Application of Variance-Based Sensitivity Analysis to a Large System Dynamics Model

TL;DR

This paper demonstrates how variance-based sensitivity analysis can be effectively applied to large, complex system dynamics models like the Biomass Scenario Model to identify key drivers and understand interactions.

Contribution

It introduces a novel two-stage sensitivity approach combining screening and variance decomposition for complex system dynamics models.

Findings

Identified key system levers in the biomass industry model.

Quantified the strength of different system interactions.

Showed the effectiveness of VBSA in complex nonlinear models.

Abstract

Variance-based sensitivity methods can provide insights into large computational models. We present a novel application of sensitivity analysis to the Biomass Scenario Model (BSM) a large and complex system dynamics model of the developing biofuels industry in the United States. We apply a two-stage sensitivity approach consisting of an initial sensitivity screening, followed by a variance decomposition approach. Identifying key system levers and quantifying their strength is not straightforward in complex system dynamics models that have numerous feedbacks and nonlinear results. Variance-based sensitivity analysis (VBSA) offers a systematic, global approach to assessing system dynamics models because it addresses nonlinear responses and interactive effects. Especially when a large model's size makes manual exploration of the input space difficult and time-consuming, the approach can help to provide a comprehensive understanding of interactions that drive model behaviors.