Navigating Uncertainty in Multi-Sector System Modeling
Enhancing Societal Resilience through Comprehensive Uncertainty Analysis in MultiSector Dynamics Models. We review approaches and opportunities for uncertainty analysis in MSD, advocating for improved transparency and uncertainty-informed research designs.
Photo by RDNE Stock project via Pexels
Simulation models of multi-sector systems are increasingly used to understand the impacts of climate and economic shocks and change. However, multi-sector systems are also subject to numerous uncertainties that prevent the direct application of simulation models for prediction and planning, particularly, when extrapolating past behavior to a nonstationary future. Methods and vocabulary for treating and discussing uncertainties can vary across disciplines, making it difficult to communicate uncertainty analyses for interdisciplinary research teams. We propose a taxonomy for varying types of uncertainties and discuss methods, challenges, and opportunities for multi-sector dynamics uncertainty analyses.
Standardization of vocabularies and methods for uncertainty analyses is essential to communicate and contextualize research results for complex systems, which can rely heavily on assumptions about models, parameters, and inputs. This is particularly important within interdisciplinary research groups and projects, such as those involved in multisector dynamics research. Our article provides an initial taxonomy and vocabulary for use in communicating multisector dynamics research uncertainties. We also highlight gaps and challenges in the existing landscape of uncertainty methods, providing guidance for future research opportunities or applications of new methods from statistics or machine learning.
We organize multisector dynamics-relevant uncertainties into sampling, parametric, and structural uncertainties and demonstrate how the same conceptual uncertainty can be treated differently depending on how it is classified. How uncertainties are treated also depends on whether they are endogenous or exogenous to the system model, as well as the degree of coupling within the model framework. We also highlight the impacts of careful design of experiments, statistical treatments, and considerations of cross-scale dynamics on resulting uncertainties, as well as the inadequacy of relying on “ensembles of opportunity” to treat uncertainties, which may combine several analysis-relevant uncertainties together without clarifying which contribute to variations in outcomes. Advances in scenario discovery techniques are needed to help disentangle the roles of uncertainties and internal systems dynamics for complex systems, which are often complicated by the dimensionalities of the parameter and outputs and the interest in extreme outcomes.