Local Domain Forecasting Support to the 1996 Atlanta Olympic Games

John S. Snook NOAA Forecast Systems Laboratory Boulder, CO 80303

1. INTRODUCTION

The Local Analysis and Prediction System(LAPS), developed at NOAA's Forecast System Laboratory (FSL), is a data assimilation system designed to generate meso-beta scale analysis and forecasts for an area roughly the size of a typical National Weather Service (NWS) office forecast domain (McGinley et al. 1991, McGinley 1995). Operational LAPS analyses are currently being provided to the Denver NWS Forecast Office (NWSFO) and plans to provide similar analyses to the Peachtree City, Georgia, NWSFO for Olympic Games support are discussed in a companion paper (Stamus 1996). Recent improvements in affordable desk-top computer workstation technology (e.g., Cotton et al. 1994) are now making it feasible to use the operational LAPS analyses as initialization to a real-time, local domain, meso-beta scale numerical weather prediction model (Snook andent{Pielke}, 1995, Snook et al. 1995). For the Atlanta Games, the prediction portion of LAPS uses the Regional Atmospheric Modeling System (RAMS) numerical model developed at Colorado State University (Pielke et al. 1992). The LAPS/RAMS system has been running quasi-operationally at FSL for about two years.

The LAPS/RAMS system was installed at Peach-tree City in May 1995. The modeling system is designed to provide enhanced meso-beta scale forecasting capabilities at the local forecast office. This paper discusses the model configuration, the efforts to take advantage of low cost parallel computing platforms to perform model calculations, and the benefits of locally generated, high-resolution numerical weather prediction.

2. MODEL DESIGN

The aim of the LAPS real-time system is to provide local-scale guidance to the individual forecast office beyond what is currently available from coarser resolution, large-domain operational products (Snook et al. 1995). This is accomplished by generating local-domain analyses and short-range predictions of meteorological variables and, very important, other derived fields that are pertinent to the local needs of any particular weather office. For these reasons, the LAPS high-resolution domain is limited in size to cover approximately the local weather offices' area of responsibility (~500 x 500 km), and the primary goal is to add value to the short-range (0-12 h) prediction of local weather phenomena such as convection, terrain-forced circulations, and mesoscale variation of precipitation.

The primary design considerations of the LAPS system are to take advantage of all local data sources at the highest possible resolution and to run the system locally on computer workstations affordable for operational weather offices. Some compromises are still required to maintain model efficiency on today's affordable computer workstations, but further advances in workstation technology are expected to reduce these limitations in the near future. The Olympics LAPS/RAMS system has three components: (1) operational analyses (Stamus 1996), (2) real-time RAMS predictions, and (3) visualization of model output. All three components are designed to perform on the same affordable computer workstation.

2.1 Operational RAMS Configuration

A unique aspect of the LAPS/RAMS operational system is the initialization of the high-resolution model with comparable high-resolution LAPS analyses. Hence, the model grid is configured to utilize LAPS analyses with as little manipulation to the initial fields as possible. The horizontal domain is equivalent to LAPS (i.e., 85 x 85 grid points with an 8-km grid increment). The vertical grid is a stretched sigma-z coordinate system with a 300-m grid spacing nearest the ground, a stretch factor of 1.1, and a maximum grid spacing of 750 m. Twenty-five levels are employed resulting in a model top elevation of approximately 15.3 km above the surface. The vertical grid spacing is designed to maintain all the available resolution in the LAPS analyses while foregoing greater resolution to allow implementation in real time. Model topography is set equal to the LAPS terrain.

The LAPS isobaric analyses are linearly interpolated in the vertical to the RAMS sigma-z coordinate system. Horizontal interpolation is not required because the model points correspond exactly with the LAPS analysis points. Separate LAPS surface analyses are blended into the three-dimensional analyses up to 500 m AGL.

RAMS model physics are selected to complement the grid scale resolution. A nonhydrostatic version of the model is employed with a full implementation of liquid and ice microphysics (Flatau et al. 1989). A radiation parameterization scheme that includes the radiative effects of clouds is utilized. A rigid lid with a five-point sponge layer is used for the top boundary condition. A time-dependent lateral boundary condition is implemented to force the model variables toward forecast values derived from the operational 29-km Eta model provided by the National Center for Environmental Prediction (NCEP), previously the National Meteorological Center. At the surface, a 0.5 m, 11-level soil model and a vegetation model are utilized. Turbulence closure is accomplished using a deformation K scheme.

During the summer of 1995, forecasts to 18 hours were generated once a day starting at 0600 UTC. Model predictions were produced using an IBM 590 computer located at FSL, and model output was sent to Peachtree City via the Internet. Approximately 10 hours of computer resources were needed to complete the 18-hour forecast. Plans are to acquire a parallel computer workstation with at least five times the speed that will be placed in Peachtree City during the Olympic Games. Parallel computer workstations are beginning to provide an enormous increase in compute power at a reasonable cost. To take advantage of these platforms, FSL has started an effort to develop a high-level library, the Nearest Neighbor Tool (NNT), that significantly enhances the ability to parallelize the numerical forecast model and provides source code portability between a large number of existing parallel platforms (Rodriguez et. al. 1995). The parallel computer workstation will have the necessary resources to provide a more timely utilization of the 18-hour forecast.

The additional resources will also allow the completion of several more prediction cycles. The daily strategy is as follows. First, the 8-km grid increment, 18-hour prediction initialized at 0600 UTC will be available to the operational forecasters when they begin the early morning forecast shift. The model output is designed to provide an early indication of any potentially significant weather during the day. Next, the forecasters will be able to initiate several daytime, very high-resolution (2-km grid increment) forecasts for short prediction cycles of six hours. Two reduced-sized model domains will be utilized and the locations will be selectable by the forecaster based on the current weather and planned Olympic activities. For example, during the yachting events, one window will likely be positioned over Savannah to provide highly detailed wind forecasts for the yachting venue. The very high-resolution window forecasts can be restarted with new LAPS analyses through the day based on forecaster requests.

2.2 Model Output Visualization

A comprehensive visualization system that allows the forecaster to rapidly peruse the enormous amounts of numerical weather model output is important because the real-time predictions are useful for only a short time span. The recent advancements in affordable workstation technology have facilitated the development of several visualization systems that are capable of meeting these requirements. The display of LAPS/RAMS output in Peachtree City is currently accomplished through the N-AWIPS meteorological visualization system (Rothfusz et al. 1996a, Rothfusz et al. 1996b), in the testing stage at NCEP. The system is X-windows based and uses color graphics to shade and contour a variety of meteorological fields. Atmospheric state variables and derived fields from LAPS/RAMS are automatically generated in real-time using a format that is directly accessible by N-AWIPS.

FSL has been very successful in utilizing the three-dimensional Advanced Visualization System (AVS) commercial software package for the display of LAPS/RAMS model predictions. The ability to visualize the model output in three dimensions and to animate the images in time provides an excellent tool to rapidly peruse the model forecasts.

Two examples of AVS products are illustrated in Figures 1 and 2. An 18-hour surface wind forecast valid at 0000 UTC 22 August 1995 ( Fig. 1) indicates a line of convergence extending from southeast Alabama, through Georgia, and into South Carolina. The 18-hour forecast of total accumulated precipitation is represented by the contours. Prior to 0000 UTC, precipitation fell on northcentral Georgia which created an outflow boundary that propagated southeastward and eventually collided with the afternoon sea-breeze front. Precipitation then began to develop along the convergence line. Color shading of the forecast surface temperature is difficult to interpret in the black and white representation, but cooler temperatures are forecast along the coast and beneath the precipitation and warmer temperatures are predicted elsewhere.

The corresponding three-dimensional cloud and precipitation forecast is depicted in Figure 2. The shaded areas represent the regions of liquid and ice water as forecast by the RAMS explicit microphysics. The relatively flat shaded regions suggest clouds with a band of precipitation indicated beneath the clouds along the convergence line. Both figures nicely illustrate the detail available from the RAMS forecasts. Plans include implementation of these and other products in Peachtree City prior to the summer of 1996.

3. BENEFITS OF LOCALLY GENERATED NUMERICAL WEATHER PREDICTIONS

It is important to understand that the locally-produced local-domain numerical weather forecast effort is not intended to replace any guidance that is available from the NMC central modeling facility. The local domain forecasting support is designed to provide an additional mesoscale forecast tool to the suite of products already available on the meteorological workstation. The experiences at FSL and at Peachtree City have demonstrated several benefits of locally generated numerical weather predictions.

A central modeling facility typically has the responsibility to provide modeling support to numerous users. This may not allow the central facility to concentrate on particular local forecast problems. Local modeling efforts can be more flexible and forecasters can focus more on local forecast problems. For example, the Olympic strategy to enable forecasters to interactively select the model domain, model start time, and frequency of model predictions is a luxury that can only occur in the local forecast office.

The ability to interactively select model grid resolution is another capability available at the local level. Selecting very high-resolution model grids (e.g., the 2-km grid increment windows planned for the Olympics support) allows for improved predictions with greater detail. This is especially true for regions where land features significantly affect the weather, such as along the coast where the sea-breeze front typically moves inland during the afternoon (e.g., Fig. 1). Greater resolution also facilitates a more accurate representation of the cloud physics and provides a more detailed prediction of precipitation (as shown in Fig. 2).

Locally produced weather analyses and predictions greatly reduce the amount of required communications. Local data sources, such as networks of automated weather stations, are often available to a local office that may not be available to a central facility. These data sources can be incorporated into the local analyses and predictions. As computer hardware capabilities expand and the number of model grid points grow, the model output becomes increasingly larger. Communication of model output over long distances to another computer platform is not necessary when the model runs locally on the same network of computers that displays the output. This also eliminates the problem of degrading the frequency and resolution of the model output that frequently occurs when disseminating model data from a central facility to a local office. Hence, the whole flow of data into the local analyses, model initialization from the local analyses, and model output into the visualization system (i.e., the three components of LAPS) occurs in one location in a timely fashion.

4. SUMMARY

Local domain forecasting support to the 1996 Atlanta Olympic Games will be provided by the LAPS/RAMS system, a meso-beta scale data assimilation and prediction system. The system generates very high-resolution (8-km grid increment and smaller) analyses and forecasts within the local NWSFO on an affordable computer workstation. The system is designed to provide enhanced meso-beta scale forecast capabilities to the NWSFO and to the Olympic Games. Benefits of running the system in the local forecast office include increased numerical model grid resolution, greater flexibility to focus on local forecast problems, and reduced data communications resulting in more timely analyses and short-range numerical weather predictions to the forecaster. Efforts to take advantage of faster, low-cost parallel computer workstations are underway which will provide the necessary resources to provide the enhanced level of forecasting support.

5. ACKNOWLEDGEMENTS

Phil McDonald and Paula McCaslin have provided support in the applications of AVS. Jim Edwards and Jerry Schmidt are providing support toward the parallel computing effort. Pete Stamus reviewed the article and Juanita Fullerton provided technical editing support. The author wishes to thank Drs. Roger Pielke and William Cotton for their permission to use RAMS for this project.

RAMS was developed under the support of the National Science Foundation (NSF) and the Army Research Office (ARO).

6. REFERENCES

Cotton, W. R., G. Thompson, and P. W. Mielke, Jr., 1994: Real-time mesoscale prediction on workstations. Bulletin Amer. Meteor. Soc., 75, 349-362.

Flatau, P. J., G. J. Tripoli, J. Verlinde, and W. R. Cotton, 1989: The CSU-RAMS cloud microphysical module: General theory and code documentation. Dept. of Atmospheric Science Paper No. 451, Colorado State University, Fort Collins, CO, 88 pp.

McGinley, J. A., 1995: Opportunities for high resolution data analysis, prediction, and product dissemination within the local weather office. Preprints, 14th Conference on Weather Analysis and Forecasting, 15-20 January 1995, Dallas, TX, Amer. Meteor. Soc., 478-485.

McGinley, J. A., S. C. Albers, and P. A. Stamus, 1991: Validation of a composite convective index as defined by a real-time local analysis system. Weather and Forecasting, 6, 337-356.

Pielke, R. A., W. R. Cotton, R. L. Walko, C. J. Tremback, W. A. Lyons, L. D. Grasso, M. E. Nicholls, M. D. Moran, D. A. Wesley, T. J. Lee, and J. H. Copeland, 1992: A comprehensive meteorological modeling system - RAMS. Meteor. Atmos. Phys., 49, 69-91.

Rodriguez, B., L. Hart, and T. Henderson, 1995: Programming regular grid-based weather simulation models for portable and fast execution. Proceedings of the 1995 International Conference on Parallel Processing, Volume III Algorithms and Applications, K. Gallivan (ed.). CRC Press, Inc., Boca Raton, FL, III-51-III-59.

Rothfusz, L. P., J. T. Johnson, L. C. Safford, M. R. McLaughlin, and S. K. Rinard, 1996a: The Olympics weather support system. Preprints, 12th International Conference on Interactive and Processing Systems (IIPS) for Meteorology, Oceanography and Hydrology, Atlanta, GA, Amer. Meteor. Soc., 1-6.

Rothfusz, L. P., J. Noffsinger, M. Peroutka, S. Schotz, and J. T. Johnson, 1996b: Weather information display, analysis and product generation tools used in support of the 1996 Olympic Games: Daily forecasts. Preprints, 12th International Conference on Interactive and Processing Systems (IIPS) for Meteorology, Oceanography and Hydrology, Atlanta, GA, Amer. Meteor. Soc., 7-12.

Snook, J. S. and R. A. Pielke, 1995: Diagnosing a Colorado heavy snow event with a nonhydrostatic mesoscale numerical model structured for operational use. Weather and Forecasting, 10, 261-285.

Snook, J. S., J. M. Cram, and J. M. Schmidt, 1995: LAPS/RAMS: A nonhydrostatic mesoscale numerical modeling system configured for operational use. Tellus, 47A, 864-875.

Stamus, P. A., 1996: Providing weather support to the 1996 Atlanta Olympic Games with the Local Analysis and Prediction System (LAPS). Preprints, 12th International Conference on Interactive and Processing Systems (IIPS) for Meteorology, Oceanography and Hydrology, Atlanta, GA, Amer. Meteor. Soc., 36-39.