| Description |
Ten separate idealized cloud-resolving model (CRM) and four separate nested limited area model (LAM) three-dimensional simulations having horizontal grid spacing of ~1 km and ~75-100 vertical levels are compared to observations during the active monsoon period of the Tropical Warm Pool - International Cloud Experiment, based in Darwin, Australia, with specific focus on a large mesoscale convective system observed on January 23-24, 2006. All simulations produce high biased convective radar reflectivity and low biased stratiform rainfall with these biases heavily modulated by bulk microphysics scheme assumptions. High biased convective radar reflectivity aloft always involves a graupel/hail component, but also includes a snow component for some two-moment schemes. Making snow particle mass proportional to ~D2 rather than D3 may lower snow reflectivity. This high bias is also related to freezing of very large simulated rain water contents in deep convective updrafts. Peak vertical velocities are greater than dual-Doppler retrieved values, especially in the upper troposphere likely due to greater latent heating from freezing and deposition in simulations. A subdomain LES simulation also produces overly intense simulated updrafts. Therefore, they may be a product of interactions between convective dynamics and parameterized microphysics that promote a different convective mode and strength than observed, while inadequately simulated instability and vertical shear variability may also be involved. Two-moment schemes do not outperform one-moment schemes in stratiform rainfall prediction. Excessive size sorting produces more large stratiform raindrops at low levels than observed in two-moment schemes. One-moment schemes produce too many small stratiform raindrops relative to observed because constant size intercepts are too high. Increasing the rain gamma shape parameter from 0 to 2.5 improves agreement with observations. Due to differences in raindrop size that create different mass sedimentation rates, low-level stratiform liquid water contents are close to observed in one-moment schemes, but lower than observed in two-moment schemes. Low biased stratiform rainfall is primarily due to an under-prediction of melting ice consistent with the lack of a large well-developed stratiform region in simulations. This may be caused by overly intense simulated convection, limited domain size in the CRM simulations, and large-scale forcing biases in the LAM simulations. |
