|AVN||Aviation Model, global spectral model that forecasts out to 72 hours|
|CLIPER||CLImate and PERsistence no-skill model|
|Eta||Eta Model, Western Hemisphere (northern portion) out to 48 hours with more resolution in the vertical than the RAFS model|
|NHC90||Statistical dynamic model that uses output from the AVN as predictors|
|RAFS||Regional Analysis and Forecast System, Western Hemisphere (northern portion) out to 48 hours|
Most track model guidance (and the NHC official forecasts based, in part, on that guidance) had a large left bias prior to about 1200 UTC or 1800 UTC on July 2 (Figure 2-1). The rightward swing of the official forecast tracks eventually required an eastward shift of the watch/warning area along the north coast of the Gulf of Mexico.
The performance of the NHC90 model was the best overall (including the depression stage of the storm), with errors comparable to the long-term average NHC official forecast errors. Another hurricane model, CLIPER, which has no dynamical input, also forecast the track fairly well. The Aviation (AVN) model track scheme produced by far the worst storm track forecasts. The NHC and model guidance intensity forecasts were quite good. Most NHC wind speed forecasts were no larger than 10 knots for all forecast periods.
Since the synoptic pattern had little or no forcing and weak steering currents, the NMC models generally did not perform well with regard to the track of the storm, especially after the storm made landfall. In addition to difficulty in forecasting storm track, current numerical models do not typically perform well forecasting precipitation amounts associated with tropical systems. Beginning Tuesday, July 5, the models began to diverge with respect to the forecast track of the remnants of Alberto. The Eta model consistently produced the preferred (and best, as it turned out) forecast of storm movement and QPFs throughout the event.
The model forecasts were good for the first two days of the event, Sunday, July 3, and Monday, July 4. However, after this time, only the Eta model performed well. Although the Eta model at times had a tendency to take the remains of Alberto a little too far north, it offered the most reliable numerical model guidance. From Tuesday, July 5, through Thursday, July 7, the RAFS model was incorrectly trying to shear the remnants of Alberto out to the northeast, while the AVN model would continually dissipate the system beyond its 24-hour forecast. Thus, much of the manual guidance issued by the WFB was based on the Eta model.
FINDING 2-1: As is generally the case with a synoptic pattern with little or no forcing and weak steering currents, the NMC and NHC models in general did not perform well with regard to the track of the remnants of Tropical Storm Alberto.
RECOMMENDATION 2-1: The NWS should continue to strive for improvements in tracking tropical systems once they make landfall. It is especially important that improvements be made in the forecasts at the surface and not just in the mid and upper levels of the atmosphere. Interactions with the research community within NOAA (such as the Office of Atmospheric Research) and other Federal agencies, as well as the academic research community, are especially encouraged.
Forecast discussions issued by the forecasters in the WFB between July 4 and July 7 highlighted the strong possibility for flooding and extremely heavy rain. A discussion issued very early Monday morning, July 4, highlighted a "Very dangerous Flash Flood and Flood situation for much of Georgia today into tonight as the remnants of Alberto drift slowly north." This same discussion mentioned isolated rains of greater than 5 inches between Monday morning and Tuesday morning. On Tuesday afternoon, July 5, an excessive rainfall discussion was issued which indicated that 5-8 inches of rain was a good possibility Tuesday night over a large portion of northern and central Georgia into eastern Alabama. On Wednesday, July 6, an excessive rainfall discussion began with "A dangerous and almost unbelievable situation remains over Georgia and is expected to spread into eastern Alabama, with additional rainfall amounts of 5 to 8 inches possible across southwestern Georgia into southeastern Alabama."
Manual QPFs issued by the WFB were better than the QPFs generated by the models. QPF graphic forecasts for July 4, 5, 6, and 7 are included as Figures 2-2, 2-3, 2-4, and 2-5, respectively. For the most part, all of the manual 24-hour QPFs were typically not quite heavy enough and were generally displaced a little too far north and east from the heaviest rain occurrence. Clearly, the manual 24-hour QPFs issued about 1000 UTC each day did not completely capture the magnitude of the event. However, quotes from many of the excessive rainfall discussions and storm summaries that were issued definitely showed that WFB forecasters were very aware of the dangers of this decaying storm.
Between 1200 UTC July 3 and 1200 UTC July 4, as Alberto was making landfall, the heaviest rains occurred over portions of the western Florida Panhandle and southeastern Alabama, just to the east of the storm track. The 24-hour manual QPF issued early Monday morning, July 4, and valid at 1200 UTC Tuesday, July 5, had greater than 5-inch rains occurring over southwestern Georgia and extreme southeastern Alabama. Model QPFs for the same time period were not nearly as good. The Eta model indicated just over 2 inches of rain for parts of the Florida Panhandle and southern Alabama and Georgia. The RAFS indicated slightly more than 2-inch rains in central Georgia, and the AVN model forecasted only 0.75 inch of rain for the Florida Panhandle.
Alberto drifted slowly north-northeastward from near Montgomery, Alabama, on Monday morning, July 4, to just southwest of Atlanta, Georgia, by Tuesday morning, July 5. Extremely heavy rains of locally greater than 10 inches occurred near the path of the storm, across parts of west-central Georgia. QPFs issued by the WFB indicated up to 5 inches of rain over central Georgia. As was the case the previous day, model QPFs were grossly underforecast compared to the manual forecasts; and they were generally in the wrong location. The Eta model forecast a little over 2 inches of rain for southeastern Alabama; the RAFS had 2 inches for southeastern Georgia; and the AVN forecast had only a paltry 0.9 inch for east-central Alabama. Despite poor QPFs, the models did a respectable job of forecasting where the storm would be for this 24-hour period, at least as interpreted from their 500-mb prognostication chart (prog). This added some confidence to the manually prepared QPF issued by the WFB forecaster.
The period between Tuesday, July 5, and Wednesday, July 6, was a great challenge for the WFB forecasters. The models had been doing a fair job in forecasting the position of the storm in their 500-mb progs, but they started to diverge on the track of the remnants of Alberto. Only the Eta model performed well with its 500-mb forecast during this period. It essentially kept the well-defined, mid-level remains of Alberto nearly stationary in the vicinity of the Alabama/Georgia border, which turned out to be a very good prediction. The RAFS, on the other hand, tried to shear the remains out to the northeast into the central and southern Appalachians, while the AVN model consistently dissipated the system with time. WFB meteorologists correctly accepted the Eta forecast as early as Monday afternoon. The discussion issued at 2:30 p.m. EDT noted that "the RAFS forecast seems too quick considering the lack of upper level winds, and the Eta appears to have the best handle on the situation with a well- organized vorticity field/upper low meandering ever so slowly northeastward."
By following the Eta solution, the 24-hour QPF issued by the WFB on Tuesday morning and ending at 1200 UTC Wednesday, July 6, hit central Georgia the hardest, with up to 5 inches of rain forecast. Subsequent shorter-range forecasts issued later Tuesday afternoon and verified Wednesday morning increased the rainfall potential. The excessive rainfall discussion issued at 2:50 p.m. EDT Tuesday, July 5, mentioned a "good possibility of 5-8 inch rains" by Wednesday morning for a large part of central and northern Georgia and eastern Alabama. The observed 24-hour precipitation ending at 1200 UTC Wednesday, July 6, showed the manual forecast was underforecast and a little too far north. Several stations in central and southwest Georgia received in excess of 10 inches of rain during this period.
QPFs produced by the models were generally less accurate than the manual forecasts for the verifying period from Tuesday through Wednesday morning, especially the RAFS and AVN. The Eta model was the best of the models, forecasting a little more than 3 inches of rain for southern Alabama.
By Tuesday night, July 5, it became increasingly apparent from upper air data that Alberto likely would be blocked from moving any farther north. The upper-level analyses from 0000 UTC Wednesday, July 6, indicated that a ridge was building north of the system from the Tennessee Valley eastward through the central Appalachians. This building ridge would essentially put a halt to any further northward progression. In addition, on Tuesday evening, satellite pictures showed Alberto had stopped moving north and could even be drifting very slowly southwestward.
Unfortunately, the numerical models couldn't agree again on where the system was going. The Eta model kept the system stationary in northern Georgia. The RAFS again sheared the system out too quickly into the Appalachians, and the AVN "lost" the feature after about 24 hours. Knowing that the Eta model had been superior in handling this storm throughout the event, and seeing what was developing synoptically, WFB forecasters again adopted the Eta solution. The excessive rainfall discussion issued very early Wednesday morning, July 6, correctly noted that "The actual center has been forced a little southward during the past 24 hours by ridging to the north and by far the Eta model has been superior in handling this system. Believe the system could continue moving slowly southward for a while longer before becoming stationary later today. Additional rains of 5 to 8 inches are possible over southwestern Georgia and southeastern Alabama."
The 24-hour manual QPF issued around 1000 UTC Wednesday morning, July 6, valid 1200 UTC on Thursday, July 7, had more than 5 inches of rain occurring over a large part of southeastern Alabama and a smaller portion of southwestern Georgia and more than 3 inches of rain for much of southwestern Georgia, the southern half of Alabama, and most of the Florida Panhandle. These amounts turned out to be underforecasted and displaced a little too far north and east. The observed rains, ending at 1200 UTC Thursday, July 7, showed maximum amounts greater than 10 inches over the Florida Panhandle, with much of the Panhandle and southeastern Alabama receiving more than 5 inches.
Model QPF guidance showed essentially the same trends as previous days, although the AVN got closer for the 24-hour period ending on July 7. The Eta again gave the best QPF guidance with a forecast of greater than 3-inch rains for east-central Alabama. The RAFS was much too far north and east with its maximum rain, having forecast a 2.5-inch maximum near the Georgia/South Carolina border. Somehow, even though the AVN continually tried to dissipate the feature, it still forecast a 3-inch rainfall maximum over central Alabama. This in itself is significant, though, since AVN rarely forecasts rains in excess of 2-3 inches in a 24-hour period.
By late Thursday, July 7, any circulation associated with Alberto was becoming increasingly hard to find and, by very early Friday, July 8, the system had dissipated completely. Between 1200 UTC Thursday, July 7, and 1200 UTC Friday, July 8, only isolated greater than 3-inch rains occurred, mostly in central and southwest Alabama.
FINDING 2-2: The QPF guidance generated by the NMC models was poor (as is common for convective situations during the warm season) and therefore of limited help to the forecasters. The national QPF guidance frequently underestimated excessive rainfall amounts and sometimes did not accurately highlight the area of maximum rainfall.
RECOMMENDATION 2-2: The NWS should continue to strive for improvements in QPFs for tropical and convective systems.
This section focuses on the acquisition and use of precipitation data. A description of the heavy precipitation that accompanied Alberto and the storm total isohyetal analysis (Figure 1-4) is contained in section 1.3.2. Section 2.1.3 discusses the NMC-issued QPF (NWS field offices did not issue QPFs).
Figure 2-6 shows that Weather Surveillance Radar-1988 Doppler (WSR-88D) umbrellas provide almost complete coverage of the SERFC area of responsibility, which includes all the areas affected by Alberto. Some of the WSR-88Ds in the SERFC area were recent acquisitions, with acceptance of the radars at NEXRAD Weather Service Forecast Office (NWSFO) Atlanta just days before Alberto and at Warner Robbins Air Force Base (AFB) in January 1994. The NWS staff at NWSFO Atlanta and NWSFO Birmingham and at SERFC were adequately trained in the use of the WSR-88D prior to installation of the equipment.
The three stages of NEXRAD precipitation processing are described below to provide some background information. Stage I Precipitation Processing is performed within the WSR-88D itself and is designed to incorporate up to 50 ground-based precipitation gages to adjust the WSR-88D precipitation estimates (see Figure 2-7 for rain gage locations under the various WSR-88D umbrellas). Subsequent steps in WSR-88D precipitation processing require interactive computer systems at NWS offices; these systems were not available at the offices affected by Alberto. In Stage II, the hourly digital precipitation data from Stage I are combined with Geostationary Operational Environmental Satellite (GOES) infrared satellite data and rain gage data to perform additional quality analysis and screen out anomalous data. Stage III Precipitation Processing involves interactive analysis and assimilation of gridded precipitation estimates from all WSR-88Ds covering a River Forecast Center (RFC) area of responsibility. The resulting hourly precipitation product is used as gridded input to the RFC's hydrologic modeling system. In the summer of 1994, the WSR-88D Stages II and III Precipitation Processing were being run on pre-AWIPS workstations at the Arkansas-Red Basin RFC and the North Central RFC for their areas of responsibility. The gridded precipitation products are then input to the NWS River Forecasting System (NWSRFS) Interactive Forecast Program (IFP) , which is also run on the workstations.
The consensus of NWS field personnel, cooperating agencies, and the media was that the WSR-88Ds did a fine job of representing the areal coverage of the precipitation. WSR-88D images appeared on television and in the newspapers. A WSR-88D storm total precipitation color graphic is shown on the cover of this report. While the representation of the areal coverage was good, the WSR-88D underestimated the amount of precipitation associated with the Alberto warm tropical event.
FINDING 2-3: The WSR-88D Stage I Precipitation Processing, which runs in the Radar Products Generator, does not currently use rain gage data to provide potentially better quantitative estimates of the precipitation.
RECOMMENDATION 2-3: Rain gage data must be included in the WSR-88D Stage I Precipitation Processing as soon as possible, so that the radar-rainfall can be adjusted to avoid underestimation of rainfall associated with warm tropical events.
FINDING 2-4: The number of automated rain gages under the umbrellas of many of the WSR-88Ds in the area affected by Alberto was inadequate to effectively incorporate rain gage data into the Stage I Precipitation Processing.
RECOMMENDATION 2-4: The rain gage data network must be expanded and the reporting characteristics of existing sites modified to provide more timely data to produce a higher quality WSR-88D precipitation estimate.
Because the only usable information available from the WSR-88D was visual, it was only used to "flavor" SERFC forecasts. The SERFC staff looked for downstream/upstream rainfall concentrations in the WSR-88D images to manually adjust timing and distribution of the precipitation. The hourly digital rainfall (HDRAIN) products from Stage I Precipitation Processing could not be used due to lack of workstations to process data and incorporate them into hydrologic models at the SERFC.
FINDING 2-5: Even though the SERFC area of responsibility has almost complete WSR-88D coverage, the SERFC was not able to quantitatively use the WSR-88D information in its forecasts. The capability to process WSR-88D digital precipitation estimates would have added value to the hydrologic forecasts.
RECOMMENDATION 2-5: Pre-AWIPS workstations must be deployed immediately to the SERFC and other RFCs so the Stages II and III Precipitation Processing can be performed and utilized in the forecasts.
The WSR-88D has known hardware and software problems that make it unreliable for archiving. Although the Atlanta NWSFO did not detect a problem with their WSR-88D Levels 3 and 4 archiving (base and derived products) during Alberto, they encountered an unrecoverable error in trying to retrieve the archived data after the event. There has only been limited success in getting archived products from the Warner Robbins (near Macon, Georgia) WSR-88D, which was the closest WSR-88D to the area of heavy precipitation.
FINDING 2-6: The Atlanta WSR-88D was not able to retrieve data from the archive for a precipitation event that set historical records.
RECOMMENDATION 2-6: The potential for losing data, for all time, that could be used for storm analysis, training, and calibration of hydrometeorologic models and calibration of the WSR-88D dictates a requirement that there be a prompt resolution of the problems with the archive media.
There were also problems acquiring data from associated Principle User Processors (PUP). Associated PUPs provide access to WSR-88Ds that are not located at the NWS office and are usually not owned by the NWS. The NWSFO at Birmingham, Alabama, noted a problem with the number of products that can be sent from associated PUPs (e.g., Maxwell AFB) when there is widespread precipitation. They had to reduce the number of Maxwell AFB products from 36 to 20 to get the subset of products in a timely manner.
FINDING 2-7: The WSR-88D was unable to provide all the products in the time required when there was a large-scale precipitation event.
RECOMMENDATION 2-7: Develop methods to increase the number of products that can be obtained by associated PUPs, especially for offices with warning responsibilities.
The SERFC reported a good flow of observed data from the NWS offices in its area. The storm precipitation totals are included in Figure 1-4. The Atlanta NWSFO experienced difficulty in getting data from cooperative observers because of the evacuation of some of the observers. The Atlanta NWSFO used the sheriffs' departments to get more reports. The cooperative observer from Plains, Georgia, was unable to call in his observation due to repeatedly busy phones at NWSFO Atlanta. All spotters and cooperative observers share one phone line which is not dedicated to those functions. The Meteorologist in Charge (MIC) at Atlanta said additional phone lines would have been helpful. NWSFO Birmingham had some problems getting people to the gages to take observations and also with backup observers adding the height above mean sea level to their gage readings. The NWSFO Birmingham frequently called their cooperative observers to obtain data. Both the Atlanta and Birmingham NWSFOs called the Limited Automatic Remote Collector (LARC) gages themselves and routinely got 6-hour data (or more frequently, as needed).
FINDING 2-8: A limitation in the number of phone lines caused problems for at least one office and a cooperative observer from a critical area who was not able to provide data to the NWS because of busy phone lines.
RECOMMENDATION 2-8: Ensure that data are not lost due to inadequate phone lines into NWS offices and have adequate automated collection systems to acquire data so that the capacity of voice lines is not a constraint.
The SERFC gets centrally decoded data and makes its model runs on the NOAA Central Computer Facility (NCCF) via a Remote Job Entry (RJE) system. There were no problems with the central data systems, and model run performance on the NCCF was good. When the dedicated RJE communications were down for 3-4 hours, the dial backup feature did not work because the dial backup phone line was not reattached to the RJE system after the office moved in April 19, 1994. The SERFC staff later successfully tested the dial backup by sharing the fax line.
FINDING 2-9: The RJE dial backup did not function because the dedicated phone line had not been connected to the system. The RJE dial backup had not been tested since the office moved.
RECOMMENDATION 2-9: RFC staffs must routinely test the RJE dial backup.
The RFCs running on the NCCF have the ability to declare a Critical Flood Situation which allows them to use the crisis job priority, increases the priority of the central data acquisition systems, and may limit the preventative maintenance that is done on the NCCF.
FINDING 2-10: The SERFC did not declare a Critical Flood Situation during the Alberto event, because job processing times were adequate.
RECOMMENDATION 2-10: The declaration of a Critical Flood Situation and use of the crisis job priority are powerful tools that should be utilized by the RFCs during any critical flood event.
The U.S. Army Corps of Engineers reported receiving good information via the RFC HYDROMET system but wanted access to the WSR-88D data. They had ports allocated on most WSR-88D systems for their use but at the time did not have their PUP emulator systems in place to interface with these ports.
The flooding in this event occurred on rivers in Georgia, Alabama, and Florida. The SERFC, located in Peachtree City, Georgia, has river forecast responsibility for all rivers affected by the event. Hydrologic Service Area (HSA) responsibility was shared by three offices:
NWSFO Atlanta has HSA responsibility for the rivers in Georgia, including the Chattahoochee River along the Georgia-Alabama border.
NWSFO Birmingham has HSA responsibility for rivers in Alabama and the Florida Panhandle west of the Apalachicola River.
NWSO Melbourne has HSA responsibility for rivers in Florida, excluding the rivers in the Florida Panhandle west of the Apalachicola River but including the Apalachicola River.
HSAs are defined for the issuance of longer-fused Flood Warnings and Statements, while the County Warning Area (CWA) determine responsibility for the shorter-fused Flash Flood Watches and Warnings. Figure 2-8 contains a map with the HSA boundaries and Figure 2-9 contains the CWA boundaries. CWA responsibility in the flooded area was shared by NWSFO Atlanta, WSO Columbus, WSO Macon, NWSFO Birmingham, WSO Montgomery, NWSO Tallahassee, and WSO Pensacola.
All the offices with CWA responsibility were involved in issuing Flash Flood Warnings and Statements for their areas. RFC flash flood guidance and NMC QPF were used to create the Flash Flood Watch and Warning products. The local offices did not refine the NMC QPF. Table 2-1 below shows the number of public products related to Alberto that were issued by each office.
The early part of this event was marked by major flash flooding, particularly in west-central Georgia and southeast Alabama. Although several counties suffered major flash flooding, the loss of life in Sumter County, Georgia, was extremely high compared to surrounding areas. Sumter County received the heaviest 24-hour rainfall during the storm with 21.10 inches falling at Americus in the period ending 7 a.m. July 6. The resultant flash flooding and flooding claimed 15 lives in this county.
Flash flooding in Sumter County was aggravated by the overtopping and failure of many small, unregulated earthen dams. According to the Georgia Department of Natural Resources (DNR), Safe Dams Program, a total of 218 dams failed in Georgia during this event, of which 35 were in Sumter County. Unregulated earthen dams are defined by Georgia DNR as small rural stock ponds and do not fall under the state dam inspection program. These dams often fail when an extreme rainfall event causes the outflow for the dam to exceed the spillway capacity of the dam.
Floodwaters swept many vehicles off roadways (two-thirds of the flood deaths here occurred in vehicles) as motorists attempted to cross flooded roads and bridges while floodwaters were rising rapidly. A number of homes were also flooded--and in a few cases swept away--by rapidly rising flood waters, which resulted in two of the deaths.
NWSFO Atlanta had that area of Georgia which includes Sumter County under a Flash Flood Watch (FFA) continuously beginning 4 p.m. July 3 through July 7. WSO Columbus, which has CWA responsibility for Sumter County, issued the first Flash Flood Warning (FFW) for Sumter County at 2:19 a.m. EDT on July 6. Flash flooding claimed its first victim in Sumter County around midnight, and the majority of the deaths in this county occurred between midnight and dawn on July 6.
Additional major flash flooding occurred in numerous counties stretching from central Georgia westward and southward into southeastern Alabama. Considerable property and road damage resulted due to this flooding, but the loss of life was restricted to vehicle-related incidents. Sumter County is out of range of WSO Macon's NOAA Weather Radio (NWR), which is 68 miles away (the tone alert range is 40 miles). All the Sumter County patrol cars are equipped to monitor NWR, but it blocks out other radio communications so is not used much. There is a similar situation in Bainbridge, Georgia. The Emergency Operation Center has NWR but doesn't use it nor rely on it. Additionally, emergency operation centers should have the National Attack Warning System (NAWAS), which is a telephone communication link. But, again, not every county in Georgia has this system. The Georgia Emergency Management Agency rebroadcasts NWS forecasts on radio and faxes hard copies of NWS watches/warnings to its six field offices across the State. Sumter County has no means of direct contact (other than phone) or automated dissemination from NWS. It does not receive NOAA Weather Wire Service (NWWS) or NAWAS and is out of range of NWR from WSO Macon. The same situation applies in Calhoun County, Alabama. Florida now has NAWAS in each county, however, it is in the Sheriff's office or connected to a 911 dispatch but not necessarily located in an emergency management agency/emergency operation center. Word does not always get through to emergency management agencies.
FINDING 2-11: The Sheriff/Emergency Management Agency Director for Sumter County, Georgia, receives weather watches and warnings from the public broadcast media. The county does not receive the NAWAS transmissions and is on the outside fringe of NWR reception (Americus is 68 miles from the nearest NWR transmitter). The NWR tone alert does not work reliably in the county because of this distance.
RECOMMENDATION 2-11: NWS should work with FEMA to ensure that every county emergency management agency/emergency operation center in Alabama, Florida, and Georgia has a communication link to NAWAS. Additionally, the Gore Initiative should be implemented as soon as possible to expand the NWR network of transmitters to reach 95 percent of the population.
FINDING 2-12: The Sheriff of Sumter County, Georgia, as with many other emergency management officials in the impacted area, expressed a high degree of frustration in making residents aware of the danger from the floodwaters and of the need to evacuate. Some of the deaths that occurred were people who had been warned (more than once) to evacuate but failed to act until it was no longer safe to do so. No flash flood/flood anywhere near the magnitude of this event had ever occurred in this area; and residents were, for the large part, unable to realize the dangers they faced until it was too late.
RECOMMENDATION 2-12: More emphasis should be placed on public awareness and preparedness training for flood and flash flood events. The continued high number of vehicle-related deaths during floods and flash floods indicates the need to educate the public of the risks involved with vehicles in flood situations. The "Hidden Danger" video currently being developed by the NWS should be used to inform the public of the dangers of low-water crossings.
FINDING 2-13: The Sheriff of Sumter County, Georgia, had high praise for NWS products and service during this event and did not think there was anything the NWS could have done to reduce the loss of life during this event. He did think it is a mistake for the NWS and the media to emphasize tropical storms only up until landfall; and then, in some cases, the public perceives that there is no danger because of a relatively weak wind-producing storm.
RECOMMENDATION 2-13: The NWS should work with the media to educate the public on the fact that heavy rains and widespread flooding from tropical storms and hurricanes may have as much, and in some cases even more, detrimental impact as winds at landfall.
FINDING 2-14: The disaster survey team found that the Flash Flood Warnings issued in this event were generally accurate and timely. However, many lacked a strong enough indication of the life-threatening nature of the flash flooding.
RECOMMENDATION 2-14: NWS offices should strive to better recognize truly extreme rainfall events and, in those events, use the strongest possible wording in the warnings and statements issued to make the public more cognizant of the life-threatening nature of the event.
The area affected by flooding from the rains of Alberto stretched from central Georgia southwest into southeastern Alabama and southward into the Florida Panhandle. The river basins with the most severe flooding included the Ocmulgee and Flint Basins in Georgia, the Chattahoochee Basin along the Georgia-Alabama border, the Choctawhatchee Basin in Alabama, and the Apalachicola Basin in Florida.
Appendix B includes the USGS listing of the crest values, previous record crests, and return interval of the crests for this event. Appendix C also includes selected hydrographs. Table 2-2 summarizes the number of locations by basin that equaled or exceeded the 100-year recurrence interval and the number of locations that exceeded the previous record crest.
The SERFC issued 238 river forecasts for locations in the impacted area during this event. On July 3, 24-hour operations began and continued through July 9 at 9 p.m. with the brief exception from midnight July 4 to 6 a.m. on July 5. The SERFC Hydrometeorological Analysis and Support (HAS) unit had not yet become operational and contained only one of the three positions. As was Southern Region policy at the time, the SERFC did not use QPF in the preparation of river forecasts since the majority of the WSFOs in their forecast area did not routinely supply QPFs. Interestingly, an analysis of the forecasts indicates the lack of any clear trend in the river forecasts issued. Some forecasts were very close to the observed crests several days in advance; some forecasts were lower than the observed crest; and a few forecasts were above the observed crests. Appendix D includes a chronological listing, by forecast point, of the forecasts issued by SERFC during this event.
|Record Flood Events|
|Basin||Number of locations >= 100 yr recurrence||Number of locations with new record flood|
|Alabama (includes all affected basins)||0||4|
The destruction and human misery wrought by the floodwaters were enormous. A brief summary of some of the major flood impacts is listed below.
Ocmulgee River - Macon, Georgia, was swamped by a record crest of 35.4 feet on July 7 (previous record was 28.00 feet on 11/29/48). The floodwaters overtopped and breached levees at Macon and flooded the water treatment plant. Freshwater was not restored for nearly 3 weeks. Two major Interstate Highways (I-75 and I-16) were closed for approximately 36 hours due to the floodwaters and required traffic detours of more than 100 miles. Several hundred homes were evacuated in Macon, most of which eventually flooded.
Flint River - Some of the most spectacular flooding occurred along the Flint River. The crest, generally 20-25 feet above flood stage and 4-6 feet above the previous record crest (January 1925), wreaked havoc as it moved downstream and caused immense damage as well as the evacuation of tens of thousands of people. Blackshear Dam, upstream of Albany, was overwhelmed; and the high pool level forced the evacuation of residents in about 1,400 homes around the lake (almost all of which were ultimately inundated) before the dam was overtopped and breached. Albany suffered major flood damage after nearly one-third of its 76,000 residents were evacuated. Further downstream, at Newton, nearly the entire town was flooded to depths of 15-20 feet. After exceeding previous record flood levels as far downstream as Newton, the Flint River at Bainbridge crested about 4 feet below record levels (although the measured discharge of 108,000 cfs exceeded the previous record discharge of 101,000 cfs). Section 2.3.3 of this report further analyzes the Bainbridge forecast problem.
Chattahoochee River - This river is somewhat more controlled by impoundment structures than the other rivers impacted by Alberto, but the volume of water still caused considerable flooding along the lower half the river. At Columbia Tailwater, a record crest more than 2 feet higher than the previous record, and over 12 feet above flood stage, was observed.
Apalachicola River - The Flint and Chattahoochee Rivers join at Lake Seminole, which is formed by Woodruff Dam. Outflow from Woodruff Dam flows down the Apalachicola River. The excessive inflow into Lake Seminole forced high discharges from Woodruff Dam (the peak discharge was 224,486 cfs on July 10th) and caused record flooding at Blountstown.
Choctawhatchee River - Major flooding occurred along this river in Alabama and Florida as a crest 15-20 feet above flood stage moved down the basin. This crest was about 4 feet below the record crests. Considerable damage resulted at Newton and Geneva, Alabama, and Caryville, Florida. Additional major damage occurred in Dale County, Alabama, and Holmes County, Florida.
FINDING 2-15: The disaster survey team found a high degree of satisfaction from emergency managers, the media, and the public with the river forecast services they received during this event. In particular, the impact statements and relationship to recent and historical flood levels were judged valuable information.
RECOMMENDATION 2-15: NWS HSA offices should make every effort to include up-to-date and informative impact statements in all Flood Warnings and Flood Statements.
FINDING 2-16: There were several suggestions from emergency managers and the media that the public river forecasts be updated more frequently. The normal procedure presently is to issue the Flood Statements once per day in late morning or early afternoon. In particular, an early morning update was suggested to provide current information so the public can make more informed decisions on commute, daily activities, or evacuation activities.
RECOMMENDATION 2-16: NWS offices should make every attempt to update Flood Warnings and Flood Statements more than once per day.
FINDING 2-17: Several users suggested that changes in crest forecast values be highlighted at the beginning of Flood Statements. An analysis by the disaster survey team of the Flood Statements issued during this event where the crest forecast was revised from the previous forecast showed that they, in general, did not call attention to the fact that a crest forecast had been revised.
RECOMMENDATION 2-17: Any significant change in the crest forecast from a previous crest forecast should be highlighted at the beginning of the Flood Warning or Flood Statement.
Forecasts were generally accurate and highly regarded with the notable exception of Bainbridge, Georgia, where the Flint River was forecast to reach a level some 8 feet higher than its eventual crest. This persistent overforecast was the subject of considerable negative attention by the media and the public and has resulted in some loss of forecast credibility for the NWS.
The SERFC crest forecast for Bainbridge was raised during the first days of the event (prior to July 7) due to additional heavy rainfall. Table 2-3 shows the crest forecast issued by SERFC for Bainbridge July 7-13.
|The Bainbridge Forecast|
|Issue Date||Forecast Crest/Date|
|7/7/94||Near 45 feet / July 13|
|7/8/94||Near 45 feet / July 13|
|7/9/94||Near 45 feet / July 13|
|7/10/94||Near 45 feet / July 13|
|7/11/94||44 - 45 feet / July 14|
|7/12/94||43 - 44 feet / July 14|
|7/13/94||37 - 38 feet / July 14|
The Flint River crested at Bainbridge July 14 at a stage of 37.20 feet. A discharge measurement taken by the USGS shortly before the crest (while the stage was 37.18 feet) indicated a flow of approximately 108,000 cfs. By comparison, the record flood of January 1925 reached a level of 40.9 feet (from high water marks) with a flow of approximately 101,000 cfs.
Based on the forecast provided by the NWS, and seeing the record crests that were occurring upstream, Bainbridge city officials determined the area that would be affected by the 45 foot forecast crest and proceeded with their evacuation plans and flood protective measures. In the end, only about half the evacuated area was flooded, causing much less damage than anticipated. As a result, the credibility of the NWS river forecast for Bainbridge was indeed damaged and needs to be restored.
There may be a variety of factors which led to the Bainbridge forecast error. It will not be a part of this report to completely analyze the hydrology relating to the Bainbridge forecast. Factors that may need further investigation include:
1. The lower portion of the Flint River Basin lies in a large Karst area. Karst areas are irregular limestone regions with sinkholes and underground caverns and streams. Such areas can have a significant and complex effect on modeling the hydrology of a basin.
2. The possible hydrologic effects of Big Slough Creek when Flint River stages exceed 32 feet needs to be investigated. Big Slough Creek joins the Flint River a few miles upstream of Bainbridge.
3. The stage-discharge relationship for Bainbridge used during this flood event was a graphical or mathematical extension above 70,000 cfs based on high water marks from the 1925 flood. The rating that resulted from measurements made during the 1994 flood is significantly different and has already been implemented in the SERFC forecast model.
FINDING 2-18: The forecast for the Flint River at Bainbridge received considerable media and public attention when the river crested well below the forecasted level.
RECOMMENDATION 2-18: The SERFC must investigate the causes for the Bainbridge forecast error and make the appropriate changes to the hydrologic forecast model as soon as possible. When the appropriate modifications to the hydrologic model are completed, NWS personnel (RFC and/or NWSFO) should make the necessary effort to brief the Bainbridge public officials (and media) on their findings.
The primary mission of the NWS is to save lives and reduce losses to property due to the weather. Generally stated, this is accomplished by the NWS in two equally important phases. The first is the generation of hydrometeorological forecasts and warnings; the second is internal and external preparedness activities.
The offices and individuals contacted by the disaster survey team are contained in Appendix E. The survey team found that the NWS's internal state of preparedness prior to and during the event was, for the most part, adequate. However, the survey team found external awareness of the NWS's hydrologic services to range from a high level to a relatively low level. For example, awareness ranged from knowing the NWS contact by name to confusion over who provides river forecasts and warnings. The variance seemed to be directly related to frequency of personal contact by NWS personnel with all levels of the emergency management community.
Overall, the survey team judged NWS preparedness activities to be acceptable. The fact that 33 people lost their lives caused much concern. The survey team felt that this high figure could perhaps have been lower had NWS preparedness activities been more frequent and comprehensive in two broad areas: (1) the number of personal visits to the user community by the Warning Coordination Meteorologists (WCM) and Service Hydrologists (SH); and, (2) greater emphasis in routine preparedness activities on the dangers posed to passengers in vehicles in flood situations and posed by heavy rains and flooding, including floods caused by decaying tropical cyclones.
The two deaths in Alabama involved (1) a man who, in the early hours of Wednesday, July 6, drove his vehicle around a barricade and then slid into the swollen Choctawhatchee River and (2) a 13-year-old boy playing in a storm drain, was subsequently swept away by the floodwater on July 5, and later died from his injuries. Table 2-4 contains a chronicle of the 31 deaths in Georgia from The Atlanta Constitution newspaper article dated July 31, 1994.
Table 2-4. Georgia flood-related deaths from Tropical Storm Alberto July 5, 1994
|Flood Related Deaths - July 5, 1994|
|John F. Peavy||Male||Age 54||Truck hydroplaned and hit a wrecker.|
|Richard Rodgers||Male||Age 20||car crashed.|
|Jack S. Shriver||Male||Age 40||trying to tie down a small bridge in Line Creek.|
|Teresa Beyahf||Female||Age 31||car hit a washed-out road.|
|Gloria Dixon||Female||Age 16||current pulled her under in a ditch after she rescued dog.|
|Monty Folsom||Male||Age 35||truck caught in whirlpool that formed in a flooded parking lot and was pulled through an 8-foot culvert.|
|Lisa Sheppard||Female||Age 25||passenger in truck with Folsom.|
|William Miller||Male||Age 62||car was swept into the Towaliga River.|
|Flood Related Deaths - July 6, 1994|
|Eugene Marner||Male||Age 40||truck and trailer swept away by wall of water.|
|Kent Marner||Male||Age 12||passenger in truck and trailer swept away by wall of water.|
|Roger Cornelius||Male||Age 40||passenger in truck and trailer swept away by wall of water.|
|Josephine S. Anderson||Female||Age 70||car went into a creek.|
|Walter Davenport Stapleton III||Male||Age 17||died stringing telephone lines on Lake Corinth when his boat overturned (upstream dam break caused log to ram boat; current pulled him over the dam).|
|Oscar Brown||Male||Age 84||mobile home crushed by water.|
|Idell Jackson||Female||Age 67||home crushed by water.|
|Gloria Tatum||Female||Age 28||car washed off bridge into flooded creek.|
|Tomeko Y. Woodham||Female||Age 20||car went into a creek.|
|Chad Jones||Male||Age 18||trying to rescue animals by using an inner tube on the Towaliga River.|
|Douglas K. Bassett||Male||Age 32||trying to cross a train trestle over the Towaliga River.|
|Hilton Howard||Male||Age 42||car went into a creek.|
|Freddie Hawkins||Male||Age 35||bridge washed out and truck was swept away.|
|Kedrick Hawkins||Male||Age 16||passenger in truck that was swept away when bridge washed out.|
|Kourtney Hawkins||Male||Age 8||passenger in truck that was swept away when bridge washed out.|
|Kathy R. Hurley||Female||Age 28||car was swept into a creek.|
|John Hurley||Male||Age 2||car was swept into a creek.|
|Flood Related Deaths - July 7, 1994|
|William Wallace||Male||Age 41||died searching for his mother (she was later found in a shelter).|
|Flood Related Deaths - July 8, 1994|
|Kason Mallory||Female||Age 4||passenger in father's car which plunged into the Flint in Albany.|
|Shabazz Mallory||Male||Age 2||passenger in father's car which plunged into the Flint in Albany.|
|Flood Related Deaths - July 10, 1994|
|Ishkabah T. Linkhorn||Male||Age 28||swept away by the Flint.|
|Flood Related Deaths - July 13, 1994|
|Pearlie Mae Brantley||Female||Age 59||drowned when Flint River floodwaters filled her home.|
|Flood Related Deaths - July 14, 1994|
|Maureen Johnson||Female||Age 71||car plunged into a creek in Terrell County.|
The survey team felt expectation levels of the NWS by the public and emergency management community were based on its past capabilities. Hence, relative to the NWS's capabilities during its transition and modernized phases, the level of expectations by the users was perhaps low. This poses a significant challenge to the NWS: Improving the public's awareness level of the impacts of both short-term and longer-term hydrometeorological events must be put on at least the same level as improvement of the NWS's scientific capabilities in the generation of forecasts and warnings.
FINDING 2-19: Some communities, and perhaps emergency managers, were not as prepared for the disastrous floods as they could have been if there were greater personal contact and education on floods by NWS WCMs and SHs.
RECOMMENDATION 2-19: NWS policy should require periodic (annual if possible) personal visits by the WCMs and/or SHs to emergency management and other action agencies from the state to the local level. These contacts should include a review of the flood threat to the local community (emphasizing the threat to vehicular passengers) and a review of the hydrologic services that the NWS provides. This educational process should specify what products are available, how they can be used, and where they can most efficiently be obtained.
Several individuals interviewed by the survey team remarked that there may have been an overemphasis on the landfall of Tropical Storm Alberto and not enough attention focused on the potential for heavy rain and flooding associated with the storm. This is probably more of a media and public perception problem than a forecast problem. For example, the Meteorological Operations Division of NMC highlighted a "very dangerous flash flood and flood situation for much of Georgia today [July 4] into tonight as the remnants of Alberto drift slowly north."
FINDING 2-20: The public's perceived threat from Alberto appeared to lessen once it made landfall.
RECOMMENDATION 2-20: The NWS and NOAA should take maximum advantage of the recommendations from the 1995 Interdepartmental Hurricane and the NOAA Hurricane Conferences, which focused on the inland effects of tropical cyclones, in order to enhance the public's perception of the dangers associated with landfalling tropical cyclones. In addition, the WCMs in all areas which might be affected by the aftermath of decaying tropical cyclones should reenforce the potential for severe flooding from such storms with the user community.
There is room for improvement in better identification of flood-prone areas. Traditionally non-NOAA agencies identified flood-prone areas, most often as part of a flood insurance study. If potential flood inundation maps were widely available and used, the NWS and emergency management personnel could coordinate more easily with local communities to communicate the potential impact a disastrous flood could have on their community.
FINDING 2-21: The disaster team believes another possible contributing factor to the high death count could be that the public was not adequately educated regarding the locations of flood-prone areas (particularly roads), safe evacuation routes, and the potential impact of their actions.
RECOMMENDATION 2-21: If funding permits, the NWS, in conjunction with FEMA and appropriate state and local agencies, should embark upon a campaign to educate the public as to their local flood-prone areas. This should include a widely distributed array of visual representations of flood-prone areas depicting roads and bridges as well as portions of communities that may be potentially inundated by floods. Additionally, the NWS should plan to issue graphical flood forecasts as well as the traditional text products.
An area that the NWS clearly needs to improve is its national-level response to disasters. FEMA has improved the timeliness and magnitude of its disaster response. During a disaster, FEMA is able to quickly establish a television communications link to the affected communities as well as to key Federal officials in Washington, D.C., and other areas across the Nation. The NWS must be prepared to participate in or establish similar radio and satellite communication links and be a part of the information superhighway.
The media showed tremendous interest in broadcasting from the NHC before and during landfall of Alberto. The NHC Director (or representative) was prominently featured on the broadcasts, which heightened the sense of urgency. Once the storm was inland, however, there was much less attention by the media and no single spokesperson from the NWS focusing attention on continued potential from weather hazards.
FINDING 2-22: The disaster team felt it was inappropriate for a single NWS office to be expected to respond to an event that covered multiple offices and to FEMA's national-level press and Federal coordination briefings. In addition, there was an imbalance in the media contacts and interest with NHC prior to landfall and WFB once the tropical cyclone had made landfall.
RECOMMENDATION 2-22: The NWS should establish a national media unit to provide beginning-to-end coverage of storm events that have national impact or interest. This unit would provide a consistent posture in front of the national media, which could emphasize the dangers associated with each phase of the storm. The unit would be headed by a public affairs specialist and supported by an ad hoc team of meteorologists and hydrologists, as appropriate for the event. Teleconferencing should be utilized to maximize participation of personnel from a variety of NWS offices.
Main Link Categories:
Home | NWS | OHD
US Department of Commerce
National Oceanic and Atmospheric Administration
National Weather Service
Office of Hydrologic Development
1325 East West Highway
Silver Spring, MD 20910
Page Author: OHD webmaster
Page last modified: November 3, 2011