The Biological Diatom Index (BDI) is a standardized method routinely used in France for the surveillance of watercourse quality. This index is based on a list of 209 key species showing different pollution sensitivities. The pollution sensitivity, or ecological profile', is determined through the species presence probability values along a seven quality classes gradient. In a sample, the presence and abundance of species from this list are used to calculate the BDI score. Recently, water managers pointed out several weaknesses of this index: (a) the BDI fails to correctly assess water quality in acidobiontic and brackish conditions, and (b) pollution sensitivity and tolerance of several key species used to calculate the index need to be improved. This paper presents and discusses the new version of the BDI, called BDI-2006. 2802 samples were used to create this new index (samples collected on all French river types during the last thirty years), where 1063 diatom species were identified. For each sample, the community was described with species relative abundance, and relevant physico-chemical parameters were collected: pH, conductivity, dissolved oxygen, biological oxygen demand, ammonium, orthophosphates and nitrates. Eight hundreds and thirty eight key species were finally kept, including species typical of acidic and brackish waters, tropical taxa and abnormal forms of many widespread taxa (as evidence of toxicological impacts). The physico-chemical and the biological datasets were explored with classical analysis (Principal Component Analysis) and neural networks (Self-Organizing Maps), combined to draw the new key species ecological profiles. Comparing to the initial version of the BDI, several species profiles (Nitzschia paleacea for example) were clearly enhanced. This new version was adopted by the French standardisation authority (AFNOR) in October 2007.