Date of Graduation

2018

Document Type

Thesis

Degree Type

MS

College

Statler College of Engineering and Mineral Resources

Department

Civil and Environmental Engineering

Committee Chair

Leslie C Hopkinson

Committee Co-Chair

Seungho Hong

Committee Member

Antarpreet S Jutla

Abstract

There has been an extraordinary growth in the production of natural gas within the Marcellus Shale region in the past two decades. This sharp increase in production corresponds to an increased risk to surface waters from spill events related to the industry. At present, there are some private organizations and government agencies which monitor water quality in the region using continuous monitoring. However, there are no protocols in place to use data from continuous monitoring stations to quickly detect contamination resulting from spills.;Previous research has found that standard water quality parameters, such as specific conductivity, pH, and dissolved oxygen, measured by "off the shelf" water quality sensors are affected by commonly spilled materials from the natural gas industry. However, it was also found that due to expected low concentrations at monitoring stations, the effects of pollutants on these parameters are not strong enough to use simple threshold value methods to detect contaminants. Thus, there is a need for detection of subtle changes in water quality data which are not congruent with background behavior. The overall objective of this work was to add to the understanding of event detection related to surface spill events resulting from the development of Marcellus Shale.;Municipal water distribution networks commonly deploy event detection systems (EDS) to detect accidental or malicious contamination. The application of EDS in natural systems is less common. In this study, a publically available EDS, "CANARY", was applied to water quality data from natural channels to evaluate its ability to detect spill events and its overall performance when applied to a natural system.;Two approaches were used to test event detection capabilities: historical events and a simulated event. First, water quality data from three streams in the Susquehanna River Basin Commission (SRBC) remote water quality monitoring network (RWQMN) which had a spill event (i.e., production water, flowback fluid, and drilling mud) in the watershed were analyzed using CANARY. Then, a contamination event was imposed on water quality data from one uncontaminated stream and analyzed. For spills of flowback and production fluid, events were detected during the week following the spills. However, the algorithms detecting these events produced too many false alarms throughout the data set to confidently attribute these events to the spills. For the drilling mud spill, the only event detected was the result of a sensor malfunction. Therefore, the historical events were not detected. The simulated flowback fluid spill was also not detected. It was therefore concluded that the contaminant concentrations at the sensors were too low for detection. This further demonstrates the need for more extensive water quality monitoring networks in watersheds of concern. Decreasing the distance from spill events to sensors, and optimizing configuration parameters could provide a viable protocol for detecting acute contamination events resulting from natural gas development in the Marcellus Shale.

Download

Share

COinS