Using TWP-ICE observations with model intercomparison to improve simulations of tropical oceanic convection

We present a detailed comparison of radar data sets obtained during the Tropical Warm Pool – International Cloud Experiment (TWP-ICE) with nine simulations of monsoonal convection performed by four cloud-resolving models (CRMs, with fully periodic boundary conditions) and one limited area model (LAM, with open boundary conditions and multiply nested domains). Using scanning precipitation radar, we first identify several glaring differences between the simulated and observed distributions of radar reflectivity. The largest differences exist in stratiform ice regions, where modeled reflectivities are too weak, too shallow, and not sufficiently widespread. This may be related to the opposite problem identified in convective regions, where modeled reflectivities are too high aloft and cover too much area. Simulated updraft size and speed are relatively realistic when compared with conservatively analyzed retrievals from a dual-Doppler radar system, suggesting that bulk microphysics scheme assumptions are likely to be the primary problem.

Before the schemes can be improved, the specific assumptions that are likely the primary source of the problem must be discovered. This is addressed via model sensitivity tests by Smith-Mrowiec et al. at this conference and via intercomparison of different models and schemes with additional observational data such as satellite data, radar-retrieved DSDs, surface station data, and soundings. A fundamental hypothesis is that microphysical characteristics in the model convective cores are having a deleterious effect on stratiform strength and extent. This hypothesis is examined through intercomparison of model microphysics assumptions and variables affecting stratiform development, and all findings are placed in the context of the foregoing differences between model fields and observations.