Handbook for Surface Flow Velocity Measurement

Acquisition of real-time hydraulic data is an essential component for flood forecasting. However, we frequently face difficulties in obtaining discharge data using classical contact methods during high magnitude floods and for systems experiencing rapid hydrogeomorphological adjustment. Therefore, we developed low-cost, non-contact sensors and platforms that are designed to overcome these difficulties. These advances enable flood flow properties to be monitored at multiple locations across a river catchment, at low-cost, and communicated in near real-time by using an image velocimetry method. This is an optics-based approach for stream flow measurement using commercially available nearinfrared digital cameras to acquire video footage in full HD (30fps). Video footage is then subjected to optical flow tracking techniques based on cross-correlation, and feature-based tracking, enabling the displacement rates of detected features (for example natural foam, seeds, woody debris, and turbulent structures) to be computed. This manual provides step by step guidance to install an image-based gauging station. It contains the list of necessary components, the calibration process of a new camera and the assembly procedure of the system.


Background
River water is a vital resource and its supply and management is fundamental for societal well-being and economic development.Globally, floods represent 47% of all weatherrelated disasters and over recent decades have seen increases in occurrence with consequent increases in losses to life and infrastructure in both developing and developed countries (UNISDR, 2015).As such, there is a pressing need to improve our ability to reduce flood risk for less developed countries where sparse hydrometric monitoring networks and poor communications hamper the ability to forecast floods and to provide real-time flood alerts.
Risk to life, infrastructure and property is considerably enhanced when flood waters are charged with sediment and floating debris.Enhanced stream powers during high magnitude floods drive erosion and deposition of sediment which in turn bring about significant changes in channel cross-sectional geometry.Within-flood processes have been responsible for major river channel avulsions into populated areas (Major et al. 2016;Wilcox et al., 2016).Increased supply of sediment to fluvial systems through natural or anthropogenic processes can also have a major impact on downstream river channel dynamics resulting in localised aggradation and the downstream passage of sediment waves.Large volumes of woody debris may also be introduced suddenly to fluvial systems via landslides or following volcanic activity.Such hydrogeomorphological activity poses a major hazard to population and infrastructure in both developed and developing nations.
Acquisition of real-time hydraulic data is an essential component for flood forecasting, allowing validation of rainfall-runoff and hydraulic modelling approaches and providing information directly to hazard managers.However, the ability to obtain discharge data during high magnitude floods and for systems experiencing rapid hydro-geomorphological adjustment and is severely reduced by high sediment loads and floating debris which frequently clog conventional stage-recording equipment and compromise the integrity of stage-discharge relationships due to changes in channel cross-sectional geometry.
Difficulties involving the instrumentation of catchments deemed to be vulnerable to extreme runoff events in Chile, and other developing countries include: (i) the technical challenges of building accurate and resilient systems; (ii) the cost of implementation and maintenance, and (iii) access to adequately trained staff.Therefore, we developed low-cost non-contact sensors and platforms that are designed to overcome these difficulties.These advances enable flood flow properties to be monitored at multiple locations across a river catchment, at low-cost, and communicated in near real-time.This allows measurement of fundamental hydraulic processes, and provides the potential for enhanced flood warnings during high magnitude events when traditional river gauging methodologies often fail (Perks et al., 2016).This new data can be used for short term, early warning (to protect lives and infrastructure such as roads, bridges, drinking water plants, hydropower plants and dams), as well as longer-term areal coverage (where the established national network cannot cover due to costs or other considerations), with direct development implications including hazard maps and infrastructure site planning.

Methodology
Image velocimetry is an optics-based approach for stream flow measurement using commercially available near-infrared digital cameras to acquire video footage in full HD (30fps).Video footage is then subjected to optical flow tracking techniques based on crosscorrelation, and feature-based tracking, enabling the displacement rates of detected features (for example natural foam, seeds, woody debris, and turbulent structures) to be computed.First, we extract video frames from the footage, then, georeference them to convert image pixels to real-world coordinates.Second, we extract the start and end position of selected surface water features, then convert them to real-word coordinates to, finally, generate vectors of water velocities.The computation of water velocity vectors are achieved through application of several methodological approaches including large scale particle image velocimetry (LSPIV) (Muste et al. 2008, LeCoz et al. 2010), particle tracking velocimetry (PTV) (Tauro and Grimaldi 2017), and Kanade-Lucas-Tomasi (KLT) flow tracking (Perks et al. 2016).Following the determination of the surface velocity, a site-specific velocity coefficient can be calculated to translate surface velocities to depth-averaged velocities.To calibrate the relationship between surface velocity and discharge or water level, site-specific flow data (e.g.ADCP data) is necessary.

Equipment
The following table lists the major components necessary for an installation when a solar panel is used for power supply and an ultrasonic level sensor added.The use of an ultrasonic level sensor is optional; however additional hydraulic information (e.g.water surface slope) can be gained by installing a pair of downward facing ultrasonic water level sensors for mounting on bridges or structures where there is a vertical view angle.The white and black cables are connected to the temperature sensor.These two cables have to be soldiered on the sensor previously.

Item
Add a capacitor to the red and black cable of the ultrasonic sensor.The positive (longer) part goes to the red cable, the negative (short) one to the black.
Finally connect the 3 cables of the ultrasonic sensor (brown, orange, black) and the green (measuring the charging level of the battery) to the connection board (see wiring diagram).
6.The camera is powered directly from the battery.Connect the camera to the battery and use the Ethernet cable to connect the camera to the Raspberry Pi.
6 Field installation 1. Set up the system as described earlier 2. Place the SD card in the Pi 3. Prepare the Ethernet cable the following way and connect the system to the power 7 Site selection criteria and advice for install  Ideally choose a site where you can detect the flow moving with your eyes.It is helpful if some floating features (e.g.white foam, standing waves, plant material) are present on the water surface. Check that you can see all the river or channel surface. Check mobile phone reception at the site. Favour installation where main power can be used instead of solar panels.It will be more reliable and will need less maintenance. Place the camera in an elevated position. The camera needs a clear view on the river; avoid for example branches within camera view and position the camera so it is not too far from the river to reduce the impacts of weather conditions which may lower visibility. If possible, set-up the camera in a way that the entire width of the channel is visible. Avoid facing the camera to the sun, reflection might cause an issue.
 Consideration is needed for solar panel installation which should be located in a secure location well above inundation level but as close to the camera as possible in order to minimise the power requirements which increase with longer cables. The distance between the camera and the grey box should not exceed 20m, the ultrasonic sensor 300m.

Access to camera from distance
You can connect to your camera to view a real-time image using iVMS-4500 app by using the IP address, the port and the password of the camera.

4 .
Connect the converter to the Raspberry Pi with the micro USB port. 5. Prepare the following wiring to power the Raspberry Pi and the ultrasonic sensor with 5V instead of 12V.Please note that the images slightly differ from the wiring diagram, follow the wiring diagram.