Measuring and Monitoring River Ice Using UAV and Satellite Imagery

Abstract:

Seasonal river ice has profound hydrologic and ecologic effects on river dynamics and nearby social and environmental systems, with economic impacts for North America estimated at more than $250 million annually. River ice is of particular concern on the Yellowstone River (Montana, USA) where ice-jam flooding damages valuable rangeland in rural communities, destroys bridges and other infrastructure, and impacts riparian ecosystems. Despite its wide-ranging importance and destructive potential, river ice is critically understudied, especially in mid-latitude continental river systems like the Yellowstone. The lack of quantitative data prevents federal and state agencies from forecasting river ice beyond seasonal outlooks of potential ice-jam flooding. Here we present results from an initial field campaign to collect aerial images of river ice using a commercially-available unmanned aerial vehicle (UAV). The imagery is processed using Structure-from-Motion (SfM) photogrammetry software to create 3-dimensional models of the ice at the sub-decimeter scale. Ice-free models of the channel created from 3-D SfM and river stage levels serve as baseline conditions to derive quantitative estimates of ice thickness and volume. This field campaign is one component of a larger project to develop a river ice mapping and monitoring tool using a neural network trained on Sentinel 1 and 2 satellite imagery. Initial efforts comparing datasets from these two remote sensing platforms indicate that UAV-based imagery and photogrammetric models can be used as validation data for river ice detected in satellite images. Ultimately this neural network-based monitoring tool works toward integrating river ice into regional flood forecast models, which will provide emergency response teams and water resource managers additional time to react to potential flood events.