Uncertain climate futures - on optimality and robustness of climate adaptation options

Climate change poses significant risk to civil infrastructure systems, exacerbated by epistemic uncertainties (also known as deep uncertainty) arising from limited knowledge about climate change.Consistently accounting for these epistemic uncertainties in developing decision support for the planning, design, and assessment/adaptation of civil infrastructure systems presents a critical challenge for society.This study introduces a risk-informed decision support framework that incorporates a new approach to represent epistemic uncertainty in decision-making.In particular, the proposed approach assigns probabilities to relevant climate scenarios based on expert opinions (e.g., from climate scientists, engineers), reflecting the degree of belief they hold about the likelihood of different future projections.This probabilistic representation of epistemic uncertainty differs from conventional robust decision-making approaches, such as the minimax regret (MMR), which typically relies on non-probabilistic methods.A case study compares the proposed approach with MMR, showing that the proposed approach is more effective on selecting decisions that are both optimal and robust based on the associated risk, whereas MMR focuses solely on robustness while neglecting the optimality. INTRODUCTIONClimate change presents growing challenges to the planning, operation, and adaptation of infrastructure systems due to the increasing intensity and variability of environmental hazards such as flooding, extreme precipitation, and storm surges.These hazards are subject to significant uncertainty, not only due to natural variability (aleatory uncertainty), but more critically due to epistemic uncertainty, i.e., uncertainty arising from limitations in knowledge, modeling assumptions, and future societal developments.Classical decision-making methods under epistemic uncertainty has been extensively studied through various robustness metrics, including the maximin criterion (which selects alternatives based on their worst-case performance) (Wald 1950), the maximax rule (based on the best-case outcome) (Savage 1954), the Hurwicz optimism-pessimism compromise