Using spectral and textural information to detect and map Parthenium hysterophorus L. in Mtubatuba, South Africa.

Abstract

Parthenium hysterophorus L. (parthenium) is an alien invasive species that has had severe environmental and human impacts in three continents. Sustainable management and control of the invasive species requires an understanding of its distribution and rate of spread. Our first study focuses on the use of spectral information of commercial sensor RapidEye and freely available Sentinel-2 imagery to detect parthenium and other land cover classes. Sentinel-2 outperformed RapidEye to classify most land cover classes, with an overall classification accuracy of 82% and 71%, respectively. This was likely due to the superior spectral resolution of Sentinel-2. However, RapidEye performed better when classifying parthenium, potentially due to the fact that there were some patches that were smaller than the Sentinel-2 spatial resolution. Nonetheless, Sentinel-2 represents a good opportunity to map larger parthenium stands and other land cover types. The second study focused on mapping parthenium using texture analysis and SPOT-6 imagery. It compared the mapping ability between the panchromatic and multispectral bands using the PLS-DA algorithm. The panchromatic band achieved a higher overall classification accuracy than the multispectral bands (77% and 73%, respectively). Furthermore, the panchromatic band achieved superior performance compared to multispectral bands for parthenium. This may be attributed to the higher spatial resolution of the panchromatic band as it has been shown that finer spatial resolution is beneficial in texture analysis. Overall texture analysis using SPOT 6 imagery was the most successful combination which allowed us to accurately map parthenium distribution.