%0 PDF
%T An Adjoint-Sensitivity-Analysis Based
Mathematical Framework: DNAPL Source Zone Characterization, Uncertainty Quantification,
and Sampling Strategy Design.
%A Tang, Tian.
%D 2019-04-24T09:17:33.431-04:00
%8 2019-04-24
%I Tufts University. Tisch Library.
%R http://localhost/files/1831cx72r
%X Subsurface
contamination by dense non-aqueous phase liquids (DNAPLs) is a continuing societal
concern due to the widespread use and improper disposal of these compounds. Due to the
heterogeneous nature of geological formations, which leads to a high degree of spatial
variability in DNAPL migration and entrapment, delineation of the distribution of source
zone mass and the associated downgradient plume presents significant challenges. This
research focused on the local estimation of source zone mass distribution metrics from
in situ Push-Pull Tracer Tests (PPTTs) and on the optimization of source zone
characterization for predictions of future plume persistence and risk to down gradient
receptors. An adjoint sensitivity analysis- and inverse theory-based mathematical model
is developed and implemented to estimate domain-averaged source zone properties (total
mass, average DNAPL saturation, and mass weighted distance from the test well) using
PPTT data, accounting for both rate-limited mass transfer and heterogeneity in both the
flow field and source zone mass distribution. This model is demonstrated to provide
reliable estimates of source zone metrics with accuracy better than 20%. Estimation
error of domain-averaged source zone metrics is shown to increase with increased
heterogeneity of the permeability field and DNAPL saturation distribution. Comparison of
the presented estimation approach with that based upon a local equilibrium partitioning
assumption demonstrates the superiority of the new approach, with equilibrium-based
estimates resulting in overestimation of the average saturation. In the second portion
of this work, an adjoint sensitivity analysis and first-order second-moment based source
zone characterization model, honoring borehole observations, is designed and validated
for prediction of downgradient flux-averaged concentration (FAC) and its variance at
distinct times. Adjoint state theory is applied to quantify the importance of local
system properties (e.g., permeability field, initial contaminant mass in aqueous,
sorbed, and DNAPL phases) on FAC. By considering the process coupling of DNAPL
dissolution, sorption, dispersion, and mass transport, results reveal that local
permeability has the greatest impact on FAC predictions. Data worth analyses are
performed to guide optimal sampling strategy to provide more accurate FAC predictions
and to maximize uncertainty reduction. It is demonstrated that areas of low permeability
and high DNAPL saturation have the largest utility for uncertainty reduction and that
optimal borehole observation locations vary with the prediction time window. The
FOSM-based optimal sampling patterns are demonstrated to outperform alternative sampling
approaches for most of the prediction goals, yielding more precise estimates and better
confidence intervals.; Thesis (Ph.D.)--Tufts
University, 2019.; Submitted to the Dept. of Civil
Engineering.; Advisor: Linda
Abriola.; Committee: Laurie Baise, John Christ, Eric
Miller, and Kurt Pennell.; Keyword: Environmental
engineering.
%[ 2022-10-11
%9 Text
%~ Tufts Digital Library
%W Institution