Fully-Constrained Remote Estimation of River Discharge

Abstract:

River discharge (flow volume) is unavailable for a significant fraction of worldwide rivers: measurement currently requires permanently installed gauges and a “rating curve” calculated from detailed historical surveys. Recent work has explored whether discharge can be estimated from overhead imagery such as LANDSAT, using the pair of relationships known as hydraulic geometry that are also the basis for rating curves. High accuracy has not yet been achieved, because of the three observable variables (width, depth, velocity) only width is directly available from imagery. This results in an under-constrained problem that requires additional assumptions and suffers from high error. We here utilize high-resolution commercial imagery (e.g. WorldView 3) which allows measuring the water surface elevation through stereo processing. We also utilize an algorithm known as RiverEye that calculates surface velocities by solving an optical flow problem on passive tracers such as sediment load. Our technique thus provides measurements of all three hydraulic variables from the same observation. By observing under different flow conditions, we will be able to solve for the hydraulic geometry coefficients and enable fully remote measurement of river discharge.