| dc.description.abstract | Numerical models of flow and transport are commonly applied for the sustainable management of water resources and for the selection of appropriate remediation techniques. However, these numerical models are not always accurate due to uncertain parameters and the disparity of scales across which observations are made, hydrological processes occur, and modeling is conducted. The modeling framework becomes further complex because hydrologic processes are coupled with chemical and biological processes. This dissertation focuses on the most widespread contaminants of surface and ground water, which are E. coli and nitrate, respectively. Therefore, this research investigates the linkages between bio-chemical and hydrologic processes for E. coli transport, explores the spatio-temporal variability of nitrate, quantifies uncertainty, and develops models for both E. coli and nitrate transport that better characterize these biogeochemical linkages.
A probabilistic framework in the form of Bayesian Neural Networks (BNN) was used to estimate E. coli loads in surface streams and was compared with a conventional model LOADEST. This probabilistic framework is crucial when water quality data are scarce, and most models require a large number of mechanistic parameters to estimate E. coli concentrations. Results indicate that BNN provides better characterization of E. coli at higher loadings. Results also provide the physical, chemical, and biological factors that are critical in the estimation of E. coli concentrations in Plum Creek, Texas.
To explore model parameters that control the transport of E. coli in the groundwater (GW) and surface water systems, research was conducted in Lake Granbury, Texas. Results highlight the importance of flow regimes and seasonal variability on E. coli transport.
To explore the spatio-temporal variability of nitrate across the Trinity and Ogallala aquifers in Texas, an entropy-based method and a numerical study were employed. Results indicate that the overall mean nitrate-N has declined from 1940 to 2008 in the Trinity Aquifer as opposed to an increase in the Ogallala Aquifer. The numerical study results demonstrate the effect of different factors like GW pumping, flow parameters, hydrogeology of the site at multiple spatial scales.
To quantify the uncertainty of nitrate transport in GW, an ensemble Kalman filter was used in combination with the MODFLOW-MT3DMS models. Results indicate that the EnKF notably improves the estimation of nitrate-N concentrations in GW.
A conceptual modeling framework with deterministic physical processes and stochastic bio-chemical processes was devised to independently model E. coli and nitrate transport in the subsurface. Results indicate that model structural uncertainty provides useful insights to modeling E. coli and nitrate transport. | en |
