Yeqian Xu

Date of Award

May 2024

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Earth System Science & Policy

First Advisor

Haochi Zheng

Abstract

Establishing and maintaining high-quality grasslands and conservation lands for pollinators can be difficult and costly. As an alternative, roadsides can provide additional forage resources. This study aims to examine the potential of roadside, road ditch, and right-of-way (ROW) land in North Dakota (ND) as alternative habitats for pollinators by quantifying the yellow flowering observed in these sites. The project encompasses a multifaceted approach to assess the viability of roadside ditches in supporting pollinator ecosystems. The research incorporates two essential components: 1) an extensive road survey through the Prairie Pothole Region (PPR) of ND in two consecutive summer seasons and 2) intensive image detection using machine learning techniques. To ensure coverage of the three main routes through the PPR, a travel plan was designed to maximize survey sites within limited time and distance over the growing season. The data analysis, powered by state-of-the-art machine learning frameworks such as TensorFlow and PyTorch, estimates the yellow floral coverage observed from the imagery data collected from the road surveys. The approach used in the study has numerous advantages and can serve as an alternative to the traditional biology and ecology survey method, which is usually time-consuming and requires a lot of effort. The study's results can provide valuable data and insights into the current state of ROWs in terms of their ability to provide habitats for pollinators, as well as suggestions for enhancing the management of ROWs to support pollinators and their pollination services.

Download

Share

COinS