Assessing Vegetation Composition and Characteristics Using Ground-Level Hyperspectral Data in Northern Virginia

Friday, 19 December 2014
Itiya Aneece, University of Virginia Main Campus, Charlottesville, VA, United States and Howard E Epstein, University of Virginia Main Campus, Environmental Sciences, Charlottesville, VA, United States
The study of ecosystem properties and processes through remote sensing allows ecological questions to be answered more efficiently for large geographical expanses than field work alone, especially in areas that are relatively inaccessible. These properties and processes are often studied at coarse spatial scales with multispectral data; however, the use of hyperspectral data to ask plant community and species-level questions is still a developing field. Many applications, such as understanding the influence of disturbances and assessing management strategies, need finer-scale information than is currently available using multispectral data. In this study, we used hyperspectral data to examine vegetation community properties in preparation for addressing these finer-scale questions. Specifically, we examined the ability to assess vegetation composition and diversity using ground-level hyperspectral data for early successional and other non-forested fields in north-central Virginia. Twelve 5m by 5m plots were established at which a vegetation survey was conducted at the ground, understory, and canopy levels at 0.5m intervals. We additionally collected twelve spectra with approximately 1m footprints at each plot. We then ran ordinations to assess clustering of plots by similarity in species compositions and assessed the spectral bands most strongly correlated with clustering. We found that plots do cluster by species composition, but the most influential wavelengths varied by year of data collection. In 2012, the most influential bands were in the near-infrared plateau region followed by some influence from the red region. The most influential bands in 2014 were in the blue-green and red regions. The correlations between species diversity and spectral diversity also differed by year; however, when an outlier was removed from each of the years, there was a weak positive correlation between species diversity and spectral diversity during both years. These results are promising in that they give us a glimpse into the potential to use hyperspectral data to assess fine-scale ecosystem properties and processes; these techniques may in the future be used to study the role of disturbance and effectiveness of management practices across landscapes.