Attributes and Application of Indicators
Microbial water quality indicators are used in a variety of ways within public health risk assessment frameworks, including assessment of potential hazard, exposure assessment, contaminant source identification, and evaluating effectiveness of risk reduction actions. The most desirable indicator attributes, and therefore the most appropriate indicators, naturally depend on their manner of use. This chapter describes desirable attributes of an indicator, typical applications of indicators, indicator attributes that are appropriate for such applications, and provides an assessment of whether current indicators and indicator approaches are meeting the needs of each application. The chapter ends with a summary of its conclusions and recommendations.
For almost 40 years, Bonde’s (1966) attributes of an ideal indicator have served as an effective model of how a fecal contamination index for public health risk and treatment efficiency should function (Box 4-1). Three of Bonde’s attributes (1, 2, and 4) address the relationship between indictor organisms and pathogens of concern, while the remaining five describe desirable properties associated with quantifying the indicator. However, Bonde’s attributes of an ideal indicator must be refined to continue their relevance to public health protection because the development and increasing availability of new measurement methods necessitates the separation of criteria for evaluating indicators and detection
An ideal indicator should
methods. Historic definitions of microbial indicators, such as coliforms, have been tied to the methods used to measure them. Newly available methods (particularly molecular methods; see Chapter 5 and Appendix C) allow more specificity in the taxonomic grouping of microorganisms that are measured. More importantly, a variety of new methods are becoming increasingly available, providing several options for measuring each indicator group. Thus, separate criteria allow one to choose the indicator with the most desirable biological attributes for a given application and then match this with a measurement method that best meets the need of the application. Box 4-2 lists desirable biological attributes of indicators and Box 4-3 lists desirable attributes of methods.
The most important biological attribute is a strong quantitative relationship between indicator concentration and the degree of public health risk. This relationship has been demonstrated primarily through epidemiologic studies for recreational exposures (Cabelli et al., 1979; Cheung et al., 1990; Seyfried et al., 1985a,b; Zmirou et al., 1987). An alternative means of demonstrating the relationship to health risk is through correlation between prospective indicator concentration and pathogen levels (Gerba et al., 1979; Labelle et al., 1980; Lipp et
al., 2001a; Robertson, 1984; Seyfried et al., 1984). The latter approach is used less frequently because assays for pathogens are specific to individual agents or classes of agents (e.g., enteroviruses) and correlation with a single pathogen, or subset of pathogens, does not establish a relationship with all illness-causing agents or their risks to human health (their health effects).
The next two desirable biological attributes are similarity in survival and transport characteristics of the indicator to those of the pathogen(s) of interest. If there is differential transport or survival, the relationship between pathogen and indicator concentrations will change at varying distances from the source and over different times in the environment, making it difficult to select a critical indicator concentration on which to make public health decisions (Griffin et al., 2001). For example, differences in viral and bacterial transport through soils and aquifers have been found to affect assessment of water quality impacts from septic systems (Harden et al., 2003). If there is differential survival, it is generally preferable that the indicator be more resilient than the pathogens so as to be protective of public health. However, exceptionally long survival of potential indicators, such as spore-forming Clostridium perfringens, may render them too over-protective or nondiscriminatory because they may be present at concentrations mistakenly considered to be indicative of a health risk long after the pathogens have declined to levels not considered a risk.
The next desirable attribute is that the indicator be present at densities that are detectable with an easily sampled volume. It is always possible to measure lower concentrations of indicators through use of high-volume collection strategies, but it is typically preferable for indicators to be present at high enough density to be detected easily in sample volumes that are convenient to collect and transport to a laboratory for analysis. Pathogens are excreted by infected individuals in numbers per gram of feces are comparable to that of coliforms (Gerba, 2001). However, domestic wastewater contains a mixture of excreta from a variety of people, many of whom are not infected with a pathogen but excrete coliforms and other microbial indicators. Thus, the indicators are present in wastewater at densities several thousand times higher than that of most pathogens, including enteric viruses and protozoa (Feachem et al., 1983; Rose et al., 2001).
The final desirable biological attribute is source-specificity. Indicators that are specific to animal digestive systems are preferable to those that occur naturally in the ambient environment, because the dichotomy of sources may lead to different risk potential depending on the nature of the source. A similar, though lesser, concern exists when the indicator occurs in the gut flora of numerous animal species, because of the difference in pathogen types and concentrations excreted among species. Some indicator microorganisms, while not source specific, have genotypic or phenotypic properties that allow distinction as to whether the fecal source is human or animal (Simpson et al., 2002). Other indicators even allow for identification of particular animal species contributing to the fecal contamination, which can be used to indicate the degree or type of risk. For example,
the proximity of cattle to a water source could indicate a concern regarding Cryptosporidium and Escherichia coli O157:H7 because these pathogens are common in cattle (LeChevallier et al., 1999a,b).
Attributes of Methods
The attributes of a method that should be considered are not independent of one another, and these relationships are described in the following text. One of the most important method attributes is specificity, or ability to measure the target indicator organism in an unbiased manner. Specificity may be directed at microorganism groups (e.g., coliforms, cultivatable enteroviruses), genera (e.g., Giardia), species (e.g., Cryptosporidium parvum), or subtypes (e.g., E. coli O157:H7). Specificity can also be described on a biochemical, antigenic, or genetic basis.
In most cases, the specificity concern is for false positives, in which a confounding organism reacts similarly in the test and yields incorrectly high results. Among newer methods, Pisciotta et al. (2002) suggest that coliform measurements can be confounded with Vibrio cholerae counts in subtropical environments when using chromogenic substrate techniques. However, there are cases in which false negatives are of concern, such as when high levels of heterotrophic plate count microorganisms may, in some instances, interfere with the detection of coliforms (Allen, 1977; Edberg and Smith, 1989).
Lack of specificity can also be introduced from matrix interferences. Many waters that are tested for microbiological quality are saline, or turbid, or have a high organic content, all of which have the potential to interfere with some indicator measurement methods (Geldreich, 1978). For example, tannic and humic acids from decaying plant material can interfere with some molecular methods. Filtration methods are particularly susceptible to high suspended solid load, which can cause clogging or clumping. Low levels of residual chlorine can produce sublethal injury to coliforms, interfering with their enumeration on highly differential media (Camper and McFeters, 1979; McFeters et al., 1986), although this will be of greater concern in treated water monitoring systems. It is also desirable for a method to have broad applicability to a number of geographic locations (tropical waters versus temperate waters), various types of watersheds (e.g., point source and nonpoint source inputs), and different water matrices.
Preferred methods will also measure indicator concentrations precisely, which is particularly important when decisions must be made on a limited number of samples. Method precision includes not only repeatability with a laboratory, but variability across laboratories. Generally, greater precision is better, but in particular the precision must meet the needs for the decision-making process. Multiple tube fermentation, which has been one of the most frequently used indicator methods, is based on a statistical approach to estimating concentrations and has a coefficient of variation equal to more than half the mean (Noble et al.,
2003a), yet interlaboratory variability has been found to be acceptable for most applications.
Sensitivity is the lower limit of detection of an indicator in a certain sample volume and has implications for precision. The needed sensitivity may be risk based, technology based or management based. Methods that amplify or concentrate the target are typically more sensitive (e.g., culture, polymerase chain reaction [PCR], filtration). Methods may be quite amenable to changes in the sensitivity (e.g., membrane filtration and fecal coliform cultivation) but at some point they become technologically limited (e.g., via clogging of the filter and masking of the bacteria). Sensitivity is also affected by the sample volume, particularly if the target indicator concentration is low relative to the volume analyzed and detection is reduced to a “Poissonian sampling” process (see Chapter 5 and Figure 5-5 for further information). Although sensitivity concerns can be overcome by processing larger sample volumes, this can affect logistical feasibility in some applications.
It may not be necessary in all cases to be quantifiable. In some applications, presence/absence information may suffice, particularly since counting can be tedious, adds expense, and typically increases the time of the assay. However, quantification increases precision and is necessary in most applications associated with assessing public health risk.
The speed of the method is an important characteristic, particularly when warning systems (discussed later) are involved and human exposure continues to occur during the laboratory analysis period. Methods vary widely in their speed; with faster molecular methods soon becoming available to replace traditional culture-based methods (see Chapter 5 and Appendix C for further discussion). Culture-based methods often take several days to complete, whereas molecular methods take hours or less. However, hybrid approaches employing brief culture periods (to ensure the culturability or infectivity of the microbe) coupled with rapid molecular detection have the potential to rapidly detect and quantify culturable microbes in environmental samples. This has been particularly useful in decreasing the time for virus detection in cell culture (Reynolds et al., 1996).
Many indicator methods that are able to produce results rapidly do so by measuring molecular properties that do not address viability or infectivity. Thus, high indicator counts may be recorded in areas where chemical or physical agents have been effective at inactivating pathogens. Viability or infectivity is an important issue because the epidemiologic studies on which current standards are based have all been conducted with culture-based methods, and it is not clear how well those epidemiologic relationships will hold if nonviable indicators are included in the counts.
Logistical feasibility will often govern the indicator method of choice. Cost concerns can be important when large numbers of samples are needed for screening purposes, but they may be less important when the consequences to be addressed have major impacts on human health risk, such as the risk of an outbreak
or a high burden of disease related to the exposure. Costs include not just labor and materials, but also capital and training costs. Many of the new measurement technologies require large initial investments because the equipment and personnel necessary to implement them are not already in place. Moreover, simpler methods with proven field utility and small volume requirements may be preferred when applications are most appropriately implemented on-site using typically less well-trained personnel, such as lifeguards.
Finally, although not considered a method attribute per se, all methods are amenable to some form of ad hoc or “official” standardization (see Chapter 5 for a full discussion of the importance of and approaches for standardizing and validating microbiological methods) over time and with increasing implementation.
Measurement-based Warning Systems
One of the most frequent applications of indicators is in public health warning systems. Warning systems include measurement of indicators to assess whether there is a likelihood that pathogenic microorganisms are present at unacceptable risk levels. Warning systems may be related to ingestion of treated drinking water, recreational water contact, or shellfish consumption. Risk levels are codified through enforceable standards, which may be based on a single sample maximum level, an average or median concentration for a specified period of time, or a maximum frequency of samples over a threshold. When a standard is exceeded, actions are taken to reduce exposure, such as increased treatment levels for drinking water, shellfish bed closures, or warnings to avoid recreational water contact. Because drinking water warning systems focus on treatment effectiveness, which is largely outside the scope of this study, this section focuses on the recreational contact warning system. Box 4-4 provides some comparisons and contrasts between recreational and drinking water warning systems.
For recreational bathing waters, the U.S. Environmental Protection Agency (EPA) recommends the use of enterococci in marine water and E. coli in freshwater, based on epidemiologic evidence (see Chapter 2; EPA, 1986). Many states follow EPA’s recommendation for freshwater, although there are considerable differences among standards for marine water (see Table 1-4), with several states still using fecal coliforms and more having no standards at all. California uses a multiple-indicator approach including enterococci, fecal coliforms, and total coliforms (see Box 4-5). Hawaii augments enterococci with the use of Clostridium perfringens, primarily because of the problem of regrowth associated with coliform bacteria in tropical environments (Fujioka, 2001). Although EPA (1986) also recommends action limits for each of these indicators, there remain considerable differences in standards among states (see Table 4-1), leading to differential levels of public health protection. The goal of the Beaches Environmental
Drinking water warning systems typically focus on treatment adequacy and integrity of the distribution system, rather than on source water quality. They differ from recreational water contact systems in three primary ways:
Assessment and Coastal Health (BEACH) Act of 2000 was to bring consistency to beach assessments; however, differences between the states continue based on the various approaches for setting standards and their use in closing impaired beaches.
Several factors limit the effectiveness of current recreational water warning systems, the most prominent of which is the delay in warnings caused by long laboratory sample processing time. Current laboratory measurement methods used to enumerate indicator bacteria (multiple tube fermentation, membrane filtration, and chromogenic substrate) require an 18- to 96-hour incubation period. By the time results are obtained, exposure has already occurred for a day or more. This inadequacy in the notification system is exacerbated because most contamination events are intermittent and indicator levels typically return below thresholds within 24 hours (Boehm et al., 2002; Leecaster and Weisberg, 2001). Thus, contaminated beaches remain open during the laboratory incubation period, but often return to acceptable levels by the time laboratory results are available and warning signs are posted.
The State of California has the most rigorous beach water quality monitoring requirements and standards in the country. Regulations implemented in response to a 1998 state law (AB411) require that three indicator species (enterococci, fecal coliforms, and total coliforms) be measured at least weekly at beaches with more than 50,000 annual visitors. State regulations also define daily and monthly average standards for each indicator, as well as a daily standard for the ratio of total to fecal coliforms. These thresholds were established based on a California-specific epidemiologic study (Haile et al., 1999), and the law requires that public warning signs be posted whenever any of the thresholds are exceeded. Implementation of AB411 requirements resulted in an eightfold increase in the number of public warnings issued. Most of the increase was due to inclusion of an enterococci standard that did not previously exist in California. More than 90 percent of the public warnings are associated with enterococci violations, which are several times higher than warnings associated with either of the other indicators (Noble et al., 2003b).
TABLE 4-1 Range of Bacterial Standards Values Used Among Statesa
Another shortcoming is the poorly established relationship between presently used indicators and health risk. Recent reviews of beachgoer epidemiology studies (Prüss, 1998; Wade et al., 2003) found that enterococci had the best relationship to health risk among presently used indicators for marine water, but less than half of the studies found a significant health relationship and the dose-response curves establishing the relationship between increased illness and indicator density were highly variable. This inconsistency among epidemiologic study results may be due to geographic variability and differences in the sources of
The use of indicators is based on the presumption that they co-occur at a constant ratio with illness-causing pathogens. This premise is flawed because indicator levels in the gastrointestinal tract may vary within a narrow range, but pathogen concentration is highly variable and dependent on which pathogens are in the population at what levels at specific times. Furthermore, upon leaving the intestinal tract, microbial indicators and pathogens degrade at different rates that are mediated by factors such as their resistance to aerobic conditions, ultraviolet radiation, temperature changes, and salinity. As a result, the epidemiological relationship between indicator density and illness patterns can differ depending on the age of the source material, as well as local meteorological and other environmental conditions. Several studies also have found that some indicator bacteria can grow outside the human or animal intestinal system (Desmarais et al., 2002; Fujioka, 2001; Hardina and Fujioka, 1991; Solo-Gabrielle et al., 2000; see also Chapter 3), further confounding the correlation between pathogens and indicators.
The underlying epidemiologic studies are also limited because many reported failures of beach water quality standards are associated with nonpoint source contamination (Lipp et al., 2001a; Noble et al., 2000; Schiff et al., 2003), but the epidemiologic studies used to establish recreational bathing water standards (EPA, 1986) have been based primarily on exposure to human fecal-dominated point source contamination (Haile et al., 1999). Since nonpoint sources generally have a higher percentage of animal fecal contributions, and animals shed bacterial indicators without some of the accompanying human pathogens, there is considerable uncertainty in extrapolating present standards to nonpoint source situations. A poor correlation between bacterial indicators and virus concentrations has been found in the study of nonpoint sources and water quality (Jiang et al., 2001; Noble and Fuhrman, 2001). However, when a human source, such as septic systems, has been present, enterococci have been significantly correlated with viruses (Lipp et al., 2001a).
A major problem with present water contact warning systems is that bacterial indicator concentrations are spatially and temporally variable and most sampling is too infrequent to transcend this granularity.1 Taggart (2002) found that sequential samples collected at the same location typically varied by a factor of two and samples 100 meters apart typically differed tenfold. Cheung et al. (1990) found
that indicator concentration at a site varied fifteenfold within a day, and Boehm et al. (2002) found that elevated indicator counts typically lasted less than two hours as water masses moved past their sampling site. Most beach monitoring occurs only weekly, and more than one-third of beaches nationally are monitored only monthly. Most of this monitoring is based on collection of a single water sample, the interpretation of which is further compromised by measurement variability. For multiple-tube fermentation, laboratory measurement error based on the 95 percent confidence interval exceeds 50 percent of the mean; more than half of the beach warnings issued in Los Angeles are within measurement error of the standard (Noble et al., 2003a). A guidance document to address these issues is needed from EPA.
Granularity and measurement error concerns are exacerbated by the all-ornone paradigm that is pervasive for beach warnings. Most water quality managers choose from only two options in response to high bacterial indicator counts: (1) close a beach because of a perceived health risk or (2) do nothing. Beach closures are usually reserved for sewage spills, with indicator measurements used primarily to help identify the likelihood that a spill has occurred. No action is typically taken based on indicator measurements alone, particularly when high counts are intermittent. Thus, efforts to inform and protect the public are supported only partially through the current use of indicator measurements.
Some locales are beginning to change this dual-action paradigm by adding additional management options. For instance, California now issues beach advisories when a sample exceeds state bacterial indicator standards and there is no apparent evidence of a sewage spill. Advisories differ from closures in that swimmers are not required to exit the water. California’s approach, though, is limited because it requires advisories based on comparison to a single-sample bacterial standard. Temporal and spatial granularity of bacterial counts, combined with the day or longer laboratory processing time, leads to frequent misinformation when warnings are based on a single sample.
Several environmental advocacy groups, such as Heal the Bay (http://www.healthebay.org) and the National Resources Defense Council (http://www.nrdc.org/water/oceans/ttw/titinx.asp), are also beginning to transcend the all-or-none and single-sample difficulties by using the magnitude and frequency of standard failures to develop “letter grades” to describe water quality of recreational beaches. Letter grades have been used successfully in some parts of the country to provide the public with information about the health quality of restaurants and are readily understandable to the public. Such grades can effectively address the granularity issue by integrating data over a longer time period but to be effective they require more frequent monitoring than the monthly sampling that is conducted in many parts of the country.
When a public health risk or water quality impairment is identified through measurement-based systems, the next step is often to conduct investigations to identify the source of contamination. There are two primary purposes of source identification. The first is to decide whether a health warning should be issued because a recreational water body closure is typically issued only after determining that a human fecal source is associated with the high bacterial indicator levels. The second is to identify the most promising approach for fixing the problem. For example, should a local agency be looking for a leaking sewage pipe or for a flock of birds as the source of the problem? From a regulatory point of view, source tracking also feeds directly into the total maximum daily load (TMDL; see also Chapter 1) requirement of the Clean Water Act (CWA) for problem characterization in impaired waters.
Four basic approaches have been used for source identification of microbial contamination, commonly referred to as microbial source tracking. The first involves spatially intensive sampling to identify the source through gradients in indicator density. The second enhances typical indicator measurement with genotypic/phenotypic examination, based on the presumption that certain strains of microorganisms have coevolved with their host and demonstrate specificity to that type of animal. The third is direct measurement of alternative indicators or pathogens that are more closely linked to human intestinal tracts than bacteria such as E. coli and enterococci. The fourth is measurement of chemical compounds that are specific to human waste streams, such as caffeine and coprostanol. Some of these methods are only able to discriminate between human and nonhuman sources, while others are able to distinguish specific animal sources contributing to fecal pollution (e.g., dogs, cattle, birds). The following sections describe and evaluate the source tracking methodologies that are currently being used.
Intensive Sampling Approaches
The most frequent practice when routine monitoring identifies a persistent bacteriological water quality problem of unknown origin is to conduct spatially and temporally intensive sampling with standard bacterial indicators, along with efforts to visually identify waste sources. These types of “sanitary surveys” are often preferred because they can be performed using existing equipment and manpower. Although an extensive discussion of sanitary surveys and criteria for their use is beyond the scope of this report, the committee believes these surveys are generally effective when the problem is a leaking pipe in which concentrations are linearly related to location. This approach has been used successfully to identify sewage and manure leaks (Burkholder et al., 1997; Mallin et al., 1997), but it is less effective when the problem is a nonpoint source. Intensive surveys are
mostly limited by the long time necessary for laboratory analysis, requiring them to proceed multidirectionally, rather than following a contamination trail in a single direction based on differential concentrations. This makes them expensive and highly impractical when there are multiple tributaries in a system.
Phenotypic and Genotypic Indicator Approaches
Phenotypic and genotypic methods are being used increasingly to track fecal contamination sources (Scott et al., 2003; Simpson et al., 2002). These approaches are based on the presumption that fecal bacteria in the intestines of animals coevolve to become host specific. Over time, this process produces both identifiable phenotypic traits and changes in gene sequences. Phenotypic and genotypic methods exploit these changes to link fecal bacteria with hosts. This coevolution may be enhanced by differences in food sources among animals. In domesticated animals and humans, it may also be enhanced by the introduction of different types of antibiotics among species.
Both phenotypic and genotypic methods can be divided further into those that require a library of bacterial isolates of known origin and those that do not (Table 4-2). A “library-dependent” method requires cataloging a large number of phenotypic or genotypic patterns from fecal bacteria of known origin. Source identification is then achieved through matching the patterns produced by bacteria obtained in ambient water samples to those in the database.
The most frequently used phenotypic method is multiple antibiotic resistance profiling (MAR; Cooke, 1976; Hagedorn et al., 1999; Harwood et al., 2000; Parveen et al., 1997). This method is based on growing isolates on selective media containing a suite of antibiotics of varying types and concentrations. The resulting resistance “fingerprints” comprise a library. Resistance patterns of indicator bacteria isolated from ambient samples are then matched against the library to determine probable sources. Carbon source utilization (CSU) is similar to MAR except that the fingerprints are based on differential growth in various carbon source growth media (Hagedorn et al., 2003). F+ RNA coliphages can also be used in this way because they belong to four groups or serotypes that differ in
TABLE 4-2 Classification of Genotypic and Phenotypic Methods
Repetitive intergenic DNA sequences (rep-PCR)
Pulsed-field gel electrophoresis (PFGE)
Host-specific molecular markers (PCR)
Enterotoxin biomarkers (PCR)
Terminal restriction fragment length polymorphism (t-RFLP) analysis of total bacterial community
Multiple antibiotic resistance Carbon source profiling
F+ RNA coliphage serotyping
their occurrence in humans and animals. F+RNA coliphage isolates can be typed by a simple serological test based on prevention of phage replication (absence of host cell lysis, called “neutralization”) in the presence of its specific antiserum (Hsu et al., 1995). F+ RNA coliphage grouping also can be done by nucleic acid hybridization as an alternative genotypic method.
The phenotypic source identification methods have the advantage of rapid processing of multiple samples, relatively low cost per sample, and use of equipment already present in most microbiological laboratories. Recently, the detection of specific antibiotic resistance traits in enteric bacteria found in environmental media and fecal waste sources has been used to identify bacterial sources without the development of an extensive reference database of isolates. In these studies, enteric bacteria impacted by anthropogenic sources, such as human sewage or animal manure, were more likely to be antibiotic resistant than the same species of bacteria from ambient sources (Chee-Sanford et al., 2001). Mathew et al. (1998, 1999) also determined that resistance patterns differed between farm types and between pigs of differing ages.
Currently, the most widely used library-dependent genotypic methods include pulsed-field gel electrophoresis (PFGE), ribotyping, and repetitive-intergenic DNA sequence polymerase chain reaction (rep-PCR), although other methods such as denaturing gradient gel electrophoresis (DGGE) are also being investigated (Simpson et al., 2002). In the future, microarray technology should lend itself to the detection of a wide variety of gene sequences in support of microbial source tracking (see Appendix C for further information). All of these methods require characterization of some bacterial genetic sequence to create a reference library and are summarized below:
PFGE involves generating a DNA “fingerprint” by digesting the complete genomic DNA from a pure culture of bacteria using restriction endonucleases (enzymes that cut DNA at specific sequences of the genetic code), with the resulting banding pattern constituting the fingerprint. Assessments of PFGE’s effectiveness as a microbial source tracking tool are contradictory and limited (Parveen et al., 2001; Scott et al., 2002).
Ribotyping is similar to PFGE, except that only DNA fragments containing ribosomal genes are identified. Ribotyping has been used for bacterial pathogens (Salmonella, Vibrio), but source tracking applications to date have focused on E. coli (Carson et al., 2001; Parveen et al., 1999). This genomic method has been used mostly to distinguish between human and other animal sources, but it also has been used to discriminate between various animal hosts (Hartel et al., 2002; Scott et al., 2003). Although it is the most widely published of the genetic methods, no standard approach has yet been developed with regard to the numbers of isolates measured, the necessary library size, the most effective restriction enzyme, or even which bacterial indicator species has the greatest level of host specificity.
Rep-PCR amplifies specific repetitive elements distributed throughout the bacterial genes to produce a complex banding pattern. This method has been shown to differentiate between human and various animal sources, but only a few studies of its use have been published (Carson et al., 2003; Dombek et al., 2000).
Non-library-dependent genotypic source tracking methods currently in use include host-specific molecular markers (PCR), terminal restriction fragment length polymorphism (t-RFLP) analysis of total prokaryotic community, and enterotoxin biomarkers (PCR) and are summarized below:
Host-specific molecular markers focusing on t-RFLP to members of Bacteriodes-Prevotella and Bifidobacterium have been used successfully for source identification in a few cases (Bernhard and Field, 2000a,b; Bernhard et al., 2003). A related method, total prokaryotic community profiling using t-RFLP, has been used to differentiate between source waters at their ocean terminus (Patricia Holden, University of California at Santa Barbara, personal communication, 2003). This latter method is limited to local applications because a unique source sample is necessary for comparison with receiving waters.
Enterotoxin biomarkers using species-specific PCR primers to amplify toxin genes found in E. coli have been found to differentiate cow, human, and pig waste (Khatib et al., 2002), but geographic and temporal stability of this method is unknown.
Although many genotypic and phenotypic methods are promising, none have been thoroughly tested or are yet widely accepted in the regulatory community. Of greatest concern is the poor understanding of the temporal and geographic stability of traits and genetic sequences (Hartel et al., 2002; Jenkins et al., 2003). High variability in either parameter could restrict use of many methods to local venues due to the time and cost constraints inherent in constructing suitable libraries. It is also unclear which bacterial species show the greatest host specificity, and different bacterial species could show greater or lesser specificity in different hosts depending on the method employed. Finally, the lack of standardized protocols for performing these methods may preclude non-research-oriented laboratories from adopting them and inhibit sharing of libraries.
Direct Measurement Approach
Direct measurement of certain bacteria or viruses that are exclusive to human waste or are rarely found in animals is another approach that has been used to distinguish human from animal sources of fecal contamination (though see Box 4-6). The advantage of direct measurement methods is that there is no need for a large database to match a “fingerprint” against. Thus, these methods can be directly incorporated into microbiological monitoring schemes with the presence of
The relationship between fecal indicators and human enteric pathogens varies with the fecal contamination source (human, animal, domestic sewage from septic systems, municipal sewage from cities, combined animal feeding operation waste, etc.). Relatively few pathogens, such as hepatitis A virus, Shigella bacteria, and the amoebic protozoan Entamoeba histolytica, are unique to humans. Most enteric bacterial and protozoan parasites of humans are harbored by a variety of mammals. Many enteric viruses were once considered unique to humans, with other animals harboring taxonomically similar viruses that do not infect humans. However, evidence for the distinctiveness of human and nonhuman enteric viruses is beginning to decline as the likelihood of cross-species transmission between human and nonhuman strains of some enteric viruses is becoming apparent. A recent example is hepatitis E virus (HEV), which is found in humans, swine, rats, and other animals. The human and swine strains of HEV are very similar at the genetic and protein levels; the human virus has been found to infect swine and the swine virus to infect nonhuman primates (Emerson and Purcell, 2003). Given the current available data, the relative risks of exposure to human pathogens from either human or nonhuman animal sources of fecal contamination are difficult to quantify for any particular source of fecal contamination.
the microbe indicative of human waste input. A number of direct measurement indicators have been used for microbial source tracking as described below and summarized in Table 4-3:
Bifidobacterium spp. are obligate anaerobic bacteria. Mara and Oragui (1983) developed a human bifid sorbitol agar that can be used with membrane filtration (Clesceri et al., 1998). There is no indication that these bacteria reproduce in the environment; however, the survival is highly variable and methods for their recovery and detection in water and other environmental samples are inefficient.
Bacteroides fragilis bacteriophage is a virus that infects anaerobic bacteria found in the intestinal tract. Application of this phage has been used primarily in Europe, where it was determined that a bacterial host-specific strain, HSP40, was found in 10 percent of humans but not in animals (Tartera and Jofre, 1987). The phage does not replicate in the environment and has good stability, but it may be found in relatively low numbers and be undetectable in areas with large dilution (McLaughlin, 2001). Furthermore, some studies have determined that the
TABLE 4-3 Advantages and Disadvantages of Select Microorganisms Used for Microbial Source Tracking
Sorbitol-fermenters may be human specific
Low numbers present in environment
Variable survival rates
Culture methods not well developed
B. fragilis HSP40 bacteriophage
Possibly human specific
Not always present in sewage or present in low numbers
Not always human specific, depending on bacterial strain and geographic location
F+ RNA bacteriophage
Groups are correlated with sources (humans versus animals and livestock) with small % of overlap
Lower numbers in warm marine and tropical waters due to variable survival rates
Human enteric viruses
Human specific if cell cultures and genome targets to amplify are chosen carefully
Addresses hazard identification component of risk assessment paradigm
Low concentrations in water and other samples
Better cultivation and molecular methods needed
Specific indicator of grazing animal fecal sources
Detected by culture for a lengthy period (weeks)
SOURCE: Adapted from Scott et al., 2002.
host bacterium also detected B. fragilis phages from animals (swine) as well as from humans, leading to questions about the human specificity of the system (Chung, 1993).
F-specific coliphages are viruses that infect male (F+) strains of Escherichia coli bacteria that produce F pili. The F pili possess the receptor that allows for detection and distinction of the F+ groups of coliphages. There are two taxonomically distinct F-specific phage groups, one containing RNA (Leviviridae) and another containing DNA (Inoviridae). Members of the two groups can easily be separated and identified by determining if they are resistant (Inoviridae) or sensitive (Leviviridae) to RNase. The F+ RNA coliphages are morphologically similar to many human enteric viruses. By means of plaque isolation techniques, phages have long been used as virus indicators due to their similar transport and fate (Havelaar, 1993; IAWPRC, 1991; Kott et al., 1974). Serotyping or genotyping the F+ RNA coliphages into their four main groups can further distinguish the origin as human or animal (Hsu et al., 1995), although there is some overlap of the groups. However, Schaper et al. (2002) found that serotypes II and
III were associated with human sewage, but human samples also contained serotypes I and IV. Animal samples contained all four serotypes, with the majority of the F+ coliphage being serotypes I and IV. They found statistical significance in the assignment of serotypes to specific human or animal sources, but the distinction may not be as definitive as previously thought since there was some overlap between the serotypes and their expected animal sources.
Human enteric viruses are potentially definitive for human fecal contamination and have been monitored directly in water since the 1960s. Originally cell culture and serology were used to isolate and identify the viruses, but PCR and reverse transcription PCR (RT-PCR) methods are now more common (Jiang et al., 2001; Lee and Kim, 2002). The specificity of these methods to human enteric viruses depends on careful choice of the cell cultures for isolation and target genomic region for PCR or RT-PCR amplification.
Rhodococcus coprophilus is a fecally excreted bacterium associated with grazing animals, and its presence in water has been suggested as an animal-specific fecal indicator (Long, 2002; Mara and Oragui, 1985). Unfortunately, methods to culture this bacterium require long incubation periods, and molecular methods for its detection in water have not yet been developed.
Although several species are specific enough to human sources of fecal contamination that they are suitable candidates for direct measurement, two challenges remain with the direct measurement approach. The candidate species generally require specialized methods, such as molecular techniques, that are available only in research laboratories. Also, direct measurement approaches are capable only of defining the presence of human material, not the percentage of fecal contamination that human source material encompasses. Without the ability to distinguish whether human material comprises a small or large fraction of the fecal contamination, managers do not have all the information necessary to determine the most effective course of corrective action.
A number of organic compounds are primarily specific to human fecal contamination. These chemical indicators can be either natural products found in human feces (e.g., fecal steroids, aminopropanone) or synthetic chemicals found in products, such as detergents, that are specific to household waste streams (e.g., linear alkylbenzenes, trialkylamines, nonylphenols, whitening agents) (Takada and Eganhouse, 1998).
Fecal steroids (i.e., sterols, stanols, stanones) have been used frequently for differentiating between human and nonhuman sources (Grimalt et al., 1990; Leeming et al., 1996; Standley et al., 2000). The most frequently used steroid is coprostanol, which is produced by catabolism of cholesterol in the intestinal tract and is abundant in human feces (Hatcher and McGillivary, 1979; Maldonado et
al., 2000; Takada and Eganhouse, 1998; Venkatesan and Kaplan, 1990; Venkatesan and Mirsadeghi, 1992). While useful, interpretation of fecal steroid data is often confounded because these compounds can be produced via both in vivo biotic and in situ abiotic processes.
Linear alkylbenzenes (LABs), which are residues of anionic surfactants used in detergents and are an example of synthetic industrial chemical indicators. They have been widely used as chemical indicators of human fecal contamination (Eganhouse et al., 1988; Gustafsson et al., 2001; Phillips et al., 1997; Takada and Ishiwatari, 1987; Zeng et al., 1997), but are extremely hydrophobic and susceptible to degradation under aerobic conditions. Thus, LAB concentrations tend to decrease rapidly away from the point of discharge, limiting their usefulness as an indicator for sewage-contaminated waters.
A third widely used chemical indicator of human fecal contamination is caffeine. Although caffeine occurs naturally in more than 60 species of plants, none of these are indigenous to the United States, and any caffeine detected in surface waters must have originated from anthropogenic sources (Seigener and Chen, 2002). While source specific, caffeine is soluble in water and is diluted rapidly once released into the water column. Moreover, it does not sequester in sediments and is rapidly transported long distances from the source, so its concentration is usually low in environmental samples.
Although chemical indicators can provide some useful information regarding fecal contamination, they are generally less promising than genotypic or phenotypic methods because they are more costly and time consuming. Chemical methods are also typically less sensitive than biological methods and generally require large sample volumes to achieve adequate detection limits. Perhaps the biggest drawback to using chemical indicators is that so little is known about their fate in aquatic systems. To achieve their full potential utility, more research is needed to gain a better understanding of the transport, transformations, and persistence of these compounds under various environmental conditions. Despite their shortcomings, chemical indicator-based methods can be effective when used to complement other methods of microbial pollution monitoring.
Status and Trends Assessment
Bacterial indicators are frequently used in ambient water quality monitoring to assess whether water bodies meet state-designated water quality standards for beneficial use(s), such as drinking water supply, water contact recreation, and shellfish harvesting (see also footnote 3, Chapter 1). Ambient water quality monitoring can also focus on trends, to determine whether conditions are degrading or whether mitigation activities are leading to improvements. These beneficial use assessments generally involve integration of data over longer time periods and larger spatial scales than public health warning systems, which typically focus on
short-term decisions about individual sites. As a result, the indicator properties that are desirable for warning systems, such as rapidity of laboratory processing and results, differ for this application. Examples of indicator use in various types of ambient water monitoring are described below.
Surface Water Assessments
As described in Chapter 1 (see also Table 1-2), EPA is required under Sections 305(b) and 303(d) of the Clean Water Act to provide a national water quality assessment and to identify those water bodies that are failing to achieve their designated ambient water quality standards and beneficial uses. Water bodies not in compliance are subject to EPA’s Total Maximum Daily Load Program, which requires that discharges to that water body be reduced to the level necessary for achieving beneficial uses. TMDLs are initiated through an assessment of multiyear data and are typically accompanied by monitoring programs to assess trends in both source inputs and receiving water quality. Coliform (total, fecal, and E. coli) bacteria have been among the most frequent indicators used for identifying microbial water body impairment. Although coliforms have been useful for the status assessment portion of these activities, their use in trends assessment is often compromised because sample processing typically focuses on quantifying only a small part of the possible response range. Whereas health managers are interested in whether indicator concentrations exceed a risk-based threshold, trend detection requires measurement of indicator concentrations at levels below public health thresholds, since trend detection is not well achieved when data at the low end of the range are classified as non-detects. Trend detection is also hampered when values are censored at the high end of the range. This problem can be addressed within the context of existing bacterial indicator methods by processing multiple dilutions of a sample, but this adds significantly to the processing costs.
Although current microbial indicators meet most trend assessment needs, they can be improved. For instance, most TMDL listing decisions are based on indicator bacteria measurements without assessment of whether the source is human, even though TMDLs are intended to limit anthropogenic discharges to the water body. Similarly, trends assessments could be improved by determining whether ambient water quality changes were due to human or nonhuman inputs. The source tracking techniques discussed above are appropriate to these applications. In addition, most marine assessments are still based on coliforms, even though enterococci have been found to be more closely associated with health risk and significantly more marine water bodies would be detected as impaired if enterococci were used instead of coliforms (Noble et al., 2003b).
Shellfish Water Assessments
As noted in Chapter 1, extensive discussion of the use of indicators for shellfish microbial water quality monitoring is beyond the committee’s statement of task. However, it is appropriate to note that comparing coliform counts to standards based on biweekly or monthly average concentration is typical for monitoring shellfishing waters. Longer-term averages are used because shellfish tissue concentrations reflect this longer exposure period. The critical factor in these programs is ensuring that collections are made either in an unbiased manner or at times when concentrations are likely to be highest. Numerous studies have demonstrated the adverse effects of rainfall on bacteriological water quality (e.g., Ackerman and Weisberg, 2003; Boehm et al., 2002, Lipp et al., 2001a; Noble et al., 2003c), and these periods need to be included in the sampling effort, despite inconveniences associated with sampling during rain events.
While shellfish monitoring programs have generally been effective at protecting public health, they are most effective when complemented with a sanitary inspection of the surrounding shoreline. These surveys describe the location of human and animal populations in the shellfish growing area, land uses in the watershed that could impact water quality, and the location and magnitude of fecal waste sources (wastewater treatment plants and their effluents, on-site septic systems, etc.). Understanding these activities can be used to better focus the monitoring and aid in the interpretation of trends.
Drinking Water Supplies
Microbiological monitoring has historically played only a minor role in the evaluation of surface and groundwater sources of drinking water. Water utilities have relied on treatment (filtration and disinfection) and focused their monitoring on treatment effectiveness, rather than monitoring indicators of waterborne pathogens in their source water. However, recent waterborne outbreaks of cryptosporidiosis have led EPA to revise the Interim and Long Term 2 Enhanced Surface Water Treatment Rule (LT2ESWTR) that will require monitoring for Escherichia coli by utilities with surface source drinking water supplies (EPA, 2002b, 2003; see also Chapter 1 and Table 1-1). Depending on E. coli levels, some utilities will also be required to monitor for Cryptosporidium. The annual average concentration of these indicators will be used to define monitoring and treatment requirements that will provide increased control of Cryptosporidium.
Although the proposed requirements of the LT2ESWTR are a step in the right direction, surface water monitoring will remain a low-frequency activity (either monthly or every two weeks) and will miss transient impairments in microbial water quality. More specifically, the proposed sampling schedule does not directly consider the impact of precipitation events on source water quality, even though the microbial quality of both surface and groundwater becomes ap-
preciably worse following precipitation events, and this is the period of greatest vulnerability to waterborne outbreaks (Curriero et al., 2001; Rose et al., 2000).
Groundwater quality monitoring is rare, despite data that show the majority of drinking water outbreaks of disease in the United States result from groundwater systems (see Chapter 1 and Figure 1-2). Although there was no final national regulation for groundwater quality at the time this report was prepared, some states have wellhead protection programs for drinking water supplies using groundwater sources. The Ground Water Rule is expected to be promulgated sometime in late 2004, and will define disinfection needs for source water based on the vulnerability of the aquifer according to its hydrogeological characteristics and bacteriological quality (EPA, 2000). More specifically, coliform bacteria have been recommended as the indicator of choice for groundwater, with an option for including coliphage or direct virus monitoring. The known risks from viruses in fecally contaminated groundwater, combined with evidence that coliphages are better indicators of viruses than are indicator bacteria, and that human enteric viruses are detectable in fecally contaminated groundwater using current technologies, suggest that coliphage or direct virus monitoring would enhance the assessment of groundwater microbiological quality and would make better indicators of human health risk (see Chapter 6 for further information).
Prediction-based Warning Systems
The typical application of indicators for public health warning systems involves measuring bacterial indicators to assess recent water quality conditions. One shortcoming of this approach is that it does not prevent exposure, since people swim in (or drink) the water prior to sampling, during sample processing, and while mitigative or warning actions are being taken. An alternative approach is to develop predictive models that prevent exposure.
One example is the use of rainfall as a predictive indicator. Rainfall is associated with elevated bacterial indicator levels on both daily (Curriero et al., 2001; Kistemann et al., 2002; Schiff et al., 2003) and seasonal (Boehm et al., 2002; Lipp et al., 2001b) time scales. These elevated levels typically result from urban runoff and combined sewage-stormwater system overflows.
Several states issue swimmer warnings based on rainfall. For example, five county health departments in southern California routinely issue warnings not to swim in the ocean for three days following a rainstorm of 0.1 inch or more (Ackerman and Weisberg, 2003). Although California’s warnings are only advisory, Monmouth County in New Jersey routinely closes two beaches that typically have elevated bacterial concentrations following runoff events for 24 hours following 0.1 inch or more rain; the closure is extended to 48 hours following 2.8 inches or more of rain (David Rosenblatt, New Jersey Department of Environmental Protection, personal communication, 2003).
While rainfall-based warnings are valuable, they are based on limited em-
Great Lakes scientists are using multiple regression techniques to develop more sophisticated models for predicting beach water quality in Chicago and Milwaukee (Olyphant and Whitman, in press). These models include rainfall during the previous 24 hours, wind, solar radiation, water temperature, lake stage, water turbidity, and pH. Rainfall, wind, and turbidity are indicative of the strong influence that storms have on E. coli concentrations. At the Milwaukee beach, storm effects result primarily from sewage overflows into tributary rivers that get pushed shoreward by easterly winds. The Chicago beach is not directly influenced by stream inflows, but storms stir up E. coli laden sand in the breaker zone. Solar radiation is a negative term in the model that reflects UV-mediated bacterial die-off during bright sunshine. Water temperature and lake stage represent conditions that lead to high bacterial concentrations during non-storm periods. Bacterial populations grow faster in warm water and bacteria become more concentrated when lake levels fall at the beach in Chicago. These models were evaluated by comparing predictions of E. coli concentration exceeding EPA’s recommended threshold of 235 CFU/100mL with measured concentrations. The model correctly predicted 66 of 90 events at the Milwaukee beach and 50 of 57 events at the Chicago beach. Model errors were evenly split between false negatives and false positives for the Milwaukee beach, but six of the seven incorrect predictions for Chicago are ones that would have led to over-protective actions.
pirical evidence. Ackerman and Weisberg (2003) found that 91 percent of storms with precipitation greater than 0.25 inches led to an increase in the number of Los Angeles beaches failing bacterial water quality standards. However, the response was more equivocal for storms with precipitation between 0.1 and 0.25 inches, when factors such as spatial coverage of the storm, antecedent rainfall, and size and type of watershed become potentially more important in determining the need for warnings. More complex models that incorporate these factors, as well as similar studies conducted in other parts of the country (see Box 4-7), will have to be developed before predictive models become widely accepted tools for public health warnings.
Another predictive-based warning system, which operates on a longer time scale, involves land use as an indicator of fecal contamination. Many recreational bathing areas, drinking water sources, and shellfishing areas are located in drain-
age basins that are undergoing development pressure. Changing land use, such as increased urbanization or conversion of rangelands to agricultural lands, can affect pathogen contributions within the drainage area. Mallin et al. (2000, 2001) have demonstrated a statistical relationship between the amount of development in a watershed and downstream bacterial concentrations, but these results are likely to be site specific. The theoretical framework for more generalizable models that predict receiving water contaminant concentrations based on land use, such as HSPF (Hydrologic Simulation Program—Fortran; Bicknell et al., 1997), are available, but the runoff relationships necessary to parameterize these models are not well developed. Further work on these models is needed before managers can use them to define the level of development at which increased mitigation activities will be necessary to ensure acceptable water quality.
Lobitz et al. (2000) have suggested that remote sensing data can also serve as a predictive tool for bacterial waterborne outbreaks. They indicate that Vibrio cholerae occurs commensally with species of phytoplankton, the density of which can be tracked through satellite imagery. Moreover, satellite imagery of circulation and sea surface temperature can be used to predict future blooms. While such modeling approaches need more empirical testing, rapid advances in remote sensing technology (e.g., Isern and Clark, 2003) will provide new opportunities for developing such models.
SUMMARY: CONCLUSIONS AND RECOMMENDATIONS
Microbial water quality indicators are used in a variety of ways within public health risk assessment frameworks, and the most desirable indicator attributes—and therefore the most appropriate indicators—naturally depend on their manner of use. Despite their importance and longevity, Bonde’s attributes of an ideal public health indicator need to be refined. These historic definitions of indicators have been tied to the methods used to measure them, but the development of new measurement methods necessitates separate criteria for evaluating the biological and method attributes of indicators. Separate criteria allow one to choose the indicator with the most desirable biological attribute for a given application and to match this with a measurement method that best meets the need of the application.
The most important biological attribute is a strong quantitative relationship between indicator concentration and the degree of public health risk. One of the most important method attributes is its specificity, or ability to measure the target indicator organism in an unbiased manner. Speed of the method (processing time and rapidity of results) is also an important characteristic in many cases, particularly when warning systems are involved and human exposure occurs during the laboratory analysis period.
Many public health applications use microbiological indicators, including public health warning systems, source identification, and status or trends assess-
ments. No single indicator or analytical method (or even a small set of indicators or analytical methods) is appropriate to all applications. A suite of indicators and indicator approaches is required for different applications and different geographies.
Several factors limit the effectiveness of current recreational water warning systems, the most prominent of which is the delay in warnings caused by long laboratory sample processing times. Current laboratory measurement methods used to enumerate indicator bacteria (multiple tube fermentation, membrane filtration, and chromogenic substrate) are too time consuming. They require an 18-to 96-hour incubation period, during which the public is exposed to potential health risks. One approach that is increasingly being used to address this problem is predictive models intended to prevent exposure.
Another shortcoming of present warning systems is the poorly established relationship between presently used indicators and health risk. Current studies do not address all sources of contamination, have not identified the etiological agents of illness, have not been conducted in enough geographical locations, and do not address chronic exposure. Many reported failures of beach water quality standards are associated with nonpoint source contamination, but the epidemiologic studies used to establish recreational bathing water standards have been based primarily on exposure to point source contamination dominated by human fecal material.
A major problem with present water contact warning systems is that bacterial indicator concentrations are spatiotemporally variable and most sampling is too infrequent to transcend this granularity. The predominant all-or-none decision framework of either closing the beach or taking no action at all, sometimes on the basis of a single sample, magnifies the errors associated with this temporal and spatial granularity.
There are many promising microbial source identification techniques that can help in deciding whether a health warning should be issued or in identifying the best approach for fixing the problem. However, these techniques are not yet standardized or fully tested.
Groundwater quality monitoring is rare, despite data showing that the majority of waterborne outbreaks of disease in the United States result from groundwater systems. Viral contamination of groundwater is a particular concern because the small size and considerable environmental persistence of viruses make it more likely they will reach and contaminate groundwater. The known risks from viruses in fecally contaminated groundwater, and evidence that human enteric viruses are detectable in fecally contaminated groundwater, suggest that coliphage or direct virus monitoring would enhance the assessment of groundwater microbiological quality and would make a better indicator of human health risk.
The discussion in this chapter and the preceding conclusions support the following recommendations:
Since it is not possible to identify a single, unique indicator or small set of indicators capable of identifying all classes of waterborne microbial pathogens, priority should be given to developing a phased monitoring approach that relies on a flexible “tool box” of indicators and indicator approaches that are used according to strategies appropriate to the specific applications (see Chapter 6).
The link between indicators and pathogens, and among indicators, pathogens, and adverse health outcomes, would be strengthened by including measurements of both indicators and pathogens in comprehensive epidemiologic studies. In particular, studies to better assess the role of nonpoint sources in occurrence of human pathogens and indicator organisms, disease outbreaks, and endemic health risks in recreational waters should be conducted. Use of alternative indicators need to be included in these studies.
Improved indicators for viruses in groundwater sources of drinking water need to be developed.
New paradigms for reporting water contact health risk, such as “letter grades” for public beaches, need to be developed. The present all-or-none closure decisions can misinform the public because of large spatiotemporal heterogeneity in indicator concentrations. Letter grades—which have been used successfully in some parts of the country to provide the public information about the health quality of restaurants—are one option that would effectively address the granularity issue by integrating data over a longer time period and are readily understandable.
Investment should be made in developing rapid analytical methods. The most commonly used warning systems involve laboratory methods that are too time consuming to achieve the best possible public health protection. New molecular methods, which do not have the long incubation time requirements of present culture-based methods, are on the near term horizon (see Chapters 5 and 6).
There are several promising source identification (i.e., microbial source tracking) techniques on the horizon that should be incorporated into monitoring systems when they have been adequately validated. Public health risk from exposure to fecally contaminated water is likely to vary depending on whether high indicator concentrations resulted from animal or human sources, and microbial source tracking tools will allow public health managers to incorporate that distinction into their decision making.
No matter how rapid measurement techniques become, they will always be retrospective. Models that predict future water quality conditions, based on factors such as rainfall, are potentially valuable tools for warning the public before exposure occurs, but the scientific foundation for these models has to be enhanced before they can be widely used.
Ackerman, D., and S.B. Weisberg. 2003. Relationship between rainfall and beach bacterial concentration on Santa Monica Bay beaches. Journal of Water and Health 1: 85-89.
Allen, M. 1977. The impact of excessive bacterial populations in coliform methodology. American Society for Microbiology Annual Conference.
Bernhard, A.E., and K.G. Field. 2000a. Identification of nonpoint sources of fecal pollution in coastal waters by using host-specific 16S ribosomal DNA markers from fecal anaerobes. Applied and Environmental Microbiology 66: 1587-1594.
Bernhard, A.E., and K.G. Field. 2000b. A PCR assay to discriminate human and ruminant feces based on host differences in Bacteroides prevotella 16S ribosomal DNA. Applied and Environmental Microbiology 66: 4571-4574.
Bernhard, A. E., T. Goyard, M. Simonich, and K.G. Field. 2003. A rapid method for identifying fecal pollution sources in coastal waters. Water Research 37: 909-913.
Bicknell, B.R., J.C. Imhoff, J.L. Kittle, Jr., A.S. Donigian, Jr., and R.C. Johanson. 1997. Hydrological Simulation Program - Fortran Users Manual for Version 11. Athens, Georgia: U.S. Environmental Protection Agency, National Exposure Research Laboratory.
Boehm, A.B., S.B. Grant, J.H. Kim, S.L. Mowbray, C.D. McGee, C.D. Clark, D.M. Foley, and D.E. Wellman. 2002. Decadal and shorter period variability and surf zone water quality at Huntington Beach, California. Environmental Science and Technology 36: 3885-3892.
Bonde, G. 1966. Bacteriological methods for estimation of water pollution. Health Laboratory Science 3: 124-128.
Burkholder, J.M., M.A. Mallin, H.B. Glasgow, Jr., L.M. Larsen, M.R. McIver, G.C. Shank, N. Deamer-Melia, D.S. Briley, J. Springer, B.W. Touchette, and E.K. Hannon. 1997. Impacts to a coastal river and estuary from rupture of a large swine waste holding lagoon. Journal of Environmental Quality 26: 1451-1466.
Cabelli, V.J., A.P. Dufour, M.A. Levin, L.J. McCabe, and P.W. Haberman. 1979. Relationship of microbial indicators to health effects at marine bathing beaches. American Journal of Public Health 69: 690-696.
Camper, A.K., and G.A. McFeters. 1979. Chlorine injury and the enumeration of waterborne coliform bacteria. Applied and Environmental Microbiology 3: 633-641.
Carson, C.A., B.L. Shear, M.R. Ellersieck, and A. Asfaw. 2001. Identification of fecal Escherichia coli from humans and animals by ribotyping. Applied and Environmental Microbiology 67: 1503-1507.
Carson, C.A., B.L. Shear, M.R. Ellersieck, and J.D. Schnell. 2003. Comparison of ribotyping and repetitive extragenic palindromic-PCR for identification of fecal Escherichia coli from humans and animals. Applied and Environmental Microbiology 69: 1836-1839.
Chee-Sanford, J.C., R.I. Aminov, I.J. Krapac, N. Garrigues-Jeanjean, and R.I. Mackie. 2001. Occurrence and diversity of tetracycline resistance genes in lagoons and groundwater underlying two swine production facilities. Applied and Environmental Microbiology 67: 1494-1502.
Cheung, W.H.S., R.P.S. Hung, K.C.K. Chang, and J.W.L. Kleevens. 1990. Epidemiological study of bathing beach water pollution and health related bathing water standards in Hong Kong. Water Science and Technology 23: 243-252.
Chung, H. 1993. F-specific coliphages and their serogroups and Bacteroides fragilis phages as indicators of estuarine water and shellfish quality. Ph.D. dissertation, University of North Carolina, Chapel Hill.
Clesceri, L.S., A.E. Greenberg, and A.D. Eaton, eds. 1998. Standard Methods for the Examination of Water and Wastewater, 20th Edition. Washington, D.C.: American Public Health Association.
Cooke, M.D. 1976. Antibiotic resistance among coliform and fecal coliform bacteria isolated from sewage, seawater, and marine shellfish. Antimicrobial Agents and Chemotherapy 9: 879-884.
Curriero, F.C., J.A. Patz, J.R. Rose, and S. Lele. 2001. The association between extreme precipitation and waterborne disease outbreaks in the United States, 1948-1994. American Journal of Public Health 91: 1194-1199.
Desmarais, T.R., H.M. Solo-Gabriele, and C.J. Palmer. 2002. Influence of soil on fecal indicator organisms in a tidally influenced subtropical environment. Applied and Environmental Microbiology 68: 1165-1172.
Dombek, P.E., L.K. Johnson, S.T. Zimmerly, and M.J. Sadowsky. 2000. Use of repetitive DNA sequences and the PCR to differentiate Escherichia coli isolates from human and animal sources. Applied and Environmental Microbiology 66: 2572-2577.
Edberg, S.C., and D.B. Smith. 1989. Absence of association between total heterotrophic and total coliform bacteria from a public water supply. Applied and Environmental Microbiology 55(2): 380-384.
Eganhouse, R.P., D.P. Olaguer, B.R. Gould, and C.S. Phinney. 1988. Use of molecular markers for the detection of municipal sewage sludge at sea. Marine Environmental Research 25: 1-22.
Emerson, S.U., and R.H. Purcell. 2003. Hepatitis E virus. Reviews in Medical Virology 13(3): 145-154.
EPA (U.S. Environmental Protection Agency). 1986. Ambient Water Quality Criteria for Bacteria -1986. Washington, D.C.: Office of Water. EPA 440-5-84-002.
EPA. 2000. National Primary Drinking Water Regulations: Ground Water Rule; Proposed Rule. Federal Register 65(91): 30193-30274.
EPA. 2002a. EPA’s Beachwatch Program: 2001 Swimming Season. Washington, D.C.: Office of Water. EPA 823-F-02-008.
EPA. 2002b. National Primary Drinking Water Regulations: Long Term 1 Enhanced Surface Water Treatment Rule. Federal Register 76: 1812-1844.
EPA. 2003. National Primary Drinking Water Regulations: Long Term 2 Enhanced Surface Water Treatment Rule; Proposed Rule. Federal Register 68(154): 47640-47795.
Feachem, R.G., D.J. Bradley, H. Garelick, and D.D. Mara, eds. 1983. Sanitation and Disease, Health Aspects of Excreta and Wastewater Management. New York: John Wiley & Sons.
Fujioka, R.S. 2001. Monitoring coastal marine waters for spore forming bacteria of faecal and soil origin to determine point from non-point source pollution. Water Science and Technology 44: 181-188.
Geldreich, E.E. 1978. Interferences to coliform detection in potable water supplies. Pp. 13-19 in Evaluation of the Microbiology Standards for Drinking Water. C.W. Hendricks, ed. Washington, D.C.: EPA-570-9-78-00C.
Gerba, C.P., S.M. Goyal, R.L. LaBelle, I. Cech, and G.F. Bogdan. 1979. Failure of indicator bacteria to reflect the occurrence of enteroviruses in marine waters. American Journal of Public Health 69: 1116-1119.
Gerba, C.P. 2001. Assessment of enteric pathogen shedding by bathers during recreational activity and its impact on water quality. Quantitative Microbiology 2: 55-68.
Griffin, D.W., E.K. Lipp, M.R. McLaughlin, and J.B. Rose. 2001. Marine recreation and public health microbiology: Quest for the ideal indicator. BioScience 51: 817-825.
Grimalt, J.O., P. Fernandez, J.M. Bayona, and J. Albages. 1990. Assessment of fecal sterols and ketones as indicators of urban sewage inputs to coastal waters. Environmental Science and Technology 24(3): 357-363.
Gustafsson, Ö., C.M. Long, J. Macfarlane, and P.M. Gschwend. 2001. Fate of linear alkylbenzenes released to the coastal environment near Boston Harbor. Environmental Science and Technology 35: 2040-2048.
Hagedorn, C.S., S.L. Robinson, J.R. Filtz, S.M. Grubbs, T.A. Angier, and R.B. Reneau, Jr. 1999. Using antibiotic resistance patterns in the fecal streptococci to determine sources of fecal pollution in a rural Virginia watershed. Applied and Environmental Microbiology 65: 5522-5531.
Hagedorn, C.S., J.B. Crozier, K.A. Mentz, A.M. Booth, A.K. Graves, N.J. Nelson, and R.B. Reneau. 2003. Carbon source utilization profiles as a method to identify sources of faecal pollution in water. Journal of Applied Microbiology 94: 792-799.
Haile, R.W., J.S. Witte, M. Gold, R. Cressey, C.D. McGee, R.C. Millikan, A. Glasser, N. Harawa, C. Ervin, P. Harmon, J. Harper, J. Dermand, J. Alamillo, K. Barrett, M. Nides, and G. Wang. 1999. The health effects of swimming in ocean water contaminated by storm drain runoff. Journal of Epidemiology 104: 355-363.
Harden, H.S., J.P. Chanton, J.B. Rose, D.E. John, and M.E. Hooks. 2003. Comparison of sulfur hexafluroride, fluorescein and rhodamine dyes and the bacteriophage PDR-1 in tracing subsurface flow. Journal of Hydrology 277(1-2): 100-115.
Hardina, C.M., and R.S. Fujioka. 1991. Soil, the environmental source of Escherichia coli and enterococci in Hawaii’s streams. Environmental Toxicology 6: 185-195.
Hartel, P.G., J.D. Summer, J.L. Hill, J.V. Collins, J.A. Entry, and W.I. Segars. 2002. Geographic variability of Escherichia coli ribotypes from animals in Idaho and Georgia. Journal of Environmental Quality 31: 1273-1278.
Harwood, V.J., J. Whitlock, and V.H. Withington. 2000. Classification of the antibiotic resistance patterns of indicator bacteria by discriminant analysis: Use in predicting the source of fecal contamination in subtropical Florida waters. Applied and Environmental Microbiology 66: 3698-3704.
Hatcher, P.G., and P.A. McGillivary. 1979. Sewage contamination in the New York Bight: Coprostanol as an indicator. Environmental Science and Technology 13: 1225-1229.
Havelaar, A.H. 1993. Bacteriophages as models of human enteric viruses in the environment. American Society for Microbiology News 59: 614-619.
Hsu, F-C., Y-S Shieh, J. van Duin, M.J. Beekwilder, and M.D. Sobsey. 1995. Genotyping male-specific RNA coliphages by hybridization with oligonucleotide probes. Applied and Environmental Microbiology 61: 3960-3966.
IAWPRC (International Association on Water Pollution Research and Control) Study Group. 1991. Bacteriophages as model organisms in water quality control. Water Research 25: 529-545.
Isern, A.R., and H.L. Clark. 2003. The ocean observatories initiative: A continued presence for inter-active ocean research. Marine Technology Society Journal 37:26-41.
Jenkins, M.B., P.G. Hartel, T.J. Olexa, and J.A. Stuedemann. 2003. Putative temporal variability of Escherichia coli ribotypes from yearling steers. Journal of Environmental Quality 32: 305-309.
Jiang, S., R. Noble, and W. Chu. 2001. Human adenoviruses and coliphages in urban-runoff impacted coastal waters of southern California. Applied and Environmental Microbiology 67: 179-184.
Khatib, L.A., Y.L. Tsai, and B.H. Olson. 2002. A biomarker for the identification of cattle fecal pollution in water using the LYIIa toxin gene from the enterotoxigenic Escherichia coli. Applied Microbiology and Biotechnology 59: 97-104.
Kistemann, T.C., C. Koch, F. Dangendorf, R. Fischeder, J. Gebel, V. Vacata, and M. Exner. 2002. Microbial load of drinking water reservoir tributaries during extreme rainfall and runoff. Applied and Environmental Microbiology 68: 2188-2197.
Kott, Y., N. Roze, S. Sperber, and N. Betzer. 1974. Bacteriophages as viral pollution indicators. Water Research 8: 165-171.
Labelle, R.L., C.P. Gerba, S.M. Goyal, J.L. Melnick, I. Cech, and G.F. Bogdan. 1980. Relationships between environmental factors, bacterial indicators and the occurrence of enteric viruses in estuarine sediments. Applied and Environmental Microbiology 39: 586-596.
LeChevallier, M.W., M. Abbaszdegan, A.K. Camper, G. Izaguirre, M. Stewart, D. Naumovitz, M. Mardhall, C.R. Sterling, P. Payment, E.W. Rice, C.J. Hurst, S. Schaub, T.R. Slifko, J.B. Rose, H.V. Smith, and D.B. Smith. 1999a. Emerging pathogens - bacteria. Journal of the American Water Works Association 91: 136-172.
LeChevallier, M.W., M. Abbaszdegan, A.K. Camper, G. Izaguirre, M. Stewart, D. Naumovitz, M. Mardhall, C.R. Sterling, P. Payment, E.W. Rice, C.J. Hurst, S. Schaub, T.R. Slifko, J.B. Rose, H.V. Smith, and D.B. Smith. 1999b. Emerging pathogens - viruses, protozoa, and algal toxins. Journal of the American Water Works Association 91: 110-121.
Lee, S.-H., and S.-J. Kim. 2002. Detection of infectious enteroviruses and adenoviruses in tap water in urban areas in Korea. Water Research 36: 248-256.
Leecaster, M.K., and S.B. Weisberg. 2001. Effect of sampling frequency on shoreline microbiology assessments. Marine Pollution Bulletin 42: 1150-1154.
Leeming, R., A. Ball, N. Ashbolt, and P. Nichols. 1996. Using fecal sterols from humans and animals to distinguish fecal pollution in receiving waters. Water Research 30: 2893-2900.
Lipp, E.K., R. Kurz, R. Vincent, C. Rodriguez-Palacios, S.R. Farrah, and J.B. Rose. 2001a. The effects of seasonal variability and weather on microbial fecal pollution and enteric pathogens in a subtropical estuary. Estuaries 24: 266-276.
Lipp, E.K., N. Schmidt, M.E. Luther, and J.B. Rose. 2001b. Determining the effects of El Niño-Southern Oscillation events on coastal water quality. Estuaries 24: 491-497.
Lobitz, B., L. Beck, A. Huq, B. Wood, G. Fuchs, A.S. Faruque, and R. Colwell. 2000. Climate and infectious disease: Use of remote sensing for detection of Vibrio cholerae by indirect measurement. Proceedings of the National Academy of Sciences 97(4):1438-1443.
Long, S. 2002. Development of Source-Specific Indicator Organisms for Drinking Water (Project #2645). Report Order Number 90911. Denver, Colorado: American Water Works Association Research Foundation.
Maldonado, C., M.I. Venkatesan, C.R. Philips, and J.M. Bayona. 2000. Distribution of trialkylamines and coprostanol in San Pedro shelf sediments adjacent to a sewage outfall. Marine Pollution Bulletin 40(8): 680-687.
Mallin, M.A., J.A.M. Burkholder, and J. Springer. 1997. Comparative effects of poultry and swine waste lagoon spills on the quality of receiving streamwaters. Journal of Environmental Quality 26: 1622-1631.
Mallin, M.A., K.E. Williams, E.C. Esham, and R.P. Lowe. 2000. Effect of human development on bacteriological water quality in coastal watersheds. Ecological Applications 10: 1047-1056.
Mallin, M.A., S.H. Ensign, M.R. McIver, G.C. Shank, and P.K. Fowler. 2001. Demographic, landscape, and meteorological factors controlling the microbial pollution of coastal waters. Hydrobiologia 460: 185-193.
Mara, D.D., and J.I. Oragui. 1983. Sorbitol-fermenting bifidobacteria as specific indicators of human faecal pollution. Journal of Applied Bacteriology 55: 349-357.
Mara, D.D., and J.I. Oragui. 1985. Bacteriological methods for distinguishing between human and animal faecal pollution of water: Results of fieldwork in Nigeria and Zimbabwe. Bulletin of the World Health Organization 63(4): 773-783.
Mathew, A.G., W.G. Upchurch, and S.E. Chattin. 1998. Incidence of antibiotic resistance in fecal Escherichia coli isolated from commercial swine farms. Journal of Animal Science 76: 429-434.
Mathew, A.G., A.M. Saxton, W.G. Upchurch, and S.E. Chattin. 1999. Multiple antibiotic resistance patterns of Escherichia coli isolates from swine farms. Applied and Environmental Microbiology 65: 2770-2772.
McFeters, G.A., J.S. Kippin, and M.W. LeChevallier. 1986. Injured coliforms in drinking water. Applied and Environmental Microbiology 5: 1-5.
McLaughlin, M.R. 2001. Application of Bacteroides fragilis phage as an alternative indicator of sewage pollution in Tampa Bay, Florida. M.S. thesis. University of South Florida, St. Petersburg.
Noble, R.T., J.H. Dorsey, M.K. Leecaster, V. Orozco-Borbon, D. Reid, K.C. Schiff, and S.B. Weisberg. 2000. A regional survey of the microbiological water quality along the shoreline of the Southern California Bight. Environmental Monitoring and Assessment 64: 435-447.
Noble, R.T., and J.A. Fuhrman. 2001. Enterovirsuses detected by reverse transcriptase polymerase chain reaction from the coastal waters of Santa Monica Bay, California: Low correlation to bacterial indicator levels. Hydrobiologia 460: 175-184.
Noble, R.T., S.B. Weisberg, M.K. Leecaster, C.D. McGee, K. Ritter, K.O. Walker and P.M. Vainik. 2003a. Comparison of beach bacterial water quality indicator measurement methods. Environmental Monitoring and Assessment 81: 301-312.
Noble, R.T., D.F. Moore, M.K. Leecaster, C.D. McGee, and S.B. Weisberg. 2003b. Comparison of total coliform, fecal coliform, and enterococcus bacterial indicator response for ocean recreational water quality testing. Water Research 37: 1637-1643.
Noble, R.T., S.B. Weisberg, M.K. Leecaster, C.D. McGee, J.H. Dorsey, P. Vainik, and V. Orozco-Borbón. 2003c. Storm effects on regional beach water quality along the southern California shoreline. Journal of Water and Health 1: 23-31.
Olyphant, G.A., and R.L. Whitman. 2004. Elements of a predictive model for determining beach closures on a real time basis: The case of 63rd Street Beach, Chicago. Environmental Monitoring and Assessment. (in press).
Parveen, S., R.L. Murphree, L. Edminston, C.W. Kaspar, K.M. Portier, and M.L. Tamplin. 1997. Association of multiple-antibiotic-resistance profiles with point and nonpoint sources of Escherichia coli in Apalachicola Bay. Applied and Environmental Microbiology 63: 2607-2612.
Parveen, S., K.M. Portier, K. Robinson, L. Edminston, and M.L. Tamplin. 1999. Discriminant analysis of ribotype profiles of Escherichia coli for differentiating human and nonhuman sources of fecal pollution. Applied and Environmental Microbiology 65: 3142-3147.
Parveen, S., N.C. Hodge, R.E. Stall, S.R. Farrah, and M.L. Tamplin. 2001. Genotypic and phenotypic characterization of human and nonhuman Escherichia coli. Water Research 35: 379-386.
Phillips, C.R., M.I. Venkatesan, and R. Bowen. 1997. Interpretations of contaminant sources to San Pedro shelf sediments using molecular markers and principal components analysis. Pp. 242-260 in Molecular Markers in Environmental Geochemistry, R.P. Eganhouse, ed. Washington, D.C.: American Chemical Society.
Pisciotta, J.M., D.F. Rath, P.A. Stanek, D.M. Flanery, and V.J. Harwood. 2002. Marine bacteria cause false-positive results in the Colilert-18 rapid identification test for Escherichia coli in Florida waters. Applied and Environmental Microbiology 68: 539-544.
Prüss, A. 1998. Review of epidemiological studies on health effects from exposure to recreational water. International Journal of Epidemiology 27: 1-9.
Reynolds, K.A., C.P. Gerba, and I.L. Pepper. 1996. Detection of infectious enteroviruses by an integrated cell culture-PCR procedure. Applied and Environmental Microbiology 62: 1424-1427.
Robertson, W.J. 1984. Pollution indicators and potential pathogen microorganisms in estuarine recreational waters. Canadian Journal of Public Health 75: 19-24.
Rose, J.B., S. Daeschner, D.R. Deasterling, F.C. Curriero, S. Lele, and J. Patz. 2000. Climate and waterborne disease outbreaks. Journal of the American Water Works Association 92: 77-87.
Rose, J.B., D.E. Huffman, K. Riley, S.R. Farrah, J.O. Lukasik, and C.L. Harman. 2001. Reduction of enteric microorganisms at the Upper Occoquan Sewage Authority water reclamation plant. Water Environment Research 73: 711-720.
Schaper, M., J. Jofre, M. Uys, and W.O.K. Grabow. 2002. Distribution of genotypes of F-specific RNA bacteriophages in human and non-human sources of faecal pollution in South Africa and Spain. Journal of Applied Microbiology 92: 657-667.
Schiff, K.C., J. Morton, and S.B. Weisberg. 2003. Retrospective evaluation of shoreline water quality along Santa Monica Bay beaches. Marine Environmental Research 56: 245-254.
Scott, T.M., J.B. Rose, T.M. Jenkins, S.R. Farrah, and J. Lukasik. 2002. Microbial source tracking: Current methodology and future directions. Applied and Environmental Microbiology 68: 5796-5803.
Scott, T.M., S. Parveen, K.M. Portier, J.B. Rose, M.L. Tamplin, S.R. Farrah, and J. Lukasik. 2003. Geographical variation in ribotype profiles of Escherichia coli isolated from humans, swine, poultry, beef, and dairy cattle in Florida. Applied and Environmental Microbiology 69(2): 1089-1092.
Seigener, R., and R.F. Chen. 2002. Caffeine in Boston Harbor seawater. Marine Pollution Bulletin 44: 383-387.
Seyfried, P.L., N.E. Brown, C.L. Cherwinsky, G.D. Jenkins, D.A. Cotter, J.M. Winner, and R.S. Tobin. 1984. Impact of sewage treatment plants on surface waters. Canadian Journal of Public Health 75: 25-31.
Seyfried, P.L., R.S. Tobin, N.E. Brown, and P.F. Ness. 1985a. A prospective study of swimming-related illness: I. Swimming-associated health risk. American Journal of Public Health 75: 1068-1070.
Seyfried, P.L., R.S. Tobin, N.E. Brown, and P.F. Ness. 1985b. A prospective study of swimming-related illness: II. Morbidity and the microbiological quality of water. American Journal of Public Health 75: 1071-1075.
Simpson, J.M., J.W. Santo-Domingo, and D.J. Reasoner. 2002. Microbial source tracking: State of the science. Environmental Science and Technology 36: 5729-5289.
Solo-Gabrielle, H.M., M.A. Wolfert, T.R. Desmarais, and C.J. Palmer. 2000. Sources of Escherichia coli in a coastal subtropical environment. Applied and Environmental Microbiology 66: 230-237.
Standley, L.J., L.A. Kaplan, and D. Smith. 2000. Molecular tracer of organic matter sources to surface water resources. Environmental Science and Technology 34: 3124-3130.
Taggart, M. 2002. Factors affecting shoreline fecal bacteria densities around freshwater outlets at two marine beaches. Ph.D. dissertation, University of California, Los Angeles.
Takada, H., and R. Ishiwatari. 1987. Linear alkylbenzenes in urban riverine environments in Tokyo: Distribution, source, and behavior. Environmental Science and Technology 21: 875-883.
Takada, H., and R.P. Eganhouse. 1998. Molecular markers of anthropogenic waste. Pp. 2883-2940 in Encyclopedia of Environmental Analysis and Remediation, R.A. Meyers, ed. New York: John Wiley & Sons.
Tartera, C., and J. Jofre. 1987. Bacteriophages active against Bacteroides fragilis in sewage-polluted waters. Applied and Environmental Microbiology 53: 1632-1637.
Venkatesan, M.I., and I.R. Kaplan. 1990. Sedimentary coprostanol as an index of sewage addition in Santa Monica Basin, Southern California. Environmental Science and Technology 24: 208-214.
Venkatesan, M.I., and F.H. Mirsadeghi. 1992. Coprostanol as sewage tracer in McMurdo Sound, Antarctica. Marine Pollution Bulletin 25: 328-333.
Wade, T.J., N. Pai, J.N.S. Eisenberg, and J.M. Colford, Jr. 2003. Do U.S. Environmental Protection Agency water quality guidelines for recreational waters prevent gastrointestinal illness? A systematic review and meta-analysis. Environmental Health Perspectives 111(8): 1102-1109.
Zeng, E.Y., A.R. Khan, and K. Tran. 1997. Organic pollutants in the coastal marine environment off San Diego, California. 3. Using linear alkylbenzenes to trace sewage-derived organic materials. Environmental Toxicology and Chemistry 16: 196-201.
Zmirou, D., J.P. Ferley, J.F. Collin, M. Charrel, and J. Berlin. 1987. A follow-up study of gastrointestinal diseases related to bacteriologically substandard drinking water. American Journal of Public Health 77: 582-584.