USGS - science for a changing world

South Atlantic Water Science Center - Georgia Office

 South Atlantic WSC  Georgia office  Information/data  Projects  Publications  RiverCam  Drought  Flood  About  Contact

Picture of the main South Atlantic Water Science Center - Georgia office.


[an error occurred while processing this directive]


USGS Water Science Centers are located in each state.

There is a USGS Water Science Center office in each State. Washington Oregon California Idaho Nevada Montana Wyoming Utah Colorado Arizona New Mexico North Dakota South Dakota Nebraska Kansas Oklahoma Texas Minnesota Iowa Missouri Arkansas Louisiana Wisconsin Illinois Mississippi Michigan Indiana Ohio Kentucky Tennessee Alabama Pennsylvania West Virginia Georgia Florida Caribbean Alaska Hawaii New York Vermont New Hampshire Maine Massachusetts South Carolina North Carolina Rhode Island Virginia Connecticut New Jersey Maryland-Delaware-D.C.

The 1977 Toccoa Flood
Report of Failure of Kelly Barnes Dam Flood and Findings

by Federal Investigative Board
December 21, 1977


1. General. No design, plans, or construction cross sections for any of the various stages of construction and/or repair have been discovered. The maximum section of the dam was destroyed by the failure. It can only be presumed that the cross sections of the dam, as seen on the two faces of the breach, continued across the creek. The dam was generally crescent shaped, concave upstream, and apparently followed a low ridge.

Though it is impossible to fully reconstruct the dam condition and configuration prior to the failure, the following is considered a reasonable reconstruction of recent conditions in the vicinity of the failure. Figure 4 is a probable configuration based on site inspections, historical data, and old photographs.

The embankment in the area of the failure was approximately 38-feet high with a crest width of about 20 feet. The upstream slopes were approximately 1V on 1H and the downstream slopes were about 1V on 1H. The main or newer embankment was constructed over portions of the old rock crib dam, all of which may or may not have remained beneath the downstream toe. There was probably a berm approximately halfway up the downstream slope.

Examination of the cross section that remains discloses four feet of different colored soil below the crest. Whether this is part of the 1939 or 1948 additions, or was added by later maintenance, is not known. Therefore, it can be concluded that the dam was built in two to four stages commencing with a rock crib dam in 1899.

There were apparently two steel plate lined masonry inlet structures. (Photo 3) One structure, at some point in time, controlled water into a welded steel penstock to the Toccoa Falls Powerhouse and was located on the right side of the embankment with the outlet pipe running sub-parallel to the dam axis. The other structure was essentially a low-level spillway inlet structure and had a pipe extending through the dam in the direction of the old creek bed. These inlet structures and pipe lines are shown on Figures 1 and 2. The embankment slopes were heavily vegetated. There is no evidence of heavy surface erosion. There is no evidence to indicate either structure was functional in recent years. Evidence indicates that the main earth spillway was nearly always in use and the low-level spillway through the dam was blocked or plugged. Photographs taken in 1973 show that the area surrounding the low-level spillway outlet pipe had experienced seepage for a number of years and that a slope failure had developed in the vicinity of the pipe on the downstream face of the dam (Photos 12 and 13). Historical accounts support the photographic evidence of long term seepage in the vicinity of this pipe.

2. Site Geology. The dam site is located in the Piedmont Province. The rock type at the site is a biotite gneiss of the Carolina gneiss series.

The bedrock is exposed over the stream channel and in the banks at, and just downstream of, the dam site. The rock exposed at the site is a strong, competent, biotite gneiss with minor, shallow weathering. The primary joint sets include two near vertically dipping sets, one striking N80E and the other striking generally North-South. Another primary joint set is nearly horizontal. A rather extensive root system has penetrated the near horizontal joint set along the creek bank near the downstream toe of the dam.

A layer of soft alluvium, from two to five feet thick, was observed in both faces of the breach of the dam. This material extends from the proximity of the old rock crib dam upstream--an undetermined distance. The layer appears to be quite extensive in the foundation of the earth dam. The lower six inches of the layer of the alluvium contains an extensive root mat. The alluvium appears to be composed principally of soft, plastic clay and/or silt with some layers of fine silty sand.

3. Meteorological Conditions. Before rain began the ground was already wet from heavy rainfall of 3-1/2 to 4-1/2 inches which fell on October 25-26. The rainfall began on Wednesday morning, November 2, and ended by midnight on Saturday, November 5. A strong high pressure area, centered over New England, was bringing Atlantic moisture into Georgia on Wednesday and this produced the rainfall at the beginning of the storm period. Meanwhile, a more important development was taking place in Texas where a strong low pressure area was developing at upper levels in the atmosphere. This intense upper level low was located near Shreveport, Louisiana on Thursday morning, November 3; New Orleans, Louisiana on Friday morning, November 4; Mobile, Alabama on Saturday morning, November 5; and Centreville, Alabama on Sunday, November 6. As this slow-moving upper level storm moved closer to Georgia, more and more moisture was brought in on southerly winds from the Gulf of Mexico. Lifting of the air by the higher terrain of north Georgia accentuated the lifting processes provided by the atmospheric storm patterns. A plentiful supply of moisture, lifting of the air, and a slow-moving storm system are the ingredients of a long, heavy rain event.

Rainfall reports are received by the National Weather Service from radio stations WLET and WNEG in Toccoa. Since these stations are located some distance from the drainage area above Kelly Barnes Dam, a rainfall survey was made. Three additional storm total values were obtained as shown on Figure 5. In addition to the storm totals, measurements of rainfall at 11:00 p.m. each night were available from one source while the 24-hour amount from 8:00 a.m., November 5 to 8:00 a.m., November 6 was measured by another local citizen. These measurements, interviews with local residents, radar echoes, and hourly values from nearby recording stations were used in estimating the rainfall distribution for the drainage basins above Kelly Barnes Dam. Although radar echoes indicated that the heaviest rainfall probably occurred between 6:30 p.m. and 7:30 p.m., on November 5, the estimated maximum one-hour value is shown between 6:00 p.m. and 7:00 p.m. since points are plotted on the hour.

As indicated on Figure 6, the rainfall was light for the first two days with slightly over one inch falling in the Toccoa area by Friday morning, November 4. Rainfall intensity increased somewhat for the next 24 hours, with a storm total of approximately 2-1/2 inches by 8:00 a.m, Saturday morning, November 5. The storm total increased to approximately 3-1/2 inches by noon on Saturday and then the rainfall was apparently light until about 6:00 p.m., when showers moved into the area. From 6:30 p.m. to 7:30 p.m., radar echoes indicated periods of torrential rainfall with frequent intense lightning and thunder. A small tornado apparently touched down about five miles southeast of Kelly Barnes Dam at approximately 8:00 p.m. The damage to trees was limited to only 100 yards in width and about one-fourth mile in length but the tornado was indicative of the severity of the weather. After 8:00 p.m. the showers were less frequent and had practically ceased by midnight. Estimated basin rainfall for the entire period is about seven inches with almost 3-1/2 inches occurring between 6:00 p.m. and midnight on November 5.

Rainfall intensities estimated for this storm are common for the Toccoa area. The estimated maximum one-hour, two-hour, three-hour, six-hour, 12-hour and 24-hour rainfall amounts indicate return periods between two and five years.

A heavy rainfall period occurred on May 27-29, 1976 in the Toccoa area. The 12-hour and 24-hour rainfall amounts were larger for that storm than for the November, 1977 rainfall period. However, rainfall amounts for six-hour and shorter durations appeared to be heavier in the November 1977 storm.

4. Hydrology and Hydraulics. Information regarding the lake capacity, drainage area, inflow and outflow to Kelly Barnes Lake during the period of November 5-6, 1977, was determined from field surveys, topographic maps, interviews with local residents, and engineering computations. Based on this information it is possible to determine, within reasonable limits, the probable inflow hydrograph for a condition assuming the dam had not failed. In addition, certain inferences can be made regarding the dam failure.

The drainage basin above Kelly Barnes Dam is heavily wooded with pines and deciduous trees (Photo 4). It is located in the Piedmont Province in northeastern Georgia. Stream slopes are fairly steep, generally exceeding 100 feet per mile. The drainage area of Toccoa Creek at the head of Kelly Barnes Lake is 4.10 square miles, and at the dam site 4.57 square miles.

The normal pool level for Kelly Barnes Lake is about 596.5 feet (assumed datum). The crest of the flood of November 5-6, 1977, (a 5- to 10-year flood) reached a level in the lake of approximately 601 feet which is about five feet below the top of the dam. A hypothetical 100-year flood, assuming no storage attenuation, would cause a maximum lake level of about 604 feet which is about two feet below the top of the dam. The storage attenuation would reduce this level.

The capacity of Kelly Barnes Lake was determined from a topographic map made by plane table survey November 25, 1977. At a level of about 596.5 feet, slightly higher than the main earth spillway, the lake contained approximately 410 acre feet of water with a surface area of about 40 acres. At a level of about 599 feet, equal to the elevation of the secondary spillway, the lake contained approximately 540 acre feet. At the highest level reached on the night of the dam failure, approximately 601 feet, the lake contained approximately 630 acre feet of water.

The peak discharge of Toccoa Creek, at the head of Kelly Barnes Lake, for the flood of November 5, 1977, was determined from a field survey of cross sections, high-water profiles, and channel roughness coefficients. The discharge was computed to be approximately 840 cfs (cubic feet per second). Based on a regional flood-frequency analysis for small drainage basins, the average return period of this flood is estimated to be five to ten years.

Additional computations were made using unit-hydrograph techniques to fit a hydrograph to the computed peak discharge. An estimated inflow hydrograph to Kelly Barnes Lake was derived by making appropriate adjustments for the additional intervening drainage area.

The relationship of the lake level to outflow discharge for the main earth spillway and the secondary spillway were defined from field surveys and hydraulic computations. The two relationships were combined to provide a composite (total) outflow rating. This rating indicates that at lake level equal to the maximum height reached on the night of the failure, the two spillways discharged a total of about 400 cfs.

Streamflow and reservoir routing techniques were used to estimate the probable lake-level hydrograph, the lake-outflow hydrograph, and the stream-flow hydrograph just below Dead Man's Branch in the Toccoa Falls College area for the night of the dam failure. The computations are based on the assumption that the dam did not fail, so that the probable normal situation could be depicted. The stream-flow hydrograph just below Dead Man's Branch was computed because testimony of local residents provided information which could be used to roughly verify the computations.

Based on these computations, the lake began a significant rise at about 6:00 p.m. to 7:00 p.m. on November 5. Outflow at the time was limited to the main earth spillway of about 50 cfs. At about 8:00 p.m. to 9:00 p.m. the lake reached a level so that outflow began through the secondary spillway. Total computed outflow at this time was about 120 cfs. Had the dam not failed, the lake would have risen to a level of about 601.5 feet and would then have started a slow decline after about 11:00 p.m. on November 5. The maximum outflow through the two spillways would have been about 500 cfs. It would have taken at least six hours following the crest for the lake to recede to a level below the level of the secondary spillway, assuming the dam had not failed.

The streamflow hydrograph computed for Toccoa Creek just below Dead Man's Branch indicates a peak discharge occurring at about 9:30 p.m. November 5. The timing is supported by testimony of local residents who state that the creek overtopped a bridge in the vicinity at about 9:00 to 10:00 p.m., and then began to recede. After this time, local residents state the stream was "up and down" several times. No information is available to support the computation of peak discharge; however, the good agreement of timing indicates the hydrograph computations are reasonable.

The actual maximum level reached by the lake was 601.1 feet based on field surveys of high-water marks. This is only about a half a foot different than the computed elevation, and is well within the range of error to be expected in such computations. The computed time at which the lake level rose above the secondary spillway is significant. It is highly probable that the water would be flowing through this spillway after 8:00 or 9:00 p.m. on November 5, and throughout the remainder of that night and the early morning hours of November 6. However, a local volunteer fireman has stated that he and another person went to the dam at about 10:00 or 10:30 p.m. and crossed the secondary spillway. He said there was no water in the spillway and the lake level was "not even close" to the secondary spillway. Newspaper accounts say the men went to the dam two hours prior to the break, or about 11:30 p.m. In either case, the lake level should have been over the secondary spillway. Considering this in light of the hydrograph computations, it would imply that outflow, in addition to that carried by the two spillways, occurred to cause the lake to recede from a crest at about 601 feet, and to a level below the secondary spillway before the men went to the dam. This would suggest a partial failure of the dam several hours prior to the main failure. Additional support for this conclusion is that field observations of bent grass in the lake bed indicate little or no velocity occurred between the maximum lake level of 601.1 feet and about 597 feet. Below a level of 597 feet, a definite pattern of high velocity was noted, indicating a large amount of flow occurred in the lake in a short time span after the lake receded to that level. A portion of this drawdown could have occurred after the main failure began. Furthermore, the statements regarding the "up and down" nature of the stream in the college area after about 9:00 p.m. suggests something unusual happened before the main break.

The major failure of Kelly Barnes Dam occurred at about 1:30 a.m. November 6, 1977. At that time a surge of water was released from the lake. Field surveys of cross sections, high-water profiles, and roughness coefficients were used to compute a peak discharge of 24,000 cfs in the channel at a site about 800 to 1000 feet downstream from the dam.

5. Stability Analysis. A steady seepage stability analysis, utilizing unconsolidated-undrained tests performed on undisturbed samples, was performed for the downstream slope of the dam. For the analysis, the phreatic line was assumed to exit at a point approximately one-half way up the downstream slope. The cross section used included the estimated 12 foot by 20 foot slope failure section shown by the 1973 photos of the downstream slope. The other assumed slopes are based on observations of the remaining slopes, topo map, and reports of witnesses. The sliding circle method of analysis was used. The stability of the assumed embankment section with the stated conditions is marginal as indicated by a factor of safety of 1.02.

Next page. Next page

Accessibility FOIA Privacy Policies and Notices

Take Pride in America logo logo U.S. Department of the Interior | U.S. Geological Survey
Page Contact Information: Howard Perlman
Page Last Modified: Friday, 02-Dec-2016 12:11:20 EST