“…These include a wide variety of linear approaches including linear regression (LR) (Lahr et al, 1999;Marchant et al, 2006), multiple linear regression (MLR) (Friday, 1987), principal component analysis (PCA) (Rundle et al, 2002;Jeffries, 2005;Kasangaki et al, 2008;Haidekker and Hering, 2008;Blocksom and Johnson, 2009) and multidimensional scaling (MDS) (Cowell et al, 2004). Also, more advanced multivariate techniques like detrended correspondence analysis (DCA) (Valladares et al, 2002) and canonical correspondence analysis (CCA) (Fernández-Aláez et al, 2002;Irmler et al, 2002;Rundle et al, 2002;Miserendino and Archangelsky, 2006;Bilton et al, 2006;Rohasliney and Jackson, 2008) have been used in the past. Despite the fact that these methods may reveal significant relationships between community composition and environmental parameters, they possess a number of features that limit their explanatory, estimation or forecasting power (James and McCulloch, 1990).…”