Combination of linear and nonlinear multivariate approaches effectively uncover responses of phytoplankton communities to environmental changes at regional scale

Abstract

The response of a community to environmental changes is either linear or non-linear, so that they can be investigated approximately by linear or nonlinear models. At community level, redundancy analysis (RDA) and canonical correspondence analysis (CCA), and Mantel test and Generalized Dissimilarity Modelling (GDM) are two pairs of fundamental multivariate approaches. Thus, it is necessary to determine how they are used for a given group of communities or a metacommunity. In the present study, we explored the applications of the two pairs of commonly used multivariate methods for the analysis of tropical phytoplankton communities. Phytoplankton were collected from 60 tropical reservoirs in southern China at two distinct regions and two hydrological seasons. Because of a short environmental gradient, response of phytoplankton communities to the environmental gradients was first explored with linear models: distance-based redundancy analysis (db-RDA) and Mantel test. Then, CCA and GDM were further applied to recognize the nonlinear relationship between phytoplankton community variation and environmental changes, and to detect the significant environmental and/or spatial variables. Our results strongly suggest that the combination of db-RDA and GDM provides a highly effective tool to uncover the linearity and nonlinearity in community responses and the important associated environmental and spatial variables, which were significantly different between flooding and dry seasons.