Enhancing Drainage Delineation Using High-resolution Terrain Data Model and Geospatial Artificial Intelligence

Abstract:

Accurate and high-resolution hydrologic drainage features are critical to support a wide range of watershed management issues, such as overland nutrient transport and aquatic species passage. Albeit the increasing availability of high-resolution Digital Elevation Models (HRDEMs), fine-scale terrain-based hydrologic drainage delineation is often hindered by widespread virtual topographic barriers near road culverts and bridges. There is a critical need for the location datasets of culverts and bridges that can be used as critical breaklines for hydrologic enforcement of HRDEMs at the field scale. At present, the common practice is to manually identify the location where a stream or drainage crosses a road on aerial orthophotos, and then use the laborious on-screen digitization to develop the data. As a departure, we used a geospatial artificial intelligence (GeoAI) approach to develop a multiband image training dataset and a deep learning model. The preliminary results show that GeoAI is a promising approach to identify the drainage structure locations. In addition, the effects of drainage structures, HRDEM resolutions, and flow direction algorithms on local drainage crossings were assessed via controlled experiments for an optimal combination of variables.