Spectral discrimination of an invasive species (Spartina alterniflora) in Min River wetland using field spectrometry

Abstract

Spartina alterniflora, an invasive plant, has been a threat to the local ecological security since it was introduced to Fujian coastal beach over 30 years ago. How to monitor its dynamic changes effectively is of great significance. Currently, hyperspectral remote sensing technology has become an important way to monitor invasive species dynamic changes. This paper investigates whether S. alterniflora could be discriminated from the other three native species using field spectrometer ranging from 350 nm to 2500 nm. In order to reduce and select the optimal bands for the potential discrimination of S. alterniflora, a hierarchical method is implemented to spectrally discriminate S. alterniflora from the other three native species. In the first level of the analysis using ANOVA, we found that there were statistically significance differences in spectral reflectance between S. alterniflora and the other three native species at different bands. The algorithm of classification and regression trees (CART) were used to further investigate in the second level of analysis to identify the most sensitive bands for spectral discrimination. We found that the greatest discrimination power for S. alterniflora is located in the red-edge, especially in the near infrared, and mid infrared regions. Subsequently, we used Jeffries-Matusita (JM) distance to assess spectral separability of bands selected by CART. Overall, results of this study offer the possibility of extending field measurements at canopy level to airborne and hyperspectral data for discriminating S. alterniflora in Min river wetland.