Using multi-frequency synthetic aperture radar to develop regional-scale groundwater maps
Babonis, Gregory S.
MetadataShow full item record
The majority of accessible, potable water exists as ground water. Understanding regional-scale, ground-water transport is therefore vital to water resource management. Considering surface water as an outcropping of shallow ground-water flow, wetlands and seepage lakes can be used to determine head measurements of subsurface water transport. However, field-based, regional-scale collection of surface water elevation measurements can be geographically and logistically prohibitive. This study involves the development of a classification system for wetlands in the Northern Highlands Lakes Region (NHLR) of Vilas County, Wisconsin, based upon multi-frequency Airborne Synthetic Aperture Radar (AIRSAR) backscatter responses. Field work conducted in October, 2007 provided the necessary groundtruth for creating training and test sets for a supervised classification. The resulting classification system was used to extract elevation data from the Shuttle Radar Topography Mission (SRTM) global coverage Digital Elevation Models (DEMs) to create a prediction map of groundwater contours. An enhanced Lee and mean filtered P/C-band ratio image was classified using training sets and a maximum likelihood supervised classifier. The classification resulted in an overall accuracy of 84.21%, an improvement over the Landsat-based WISCLAND (Wisconsin Initiative for Statewide Cooperation on Landscape Analysis and Data) classification accuracy of 78%. The AIRSAR classification of closed wetlands was greatly improved over the WISCLAND because of radar's ability to penetrate canopy cover. The AIRSAR classification achieved a user's accuracy 85.17% for closed wetlands, 87.62% for open wetlands, and 74.1% for mixed wetlands. The resulting classification was processed according to the Upper Midwest Gap Analysis Program (UMGAP) protocol, the same protocol used for WISCLAND processing. Constraints within the processing software made it impossible to sieve the AIRSAR classification to the same degree as the WISCLAND classification. The resulting AIRSAR UMGAP processed layer had many more pixels than the WISCLAND wetlands layer. Using the WISCLAND wetlands layer, the WISCLAND open wetlands layer, and the AIRSAR classification open wetlands layer, elevation data was extracted from a Shuttle Radar Topography Mission Digital Elevation Model (SRTM DEM) of the study area. These elevation data were compared to the National Elevation Dataset (NED) DEM of the same area to determine the accuracy of SRTM for the different wetland layer areas. The results show that using only the open wetlands layers to extract elevations improves SRTM elevation accuracy over using the entire wetlands layer to extract elevations. However, both the WISCLAND and AIRSAR classification open wetlands layers had approximately the same accuracy for extracted SRTM elevation data. Elevation data extracted from the SRTM DEM using the AIRSAR classification of open wetlands were interpolated through Ordinary Kriging techniques to produce a groundwater contour prediction map. This map is shown to be comparable to a groundtruthed water contour map from a previous study (Fredrick et al., 2007), with discrepencies caused by registration errors between AIRSAR and SRTM data, suggesting that remotely-sensed open wetland elevations are suitable for the calibration of regional-scale groundwater models. References. Fredrick, K.C., Becker, M.W., Matott, L.S., Daw, A., Bandilla, K, and Flewelling, D.M., 2007, Development of a numerical groundwater flow model using SRTM elevations. Hydrogeology Journal, 15:171-181.