Diatom typology of low-impacted conditions at a multi-regional scale: Combined results of multivariate analyses and SOM

Some principal characteristics qualify algae as excellent indicators of the ecological status of water bodies: (1) algae are the basis of many food webs and (2) are among the first organisms to respond to environmental changes (Lowe and Laliberte 1996, McCormick and Stevenson 1998), (3) they are very rich in species compare to other communities, and autoecology of species is well documented (Lowe and Laliberte 1996), and (4) most species are widely distributed among ecosystems and geographic areas (McCormick and Cairns 1994). Intensive work has been carried out since the last century in order to recognize, describe and classify benthic diatom assemblages. In Europe, the first classification of benthic diatom assemblages, using the methods of phytosociology, has been established for the Ardennes (Belgium) and surrounding regions by Symoens (1957), who defined three watercourse types characterized by specific algal "associations". Then, most studies dealing with diatom communities were conducted in the framework of water quality monitoring programs (Descy 1976a, b, c 1979, Coste 1976, 1978). Following those investigations, some regional biotypologies were defined (Descy 1980, Fabri and Leclercq 1984, 1986, Symoens et al. 1988, Ector et al. 1997), and methods based on the global sensitivity of diatoms to pollution were developed (Prygiel and Coste 1996, 1999). In the US, variation of diatom composition along various gradients, at a continental scale, has also been studied (e.g. Pan et al. 1996, 2000, Potapova and Charles 2002). Those studies, however, considered both low-impacted and disturbed conditions at the same time. Research on diatom communities along river gradients, either at a local, regional or continental scale, has shown the prevalence of (1) downstream gradients (from fast-flowing oligotrophic highland rivers to eutrophic rivers of low-elevation plains; e.g. Symoens et al. 1988, Leland and Porter 2000, Potapova and Charles 2002), (2) chemical factors related to catchment geology, mostly alkalinity and pH (e.g. Descy 1980, Symoens et al. 1988, Fabri and Leclercq 1984, Pan et al. 1996, Potapova and Charles 2002), and (3) latitudinal and altitudinal variation of temperature (Potapova and Charles 2002). These studies represent more than 30 years of investigations and have contributed to significant progress in the knowledge of benthic diatom auto-ecology and to the development of monitoring methods using these algae. However, there have been few studies dealing with natural or near-natural diatom assemblages and with the natural factors that determine community composition and structure, all carried out at a local or regional scale (Sabater and Roca 1992, Aboal et al. 1996, Pan et al. 2000, Cantonati et al. 2001), and never at a large multi-regional scale. In a context of increasing environmental change, there is a need to better distinguish community changes due to natural factors from those driven by changes from human activities. In order to set up a typology of benthic diatom assemblages in rivers from several regions of Western Europe, 467 diatom records from streams with minimal human impact were examined in relation to water chemistry, watershed characteristics, geology, and stream habitat. Ordination techniques were used to determine the major variation in species composition data and to explore relationships between diatom taxa distributions and measured environmental variables. Artificial neural networks (self-organizing maps and multilayer perceptron with back-propagation algorithm) were computed to define and predict diatom assemblages using environmental parameters. Several methods were used: results of multivariate analyses, artificial neural networks, indicator species analysis and expert knowledge were combined to define the benthic diatom typology at a multi-regional scale in Western Europe and are presented in this paper.