Author ORCID Identifier

https://orcid.org/0009-0007-7970-0622

Semester

Spring

Date of Graduation

2025

Document Type

Thesis

Degree Type

MS

College

Davis College of Agriculture, Natural Resources and Design

Committee Chair

Paul Kinder

Committee Co-Chair

Michael Strager

Committee Member

Michael Strager

Committee Member

Walter Veselka

Abstract

This research introduces an innovative approach for managing autumn olive (Elaeagnus umbellata) on reclaimed surface mines through the integration of machine learning algorithms and unmanned aerial systems. The proliferation of autumn olive presents significant ecological challenges on disturbed landscapes, necessitating more efficient detection and control methods. This study evaluated Random Forest classification and deep learning approaches using the U-Net architecture for identifying autumn olive across four phenological stages using high-resolution imagery. The RGB U-Net model achieved F1-scores of 0.927, 0.912, and 0.871 for autumn olive detection during peak, late, and senescence stages respectively, outperforming both multispectral and vegetation index composite models. Temporal analysis revealed optimal detection windows during peak and late growing seasons when autumn olive characteristics diverge the most from native vegetation. Feature importance analyses identified NIR and red-edge indices as valuable for early and peak detection, while visible-band and soil-adjusted indices became important during senescence. Unmanned aerial herbicide applications demonstrated effective control with minimal impact on surrounding vegetation and reduced labor requirements. This integrated technological approach represents a promising framework for invasive species management on reclaimed mine lands, offering enhanced accuracy, efficiency and cost-effectiveness compared to traditional management practices.

Download

Share

COinS