DIATMOD: diatom predictive model for quality assessment of Portuguese running waters
- First Online:
- Cite this article as:
- Almeida, S.F.P. & Feio, M.J. Hydrobiologia (2012) 695: 185. doi:10.1007/s10750-012-1110-4
A predictive model for diatoms based on an adaptation of the River Invertebrate Prediction and Classification System/Australian River Assessment System approaches was evaluated as an effective tool for measuring stream ecological quality. This type of model was originally developed in UK and later in Australia and is extensively used to obtain ecological quality assessments with macroinvertebrates. The first step for the model construction was the definition of six consistent reference biological groups (ANOSIM: Global R = 0.77; P < 0.001) after classification (UPGMA) and ordination (nMDS) of 120 reference sites containing 254 different diatom taxa (species and infra-specific rank). A set of five environmental variables (slope, hydrological regime, mean annual temperature, mean annual precipitation and alkalinity) correctly discriminated 67% of reference sites (stepwise forward discriminant analysis, Jackknifed classification). The model was statistically accurate (slope = 1.07, intercept = −0.68, R2 = 0.65) and was validated by an independent set of reference data (13 reference sites; 70% correct answers). In addition, the model was tested by running data from 113 potentially disturbed sites. The model (DIATMOD) was well correlated with a general abiotic degradation gradient (Spearman correlations, R2 = 0.53, P < 0.001; and PCA analysis) and also with several specific pressure variables such as nitrates, phosphates, urban area, connectivity and land use (P < 0.001). Most diatom indices assess chemical contamination and we showed here that through predictive modelling the potential of diatoms as bioindicators increases as they also responded to hydromorphological changes. Further investigation on model potential consists in: testing different probability levels for taxa inclusion (here it was >0.5 as the most common models); comparing with alternative classification systems; assessing the influence of substrate type and seasonal variation in assessments.