Beyond Single--Component Perturbed Parameter Ensembles

Perturbed parameter ensembles (PPEs) are widely used to quantify parametricuncertainty in climate models, and now increasingly paired with statistical and machine learning emulators. Most such PPE studies are conducted within a single component of the climate system (atmosphere, land, ocean, or ice), but climate responses of interest arise from interactions across these components. Single-component PPEs by design cannot quantify these cross-component effects (e.g., how atmospheric parameters affect evapotranspiration, typically considered a land response), or how they depend on coupling configuration (e.g., ocean coupling modulating these effects). We take precipitation minus evapotranspiration (P-ET) over land as a case study, perturb fourteen land and atmospheric parameters across three coupling configurations of CESM2, and show the relative importance of cross-component parameter effects. We also show that cross-component effects can be anticipated through physical understanding of the climate system.