High-Resolution Satellite Imagery: An Alternative Method for the Detection and Monitoring of Rapid ‘Ōhi‘a Death in Hawai‘i













Journal Title

Journal ISSN

Volume Title




Starting Page

Ending Page

Alternative Title


Rapid ‘Ōhi‘a Death has caused extensive mortality of endemic ‘ōhi‘a (Metrosideros polymorpha) across the Hawaiian Islands. First detected on Hawai‘i Island, the responsible Ceratocystis pathogens have now spread to Kaua‘i, O‘ahu, and Maui. Multiple remote sensing efforts such as RGB aerial imagery, laser spectroscopy, and Digital Mobile Sketch Mapping have been successful in identifying mortality outbreaks, although each method has temporal, spatial, or fiscal limitations. In this thesis I explore the utilization of high-resolution satellite imagery with a spatial resolution of £ 0.5 m for individual symptomatic tree identification for field sampling and laboratory testing of the Ceratocystis pathogen. I also analyze landscape scale annual patterns of mortality across East Hawai‘i and compare my findings to annual helicopter- based Digital Mobile Sketch Mapping (DMSM) surveys. Lastly, I created an object detection model using NV5’s ENVI Deep Learning for suspect tree identification and compared the results to DeepForest object detection using the same datasets for training. Results showed that of the individual suspect trees visited in the field (n = 55), 94.5% were successfully located with imagery. Of the trees located (n = 52) using satellite imagery suspects, 98.1% were correctly identified as ‘ōhi‘a. From the subset sampled (n = 35), 40% resulted in a Ceratocystis lukuohia detection. Annual patterns of mortality were found to be similar between satellite-identified and DMSM survey datasets and showed rapid increases in mortality patterns in initial infection years. Temporal and spatial differences between datasets limit direct comparisons due to the short duration of the red phase that strongly signifies ROD-related mortality. Used jointly, these two datasets provide a comprehensive picture of annual mortality trends across East Hawai‘i. Finally, DeepForest models (F1-score = 0.32 and mAP = 0.73) were significantly better overall compared to ENVI models (F1-score = 0.04 and mAP = 0.017) and were promising for automatically identifying tree mortality in new satellite imagery. These results indicate that high-resolution satellite imagery complements existing remote sensing efforts and can help to create a robust and comprehensive monitoring system. This research documents the utilization of high-resolution satellite imagery for landscape-scale ecological monitoring of forest pathogens in Hawai‘i. These results provide annual mortality trends of Hawai‘i Island and share a pathway for future automated identification of new outbreaks which can be inform land management decisions and be implemented on other islands.



Remote sensing, Geography, Geographic information science and geodesy, Ceratocystis lukuohia, deep learning, Metrosideros polymorpha, visible imagery



76 pages


Geographic Location

Time Period

Related To

Related To (URI)

Table of Contents


All UHH dissertations and theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission from the copyright owner.

Rights Holder

Local Contexts

Email libraryada-l@lists.hawaii.edu if you need this content in ADA-compliant format.