Preprint, 11th Conf. on Num. Wea. Prediction
19-23, August 1996, Norfolk, VA
Ameri. Metero. Soc., 294-296.

THE 1996 CAPS SPRING OPERATIONAL FORECASTING PERIOD: REALTIME STORM-SCALE NWP, PART I: GOALS AND METHODOLOGY

Kelvin K. Droegemeier^+,1,2, Ming Xue¹, Keith Brewster¹, Yuhe Liu¹, Seon Ki Park¹, Frederick H. Carr^1,2, John Mewes¹, Jinxing Zong¹, Adwait Sathye¹, Gene Bassett², Min Zou¹, Richard Carpenter³, Dennis McCarthy⁴, David Andra⁴, Paul Janish⁵, Rich Graham⁷, Sergiu Sanielvici⁶, Janet Brown⁶, Bruce Loftis⁶, and Ken McLain⁶

¹Center for Analysis and Prediction of Storms, ²School of Meteorology, and
³Center for Computational Geosciences,
University of Oklahoma
Norman, OK 73019

⁴NOAA National Weather Service Foreast Office and ⁵National Storm Prediction Center
Norman, OK 73069

⁶Pittsburgh Supercomputuing Center
Pittsburgh, PA 15213
⁷Cray Research, Inc.
Eagan, MN 55121

1. INTRODUCTION

The Center for Analysis and Prediction of Storms was established at the University of Oklahoma in 1988 as one of the National Science Foundation's first 11 Science and Technology (S&T) Centers. Its mission is to demonstrate the practicability of storm-scale numerical weather prediction and to develop, test, and validate a regional storm-resolving forecast system appropriate for operational, commercial, and research applications. Its ultimate vision is to make available a fully functioning, multi-season storm-scale NWP capability around the turn of the century.

As a partial means to achieving these goals, CAPS began in 1993 a series of daily, realtime operational tests of its Advanced Regional Prediction System (ARPS; Xue et al., 1995) over the southern Great Plains in collaboration with the National Weather Service and other local NOAA elements (Janish et al., 1994; Droegemeier et al., 1996; Xue et al., 1996a). Since that time the tests have grown in sophistication and realism until, during the 1996 spring season predictions, CAPS for the first time utilized realtime wideband WSR-88D Doppler radar data.

We describe herein the goals and methodology of the 1996 "Spring Operational Period (SOP '96)", which ran from 15 May through 6 June, and refer the reader to a companion paper in this volume (Xue et al. 1996b) for the operations summary and results from selected cases.

2. GOALS AND DIFFERENCES FROM THE 1995 OPERATIONAL TESTS

The SOP '96, performed in collaboration with the National Weather Service Forecast Office and National Storm Prediction Center in Norman, OK, differed from the 1995 tests in a number of key ways, including: improvements in model numerics, radiation and surface physics; use of realtime broadband (Level II) data from the KTLX (Twin Lakes, Oklahoma) WSR-88D radar; use of NIDS (Level III) digital wind data and reflectivity; application of the new ARPS Data Analysis System (ADAS), a continuously-operated data ingest, quality control, and objective analysis package designed to provide the background state for Doppler radar data; use of realtime Oklahoma Mesonet data; and application of the ARPS single-Doppler velocity retrieval and forward-variational data assimilation system (Shapiro et al., 1996, this volume) in non-realtime.

The collaboration with NOAA elements in Norman provided an opportunity to address a number of key operational and scientific questions and issues during the SOP '96, including but not limited to: a) can a storm-scale (e.g., 1 to 3 km resolution) model add value to products currently available from the NWS? b) what model products are most valuable to operational forecasters? c) is it possible to predict convective initiation? d) do single-Doppler radar and dense surface mesonet data improve storm-scale forecasts? e) what computational strategies are most appropriate for realtime storm-scale NWP?

Data collected during these and previous operational tests are being used in a number of studies at CAPS that seek to address key issues associated with storm-scale NWP, including the applicability of ensemble forecasting, the impact of various types of data on forecast quality and small-scale predictability, the use of satellite data for retrieving ambient water substance fields, convective initiation, quantitative precipitation forecasting, the practical predictability of wintertime weather, and the development of a continuous or semi-continuous data assimilation cycle.

3. OPERATIONAL CONFIGURATION

Two 7-hour forecasts (17Z to 00Z) were produced each day during the SOP '96 using the two computational domains shown in Figure 1: a 9 km horizontal resolution outer grid that was fixed in location over an area 864 x 864 km², and an inner, relocatable, storm-resolving 3 km resolution grid of size 288 x 288 km². When convection was anticipated in central Oklahoma, the latter was positioned over the KTLX (Twin Lakes, Oklahoma) WSR-88D radar. The vertical grid resolution for both domains varied over 43 levels from 20 m near the ground to around 1 km at the top of the model (20 km), and the inner grid was nested one-way within the outer grid.

Also shown in Fig. 1 is the 9 km resolution ARPS Data Analysis System (ADAS) domain. ADAS (Brewster, 1996) is a Bratseth successive correction scheme (Bratseth, 1986) designed to be modular with ARPS. It is run continuously with graphics products made available on the World Wide Web. When active weather was anticipated over central Oklahoma, ADAS was also run at 3 km resolution to provide analyses more compatible in resolution with the WSR-88D data.

Each forecast during the SOP '96 was produced as a "cold start," i.e., no continuous assimilation cycle was used whereby a previous ARPS forecast could serve as the first guess for the next cycle. This was done as a first step toward a continuous or semi-continuous cycle, the testing for which is underway and may be evaluated in realtime during the spring 1997 operational period. The background fields for the outer grid (at 17Z) were provided by the 12Z run of NCEP's 60 km resolution rapid update cycle (RUC) model (Benjamin et al., 1994). These fields were interpolated to the ARPS grid for use by ADAS, to which were added all other available data, particularly those from the Oklahoma Mesonet.

Figure 1. Configuration of the ARPS prediction grids used in the SOP '96 operational tests. The 9 km resolution ADAS and outer ARPS grid covered essentially the same area. Within the 9 km grids was a one-way nested 3 km resolution, relocatable inner ARPS grid that was typically positioned over the KTLX WSR-88D radar (dot with cross hairs).

Because realtime broadband WSR-88D data were available only from the KTLX radar, alternative methods were sought to improve the wind field over a larger portion of the ADAS and ARPS domains. As a result, toward the end of the operational period, the lowest four tilts of digital Level III WSR-88D radial wind data were brought into the ADAS from several other radars. The reflectivity fields were also used to enhance the humidity analysis in rainy areas.

The lateral boundaries of the outer ARPS grid were forced by the RUC forecast using 3-hourly data linearly interpolated in space and time.

The initial and one-way nested lateral boundary conditions for the inner fine grid were provided by the ARPS coarse grid forecast. When active weather was anticipated or occurring within the scan volume of the KTLX radar, wideband data from it were added to the fine grid by ADAS.

All predictions were made using version 4.1.6 of the non-hydrostatic ARPS, and identical numerics and physics parameterizations were used for both the inner and outer grids, including terrain based on the RUC model, fourth-order horizontal and second-order vertical advection, a 1.5-order TKE representation for sub-grid scale processes, explicit Kessler warm-rain microphysics, and the Tao et al. (1996) radiation physics coupled with a surface energy budget and 2-layer soil model that considers soil and vegetation type.

4. COMPUTATIONAL LOGISTICS

In an effort to mimic, as closely as possible, a true operational forecast environment, the Pittsburgh Supercomputing Center dedicated to CAPS, for about 6 hours each day, 6 processors on its Cray C90 and 256 processors on its Cray T3D supercomputers. The coarse grid prediction, run on the C90, required approximately 45 minutes of CPU (wallclock) time for a 7 hour forecast with a memory requirement of 29 million words. The fine grid, run on the T3D using PVM, took about 1 hour and 45 min of CPU (wallclock) time for a 7 hour prediction and consumed 41 million words of memory. Most aspects of model execution and product generation for the entire forecast cycle were automated through the use of UNIX shell scripts and cron tables.

5. PRODUCT GENERATION AND FORECAST EVALUATION

CAPS once again used the World Wide Web as a primary means for making its prediction results available to interested users. Working with NWS forecasters, CAPS developed several graphical products that were automatically posted to the Web as they were being generated during the model forecast cycle. Accumulated post-processing time was about 20 minutes for both domains. An electronic evaluation form on the Web allowed users, particularly operational forecasters at the Norman NWSFO, to submit detailed comments on model performance for later analysis.

Considerable emphasis was placed during the SOP '96 on developing techniques for the quantitative evaluation of ARPS' forecasts (see Carr et al., 1996). Specifically, Web pages were set up for comparing point ARPS forecasts with observed soundings, profiler data, and meteograms both in realtime and after the fact. Additionally, difference fields between the ARPS and ADAS, as well as level-by-level statistics, were generated for numerous basic and derived quantities at hourly intervals. To examine any of these products, please visit the CAPS home page at http://wwwcaps.uoknor.edu.

6. ACKNOWLEDGMENTS

This research was supported by the Center for Analysis and Prediction of Storms (CAPS) at the University of Oklahoma. CAPS is funded by Grant ATM91-20009 from the National Science Foundation, and by a supplemental grant through the NSF from the Federal Aviation Administration. Computer resources were provided by the Pittsburgh Supercomputing Center, which is also sponsored by the NSF. The authors gratefully acknowledge Sue Weygandt for drafting the figure.

7. REFERENCES

Benjamin, S.G. et al., 1994: The Rapid Update Cycle at NMC. Preprints, 10th Conf. on Num. Wea. Pred., 17-21 July, Amer. Meteor.. Soc., Portland, OR, 566-568.

Bratseth, A., 1986: Statistical interpolation by means of successive corrections. Tellus, 38A, 439-447.

Brewster, K. et al., 1995: Initializing a nonhydrostatic forecast model using WSR-88D data and OLAPS. Preprints, 27th Conf. on Radar Meteor., 9-13 October, Amer. Meteor. Soc., Vail, CO.

_______ 1996: Application of a Bratseth analysis scheme including Doppler radar data. Preprints, 15th Conf. on Wea. Analysis and Forecasting, 19-23 August, Amer. Meteor. Soc., Norfolk, VA.

Carr, F.H. et al., 1996: Quantitative verification of non-hydrostatic model forecasts of convective phenomena. Preprints, 18th Conf. on Severe Local Storms, 19-23 Feb., Amer. Meteor. Soc., San Francisco, CA, 174-177.

Droegemeier, K.K. et al., 1995: Realtime numerical prediction of storm-scale weather during VORTEX '95, Part I: Operations summary and example predictions. Preprints, 18th Conf. on Severe Local Storms, 19-23 Feb., Amer. Meteor. Soc., San Francisco, CA, 6-10.

Janish, P.R. et al., 1994: Evaluation of the Advanced Regional Prediction System (ARPS) for storm-scale operational forecasting. Preprints, 14th Conf. on Wea. Analysis and Forecasting, 15-20 January, Amer. Meteor. Soc., Dallas, TX, 224-229.

Shapiro, A. et al., 1996: Initial forecast fields from single-Doppler wind retrieval, thermodynamic retrieval, and ADAS. This volume.

Tao, W.-K. et al., 1996: Mechanisms of cloud-radiation interaction in the tropics and mid-latitudes. Submitted to J. Atmos. Sci.

Xue, M., K. K. Droegemeier, V. Wong, A. Shapiro, and K. Brewster, 1995: ARPS Version 4.0 User's Guide. Center for Analysis and Prediction of Storms, Univ. of Oklahoma, 380pp. [Available from CAPS, 100 East Boyd, Room 1110, Norman, OK, 73019.]

_______et al. 1996a: Realtime numerical prediction of storm-scale weather during VORTEX '95, Part II: Operations summary and example predictions. Preprints, 18th Conf. on Severe Local Storms, 19-23 Feb., Amer. Meteor. Soc., San Francisco, CA, 178-182.

_______, 1996b: The 1996 CAPS spring operational forecasting period: Realtime storm-scale NWP, Part II: Operational summary and sample predictions. This volume.