Discriminating land cover types using SPOT 4 imagery in the mixed grassland ecosystem
This study was conducted at a mixed grass prairie in southern Saskatchewan. The main objective of this study is to evaluate spectral differentiations of different land cover types and grassland communities, and thus find the way of improving the utility of the land cover data product with remote sensing techniques for the pasture insurance program. Data for this study were field measurements and two SPOT 4 images acquired in the summer of 2005. Using discriminant analysis, the study extracted and analyzed spectral signals from different land cover types. The results showed that a hierarchical classification method was necessary as different level of classification uses different spectral properties. Both June and July imagery can separate the seven major classes with high accuracy and July image is 1% more accuracy than June image. The images from different dates have different advantages for separating classes based on their discriminant functions. It is easier to differentiate land cover types, such as vegetation covered area and non-vegetation covered area, however, the accuracy will be lower when separating the classes of crop and shrub, and fallow and badland due to their similar spectral properties. Three common grassland management practices (cropland, grazed grassland, and ungrazed grassland) can be spectrally discriminated, but it is difficult to separate grasslands from different topography patterns because they have similar spectral features.
The following license files are associated with this item: