An evaluation of remotely sensed wetland mapping
Gluck, Michael J.
Master of Science
DisciplineForestry and the Forest Environment
SubjectWetlands remote sensing
Development of wetland mapping techniques
MetadataShow full item record
Landscape management is based on the maintenance of natural ecosystems and recognizes the importance of maintaining the habitat diversity of all ecosystem types. Acquiring information about the size, distribution and location of wetlands is the first step towards evaluating their habitat value in a landscape perspective. An explicit review about the strengths and limitations of any landcover database is critical prior to input into the decision making process. Techniques were developed for characterizing wetland habitat components in a landscape context utilizing remote sensing and geographic information system technologies. A hierarchy of remotely sensed data ranging from 1:5000 colour infrared aerial photography to LANDSAT Thematic Mapper satellite data was employed to compare detail of information available at each scale of data. These techniques included evaluation of ground-based wetland classification systems, air photo interpretation, investigation of approaches to image classification, and development of accuracy assessment techniques. The developed techniques were applied to a Northwestern Ontario landscape to produce a thematic layer of wetland habitat information. The effectiveness of these techniques was evaluated by assessing the accuracy of each remote sensing scale for mapping the broad scale wetland habitat at the physiognomic group level. 1:5,000 and 1:10,000 scale colour infrared aerial photography provided the best thematic accuracy at 94 percent, whereas 1:20,000 scale allowed wetland mapping at 84 percent accuracy. Satellite based mapping using Landsat Thematic Mapper integrated with digital Forest Resource Inventory map data allowed wetlands to be mapped with 72 percent accuracy. Combining physiognomically similar wetland classes increased satellite based mapping accuracy to 81 percent.