Scientific Investigations Report 2006–5290
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2006–5290
Back to Table of Contents
This section describes the establishment of the NAWQA study networks and the procedures used to select existing wells or construct new wells that were sampled for chemical and microbial quality. Procedures for the collection and handling of the ground-water samples, as well as the different microbiological analytical techniques used during the first and second cycles of the NAWQA program are described. Other parts of the section discuss quality assurance and quality-control procedures used for microbial-sample analyses and the statistics that were applied in the summaries of the microbiological data.
The selection or installation of wells for study networks followed NAWQA protocols that guide choosing site locations relative to land use, describe the suitability of existing wells with respect to plumbing and type of pump, and detail the construction of new monitoring wells regarding materials and drilling methods to meet the program’s study objectives (Scott, 1990; Koterba and others, 1995; Lapham and others, 1995). Selection of wells for the SWQA networks followed NAWQA guidelines described by Gregory Delzer (U.S. Geological Survey, written commun., June 2002, rev. March 2003).
The water from these wells is used for various purposes, including irrigation, stock watering, and recreation, as well as domestic and public drinking-water supply. Water use is designated as “unused” or “test” for wells used only for monitoring purposes such as water-level or water-quality measurements.
Within 27 study units, samples were collected from 1,244 wells in 16 principal aquifers of the United States. Through combinations of well networks (MAS, LUS, SWQA), most study units sampled more than 20 wells; however, the numbers sampled ranged from 2 wells in the Red River of the North (REDN) to 155 in the Lower Susquehanna (LSUS) (table 2). From the 1,244 wells, 1,784 samples (including replicate samples for quality control) were analyzed for concentrations of fecal-indicator bacteria and the presence or absence of coliphages. Field measurements of water temperature, pH, specific conductance, and concentrations of dissolved oxygen were taken at the time of sample collection. Water samples were analyzed for a broad range of constituents including nutrients, major ions, trace elements, and pesticides.
For describing the microbial quality of the Nation’s ground-water resources, 15 wells in each study unit were considered to be the minimum for representing or describing the quality of the ground water of a region or a principal aquifer. Therefore, data for individual study units that sampled less than 15 wells total from all their networks were removed from this analysis. This criterion reduced the number of study units from 27 to 22 and the number of wells sampled from 1,244 to 1,205. The five study units removed from the data set were Apalachicola-Chattahoochee-Flint (ACFB), Albemarle-Pamlico (ALBE), Ozark Plateaus (OZRK), Red River of the North (REDN), and South Platte (SPLT) (table 2).
The collection, handling and processing, and laboratory analyses of the samples, and field measurements made during their collection, followed USGS protocols and the guidelines of the NAWQA program during the early years of Cycle I (Koterba and others, 1995). The USGS National Field Manual (NFM), which in the late 1990s consolidated numerous guidance documents, technical memoranda, and sampling protocols, became the primary guide for the collection of water-quality data in the USGS. Chapters A1-A6 of the NFM guide the selection and cleaning of equipment, the collection and processing of water samples, and methods of measurements performed in the field (U.S. Geological Survey, variously dated); chapter A7 specifically guides the collection and processing of water for biological indicators (Myers and Wilde, 2003).
All equipment used in the collection and processing of water samples for microbiological analyses was thoroughly cleaned with laboratory-grade detergent followed by several rinses with deionized or reagent-grade water. Following the cleaning, all equipment for microbiological sampling and processing was sterilized before use by means of steam heat and pressure (autoclaving) or chemical disinfection (bleach or alcohol). Exceptions to autoclaving or chemical disinfection might apply to the submersible pumps used to collect water from monitoring wells (wells without an in-place pump). The cleaning of submersible pumps and sample tubing along with recommendations for additional quality-control samples are documented in chapter A7 of the NFM (U.S. Geological Survey, variously dated). Depending on the study network, wells selected for study were existing wells with in-place pumps and in use for domestic or public supply, or for purposes such as irrigation and livestock watering; or unused wells with or without an in-place pump. These different types of wells and conditions call for slightly different techniques for obtaining water to test for chemical or microbiological constituents.
Water samples were collected at a point as close as possible to the well head, thereby avoiding the effects of distribution lines, pressure tanks, filtration, or other treatment. Thus, the microbiological and other water-quality data reported here represent analyses of raw, untreated water only. None of these data resulted from analyses of treated finished water from distribution systems or at the point of use. As the water was withdrawn from the wells, temperature, pH, specific conductance, and concentrations of dissolved oxygen were monitored until stable values of these properties indicated that aquifer water had replaced “old,” standing water within the well casing. On reaching stable conditions, final readings of the field measurements were recorded and sterile sample-collection bottles were filled for subsequent microbiological analyses.
Water samples for analyses of most fecal-indicator bacteria were processed within 6 hours of collection. Samples for the analyses of coliphage or C. perfringens were shipped to the USGS Ohio Water Microbiological Laboratory (OWML) within 24 hours of collection or on the same day of collection to meet the laboratory’s hold time criterion of 48 hours to the start of analysis (Bushon and others, 2003).
The samples were analyzed for fecal-indicator bacteria using membrane-filtration methods, which involve filtering known volumes of sample water through a membrane filter, placing the filter on a Petri dish of growth medium, and incubating the dish for a specified length of time and temperature. After incubation, colonies of target organisms on the filter were counted and converted to a concentration expressed in terms of numbers of organisms per 100 mL of sample. During the first cycle of the NAWQA program, long-established methods commonly employed were for the total-coliform group on mENDO medium (method 9222B, American Public Health Association, 1992); the fecal-coliform group on mFC medium (method 9222D, American Public Health Association, 1992); E. coli on nutrient agar with 4-methylumbelliferyl-β-D-glucuronide (NA-MUG) medium (method 9222G, American Public Health Association, 1992) or on mTEC medium (method 1103.1, U.S. Environmental Protection Agency, 2002a); and the fecal streptococcus group on mKF medium (American Public Health Association, 1992; Britton and Greeson, 1988). During Cycle I, investigators in some study units also collected samples for analysis of concentrations of enterococci bacteria on mE/EIA media (method 1106.1, U.S. Environmental Protection Agency, 2002b) and C. perfringens, which were determined at the OWML using membrane-filtration on mCP substrate as described in the Information Collection Rule microbial laboratory manual (U.S. Environmental Protection Agency, 1996).
As approved methods using new growth media and culturing techniques became available during the Cycle II phase of studies, the NAWQA program adopted those methods for analysis of fecal-indicator bacteria and viruses in ground water. The new methods included use of the single-growth medium for simultaneous recovery of the total coliforms and E. coli on MI medium (method 1604, U.S. Environmental Protection Agency, 2002c) in the field. At the OWML, updated methods included testing for the presence or absence of somatic and male-specific coliphages by the 2-step enrichment procedure (method 1601, U.S. Environmental Protection Agency, 2001b).
The fecal-streptococcus, enterococci, and C. perfringens data were included as part of the inventory of data (fig. 2), but they are not part of the discussion of the quality of water in principal aquifers or drinking-water supplies because of the small number of tests. In the final data set that included replicate quality-control samples for the 22 study units, 1,435 analytical results for the total-coliform group, 1,338 for E. coli, and 836 for coliphage were available for assessing the occurrence of fecal-indicator bacteria and virus in ground water of different principal aquifers (fig. 2). The water from these aquifers was used for various purposes.
Microbiological data collected for the NAWQA program and ancillary data including chemical analyses and well characteristics can be accessed from the USGS National Water Information System (NWIS) data base (http://waterdata.usgs.gov/nwis), the Data Warehouse (http://infotrek.er.usgs.gov/traverse/f?p=NAWQA:HOME:9108424999420775073), or individual study units (http://water.usgs.gov/nawqa/).
The raw data were simplified and consolidated to create, to the extent possible, a consistent and comprehensive final data set by study unit, by sampling network, and by principal aquifer (table 3). For example, water samples collected from wells in the Snake River aquifer in the Upper Snake study unit (USNK) were analyzed for only the fecal-coliform group. To include discussion of the Snake River aquifer in conjunction with other principal aquifers, the results for USNK, with all but one sample testing negative for fecal coliforms, were compared to E. coli results on the basis that the predominant member of the fecal coliform group is the species E. coli (U.S. Environmental Protection Agency, 1978). Some actions involved assigning a well to a principal aquifer or to a dominant lithology. These actions were based on several sources, including ancillary information available in the USGS data bases, the geographic location of the wells and study units overlaid on the national maps of principal and major aquifers, and the primary lithologies of the principal aquifers as assigned by the NAWQA Principal Aquifer regional assessment teams (Wayne W. Lapham, U.S. Geological Survey, written commun., 2005).
Another consolidation of data involved reducing analyses of multiple samples collected at a well to a single analytical result so that statistical summaries for all study units and principal aquifers were equally weighted on the basis of a single value (table 3). These reductions involved data collected in the High Plains Ground Water (HPGW), Kanawha (KANA), Potomac-Delmarva (PODL), Puget Sound Basin (PUGT), Upper Colorado (UCOL), USNK, and lower and upper Tennessee (TENN) study units. One approach to data consolidation was to select data for a group of wells, such as a study network, collected during a particular period, and reject samples for the same wells collected outside that period. The selected period was chosen to maximize numbers of wells and reported results for the study unit or principal aquifer. Another approach was to select the data that represented the first visit and sampling of a well in a particular network. The first visits to wells often equated to the greatest number of wells and samples for a study network. In addition, for discussing the quality of ground water by network, water use, or principal aquifer, only one result from the pair of samples was used. Therefore, if one of the pair tested negative and one tested positive for the target organism, the entire well was considered to test positive for that organism.
Because of quality-control issues, some analytical results for the coliform bacteria and coliphage were excluded from the final data set. No coliphage samples collected from wells in the KANA, Lake Erie (LERI), Santee (SANT), and UCOL study units during 1997–98 and analyzed by quantitative methods were used in this report.
Prior to water-year 2003, quality-assurance (QA) procedures and the collection or preparation of quality-control (QC) samples generally followed the guidelines presented in Britton and Greeson (1988) and Myers and Sylvester (1997). Those guidelines suggest that at a minimum, QC samples should include processing filter blanks—one blank at the beginning of filtration (an equipment blank) and one blank at the end of filtration (a procedure blank), bracketing the processing of a water sample. The guidelines also specified the use of heat-sensitive indicator tape on all autoclaved sample-collection and filtration equipment for monitoring the effectiveness of the autoclave and ensuring sterilization of equipment.
In water-year 2003, when the NAWQA program incorporated microbiology as part of routine sampling schedules for Cycle II, additional guidelines that list the types of QA procedures and numbers of QC samples appropriate for the NAWQA program were put in place (http://oh.water.usgs.gov/micro/nawqa.html). For fecal-indicator bacteria in ground water, the guidelines specify a pair of samples be collected at each well and a set of filter blanks be processed to accompany every pair of water samples. The filter blanks document sterility of the filtration equipment, lack of carryover of cells between filtrations, and whether cells are lost to the walls of the filtration equipment due to incomplete rinsing. The guidelines also specify processing of field blanks and of positive and negative control samples at frequencies depending on numbers of wells sampled. Processing of field blanks involves passing sterile water through all parts of collection equipment, from the pump (if applicable) through the tubing and into the sample bottles. Positive and negative control samples are pure cultures of target organisms that ensure the quality of media and reagents for use. The control samples are processed by the membrane filtration method and the results are evaluated for proper recovery and growth of the positive-control organism and lack of growth by the negative-control organism. In addition to the QC samples for the fecal-indicator bacteria, field blanks are prepared for coliphage analysis, and replicate samples are collected and spiked with target organisms at the laboratory to determine if there are matrix interferences in the analysis.
At the time of this data compilation, selected QC data were obtained from nine Cycle II study units for evaluating the set of microbiological data for the purposes of this report. QC data were not requested from the Cycle I study units because these data do not exist in electronic data bases and so are not easily constructed from archived files. For the Cycle II data, target organisms were detected in only 13 of 379 filter blanks (3.4 percent). For the individual study units reporting filter-blank data, detection rates ranged from 1 to 12 percent of the blanks prepared. The QC guidelines state that if target colonies in the filter blank are less than 5 percent of the colony count of the environmental sample, the results for the sample are accepted. If target colonies on the filter blank are greater than 20 percent of the target colonies of the sample, the results are rejected. For target colonies that amount to 5 to 20 percent of a sample’s colony count, the concentration value in NWIS is assigned a remark code of V (value affected by contamination), a value-qualifier code of W (high variability), and a result-level comment indicating the number of target colonies in the filter blank. Evaluation of the filter blank data resulted in rejecting five sample results (including any replicates) for the total coliform group. During their well-network sampling, five Cycle II study units processed field blanks, all of which were negative for target organisms. Analyses of these filter- and field-blank samples indicated that equipment sterilization, sampling procedures, and sample handling and filtration were effectively controlled and thus resulted in a set of high-quality data for describing the quality of the ground-water resources.
NAWQA staff in eight of the Cycle II study units consistently collected and processed duplicate samples for each well. The relative percent differences in the analytical results for 402 pairs of samples ranged from 0 to 200 percent. The average and median relative percent differences were 13 and 0, respectively. Most pairs (338) had the same value, which results in percent differences of zero.
Graphical analyses of the data set through the use of box plots, notched box plots, scatter plots, and bar charts were used to evaluate the occurrence, distribution, and potential relations of the fecal-indicator bacteria in ground water with other constituents or with well characteristics. Box plots display summaries of the distribution of a group of data by showing the center location (median), the variation through the height of the box (interquartile range of values), the skewness through the relative size of the box halves, and extreme values or outliers of the data (Helsel and Hirsch, 1992). Notched box plots portray the same information and also show the 95-percent confidence interval about the median. Boxes with nonoverlapping notches indicate significant differences between the groups of observations (McGill and others, 1978). Scatter plots can indicate patterns in a group of data, such as correlation or trend relations. Bar charts are mostly used in this report to display numbers of observations or frequencies of bacterial detections.
Nonparametric statistical testing by the Kruskal-Wallis test and Wilcoxon rank-sum test was used to examine the data for significant differences between groupings of coliform bacteria concentrations or detections within study networks, classes of water use, principal aquifers, or types of lithology. The null hypothesis of the Kruskal-Wallis test states that the rank of the median (or mean) of the data for each group is identical, and the alternative hypothesis is that at least one group differs in its distribution, which tends to produce observations (concentrations) different in value (Helsel and Hirsch, 1992). The Kruskal-Wallis test does not determine which group or groups are different. For this, the Wilcoxon rank-sum test was used with significance levels adjusted for the total number of tests performed within each group of data, or notched box plots helped to identify statistically different groups within the groupings of interest.
For statistical testing of concentration data, values reported as < (less than) 1 CFU/100mL (colony forming unit per 100 milliliters) in the USGS NWIS database were treated as zeroes. This assumption was because no organisms were found on the filter and media; therefore the count for that test was zero organisms. Statistical tests were also performed with coliform bacteria data expressed as detections (with positive results set to equal 1) or nondetections (with negative results set to equal 0), and with coliphage data, which are coded in NWIS as a 1 (for present in the sample) or a 2 (for absent in the sample).
Back to Table of Contents