Using satellite cloud products to evaluate cloud types simulated by a cloud resolving model and a single column model

Cloud radiative forcing (CRF) is determined by the occurrence frequency of cloud types and their diurnal variation. We determined the occurrence frequencies and CRFs of various cloud types defined by cloud-top pressure and cloud optical depth using the pixel-level VISST/SIRS GOES satellite cloud products from Pat Minnis' group (NASA Langley Research Center) at the Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP) site during the summer 1997 single column model (SCM) Intensive Operational Period (IOP). We ran a SCM version of the National Centers for Environment Prediction (NCEP) Global Forecast System (GFS) model and the UCLA/CSU cloud resolving model (CRM) using the ARM variational analysis for the summer 1997 SCM IOP. We diagnosed the SCM and the CRM simulated cloud type frequencies using the ISCCP cloud type simulator. We further analyzed the SCM and CRM simulated cloud radiative forcings. These results are compared to those estimated from the satellite observations.

During the ABC subperiods, the CRM and SCM had comparable mean errors in LW CRF (-7.1 and -7.6 W/m2). However the rms error in the LW CRF for each cloud type in the CRM was about one third of that in the SCM (4 vs 13 W/m2). Compared to the CRM, the SCM had larger mean errors for the SW CRF (17 vs 12 W/m2) and the net CRF (9 vs 5 W/m2), and had about three times greater rms error by cloud types for the SW CRF (22 vs 8 W/m2) and twice the rms error for the net CRF (11 vs 5 W/m2).

Compared to the satellite observation, the SCM had the same cloud amount when averaged over the 14-day subperiods, while the rms error by cloud types are larger than the CRM (0.15 vs 0.06).The SCM COF for low clouds is close to the observation in daytime but overestimated during night, while the CRM's is too low in both day and night due to low horizontal resolution and the omission of its SGS cloud scheme.

The SCM errors in CRF and cloud amount are due to: 1) the SCM simulated too few optically thick high-top clouds in daytime, and underestimated the optical depth for this cloud type. As a result, the TOA CRF of these clouds had a warming effect in the SCM, which is contrary to the observations and the CRM results. 2) More optically thin-to-moderate high clouds were simulated in the SCM than observed. 3) Too few optically thick clouds were simulated by the SCM at mid- and low-levels. 4) The SCM thin mid-level clouds had CRF close to the observation and the CRM due to compensating errors: larger optical depth and less occurrence frequency.