A new sediment budget for the Congo River Basin reveals underestimated tributary contributions and large-scale deposition

Carlier, Nathan; Barthel, Matti; de Clippele, Antoine; de Groot, Lissie Willemijn; Drake, Travis William; Hemingway, Jordon Dennis; Hou, Yi; Wabakanghanzi, José Nlandu; Mandea, Joseph Zambo; Zhao, Pengzhi; Six, Johan; Van Oost, Kristof

doi:10.5194/egusphere-2026-247

Preprints

https://doi.org/10.5194/egusphere-2026-247

Preprints

03 Feb 2026

| 03 Feb 2026

A new sediment budget for the Congo River Basin reveals underestimated tributary contributions and large-scale deposition

Nathan Carlier, Matti Barthel, Antoine de Clippele, Lissie Willemijn de Groot, Travis William Drake, Jordon Dennis Hemingway, Yi Hou, José Nlandu Wabakanghanzi, Joseph Zambo Mandea, Pengzhi Zhao, Johan Six, and Kristof Van Oost

Abstract. Sediment yields from fluvial networks to the global oceans impact land carbon and nutrient cycles and are susceptible to climate, population, and vegetation changes. The Congo River Basin is a frontier for population and land use change, but sediment yield dynamics are poorly constrained within its basin, in particular within its largest tributary, the Kasaï River. To address this, we aimed to (1) introduce a spatially flexible methodology for estimating sediment yield from remote sensing monitoring, (2) establish a budget for the Congo and Kasaï river basins from both major and secondary tributaries, and (3) better constrain depositional losses of sediment during transit in the mainstem. A random forest model was calibrated on Landsat-8 spectral data, and total suspended sediment was accurately predicted (measured vs. predicted R² = 0.79), though predictions degraded in highly turbid waters due to spectral saturation. A sediment budget of the Congo River revealed that 33.0 Tg yr⁻¹ are exported to the coastal ocean. Most sediment is derived from Congo River headwaters, the Kasaï River, the Oubangui River, and the Aruwimi River, whereas lower-order tributaries contributed 10 % of all sediment inputs. Meanwhile, major contributors to the Kasaï budget (export = 11.1 Tg yr⁻¹) were the Kasaï headwaters, Sankuru, and Kwango-Kwilu rivers. Finally, we found the Cuvette Centrale, a peatland-dominated depression, to be a depositional hotspot, and we estimated its net sediment deposition to be between 5.96 Tg yr⁻¹ and 9.4 Tg yr⁻¹. By monitoring Congo River Basin sediment transport, we aim to provide a better understanding of the Earth-surface processes occurring in a globally significant and rapidly changing watershed that lacks crucial baseline information on its sediment and carbon cycles.

Received: 16 Jan 2026 – Discussion started: 03 Feb 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Status: final response (author comments only)

RC1: 'Comment on egusphere-2026-247', Anonymous Referee #1, 17 Apr 2026

Title: A new sediment budget for the Congo River Basin reveals underestimated tributary contributions and large-scale deposition

overall evaluation

The paper presents a compelling and timely exploration of a crucial Earth-surface system. Its primary strengths lie in its basin-scale scope, the integration of remote sensing with a “sediment-budget” framework, and the provision of new quantitative insights into tributary contributions and net deposition in the Cuvette Centrale. It is relevant to the Journal scope and has the potential to make a significant contribution.

Overall, my assessment is positive, but I believe the manuscript requires more refinement before it is ready for publication. The central conclusions are plausible and generally well-supported, but there are methodological and interpretive issues that need to be elaborated on and presented more clearly. Specifically, the authors should show how uncertainty is propagated, the effects of attrition in calibration data, the methods used for spatial and temporal validation and support the inferred deposition with some evidence. I am also apprehensive about the use of the term “sediment budget,” given that the study quantifies washload-dominated suspended fluxes.

Review against the detailed ESurf criteria

S/N	CRITERIA
1	Does the paper address relevant scientific questions within the scope of ESurf?
	Yes, the paper falls within the journal's scope. It examines the sediment budget of the world's second largest river basin and directly contributes to ongoing research in geomorphology, sediment connectivity, and source-to-sink dynamics.
2	Does the paper present novel concepts, ideas, tools, or data?
	Yes, it does: It develops a region-specific remote-sensing/machine-learning workflow for TSS in the Congo Basin, outputs an updated “sediment budget” for the entire Congo Basin, including secondary tributaries, and presents the first “sediment budget” for the Kasai Basin. However, the methodological novelty is more applied than conceptual, and the authors should avoid overstating the approach's broad transferability without clearer validation across diverse hydro-sedimentary conditions.
3	Are substantial conclusions reached?
	1. The Congo exports approximately 33 Tg/year. This is likely the minimum exported given the study’s focus on washload-dominated suspended fluxes. 2. The Kasaï River contributes more to these exports than previously estimated. 3. Secondary tributaries play a notable role and should not be overlooked. 4. The Cuvette Centrale is likely a major zone for net deposition. Again, this is likely true, but it's important to approach it with caution, as the paper addresses net depositional losses rather than gross deposition, and the supporting evidence is based on inference rather than direct observation.
	Are the scientific methods and assumptions valid and clearly outlined?
	Partly. The general framework is valid, but some assumptions need more justification The cloud cover problem in the Congo is well known. However, the 11-day matching window between field observations and satellite images may be too long, even for larger, less dynamic rivers, leading to temporal mismatch. Additionally, because sediment transport is event-dominated, with a small number of high-flow events carrying the bulk of the annual sediment load, peak sediment loads are missed. The treatment of spectral saturation is somewhat ad hoc. The authors remove values above 100 mg/l a priori, yet later suggest that saturation may begin around 70 mg/l in the Congo and possibly around 45 mg/l in some tributaries. This materially affects interpretation in high-TSS rivers and needs a more formal analysis. The independence of the validation in space and time is unclear. If training and test data overlap across stations or dates, the performance results may be overly optimistic. Conducting a grouped validation by station, tributary, or campaign would enhance confidence in the results. The process of converting predicted TSS and discharge into sediment yield is simple, but it’s important to explicitly describe the uncertainty propagation throughout the entire workflow.
5	Are the results sufficient to support the interpretations and conclusions?
	Mostly, but only after more careful qualification The sediment yield estimates are derived from satellite-based TSS predictions combined with discharge and temporally integrated; however, the methodology is inherently limited in its ability to capture event-scale sediment dynamics. Given the reliance on Landsat imagery, persistent cloud cover, and the use of a ±11-day matching window, high-flow, high-sediment events, likely responsible for a substantial proportion of annual sediment transport, are systematically undersampled or excluded. As a result, the derived sediment yields may be biased toward lower values, reflecting temporally averaged conditions rather than true flux magnitudes. This limitation should be more explicitly acknowledged, and its implications for both total yield and inferred deposition should be discussed Further, the precision of the quantified depositional estimates and the confidence with which along-river decreases can be attributed specifically to deposition rather than to compounded model error, changes in bias along the mainstem, floodplain exchange, or unresolved tributary/bank inputs is unclear. The paper acknowledges some of this, but the discussion should be strengthened.
6	Is the description of experiments and calculations sufficiently complete and precise to allow reproduction?
	Not yet fully, conceptually, yes, but they could include: fuller details on train/test splitting and cross-validation, hyperparameter settings for the random forest model, the exact variable set retained after RFE in a machine-readable table, a more explicit uncertainty propagation workflow, criteria for defining virtual stations and extracting cross-sectional reflectance, clearer documentation of discharge inputs and how missing data were handled.
7	Do the authors give proper credit to related work and clearly indicate their own contribution?
	Yes
8	Does the title clearly reflect the contents of the paper?
	The title broadly reflects the scope and findings of the manuscript; however, the use of the term ‘sediment budget’ may overstate the completeness of the analysis. Given the methodological limitations, more qualified wording would improve alignment between the title and the underlying analysis, e.g., “Satellite-derived suspended sediment fluxes in the Congo River Basin indicate underestimated tributary contributions and net downstream sediment losses.”
9	Does the abstract provide a concise and complete summary?
	Mostly yes. A minor concern is that it reads slightly more certain than the body of the paper justifies, especially regarding model accuracy and depositional interpretation
10	Is the overall presentation well-structured and clear?
	Yes.
11	Is the language fluent and precise?
	Yes
12	Are mathematical formulae, symbols, abbreviations, and units correctly defined and used?
	Mostly yes.
13	Should any parts of the paper be clarified, reduced, combined, or eliminated?
	a more comprehensive presentation of the model calibration workflow, a more explicit limitations subsection focused on validation, saturation, and uncertainty, The conclusion should frame results as indicative rather than conclusive. In particular, statements regarding tributary contributions and large-scale deposition appear more definitive than the underlying data support.
14	Are the number and quality of references appropriate?
	Yes. The references appear appropriate and sufficient.
15	Are the amount and quality of supplementary material appropriate?
	They are ok.

Citation: https://doi.org/10.5194/egusphere-2026-247-RC1

RC2: 'Comment on egusphere-2026-247', Joshua Himmelstein, 04 Sep 2026

The authors use a random forest model trained on 148 in-situ samples (and validated on 36 additional samples) to refine estimates of total suspended sediment in the surface waters of the Congo River Basin, which has historically lacked local calibration in similar analyses. The basin-wide geographic distribution of these samples is a strength, especially compared to studies trained on only one or a few monitoring stations. The authors then combine predicted TSS with modeled discharge to find evidence of sediment surpluses and deficits in different reaches of the river.
The presentation of the findings are clear and concise, though at the cost of (1) reproducibility, as the methods lack detail in some crucial parts, like how confidence intervals are calculated and error propogated, and (2) explaining the broader relevance of the results to earth surface processes. Several of the main findings depend on choices about the spatial buffer, temporal matching window, interpolation of TSS, and the modeled discharge dataset, and more explanation of these choices and their associated uncertainty is needed. There is also an important limitation from the underrepresentation of high-TSS conditions due to spectral saturation. This deserves more emphasis because these may be the conditions when sediment transport is greatest and therefore most important to the resulting sediment budgets.
In a revised version, I suggest the authors include more specifics in the methods and spend more time explaining how their findings can benefit our understanding of sediment transport, sediment budgets, depositional processes, and carbon delivery to the coast. Some of the proposed mechanisms behind the observed patterns are interesting, but should be more clearly separated into what is shown by the model versus what remains a hypothesis. It would also be helpful to end with more specific recommendations for what future field data would be needed to better constrain sediment and carbon budgets in the basin. Finally, I suggest that consistent styling is used throughout the figures and attention is paid to their readability.
Overall, I think the paper is a welcome contribution in a region that lacks high spatiotemporal resolution estimates of riverine sediment yield.

Below are comments provided in the attached review PDF.

Main-text comments C01 | Lines 13-15 | PDF p. 1

Referenced text: By monitoring Congo River Basin sediment transport, we aim to provide a better understanding of the Earth-surface processes occurring in a globally significant and rapidly changing watershed that lacks crucial baseline information on its sediment and carbon cycles.

Reviewer comment: Instead of aim to provide better understanding, perhaps state what you found and why it matters beyond just lacking local data.
C02 | Lines 36-38 | PDF p. 2

Referenced text: this will drive a rapid land-use transition to agriculture, which in turn will increase soil erosion and is likely to substantially impact sediment yields.

Reviewer comment: Must cite this or soften the statement.
C03 | Line 54 | PDF p. 2

Referenced text: no large depositional hotspot in the middle of its course

Reviewer comment: if virtually unstudied, how can you know there is no depositional hotspot in its course? Consistent river width, I imagine? Bring the reader along with your reasoning or cite the few geomorphic studies that may exist.
C04 | Line 64 | PDF p. 3

Referenced text: total suspended sediment (TSS)

Reviewer comment: Include units here as this is first use. And perhaps a more formal definition of what TSS is can come in the methods.
C05 | Line 74 | PDF p. 3

Referenced text: To address the lack of unified, cohesive, long-term records of sediment transport in the CRB, we first aimed to leverage state-of-the-art techniques

Reviewer comment: This is a nice paragraph to end the intro and set up the methods with. I understand trying to be general as to not overwhelm the reader with technical terms before methods, but "leverage state of the art techniques" feels overly broad for the audience here, who likely is quite familiar with many of those techniques. In general, I think more specificity would help the reader know how you produced the results in what is a rather methods dependent predictive model.
C06 | Line 89 | PDF p. 4

Referenced text: 1511 mm yr-1 (Fick and Hijmans, 2017)

Reviewer comment: report +/- or error if provided, throughout
C07 | Line 93 | PDF p. 4

Referenced text: 0.9 10^6 km^2

Reviewer comment: formatting
C08 | Line 100 | PDF p. 4

Referenced text: 2.2.1 In situ data

Reviewer comment: What range of concentrations are covered by the in-situ data? When you later present the predicted results (2.8-66.4 mg/L), it would be nice to see if estimates are within the bounds of training data.
C09 | Line 101 | PDF p. 4

Referenced text: TSS monitoring data was compiled from multiple sources

Reviewer comment: Are these sources consistent in their collection and calculation of TSS? All surface collected water bottle samples? Or aliquots of samples? Non-combusted (including organic and inorganic, with implications for carbon quantities). Specify as possible.
C10 | Lines 112-113 | PDF p. 4

Referenced text: -0.1 (i.e., NDWI > -0.1 is considered water) (Fig. 2a).

Reviewer comment: justification for this NDWI cutoff? A citation or at least a phrase as to why.
C11 | Line 114 | PDF p. 4

Referenced text: with a buffer (500 m or 100 m for large and small rivers respectively) around each point.

Reviewer comment: Why 100m and not 90m (if Landsat is at a nominal 30m resolution?) And why that much larger for large rivers?
C12 | Lines 117-118 | PDF p. 4

Referenced text: Second, a maximum time window of 11 days (Fig. A1) for matches between in situ observations and their closest L8 image was allowed.

Reviewer comment: In looking at A1, it does seem there is sufficient temporal autocorrelation to justify up to 11 day time window. But why not even more? Day 15 r-squared is ~0.7 and recovers more samples, which in a sparsely ground-truthed area might be worth it. Was a sensitivity analysis performed to motivate this decision?
C13 | Line 145 | PDF p. 5

Referenced text: relative error (%) (Dethier et al., 2020), and % bias

Reviewer comment: Not sure what the citation is referring to. Just relative error? As it is placed within a list of evaluations of model performance and not at the end.
C14 | Line 166 | PDF p. 6

Referenced text: assuming TSS concentrations exhibit low weekly to monthly variability

Reviewer comment: And are well-mixed vertically. Any reason to believe there is not vertical stratification or gradients in TSS in the flow here?
C15 | Line 167 | PDF p. 6

Referenced text: (8.64 x 10^-8)

Reviewer comment: What type of unit conversion factor is this? And is it unitless? perhaps place units of SY daily somewhere here.
C16 | Line 168 | PDF p. 6

Referenced text: Q (m^3 s^-1) is daily discharge obtained from Wongchuig et al. (2023)

Reviewer comment: Lots of the remaining analysis here depends on this discharge model. Due to the importance, another sentence or two on how that was calculated in the body of this articles text will be helpful.
C17 | Line 168 | PDF p. 6

Referenced text: SYdaily is daily sediment yield (Tg day^-1)

Reviewer comment: Confidence Intervals are reported in the results but their contributed sources of error should at least be mentioned if not formulated here. Unclear how the uncertainty from the Wongchuig discharge dataset (which has large margins of error) is forward propogated if at all. I commend the effort to get the data into mass terms, which helps their general applicability.
C18 | Lines 198-199 | PDF p. 7

Referenced text: five variables were contextual (discharge, day number of the year, longitude, Euclidean distance to Congo outlet, and Strahler order 8).

Reviewer comment: If you list all five contextual variables here (which I think is helpful), I suggest you do the same for the visible bands.
C19 | Lines 218-219 | PDF p. 8

Referenced text: Main contributors to this budget were from the mainstem (Congo headwaters) with 51% of inputs, as well as the Kasai (34%), Oubangui (13%), and Aruwimi rivers (11%) (Fig. 5a).

Reviewer comment: These percentages sum to more than 100.?
C20 | Line 228 | PDF p. 8

Referenced text: displayed a significant monotonic positive trend

Reviewer comment: Will this be discussed later? It seems interesting. And what is the trend, if significant?
C21 | Line 234 | PDF p. 8

Referenced text: River showed that an extra 8% (0.916 Tg yr-1, CI = -4.2, 6.1) of sediment was exported by the system compared to its inputs.

Reviewer comment: Dredging, mining, increasing erosion? Might you expand on this in the discussion? These diverging trends are exciting.
C22 | Lines 260-261 | PDF p. 9

Referenced text: which can introduce bias due to lack of spatial representativity.

Reviewer comment: This is an important point, glad to see it included here.
C23 | Lines 265-266 | PDF p. 9

Referenced text: heteroskedasticity at TSS >= 70 mg L-1, which we interpreted as spectral saturation.

Reviewer comment: Any other studies which find spectral saturation at such low TSS? Any suspects like dissolved organic matter which you could fold into the discussion ? Seems like we would want to capture higher TSS (when sediment yields are high and flashy)
C24 | Line 267 | PDF p. 10

Referenced text: abnormally low TSS prediction standard deviations were observed in the Lwange River and Kasai headwaters

Reviewer comment: Could save observed for in-situ data, as it is easy to get confused in this sentence with what is predicted versus ground truthed.
C25 | Lines 306-307 | PDF p. 11

Referenced text: reveals that inputs exceed outputs, implying a substantial net deposition along the mainstem.

Reviewer comment: Can this continue? Is accommodation space being filled? Feel free to propose some reasons in the discussion so long as it is clear they are not proven by your predictions.
C26 | Lines 345-347 | PDF p. 12

Referenced text: due to the relationship between OC and surface reflectance being less direct. However, a successfully calibrated model for POC and DOC would lead to a better understanding and OC budgets and transport pathways in the CRB.

Reviewer comment: I am glad you got to carbon, as was alluded to or motivated by the intro. I think more can be said here. What data needs to be collected? I think it could be nice to end this discussion section with a couple so-what sentences; why does this all matter? And also recommendations for future studies / data collection efforts (for example if carbon budgets are a goal of future work how do we get there?).
C27 | Line 369 | PDF p. 13

Referenced text: Code availability. All code used for generating the data in this paper will be made available before publication.

Reviewer comment: Plans for when/where code will be provided should be described.
Figure comments C28 | Figure 1 | PDF p. 20 | map
Reviewer comment: North Arrow. Also include coordinates around the edges (graticules in ArcGIS if that is whats used here)
C29 | Figure 1 | PDF p. 20 | A-E site markers
Reviewer comment: The most important part of this figure to me is seeing where the A-E sites are, so perhaps increase them in size and/or annotate them directly on the map.
C30 | Figure 1 | PDF p. 20 | map sampling locations
Reviewer comment: I also would like to see where the in-situ sites are located (especially if <200). Any way to place those as small dots here?
C31 | Figure 2 | PDF p. 21 | flowchart and plot axes
Reviewer comment: Increase font size of numbers on axes, and perhaps bold the words in the flowchart. Just feels too light to me.
C32 | Figure 3 | PDF p. 23 | ordering of basin outlets
Reviewer comment: Mention in caption, but is left-to-right sorted by outlet size?
C33 | Figure 4 | PDF p. 24 | figure fonts and colorbar
Reviewer comment: Suggest using similar (non-serif) fonts throughout your figures. And numbers on the colorbar are tiny.
C34 | Figure 6 | PDF p. 26 | caption / panel scales

Referenced text: for each year), for each basin. Dotted black lines are mean sediment yield, and CV is mean yearly coefficient of variation for each basin.

Reviewer comment: Add a comment highlighting that each is scaled differently. Also suggest all should start at 0 and go to their max.
C35 | Figure 8 | PDF p. 28 | axes and river-outlet labels
Reviewer comment: Increase font size of axes and river outlets. Nice plot!

Citation: https://doi.org/10.5194/egusphere-2026-247-RC2

Viewed

Total article views: 2,761 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
1,586	1,062	113	2,761	106	204

HTML: 1,586
PDF: 1,062
XML: 113
Total: 2,761
BibTeX: 106
EndNote: 204

Views and downloads (calculated since 03 Feb 2026)

Month	HTML	PDF	XML	Total
Feb 2026	801	234	68	1,103
Mar 2026	414	560	16	990
Apr 2026	145	118	3	266
May 2026	77	83	0	160
Jun 2026	23	7	1	31
Jul 2026	42	22	4	68
Aug 2026	21	19	6	46
Sep 2026	63	19	15	97

Cumulative views and downloads (calculated since 03 Feb 2026)

Month	HTML	PDF	XML	Total
Feb 2026	801	234	68	1,103
Mar 2026	414	560	16	990
Apr 2026	145	118	3	266
May 2026	77	83	0	160
Jun 2026	23	7	1	31
Jul 2026	42	22	4	68
Aug 2026	21	19	6	46
Sep 2026	63	19	15	97

Viewed (geographical distribution)

Total article views: 2,693 (including HTML, PDF, and XML) Thereof 2,693 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 20 Sep 2026

Short summary

The Congo River Basin, a region of global significance, is undergoing a strong increase in population, which will in turn increase soil erosion, impacting soil fertility and the carbon cycle. We examined river sediment (eroded soil) transport, whereby we provided information on the main contributors of sediment to the Congo River. We also highlighted the role of a swampy region which removes large amounts of sediment from the water, acting as a regulator sediment transport to the Congo River.


Total:	0
HTML:	0
PDF:	0
XML:	0