LLIN, long-lasting impregnated nets. Remaining candidate models had ΔBIC > 6.5 and wi < 0.025.
Model
Cattle + Cooked + LLIN use + Larval habitat prox.
Cattle + Cooked + LLIN use + Larval habitat prox. + Goats
Cattle + Cooked + LLIN use + Larval habitat prox. + Number of people
Cattle + Cooked + LLIN use + Larval habitat prox. + Roof type

100

df
7
8
8
8

BIC ΔBIC
862.0
0.0
865.7
3.7
868.2
6.2
868.2
6.2

wi
0.745
0.116
0.034
0.033

Table 3.6: Relative rate ratios of the number of An. gambiae s.l. per house, according to
the variables in the leading negative binomial regression model. Note that all models
were analyzed with larval habitat indices as continuous variables; mean probability of
habitat, P(habitat), within 500 m is presented here as a factor for a more intuitive
interpretation. LLIN use, long-lasting impregnated nets.
Relative Rate Ratio
Variable
(95% Confidence Interval)
Mean P(habitat) within 500 m
0.000 - 0.020
1.00
0.021 - 0.040
1.68 (1.01, 2.79)
0.041 - 0.070
2.01 (1.24, 3.26)
0.071 - 0.113
3.99 (2.10, 7.59)
LLIN use
all
1.00
some
2.13 (1.29, 3.51)
none
2.46 (1.63, 3.72)
Cooked in house
no
1.00
yes
0.32 (0.20, 0.52)
Cattle in compound
no
1.00
yes
2.44 (1.52, 3.92)

101

Figure 3.1: (A) Map of Kenya with red square indicating location of the study region in western Kenya. (B) Map showing the
boundaries of Asembo and the streams within the community, with black dots representing all compounds in the community. (C)
Probability of larval habitat presence as predicted by a random forest model across a 10 by 10 km area in Asembo. (D) 8 by 8 km
border for pyrethrum spray catch sampling shown as dashed black line, with red dots showing the 525 houses sampled for adult
Anopheles and solid red line representing the extent of the larval habitat model.

102

Figure 3.1 (cont’d)

103

Figure 3.2: Estimated number of adult female An. gambiae s.l. per house according to the
leading model. Shading bounded by dashed lines indicates 95% confidence intervals.
Levels of LLIN use refer to whether all, some, or none of the residents used LLINs in the
house the previous night. Estimates are shown for houses where cattle are present in the
compound (i.e. cattle = yes), but residents did not cook in the house (cooked = no).

104

Figure 3.3: Scatterplot of the mean probability of larval habitat presence within 500 m of
a house and the distance to the nearest larval habitat to the house (r = 0.22) for the 525
houses sampled in this study.

105

REFERENCES

106

REFERENCES

Bayoh, M. N., D. K. Mathias, M. R. Odiere, F. M. Mutuku, L. Kamau, J. E. Gimnig, J.
M. Vulule, W. A. Hawley, M. J. Hamel, and E. D. Walker. 2010. Anopheles
gambiae: historical population decline associated with regional distribution of
insecticide-treated bed nets in western Nyanza Province, Kenya. Malaria Journal
9:62.
Beier, J. C., P. V. Perkins, F. K. Onyango, T. P. Gargan, C. N. Oster, R. E. Whitmire, D.
K. Koech, and C. R. Roberts. 1990. Characterization of malaria transmission by
Anopheles (Diptera: Culicidae) in western Kenya in preparation for malaria vaccine
trials. Journal of Medical Entomology 27:570–577.
Biran, A., L. Smith, J. Lines, J. Ensink, and M. Cameron. 2007. Smoke and malaria: are
interventions to reduce exposure to indoor air pollution likely to increase exposure to
mosquitoes? Transactions of the Royal Society of Tropical Medicine and Hygiene
101:1065–1071.
Bockarie, M. J. M., M. W. M. Service, G. G. Barnish, W. W. Momoh, and F. F. Salia.
1994. The effect of woodsmoke on the feeding and resting behaviour of Anopheles
gambiae s.s. Acta Tropica 57:337–340.
Bogh, C., S. W. Lindsay, S. E. Clarke, A. Dean, M. Jawara, M. Pinder, and C. J. Thomas.
2007. High spatial resolution mapping of malaria transmission risk in the Gambia,
west Africa, using LANDSAT TM satellite imagery. American Journal of Tropical
Medicine and Hygiene 76:875–881.
Boyd, M. F. 1949. Historical overview. Pages 3–25 in M. F. Boyd, editor. Malariology:
A comprehensive survey of all aspects of this group of diseases from a global
standpoint. W.B. Saunders Co., Philadelphia.
Breiman, L. 2001. Random forests. Machine learning 45:5–32.
Burnham, K. P., and D. R. Anderson. 2004. Multimodel inference understanding AIC and
BIC in model selection. Sociological Methods & Research 33:261–304.
Burnham, K., and D. Anderson. 2002. Model selection and multimodel inference: a
practical information-theoretical approach, 2nd edition. Springer Verlag, New York.
Carter, R. R., K. N. K. Mendis, and D. D. Roberts. 2000. Spatial targeting of
interventions against malaria. Bulletin of the World Health Organization 78:1401–
1411.
Clark, T. D., B. Greenhouse, D. Njama Meya, B. Nzarubara, C. Maiteki Sebuguzi, S. G.
Staedke, E. Seto, M. R. Kamya, P. J. Rosenthal, and G. Dorsey. 2008. Factors

107

determining the heterogeneity of malaria incidence in children in Kampala, Uganda.
The Journal of Infectious Diseases 198:393–400.
Clarke, S. E., C. Bogh, R. C. Brown, G. E. L. Walraven, C. J. Thomas, and S. W.
Lindsay. 2002. Risk of malaria attacks in Gambian children is greater away from
malaria vector breeding sites. Transactions of the Royal Society of Tropical Medicine
and Hygiene 96:499–506.
De Meillon, B. 1930. Anopheles funestus (Giles) in Smoke-Filled Native Huts. Journal of
the Medical Association of South Africa 4.
Eisele, T. P., J. Keating, M. Littrell, D. Larsen, and K. Macintyre. 2009. Assessment of
insecticide-treated bednet use among children and pregnant women across 15
countries using standardized national surveys. American Journal of Tropical
Medicine and Hygiene 80:209–214.
Foster, W. A. 1995. Mosquito sugar feeding and reproductive energetics. Annual review
of entomology 40:443–474.
Gamage-Mendis, A. C., R. Carter, C. Mendis, A. P. De Zoysa, P. R. Herath, and K. N.
Mendis. 1991. Clustering of malaria infections within an endemic population: risk of
malaria associated with the type of housing construction. American Journal of
Tropical Medicine and Hygiene 45:77–85.
Ghebreyesus, T. A., M. Haile, K. H. Witten, A. Getachew, A. M. Yohannes, M.
Yohannes, H. D. Teklehaimanot, S. W. Lindsay, and P. Byass. 1999. Incidence of
malaria among children living near dams in northern Ethiopia: community based
incidence survey. BMJ 319:663–666.
Gibbins, E. G. 1933. The domestic Anopheles mosquitoes of Uganda. Ann. trop. Med.
Parasit 27:15–25.
Gillies, M. T., and M. Coetzee. 1987. A supplement to the Anophelinae of Africa south
of the Sahara (Afrotropical Region). Pages 1–143. South African Institute for
Medical Research, Johannesburg, South Africa.
Gimnig, J. E., J. M. Vulule, T. Q. Lo, L. Kamau, M. S. Kolczak, P. A. Phillips-Howard,
E. M. Mathenge, F. O. ter Kuile, B. L. Nahlen, A. W. Hightower, and W. A. Hawley.
2003a. Impact of permethrin-treated bed nets on entomologic indices in an area of
intense year-round malaria transmission. American Journal of Tropical Medicine and
Hygiene 68:16–22.
Gimnig, J. E., M. S. Kolczak, A. W. Hightower, J. M. Vulule, E. Schoute, L. Kamau, P.
A. Phillips-Howard, F. O. ter Kuile, B. L. Nahlen, and W. A. Hawley. 2003b. Effect
of permethrin-treated bed nets on the spatial distribution of malaria vectors in
western Kenya. American Journal of Tropical Medicine and Hygiene 68:115–120.

108

Greenwood, B. M. 1989. The microepidemiology of malaria and its importance to
malaria control. Transactions of the Royal Society of Tropical Medicine and Hygiene
83 Suppl:25–29.
Hamel, M. J., K. Adazu, D. Obor, M. Sewe, J. M. Vulule, J. M. Williamson, L. Slutsker,
D. R. Feikin, and K. F. Laserson. 2011. A reversal in reductions of child mortality in
western Kenya, 2003-2009. American Journal of Tropical Medicine and Hygiene
85:597–605.
Hightower, A. W., M. Ombok, R. Otieno, R. Odhiambo, A. J. Oloo, A. A. Lal, B. L.
Nahlen, and W. A. Hawley. 1998. A geographic information system applied to a
malaria field study in western Kenya. American Journal of Tropical Medicine and
Hygiene 58:266–272.
Hiscox, A., P. Khammanithong, S. Kaul, P. Sananikhom, R. Luthi, N. Hill, P. T. Brey,
and S. W. Lindsay. 2013. Risk Factors for Mosquito House Entry in the Lao PDR.
PLoS ONE 8:e62769.
Kirby, M. J., C. Green, P. M. Milligan, C. Sismanidis, M. Jasseh, D. J. Conway, and S.
W. Lindsay. 2008. Risk factors for house-entry by malaria vectors in a rural town and
satellite villages in The Gambia. Malaria Journal 7:2.
Korenromp, E. L., J. Miller, R. E. Cibulskis, M. K. Cham, D. Alnwick, and C. Dye. 2003.
Monitoring mosquito net coverage for malaria control in Africa: possession vs. use
by children under 5 years. Tropical Medicine and International Health 8:693–703.
Lorenz, L. M., A. Keane, J. D. Moore, C. J. Munk, L. Seeholzer, A. Mseka, E. Simfukwe,
J. Ligamba, E. L. Turner, L. R. Biswaro, F. O. Okumu, G. F. Killeen, W. R.
Mukabana, and S. J. Moore. 2013. Taxis assays measure directional movement of
mosquitoes to olfactory cues. Parasites & Vectors 6:131.
Macintyre, K., M. Littrell, J. KEATING, B. Hamainza, J. Miller, and T. P. Eisele. 2011.
Determinants of hanging and use of ITNs in the context of near universal coverage in
Zambia. Health Policy and Planning.
Mackinnon, M. J., D. M. Gunawardena, J. Rajakaruna, S. Weerasingha, K. N. Mendis,
and R. Carter. 2000. Quantifying genetic and nongenetic contributions to malarial
infection in a Sri Lankan population. Proceedings of the National Academy of
Sciences of the United States of America 97:12661–12666.
Malima, R. C., S. M. Magesa, P. K. Tungu, V. Mwingira, F. S. Magogo, W. Sudi, F. W.
Mosha, C. F. Curtis, C. Maxwell, and M. Rowland. 2008. An experimental hut
evaluation of Olyset® nets against anopheline mosquitoes after seven years use in
Tanzanian villages. Malaria Journal 7:38.

109

Mathenge, E. M., J. E. Gimnig, M. Kolczak, M. Ombok, L. W. Irungu, and W. A.
Hawley. 2001. Effect of permethrin-impregnated nets on exiting behavior, blood
feeding success, and time of feeding of malaria mosquitoes (Diptera: Culicidae) in
western Kenya. Journal of Medical Entomology 38:531–536.
Minakawa, N., P. Seda, and G. Yan. 2002. Influence of host and larval habitat
distribution on the abundance of African malaria vectors in western Kenya. American
Journal of Tropical Medicine and Hygiene 67:32–38.
Nayar, J. K., and E. Van Handel. 1971. The fuel for sustained mosquito flight. Journal of
Insect Physiology 17:471–481.
Odiere, M., M. N. Bayoh, J. E. Gimnig, J. M. Vulule, L. Irungu, and E. D. Walker. 2007.
Sampling outdoor, resting Anopheles gambiae and other mosquitoes (Diptera:
Culicidae) in western Kenya with clay pots. Journal of Medical Entomology 44:14–
22.
Ombok, M., K. Adazu, F. Odhiambo, M. N. Bayoh, R. Kiriinya, L. Slutsker, M. J.
Hamel, J. Williamson, A. Hightower, K. F. Laserson, and D. R. Feikin. 2010.
Geospatial distribution and determinants of child mortality in rural western Kenya
2002-2005. Tropical Medicine and International Health 15:423–433.
Pates, H. V., W. Takken, K. Stuke, and C. F. Curtis. 2001. Differential behaviour of
Anopheles gambiae sensu stricto (Diptera: Culicidae) to human and cow odours in
the laboratory. Bulletin of Entomological Research 91:289–296.
Phillips-Howard, P. A., B. L. Nahlen, J. A. Alaii, F. O. ter Kuile, J. E. Gimnig, D. J.
Terlouw, S. P. Kachur, A. W. Hightower, A. A. Lal, E. Schoute, A. J. Oloo, and W.
A. Hawley. 2003. The efficacy of permethrin-treated bed nets on child mortality and
morbidity in western Kenya I: development of infrastructure and description of study
site. American Journal of Tropical Medicine and Hygiene 68:3–9.
Po, J. Y. T., J. M. FitzGerald, and C. Carlsten. 2011. Respiratory disease associated with
solid biomass fuel exposure in rural women and children: systematic review and
meta-analysis. Thorax 66:232–239.
Ribeiro, J. M. C., F. Seulu, T. Abose, G. Kidane, and A. Teklehaimanot. 1996. Temporal
and spatial distribution of anopheline mosquitos in an Ethiopian village: Implications
for malaria control strategies. Bulletin of the World Health Organization 74:299–305.
Silver, J. B. 2008. Mosquito ecology: field sampling methodsThird. Springer, New York.
Smith, M. W., M. G. Macklin, and C. J. Thomas. 2013. Earth-Science Reviews. Earth
Science Reviews 116:109–127.
Symes, C. B. 1930. Anophelines in Kenya. Kenya and East African Medical Journal 7:2–

110

11.
Takken, W., and N. O. Verhulst. 2013. Host Preferences of Blood-Feeding Mosquitoes.
Annual review of entomology 58:433–453.
Taylor, K. A., J. K. Koros, J. Nduati, R. S. Copeland, F. H. Collins, and A. D. BrandlingBennett. 1990. Plasmodium falciparum infection rates in Anopheles gambiae, An.
arabiensis, and An. funestus in western Kenya. American Journal of Tropical
Medicine and Hygiene 43:124–129.
Thompson, R. R., K. K. Begtrup, N. N. Cuamba, M. M. Dgedge, C. C. Mendis, A. A.
Gamage-Mendis, S. M. S. Enosse, J. J. Barreto, R. E. R. Sinden, and B. B. Hogh.
1997. The Matola malaria project: a temporal and spatial study of malaria
transmission and disease in a suburban area of Maputo, Mozambique. American
Journal of Tropical Medicine and Hygiene 57:550–559.
Trape, J.-F., E. Lefebvre-Zante, F. Legros, G. Ndiaye, H. Bouganali, P. Druilhe, and G.
Salem. 1992. Vector density gradients and the epidemiology of urban malaria in
Dakar, Senegal. American Journal of Tropical Medicine and Hygiene 47:181–189.
van der Hoek, W., F. Konradsen, P. H. Amerasinghe, D. Perera, M. K. Piyaratne, and F.
P. Amerasinghe. 2003. Towards a risk map of malaria for Sri Lanka: the importance
of house location relative to vector breeding sites. International Journal of
Epidemiology 32:280–285.
Walker, M., P. Winskill, M.-G. Basáñez, J. M. Mwangangi, C. M. Mbogo, J. C. Beier,
and J. T. Midega. 2013. Temporal and micro-spatial heterogeneity in the distribution
of Anopheles vectors of malaria along the Kenyan coast. Parasites & Vectors 6:311.
World Health Organization. 2012. World malaria report 2012. Pages 1–276. World
Health Organization, Geneva.
Zhou, G., S. Munga, N. Minakawa, A. K. Githeko, and G. Yan. 2007. Spatial relationship
between adult malaria vector abundance and environmental factors in western Kenya
highlands. American Journal of Tropical Medicine and Hygiene 77:29–35.

111

CHAPTER IV. Reemergence of Anopheles funestus as a vector of
Plasmodium falciparum in western Kenya after long-term implementation of
insecticide-treated bed nets
Abstract
Historically, the malaria vectors in western Kenya have been Anopheles funestus,
Anopheles gambiae s.s., and Anopheles arabiensis. Of these species, An. funestus
populations declined the most after the introduction of ITNs in the 1990s in Asembo, and
collections of An. funestus in the region remained low until at least 2008. Contrary to
findings during the early years of ITN use in Asembo, the majority of the Anopheles
collected here in 2010 and 2011 were An. funestus. Female An. funestus had
characteristically high Plasmodium falciparum sporozoite rates and showed nearly 100%
anthropophily. Female An. funestus were found more often indoors than outdoors and had
relatively low mortality rates during insecticide bioassays. Together, these results are of
serious concern for public health in the region, indicating that An. funestus may once
again be contributing significantly to the transmission of malaria in this region despite the
widespread use of ITNs/LLINs.
Introduction
Significant regional reductions in malaria-related morbidity and mortality have
occurred globally since 2001 (World Health Organization 2012), but these outcomes may
reverse if lessons from the past are not heeded. The Global Malaria Eradication Program
(GMEP) of the 1950-60s succeeded in eliminating malaria from many countries and
reducing malaria vector populations and malaria transmission substantially in others
where elimination was not achieved (Feachem et al. 2010, Nájera et al. 2011).
Unfortunately, resurgences in vector populations and transmission occurred in many of

112

these countries, likely due to a combination of reductions in program funding, economic
and political changes, reintroductions of the malaria parasite by movement of people,
development of drug resistance in malaria parasites, and evolution of insecticide
resistance in vector populations (Feachem et al. 2010, Nájera et al. 2011). As the global
health community strives once again toward the goal of malaria elimination and
eradication, it is therefore vital to quantify and understand the causes of any resurgence in
malaria transmission or malaria vector population size.
In western Kenya, malaria continues to cause significant morbidity and mortality
despite public health efforts to reduce malaria prevalence (Zhou et al. 2011, Hamel et al.
2011). These efforts include the widespread distribution of long-lasting impregnated nets
(LLINs), which target indoor biting malaria vectors (Bayoh et al. 2010). While high
community-level coverage with LLINs reduces morbidity and mortality caused by
malaria in the short term through impacts on malaria vector populations (Hawley et al.
2003), the long-term effectiveness of this intervention depends, in part, on whether
insecticide resistance evolves in vector populations (Ranson et al. 2011, Trape et al.
2011).
Historically, the primary malaria vector mosquito species in western Kenya were
Anopheles gambiae sensu stricto (s.s.) and Anopheles funestus (Beier et al. 1990, Taylor
et al. 1990). Anopheles arabiensis was historically considered a secondary malaria vector
in the region (Beier et al. 1990, Taylor et al. 1990), though the role of An. arabiensis in
malaria transmission has become increasingly important in the last decade as
conventional insecticide treated bed nets (ITNs) and LLINs have reduced the abundance
of the other vector species (Bayoh et al. 2010). Following the introduction of ITNs in the

113

Asembo region of western Kenya in 1997 during a large scale, randomized trial, the
population of An. funestus declined the most of these three species (Gimnig et al. 2003).
The population of An. funestus remained low for several years after the trial ended and as
ITNs were distributed nationally (Lindblade et al. 2004, Bayoh et al. 2010).
In Kenya resistance to the pyrethroid insecticides used in ITNs has been reported
largely in An. gambiae sensu lato (s.l.) (Vulule et al. 1994, Mathias et al. 2011, Ochomo
et al. 2013), a species complex of at least eight closely related species including An.
gambiae s.s. and An. arabiensis (Coetzee et al. 2013). There are few published studies on
insecticide resistance in An. funestus from Kenya, although reports of resistance to
pyrethroids, carbamates, and DDT in An. funestus from other regions of Africa have
increased over the last decade (Coetzee and Koekemoer 2013). Notably, pyrethroid
resistance in An. funestus was associated with an increase in malaria cases in South
Africa in the 1990s (Hargreaves et al. 2000).
The purpose of this study was several-fold. First, we investigated relative
abundances, human biting rates, and malaria infection rates of An. funestus and other
malaria vectors in the same western Kenyan region (Asembo) where vector abundances
had decreased previously due to ITN distribution programs (Gimnig et al. 2003,
Lindblade et al. 2004), to determine if vector abundances and malaria transmission
remained similarly low or had increased despite continued high coverage of ITNs and
LLINs. Additionally, the sensitivity of the vector populations to pyrethroid insecticides in
ITNs and LLINs was measured through standardized bioassays to investigate the
possibility that pyrethroid resistance could explain changes in the malaria vector
populations.

114

Methods
Study site. Adult Anopheles mosquitoes were sampled in the Asembo region of
Rarieda District (Figure 4.1). Numerous studies of malaria vectors have been conducted
here since the 1970s, and the region has been described in detail (Phillips-Howard et al.
2003, Gimnig et al. 2013). Malaria is holoendemic in Asembo and is caused chiefly by
Plasmodium falciparum. Rainfall occurs year-round but is seasonally bimodal, with
peaks occurring from March through May and in November and December. Mosquito
sampling intervals for the indoor resting and human landing studies in 2010 and 2011
(below) were chosen to coincide with seasonal peaks of Anopheles populations, which
occur after the March-May rainy season (Beier et al. 1990, Taylor et al. 1990, Odiere et
al. 2007).
As mentioned previously, Asembo was the site of a randomized, controlled trial
of ITNs from 1997 to 1999 (Phillips-Howard et al. 2003). Following the trial, ITNs were
distributed to control villages, and the ITNs were retreated every 6 to 9 mo with
permethrin until 2002, then with alphacypermethrin until 2007 (Lindblade et al. 2004,
Bayoh et al. 2010). This program in Asembo covered a population of approximately
55,000 persons as of 1997 (Phillips-Howard et al. 2003), which grew to 66,727 by 2011
(F. Odhiambo, personal communication). Nationwide, ITNs became available at partially
subsidized rates through the retail sector in 2002, and at heavily subsidized rates through
health clinics in 2004 (Noor et al. 2007). Furthermore, the Kenya Division of Malaria
Control provided LLINs in mass distribution campaigns in 2006 and in June 2011,
resulting in high ownership with the goal of reaching universal coverage, or one LLIN for
every two people (Division of Malaria Control 2009).

115

Sampling the indoor resting populations. Indoor-resting Anopheles were
sampled from 14-18 June 2010 and 16 May to 24 June 2011 using the pyrethrum spray
catch method (PSC) as described by Gimnig and colleagues (2003). Sampling was done
2

in 416 houses across an area of 15 km in 2010 and in 806 houses across an area of 100
2

km in 2011, and each house was sampled once. Differences between years were
examined to determine whether grouping the data across the two sampling periods was
appropriate, with analyses performed using R version 2.14.2 (R Development Core
Team, Vienna, Austria). Differences between years in the number collected per house
within each Anopheles species within each sex were determined using non-parametric
Mann-Whitney tests. Differences between years within each Anopheles species in the
proportions of blood meals from humans and cattle, and in the ELISA positivity rate (see
below), were determined using Fisher’s exact test.
Sampling the human landing populations. Host-seeking Anopheles were
2

sampled indoors and outdoors in 75 villages (150 houses) covering about 200 km from
13 June to 22 July 2011 using the human landing collection method (HLC) as described
by Gimnig and colleagues (2013). Local adult men were trained as collectors and
organized into 38 teams of 4 persons from two neighboring villages with the exception of
1 team that consisted of only 2 collectors. Each team rotated among four collection sites,
sampling for four nights every week. Collectors working outdoors were given discretion
to stop collections during rainfall, indicating any hour for which collections were stopped
because of rain. Collectors working indoors were instructed to continue regardless of

116

rainfall. All collectors were provided atovaquone proguanil as malaria prophylaxis during
the sampling period and for 7 days after the last night of collection.
Historical comparison. Anopheles females were sampled weekly in 19 villages
in southern Asembo from 1993 through September 1997 using the bed net trap sampling
method (BNT). From December 1996 to September of 1997, collections were done in 10
of the villages as the remaining villages had received ITNs as part of the large-scale trial.
BNTs consisted of untreated nets hung above the sleeping spaces of volunteers so that the
bottom edge of the net was approximately 6 cm above the bed. Mosquitoes that fed on
the sleeper and remained resting inside the net were collected the following morning
using mouth aspirators. The sleepers were not at an additional risk for malaria during
BNT sampling as ITNs were not the standard of care at the time.
From 2002 through 2008, Anopheles females were sampled monthly using CDC
light traps (LT). Light traps were set inside houses next to persons sleeping under their
own ITNs. Each person was instructed to turn on the trap just before they went to sleep
and mosquitoes were collected from traps in the morning. The traps were set in 30-60
houses each month and run for two consecutive nights in each house.
Collection rates of LT relative to HLC from Wong and colleagues (2013) were
used to compare the LT data from 2002 through 2008 to the HLC data from 2011.
Collection rates of BNT relative to HLC, calculated as described below, were used to
compare the BNT data from 1993 through 1997 to the HLC data from 2011. Only data
from June and July of each year of BNT and LT sampling were used in these
comparisons to avoid any potential bias caused by seasonal variation in the abundance of
either An. gambiae s.l. or An. funestus.

117

A comparison of the relative collection rates of HLC and BNT was done using
weekly samples of Anopheles from a pool of 91 houses in 15 villages covering about 70
2

km in Asembo from 1 Nov 1992 to 1 May 1993. BNT and HLC methods were carried
out as described above. HLC and BNT were done in the same houses, with HLC always
done 1-3 nights after BNT for a given house. The collection rate of BNT relative to HLC
was calculated using a generalized linear model with a log link function and negative
binomial error distribution. To account for repeated measures within houses, an
autoregressive correlation structure was used, allowing for correlation between
collections to decrease with increasing time. Analyses were performed using PROC
GENMOD in SAS version 9.2 (SAS Institute, Cary, NC, USA).
Mosquito identification, sporozoite rates and blood meal analysis. All
Anopheles were morphologically identified to species according to Gillies and Coetzee
(1987). Specimens from PSC and HLC sampling identified morphologically as part of the
An. gambiae s.l. species complex were further differentiated to the species level by PCR
using primers specific to Anopheles gambiae s.s. and An. arabiensis (Scott et al. 1993).
All females collected during PSC and HLC were separated at the line of the thorax and
abdomen to allow for separate diagnostic tests on each specimen. The heads and thoraces
of those females were tested for P. falciparum sporozoite proteins by ELISA (Wirtz et al.
1987) using the P. falciparum sporozoite ELISA reagent kit (MRA-890, MR4, ATCC®
Manassas, VA). The blood meal hosts of all fed and half-gravid females collected during
PSC were identified by direct sequencing of the vertebrate mitochondrial cytochrome B
gene (Hamer et al. 2008). Mosquito voucher specimens were deposited in the Laboratory

118

of Entomology at the Center for Global Health Research, Kenya Medical Research
Institute/Centers for Disease Control and Prevention in Kisian, Kenya (Table 4.1).
Insecticide bioassays. Insecticide susceptibility assays were carried out following
the World Health Organization protocol (World Health Organization 1998) using An.
gambiae s.l. and An. funestus adults collected from houses in Asembo using backpack
aspirators from May through October 2012. Only unfed females not injured during
collections were retained for bioassays. These mosquitoes were held for 1 h with access
to 10% sugar solution, then exposed in the field to insecticide-impregnated filter paper
for 1 h. Following exposure the mosquitoes were transferred to a clean holding tube with
10% sugar solution, and mortality was determined 24 h later. The mosquitoes were
exposed to the pyrethroids permethrin (at a concentration of 0.75%) and deltamethrin
(0.05%), and the carbamate bendiocarb (0.01%).
Ethical approval. This work was approved by the Institutional Review Boards of
the U.S. Centers for Disease Control and Prevention and Michigan State University and
by the Ethical Review Committee of the Kenya Medical Research Institute.
Results
Indoor resting samples. A total of 1,897 Anopheles females and 1,033 Anopheles
males were collected indoors during PSC sampling, and greater than 99% of the
Anopheles collected were identified morphologically as either An. funestus or An.
gambiae s.l. The only other anopheline species found indoors during PSC sampling was
Anopheles rufipes. A total of 830 (85%) of the An. gambiae s.l. specimens were
successfully differentiated into An. gambiae s.s. and An. arabiensis, while PCR
amplification failed or was not done in 146 An. gambiae s.l. specimens. The number of

119

Anopheles collected per house differed between the two PSC sampling periods, though
An. funestus made up a large proportion of the Anopheles collected in both years (Table
4.2). In 2010 three quarters (75.2%) of all female Anopheles collected were An. funestus,
and in 2011 most of the female Anopheles collected were either An. funestus (37.9%) or
An. arabiensis (37.0%).
Blood meal hosts were identified in 946 of the 1,328 fed or half-gravid Anopheles
females collected during PSC (Figure 4.2). The proportions of blood meals from humans
and cattle in An. funestus did not differ between years (Fisher’s exact test, p = 1.00).
Nearly all of the 715 An. funestus blood meal hosts identified were from humans, though
2.5% were from cattle. The proportions of blood meals from humans and cattle in An.
gambiae and An. arabiensis differed between years (Fisher’s exact test, p = 0.017 and p =
0.037, respectively). In 2010, 75.7% of the 33 An. gambiae s.s. blood meals and 51.3% of
the 115 An. arabiensis blood meals identified were from humans. In 2011, 94.5% of the
55 An. gambiae s.s. blood meals and 35.6% of the 73 An. arabiensis blood meals
identified were from humans. The remaining blood meals in both species from both years
were from cattle, except for a total of five blood meals from goats in An. arabiensis.
With the exception of An. rufipes, a proportion of individuals of all Anopheles
species collected during PSC sampling were positive by ELISA for P. falciparum
sporozoite infection, but the percentage positive varied by mosquito species (Table 4.3).
The ELISA-positivity rate varied between years only for An. funestus (3.4% in 2010 and
8.9% in 2011, p < 0.001). The highest rate across both years was in An. gambiae s.s., and
the lowest rate was in An. arabiensis.

120

Human landing samples. HLC sampling was done indoors for a total of 888
collector-nights (12,432 collector-hours) and outdoors for 889.4 collector-nights (12,451
collector-hours). During HLC sampling, 1,936 female Anopheles were collected. Greater
than 98% of these were morphologically identified as either An. funestus or An. gambiae
s.l., while the remaining specimens were Anopheles coustani and An. rufipes. A total of
380 (71%) of the An. gambiae species complex specimens were differentiated into An.
gambiae s.s. (43%) and An. arabiensis (28%), while PCR amplification failed in 159
(29%) of the specimens. Most of the Anopheles females collected during HLC sampling
were An. funestus, and more females of each species were collected indoors than
outdoors (Figure 4.3). Similar to specimens from PSC sampling, ELISA-positivity rates
varied among the three malaria vector species during HLC sampling. Overall ELISApositivity rates for these specimens were highest in An. gambiae s.s. and An. funestus,
while the lowest rate was found in An. arabiensis (Figure 4.3).
Historical comparison. In our analysis of relative collection rates, BNT sampling
produced similar numbers of An. funestus relative to HLC sampling, but fewer An.
gambiae s.l. were collected from BNT relative to HLC (Table 4.4). According to Wong
and colleagues (2013), collections of both An. funestus and An. gambiae s.l. were lower
from LT relative to HLC in Rarieda District. The collection rate of An. funestus from LT
relative to HLC was 0.69 (95% CI: 0.49-0.98), while the relative collection rate of An.
gambiae s.l. was 0.76 (95% CI: 0.61-0.96) (Wong et al. 2013).
Prior to the introduction of ITNs in Asembo, the populations of both An. funestus
and An. gambiae s.l. were large according to BNT sampling (Figure 4.4). Few An.
funestus were collected in Asembo using LT from 2002 through 2008. For the months of

121

June and July, which historically were the months when An. funestus populations peaked,
zero An. funestus were collected in two years from the period, and never were more than
0.17 An. funestus per trap collected. During the same sampling period, the monthly mean
of An. gambiae s.l. collected per trap ranged from 0.13 to 1.53 with a mean of 0.75
(Figure 4.4).
Insecticide bioassays. Figure 4.5 depicts mortality to bendiocarb, deltamethrin,
and permethrin in wild caught An. gambiae s.l. and An. funestus adults from Asembo. A
total of 78 An. funestus were exposed to deltamethrin, 38 to permethrin, and 34 to
bendiocarb. A total of 159 An. gambiae s.l. were exposed to deltamethrin, 6 to
permethrin, and 48 to bendiocarb. While mortality rates following exposure to
bendiocarb were high, much lower rates were seen for An. funestus samples exposed to
deltamethrin and permethrin, indicating resistance in these specimens to the insecticides
used in ITNs and LLINs.
Discussion
Anopheles funestus was the dominant species collected in all three of our
sampling efforts in 2010 and 2011, comprising between one third and three quarters of
the female Anopheles collected (Table 4.2; Figure 4.3). Historically, An. funestus was
common in our study region (Evans and Symes 1937, Beier et al. 1990, Taylor et al.
1990), and it remained abundant until the introduction of ITNs in the late 1990s (Figure
4.4). Following ITN distribution, An. funestus became rare and likely played a minor role
in malaria transmission in the period thereafter as ITN coverage increased (Gimnig et al.
2003, Lindblade et al. 2006). The population of An. funestus in Asembo remained low
through 2008 according to the LT sampling reported here (Figure 4.4), but our sampling

122

in 2010 and 2011 indicates a marked reversal of this trend. The reemergence of this
species has important implications for malaria transmission, which has increased recently
despite widespread distribution of ITNs and LLINs (Bayoh et al. 2010, Zhou et al. 2011,
Hamel et al. 2011).
The low mortality rate of An. funestus collected from Asembo when exposed to
the pyrethroids used in ITNs and LLINs (Figure 4.5) supports one possible mechanism by
which the An. funestus population in the region has been able to increase, though a more
extensive analysis is necessary to determine the scope of insecticide resistance in An.
funestus populations within western Kenya. The potential for insecticide resistance to
reduce the effectiveness of ITNs is widely recognized (Ranson et al. 2011), and the
appearance of pyrethroid resistance in populations of An. funestus after indoor residual
spraying in South Africa (Hargreaves et al. 2000) underscores that potential for this
malaria vector. A recent review by Coetzee and Koekemoer (2013) covers the extent to
which insecticide resistance in An. funestus is currently understood. Varying degrees of
resistance have been reported across many regions of Africa, notably including Tororo
District in eastern Uganda near the Kenyan border (Morgan et al. 2010). In addition to
the reemergence of An. funestus in the 1990s after evolving pyrethroid resistance in South
Africa (Hargreaves et al. 2000), the development of insecticide resistance in other malaria
vectors during the Global Malaria Eradication Program provides numerous examples of
the detrimental effects of insecticide resistance on malaria control (Busvine and Pal
1969). Thus, the reemergence of An. funestus in this region of high ITN and LLIN
coverage indicates the need to further investigate the resistance characteristics of this
population.

123

Anopheles funestus and An. gambiae s.s. were found here to have taken their
blood meals almost exclusively from humans, highlighting one factor contributing to
their high vectorial capacity, a measure of the rate at which a vector population transmits
malaria (Garrett-Jones 1964). Similarly to An. gambiae s.s. (Bayoh et al. 2010), the
presence of ITNs or LLINs appears not to shift An. funestus host selection away from
humans. Anopheles arabiensis took a larger proportion of blood meals from cattle than
did either An. gambiae s.s. or An. funestus, but An. arabiensis also took blood meals from
humans, demonstrating broader variation in feeding behaviors typical of An. arabiensis
(Takken and Verhulst 2013). Overall, the results from this blood meal analysis emphasize
the need for additional vector control methods beyond the indoor, insecticide-based
strategies that target only highly anthropophilic vectors preferring late-night, indoor
feeding.
The high rate of sporozoite ELISA-positive females collected during both PSC
and HLC sampling implies a substantial infective human reservoir of malaria in Asembo,
corresponding with recent measures of P. falciparum prevalence in the region (Hamel et
al. 2011). We found the highest ELISA-positivity rates in An. gambiae s.s. females,
implying the continued importance of this vector to malaria transmission in the region.
Similarly high ELISA-positivity rates were found in An. funestus, further indicating its
reemergence as one of the primary malaria vectors in Asembo. The relatively lower rates
of ELISA-positive An. arabiensis females during this study agree with findings from
before the introduction of ITNs suggesting it has a lower vectorial capacity in the region
than An. gambiae s.s. and An. funestus (Taylor et al. 1990).

124

The high proportion of An. funestus relative to An. arabiensis and An. gambiae in
our indoor resting and human landing samples is an important indication of changes to
malaria transmission in this region, and demonstrates how our most efficient and widely
implemented malaria control interventions may have important limitations. Anopheles
funestus is well known as an efficient malaria vector due to near 100% anthropophily,
which was confirmed in this study along with high sporozoite rates. The distribution of
ITNs in Asembo in the 1990s significantly impacted An. funestus, and collections of An.
funestus in the region had remained low until at least 2008. The increased abundance of
An. funestus reported in this study suggests a reemergence of this species as a primary
malaria vector in western Kenya, and might be explained by the development of
insecticide resistance. Public health officials here and throughout regions where ITNs or
IRS are deployed should be aware of changes in malaria vector populations, especially as
our findings indicate reduced effectiveness of ITNs as a long-term, stand-alone method
for malaria control. The GMEP of the 1950-60s revealed the potential dangers of relying
on a single method of vector control, as malaria transmission resurged in many areas
where significant gains had been made previously. Strategies to attenuate insecticide
resistance development and complementary control methods for integrated vector
management (World Health Organization 2004) deserve a renewed focus as the global
health community continues to fight malaria.

125

APPENDIX

126

Chapter IV Tables and Figures

127

Table 4.1: Record of deposition of voucher specimens for Chapter IV. The specimens
listed below have been deposited in the Laboratory of Entomology at the Center for
Global Health Research, Kenya Medical Research Institute/Centers for Disease Control
and Prevention in Kisian, Kenya, as samples of those species, which were used in this
research. Voucher recognition labels bearing the voucher number 2013-14 have been
attached.
Species
Anopheles gambiae
Anopheles gambiae
Anopheles arabiensis
Anopheles arabiensis
Anopheles rufipes
Anopheles rufipes
Anopheles coustani
Anopheles funestus
Anopheles funestus

Life Stage Quantity Preservation
Adult female
5
Pinned
Adult male
5
Pinned
Adult female
5
Pinned
Adult male
5
Pinned
Adult female
5
Pinned
Adult male
4
Pinned
Adult female
5
Pinned
Adult female
5
Pinned
Adult male
5
Pinned

128

Table 4.2: Number of Anopheles collected per house during pyrethrum spray catch sampling shown by species, sex, and year.
Percentages of the total for each column are shown in parentheses. P-values are from non-parametric Mann-Whitney tests to
determine differences between years for each species within each sex. NA, not applicable.
*
These specimens were identified as members of the Anopheles gambiae species complex according to morphological methods but
could not be further differentiated by polymerase chain reaction.

Species
An. funestus
An. arabiensis
An. gambiae s.s.
*
An. gambiae s.l.
An. rufipes

Females
Males
2010
2011
p-value
2010
2011 p-value
2.305 (75.2%) 0.293 (37.9%) < 0.001 1.327 (80.0%) 0.231 (54.2%) < 0.001
0.498 (16.2%) 0.285 (37.0%)
0.007 0.127 (7.7%) 0.104 (24.5%)
0.374
0.099 (3.2%) 0.122 (15.8%)
0.230 0.149 (9.0%) 0.083 (19.5%)
0.052
0.163 (5.3%) 0.066 (8.5%) < 0.001 0.055 (3.3%) 0.002 (0.6%) < 0.001
0.000 (0.0%) 0.006 (0.8%)
NA 0.000 (0.0%) 0.005 (1.2%)
NA

129

Table 4.3: Plasmodium falciparum sporozoite-negative and positive female mosquitoes,
according to enzyme-linked immunosorbent assay (ELISA), from pyrethrum spray catch
(PSC) sampling, shown by species with percent positive in parentheses.
*
ELISA positivity rates for Anopheles funestus were significantly different between PSC
sampling periods in 2010 and 2011 (3.4 and 8.9% respectively; Fisher’s exact test, p <
0.001).
†
These specimens were identified as members of the Anopheles gambiae species
complex according to morphological methods but could not be further differentiated by
polymerase chain reaction.
Species
An. funestus
An. arabiensis
An. gambiae s.s.
†
An. gambiae s.l.

Negative Positive (%)
*

1140
433
127

54 (4.5)
4 (0.9)
12 (8.6)

119

2 (1.7)

130

Table 4.4: Comparison of human landing catch (HLC) and bed net trap (BNT) sampling
methods for Anopheles funestus and Anopheles gambiae s.l., Nov 1992 to May 1993.
Relative rates of collection and associated 95% CI and p-value were calculated using
negative binomial regression. P-values indicate the probability that the relative collection
rate of BNT sampling was different from that of HLC. NA, not applicable.
Method
Total Mean per night (95% CI) Relative rate (95% CI) p-value
An. funestus
HLC
730
1.25 (1.05, 1.45)
1.00
NA
BNT
917
1.57 (1.34, 1.79)
1.27 (0.92, 1.76)
0.145
An. gambiae
HLC
2408
BNT
1389

4.12 (3.47, 4.78)
2.38 (2.01, 2.74)

131

1.00
NA
0.57 (0.44, 0.74) < 0.001

Figure 4.1: (A) Map of Kenya with box indicating location of the study region in Nyanza
Province. (B) Map showing the location of Asembo on the shores of Lake Victoria, about
50 km west of the city of Kisumu.

132

Figure 4.2: Proportion of blood meals taken from humans and cattle by Anopheles species collected during pyrethrum spray catch in
Asembo. Non-amplifiers to polymerase chain reaction for blood meal identification are not shown (29%).
*
Data pooled across years for Anopheles funestus because the proportions of blood meals from humans and cattle did not differ
between years (Fisher’s exact test, p = 1.00).

133

Figure 4.3: Number (No.) of Anopheles females collected per collector per night during human landing catch sampling shown by
species and location (indoors/outdoors). Error bars are 95% confidence intervals. Plasmodium falciparum sporozoite rates as
determined by enzyme-linked immunosorbent assays are shown as percentages above each species. Anopheles gambiae s.l. specimens
were identified as members of the An. gambiae species complex according to morphological methods but could not be further
differentiated by polymerase chain reaction.

134

Figure 4.4: (A) Mean number of Anopheles females collected per sample in June through July from 1993 through 2011 using bed net
traps (BNT; 1993-1997), light traps (LT; 2002-2008), and human landing catch (HLC; 2011). (B) Mean number of Anopheles females
collected, adjusted for comparison to HLC sampling using the relative rates of BNT to HLC (1993-1997) and LT to HLC (20022008). (C) Proportion of total Anopheles females collected identified as either Anopheles funestus or Anopheles gambiae s.l. Error bars
indicate 95% confidence intervals. Data were not available from 1998-2001, 2009, and 2010.

135

Figure 4.4 (cont’d)

136

Figure 4.4 (cont’d)

137

Figure 4.5: Percent mortality after 24 hours of wild collected Anopheles funestus and Anopheles gambiae s.l. adults from Asembo
when exposed to permethrin, deltamethrin, or bendiocarb in one-hour World Health Organization tube bioassays.

138

REFERENCES

139

REFERENCES

Bayoh, M. N., D. K. Mathias, M. R. Odiere, F. M. Mutuku, L. Kamau, J. E. Gimnig, J.
M. Vulule, W. A. Hawley, M. J. Hamel, and E. D. Walker. 2010. Anopheles
gambiae: historical population decline associated with regional distribution of
insecticide-treated bed nets in western Nyanza Province, Kenya. Malaria Journal
9:62.
Beier, J. C., P. V. Perkins, F. K. Onyango, T. P. Gargan, C. N. Oster, R. E. Whitmire, D.
K. Koech, and C. R. Roberts. 1990. Characterization of malaria transmission by
Anopheles (Diptera: Culicidae) in western Kenya in preparation for malaria vaccine
trials. Journal of Medical Entomology 27:570–577.
Busvine, J. R., and R. Pal. 1969. The impact of insecticide-resistance on control of
vectors and vector-borne diseases. Bulletin of the World Health Organization
40:731–744.
Coetzee, M., and L. L. Koekemoer. 2013. Molecular systematics and insecticide
resistance in the major African malaria vector Anopheles funestus. Annual review of
entomology 58:393–412.
Coetzee, M., R. H. Hunt, R. C. Wilkerson, A. della Torre, M. B. Coulibaly, and N. J.
Besansky. 2013. Anopheles coluzzii and Anopheles amharicus, new members of the
Anopheles gambiae complex. Zootaxa 3619:246–274.
Division of Malaria Control. 2009. National malaria strategy 2009–2017. Pages 1–116.
Ministry of Public Health and Sanitation, Nairobi, Kenya.
Evans, A. M., and C. B. Symes. 1937. Anopheles funestus and its allies in Kenya. Annals
of Tropical Medicine and Parasitology 31:105–112.
Feachem, R. G., A. A. Phillips, J. Hwang, C. Cotter, B. Wielgosz, B. M. Greenwood, O.
Sabot, M. H. Rodríguez, R. R. Abeyasinghe, T. A. Ghebreyesus, and R. W. Snow.
2010. Shrinking the malaria map: progress and prospects. Lancet 376:1566–1578.
Garrett-Jones, C. 1964. The human blood index of malaria vectors in relation to
epidemiological assessment. Bulletin of the World Health Organization 30:241–261.
Gillies, M. T., and M. Coetzee. 1987. A supplement to the Anophelinae of Africa south
of the Sahara (Afrotropical Region). Pages 1–143. South African Institute for
Medical Research, Johannesburg, South Africa.
Gimnig, J. E., E. D. Walker, P. Otieno, J. Kosgei, G. Olang, M. Ombok, J. Williamson,
D. Marwanga, D. Abong'o, M. Desai, S. Kariuki, M. J. Hamel, N. F. Lobo, J. M.
Vulule, and M. N. Bayoh. 2013. Incidence of malaria among mosquito collectors

140

conducting human landing catches in western Kenya. American Journal of Tropical
Medicine and Hygiene 88:301–308.
Gimnig, J. E., J. M. Vulule, T. Q. Lo, L. Kamau, M. S. Kolczak, P. A. Phillips-Howard,
E. M. Mathenge, F. O. ter Kuile, B. L. Nahlen, A. W. Hightower, and W. A. Hawley.
2003. Impact of permethrin-treated bed nets on entomologic indices in an area of
intense year-round malaria transmission. American Journal of Tropical Medicine and
Hygiene 68:16–22.
Hamel, M. J., K. Adazu, D. Obor, M. Sewe, J. M. Vulule, J. M. Williamson, L. Slutsker,
D. R. Feikin, and K. F. Laserson. 2011. A reversal in reductions of child mortality in
western Kenya, 2003-2009. American Journal of Tropical Medicine and Hygiene
85:597–605.
Hamer, G. L., U. D. Kitron, J. D. Brawn, S. R. Loss, M. O. Ruiz, T. L. Goldberg, and E.
D. Walker. 2008. Culex pipiens (Diptera: Culicidae): a bridge vector of West Nile
Virus to humans. Journal of Medical Entomology 45:125–128.
Hargreaves, K., L. L. Koekemoer, B. D. Brooke, R. H. Hunt, J. Mthembu, and M.
Coetzee. 2000. Anopheles funestus resistant to pyrethroid insecticides in South
Africa. Medical and Veterinary Entomology 14:181–189.
Hawley, W. A., P. A. Phillips-Howard, F. O. ter Kuile, D. J. Terlouw, J. M. Vulule, M.
Ombok, B. L. Nahlen, J. E. Gimnig, S. K. Kariuki, M. S. Kolczak, and A. W.
Hightower. 2003. Community-wide effects of permethrin-treated bed nets on child
mortality and malaria morbidity in western Kenya. American Journal of Tropical
Medicine and Hygiene 68:121–127.
Lindblade, K. A., J. E. Gimnig, L. Kamau, W. A. Hawley, F. Odhiambo, G. Olang, F. O.
ter Kuile, J. M. Vulule, and L. Slutsker. 2006. Impact of sustained use of insecticidetreated bednets on malaria vector species distribution and culicine mosquitoes.
Journal of Medical Entomology 43:428–432.
Lindblade, K. A., T. P. Eisele, J. E. Gimnig, J. A. Alaii, F. Odhiambo, F. O. ter Kuile, W.
A. Hawley, K. A. Wannemuehler, P. A. Phillips-Howard, D. H. Rosen, B. L. Nahlen,
D. J. Terlouw, K. Adazu, J. M. Vulule, and L. Slutsker. 2004. Sustainability of
reductions in malaria transmission and infant mortality in western Kenya with use of
insecticide-treated bednets: 4 to 6 years of follow-up. JAMA : the journal of the
American Medical Association 291:2571–2580.
Mathias, D. K., E. Ochomo, F. K. Atieli, M. Ombok, M. N. Bayoh, G. Olang, D. Muhia,
L. Kamau, J. M. Vulule, M. J. Hamel, W. A. Hawley, E. D. Walker, and J. E.
Gimnig. 2011. Spatial and temporal variation in the kdr allele L1014S in Anopheles
gambiae s.s. and phenotypic variability in susceptibility to insecticides in western
Kenya. Malaria Journal 10:10.

141

Morgan, J. C., H. Irving, L. M. Okedi, A. Steven, and C. S. Wondji. 2010. Pyrethroid
resistance in an Anopheles funestus population from Uganda. PLoS ONE 5:e11872.
Nájera, J. A., M. González-Silva, and P. L. Alonso. 2011. Some lessons for the future
from the Global Malaria Eradication Programme (1955–1969). PLoS Medicine
8:e1000412.
Noor, A. M., A. A. Amin, W. S. Akhwale, and R. W. Snow. 2007. Increasing coverage
and decreasing inequity in insecticide-treated bed net use among rural Kenyan
children. PLoS Medicine 4:e255.
Ochomo, E., M. N. Bayoh, W. G. Brogdon, J. E. Gimnig, C. Ouma, J. M. Vulule, and E.
D. Walker. 2013. Pyrethroid resistance in Anopheles gambiae s.s. and Anopheles
arabiensis in western Kenya: phenotypic, metabolic and target site characterizations
of three populations. Medical and Veterinary Entomology 27:156–164.
Odiere, M., M. N. Bayoh, J. E. Gimnig, J. M. Vulule, L. Irungu, and E. D. Walker. 2007.
Sampling outdoor, resting Anopheles gambiae and other mosquitoes (Diptera:
Culicidae) in western Kenya with clay pots. Journal of Medical Entomology 44:14–
22.
Phillips-Howard, P. A., B. L. Nahlen, J. A. Alaii, F. O. ter Kuile, J. E. Gimnig, D. J.
Terlouw, S. P. Kachur, A. W. Hightower, A. A. Lal, E. Schoute, A. J. Oloo, and W.
A. Hawley. 2003. The efficacy of permethrin-treated bed nets on child mortality and
morbidity in western Kenya I: development of infrastructure and description of study
site. American Journal of Tropical Medicine and Hygiene 68:3–9.
Ranson, H., R. N’Guessan, J. Lines, N. Moiroux, Z. Nkuni, and V. Corbel. 2011.
Pyrethroid resistance in African anopheline mosquitoes: what are the implications for
malaria control? Trends in Parasitology 27:91–98.
Scott, J. A., W. G. Brogdon, and F. H. Collins. 1993. Identification of single specimens
of the Anopheles gambiae complex by the polymerase chain reaction. American
Journal of Tropical Medicine and Hygiene 49:520–529.
Takken, W., and N. O. Verhulst. 2013. Host Preferences of Blood-Feeding Mosquitoes.
Annual review of entomology 58:433–453.
Taylor, K. A., J. K. Koros, J. Nduati, R. S. Copeland, F. H. Collins, and A. D. BrandlingBennett. 1990. Plasmodium falciparum infection rates in Anopheles gambiae, An.
arabiensis, and An. funestus in western Kenya. American Journal of Tropical
Medicine and Hygiene 43:124–129.
Trape, J.-F., A. Tall, N. Diagne, O. Ndiath, A. B. Ly, J. Faye, F. Dieye-Ba, C. Roucher,
C. Bouganali, A. Badiane, F. D. Sarr, C. Mazenot, A. Touré-Baldé, D. Raoult, P.
Druilhe, O. Mercereau-Puijalon, C. Rogier, and C. Sokhna. 2011. Malaria morbidity

142

and pyrethroid resistance after the introduction of insecticide-treated bednets and
artemisinin-based combination therapies: a longitudinal study. The Lancet Infectious
Diseases 11:925–932.
Vulule, J. M., R. F. Beach, F. K. Atieli, J. M. Roberts, D. L. Mount, and R. W. Mwangi.
1994. Reduced susceptibility of Anopheles gambiae to permethrin associated with the
use of permethrin-impregnated bednets and curtains in Kenya. Medical and
Veterinary Entomology 8:71–75.
Wirtz, R. A., F. Zavala, Y. Charoenvit, G. H. Campbell, T. R. Burkot, I. Schneider, K. M.
Esser, R. L. Beaudoin, and R. G. Andre. 1987. Comparative testing of monoclonal
antibodies against Plasmodium falciparum sporozoites for ELISA development.
Bulletin of the World Health Organization 65:39.
Wong, J., M. N. Bayoh, G. Olang, G. F. Killeen, M. J. Hamel, J. M. Vulule, and J. E.
Gimnig. 2013. Standardizing operational vector sampling techniques for measuring
malaria transmission intensity: evaluation of six mosquito collection methods in
western Kenya. Malaria Journal 12:143.
World Health Organization. 1998. Test procedures for insecticide resistance monitoring
in malaria vectors, bio-efficacy and persistence of insecticides on treated surfaces:
report of the WHO informal consultation. Pages 1–46. World Health Organization,
Geneva.
World Health Organization. 2004. Global strategic framework for integrated vector
management. Pages 1–12. World Health Organization, Geneva.
World Health Organization. 2012. World malaria report 2012. Pages 1–276. World
Health Organization, Geneva.
Zhou, G., Y. A. Afrane, A. M. Vardo-Zalik, H. Atieli, D. Zhong, P. Wamae, Y. E.
Himeidan, N. Minakawa, A. K. Githeko, and G. Yan. 2011. Changing patterns of
malaria epidemiology between 2002 and 2010 in western Kenya: the fall and rise of
malaria. PLoS ONE 6:e20318.

143

CHAPTER V: Conclusions
Within the last decade, there has been renewed interest in the regional elimination
and global eradication of malaria (Roberts and Enserink 2007, Feachem et al. 2010,
Alonso et al. 2011, World Health Organization 2011). For this goal to be attainable, we
must understand the ecology of the local malaria vectors and apply vector control
methods in ways that acknowledge and exploit their biological complexity (Ferguson et
al. 2010). This dissertation contributes to such an understanding by describing the use of
microdam-associated aquatic habitat by malaria vector larvae, which had not been
previously investigated in western Kenya. Given that larvae of the primary malaria vector
species in the region used these habitats during the dry season, microdam reservoirs could
be important for the persistence of malaria vector populations through the dry season.
Malaria control programs could exploit this finding by implementing larval source
management (Fillinger and Lindsay 2011) at microdams in the dry season. Indeed, larval
control could have the desirable effect of delaying the seasonal build up of adult
populations as the rainy season commences (J. Miller, unpublished). Approaches could
include habitat modification to make the habitats less suitable for larvae. For example,
vegetation could be cleared away from the reservoir perimeters to discourage egg-laying
and reduce survival of An. funestus. Additionally, microdam perimeters could be
modified to reduce the formation of hoof print aggregations, the preferred habitat of An.
gambiae s.l. near microdams. For example, the entrance to microdam reservoirs used by
livestock could be covered with grass cuttings to prevent hoof prints from forming and
filling with water. Furthermore, the application of a larvicide could complement habitat
modifications at microdams. Finally, larval source management in the rainy season in

144

western Kenya should also include microdams, because An. gambiae s.l. larvae were
relatively abundant during the rainy season in microdam-associated habitats sampled
here.
This dissertation also provides detailed information about the relationship
between malaria vector larval habitat locations and both landscape variables and
accumulated precipitation. This information was used to model the locations of larval
habitats across a relatively large area and throughout seasons with variation in
precipitation. This has two potential applications. First, accurate larval habitat models can
be used to investigate the relationship between larval habitat proximity to houses and the
number of adult malaria vectors in those houses, across large areas where mapping all
larval habitats is infeasible (Bogh et al. 2007). Such models can provide indirect evidence
for determining the dispersal distances of adult malaria vectors (Thomas et al. 2013),
which is one of the major gaps in our understanding of malaria vector ecology (Ferguson
et al. 2010). Second, predictive models of larval habitat locations could potentially be
used to assist in larval control efforts by allowing targeted management (Li et al. 2011,
Nmor et al. 2013). A risk for this approach is that models fitted to data from a single
geographic location may have limited generalizability (Strauss and Biedermann 2007).
However, the sampling strategy employed here for model parameterization can be used in
other geographic locations to allow malaria control programs to build models that are
useful over larger areas.
Finally, this dissertation addresses two significant limitations of insecticidetreated bed nets (ITNs), including long-lasting impregnated nets, as a long-term, standalone method for malaria control. First, differences between ITN ownership at the

145

household-level and individual ITN use could have important implications for malaria
transmission. ITNs have been shown to provide community-wide benefits, even for those
not sleeping under ITNs, by reducing malaria vector population levels (Hawley et al.
2003b). However, the results presented here indicate that individuals living in houses
with ITNs but not sleeping under ITNs are exposed to as many malaria vectors as
individuals in houses without ITNs. Thus, while ITNs serve as a public good (Hawley et
al. 2003a), they do not confer equal household-wide protection to those not sleeping
under them. Second, the risk of malaria vector populations developing insecticide
resistance is of serious concern. Resistance to the pyrethroids used in ITNs has been
reported in multiple regions for Anopheles gambiae, Anopheles arabiensis, and
Anopheles funestus, the three most important malaria vectors in Africa (Ranson et al.
2011). Insecticide resistance in one potential cause of the reemergence reported here of
An. funestus to the highest relative abundance since the introduction of ITNs in an area of
long-term ITN use. If ITNs are to remain a useful tool for malaria control, strategies to
attenuate insecticide resistance development and complementary control methods for
integrated vector management (World Health Organization 2004) must be employed
immediately.

146

REFERENCES

147

REFERENCES

Alonso, P. L., G. Brown, M. Arevalo-Herrera, F. Binka, C. Chitnis, F. H. Collins, O. K.
Doumbo, B. Greenwood, B. F. Hall, M. M. Levine, K. Mendis, R. D. Newman, C. V.
Plowe, M. H. Rodríguez, R. Sinden, L. Slutsker, and M. Tanner. 2011. A research
agenda to underpin malaria eradication. PLoS Medicine 8:e1000406.
Bogh, C., S. W. Lindsay, S. E. Clarke, A. Dean, M. Jawara, M. Pinder, and C. J. Thomas.
2007. High spatial resolution mapping of malaria transmission risk in the Gambia,
west Africa, using LANDSAT TM satellite imagery. American Journal of Tropical
Medicine and Hygiene 76:875–881.
Feachem, R. G., A. A. Phillips, J. Hwang, C. Cotter, B. Wielgosz, B. M. Greenwood, O.
Sabot, M. H. Rodríguez, R. R. Abeyasinghe, T. A. Ghebreyesus, and R. W. Snow.
2010. Shrinking the malaria map: progress and prospects. Lancet 376:1566–1578.
Ferguson, H. M., A. Dornhaus, A. Beeche, C. Borgemeister, M. Gottlieb, M. S. Mulla, J.
E. Gimnig, D. Fish, and G. F. Killeen. 2010. Ecology: a prerequisite for malaria
elimination and eradication. PLoS Medicine 7:e1000303.
Fillinger, U., and S. W. Lindsay. 2011. Larval source management for malaria control in
Africa: myths and reality. Malaria Journal 10:353.
Hawley, W. A., F. O. ter Kuile, R. S. R. Steketee, B. L. B. Nahlen, D. J. D. Terlouw, J. E.
J. Gimnig, Y. P. Y. Shi, J. M. J. Vulule, J. A. J. Alaii, A. W. A. Hightower, M. S. M.
Kolczak, S. K. S. Kariuki, and P. A. P. Phillips-Howard. 2003a. Implications of the
western Kenya permethrin-treated bed net study for policy, program implementation,
and future research. American Journal of Tropical Medicine and Hygiene 68:168–
173.
Hawley, W. A., P. A. Phillips-Howard, F. O. ter Kuile, D. J. Terlouw, J. M. Vulule, M.
Ombok, B. L. Nahlen, J. E. Gimnig, S. K. Kariuki, M. S. Kolczak, and A. W.
Hightower. 2003b. Community-wide effects of permethrin-treated bed nets on child
mortality and malaria morbidity in western Kenya. American Journal of Tropical
Medicine and Hygiene 68:121–127.
Li, L., L. Bian, L. Yakob, G. Zhou, and G. Yan. 2011. Analysing the generality of
spatially predictive mosquito habitat models. Acta Tropica 119:30–37.
Nmor, J. C., T. Sunahara, K. Goto, K. Futami, G. Sonye, P. Akweywa, G. Dida, and N.
Minakawa. 2013. Topographic models for predicting malaria vector breeding
habitats: potential tools for vector control managers. Parasites & Vectors 6:14.
Ranson, H., R. N’Guessan, J. Lines, N. Moiroux, Z. Nkuni, and V. Corbel. 2011.
Pyrethroid resistance in African anopheline mosquitoes: what are the implications for

148

malaria control? Trends in Parasitology 27:91–98.
Roberts, L., and M. Enserink. 2007. Did They Really Say... Eradication? Science, New
Series 318:1544–1545.
Strauss, B., and R. Biedermann. 2007. Evaluating temporal and spatial generality: How
valid are species–habitat relationship models? Ecological Modelling 204:104–114.
Thomas, C. J., D. E. Cross, and C. Bogh. 2013. Landscape Movements of Anopheles
gambiae Malaria Vector Mosquitoes in Rural Gambia. PLoS ONE 8:e68679.
World Health Organization. 2011. Eliminating Malaria: Learning From the Past, Looking
Ahead. Pages 1–86. World Health Organization, Geneva.

149