Trends in sensitivity analysis practice in the last decade.

The majority of published sensitivity analyses (SAs) are either local or one factor-at-a-time (OAT) analyses, relying on unjustified assumptions of model linearity and additivity. Global approaches to sensitivity analyses (GSA) which would obviate these shortcomings, are applied by a minority of researchers. By reviewing the academic literature on SA, we here present a bibliometric analysis of the trends of different SA practices in last decade. The review has been conducted both on some top ranking journals (Nature and Science) and through an extended analysis in the Elsevier's Scopus database of scientific publications. After correcting for the global growth in publications, the amount of papers performing a generic SA has notably increased over the last decade. Even if OAT is still the most largely used technique in SA, there is a clear increase in the use of GSA with preference respectively for regression and variance-based techniques. Even after adjusting for the growth of publications in the sole modelling field, to which SA and GSA normally apply, the trend is confirmed. Data about regions of origin and discipline are also briefly discussed. The results above are confirmed when zooming on the sole articles published in chemical modelling, a field historically proficient in the use of SA methods.

Moment independent importance measures: new results and analytical test cases.

Moment independent methods for the sensitivity analysis of model output are attracting growing attention among both academics and practitioners. However, the lack of benchmarks against which to compare numerical strategies forces one to rely on ad hoc experiments in estimating the sensitivity measures. This article introduces a methodology that allows one to obtain moment independent sensitivity measures analytically. We illustrate the procedure by implementing four test cases with different model structures and model input distributions. Numerical experiments are performed at increasing sample size to check convergence of the sensitivity estimates to the analytical values.

Local and global uncertainty analyses of a methane flame model.

Local and global uncertainty analyses of a flat, premixed, stationary, laminar methane flame model were carried out using the Leeds methane oxidation mechanism at lean (phi = 0.70), stoichiometric (phi = 1.00), and rich (phi = 1.20) equivalence ratios. Uncertainties of laminar flame velocity, maximal flame temperature, and maximal concentrations of radicals H, O, OH, CH, and CH(2) were investigated. Global uncertainty analysis methods included the Morris method, the Monte Carlo analysis with Latin hypercube sampling, and an improved version of the Sobol' method. Assumed probability density functions (pdf's) were assigned to the rate coefficients of all the 175 reactions and to the enthalpies of formation of the 37 species. The analyses provided the following answers: approximate pdf's and standard deviations of the model results, minimum and maximum values of the results at any physically realistic parameter combination, and the contribution of the uncertainty of each parameter to the uncertainty of the model result. The uncertainty of a few rate parameters and a few enthalpies of formation causes most of the uncertainty of the model results. Most uncertainty comes from the inappropriate knowledge of kinetic data, but the uncertainty caused by thermodynamic data is also significant.