Scott M. Spratt
NOAA/NWSO Melbourne, FL
Andrew J. Nash
NOAA/NWSO Tampa Bay Area - Ruskin, FL
Three storms, Alberto, Beryl and Gordon, affected central Florida during the 1994 tropical season. With the installation of the WSR-88D at Melbourne in 1991 and Tampa during early 1994, complete doppler radar coverage of central Florida recently became a reality. Thus, the past hurricane season provided the first opportunity to collectively utilize the new radars for tropical cyclone interrogation (Fig. 1). As a result, forecasters at both the Melbourne (MLB) and Tampa (TBW) Nexrad Weather Service Offices (NWSO) had realtime access to reflectivity and velocity data, as well as radar-based rainfall estimates. This brief paper will discuss the operational benefits and deficiencies experienced while using the WSR-88D within a tropical cyclone environment. Several recommendations will be offered which may be useful for WSR-88D operators during future tropical cyclone events.
As part of the ongoing Modernization And Restructuring (MAR) of the NWS, several new information products have been developed to help disseminate detailed weather data using the most advanced technology, such as the WSR-88D. One new product, the "nowcast" (NOW), allows forecasters to issue frequent updates to the public on short term weather activity. This product allows for much more literary freedom than in the older Radar Narrative Summaries (RNS) or than can be expressed in State/Zone/Local Forecast Products (SFP/ZFP/LFP's). Florida forecasters have also recently begun issuing twice daily Area Thunderstorm Outlooks (ATO's), which detail the expected threats from thunderstorm activity in the 0-12 hour time frame, for use by the public, emergency managers and weather spotters. Both of these products proved effective during the 1994 tropical events.
The major impact of Alberto upon Florida was excessive rainfall across the panhandle and the subsequent river and lowland flooding. NWSO MLB has hydrologic responsibility (e.g. river flood warnings) for all of the state except for the westernmost portion of the panhandle. The remote dial in capability of the WSR-88D was used to request precipitation products from the Eglin Air Force Base radar (KEVX) in the panhandle. Although temporary communication problems severely limited the availability of data from KEVX, a few Storm Total Precipitation (STP) products were obtained. These products delineated a sharp precipitation gradient along the western portion of the hydrologic area, with the heaviest rainfall properly displayed just west of the area of responsibility. However, comparisons between the radar products and area cooperative observer reports indicated a general underestimation of the radar based accumulations. Korotky et. al. (1995) and Choy et. al. (1995) noted similar WSR-88D estimation errors in association with tropical storms Alberto and Gordon.
After examining the location of heaviest rainfall on STP products, and analyzing data from cooperative observers, model output, and using local historical knowledge, close collaboration between the MLB Service Hydrologist and the SouthEast River Forecast Center resulted in the issuance of credible flood forecasts (L. Mazarowski 1995, personal communication).
As far as Central Florida was concerned, the heaviest of the rainfall activity remained offshore of the west coast as Alberto's center came no closer than 277 km (150 nm) of the coast. However, as several of the far outer rainbands approached the coast, forecasters at TBW, using velocity products in combination with reflectivity data, issued marine weather statements (MWS) and warnings (SMW) directed at mariners and coastal residents.
The continuous presence of convective activity a week prior to Alberto resulted in an accumulation of the STP. TBW forecasters thus dealt with a "contaminated" STP as a manual reset was, at the time, a somewhat involved process. Comparing pre-Alberto to post-Alberto STP's, approximately 76 to 152 mm (3-6 inches) was estimated to have fallen over the eastern Gulf during the period.
TBW forecasters had to also contend with range folding in the velocity data due to the large areal coverage of precipitation. This resulted in large gaps in the velocity fields and made it difficult to fully interpret winds over the Gulf of Mexico where few surface based observations were available.
During the middle of August, Tropical Storm Beryl followed a similar path to that of Alberto, making landfall along the Florida panhandle (Fig. 1). Although Beryl did not produce rain totals as large as Alberto, 152 to 229 mm (6-9 inches) did occur across northwest Florida, further aggravating high water levels. Numerous flood warnings and statements were issued by MLB for rivers across the panhandle. Communication problems between KMLB and KEVX again restricted product retrieval.
MLB ATO's addressed the possibility of a few strong to severe thunderstorms, indirectly associated with Beryl. These predictions were based on an intrusion of a dry layer aloft into the tropical airmass, and the persistence of strong low level wind fields as evidenced in radar VAD Wind Profile (VWP) data. The forecasts verified as two microbursts produced severe wind gusts at the Kennedy Space Center and three large waterspouts were reported just offshore of Daytona Beach Shores.
Based on reflectivity time lapses, hourly NOW's issued by MLB throughout the event accurately stressed that the outermost rainbands would dissipate prior to reaching the interior of east central Florida. Meanwhile, the TBW NOW's highlighted the offshore rain bands and their effects upon mariners and coastal residents.
From experience gained during Alberto, a forced reset of the STP product was performed as Beryl began to move into range of the KTBW 88D. The STP indicated that heavy rainfall in excess of 254 mm (10 inches) fell over a large area of the offshore Gulf waters during this period.
Although the majority of the weather remained well offshore of west central Florida, an outer rainband did approach the coast mid-day on the 15th. KTBW reflectivity data around 1850 UTC indicated the strongest activity within this band was over 139 km (75 nm) offshore with what appeared to be weaker storms within 37 km (20 nm) of the coast. In contrast, velocity data indicated the "weaker" storms to be potentially severe. Forecasters recognized signals in the Storm Relative Motion (SRM) velocity product of an operator-defined mesocyclone with one of these storms and issued a SMW. Post analyses indicated the mesocyclone algorithm did not identify the circulation until 1916 UTC, and then again 15 minutes later. As the cell approached the coast, a severe thunderstorm warning was issued, however the circulation rapidly weakened at the coast and no damage was reported.
A useful product during typical "summertime" convective activity is the Vertically Integrated Liquid (VIL). As a general rule across central Florida, VIL's above 55 kg/m2 often correspond to hail/strong wind producing thunderstorms, yet due to the low topped, tropical nature of the rainband convection, most VIL's were in the 10 to 15 kg/m2 range. These values ordinarily would not concern forecasters as being indicative of potentially severe weather, although in this case the values were misleading.
Gordon was a unique late season storm and the third tropical system to affect the area. While Gordon was still southeast of Miami, MLB and TBW forecasters ran User Functions which constantly requested numerous products from the Miami radar. This allowed the storm to be tracked while still far beyond local radar range.
Before rainbands approached central Florida, the STP product was again reset at both KMLB and KTBW. Total estimated accumulations by KMLB averaged 76 to 178 mm (3-7 inches) across east central Florida with a large area of greater than 330 mm (>13 inches) just offshore. KTBW totals for this same area were near 127 mm (5 inches). Similar to the previous cases, the overall rainfall pattern compared well with gage measurements, however actual values were substantially underestimated (see Choy et. al. 1995) as 203 to 279 mm (8-11 inch) totals were common across the coastal counties.
Late on 15 November and early on the 16th, a rainband detached from the circulation of Gordon and rotated inland across east central Florida from the Atlantic. Initially, the band consisted of areas of moderate to heavy rain. However, during the evening the KMLB VWP indicated a transition of the low level wind field from a uni-directional pattern to one of pronounced veering, resulting in the formation of five persistent mesocyclones within a 12 hour period. The first four circulations were initially detected manually by the MLB radar operator 37 km (20 nm) or more offshore, with all but one weakening upon approaching the coast. The non-weakening cell actually intensified near the coast and eventually produced a deadly tornado. Velocity range folding severely limited the ability of the KTBW radar to identify these shallow systems as they approached the east coast.
The last and longest lived of the five mesocyclones developed just inland from the east coast and moved northwest to between the two radar sites. Throughout the two hour lifecycle of the system, the lowest elevation angle of the KTBW radar intersected the system at about 3350 m (11 kft), while KMLB interrogation varied between 1520 and 4570 m (5 and 15 kft) as the cell moved further inland. Manual recognition of a circulation was again observed within SRM products by radar operators well before identification by the mesocyclone algorithm. In fact, both the KMLB and KTBW radar algorithms did not initially detect presence of a mesocyclone until a tornado had already formed.
As was the case during Beryl, reflectivity and VIL data were very misleading when attempting to pinpoint local severe storms. Conventional severe weather characteristics were not evident within reflectivity signatures and corresponding VIL's rarely exceeded 15 kg/m2. Although performance of the 88D mesocyclone algorithm was considered poor during the events, manual recognition of cyclonic circulations within SRM products allowed KMLB radar operators to differentiate between potentially severe and non-severe convection. Significant lead-time warnings were therefore possible.
When in a tropical cyclone situation, NWS forecasters should consider the following suggestions: Use dial-in User Functions to adjacent radars to monitor the storm before it moves into the local area.
Choy, Mazarowski, and Glitto, 1995: Tropical Storm Gordon: 72-hr Rainfall Totals over East Central Florida and WSR-88D Comparisons. Submitted as NOAA Tech. Memo. NWS SR.
Korotky, J., J.R. Michael, and C.E. Williford, 1995: Tropical Storm Alberto: A WSR-88D Radar Overview, 14th Conference on Weather Analysis and Forecasting, Dallas, TX, 559-564.
NHOP, 1995: National Hurricane Operations Plan, FCM-p12-1995, US Dept. of Commerce, NOAA, Washington DC.
