Matlab/R workflows to assess critical choices in Global Sensitivity Analysis using the SAFE toolbox

Valentina Noacco,Fanny Sarrazin,Francesca Pianosi,Thorsten Wagener

doi:10.1016/j.mex.2019.09.033

Valentina Noacco, Fanny Sarrazin + Show 2 more

Open Access

https://doi.org/10.1016/j.mex.2019.09.033

Copy DOI

Journal: MethodsX	Publication Date: Jan 1, 2019
Citations: 31	License type: cc-by

Affiliation: University of Bristol

Abstract

Global Sensitivity Analysis (GSA) is a set of statistical techniques to investigate the effects of the uncertainty in the input factors of a mathematical model on the model’s outputs. The value of GSA for the construction, evaluation, and improvement of earth system models is reviewed in a companion paper by Wagener and Pianosi (2019). The present paper focuses on the implementation of GSA and provides a set of workflow scripts to assess the critical choices that GSA users need to make before and while executing GSA. The workflows proposed here can be adopted by GSA users and easily adjusted to a range of GSA methods. We demonstrate how to interpret the outcomes resulting from these different choices and how to revise the choices to improve GSA quality, using a simple rainfall-runoff model as an example. We implement the workflows in the SAFE toolbox, a widely used open source software for GSA available in MATLAB and R.•The workflows aim to contribute to the dissemination of good practice in GSA applications.•The workflows are well-documented and reusable, as a way to ensure robust and reproducible computational science.

Highlights

Global Sensitivity Analysis (GSA) is a set of statistical techniques to investigate the effects of the uncertainty in the input factors of a mathematical model on the model’s outputs
Global Sensitivity Analysis (GSA) is a set of statistical techniques that allow to assess the effects of the uncertainty and variability in the input factors of a mathematical model on the model’s output(s) [2]
GSA has been shown to improve the construction and evaluation of earth system models and to maximise the information content that is extracted from model predictions [1]

Summary

Introduction

Global Sensitivity Analysis (GSA) is a set of statistical techniques to investigate the effects of the uncertainty in the input factors of a mathematical model on the model’s outputs. Global Sensitivity Analysis (GSA) is a set of statistical techniques that allow to assess the effects of the uncertainty and variability in the input factors of a mathematical model on the model’s output(s) [2]. Implications The choice of the GSA method is critical as different methods focus on different aspects of the model input-output response and may lead to different sensitivity estimates.

Results

Conclusion