Because Doubt Is A Sure Thing: Incorporating Uncertainty Characterization Into Climate Change Decision-Making

This presentation describes the results of new research to develop a stakeholder-driven uncertainty characterization (UC) process to help address the challenges of regional climate change mitigation and adaptation decisions. Integrated regional Earth system models are a promising approach for modeling how climate change may affect natural resources, infrastructure, and socioeconomic conditions at regional scales, and how different adaptation and mitigation strategies may interact. However, the inherent complexity, long run-times, and large numbers of uncertainties in coupled regional human-environment systems render standard, model-driven approaches for uncertainty characterization infeasible. This new research focuses on characterizing stakeholder decision support needs as part of an overall process to identify the key uncertainties relevant for the application in question. The stakeholder-driven process reduces the dimensionality of the uncertainty modeling challenge while providing robust insights for science and decision-making. This research is being carried out as part of the integrated Regional Earth System Model (iRESM) initiative, a new scientific framework developed at Pacific Northwest National Laboratory to evaluate the interactions between human and environmental systems and mitigation and adaptation decisions at regional scales. The framework provides a flexible architecture for model couplings between a regional Earth system model, a regional integrated assessment model, and highly spatially resolved models of crop productivity, building energy demands, electricity infrastructure operation and expansion, and water supply and management. In an example of applying the stakeholder-driven UC process, the presentation first identifies stakeholder decision criteria for a particular regional mitigation or adaptation question. These criteria are used in conjunction with the flexible architecture to determine the relevant component models for coupling and the simulation outputs that will be relevant to the decision criteria. The next step is to perform uncertainty source identification for the chosen model couplings and simulation outputs. Uncertainty sources include input quantification, model skill, model completeness, and integration/implementation. Once these sources are identified, the process moves into a sensitivity analysis phase to identify the uncertainties with the largest impact on the relevant simulation outputs. The final steps in the process are to develop probability distributions on the most sensitive uncertainties (e.g., through expert elicitation) and then to propagate those uncertainties across the relevant model couplings.