Global storm-resolving models (GSRMs) are expected to provide greater accuracy and have less uncertainty in their predictions than coarser models participating in the Coupled Model Intercomparison Project (CMIP). We present radiative feedback estimates from a pair of 13-month simulations using the US Department of Energy’s Simple Cloud Resolving E3SM Atmosphere Model (SCREAM) and compare them with feedbacks from CMIP-class models. With a horizontal resolution of ~3.25-km, SCREAM is a global atmosphere model written in C++ Kokkos to run efficiently on the world’s fastest supercomputers. The simulations are forced with present-day sea-surface temperatures (SSTs) and with SSTs increased uniformly by 4K to allow the calculation of Cess feedbacks. To better characterize the uncertainties associated with just running a pair of one-year experiments, we also run five pairs of simulations with a lower-resolution (~12km) version of SCREAM.

In the present-day, SCREAM produces a reasonable mean-state climate, similar in skill to 100-km scale models. Its estimated climate feedback, however, is -1.2 Wm^-2K^-1; its strong positive cloud feedback puts it among CMIP models with the weakest overall radiative damping, implying a larger than average equilibrium climate sensitivity. At the same time, all five pairs of the 12-km SCREAM ensemble produce significantly more negative total feedbacks than the 3-km SCREAM, suggesting that resolution sensitivity exists even in the 1-10 km dx range. Given recent GSRM studies showing more moderate estimates of radiative feedback, running at ultra-high resolution without a deep convection scheme does not appear to narrow the uncertainty in climate feedbacks.

This work was conducted under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. LLNL-ABS-867478