e , the number of taxa); P i = the relative abundance of each tax

e., the number of taxa); P i = the relative abundance of each taxon, calculated as the proportional contribution of the number of individuals of that taxon to the total number of individuals within the dataset; E = evenness. The environmental variables flooding duration, median grain size (d50) and average herb height showed right-skewed distributions and were log-transformed before further analyses.

The relations between the arthropod assemblages and the different environmental variables selleck compound (Table 1) were assessed with variance partitioning (Borcard et al. 1992; Peeters et al. 2000). Prior to the variance partitioning, the total amount of variation in each arthropod dataset was assessed by determining the sum of all canonical eigenvalues with detrended correspondence analyses (DCA; CANOCO 4.0; Ter Braak and Šmilauer 1998). DCA was also used to assess whether the arthropod assemblages followed linear or unimodal response models. The DCA was based

on logarithmically transformed arthropod numbers (log (N + 1)) and revealed short to moderate gradients for each of the four arthropod datasets LGK 974 (PXD101 mouse gradient length <3 SD). Hence, the variance partitioning was based on the linear method of redundancy analysis (RDA; CANOCO 4.0; Ter Braak and Šmilauer 1998). For each environmental variable in a canonical analysis, a so-called variance inflation factor (VIF) is calculated which expresses the (partial) multiple

correlation with other environmental variables. A VIF >20 indicates that a variable is almost perfectly correlated with other variables, which results in an unstable canonical coefficient for this variable (Ter Braak and Šmilauer 1998). Initial analyses revealed high VIFs for Racecadotril the grain size distribution parameters, i.e. clay fraction, silt fraction, sand fraction and median grain size. Of these, the median grain size was selected as representative grain size distribution parameter and the others were excluded from further analysis. Similarly, the total soil concentrations of As, Cd, Cr, Cu, Ni, Pb, and Zn were characterized by high VIFs in the initial ordinations. A principal component analysis (PCA; SPSS 16.0) was executed on the soil metal concentrations in order to reduce the amount of variables while preserving the main part of the variation. As the first principal component accounted for over 92% of the variation in the soil metal concentrations, the remaining components were discarded and for each sampling site the soil metal concentrations were replaced by the site score on the first component (Schipper et al. 2008b).

Comments are closed.