Atmospheric models are getting progressively more sophisticated as computational power has grown exponentially over the past few decades. The module responsible for predicting the evolution of hydrometeors, the microphysics scheme, has also gone through a series of updates. However, in the operational forecasting world, its moment-resolving structure has remained largely unchanged due to its high computational efficiency compared to a bin scheme or the more recently developed Lagrangian superdroplet method, without losing too much accuracy. In this dissertation, we seek to improve the bulk scheme in its current state and also give guidance to its future development, from an improved shape parameter diagnostic equation, to understanding the structural limitation of bulk schemes in general, and to exploring a more natural representation of a droplet size distribution that could potentially be the beginning of a paradigm shift for bulk schemes in numerical weather prediction.
Observational records from five in-situ flight campaigns were used to improve our understanding of aerosol-dispersion effect and the shape parameter diagnostic equation in a bulk scheme. We identify the reason why past studies found inconsistent sign of the aerosol-dispersion correlation by distinguishing local correlation from cloud-mean correlation. No clear correlation is found between cloud-mean aerosol concentration and cloud-mean dispersion, while in-cloud regions with high droplet dispersion are typically associated with low local droplet concentration. A new equation is proposed to diagnose the shape parameter in marine warm stratocumulus clouds for double moment bulk schemes in cloud-resolving models or large-eddy simulations.We use a special bulk scheme called Arbitrary Moment Predictor (AMP) to understand the effect of structural differences between bin and bulk scheme. One- dimensional kinematic simulations show that overall differences between AMP (bulk) and bin schemes are small. They produce similar mean liquid water path, but AMP has a significantly lower mean precipitation rate due to slower precipitation onset. The primary reason for the divergence between the two schemes is found to be the collision-coalescence process, particularly autoconversion. By adjusting the diameter cutoff between cloud and rain categories in AMP, the largest difference between the two schemes can be reduced to about 10%, making the structural differences between AMP and bin schemes smaller than the parameterization differences between two bin schemes.
To explore the feasibility of a bulk scheme with unified liquid water category, we upgrade AMP to be configurable as a unified category bulk scheme (U-AMP) in addition to its original separate category setting (S-AMP). Due to our lack of knowledge of such a novel scheme, we first perform an extensive search for the best configuration regarding the optimal moments to predict and the best assumed shape parameter under a variety of initial conditions. When configured optimally, U-AMP far outperforms S-AMP in estimating precipitation rate, with which S-AMP struggles the most. U-AMP is also on par with S-AMP in estimating other quantities such as CWP, RWP, and LWP, where S-AMP already makes near-perfect predictions compared to bin schemes. U-AMP's superiority over S-AMP in simulating certain precipitation patterns is mostly due to its ability to situations where multiple modes exist within the traditional rain category (D > 80 μm), which can occur when large raindrops fall through a layer of small raindrops. Such structural improvements by U-AMP could mark the beginning of a shift in the paradigm for bulk scheme development.
