First Advisor

Geoffrey Duh

Term of Graduation

Spring 2021

Date of Publication


Document Type


Degree Name

Master of Science (M.S.) in Geography






Hyperspectral imaging, Trees -- Classification -- Remote sensing



Physical Description

1 online resource (vi, 57 pages)


Hyperspectral imagery has become a common remote sensing data type used in tree species classifications because of its rich spectral signals that allow the detection of the variations in canopy reflectance. While high spatial resolution hyperspectral imagery provides fine spatial resolution for discerning surface objects, it has the inherent drawbacks of expensive acquisition costs, large data sizes, and can be computationally taxing to use. This study attempts to determine a relationship between crown level tree species classification accuracy and hyperspectral spatial resolution. Future tree species classification projects can make use of this relationship by targeting a spatial resolution that best avoids the drawbacks of hyperspectral imagery. I processed a 37-band hyperspectral mosaic that has a 0.3 meters resolution and resampled it to 0.5, 1.0, 2.0, 3.0, and 5.0 meters mosaics and used a support vector machine (SVM) classifier to create tree species classifications for each of the resampled mosaics to examine the relationship between spatial resolution and classification accuracies. The mosaic covers a 50 sq-km study site in El Dorado County, California. The classifier used tree species data that I collected in the field as training and validation data. The results show that there was no significant classification accuracy difference between the resolutions. The averaged overall accuracies were highest when using the 1.0 meters mosaic (73.23%) and dropped when increasing or decreasing the spatial resolutions. The 5.0 meters mosaic yielded a minimum overall accuracy of 64.42%. The finding suggests that spatial resolution is not a critical factor in classification accuracy, indicating that reasonable classification results can be achieved using either coarser resolution imagery, such as those collected with satellite or airborne sensors, or fine resolution imagery, such as those collected with unmanned aerial vehicles (UAV). The crown size of the trees appears to be an important factor mediating classification accuracy and image resolution. The knowledge gained in this study could help remote sensing project managers to determine a resolution that best fits their budget and computational power.


In Copyright. URI: This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).

Persistent Identifier