Electronic Theses and Dissertations

Date of Award

1-1-2025

Document Type

Dissertation

Degree Name

Ph.D. in Engineering Science

First Advisor

Mohammad M. Al-Hamdan

Second Advisor

Mohammad M. Al-Hamdan

Third Advisor

Hakan H. Yasarer

School

University of Mississippi

Relational Format

dissertation/thesis

Abstract

Large-scale watershed simulation platforms, such as the Agricultural Integrated Management System (AIMS), are essential for water resource management but are often limited by a critical lack of detailed in-stream channel data required for comprehensive river analysis. This study addresses this gap by developing and validating an integrated, data-driven framework that couples upland process models with in-stream hydraulic geometry estimation at a national scale. The research was executed through two complementary thrusts founded on the National Hydrography Dataset Plus (NHDPlus).

First, a geospatial workflow was developed to create NHDPlus-driven input files for the Annualized Agricultural Non-Point Source (AnnAGNPS) model—a tool for simulating watershed processes. To evaluate this new method, the AnnAGNPS model was executed using both the NHDPlus-driven inputs and a set of baseline inputs generated by the traditional DEM-based TopAGNPS pre-processor. A comparative analysis of the outputs from these two model runs revealed that while runoff simulations were consistent, sediment yield predictions diverged significantly. This divergence was statistically attributed to differences in the topographic Length-Slope (LS) Factor of the Revised Universal Soil Loss Equation (RUSLE), an empirical soil erosion model, derived from NHDPlus’s hydro-enforced topography.

Second, a machine learning framework was engineered to estimate river channel geometry (width, depth, area). This was achieved by integrating in-situ measurements from the Hydrology for the Surface Water and Ocean Topography (HYDROSWOT) database, a collection of ground-based data supporting satellite missions, with NHDPlus watershed attributes. The developed Artificial Neural Network (ANN) demonstrated high predictive accuracy at a national scale for cross-sectional area under mean flow conditions. Feature importance analysis using SHapley Additive exPlanations (SHAP), a method for explaining the outputs of machine learning models, confirmed the model’s physical interpretability. The analysis revealed that the model learned to prioritize physically intuitive drivers such as flow metrics and network hierarchy.

Together, these research thrusts establish a validated, spatially consistent, and scalable proof-of-concept that bridges the divide between watershed-scale hydrology and instream hydraulics. This provides the foundational data required to enable advanced one-dimensional hydrodynamic modeling applications within AIMS and other national-scale water resource assessment platforms.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.