Long-term hydro-sediment dynamics of the Ucayali River (Amazon Basin) revealed through combined observations, remote sensing, and SWAT-Amazon modelling

Santini, William; Delort-Ylla, Alexandre; Martinez, Jean Michel; Lavado-Casimiro, Waldo; Camenen, Benoît; Le Coz, Jérôme; Roussillon, Joana; Pérez Arévalo, Jhonathan Junior; Molina-Carpio, Jorge

doi:10.5194/egusphere-2025-4101

Preprints

https://doi.org/10.5194/egusphere-2025-4101

Preprints

24 Oct 2025

| 24 Oct 2025

Status: this preprint is open for discussion and under review for Hydrology and Earth System Sciences (HESS).

Long-term hydro-sediment dynamics of the Ucayali River (Amazon Basin) revealed through combined observations, remote sensing, and SWAT-Amazon modelling

William Santini, Alexandre Delort-Ylla, Jean Michel Martinez, Waldo Lavado-Casimiro, Benoît Camenen, Jérôme Le Coz, Joana Roussillon, Jhonathan Junior Pérez Arévalo, and Jorge Molina-Carpio

Abstract. Since the early 1970s, the Amazon basin has experienced growing local and global changes, potentially reaching a climatic tipping point in the coming decades. However, due to cost constraints and limited access, conventional hydrological networks in the basin struggle to provide the spatial resolution and temporal extent required for accurate quantification of water and sediment budgets, which are essential for understanding biogeochemical cycles.

Focusing on the Ucayali River, a major Amazonian foreland tributary, this study provides the first long-term hydro-sediment balances in this region at sub-basin scale, distinguishing fine sediments from sand loads (37 years for water and sands, 20 years for fine sediments). It is achieved by the integration of remote sensing and hydrological-hydraulic modelling using a modified SWAT model, SWAT-Amazon. A new hydraulic module for water routing was implemented in SWAT-Amazon to suit the Amazon diffusive flood wave, representing floodplains as reservoirs. Fine sediment loads were estimated using satellite-derived concentrations and simulated discharges, while suspended sand loads were simulated within SWAT-Amazon.

Results indicate that the Andean Ucayali River exports 455 10⁶ t yr⁻¹ of suspended sediment (40 % sand). As the floodplain traps 36 % of the Andean sediments (65 % sand), mostly by tectonic subsidence, the Ucayali delivers 290 10⁶ t yr⁻¹ of total suspended sediment to the Amazon River, 26 % as sand. Floodplain recycling plays a crucial role as a secondary sediment source (22 % of the Ucayali load), with a water storage that peaks at 19.1 km³ in March (38 % of discharge). A previously undocumented sand sedimentation process is identified during the flooding period, capturing 14 % of the sand flux at peak discharge and thus decorrelating sediment transport from water discharge. No significant long-term trends in flood duration, discharge, or sediment fluxes were detected, suggesting contrasted evolution patterns of the precipitations in the basin due to its particular position in the Amazon Basin. This study emphasizes the need to rethink hydrological network management with robust and long-term conventional data at ‘super’ stations to support the calibration of remote sensing and modelling at ‘virtual’ stations. Extending this approach to other Amazonian basins could significantly enhance hydro-sediment and biogeochemical cycle research in large river systems. Additionally, it highlights the importance of regionally focused over large-scale assessments, which often carry high uncertainties and may mislead mitigation strategies.

Received: 22 Aug 2025 – Discussion started: 24 Oct 2025

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

Preprint (PDF, 5999 KB)

Supplement (211 KB)

Download & links

William Santini, Alexandre Delort-Ylla, Jean Michel Martinez, Waldo Lavado-Casimiro, Benoît Camenen, Jérôme Le Coz, Joana Roussillon, Jhonathan Junior Pérez Arévalo, and Jorge Molina-Carpio

Status: open (extended)

Post a comment Subscribe to comment alert

RC1:
'Comments on egusphere-2025-4101', Anonymous Referee #1, 18 Dec 2025 reply
This manuscript presents an ambitious and innovative integration of long-term in situ observations, satellite remote sensing, and a modified SWAT framework to quantify multi-decadal hydro-sediment dynamics in the Ucayali River. The development of SWAT-Amazon and the explicit treatment of floodplain hydraulics and sand routing represent a substantial methodological advance for large, low-gradient tropical rivers, and the assembled observational dataset, particularly the long-term “super station” records and dedicated field campaigns, is a clear strength of the study. The results are compelling and potentially impactful for understanding sediment trapping, recycling, and floodplain controls in the Amazon Basin.
My comments below are intended to strengthen the robustness, generalisability, and interpretability of the findings. In particular, they focus on clarifying the separation between calibration and validation, quantifying uncertainty and equifinality in key inferred budgets, and providing additional evidence that internal floodplain processes are realistically represented rather than inferred solely from outlet behavior. I examined the Supplementary Material, while it provides valuable methodological detail, it does not address the validation, uncertainty, or equifinality issues raised below. Addressing these points would substantially increase confidence in the reported trapping and recycling fractions and enhance the value of the framework for application to other Amazonian basins.
The calibration period for discharge and sand overlaps almost entirely with the reported validation; I am wondering whether the authors could provide performance metrics for an independent hold-out period to demonstrate that the high NSE values are not artifacts of overfitting.

The daily timestep is justified qualitatively by slow, subcritical flow conditions, but I cannot find any formal numerical analysis; could the authors present Courant number ranges for representative reaches or demonstrate via timestep halving that peak attenuation and timing errors remain acceptably small?

The authors state that rectangular main channels and simple-reservoir floodplains were chosen after testing other configurations, but results from these tests are not shown; could the authors include a supplementary comparison of discharge, water level, and sand flux performance under alternative geometries such as trapezoidal channels or compound sections?

Validation focuses on outlet hydrographs, rating curves, and sediment fluxes, yet floodplain processes drive the key findings; I am missing independent validation of internal floodplain behavior using, for example, SAR-derived inundation extent time series, floodplain lake stage records, or tracer-based residence time estimates.

The Sobol analysis identifies hf, Cnfp, kfp, and db as influential parameters, but does not explore whether multiple distinct parameter combinations yield equivalent outlet performance; could the authors present dotty plots or behavioral parameter ranges to assess equifinality and discuss how it affects the certainty of inferred trapping and recycling fractions?

The reflectance–concentration relationship is calibrated on a limited number of samples and applied basin-wide for two decades; given the acknowledged potential for hysteresis and grain-size variability at the surface, I am wondering whether the authors could quantify the bias or uncertainty bounds under contrasting conditions such as rising versus falling limb or low-water resuspension events.

The bimodal fine/sand assumption is conceptually justified, but I cannot find any sensitivity test of budget estimates to the threshold grain size separating the two classes; could the authors show how trapping and recycling percentages change if the cutoff is shifted?

The paper advocates extending the approach to other Amazonian basins but does not specify minimum data requirements; could the authors indicate the minimum gauging frequency at a super station, the required number of sediment samples for calibrating the reflectance model, and whether performance degrades gracefully as data density decreases?

Channel geometry parameters are held constant over the full simulation period in a dynamically meandering system; could the authors test stationarity of the calibrated relationships using moving-window recalibration or change-point analysis to confirm that morphological evolution has not introduced time-varying bias?

The MGB model is critiqued for underestimating Ucayali sediment loads, but no controlled comparison using equivalent forcing and calibration effort is provided; I would encourage the authors to clarify whether this difference reflects model structure or calibration intensity, or alternatively defer detailed benchmarking to future work while softening the current language.

Key budget figures such as floodplain trapping, recycling contribution, and sand capture at peak discharge are presented without confidence intervals; I strongly encourage the authors to propagate observational uncertainty and parameter uncertainty through to these final percentages using Monte Carlo or ensemble methods.

Several figures use cyan for SWAT-Amazon simulations and marine blue for observations, which are difficult to distinguish; some figures also use slightly different shades of cyan for multiple plots. I suggest that the authors consider using higher-contrast colour pairs or different line styles to improve readability.

Reply
Citation: https://doi.org/10.5194/egusphere-2025-4101-RC1
CC1:
'Comment on egusphere-2025-4101', Dhruv Sehgal, 18 Dec 2025 reply
Thank you for the opportunity to review this manuscript. The paper presents an integrative framework to quantify fine and sand-sized sediment fluxes in a large floodplain river system by combining discharge modelling, floodplain-aware routing, and satellite-derived suspended sediment estimates. The topic is timely and potentially high impact. I have listed my section based comments below.
Abstract
While I understand the large goal of the paper is to highlight the role of floodplain in sediment transport, and the paper is attempting to build an integrated approach using discharge estimates to calculate fine sediment load and land surface model to simulate sand load. I cannot understand from the latter half of the Abstract what critical research questions this paper attempts to solve, except the role of floodplain in sediment transport.

The abstract misses in providing uncertainty or targeted application of this study, as the last three lines reads very generic. Until the reader reads the full paper, he/she may not understand the latter part of Abstract.

Introduction: section 1.4
What are the key research questions that the authors are trying to address in this research?Role of floodplain in sediment transport? The paper should state 2–3 explicit objectives/hypotheses early (end of Abstract + end of Intro). Kindly add a 1–2 sentence novelty statement distinguishing this from prior Amazon basin sediment studies

Figure 1
Figure 1 is scattered. How a), b) & c), and d) are linked with each other is not easy to comprehend with the figure. Kindly sync them.

Integrative strategy
I doubt the assumption of assuming mean diameter for fine sediment in the range of 10-20 micrometer. This because

Firstly, are there any records of PSD profile of the basin.
Secondly, 10- 20 micrometer sits at the lower range of very fine silt. Thus, underrepresenting coarse silt or large composition of fine sediments in the range of 20- 100 Micrometer. The sediment transport of both the group large differs, with former influenced by flocculation, organic matter, and high suspension. And latter having conventional particle dynamics. To simplify, why don’t the authors use at least 3 particle size to better represent the PSD, clay, silt and sand.
I feel the current setup may underrepresent sand load in the model. Could you please clarify and provide support?

3.1 Conventional data
Kindly introduce the term “super stations” in this section as you talk about long data stations, the virtual stations and low data station in the next station.

Figure 2.
The blue color interferes with the color used in the map. Kindly use some other color to mark the stations. Perhaps red. It is not needed to write the years in the map.

“The text box details observation periods for water level (𝒉) discharge (𝑸), and suspended sediment concentration (𝑪).” It is not easy to comprehend.

River mapping is not clear

3.3.2
Could you please clarify, how total, fine, and sand concentration was determined. Were the samples dry or wet sieved. Was gravimetric filtration used or else?

3.3.3
How the vertical gradient in sediment concentration was considered in the field measurements.

Does the site also have lateral change in the suspended sediment concentration? Was the mean concentration representative vertically and/or laterally? How would author try to convince this?

4.1
Please explain how reach-wise inputs are derived and how this can be reproduced in other basins. Additionally, given floodplain activation is event-scale, please discuss how a daily routing framework affects peak timing/magnitude and floodplain exchange, and whether it biases sediment flux estimation.

4.2
The approach relies on an effective bed-material sand diameter; please justify its spatial representativeness (Amazon has strong upstream–downstream contrasts) and provide a sensitivity test to plausible diameter ranges (or consider a simple multi-class sand representation).

For Sections 5–6,
13. I recommend strengthening the separation of calibration and validation (including timing metrics), making ‘not shown’ key comparisons available in Supplement, and adding a clearer treatment of uncertainty.
The Sobol analysis is performed on 1 sub-basin. Would the authors clarify how this is representative and ideally include at least one additional contrasting sub-basin.

In sand routing sensitivity, I would like to clarify whether the PSD is stationary or seasonally varying.

They state discharge calibration is insensitive to and under some conditions. kindly specify it is sensitive to which regimes (low flow vs floodplain-active) and show quantitatively.

The authors argue that sand resuspension is driven by floodplain recession and banks/bars erosion and that bed erosion is negligible because allowing bed erosion does not fit peaks as well. Isn’t it a strong process statement to make, just based on calibration.

Reply
Citation: https://doi.org/10.5194/egusphere-2025-4101-CC1

William Santini, Alexandre Delort-Ylla, Jean Michel Martinez, Waldo Lavado-Casimiro, Benoît Camenen, Jérôme Le Coz, Joana Roussillon, Jhonathan Junior Pérez Arévalo, and Jorge Molina-Carpio

Supplement

https://doi.org/10.5194/egusphere-2025-4101-supplement

Model code and software

SWAT-Amazon William Santini https://github.com/william-santini/SWAT-Amazon

William Santini, Alexandre Delort-Ylla, Jean Michel Martinez, Waldo Lavado-Casimiro, Benoît Camenen, Jérôme Le Coz, Joana Roussillon, Jhonathan Junior Pérez Arévalo, and Jorge Molina-Carpio

Viewed

Total article views: 462 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
273	161	28	462	24	15	16

HTML: 273
PDF: 161
XML: 28
Total: 462
Supplement: 24
BibTeX: 15
EndNote: 16

Views and downloads (calculated since 24 Oct 2025)

Month	HTML	PDF	XML	Total
Oct 2025	119	14	8	141
Nov 2025	68	58	7	133
Dec 2025	86	89	13	188

Cumulative views and downloads (calculated since 24 Oct 2025)

Month	HTML	PDF	XML	Total
Oct 2025	119	14	8	141
Nov 2025	68	58	7	133
Dec 2025	86	89	13	188

Viewed (geographical distribution)

Total article views: 452 (including HTML, PDF, and XML) Thereof 452 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 27 Dec 2025

Short summary

This study presents the first long-term hydro-sediment balance of the Ucayali River using observations, remote sensing, and SWAT-Amazon. The river exports 455 Mt yr⁻¹ of sediment, 36 % trapped in the floodplain. Floodplain recycling contributes 22 %, while a newly identified sand deposition process during floods decouples sediment from water fluxes. The study highlights the need for robust hydrological networks and regional assessments to better capture Amazonian sediment dynamics.


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