Uncertainty, Sensitivity Analysis, and Causal Identification in the Arctic using a Perturbed Parameter Ensemble of the HiLAT Climate Model

• Main Authors: Hunke, Elizabeth Clare, Urrego Blanco, Jorge Rolando, Urban, Nathan Mark
• Language: unknown
• Published: 2021
• Subjects: 54 ENVIRONMENTAL SCIENCES; Arctic; Sea ice
• Online Access: http://www.osti.gov/servlets/purl/1422910; https://www.osti.gov/biblio/1422910; https://doi.org/10.2172/1422910

Coupled climate models have a large number of input parameters that can affect output uncertainty. We conducted a sensitivity analysis of sea ice proper:es and Arc:c related climate variables to 5 parameters in the HiLAT climate model: air-ocean turbulent exchange parameter (C), conversion of water vapor to clouds (cldfrc_rhminl) and of ice crystals to snow (micro_mg_dcs), snow thermal conduc:vity (ksno), and maximum snow grain size (rsnw_mlt). We used an elementary effect (EE) approach to rank their importance for output uncertainty. EE is an extension of one-at-a-time sensitivity analyses, but it is more efficient in sampling multi-dimensional parameter spaces. We looked for emerging relationships among climate variables across the model ensemble, and used causal discovery algorithms to establish potential pathways for those relationships.