Process-level metrics for evaluating the realism of CMIP6 models’ climate sensitivity based on multiple lines of observational evidence

In this proposal for MAPP competition 5 (Constraining Models’ Climate Sensitivity) we will develop diagnostics of the radiative feedbacks in CMIP6 models contributing to equilibrium climate sensitivity (ECS). We will then assess the realism of these feedbacks against observational constraints using multiple lines of evidence. Our focus will be on previously-proposed metrics of ECS that permit like-with-like comparisons between models and observable quantities. Research has demonstrated that such like-with-like comparisons are essential for ensuring that equivalent physical processes are compared so that unbiased estimates of ECS are made. We will also evaluate the ability of the observable metrics to constrain ECS and future warming. The CMIP6 model ensemble offers a unique opportunity to perform an out-of-sample test of proposed ECS metrics by assessing their ability to predict ECS in a different model ensemble from the one in which they were identified (generally CMIP5). We will then devise approaches for constraining models’ ECS using multiple key observational metrics at once. The specific objectives of this proposal are fourfold:
(1) Identify the processes that have driven the increase in ECS in CMIP6 models relative to CMIP5 and the processes most responsible for uncertainty in ECS in both model ensembles.
(2) Assess a suite of observational metrics to determine which most strongly constrain ECS and 21st century warming within CMIP5/6 models, with a particular focus on observational metrics that permit like-with-like comparisons with models.
(3) Use observations of the key metrics we identify to evaluate the realism of CMIP5/6 climate
models’ ECS.
(4) Develop process-oriented analysis tools, designed to run on generic CMIP6 output formats, that permit a given models’ ECS to be benchmarked against multiple observational metrics to facilitate model improvement. We are enthusiastic about working with the NOAA Model Diagnostics Task Force (MDTF) to create new process-oriented analysis tools for evaluating the realism of climate models’ ECS using multiple lines of observational evidence. Our analysis scripts will be designed to run on generic CMIP6 output formats and will be implemented as process oriented diagnostic module within the MDTF framework. Our NOAA-GFDL collaborator Dr. Yi Ming, co-lead of the MDTF, will assist us with these efforts and with other components of the proposed research. We also plan to work with our NCAR collaborator Dr. Daniel Amrhein and with our LLNL collaborator Dr. Mark Zelinka on various aspects of the research, as outlined in the proposal.

Relevance to the NOAA MAPP Competition and NOAA’s long-term goal
Our proposed research aims to constrain climate models’ ECS using multiple lines of evidence that permit like-with-like comparisons between models and observations. By identifying the processes relevant to ECS in models and providing observational constraints on those metrics, this work will provide new benchmarks for model development and improve the accuracy with which future climate change can be predicted. This project is aligned with MAPP’s mission to enhance the Nation’s capability to understand and predict natural variability and changes in Earth’s climate system and with NOAA’s long-term goal of providing the essential and highest quality environmental information vital to our Nation’s safety, prosperity and resilience. We propose to develop systematic process-oriented analysis methods and scripts for community use within the MDTF framework that can be used by modeling centers to compare model diagnostics with observational constraints.