Uncertainty-Informed Selection of CMIP6 Earth System Models for Use in Multisectoral and Impact Models
Earth System Models produce large volumes of data, used as inputs to multisectoral dynamic and impact modeling studies. Earth System Models are crucial for understanding how the human and Earth systems interact and co-evolve across sectors like energy and land use. However, using all available data from exercises such as CMIP6 with a large number of participating models is often too resource-intensive for a multisector dynamics modeling application, and the number of models participating in future CMIP eras is only expected to grow. In this work, we develop a method to select a manageable subset of Earth System Models that still captures the full range of uncertainties present in the full collection of CMIP6 data. This approach helps researchers efficiently utilize CMIP outputs for multisector dynamic and impact studies without needing overwhelming computational resources.
This research helps us better utilize the outputs of Earth System Models for understanding the dynamics of the coupled human-Earth system by selecting a manageable subset of climate models from the larger CMIP6 collection. This allows a dramatic reduction in the computational resources needed by multisector dynamic and impact modelers to utilize CMIP6 outputs without losing the breadth of uncertainties, something not prioritized by other selection methods in the literature. This efficiency will become increasingly important with the expected growth in the number of models participating in CMIP. This article was selected to appear on the journal Editor's Choice Highlights page with brief commentary from a Chief Editor.
Our research addresses the challenge of efficiently selecting a subset of Earth System Models (ESMs) from the extensive CMIP6 ensemble, which is essential for multisectoral and impact modelers who may face computational constraints. We introduce a method that retains the full range of model, scenario, and interannual variability uncertainties, with a focus on temperature and precipitation outputs. By utilizing Principal Component Analysis (PCA), we identify a new coordinate basis that maximizes total variance, enabling us to select a representative subset of models in that space. This approach preserves the variance characteristics of the full ensemble, and it can be supplemented with other criteria for selection, such as addressing the ‘hot model problem’ by ensuring that the selected subset reflects the very likely equilibrium climate sensitivity distribution defined in IPCC AR6. Adaptable to other ESM output variables, this method offers a flexible framework for designing efficient experiments using CMIP6 data for multisector dynamic and impact studies. This capability is particularly advantageous for large ensemble experiments with multisector dynamics models that require global climate information. Additionally, the method provides insights into the properties of existing CMIP6 models and can be adapted to prioritize different regions or variables, making it a valuable tool for researchers focused on ESM emulation.