program [15]. This canonical analysis is the simultaneous analysis of two or eventually several data tables. In our case,
the two tables were constructed with fauna data (only for the taxa where total number is higher than three individuals)
and soil parameters (heavy metal, organic matter, carbon,nitrogen and soil humidity). Redundancy analysis combines
two families of methods: regression and ordination. The regression is performed between the table of response
variables (fauna), the table of explanatory variables, and ordination (PCA). After the regression, principal component
analysis (PCA) of the matrix of fitted values is carried out to obtain the eigenvalues and eigenvectors of the PCA. Contrary
to the simple ordination method, eigenvalues can be tested for significance in canonical analysis, using the
permutation method [9]. This test is useful, particularly in
our case, when we consider a gradient of pollution where
data were auto-correlated.