Theses and Dissertations

Issuing Body

Mississippi State University

Advisor

Pricope, Narcisa

Committee Member

Leyton, Javier Osorio

Committee Member

Dash, Padmanava

Date of Degree

8-7-2025

Original embargo terms

Visible MSU Only 1 year

Document Type

Graduate Thesis - Campus Access Only

Major

Geoscience (Geospatial Science)

Degree Name

Master of Science (M.S.)

College

College of Arts and Sciences

Department

Department of Geosciences

Abstract

This study presents a multi-scale remote sensing framework to classify fractional vegetation cover (FVC) across rangelands by integrating UAV multispectral imagery, LiDAR structural data, and multisource satellite inputs. Four machine learning models, ranging from UAV-only to fused UAV-LiDAR-Sentinel models, were developed and evaluated using Random Forest (RF) and Extreme Gradient Boosting (XGBoost) algorithms. The research was conducted at Martin Ranch, Texas, using Uncrewed Ariel Vehicle (UAV)-collected multispectral imagery, LiDAR-derived canopy height, and Sentinel-1/2 indices. Model 2 (UAV + LiDAR) emerged as the optimal model for balancing accuracy, processing time, and compatibility with ecological simulation tools. This model captures both spectral and structural vegetation features essential for biomass estimation and hydrologic modeling. Results highlight XGBoost’s superior classification performance, particularly for complex vegetation classes. The framework enables scalable high-resolution FVC mapping, experiments on how fused machine learning outputs can directly enhance ecological modeling and supports decision-making for sustainable rangeland management.

Sponsorship (Optional)

Texas A&M AgriLife Research, Texas A&M AgriLife Blackland Research & Extension Center, Mississippi State University Office of Research and Economic Development

Download

Share

COinS