An object-based approach to delineate wetlands across landscapes of varied disturbance with high spatial resolution satellite imagery ☆

摘要: Mapping wetlands across both natural and human-altered landscapes is important for the management of these ecosystems. Though they are considered landscape elements providing ecological socioeconomic benefits, accurate wetland inventories do not exist in many areas. In this study, a multi-scale geographic object-based image analysis (GEOBIA) approach was employed to segment three high spatial resolution images acquired over varying heterogeneity due human-disturbance determine robustness method changing scene variability. Multispectral layers, digital elevation layer, normalized-difference vegetation index (NDVI) first-order texture layer were used segmentation scales with focus on delineation boundaries components. Each ancillary input contributed improving at different scales. Wetlands classified using nearest neighbor relatively undisturbed park site an agricultural GeoEye1 imagery, urban WorldView2 data. Successful classification achieved all study sites accuracy above 80%, though results suggest that overall higher degree may negatively affect classification. The suffered from greatest amount under segmentation, lowest map (kappa: 0.78) which partially attributed confusion among greater proportion mixed vegetated classes uplands. Accuracy individual based Canadian Wetland Classification system varied between each site, kappa values ranging 0.64 swamp class 0.89 marsh class. This research developed unique mapping various degrees disturbance GEOBIA, can be applied other similar settings.