The Validation of the Simulated Microphysical Structure within the Midlevel Inflow Region of a Tropical, Oceanic Squall Line

Barnes and Houze [2014] demonstrated that microphysical processes within mesoscale convective systems (MCSs) observed over the Indian Ocean during the Dynamics of the Madden-Julian Oscillation / ARM MJO investigation experiment (DYNAMO/AMIE) were organized in a repeatable, systematic manner with respect to the kinematic structure of these storms using the dual-polarimetric NCAR S-PolKa radar. A version of the NCAR particle identification algorithm (PID), modified for the central Indian Ocean, was composited around the kinematic structure of these storms, which was identified using radial velocity. Using this spatial compositing technique, Barnes and Houze [2014] showed that the region of the containing the midlevel inflow is characterized by decreasing rain intensity towards the anvil and wet aggregates, dry aggregates, and small ice crystals layered sequentially above the melting layer. Interestingly, it was also noted that shallow pockets of rimed particles are found just about the melting layer. Given that these microphysical structures were observed in more than twenty MCSs observed during DYNAMO/AMIE, the current study investigates if the microphysical processes simulated by the Weather Research and Forecasting Model (WRF) have a similar structure.

The current study focuses on a series of intense leading-line trail stratiform MCSs that were observed by the NCAR S-PolKa radar on Addu Atoll in the Maldives from 1000 UTC 23 December 2011 through 0000 UTC 24 December 2011. In order to compare the structure of the simulated microphysical processes to the structure observed during DYNAMO, it is essential that the simulated midlevel inflow matches the observed midlevel inflow as close as possible. To achieve this requirement, we use the Pennsylvania State University Ensemble Kalman Filter Version of WRF (PSU EnKF WRF) to assimilate radial velocity data obtained using the NCAR S-PolKa radar every fifteen minutes from 1800 UTC December 2011 until 0000 UTC 24 December 2011. The simulation has an outer domain at 3 km resolution and an inner domain at a 1 km resolution.

By assimilating only radial velocity data, we are able to greatly improve WRF's representation of the midlevel inflow in these squall lines. Given that we do not assimilate radar reflectivity, the microphysical structure of these storms are allowed to evolve freely in response to the simulated airflow. Thus, our experimental design allows us to specifically investigate how different microphysical parameterization schemes organize microphysical processes around the midlevel inflow of this simulated squall line. Nine simulated microphysical processes are analyzed in this study: melting, aggregation, riming, sublimation, deposition, evaporation, condensation, ice nucleation, and rain auto-collection. These fields were obtained by modifying the microphysical parameterization schemes so that these variables, which are already calculated within the parameterization, are output by WRF. The microphysical parameterization schemes analyzed in this study include the Goddard, Morrison 2-moment, Milbrandt-Yau double moment, WRF single moment 6-class, and WRF double moment 6-class schemes.

Preliminary results from the Goddard scheme indicate that simulated microphysical processes also have a systematic organization around the midlevel inflow of this oceanic squall line. Some of the simulated microphysical structures are consistent with observations. For example, melting particles, aggregations, and newly formed ice occur in subsequent layers above the melting layer. However, there are some discrepancies between the structure of the simulated microphysical processes and observations. For example, riming in the WRF simulation occurs from the melting layer through 12 km, which is much deeper than observations.

Analyses such as these are important since they provide a new way to use radar data to valid numerical simulations. Directly comparing simulated snow, ice, and graupel to PID data is not possible since simulated and radar-observed hydrometeors are defined in fundamentally different ways. However, by interpreting the PID fields as microphysical processes and comparing those fields with the simulated microphysical processes, we can more directly validate the microphysical parameterizations.