EN :

With the aim of suggesting some practical rules for the use of hydrological models, G. De MARSILY in his "free opinion" (Rev. Sci. Eau 1994, 7(3): 219-234) proposes a classification of hydrologic models into two categories:

- models built on data (observable phenomena) and ;

- models without any available observations (unobservable phenomena).

He claims that for the former group of observable phenomena, models developed through a learning process as well as those based on the underlying physical laws are of the black box type. For the latter group of unobservable phenomena, he suggests that physically-based hydrologic models be developed.

Physically-based hydrologic models should introduce to the phenomenological laws the correct empirical coefficients, which correspond to the proper time and space scales (GANOULIS, 1986). Well-known examples are Darcy's permeability coefficient on the macroscopic scale as derived from the Navier-Stokes equations on the local scale and the macroscopic dispersion coefficients in comparison with the local Fickian diffusion coefficients. Misuse of these models by confusing the proper time and space scales and determining the coefficients by calibration is not a sufficient reason to consider them as belonging to the black box type. Black box type hydrologic models, although very useful when data are available, remain formally empirical. They fail to give correct answers when serious constraints of unity in place, time and action are not fulfilled.

Concerning the second class of models, we may notice that purely unobservable phenomena without any available data do not really exist in hydrology. In the case of very rare events and complex systems, such as radioactivity impacts and forecasting of changes on a large scale, physically-based models with adequate parameters may be used to integrate scarce information from experiments and expert opinions in a Bayesian probabilistic framework (APOSTOLAKIS, 1990).

The most important feature of hydrologic models capable of describing real hydrologic phenomena, is the possibility of handling imprecision and natural variabilities. Uncertainties may be seen in two categories: aleatory or noncognitive, and epistemic or cognitive. Probabilistic hydrologic models are more suitable for dealing with aleatory uncertainties. Fuzzy logic-based models may quantify epistemic uncertainties (GANOULIS et al., 1996). The stochastic and fuzzy modeling approaches are briefly explained in this free opinion as compared to the deterministic physically-based hydrologic modeling.

FR :

Afin de suggérer quelques règles pratiques sur l'utilisation des modèles hydrologiques, G. De MARSILY dans sa tribune libre (Rev. Sci. Eau 1994, 7 (3): 219-234) propose une classification des modèles hydrologiques en deux catégories :

- modèles construits sur des données (phénomènes observables)

- et modèles sans observations disponibles (phénomènes non-observables).

Il préconise que pour la première catégorie des phénomènes observables, les modèles développés selon une procédure d'apprentissage ainsi que ceux basés sur les lois physiques sont du type boîte noire. Pour le deuxième groupe de modèles, il propose l'utilisation de ceux à base physique.

Les modèles hydrologiques à base physique doivent introduire dans les lois phénoménologiques des coefficients empiriques corrects vis-à-vis des échelles correspondantes d'espace et de temps (GANOULIS, 1986). Des exemples bien connus sont le coefficient de la perméabilité de Darcy à l'échelle macroscopique, dérivé à partir des équations Navier-Stokes au niveau microscopique et les coefficients de dispersion macroscopique provenant des coefficients de diffusion locale de Fick.

Une utilisation abusive de ces modèles en introduisant des échelles de temps et d'espace inapropriées et en déterminant les coefficients par étalonnage, n'est pas une raison suffisante pour les considérer de type boîte noire. Les modèles hydrologiques du type boîte noire sont très utiles lorsque des données sont disponibles, mais restent essentiellement empiriques. Ils ne donnent pas de réponses correctes lorsque des restrictions sévères d'unité de lieu, de temps et d'action ne sont pas valables.

Concernant la deuxième classe des modèles, nous pouvons remarquer que dans la réalité des phénomènes purement non observables et sans aucune donnée, n'existent pas en hydrologie. Dans les cas d'évenements rares et de systèmes hydrologiques complexes tels que les impacts dus à la radioactivité et les prévisions à grande échelle, les modèles à base physique avec des paramètres adéquats peuvent être utilisés afin d'intégrer les informations rares et les opinions des experts, dans un cadre probabiliste Bayesien (APOSTOLAKIS, 1990).

La propriété la plus intéressante des modèles hydrologiques est leur capacité éventuelle à tenir compte des imprécisions et des variabilités naturelles. Les incertitudes peuvent être classées en deux catégories : aléatoires (liées à la variabilité naturelle) et épistémiques (liées à l'approche scientifique). Des modèles hydrologiques probabilistes sont plus convenables pour tenir compte des incertitudes aléatoires. Des modèles basés sur la logique floue peuvent quantifier les incertitudes épistémiques (GANOULIS et al., 1996). Les approches stochastiques et flous sont brièvement expliquées dans cette tribune libre en comparaison avec les modèles hydrologiques déterministes à base physique.

FR :

La méthode standard de détermination de la Demande Biochimique en Oxygène (DBO-5 jours par dilutions) présente certains inconvénients bien connus dont celui de ne pas être exploitable pour la gestion en temps réel d'un procédé d'épuration. De ce fait, divers types d'appareils d'estimation rapide de la DBO, basés sur des méthodes respirométriques, ont été imaginés dans le passé et même commercialisés pour certains.

Dans cette étude, des appareils d'estimation rapide de la DBO de types bioélectrode et réacteurs à biomasse fixée (écoulement piston et parfaitement mélangé) ont été mis en oeuvre, après avoir été développés ou modifiés sur base de concepts déjà existants. Une attention particulière a été portée sur la validité de ces appareils, la crédibilité de leurs mesures et la définition de leur champ d'action.

De par leur principe de fonctionnement (injection en flux), la bioélectrode et le bioréacteur piston sont des appareils qui doivent être calibrés et dont la biomasse immobilisée doit être préalablement adaptée à l'échantillon à analyser. La solution de calibration est essentielle et doit être qualitativement la plus proche possible de l'échantillon à analyser.

Le bioréacteur parfaitement mélangé se distingue des deux premières méthodes car il travaille théoriquement en consommation totale du substrat. La méthode ne nécessite pas de calibration mais implique, avant toute mesure de respirogramme, la détermination d'un paramètre supplémentaire, le coefficient de transfert de l'oxygène.

De manière générale, la principale conclusion de l'étude réside dans la difficulté de ces appareils à fournir une DO (Demande en oxygène) corrélable à la DBO5 conventionnelle. Leur champ d'application reste en effet limité aux substrats rapidement biodégradables.

Leurs potentialités restent cependant conformes aux besoins de contrôle rapide des charges organiques polluantes en entrée et sortie des stations d'épuration.

EN :

BOD (Biochemical Oxygen Demand) is an important parameter to characterize organic pollution in aquatic environments and sewage. The five-day period required by the classical dilution method (BOD5) is incompatible with real-time control of a sewage treatment plant. Moreover, the assay procedure (closed respirometer, very diluted samples) is not only far from real growing conditions but also far from conditions in sewage treatment plants.

Several devices for rapid BOD estimation, all based on respirometric methods, have been developed and tested. These devices can be grouped into 2 categories:

- an immobilized biomass plug-flow reactor and a bioelectrode, both based on a flow injection principle;

- an immobilized biomass perfectly mixed reactor, based on an open respirometer principle.

We have focused mainly on validating the principles, checking the measurement reliability, and defining more precisely the scope of the various devices.

The bioelectrode

The BOD bioelectrode that we developed relies on a classical configuration that uses a yeast strain (Trichosporon cutaneum) as the biological receptor and a Clark probe as the transducer.

The main changes made in this system are as follows:

- a second biomembraneless Clark probe was added to the 16-ml measuring cell to serve as a reference probe in order to remove experimental disturbances (temperature, oxygen transfer coefficient, dissolved oxygen concentration, etc.).

- the second change was to include the respirogram area among the data available for processing. The signal utilized in this set-up is the difference between the signals provided by the two probes.

Our conclusions are as follows:

- Concerning the signal processing, manufacturers are currently guided by the desire to develop devices able to estimate the BOD of a large range of substrates or effluents in a very short time (a few minutes for the newest devices). However, it seems useful, even necessary in many cases (complex mixtures of components that are oxidized at variable rates), to use information provided by the respirogram shape and area. This approach allows one to maximize the BOD bioelectrode's range for a given immobilized strain, although the trade-off is a longer total run time.

- Concerning the adaptation period for microorganisms, it is impossible to correctly estimate BOD from various effluents without first adapting the biomaterial to the type of substrate to be analyzed. Therefore it is dangerous to consider a BOD bioelectrode as an analytical instrument, because an adaptation period is required after any change in the composition or, even more so, type of effluent.

- Concerning the correspondence between rapid BOD and BOD5, the BOD sensor can detect only the BOD of soluble compounds that can diffuse through the biomembrane and that will be metabolized during the time of analysis. The difference observed between BOD sensor and BOD5 depends on the calibration solution but also and even more on the structure and size of the molecules constituting the sample to be analyzed.

- Concerning the choice of calibration solution, choosing the right calibration solution is crucial. The calibration solution should therefore be qualitatively as close as possible to the test sample.

The immobilized biomass bioreactors

The plug-flow reactor

The plug-flow reactor design was validated for simple substrates; its working principle is similar to the bioelectrodes, since it relies on flow-injection analysis (FIA). In the case of the plug-flow reactor, the only usable information for BOD estimation is the respirogram area, as the peak height quickly reaches a rather constant value due to saturation of the immobilized microorganisms. However, the importance of the many physical and biological processes that occur concomitantly in the system (transfer, adsorption, substrate consumption, substrate saturation phenomena, dilution rate, etc.) makes a theoretical mathematical model of the reactor more difficult to establish. A long-term, more fundamental study of various natural or artificial substrates might ultimately enable us to reach such a goal.

On the other hand, a variant of this reactor that recirculates the partially-degraded effluent until it is completely consumed yielded a linear correlation between system response (respirogram area) and substrate amount. In this system, oxidation of the rapidly biodegradable substrates is total under our operating conditions. This alternative reactor seems to have some very interesting possibilities, especially with regard to the automation of the system.

The perfectly mixed reactor

The utilization of the perfectly mixed reactor for rapid BOD estimation is based on two sequential experiments - although this has the disadvantage of increasing the total run time - to obtain the respirogram area (S) and oxygen transfer coefficient (KL.a), as the units of the product (KL.a x S) of these parameters are equivalent to those of oxygen demand and their product is the variable that best correlates with the substrate injection volume. This correlation was observed for a large range of substrates.

In the case of the perfectly mixed reactor, unlike bioelectrodes,

- diffusion processes have no effect on measurement, as the substrate is consumed completely during the experiment;

- the substrate consumption rate does not affect the measurements, thereby freeing the method from the influence of various experimental parameters (air flow-rate, the quantity of biomass, the liquid volume in the reactor, etc.), as our tests have shown.

For a given substrate, the linearity of the correlation between the product of KL.a x S and the amount of substrate is generally excellent. The attempt to correlate the system's response with the BOD5 measurements for a variety of substrates proved to be difficult, however. Although the bioreactor's analytical range is wider than that of bioelectrodes, the system still fails to give any measurable responses for complex substrates such as starch, cellulose, proteins, etc., and the KL.a x S values estimated by the bioreactor remained much lower than the corresponding BOD5 values. KL.a x S is probably characteristic of the portion of the substrate that is degraded rapidly by the microorganisms to meet their immediate energy needs, whereas the remaining substrate is probably kept for reserve and biomass synthesis.

FR :

L'implantation de douze pluviographes, de 220 à 2490 mètres d'altitude, dans des sites difficilement accessibles des versants sud et est du massif du Piton de la Fournaise à la Réunion, apporte de nouveaux éléments pour le tracé des isohyètes annuelles. Un pluviographe situé à 1600 mètres d'altitude, à l'amont d'un alignement orienté sud-nord dans la zone du Baril sur la planèze sud-est du Volcan, a enregistré une précipitation annuelle supérieure à 18 000 mm en 1993. Dans cette zone, les cartes d'isohyètes moyennes inter-annuelles admises jusqu'alors présentaient des lames d'eau de lordre de 5000 à 6000 mm.an-¹. Les enregistrements obtenus en 1994 puis en 1995 sur le même pluviographe et sur des stations situées à l'est du massif à 1400 et à 1940 mètres d'altitude confirment qu'une très forte pluviométrie affecte cette zone.

Une couche d'inversion thermique est fréquemment observée par ballon-sondage entre 2000 et 2500 mètres d'altitude, principalement en hiver austral. Cette inversion est l'une des causes principales de la présence d'une zone de maximum pluviométrique située entre 1400 et 1940 mètres d'altitude sur les planèzes du Volcan. Selon la position de l'inversion, la Réunion est soit dans la situation des îles très hautes, avec des sommets émergeant des nuages, soit dans la situation des îles hautes, pour lesquelles les précipitations sont fortes près des sommets. Les fortes pentes des versants du massif et la présence de grands encaissements contribuent également à expliquer l'emplacement de la zone de maximum pluviométrique et les lames d'eau importantes qui y sont enregistrées.

Une nouvelle carte des isohyètes annuelles est proposée pour prendre en compte les nouvelles données. Une zone de maximum pluviométrique définie par l'isohyète 12 000 mm de moyenne annuelle est localisée entre 1300 et 1800 mètres d'altitude, localement jusqu'à 2000 m, sur le versant oriental du massif exposé aux vents dominants. De part et d'autre de cette zone la pluviométrie diminue: les lames d'eau moyennes annuelles décroissent jusqu'à 4000 mm au niveau de la mer, elles représentent 7000 mm au sommet à 2632 mètres d'altitude. L'atténuation rapide des précipitations sur les versants nord-ouest et sud-ouest est liée au changement de l'orientation des versants par rapport aux flux générateurs de précipitations mais sans doute également à l'influence de grandes discontinuités topographiques orientées perpendiculairement aux vents dominants.

EN :

Reunion Island, located in the south-west Indian Ocean (21° S / 55° E), is composed of 2 volcanoes linked with a hot spot activity: the Piton des Neiges massif (3069 m) in the north-west and the Piton de la Fournaise massif (2632 m) in the south-east (Fig. 1). Climate is intertropical and characterized by two distinct periods: the hot rainfall season, from December to April, and the rather temperate dry season from May to November. South-east to east trade winds prevail during austral winter. However, polar atmospheric disturbances occasionally affect the island. Ascending wet air masses over a rugged topography with high summits and strong slopes release intense orographic rainfalls on the windward eastern slopes of the island. This phenomenon has been reported for other islands where rainfall was shown to be correlated with altitude (Rossignol, 1990; Oki et al., 1991). For trade winds and polar disturbances rainfalls, the maximum zone lies below the thermal inversion layer (Table 1). This layer, present 80% of the time, is found between an altitude of 2000 to 2500 metres in one third of the cases (Fig. 6 and Fig. 7). Cyclonic rainfalls, frequent during the austral summer, increase with altitude. Cyclones generate heavy rains on the eastern slopes of the Piton de la Fournaise massif while they approach the island from the north-east to east sectors (Malick and Mercusot, 1976; Fig. 12).

Available isohyet maps (Fig. 3) show a maximum rainfall zone at 860 metres, centred on les Hauts de Sainte-Rose rain gauge, on the north-eastern slope of the Piton de la Fournaise massif. There, the average rainfall is higher than 10 000 mm.year-1, and decreases in the west and south directions: at the same altitude in Baril, on the south side of the massif, annual rainfall totals were estimated between 5000 and 6000 mm (Anonyme, 1975; Bargeas et al., 1984; Robert, 1986).

In 1993, twelve tipping bucket rain gauges connected to data loggers or remote transmitters were installed between 220 and 2490 metres to precisely determine isohyets and define the presence of an altitude-dependent maximum rainfall zone (Fig. 2). On the south side of the massif, four rain gauges (altitude 650, 900, 1200 and 1600 m) form the north to south Baril transect along the main slope of the mountain.

During the first sampling year at Baril 1600, where water totals are maximum, the annual rainfall was higher than 18 000 millimetres (Table 2; Fig. 4). This constitutes the highest amount ever recorded on Reunion Island. A previous record was set at les Hauts de Sainte-Rose rain gauge (15 381 mm, from 1 August 1979 to 31 July 1980). Return periods over the 27 February 1993 - 26 February 1994 time interval calculated from Météo France reference rain gauge data show that the studied year was not exceptionally wet (Table 3 and Fig. 5). Since 1993, other measurements from the Baril transect, as well as from Bois Blanc and Enclos transects on the east side of the massif, suggest an under-evaluation of isohyet estimations, principally on the east and the south sides of la Fournaise volcano (Fig. 8, Fig. 9 , Fig. 10 and Fig. 11; Table 4).

A comparison between the climatologic and topographic environment of Reunion Island and other similar islands such as Hawaii, Maui, Kauai or Oahu in the Hawaiian archipelago (Bean et al., 1994; Ekern et al., 1971; Giambelluca et al., 1986; Giambelluca and Nullet, 1992; Giambelluca and Sanderson, 1993; Juvik and Nullet, 1994; Nullet and Juvik, 1994; Nullet et al., 1995; Peterson, 1972; Schroeder, 1993), Tenerife in the Canary islands (Custodio et al., 1991), Fogo in the Cape Verde islands (Juvik et al., 1995), or Guadeloupe in the Caribbean Sea (Robin and Rossignol, 1988) is presented (Table 5).

In Hawaii a distinction has been made between high and very high islands. High islands (such as Oahu, Kauai, east Maui, Guadeloupe or Gran Canaria) have an elevation lower or equal to the thermal inversion layer mean altitude. Water totals are very high and located near the summits. In very high islands (such as Hawaii, west Maui, Tenerife or Fogo), with summits often higher than the thermal inversion layer, rainfalls are more moderate and located lower on windward slopes.

La Fournaise volcano experiences these two situations alternately, depending on the presence, altitude and strength of the thermal inversion layer. When it is strong and low, during austral winter, rainfalls are restricted to low elevation zones. The frequent lack of thermal inversion during the hot season, or its high position when present, and the occurrence of tropical depressions cause heavy rains near the summits. These elements and the presence of great embankments and steep slopes, which are increasing relief effects, contribute to give to Reunion Island an original climatic and hydrologic environment. The maximum rainfall zone has a large extension in altitude, along the upper part of steep windward slopes. Recorded water totals and rainfall gradients count among the highest recorded values.

Subsequently, atmospheric and topographic elements defining spatial rainfall variations are discussed, and a new isohyet map is proposed (Fig. 13). A maximum rainfall zone over 12 000 mm.year-¹ is defined on the whole eastern side of the massif between 1300 and 1800 metres, and locally up to 2000 metres. Above and below this zone, rainfall decreases to 7000 mm.year-¹ at the summit of the Volcano, and to 4000 mm.year-¹ at sea level respectively. Isohyets are nearly parallel to contour lines on all sides of the Volcano except along the north-western and south-western slopes. On the north-western part of the massif, rainfall decreases to 4000 - 5000 mm.year-¹ since slope exposure to prevailing trade winds diminishes. South-western slopes rapidly become leeward, and rainfall amounts to less than 2000 mm.year-¹.

FR :

Les flottateurs à air dissous sont classiquement dimensionnés à partir de deux paramètres: le rapport mA/mS des masses d'air et de solides en présence et le temps de contact entre phases. Une série d'essais effectués sur cinq différentes unités continues ou discontinues montre que ces seules variables opératoires ne suffisent pas à déterminer l'efficacité. De plus, l'extrapolation des données obtenues sur un floculateur discontinu conduirait à des besoins en air dissous considérables pouvant limiter le développement industriel. La dissipation d'énergie, habituellement négligée, peut être quantifiée par le gradient de vitesse tel qu'il a été introduit en théorie de la floculation. De plus, une analogie entre la capture bulles-particules et le processus de floculation des particules primaires sur les flocs déjà formés permet d'étendre les équations de vitesse de la floculation et d'obtenir un modèle cinétique où interviennent seulement le gradient de vitesse et la concentration de particules; ce modèle remplace avantageusement l'approche classique qui considère la flottation comme un processus du premier ordre par rapport aux particules. L'efficacité d'une cellule discontinue ou d'un floculateur piston est alors fonction du seul nombre de Camp. Les résultats montrent l'existence d'un intervalle optimal pour le gradient de vitesse, 3000 à 4000 s-¹, et pour le nombre de Camp 105 à 106. Le modèle devra être amélioré par introduction de la tension critique de mouillage des particules.

EN :

Dissolved air flotation units are generally designed on the basis of two parameters: the mA/mS ratio of the air mass to the solid mass in reaction, and the contact time between the gas phase and the solid phase. The insufficiency of this approach, which neglects energy dissipation, is demonstrated.

Five units, the efficiencies of which were quantified by turbidimetry, were operated with a bentonite suspension previously flocculated with WAC or ferric chloride. Batch flotator 1 was a commercial unit designed to evaluate flotation feasibility (Fig.1). Flotators 2 and 3 were used to establish flotation efficiency as a function of the mA/mS ratio in continuous operation (Figs. 2 and 3). The influence of contact time was determined with batch flotator 4 (Fig. 4). Continuous flotators 3 and 5 were identical rectangular reactors but the latter was designed to allow the injection of pressurized water through five different points (Fig. 3).

Turbidity abatement increases as a function of mA/mS, reaching a plateau, the curve having a classical sigmoidal shape in batch or in continuous operation (Fig. 5). However the important air requirement (mA/mS=1) to attain 70% abatement would hamper industrial applications. The contact time is the residence time of the gas phase through a batch cell or the residence time of the solid phase through a continuous flotator. Its influence is displayed in Fig. 6 where a sigmoidal curve shows that a 100 second contact time is required to reach a significant abatement even with a low mA/mS of 0.1. However, flotator 3 operated with a 108 second contact time and 0.1 mA/mS ratio afforded only 40% abatement (Fig. 7). Efficiency is not therefore determined by the two classical parameters only but also by energy dissipation. The energetic conditions can be quantified by velocity gradient measurements, of classical use in flocculation; this parameter is 3100 s-¹ in flotator 3 and between 590 and 1670 s-¹ in flotator 4.

Flotation kinetics are classically considered first-order with respect to the particle concentration (Eqn. 3). In fact there is an analogy between flotation and flocculation which allows one to extend the well-known flocculation kinetics (Eqn. 4) to the flotation process (Eqn. 5). The steadiness of the bubble concentration permits the derivation of Eqn. 6, which enables one to calculate the efficiency of a batch or a plug-flow reactor as a function of the Camp number Gt (Eqn. 9). In fact there is an optimum range of velocity gradients between 3000 and 4000 s-¹ and an optimum range of Camp number between 105 and 106 (Fig. 9). The difference with the range currently observed in flocculation could be explained by the contact efficiencies in each process and by the probable existence of two ranges of optimal conditions. The model accuracy can be verified and the rate constant calculated (Figs. 8 and 10). This approach should be extended by testing particles exhibiting different degrees of hydrophobicity.

FR :

La stabilité biologique de l'eau potable dans un réseau est à l'heure actuelle un souci majeur pour tout responsable du traitement et de la distribution de l'eau. Les nouvelles filières de traitement introduisant l'ozonation ont induit la formation de molécules biodégradables. La filtration sur Charbon Actif en Grains (CAG) après l'étape d'ozonation a permis d'améliorer la qualité de l'eau. Toutefois, est apparue plus clairement la notion de risque de prolifération bactérienne sur réseau. Le Carbone Organique Dissous Biodégradable (CODB) représentant ces composés devient alors un paramètre indispensable à contrôler.

L'objectif de notre étude est de caractériser sur un réseau réel les facteurs influençant la stabilité biologique de l'eau dans les réseaux; notamment ceux limitant (le désinfectant bactériostatique) ou favorisant (le CODB) le développement bactérien.

Le réseau étudié, situé dans l'ouest de la France en Bretagne, totalise 50 km de conduites. Il est alimenté par une station traitant une eau de retenue selon une filière classique comportant une ozonation suivie d'une filtration sur CAG. Le temps de transit maximal de l'eau est proche de 8 jours. Nos résultats complètent et confirment ceux de CAPELLIER et al. (1992) instigateurs de notre étude, indiquant que ce réseau est le siège d'une vie biologique active.

La cinétique de trois critères de qualité: le désinfectant, le CODB et la flore bactérienne, a été étudiée en fonction du temps de transit, ce qui a permis de tirer les conclusions suivantes:

- la consommation du désinfectant est totale après 100 h de transit;

- particulièrement pendant les saisons les plus chaudes, le CODB est consommé après un pic observé en début de réseau;

- la flore bactérienne en suspension, revivifiable sur gélose et totale, croît pendant 100 à 150 h de transit, après quoi elle tend à se stabiliser, voire décroître.

Une relation entre le CODB et la flore bactérienne en suspension n'a pu être établie sur ce réseau pour les périodes étudiées. Par contre, les bactéries libres et le désinfectant résiduel semblent bien liés. Les influences de la saison et du temps de transit ont pu être démontrées. Néanmoins, d'autres paramètres de l'eau ou des facteurs comme la structure du réseau et l'hydraulique du système ont probablement une influence non négligeable sur l'évolution de la qualité de l'eau et donc sur nos résultats.

EN :

The biological stability of drinking water in a network is of major concern for every water supplier. New treatment processes induding ozonation have increased the formation of biodegradable molecules. Filtration on Granular Activated Carbon (GAC) after an ozonation step improves water quality. However, the concept of bacterial regrowth risk in water distribution networks is now being considered. The Biodegradable Dissolved Organic Carbon (BDOC) representative of these compounds becomes an essential parameter to monitor. The purpose of this study was to investigate, in an existing distribution system, the factors influencing this stability, such as disinfection which limits bacterial regrowth and presence of nutrients such as BDOC which promotes it.

The network studied, located in the west of France in Brittany, comprises 50 km of pipes and is supplied by a surface water originating from a conventional treatment plant including inter- ozonation followed by GAC filtration. As the source is a reservoir water, quality variations are seasonal. The characteristics of this kind of water are a low level of minerals (low alkalinity and hardness) and a high level of organics (high Total and Dissolved Organic Carbon; TOC, DOC). Treated and distributed waters have been studied and analysed during each season of the year, between 1993 and 1994. Sampling and measurements in the network have been carried out around distribution reservoirs. Several water quality parameters, including physico-chemical parameters (temperature, pH, turbidity, free chloride dioxide, organic matter as TOC, DOC and BDOC), and bacteriological parameters (viable and total suspended bacteria), have been measured. The residence time of water, known to have an influence on the evolution of water quality, has been estimated for each sampling point.

Our results confirm and complete those of CAPELLIER et al. (1992) who instigated our study. They indicate that an active biomass occurs in this network. The dynamics of three water quality parameters: residual disinfectant, BDOC and suspended bacterial density, were studied as a function of water residence time, for which the maximum value is around 8 days, in this distribution system.

This analysis yielded the following conclusions:

- consumption of disinfectant is complete aller 100 transit hours;

- particularly during warmer seasons, BDOC is consumed aller an initial peak observed at the beginning of the network;

- suspended bacterial density, as well as viable and totalbacterial numbers increase for 100 to 150 transit hours, and then tend to become stable and even to decrease.

No relationship between BDOC and suspended micro-organisms was observed in this network for the periods studied, whereas free disinfectant concentrations and free germs appeared to be linked.

However, the influence of the season and of the residence lime on the evolution of these parameters has been demonstrated. In spring and summer, organic marrer appears to be more biodegradable, because of the presence of BDOC in treated water, probably due to the limited retention of GAC subjected to higher levels of this parameter. This latter is distributed in the network, as DOC, and after a release in the first transit hours, is consumed. This peak of organic carbon is probably due to a release of biofilm because of the temperature increase and hydraulic variations. In the treated water, the density of the suspended viable bacteria is higher in spring than in winter, probably due to a higher contamination level in raw water in warmer seasons. However, the growth rate is higher in winter than in spring, thus the relation between residual disinfectant and viable bacteria permitted to show that bacterial inhibition was better in winter than in spring. No influence of the season was observed for the evolution of total bacteria.

Influences are numerous, sometimes in opposition and difficult to evaluate. This study show that BDOC was not a factor controlling the biological stability of water in our distribution network during the periods studied. The low level of this nutrient in the treated water is probably the reason, because even without it, our results showed bacterial growths within the network. Other factors such as seasonal variations, network structure and hydraulic conditions have certainly an influence on the evolution of water quality, and therefore on our results.

Other water parameters such as phosphorus or nitrogen could have an influence on the bacterial growth. The impact and evolution of biofilms is difficult to understand and monitor. Further investigations on all these points should be undertaken. They could permit to define physico-chemical, hydraulic and biological limits to be respected in order to maintain a biologically stable system.

FR :

Le tritium, isotope radioactif de l'hydrogène, est considéré comme un élément dateur des eaux souterraines. Il permet à l'hydrogéologue d'estimer le temps de séjour moyen des eaux d'un aquifère.

Produit naturellement dans l'atmosphère, les teneurs dans les précipitations ont tout d'abord augmenté, suite aux essais thermonucléaires aériens de 1952 à 1963, pour ensuite diminuer et se stabiliser des industries nucléaires, supérieure à la teneur naturelle de 5 UT. Actuellement, la teneur moyenne se situe entre 10 et 30 UT.

Grâce au suivi des teneurs en 3H des précipitations, il est possible de déterminer le temps de séjour moyen des eaux souterraines d'un aquifère à l'aide de modèles mathématiques qui, à partir d'un signal d'entrée (teneurs en 3H dans les précipitations) et d'un signal de sortie (teneurs en 3H dans les eaux souterraines d'un aquifère), calculent le meilleur ajustement possible entre les sorties calculées et les valeurs mesurées. De cet ajustement, on en déduit le temps de séjour moyen.

Ces études isotopiques utilisent classiquement trois types de modèles suivant la façon dont les eaux se mélangent en traversant un milieu poreux: le modèle mélange ou exponentiel, le modèle piston et le modèle dispersif. Cet ensemble de modèle peut être généralisé en différenciant temps de séjour moyen et age d'une molécule d'eau.

Ces modèles ont été utilisés pour étudier les systèmes des sources minérales de la Versoie et d'Evian (Haute-Savoie, France) en appliquant un modèle mélange à la source de la Versoie et un modèle dispersif à la source d'Evian, ce qui a permis d'estimer leur temps de séjour.

EN :

Tritium, the radioactive isotope of hydrogen, is considered as a dating element for ground-water and allows the hydrogeologist to evaluate the aquifer waters average transit time T. This parameter is important for questions about the evaluation, exploitation and protection of water resources.

Naturally produced in the atmosphere by the action of the neutronic component of cosmic radiation on nitrogen, 3H is present in precipitation at levels of about 5 UT. As a result of thermonuclear aerial tests from 1952 to 1980, the 3H content of precipitation increased markedly, reaching a maximum in 1963 of about 2500 UT and decreasing thereafter to about 50 UT in 1980. Since 1980, thermonuclear aerial tests have stopped but civil thermonuclear stations continue to influenced the 3H content of precipitation, which currently lies between 10 and 30 UT.

Thanks to the regular analysis of the tritium content of precipitation, it is possible to evaluate aquifer ground-water transit times, T, with mathematical models. This approach uses an "input signal" (3H contents in precipitation) and an "output signal" (3H contents in aquifer ground-water) and calculate the best fit between the calculated output and the measured values. This best fit yields an estimate of the transit time T.

According to the manner in which the waters mix as they pass through a porous media, three kinds of models are used: the well-mixed or exponential model, which supposes that water introduced into the system is completely and uniformly mixed with the aquifer water, such that the water output from the system is representative of the aquifer waters; the piston-flow model, which supposes that the incoming water traverses the aquifer with a constant velocity; and the dispersion model, which supposes a dispersion phenomenon due to the heterogeneity of the aquifer material. The piston-flow model and the exponential model represent the two possible extreme cases of this dispersion model: in effect, as dispersion increases the calculated output will tend towards the exponential model output. Similarly, for low dispersion systems, the calculated output will tend towards the piston-flow model output.

To generalise these models, a distinction can be introduced between the age of a water molecule, equivalent to the residence time, and the transit time. The residence time is the time elapsed since any element has entered into the system, as opposed to the transit time which is the time spent by an element between entry and outflow from the system. This distinction implies three cases:

- the residence time is smaller than the transit time, so depending on whether or not there is mixing, we have the dispersion or piston-flow model;

- the residence time is equal to the transit time, yielding the exponential model;

- the residence time is greater than the transit time, which is an unutilized case in classical modeling and which may be qualified as a short-circuit model.

These models have been applied to the Versoie and Evian mineral water systems (Haute-Savoie, France), for which we have long time-series of 3H measurements. The Evian mineral water system uses a dispersion model, for which the average transit time is between 40 and 80 years. The Versoie mineral water system uses a mixing model, for which the average transit time is between 2 and 9 years.

In conclusion, the long series of 3H measurements realized in France, Switzerland, Italy and Spain ground-waters from 1990 to 1993 allow us to distinguish four categories of results: 3H contents less than 2 UT, which characterizes an old ground-water with an average transit time greater than 2000 years; 3H contents between 2 and 10 UT, which characterizes a ground-water with an average transit time between 200 and 300 years, or a mixing between an old and a recent water; 3H contents between 10 and 40 UT, which characterizes a recent ground-water; 3H contents greater than 40 UT, which characterizes a groundwater near a nuclear station.

FR :

Pour limiter la formation de composés organohalogénés des eaux traitées et la reviviscence bactérienne des réseaux, il est important d'éliminer la majeure partie du carbone organique dissous (COD) et du carbone organique dissous biodégradable (CODB) contenus dans les eaux naturelles. Des travaux récents nous ont permis de montrer que la nanofiltration est une technologie de choix pour répondre à ces contraintes.

L'objectif de cet article est de présenter à partir de travaux de laboratoire un inventaire détaillé du carbone organique résiduel d'un perméat prélevé le 21/04/93 sur le prototype industriel de nanofiltration de Méry/Oise en banlieue parisienne. Pour atteindre cet objectif il a été nécessaire de mettre en œuvre des 'techniques analytiques impliquant l'utilisation de la chromatographie liquide haute performance (CLHP) et de la chromatographie en phase gazeuse (CG) soit équipée d'un détecteur à ionisation de flamme (FID) ou d'un détecteur à capture d'électrons (ECD), soit couplée à la spectrométrie de masse (SM).

Les résultats obtenus ont montré que le COD du perméat étudié est constitué d'environ 60% d'acides aminés libres et combinés, de 7% d'aldéhydes et de 10 à 20% de composés divers identifiables en CG/SM. Ces derniers composés comprennent majoritairement des acides gras aliphatiques et des acides aromatiques de faibles masses. La concentration de chacun de ces composés a été estimée à 0,3 µg l-¹ C. On peut raisonnablement penser, d'après la bibliographie que les hydrates de carbone (non analysés dans cette étude) représenteraient une part importante de COD du perméat. En outre, cette étude a montré que la part prise par les acides aminés totaux dans le CODB du perméat est importante.

Seul le tiers des potentiels de formation d'organohalogénés totaux (PFTOX) a été identifié comme étant des trihaloméhanes (THM) et des acides haloascétiques. Toutefois, étant donné que les acides aminés totaux représentent à eux seuls la quasi totalité de la demande en chlore du perméat, les autres sous-produits de chloration non identifiés seraient probablement des nitriles chlorés, des chloramines et des chloroaldéhydes qui sont parmi les principaux intermédiaires réactionnels de la coloration des acides aminés.

EN :

Removal of dissolved organic carbon (DOC) and biodegradable dissolved organic carbon (BDOC) is one of the most important means to prevent disinfection by-products (DBPs) formation during water treatment and bacterial regrowth in distribution systems. In previous investigations, the authors have shown that nanofiltration, over nine months of operation at industrial scale in Paris suburbs, was an effective technology to meet the new guidelines concerning chlorine DBPs.

This paper reports laboratory investigations aimed to identify and quantify the main organic components included in the low DOC, BDOC and TOXFP (Total - Organohalides Formation Potential) residuals of a nanofiltration permeate sampled on April 21, 1993.

Details on DOC, BDOC, organohalides, amino acids and aldehydes analysis procedures were described elsewhere (AGBEKODO et al., 1994). Chlorination was undertaken in potential conditions k: 2.5 mg Cl2/mg DOC, pH=7.5 (phosphate buffer), 72 hours contact time, in dark at 20°C. Haloacetic acids determination consisted (after chlorination in potential conditions) in liquid-liquid extraction, methylation with diazomethane and gas chromatography analysis. Extraction procedure based on XAD8/XAD4 adsorption prior to gas chromatography/Mass Spectrometry (GC/MS) analysis (Fig. 3), allowed a 19000 fold concentration of the permeate. To prevent possible contamination of the permeate, the extraction system consisted of four glass columns and teflon materials. The flow through the columns was performed under high purity nitrogen gas pressure.

Analysis involving high performance liquid chromatography (HPLC) and gas chromatography (GC) in combination with Mass Spectrometry (MS), showed that the studied permeate DOC (- 0.15 mg l-¹ c) consisted of amino acids at an average of 60% of DOC, aldehydes (7%) and 10 to 20% of several other compounds (analyzed in GC/MS) including primarily fatty and aromatic acids of low molecular weight (Table 4a and Table 4b). The maximum concentration of each compound (identified hy GC/MS) has been roughly assessed to 0.3 µg l-¹ C. According to literature, sugars represent probably an important portion of the remaining DOC of the permeate. Moreover, the authors have shown that amino acids represented a large portion of permeate BDOC.

Only 34% of the total organohalide potentials were identified as trihalomethanes and haloacetic acids. However, since amino acids represent almost the entire chlorine demand of the permeate, the non- identified chlorination DBPs are likely chlorinated nitriles, chloramines and chloraldehydes which are known as the main reactionnal intermediates of aminoacid chlorination.