Projections of future climate change to support decision-making require Earth system models (ESMs) running at high spatial resolution, but this is computationally prohibitive. A major challenge is the calibration (parameter tuning) during the development of ESMs, which requires running large numbers of simulations to identify the optimal values for parameters that are poorly constrained by observations. Here we train a convolutional neural network (CNN) on perturbed parameter ensembles from two lower-resolution (and thus much less expensive) versions of the same ESM, and a smaller number of higher-resolution simulations. Cross-validated results show that the CNN’s skill exceeds that of a climatological baseline for most variables with as few as 5-10 examples of the higher-resolution ESM, and for all variables (including precipitation) with at least 20 examples. This proof-of-concept study offers the prospect of significantly more efficient calibration of ESMs, by reducing the required CPU time for calibration by 20-40 \%

To assess impacts of solar geoengineering, the GeoMIP G1 experiment forces Earth System Models with prescribed reductions in solar radiation to balance increases in atmospheric CO2. One key source of uncertainty is the magnitude of solar constant reduction required to offset a CO2 quadrupling. Here, we decompose the G1 experiment in the Community Earth System Model with solar only and CO2 only forcing experiments to quantify single forcing rapid radiative adjustments. We find that radiative adjustments to both single forcings have a net positive effect on top of atmosphere energy balance such that they both increase the net G1 forcing. Stratospheric temperature and shortwave cloud adjustments are the main sources of positive adjustment in both Solar and CO2 scenarios. We also show that net G1 radiative adjustment cannot be represented linearly with CO2 and solar forcing adjustments, which is primarily traced to further reductions in boundary layer clouds.