Work

Sensitivity Analysis: An Operational Picture

Year: 2023

Type: article

Abstract: Modeling is crucial to understand the behavior of environmental systems. A deeper comprehension of a model can be aided by global sensitivity analysis (GSA). Variability ascribed to model variables could have a stochastic (i.e., lack of knowledge) or an operational (i.e., possible design values) origin. Despite the possible different nature in the variability , current GSA strategies do not distinguish the latter in their formal derivations/developments. We propose to disentangle the variability in the operational and stochastic variables while assessing the model output sensitivity with respect to the former. Two operational sensitivity indices are introduced that serve to characterize the sensitivity of a model output of interest with respect to an operational variable in terms of (a) its average (with respect to the stochastic variables) intensity and (b) its degree of fluctuation (across the set of possible realizations of the stochastic variables), respectively. We exemplify our developments considering two scenarios. Results highlight the relevance of employing an operational GSA when the focus is on the influence of operational variables on model output.

Source: Water Resources Research

DOI: 10.1029/2022wr033780

Author: A. Dell’Oca

Institutions: Consejo Superior de Investigaciones Científicas, Institute of Environmental Assessment and Water Research

Language: English

Topic: Probabilistic and Robust Engineering Design

Subfield: Statistics, Probability and Uncertainty

Field: Decision Sciences

Domain: Social Sciences

Open Access status: bronze

Funders: European Commission, H2020 Marie Skłodowska-Curie Actions

Awards: G434337789, Quantification of mixing and dynamic uncertainty for transport in heterogeneous porous media, Quantification of mixing and dynamic uncertainty for transport in heterogeneous porous media