Content uploaded by Praveen Kumar Rai
Author content
All content in this area was uploaded by Praveen Kumar Rai on Jul 23, 2015
Content may be subject to copyright.
Informatics in Primary Care Vol 21, No 1 (2013)Informatics in Primary Care Vol 21, No 1 (2013)
Using the information value method in
a geographic information system and
remote sensing for malaria mapping: a
case study from India
Praveen Kumar Rai
Assistant Professor (PGDRS and GIS), Department of Geography, Banaras Hindu University, Varanasi,
Uttar Pradesh, India
Mahendra Singh Nathawat
Professor and Head, School of Sciences, Indira Gandhi National Open University, New Delhi, India
Shalini Rai
SRF, National Bureau of Agriculturally Important Microorganism (ICAR), Mau, Uttar Pradesh, India
ABSTRACT
Background This paper explores the scope of malaria-susceptibility modelling to
predict malaria occurrence in an area.
Objective An attempt has been made in Varanasi district, India, to evaluate the
status of malaria disease and to develop a model by which malaria-prone zones
could be predicted using ve classes of relative malaria susceptibility, i.e. very low,
low, moderate, high and very high categories.
The information value (Info Val) method was used to assess malaria occur-
rence and various time-were used as the independent variables. A geographical
information system (GIS) is employed to investigate associations between such
variables and distribution of different mosquitoes responsible for malaria trans-
mission. Accurate prediction of risk depends on a number of variables, such as
land use, NDVI, climatic factors, population, distance to health centres, ponds,
streams and roads etc.,, all of which have an inuence on malaria transmission or
reporting. Climatic factors, particularly rainfall, temperature and relative humidity,
are known to have a major inuence on the biology of mosquitoes. To produce a
malaria-susceptibility map using this method, weightings are calculated for various
classes in each group. The groups are then superimposed to prepare a Malaria
Susceptibility Index (MSI) map.
Results We found that 3.87% of the malaria cases were found in areas with a low
malaria-susceptibility level predicted from the model, whereas 39.86% and 26.29%
of malaria cases were found in predicted high and very high susceptibility level
areas, respectively.
Conclusions Malaria susceptibility modelled using a GIS may have a role in
predicting the risks of malaria and enable public health interventions to be better
targeted.
Keywords: epidemiology, geographical mapping, malaria, public health,
spatial analysis
Cite this article: Rai PK., Nathawat MS., Rai S.
Using the information value method in a geographic
information system and remote sensing for malaria
mapping: a case study from India. Inform Prim Care.
2013; 21(1):43–52.
http://dx.doi.org/10.XXXX
Copyright © 2013 The Author(s). Published by
BCS, The Chartered Institute for IT under Creative
Commons license http://creativecommons.org/
licenses/by/2.5/.
Author address for correspondence:
Assistant Professor (PGDRS and GIS),
Department of Geography,
Banaras Hindu University,
Varanasi-221005, Uttar Pradesh, India
Email: rai.vns82@gmail.com
Accepted December 2013
Informatics in Primary Care Vol 21, No 1 (2013)
Rai Geographic information system and remote sensing for malaria mapping in Varanasi, India 44
INTRODUCTION
Malaria remains one of the greatest killers, particularly
in the developing world.1 Malaria transmission depends
on diverse factors that all have an inuence on the vec-
tors, parasites, human hosts and the interactions among
them. These factors may include, among others, meteoro-
logical and environmental conditions.2 The most apparent
determinants are observed to be the meteorological and
environmental parameters such as rainfall, temperature,
humidity and vegetation type and cover.3,4 There are only a
few examples of the application of epidemiological maps in
malaria control and this may be explained by a lack of suit-
able, spatially dened data and also by a relatively incom-
plete understanding of how epidemiological variables relate
to disease occurrence. Recent evidence suggests that
clinical manifestations of infection are determined by the
intensity of parasite exposure. Developments in the area
of geographical information systems (GISs) can provide
new ways to represent epidemiological data spatially.5 GIS
software is being used to correlate the climatic attributes
of the collection localities with the presence or absence
of various mosquitoes.6 This technology has been in exis-
tence for a number of years, but it is only recently that it has
been widely accepted as a powerful tool to augment exist-
ing monitoring and evaluation methods.3,7 A GIS technique
integrated with remote sensing can play a variety of roles in
the planning and management of a dynamic and complex
healthcare system, disease mapping, public health and epi-
demiology,8,9 although it is still at an early stage of integra-
tion into public healthcare planning.20
Vegetation has often been found associated with the vector’s
breeding, feeding and resting locations. A number of vegeta-
tion indices have been used in remote sensing and earth sci-
ence disciplines. The most widely used index is the Normalized
Difference Vegetation Index (NDVI). It is dened as the dif-
ference between red and near-infrared bands normalised by
twice the mean of these two bands.10 For green vegetation,
the reectance in the red band is low because of chlorophyll
absorption, while the reectance in the near-infrared band is
high because of the spongy mesophyll leaf structure.11
Mathematical and statistical modules embedded in GISs
enable the hypotheses to be tested and the estimation, explana-
tion and prediction of spatial and temporal trends.12,13 Statistical
techniques model the relation between parasite exposure risk
and environmental risk factors via a multivariate linear regres-
sion model. Such a model can also be used for prediction.14,15
The purpose of malaria-susceptibility modelling is to exam-
ine the long-term parameters, and to dene their relations to
malaria occurrence in the studied area. The primary goal has
been set for assessment and appraisal of important param-
eters for incidence of malaria disease in the study area and,
then, selection of the most determinant parameters. The study
area is Varanasi district, Uttar Pradesh, India, extending
between 25°10′ N and 25°37′ N latitude and 82°39′ E and
83°10′ E longitude. The main expanse is towards the west and
north of Varanasi city and spreads over an area of 1454.11 km2
(Figure 1). Administratively, the study area comprises two tah-
sils, namely Pindra and Varanasi Sadar, which are further
sub-divided into eight development blocks, namely Baragaon,
Pindra, Cholapur, Chiraigaon, Harhua, Sevapuri, Araziline and
Kashi Vidapeeth, consisting of 1336 villages altogether.
Figure 1 Study area as viewed on satellite data (IRS-1C LISS III)
Informatics in Primary Care Vol 21, No 1 (2013)
Rai Geographic information system and remote sensing for malaria mapping in Varanasi, India 45
OBJECTIVE
The main aim of the study was to develop a malaria-suscep-
tibility model using the information value (Info Val) method
with the help of remote-sensing data and GIS techniques.
METHODS
A number of thematic maps (referred to as data layers in GIS)
were generated on specic parameters related to the conse-
quences of malaria, that is land use, NDVI, distance to water
bodies (such as ponds, rivers etc.), roads and health centres,
rainfall and temperature data and projected population den-
sity in the year 2009. In this study, Ilwis Version 3.4, ArcGIS
Version 9.3 and ERDAS Imagine Version 9.1 software as
well as the statistical software SPSS Version 16 were used
to produce the layer maps, which help in the production of
the malaria-susceptibility maps. We used a 1:50,000 scale
topographical map of the study area to digitise the district and
development block boundaries. The coordinates of important
geo-reference points such as road junctions, malaria-prone
areas and existing healthcare facilities were measured dur-
ing eld surveys using global positioning system (GPS) tech-
nology. During measurement, one receiver served as a base
station, while the other collected GPS data at the selected
ground control points. To establish the relationship between
object space and image space, ground control points selected
in the model area to conduct all measurements in the National
Coordinate System. The vector maps were developed from the
IRS-1C LISS-III 2008 remote sensing data and Survey of India
(SOI) topographical map. A land use map, NDVI and vector
layers of water bodies and other important parameters were
delineated in ERDAS Imagine Version 9.1 and ArcGIS Version
9.3 software. A geo-referenced digital map of the development
blocks of Varanasi district was used on the GIS platforms.
The Info Val method was used in this study to produce a
malaria-susceptibility zone (MSZ) and malaria-susceptibility
index (MSI). Other inuential parameters considered in this
study for optimum zonation and modelling of the study area
were rainfall (Rf), temperature (Temp), population density
(Pd), distance to river (Dri), distance to road (Dro), distance
to health facilities (Dhf), land use/land cover (Lc) and NDVI.
These parameters have been used in all three of the methods
mentioned above to produce MSZ of the study area.
Info Val
The Info Val method for MSZ considers the probability of
malaria occurrence within a certain area of each class of a
thematic.16 In this model, weights of a particular class in a
thematic re determined as
Wi= ln
(
Densclas
Densmap
)
= ln
Npix(S
i
)/Npix(N
i
)
Σ
n
i= 1 Npix(Si)/Σ
n
i= 1 Npix(Ni)
(1)
where Wi is the weight given to the ith class of a particular
thematic layer, Densclas is the malaria density within the
thematic class, Densmap is the malaria density within the
entire thematic layer, Npix(Si) is the number of malaria pix-
els in a certain thematic class, Npix(Ni) the total number of
pixels in a certain thematic class and N is the number of
classes in a thematic map. The natural logarithm is used
to take care of the large variation in the weights. Thus, the
weight is calculated for various classes in each thematic.
The thematic is overlaid and added to prepare a MSI map.
For near-equal subdivision of the MSI, the cumulative fre-
quency curve was categorised into ve zones based on
malaria susceptibility (that is very high, high, moderate, low
and very low). Information analysis includes two specic
steps, that is bivariate analysis and multivariate analysis
(Figure 2).
Bivariate analysis
In this analysis, two ratios are used:
a) P is the ratio of malaria area and study area
P=
M
N
(2)
where M is the number of the malaria cases in the
study area and N is whole of the study area;
b) Pi is the ratio of the malaria area in the individual
parameters:
Pi =
Mi
Ni
(3)
where Mi is the number of the malaria cases into
the ith variable and Ni is whole of the study area
including the ith variable.
Data Gathering Bivariate AnalysisMulti-Variate Analysis
Malaria Zonation
Malaria Map
Making Factors Maps
Figure 2 Different stages involved in constructing the malaria-susceptibility zone (MSZ) map using the
information value (Info Val) method
Informatics in Primary Care Vol 21, No 1 (2013)
Rai Geographic information system and remote sensing for malaria mapping in Varanasi, India 46
Then the relation between these two values, the Pi/P infor-
mation value, can be calculated from the ith variable in the
predicted malaria potential characterised with Mi:
Mi =
Pi
P
=
Mi/Ni
M/N
(4)
Then for each variable and its sub-group, the Mn of each
Mi, which indicates positive and negative zones, has been
calculated. If the calculated Mn are positive, this indicates
that the pixels including the i variable have a greater inci-
dence of malaria than mean of the study area. This indi-
cates the susceptibility of these parameters to instability.
Negative values indicate stability, meaning no presence of
malaria pixels.
Multivariate analysis
After producing thematic maps by interpolating the results
of each variable information value, we divided the maps into
sample areas of 200 × 200 pixels. Then the numerical values
results, which included 38,622 samples, have been trans-
ferred to EXCEL software and the nal information value has
been calculated as follows:
Ij =Σ xji
·
li =Σ xji
·
Mi/Ni
M/N
(5)
where j = 1, 2, …, n indicate the number (area) of networks,
i = 1, 2, …, n indicate the number of variables, xji is the quan-
tity of the ith variable in the jth indicator, where i = 1 means
malaria is present i = 0 indicates no malaria, and Ii is the
information values resulting from the ith variable.
After statistical calculation of the model, information result-
ing from the model is transferred to the GIS (ILWIS 3.4) and
the MSI map was created.
The next step in this method is to determine the quantities
of crucial information values and to divide the map according
to the degree of susceptibility, which is based on the calcu-
lated values.
Malaria inventory map
A malaria inventory map identies denite and probable
areas of existing malaria prevalence and is a basic require-
ment for MSZ. The malaria inventory map shows the spatial
distribution of malaria as points or to scale.
Malaria inventory maps are often used as the basis for
other MSZ techniques or as an elementary susceptibility map.
Village-wise malaria location data were collected from each
primary health centre (PHC) and then the locations were deter-
mined using GPS. GPS location data related to malaria preva-
lence were imported into the GIS platform and 500 m buffer
zones were created around each point (Figure 3). On the basis
of these malaria pixels falling in the study area, the pixels from
the whole study area were assigned one of two values, that is
0 (no malaria pixels) 1 (where malaria pixels are present).
RESULTS
Malaria-inuencing data layers and their map
preparations
A representation of the malaria database is shown in Figure 4.
The following layers, that is rainfall (Figure 4(a)), average
Figure 3 Buffer zones (500 m) around the location of each point of malaria prevalence
Informatics in Primary Care Vol 21, No 1 (2013)
Rai Geographic information system and remote sensing for malaria mapping in Varanasi, India 47
Figure 4 Representation of the malaria database: (a) rainfall; (b) average temperature; (c) population density;
(d) distance to rivers/streams; (e) distance to roads; (f) distance to healthcare facilities; (g) distance to ponds;
(h) NDVI; (i) land use; (j) and status of malaria in 2009
Informatics in Primary Care Vol 21, No 1 (2013)
Rai Geographic information system and remote sensing for malaria mapping in Varanasi, India 48
temperature (Figure 4(b)), population density (Figure 4(c)),
distance to rivers/streams (Figure 4(d)), distance to roads
(Figure 4(e)), distance to healthcare facilities (Figure 4(f)),
distance to ponds (Figure 4(g)), NDVI (Figure 4(h)), land use
(Figure 4(i)) and status of malaria in 2009 (Figure 4(j)), are
used to produce the malaria-susceptibility model map.
Rainfall
Rainfall is considered to be the most important malaria-trig-
gering parameter causing soil saturation and a rise in pore
water pressure. However, there are not many examples of
the use of this parameter in stability zonation, probably due
to the difculty in collecting rainfall data for long periods over
large areas.
After interpolation between the amounts of annual rainfall
in the study area stations, the isohyets map was created.
Finally, this map has been grouped into ve classes to pre-
pare the rainfall data layer (Figure 4(a)). It is veried that
approximately 57.77% of the malaria cases occurred in areas
with >984 mm rainfall, but in areas with <970 mm rainfall only
very low and moderate zones of malaria were found, with only
3.19% of cases. From this we can conclude that increasing
amounts of rainfall increases the malaria breeding sources
(Table 1, under column A).
Temperature
To take into account the relationship between temperatures
and malaria transmission, the temperature data are gathered
for different periods. The temperature distribution map has
been grouped into three main classes, that is 35.44–35.46°C,
35.47–35.49°C and 35.50–35.52°C (Figure 4(b)). In Table 1,
under column B, it is found that in the study area malaria vec-
tors were highly developed in the 35.44–35.46°C tempera-
ture category: 56.50% of the malaria-prone area pixels were
found in this category, whilst only 20.04% of malaria-prone
area pixels were found in the 35.47–35.49°C category.
Population density
The overall population distribution in the district is closely
related to the physical and sociocultural factors. Population
distribution is a dynamic process which manifests the varying
nature of man’s adjustment to physical resources. Population
density has been encountered under various typological pur-
views to reveal different aspects of population distribution.
Census data from the year 2001 is also used and using this
population data, the projected population for the year 2010
was calculated which used to calculate population density,
with the area divided into ve categories on the basis of this
population density, that is very low, low, medium, high or very
high (Figure 4c). In Table 1, under column C it is found that
in very low and low population density areas especially in the
rural parts of Varanasi district very high malaria prevalence
(33.74%) is identied, but where population density is high
and very high, only 4.83% and 14.23% respectively of the
malaria area pixels were found. Most of the very high malaria
zone is found in the Varanasi city area where the projected
population density is very high.
Distance to rivers/streams
One of most important parameters that contribute to an
increase in malaria parasites and malaria disease is the dis-
tance to rivers/streams. The proximity of the populated area to
drainage structures is an important parameter for malaria vec-
tor breeding sources. Streams may also adversely affect those
in low-lying areas, especially villages and settlement areas
near the Varuna River. A thorough eld investigation was car-
ried out to determine the effects of rivers/streams on malaria
prevalence. The malaria area percentage in each buffer zone
is given in Table 1, under columnD. This shows that 23.23%
of the malaria area is closely located within the <1000 m buffer
zone (Figure 4(d)). We identied 31.32% of the malaria area
pixels in the buffer zone of 3000–6000 m. In this study, we
found that 3.56% of the malaria area pixels are found >10,000
m from a river/stream and in this zone only very low and mod-
erate categories are available, which is mainly because of
the inuence of some of the other indicators/variables, so at
this distance malaria indicators or breeding sources have little
inuence on people.
The important thing result found in this study is that as
the distance to rivers/streams increases, the percentage of
malaria-affected area pixels decreases.
Distance to roads
The distance to roads is also an important parameter as it
can be used as an estimate of the access to existing health-
care facilities in the study area. Different buffer areas are cre-
ated on the path of the road to determine the effect of the
road on malaria prevalence (Figure 4(e)). The malaria area
pixel percentage in each buffer zone is given in Table 1, ,
under columnE and shows that 65.84% of the malaria area
pixels are closely located within the <300 m buffer category,
whereas only a very nominal 0.92% of the malaria area is
found in the buffer zone of 2000–3000 m. Only 0.41% of
malaria pixels are located in the >3000 m buffer category.
Here it can be seen that as the distance to roads increases,
the malaria area percentage shows a decreasing trend.
Distance to health facilities
The health facilities of Varanasi district are based on mainly
modern allopathic treatment methods. To nd the distribu-
tional pattern of healthcare facilities, data has been collected
from the chief medical ofcer (CMO) ofce and government
hospitals located in rural areas of Varanasi district. The
existing health facilities both in rural and urban areas were
surveyed with the help of a differential global positioning sys-
tem (DGPS). There are different categories of health centre
providing infrastructure and treatment in the district. The
PHC’s are dotted around the district located at an interval of
10–20 km and the tahsil hospitals are located about 50 km
apart. The hierarchical distribution of medical centres in the
district bears a close relationship with the hierarchy of central
locations and the population size of the settlement. In addi-
tion, the transport network has also inuenced the growth of
healthcare facilities. The percentage of malaria area pixels is
very much related to the distance from healthcare facilities.
Informatics in Primary Care Vol 21, No 1 (2013)
Rai Geographic information system and remote sensing for malaria mapping in Varanasi, India 49
Parameters for malaria mapping Number of malaria pixels Total number of pixels Malaria pixel area (%)
A. Rainfall class
<970 7897 19,519 6.38
970–973 19,682 68,937 15.89
973–976 9542 99,487 7.7
976–979 15,175 143,285 12.25
>984 71,547 280,162 57.77
B. Temperature class
35.44–35.46°C 69,974 266,938 56.5
35.47–35.49°C 24,812 234,152 20.04
35.50–35.53°C 29,057 110,300 23.46
C. Population density
Very low 41,787 222,199 33.74
Low 41,584 212,569 33.58
Moderate 16,861 53,352 13.61
High 5985 84,805 4.83
Very high 17,626 38,460 14.23
D. Distance to rivers/streams
<1000 m 28,766 132,325 23.23
1000–3000 m 36,354 173,300 29.35
3000–6000 m 38,783 184,054 31.32
6000–10,000 m 15,533 80,756 12.54
>10,000 m 4407 40,955 3.56
E. Distance to roads
<300 m 81,541 390,258 65.84
300–1000 m 34,954 177,295 28.22
1000–2000 m 5706 38,918 4.61
2000–3000 m 1134 3499 0.92
>3000 m 508 1420 0.41
F. Distance to healthcare facilities
0–1000 m 8158 29,136 6.59
1000–3000 m 21,720 145,065 17.54
3000–6000 m 43,515 242,719 35.14
6000–10,000 m 39,376 147,357 31.8
>10,000 m 11,074 47,113 8.94
G. Distance to ponds
<500 m 67,032 27,0681 50.45
500–1500 m 48,608 283,869 36.58
1500–3000 m 15,909 39,612 11.97
>3000 m 1319 17,228 0.99
H. NDVI
–0.288 23,909 115,299 19.31
0–0986 99,934 496,053 80.69
I. Land use class
Water bodies 2690 19,282 2.17
Agriculture 65,188 330,581 52.64
Settlement 25,810 114,660 20.84
Vegetation 21,595 17.44
Fallow land 8544 46,601 6.9
Table 1 Malaria database showing the characteristics of malaria based on different parameters
Informatics in Primary Care Vol 21, No 1 (2013)
Rai Geographic information system and remote sensing for malaria mapping in Varanasi, India 50
In Table 1, under column F, it is found that 6.59% of malaria
area pixels belong in the 0–1000 m buffer distance to health-
care facilities and 35.14% of malaria area pixels are found in
the 3000–6000 m buffer distance (Figure 4(f)).
Table 1, under column F shows that as the distance to
healthcare facilities increases, the malaria prevalence also
increases, except in >10,000 m buffer zone (8.94% malaria
area pixels). In these areas it may be that malaria breeding
sources are not as developed as in other areas.
Distance to water ponds
The locations of water ponds in the study area were extracted
with the help of IRS-1C LISS III satellite data from 2008.
Five different buffer areas were created for the water ponds to
determine the effect of the distance to water ponds on malaria
prevalence (Figure 4(g)). In Varanasi, there used to be many
ponds and tanks dating back to ancient times. In addition to
serving as the holy places for holding Hindu religious rituals,
they also played an important role in rainwater collection and
thereby served as sources for ground water replenishment.14
However, due to the rapid increase of the population, most of
these ponds have been wiped off the map of Varanasi com-
pletely or are rapidly deteriorating. The main source of pol-
lution in the ponds is heaps of garbage. The solid and liquid
wastes generated from household and industrial activities
are dumped and released into uncontrolled sites. These sites
leak into the low-lying areas where the tanks and ponds are
located and due to this malaria vectors develop very easily
and many cases of malaria are found near to these polluted
ponds. We found that 50.45% of malaria area pixels occurred
in the <500 m buffer category of ponds and only 0.99% of
malaria area pixels were found in the zone of >3000 m buffer
category of ponds. Table 1, under column G reveals that as
the distance to ponds or water bodies increases, the percent-
age of malaria area pixels decreases.
NDVI
Vegetation is often associated with vector breeding, feed-
ing and resting locations. Because malaria is vector-borne,
there are many remotely sensed abiotic and biotic environ-
mental variables that are relevant to the study of malarial
transmission and habitat niches of the vector. A number of
vegetation indices have been used in remote sensing, but
the most widely used index to enhance the vegetation areas
and crop elds is NDVI. NDVI values range from -1 to +1,
with higher values indicating denser vegetation. The higher
the NDVI value, the denser the vegetation. Many diseases
and their causative agents possess environmentally linked
attributes that must be present for transmission or infection
to occur. NDVI and remotely sensed variables provide addi-
tional methods of exploring and better dening these attri-
butes. Distributions of diseases associated with arthropod
and gastropod vectors, classiable as either intermediate or
denitive hosts depending upon the presence or absence of
sexual reproduction of the agent while hosted, are denable
by their landscape features such as land use, land cover and
proximity to aquatic habitats.
The NDVI map has been grouped into three main classes
and in this study it is found that 19.39% of malaria area pixels
are found in the –0.288 to 0 categories (Figure 4(h)). In Table
1, under column H, it is shown that 80.69% of malaria area
pixels are found in the 0–0.986 category, which is the class of
agriculture, vegetation and fallow land.
Land use/land cover
Land use/land cover information is also a very important
parameter used to calculate the malaria-susceptibility map and
calculate the MSZ using a multilinear regression model.14 The
land use/land cover map of the study area has been prepared
from the IRS-1C LISS III remote-sensing data from 2008. The
land use map is prepared in the image processing platform to
highlight ve main classes that is agricultural elds, settlement,
vegetation, water bodies and fallow land. In this study we found
that agriculture and vegetation are very important parameters
and play important roles as malaria vector breeding sources.
Areas which include dense vegetation provide favourable
conditions for malaria vectors. The presence of crop elds,
especially in those areas where rice cultivation is dominant, is
also crucial as a malaria vector breeding source. Many parts
of Varanasi district have fertile agricultural elds and wherever
irrigation facilities are very good, farmers cultivate rice.
This area, with good crop elds, should be prone to the
occurrence of malaria in some cases (Figure 4(i)). When
using land cover and malaria maps to determine the distribu-
tion of malaria, according to the land cover classes, 52.64%
of malaria affected pixels out of the total pixels occurred in
the agriculture class whereas 20.84%, 17.44%, 6.90% and
2.17% of the malaria area pixels occurred in the settlement,
vegetation, fallow land and water bodies categories, respec-
tively (Table 1, under column I).
Summary of results
As shown in Figure 5(A), the distribution of malaria area pixels
in the information value 0.1–0.6 is looking sensitive: 29.73%
of the pixel area has quantities greater than this amount, so
this value can be dened as a crucial value for malaria. Pixel
networks having information values of more than 0.1 based
on malaria area percentages have been divided into two
groups, that is high and very high susceptibility, and pixels
having lower than 0.1 information value are divided into three
levels of low, very low and moderate susceptibility (Table 2
and Figure 5(B)). Table 2 also highlights that 3.87% of the
malaria area pixels are in very low malaria susceptibility level
whereas 39.86% and 26.29% of the malaria area falls into the
high and very susceptibility levels respectively.
DISCUSSION
Principal ndings
By applying and integrating the Info Val weights using ArcGIS
and ILWIS software, a continuous scale of susceptibility index
is generated with which the study area can be divided into ve
classes of malaria susceptibility. A reliable and accurate suscep-
tibility map depends on the inclusion and proper determination
Informatics in Primary Care Vol 21, No 1 (2013)
Rai Geographic information system and remote sensing for malaria mapping in Varanasi, India 51
of the role of these parameters. Info Val is used for malaria
modelling and to calculate the optimum model for the MSI and
for identifying MSZs. The distribution of malaria areas with
the information values of 0.1–0.6 is sensitive, as 29.73% of
malaria areas have quantities greater than this amount, so this
value can be dened as a crucial value for malaria. Also, whilst
3.87% of the malaria pixel area has very low malaria suscepti-
bility, 39.86% and 26.29% of the malaria pixel area falls in the
high and very susceptibility classes respectively.
Implications of the ndings
The aim of malaria susceptibility modelling is to use known
risk factors and their relations to dene malaria occurrence
in the study area. This method may be helpful in improving
access the healthcare facilities in areas affected by malaria.
Early detection and prompt response measures may be facili-
tated though improved surveillance and allow timely remedial
measures to be used.
Comparison with the literature
GIS will continue to play a signicant role in the reorgan-
isation of public health and disease planning, especially in
response to the sweeping changes taking place in the han-
dling of the health-related information.17 It has been noted
that GIS and remote sensing play important roles in the clear
interpretation of the malaria-related parameters used in this
study. GIS has shown its capability to answer a diverse range
of questions relating to the key goals of efciency, effective-
ness and equity in the provision of public health services.18,19
Limitations of the study
Accurate data collection related to healthcare facilities and
disease data from the government hospitals are the major
problems faced in this study. The probability values estimated
in these kinds of predictive methods are not absolute and rep-
resent a relative degree of susceptibility. However, they can
provide an appropriate and valid measure of malaria with the
limitation that knowledge of past malaria information affects
the nal probability values calculated by this method. It is
possible to use other statistical methods for model validation.
Call for further research
There is clear evidence that the application of GIS and
remote sensing integrated with statistical methods for health
and disease mapping can play a major role in public health
management and disease surveillance. Statistical methods
like multiple linear regression, heuristic approach, and logis-
tic regression methods can also be used for further malaria
mapping and to calculate the optimum model for MSI and
MSZ. Based on the optimum model for MSI and MSZ, we
can also calculate the hospital requirement index (HRI) and
hospital requirement zones (HRZ).
Malaria level Total number
of pixels Pixel (%) Malaria area
(km2)Malaria area
(%)
Very Low 58,782 9.61 47.95 3.87
Low 143,609 23.49 149.61 12.08
Moderate 149,494 24.45 221.53 17.89
High 181,759 29.73 493.70 39.86
Very High 77,746 12.72 325.64 26.29
Total 611,390 100 1238.43 100
Table 2 Status of malaria area percentage versus malaria level based on the information value method
Figure 5 (A) Malaria-susceptibility index (MSI) and (B) malaria susceptibility zone (MSZ) based on the information
value method
Informatics in Primary Care Vol 21, No 1 (2013)
Rai Geographic information system and remote sensing for malaria mapping in Varanasi, India 52
CONCLUSION
GIS tools can help predict where malaria is more or less likely to
occur. The results provide face validity for known predictors of
the disease. Such tools might be used by public health agencies
to focus on where interventions are most likely to be needed.
Acknowledgements
The authors are grateful to Mr Mohhamad Onagh, Research
Scholar, Department of Geography, BHU, Mr M. A. Khan,
District Malaria Ofcer, Varanasi, India for supporting the
data collection and nally Dr Kshitij Mohan, Department of
Geography, Banaras Hindu University.
1. Kaya S, Pultz TJ, Mbogo CM, Beier JC and Mushinzimana
E. The use of radar remote sensing for identifying environ-
mental factors associated with malaria risk in coastal Kenya.
International Geoscience and Remote Sensing Symposium,
Toronto, ON, 2002.
2. Saxena R, Nagpal BN, Srivastava A, Gupta SK and Dash AP.
Application of spatial technology in malaria research and con-
trol: some new insights. Indian Journal of Medical Research
2009; 130(2): 125–32.
3. Connor SJ, Flasse SP, Perryman AH and Thomson MC. The
contribution of satellite derived information to malaria stratica-
tion, monitoring and early warning. World Health Organization
mimeographed series, WHO/MAL/1997; 1079.
4. Craig MH, Snow RW and Le Sueur D. A climate-based distri-
bution model of malaria transmission in sub-Saharan Africa.
Parasitology Today 1999, 15: 105–11.
5. Omumbo JA, Hay SI, Snow RW, Tatem AJ and Rogers DJ.
Modelling malaria risk in East Africa at high spatial resolution.
Tropical Medicine and Internal Health 2005; 10(6): 557–66.
6. Mullner RM, Kyusuk C, Croke KG and Menash EK. Geographical
information systems in public health and medicine. Journal of
Medical Systems 2004; 28(3): 215–21.
7. Sudhakar S, Srinivas T, Palit A, Kar SK and Battacharya SK.
Mapping of risk prone areas of kala-azar (Visceral leishmaniasis)
in parts of Bihar state, India: an RS and GIS approach. Journal of
Vector Borne Disease 2006; 43: 115–22.
8. Joyce K. To me it’s another tool to help understand the evi-
dence: Public health decision-makers’ perceptions of the value
of geographical information system (GIS). Health and Place
2009; 15: 831–40.
9. Rytkönen Mika JP. Not all maps are equal: GIS and spatial
analysis in epidemiology. International Journal of Circumpolar
Health 2004; 63: 9–24.
10. Messina JP and Crews-Meyer KA. The Integration of remote
sensing and medical geography: process and application. In:
Albert DP, Gesler WM and Levergood B (Eds), Spatial Analysis,
GIS, and Remote Sensing Applications in the Health Sciences.
Chelsea, MI: Ann Arbor Press, 2005, 156.
11. Hoek W, Konradson F, Amersinghe PH, Perara D, Piyaratne
MK and Amerasinghe FP. Towards a risk map of malaria in Sri
Lanka: the importance of house location relative to vector breed-
ing sites. International Journal of Epidemiology 2003; 32: 280–5.
12. Kulldorff M. Spatial scan statistics: model, calculations and
applications. Scan Statistics and Applications 1999; 15:
303-302.
13. Lawson AB. Disease Mapping: Basic approaches in new devel-
opments. In: Maheswaran R and Cragilla M (Eds), GIS in Public
Health Practice. New York: CRC Press, 2001; 31-49.
14. Rai PK, Nathawat MS and Onagh M. Application of multiple
linear regression model through GIS & Remote Sensing for
malaria mapping in Varanasi district, India. Health Science
Journal 2012; 6(4): 731–49.
15. Riedel N, Vounatsou P, Miller JM, Gosoniu L, Chizema-Kawesha
E, Mukonka V and Steketee RW. Geographical patterns and
predictors of malaria risk in Zambia: Bayesian geostatistical
modelling of the 2006 Zambia national malaria indicator survey
(ZMIS). Malaria Journal 2010; 9: 37.
16. Saha AK, Gupta RP and Arora MK. GIS based landslide hazard
zonation in Bhagirathi, Ganga valley, Himalayas. International
Journal of Remote Sensing 2005; 23: 357–69.
17. Donald PA, Wilbert MG and Barbara L. Spatial Analysis, GIS
and Remote Sensing Application in the Health Sciences.
Chelsea, MI: Ann Arbor Press, 2006; 185.
18. Boscoe FP, Ward MH and Reynolds P. Current practices in
spatial analysis of cancer data: data characteristics and data
sources for geographic studies of cancer. International Journal
of Health Geography 2004; 97: 14041–3.
19. Kleinschmidt I, Bagayoko M, Clarke GPY, Craig M and le
Sueur D. A spatial statistical approach to malaria mapping.
International Journal of Epidemiology 2000; 29: 355–61.
20. Sweeney AW. A spatial analysis of mosquito distribution. GIS
Use 1997; 21: 20–21.
REFERENCES