Use of classification and regression tree (CART) analysis with chemical faecal indicators to determine sources of contamination

Abstract

Both human and non-human faecal contamination of environmental waters pose a risk to human health that can be reduced if the source can be identified and responsibility for mitigation accepted. Traditional microbiological indicators are rarely adequate for source identification but some chemical indicators offer promise. In this work, the similarities and differences of selected chemical profiles of different faecal effluent types have been modelled using classification and regression tree (CART) analysis. Human effluents can be distinguished from non-human effluents by the presence of fluorescent whitening agents, and the concentrations or ratios of concentrations of the faecal stanols coprostanol and 24-ethylcoprostanol can distinguish most of the non-human effluent types considered. To allow for the effect of variable dilution when effluents enter environmental waters, ratios of concentrations will be required to establish a model, and a model more complex than CART will be required to apportion multiple effluent-type contamination.