1027 casa dallas fort worth urban testbed observations

Pdf File 2,959.10 KByte, 11 Pages


Presented at the Special Symposium on Meteorological Observations and Instrumentation, 97th AMS Annual Meeting, Seattle, WA, January 22-26, 2017, Amer. Meteor. Soc., Paper 1027


Keith A. Brewster1, Apoorva Bajaj3, Brenda J. Philips3, David L. Pepyne3, Eric Lyons3 and Frederick H. Carr1,2

1Center for Analysis and Prediction of Storms 2School of Meteorology University of Oklahoma Norman, OK 73072

3University of Massachusetts Amherst, MA 01003


Figure 1. Schematic of CASA concept of gap filling radars covering lower levels of the atmosphere.

The Collaborative-Adaptive Sensing of the Atmosphere (CASA) project was originally established as a National Science Foundation Engineering Research Center (ERC) with the idea of deploying low-cost X-band (3 cm wavelength) dualpolarization Doppler weather radars that could scan the low-levels of the atmosphere in an adaptive fashion to precisely identify and track weather hazards such as tornadoes, strong winds and localized heavy rainfall (McLaughlin et al. 2009). The CASA ERC was based at the University of Massachusetts (UMass) and originally included three other academic institutions (University of Oklahoma, Colorado State University and the Universi-

ty of Puerto Rico-Mayaguez) with two others added later (University of Delaware and University of Virginia).

CASA's Integrated Project-1 (IP1) was established in 2006 in southwestern Oklahoma with four X-band dual-polarization Doppler weather radars. The radars were sited to maximize lowlevel coverage under the WSR-88D surveillance umbrella while providing overlapping coverage to provide good dual-Doppler winds (Brewster et al., 2005). CASA IP1 had a number of notable successes in Oklahoma including identification and tracking of several tornadoes, most notably the Chickasha-Newcastle tornado of 24 May 2011.


Toward the end of the ten-year CASA ERC it was decided to relocate the radars to the DallasFort Worth Metroplex, an urban area with more than 6 million residents, several key transportation hubs, and a number of large sports and entertainment venues. The area experiences numerous severe thunderstorms, including some hail, flash flooding and tornadoes, in a typical year, as well as an occasional visit from a tropical storm.

The North Central Texas Council of Governments (NCTCOG) and CASA have developed a multisector partnership to install and operate the radars, provide radar sites, communication

*Corresponding author address: Keith Brewster, 120 David L. Boren Blvd., Suite 2500, Norman, OK 73072, kbrewster@ou.edu

Figure 2. Schematic illustrating partnerships supporting the CASA DFW Testbed, including several private companies, academic institutions, federal, state and local governments. Trademarks property of respective companies.

Figure 3. CASA X-band radar network. 40-km range rings for planned (green) and installed (blue) radars. Map of counties of NCTCOG, including urban areas in beige shading. Map courtesy of NCTCOG.

and other support for the radar network (Bajaj and Philips, 2012). At the same time the U.S. National Weather Service (NWS) began exploring a National Mesonet that would include data from several government and private weather data sources based on the Network of Networks concept, spawned by a National Research Council report (NRC, 2009). These ensuing public-private partnerships (Fig. 2) comprise the CASA Dallas-Ft Worth Urban Testbed (CASA DFW Testbed, hereafter).

Siting for the radar network followed the principles discussed in Brewster et al. (2005b) while being constrained by more complex logistical issues than in Oklahoma. Site locations, power and communication had to be identified and arranged with various hosting institutions that included the University of North Texas, University of TexasArlington, Johnson County and the cities of Fort Worth, Addison, Midlothian, and Mesquite. Radars deployed included the original CASA radars from UMass, plus radars from EWR Weather Radar, Ridgeline Instruments, Furuno, and Enterprise Electronics Corporation. Figure 3 shows the layout of the X-band radar network overlaid on the county boundaries of the NCTCOG. Seven of eight planned radars have been deployed to date, and the urbanized area of the Metroplex is well covered by the network.

Figure 4. Dual-Doppler crossing angles (color fill, scale in degrees at right) for federal S-band (NEXRAD: KFWS) and C-band (TDWR: TDAL and TDFW) based in the Dallas-Ft Worth Metroplex.

Figure 4 shows the dual-Doppler intersection angle for the best pair of radars at each point considering the three federal radars in the testbed, namely the KFWS WSR-88D NEXRAD radar and the two FAA Terminal Doppler Weather Radars (TDWR) covering Dallas Love Field and DFW International Airport, respectively. The quality dualDoppler coverage (orange and red colors in Fig. 4) is limited because the radars happen to be aligned in nearly a straight line.


Figure 5. As in Fig. 3 but adding 8 X-band radars in the

planned CASA radar network.

Figure 6. Sample observation distribution in the CASA

DFW Testbed. Blue squares: ASOS & AWOS stations,

Green Dots: CWOP sites, Magenta Dots: WxBug sites,

Red Dots: MoPED truck observations, Black Dots:

Other surface sites from MADIS, Orange Crosses:




Table 1. Observations and their sources for the CASA DFW Testbed

Observation NEXRAD TDWR CASA ASOS/AWOS CWOP Oklahoma West Texas WxBug Other Mesonets High Freq. Pressure MoPED Understory SODAR Radiometer MDCRS/ACARS TAMDAR

Type S-Band Radar C-Band Radar X-Band Radar Surface Surface Surface Surface Surface Surface Surface Surface Mobile Surface & Hail Wind Profiles Temp & RH Aircraft Aircraft

Provider NWS, FAA & DoD FAA CASA, EWR, EEC, Ridgeline & Furuno NWS & FAA NWS Oklahoma Climatological Survey Texas Tech University EarthNetworks Misc Federal & State Agencies ParoScientific GST Understory WeatherFlow Radiometrics Airlines via MDCRS/ACARS Airlines via Panasonic

Source IRADS via NWS FAA via NWS & NSSL AWIPS-II via CASA MADIS (Natl Mesonet) MADIS (Natl Mesonet) MADIS (Natl Mesonet) MADIS (Natl Mesonet) MADIS (Natl Mesonet) MADIS (Natl Mesonet) CASA GST Understory MADIS MADIS MADIS MADIS


Figure 5 shows the dual-Doppler crossing angles for the same area including the CASA radar network. The urban area within in the testbed is covered quite well with quality dualDoppler crossing angles. In addition, the lowest radar beam height above ground is about 500 m compared to the NEXRAD coverage that has an average height of 1000 m in this domain.


In addition to the X-band radar network there are other observation networks that cover the CASA DFW Testbed. Table 1 summarizes the observation networks within the testbed. Many are surface observation networks included in the NWS National Mesonet Program distributed via NOAA Meteorological Assimilation Data Ingest System (MADIS), but also includes truck data from the GST MoPED, surface and hail data from Understory Weather, wind profilers from WeatherFlow, surface-based radiometers from Radiometrics, aircraft data from MDCRS, as well as high-frequency pressure and GPS-Met data from CASA installed weather stations. Figure 6 shows a sample distribution of data (as indicated in the caption) for a selected 5-minute window.


CASA observations are used by the public safety community in North Texas as severe weather impacts North Texas. CASA has tools to display and animate the X-band radar data that are available to partners and subscribers as a way to support continuing operations. In order to better integrate the CASA data into the realtime warning and forecasting operations of the NWS, it was desirable to integrate the data into the AWIPS-II workstations used at the NWS Forecast Office in Fort Worth. To accomplish this, the UMass team worked with the NWS Southern Region to reformat the radar data and transmit them to appropriate NWS servers.

Figure 7. AWIPS-II display of Fort Worth X-band radar reflectivity (dBZ, color scale at top) of the Haslet tornado, 23 March 2016.

Figure 8. AWIPS-II display of Fort Worth X-band Doppler radial velocity (ms-1, color scale at top) of the Haslet tornado, 23 March 2016.

Figure 9. NWS forecaster viewing the CASA radar data on AWIPS-II display in the Fort Worth NWS Forecast Office.

Figure 10. Reflectivity (dBZ, color scale at top) from the Midlothian X-band radar for the tornado of 15 January 2017.

Figure 11. Doppler radial velocity (ms-1, scale at top) from the Midlothian X-band radar on 15 January 2017.

As an example of the AWIPS-II display, the reflectivity and Doppler radial velocity from the X-band radar at Fort Worth for the 23-March2016 Haslet tornado are shown in Figs 7 and 8, respectively. Figure 9 shows how the data are used on the AWIPS-II workstation at the Fort Worth NWS Forecast Office.

A notable success of this effort was when the CASA X-band radar data were used to pinpoint the location and movement of a compact F0 tornado that tracked near the border of Tarrant and Dallas Counties on the evening of 15 January 2017 (Figures 10 and 11). Knowing the precise location and movement of this feature was of increased importance on this evening because 100,000 NFL football fans were dispersing from AT&T Stadium in Arlington, Texas,

Figure 12. Tornado warning text from Fort Worth showing the use of CASA radar to pinpoint tornado location and movement on 15 January 2017.

after the NFC Championship game that had just concluded. The warning product issued (Fig. 12) included those details from the CASA Xband radar.


In addition to the raw data displays, a real time high resolution analysis system (400-m grid spacing) and an efficient assimilation and Numerical Weather Prediction (NWP) system with 1-km grid spacing producing 0-to-2 hour forecasts with low latency are being run over the CASA DFW Testbed by the Center for Analysis and Prediction of Storms (CAPS) at the University of Oklahoma (OU). Initially designed for the CASA IP1 domain (Brewster et al., 2007, Brewster et al., 2008, Brewster et al., 2010), the domain has since been moved to Texas.

The NWP system achieves very low latency on modest computing resources by utilitizing a 3DVAR analysis with complex cloud analysis (Gao et al., 2004, Brewster et al., 2005a, Hu et al., 2006, Brewster and Stratman, 2015). The


analysis increments are assimilated into the ARPS (Xue et al, 2001, Xue et al., 2003) model using Incremental Analysis Updating (IAU, Bloom et al., 1996) with the recently-developed variable-dependent timing (IAU-VDT, Brewster et al., 2016). This allows production of the complete 2-hour thunderstorm-resolving forecast in about 20 minutes of wall-clock time on fewer than 200 cores of a Linux server cluster.

Data assimilation and short-term forecasting are run on a 350 x 320 km domain with 1-km grid spacing. 53 vertical grid levels are used with domain top at 20 km and enhanced vertical resolution near the ground (20 m minimum vertical grid spacing). Further details of the forecast and assimilation system can be found in Brewster et al., 2016.


Example products from the assimilation and forecast system are presented from two recent tornado events in the Dallas-Ft. Worth region.

In the late afternoon and early evening of 26 December 2015 a total of 13 tornadoes were observed and damage tallied at more than $40 M (NWS, 2016, Marshall et al., 2016). Among the tornadoes in the Dallas area were an EF3 tornado near Ovilla around 0000 UTC 27 Dec., and 45 minutes later a large long-track tornado (approximately 20 km long, 500 m wide) with EF4 damage rating that touched down just south of Interstate-30 in Sunnyvale and passed through the portions of Garland and Rowlett in the northeast part of the metro area (Fig. 13). Following that, there were additional tornadoes to the north-northeast of Rowlett as the parent storm continued tracking in that direction.

The forecast system on this day was being run at 30 minute intervals. At the time five of the X-band radars in the DFW Testbed were installed; the radars at Ft. Worth and Mesquite were added later.

Figure 13 Tornado tracks near Dallas on 26-Dec2015 (UTC Times 27-Dec-2015). Dallas and Rockwall Co. are highlighted to aid reader orientation with model output figures. From NOAA Damage Survey Viewer.

The forecasts were very successful in maintaining the storms and producing very strong rotation as indicated in the 1-6 km updraft helicity (UH) plots. Figure 14 shows four successive real-time forecasts (initialized 2300 to 0030) at the valid time for each where the forecast was indicating a strong UH field near the starting point of the Garland-Rowlett tornado. The tornado damage survey points are shown as the triangles in the plot, the contours are reflectivity in 10 dBZ intervals with the UH indicated in color contours, non-linear scale at right. The timing of the rotation center was a bit fast in each, with successive forecasts asymptotically approaching the 0045 actual estimated time of touchdown. Given that the latency for the 2-hour output is about 20-25 minutes the 2330 UTC forecast had nearly 1 hour actual lead time on the observed 0045 UTC touch down time.

As mentioned previously there was another tornadic storm 15 January 2017. On this date there was an issue with the real-time surface data, but after-the-fact the real-time system was rerun with all data sources and it was able to successfully predict the location and track of this storm to within 10 km. Figure 15 has sample forecast plots from the 15 January case.


Download Pdf File