Volume 5, numéro 4, 1992

FR :

Afin d'étudier la reviviscence bactérienne dans les réseaux de distribution, des méthodes de mesures de la biomasse et de l'activité bactérienne ont été investiguées sur des eaux provenant d'un réseau de distribution. Trois méthodes d'estimation de la biomasse bactérienne ont été comparées : le comptage sur gélose, selon la norme française d'examen bactériologique des eaux de consommation, le dosage de l'ADN contenu dans les particules retenues sur une membrane de porosité de 0.2 µm et le comptage direct au microscope à épifluorescence après coloration des bactéries à l'acridine orange. Les comptages sur gélose, tout comme en milieu aquatique naturel, sous-estiment très largement le nombre de bactéries; ceci semble principalement lié à la présence de bactéries viables mais non cultivables. Le dosage de l'ADN et les comptages directs corrèlent assez bien avec en moyenne un contenu en ADN par bactérie de 4,1 x 10^-15 g d'ADN, mais la première méthode semble moins précise. Le comptage direct semble donc la méthode la plus adaptée à l'estimation du nombre total de bactéries dans ce type de milieu.

Afin d'estimer l'activité bactérienne, les protocoles expérimentaux de deux méthodes utilisées en écologie bactérienne ont été adaptés aux conditions particulières de l'eau de distribution : l'incorporation de thymidine tritiée dans l'ADN bactérien et l'incorporation de leucine tritiée dans les protéines. La comparaison des deux méthodes sur une série d'échantillons montre une bonne corrélation, avec un rapport molaire entre incorporation de leucine et de thymidine compatible avec les facteurs de conversion des deux méthodes cités dans la littérature et établis pour les milieux aquatiques naturels. Les deux méthodes sont utilisables pour mesurer l'activité bactérienne dans l'eau potable, néanmoins l'incorporation de thymidine est plus aisée à mettre en oeuvre, car elle ne nécessite de travailler qu'à une seule concentration en traceur radioactif.

EN :

Bacterial regrowth in distribution systems is an important problem for drinking water producers. It is linked to the more and more frequent utilization of low quality surface waters, containing high concentration of organic matter, as raw water, and also to the increase in size and complexity of the distribution networks with high residence time of the water between its production and utilization. At the present time chlorination of treated water, with sometimes rechlorination in the network, is the usual way to limit growth in distribution systems. This solution however presents disadvantages, the major one is the formation of unpleasant organochlorine compounds which are responsible for tastes and odours of water. An alternative strategy consists of developing treatment lines in which biodegradable dissolved organic carbon is removed. It allows through a reduction of the chlorine demand of the water to increase the stability of the chlorine residual of the water. In this context, it is important to get a good knowledge of the factors controlling bacterial development in distribution networks. Up to now, studies on this subject have met some methodological problems linked to the fact that classical bacteriological methods are inadequate to study this kind of systems.

In this paper, various methods have been investigated to estimate bacterial biomass and activity in tap water. For this study, the analyzed water samples have been collected in the distribution system of the Parisian suburbs.

Three methods have been tested for the determination of bacterial biomass : plate count, measurement of DNA associated with particles with a size higher than 0.2 µm and direct microscopic enumeration. Heterotrophic plate counts have been performed following the French standard and results are expressed in CFU (Colony Forming Units) per ml; the DNA collected after filtration of 500 ml to 1500 ml of water on a 0.2 µm pore size membrane was estimated using a fluorimetric method, as proposed by Mc COY and OLSON (1985); direct enumerations were performed by epifluorescence microscopy after acridine orange staining (AODC) following the procedure proposed by HOBBIE et al. (1977), the comparison between plate counts and AODC (fig. 1) shows the important underestimation of the bacterial numbers when estimated by the CFU (up to 3 orders of magnitude). Such discrepancy has already been observed in natural aquatic ecosystems and is usually explained by the presence of numerous dead cells enumerated by microscopy. Now, it seems that the difference between, plate counts and direct counts may rather be explained by the presence in water of « viable but non culturable » bacteria.

A comparison between DNA estimation and direct counts have also been performed. Figure 2 shows the results of this comparison. In spite of the dispersion, the correlation between both methods is significant and the correlation straight line indicates an average DNA content per bacteria of 4.1 x 10-15g DNA in good accordance with the values quoted in the literature. The dispersion of the data around this average can be explained by various ways : the variability of DNA per cell content for the different bacterial strains present in the water samples, the precision of the DNA method which is not higher than 20 % and possible contamination by other organisms than bacteria, as flagellates or ciliates, which are retained on the 0.2 µm pore size membrane.

On the basis of these tests, it seems that the direct count by epifluorescence microscopy is the most adapted method for studying the bacterial regrowth in distribution system.

The understanding of bacterial dynamics in a distribution system requires measurements of bacterial activity. Various methods have been developed in order to estimate bacterial activity in natural aquatic ecosystems. They are primarily based on the use of radioactive tracers. At the present time, the tritiated thymidine incorporation method, which measures the replication of bacterial DNA, is the most usually used one, but the incorporation of tritiated leucine into proteins, which measures increase in bacterial biomass, seems to be also an interesting method. These methods have been selected, on one hand, because of their specificity towards bacteria and, on the other hand, because of their high sensibility which is required for measurements of bacterial activity in the conditions of drinking water. Up to now, these methods have never been applied to drinking water. We have modified the experimental procedure of both methods : incubation time, radioactive tracers concentrations and volume of the sample have been tested and adapted in order to allow measurement in the conditions of drinking water samples. For thymidine incorporation, the volume of sample, incubated during 20 hours in the presence of 20 nM concentration of ³H-thymidine, was 100 ml. The incorporation was measured in the DNA, using the biochemical procedure proposed by WICKS and ROBARTS (1977), rather than in the total macromolecules. For leucine incorporation, we measured the incorporation rate at four leucine concentrations (2, 27, 52, 77 nM : 2 nM of ³H-leucine + non radioactive leucine) in 25 ml samples and the incubation lasted 3 to 4 hours. The incorporation rate was calculated as the reciprocal of the angular coefficient of the correlation straight fine obtained when the reciprocal of the fraction of leucine incorporated per hour was plotted against leucine concentration (fig. 3). Comparison of both methods on samples of drinking is presented at figure 4, a good linear correlation was found. The equation of the correlation straigth line is :

log [Inc.leu (pmol/l.h)] = 0,97 log [Inc.thy (pmol/l.h)] + 1.35(n = 69, r = 0.84)

The molar ratio between leucine and thymidine incorporation found in these samples (20 to 25) seems to be in good agreement with the usual conversion factors found for both methods in natural aquatic ecosystems. Bath methods seem to be available to bacterial activity estimations in drinking water, the triatiated thymidine incorporation method which requires working with only one concentration of radioactive tracer seems easier to use.

FR :

La toxicité du phytoplancton est un problème dont l'importance est grandissante en France comme en Europe. Cette toxicité se manifeste surtout lors de l'ingestion de cyanobactéries formant des fleurs d'eau superficielles liées à l'eutrophisation. Microcystis aeruginosa est l'espèce la plus fréquemment incriminée, mais 75 % des souches de cyanobactéries d'eau douce seraient des toxiques potentielles. Lorsque des clones sont isolés d'un plan d'eau dans lequel des manifestations de toxicité ont été observées, des souches toxiques et non toxiques sont obtenues. La connaissance des conditions d'expression de la toxicité représente un sujet important actuellement peu étudié. Un mammifère peut mourir s'il passe dans son sang 0,07 mg de toxine de Microcystis par kg de son poids. En pratique, de nombreux cas de morts de bétail ont été recensés. Les toxines qui causent des accidents sont des endotoxines non rejetées par les cellules vivantes, mais libéré au cours de leur lyse. Ceci explique que des cas d'intoxication humaine par l'intermédiaire de l'eau de boisson ont pu être observés. On distingue principalement : des microcystines, hépatotoxines produites par diverses cyanobactéries dont Microcystis et Oscillatoria, et des anatoxines, neurotoxines produites par des Anabaena. Certaines cyanobactéries produiraient un mélange des deux formes. En sus du test de toxicité classique sur la souris, plusieurs autres tests existent. L'analyse est généralement réalisée par chromatographie liquide à haute pression. Des produits actifs synthétisés par des cyanobactéries possèdent une fonction antibiotique et sont susceptibles de jouer un rôle dans le comportement des espèces et leur dominance ou de les protéger contre le broutage. Ces produits seraient des exotoxines différentes des précédentes.

EN :

The aim of the paper is to present in the French language the international knowledge related to freshwater cyanobacterial toxins, a problem of great significance in our European countries but largely unknown by people in France. An analytical review of a selection of works chosen from the extensive existing literature is presented. At the present time, the mains works come from : U.S.A., with CARMICHAEL and coworkers, continuing the researchs from CORHAM; Scotland, with CODD; Scandinavia with ERIKSSON, LINDHOLM; and Japan with WATANABE, HARADA, but also many others scientists.

In freshwater, poisoning is typically associated with the ingestion of the cyanobacteria appearing in large amount, at the surface of some water bodies, and called water blooms. Many cases of livestock death (concerning sheeps, calves but also adult oxen and horses), associated with the consumption of such water blooms are reported, also, deaths of wild mammals (muskrats, and hogs), birds (ducks, geese), fishes, invertebrate, and human illness following bathing are known. But, because toxins are not destroyed by conventional sand filtration treatment, human illness may also arise from drinking water taken out from an impoundment with a cyanobacterial bloom. Cases are known from United States, Scandinavia and Sardinia. This tap water problem is serious because probabilities of long term diseases, such as tumors promotion, are now considered high.

Seventy-five percent of fresh water cyanobacterial strains are potentially toxic, but, on the whole, only some clones from a single species, simultaneously isolated out of an unique body of water, are toxic. However, there are no evidence that the nontoxic strains could never become toxic.

Cyanobacteria are also known as blue-green algae, Myxophyceae or Cyanophyta, and are typically microscopic prokaryotes, but with chlorophyll a. Toxic clones belong to : a) the Chroococcales, single coccoid cells embedded in a gelatinous matrix, represented by species of the genera Coelosphaerium, Gomphosphaeria, and Microcystis whose the species M. aeruginosa is the most frequently quotted toxic Cyanobacteria. b) the Nostocales, filamentous forms, some of them with exocellular sheath. Many are nitrogen fixing species belonging to the genera Nodularia, Anabaena, Aphanizomenon and Nostoc, others belong to the genera Oscillatoria, and are sot known as nitrogen fixers.

Most often the mice toxicity test is used to identify toxic water blooms, it allows to define the lethal doses of the toxins or of the toxic organisms. But some others tests have been applied or suggested, they use others animals : fishes, zooplankton or microorganisms, and isolated organs or cells cultures, many are not specific. Currently the modern immunological tests are not yet adapted to identify fresh-water cyanobacterial toxins.

The main toxins that can be distinguished are the microcystins and the anatoxins. The microcystins are hepatotoxins from various cyanobacteria belonging principally to the genera Microcystis and Oscilatoria; they promote liver haemorrhages. The anatoxins are neurotoxins from Anabaena and death occur by breath arrest. Some cyanobacteria simultaneously produce the two forms. A mammal may be killed by a blood level content of 0.07 mg of Microcystis toxin per kg of body weight. Cyanobacterial toxins are endotoxins which diffuse during cell lysis. This explain why toxins can be found in water from impoundments with cyanobacterial blooms. In this case, the toxins can possibly originate in a thick metalimnic plankton layer, not seen from surface.

Toxin analysis is usually performed using high pressure liquid chromatography but also thin layer chromatography, and particularly the high performance modem technique. Molecular structures are eluciated using, fast atoms bombardment spectrometry, mass spectrometry and nuclear magnetic resonance. The microcystins are cyclic heptapeptides of low molecular weight, they differed in amino-acids composition; a characteristic one, built up with 20 carbons, is called ADDA. Microcystins toxicity ensue as they act as strong inhibitors upon phosphatase activities. The anatoxins-a are alkaloids, others anatoxins are peptides or currently unknown.

The causes of the expression of the toxicity remain to be elucidated. In the years to come, much progress can be achieved by using new genetic tools. Nevertheless, as the largest problems occur always associated with water-blooms, and rise under high sunlight in hot periods, toxicity appears in the whole, associated with eutrophication, and as many toxic species are nitrogen fixers which do not need inorganic nitrogen to grow, problems follow generally a plentiful phosphorous load of water bodies due to human operations. As toxins accumulate in the cells, they could be actives either only after cells consumption, or after toxins discharge during cell lysis. Cyanobacterial toxicity is not due to bacteria associated with the cyanobacteria and can appear in pure axeniccultures. Toxicity is not associated with the presence of cell plasmids. It was shows that optimal conditions for growth did sot coincide with those for toxin production and vary with the growth phase. On the whole, the optimum temperatures, for toxin maximum production, were at about 25 °C for different cultures. Light intensity would be the primary important factor for the production of the toxin, but, in cyanobacteria cultures, this production can occur at relatively low light intensity.

Some related cyanobacterial products are not true toxins but are exotoxins acting as antibiotics and can affect species behaviour or dominance and help deter grazers. Since LEFEVRE and coworkers studies, and after a long quiescent period, ecologists take these products anew into account when studying plankton ecology and successions. Only some phytoplankton responds to cyanobacterial extracellular products. Among zooplankton there are species avoiding actively the cyanobacteria and insensitive ones. Chemists have already started search for antitoxins chemicals and found promising curative results. On the other band, biotechnology could take advantage of all these various cyanobacterial products to obtain new drags having pharmacological or agronomical uses. Some true toxins could be used as anti-neoplastics, and products, involved in allelopathic reactions, have antibiotic and antivirus activities. The use of toxins as commercial algicide for chlorophycean waterbloom control had been suggested, but the action spectra of the toxins must be precisely known before extensive implementation. From another point of view, « microalgal » by-products, used as food additives have to be carefully checked for possible toxins.

As SKULBERG et al. could rite in 1984, toxic blue-green algal blooms is a growing problem in Europe. Scientists involved in health supervision had to be watchful to it in a way to prevent people from possible major accidents.

FR :

Le peuplement ichtyologique d'un marais littoral endigué, géré en eau douce, a été étudié sur une zone de 2 700 ha, située au nord du marais Breton-Vendéen (Loire-Atlantique, France). Le réseau hydraulique, qui représente près de 16 % de cette surface, se compose, d'environ 91 m de linéaire de fossés par ha (en tout 234 km) et de bassins présents uniquement dans la partie d'origine salicole du marais. Les hauteurs d'eau, l'envasement et le recouvrement par la végétation aquatique dépendent de la gestion humaine et sont très variables (moyennes respectives : 42 cm; 43 cm; 70 %). Cela se traduit par un morcellement spatial de l'habitat pour les poissons. La stratégie d'échantillonnage adoptée, qui tient compte de cette hétérogénéité, a permis de décrire un peuplement comportant 21 espèces. Dominé par les poissons-chats et par les anguilles, ce dernier est caractéristique de la zone à brèmes des cours d'eau. Les abondances sont relativement élevées (en moyenne 315 kg/ha et 11 460 poissons/ha), mais elles sont très hétérogènes. L'évolution qualitative et quantitative de la répartition spatio-temporelle est décrite à l'aide d'une analyse factorielle des correspondances portant sur 74 échantillons prélevés par pêche électrique entre 1987 et 1989. Bien que l'approche de ce milieu soit complexe et les références bibliographiques relativement rares, l'analyse des premières données permet d'ores et déjà d'identifier quelques problèmes de gestion ayant des répercussions directes sur le peuplement piscicole de cette zone.

EN :

The dammed up marshes of the French Atlantic coast cover about 200 000 ha between River Vilaine and the « Bassin d'Arcachon ». Eighty eight % are managed with freshwater. They constitute original environments initially created for agriculture or for salt production, and they are now threatened by land abandonment within the next decade. Concurrently to aquaculture (in created or existing ponds), exploitation of the fish stocks in the ditchweb is likely to encourage a diversification of agricultural activities. Unfortunately, bibliographic analysis reveals the relative scarceness of research about sampling methods and qualitative or quantitative characteristics of these fish communities. This is quite surprising considering the importance of the ditchwed of this kind of environment outlined by several authors. In the Netherlands, BELTMANN (1984) assessed that there is a total of 400 000 km of ditches. In France, the littoral dyked marshes of the Atlantic coast couid comprise 20 000 km of ditches and about 24 000 ha of open water. The present work provides for the first data on the fish community of Bourgneuf marsh.

The northern part of the marsh of Bourgneuf, 2 700 ha provided whith f reshwater, contains nearly every kind of landscapes found throughout the whole Breton-Vendéen marsh. The pattern of the ditch network strongly changes from a zone to another (fig.1) : presence of former salt pans in the western part, regular geometric shapes in the recently created polders next to the River Falleron, irregular ditchweb pattern in the eastern part. The average density of the ditch network is 91 m of ditches per ha, totalizing 234 km in the study area. The total surface of open water, composed of ditches and basins (former salt pans), covers 411 ha (over 15 % of the study site). Diversity of ditch types occurs at fine scales (<1 000 m²), they vary according to their widths (0,3 to 7 m), depths (average, 42 cm; SD, 20,4), thickness of silt layer (average, 43 cm; SD, 42) and their hydrophyte vegetation cover (average, 70 %; SD, 60 %). As a consequence of this heterogeneity, available habitats are scattered over the marsh (mosaïc distribution). A nested sampling (FRONTIER, 1983) was carried out to take into account this high heterogeneity : 5 sampling areas were selected randomly. In each one, 3 to 5 ditches were chosen according to their characteristics (see above). Sampling stations were delimited by 2 stop nets (5 mm mesh) settled 30 m apart, in order to avoid fish migration. Field work was conducted using « Heron » electric fishing material (see LAMARQUE et al., 1978). In each ditch-section, we carried out as many successive catches as necessary to apply the maximum likelihood weighted estimation method of CARLE and STRUB (1978). Nine to 19 stations were sampled at 5 periods, between 1987 and 1989. A total of 74 samples were collected.

The fish community was composed of 21 species (table 1) and corresponded to the bream zone of Verneaux's classification (1977). The densities and biomass were quite high (on average 315 kg/ha and 11 460 fishes/ha) but very variable (0 to 2 120 kg/ ha and 0 to 39 300 fishes/ha). The catfish, Ictalurus nebulosus (170 kg/ha), the eel, Anguilla anguilla (47 kg/ha) and the tench, Tinca tinca (28 kg/ha), represented on average 77,5 % of the standing crop, but their spatial distribution was very irregular. These estimates are assumed to be reliable considering that the data used for the calculations were provided by a sampling design which permits to respect the basic assumptions of the removal method. (f) The population size could only change because of the fishings (no migration because of the stop nets; no recruitment/death because of the short duration of the fishing sequences). (ii) The standard sampling design permitted to reduce the variations of the catch probabilities between the successive removals. Several studies have shown that this removal method under-estimates by about 20 % the true size of the fish populations (e. g. BOHLIN and SUNDSTROM, 1977; MAHON, 1980). But they were based on Zippin's method, and the estimator of CARLE and STRUB (1978), that we used, was shown to be more robust (COWX, 1983; GERDEAUX, 1987). Nevertheless, we assume that the values presented in this paper provide for an approached information on the sizes of the studied fish populations.

To assess the fish-habitat relationship, a correspondence analysis (fig. 2) was performed on the 74 samples X 17 species matrix (excluding the sticklebacks, Gaslerosteus aculeatus and Pungitius pungitius, which population size estimations failed because of their low catchabilities). Four groups of samples were ordinated according to their specific richness and the species they contained. Several habitat parameters were projected on F1-F2 factorial map (fig. 3). Hydrophyte cover, thickness of silt layer, water depth (fig. 3 and 4), which are directly controlled by human maintenance, appeared to be the major structuring habitat parameters for the fish community. In the deepest and less silted stations, the communities were rich (on average 11 species; group 4, fig. 2). Predators such as pike-perch, Stizostedion lucioperca, and perch, Perca fluviatilis, occurred, and cold water species were found, such as minnow, Phoxinus phoxinus, or chub, Leuciscus cephalus. When the silt layer was thicker and the water level was intermediate, the specific richness decreased (average, 6,2 species) and the community was either dominated by the cattish (group 2, fig. 2) or by the rudd, Scardinius erythrophtalmus (group 3, fig. 2), according to the importance of the aquatic vegetation cover. Habitats with thickest silt layers, shallowest waters and maximum aquatic vegetation cover contained the poorest communities (average 3,9 species) dominated by eel (group 1, fig. 2). There is also evidence that the diversity of the community has progressively decreased since 1987 (fig. 5). The most stenothermous species disappeared, and the importance of the catfish increased : it doubled between May 1987 and September 1989 (fig. 6). Although the eel is the species most adapted to this environment, we emphasize the diminution of its biomass (fig. 7). These phenomena could be partly due to the climate (cold winter in 1987, important swelling in January 1988 and 2 droughts in summers 1988 and 1989). But they are mainly caused by the water management policy which is intended to favour agriculture by keeping stable water levels (evacuation of swellings) and by preventing the freshwater part from the marine influence (collective sluice gates). This does mot permit an optimal breeding of the species that have to spawn on flooded meadows, neither a proper colonisation of the marsh by elvers.

FR :

Au cours de ces dernières années, beaucoup d'efforts ont été consacrés au développement de formules de probabilité empirique (FPE) non biaisées. En raison même de leur définition, les FPE non biaisées sont dépendantes de la distribution parente des échantillons considérés, et une formule doit donc être établie pour chaque distribution. Dans cette étude on passe en revue les différentes approches pour le développement des FPE, et on montre que la FPE basée sur la médiane des statistiques d'ordre peut constituer un compromis acceptable entre les FPE non biaisées (i.e. correspondant à la moyenne des statistiques d'ordre), et les FPE basées sur le mode des statistiques d'ordre.

Contrairement à ces dernières, la FPE basée sur la médiane des statistiques d'ordre est indépendante de la distribution parente des échantillons, et peut donc être utilisée de façon standard. Par ailleurs, la FPE basée sur la médiane des statistiques d'ordre est moins biaisée que la FPE de Weibull, qui est également indépendante de la distribution parente des échantillons. Bien qu'il n'existe pas d'expression analytique exacte pour la FPE basée sur la médiane des statistiques d'ordre, BENARD et BOS-LEVENBACH en ont proposé une très bonne approximation, pm= (m - 0.3)/(n + 0.4). Cette formule est aussi connue sous le nom de formule de Chegodayev.

EN :

Plotting position formulae (PPFs) maintain an important role in engineering practice. In the area of flood frequency analysis they are used to assign exceedance probabilities to observed floods. In the present study we review various principles for the choice of PPFs. These can be divided into three main categories : (1) formulae based on the observed sample frequencies, (2) formulae based on the distribution of sample frequencies, and (3) formulae based on the distribution of order statistics. PPFs in the first two categories are distribution-free, meaning that no assumption needs to be made regarding the form of the parent distribution of events. The Hazen PPF is an example of a formula belonging to the first category.

The Weibull PPF, which is probably the most used formula in practice, belongs to the second category. It can be shown that the frequency corresponding to a particular order statistic is beta distributed regardless of the form of the parent distribution. Being equivalent to the expected value in the beta distribution the Weibull plotting position, pm = m/(n + 1), therefore corresponds to the mean value of sample frequencies. In his book on statistical extremes GUMBEL (1958) recommended the Weibull formula, because it fulfils a set of criteria which he found important. Some of these criteria have later been questioned, for instance by CUNNANE (1978). Various studies have demonstrated that the Weibull PPF is significantly biased in the event domain for most common distributions (an exception is the uniform distribution, where the Weibull PPF is the exact unbiased plotting position).

In recent years most attention has been paid to PPFs of the third category. CUNNANE (1978) strongly recommended the use of unbiased PPFs, i.e. formulae for the exceedance probabilities of the expected values of order statistics. In the last decade much effort has been devoted to the development of unbiased plotting positions. As unbiased PPs in virtue of their definition are related to the parent distribution, they will differ for each individual distribution. In general, it is not possible to derive exact unbiased PPFs, but good approximations have been developed for most common distributions such as the normal, the Gumbel, the generalized extreme value, and the Pearson type III distributions (see table 1). If a distribution contains a shape parameter, this must be reflected in its unbiased PPF. Approximate formulae for such distributions are therefore in general of a more complex form. This fact along with the need for distribution-dependent formulae is probably the reason why the Weibull formula is still the most used PPF in practice.

In the present study we emphasize the practical convenience of having a distribution-free PPF which at the same time has a statistical interpretation and is related to the distribution of order statistics. The median PP fulfils these points. It is easily seen to be distribution-free by observing that the median in the distribution of order statistics corresponds to the median in the distribution of frequencies (beta distribution). In general, one of two different principles is commonly adopted when developing estimators : either the choice of the modal value (maximum likelihood principle) or the choice of the mean value (principle of unbiasedness). Unfortunately these distinct principles usually lead to different estimators. For continuous distributions used in flood frequency, the median of order statistics is located in between the modal value and the mean value. In general, it is much less biased thon the Weibull PPF, which for the EV1 distribution is a reasonable approximation to the modal PP. Although no analytical expression exists for the median plotting position, an approximation has been deduced by BENARO and BOS-LEVENBACH (1953), namely pm = (m - 0.3)/(n + 0.4). This formula is also known as the Chegodayev PPF.

Various PPFs are exemplified in the case of three different parent distributions, namely the normal, the log-normal, and the Gumbel distributions. The sample size n = 10 is considered. The results of applying different PPFs are presented in table 2 (for the largest order statistic in the sample). Several observations can be made : 1) The Benard and Bos-Levenbach-formula is a good approximation to the median PP; 2) The Weibull-formula is close to the modal value PP when the parent is EV1, but differs significantly in the case of other parents; 3) The unbiased PPs depend strongly on the underlying distribution, and an unbiased PPF suitable for all distributions can therefore not be found; 4) The median PP is a fair compromise between the mean and the modal values of the order statistics in all three cases, and it is therefore recommended as a good choice of a standard PPF.

FR :

Dans les baies presque fermées de la Méditerranée, le taux de renouvellement et le pouvoir dispersif du milieu marin sont faibles. Ceci amène à des concentrations souvent élevées de bactéries califormes à la sortie des stations d'épuration, même munies de filières biologiques. L'élevage de coquillages dans ces baies constitue une contrainte supplémentaire aux concentrations maximales de bactéries permises selon les normes de la CEE.

Le problème a été posé pour le fonctionnement de la station d'épuration de la ville de Thessaloniki. Le but de cette étude a été l'analyse quantitative de la dispersion et de la mortalité des bactéries coliformes. Pendant le premier semestre 1990, plusieurs campagnes de mesures ont permis l'échantillonnage et la détermination des concentrations de bactéries coliformes à quatre stations, situées au voisinage d'une source de contamination bactérienne. Parallèlement un modèle mathématique basé sur la simulation de la marche au hasard a été mis au point.

Après étalonnage, ce modèle a servi comme outil pour simuler les impacts, sur les sites de conchyliculture, des eaux usées de la ville.

EN :

In this paper, the contamination of coastal waters by coliform bacteria is considered. The problem is studied in the bay of Thessaloniki (N. Greece) using sampling, laboratory analysis and computerized mathematical modelling. The case study is typical for semi-enclosed bays in the Mediterranean sea. The water renewal and dispersion capacity of the sea are very low and high concentrations of coliform bacteria can be expected. This is the case when modern technology (biological treatment) is used in the sewage treatment stations and chlorination or other types of disinfection are kept low in order to avoid the formation of THM (Tri-Halo-Methanes). The problem of meeting the water quality standards is more difficult, when shellfish growing waters are to be protected.

Impacts to marine environment from sewage and industrial effluents depend on 1) the degree of wastewater treatment, 2) the location of the disposal site and 3) the receiving capacity of coastal waters. The later means the maximum quantity of pollutants that a given area can receive without adversing effects (e. g. eutrophication, change of colour, odours). In the marine environment the receiving capacity is very difficult to assess, as it varies with very irregular way both in space and time. For coliform bacteria, the receiving capacity of a coastal area depends on the dispersive characteristics of the bay and the morality conditions of the bacteries. These are studied here for the case of the bay of Thessaloniki.

The bay of Thessaloniki is located in the NW Aegean sea (Eastern Mediterranean). It is a shallow, semi-enclosed basin communicating with the open sea from the south boundary only. The northern part of the bay has a total area of 300 Km² and a maximum depth of 30 m. it is actually heavily polluted by untreated sewage coming from the city of Thessaloniki. Pollutant loads are estimated as 150 000 m³/d of sewage and 60 000 m³/d of industrial effluents.

The sewage treatment station of the city has started now operating. The provisional disposal site is located in the west toast, three Km from Paliomana, where marine farms for mainly mussels and other shelffishes have been developed. In view of the economic importance of these activities and the need to protect the public health, it is important to assess the impacts from wastewaters to the coastal area.

The main objective of the study is the quantitative evaluation of the dispersion and mortality conditions of coliform bacteria in the bay of Thessaloniki. The ain of the project is the choice of the disposal site of the sewage after treatment, by predicting the water impacts from sewage discharges. During the tire semester of 1990 (February-July 90) sampling and laboratory analyses of E. coli concentrations have been made every 15 days in 4 stations. Some of the samples have been taken during the night. Statistical analyses of the results gave the distribution in space of the max, min, median and C80 values of E. coli (C₈₀ is the concentration which is not exceeded for 80 % cases).

The tale of bacteries in the marine environment is described by the convective-dispersive equation, including the decay term. A linear dependance between bacterial morality and bacterial concentration is assumed. Over the years the advective dispersive equation has been extensively investigated and numerically approximated by numerous methods. Finite differences and finite elements have been used and produced stable numerical results. However, significant errors are introduced in ail these numerical simulations. These are due to the fact that only a limited number of terms in the Taylor series expansions are taken into account. Explicit algorithms suffer from the so-called numerical diffusion. This is an artificial diffusion related to the truncation errors. It is superimposed on the physical diffusion and leads to an excessive attenuation of the input signals. Implicit finite difference algorithms introduce trading effects because the initial signals are propagated at velocities that differ from the physical ones. It seems that particle methods based on random walks are more flexible and easy to use and lead to relatively accurate results.

A random walk computerized mathematical algorithm is developed to simulate the dispersion and mortality of coliform bacteria in the bay. By use of a large number of particles (10³ - 10⁴) which move with the current velocities and by random dispacements following a Gaussian distribution, the contour lines of equal concentrations are obtained. The couple of values for the dispersion coefficient D and the mortality time T₉₀, which simulate better the space distribution of C₈₀ values is : D = 4 m²/s, T₉₀ = 5 h.

The same value of the dispersion coefficient has been independently found by tracking flotting drogues in similar wind conditions (moderate wind). It is concluded that the above values of dispersion and bacterial mortality reflect the characteristic conditions of the bay and can be used to predict the impacts from sewage discharges.

FR :

Pour faire face au problème posé par les rejets aqueux chargés en phénol, deux procédés d'épuration par voie chimique sont proposés. Les deux méthodes font appel au peroxyde d'hydrogène. Celui-ci joue le rôle de promoteur de radicaux lors de l'oxydation de la charge organique par l'oxygène moléculaire dans le premier procédé qui s'inspire de la technique « Wet Air Oxidation » et constitue l'agent oxydant dans le second procédé intitulé « Wet Peroxide Oxidation ».

L'introduction en continu de peroxyde d'hydrogène permet d'initier la réaction d'oxydation du phénol par l'oxygène moléculaire et de réduire considérable-ment les conditions de température et de pression de fonctionnement de la technique WAO classique. La réduction de la Demande Chimique en Oxygène de l'effluent dépasse 95 % à 160 °C en introduisant du peroxyde d'hydrogène à raison de 10 % de la quantité stoechiométrique nécessaire pour l'oxydation complète du phénol. Le second procédé consiste à utiliser l'oxydation par le peroxyde d'hydrogène en présence de fer ferreux (réactif de Fenton) dans des conditions de température (environ 120 °C) conduisant à un abattement important de la charge organique de l'effluent. A température élevée, la compétition entre la réaction de décomposition du peroxyde en oxygène moléculaire inactif et celle de décomposition en radicaux qui développent le processus d'oxydation engendre des conditions opératoires optimales pour lesquelles l'efficacité du procédé est maximale.

Ces deux procédés apportent une solution technique satisfaisante pour traiter, avec un abattement important de la demande chimique en oxygène et du carbone organique, les effluents aqueux assez fortement chargés en composés phénolés.

EN :

Despite of a growing concern about the problems of wastes elimination during the previous years, there is still a lack of processes in order to treat industrial aqueous wastes. Organic aqueous wastes and specially phenolic wastes, that can be either nonbiodegradable or toxic, give rise to one of the main problems. Landfilling disposal and related methods are a priori rejected as they appear to leaving the legacy of a problem we have net been able to solve rather than to considering our environment as being borrowed from the future mankind. Various oxidation techniques are suited for the elimination of this class of wastes. But, because of the environmental and economical drawbacks of incineration, it seems that liquid phase oxidation techniques should be preferred.

The paper reviews : two liquid phase purification techniques using the chemical oxidation route; phenol being used as a test compound. The first technique is adapted from the wet air oxidation (WAO) process and uses molecular oxygen as the oxidizing agent. In the meantime, hydrogen peroxide is added at a low dosage and promotes the radicle reactions. Thus, the reaction temperature and pressure can be set at lower values (typically 160 °C, 25 bar) than usually. In this way, the conventional WAO process, which is very capital intensive because of temperature and pressure constraints is turned into a more affordable process. The second technique uses hydrogen peroxide as the oxidizer. it is associated to a ferrous salt as in the Fenton's reagent but it is run out under temperature (about 120 °C) so that a very important total organic carbon (TOC) removal efficiency con be obtained. This technique was named wet peroxide oxidation (WPO) process. As opposed to WAO, WPO needs only limited capital but generates higher running colts. Yet, both techniques can be regarded as efficient and economically satisfying in order to treat organic aqueous wastes containing fair amounts of phenol or phenolic compounds.

The test compound was selected considering the frequent occurrence of phenol within the wastewaters of refineries, steel works and chemical industries. Their biological treatment is still very difficult for high concentrations despite of an important research activity. Treatment times and efficiencies of physicochemical methods are not but seldom satisfactory. Then, liquid phase oxidation methods have their whole interest. As it was reported that phenolic compounds (methylphenols, chloro-phenols) oxidation proceeds in a similar way than for phenol, the last molecule was considered for assessing the efficiency of both oxidation methods.

The first method (WAO) was tested using a completely mixed batch reactor (stirred autoclave): The cold reactor was loaded with a phenol (2100 mg. 1^-1) and ferrous sulfate (10 mg. l^-11) solution al the convenient pH value (3.5). After heating at the rated temperature, the run was started by injecting instantaneously a large amount of oxygen (10 times the amount necessary). At the same time, a dosing pump was started and fed continuously hydrogen peroxide within the reactor all along the run (90 minutes). The total amount injected was usually 10 % of the amount necessary for a stoechiometric oxidation. The promoting effect of hydrogen peroxide on molecular oxygen is evidenced on figure 2 where the initiating period is shortened and on figure 3 where the oxidation efficiency actually obtained (curve 3) is greater than expected by adding the efficiencies of molecular oxygen and hydrogen peroxide oxidations if separated (curve 2). WAO promoted with hydrogen peroxide gave after 90 minutes better oxidation efficiencies at 160 °C than conventional WAO at 220 °C, then turning into a medium pressure process a high pressure one. The promoting effect of the peroxide is more marked at 160 °C than above 200 °C where a rapid decomposition occurs; dosages greater 15 % do not significantly increase the efficiency and dosages as small as 0,2 % have already a significant affect (see figure 5). Various compounds have been identified and the oxidation sequence is as follows : phenol -> dihydroxy-benzenes -> maleic acid -> oxalic, formic, acetic acids. Most of the remaining chemical oxygen demand (COD) of the oxidized solutions is acetic acid. Only more drastic experimental conditions allow its total removal.

The WPO runs (second oxidation method) were conducted into a similar reactor. It was batch loaded with the phenol (2300 mg. 1^-1) and ferrous sulfate (30 mg. l^-1) solution at pH 3.5. After heating at 120 °C, the run was started and hydrogen peroxide was continuously fed using a dosing pump. The total amount injected all along the run (60 minutes) was the amount necessary for a stoechiometric oxidation. A similar oxidation sequence than reported hereon was observed; pyrocatechol, bydroquinone and oxalic acid were evidenced (figure 9) but, in this case, only very limited amounts of formic and acetic acids were detected. For the two processes, tables 2 and 3 summarize the material balances of the various products as a function of the oxidation time. A 90 % COD removal efficiency and a 70 % total organic carbon (TOC) removal efficiency is reported on figure 10. This result has to be compared with the TOC removal efficiencies (< 25 %) reported for the usual Fenton’s reagent at room temperature. The changes of the pH value and of the COD/TOC ratio (figure 11) during the run are easily explained by considering that oxalic acid is quite the sole product remaining after oxidation contrarily to promoted WAO where acetic acid is the major remaining product. Besides the production of radicles that bring on the oxidation process, a side-reaction decomposes hydrogen peroxide into molecular oxygen which is net active at such a low temperature. The competition between the two reactions makes optimum operating conditions to exist and to lead to a maximum efficiency of the process.

Both processes bring on new methods in order to treat fairly concentrated phenolic solutions with a typical 90 % COD removal efficiency. The products remaining after oxidation (mainly acetic acid or oxalic acid) should not be regarded as a drawback of these processes. In actual fact, such compounds can be easily treated by adding a biological post-treatment unit to the chemical oxidation.

FR :

L'effet du pH (6.2, 7.2, 8.4, 9.6), de la température (4 °C, 12 °C, 23 °C) et du rayonnement solaire a été étudié expérimentalement sur les évolutions des abondances d'Escherichia coli 0126 : B16 et d'autres bactéries pathogènes d'intérêt sanitaire : Salmonella typhimurium et Aeromonas hydrophila. En eau usée épurée, les résultats obtenus montrent que la température et le rayonnement solaire sont parmi les facteurs responsables des variations des abondances de ces bactéries dans les milieux aquatiques. Les faibles valeurs de température (4 °C) favorisent la survie d'E. coli et de S. lyphimurlum et réduisent celle d'A. hydrophila. Les faibles valeurs de température (4 °C) augmentent non seulement la survie bactérienne d'E. coli et de S. typhimurium mais limitent les effets nocifs des pH alcalins (pH 9.6) sur la diminution des abondances de ces bactéries. Parmi les pH alcalins étudiés, le pH 9.6 entraîne la plus forte diminution du temps de survie vis-à-vis d'E. coli, de A. hydrophila et de S. typhimarlam. A pH 9.6, les T₉₀ obtenus à une température de 12 °C sont respectivement de 23, 20 et 33 heures. Le rayonnement solaire joue également un rôle important dans la réduction des abondances bactériennes et ce d'autant plus que le pH est élevé. Dans l'eau usée épurée par lagunage, ajustée à pH 9.6 et exposée au rayonnement solaire, les T₉₀ d'E coli, d'A. hydrophila et de S. typhimarium sont respectivement de 6, 4 et de 6 heures. L'effet combiné du pH et du rayonnement est beaucoup plus important sur la réduction des abondances bactériennes que si l'un des facteurs agit isolément.

EN :

The behaviour of coliforms especially E. coli, pathogenic bacteria (Salmonella typhimurium), and occasionally pathogenic bacteria (Aeromonas hydrophila) in different aquatic environments has often been studied, but rather than a medical than an environmental point of view. However, studies on the affect of various environmental factors such as temperature, pH and solar radiation on their evolution have seldom been carried out. Their action on pathogenic and occasional pathogenic bacteria remains fragmentary.

The aim of this survey is analyse the effect (in laboratory experiment) of certain environmental factors such as temperature, pH and solar radiation on the behaviour of E. coli 0126 : B16, S. typhimurium and A. hydrophila, in waste waters treated in a sewage treatment lagoon. The values of these tested factors are comparable to those which exist in aquatic environments at different seasons.

Experimentation has been realized in a variable and complex medium in this case water than the basin of lagoon, in order to study the varying effects of differents factors on evolution of abundance bacteria. Hierarchization of effects of environmental factors or of certain of their associations has been proposed.

In order to study the action of the temperature and the combined action of temperature-pH on the bacterial evolution, four sterile 1 litre flasks were filled with 500 ml of waste water treated in lagoon. The pH of each flask was adjusted by acid (H₂SO₄) or alkaline (NaOH) solutions at 6.2; 7.2; 8.4 and 9.6. Each flask was seeded with a bacterial inoculum of each of the tested strains (E. coli, A. hydrophila and S. typhimurium). The temperature used for the incubation of the flasks corresponded to the different seasons : 4 °C, 12 °C and 23 °C. The temporal evolution of the abundance of each tested strain is followed up by the indirect counting on specific media : TTC and Tergitol Lactose Agar (Institut Pasteur Production) and incubation at 44.5 °C for 24 hours (E. coli), Salmonella-Shigella Agar (BioMérieux) and incubation at 37 °C for 24 hours (S. typhimurium 4,5 H 1,2 i), Pril-Xylose-Ampicilin agar (ROGOL et al., 1979) and incubation at 37 °C for 48 hours (A. hydrophila ATCC 7966).

The action of the solar radiation and the combined action of solar radiation-pH on E. coli, S. typhimurium and A. hydrophila was carried out as follows :

i) 3 series of glass crystallizers were filled with 500 ml of waste water treated in lagoon (height water : 55 mm). Three reactional mediums were thus prepared by adjusting the pH to 7.2; 8.4 and 9.6. Each pH corresponds to two crystallizers which were seeded with a bacterial inoculum of E. coli; S. typhimurium; and A. hydrophila. For each pH one of the crystallizers is placed in front of a window receiving plenty of solar radiation, the other being protected from all radiation by aluminium paper and used as control. The number of bacteria is verified by counting the colony forming units (c.f.u.) through a dilution-spreading technique on the same media as those used to estimate the action of temperature and pH on the survival of E. coli, A. hydrophila and S. typhimurium.

ii) In order to understand the night and day variations in the number of E. coli according to the solar radiation, another method was used. A standard E. coli inoculum is added to some water of the outflow of lagoon filtered on 0,45 µm (Millipore) and exposed to solar radiation. The temporal evolution of the number of E. coli is then followed by daily counts at 6:00 AM, at 2:00 PM and at 19:00 PM (GMT hour).

Comparison of the different results is made with T₉₀ (the time needed to reduce the initial bacterial population by 90 %) and with an equality test of 2 regression coefficients (FRONTIER, 1981).

Obtained results show that temperature and solar radiation could be considered among factors responsible for the variable abundance of E. coli, A. hydrophila and S. typhimurium. At pH 6.2; 7.2 and 8.4 and effect incubation at 23 °C, the survival time (T₉₀) of E. coli and A. hydrophila is almost the same (T = 53 h) and is anyway inferior to the one for S. typhimurrum which survived nearly twice as long (T₉₀ =100 h). On the other hand, at pH 9.6 and with the same incubation temperature of 23 °C, the survival time of different tested bacteria is reduced. The T₉₀ for E. coli is the same as the one for A. hydrophila (T₉₀ = 20 h) and is 24 hours for S. typhimuriun (table 1). It is almost 4,3 times lower than the one with pH 6.2 and 7.2.

The evolution in number of the same bacteria with similar pH but at a temperature of 12 °C are shown in figure 2. Compared with results obtained at 23 °C, one notices that the arrival of different tested bacteria is longer. The survival times for E. coli and A. hydrophila are almost the same (T₉₀ = 73 h), but differ than the one for S. typhimurium (T₉₀ =142 h), a bacteria which always survives better than other bacteria under the same experimental conditions. Low temperatures (4 °C) increase the survival et E. coli and Salmonella typhimurium and reduce that of Aeromonas hydrophila. Not only do these low temperatures (4 °C) reduce survival of E. coli and S. typhimurium but they also reduce negative effects of alkaline pH (pH 9.6) which decreases the abundance of these bacteria. Among the alkaline pH values studied, pH 9.6 caused the highest reduction of the survival of E. coli, A. hydrophila and S. typhimurium.

We can conclude that at 4 °C the survival time of A. hydrophila significantly decreases in comparison with that of E. coli and S. typhimurium. At 12 and 23 °C, the survival time of E. coli and A. hydrophila is the same. S. typhimurium survived better than both E. coli and A. hydrophila. Solar radiation is also considered an important factor in the decrease of bacterial abundance especially when the pH is alkaline. Under solar radiation and pH 9.6, the survival time of E. coli, A. hydrophila and S. typhimurium is respectively 6, 4 and 6 hours (table 2). The decline in the numbers of these bacteria in effluent lagoon samples was found to be significantly greater in the presence of both pH-temperature or pH-solar radiation than when each of these factors was acting independently.

In waste waters treated in lagoon and under sunlight, E. coli 0126 : B16 undergoes a successive evolution (night growth phase-diurnal decrease phase : fig. 5). According to all count values, we can plot the regression straight line (log 10 bacteria = vs (time)) and estimate the death coefficient which is the slope of this regressional straight line. This model expresses the tendency of the linear decrease of E. coli. Predictions of this model are satisfactory for counting values corresponding to periods of 24 hours or their multiple. The decrease phase or stable night phase cannot be predicte by this model owing to the fact that it is linear and presents the negative slope.

FR :

Des souches appartenant aux espèces : Klebsiella oxytoca, Serratia marcescens et Pseudomonas putida, isolées d'un biofiltre utilisé pour le traitement d'effluents urbains ont été choisies parmi une centaine d'autres pour être réimplantées dans un réacteur du même type. Dans le but de suivre leur fixation en réacteur ouvert, une méthode spécifique de sélection a été développée. Des clones de ces souches résistant naturellement à des antibiotiques (rifampicine, streptomycine, acide nalidixique) et à des substrats suicides (chlorate, bromoacétate, fluorouracile) ont été recherchés. Cette sélection a permis d'obtenir des clones de Klebsiella et de Serratia résistants à 2 g/l de streptomycine, 1 g/l de rifampicine et à 2 g/l de chlorate ainsi que des clones de Pseudomonas résistants à 0,5 g/l d'acide nalidixique et à 2 g/l de bromoacétate ou à 40 mgll de fluorouracile.

Les clones résistants dont les caractéristiques de croissance et les activités enzymatiques sont identiques à celles de la souche sauvage et dont la stabilité génétique a été maintenue après de nombreux repiquages ont été retenus. Afin de valider notre méthode de reconnaissance, une numération de la flore indigène d'un effluent urbain a été réalisée sur les milieux spécifiques des clones résistants : seule une faible proportion de cette flore, à savoir 0,02 % est capable de s'y développer. Des essais préliminaires d'ensemencement du biofiltre avec les souches sélectionnées ont été réalisés, ils montrent que celles-ci s'implantent puisqu'elles sont retrouvées sur les grains de matériau de garnissage et que chacune d'elle représente 1 % de la flore totale.

EN :

Comparison with free tell system, fixed process applied for biological wastewater treatment have been shown to offer numerous advantages. The Biocarbone process, an aerobic down flow immersed bed reactor (ODA patent n° 78-30246), has been selected for many industrial and municipal wastewater treatment facilities.

From this type of aerobic fixed-bed reactor, made of expanded schist as a granular support and fed with clarified domestic wastewater, eigthy-eigth strains were isolated (ZINEBI et al., 1992). Three of the bacterial strains were chosen for their abilities to express high levels of glucidolytic, proteolytic or lipolytic activities and to grow on the granular support as microcolonies which developed into a film of organisms over the whole surface.

Our objective was to initiate biofilm formation by feeding the clean support with thon selected strains named : Klebsiella oxytoca, 501; Serratia marcescens, 532 and Pseudomonas putida, 601. In order to follow attachment kinetics of these selected strains of this biofilter, and to verify their perenity within the biofilm in non sterile conditions (mixed with indigeneous flora from the influent), a specific labelling method was required.

As antibiotic-resistant mutants are easily isolated and the resistances can often serve as convenient genetic markers for use in characterizing bacterial strains, a direct selection of tells acquiring resistance to various antibiotics (ampicillin, streptomycin, nalidixic acid and rifampicin) bas been performed. Selected antibiotic-resistant strains were further incubated in presence of growth inhibitors or suicide substrates in order to select again spontaneous arising mutants well characterized by two distinct markers. From the two bacteria belonging to the Enterobacteriaceae family, mutants having lost the nitrate reductase have been isolated under anaerobic growth conditions in the presence of chlorate. In the case of Pseudomonas strain, mutants resistant towards substrate halogen analogues were obtained.

Colonies resistant to antibiotics and resistant to lethal substrates were isolated : thus, colonies of Klebsiella resistant to streptomycin at 2 g/l, to rifampicin al 1 g/l and chlorate al 2 g/l ; colonies of Serratia resistant to streptomycin at 2 g/l or to rifampicine at 1 g/l and chlorate at 2 g/l and colonies of Pseudomonas resistant to nalidixic acid at 0.5 g/l and to bromoacetate at 2 g/l or to fluorouracil at 40 mg/l, were obtained. We have selected : trains showing the same doubling time as well as the same final population titan the parental strains when growths were performed with or without the markers. The three strains retained were : Klebsiella oxytoca, 501 R₁S₂Cl₂ which grew on the Mac Conkey medium added with 1 g/l of rifampicin, 2 g/l at streptomycin and 2 g/l of chlorate; Serratia marcescens, 532 S₂Cl₂ (on Mac Conkey plus 2 g/l of streptomycin and 2 g/l of chlorate) and Pseudomonas putida, 601 NB₂ (on King plus 0,5 g/l of nalidixic acid and 2 g/l of bromoacetate). These specific media for the detection of selectionned clones were selective toward a fixed indigenous flora since only 0,02 % of total heterotrophic population can grow.

A column filled with grains of « Biodagen » either colonized by natural, microbial populations or with clean grains of « Biodagen » was fed with a population of the : train Klebsiella 501 R₁S₂Cl₂. The strain colonized virgin « Biodagen » and maintained population of 4.10⁶ CFU per grain for 9 days with new material and 10⁵ CFU for 7 days with precolonized material.

Experiment with a mixed population resulting from the three identified microbial species have been conducted with clean grains of « Biodagen », a whole population of 10⁷ CFU per grain was obtained after two days and each identified strain corresponded to 1 % of the entire bacterial population. The relative concentrations of the three : trains did not decrease feeding the column with a mixture of the three : trains and of wastewater but slightly decreased when the column was fed with wastewater only.