Amir Abdollahian, Mahsa (2016) Development of integrated methodologies for optimal monitoring and source characterization in contaminated groundwater systems under uncertainty. PhD thesis, James Cook University.

Preview

PDF (Thesis) Download (3MB) | Preview

Abstract

Groundwater is a major source of water in many parts of the world. Due to anthropogenic activities it is subjected to various sources of contamination. Effective groundwater pollution management and remediation relies on accurate identification of unknown pollution source characteristics. The pollution source should be defined in terms of location, flux magnitude, and time of release. The source identification procedure is a challenging task due to uncertainties in model definition, parameter estimation, hydrogeologic parameters and field measurements. Availability of adequate data is vital due to the nature of the contaminant source identification procedure; however, acquiring accurate and extensive field measurement data is a very costly and time intensive task. The source identification procedure remains a challenging task due to sparse measurement data, uncertainties in model definition, parameter estimation, and field measurements. Therefore the source identification problem is often characterized by very little information, and is considered an ill-posed, complex, and non-unique inverse problem.

In the recent past, the contaminant source identification problem has been addressed using different approaches. The linked simulation-optimization approach is capable of incorporating complex real-life scenarios and large study areas. In this approach, the numerical flow and transport simulation models are linked to an optimization model. A typical optimal source identification model minimizes the difference between estimated pollutant concentration and measured contamination values at monitoring locations. Due to the complex nature of the contaminant source identification process, various issues are needed to be addressed in the proposed methodologies in order to achieve an acceptable level of accuracy in recovering contaminant source histories. This study focuses on improving the accuracy and efficiency of contaminant source identification procedures in the presence of hydrogeologic parameter and measurement, uncertainties. This study specifically focuses on the following issues:

i. The utilized simulation model should include an accurate description of the study area in terms of hydrogeologic parameters. The uncertainty in hydrogeologic parameter values must be quantified and incorporated into the contaminant source identification methodology.

ii. Often the recorded concentration measurements are erroneous. The measurement error/uncertainty is required to be explicitly incorporated into the optimal source identification objective functions.

iii. Selection of the monitoring locations for measuring concentrations is vital for accurate source identification, and efficient selection of these locations can increase the accuracy of recorded source histories. Therefore, an optimal monitoring network is required to be designed, intended to reduce uncertainty and increase efficiency in recovering source release histories.

This study focuses on improving the accuracy and efficiency of contaminant source identification/characterization procedures in the presence of hydrogeologic parameter, and measurement, uncertainties. The main features of this study are:

i) The hydrogeological parameter uncertainty is incorporated in the optimal source identification methodology, using a new parameter for uncertainty quantification based on various realizations of the flow field. The new parameter, called the Coefficient of Confidence (COC), is estimated for each available contaminant observation data and incorporated into the optimal source identification model. Although such quantification cannot be designated as fuzzy quantification, in a strict sense, it differs from the crisp approach generally adopted in earlier developed methodologies for source characterization. This approach can actually eliminate the possibility of incorporating additional uncertainties due to inappropriate use of so called expert judgments. Adaptive simulated Annealing (ASA) is used as the optimization algorithm.

ii) A two-objective contaminant source identification model is developed which can improve the accuracy of recovered source histories using erroneous contaminant measurement data. Non-dominated Sorting Genetic Algorithm (NSGA) II is used as the optimization algorithm.

iii) A new optimal monitoring network design methodology is developed using redundancy reduction and uncertainty reduction objective functions. This new two-objective monitoring network design model is solved using NSGA-II and integrated into the source identification methodology by sequential implementation of the optimal monitoring network design, and solution of the optimal source identification model, in order to increase the efficiency of both methodologies.

This study includes the following steps. An ASA based simulation-optimization source identification model is developed which is externally linked with numerical flow and transport simulation models. To address the hydrogeologic parameter uncertainty in the source identification model, an uncertainty quantification method is developed. The hydraulic conductivity uncertainty is quantified using a quantification parameter based on various realizations of the flow field, and using a new parameter called the Coefficient of Confidence (COC). Incorporating the estimated COC values in the contaminant source identification methodology results in the uncertainty-based linked simulation optimization model for contaminant source identification. The performance of this proposed methodology is evaluated for both illustrative and experimental contaminated aquifers. Obtained solution results show that the proposed methodology is capable of recovering source release histories more accurately compared with those obtained using earlier developed crisp methodologies in which the hydrogeologic parameter uncertainty is not considered explicitly.

In the next stage of this thesis, a two-objective optimal source identification methodology is developed which focuses on measurement error/uncertainty. Twoobjective optimal contaminant source identification models can improve the accuracy of estimating the recovered source histories using erroneous contaminant concentration measurement data. NSGA-II is linked to the flow and transport simulation models to determine the contaminant source characteristics.

When the measurements are erroneous, the source identification problem becomes non-unique. Therefore, various solutions with a possibility of being the true source characteristics may be achieved. When the contaminant concentrations are erroneous, it is not possible to match all the observed and simulated concentrations. In the two-objective approach, the first objective function is normalized using observed concentrations. Therefore, this objective function emphasizes matching smaller observed concentrations (for larger objective function improvement); however, the second objective function, which is not normalized, tries to match the high concentrations. Therefore, the two-objective approach generally finds the possible solutions as a Pareto-front.

The performance of the developed methodology using two different objectives for optimal source characterization is evaluated for an illustrative study area considering both point and distributed contaminant sources. The obtained solution results demonstrate that the two-objective model is capable of finding more accurate source characteristics in the presence of measurement error, compared with those obtained using each objective function separately in a single-objective model.

In the next step, a new sequential monitoring network design methodology is developed which selects new monitoring locations based on sequential characterization of pollutant sources and feeds back new measurement information from the newly designed and implemented monitoring network. Integrated source identification and monitoring network design sequences are carried out to reach an acceptable level of accuracy in characterization of the contaminant source properties. A new two-objective monitoring network design model is proposed which minimizes the uncertainty in recovered source histories and minimizes the redundancy in measured contaminant concentrations.

The performance of the integrated sequential contaminant source identification and monitoring network design methodologies are evaluated using illustrative data, and data from real-life contaminated groundwater aquifers. Performance evaluation results show that the developed methodologies can improve the accuracy of recovered contaminant source histories in the presence of uncertainty, and provides insight into the reliability of recovered source histories. The new sequential monitoring network design methodology can reduce the cost and time required to achieve relatively accurate characterization of polluted aquifers.

Item ID:	46026
Item Type:	Thesis (PhD)
Keywords:	contaminants; contamination; groundwater; hydrogeology; identification; methodologies; network design; pollution; source of contamination; source of pollution; uncertainty
Date Deposited:	13 Oct 2016 00:19
FoR Codes:	09 ENGINEERING > 0905 Civil Engineering > 090509 Water Resources Engineering @ 100%
SEO Codes:	96 ENVIRONMENT > 9609 Land and Water Management > 960999 Land and Water Management of Environments not elsewhere classified @ 100%
Downloads:	Total: 189 Last 12 Months: 7
	More Statistics