Mapping the invasive species Phragmites australis in linear wetland corridors

The detection and monitoring of invasive species at the initial stage of invasion is often critical to control/eradication efforts. In the case of Phragmites australis, anthropogenic linear wetlands such as roadside and agricultural ditches are believed to play a key role in invasion patterns. Accurate remote sensing of an aquatic macrophyte in such narrow habitats, however, remains a challenge. We used large-scale (1/8000) panchromatic and color aerial photographs to produce different distribution maps of P australis in a network of linear wetlands. Accuracy assessments were conducted to compare the two classifications and sources of errors were identified using logistic regressions. Different thresholds of stein abundance (1%, 5%, 20%, and 40%) were used in the error matrices to determine the stein abundance at which our classification is optimized. Results show that color images are much better in enabling the detection of P. australis. Producer's accuracy ranges from 44% to 71% (depending on the selected threshold of stem abundance) for color images and from 16% to 28% for panchromatic images. User's accuracy ranges from 84% to 55% for color photographs and from 51% to 28% for panchromatic photographs. Generally, the mapping of vigorous populations is more accurate. The presence of Typha sp. is the main source of commission errors. Landscape context also affects the mapping accuracy. We discuss the relevance of our results for mapping invasion patterns in narrow linear wetlands. (C) 2005 Elsevier B.V. All rights reserved.