Jeff Neal (University of Bristol, United Kingdom); Stephen Chuter (Fathom, United Kingdom); Paul Bates (University of Bristol, United Kingdom)

Estimating river discharge using a SWOT-based statistical approach

Authors: Izzy Probyn, Jeff Neal, Stephen Chuter, Paul Bates

Accurate estimates of river discharge are critical for a variety of applications including global hydrological and hydraulic models, flood forecasting, and water resources management. A key limitation for many of these models is the limited availability of observations, such as data records from gauging stations. The availability and reliability of discharge estimates, particularly in data-scarce regions, could be substantially improved by increased hydrological observations. The SWOT mission provides an opportunity to advance these models due to the unpreceded global data on water surface elevation (WSE), and the resolution and accuracy of the measurements taken. Releasing WSE measurements from all rivers exceeding 100m (and in many cases 50m) in width, with a repeat cycle of 21 days, it allows the incorporation of data into models as forcing or prior information in many locations where there has never been access to these measurements before, as well as providing regular, and consistent, data to add to what is already available.
The Discharge Algorithm Working Group has been using SWOT RiverSP data to calculate discharge for these rivers using a variety of different physical methods (Durand et al., 2023). All these algorithms involve solving hydrodynamical equations with SWOT observations as the driving data. However, the problem is ill-posed and unconstrained, and there is no exact solution to the equations that will fully satisfy the input observations. The algorithms are always going to need additional prior data to converge on a realistic discharge timeseries. Subsequently, the different algorithms have had varying degrees of success, with each sensitive to the input data in different ways. This is due to several reasons; the present unreliability of the width variable produced by SWOT, uncertainty in the WSE variable, but also in the conditions that must be met by the input data for these algorithms to be run.
Here we present a complimentary methodological approach to the problem using a statistical as opposed to empirical methodology. This statistical method produces a level-duration curve for each SWORD reach by linking quantiles of SWOT WSEs to global discharge distributions constructed from GRADES-HydroDL (Yang et al., 2023), essentially creating a level-discharge rating curve for each SWORD reach. By generating pairs of level-duration and flow-duration curves, one for each reach, a simple look-up from a SWOT reach observation to produce discharge should provide as much insight as a physical model, reducing compute time and resources without compromising on accuracy.
With promising results, we believe this method could be a useful addition to the Confluence workflow, with differing strengths from the current set of algorithms. Ultimately, this could be scaled to form an extensive and reliable series of virtual network of gauges to better inform models across hydrology, hydraulics, forecasting and many other areas.

References

Durand, M. et al. (2023) ‘A Framework for Estimating Global River Discharge From the Surface Water and Ocean Topography Satellite Mission’, Water Resources Research, 59(4), p. e2021WR031614. Available at: https://doi.org/10.1029/2021WR031614.
Yang, Y. et al. (2023) ‘Global Daily Discharge Estimation Based on Grid-Scale Long Short-Term Memory (LSTM) Model and River Routing’, ESS Open Archive. Available at: https://www.authorea.com/users/673109/articles/671985-global-daily-discharge-estimation-based-on-grid-scale-long-short-term-memory-lstm-model-and-river-routing.