Theses and Dissertations

ORCID

https://orcid.org/0009-0005-4706-5025

Advisor

Dash, Padmanava

Committee Member

Mercer, Andrew

Committee Member

Ambinakudige, Shrinidhi

Committee Member

Bhushan, Shanti

Date of Degree

12-12-2025

Original embargo terms

Visible MSU Only 2 Years

Document Type

Dissertation - Campus Access Only

Major

Earth and Atmospheric Sciences

Degree Name

Doctor of Philosophy (Ph.D.)

College

College of Arts and Sciences

Department

Department of Geosciences

Abstract

Turbidity is a vital indicator of water quality, influencing light penetration and the ecological function of coastal environments. Traditional monitoring methods often lack the spatial and temporal resolution needed for effective environmental management. This dissertation introduces a multi-scale framework that integrates Uncrewed Aircraft Systems (UAS), Autonomous Surface Vessels (ASV), satellite imagery, and machine learning to produce high-resolution turbidity maps and long-term trend analyses for the Mississippi Sound. Chapter I presents a fine-scale turbidity mapping method using UAS multispectral imagery calibrated with in situ ASV measurements. This hybrid UAS-ASV system enabled the generation of detailed turbidity estimates. Among various machine learning models tested on radiometer-derived sensor-specific remote sensing reflectance, the Support Vector Machine (SVM) performed best (R² = 0.943, RMSE = 0.454 NTU), effectively capturing nonlinear relationships between turbidity and remotely sensed data. Chapter II expands the analysis to broader spatial and temporal scales by developing a deep neural network (DNN) regression model. Trained on datasets from six ASV field campaigns and Landsat 8 and 9 surface reflectance imagery, the DNN outperformed RandomForest, SVM, and XGBoost models, achieving an R² of 0.864 and RMSE of 1.794 NTU. The model was then used to generate a 22-year turbidity time series (2002–2024), revealing seasonal trends and peak turbidity events in 2013, 2016, and 2020—linked to meteorological disturbances and Bonnet Carré Spillway openings. Chapter III investigates the influence of land use and land cover (LULC) changes on turbidity. Using harmonized annual LULC maps and spatial correlation analysis, the study found a significant negative correlation between barren land and turbidity at the HUC-10 watershed scale, while cropland and urban areas had minimal effects. Finer-resolution HUC-10 analysis revealed slightly stronger land–water relationships than coarser HUC-8 scales. As a preliminary study focused on a single LULC factor, this chapter provides a foundation for more comprehensive future research. This research underscores the value of integrating remote sensing with machine learning to monitor complex coastal systems. The proposed framework is adaptable across regions and scales, offering a flexible tool for water quality assessment, habitat restoration, and informed environmental decision-making

Download

Share

COinS