The combined sewer system of the urban area of the city of Pau (south-west France) drains about 50 km² with 150 000 inhabitants. The majority of the surface is composed of residential neighborhoods; several areas of commercial and industrial activities are also present. The ground impervious ratio is about 80%. The sewer network is 800 km long and more than 75% of the population is connected to a combined sewer. This study area covers the entirety of the catchment system as well as the sewer network, the CSO, the wastewater treatment plant and the receiving water (river “Gave de Pau”). That makes it possible to have a global comprehensive view of wastewater management from the sources until the discharge in the receiving environment.

Since 2009, the study area is equipped with four tipping-bucket rain gauges, seven flow meters in the mains parts of the network and 27 CSOs have been also equipped with flow meters. For the monitoring of pollutant concentrations, one turbimeter and one conductimeter were installed in March 2012 downstream from the junction of the two main pipes just before the WWTP. This site was chosen because it allows the monitoring, in the same time, of wastewater discharged at the main CSO of the study area and wastewater directed towards the WWTP. This probe is a nephelometric turbidimeter with a 0 to 4 000 NTU range. Turbidimeter is calibrated with formazine solution. The turbidity measurements can be disturbed by a waste in front of the probe; a technique for the removal of these outliers has been implemented. The conductivity probe is an inductive condictimeter with 0 to 10 000 µS/cm range calibrated with NaCl solution. The two probes were previously calibrated at the laboratory and monthly in situ verification was performed. No derivation was observed during the entire course of the study period.

For all samples, automatic samplers are used; the samples are systematically kept cold (4°C) and analyzed in the laboratory within 24 h following the protocols described below. All analysis are performed on the raw sample using the following methods. The measurement of turbidity and conductivity is made with the probes described above. Total Suspended Solids (TSS) are analyzed according to the NF T 90-105 standard. Chemical Oxygen Demand (COD) is determined with Aqualytic colorimetric kits after validation according to the normalized standard NF T 90-101. For total nitrogen measurements, a Shimadzu TOC-VCSN is used. For all these analysis, triplicates, standards and blanks are done.

108 samples of wastewater were analyzed to establish the correlation between turbidity/TSS and COD. 70 samples were used for the correlation between total nitrogen and conductivity. The samples were taken from the entire network during dry and wet weather. Positive correlations between TSS/turbidity, COD/turbidity and total nitrogen/conductivity were found to be consistent (Figure 1). The corresponding correlation functions were established:

• TSS = Turbidity^1.0124 + e^0.2894

• COD = Turbidity^0.917 + e^1.559

• Total nitrogen = Conductivity^1.089 + e^3.8501

-> Voir la liste des figures

Correlation coefficients are between 0.83 and 0.92 and allow a good estimation of concentrations of pollutants. It has been demonstrated that these correlations are not statistically different between dry and wet weather. Indeed “a” and “b” coefficient are not statistically different for the dry and wet weather conditions for all of these correlations (in a 95% confidence interval).

Figure 2 gathers all the data recorded during October 2012. From the top to the lower part of the figure, rainfalls, treated flow at the WWTP, wastewater discharged at the CSO, conductivity and turbidity are showed. Data monitoring performed each five minutes allows a good understanding of the sewer system. This month starts with a period of ten days with dry weather. It shows that during this period, flow, turbidity and conductivity evolved in the same time. This is due to human and industrial activities. For these parameters, a minimum is observed during the night between 3:00 and 6:00 am. Then a maximum is observed around 11:00 am. Four periods of rain occurred during the month. For each one, although the treated flow goes up by a factor 2 compare to the dry weather period discharge by CSO occurred. During rain events, a turbidity peak is observed as soon as the flow goes up. Nevertheless, these peaks are different from one event to the other. It’s particularly important for the first rain event (October 11 and 12, 2012) with a maximum around 450 NTU. But smaller for all the other events: between 150 and 250 NTU. Evolutions of conductivity during wet weather periods are really different from turbidity. For all the rain events, a decrease is observed and conductivity goes up at the end of the event. This decrease can reach a factor 7 compare to dry weather period and seems to be correlated with the flow increase. Using correlation functions presented before, these data can be converted in TSS, COD and total nitrogen concentrations and will be used to calculate pollutant fluxes.

To translate the large amount of data produced by the probes into comprehensive scheme of the network functioning over a long time period (a month or more), we integrated data at a daily scale by calculating daily fluxes of pollutants. Figure 3 shows the results obtained for COD fluxes, flow and COD concentrations values. Because TSS and COD values result of the same measure (turbidity), they follow exactly the same trend, here just COD is presented. To make the discussion easier, characteristics of rain event of each day are presented in table 1. It is interesting to compare the three parameters studied: fluxes, flow and COD concentration. The first rainfall event that occurred on October 11, 2012 generates a COD fluxes increase by a factor 3.5 compare to the dry weather fluxes. This is due to the simultaneous increase of flow and COD concentration of respectively 2.7 and factor 1.3. During this event, stormwater contributes to increase the fluxes of pollutants probably in two ways, first by mobilizing pollutants on impermeable surfaces and secondly by sediment erosion in the network. The latter is probably the predominant source for this increase. Indeed, this deposits resuspension phenomenon during rainfall events has been demonstrated in various studies (CHEBBO et al., 2003; GROMAIRE et al., 2006; HANNOUCHE et al., 2011). The dry weather period particularly long (11 days) preceding this event is certainly the cause of the important increase in the fluxes due to a large stock of sediments rich in COD. The following rainfall events of 12 and 14 October 2012, although less intense, also generate a flow increase in the network of the same order of magnitude as that of October 11^th. However, they differ in that the COD concentrations for these events are two times lower than during a dry spell. These events probably eroded less sediment, sediment stock erosion removed during the event of 11th have not been restored yet. Another phenomenon may contribute to the decrease in COD concentrations; it is the effect of dilution of wastewater by stormwater lightly loaded in COD. This dilution effect appears particularly marked during the subsequent events of 19 and 20 October. COD fluxes are 3.5 times greater the 19^th than the 20^th although the flows are higher the 20th (by a factor 1.5 with respect to 19). It is likely that the relatively low fluxes observed on the 20^th are due to the fact that the sediments were eroded during the previous event and thus the amount of material available for this event is low. These examples highlight the complexity and the importance of erosion phenomena in wastewater and their importance in the transport of pollutants to the WWTP or the receiving environment. During this month, eight days of CSO (Combined Sewer Overflow) discharge were observed. The spilled volume varies between 130 000 and 2 000 m³/day in relation with the intensity and the depth of water precipitated. COD fluxes discharged into receiving waters vary in a wide range: from 400-28 000 kg O₂/d. The most critical events occurred on 11 and 19 October with a quantity of COD discharged close to 25 000 kg O₂ every day. However, spilled volumes are different on the two days with 35 500 and 83 000 m³ the 11^th and 19^th respectively. These data show the importance of not only considering the volume spilled but the flow of pollution in order to evaluate the impact on the receiving environment. Online measurement of pollutant concentrations is interesting because it allows to obtain, almost instantaneous, information and thus makes possible a dynamic management of sanitation.

Figure 4 shows the total nitrogen fluxes, the rain, the total volume and the concentration of total nitrogen for the study period. Total nitrogen fluxes evolution is different to that observed for COD. Nitrogen fluxes vary relatively little over the period. During the largest rainfall event, they increase by a factor of 1.6 compared to dry weather fluxes. In dry weather, no significant difference could be observed. This behavior can be explained because in the sewerage, total nitrogen is mainly present in dissolved form (from 95 to 70%) (KAFI et al., 2008). Therefore it is weakly influenced by the phenomena of deposits and sedimentation as it is the case for TSS and COD. Therefore, no significant phenomenon and resuspension during storm events accumulation is observed. Similarly, it seems that the contribution of runoff is low, nitrogen sources are mainly due to domestic and industrial inputs. According to KAFI-BENYAHIA (2006), the nitrogen due to runoff is consistently below 10%. The amount of nitrogen is slightly affected by rainfall events and remains essentially constant during periods of dry weather and wet weather. During rainy periods, the flow increases at the same time a reduction of the total nitrogen concentration in proportion to the increase of flow was observed, and the fluxes are relatively steady. In the case of particularly significant rainfall events (that of October 20, for example), a small increase in fluxes is however observed (1.5 fold). According to GASPERI et al. (2010), in rain 30% of the nitrogen derived from resuspension of sediments and 6% of runoff which would explain the increase in flow. In contrast to what is observed for COD, following a rainfall event, nitrogen concentrations found their level of dry weather when the flow has recovered its level of dry weather which usually takes about three days.