Near-field Remote Sensing Of Riverine Hydrodynamic Processes With 3D Large Scale Particle Image Velocimetry

For mountain areas, flash flood is common natural disaster that occurs frequently in rainy seasons, however, our current knowledge on the formation and development is largely based on theoretical derivation and experimental studies with normal conditions, lacking of field measurement and data under extreme conditions. The continuous monitoring of medium- and small-scale rivers, especially for gaining the essential flow information, such as water stage, velocity field and discharge of flood, is key to forecasting and management work. Nevertheless, this kind of measurement is of difficulties to be conducted during flood and confronted with great challenges to deal with.In this research, we aimed to develop a set of near-field remote sensing system for natural rivers utilizing simple, low-cost components and conduct hydrologic measurements where traditional techniques cannot work. Based on the Large Scale Particle Image Velocimetry (LSPIV) and stereo imaging technologies, the non-intrusive system was raised and deployed successfully in the ungauged catchment, Dujiangyan, China. This system is built with low-cost, off-the-shelf components, which can acquire continuous images/videos automatically at programmed intervals. Hydraulic parameters, such as three-dimensional bathymetry, water stage, velocity field distribution, discharge, Froude number and etc., could be obtained with data analysis. The results of the proposed study could provide valuable data that significantly help the understanding on the dynamic processes of flash flood and debris flows in mountain areas, and support the development of physical-based numerical models. In the future, this method could also serve as a technological foundation for watershed-scale flood flow measurements via remote sensing.The main conclusions and highlights of this research are as follows:(1) Mutiple subjects, like hydrology, image processing and field surveying, have been combined to explore the theories and methods for flow measurements in remote ungauged mountainous catchment. The equipment called’near-field remote sensing of riverine hydrodynamic processes with 3D Large Scale Particle Image Velocimetry’ has been developed, solving measurement and survey issues in a way. While this system might be applied in China for the first time, it performs to be helpful for obtaining hydrologic data in ungauged catchment and has great potentials to be widely used;(2) We applied binocular stereo imaging algorithm for three-dimensional reconstruction of static topography, dynamic water stage and water surface morphology and proposed two algorithms for different measurements. Point cloud method is recommended for wave and water surface morphology calculation, and virtual pole is better used for specific cross sectional bathymetry. Meanwhile, uncalibrated rectification method should be applied for image pre-processing, dealing with the matching issues resulted from the larger parallax and improving the epipolar search process;(3) Previously, water surface is usually simplified as two-dimensional horizontal plane or oblique plane with a fixed angle. Supported by real-time water stage and surface data via three-dimensional reconstruction, the image ortho-rectification has established more accurate relationship between physical coordinates and image gray-values, improving the accuracy of surface velocity field calculation. Instead of traditional rectangle meshgrid, we found that triangle meshgrid is more flexible and suitable for sinuous channel, in case of wasting time on numerous null nodes;(4) A relative robust algorithm Minimum Quadratic Difference (MQD) has been added in LSPIV core procedure for tracking the natural surface patterns. Under the non-uniform conditions, MQD performs better for more reliable results. And a comprehensive sensitivity analysis shows that the calculation time and accuracy of LSPIV have complicated relations to many factors, including tracer density, interrogation window size, time interval and etc;(5) The whole procedure has been validated by a small-scale flash flood occurring in July,2014, Longxi River. The cross-sectional bathymetry has been measured by virtual pole method, which could help determining real-time water stage together with image data. Discharge, flow depth, Froude number and manning coefficient have been obtained simultaneously.This research verified that this system worked well, and also gained much experience both on hardware and software, which could help system sustain updated in the future. It makes non-intrusive hydrologic equipment developing towards more automated, intelligent and Internet-based, which provides raw data for disaster warnings and hydraulic construction projects.