In Alsace (France) drinking water is supplied from two different sources. This situation is favourable to study the effects on human health of various water characteristics, among which hardness. However it would be misleading to consider the 947 towns or villages as independent units of observation because of their geographical adjacencies. An aggregation algorithm of these units has been set up. It consists in a progressive regrouping of adjacent areas, similar "enough" from the point of view of the parameters studied, the aim of the procedure being to get final areas with contrasted values of water characteristics. This partition is then used to compute rank correlation coefficients between water parameters and various mortality rates. With this method, no significant correlation was found between water hardness and mortality, in spite of the existence of a sharp gradient of hardness.