Identifying invasive plant species using field spectroscopy in the VNIR region in successional systems of north-central Virginia. Issue 1 (2nd January 2017)

• Record Type: Journal Article
• Authors: Aneece, Itiya; Epstein, Howard
• Is Part Of: International journal of remote sensing. Volume 38:Issue 1(2017)
• Page Start: 100
• Page End: 122
• Publication Date: 2017-01-02
• Subjects: Remote sensing -- Periodicals; Télédétection -- Périodiques; 621.3678
• DOI: 10.1080/01431161.2016.1259682
• Languages: English
• ISSNs: 0143-1161

Abstract

ABSTRACT: Remote sensing can potentially be used to monitor the extent and distribution of invasive species across landscapes and regions, thus aiding conservation efforts. We collected ground-level hyperspectral data of six exotic invasive plant species in abandoned agricultural fields at the Blandy Experimental Farm in northern Virginia to determine the degree to which species could be identified using visible and near-infrared wavelengths. The spectral profile from 350 to 1025 nm was used in support vector machine analysis to determine separability of these species. We used sensitivity analyses to determine which spectral regions were most influential to identifying species by removing 50 nm regions and comparing species identification to that using the full spectral profile. Ailanthus altissima, Carduus acanthoides, and Cirsium arvense had high ability to be identified (75%, 87.5%, and 75%, respectively). Galium verum had low ability to be identified (44.4%), perhaps due to high spectral contamination from soil. Celastrus orbiculatus and Rhamnus davurica had low ability to be identified (27.3% and 30.8%, respectively); however, they were often misclassified as each other, due to their physical overlap in the field. The sensitivity analysis revealed that the 350–399, 500–549, 700–749, and 900–949 nm regions were most useful for species identification, while 550–599 and 650–699 nm regions were detrimental, perhaps due to greater intraspecific variability than interspecific variability in these regions. These most influential regions for identification were similar to those found in other studies. Thus, it is possible to identify species using ground-level hyperspectral data.