Tài liệu hạn chế xem trước, để xem đầy đủ mời bạn chọn Tải xuống
1
/ 12 trang
THÔNG TIN TÀI LIỆU
Thông tin cơ bản
Định dạng
Số trang
12
Dung lượng
3,44 MB
Nội dung
A Free and Simple Satellite‐based Mapping Technique for Chinese Privet James S. Cash Graduate Research Assistant, School of Forestry and Wildlife Sciences, Auburn University Photo credit: ESA Chinese Privet (Ligustrum sinense) • Invasive semi‐evergreen shrub common in hardwood forests of the southeastern US • Dense thickets outcompete native plants • Negative effects on forest regeneration • Potential for wildlife habitat degradation How Much and Where? • Mapping privet on a property can aid in planning control operations • This technique is a simple and free first step in the planning process Invaded acres: 1,371 Percent of property: 24.06% Satellite Imagery Sentinel‐2 Moderate Resolution Multispectral Imagery • 10m spatial resolution • 13 spectral bands • Free downloads from earthexplorer.usgs.gov March 9, 2017 Infrared false color composite of study site in western AL Software Free downloads at qgis.org Semi‐Automatic Classification Plugin Supervised Classification Step 1: Create Training Sites Train software to recognize all major land cover types in area of interest Ex: Invaded hardwoods (privet), uninvaded hardwoods, pine, field, swamp, water 2‐3 training sites per cover type to capture variation in spectral signatures Spectral Signatures Image credit: seos‐project.eu Supervised Classification Step 2: Sort Pixels Algorithm Choices • Minimum Distance • Maximum Likelihood • Spectral Angle Mapping • Land Cover Signature Classification Supervised Classification Step 3: Accuracy Assessment Informal • • • • How does the map look? Adjust training polygons as necessary Try different classification algorithms Manually edit map Formal • Survey vegetation at random points (250 used in study) • Calculate accuracy statistics Accuracy Results Assuming privet cover threshold of 40% Overall Accuracy: 91.6% Producer’s Accuracy: 75.6% User’s Accuracy: 77.3% Overall Accuracy: (True Positives + True Negatives) / Total # Plots Producer’s Accuracy: True Positives / (True Positives + False Negatives) User’s Accuracy: True Positives / (True Positives + False Positives) Additional Applications • Other Invasive Plant Species • Needs to reach high enough densities to register on 10m pixel • Must be spectrally distinct from non‐target species/ land cover • Evergreen vs. deciduous forest cover • Impervious cover within a watershed • Forest cover loss due to harvest, wildfire, insect damage, etc • Agricultural crops • Much more! Photo credit: ESA and Airbus Defense and Space Acknowledgements I’d like to thank the following organizations for providing financial and/or logistical support for this and related research. I’d also like to thank the Harry Murphy Dean’s Enhancement Fund for Excellence in The School of Forestry and Wildlife Sciences at Auburn University for providing assistantship funding Major Advisor Dr. Christopher J. Anderson Committee Members Dr. Luke Marzen Dr. John Kush Dr. William Gulsby