To assess marine boundary layer (MBL) cloud simulations in three versions of the Community Atmosphere Model (CAM), three sets of short-term global hindcasts are performed and compared to Atmospheric Radiation Measurement (ARM) program observations on Graciosa Island in the Azores from June 2009 to December 2010. The three versions consist of CAM5.3 with default schemes (CAM5.3), CAM5.3 with Cloud Layers Unified By Binormals (CLUBB-MG1), and CAM5.3 with CLUBB and updated microphysics scheme (CLUBB-MG2).
Our results show that relative to CAM5.3 default schemes, simulations with CLUBB better represent MBL cloud-base height, the height of the major cloud layer, and the daily cloud cover variability. CLUBB also better simulates the relationship of cloud fraction to cloud liquid water path (LWP) most likely due to CLUBB’s consistent treatment of these variables through a probability distribution function (PDF) approach. Sub-cloud evaporation of precipitation is substantially enhanced in simulations with CLUBB-MG2 and is more realistic based on the limited observational estimate. Despite these improvements, all model versions underestimate MBL cloud cover.
CLUBB-MG2 reduces biases in in-cloud LWP (clouds are not too bright) but there are still too few of MBL clouds due to an underestimate in the frequency of overcast scenes. Thus combining CLUBB with MG2 scheme better simulates MBL cloud processes, but because biases remain in MBL cloud cover CLUBB-MG2 does not improve the simulation of the surface shortwave cloud radiative effect.