Document Type

Thesis - Open Access

Award Date

2018

Degree Name

Master of Science (MS)

Department / School

Natural Resource Management

First Advisor

Michael C. Wimberly

Keywords

Eastern redcedar, Great Plains, Juniperus virginiana, matched filtering, remote sensing, woody encroachment

Abstract

Juniper encroachment is a considerable threat to the prairie ecosystems of the Great Plains because it has the potential to alter native grasslands by changing soil characteristics, limiting herbaceous biomass, and hindering native community regeneration. Accurate maps of juniper cover and predictions of areas at risk for future expansion are needed to support proactive management measures. Therefore, our objectives are to: (1) Develop a practical workflow for large-scale juniper mapping using Landsat 8 Operational Land Imager (OLI) imagery and partial unmixing techniques, (2) Compare the classification accuracies from the resulting map based on different juniper density thresholds and different types of imagery, (3) Develop a predictive spatial model for the distribution of low-density juniper based on distance to seed source and environmental covariates and determine the prediction accuracy, and (4) Use the resulting maps to evaluate the extent of current juniper establishment and the risk of future encroachment. The study area encompasses counties bordering the Missouri River in southeastern South Dakota and northeastern Nebraska and covering approximately 23,000 km2. We applied a matched filtering technique to classify juniper with snowcovered and snow-free winter imagery (December-March) and snow-free spring imagery (April-June). We found that using the snow-covered winter images suppressed background spectral signatures and resulted in a higher overall classification accuracy of 93.7% for juniper densities above 15 percent, compared to snow-free winter imagery and spring imagery. When characterizing juniper densities below 10 percent our 30-meter pixel level classification map was unreliable, with an 11% probability of correctly classifying juniper. Therefore, we used Random Forests, a machine-learning algorithm, to develop a model of low-density (≤ 15%) juniper based on classified juniper cover and other ecological factors. We used the receiver operating characteristics (ROC) curve to evaluate model predictions; accuracy was high with an area under the curve (AUC) of 0.884. Our susceptibility map indicated that an additional 7.7% of the study area currently contained low densities of juniper and had high to very high risk of future encroachment. This study will provide agencies and land managers with information and techniques needed to address juniper encroachment in the Northern Great Plains.

Library of Congress Subject Headings

Eastern redcedar -- Great Plains.
Junipers -- Great Plains.
Invasive plants -- Great Plains.
Remote sensing.

Description

Includes bibliographical references

Format

application/pdf

Number of Pages

103

Publisher

South Dakota State University

Download

Share

COinS
 

Rights Statement

In Copyright