| Abstract: |
Wetlands are recognized as one of the world’s most valuable natural resources. With the increasing world population, human demands on wetland resources for agricultural expansion and urban development continue to increase. In addition, wetlands around the world are facing unprecedented threats due to climate change. To better manage and conserve wetland resources, there is a critical need to document the spatial and temporal distribution of wetlands and monitor their dynamic changes. The National Wetlands Inventory (NWI) developed by the U.S. Fish and Wildlife Service in the 1980s is considered the most spatially and categorically detailed wetland inventory for the conterminous United States. The NWI maps were primarily produced by manually interpreting the mid-1980’s aerial photographs with subsequent support from soil surveys and field verifications. NWI is a static dataset that might not reflect current wetland conditions, especially in areas where profound changes have occurred over the past decades due to natural and anthropogenic impacts. Satellite data are increasingly being used to map and characterize wetlands. However, few satellite sensors have the capability to map wetlands at the spatial, temporal, and thematic resolutions required to meet NWI mapping standards. The objective of this project is to develop a geospatial framework for operational wetland mapping using a combination of fine-resolution aerial imagery, high-performance computing, and state-of-the-art machine learning algorithms. This project will focus on mapping wetland dynamics in the Prairie Pothole Region, which covers five U.S. states and is characterized by millions of depressional wetlands. The expected results and data products from this project will contribute to updating the National Wetlands Inventory as well as wetland science and conservation. Training deep neural network models on fine-resolution aerial imagery at a large geographic extent is extremely computationally intensive, and beyond what can be done with the workstation I have at the University of Tennessee, Knoxville. |
0000-0001-5437-4073