Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Hydrology and Remote Sensing Laboratory » Research » Publications at this Location » Publication #374463
Research Project: Integrating Remote Sensing, Measurements and Modeling for Multi-Scale Assessment of Water Availability, Use, and Quality in Agroecosystems

Location: Hydrology and Remote Sensing Laboratory

Title: Long-term trends in root-zone soil moisture across CONUS connected to ENSO

Author
item TOBIN, K. - Texas A&M University
item TORRES, R. - Texas A&M University
item BENNETT, M.E. - Texas A&M University
item DONG, J. - US Department Of Agriculture (USDA)
item Crow, Wade

Submitted to: Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/30/2020
Publication Date: 7/15/2020
Citation: Tobin, K., Torres, R., Bennett, M., Dong, J., Crow, W.T. 2020. Long-term trends in root-zone soil moisture across CONUS connected to ENSO. Remote Sensing. 12:2037. https://doi.org/10.3390/rs12122037.
DOI: https://doi.org/10.3390/rs12122037

Interpretive Summary: Forested wetlands are most common along the Atlantic Coastal Plain including the Delmarva peninsula. The inundation status of those wetlands provides a key indicator of climate variability and shifts in hydrological (e.g. floodwater storage), biogeochemical (e.g. carbon sequestration), and biological (e.g. habitat) functions. However, many of these forested wetlands occur in small, shallow depressions and possess an overstory of forests and many wetlands are only inundated for a short period throughout the year, usually in early spring. These characteristics are a challenge to mapping wetland inundation using commonly available optical imagery. In this study, we used artificial intelligence (AI) models based on deep neural networks to map wetland inundation using optical imagery and demonstrate that AI has an advantage over other statistical approaches for mapping inundation. This use of AI will greatly improve accuracy of wetland maps and improve estimates of ecosystem services provision by wetlands in agricultural landscapes.

Technical Abstract: The Delmarva peninsula in the eastern United States is partially characterized by thousands of small, forested depressional wetlands that are highly sensitive to weather variability and climate change, but provide critical ecosystem services. Due to the relatively small size of these depressional wetlands and their occurrence under forest canopy cover, it is very challenging to map their inundation status based on existing remote sensing data and traditional classification approaches. In this study, we applied a deep convolutional neural network to map forested wetland inundation in the Delmarva area by integrating leaf-off Worldview-3 (WV3) multispectral data with fine spatial resolution light detection and ranging (lidar) intensity and topographic data, including a digital elevation model (DEM) and topographic wetness index (TWI). Wetland inundation maps generated from lidar intensity were used for model calibration and validation. The wetland inundation map results were also validated using field data and compared to the U.S. Fish and Wildlife Service National Wetlands Inventory (NWI) geospatial dataset and a random forest output from a previous study. Our results demonstrate that deep learning models can accurately determine inundation status with an overall accuracy of 95% compared to field data and high overlap with lidar mapped inundation. The integration of topographic metrics in deep learning models can improve classification accuracy for depressional wetlands. This study highlights the great potential of deep learning models to improve the accuracy of wetland inundation maps through use of high resolution optical and lidar remote sensing datasets.