| Summary: | dissertation Three studies were performed which, in different ways, evaluated System for Atmospheric Modeling (SAM) model performance in a variety of cases: 1) CONSTRAIN, a North Atlantic marine cold air outbreak case, 2) radiative convective equilibrium (RCE) simulations, and 3) Dynamics of the Madden-Julian Oscillation (DYNAMO) shallow cumulus. In the CONSTRAIN study, a dozen different model physics setups were used, some of which were compared with sets of runs with varying turbulence parameterization scheme and varying grid spacings. In the SHOC vs NOSHOC comparisons, the choice of parameterization scheme had little influence on runs without cloud ice. LES-scale run comparisons between the different model physics showed runs with radiation increased the precipitation and cloud cover but more precipitation reduced cloud cover causing these effects to largely cancel. Ice sedimentation increased precipitation while decreasing cloud amount and entrainment. Double-moment microphysics runs resulted in more supercooled water and less ice. For the RCE simulation, model runs were performed varying in sea surface temperature, turbulence parameterization scheme, microphysics scheme, and grid spacing. Grid spacing had a large influence on cloud water path and SW radiation. The microphysics scheme selection had a large influence on CWP and IWP, shifting to greater CWP and less IWP in the double-moment runs. SHOC and double-moment microphysics produced a higher upper-tropospheric cloud fraction. Using radiative kernels to evaluate cloud feedbacks the runs with SHOC had a negative net cloud feedback while the single-moment NOSHOC run had a positive net cloud feedback. For DYNAMO, shallow cumulus in calmer periods of the Madden-Julian oscillation (MJO) at Gan Island were selected as case studies based on satellite imagery, ground-based sky imagers, and combined KAZR/S-Pol radar data. Model runs at 2, 1, 0.5, and 0.1 km grid spacing were performed for the two cases. The model tended to underestimate midlevel cloud in Case 1 and overestimate shallow cumulus in Case 2 except for the high resolution runs which overestimated shallow cumulus for both. For larger grid spacings, the observation cloud profile generally stayed within one standard deviation of the model in the lowest 4 km.
|
|---|