Marine low-level clouds have long been poorly simulated in models despite many studies and field experiments devoted to their improvement. The marine subtropical stratocumulus decks simulated in E3SMv1 as manifested in satellite simulator low-level cloud cover (LCC) are evaluated by comparing with that simulated by three U.S. contributions to CMIP5. The stratocumulus decks are displaced from where they are observed from satellite within the classical 10° × 10° core regions. Because of this displacement, there is a wide spread in model-simulated LCC and in the well-known strong relationship between LCC and lower tropospheric stability (LTS).

To gain further insights for model improvement, we consider larger 30° latitude × 35° longitude extended regions, focusing on those parts where low-level stratiform clouds are dominant (with LCC > 45%). In these regions, the spread in model-simulated LCC and in the LCC-LTS relationship is reduced. The displacement of the marine subtropical stratocumulus decks point to model biases in the interaction between model physical parameterizations and the large-scale dynamics, since simulated LTS is very similar to ERA-Interim. However, large-scale dynamical errors are likely to play a role in the lack of sensitivity to ENSO in E3SMv1 and the other models and in the lack of sensitivity of LCC to LTS in E3SMv1’s fully coupled historical simulations in the northeast Pacific stratocumulus deck.

Furthermore, simulated LWP in the decks are too low and insensitive to LTS. This contradicts the decrease in LWP with increasing LTS measured by Moderate Resolution Imaging Spectroradiometer (MODIS). Also, model LWP is much more sensitive to LCC than in MODIS in which cloud thickness is an additional important determinant of cloud water besides cloud fraction. Based on the additional insights provided by this analysis, paths forward for research into the improvement of modeling marine low-level clouds will be offered.