Water samples were collected for bacteriological testing during three summer seasons: July 8, 1993, through August 11, 1993, in the Piedmont carbonate and the Appalachian Mountain sandstone and shale subunits; June 28, 1994, through August 16, 1994, in the Piedmont crystalline and the Appalachian carbonate subunits; June 26, 1995, through August 9, 1995, in the Great Valley carbonate subunit. Water from each of the 146 wells from these five subunits was sampled once for this study. The wells were chosen on the basis of both subunit criteria and construction criteria. Owner permission to use the well was an additional criterion.

Subunit criteria included the physiographic province, bedrock type, and land use. Subunit boundaries were established with a geographical information system (GIS). A computerized random-selection program was then used to select potential sampling locations within each subunit (Scott, 1990). Field visits were conducted to confirm land-use criteria, and the bedrock type was confirmed by geologic maps and drilling records. Sites were chosen such that there were no feedlots within a half mile radius.

Once a potential sampling location was selected, homeowners were queried and drilling records were checked to determine if their well met the well-construction criteria. The well-construction criteria were established to obtain samples representative of shallow ground water. These criteria dictated that the wells were open boreholes, drilled and completed in the bedrock, less than 200 ft deep, and less than 25 years old. Most wells sampled met all of the sampling criteria; in few cases where these criteria could not be met, deeper or older wells were sampled (table 2). All wells sampled had similar household plumbing designs. A detailed plumbing inspection ensured that the sample was raw water. Most samples were collected from an outside faucet on the house. Since it was not possible to fill the bottles directly under the spigot in many cases, a Teflon sampling hose was used to fill the sample bottles. If a raw water spigot was not available outside the house, the Teflon sampling hose was connected directly to a spigot at the pressure tank. The sampling hose was cleaned with a Liquinox soap solution and rinsed with 5 gallons of deionized water between sites. Hoses were not autoclaved or disinfected; the amount of water that passed through the hose prior to sampling was 100 to 300 gallons.

A minimum of one well volume of water was purged prior to sampling. The well volume was estimated from the static water level, the casing diameter, and the total depth of the well. Latex gloves were worn during sample collection. Water samples for bacteriological analyses were collected in 1-L, amber-colored, glass bottles that had been sterilized at 121^o C and 15 psi for 15 minutes in an autoclave. The samples of untreated ground water were collected in a plastic enclosure designed to reduce the risk of airborne contaminants entering the sample bottles. Water samples were kept on ice until they were processed for analysis.

Water samples were analyzed for concentrations of indicator bacteria of two taxonomic groups. Total coliform, fecal coliform, and E. coli are members of the coliform group. E. coli are a subset of fecal coliforms and fecal coliforms are a subset of total coliforms. Total coliforms are those organisms that produce a golden-green metallic sheen within 24 hours on m-Endo media when incubated at 35^oC. Fecal coliforms are organisms that produce blue colonies within 24 hours on m-FC media when incubated at 44.5^o C. E. coli are organisms that produce a bright blue fluorescent perimeter around a darker colony center within 4 hours when incubated at 35^o C on NA-MUG media after primary culturing as total coliform bacteria on m-Endo media. The fecal streptococcus group also was studied and is defined as all organisms that produce red or pink colonies within 48 hours on KF media when incubated at 35^o C. A summary of processing methods is given in table 3.

All water samples were processed by use of membrane filtration techniques within 6 hours of collection (Britton and Greeson, 1989); most samples were processed within 1 hour. From the 1-L sample of ground water, 100-mL aliquots were measured in sterilized, glass graduated cylinders. The aliquots were filtered with a hand-operated vacuum pump through 0.45 or 0.65 µm pore-size membrane filters mounted in sterilized plastic funnels. The membrane filters were placed on the media in petri dishes. The petri dishes were immediately placed into incubators in an inverted position.

Table 3. Summary of sample processing methods for determination of bacterial concentrations in well water --------------------------------------------------------------------------------- Bacteria type Media Processing method Reference type --------------------------------------------------------------------------------- Total Coliform m-Endo membrane filtration (0.45 µm) Britton and Greeson immediate incubation (1989, p. 13-16) Fecal Coliform m-FC membrane filtration (0.65 µm) Britton and Greeson immediate incubation (1989, p. 37-40) E. coli NA-MUG membrane filtration (0.45 µm) U.S. Environmental First incubate on m-Endo Protection Agency media for 24 hours and then (1991, p. 1) transfer to NA-MUG media Fecal KF membrane filtration (0.45 µm) Britton and Greeson Streptococcus immediate incubation (1989, p. 47-50) ----------------------------------------------------------------------------------

RETURN TO TOP OF SECTION

Quality-assurance measures were practiced throughout the study. All samples tested for bacteria were processed and analyzed by USGS personnel from the Pennsylvania Water Science Center Office in Lemoyne, Pa. The effectiveness of sterilization procedures was checked by processing a sterile-water blank at each sampling site. Water used for the blanks was a sterile phosphate buffer solution with peptone. Instructions on how to prepare buffered dilution water are found in Britton and Greeson (1989, p. 18). Blanks were
30 mL of sterile buffered water that were processed with the sterilized equipment before the sample water was processed. If colonies formed on the blank plates, the corresponding site sample results were disregarded. This occurred one time for fecal streptococcus out of 438 blank plates for total coliform, fecal coliform and fecal streptococcus. Duplicate samples were filtered and analyzed for each bacteria type at each site. The average bacteria colony count of the two plates was recorded. If either of the plates was unreadable, the colony count from the remaining plate was recorded.

Because the sampling hose was not sterilized, and no blanks were collected through the sampling hose, the data were analyzed to determine if cross-contamination from site to site could have occurred. All detections were compared to the detections at the previous site. At 67 percent of the sites, cross contamination was not possible because either the previous site or the following site had no detections of bacteria. At an additional 11 percent of the sites cross contamination was highly unlikely because a high bacteria count was preceded by a low bacteria count. At the remaining sites it is numerically possible that cross contamination occurred, because a low bacteria count was preceded by an equal or higher bacteria count. It is considered to be very unlikely that this occurred because of the cleaning of the hose and the large volume of water (100 to 300 gallons) that passed through the hose before the sample was collected. Also, at 16 percent of the sites a detection of bacteria was followed by a nondetection of bacteria, which indicates the effectiveness of the cleaning process in those cases.

All sample media and sterile buffered water were obtained from the Quality of Water Service Unit (QWSU) in Ocala, Fla. The bacteria kits that QWSU supply must pass quality-assurance tests performed by the USGS National Water Quality Laboratory (Horowitz and others, 1994). All media were fresh and used prior to the expiration date.

Statistical tests were selected to determine 1) relations between concentrations of bacteria and categorical variables such as land use or lithology and 2) relations between concentrations of bacteria and continuous variables such as well depth or nitrate concentrations. Histograms and Wilk-Shapiro tests were used to determine the normality of the distribution of the bacterial concentrations. The data are not normally distributed. Therefore, nonparametric statistical methods were used to analyze the data. The degree of censoring was quantified by determining the percentage of nondetects for each bacteria type. Some data also are censored at an upper detection limit (> 80 colonies/100 ml) because dilutions were not conducted.

The Kruskal-Wallis test was used to make comparisons in the ranks of concentrations of bacteria between categorical variables. This tests for differences in the mean ranks of two or more groups. If the Kruskal-Wallis tests on the entire population showed significant differences among categories, a Multi-Stage Kruskal-Wallis (MSKW) test was performed on the ranked data to show how the categories differed with an overall alpha value of 0.05. A probability was calculated for each statistical test conducted. If the probability is less than the alpha value (0.05) for the Kruskal-Wallis test, there is a 95 percent probability that categories are significantly different, or less than 5 chances out of 100 that the categories being tested are from the same population. This test was used on data sets in which less than 50 percent of the data were below the detection limit, which included the total coliform and fecal streptococcus data sets.

If more than 50 percent of the data are censored, the nonparametric tests based on ranks have less power to detect differences in central tendencies (Helsel and Hirsch, 1992, p. 367). In these cases, the response variable, bacteria, was converted to a categorical variable (detect/nondetect). The Kruskal-Wallis test was then used to test for a shift in the distribution of detects and nondetects instead of testing for differences in the medians of continuous data. (Helsel and Hirsch, 1992, p. 382). If significant differences existed for the categories, MSKW tests were performed on the subcategories to determine which subcategories differed, again with the overall alpha value being set equal to 0.05. This test was used for the fecal coliform and E. coli data sets. Categories that were not significantly different were assigned a common letter code (tables 4-7) Categories could be assigned more than one letter.

Answers to many common questions can be found on our
Frequently Asked Questions (FAQ) page. Please direct content comments
or questions to the PA District Information Specialist or contact:

District Chief
U.S. Geological Survey
840 Market Street
Lemoyne, PA 17043-1584

Phone: (717) 730-6912
FAX: (717) 730-6997
Email: dc_pa@usgs.gov

Please direct web related comments to webmaster@pa.water.usgs.gov

METHODS OF STUDY

Microbiology

Quality Assurance

Statistics