Propagation of Uncertainty in Aqueous Equilibrium
Calculations: Non-Gaussian Output Distributions
Stephen E. Cabaniss
10.1021/ac961290y.s001
https://acs.figshare.com/articles/journal_contribution/Propagation_of_Uncertainty_in_Aqueous_Equilibrium_Calculations_Non-Gaussian_Output_Distributions/3583497
The propagation of uncertainty in aqueous equilibrium
calculations is examined using a derivative method and
Monte Carlo simulations. Simulations of 10<sup>4</sup> trials
provide both good reproducibility and reasonably short
simulation times (<100 s on a 90 MHz Pentium microcomputer) for simple systems of up to seven components.
Independent Gaussian uncertainty distributions of
input
constraints can lead to bimodal and/or skewed output
distributions of pH, pM, and species concentrations.
Gaussian input uncertainties of ≤10% can lead to
much
larger output uncertainties (95% confidence interval in
pH or pM over 2 log units). While derivative methods
of
uncertainty prediction are faster than Monte Carlo simulations and reasonably accurate for some solution conditions, they are inappropriate if the output distribution is
non-Gaussian. Consequently, Monte Carlo simulations
are an essential complement to derivative methods for
evaluating the uncertainty of calculated equilibrium concentrations.
1997-09-15 00:00:00
output distributions
input constraints
uncertainty prediction
Monte Carlo simulations
equilibrium calculations
solution conditions
output distribution
Gaussian input uncertainties
uncertainty distributions
species concentrations
method
equilibrium concentrations
10 4 trials
2 log units
90 MHz Pentium microcomputer