Impact of Uncertainty Parameter Distribution on Robust Decision Making Outcomes for Climate Change Adaptation under...

Area of Science:

• Environmental Science
• Climate Change Adaptation
• Water Resource Management

Background:

• Developing long-lasting infrastructure like water reservoirs faces deep uncertainty from future climate and economic conditions.
• Robust decision frameworks (RDFs) aid in identifying resilient investments under uncertainty, but rely on assumed probabilistic distributions for variables.
• The impact of these distributional assumptions on RDF outcomes is poorly understood, lacking guidance for selection.

Purpose of the Study:

• To evaluate the impact of different probabilistic distribution assumptions on the outcomes of robust decision-making procedures.
• To provide guidance on selecting appropriate distributions for exogenous uncertainties in exploratory modeling within RDFs.
• To assess the robustness of water reservoir management strategies under varying climatic and economic scenarios.

Main Methods:

• Applied the robust decision-making (RDM) procedure to a water reservoir case study in Ethiopia.
• Sampled climatic and economic parameters using four distributions: uniform, normal, extended uniform, and extended normal.
• Computed regret and satisficing robustness for agricultural water demand coverage and net present value, followed by scenario discovery.

Main Results:

• Extended parameter distributions resulted in lower robustness scores compared to standard distributions.
• The choice of parameter distributions significantly influenced the identification of system vulnerabilities through scenario discovery.
• Different distributions led to varying assessments of reservoir alternative robustness.

Conclusions:

• Exploratory modeling within robust decision frameworks should prioritize sampling from extended, uniform parameter distributions.
• Careful consideration of distributional assumptions is essential for accurate vulnerability assessment and robust infrastructure planning.
• The study highlights the sensitivity of RDM outcomes to input uncertainty characterization.