“…Derivative methods of predicting uncertainty typically assume that the input constraints have a Gaussian distribution of uncertainty, that these uncertainties are mutually independent, and that the constraint/concentration relationships are nearly linear. Under these conditions, the uncertainty in calculated concentrations can be predicted directly from the uncertainties in the inputs and the partial derivatives of the relevant equation(s), and the predicted concentrations will have a Gaussian uncertainty distribution. , (A more complete discussion of the necessary conditions can be found in standard texts, e.g., ref ). Several authors have used Monte Carlo methods to calculate standard deviations of output concentrations for Al speciation in natural waters, − metal complexation, pH in natural waters, and carbonate equilibria …”