Equilibrium Climate Sensitivity (ECS) is defined as the equilibrium global mean surface temperature response to a doubling of atmospheric CO2. While an idealized metric, ECS is closely related to several societally relevant aspects of climate change. For example, the spread in model estimates of ECS explains about 90% of the total variance in model projections of global temperature change over the 21st century, and 60-90% in regional temperature change over land in the tropics and mid-latitudes.

The primary source of uncertainty in the ECS lies with our ability to model how clouds, particularly low clouds, respond to surface temperature changes. Changes clouds in turn lead to changes in radiation that feed back onto surface temperature. Despite substantial advances in the ability of general circulation models (GCMs) to represent cloud processes the uncertainty range in ECS has not narrowed. 

Especially difficult to constrain has been the magnitude of the so-called “pattern effect”: how time evolving patterns of sea surface temperatures (SSTs) alter the circulation and structure of the atmosphere, leading to generally more positive (i.e. more destabilizing) cloud response. This effect is both very pronounced and very uncertain in General Circulation Models (GCMs), where simulations suggest that future changes in SST patterns may lead to a change in the net radiative feedback - i.e. the increase in outgoing radiation per unit increase in surface temperature - by as much as a factor of two. 

So far, no observational constraints have been brought to bear on the large model spread in the magnitude of the pattern effect, limiting our forecast ability for both near-term and long-term climate change. 

Here we propose to constrain the model spread in the magnitude of the pattern effect by leveraging a scale separation, in both time and space, between low clouds and atmospheric circulation. We will separate the pattern effect into the cloud response to changing meteorology and the response of the meteorology to time-evolving patterns of SSTs. The first term acts on small spatial scales and fast time scales, while the second involves the slow variations in the large scale structure of the atmosphere. Recent DOE-led work has helped constrain the response of clouds to meteorology. We will use a combination of GCM simulations and data analysis to constrain the response of the large scale atmosphere to changing SST patterns.

We will do this in four steps. First, we will use the available multi-model archive as well as new simulations to understand how structural uncertainty due to different model formulations impact how clouds depend on SST patterns. We will then explore parametric uncertainty within a single model to map out how parameter choices that yield specific results in the current climate relate to cloud feedbacks in the future climate.  Third, perform a dimensional reduction on the very large parameter space of climate model output, determining the geographical regions and cloud controlling factors that are most important for constraining future warming and ECS. Finally, we will use the relationship between spread in current climate and spread in future climate, along with information about dominant regions and processes, to construct metrics with high predictive power. These metrics will be used to constrain the pattern effect and ECS and guide future model development.