From climate risk to action: Analysing adaptation decision robustness under uncertainty

Highlights

• Adaptation decision-making is based on uncertain climate risk and decision attributes.
• Uncertainty and sensitivity analyses typically focus only on climate risk.
• We present a framework to analyse adaptation decision robustness and sensitivity.
• Decision-makers must explore uncertainty and sensitivity of decisions, not only risk.

Abstract

Climate adaptation decisions are made under great uncertainty, arising from uncertainties about both the level of climate risk and the attributes of decision options. Decision-makers must understand how uncertainties in the input factors of risk assessment and decision models affect the ultimate adaptation decision, and whether the modelling yields a robust decision, i.e. one that is consistently identified as optimal over a range of uncertain input factors. Here, we present a framework for analysing the robustness of climate adaptation decisions. We apply a Bayesian Decision Analysis framework to determine the optimal output decision in a region based on both climate risk and decision-related attributes. Then, we present an approach for performing global uncertainty and sensitivity analysis on the optimal adaptation decision itself to assess robustness and understand which input factors most influence the decision in a particular region. We demonstrate this framework on an idealised example of adaptation decision-making to mitigate the risk of heat-stress on outdoor physical working capacity in the UK. In this application, we find that regions with high uncertainty in climate risk can still exhibit greater robustness in the optimal decision, and the decision is often more sensitive to variations in decision-related attributes rather than risk-related attributes. Previous research often stops short at assessing uncertainty and sensitivity in climate risk alone. These results highlight the necessity of conducting uncertainty and sensitivity analysis on the ultimate decision output itself in order to understand what factors drive decision robustness.