1st step : urban rural dichotomy

At the beginning, the SIFP database had two separate layers: one containing the villages of South India, the other containing the towns. However, small towns were too numerous to be mapped in a proper way and it was decided to merge them with surrounding rural areas.
The first step consisted in separating the cities (towns with more than 100.000 inhabitants) from the other towns. Cities have been mapped under squared form (if you roll over a square with your mouse, the name of the city will appear).
The other towns (with less than 100.000 inhabitants) have been merged in the village database. Therefore we now have 2 new superimposed layers, which will be mapped separately, but following the same classification, so that we can use the same legend for both.
As a result, only cities appear in the foreground, the other towns are embedded in the background map.

2nd step : clustering

The entire database contains more than 75.000 units. Hence it is impossible to map out all these data in a visually useful manner. Furthermore, the variation of population and surface from one village to the other is so huge that the Modifiable Areal Unit Problem forbids any interpretation of the maps (for details about the MAUP see Openshaw, 1984).
That's why we decided to aggregate the data, in order to obtain less units to map. This method, called "clustering", was designed by Christophe Z Guilmoto and is described in Guilmoto et al. (2003 & forthcoming). This process really diminishes the number of units, but respects each demographic weight (as it is an aggregation) and the geographical location of the units (as it is spatially restricted). Using the clusters has a direct impact on the statistical robustness of the data, particurlaly because of the decrease in the data variability (in particular, the method reduces drastically the role of the outliers).

3rd step : kriging

To map the clusters, we used a method of spatial smoothing called ordinary kriging (for further details see Burrough & McDonnell, 1998, page 132 and following). Kriging is an exact estimation method, which means that it retains original observed values without smoothing them out. The interpolation is only on the unknown surface. The spatial repartition of the phenomena is then not modified, unlike other methods like modelling based on potentials. Last, the procedure is based on a spatial autocorrelation analysis of the phenomena, since the modelling is based on the semivariance of the data. An example of cartography using kriging is proposed by Oliveau (2003).

4th step : contouring

The map we obtain is then contoured to get classified color areas, which are easier to read than a continuous range of color.

A brief note about the names

Names given on the maps are from different sources. As the spelling of names is often changing in India, we tried to put the official one or the most commonly used. For example, Madras is officially called Chennai, and everyone in South India knows it, so we kept Chennai. The river Cauvery may be found as Kavery, but the former name is most commonly used.
Names are from census maps (1991), Lonely Planet road atlas (one of the best sources...) and the maps of the Survey of India (National Atlas of India, Physical, (1961), plates 33, 34, 35, 36, 37).

Bibliography

• BURROUGH, Peter A., MC DONNELL, Rachael, (1998), Principles of Geographical Information Systems, Oxford University Press, Oxford, 333 p.
• GUILMOTO, Christophe Z., OLIVEAU, Sébastien, VINGADASSAMY, Sattianarayanin (2002), " Un système d'information géographique en Inde du Sud : Théorie, mise en œuvre et applications thématiques", Espace, Populations et sociétés, Lille, pp. 147-163
• GUILMOTO, Christophe Z., OLIVEAU, Sébastien, (forthcoming), " Un système d'information géographique en Inde du Sud : Théorie, mise en œuvre et applications thématiques ", Pondy paper in Social Sciences, French Institute of Pondicherry, Pondicherry.
• OLIVEAU, Sébastien, (2003), "Mapping out fertility in South India: methodology and results", in Guilmoto, C.Z., RAJAN, I., pp. 147-163
• OPENSHAW, S., (1984), The modifiables Areal Unit Problem, Concepts and techniques in Modern Geography 38, Norwich : Geo books.