Scott M. Spratt
and David W. SharpThe first Doppler Weather Surveillance Radar (WSR-88D) to be situated in a sub-tropical marine environment was installed at Melbourne, Florida (MLB) in October 1991. Since that time, numerous tropical cyclone (TC) data sets have been examined in real-time for short-term forecast and warning preparation and archived for post-event applied research. This paper will highlight several local studies and projects and examine how their undertaking have resulted in direct operational benefits via improved public forecasts and warnings.
Several studies have documented that WSR-88D's often significantly underestimate rainfall during tropical events (e.g. Choy et al. 1996; Glitto and Choy 1997). Results from these studies provided a justification for local (MLB) modification of adaptable parameters within several WSR-88D precipitation algorithms, allowing for more accurate rainfall accumulation estimates.
Specifically, MLB currently changes the default Z/R relationship (300R^1.4) to a "tropical" (Rosenfeld et al. 1993) Z/R (250R^1.2) and the maximum rainfall rate from a summer default (144.3 mm/h) to a "tropical" rate (262.1 mm/h) prior to the onset of TC rainbands. These algorithm changes (along with others discussed below) were formulated into a "local WSR-88D TC operations plan" , enhancing the original plan "WSR-88D use during TC events" suggested by the Tropical Prediction Center (Sharp 1995). An application of these changes will be discussed in section 5.1.
A local climatological study by Hagemeyer and Hodanish (1995) identified 214 tornadoes which occurred across Florida between 1882 and 1994 in association with 61 tropical and hybrid (tropical and extra-tropical characteristics) systems. The tornadoes in 87 percent of these systems occurred in the right-front quadrant with respect to system motion, meaning the systems approached Florida from the Gulf of Mexico.
More recently, Spratt et al. (1997) examined WSR-88D data from two tropical cyclones which produced multiple tornadoes within Florida, namely, Gordon (1994) and Allison (1995). The radar observations indicated that the TC mesocyclones were often more physically compact compared to "traditional" (US Southern Plains) mesocyclones, having reduced rotational depths (3-3.5 km) and smaller horizontal diameters (1.75-2.0 km).
Likewise, the average magnitude of rotational velocity (12.5 m/s) and shear (.015 /s), at the time of tornado occurrence, were also less. However, the ratio of the average depth of rotation to average depth of the parent cells reflectivity core (12-km) were comparable with that of traditional mesocyclones.
Although traditional severe weather reflectivity signatures were often subtle, in each case the tornado- producing cell was associated with a reflectivity maxima greater than those of surrounding cells. Additionally, high spectrum width values were found to be a reasonable way to track suspect cells as they moved beyond the range of optimal velocity sampling.
Cells containing persistently high reflectivity and weak rotation (especially for core diameters less than 2-km) often later produced tornadoes. Assessing the trend of a mesocyclone's shear proved to be more useful than simply assessing the magnitude of its rotational velocity. A shear value of .010 /s within 2-km of the surface was found to often precede tornado occurrence by 5-15 minutes and was suggested as a minimum threshold for categorizing cells having a high potential for tornadogenesis, especially if contraction of the mesocyclone diameter subsequently occurred.
Due to the small, shallow nature of most TC-mesocyclones, it is not surprising that the current WSR-88D mesocyclone algorithm often resulted in a low probability of detection. While local research indicated that reducing the threshold pattern vector (TPV) adaptable parameter within the mesocyclone algorithm (see MLB WSR-88D ops plan) improved detection somewhat, dramatic improvements were not noted. Therefore, until a more robust mesocyclone algorithm is employed, manual recognition must be stressed. An example of the manual recognition technique resulting from this study will be shown in section 5.2.
During production of Spratt et al. (1997), personnel from three adjacent Florida NWSO's collaborated to establish preliminary critical criteria for use in the TC-tornado detection and warning process. In addition to applying results from this research into local applications at NWSO Melbourne and the adjacent offices, external training has also been conducted through oral and poster presentations at several national conferences and NWS workshops. For example, presentations were made at the Science Operation Officers (SOO) workshop at the Florida State University Cooperative Institute of Tropical Meteorology (FSU-CITM) in April 1996 and at the Advanced Tropical Weather Workshop (January 1997) sponsored by the Operational Support Facility (OSF) at Norman, OK. Additionally, a laboratory scenario was conducted at the OSF workshop which utilized the TC Josephine (1996) data from KMLB. During the scenario, participants analyzed WSR-88D products in PUP "play-back mode" (simulated real-time operations) and attempted to issue severe weather warnings based on techniques outlined in the earlier presentations.
Another ongoing endeavor involves the participation by both authors in a Tropical Cyclone Working Group (TCWG). The goal of the TCWG is to develop operational tropical cyclone WSR-88D algorithms. Initially, a working group of scientists (both research and operational) will be organized. At a later time, proposed algorithms will be put on a common workstation at the TPC to undergo detailed testing. The TCWG will then present a short list of algorithms to the OSF for implementation.
Personnel at MLB expect to continue examining tropical cyclone data sets in hopes of offering additional improvements to precipitation estimates and mesocyclone detection and warning strategies. Future work is expected to concentrate on identifying TC wind maxima and improving eye fixes. Furthermore, the recent implementation of a Warning Decision and Support System (WDSS) at MLB will likely provide an important platform for future TC algorithm testing in real-time.
5.1 Tropical Storm Jerry (1995) - Heavy Rainfall
As the center of TS Jerry reached a position west of MLB during the evening of 23 August 1995, bands of heavy rainfall rotated onshore from the Atlantic. The local WSR-88D operations plan was implemented earlier in the day, giving forecasters added confidence in radar rainfall estimates.
By late evening, storm total precipitation (STP; 12-h accumulation) estimates of 76-127 mm were noted over the extreme southeast portion of MLB's county warning area (CWA). Spot real-time cross-checks between radar accumulations and gage reports indicated accurate estimates were being made. Given reports of the onset of minor roadway flooding and the expected persistence of rain rates of 25 mm/h, a flood warning was issued for the southernmost county in the CWA. During the next 4 h, flood warnings were issued for 3 downwind counties as a band with 25-50 mm/h estimated rain rates lifted slowly northward. Around daybreak, additional flood warnings were issued for 2 inland counties as a new band of heavy rain developed.
All flood warnings verified as widespread urban/street flooding occurred, water entered several homes, and a bridge was washed away. Twenty-four hour rainfall estimates and gage reports confirmed that a maximum of 250-305 mm fell in the area of most severe flooding. By using accurate STP products and one hour precipitation (OHP) and reflectivity time lapses, forecasters were able to issue warnings in advance of all flood events.
5.2 Tropical Storm Josephine (1996) - Tornadoes
As TS Josephine made landfall over northwest Florida on 7 October 1996, outer
rainbands spiraled across the MLB CWA. Embedded within these rainbands were several
isolated high reflectivity cells. These cells were monitored closely by radar for
signatures similar to those observed prior to past TC-tornadoes (Spratt et al. 1997). Over a 9-h period,
numerous tornado warnings were issued as 5 long-lived cells acquired low-level rotation,
decreasing core diameters, and increasing shear values to near or greater than .010 /s.
This strategy (section 3) proved to be extremely useful as all tornadoes (6) were preceded
by warnings, with an average lead-time of 26 min. Five of the tornadoes were rated F0, but
one produced a long damage track of F2 intensity. During the event, only 2 warnings were
not "officially verified", although one of these was associated with a confirmed
waterspout. The performance illustrated by this event highlights that favorable warning
results can be realized by anticipating rapid mesocyclogenesis and providing detailed
radar analyses of suspicious cells while accounting for inherent sampling limitations
(i.e. aspect ratio, overshooting beam, etc.)
Since WSR-88D placement at MLB in 1991, a considerable amount of research has been conducted using locally archived TC data sets. Specifically, precipitation studies lead to improved accumulation estimates and lead-time flood warnings during TC Jerry (1995) and a mesocyclone analysis technique furnished increased lead-time tornado warnings during TC Josephine (1996). During Hurricane Erin (1995), accurate and timely products provided an improved public service to the residents of east central Florida, in large part due to the implementation of applied research results and close coordination with the TPC. In fact, NWSO MLB received recognition via a "Service Enhancement" award for forecast and warning operations during Hurricane Erin.
MLB personnel plan to continue performing research to further optimize tropical rainfall estimates and precursor tornado signatures, as well as to search for additional methods to improve real-time TC forecasts and warnings.
Choy, B.K., L.M. Mazarowski, and P. A. Glitto, 1996: Tropical Storm Gordon: 72-hr rainfall totals over east-central Florida and WSR-88D comparisons. NOAA Technical Memorandum, NWS-SR 174, 14 pp. [NTIS PB 96-163522].
Glitto, P.A, and B.K. Choy, 1997: A comparison of WSR-88D storm total precipitation performance during two tropical systems following changes to the multiplicative bias and upper reflectivity threshold. Submitted to Wea. Forecasting.
Hagemeyer, B.C. and S.J. Hodanish, 1995: Florida tornado outbreaks associated with tropical cyclones. 21st Conf. on Hurricanes and Tropical Meteor., Miami, FL., Amer. Meteor. Soc., 312-314.
Rosenfeld, D., D.B. Wolff, and D. Atlas, 1993: General probability-matched relations between radar reflectivity and rain rate. J. Appl. Meteor., 32, 50-71.
Sharp, D.W., 1995: Report on the 1995 southeastern United States WSR- 88D Users' Workshop, February 22-23, 1995, Patrick Air Force Base, Florida. NOAA Technical Memorandum, NWS-SR 168, 25 pp. [NTIS PB 95- 229829].
Spratt, S.M., D.W. Sharp, P. Welsh, A. Sandrik, F. Alsheimer, and C. Paxton,
1997: A WSR-88D assessment of tropical
cyclone outer rainband tornadoes. Submitted to Wea. Forecasting.