Convection and cloud processes are examined in a hierarchy of two-dimension
al numerical realizations of cloud systems observed during the 19-26 Decemb
er 1992 period of the Tropical Ocean Global Atmosphere Coupled Ocean-Atmosp
here Response Experiment. The hierarchy consists of cloud-resolving simulat
ions at a 2-km resolution, and two sets of 15-km resolution simulations; on
e attempts to treat convection explicitly and the other parameterizes conve
ction using the Kain-Fritsch scheme.
The Kain-Fritsch parameterization shows reasonable results but shortcomings
are found in comparison with the cloud-resolving model. (i) The entraining
plumes in the parameterization excessively overshoot the tropopause, which
produces a cold bias mostly through adiabatic cooling. The attendant moist
ure detrainment over-produces cirrus cloud. (ii) Because parameterized down
drafts detrain at the lowest level they generate a surface cold bias. (iii)
The scheme fails to represent the trimodal convection (cumulonimbus reachi
ng the tropopause, cumulus congestus around the melting level, and shallow
convection regimes) realized by the cloud-resolving simulation and also see
n in observations. The lack of shallow convection and cumulus congestus lea
ds to an overprediction of the low-level moisture. (iv) The simulations are
sensitive to the magnitude of moisture feedback from the convective parame
terization to the grid scale but less sensitive to whether the moisture is
in vapor or condensed phase.
These deficiencies are mostly a consequence of the single-plume model that
represents updrafts and downdrafts in the parameterization scheme, along wi
th the lack of a shallow convection scheme. A more realistic model of entra
inment and detrainment that reduces overshoot and represents the cumulus co
ngestus is required. Realistic downdraft detrainment and relative humidity
are needed to improve the downdraft parameterization and alleviate the surf
ace temperature bias.