Representative ice thickness data is essential for accurate hydraulic modelling, assessing the potential for ice induced floods, understanding environmental conditions during winter and estimation of ice-run forces. Steep rivers exhibit complex freeze-up behaviour combining formation of columnar ice with successions of anchor ice dams to build a complete ice cover, resulting in an ice cover with complex geometry. For such ice covers traditional single point measurements are unrepresentative. Gathering sufficiently distributed measurements for representativeness is labour intensive and at times impossible with hard to access ice. Structure from Motion (SfM) software and low-cost drones have enabled river ice mapping without the need to directly access the ice, thereby reducing both the workload and the potential danger in accessing the ice. In this paper we show how drone-based photography can be used to efficiently survey river ice and how these photographic surveys can be processed into digital elevation models (DEMs) using Structure from Motion. We also show how DEMs of the riverbed, riverbanks and ice conditions can be used to deduce ice volume and ice thickness distributions. A QGIS plugin has been implemented to automate these tasks. These techniques are demonstrated with a survey of a stretch of the river Sokna in Trøndelag, Norway. The survey was carried out during the winter 2020–2021 at various stages of freeze-up using a simple quadcopter with camera. The 500 m stretch of river studied was estimated to have an ice volume of up to 8.6 × 103 m3 (This corresponds to an average ice thickness of ∼67 cm) during the full ice cover condition of which up to 7.2 × 103 m3 (This corresponds to an average ice thickness of ∼57 cm) could be anchor ice. Ground Control Points were measured with an RTK-GPS and used to determine that the accuracy of these ice surface geometry measurements lie between 0.03 and 0.09 m. The ice thicknesses estimated through the SfM methods are on average 18 cm thicker than the manual measurements. Primarily due to the SfM methods inability to detect suspended ice covers. This paper highlights the need to develop better ways of estimating the volume of air beneath suspended ice covers.
Read full abstract