EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Sources of uncertainty in video-based flow observations, revealed by co-location experiment

Hessel Winsemius1,2,3, Salvador Peña-Haro4,5, Frank Annor3,6, Rick Hagenaars3,6, Wim Luxemburg3, Gijs Van den Munckhof7, Felix Grimmeisen5, and Nick Van de Giesen3,6
Hessel Winsemius et al.
  • 1Deltares, Delft, Netherlands (
  • 2Rainbow Sensing, Den Haag, Netherlands
  • 3Faculty of Civil Engineering and Geosciences, Delft U/niversity of Technology, Delft, Netherlands
  • 4Photrack AG, Zürich, Switserland
  • 5SEBA Hydrometrie GmbH, Kaufbeuren, Germany
  • 6Trans-African Hydrometeorological Observatory (TAHMO), Nairobi, Kenya
  • 7Waterboard Limburg, Roermond, Netherlands

In the last years, several methods to establish surface flow velocities and river flow from camera videos have been developed and codified into software. Together with a hardware setup, these may be used to establish near real-time observations of river flow. The hardware setup used and associated quality of the camera, methods to pre-process, process and post-process the videos may all result in errors, and uncertainties. In this contribution we assess what the main sources of uncertainty are, and under what conditions these may appear, focusing on both hardware and processing methods. We do this by co-locating two different camera setups, and using two different software processing methods. For camera setups we use a very simple and low cost FOSCAM FI9900EP running at its maximum of 4Mbps and a much better quality Vivotek IB9367-EHT running at 20Mbps. As systems we use the DischargeKeeper and pyOpenRiverCam.

The cameras were co-located over a significantly long period at a site in Limburg in The Netherlands, and footage analyzed with 15-minute intervals. Videos were treated with as much as possible the same settings, reprojection resolution and window. Results were compared in terms of the ability to resolve velocities (amount and quality) and the impact of post-processing. Integrated flow over a cross-section is also compared. We assess under what conditions flow and velocity estimates are robust and similar and under what conditions these diverge focusing on the platform used, light conditions, and flow conditions.

Keywords: River flow monitoring, stage-discharge relationships, OpenRiverCam, DischargeKeeper, computer vision

The work leading to these results has received funding from the German Federal Ministry of Education and Research (BMBF) and the CLIENT II program (Drought-ADAPT, FKZ: 01LZ2002B) and the European Horizon Europe Programme (2021-2027) under grant agreement no. 101086209 (TEMBO Africa). The opinions expressed in the document are of the authors only and no way reflect the European Commission’s opinions. The European Union is not liable for any use that may be made of the information.

How to cite: Winsemius, H., Peña-Haro, S., Annor, F., Hagenaars, R., Luxemburg, W., Van den Munckhof, G., Grimmeisen, F., and Van de Giesen, N.: Sources of uncertainty in video-based flow observations, revealed by co-location experiment, EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-14234,, 2023.