the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Jun 2026
High accuracy river discharge estimation using UAS-based hydrometry and SWOT-derived WSE
Abstract. River discharge remains critically ungauged across much of the globe, limiting the accuracy of flood forecasting and constraining climate adaptation strategies. To address this, we propose a novel framework that integrates occasional Unoccupied Aerial Systems (UAS) with satellite Earth observations. Specifically, we construct a high-resolution hydraulic model of the Torne River in northern Scandinavia by combining a steady gradually varied flow (SGVF) solver with riverbed geometry extracted from UAS-based water-penetrating radar (WPR). The model is calibrated using four in-situ measured discharge–water surface elevation (WSE) snapshots from the Surface Water and Ocean Topography (SWOT) mission to estimate spatially variable, depth-dependent Manning’s roughness coefficients via automated optimization. This calibration enables river discharge to be estimated solely from satellite altimetry data (e.g., SWOT, Sentinel-3, and ICESat-2). Our approach demonstrates high accuracy and operational feasibility, achieving a mean absolute relative error of only 6.15 % when validated against in situ gauge measurements. Remarkably, the model successfully reconstructed an extreme 100-year flood event observed by ICESat-2, with an error of just 2.59 %. This framework provides a scalable and transferable approach for accurately estimating river discharge in virtually any reach observable by SWOT, in combination with one-off or occasional UAS hydrometry surveys.
Status: open (until 26 Sep 2026)
- RC1: 'Comment on egusphere-2026-3535', Anonymous Referee #1, 23 Aug 2026 reply
Viewed
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 42 | 0 | 1 | 43 | 0 | 0 |
- HTML: 42
- PDF: 0
- XML: 1
- Total: 43
- BibTeX: 0
- EndNote: 0
Viewed (geographical distribution)
Since the preprint corresponding to this journal article was posted outside of Copernicus Publications, the preprint-related metrics are limited to HTML views.
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General Comment
This paper uses SWOT water surface elevation data and observed discharge data to calibrate Manning’s coefficients in a 1D hydraulic model , with UAS bathymetric water penetrating radar data used to obtain river depths for the hydraulic modelling. The 1D hydraulic model is then used with a range of discharge values to estimate water surface elevations, from which a rating curve is generated. These rating curves are then used with additional SWOT water surface elevation data to estimate discharge. Overall, I found this to be a useful paper, employing data from several newly available technologies for river modelling. However, as outlined below, in my opinion the authors have oversold their claim that they have developed a method to estimate discharge in ungauged basins. I also had a few other comments that may help further strengthen the paper. Accordingly, I suggest the paper be returned to the authors for major revisions.
Specific Comments
L62 I would appreciate citations here for all (any?) previous studies that have used UAV-WPR for river bathymetry. In my opinion this is an interesting and possibly unique aspect of the present paper, and the novelty of this approach should be identified by full review of previous literature. Similarly, it would be informative and helpful to review in detail previous papers describing UAS-greenLIDAR for bathymetry.
L70 I appreciate that the intent of the methodology is to be applicable in ungauged basins. In Figure 1d it is not clear how discharge is selected during calibration, but L68 and L118 indicate this was the measured discharge at the time of the SWOT-observed water surface elevation. This begs the question: what would be the accuracy of eventual discharge estimation without the gauge station data? An answer to this question would be necessary to support the claim within the paper title that the methodology can produce discharge in ungauged basins. Without this, then I think the paper title, objectives and conclusions should be rephrased. Ideally, the authors would test their approach by modelling discharge (without use of the reference discharge for hydrological model calibration) and then test the accuracy of the eventual discharge estimation. However, the authors did not test if the model could be calibrated without the reference gauge station discharge data. Instead, the authors apply a Monte Carlo analysis (Section 4.5, L412) to test the outcome if random error is added to the measured discharge (with 30% error in discharge assumed for hydrological modelling). They argue (Line 430, Figure 11d) that the central tendency of the resulting confidence interval indicates that the method is applicable to ungauged basins. However, in my opinion, the central tendency is entirely the result of the uncertainty analysis employed, since only random error without bias was introduced. It is challenging to calibrate a hydrological model without bias in the absence of reference discharge data for calibration, and any bias in the hydrological model estimates of discharge used for calibration of their method would eventually propagate to their discharge estimation. I suggest that the large width of the confidence band shown in Figure 11d for a 30% relative error provides a better estimation of the outcome of a 30% bias in modelled discharge used for method calibration than the central tendency. This issue should be discussed thoroughly.
L123 Similarly, SWOT elevations are bias corrected with gauge water levels, and there is a statement that for ungauged stations overlapping SWOT elevations could be used for bias correction. I can see how overlapping SWOT elevations could be used to reduce inconsistencies, but I do not see how this is equivalent to absolute bias correction.
L173 I also would appreciate greater clarity as to which reaches were calibrated with discharge data measured at gauges versus obtained by hydrological model predictions.
Figure 1. It may just be my pdf, but the resolution of Figure 1a is poor, so the scale bar is not legible. This makes it difficult for me to judge the study site river width. Can the resolution be improved? Also, please state the mean river width in the text Section 2.
L123 I appreciate the citation to Liang et al. (2025) for the point cloud filtering. For the benefit of the reader, please briefly describe how this filter works. Also, please clarify if all the red points in Figure 2 are included in and thus “selected” from the raw data set, or rather they are derived values obtained during the filtering process.
Table 1. Please add a column with the actual gauge discharge during each SWOT measurement. This would clarify the range of discharges included within each discharge class. For example, what were the actual discharges for each day included in the Q=300 m3/s discharge class?
Table 1. How were the discharge classes selected?
Figure 4a. Which WPR return was deemed to be the river bed? Please explain how this image was assessed to determine bathymetric depths.
Figure 4d, L306. The measured section (Figure 4d) is for a different section than the constructed section (Figures 4a-c), The text says Figure 4 is provided to “compare” the measured and virtual sections, which I think is confusing. Instead, of “We compare” I suggest the sentence should be “Figure 4 presents both an in-situ surveyed section (XS9, chainage 99226.4 m) and a virtual XS (chainage 82043 m) using the same bathymetry.”
L310-314. Despite this explanation it is still not clear to me how the virtual sections were constructed. How exactly were the vertical and horizontal offsets selected to place the WPR-measured bathymetry from a different location fit within the DEM? In Figure 4d there are red lines that indicate modelled portions of the near-shore section at each bank, which indicates to there was subjectivity in the offset selection. Greater clarity/detail in the offset selection would be appreciated.
L197. The water surface slope measured over ~15km of SWOT data (0.0683 cm/km) implies a water elevation difference of ~ 1 cm. Are SWOT WSE data sufficiently internally accurate to measure this elevation difference?
L210-219. I understand this section to mean that Manning’s coefficient for a given location in a section is constant as discharge changes, but distinct Manning’s coefficients are selected for near-bank zones as they become inundated as water level rises. In other words, each subsection corresponds to an increment in wetted width as discharge increases. Is this correct? If not, then please clarify this section. A conceptual figure showing the subsections across a section would be helpful.
Figure 11a-c. I do not understand Figures 11a-c. These are supposed to show 95% confidence intervals, but I see only single lines. Are these lines the mean value of the estimate or the confidence interval? Is the line thickness supposed to represent the confidence interval? If the reference discharge was varied (by 5%, 15%, 30%) randomly during Monte Carlo, then it is not surprising that the mean values of the estimates do not change. What matters is the standard deviation. What were the standard deviations of the estimates from the Monte Carlo runs? Please clarify.
Technical/editorial
L161 typo “triangles” not “triagles”