Saltcedar is commonly recognized as one of the most threatening invasive species in the United States and has the potential to cause great environmental harm over the coming decade. Accurate mapping of saltcedar distribution and abundance in a timely manner plays a central role in assisting with effective control. Current studies have mostly concentrated on large-area detection with coarse-resolution remote sensing data. In this study, a comprehensive test was designed and carried out to examine the ability to integrate multitemporal and multiresolution imagery for differentiating saltcedar from other riparian vegetation types in the Rio Grande basin of Texas, including very high spatial resolution (QuickBird), hyperspectral resolution imagery (AISA), and moderate resolution satellite imagery (Landsat TM). Two types of analyses were fulfilled. First, five pixel-based classification methods were adopted for assessing the effectiveness of QuickBird and AISA for discerning saltcedar, respectively; that is, the maximum likelihood classifier (MLC), neural network classifier (NNC), support vector machine (SVM), spectral angle mapper (SAM), and maximum matching feature (MMF). Second, Landsat TM imagery was synthesized from AISA and tested for mapping the abundance of saltcedar with four linear spectral unmixing methods and three back-propagation neural network methods. Results indicate that AISA outperformed QuickBird imagery in differentiating saltcedar from other riparian vegetation species. SVM achieved the highest classification accuracy among the five classifiers. Linear spectral unmixing methods exhibited similar mapping accuracy to neural network methods in estimating the abundance of saltcedar at a spatial resolution of 30 by 30 m2 but with significantly better computing efficiency.
- spectral unmixing
ASJC Scopus subject areas
- Geography, Planning and Development
- Earth-Surface Processes