Testing Multiple Analyses of Remotely Sensed Data for the Automated Detection of Anoxic Groundwater Seepages
Abstract
Characteristic rust-colored staining is often observed around wetland margins and along the river corridor in discrete patches coincident with hyporheic and groundwater seepage. This staining specifically occurs when seepage water low in dissolved oxygen and high in reduced (dissolved) iron reaches the oxygenated land surface, causing precipitation of iron oxides that often correspond with biogeochemical hot spots or zones of contaminant loading. Remote sensing of exposed bank soils and the streambed of shallow, clear streams can potentially be used to efficiently map deposits of iron oxides and represent a new, scalable method of assessing reactive groundwater/surface water exchange processes. The East River watershed located in Western Colorado, USA has numerous iron oxide deposits along the mainstem river and tributary corridors, some of which are influenced by legacy mine contaminants and others by beaver activity. Using aerial hyperspectral data collected in 2018 by the National Ecological Observatory Network (NEON) along with orthomosaics derived from photos taken by an unmanned aerial system (UAS) in 2017 and 2018 by U. S. Geological Survey, we tested a supervised classification model to automatically delineate iron oxide staining present in the study area. We used a beaver-impacted reach of Coal Creek, a tributary to the East River, as a training dataset and portions of the East River to test the trained model. The classification model accurately delineated known iron oxide staining when tested on reaches outside of the training data Additionally, the hyperspectral data was used to create visual indices unique to specific species of iron oxide to highlight their presence or absence in the study area, and initial results were positive for the oxide minerals goethite and hematite. A similar approach was tested for more limited multi-spectral data collected by UAS, but those spectral bands did not align well with expected oxide mineral reflectance. This work investigates multiple, scalable methods to automatically delineate anoxic groundwater seepages from remotely sensed data using iron oxide deposits as a proxy. Additionally, results indicate the possibility of using visual indices derived from hyperspectral data to identify other materials with unique optical signatures at a variety of scales.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.H35A..03C