The data were analyzed using the generalized linear model (GLM) procedure of SAS 8.1 (SAS Institute, Cary, NC). Means separation was conducted using the Student–Newman–Keuls test of SAS, with significance declared at a P ⩽ 0.05. The absolute abundance was calculated as copies of 16S rRNA genes ml−1 sample, while relative abundance was calculated as percent of total bacteria (for bacterial species), or ratio over total bacteria (for total archaea).