Climate and malaria mapping in East Africa

1
Climate and malaria mapping in East Africa
2
Malaria transmission and climate
A complex interaction exists between man, the
parasite, the vector and the environment and this
interaction determines malaria's endemicity
Climate has a major impact on vector and parasite
development and also determines the environment
in which these transmission interactions occur
3
Temperature dependence of parasite and vector
development and survival
(Detinova, 1962 Martens, 1997)
Optimum temperature for sporogony for P.
falciparum is 25-30oC
4
Geographical limits of malaria endemicity in
relation to the 60F (15.6C) 70F (21C)
summer isotherms
Holoendemic Hyperendemic Mesoendemic Hypoendemic E
pidemic Malaria free
(Lysenko Beljaev, 1969)
Summer isotherms representing ideal temperatures
for sporogony in P. falciparum (gt18 ºC), P. vivax
and P. malariae (15 ºC)
5
Altitude ranges of highland malaria epidemics in
Eastern and Southern Africa
6
Historical classifications of malaria
transmission are available for several African
countries dating from the malaria eradication
era (1950s-1960s)
Based largely on expert opinion, some
epidemiological observations and knowledge of
rainfall patterns Settlement patterns,
urbanization, localized control efforts and
development of drug resistance have modified the
distribution of malaria
Very little empirical data of malaria are
available for SSA and so more recent maps are
based on predictive models of climate-based
surrogate measures of malaria transmission
malaria free malaria near water
transmission for 3-6 months in a year
transmission for gt6 months in a year
Butler, 1959 Government of Uganda, 1962
Government of Tanganyika, 1956
7
Determinants of malaria transmission intensity

• Temperature
• Rainfall and humidity
• Seasonality of climate
• Altitude and frost
• Soils and agricultural land-use
• Man-made changes to the environment
• Availability of permanent breeding sites
• Population, settlement and urbanisation
• Malaria control interventions

Climatic
Non-climatic
8
Advances in Geographic Information Systems (GIS)
and Remote Sensing (RS) have fuelled a
renaissance in malaria risk mapping in Africa
using climate data Such climate and
environmental data enable the stratification of
the varied epidemiology of malaria across the
areas of sub-Sahara Africa where the disease is
endemic
9
The geographical distribution of malaria
infections
Pampana Russel, 1955 World Health
Organization, 1966 WHO, 1997 Hay et al., 2004
2002 1994 1975 1965 1946 1900 Malaria free
Intensity and seasonality of infection are not
captured
10
Distribution of PR surveys across East Africa
(19072003)
Climate information is used to model malaria
transmission in data poor areas
Kenya n992, Tanzania n 522, Uganda n 489
11
Distribution of An. gambiae mapped against total
annual rainfall
Field records of 2537 entomological studies
dating from 1944) have been transferred to a GIS
Data of vector distributions across Africa are
also sparse but these data have been used to
train RS climate data in predicting species
distribution in previously un-surveyed locations
Coetzee et al (2000).
12
GIS based models have been developed to map
malaria in data poor areas using combinations of
climate data and malariometric data
Climate models of vector-borne disease are
essentially models of environmental suitability
for a vector. Areas where environmental
conditions are unsuitable are classed as outside
the distributional limits of the vector.
Two main sources of climate data have been used
in malaria mapping a) Synoptic weather station
data b) Remotely sensed surrogate measures of
climate
13
Malaria maps based on remotely sensed data
14
Using remote sensing (NDVI, LST) to predicting
malaria transmission
RS based predictive map

• The study looked at the relationship between
  historical stratification of malaria and NDVI and
  LST
• Demonstrates the utility of RS for malaria
  predictive mapping

Historical map

• The satellite data were able to predict
  historical classifications of malaria with an
  overall accuracy of 89.9 (kappa0.775) with
  areas of the least intense transmission being
  predicted with greatest accuracy (96).

malaria free malaria near water 3-6 months
in a year gt6 months in a year
15
Predicting vector distribution using satellite
imagery
Predictions of the distributions of An. gambiae
s.s., An. arabiensis, An. merus, An. melas and
An. quadriannulatus have been made for Africa
between latitudes 18oN30oS using data from
NOAA-AVHRR and Meteosat satellites
Rogers et al., (2002)
16
Predicting malaria seasonality using NDVI
thresholds
Duration of a malaria season was computed by
counting the number of months during which NDVIs
were at or above a certain threshold related to a
rise in number of admissions
Hay et al. (1998)
17
Satellite-derived predictions of EIR for Africa
Satellite-based environmental data were able to
discriminate between 5 equal-sized classes of EIR
giving a high index of agreement (kappa
statistic0.77) between predicted values and
training data. The satellite data included
temperature variables i.e. reflectance in the
middle infra-red (MIR) channel and land surface
temperature (LST) and the cold cloud duration
(CCD), an index of precipitation
Rogers et al., (2002).
18
A model of malaria risk for East Africa using
parasite prevalence data
Statistical techniques applied to high spatial
resolution (1x1 km) remotely sensed climate
surrogate data urbanization and land use data
were used to predict the intensity of malaria
transmission as defined categorically according
to the childhood parasite ratio in East Africa
(330 PR studies). Discriminant analysis was used
to train the environmental and urbanization
predictor variables to distinguish between four
classes of PR risk. Surrogates of precipitation
(NDVI) and temperature (LST) were among the top
predictors of PR class
0lt5, PR 5-lt25, PR 25-lt75, PR gt75,
no prediction, water body
Omumbo et al., (2005).
19
Malaria maps based on synoptic weather station
data
20
Climate Suitability for malaria transmission
Most ecological measures have some uncertainty
and do not necessary fall into distinct Boolean
categories. Climatic conditions, for example, can
be suitable for malaria transmission along a
gradient of suitability relative to 2 extremes,
suitable (S) and unsuitable (U). Such data
can be described as fuzzy sets. The MARA model
uses a sigmoidal fuzzy membership curve to rank
climate variables derived from spatially
interpolated weather station data (Hutchinson et
al., 1996) according to a scale of suitability
of climate for malaria transmission.
21
Climatic suitability for malaria transmission in
Africa
Optimum temperatures for sporogonic development
in the mosquito were defined as 22oC - 32oC
(suitable fuzzy suitability 1) while lt18oC and
gt40oC was considered unsuitable (fuzzy
suitability0). A mean monthly rainfall of 80
millimetres for at least 5 months in a year was
defined as suitable.
Each map pixel was assigned a climate
suitability value based on these thresholds
Craig et al., 1999
22
Climate Suitability for Malaria Transmission map
tool
Data Monthly precipitation over land areas on a
0.5 x 0.5 lat/long grid Base Period for
Climatology 1951-2000
1961- wet
1987- dry
23
Climate Suitability for Malaria Transmission
24
Malaria Atlas Project (MAP)
Global distribution of P. falciparum
malaria Estimating populations at risk
Estimated 2.37 billion people are at risk
worldwide

• Input data
• Case incidence data
• medical intelligence
• aridity mask

25
Using maps to define epidemic areas
26
Epidemics present a special problem for malaria
control
Epidemics occur within fringe areas of
transmission typically under 2 types of
environmental conditions
a) Where conditions are arid and vector breeding
sites unavailable for most of the year
(transmission is limited by rainfall) OR b) At
high altitude where temperatures are too low for
most of the year to support vector and parasite
development (transmission limited by temperature)
The recent emergence of malaria at higher
altitudes causing epidemics of highland malaria
has sparked much debate on the possible role of
climate change and global warming in the
aetiology of re-emergent malaria
27
In summary
Rainfall temp
Rainfall (expert opinion)
Temp (altitude)
Climate based predictions
Malaria prevalence ratios
Entomological Inoculation Rates
Malaria admissions
Predictions based on climate empirical data
Lighter shades on the maps show lower malaria
endemicity. For EIR map yellow EIRlt30
greenEIRgt90