Date of Award

Summer 8-20-2021

Level of Access Assigned by Author

Open-Access Thesis

Degree Name

Master of Science (MS)


Forest Resources


Parinaz Rahimzadeh-Bajgiran

Second Committee Member

Aaron Weiskittel

Third Committee Member

John Daigle

Additional Committee Members

Kara Costanza


North American ash species (Fraxinus spp.) are under dire threat from the invasive pest, emerald ash borer (Agrilus planipennis, EAB). Black ash (F. nigra) has shown no resistance to EAB while its cultural and ecological importance render it irreplaceable. Traditional field forestry techniques are not suitable for the large-scale identification of individual black ash trees to facilitate conservation, thus necessitating the need for other identification and classification techniques. The objective of this research is to develop remote sensing techniques that can be used to identify ash trees, in particular black ash, at the individual tree level using both hyperspectral and multispectral data. Both general ash species identification and black ash tree identification in a low-density mixed forest using hyperspectral data have not been reported in the literature. Specifically, this study aims to use optical remote sensing data to: 1) create a pixel-based classification model for ash tree identification, 2) develop an object-based classification model for ash tree identification, and 3) use the most accurate ash tree classification model as a basis for a black ash tree classification model. Analysis of spectral differences between classes suggests that both ash in general and black ash specifically can be successfully separated from co-occurring species. Where classification models were significantly different, object-based methods performed better than pixel-based methods and Support Vector Machine (SVM) models generally outperformed Random Forest (RF). The highest accuracies were achieved using object-based methods and hyperspectral data, although multispectral data were able to successfully differentiate ash as well. Using object-based, SVM methods, black ash was successfully differentiated from co-occurring hardwood species using both hyperspectral and multispectral data, with hyperspectral data achieving 70% Producer’s and 70% User’s Accuracy for black ash and multispectral data achieving 57% and 50%, respectively. Despite relatively low sample sizes, this research presents a viable path forward with respects to black ash mapping. As this study shows, black ash can be successfully differentiated from closely related species using remotely sensed optical data. While capturing hyperspectral data is likely cost prohibitive for large-scale mapping efforts, multispectral sensors are more viable and can achieve similar results. At a minimum, the techniques presented in this research can be used to assist and guide field conservation work to locate areas of high likelihood of black ash presence so that they can be identified and informed decisions made about their preservation.