(The following is excerpted from the final report on the second year of Dr. Doggett's fellowship)
Following the successful development of forecasting techniques for general convection and elevated convection during the first year of research, the focus of research in the second year of the fellowship turned to more subtle mechanisms of convective initiation In particular, forecasting convective initiation along the West Texas was of interest. As laid out in the project proposal this would be approached in a manner that utilized both available model data and real time data analysis techniques.
Initially, a suite of surface analyses were put together to aid in the identification of the dryline. These products included the standard parameters of dewpoint and mixing ratio that are typically used to identify dryline position. In addition, new analyses were also examined, including the parameters of moisture convergence, virtual temperature, streamline analyses, and three-hour dewpoint and mixing ratio changes. These fields were developed through interactions with operational forecasters, who provided input both on what parameters they have found useful in the past and on parameters they would like to have but were currently unavailable.
From these analyses it was found that several of these products showed promise in identifying the position, movement, and relative intensity of the dryline. In particular, examination of gradients in the difference fields between temperature and virtual temperature was useful in delineating between the hot dry air west of the dryline and the cooler, more humid air east of the dryline. In addition, moisture convergence, streamlines, and three-hour moisture change fields were important in tracking dryline motions.
Attempts to use these products to identify regions of convective initiation quickly showed the limitations of the data set. While useful in defining the general properties of the dryline, available data are not of a fine enough resolution to accurately define forcing of the scale of convective initiation. To pursue this research further, two sets of actions were taken. First, an attempt was made to gather data on a finer scale then was currently available. Second, applications of mesoscale modeling in forecasting the dryline and other convective activity were examined.
Coincident to this project, though independent of the fellowship investigation, two new data sets were identified that increased the resolution of observations in the Lubbock area. Permission to access the Texas Department of Transportation highway observation network was gained, which provided twelve new observation sites for the Texas South Plains and Panhandle area. These data, through careful supervision and quality control by the forecasters, proved useful in better defining dryline location. Unfortunately, access to these data was limited and these data were not able to be included in the surface analysis products directly.
The second data source was an additional automated station that was acquired from Southern Region Headquarters and placed in a data poor region to the east of Lubbock. Data from this single station improved the quality of the surface analyses in many instances. The success of this station led to the investigation of developing a full mesonetwork in the region. As a result of several meetings with individuals at Texas A&M University, both Texas Tech University and the Lubbock National Weather Service Forecast Office were included in the development of the Texas MesoNet project. In particular, plans were made over the summer to use the Lubbock area as a pilot region for a statewide network. The pilot network would instrument between 30 and 40 sites across the region at a resolution of 50 km and would include both surface observations and atmospheric profiler sites. Although no network was established during the time period of this research project, participation of all parties continues and a regional mesonet appears to be imminent.
The second focus of research this year was the implementation of a mesoscale model. The RAMS model was chosen due to its availability and access to support at FSL. Initially the model was set up to run as a single 27-km grid centered on Lubbock and covering an 800-km by 800-km area. The model used Eta grib files for both initialization and boundary conditions. RAMS was run twice a day upon the arrival of the Eta grib files, and one-hour forecast fields were produced out to eighteen hours ahead of the initial time. Typically forecasts were available to the forecasters six-hours into the forecast cycle, so that an eighteen-hour forecast derived from 1200 UTC data would be available by 1800 UTC. The major limiting factor in running RAMS in an operational setting was the availability of computing power in the office. With two HP workstations, an HP-715 and an HP-755, the Lubbock office is in a better position then many NWS offices to run a mesoscale model. However, with all the real time data processing required in an operational setting, there was still a limit to the resolution, domain size, and forecast duration that could be obtained.
Running in this mode RAMS showed a significant level of skill in predicting the development and nature of larger scale convective activity but did not have the resolution necessary to model convective initiation at the dryline scale. Feedback from the forecasters indicated that even at this scale the model output was an extremely useful forecast tool to have available. GEMPAK analyses and GARP scripts were developed to make the RAMS output easily accessible to the operations staff.
To improve the resolution of the model, RAMS was configured to run nested grids, with a coarser outer grid having a resolution of 36 km and a finer inner grid of 12 km. The domain of the outer grid remained the same as the previous configuration, while the inner grid covered about a 90 km by 90-km region centered within the outer domain. Otherwise, the model configuration remained the same and still operated under the same computing constraints. While the new configuration seems to provide more detail in defining precipitation events, the configuration was not implemented in time to run during the convective season. A better determination of the effects of the new configuration will be able to be demonstrated during the spring of 1998.
During meetings with Conrad Zeigler of NSSL on using RAMS to research dryline forcing several issues were discussed. The first concerned the manner in which vegetation is handled by the model and the importance this might have on model forecasts. The second issue dealt with better initialization of soil moisture, especially in terms of rainfall accumulation during the preceding day. It was hoped that WSR88D data could be used to better define this rainfall in near real time. Neither of these topics was ever acceptably dealt with and will limit the reliability of these mesoscale forecasts. This was due in part to the computing constraints mentioned and the early conclusion of the fellowship.
In conclusion, although much work was done to help forecasters deal with the problem of convective initiation, the problems dealt with in the second year were far more difficult to approach. The resolution of available data and limitations of computing power at local NWS offices limited the ability to fully examine the problem of fine mesoscale forcing. It is hoped that this work has set the foundation for further investigations on the subject, especially in regards to utilizing any mesonet data that may be obtained in the near future.