Distributiondes niches écologiques le long d'un gradient de toxicité métallique

Why species do not occur everywhere? The concept of ecological niche is central to understand relationships among biotic/abiotic factors and species distribution. Gradients of environmental stress associated to interspecific interactions generate testable patterns of specie’s response curves. Distribution of response curves have been seldom tested along toxicity gradient, in contrast to resource gradients. On Katangan copper hills (south R.D.C.), natural copper and cobalt concentrations span a large range: 2 - 1000 mg kg-1 for cobalt and 30 - 10 000 mg kg-1 for copper. In this study, we evaluate three hypotheses related to niche distributions and shapes along metal toxic gradients: (1) Species optima are uniformly distributed over copper-cobalt gradient. (2) Realized niche width varies in relation to the niche optimum along copper and cobalt gradient. (3) Absolute values of skewness coefficient are higher when plant species optima are in the extremes of cobalt gradient with a niche tails toward mesic conditions. Realized niches of 80 taxa were modeled with general additive models (GAM) using presence/absence data in 172 1m² plots. Niche optima, niche widths and skewness coefficients were estimated from species response curves. The three hypotheses were globally validated. Three groups of species were distinguished according to their optima position along metal gradient with packed optima on intermediate concentrations suggesting higher interspecific competition in low metal concentrations. Niche width increased with metal concentrations. Highest metal tolerant plant species had broader niches. Skewness coefficient was inversely related to niche optima positions. Our study demonstrates that species distribution pattern on toxicity gradient presents similarities with resource gradient. Broad realized amplitude of species adapted to high levels of copper cobalt concentrations suggests that fundamental niche of specialist metallophytes may be larger than expected. This should be further tested in controlled conditions in association to competition tests