Mapping an invasive bryophyte species using hyperspectral remote sensing data

详细信息    查看全文

• 作者：Sandra Skowronek ; Michael Ewald ; Maike Isermann…
• 关键词：Campylopus introflexus ; Dunes ; Heathland ; Imaging spectroscopy ; Maxent ; Moss
• 刊名：Biological Invasions
• 出版年：2017
• 出版时间：January 2017
• 年：2017
• 卷：19
• 期：1
• 页码：239-254
• 全文大小：
• 刊物类别：Biomedical and Life Sciences
• 刊物主题：Ecology; Freshwater & Marine Ecology; Plant Sciences; Developmental Biology;
• 出版者：Springer International Publishing
• ISSN：1573-1464
• 卷排序：19

文摘

Reliable distribution maps are crucial for the management of invasive plant species. An alternative to traditional field surveys is the use of remote sensing data, which allows coverage of large areas. However, most remote sensing studies on invasive plant species focus on mapping large stands of easily detectable study species. In this study, we used hyperspectral remote sensing data in combination with field data to derive a distribution map of an invasive bryophyte species, Campylopus introflexus, on the island of Sylt in Northern Germany. We collected plant cover data on 57 plots to calibrate the model and presence/absence data of C. introflexus on another 150 plots for independent validation. We simultaneously acquired airborne hyperspectral (APEX) images during summer 2014, providing 285 spectral bands. We used a Maxent modelling approach to map the distribution of C. introflexus. Although C. introflexus is a small and inconspicuous species, we were able to map its distribution with an overall accuracy of 75 %. Reducing the sampling effort from 57 to 7 plots, our models performed fairly well until sampling effort dropped below 12 plots. The model predicts that C. introflexus is present in about one quarter of the pixels in our study area. The highest percentage of C. introflexus is predicted in the dune grassland. Our findings suggest that hyperspectral remote sensing data have the potential to provide reliable information about the degree of bryophyte invasion, and thus provide an alternative to traditional field mapping approaches over large areas.