High resolution channel geometry from repeat aerial imagery

Abstract:

River channel cross sectional geometry is a key attribute for controlling the river energy balances where surface heat fluxes dominate and discharge varies significantly over short time periods throughout the open water season. These dynamics are seen in higher gradient portions of Arctic rivers where surface heat fluxes can dominates river energy balances and low hillslope storage produce rapidly varying hydrographs. Additionally, arctic river geometry can be highly dynamic in the face of thermal erosion of permafrost landscape. While direct in-situ measurements of channel cross sectional geometry are accurate, they are limited in spatial resolution and coverage, and can be access limited in remote areas. Remote sensing can help gather data at high spatial resolutions and large areas, however techniques for extracting channel geometry is often limited to the banks and flood plains adjacent to river, as the water column inhibits sensing of the river bed itself. Green light LiDAR can be used to map bathymetry, however this is expensive, difficult to obtain at large spatial scales, and dependent on water quality. Alternatively, 3D photogrammetry from aerial imagery can be used to analyze the non-wetted portion of the river channel, but extracting full cross sections requires extrapolation into the wetted portion of the river. To bridge these gaps, an approach for using repeat aerial imagery surveys with visual (RGB) and near infrared (NIR) to extract high resolution channel geometry for the Kuparuk River in the Alaskan Arctic was developed. Aerial imagery surveys were conducted under multiple flow conditions and water surface geometry (elevation and width) were extracted through photogrammetry. Channel geometry was extracted by combining water surface widths and elevations from multiple flights. The accuracy of these results were compared against field surveyed cross sections at many locations throughout the study reach and a digital elevation model created under extremely low flow conditions. These extrapolation methods have shown to be promising for estimating detailed channel geometry at large scales.