Classification of wetland ground-water interactions using remote sensing
Piurek, Robert B
MetadataShow full item record
The application of remote sensing technology can potentially be used as a cost-effective method for locating regions of ground-water discharge and recharge. The overarching objective of this thesis is to determine if, under certain hydrologic conditions, ground-water discharge to wetlands can be estimated using remote sensing methods. The specific hypothesis to be tested here is that certain ground-water indicators are significantly correlated with vegetation populations. Ground-water indicators include wetland water pH, electrical conductivity, silica concentration, as well as landscape position. If these groundwater indicators are related to wetland vegetation population, then it is possible that remotely sensed vegetation population could be used to estimate ground-water discharge to wetlands. Visible and near infrared data are collected from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) instrument. Wetlands in the Northern Highlands region of Wisconsin are considered for this study. Previous research in this region have shown that there is a clear relationship with ground-water discharge and lake water chemistry, and it is assumed that the same relationship holds true for wetlands. A method called "wetland ordering", similar to stream and lake ordering, is defined here for classifying the relationship of a wetland to its surface hydrology and landscape position. Wetland vegetation sampling included releve sampling in specific wetlands. Wetland water sampling included in-situ pH and electrical conductivity, as well as the collection of water samples for silica concentration analysis. ANCOVA analyses of these data show that wetland water pH, electrical conductivity, and silica concentration are not correlated to wetland order as was seen in previous research with regard to lake order, and are therefore poor indicators of ground-water discharge. However, the results of the ANCOVA analyses indicate that wetland order as a ground-water indicator is strongly correlated to certain vegetation types. Vegetation cover was categorized using end-member classification. Two different types of wetland vegetation end-member supervised classifications were performed (parallelepiped and maximum likelihood classifications) and were shown to be 47% to 90% accurate when compared with ground-truthed vegetation coverage. While the ANCOVA results show that wetland order is correlated to certain wetland vegetation, wetland order and vegetation type may alternatively be related by surface water or other landscape dependant variables. We cannot test with certainty which is the case without a reliable indicator of ground-water discharge to wetlands. This study suggests that visible/near-infrared satellite data should be further investigated as a means to describing the subsurface component of the hydrologic cycle.