A comparison of SOFM ordination with DCA and PCA in gradient analysis of plant communities in the midst of Taihang Mountains, China

The self-organizing feature map (SOFM) neural network is attractive for ecological investigations for its power in analyzing and solving complicated and non-linear matters and for its freedom from restrictive assumptions that underlie many ordination techniques. The SOFM ordination were described and compared with DCA and PCA, the most common ordination methods, in analysis of plant communities in the midst of Taihang Mountains in China. The dataset consisted of importance values of 88 species in 68 quadrats of 10 in x 20 in. The SOFM ordination successfully displayed quadrats in species space and revealed ecological gradients. The distribution of quadrats and community types on SOFM ordination diagram was fully interpreted. SOFM, DCA and PCA produced consistent results, i.e. their axes were significantly correlated with elevation, soil organic matter, N, P, K and slope. These variables are important to development and distribution of plant communities in the Tailiang Mountains. SOFM ordination is effective for analysis of large-dataset of plant communities and has some advantages compared with DCA and PCA. And SOFM is conduced for combination of ordination and classification in vegetation study. (C) 2008 Elsevier B.V. All rights reserved.