Reply on AC1

Trude Eidhammer, Andrew Gettelman, Katherine Thayer‐Calder, Duncan Watson‐Parris, Gregory S. Elsaesser, Hugh Morrison, Marcus van Lier‐Walqui, Ci Song, Daniel T. McCoyOpenalex Unknown 2024 · DOI · OpenAlex

Abstract

This paper documents the methodology and preliminary results from a Perturbed Parameter Ensemble (PPE) technique, where multiple parameters are varied simultaneously and the parameter values are determined with Latin hypercube sampling. This is done with the Community Atmosphere Model version 6 (CAM6), the atmospheric component of the Community Earth System Model version 2 (CESM2). We apply the PPE method to CESM2-CAM6 to understand climate sensitivity to atmospheric physics parameters. The initial simulations vary 45 parameters in the microphysics, convection, turbulence and aerosol schemes with 263 ensemble members. These atmospheric parameters are typically the most uncertain in many climate models. Control simulations are analyzed and targeted simulations to understand climate forcing due to aerosols and fast climate feedbacks. The use of various emulators is explored in the multi-dimensional space mapping input parameters to output metrics. Parameter impacts on various model outputs, such as radiation, cloud and aerosol properties are evaluated. Machine learning is also used to probe optimal parameter values against observations. Our findings show that using PPE is a valuable tool for climate uncertainty analysis. Furthermore, by varying many parameters simultaneously, we find that many different combinations of parameter values can produce results consistent with observations, and thus careful analysis of tuning is important. The CESM2-CAM6 PPE is publicly available, and extensible to other configurations to address questions of other model processes in the atmosphere and other model components (e.g. coupling to the land surface).

Introduction: The Subgrid-Scale Challenge

Global Climate Models (GCMs) are essential for predicting future climate states, yet they face significant challenges due to "subgrid-scale" processes. Phenomena such as cloud droplet formation or turbulent eddies occur at scales much smaller than the typical model grid (~100 km). To represent these, scientists use parameterizations—simplified mathematical proxies governed by specific numerical parameters.

Historically, these parameters are poorly constrained and often tuned using "One-At-a-Time" (OAT) methods. However, OAT is inefficient and fails to capture non-linear interactions between parameters. This study introduces a systematic Perturbed Parameter Ensemble (PPE) for the Community Atmosphere Model version 6 (CAM6) to objectively explore the vast uncertainty space of atmospheric physics.

Methodology: Sampling and Emulation

The researchers varied 45 parameters across four critical physics packages: microphysics, aerosols, convection, and turbulence. Using Latin Hypercube Sampling (LHS), they generated 263 unique parameter combinations to ensure efficient coverage of the multi-dimensional space.

Distribution of the autoconversion factor across the 263 ensemble members, demonstrating the uniform coverage achieved via Latin Hypercube Sampling.
Distribution of the accretion enhancement factor, illustrating the sampling strategy for microphysical parameters.

For each parameter set, three climate scenarios were simulated: Present Day (PD), Pre-Industrial (PI), and a +4K sea surface temperature increase. To overcome the computational cost of full GCM runs, the team developed machine learning emulators, including Neural Networks (NN), Gaussian Processes (GP), and Random Forests (RF), to map input parameters to output metrics like radiation and cloud properties.

Key Insights: Equifinality and Sensitivity

The PPE approach revealed several critical insights into model behavior:

  • The Equifinality Problem: Multiple, distinct parameter combinations can produce results that align equally well with satellite observations (e.g., CERES and MAC-LWP). This suggests that matching present-day observations does not guarantee the underlying physics are accurately represented for future projections.
  • Dominance of Microphysics: Parameters related to cloud microphysics, such as ice fall speed and accretion rates, were found to be the primary drivers of model uncertainty across most regions.
  • Regional Variability: Sensitivity is geographically dependent; microphysics dominates the Arctic and midlatitudes, while deep convection parameters drive uncertainty in the tropics.
  • Structural vs. Parametric Error: In some instances, optimal parameter values were found at the extreme edges of plausible ranges, indicating potential structural errors in the model's fundamental equations.

Results and Performance

The PPE revealed massive variations in key metrics; for example, the Ice Water Path (IWP) and Liquid Water Path (LWP) varied by a factor of 2 to 4 across the ensemble. Gaussian Processes (GP) and specialized Neural Networks achieved high accuracy in predicting global mean values, with $R^2$ values up to 0.90. Using these emulators to sample 20 million possible parameter sets, the authors identified "optimal" regions that satisfy multiple observational constraints simultaneously—a task nearly impossible with manual tuning.

Conclusion and Future Directions

The CESM2-CAM6 PPE provides a powerful, open-source framework for climate uncertainty analysis. By moving beyond manual tuning, researchers can better identify optimal parameter regions and highlight areas where structural model improvements are necessary. Future work will extend this framework to fully coupled Earth System Models to investigate how atmospheric uncertainties propagate through the entire climate system.

Summary updated

Abstract

<strong class="journal-contentHeaderColor">Abstract.</strong> This paper documents the methodology and preliminary results from a Perturbed Parameter Ensemble (PPE) technique, where multiple parameters are varied simultaneously and the parameter values are determined with Latin hypercube sampling. This is done with the Community Atmosphere Model version 6 (CAM6), the atmospheric component of the Community Earth System Model version 2 (CESM2). We apply the PPE method to CESM2-CAM6 to understand climate sensitivity to atmospheric physics parameters. The initial simulations vary 45 parameters in the microphysics, convection, turbulence and aerosol schemes with 263 ensemble members. These atmospheric parameters are typically the most uncertain in many climate models. Control simulations are analyzed and targeted simulations to understand climate forcing due to aerosols and fast climate feedbacks. The use of various emulators is explored in the multi- dimensional space mapping input parameters to output metrics. Parameter impacts on various model outputs, such as radiation, cloud and aerosol properties are evaluated. Machine learning is also used to probe optimal parameter values against observations. Our findings show that using PPE is a valuable tool for climate uncertainty analysis. Furthermore, by varying many parameters simultaneously, we find that many different combinations of parameter values can produce results consistent with observations, and thus careful analysis of tuning is important. The CESM2-CAM6 PPE is publicly available, and extensible to other configurations to address questions of other model processes in the atmosphere and other model components (e.g. coupling to the land surface).
BibTeX
@inproceedings{eidhammer2024reply,
  author = {Trude Eidhammer and Andrew Gettelman and Katherine Thayer‐Calder and Duncan Watson‐Parris and Gregory S. Elsaesser and Hugh Morrison and Marcus van Lier‐Walqui and Ci Song and Daniel T. McCoy},
  title = {{Reply on AC1}},
  booktitle = {Openalex Unknown},
  year = {2024},
  doi = {10.5194/egusphere-2023-2165-cc2},
}
Figure