Evaluation and Comparison of Preliminary Meteorological Modeling for the August 2000 Houston- Galveston Ozone Episode
Institutional Repository Document
Overview
Identity
View All
Overview
abstract
This report is an evaluation of the quality of MM5 simulations of weather phenomena during the August 2000 Houston-Galveston Ozone episode. The report serves two purposes: first, to help guide final selection of a model configuration, and second, to evaluate the viability of MCNCs real-time forecasting system as an alternative meteorological model for regulatory work. The TAMU and MCNC modeling efforts are both based on the MM5 model. Primary differences involve the incorporation of analysis information, the size of the domains, the boundary layer parameterizations, and the soil moisture specifications. Data used for the model evaluation include profiler data, surface meteorological data, radiosonde data, radar and satellite imagery, and doppler lidar data. The profiler and doppler lidar data sets are still evolving, and subsequent use of profiler data in this modeling effort will require consideration of both quality-controlled and non-qualitycontrolled data. The analysis of weather phenomena is conducted using the GEMPAK software package and TAMU-written converters and scripts. Overall, the TAMU simulations using the MRF PBL scheme (dec6grid4 and dec30grid4) performed best, with the Gayno-Seaman (dec16grid4) and MCNC simulations deficient in various critical errors. Precipitation was simulated remarkably successfully with the MRF PBL schemes; seven to eight days out of ten had no significant precipitation errors. The Gayno-Seaman run had the proper temporal variation but produced too much precipitation. The MCNC model did not properly simulate the squall line on August 24, produced rain in the wrong place on August 25, and failed to produce any rain at all on succeeding days. None of the model runs produced the observed outflow boundary on the evening of September 1, and all but the MCNC model produced an erroneous outflow boundary on the previous evening. Most clouds during the ozone episode were fair weather cumulus which formed at the top of the boundary layer in the morning and dissipated in late afternoon. None of the model runs can resolve these clouds, and as a result the models produce clouds with too large a horizontal extent. Under such circumstances, the simulation with the fewest clouds is usually the best, and in this case it was the runs with the MRF PBL scheme. None of the models had the proper day-to-day variations in cloudiness. The three TAMU runs had essentially zero maximum temperature bias, while the bias for the MCNC run was on the order of 3 C. This bias likely originates from the use of the default soil moisture. All model runs were able to track day-to-day temperature variations. It is recommended that temporally-varying soil moisture be used in subsequent model runs. Large-scale temperature patterns were generally forecasted well, except for the MCNC model in some cases. The land-sea contrast was well-simulated, except for MCNC, which had too small a contrast. Another major source of variability was the urban heat island, which during this episode was observed to be essentially nil during the day and significantly warmer than its surroundings at night. No model simulation produced a warm nighttime heat island, possibly because no simulation was able to get surrounding areas cool enough. Runs with relatively dry urban soil produce too strong a 3 daytime heat island; runs with relatively moist urban soil produce a late afternoon and evening urban cool island. Comparative analysis of soundings confirmed that the models nighttime temperature inversions were too weak or nonexistent. The MCNC runs were particularly deficient in that regard. During the day, it was found that the MRF PBL runs produced bias-free mixed-layer depths, while the Gayno-Seaman and MCNC simulations were too shallow by nearly 20% on average. Wind simulations on most days were good; frequently the wind field will evolve into a more accurate configuration as the model atmosphere responds to local forcing. A close look at August 25 and August 30 suggests that one or both might be successfully simulated with a photochemical model using the current, preliminary TAMU grids. The MCNC model had erroneously weak sea breeze winds because land areas were not heating up sufficiently. The vertical structure of the wind was dominated by a diurnally-varying component which seems to have large vertical extent on some days and very shallow extent on others. Further analysis of the diurnal wind cycle and the nocturnal low-level jet will be described in a subsequent report.
