Pathogen contamination, typically quantified by elevated levels of fecal indicator bacteria (FIB), remains the leading cause of surface water-quality impairments in the United States. Continuous watershed-scale models are typically employed to facilitate Total Maximum Daily Load (TMDL) restoration efforts. Due to limited understanding of microbial fate and transport, predictions of FIB concentrations are associated with considerable uncertainty relative to other water-quality contaminants. By focusing on a data-rich instrumented urban watershed, this study aims to improve understanding of microbial fate and transport processes. Weekly FIB concentrations in both the water column and streambed sediments were monitored for one year, and statistical correlations with hydrometeorological and physicochemical variables were identified. An intensive six storm intra-sampling campaign quantified and contrasted loading trends of both traditional regulatory FIB and emerging Microbial Source Tracking (MST) markers. Together, these intensive monitoring efforts facilitated evaluation of the impacts of bacteria-sediment interactions on the predictions of daily FIB concentrations in Hydrological Simulation Program-Fortran (HSPF) over multiple years. While superior overall model performance was demonstrated as compared to earlier efforts, the inclusion of bacteria-sediment interactions did not improve performance. Large wet-weather microbial loading appears to have dwarfed the effects of FIB release and resuspension from sediment. Although wet-weather loading is generally considered as a primary source of waterbody microbial loads, dry-weather periods are more directly associated with public health concern, which may be a more suitable area for future model-refinement efforts. Site evaluation is critical to determine whether the added model complexity and effort associated with partitioning phases of FIB can be sufficiently offset by gains in predictive capacity. Finally, a stochastic framework to translate simulated daily FIB concentrations into estimates of human illness risks is presented that can be can be readily integrated into existing TMDLs. As even small concentrations of FIB from human sources are associated with great risk, and monitoring efforts indicated moderate/high levels of human-associated MST marker in this watershed, remediation efforts to protect public health would be best directed toward infrastructure improvements. Uncertainty analysis indicates more site-specific knowledge of pathogen presence and densities would best improve the estimation of illness risks. / Ph. D.

E. coli and Enterococcus were used to determine the fate and survival of fecal indicator bacteria (FIB) in sand and sediments. The microbial source tracking (MST) methods antibiotic resistance analysis (ARA), Bacteroides human-specific primer test, and fluorometry were compared against the FIBs to determine how reliable each method was in detecting the presence of human fecal contamination. Two phases (Summer 2009 and 2010) were evaluated based on the type of contamination event. A combined sewage overflow (CSO) event was simulated in Phase I, where large amounts of influent were added to sand and bay water columns over 1 to 4 days. In 2010, a low volume sewage leak was simulated in which smaller doses of influent were added to sand and bay water columns over a period of 5 to 15 days. Within each of the phases, both non- and re-circulated columns were also evaluated. Evaluation of FIB survival indicated that Enterococcus was able to stabilize and re-grow in the water and at the sediment/water interface within the Phase I non-circulated columns. E. coli was unable to re-grow and/or stabilize within any environment. Comparisons between the ARA and the FIBs revealed a large majority of isolates identified as coming from either bird or wildlife sources. Human sources were identified but at much lower concentrations than expected. Bacteroides results indicated strong relationships between the increase of FIB concentrations and the presence of the human-specific Bacteroides. Fluorometry results did not indicate any relationship with the FIBs. Unexpectedly, fluorometry readings increased as time progressed indicating that another compound was present that fluoresced at the same wavelength as optical brighteners (OBs). This project was one of the first to study the differences related to two different pollution events (CSO vs. sewage leak) while also evaluating what happens to pollution as it settles into the sediment. It was also unique because it compared bacterial (ARA), molecular (Bacteroides), and chemical (fluorometry) MST methods. / Master of Science

Between 1970 and 2010 almost one-third of drinking water related waterborne disease outbreaks reported to the US Centers for Disease Control and Prevention were associated with systems dependent on untreated groundwater (i.e., most commonly, household wells). This is unsurprising, given that numerous past efforts to monitor household well water quality have indicated a high prevalence of fecal coliforms and/or E. coli at the point of use. Non-point sources of pollution, including septic tank leakages and poorly constructed drain fields, have been identified as the leading risk factors associated with outbreaks in households dependent on groundwater. Ideally, the integration of emerging source tracking (ST) analyses in well monitoring programs could be used to identify whether the presence of fecal indicator bacteria (FIB) is associated with human or non-human sources in order to inform remediation strategies. However, the application of ST to groundwater has been limited, and the interpretation of data is consequently difficult. This research compares the fate and transport of FIB (E. coli and enterococci) with a chemical (optical brighteners, OB) and a molecular (Bacteroides HF183) ST target in order to evaluate their potential use as indicators of water quality issues in private drinking water systems. Eighteen PVC soil columns were constructed in an outdoor soil column facility to represent small-scale septic drainfield models; they received synchronized doses of primary-treated wastewater twice daily and were monitored bi-weekly over a 7-month period. Columns were subject to variable influent loading rates of wastewater effluent, and differing degrees of soil compromisation (i.e. synthetic solution channels). Results show that while column effluent volume and constituent levels were related to dosage, they were not always related to soil compromisation (ANOVA, p < 0.05). E. coli and enterococci concentrations were associated with effluent volume and OB levels (Spearman's rank, p < 0.05). The presence of Bacteroides HF183 was not strongly associated with the other measured ST target levels (Point-biserial correlation, p < 0.05). Findings from this study suggest surface water ST methodologies may have a role in groundwater quality monitoring efforts. Quantifying the relative recovery of ST targets and FIB from controlled groundwater simulations will assist in the development of strategies to identify non-point sources of human wastewater pollution efficiently and effectively to inform remediation. / Master of Science

Over 1.7 million Virginians rely on private water systems to supply household water. The heaviest reliance on these systems occurs in rural areas, which are often underserved in terms of financial resources and access to environmental health education. As the Safe Drinking Water Act (SDWA) does not regulate private water systems, it is the sole responsibility of the homeowner to maintain and monitor these systems. Previous limited studies indicate that microbial contamination of drinking water from private wells and springs is far from uncommon, ranging from 10% to 68%, depending on type of organism and geological region. With the exception of one thirty-year old government study on rural water supplies, there have been no documented investigations of links between private system water contamination and household demographic characteristics, making the design of effective public health interventions, very difficult. The goal of the present study is to identify potential associations between concentrations of fecal indicator bacteria (e.g. coliforms, E. coli) in 831 samples collected at the point-of-use in homes with private water supply systems and homeowner-provided demographic data (e.g. homeowner age, household income, education, water quality perception). Household income and the education of the perceived head of household were determined to have an association with bacteria concentrations. However, when a model was developed to evaluate strong associations between total coliform presence and potential predictors, no demographic parameters were deemed significant enough to be included in the final model. Of the 831 samples tested, 349 (42%) of samples tested positive for total coliform and 55 (6.6%) tested positive for E. coli contamination. Chemical and microbial source tracking efforts using fluorometry and qPCR suggested possible E. coli contamination from human septage in 21 cases.  The findings of this research can ultimately aid in determining effective strategies for public health intervention and gain a better understanding of interactions between demographic data and private system water quality. / Master of Science

The microbial source tracking (MST) methods of antibiotic resistance analysis (ARA) and fluorometry (to detect optical brighteners in detergents) were used in the summers of 2004 and 2005 to determine the origins of fecal pollution at beaches with a past history of, or the potential for, high enterococci counts and posted advisories. At Hilton and Anderson beaches, ARA and fluorometry in the summer of 2004 detected substantial human-origin pollution in locations producing consistently high counts of Enterococcus spp. Investigations by municipal officials led to the fluorometric detection and subsequent repair of sewage infrastructure problems at both beaches. The success of these mitigation efforts was confirmed during the summer of 2005 using ARA and fluorometry, with the results cross-validated by pulsed-field gel electrophoresis (PFGE). Results at other beaches indicated that birds and/or wildlife were largely responsible for elevated enterococci levels during 2004 and 2005. The application of fluorometry proved difficult in opens waters due to high levels of dilution, but showed potential for use in storm drains. An additional study developed and tested a new library-based MST approach based on the pattern of DNA band lengths produced by the amplification of the 16S-23S rDNA intergenic spacer region, and subsequent digestion using the restriction endonuclease MboI. Initial results from small known-source libraries yielded high average rates of correct classification (ARCC). However, an increase in the library size was accompanied by a reduction in the ARCC of the library and the method was deemed unsuccessful, and unsuitable for field application. A final study focused on the potential for classification bias with disproportionate source category sizes using discriminant analysis (DA), logistic regression (LR), and k-nearest neighbor (K-NN) statistical classification algorithms. Findings indicated that DA was the most robust algorithm for use with source category imbalance when measuring both correct and incorrect classification rates. Conversely k-NN was identified as the most sensitive algorithm to imbalances with the greatest levels of distortion obtained from the highest k values. Conclusions of this project include: 1) application of a validation set, as well as a minimum detectable percentage to known-source libraries aids in accurately assessing the classification power of the library and reducing the false positive identification of contributing fecal sources; 2) the validation of MST results using multiple methods is recommended for field applications; 3) fluorometry displayed potential for detecting optical brighteners as indicators of sewage leaks in storm drains; 4) the digestion of the 16S-23S rDNA intergenic spacer region of Enterococcus spp. using MboI does not provided suitable discriminatory power for use as an MST method; and 5) DA was the least, and k-NN the most, sensitive algorithm to imbalances in the size of source categories in a known-source library. / Ph. D.

Prince William County, located in the rapidly developing Northern Virginia region, contains watersheds of mixed rural and urban/suburban uses. As part of Virginia regulations, recreational waters must be tested and remain under a certain standard for levels of fecal indicator bacteria (FIB). The sources of fecal pollution in neighboring watersheds within the county were determined over the 12 months previous to this project by performing Antibiotic Resistance Analysis (ARA, a microbial source tracking protocol) on Enterococcus and Escherichia coli (E. coli). This study indicated that multiple sources of pollution were present at all sampling locations and that the dominant sources of contamination were related to the land-use patterns and human activities that were adjacent to each location. The goal of the current project was to monitor and identify the sources of fecal pollution in eight streams in the Occoquan Basin (OQB) that have been classified as impaired waters due to high E. coli concentrations. Project objectives were i) employ microbial source tracking technology to identify the categories of sources that were responsible for the bacterial impairments; ii) develop and analyze appropriate Known Source Libraries (KSL's) to determine the best design for identifying the sources of water-sample isolates; and iii) evaluate the use of optical brighteners in freshwater by fluorometry as an indicator for human-origin pollution. One site on each of six streams and two sites on the remaining two (ten total) were selected for E. coli and Enterococcus monitoring and microbial source tracking. Repeated sampling of the ten locations for thirteen months assessed the concentrations of the bacteria over time, while comparison of monthly bacterial concentrations to the U.S. standards was used to verify the impaired water designation. Three thousand, four hundred and eighty-eight Enterococcus and 969 E. coli water-sample isolates were collected and evaluated to determine their sources. These isolates were compared to several known source libraries (KSL's) comprised of host-origin isolates collected from the Northern Virginia region. Linear discriminant analysis (LDA) using a KSL of unique isolates determined wildlife were the dominant source of fecal pollution. Results based on ARA were cross-validated through fluorometry of the water samples (to detect optical brighteners in detergents as human-derived pollution) and pulsed-field gel electrophoresis (PFGE, a DNA fingerprinting technique) of select E. coli isolates. In order to determine the best method to classify the water-sample isolates, variation in antibiotic resistance data representation, known source isolate inclusion, and LDA processing were compared. The KSL that used the most antibiotic resistance datapoints, contained no conflicting data, and performed most of the parameters associated with standard LDA, classified water-sample isolates the most successfully. This project involves the first thorough testing of fluorometry for the detection of human signatures in freshwaters. Monitoring results showed consistent Enterococcus and E. coli contamination in all eight streams, demonstrating that each had been correctly placed on Virginia's impaired waters list by state regulatory agencies. Counts between Enterococcus and E. coli did not correlate well, although concentrations of both indicator organisms were higher during dry months. Source tracking results determined a dominant wildlife signature at all sites. Few Enterococcus water-source isolates were classified as human and fluorescence at all sites was consistently low. KSL's with antibiotic resistance data represented as binary values classified isolates the best. Removal of conflicting isolates improved the KSL's rate of correct classification (RCC). Creation of an unknown category, clustering of the KSL, and only accepting results above a threshold did not appreciably improve the RCC. The KSL with the binary representation was not used to classify isolates because it violates the normal distribution assumption of LDA. Differences in the results of Enterococcus and E. coli source classifications indicated that contributing sources vary in frequency. Human fecal matter was shown to be of little concern because both Enterococcus ARA and fluorometry indicated low presence. The positive predictive value (PPV) statistic was found to be preferable to the minimum detectable percentage (MDP) because it does not depend on KSL size. Establishing confidence intervals to determine completeness of KSL allows one to determine whether particular methods to refine the KSL will be helpful. This project was successfully completed and the monitored streams were correctly identified by state authorities as impaired waters. Source tracking results often conflicted, although wildlife and pets were indicated as the major sources of impairment by ARA. More local source samples need to be taken to verify this result. The best ARA library design used only unique isolates, all pattern data points, and removed conflicting isolates. Continuing examination of the representation of library data as binary is necessary to determine whether the statistical assumptions in LDA prevent meaningful results. Evaluation of fluorometry was partially successful as the absence of "hotspots" of high fluorescent brighteners agreed with ARA results that indicated little contamination form human sources. The fluorometer continues to have potential as a metric of waste in freshwater although more work needs to be done to fully prove its utility. / Master of Science

Prince William County, located in the rapidly developing Northern Virginia region, contains watersheds of mixed rural and urban/suburban uses. The project goal was to monitor and evaluate 21 stream locations, over 13 months, in the Occoquan Basin identified as impaired due to high E. coli densities. One site on each of eight streams, two sites on each of five streams, and three sites on the remaining stream were chosen for E. coli monitoring and microbial source tracking (MST). MST was performed using antibiotic resistance analysis (ARA) and fluorometric analysis. Escherichia coli was chosen as the indicator bacterium for purposes of comparison with previous project data and because a large body of evidence supports its use in freshwater systems. This study involved the only known MST project to incorporate data from five or more consecutive years. A total of 2854 environmental isolates were collected for analysis with ARA. These isolates were classified using a known source library (KSL) that consisted of 1003 unique resistance patterns. The resistance patterns of the KSL came from known fecal sources (human, pets, livestock, wildlife) in Prince William County. The KSL included isolates from previous years but was also updated with fresh isolates. The accuracy of the KSL was assessed through the use of a challenge set. The challenge set was classified against the KSL using discriminant analysis, verified by logistic regression. The average rate of correct classification was 93% for discriminant analysis and 96% for logistic regression. Results indicated that multiple sources of contamination were present at all sampling locations and that the major source(s) (human, pets, livestock, wildlife) of contamination were generally related to the land-use patterns and human activities at each location. Although no major or minor human signatures were found, all but two locations had either pet or livestock as the major signature, suggesting that human-related activities are playing a key role in contamination of the streams. Pets were the single most frequent major signature and wildlife was the most common minor signature. Fluorometric analysis was used to corroborate human-derived contamination. Fluorometric analysis has the ability to detect the presence of optical brighteners, synthetic compounds added to such household items as laundry detergent, dishwashing detergent and other washing agents. Despite having an undesirably high rate of false negatives (negative fluorometry readings not supported by ARA), fluorometric analysis maintained a low rate of false positives (positive fluorometry readings not supported by ARA) and continued to demonstrate its potential for source tracking. This project represented one of the first attempts at applying a full suite of performance criteria now recommended by the source tracking community for all MST projects. Such concepts as experimental design, toolbox approach, minimum detectable percentage, quantification, accuracy, specificity, robustness, range of applicability, and practicality were successfully incorporated. These performance criteria have in effect set a new standard to which all subsequent MST projects should adhere. / Master of Science

With the advent of computers and the information age, vast amounts of data generated in a great deal of science and industry fields require the statisticians to explore further. In particular, statistical and computational problems in biology and medicine have created a new field of bioinformatics, which is attracting more and more statisticians, computer scientists, and biologists. Several procedures have been developed for tracing the source of fecal pollution in water resources based on certain characteristics of certain microorganisms. Use of this collection of techniques has been termed microbial source tracking (MST). Most of the current methods for MST are based on patterns of either phenotypic or genotypic variation in indicator organisms. Studies also suggested that patterns of genotypic variation might be more reliable due to their less association with environmental factors than those of phenotypic variation. Among the genotypic methods for source tracking, fingerprinting via rep-PCR is most common. Thus, identifying the specific pollution sources in contaminated waters based on rep-PCR fingerprinting techniques, viewed as a classification problem, has become an increasingly popular research topic in bioinformatics. In the project, several statistical methods for classification were studied, including linear discriminant analysis, quadratic discriminant analysis, logistic regression, and $k$-nearest-neighbor rules, neural networks and support vector machine. This project report summaries each of these methods and relevant statistical theory. In addition, an application of these methods to a particular set of MST data is presented and comparisons are made.

Auditory stream denotes the abstract effect a source creates in the mind of the listener. An auditory scene consists of many streams, which the listener uses to analyze and understand the environment. Computer analyses that attempt to mimic human analysis of a scene must first perform Audio Scene Segmentation (ASS). ASS find applications in surveillance, automatic speech recognition and human computer interfaces. Microphone arrays can be employed for extracting streams corresponding to spatially separated sources. However, when a source moves to a new location during a period of silence, such a system loses track of the source. This results in multiple spatially localized streams for the same source. This thesis proposes to identify local streams associated with the same source using auditory features extracted from the beamformed signal. ASS using the spatial cues is first performed. Then auditory features are extracted and segments are linked together based on similarity of the feature vector. An experiment was carried out with two simultaneous speakers. A classifier is used to classify the localized streams as belonging to one speaker or the other. The best performance was achieved when pitch appended with Gammatone Frequency Cepstral Coefficeints (GFCC) was used as the feature vector. An accuracy of 96.2% was achieved.

The overall goal of this thesis is to develop non-linear Bayesian methods for three-dimensional tracking of a moving acoustic source in shallow water despite environmental uncertainty, with application to data from a horizontal line array (HLA) of hydrophones. As a precursor, Bayesian geoacoustic inversion is applied to estimate seabed model parameters and their uncertainties. A simulation study examines the effect of source and array factors on geoacoustic information content in matched-field inversion of HLA data, as quantified in terms of model parameter uncertainties. Bayesian geoacoustic inversion is applied to both controlled-source and ship-noise data from a HLA deployed on the seafloor in a shallow-water experiment conducted in the Barents Sea. A new approach is introduced to account for data error reduction due to averaging data over time-series subsegments (snapshots), based on empirically apportioning measurement and theory error, with effects on inversion results compared to those of existing approaches. It is further demonstrated that combining data from multiple, independent time-series segments (for a moving source) in the inversion can significantly reduce geoacoustic parameter uncertainties. Geoacoustic uncertainties are also shown to depend on ship range and orientation, with lowest uncertainties for short ranges and for the ship stern/propeller oriented toward the array. Sediment sound-speed profile and density estimates from controlled-source and ship-noise data inversions are found to be in good agreement with values from geophysical measurements. Two non-linear Bayesian matched-field inversion approaches are developed for three-dimensional source tracking despite environmental uncertainty. Focalization-tracking maximizes the posterior probability density (PPD) over track and environmental parameters. Synthetic test cases show that the algorithm substantially outperforms tracking with poor environmental estimates and generally obtains results close to those achieved with exact environmental knowledge. Marginalization-tracking integrates the PPD over environmental parameters to obtain joint marginal distributions over source coordinates, from which track uncertainty estimates and the most probable track are extracted. Both approaches are applied to data from the Barents Sea experiment. Focalization-tracking successfully estimates the tracks of the towed source and a surface ship in cases where simpler tracking algorithms fail. Marginalization-tracking generally outperforms focalization-tracking and gives uncertainty estimates that encompass the true tracks.