| 01 Oct 2026
A priori diagnostic for acoustic backscatter inversion to suspended sediment concentration in hyperturbid estuaries
Abstract. Accurate quantification of sediment transport in fluvial, estuarine and coastal environments depends on reliable estimates of suspended sediment concentrations (SSCs). While Acoustic Doppler Current Profilers (ADCPs) provide depth-resolved backscatter observations that have potential to become translated to SSC, conventional inversion techniques fundamentally depend on the assumption of homogeneous sediment concentration and acoustic scattering/attenuation properties along the acoustic beam. In stratified and hyperturbid conditions, this homogeneity may be compromised, even when the particle size distribution itself remains vertically stable. Strong sediment gradients trigger stratification that either amplifies signal strength at the lutocline or diminishes it by fine-sediment-induced viscous attenuation, possibly invalidating this assumption. To assess whether the assumption of homogeneous sediment properties remains valid under hyperturbid conditions, we propose an a priori diagnostic based on the shape of the acoustic backscatter profile. Using changepoint detection, we identify slope breaks in acoustic profiles, which indicate transitions in the effective acoustic regime. These transitions likely indicate a breakdown of the homogeneity assumptions underlying acoustic backscatter inversion techniques. We apply the diagnostic to ADCP measurements (approximately 2 m above the bed) collected in an engineered navigation channel of the Ems Estuary, spanning SSCs of 0.1–100 g L-1, this upper limit corresponding to hyperturbid conditions. Validation against in-situ water samples confirms that profiles without detected changepoints (Type 1) yield reliable SSC estimates, even at concentrations up to ~20 g L-1. In contrast, Type 2 profiles exhibit pronounced spikes or slope breaks in the backscatter profiles that may reflect range-dependent acoustic sediment properties. This variation decouples the observed backscatter from a simple, range-invariant relationship with concentration. As a result, the inversion algorithm fails to accurately invert backscatter to concentration, leading to erroneous and unphysical SSC estimates that deviate significantly from in-situ water samples. The proposed diagnostic provides a simple step to flag unreliable profiles and support targeted sampling and interpretation in challenging hyperturbid environments, enabling more reliable estimation of suspended sediment concentrations and associated sediment transport fluxes.
This is a really interesting and useful paper, and I particularly like the elegance of the approach. Taking a rather complex problem in acoustic inversion and developing a relatively straightforward diagnostic for identifying when the underlying assumptions begin to break down is a valuable contribution to the field with broad application. The field observations are compelling, and the distinction between high sediment concentration itself and the effects of structural heterogeneity is particularly important.
I have a few suggestions that might help strengthen the discussion and draw out the wider significance also. In particular, I wonder whether there is more to say about the diagnostic as a means of identifying lutocline development, rather than simply inversion failure; such as whether the relationship between stratification and the underlying acoustic mechanisms could be distinguished a little more clearly; and whether the approach could eventually provide a graded measure of confidence in derived SSC??
Perhaps most importantly, I think there is an interesting wider implication for sediment flux estimation. If the profiles where inversion becomes unreliable are also those carrying the highest sediment loads, how should we deal with the potential bias introduced by excluding them? That seems a particularly worthwhile point to explore in the discussion.
Here are some considerations and possible minor corrections that should be considered:
1. For me, one of the strongest findings is that high SSC itself isn't necessarily what defeats acoustic inversion. It's actually the changing acoustic regime along the beam and what that can yeild in understanding. - turn it into a positibve as well as a filter. The fact that Type 1 profiles retain excellent agreement with physical samples up to around 20 g/L is important. This could arguably be elevated further in the abstract and discussion. So I suggest that the authors can make this conceptual distinction more prominent? It challenges the tendency to regard highly turbid environments as inherently unsuitable for acoustic inversion.
2. The interpretation of sharp backscatter transitions as potentially indicating a lutocline is particularly interesting ( pp. 20–21). So, there might be a second contribution here in that rather than treating changepoints exclusively as a warning that inversion will fail, could their occurrence and position provide useful information about the development and movement of lutoclines? So, even without developing this further in the present paper, it would be worth discussing whether changepoint detection might ultimately offer a tool for monitoring fluid-mud dynamics, particularly through tidal cycles.
3. I find the relationship with the sediment-modified Richardson number (Figure 10) persuasive, but there is a subtle distinction between demonstrating association and establishing causality of course. Could the authors acknowledge that flocculation, multiple scattering and particle–fluid interactions may also affect the effective acoustic properties. There is thus a need to perhaps distinguish more clearly between stratification as the hydrodynamic condition associated with changepoints and the particular acoustic mechanisms responsible for them. This would strengthen rather than weaken the physical interpretation in my view.
4. Could the diagnostic be expressed as a practical measure of confidence and can this be drwan further. The Type 1/Type 2 classification is elegant and straightforward, and Figure 7 shows a striking difference in inversion performance (Pearson correlations of 0.98 and 0.28). wonder whether there is scope to move beyond a binary classification here however in that could changepoint magnitude, slope contrast or deviation from linearity provide a continuous indication of inversion confidence and or flocculation and or something else? Even a short discussion of this possibility would add to the methodological significance. As an aside I also think it might also be useful to report the absolute inversion errors for each type, alongside the correlations, to make the implications for sediment flux estimation more tangible.
5. The practical implications are already substantial in terms of methodological advance in giving QA to these sorts of datasets. Given that ADCP-derived velocity and SSC are commonly combined to estimate sediment fluxes, I wonder if they could draw out how this can result in systematic underestimation of SSC under particular conditions could lead to seriously biased transport estimates. I think this is especially relevant in estuaries where short periods of very high concentration can contribute disproportionately to sediment transport. I therefore think that the the authors could say a little more about what the diagnostic means for integrated sediment flux calculations? Flagging unreliable profiles improves data quality, but excluding those profiles could itself introduce a sampling bias if they correspond to the most sediment-laden conditions. How should practitioners manage that would be useful to add to the discussion. They currently say remove the data when things get interesting! :-)
Thanks and nice work, Dan Parsons