Abstract
In a complex environment, it can be difficult to assess the degree to which a continuous variable influences microbial community structure. We propose a method that involves using the community data to "predict" the value of the presumed dominant variable. The assumption is that in order to "predict" the variable the community composition must be sensitive to or affected by the variable in question. The concentration range over which the prediction is accurate should thus provide information on the concentrations that influence community structure. We explored this approach using T-RLFP data on at a site polluted by tannery wastes. We were able to use the microbial community structure measures to predict Cr concentration over a surprisingly wide range of concentration. Although, it appears from this work that this approach can give useful information about the relationships between microbial community structure and specific environmental conditions much further testing is required.