| | The objectives of the two-year program were to (1) Assess and address the public health service needs generated by the flood disaster related to environmental health, infectious diseases, child health, chronic conditions, and injury of the 320,000 residents in the 14 counties of the Albany health district, (2) Undertake environmental health activities to identify the remaining concerns produced by flood damage and its aftermath related to well and septic contamination, damage and displacement at known hazardous waste sites, and breeding of disease-causing insects throughout the 55 affected counties, (3) Produce a comprehensive state disaster response field operations manual and conduct training, and (4) Develop the skill capacity of district and local level staff in all parts of Georgia to undertake disaster-response activities related to overall crisis management and emergency response, environmental health concerns, and epidemiologic surveillance as well as provide these staff with both primary data and necessary equipment needed in such activities. |
| Figure 1. Flood Boundary and Declared Counties in Georgia. |
| Figure 2. Area of Concentration. | |
 |
In Calhoun County 543 wells were located by GPS (see Figure 3). Data collection has been completed for 387 of these wells. Of these wells, 173 (45%) tested positive for coliform bacteria, 205 (53%) showed no evidence of contamination, and 9 (2%) were removed from the analysis due to incomplete data.
Figure 3. Well Locations in Calhoun County.
Of the 173 wells testing positive for total coliform bacteria, 136 wells were chlorinated and subsequently tested again. These 136 wells serve as the study group for the analyses reported here. Figure 4 shows the number of wells failing chlorination. Of the 136 contaminated wells, 81 (59.6%) wells tested positive for total coliform bacteria after chlorination and 55 (40.4%) showed no contamination after chlorination. Eight (5.9%) of the wells tested positive for fecal coliforms. Twelve (8.8%) of the 136 wells were located in the tributary flooding zone. Eight (9.88%) of the 81 contaminated wells were located in the tributary flooding zone and four (7.27%) of the 55 uncontaminated wells fell in the flooded zone. Information obtained on the entire study area should give sufficient data to determine if the relationship between tributary flooding and well contamination is statistically significant.
Figure 4. Wells Failing Chlorination.
In order to determine if characteristics associated with poorly constructed or maintained wells were associated with well contamination, questionnaires were administered at each site in addition to collection of water samples. These data are summarized in Table 1. This table shows the frequency of occurrence of each variable, the frequency of each according to comtamination status after chlorination, and the probability that an association at least as strong as that seen in the data might have arisen by chance alone. All 136 wells were drilled and 135 (99.3%) had a depth of 100 ft. Because of the lack of variability in these two well characteristics, it could not be determined whether well type or depth were related to well contamination after chlorination. Because well depth tends to vary spatially according to the underlying geologic characteristics, similar well depths for a localized county area can be expected. As more data become available from the other 10 counties, well depth and its relationship with contamination will be able to be analyzed further. Approximately 80% of wells in Georgia are drilled and 90% of wells in this 11 county area are drilled (3). Because such a high percentage of wells are drilled, it is unlikely that that this study will allow an adequate analysis of well type and its possible associations with well contamination. However, information obtained on this variable should give sufficient information to generate hypotheses concerning well type as well as provide a source for further investigation if deemed necessary.
Table 1. Frequencies of Well Characteristics
| Well Characteristic | Frequency of Occurence (N=136) | Percentage | # Contaminated Wells with Well Characteristic (N=81) | # Uncontaminated Wells with Well Characteristic (N=55) | Prob. |
| No Seal | 6 | 4.4% | 4 (4.94%) | 2 (3.64%) | 0.779 |
| No Slab | 73 | 53.7% | 42 (51.9%) | 31 (56.4 %) | 0.606 |
| No Well Cover/House | 57 | 41.9% | 39 (48.2%) | 18(32.7%) | 0.075 |
| Cracked | 2 | 1.5% | 2 (2.47%) | 0 (0.0%) | 0.353 |
| No Grout | 8 | 5.9% | 6 (7.41%) | 2 (3.64%) | 0.300 |
| Upgrade Septic Tank | 90 | 66.2% | 55 (67.9%) | 35 (63.6%) | 0.607 |
| Unprotected from Surface Drainage | 8 | 5.9% | 4 (4.94%) | 4 (7.27%) | 0.415 |
| No Disinfection Unit | 131 | 96.3% | 77 (95.1%) | 54 (98.2%) | 0.325 |
| Construction Material | |||||
| Steel | 108 | 79.4% | 65 (80.3%) | 43 (78.2%) | 0.771 |
| PVC | 28 | 20.6% | 16 (19.8%) | 12 (21.8%) | |
| Diameter | |||||
| Less than 4 Inches | 8 | 5.9% | 6 (7.41%) | 2 (3.64%) | 0.300 |
| 4 or More Inches | 128 | 94.1% | |||
| Pump Type | |||||
| Jet Pump | 33 | 24.3% | 22 (27.2%) | 11 (20.1%) | 0.341 |
| Sumersible | 103 | 75.7% | 59 (72.8%) | 44 (80.0%) | |
| Well Depth | |||||
| 100 Feet | 135 | 99.3% | |||
| 300 Feet | 1 | 0.7% | |||
| Well Type | |||||
| Drilled | 136 | 100% |
Wells that remained contaminated after chlorination were more likely than uncontaminated wells to have no seal, no grout, an upgrade septic tank, and a jet pump. However, none of these associations showed statistical significance. The variable "No well cover/house" was marginally significant with a p-value of 0.07. No significant differences were seen between type of construction material or diameter. The small numbers associated with several of these variables could have prevented real associations from resulting.
In order to determine if land type was associated with continued well contamination, land data was integrated into the well characteristic data base using ArcView GIS 3.0. Figure 5 shows a portion of Calhoun County demonstrating well locations in relation to land type. These data are also summarized in Table 2.
Figure 5. Example of Well Locations in Relation to Land Type in Central Calhoun
County.
Table 2. Frequencies of Wells for Each Land Type
| Land Type | Frequency (N=136) | Percent | # Contaminated Wells in Each Land Classification | # Uncontaminated Wells in Each Land Classification | Prob. |
| Clear Cut/Young Pine | 6 | 4.3% | 4 (4.82%) | 2 (3.64%) | 0.547 |
| Pasture | 28 | 20.3% | 18 (21.7%) | 10 (18.2%) | 0.616 |
| Cultivated/Exposed Earth | 76 | 55.1% | 47 (56.6%) | 29 (52.7%) | 0.652 |
| Urban | 3 | 2.2% | 1 (1.2%) | 2 (3.64%) | 0.349 |
| Wetland | 3 | 2.2% | 2 (2.41%) | 1 (1.82%) | 0.651 |
| Forest | 20 | 14.5% | 9 (10.9%) | 11 (20.0%) | 0.135 |
The majority of the well points were located in areas classified as cultivated/exposed earth, pasture, and forest. Contaminated wells were not significantly more frequent in any one land classification compared to all others combined. Once data collection is completed for the remaining ten counties in the study area, spatial analyses can be done comparing well locations with land type as well as with other geologic characteristics such as soil type and aquifer location.
Univariate and multiple logistic regression were used to determine odds ratios measuring the relationship between the predictor variables and chlorination failure. In the univariate analysis, none of the predictor variables were significantly associated with continued well contamination. For the multivariate analysis, models were developed to control for bias due to confounding. Results of this analysis are shown in Table 3. An initial model was fit which contained all predictor variables related to well construction, flooding, and land type. The inclusion of the variables no seal, cracked, no grout, diameter, and no disinfection unit did not change the outcome of the model and were removed. Similarly, all land type variables except pasture were removed from the model. The rest of the predictor variables were retained in the model in order to control for confounding. Controlling for all other variables, wells that remained contaminated after chlorination were 3.2 times more likely to have no well cover/house than those which were not contaminated (p=0.008). Having a jet pump was marginally significant with a p-value of 0.077. Contaminated wells were 2.3 times as likely to have a jet pump rather than a submersible pump. None of the other variables were significantly associated with failure of wells to clear after chlorination.
Table 3. Multivariate Logistic Regression Analysis.
| Predictor Variable | Contaminated Wells | Uncontaminated Wells | Odds Ratio | 95% Confidence Interval | P-value |
| No Well Cover/House | 48% | 33% | 3.196 | 1.36 to 7.52 | 0.0078 |
| Pasture | 22% | 19% | 1.813 | 0.68 to 4.84 | 0.235 |
| Tributary Flooded Zone | 10% | 7% | 0.964 | 0.24 to 3.82 | 0.959 |
| Pump Type (Jet pump) | 27% | 20% | 2.344 | 0.91 to 6.03 | 0.077 |
| No Slab | 52% | 57% | 0.546 | 0.24 to 1.24 | 0.146 |
| No Disinfection Unit | 95% | 98% | 0.234 | 0.02 to 2.47 | 0.227 |
| Upgrage Septic Tank | 68% | 64% | 1.30 | 0.55 to 3.09 | 0.548 |
| No Drainage Protection | 5% | 7% | 0.407 | 0.068 to 2.43 | 0.324 |
| Construction Material (Steel) | 80% | 78% | 1.25 | 0.48 to 3.20 | 0.649 |
The preliminary results of this study suggest that continually contaminated wells are significantly more likely to be uncovered (without a shelter or well house) than are wells easily cleared of contamination by chlorination. When data collection is completed for the entire study area, the data should allow a more precise evaluation of how well characteristics and tributary flooding are associated with continued contamination of wells. Only flooding along rivers and tributaries can be easily measured. Many areas in the flooded counties received large amounts of rainfall and were subject to standing water and surface run-off and are not included in the tributary flooding zone. If results show that wells in the tributary flooding zone are no more likely to be continually contaminated than wells not in the tributary flooding zone, then precautions to secure and protect wells may be needed in these areas as well as in areas along rivers and tributaries. This could add evidence of aquifer contamination as well.
One limiting factor in the study concerns confounding factors that could not be controlled for in the present analysis. Individual wells are subject to varying local sources of contamination. Because this study developed in the aftermath of a serious and immediate public health need to decontaminate wells, this data could not be collected. The wells sampled were also not a random sample of wells in the area but were sampled because of fears local residents had of contamination. This could have resulted in an artificially high percentage of contaminated wells, however, large numbers of wells located throughout the study area were sampled. It is not likely that these wells were different with regard to well construction and maintenance from wells that were not tested. Because wells were sampled in all areas of the counties, it is also unlikely that flooding differed greatly between sampled and unsampled wells.
Natural disasters such as floods can cause severe devastation and affect virtually everyone in some way. Because large numbers of persons in rural areas use private wells as their primary source of drinking water, contamination caused by flooding is a public health problem. Data obtained from this study and from the state-wide survey should give needed information on the relationships of well construction and flooding with well contamination and chlorination failure. This knowledge could be used to prevent future flood-induced well contamination and to indicate areas in which prevention resources can be used most efficiently.
2. Hicks DW, Gill HE, and Longsworth SA. "Hydrogeology, chemical quality, and availability of ground water in the Upper Floridan aquifer, Albany area, Georgia." U.S. Geological Survey, Water-Resources Investigations Report 87-4145. 1987.
3. U.S. Census Data, 1990
4. Freedman B. Sanitarian's Handbook - Theory and Administrative Practice for Environmental Health. 1977.
5. Landcover classification of Georgia 1988-1990. Georgia Department of Natural Resources. U.S. Geological Survey, 1995.