Sensitivity study of large-scale particle image velocimetry measurement of river discharge using numerical simulation
Résumé
This study deals with the uncertainty of large-scale particle image velocimetry (LSPIV) measurements in rivers. LSPIV belongs to the methods of local remote sensing of rivers, like Radar- and Lidar-based techniques. These methods have many potential advantages, in comparison with classical river gauging, but they have a fundamental drawback: they are indirect measurements. As such they need to be assessed in reference to direct measurements. A first validation method consists in the comparison of LSPIV measurements with classic gauging results, in field and laboratory experiments. Unfortunately, in both cases, it is impossible in practice to control all the parameters and to distinguish the impact of the various error sources. In the present study we propose a more theoretical assessment of LSPIV potential through numerical simulation. The idea is simply to mathematically formulate the present state of knowledge of the measurement including both the physics of the phenomenon (the illuminated river) and the physics of the sensor (the camera and the PIV tracking). The dilemma about when to start this type of simulation is the following: – The simulation is satisfactory if we can validate it which means to be able to compare simulations and observations over a wide range of conditions. – The simulation is useful to get preliminary insights about the most important measurement conditions to organize validation studies. Our simulator is composed of three blocks: (1) The river block represents the unidirectional river flow by the association of the EDM model and a theoretical vertical velocity profile giving a 3D velocity distribution. This hydraulic model is complemented by features representing free surface tracers, the illumination of the free-surface (shadows and sun reflection) and the effect of the wind. (2) The camera block transforms the river state parameters into raster images according to the intrinsic and extrinsic parameters of the camera. (3) The LSPIV analysis block performs a classical LSPIV analysis, including geometric transformation of the images, PIV analysis to obtain a surface velocity field, and discharge computation. We tried to keep a good balance between the different blocks of the simulator (i.e. not to make one component much more sophisticated than the others). The simulator was partly tested during the development of its different blocks, and then globally validated. It reproduced well the variability observed in the field LSPIV experiments conducted with the real-time continuous system of Hauet et al. [Hauet, A., Kruger, A., Krajewski, W., Bradley, A., Muste, M., Creutin, J.D., Wilson, M., 2008. Experimental system for real-time discharge estimation using an image-based method. Journal of Hydrologic Engineering]. The simulator can also be used to check different scenarios and to assess relative importance of the different sources of error. With two examples, we illustrate this capability of the simulator to assess the relative weight of a given error source and to test a new configuration of measurement.