Vector-Borne Disease Monitoring
Although some remarkable successes have been achieved in the battle against vector-borne diseases,
such as the Onchocerciasis Control Program in West Africa, the threats to health from
vector-borne parasitic diseases are at least as important now as at any time in history. This is
illustrated by the following two examples.
The incidence of visceral leishmaniasis has risen worldwide, and southern Sudan is currently experiencing a major epidemic (WHO 1993). The disease's sand-fly vector is associated with a particular woodland type. Changes in the distribution of tree cover in Sudan as a result of ecological and/or developmental processes may have important consequences for the spread of this disease.
Malaria remains one of the greatest killers of young children in the developing world. It also imposes a considerable economic burden resulting from high morbidity levels within the adult population. Although statistics on the scale of the disease vary considerably, it is inevitably the poorer sections of the poorest nations that suffer most. The World Health Organization estimates some 110 million clinical cases of malaria worldwide per year - over 80% of these occur in Africa south of the Sahara. Malaria is transmitted by anopheline mosquitoes, which breed in surface water pools where environmental conditions are suitable for both vector and parasite development.
Both spatial and temporal changes in environmental conditions may be important determinants of vector-borne disease transmission. Capable of identifying these changes, satellite imagery may be able to define and predict areas and periods of high transmission. Thus the potential for using remotely sensed images for monitoring and evaluating the factors associated with tropical diseases clearly merits further investigation and research.
In 1985, the National Aeronautics and Space Administration (NASA) initiated the Biospheric Monitoring and Disease Prediction Project, the aim of which was to determine if remotely-sensed data could be used to identify and monitor environmental factors that influence malaria vector populations. Initial studies supported by this project used high resolution images from the LANDSAT (American commercial Earth observation satellite) TM sensor to monitor the development of canopy cover in Californian rice fields. Changes in rice canopy cover over the season were successfully used to predict fields with high or low mosquito densities (Wood et al. 1991).
Unfortunately, the use of high resolution satellite imagery is limited by its cost (some US$4000-5000 per image at a scale of 60-180 km) and temporal resolution (16 days per repeat for the LANDSAT, and 26 days for SPOT [French commercial Earth observation satellite]). Poor temporal resolution is a serious limitation as cloud cover may obscure a "scene" (area of the image) frequently enough as to allow only one good image of the region per season. Although high resolution images may play an important role in producing baseline land use maps, their high cost and infrequent occurrence limit their use for large-scale temporal monitoring of vector habitats.
Image availability is an important requirement for the use of satellite data in monitoring the environmental variables associated with fluctuations in vector populations. Images must be obtainable with sufficient frequency to allow for comparison with changing vector bionomics, within a biologically meaningful time frame. This means that the images should be available at a cost low enough to make it possible to use a large number of sequential images over the area of interest.
The lower spatial resolution images from the geostationary and polar orbiting satellites (METEOSAT [European meteorological satellite] and NOAA [National Oceanic and Atmospheric Organization]) have been useful for a variety of relevant applications. These include famine early warning systems (FAO 1990), monitoring changes in rainfall conditions and vegetation associated with desert locust swarms (Heilkema and van Herwaarden 1993), assessing deforestation in tropical forests (Malingreau 1991), identifying environmental conditions known to favour tsetse fly reproduction (Rogers and Randolph 1991), and detecting flooding of the breeding sites of Rift Valley Fever vectors (Linthicum et al. 1990). Therefore, any disadvantage of the lower spatial resolution of these satellites may be offset by the benefits of their frequency of observation, low cost, and ease of availability.
Images as Information
Even where satellite images are readily accessible, the data cannot be used immediately. They must
first be analyzed to produce information, which in turn must be presented in such a way to be able
to influence decision-making. The most appropriate facility for achieving this is perhaps a
geographic information system (GIS), which is able to read, process, analyze, and present
spatially-related data for effective interpretation and use for a variety of environmental and
resource management purposes. Within the context of tropical vector-borne disease forecasting
and control, such a system would enable the presentation of temporal and spatial dynamics of the
disease, in a meaningful way, to planners responsible for national and regional control strategies.
Real-time information relevant to potential surges in disease transmission could be included,
enabling the initiation of rapid response strategies.
Africa Real Time Environmental Monitoring Information Systems (ARTEMIS)
One valuable source of satellite imagery has recently become available through the Food and
Agricultural Organisation's (FAO) ARTEMIS program. This program has compiled an archive
data set of NOAA-AVHRR NDVI images for the African continent for the period 1981-1991.
These NDVI (Normalised Differentiated Vegetation Index) images are the product of the red and
near infrared sensors aboard the satellite, and indicate the amount of photosynthetic activity
taking place in the image area with an initial resolution of 1.1 km. The NOAA satellites pass over
the same scene twice daily but have limited data storage available on board. They use the peak
NDVI value to create an "optimal" composite image, known as global area coverage (GAC),
every 10 days, which is then downloaded to the NOAA database. It has been suggested that this
largely removes the problem of cloud cover over a scene.
The data are transformed from GAC (in a process including further subsampling, cloud clearing, and reprojection) into a 10-day maximum value composite (MVC), which minimizes atmospheric interference, sun angle, and viewing geometry effects (Holben 1986). They then have a resolution of 7.6 km on the Hammer-Aitoff map projection.
It has been difficult to obtain reliable NOAA images, for certain areas, following the eruption of Mount Pinatubo in June 1991. This eruption produced tons of aerosol dusts which have seriously affected the transmission properties of the upper atmosphere. Workers at NASA have been busy correcting the data, which will become available in due course.
The NOAA imagery from ARTEMIS, although of spatially coarse resolution, is very useful in terms of its low cost (currently free) and high temporal resolution, allowing for seasonal monitoring of vegetation growth. The ARTEMIS data set is available from FAO's Remote Sensing Centre (Via delle Terme di Caracalla, I 001000 Rome, Italy) on a CD-ROM, and comes complete with both Windows and PC-DOS software, allowing graphic display, preliminary analysis, and subextraction of the data set.
Local Application of Remote Sensing Techniques (LARST)
The LARST program was developed by the Natural Resources Institute (Chatham Maritime, Chatham, Kent, UK ME4 4TB)
of the Overseas Development Administration in collaboration with other scientific groups. The aim of this
program is to promote improved management of natural resources through cost-effective remote
sensing. LARST develops robust, low-cost systems for local reception, processing, and use of
satellite data. Within developing countries, it can be used to inform and assist with practical
decision-making (Sear et al. 1993; Williams and Rosenberg 1993). LARST is currently active in
several countries, such as Algeria, Ethiopia, Indonesia, Kenya, Namibia, Tanzania, Zambia, and Zimbabwe.
These countries are exploring new ways of converting data from the METEOSAT
and NOAA satellites to provide useful environmental information for such areas as agriculture,
fisheries, forestry, water resources, and pest and vector control.
The LARST program is particularly valuable for those interested in using low-cost satellite data to monitor vector habitats, as it helps to provide real-time local-access NOAA NDVI imagery at full spatial and temporal resolution. LARST is able to do this by installing a ground receiving station within range of the scene of interest, enabling direct reception of the full satellite data stream. The local area coverage (LAC) images have a spatial resolution of 1.1 km directly beneath the satellite, and can usually be received at least twice daily (the size of the image pixel is not constant; it increases as the viewing angle increases). The data received from the satellite can then be converted to whichever specific mapping projection is required, a process that involves resampling the data, usually to an image pixel size of 1 km.
It may be possible to set up a LARST system "in country" or in a neighbouring country, through the national meteorological service. Where these facilities are not yet available, it may be possible to put together a case for establishing a minimal operation through donor assistance. The Overseas Development Administration (UK) has to date supported a number of these systems in sub-Saharan Africa.
Rainfall and Vegetation Monitoring
The NOAA-AVHRR NDVI images have been shown to be highly correlated with the Cold Cloud
Duration (CCD) images from METEOSAT. The combination of these two images provides a
good, proxy indicator of effective rainfall in areas receiving less than 1000 mm rainfall each year
(Dugdale and Milford 1988; Bonifacio et al. 1993).
They are particularly useful where ground-based meteorological stations are few and/or poorly dispersed. The availability of such images would be very useful for seasonal studies of the environmental factors associated with vector-borne diseases, such as rainfall, river flooding, changes in the water table, vegetation growth, and the location of vector breeding sites.
Geographic Information Systems
Geographic information systems (GIS) are defined here as computer-based systems for entering,
storing, analyzing, and displaying digital geo-referenced data sets. The problems arising from the
use and development of GIS in developing countries, and some
advice on their solutions, are outlined by Hastings and Clark (1991), in a special issue of the
International Journal of Geographical Information Systems devoted to the
consideration of GIS within a developmental context. Rajan (1991) has, with the support of the
Asian Development Bank, produced a valuable overview of the use of remote sensing in GIS.
The document includes a useful section covering some of the key issues and problems
surrounding their use, such as increasing commercialization, human resources, institutional
capacity, training, and technology transfer. A consultation report to FAO on the use of GIS in
strengthening information systems for veterinary services in developing countries offers a
comprehensive outline of hardware requirements and software considerations (Perry and Kruska
1993).
At present, GIS are seen primarily as research tools in the field of vector-borne disease; indeed, they will become an increasingly important research tool as geographic databases, models, and analysis procedures continue to develop at a rapid pace. However, the use of GIS in decision support has also become an area of growing interest. Spatially referenced, interactive models have been developed to simulate the broader effects of development policy in Senegal (Engelen et al. 1992; Connor and Allen 1994). A system for the national control of foot and mouth disease in New Zealand (Morris et al. 1993) is probably the best example of how geographic databases and disease epidemiology models can be integrated into a decision support system. GIS designers are also beginning to provide analytical decision support tools as part of their options, and such systems are being promoted for a more participatory planning process (Eastman et al. in press; Hutchinson and Toldano in press).
Of the many GIS available, perhaps the most obvious low-cost choice for the study of the environmental factors associated with vector-borne disease transmission is the raster-based system developed by the IDRISI Project at the Clark Laboratories for Cartographic Technology and Geographic Analysis (Clark University, Worcester, MA, USA 01610-1477). The IDRISI Project is run on a non-profit basis and maintains close developmental links with the United Nations Institute for Training and Research (UNITAR) and the United Nations Environment Programme's Global Resource Information Database (UNEP/GRID). Continuity and compatibility of resource formats shared between different databases is therefore possible. Despite having a high degree of functionality, the software is inexpensive and easy to learn, and consequently has found much use in research, teaching, and training establishments worldwide. The project has a continuing commitment to provide a low-cost, professional-level, modular GIS which is regularly updated in response to user needs, and which enjoys good technical support and a user network. The strengths and weaknesses of the IDRISI software for the following studies are currently being reviewed.
The "Eco-Epidemiology" of Leishmaniasis in Southern Sudan
The Nuer and Dinka people of southern Sudan are currently experiencing what is thought to be
the worst epidemic of visceral leishmaniasis (kala azar) in history. Environmental and
demographic changes and the problems associated with regional conflict exacerbate the conditions
under which this disease spreads, and there is probably nowhere that this is more apparent. The
epidemic began in 1984; the most recent reports have put the death toll at some 200,000.
The sandfly vector of kala azar in southern Sudan is known to be associated with mature Acacia- Balanites woodland (Quate 1964; Schorscher 1991). Howell et al. (1988) found that vegetation maps from 1952, of an area close to the epidemic area, showed a much more extensive distribution of forests than later maps produced in 1980. This loss of woodland was attributed to extensive and prolonged local flooding during the early 1960s, caused by rainfall anomalies in Kenya and Uganda. During a field visit by staff from the Liverpool School of Tropical Medicine, the relatively uniform and young age of the present woodland was noted. It has been suggested that the current epidemic of kala azar is, in part, associated with the recent local maturation of Acacia-Balanites woodlands (Ashford and Thomson 1991).
Operational medical services in the epidemic area require maps showing the current distribution of Acacia-Balanites woodland throughout southern Sudan. These maps are needed to investigate the link between the Acacia-Balanites woodland and known occurrences of the kala azar vector and parasite. Preliminary explorations of the ARTEMIS images from the area, as well as the 1 km NOAA-AVHRR images obtained through the LARST facility in Ethiopia, have shown clear, delineated features believed to be associated with wooded areas observed on the ground.
It is hoped that these features can be identified and defined within the GIS and used as "training sites," which would allow the computer to extrapolate their distributions on a wider scale. Maps produced in this way could then be compared with historical vegetation maps of the same area and changes in canopy cover quantified. The maps could also be used, along with other epidemiological information, to help predict the spread of the current epidemic and to indicate where resources might be best mobilized (Connor and Thomson 1994).
Malaria Transmission Dynamics in the Gambia
The value of LAC NDVI data in explaining spatial variation in malaria transmission in the
Gambia is currently being evaluated. The results of an extensive epidemiological and
entomological survey in 1991 showed that malaria prevalence and intensity varied considerably
from area to area, with children from eastern villages showing the highest malaria parasite and
spleen rates. Significant regional differences were also found in mosquito species composition,
mosquito abundance, sporozoite rate, and human blood index.
Mosquito abundance was highest in the west and central regions; the sporozoite rate and human blood index were highest in the east, where malaria transmission was also more intense. The results of this survey indicated that soil type and proximity of villages to the River Gambia were not correlated with malaria prevalence levels, although they were highly correlated with vector abundance. It is postulated instead that the length of the transmission season is a more important determinant of malaria transmission (Thomson et al. in press). Both the historical GAC NDVI data sets now available and the 1991 3-day composite LAC NDVI data set (captured by the Centre Suivie de Écologique in Dakar, Senegal) can now be used to explore this hypothesis. This information will be incorporated into a model of malaria transmission dynamics for the Gambia the components of which will be identified and analyzed within an appropriate GIS framework.
The ODA's Natural Resources Institute is keen to encourage the wider use of satellite data through their LARST initiatives in the Gambia, which currently receive METEOSAT primary products through a METEOSAT primary data user system. This will permit the production of 10-day cold cloud duration images (a proxy for rainfall). A METEOSAT meteorological data distribution (MDD) receiver system was installed in June 1994, courtesy of the UK Meteorological Office and NRI/ODA. NRI is seeking ways to establish facilities which will enable real-time reception of the NOAA-AVHRR satellite data to assist environmental monitoring and resource management activities in the country (including weather, livestock, tsetse habitat, and factors relating to fisheries resources). Although the METEOSAT data has potential value for predicitng changes in malaria prevalence throughout the country, its value would be increased with the availability of the NOAA-AVHRR imagery.
If a model of malaria transmission dynamics can be developed here, based on information from the satellite images, its application in other semi-arid regions where there is little information from ground-based studies may be particularly useful.
How can such tools inform malaria control policy within the primary health care system? Research on the delivery of a decision support facility using RS and GIS must be simultaneously linked to the process of developing the malaria model. Our concordant project aims to review the potential for using a LARST system in malaria monitoring and control for use by the Ministry of Health and nongovernmental organizations, in planning the most appropriate malaria control practices.
Using RS/GIS for Assessing the Health Impact of Development Projects
The application of remote sensing images to individual development projects or programs raises
further considerations of scale. High resolution images may be useful in mapping the project area
and determining potential hazards (Walter 1994). Low resolution images will be of limited value
over a small area, but it is important to realize that environmental and health impacts of
development programs may not be confined within the project perimeter. They will invariably
have some "upstream" and "downstream" effects.
At a macro level, rainfall throughout the Nile subbasins and Sudan is of crucial importance to natural resources development along the whole course of the river. Studies on the paleobotany of Sudan (Wickens 1975), and historical analysis of the changing nature of rainfall in the region, suggest that changes in rainfall in the middle of this century represent 25% of the total variation over the past 20,000 years (Hulme 1989). Such increasing variability and the resulting ecological instability could have very significant effects on the epidemic nature of vector-borne disease. For example, the anomalous rainfall which destroyed the existing Acacia-Balanites woodlands in the south can be linked to the current Sudanese epidemic. Such rapid climatic change, however, is unlikely to be causally divorced from the regional development process. Arguments have recently been made that climate should no longer be seen as a fixed boundary condition unaffected by social and economic change, but rather should be viewed as a renewable resource, subject to management criteria, and incorporated into the planning cycle.
A broader perspective view of water shed management must then be taken. However, the need to do so has come at a time when, in the case of Sudan, access to information from ground-based rainfall stations has declined drastically. In the south, data are available from only two sites, Juba and Malakal, compared to over 100 sites in the 1950s. Fortunately, experiments with satellite-derived areal rainfall estimates have been shown to be preferable to rain-gauge data for inputs into hydrological models of a whole catchment (Milford 1989).
Back to Contents -
To next section
This file was created 23 February 1996