Machine learning + non-contact radar change hydrography
January 18, 2020
Advances in computing and contactless water measurements are revolutionizing hydrology and hydrography. Non-contact radar sensors have the ability to accurately measure velocity and level — and compute ﬂow — in real time. Mounted high above the water, away from debris and fouling, RQ-30 sensors empower public works staff to safely collect reliable data from sites previously thought to be too challenging or dangerous. These capabilities are a game-changer for water authorities around the world.
Smarter delivery of real-time data
The availability of real-time discharge data is crucial to flood control districts who partner with public safety agencies. The information is also key to water utilities, especially those in drought-stricken areas.
From deployment, the RQ-30 begins to provide real-time, accurate water level, velocity and discharge data. A water level sensor and a velocity sensor are housed in one enclosure. The device simultaneously and continuously measures said parameters. Native software then uses known cross-section information to compute discharge data in real time.
In addition, machine learning helps the sensor gather data at an unprecedented rate. It performs advanced pattern recognition, accelerating the building of a highly reliable internal rating curve. The computing can shave months of time off the conventional method of rating curve development at a specific monitoring site.
The conventional method for developing a “stage-discharge rating” commonly takes two years. Technicians periodically visit sites to record discharge measurements using mechanical current meters or acoustic doppler current profilers (ADCP) for a range of stages. Complex flow conditions may negate stable stage-discharge ratings and make conventional methods impractical or impossible. Adverse conditions include flow reversals, backwater and hysteresis effects.
Overcome discharge measuring challenges
At low flow rates, wind can create surface waves by accelerating or decelerating the water surface. The interference can impact the accuracy of velocity and discharge measurements — especially in wide, deep, slow moving rivers. Deep channels with a calm, glassy surface tend to easily pick up wind effects. Machine learning offers the ability to remove wind noise from raw data.
To compensate for wind and other environmental influences, the RQ-30 and its machine learning continuously build an internal relationship between water level and velocity. The more often a velocity or velocities are measured at a specific water level, the stronger the stage / discharge relationship for that specific level becomes.
The more often a relationship is “learned,” the stronger and more accurate the learning becomes. Thus, the RQ-30 can provide accurate flow data — even in challenging applications.
In addition to the noncontact radar sensor, Water Information System by KISTERS (WISKI) equips hydrologists with advanced discharge measurement and rating curve tools to check the accuracy of the RQ-developed rating curves. This combination of sensor and software enhances situational awareness of risk for emergency management.