A common goal of next-generation Global Circulation Models (GCMs) is that they should be "scale-aware", which typically implies that such models should not be excessively sensitive to grid spacings, and that they should in some sense converge monotonically towards a result as grid spacings decrease. While both horizontal and vertical resolution have been treated in this manner, time resolution is typically viewed differently. Specifically, a decrease in time step size is often viewed as a "necessary evil", being decreased only in cases where spatial resolution is also decreased, requiring a change to the time resolution to satisfy a CFL condition.
Our experiments with the E3SM Atmosphere Model suggest that cloud physics and precipitation in GCMs is in fact quite sensitive to process coupling time step size, and that the biases affected by time integration error are independent from (and of comparable size to) biases due to other common sources of error, such as grid spacing and choice of sub-grid-scale physics parameterizations. This suggests that process coupling frequency is a key feature that should be adjusted for future models.