Graduation Semester and Year
Summer 2026
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Computer Science
Department
Computer Science and Engineering
First Advisor
Yonghe Liu
Second Advisor
Hao Che
Third Advisor
David Levine
Fourth Advisor
Jia Rao
Abstract
Coastal communities face increasing environmental and infrastructural risks associated with extreme weather, flooding, industrial activity, shipping, and rapid development. Continuous monitoring of air and water quality is therefore important for environmental management, community planning, and public safety. Conventional monitoring systems provide high-quality observations but are often expensive to deploy densely and difficult to maintain at remote or harsh coastal sites. Low-cost sensing and Internet of Things (IoT) technologies can improve spatial and temporal coverage, but practical long-term deployments must address power variability, intermittent communication, local data loss, and uncertainty in low-cost sensor measurements. This dissertation investigates scalable distributed systems for intelligent coastal environmental monitoring by integrating low-cost sensing, edge computing, hybrid communication, local data management, and machine-learning-assisted calibration.
The first contribution is the design and deployment of the Coastline Environmental Monitoring System (CEMS), a modular platform that integrates heterogeneous air- and water-quality sensors, LoRaWAN and Wi-Fi communication, renewable and grid-assisted power, cloud-based storage, and visualization services. A proof-of-concept deployment in Ingleside on the Bay, Texas demonstrates the feasibility and cost advantages of collecting multiple environmental variables with low-cost distributed sensing nodes. This platform establishes the sensing, communication, and server foundation used by the subsequent chapters.
The second contribution extends CEMS for long-term field operation. The enhanced edge-IoT design combines energy-aware operating modes, store-first local persistence, acknowledgement-based status management, automatic retransmission, and rule-based selection among LoRaWAN, Wi-Fi, cellular communication, and deferred upload. During a four-day field evaluation containing an intentional 12~h 56~min LoRaWAN interruption, 3,622 of 3,633 locally collected records were delivered, corresponding to 99.7\% data completeness. A counterfactual immediate-transmit LoRaWAN-only condition derived from the same outage would have delivered 71.0\% of the records. The local pending queue absorbed the interruption and was cleared after connectivity restoration and Wi-Fi-assisted backfill. The estimated average communication-subsystem power was approximately 0.26~W, about 86\% lower than a continuously active 1.8~W Wi-Fi interface; this comparison does not represent whole-node energy savings.
The third contribution develops a data-quality and AI-based calibration workflow for low-cost air-quality sensors. The workflow aligns low-cost observations with reference-station measurements, applies rule-based checks, constructs sensor, environmental, and temporal features, and evaluates Linear Regression, Random Forest, XGBoost, and multilayer perceptron models. In the reported exploratory 8:2 evaluation of a representative NO$_2$ dataset, Random Forest achieved an $R^2$ of 0.7112 and substantially reduced mean absolute error and root mean squared error relative to the raw sensor output. The analysis also identifies the need for longer co-location periods, chronological validation, and drift-aware model maintenance before the calibration can be considered deployment-ready.
Together, the three contributions form a coherent system progression: CEMS establishes a low-cost monitoring platform, the edge-IoT enhancements preserve and deliver observations under field constraints, and the calibration workflow addresses the credibility of the resulting measurements. The dissertation provides practical design principles and evaluation methods for scalable, resilient, and intelligent coastal monitoring systems.
Keywords
Internet of Things; adaptive sensing; hybrid communication; low-power systems; sensor data reliability; AI-based calibration; Environmental monitoring
Disciplines
Computational Engineering | Computer and Systems Architecture
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Li, Jian, "SCALABLE DISTRIBUTED SYSTEMS FOR INTELLIGENT MONITORING" (2026). Computer Science and Engineering Dissertations. 19.
https://mavmatrix.uta.edu/cse_dissertations2/19