Using microbial source tracking as a fecal pollution source identifier
Fecal pollution remains a problem worldwide. Government authorities, environmental groups and various stakeholders are dedicated to reducing and eliminating fecal pollution from water systems. However, achieving these goals is a struggle because traditional water quality monitoring methods do not reveal fecal pollution sources. How can remediation be effective if the source is not known?
Bacterial contamination in storm water runoff within urban settings can originate from multiple sources. Examples are human sources, which can come from leaking septic tanks, illicit sewer connections and transient human populations; and animal sources, such as dogs, birds, or cattle fecal material from fertilizers or cattle found in peri-urban towns or slaughterhouses.
A DNA-based test method called microbial source tracking can determine the origins of bacterial pollution. Visual observations are subjective and non-repeatable. Monitoring for fecal indicator bacteria (FIB) can indicate elevated fecal matter but will not discriminate between the sources. FIB monitoring also will show readings from non-fecal sources, such as soil or decaying plant matter. These readings complicate the situation further and make it more difficult to find the source to remediate. Non-fecal natural sources have a small chance of containing pathogens and should not need remediation.
DNA testing has become a favorable method to identify bacteria sources. It is used in forensics, water quality and even in the food quality industries. The underlying premise for genetic testing is that all warm-blooded animals have gut bacteria and each individual species has its own unique gut bacteria. The distinction is influenced by different body temperatures, physiology and diet. Microbial source tracking (MST) methods based on quantitative real-time polymerase chain reaction (qPCR) technologies are powerful tools to identify the unique gut bacteria of various host sources. A positive reading denotes that the species is one of the causes of contamination.
It is difficult to compare MST data without manipulation. qPCR machines analyze MST samples, and qPCR data is either positive for a particular marker or negative if the marker is not present. In the case of the latter, a non-detect (ND) is reported. If the result is positive, then either the number of genetic copies per 100 mL is quantified or, in many situations, if the concentrations are below the sensitivity of the qPCR machine, then a did not quantify (DNQ) is reported. It is important to note that absence of detection in one sample is not scientific proof of detection of absence from the entire study. An accurate study must include all data points, including the ND and DNQ results. Ignoring the ND or DNQ results, or substituting with a zero, is not scientifically sound.
Standard laboratory protocols exist to analyze samples and an internationally accredited lab is available to generate high-quality data. However, data interpretation is a major obstacle. Many MST projects hire an individual or professionals to analyze PCR data. This is called Best Professional Judgment (BPJ).
A team of researchers sought to elucidate if inconsistencies were a concern in BPJ data interpretation. The team tested this theory with the help of 10 experts that are published water quality managers and scientists from the federal government, a public research agency, academic institutions and a wastewater treatment agency.
The experts were given 26 sets of simulated data representing 26 fake sites and were asked to rank the sites from 1 to 26 based on their interpretation of the extent of human fecal pollution at these sites. Rank 1 represents the most contaminated and 26 the least. Consistency between interpretations was evaluated by correlation between the ranks. A larger correlation coefficient indicated a higher level of consistency among experts, and a smaller correlation coefficient indicated a lower level of consistency.
The experts did the ranking exercise twice. During the first iteration, the experts did not collaborate when they ranked the sites. Once all sites were ranked, the experts explained their ranking rationale and established descriptive consensus principles in ranking the sites. They then ranked again all 26 sites based on the consensus principles.
Results indicated a high level of inconsistency among experts, even when they were following the same set of consensus principles for ranking. Figure 1 shows the correlation coefficients among experts ranking the 26 sites. The correlation coefficient ranged from -0.33 to 0.97 (average 0.41) among experts during iteration one, when the experts never discussed the project. The negative correlation indicated a high level of inconsistency. When the experts reached a common set of principles on how to rank the sites, the correlation improved but still ranged from negative to positive (–0.14 to 0.98, average 0.47).
A New Approach
Following the findings from the BPJ study, a team of researchers from the U.S. EPA, Stanford University and Southern California Coastal Water Research Project Authority sought to develop an objective, mathematically defined standardized approach to interpret qPCR MST data. The researchers developed an algorithm based on a novel Bayesian weighted average approach that produces an estimate of the level of fecal contamination at a given site based on the average concentration of human fecal-associated marker called HF183 in water samples collected over a defined period of time.
The Human Fecal Score (HFS) is a number indicating the extent of human fecal pollution at a site. The higher the score, the more polluted the site is. This significantly improves communication of MST data to non-experts and allows for easy comparison among sites. HFS also uses all data, regardless if it is ND, DNQ or quantifiable data. As all data contains information about the site, this allows a more accurate depiction of the site condition without the bias introduced by ignoring ND, DNQ or arbitrary data substitution. Finally, because HFS is mathematically defined instead of descriptive, its implementation would not be standardized.
Bacterial pollution abatement works best when comparisons are made between different data from different projects or different data from the same project over time. With HFS, data is accurately compared to each other in various water quality management applications.