UROP Project

PFAS, Laboratory work, Machine Learning, Deep Learning,

Research Mentor: Mr. Shahin Alam , Shaheen
Department, College, Affiliation: Civil & Environmental Engineering, FAMU-FSU College of Engineering
Contact Email: ma23ch@fsu.edu
Research Assistant Supervisor (if different from mentor):
Research Assistant Supervisor Email:
Faculty Collaborators:
Faculty Collaborators Email:

Looking for Research Assistants: Yes
Number of Research Assistants: 2
Relevant Majors: Environmental Engineering, Environmental Science, Chemistry, Chemical Engineering, Data Science, Computer Science
Project Location: FAMU-FSU College of Engineering
Research Assistant Transportation Required: University Bus
Remote or In-person: In-person
Approximate Weekly Hours: 10,
Roundtable Times and Zoom Link: Tuesday, September 3:00 to 3:30
Zoom Link:
https://fsu.zoom.us/j/92107461404
https://fsu.zoom.us/j/92107461404

Project Description

Introduction
Per- and polyfluoroalkyl substances (PFASs) represent a class of synthetic chemicals that have emerged as pervasive contaminants across various environmental media globally (Alam & Chen, 2024). Their persistence, bioaccumulative nature, and toxicity present substantial risks to ecosystems and human health (Hamid et al., 2023). The Lower Suwannee River Basin of Florida, which provides essential water resources for nearby communities, is threatened by PFAS contamination owing to intensified industrial and agricultural activities. This study aims to characterize PFAS spatial distribution in the basin and potential ecological impacts. Owing to limited available data, machine learning (ML) will be used to identify PFAS contamination patterns, and trace PFAS sources (Kibbey et al., 2020, 2021). Deep learning (DL), a subset of machine learning, which excels at modeling complex relationships in datasets, will be used to predict future PFAS contamination scenarios, enabling proactive management and mitigation strategies. This goal of this proposed research outlines a comprehensive study to investigate PFAS contamination in the Lower Suwannee River Basin by integrating field sampling and laboratory analysis with advanced ML and DL techniques to identify PFAS sources, track their distribution, and predict future contamination patterns on a watershed scale. It is hypothesized that PFAS contamination in the Lower Suwannee River Basin has identifiable source footprints and exhibit spatial distribution patterns corresponding to related industrial and agricultural activities, which can be predicted and modeled using ML and DL techniques.
Objective
The research questions to be answered in this proposed research include: 1) What are the concentration levels and distribution patterns of PFAS in the surface water and sediments of the Lower Suwannee River Basin? 2) How can machine learning models be utilized to identify and trace the sources of PFAS contamination using non-targeted PFAS data? And 3) Can deep learning models accurately predict PFAS contamination scenarios based on current and historical data? Corresponding to the research questions, the Objectives of this proposal research are to: 1) quantify PFAS concentrations and distribution in the surface water and sediments of the Lower Suwannee River Basin by assessing PFASs in surface water and sediments through field sampling and laboratory analysis, 2) apply machine learning to track PFAS sources by using non-targeted PFAS data and developing machine learning models to identify contamination footprints, and 3) develop deep learning models to predict future PFAS contamination scenarios, enabling proactive management and mitigation strategies.
Methodology
Water and sediment samples will be collected from locations along Suwannee River in the Lower Suwannee River Basin, covering upstream, midstream, and downstream sites. PFASs in the surface water and sediments will be identified and quantified by liquid chromatography-tandem mass spectrometry (LC-MS/MS) following EPA methods 1633. Statistical analyses will be performed to determine PFAS concentration levels, distribution patterns, and potential sources. Non-targeted PFAS data from various sources, including environmental samples, industrial discharge records, and historical monitoring data, will be compiled from US Environmental Protection Agency (EPA), Florida Department of Environmental Protection (FDEP), and published literatures, and integrated to provide a comprehensive overview of PFAS contamination in the region. ML algorithms will be used to extract meaningful features from the non-targeted PFAS data by focusing on identifying unique fingerprints that can indicate specific contamination sources, enhancing the accuracy of source tracking. Supervised and unsupervised ML models, such as random forests and clustering algorithms, will be applied to identify and classify PFAS sources. These models will help distinguish between different contamination sources and provide insights into their contributions to the overall PFAS contamination levels. The ML models will be validated using potential known PFAS sources (e.g., airport location, military bases, wastewater treatment plant, biosolids application location) and cross-referenced with historical data to ensure the model reliability and accuracy. A comprehensive dataset will then be established by incorporating PFAS concentration levels, environmental parameters (e.g., temperature, pH, flow rate), rainfall, soil, and land use data, which will serve as the foundation for developing predictive models. Deep learning models, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), will be modified and trained to predict future PFAS contamination levels. These models will leverage the comprehensive dataset to learn complex relationships and patterns and be trained to simulate various contamination scenarios under different environmental and anthropogenic conditions. The predictions of the deep learning models will be validated using a subset of the dataset, and model parameters will be optimized to improve accuracy and reliability, ensuring robust predictions. The results of the proposed research will provide valuable insights into potential future contaminations and inform proactive management strategies.
References
Alam, M. S., & Chen, G. (2024). PFAS Regulatory Frameworks, Sources, Occurrence, Fate, and Exposure: Trend, Concern, and Gaps. Manuscript submitted for publication.
Hamid, N., Junaid, M., Manzoor, R., Sultan, M., Chuan, O. M., & Wang, J. (2023, Dec 20). An integrated assessment of ecological and human health risks of per- and polyfluoroalkyl substances through toxicity prediction approaches. Science of the Total Environment, 905.
Kibbey, T. C. G., Jabrzemski, R., & O'Carroll, D. M. (2020, Aug). Supervised machine learning for source allocation of per- and polyfluoroalkyl substances (PFAS) in environmental samples. Chemosphere, 252.
Kibbey, T. C. G., Jabrzemski, R., & O'Carroll, D. M. (2021, Jul). Source allocation of per- and polyfluoroalkyl substances (PFAS) with supervised machine learning: Classification performance and the role of feature selection in an expanded dataset. Chemosphere, 275.


Research Tasks: Water and Sediment sampling; Laboratory analysis; Literature Review; Computational Modeling

Skills that research assistant(s) may need: Recommended