A fuzzy classification of the hydrodynamic forcings of the Rhone River plume: An application in case of accidental release of radionuclides
Introduction
The Rhone River catchment extends over 98,000 km2 and covers one fifth of the French metropolitan territory. It is the main source of particles and freshwater for the Gulf of Lion in the North Western Mediterranean sea (Durrieu De Madron et al., 2000), and all together one of the most important input to the Mediterranean sea (Ludwig et al., 2009). The Rhone valley also hosts the largest concentration of nuclear power plants in Europe with 4 nuclear power plants in process and a spent fuel reprocessing center, under dismantlement since 1997. Eyrolle et al. (2020) recently synthesized the studies showing that this river carries artificial radionuclides from decades, resulting from authorized releases of low level radioactive liquid wastes and from the export of atmospheric deposits on watersheds consequently to nuclear weapons testing and Chernobyl accident.
France is presently ranked second in the world for the production of nuclear energy, and the total electricity production in the combined regions of Northern, Western and Southern Europe is projected to increase by 2050 ((IAEA, 2019)). Also, the risk of incident on any kind of nuclear installations is still of concern in France and must be taken into account. As for any river, the transport of artificial radionuclides in case of accidental release occurs both in dissolved and particulate form, depending on the amount of suspended particulate matter and on the chemical properties of the radionuclides, and particularly their distribution coefficient (Tomczak et al., 2019). For the Rhone River, the prediction of dissolved vs particulate fluxes and the associated time scale for transit can be evaluated through numerical modeling (Launay et al., 2019), but the behavior of radionuclides once at sea is clearly less constrained, because it will primarily depend on the forcings governing the shape of the Rhone River plume.
The area of the Rhone river mouth is characterized by a very small tidal amplitude about 30 cm inducing the formation of a sedimentary delta. As usual in this case, the freshwater input forms a thin stratified plume of low salinity water (and higher turbidity) overlying the seawater and extending between 4 and 1000 km2 (Estournel et al., 1997; Gangloff et al., 2017) with a thickness decreasing seaward (Pairaud et al., 2011; Gangloff et al., 2017). It is preferentially deflected westward in a clockwise orientation running East to West (Reffray et al., 2004) due to the general circulation induced by the Northern Current along the continental slope. Under north-northwest winds, the plume extends offshore towards the southwest, whereas it is pushed to the coast west of the river inlet in case of southeastern winds. Satellite and modelling results have also shown that the plume size increases with river discharge (Fraysse et al., 2014; Gangloff et al., 2017). More episodic processes impact the plume pattern such as dense water formation and cascading (Ulses et al., 2008), upwelling cells and marine storms (Millot, 1990, 1999). As a result, this plume extends far beyond the coastal areas and may covers a large area in the GoL extending from the vicinity of the mouth up to the Cap de Creus at the french-spanish border (Sanchez-Cabeza et al., 1992). It can also reach the Gulf of Fos (Gontier et al., 1992; Charmasson et al., 1999) or the Bay of Marseille (Pairaud et al., 2011; Fraysse et al., 2014) on the eastern side of river mouth. The Gulf of Fos is an important economic area with one of the biggest commercial port in Europe and a large shellfish area, and Marseille is one of the biggest Mediterranean coastal city with one million inhabitants. Due to the oligotrophic nature of the Mediterranean Sea, the region of freshwater influence (ROFI) of the Rhone river has a major influence on the distribution of plankton groups (Diaz et al., 2019) and thus pelagic catches on the GoL. Obviously, the inputs of chemical contaminants from the Rhone River can greatly affect the fishery activity.
The combination of meteorological and hydrodynamic forcings with the dynamics of the Rhone River discharge results in a large spatio-temporal variability in freshwater and associated pollutants delivery to the GoL (Martin et al., 2019). If an accidental release occurs in the Rhone River, the dissolved radionuclides may reach the estuary within 48 h h to few days depending on the source location and water discharge [unpublished results]. Once at sea, the different shapes that the plume may present will depend on hydrodynamic and weather conditions and will lead to contaminate different areas. Since one goal of radioprotection is to predict the transfer of radionuclides in the environment, there is a need to anticipate their dispersion at any time and in any kind of meteorological and hydrodynamical conditions.
Different numerical hydrodynamic models have been set up in the GoL, including the river mouth (Pairaud et al., 2011; Duffa et al., 2016), and they could be used actually in case of accidental release in order to predict the behavior of the freshwater input. However, the delay necessary for their implementation will range from few hours to few days, whereas very quick and concise information should be provided to experts and decision-makers as a first picture of the local issues. Alsothe potentially impacted zones will be better defined by performing a fine spatial scale simulation adequately centered, compared to a large-scale simulation.
As a result, a preliminary study embraces all possible plume patterns is necessary, and a first step is to target the general behavior of the estuarine-plume system. Bárcena et al. (2015) explained that two approaches may be conducted for that: simulating several scenarios using constant conditions of hydrodynamic forcings or simulating few scenarios using the most frequent or extreme real hydrodynamic forcings during short-medium term periods (month to year).
These authors demonstrated that the first approach is not complete because real forcings cannot be deduced from the combination of simple idealized scenarios. The second approach relies on a subjective selection of scenarios by an expert and it will have an expensive computational cost for simulations if the need is to get on overview of the different kind of realistic responses of the estuarine-plume mean behavior. In this case, and to minimize subjectivity, a methodology based on data mining should be able to select the most relevant condensed hydrodynamic scenarios, taking into account the time evolution and the occurrence probability of the forcings.
Plume classifications based on satellite observations or hydrodynamic model output have been defined in several river-sea systems using Empirical Orthogonal Function or Self-Organizing-Map (Falcieri et al., 2014; Xu et al., 2019). Such classification method deals with large spatial scale but implies a heavy data pretreatment like « masking » to treat the satellite data or for the computation of the model. In addition, the need of long-term environmental databases (e.g., 10–20 years) to assess probabilities implies significant computational costs as well as long and multiple series of data to be used as boundary conditions and climatic forcings. Another approach is to classify the main hydrodynamics drivers by looking for example at the catchment discharge and the winds intensities and directions (Kaufmann and Whiteman 1999; Zhang et al., 2011). Since the plume response to these forcings can be longer than 24h (Demarcq and Wald, 1984; Estournel et al., 1997), the classification should work observation by observation but must also keep consistency over longer temporal scales of few days in order to be accurate. Clustering performed on temporal series helps to assess the consistency of a trend over time, and a fuzzy clustering algorithm provides a continuous cluster membership function allowing to spot significant trend changes.
In this context, this paper presents a methodology based on statistical analysis and numerical modelling that was developed to address the limitations of the previously mentioned approaches. Firstly, we used a fuzzy c-mean algorithm to identify and classify combinations of winds and discharge at the mouth of the Rhone river in order to define “model scenarios” of realistic forcings. Secondly, the consequences for sea surface currents will be assessed and the resulting plume pattern will be modelled for each scenario, as well as the distribution of dissolved radionuclides due to a hypothetical and episodic release on the Rhone River. These plumes scenario can be used as a support for operational tools improvement and decision.
Section snippets
Field study and data
The Rhone River hourly discharges have been provided by the C.N.R (Compagnie Nationale du Rhône) thanks to the Rhone Sediment Observatory (OSR program). They were measured at the SORA station, in the city of Arles located 47 km upstream of Rhone River mouth (Fig. 1). It must be noted that the Rhone River splits in two branches upstream of this station: the Grand Rhone and Petit Rhone. The station reports the discharge for the Grand Rhone River only, which represents about 90% of the total Rhone
Principal Component Analysis
PCA successfully reduced the five original variables (Wind speed toward North and East, Gust wind speed toward North and East and Rhone discharge) into three components and gave a summed variance of 96.4% (Fig. 1 supplementary material). This is not surprising since the gust wind speed and mean wind speed are correlated due to same direction (Fig. 2 supplementary material. The first axis contains 51% of variability with the information on wind direction. The second axis with 25% of variability
Conclusion
In this paper, a 10-year period was considered in order to identify the main combinations of hydrodynamic forcings (wind and Rhone River discharge) using a fuzzy c-mean clustering. These combinations, called scenario summarized mean shelf behavior providing a very important information to estimate and understand the Rhone plume patterns in case of accidental release. In addition, existence of observations memberships allowed to spot the best temporal windows to run simulations covering all
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The authors are indebted to the Institute for Radiological Protection and Nuclear Safety (IRSN) and to Region Sud (Provence-Alpes-Côte d’Azur) authorities for the PhD funding. This study was conducted within the Rhône Sediment Observatory (OSR) program, a multi-partner research program funded through Plan Rhône of the European Regional DevelopmentFund (ERDF), Agence de l’Eau Rhône Méditérranée Corse, CNR, EDF and three regional councils (Région Auvergne-Rhône-Alpes, PACA and Occitanie). The
References (67)
- et al.
Selecting model scenarios of real hydrodynamic forcings on mesotidal and macrotidal estuaries influenced by river discharges using K-means clustering
Environ. Model. Software
(2015) - et al.
Modelling of sediment transport pattern in the mouth of the Rhone delta: role of storm and flood events
Estuar. Coast Shelf Sci.
(2017) - et al.
A fuzzy extension of the silhouette width criterion for cluster analysis
Fuzzy Set Syst.
(2006) - et al.
Long-term variations of man-made radionuclide concentrations in a bio- indicator Mytilus gallopro incialis from the French Mediterranean coast
Sci. Total Environ.
(1999) - et al.
The ESCOMPTE program: an overview
Atmos. Res.
(2004) - et al.
A review on cesium desorption at the freshwater-seawater interface
J. Environ. Radioact.
(2020) - et al.
Implementation of an end-to-end model of the Gulf of Lions ecosystem (NW Mediterranean Sea). II. Investigating the effects of high trophic levels on nutrients and plankton dynamics and associated feedbacks
Ecol. Model.
(2019) - et al.
Development of emergency response tools for accidental radiological contamination of French coastal areas
J. Environ. Radioact.
(2016) - et al.
Particulate matter and organic carbon budgets for the Gulf of Lions (NW Mediterranean)
Oceanol. Acta
(2000) - et al.
The plume of the Rhone: numerical simulation and remote sensing
Continent. Shelf Res.
(1997)
Radionuclides in Waters and Suspended Sediments in the Rhone River (France) - Current Contents, Anthropic Pressures and Trajectories
Po River plume pattern variability investigated from model data
Continent. Shelf Res.
A toolbox for fuzzy clustering using the R programming language
Fuzzy Set Syst.
Investigating Rhône river plume (gulf of lions, France) dynamics using metrics analysis from the MERIS 300m ocean color archive (2002–2012)
Continent. Shelf Res.
The contribution of Mytilus sp. in radionuclide transfer between water column and sediments in the estuarine and delta systems of the Rhône river. Estuarine
Coastal and Shelf Science
Numerical modelling of the suspended particulate matter dynamics in a regulated river network
Sci. Total Environ.
River discharges of water and nutrients to the Mediterranean and Black Sea: major drivers for ecosystem changes during past and future decades?
Prog. Oceanogr.
Glider and satellite monitoring of the variability of the suspended particle distribution and size in the Rhône ROFI
Prog. Oceanogr.
A numerical study of the formation of the Rhone River plume
J. Mar. Syst.
Long-distance influence of the Rhône River plume on the marine benthic ecosystem: integrating descriptive ecology and predictive modelling
Sci. Total Environ.
The gulf of lions’ hydrodynamics
Continent. Shelf Res.
Circulation in the western Mediterranean sea
J. Mar. Syst.
Dynamics of the semi-diurnal and quarter-diurnal internal tides in the Bay of Biscay. Part 1: barotropic tides
Continent. Shelf Res.
Hydrology and circulation in a coastal area off Marseille: validation of a nested 3D model with observations
J. Mar. Syst.
A new cluster validity index for the fuzzy c-mean
Pattern Recogn. Lett.
Kd distributions in freshwater systems as a function of material type , mass- volume ratio , dissolved organic carbon and pH
Appl. Geochem.
On fuzzy cluster validity indices
Fuzzy Set Syst.
A numerical analysis of the summertime Pearl River plume from 1999 to 2010: dispersal patterns and intraseasonal variability
J. Mar. Syst.
Fuzzy Sets * Information and Control
Clustering gene expression patterns
J. Comput. Biol.
Pattern Recognition with Fuzzy Objective Function Algorithms
Fukushima daiichi–derived radionuclides in the ocean: transport, fate, and impacts
Annual Review of Marine Science
Cited by (5)
PEF-MODFLOW: A framework for preliminary soil profile horizon delineation based on soil color captured by smartphone images
2022, Environmental Modelling and SoftwareHydrological classification by clustering approach of time-integrated samples at the outlet of the Rhône River: Application to Δ<sup>14</sup>C-POC
2022, Water ResearchCitation Excerpt :However, many algorithms exist and present different results and convergence speeds (Jain et al., 1999). Fuzzy c-mean algorithms are usually considered as reliable algorithms in environmental sciences (Delaval et al., 2021, Kim et al., 2011). Their success is due to their simple geometric criteria and their robustness without any prior information on the cluster structure such as possible overlap or sphericity (Kamel and Selim, 1994).