For over three decades, cloud feedback has remained the major source of uncertainty in model predictions of climate sensitivity. While progress has been made in identifying processes the regulate cloud feedback, the intermodel spread in cloud feedback has not been reduced. However, this interpretation is complicated by the fact that the various coupled ocean-atmosphere models used to make these projections have two types of emergent differences in their SSTs: First, they start out with widely varying base states due to different climatological SST biases; Second, their projections of SST changes and how they map onto the regional biases in SST also differ between models. Although these emergent large-scale SST differences ultimately arise from differences in model formulation, we hypothesize that the consequence of inter-model SST differences have impacts that are independent from the model formulation. The objective of this research is to better understand how biases in the base state and inter-model differences in the regional patterns of SST warming both influence and are influenced by cloud feedback.
We propose to better understand the contributions of regional biases in both climatological SST
and patterns of SST change from CMIP6 on their projections of cloud feedback through an analysis of CMIP6 and CFMIP.v3 simulations, and a series of coordinated modeling experiments designed to fill gaps in the CFMIP.v3 framework.
The initial analysis will exploit the suite of CFMIP.v3 atmosphere-only simulations to better
understand how the pattern of SST change impacts cloud feedback in CMIP6 models. The second
phase of this work will then perform a series of experiments using a GFDL atmospheric GCM
with climatological SSTs from other CMIP6 coupled models to investigate how biases in the mean
state of each of the CMIP6 models impact the cloud feedback in the GFDL model. These
experiments will be done with two different versions of the GFDL GCM that have both low and
high sensitivity (GFDL AM2, AM4).
The final set of experiments will use surface flux adjustments to artificially force the model’s
climatological SST to that of an exogenous model from CMIP6 (the “target” model). The GFDL
coupled model will then be integrated under both control and 4xCO2 conditions to determine how
the pattern of SST change is modulated by the climatological bias of that particular “target” model.
Our hypothesis is that climatological SST biases contribute to inter-model differences in the
pattern of SST change and that these, in turn, influence the cloud feedback. Our prediction is that
the cloud response and feedbacks of the GFDL model will move towards those of the “target”
model simply due to the artificial adjustment of the climatological SSTs.
Through these sets of experiments, we aim to both quantify and better understand the role of
climatological SST biases and patterns of SST response on the cloud feedbacks in CMIP6 models.
This will directly support the MAPP program goal to “develop new/improved methodologies and
integrate new understanding and data to improve our understanding and assessment of climate
sensitivity, explaining differences in sensitivities among models and generations of models.”