“…Each tree and each new split are made from random data selected by the bagging method (Breiman 2001). From the variable selection measures in tree development, we used the Gini index that indicates the frequency of the selection for a split for each variable and their overall discriminative value for the classification problem (Breiman et al 1984). First, we ranked several variables (pH, alkalinity, N emissions from point sources, P emissions from point sources, zinc, DOC, BOD 5 , HCO 3 , COD Cr , calcium, chloride, chromium, manganese, magnesium, inorganic nitrogen, organic nitrogen, sodium, NH 4 -N, nickel, silicon, sulphide, sulphate, iron, nitrate, nitrite and phosphate concentrations, hardness, conductivity, TP and TN concentrations, temperature, total dissolved solids, slope, velocity, transparency, morphological status, hydrological status, status of longitudinal and transversal continuity, hydromorphological status, land use (percentages of urban, agricultural, and forest areas), category (natural, artificial, and heavily modified) water level, discharge, orthophosphate concentration, and oxygen saturation) that we considered having a big impact on the biota than we choose the five variables that have the biggest importance according to the Gini index for each watercourse type.…”