Each spring since 1987, FWS staff have conducted the Four-Square-Mile Breeding Waterfowl Survey across five states and two FWS regions in the Prairie Pothole Region of the northern Great Plains. Part of the study includes the manual digitization of surface-water extent for approximately 75,000 wetlands on 2 mi x 2 mi sample plots to monitor wetland conditions and provide baseline data for modeling annual waterfowl distribution and abundance. In 2014, Habitat and Population Evaluation Team (HAPET) researchers began integrating new camera systems into the remote sensing protocol to collect submeter color infrared imagery at 380 sample plots. These changes were implemented with the intent of introducing automated water classification algorithms into the workflow to reduce annual effort and increase accuracy.
Using images collected in spring 2015, researchers used eCognition software to produce an automated surface-water classification. Although the classification outputs continue to require some manual interpretation, the average time required to process each remote sensing plot has been reduced approximately 60 percent (from 2 hours to 45 minutes per plot, or ~500 hours less effort per year). With continued enhancement to the classification algorithm, the workload could be reduced further. Additionally, changes were made to the image acquisition protocols that also reduced pilot interaction with the system. Efforts are currently underway to introduce a more robust flight management system (incorporating highly accuracte navigational tools and automated camera control) to further reduce the burden on pilots and increase operational safety.
This protocol has also been used to provide input data for studies identifying priority habitat for the rare Wyoming toad (Anaxyrus baxteri), provide high-resolution (3-cm ground sample distance (GSD)) orthophotos to inventory three American white pelican nesting colonies (approximately 25,000 individuals), and investigate effects of oil/gas exploration on waterfowl abundance and productivity.
This series demonstrates the utility of color infrared (and specifically near-infrared light) for mapping surface water and ultimately creating automated classification of water.