Concentrations and fluxes of suspended particulate matters and associated contaminants in the Rhône River from Lake Geneva to the Mediterranean Sea

. The Rhône River is amongst the main rivers of Western Europe and the biggest by freshwater discharge and sediment 15 delivery to the Mediterranean Sea. Its catchment is characterized by distinct hydrological regimes that may produce annual sediment deliveries ranging from 1.4 to 18.0 Mt y -1 . Furthermore, the course of the Rhône River meets numerous dams, hydro- and nuclear power plants, and agricultural, urban or industrial areas. Thus, suspended particulate matters (SPM) have been involved in the fate of hydrophobic contaminants such as polychlorobiphenyls (PCB), mercury (Hg) and other trace metal elements (TME), and radionuclides for decades. To investigate the concentrations and the fluxes of SPM and 20 associated contaminants, as well as their sources, a monitoring network of 15 stations (three on the Rhône River and 12 on tributaries, from Lake Geneva to the Mediterranean Sea) has been set up in the past decade within the Rhône Sediment Observatory (OSR). Its course meets numerous dams, samples then sealed in quartz tubes under a vacuum with an excess of CuO and silver wire. Organic carbon was converted into CO2 by introducing the tubes into a furnace at 835 °C for 5 h and then released, dried, measured, and collected after broking the tube under a vacuum. The graphite target is obtained with a direct catalytic reduction of the CO 2 , using iron powder as a catalyst (Merck® for analysis reduced, 10 μm particles). The reduction reaction occurs at 600 °C with excess H2 (H2/CO2 = 325 2.5) and is complete after 4–5 h. The iron-carbon powder is pressed into a flat pellet and stored under pure argon in a sealed tube. All quartz and glass dishes are burned for at least 5 h at 450 °C to reduce contamination. To evacuate the vacuum lines, a turbo-molecular pump reaching 10−6 mbar is used. Measurements are performed using a 3 MV NEC Pelletron Accelerator coupled with a spectrometer dedicated to radiocarbon dating, measuring 12C, 13C and 14C contents and counting the 14C ions by isobaric discrimination. Analysis require 1 to 100 mg of dry sample (to obtain 1 mg of carbon). The specific 14C activity is expressed as Becquerel of 14C per kilogram of total organic carbon (Bq kg −1 of C). The detection limit is 0.8 Bq kg -1 of C and the uncertainty is 0.1% for modern samples (k= 2).

especially farming and grazing. The Rhône watershed covers about a 5 th of the surface of metropolitan France, which implies the transport of eroded material from a wide variety of land uses. Therefore, anthropic contamination of the Rhône River by hydrophobic organic contaminants such as polychlorobiphenyls (PCBs) and polycyclic aromatic hydrocarbons (PAHs), or by trace metal elements (TME), or radionuclides has been observed for many years (Radakovitch et al., 2008;Mourier et al., 70 2014;Delile et al., 2020;Eyrolle et al., 2020). For these substances, SPM transport represents the main driver of contaminants from rivers to coastal areas, leading to an alteration of bio-geochemical cycles and water quality (Horowitz, 2009).
In this watershed, studies conducted on sediment dynamics and associated contaminants are unfortunately scarse (Antonelli et al., 2008;Radakovitch et al., 2008;Panagiotopoulos et al., 2012;Delmas et al., 2012) and do not allow to understand the 75 observed changes over the long term. On this basis,T the monitoring of spatial and temporal distribution of SPM and associated contaminants in the Rhône River has been conducted within the Rhône Sediment Observatory (OSR) since 2009 (Le Bescond et al., 2018). Fifteen monitoring stations have been installed with 3 of them along the Rhône River channel and 12 on the main tributaries. This monitoring network was designed to improve our understanding on SPM transfer processes in rivers exposed to anthropogenic contamination and extreme hydro-sedimentary events (flood, low-water, dam regulation), 80 and to provide stakeholders with precise values of SPM and contaminants fluxes. The current database provides time series of Suspended Solid Concentrations (SSC), SPM fluxes, Particle Size Distributions (PSD), Particulate Organic Carbon (POC) contents, and the concentrations and fluxes of several contaminants of interest (Thollet et al., 2021).
The hydro-sedimentary and contaminant monitoring data from the Rhône River and the main tributaries are useful for the assessment of: 85 • The annual and inter-annual fluxes of SPM and hydrophobic organic contaminants and TME and radionuclides; • The spatial and temporal variations of the contaminant concentrations; • The impact of extreme events (floods or dam flushing operations) in term of sediment budget and contaminant concentrations.

Description of the data and the functionality of the database (BDOH/OSR) 90
In the database, two types of time series are stored. Discontinuous time series are used for measurements on SPM samples that are collected from a start time to an end time (with no information in between sampling periods). Calculated time series are obtained by several transformations (including data completion) of the discontinuous time series and the SSC (Fig. 1).
The dataset includes the following calculated parameters: • Suspended Solid Concentration -SSC (mg L -1 ) either derived from in-situ turbidity measurements or by filtration of 105 water samples as explained in the Method section. The procedure for data completion are presented in the Method section, • SPM flux (t s -1 ) was calculated in the database by multiplying water discharge (not made available in the dataset because it is released separately by data producers -in m 3 s -1 ) and SSC (Fig. 1). The owners of the water discharge data are reported in Tables 2 and 3,  110 • Contaminant flux (g s -1 or Bq s -1 ) was calculated in the database by multiplying the SPM flux and the contaminant concentration (Fig. 1). The procedures for flux computation and data completion of the discontinuous contaminant time series are presented in the Method section. Data are organized by station with information on the provider, the period of availability (date in Coordinated Universal Time -UTC) and the number of data ( Fig. 2A). For particulate contaminant concentration, the name of the time series is related to the sampling method, as illustrated in Fig. 2A for Hg, with: -CX = discontinuous concentration of particulate contaminant X in SPM samples collected by continuous flow 120 centrifuge, -CX-2 = discontinuous concentration of particulate contaminant X in SPM samples collected by particle trap. the period of the time series by selecting start and end dates; the file format: BDOH (raw) or Hydro2-QTVAR (specific format used by the French national hydrological services); the type of transformation and time steps: identical (raw data, constant or variable time steps), linear interpolation at 135 constant time steps, mean (1 or 6 hours, daily, monthly, yearly, event scale) or accumulation (1 or 6 hours, daily, monthly, yearly, event scale). This parameter is only available for calculated and continuous time series; the time zone (by default in the database: UTC +00).
Data are sent by e-mail in a compressed folder (export.zip) containing the time series as flat text files (STATION_PARAMETER.txt) and another file (Report.txt) that contains all the necessary metadata (producer, parameter, 140 genealogy , time zone, conversion factors). The file of the time series can be processed by any software (Excel, R, Matlab, etc.). Each value of the time series is associated with a quality code ( Fig. 2D) according to Table 1. the value (coded "l" or "i") was estimated or modeled following the method described in the Method section .
lq limit of quantification the value is lower than the limit of quantification ld limit of detection (used for radionuclides only) the value is lower than the limit of detection 3 Methods

Sampling location 145
The monitoring conducted within the OSR is located in the Rhône River catchment downstream of Lake Geneva in Switzerland (~95 000 km²) (Fig. 3). Due to its large volume and high trapping efficiency, Lake Geneva is a barrier to SPM coming from the upstream catchments so that its sedimentary output can be neglected. Fifteen monitoring stations have been installed ( Fig. 3) with three of them along the Rhône River (Table 2) and 12 on the main tributaries (Table 3).

Suspended Solid Concentration (mg L -1 )
The SSC at most stations are derived from in-situ turbidity measurements conducted every 10 minutes (Le Bescond et al., 165 2018). Universal Controller SC100 or SC200 (HachLange, Germany) are used in addition to numerical Solitax SC optical turbidity probes (Fig. 4A), all equipped with a mechanical cleaning system (wiper). Sensors use the infrared scattered light method with the optical response being dependent on sediment characteristics in the water. The turbidity meter is usually immersed at a fixed position along the riverbank near the station, avoiding dead zones or effluents so that the measured turbidity is representative of the average turbidity throughout the river cross-section. Exceptions: at Jons, river water is 170 pumped and circulated to an in-door turbidity meter; at Arles, there is no turbidity meter. The SSC is then calculated through the site-specific turbidity-SPM rating curve (Navratil et al., 2011), which is determined on each site for a wide range of concentrations (Table 4). The curves are established using water samples collected manually or by automatic samplers (Fig.   4B) triggered hourly during flood events. Water samples are collected regularly to ensure there is no change in the relationship between turbidity and SPM. A new turbidity-SPM rating curve is systematically built when a turbidity probe is 175 replaced. For the Isère, the Durance and the Andancette stations, the conversion is computed by the external provider (Table   2 and 3). In order to determine SPM values from these samples, they are filtered through pre-weighed glass fiber filters, dried (105°C during 2 hours) and weighted according to the standard method NF EN 872 (AFNOR, 2005). Relative uncertainty on SPM concentrations is estimated to 9% (at 95% uncertainty level, coverage factor k=2).  (Raimbault et al., 2014). Water intake is located on a floatable structure at a distance of 7 m from the bank and 0.5 m under 185 the surface. Sampling for SSC is achieved using a cooled automatic water sampler that fills a daily bottle with 150 mL every 90 minutes (Eyrolle et al., 2010). At Arles, near the outlet of the Rhône River to the Mediterranean Sea, SSC are measured by the MOOSE network (Mediterranean Ocean Observing System for the Environment) (Raimbault et al., 2014). Sampling is achieved using a cooled -down automatic water sampler that fills a daily bottle with 150 mL every 90 minutes (Eyrolle et al., 2010). During flood events (water discharge greater than 3000 m 3 s -1 ), 150 mL samples are collected every 30 minutes to 190 constitute a composite sample every 4 hours. Samples are poisoned with HgCl2 and kept at 5°C until they are filtered on GF/C Whatman pre-conditioned glass fiber filters (dried at 500°C for 4 hours). The filtered volumes are adapted to the charge of SPM.

Sampling of SPM for analysis
The SPM used to measure POC, PSD and contaminant concentrations are mainly collected using particle traps (PT, Fig. 4C), while continuous flow centrifuges (CFC, Fig. 4D) are used to monitor specific events such as fast flood event. The PT are 200 rectangular stainless-steel boxes, whose internal flows circulate in two distinct parts separated by plates (Schulze et al., 2007;Masson et al., 2018). Such integrated sampling conducted near the riverbank at a depth of 0.5 -1 m allows the collect of sufficient amounts of SPM for contaminants analysis. The PT are immersed near the riverbank (Fig. 4C) avoiding dead zones or effluents so that the sampled material is representative of the river fine suspension throughout the cross-section. For Andancette and the Saône river monitoring stations, the PT are suspended from a chain and kept immersed at a depth of 0.5 -205 located inside the SORA monitoring station (Eyrolle et al., 2010) and supplied by a pipe. These devices allow to take into account the fluctuations of the SPM flux (Le Bescond et al., 2018;Delile et al., 2020). The PT is generally collected every month but can be collected at shorter time intervals in order to monitor specific events such as a flood or dam regulation.
Unfortunately, PT were sometimes not recovered due to logistic constraints including high level of water or vandalism. The 210 measurements conducted on PT samples are considered as time-averaged over its sampling period. The purpose of the PT is to obtain an integrative response over a period, which does not allow for the assessment of variation that may occur within that sampling period.At Arles, water intake is located on a floatable structure at a distance of 7 m from the bank and 0.5 m under the surface regardless the water discharge and continuously supplies several devices including a PT and a CFC (Eyrolle et al., 2010). The measures conducted on PT samples are considered as a mean of its sampling period. 215 In addition, SPM samples for contaminants analysis are occasionally collected using a high speed CFC (Masson et al., 2018), especially for the TME at Arles and at Jons. The duration of pumping can be regulated from 10 min to 8 hours in order to collect sufficient amount of SPM. No significant differences in Hg and PCB concentrations were found in samples collected by the two methods although particles in PT are slightly coarser than particles collected using CFC (Masson et al., 2018).
The analyses are carried out on the total samples without separation of the organic part because it is negligible in the samples 220 (see the POC measurements). Prior to chemical analysis, SPM collected with the two sampling techniques are transferred to Clean brown glass bottles (250 mL) are used to transfer the SPM. Prior to chemical analysis, samples were deep-frozen (-18°C) and freeze-dried before being homogenized in an agate mortar and stored in the dark at ambient temperature.
In addition, excess SPM samples are stored in a chamber at -80°C. This will allow, according to the needs and the development of new analytical techniques, to carry out later analysis without aging of the samples. More than 1300 samples 225 are currently stored this way. Meta-data and location of the samples inside the chamber are saved within the software Collec-Science (Quinton et al., 2020).

Physico-chemical analyses
All the physico-chemical analyses information are briefly recalled during visualization and in an additional file (Report.txt) when downloading the data. 230

Particle size distribution (PSD)
The PSD is measured by laser diffraction with a Cilas 1190 particle size analyzer (Cilas SA, Orléans, France). The volumetric particle size distribution of SPM (measuring range: 0.04 to 2500 µm) was assessed by INRAE-RiverLy Aquatic chemistry laboratory (AFNOR, 2009). During measurement (obscuration rate typically 15%), mechanical agitation in the tank (at 350 rpm) and circulation with a peristaltic pump (at 120 rpm) were used in order to homogenize the sample. A 235 refractive index in the range of kaolinite was used for the solid phase (RI=1.55). Ultrasounds were used during dispersion and during measurement in order to avoid particle aggregation (20 seconds at 38kHz). The PSD was also measured without ultrasound for comparison. The volumetric particle size distribution of the sample was computed using the Fraunhofer optical model. A quality control sample (made by INRAE) was systematically used to control the device.
Before 2018, the PSD of Ardèche, Durance, Gardon and Arles stations were measured without ultrasounds by a Beckman 240 Coulter LS 13 320 (Beckman Coulter, Fullerton, CA, USA) at the CEREGE laboratory. Prior to measurement with this device (measuring range: 0.04 to 2000 µm), the organic component of the water sample was oxidized using a solution of 30% hydrogen peroxide (H2O2) at 200 °C. The remaining fraction was then resuspended in a 0.3% hexametaphosphate solution and sub-sampled under stirring (800 rpm) to match the optimal obscuration windows of the laser and of the light polarization system, between 8 and 16 % and between 50 and 70 % respectively. The volumetric particle size calculation 245 model was performed in accordance with the Fraunhöfer and Mie theory. A refractive index in the range of kaolinite was used for the solid phase (RI=1.56). Each sample was analyzed 6 times (90 seconds each) with water circulation at 80% and the result finally recorded is an average of the 5 last runs, because some air bubbles sometime alter the first run just after the rinsing phase. Sample size reproducibility based on independent replicate measurements of the same sample did not exceed 2% residual (Psomiadis et al., 2014). 250 The database provides information on characteristic diameters D10, D50 and D90 of the particle size distribution. Non negligible differences between these parameters were found during an intercomparison of the two devices (Lepage et al., 2019). The comparison of the results must therefore be done with caution. Finally, the entire information on particle size distribution of each sample is stored by INRAE and CEREGE laboratories.

Particulate organic carbon (POC) 255
At Arles, POC was measured in filtered water samples within the MOOSE observatory (Raimbault et al., 2014). The SPM was collected on GF/C Whatman glass fiber filters (25 mm in diameter) precombusted at 500°c during 4 hours. Filters were dried at 60°C and stored dried until analysis. Before analysis the filters were placed in tin capsules and acidified with sulfuric acid (0.25 N) to remove inorganic carbon, and dried at 60°C. The POC measurements were performed using high combustion procedure (950° C) on a CN Integra mass spectrometer (serCon Ltd, Crewe, UK) according to Raimbault et al. 260 (2008). Reference materials (glycine and casein) were systematically used to control analytical uncertainty.
For all the other stations, the determination of POC before December 2014 was performed using a CHN Flash 2000 carbon analyzer (ThermoFisher-Scientific, USA) by INRAE-RiverLy Aquatic chemistry laboratory. Prior to analysis, samples were decarbonated using hydrochloric acid HCl (AFNOR, 1995). Depending on the POC concentration, the analytical uncertainty ranged between ~3% and ~6% (k=2) while the limit of quantification (LQ) was estimated to be 0.1 g kg −1 . After December 265 2014, a different device was used (CNS Flash 2000, ThermoFisher Scientific, USA) following manufacturer recommendations and the above method. Analytical accuracy (93%) and uncertainty (8%; k = 2) were controlled using a reference material (AGLAE, 15 M9.1; 40 g kg −1 ) and the LQ was estimated to be 0.5 g kg −1 .
To avoid sulfur interferences, small amount of copper powder (10 mg mL 1 ) was added prior to gas chromatography 295 analysis with a 63 Ni electron capture detector (GC-ECD). Two columns were used to analyze the samples (RTX®-5 and RTX®-PCB) and the accuracy was checked via the analysis of a certified reference material (BCR 536) and intercomparison exercises. The LQs were estimated between 0.5 and 1 mg kg 1 depending on the congeners while the analytical uncertainties was determine using a sediment sample from the Bourbre River as no certified reference material exists for such low levels of PCBi in equivalent matrix. Analytical uncertainties were estimated to be 60% (k = 2) for 300 concentrations lower than 3-times the LQ, and to be 30% (k = 2) for concentrations higher than 3-times the LQ.

Radionuclides (cesium-137, organically bound tritium and radiocarbon)
For 137 Cs activity, the SPM samples were ashed and put into tightly closed plastic boxes (17 mL or 60 mL) for gamma-ray spectrometry measurements (20-60 g) using low-background and high resolution High Purity Germanium detectors at the IRSN/LMRE laboratory (Eyrolle et al., 2020). These measurements being performed under 17025 accreditation, the 305 laboratory participates each year in proficiency tests organized mainly by the ALMERA network (Analytical Laboratories for the Measurement of Environmental Radioactivity) of the IAEA. Each sample was measured for 3 days to achieve detection limits around 0.5 Bq kg -1 d.w. for 137 Cs, after waiting for 30 days for the radioactive equilibrium of the Ra-226's progeny. Efficiency calibrations were constructed using gamma-ray standard sources in a 1.15 g cm 3 density solid resinwater equivalent matrix. Activity results were corrected for true coincidence summing (TCS) and self-attenuation effects 310 (Lefèvre et al., 2003). Measured activities, expressed in Bq kg -1 (d.w.), are decay-corrected to the date of sampling. The activity uncertainty (k=2) was estimated as the combination of calibration uncertainties, counting statistics, and summing and self-absorption correction uncertainties.
Organically bound tritium (OBT in Bq kg −1 dry) analysis was performed at the IRSN/LMRE laboratory by the Helium-3 (3He) ingrowth method (Cossonnet et al., 2009). The samples were put under vacuum (10 −6 mbar) and stored up to 4 months 315 prior to analysis. The 3He/4He ratio was measured by mass spectrometry after correction of the radiogenic 4He levels contained in the sample (gaseous inclusions) and normalization of the values to ambient atmospheric levels (3He/4He).
Radiocarbon ( 14 C) contents was analyzed by the IRSN/LMRE laboratory using an accelerator mass spectrometer (LMC14 laboratory, Saclay, France). Prior to analyze the organic part of the sample, carbonates were eliminated by washing the 320 sample (0.5MHCl, 0.1MNaOH) and drying it under vacuum as describe in Eyrolle et al. (2018). Decarbonated samples were then sealed in quartz tubes under a vacuum with an excess of CuO and silver wire. Organic carbon was converted into CO2 by introducing the tubes into a furnace at 835 °C for 5 h and then released, dried, measured, and collected after broking the tube under a vacuum. The graphite target is obtained with a direct catalytic reduction of the CO2, using iron powder as a catalyst (Merck® for analysis reduced, 10 μm particles). The reduction reaction occurs at 600 °C with excess H2 (H2/CO2 = 325 2.5) and is complete after 4-5 h. The iron-carbon powder is pressed into a flat pellet and stored under pure argon in a sealed tube. All quartz and glass dishes are burned for at least 5 h at 450 °C to reduce contamination. To evacuate the vacuum lines, a turbo-molecular pump reaching 10−6 mbar is used. Measurements are performed using a 3 MV NEC Pelletron Accelerator coupled with a spectrometer dedicated to radiocarbon dating, measuring 12C, 13C and 14C contents and counting the 14C ions by isobaric discrimination. Analysis require 1 to 100 mg of dry sample (to obtain 1 mg of carbon). The specific 14C 330 activity is expressed as Becquerel of 14C per kilogram of total organic carbon (Bq kg −1 of C). The detection limit is 0.8 Bq kg -1 of C and the uncertainty is 0.1% for modern samples (k= 2).

Data completion and flux calculation
The SPM fluxes are the product of water discharges and SSC, and contaminant fluxes are the product of the SPM fluxes and the contaminant concentrations ( Fig. 1). At most stations, water discharge is provided by a collocated or neighbouring 335 hydrometric station. Most often, water levels are measured using pressure sensors, pneumatic probes (bubblers) or radar gauges, and the stage records are converted to discharge using a stage-discharge rating curve (Le Coz et al., 2014;Kiang et al., 2018), or a stage-fall-discharge rating curve (Mansanarez et al., 2016) for stations affected by variable backwater upstream of a dam. Hourly averaged water discharge data are generally calculated by conversion of water level measurements through stage-discharge rating curves, otherwise through numerical modelling. At Jons, the closest 340 hydrometric station is relatively far upstream, and two tributaries bring significant amounts of water between the hydrometric and the turbidity station. Therefore, a 1-D hydrodynamic model (Dugué et al., 2015;Launay et al., 2019) is used to compute the discharge time series at Jons from the three discharge times series measured upstream on the Rhône River (at Lagnieu) and on the two tributaries (Ain and Bourbre Rivers). Hourly averaged water discharge data are calculated at all the stations, generally by conversion of water level measurements through stage-discharge rating curves, otherwise through 345 numerical modelling. Prior to calculate these fluxes, completion of missing values (gaps in the measurement or nonmonitored periods) and time step transformation are required.

Completion of missing values of SSC
Missing SSC values are estimated using the empirical relations between water discharges (Q in m 3 s -1 ), and SSC (Cs in mg L -1 ) also known as sediment rating curves (Horowitz, 2003;Sadaoui et al., 2016). The Q-SSC relations (Eq. (1)) were 350 improved by considering a low/moderate water discharges segment (a1 and c1) and a high water discharges segment (a2, b2 and c2) (Sadaoui et al., 2016): The BaRatin method and the BaRatinAGE software (Le Coz et al., 2014) was used to fit the regressions and detects the 355 breakpoints (k in m 3 s -1 ), above which the regression coefficients change significantly. Discharge-SPM rating curves are too uncertain to allow the detection of potential temporal changes. We therefore assume that they are constant over the monitoring period. Estimated parameters of Eq. (1) are listed by station in table 4 according to Poulier et al. (2019) and updated in 2021 with new data. Estimated values take the code "e" (Table 1).  (Fig. 5 and 6). 365 Missing values were replaced by the median value of the contaminant concentration depending on the hydrological conditions (baseflow or flood) during the integrated sampling period as described in Delile et al. (2020) (Fig. 5). In brief, samples were considered as taken in flood when more than 50% of the SPM cumulative flux occurred while the water discharge was higher than the flood threshold (defined as half of the 2-year flood peak discharge). For gap periods greater than the usual time period of sampling (28 days), the gap periods were split in two to avoid gap periods greater than 1.5 370 times this usual sampling period.
For CFC, missing values are considered when the period between two samplings is longer than the usual time period of sampling. For gaps shorter than the usual sampling period, discontinuous time series are transformed to continuous time series by considering the concentration of the last sample until the half of the gap period, while the other half was filled with the next sample concentration (Fig. 5). Like with PT, median values for hydrological conditions were calculated to estimate 375 the missing values (Fig. 5). The hydrological condition of a sample was considered as flood if the mean daily water discharge value was higher than the flood threshold. For gaps longer than the usual sampling period, the same rule as for PT was applied.
For both methods, values lower than the LQ were replaced by this LQ divided by 2. Estimated values take the code "e" (Table 1). 380

Time step transformation
After their completion, the time series are set to the same time step by linear interpolation to the nearest second between two points, so as not to lose any information. For samples collected by PT, the concentration measured is considered as a mean of its sampling period. The SPM (g s -1 ) and contaminant fluxes (g s -1 or Bq s -1 ) calculation is then carried out following Eq. (2) 390 and Eq. (3), respectively. The lowest quality code of both values used is kept by respecting the following ranking: v>a>d>e. FMES = Q * CMES * 1000 with Q in m 3 s -1 and CMES in g L -1 .

Examples of applications using the dataset
With the data acquired by this network, it was possible to improve the calculation of the SPM and associated contaminant 395 fluxes near the outlet of the Rhône River, and also to evaluate the fluxes coming from the upper Rhône River and the tributaries (Poulier et al., 2019;Delile et al., 2020). It was estimated that on average 6.6 Mt of SPM transited each year in the Rhône at Beaucaire, with strong variation ranging from 1.4 to 18.0 Mt y -1 . The Durance and Isère tributaries were found to be the main contributors to SPM fluxes.
Through this long-term continuous monitoring and its spatial resolution with sampling on the main tributaries, seasonal 400 variations in several contaminant concentrations have recently been highlighted (Delile et al., 2020) as well as the impact of dam flushing operations (Lepage et al., 2020). Nearly two-thirds of the annual contaminant fluxes are released into the Mediterranean Sea during three short term periods over the year: 24% during a Mediterranean component in November, 15% during oceanic rainfall component in January and 24% during nival component in May-June. During flushing operations recorded from 2011 to 2016, the mean SPM concentrations were 6 to 8 times higher than during flood events at equal water 405 discharge (Lepage et al., 2020).
At the Jons station, an original fingerprinting method was conducted based on the TME residual fraction in SPM (Dabrin et al., 2021). This approach demonstrated that under base flow conditions and during dam flushing operations, SPM originated mainly from the Arve River, while the origin was more contrasted during flood events.
Finally, within the OSR and using this database, a 1-D hydrodynamic model of SPM dynamics was developed (Launay et 410 al., 2019). This model simulated the concentration of SPM at Arles during a flood event occurring in the Isère and the Durance tributaries. This model was also applied by Dabrin et al. (2021) for the fingerprinting approach and the combination of these two approaches demonstrated that a large portion of SPM from the Arve River was old sediment stored behind the Verbois dam and re-suspended during the dam flushing operation.

Data availability 415
All the data are made publicly available in French and English through the BDOH/OSR database (Thollet et al., 2021) at https://doi.org/10.15454/RJCQZ7 (Lepage et al., 2021). The BDOH (Base de Données pour les Observatoires en Hydrologie) application is managed by INRAE. The data is freely available for visualization, and for download after registration of a personal account. In order to respect the anonymity of the reviewers, we have created a specific account to access this feature : -login : review.bdoh@gmail.com -password : bdohreview. 420

Perspectives
The database presented in this paper will be continuously updated in the coming years, at least until the end of the current OSR program (2021)(2022)(2023)(2024). Discharge, SSC, and particulate contaminants are continuously measured at the permanent stations and data are regularly updated online. The following improvements are planned: -Uncertainties on the SPM and contaminants fluxes will be calculated and published; 425 -Other TME and new contaminants such as gadolinium will be included in the dataset depending on the authorities's concern and scientific purpose. These contaminants might also be analyzed on the SPM samples already collected and stored in the chamber at -80°C; -Additional parameters to describe the particle size distribution will be added such as the percentage of clay/silt/sand; 430 -Link between the different contaminants measured on the same SPM sample will be added as a common sample name; -The values lower than LQ or LD will be replaced by a more precise statistical method when feasible.