“…Decision trees: (i) handle non-parametric data where the predictors are not characterized as having a specific distribution, (ii) are insensitive to missing data, to inclusion of irrelevant predictors, or to the presence of outliers, (iii) effectively operate Nelson andSommers (1982, 1996) pH probe Peech (1965), Thomas (1996) Electrical conductivity probe Lal, (1996), Peech (1965) P, K, Ca, Mg, Al, B, Cu, Fe, Mn,,Ni, Pb, S Inductively coupled plasma analysis Kovacs et al (2000) F, Cl, Br, Cl − , NO 3 -N, PO 4 -P, SO 4 -S Ion chromatograph Zhang et al (2013) using numerical, ordinal, binary, and categorical classes; and (iv) identify complex hierarchical relationships between predictors and response variables (Heung et al 2014). Practically, the decision tree algorithms iteratively split the data into mutually exclusive subsets, minimizing the sum of squares, and creating homogeneous groups (Breiman et al 1984).…”