This dissertation presents an examination of the causes and mechanisms underlying the widespread contamination of a shallow groundwater aquifer by fecal bacteria. The context for this study is a field site located in a rural area of Bangladesh that represents a microcosm for the many challenges facing the approximately 2 billion people worldwide who rely upon groundwater for their daily needs. The unique contributions of this work include an improved numerical model for fitting column test results, a conceptual model to explain seasonal patterns of well contamination based on the hydraulic interaction of ponds and irrigation/drainage canals and a new understanding of the important role that such canals might play in predicting the microbial contamination of shallow aquifers in flood-protected areas. The mechanisms responsible for filtration of the fecal indicator bacteria, Escherichia coli, during passage through the fine sand aquifer were first investigated through laboratory column experiments using intact sediment cores from the field site as well as repacked sediment that had been dried and, in some experiments, chemically cleaned. To fit the hyper-exponential spatial profiles of attached bacteria in one third of the experiments, a finite difference two-population model with reversible and irreversible attachment modes incorporating bacterial die-off was developed. Where the two-population phenomenon was observed, one population typically was highly irreversible while the other was reversible with a smaller irreversible attachment rate. When applied to transport in the field, this model predicted only a two-fold reduction in bacterial concentrations over a distance of 10 m and transport was limited mainly by the bacterial die-off rate, which was also measured using microcosm experiments. The occurrence of the second population was associated with larger grain size and lower percentage of fine particles and the attachment rates in general increased linearly with increasing percentage of fines. Transport from contaminated surface water to nearby tubewells was studied in the field through measurements of bacterial infiltration below canals and ponds both inside and outside of the flood control embankment. A two-dimensional finite element model of the field-pond-canal system was built and fitted to heads measured at three monitoring wells and 2 surface water bodies. Using parameters from the field measurements, the model was not able to explain the seasonal pattern of E. coli concentrations in tubewells, even when reversible attachment assumptions from the column test results were applied. An alternative conceptual model that incorporates the seasonal shift in flow direction caused by the canal network was developed using the fitted finite element model and could explain the observed pattern of well contamination. The importance of the irrigation/drainage canals in determining the frequency of tubewell contamination by E. coli at the site was further demonstrated by applying a logistic regression model using the intensity of latrines, canals and ponds as predictors, after applying spatial decay rates drawn from the infiltration literature. The resulting Intensity Model found that population density, unsanitary latrines and canals together could explain 48% of the variation in the frequency of E. coli detection in tubewells, but these parameters were only significant at a low spatial decay rate (0.01 m-1). A less complex Proximity Model provides nearly the same explanatory power but only required population with 25 m and the distance to the nearest canal as predictors. These models could be useful in predicting water-related health risks, evaluating contamination risk for groundwater sources based on the sanitary environment around the well or estimating the potential benefits from improvements to sanitation infrastructure in a given region.
Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/D8J1012N |
Date | January 2013 |
Creators | Feighery, John E. |
Source Sets | Columbia University |
Language | English |
Detected Language | English |
Type | Theses |
Page generated in 0.1499 seconds