Evaluating the applicability of SWOT discharge inversion algorithms in multichannel rivers using satellite imagery-derived timeseries of water surface height, slope, and river width
Abstract
The Surface Water and Ocean Topography (SWOT) mission will produce discharge estimates at world rivers wider than 100m and possibly as narrow as 50m using simple flow laws such as the Manning-Strickler equation. While SWOT will measure the surface water elevation, slope, and river width, the satellite cannot directly quantify the channel friction and the total cross-sectional area of flow, which must be estimated by discharge inversion algorithms or through data assimilation. When the current discharge inversion methods such as the Metropolis Manning and the Bayesian At-many-stations-hydraulic-geometry Manning algorithms are applied to a multichannel river, they assume that the multiple channels can be aggregated into a single equivalent channel, over which one-dimensional flow laws can be applied without incurring excessive errors. Here, we evaluate the applicability of SWOT discharge algorithms to a multichannel section of the Tanana river near Fairbanks, Alaska. We conduct the evaluation using 18 sets of measurements of water surface elevation, slope, and equivalent river width covering 3 contiguous reaches each spamming 5km in length. We created the dataset with high resolution surface elevation models and imagery acquired by the WorldView-2 satellite. Prior information on discharge needed by the algorithms are extracted from the Global Reach Level A Priori Discharge Estimates for SWOT. Resulting discharge is evaluated against the nearest United States Geological Survey streamgage located near Fairbanks. Additionally, we explore the dependency of the discharge error metrics on the quality of the prior estimate of mean annual flow by systematically varying the prior biases from an underestimation of 50% to an overestimation of 100% in increments of 25%. Preliminary results show that standard errors remained fairly constant over the seven runs (0.31 to 0.35), however, normalized posterior biases varied from -69% to 8% and were highly correlated to prior biases (r=0.99) with an encouraging regression slope between prior and posterior biases equal to 0.48. Our results allow us to partition discharge errors into structural errors, propagated measurement errors and parameter errors, providing new insight on the performance of SWOT discharge inversion algorithms when applied to multichannel rivers.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMH040.0002F
- Keywords:
-
- 1855 Remote sensing;
- HYDROLOGY;
- 1857 Reservoirs (surface);
- HYDROLOGY;
- 1860 Streamflow;
- HYDROLOGY;
- 4512 Currents;
- OCEANOGRAPHY: PHYSICAL