Fourth Conference on Coastal Atmospheric and Oceanic Prediction and Processes

Friday, 9 November 2001: 10:45 AM
SENSITIVITY OF MARINE BOUNDARY LAYER SIMULATIONS IN A MESOSCALE MODEL TO INITIAL CONDITIONS AND MODEL PHYSICS
Jeremy D. Ross, Penn State University, University Park, PA; and N. L. Seaman and D. R. Stauffer
Poster PDF (1.6 MB)
The non-hydrostatic Pennsylvania State University / National Center for Academic Research (PSU/NCAR) mesoscale model, MM5, is used to generate short-range mesoscale simulations of coastal conditions off of Southern California during the Variability of Coastal Atmospheric Refractivity (VOCAR) experiment. The large vertical gradients in temperature and moisture near the top of the marine boundary layer (MBL) affect the refractivity structure resulting in anomalous electromagnetic (EM) propagation. The ability to accurately forecast the refractivity structure is of vital importance for radar and communications. It is therefore necessary to obtain high quality temperature and moisture data from mesoscale model simulations. The goal of this study is to generate several sets of model solutions that are able to capture the highly variable structure of the MBL during the period 23 August – 4 September 1993. This sensitivity study contains 14 members created from using different combinations of model physics and initial/lateral boundary conditions (IC/LBC’s): two 1.5-order (TKE) turbulence parameterizations, two 1st-order turbulence parameterizations, and initial and lateral boundary conditions from the MRF, ECMWF, and NOGAPS global models. Results from the U.S. Navy’s Coupled Ocean Atmosphere Prediction System (COAMPS) mesoscale model are also included for independent comparison. In some experiments, the MM5 uses a marine boundary layer initialization (MBLI) where a summertime inversion-base climatology of the eastern Pacific is used to produce a more realistic MBL structure in the oceanic data-void region.

The MM5 is configured with two nested domains that have resolutions of 36 km and 12 km. The 36-km domain extends from the central Pacific to the eastern United States, and north from Central America to Canada. The 12-km domain is much smaller, focusing on the Southern California bight. The model contains 53 vertical sigma layers, separated by approximately 40 meters in the lowest 600 meters. During the 12-day period, the model is initialized four times, with each simulation lasting 72-108 hours. The length of the runs allows the model to simulate changes in the marine-layer structure caused by synoptic scale and mesoscale forcing. Four dimensional data assimilation (FDDA) based on the global analyses and conventional data is performed above the boundary layer throughout each period on both the 36-km and 12-km domains. This helps provide an accurate representation of the synoptic scale, while allowing the mesoscale processes affecting the MBL to be simulated with different turbulence parameterizations.

Model results are compared with special rawinsonde data acquired during the VOCAR period. These data were excluded from the model initialization and FDDA, and therefore represent an independent data source for verification. Results suggest that the offshore data-void regions have a significant impact on the complex structure of the MBL. The use of different IC/LBC’s produces variations as large as those found from changing the model physics. Although the sensitivity experiments maintain consistent synoptic-scale trends, varying the model physics influenced the mesoscale structure of the MBL, with the higher-order turbulence parameterizations producing more accurate representations of the MBL. The potential value of using this set of model experiments as a mesoscale ensemble will also be investigated.

Supplementary URL: