Comparative analysis of urban wetland classification from multispectral, hyperspectral, and fused satellite imagery

Abstract

Protection and inventory of current wetlands is important for critical habitats, education, recreation, flood control, and storm water filtration. Over the years, the United States has lost over one-half of its wetlands due to natural processes and urban development. Satellite imagery was used to test the possibility of accurately mapping existing wetlands located in a subset study site along the confluence of the Missouri River and the Blue River in Kansas City, Missouri, following the National Wetlands Inventory (NWI) classification scheme. Testing was performed using hyperspectral (Hyperion) and higher spatial resolution multispectral (SPOT 5) imagery. Additionally, testing was performed by fusing the hyperspectral with multispectral imagery. Results indicated that the fused imagery produced a classified landscape with higher overall accuracy, as well as increased accuracy within the individual wetland classes. These results are significant because they indicate the possibility of an inexpensive, accurate approach for classifying wetlands to obtain metrics for wetlands within even larger study areas.