Use of satellite imagery in constructing a household GIS database for health studies in Karachi, Pakistan
© Ali et al; licensee BioMed Central Ltd. 2004
Received: 16 August 2004
Accepted: 28 September 2004
Published: 28 September 2004
Household-level geographic information systems (GIS) database are usually constructed using the geographic positioning system (GPS). In some research settings, GPS receivers may fail to capture accurate readings due to structural barriers such as tall buildings. We faced this problem when constructing a household GIS database for research sites in Karachi, Pakistan because the sites are comprised of congested groups of multi-storied building and narrow lanes. In order to overcome this problem, we used high resolution satellite imagery (IKONOS) to extract relevant geographic information.
The use of IKONOS satellite imagery allowed us to construct an accurate household GIS database, which included the size and orientation of the houses. The GIS database was then merged with health data, and spatial analysis of health was possible.
The methodological issues introduced in this paper provide solutions to the technical barriers in constructing household GIS database in a heavily populated urban setting.
Geographic data are increasingly being employed in health studies . By studying disease patterns in space, we can understand the relationships between socioecological exposure and illness [2, 3]. Such understanding may help the formulation of need based healthcare systems and health intervention programs. Geographic methods provide a wide spectrum of geographic scales from local to global for analyzing health and health-related data. Regional variation in disease incidence be attributed to regional or global differences in ecological or socio-environmental phenomena . Local-level geographic variation of disease obtained from fine resolution geographic data can provide clues about the spatial variability , and may pinpoint areas where health interventions are needed.
One way to facilitate the measurement of local variation in health outcomes is to create household-level geographic information systems (GIS) database. Household locations can be captured by using GPS (global positioning system) receivers [6, 7]. However precise geographic data on households are an absolute requirement for critical examination of local variation of the disease and its association with the environment . A large variety of GPS receivers are available in the market and different GPS receivers provide different levels of accuracy. A low cost receiver can capture data with an accuracy of 5 to 10 meters provided that they are configured properly and the satellites have good alignment at the time the data are collected . The alignment of the GPS satellite constellation at a particular time can be measured using GPS trip planning software. It is essential that the GPS receiver has a clear "view" of at least four GPS satellites which can be obstructed by large structures such as buildings or mountains. In congested urban settings, collecting household locations in narrow lanes using the GPS can be challenging.
Faced by such challenges we explored satellite imagery in order to acquire household GIS data in urban slums in Karachi, Pakistan. This paper describes the methods used to construct the household GIS database and the technical barriers one might encounter during the construction of a database.
The household geographic information systems project
Geographic studies have been considered as one of the research disciplines of large Vi (antigen) typhoid vaccine effectiveness trials as well as typhoid disease burden studies . The Vi typhoid vaccine provides a comparable degree of protection to the whole-cell type but with less severe side effects. Only one dose is required for a course of vaccination. The studies are part of the Diseases of the Most Impoverished (DOMI) program, a multi-country, multi-disciplinary health research program aimed to accelerate the development and introduction of a new generation vaccines against cholera, typhoid fever, and shigellosis in several Asian countries. The program involves a number of parallel activities including epidemiological studies, social science studies, and vaccine technology transfer. The local collaborator of the household GIS project in Karachi is the Pediatric Department of the Aga Khan University Hospital, Karachi, Pakistan. Technical support for the project was provided by Techno-Consult International, Karachi, Pakistan.
The base map
A commercially available map of Karachi was used as the base map for this GIS project. The base map was georeferenced with four identifiable landmarks using handheld GPS receivers with accuracy of approximately five meters. This accuracy was considered sufficient to identify the study areas, to order satellite imagery, and to conduct subsequent ground surveys. After georeferencing the map, the main geographic features such as roads, hospitals/healthcare centres, and other city landmarks were digitized and incorporated into the baseline GIS database (Figure 1).
The satellite imagery
Image processing and georeferencing
The satellite images were enhanced using an image processing software package (ERDAS Imagine, Atlanta, USA) to facilitate the digitization of house parcel boundaries. High precision, dual frequency GPS units (Trimble 4000 ssi) were used to capture data at several identifiable points on the images to be used as ground control points (GCPs). To transfer images into a GIS database, it must be geometrically rectified to a known coordinate system on the basis of a number of GCPs . Most of the GCPs were selected from the periphery of the study area so that possible errors would converge towards middle of the area. After locating GCPs on the satellite image and identifying them on the ground, GPS readings were obtained at centimeter level accuracy. The GPS data were collected in the WGS-84 (World Geodatic Systems-84) datum in the latitude/longitude system and were subsequently transformed into the Universal Transverse Mercator (UTM) Zone 42-North system. The GCP coordinates within the UTM projection were then integrated with the satellite images using the ERDAS Imagine software for georeferencing. The resultant root mean square (RMS) errors were approximately two meters, which was considered sufficiently accurate for the purpose of constructing the GIS database.
Digitization of house parcels
After completing the ground surveys, the maps were updated using AutoCAD, and the address IDs were added to household parcels in the database. Finally, the household parcel AutoCAD files were imported as polygons into the ArcGIS software package (ESRI Inc., USA). The process included several checks for missing households, duplicate address IDs, and misplacement of address IDs and data were corrected when an error was found. The corrected data were validated by randomly selecting several household parcels (about 2%) from different zones of the study area and verifying their position on ground. At this stage, we observed no discrepancies in the data between ground verification and the satellite based maps suggesting that the household level GIS database is highly accurate.
Implementation of the health GIS (HGIS) study
Entity relationships between data tables are shown as lines, and logical relationships between entities have parentheses around them (Figure 5). In the logical relationship (1,N), "1" indicates each entity should be linked to an entity on the other end, and "N" indicates multiple entities can be linked to an entity at the other end. Similarly, the "0" in (0,1) indicates not all entities will be linked to an entity at the other end. The "1" of the relationship indicates not more than one entity will link to an entity at the other end. For example, the relationship of "member" towards "patient" is shown as (0,1). Here "0" indicates not all records in "member" are to be linked in "patient," and "1" indicates not more than one record of the "member" can be linked to a record in "patient". Similarly, "1" in the relationship (1,N) of "patient" towards "member" indicates all records in "patient" should be linked to "member', and "N" indicates multiple records in "patient" can be linked to a record in the "member".
In this paper, we have outlined methodological issues involved in the construction of a household GIS database using satellite-based technology in a situation where the GPS was not appropriate. To our knowledge, this approach has never been reported, but may offer greater value in constructing household GIS databases compared to that based on GPS. Our household GIS offers size and orientation of individual houses in dense urban environment. Such database can be instrumental in health and disease studies because they facilitate the integration of socioecological and environmental factors that may influence health. Future health studies may benefit by using satellite-based technology to construct household GIS databases.
This work was supported by the Diseases of the Most Impoverished Program, funded by the Bill and Melinda Gates Foundation and coordinated by the International Vaccine Institute. The authors wish to thank Drs. Michael Emch and Lorenz von Seidlein for their valuable comments and suggestions in improving this paper.
- McGlashan ND: Medical geography: an introduction. In Medical geography: techniques and field studies. Edited by: McGlashan ND. 1972, London: Methuen, 3-15.Google Scholar
- Hunter JM: The geography of health and disease. In The challenge of medical geography. Edited by: Chapel Hill. 1974, NC: University of North Carolina, 1-3.Google Scholar
- Jones K, Duncan C: Individuals and their ecologies: analysing the geography of chronic illness within multilevel modeling framework. Health & Place. 1995, 1 (1): 27-40. 10.1016/1353-8292(95)00004-6.View ArticleGoogle Scholar
- Kennedy S: A geographic regression model for medical statistics. Soc Sci Med. 1988, 26 (1): 119-129. 10.1016/0277-9536(88)90051-2.PubMedView ArticleGoogle Scholar
- Mayer JD: The role of spatial analysis and geographic data in the detection of disease cauation. Soc Sci Med. 1983, 17 (16): 1213-1221. 10.1016/0277-9536(83)90014-X.PubMedView ArticleGoogle Scholar
- Ali M, Emch M, Ashley C, Streatfield PK: Implementation of a Medical Geographic Information System: Concepts and Uses. Journal of Health, Population, and Nutrition. 2001, 19 (2): 100-110.PubMedGoogle Scholar
- Wilkinson D, Tanser F: GIS/GPS to document increased access to comminity-based treatment for tuberculosis in Africa. The Lancet. 1999, 354: 394-395. 10.1016/S0140-6736(99)01893-0.View ArticleGoogle Scholar
- Wilkinson P, Grundy C, Landon M, Stevenson S: GIS in Public Health. In GIS and Health (GISDATA Series 6). Edited by: Gatrell AC, Löytönen M. 1998, Taylor & Francis, 179-189.Google Scholar
- Kennedy M: The global positioning system and GIS. 2002, Taylor & Francis, SecondGoogle Scholar
- Camilo JA, Claudia MG, Ochiai RL, Danovaro-Holliday MC, Page AL, Thiem VD, Park JK, Koo HW, Park E, Wang XY, Elyazeed RA, Ali M, Ivanoff B, Pang T, Xu ZY, Clemens JD: The role of epidemiology in the introduction of typhoid fever Vi polysaccharide vaccines in Asia. Journal of Health, Population and Nutrition.Google Scholar
- Jensen JR: Introductory Digital Image Processing: A Remote Sensing Perspective. 2004, Prentice Hall, 3Google Scholar
- Gao J: Non-differential GPS as an alternative source of plainmetric control for rectifying satellite imagery. Photogrammetric Engineering & Remote Sensing. 2001, 67 (1): 49-55.Google Scholar
This article is published under license to BioMed Central Ltd. This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.