Robust decision outcomes with induced correlations in climatic and economic parameters

Abstract

Robust decision making, a growing approach to infrastructure planning under climate change uncertainty, aims to evaluate infrastructure performance across a wide range of possible conditions and identify the most robust strategies and designs. Robust decision making seeks to find potential weaknesses in systems in order to gird these through a combination of policy, infrastructure, and, in some cases, resilient or recovery strategies. A system can be explored by simulating many combinations of uncertain climatic and economic parameters; statistical clustering can identify parameter thresholds that lead to unacceptable performance. Often, however, uncertain variables are correlated, complicating the robustness analysis and casting doubt upon the thresholds identified. Here, we evaluate the impact of ordinary, hidden correlations in uncertainty parameters that drive simulation in robust decision making. We induced correlations between temperature and key climatic and economic parameters. We tested correlations of 0%, 30%, 60%, and 90% between temperature and the absolute value of precipitation, coefficient of variation, and downward surface solar radiation, and negative correlations between temperature and net variable benefit and the discount rate. We used a calibrated simulation model of a dam system regulating Lake Tana, Ethiopia, to compute the agricultural supply and net present value of the reservoirs. As the correlation strength increased, the results converged in a smaller region. We found that strong correlations depressed robustness scores of lower-performing alternatives and conversely increased results of the higher-performing alternatives. As the correlations increased in favorable alternatives, the failure thresholds became more extreme, speciously suggesting that only intense changes would result in poor performance. This overall analysis highlights the degree to which correlations of an interconnected climatic and economic system can impact outcomes of robust decision making and suggests methods to avoid confounding results.

Uncertainty Analysis in Multi‐Sector Systems: Considerations for Risk Analysis, Projection, and Planning for Complex Systems

“…However, whether we are engaged in UC, UQ, or SA, we necessarily make some assumptions about parameter ranges and distributional forms (particularly in the case of UQ). These assumptions have implications for which variables we find to be most influential on the outputs and which decision alternatives we find to be most robust to that uncertainty (McPhail et al, 2020;Quinn et al, 2020;Reis & Shortridge, 2022). Moreover, a model calibrated to match observations with respect to one output may not sufficiently capture the dynamics of another (Efstratiadis & Koutsoyiannis, 2010).…”

Uncertainty analysis in multi-sector systems: Considerations for risk analysis, projection, and planning for complex systems

“…However, whether we are engaged in UC, UQ, or SA, we necessarily make some assumptions about parameter ranges and distributional forms (particularly in the case of UQ). These assumptions have implications for which variables we find to be most influential on the outputs and which decision alternatives we find to be most robust to that uncertainty (Quinn et al, 2020;McPhail et al, 2020;Reis & Shortridge, 2022). Moreover, a model calibrated to match observations with respect to one output may not sufficiently capture the dynamics of another (Efstratiadis & Koutsoyiannis, 2010).…”