University of North Dakota
Assessing The Biophysical Naturalness Of Grassland In Eastern North Dakota With Hyperspectral Imagery
Abstract
dc:description.abstract<p>Over the past two decades, non-native species within grassland communities have quickly developed due to human migration and commerce. Invasive species like Smooth Brome grass (Bromus inermis) and Kentucky Blue Grass (Poa pratensis), seriously threaten conservation of native grasslands. This study aims to discriminate between native grasslands and planted hayfields and conservation areas dominated by introduced grasses using hyperspectral imagery.</p> <p>Hyperspectral imageries from the Hyperion sensor on EO-1 were acquired in late spring and late summer on 2009 and 2010. Field spectra for widely distributed species as well as smooth brome grass and Kentucky blue grass were collected from the study sites throughout the growing season. Imagery was processed with an unmixing algorithm to estimate fractional cover of green and dry vegetation and bare soil. As the spectrum is significantly different through growing season, spectral libraries for the most common species are then built for both the early growing season and late growing season. After testing multiple methods, the Adaptive Coherence Estimator (ACE) was used for spectral matching analysis between the imagery and spectral libraries. Due in part to spectral similarity among key species, the results of spectral matching analysis were not definitive. Additional indexes, "Level of Dominance" and "Band variance", were calculated to measure the predominance of spectral signatures in any area. A Texture co-occurrence analysis was also performed on both "Level of Dominance" and "Band variance" indexes </p> <p>to extract spatial characteristics. The results suggest that compared with disturbed area, native prairie tend to have generally lower "Level of Dominance" and "Band variance" as well as lower spatial dissimilarity.</p> <p>A final decision tree model was created to predict presence of native or introduced grassland. The model was more effective for identification of Mixed Native Grassland than for grassland dominated by a single species. The discrimination of native and introduced grassland was limited by the similarity of spectral signatures between forb-dominated native grasslands and brome-grass stands. However, saline native grasslands were distinguishable from brome grass.</p>
Degree
thesis:*- Name thesis:degree_name
- Master of Science (MS)
- Level thesis:degree_level
- Thesis
- Discipline thesis:degree_discipline
- Earth System Science & Policy
- Year
- 2012
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Zhou, Qiang
- Contributors dc:contributor
-
- Michael Hill
Subjects
dc:subject × 1Identifiers
dc:identifier.*- Repository record dc:identifier
- https://commons.und.edu/theses/1218
- OAI identifier oai:identifier
- oai:commons.und.edu:theses-2219