Analysis of the accuracy and uncertainty of flow measurements taken by large-scale particle image velocimetry

Information on river flows is essential in water management for the calibration of hydrological models, the estimation of the annual exceedance probabilities and the calibration of the hydraulic models required to delimit flood-prone areas or maintain the safety of the population. Flow measurements are generally carried out using an acoustic Doppler current profiler (ADCP). However, it is sometimes difficult or even dangerous to take measurements using an ADCP, particularly during high flow, when objects (trees, ice) can drift in the river.

In addition, a certain water depth is required for taking measurements with an ADCP. Thus, extreme events are often poorly characterized. Extreme events will occur more frequently with climate change, making better characterization necessary. Large-scale particle image velocimetry (LSPIV) is a non-contact method for measuring river flow. It can be applied during low and high flow. However, the field of application, the acquisition method (drone, camera), and the method’s sources of error and uncertainty must be assessed.
 

• Analyze the performance of the LSPIV method for flow measurements made in Quebec under different conditions

• Identify the LSPIV method’s various sources of uncertainty

• Provide recommendations on LSPIV data acquisition and processing for possible implementation by the MELFFCP

• Data acquisition at different sites representing different hydraulic and morphological conditions 

• Image processing using open access software

• Analysis of results according to acquisition conditions and data processing  

Of the 51 LSPIV measurements acquired, flow calculations for 32 measurements yielded a deviation of less than 10% from the reference measurement. Some rivers showed consistently good results, while others did not appear to present favourable conditions for the LSPIV method.

Figure 1: Comparison of the difference between flows measured (ADCP or station) and calculated using the LSPIV method.

The presence of tracers is the main factor in obtaining good results. A high level of tracers favours good results, as does more diffuse illumination. The distribution of the tracers is of little importance if their concentration is sufficient. Tracers in the form of foam (or air bubbles) appear to be the most suitable for calculating flow using the LSPIV method.

Other sources of uncertainty, such as the position of the cross-section in the image, the accuracy of bathymetric data and water level measurements, the presence of wind or rain/snow, and processing parameters also have an impact.

The use of a drone is recommended, in particular because of the greater flexibility in the choice of measurement location and the possibility of data acquisition at the lowest level, which facilitates image processing.

703900