par Godefroid, Sandrine ;Dana, Elias
Référence Journal of biogeography, 34, 1, page (62-68)
Publication Publié, 2007-01
Article révisé par les pairs
Résumé : Aim  The aims of this study are: (1) to explore the relationships between the Ellenberg indicator values (light, temperature and moisture) recently developed for two Mediterranean regions (Italy and Greece); and (2) to evaluate the possibility of using these values in other Mediterranean areas.Location  Southern Greece and the Mediterranean part of Italy.Methods  A global matrix containing 966 items of information (161 species × 3 indicator variables × 2 values assigned in each study, one in Italy and one in Greece) was constructed. A test of the accuracy of the values in predicting actual environmental conditions was provided using a detrended correspondence analysis (DCA) on published vegetation relevés from the Mediterranean region. The gamma statistic was used to express relationships of ecological indicator values between the two regions, and pairs of indicator values calculated for each species were compared using Wilcoxon matched pairs tests.Results  The results showed that indices developed for Greece and Italy were not similarly correlated to sample scores along DCA axis 1. Species’ indicator values for Italy and for Greece were highly significantly correlated for light and moisture and significantly correlated for temperature, but the correlations were weak (0.20 ≤ gamma ≤ 0.31). Pairwise testing gave significantly different indicator values for the two regions. Discrepancies below or equal to 2 units on 9-point scales (12 points for moisture) were found for 88%, 70% and 54% of the species respectively for light, temperature and moisture indices. A substantial number of species showed higher discrepancies, from 3 up to 6 units.Main conclusions  In the light of the present work, it seems clear that the indicator values developed for Italy and for Greece should not be used outside the region for which they were defined. This constitutes additional evidence that indicator values can be influenced by the identity of potential competitors. It also underlines the need to use standardized methods for calibrating indicator values against measured variables in such a way that indicator systems can serve as general reference systems.