Eroded sewer sediments are a significant source of organic matter discharge by combined sewer overflows. Many authors have studied the erosion and sedimentation processes at the scale of a section of sewer pipe and over short time periods. The objective of this study was to assess these processes at the scale of an entire sewer network and over 1 month, to understand whether phenomena observed on a small scale of space and time are still valid on a larger scale. To achieve this objective the continuous monitoring of turbidity was used. First, the study of successive rain events allows observation of the reduction of the available sediment and highlights the widely different erosion resistance for the different sediment layers. Secondly, calculation of daily chemical oxygen demand (COD) fluxes during the entire month was performed showing that sediment storage in the sewer pipe after a rain period is important and stops after 5 days. Nevertheless, during rainfall events, the eroded fluxes are more important than the whole sewer sediment accumulated during a dry weather period. This means that the COD fluxes promoted by runoff are substantial. This work confirms, with online monitoring, most of the conclusions from other studies on a smaller scale.
The European Water Framework Directive has the aim of achieving ‘good ecological and chemical status’ for all of Europe's aquatic environments by 2015 (WFD 2000). In this context, reduction of the impact of wastewater effluents of urban areas from combined sewer overflows (CSOs) and wastewater treatment plants (WWTPs) has become a priority. This directive implies that sanitation managers must take measures to decrease pollutant transfer into the receiving environments. This challenge requires a perfect knowledge of pollutant concentrations and fluxes, as well as the dynamics of the wastewater sewer system. Continuous monitoring of wastewater thus becomes a key issue for wastewater treatment management and for protection of the receiving environment.
Recent studies have demonstrated the usefulness of measuring turbidity as a substitute for conventional analysis (Lacour et al. 2009a; Hannouche et al. 2011; Métadier & Bertrand-Krajewski 2012). Because good correlations are found between turbidity and total suspended solids (TSS) or chemical oxygen demand (COD), online turbidity measurements allow instantaneous monitoring of global pollutant loads such as TSS and COD (Lawler et al. 2006; Lacour et al. 2009b; Métadier & Bertrand-Krajewski 2012). The dynamics of the studied systems can then be obtained with a better temporal resolution than with discrete analyses at laboratory scale. Better management of CSOs and WWTPs can be achieved (Boutayacht et al. 2010; Blumensaat et al. 2011) in this way.
Many studies point out the importance of sedimentation in combined sewers during dry weather periods and then erosion processes during wet weather periods (Ahyerre et al. 2000; Oms et al. 2005). Most of them have shown that between 30 and 65% of the wet weather fluxes of TSS and COD come from sewer sediments (Chebbo et al. 2001; Gasperi et al. 2010). Deposition/erosion processes contribute to increasing the sewer pollutant fluxes and the total pollution load discharged by CSOs during rain events. Indeed, 40–70% of the total pollution load discharged by CSOs comes from sewer deposits (Laplace et al. 2003). Because many big cities in the world use a combined sewer, it is important to have more information about these phenomena in order to propose technical means to reduce the discharge of pollutant load by CSOs.
Usually, studies about erosion and sedimentation processes are focused on a short portion of the sewer (less than 1 km) and during short periods (less than 1 week). In such a context, we propose to use continuous turbidity monitoring to assess the erosion and sedimentation on the scale of an entire sewer network of a city (Pau, south-west France) and during a long period (1 month) in order to understand whether phenomena observed on a small scale (space and time) are still valid on a larger scale.
MATERIALS AND METHODS
The combined sewer system of the urban area of the city of Pau (south-west France) covers about 50 km2 and has 150,000 inhabitants. The site is mainly composed of residential areas with several commercial and industrial areas. The average level of imperviousness is 80%, ranging from 60% for residential to 90% for highly urbanized areas (town centre). The sewer network is 800 km long and more than three-quarters of the population are served by a combined sewer system, the other 25% being connected to a separate sewer. The WWTP treats on average 2,000 m3/h. The activated sludge WWTP has a designed capacity of 190,000 population equivalents. In the study area, the sewer network is equipped with 44 CSOs; most of them are devices with a single side weir. The average slope of the sewer network is close to 0.8% and varies between 0.4 and 1.3%. Pipe diameter varies a lot: from 200 mm to 2 m.
Sensors and online equipment
The study point is located after the junction of the two main combined sewer pipes of the city, which are directly linked to the WWTP (Figure S1, available online at http://www.iwaponline.com/wst/072/350.pdf). This location was chosen because it allows study of the entire sewer system dynamics. For this study, one turbidimeter, one rain gauge and one flow meter were installed.
The turbidity sensor is a Hach Lange SOLITAX sc built to measure turbidity over a 0–4,000 FNU (formazine nephelometric unit) range (NF EN ISO 7027 2000). Every 5 min, the mean turbidity is calculated by integration of 15 s of data. To calculate the hourly pollutant fluxes (Fhx), an average turbidity is calculated each hour. Data are stored on a Hach Lange SC200 recorder and downloaded to a laptop every week using a SD card. The turbidity sensor was calibrated at the laboratory with formazine standard solutions. However, as formazine is toxic and not stable over time, on-site calibration is checked using kieselguhr. Both calibrations were consistent and the relationship between the two calibration processes was established. No sensor drift was observed during the experimental period.
At the study point, a flow meter based on acoustic Doppler principles (Mainstream, Hydreka) records the average flow each hour. The data are transmitted automatically to an acquisition station where they are analyzed with WINFLUID software.
To monitor the rain, a PULSONIC tipping-bucket rain gauge is used. It is situated at the center of the urban area, and the measurements range between 0 and 450 mm/h with 0.2 mm accuracy. Results are sent daily to an acquisition unit and are examined using PULSOWIN software.
Sampling and analyses
To establish the correlation functions between turbidity, TSS and COD, nine sampling sessions were performed between March 2010 and November 2011 at the study site. Four were conducted during dry weather, and the remaining five during different rainfall events. Indeed, Lacour et al. (2009a) and Hannouche et al. (2011) showed that using five rain events is sufficient to build a TSS or COD/turbidity relationship and have uncertainty less than 20%. The rainfall events varied from 15.3 to 43 mm with durations from 5 to 23 h. The maximum intensity and the previous dry weather periods varied from 2.8 mm/h to 8.8 mm/h and from 0 to 13 days, respectively.
For all this sampling, automatic grab samplers (ISCO 3700 Neotek) were used. The collected samples were refrigerated at 4 °C and analyzed at the laboratory within the following 24 h. During these sessions, wastewater was collected over 24 h periods by collecting 125 mL every 30 min. For each session, 12 samples of 500 mL of bulk samples were analyzed. At the laboratory, turbidity and conductivity measurements were carried out using the same sensors as those used in situ and already described. For the COD measurement, the colorimetric method was used with Aqualytic kits (0–1,500 mg O2/L and 0–150 mg O2/L). Validations were carried out to compare results with the normalized method (NFT 90-101/ISO 6060:1989 2001). Variations of less than 10% were observed. For each session of analyses, blanks and triplicate measurement were carried out. Moreover, a certified standard sample was analyzed (ChemLab 1,000 mg O2/L) during every set of analyses. For the TSS measurement, the standard method (NF EN 872 2005) using filtration and weight determination was used with blanks and triplicate measurements.
where TSSi and CODi are the concentrations in mg/L at time i and Ti is the turbidity (NFU) recorded at time i.
Correlation coefficients are above 0.9, showing statistically acceptable correlations in accordance with the statistical test of residues and allowing the evaluation of pollutant parameters such as TSS and COD with this online sensor. In our study area, dry weather and wet weather data were pooled because slope and intercept were not statistically different (P < 0.05) (Figure S2, available online at http://www.iwaponline.com/wst/072/350.pdf). A slightly higher correlation coefficient is observed in this study compared to the literature for the COD/turbidity relationship, whereas it is similar for the TSS/turbidity relationship (Lacour et al. 2009a). During the study period, wastewater samples were analyzed and no modification of these correlations was observed. Before using recorded data, a validation test was done according to Métadier & Bertrand-Krajewski (2011b) in order to remove incorrect values. The uncertainty in the result of the pollutant concentration parameter was estimated according to the law of propagation of uncertainties (LPU) at 15%.
RESULTS AND DISCUSSION
All rain events for the month of October 2012 were studied for this paper because they were representative of the diversity of weather characteristics (dry weather period; intensity, duration, frequency and overall height of rain events) encountered during a year in this region. All the data collected in October 2012 are shown in the Supplementary Information (Figure S3, available online at http://www.iwaponline.com/wst/072/350.pdf). Recorded data were processed using the correlation functions in order to obtain COD and TSS concentrations. For October as well as for the rest of the year, COD measurements showed exactly the same behavior as TSS. So to make the results clearer, COD values only are shown in this paper. Data obtained during this month will be used in two ways, firstly at the scale of a rain event and then at the scale of the entire month.
Assessment of erosion processes at the event scale
To understand precisely the dynamics of COD during a period of successive rainfall events in the combined sewer system, the study focused on the first group of rain events of the month (called event A) occurring on 11th and 12th October. Following a dry weather period of 11 days, rain event A was constituted by a series of four sub-rainfall events (1, 2, 3 and 4) close to each other. Their characteristics and the monitored data are presented in Figure 2.
The first sub-rainfall event (1) (1.6 mm) follows a dry weather period of 11 days. The flow and the flux remain low, just 1.5 times that during a dry weather period, nevertheless, this event leads to a large increase in COD concentration (approximately 1,000 mg O2/L). The three other sub-rainfall events take place after a short dry period (less than 1 day long). They have different characteristics in terms of rain intensity, COD fluxes and concentrations but they have a high flow (close to six times more than the dry weather flow). The second sub-rainfall event (2) is particularly short and intense, with 6.8 mm of water falling in 1 h, and it exhibits the highest peak concentration (1,300 mg O2/L) and peak flux (15,000 kg O2). The third one (3) has a low intensity (maximum of 2.6 mm in 1 h) but occurred over 7 h resulting in 10.7 mm of rain. During the last sub-rainfall event (4), 9.6 mm fell in 5 h with a maximum intensity of 5.7 mm/h. These last two events show moderate increases in COD concentrations, 600 mg O2/L and 230 mg O2/L respectively. In accordance with low COD concentrations, fluxes also decreased. Concentrations found in this study are of the same order of magnitude as those described by Métadier & Bertrand Krajewski (2012), although the watershed studied in this paper is larger.
In a combined sewer, wet weather COD fluxes are composed of wastewater from houses and factories, and this part is supposed to be identical to that during dry weather periods. They are also composed of COD from runoff on all the impermeable surfaces (roofs, streets, yards, etc.), and COD coming from exchanges with the pollution stocks existing in the sewer system (Chebbo et al. 2001; Gromaire et al. 2001).
Intake of COD by runoff is variable depending on the study. It would only account for 10–20% (Gasperi et al. 2010) to around 45% (Chebbo 1992) of the total fluxes during the wet weather period, depending on the watershed characteristics. Several studies have emphasized that the erosion of in-sewer pollution stocks should be the main source of particles and organic matter during wet weather periods (Chebbo et al. 2001; Gromaire et al. 2001; Hannouche et al. 2011).
For each rain event, the maximum flow is observed around 1 h after the rain event peak due to water transfer in pipes. The maximum flow is well correlated with the total rainfall height (r² is around 0.92). For each event, as soon as the flow increases, the COD concentration increases and their maxima are achieved at the same time. In consequence, highest COD fluxes are correlated with highest flows. This behavior is due to the contribution of particles from erosion processes. Erosion capacity is related to the increase in shear stress, due in part to the water velocity but also to the flow (Laplace et al. 2003). Similar observations about flow and concentration behavior were observed by Oms et al. (2005) on a section of the sewer network.
Because the first sub-rainfall event (1) occurs after a long preceding dry weather period (11 days), a large amount of TSS and COD are accumulated on impervious surfaces and in the sewer network. Thus, the high COD concentration of this event is due to the combination of in-sewer erosion and leaching of impervious surfaces. With only 1.6 mm of rainfall during this first event, the erosion phenomenon was significant, suggesting that a proportion of these sediments are easily erodible. Indeed this low rain event leads to a moderate increase in shear stress. The study done by Ahyerre et al. (2001) in a section of combined sewer in Paris demonstrated as well that some sediment is easily erodible by a low flow. Nevertheless, for this first sub-rainfall event the maximum COD flux remained low (2,000 kg O2/h) due to low flow (close to dry weather).
In-sewer and on-catchment stocks are also largely available for the second rain event because it is for this event that the highest COD concentration and fluxes were observed. Nevertheless, the short dry weather period between these two sub-rainfall events (less than 1 day) does not allow a renewal of the sediment stocks (as shown in Figure 3, the reconstitution of sewer sediments takes around 5 days). The intense rain input over a very short period of time during sub-rainfall event 2, leads to a greater flow than for event 1 (by around a factor of 6), and this results in a greater mobilization of stocks still available due to higher shear stress in pipes.
For the last two sub-rainfall events the total flow reaches the same level as for the second rain event; nevertheless lower COD concentration and fluxes peaks were observed, especially for the last event. This is due to the reduction of the available stock in the sewer because of the first two rain events and the too short dry weather period between all these events. However, despite small increases in COD fluxes for the last two events, COD stocks in the sewer are still available because COD fluxes are between six and four times higher than during dry weather. This emphasizes that sediments are not eroded at each rain event but that the stock is constituted by different layers increasingly difficult to erode. This sediment stratification with different erosion resistance has already been shown by Ahyerre et al. (2001), Oms (2003) and Banasiak et al. (2005).
The dynamics of COD in sewer systems are complex because of the diversity of possible sources and the complexity of processes that can be involved (Gasperi et al. 2010). A similar COD dynamic was observed repeatedly over the year 2012–2013 during different rain events. The observations made in previous studies at the scale of a section of sewer are thus confirmed for successive rain events at the scale of an entire sewer system. Nevertheless, to confirm this trend on a larger time scale, the second part of this paper will deal with the use of continuous monitoring data. This will be achieved by daily COD flux estimations available from COD/turbidity correlations.
Assessment of erosion and sedimentation phenomenon in combined sewer over 1 month
To make an assessment of the quantity of COD accumulated and eroded in the sewer network, COD daily fluxes were calculated during the month of October. Figure 3 presents the difference between measured COD fluxes and the average dry weather fluxes (ΔFdCOD) and highlights the flux variation for each day of October 2012. The average dry weather flux for the month of October was not significantly different from the other dry weather periods for the year 2012–2013.
During the dry weather period at the beginning of the month (from 3rd to 10th) the ΔFdCOD is close to zero. This means that the COD fluxes are constant during this period, and the sedimentation and erosion processes are in equilibrium to reach a steady state. Observations of the sediment layer done by Oms et al. (2005) with an in-sewer observation box also showed a steady state of the sediment layer during a dry weather period. The observation done at the scale of a section of sewer is thus confirmed by this result at the scale of the entire network.
Each rainfall event group (A, B, C and D) brings positive ΔFdCOD ranging between 3,000 and 40,000 kg O2/day. These large positive fluxes were also observed for the following months of the year 2012–2013. They are mainly due to runoff and sewer sediment erosion. Sediment sources in sewers can be highly variable (Gasperi et al. 2010). According to these authors, in-sewer deposits could be related to three main categories: biofilm, organic layer and gross bed sediments. Our results confirm previous investigations showing that eroded deposits during rain events are highly organic (Gasperi et al. 2010). Furthermore, Hannouche et al. (2014) have shown that in-sewer deposit contributions can be important for sewers not expected to accumulate sediment (slope around 2.7%).
Conversely, after each rainfall event, ΔFdCOD are negative, ranging between –6,000 and –1,000 kg O2/day. Figure 3 shows that negative fluxes are maximal just after a rain event and then increase each day to reach zero. These negative fluxes are explained by the renewal of the sediment sewer stocks. After each rainfall event the flow in pipes slows down, allowing sedimentation of suspended solids produced on the entire watershed, leading to an accumulation of sediments in the sewer. Nevertheless, as proposed by Gasperi et al. (2010), the stock renewal is not only due to sedimentation processes. These authors suggested that the sewer network behaves as a bio-physicochemical reactor leading to in-sewer transformation of deposits. Reconstitution of the sediment stocks occurs in the free spaces made available by the previous rainfall event. The observation of the stock reconstitution can be observed from 15th to 18th and from 28th to 31st October. Nevertheless because of rain events at the end of these two periods it is not possible to observe the return of ΔFdCOD to zero. This is possible at the beginning of the month from the 3rd to the 9th. It appears that to renew the sewer stocks in this study, 5 days of dry weather are required after a rainfall event (Figure 3).
These observations could allow evaluation of the sewer sediment stocks in two ways. First, it is possible to estimate the storage capacity of the entire network by summing all negative fluxes stored in the sewer system between the end of one rain event and the day where ΔFdCOD returns to zero. After rain events B and D, four dry weather days are necessary to obtain ΔFdCOD close to zero, meaning that little loss occurs in pipes and that the sewer system is close to being balanced. The storage capacity of the network estimated during the four dry weather days after rain event B is about 15,000 kg and close to 20,000 kg after rain event D. It is difficult to compare this result to other studies because of very different space and time scales. Indeed, Ahyerre et al. (2000) observed a COD accumulation of 23.6 kg/day on 150 m of sewer. Nevertheless, the transfer of this result at the scale of the entire network is not possible because of the heterogeneity of the sewer in terms of diameter, slope and roughness.
Secondly, the estimate can be done by summing the positive ΔFdCOD during a rainfall event. For example, summing all the positive ΔFdCOD of the largest rainfall event of October (C) gives a total of about 40,000 kg. This result is around twice the estimated storage capacity of the pipes using negative ΔFdCOD. This difference can be explained by the input of runoff on impervious surfaces, indeed this source can be consequent (45% for Chebbo (1992)) but highly variable depending on the studies (10–20% for Gasperi et al. (2010)). Our results show that on this study site, input of COD by runoff is quite important and closer to the highest observed level (Chebbo 1992).
The calculated mass balance processed using positive and negative fluxes over the entire month of October leads to a positive result of +70,000 kg O2, highlighting the importance of runoff processes in the input of COD. Taking into account that 750,000 m3 of water were brought by the runoff at the study point during this month, the average concentration of the runoff water should be about 90 mg O2/L to explain the +70,000 kg O2. This estimation is in accordance with other studies like Gromaire et al. (2001) and Bressy (2010), but also with some analysis done on the study site but not presented in this paper.
This study has demonstrated that using continuous monitoring data to assess erosion and sedimentation processes is possible over a large scale of time and space. Monitored data were used in two complementary ways. First, detailed study of a particular rainfall event shows the importance of the strength of erosion due to the flow level and available stock of sediments in the sewer. This affects the level of COD and TSS concentrations during consecutive rainfall events. Nevertheless different layers of available stocks in the sewer pipes are increasingly more difficult to remobilise.
Secondly, estimation of the daily COD fluxes brought about by the different rainfall events of the month allowed calculation of the sediment stocks in the sewer network during dry weather periods after a rainfall event. The sediment storage decreased each day and becomes negligible after 5 days, which shows a potential saturation of the accumulation sites. Nevertheless, during rainfall events, the eroded fluxes were higher than the sum of the sewer sediment stored during dry weather periods meaning that the fluxes brought by runoff are consequent.
Most of the conclusions about erosion and sedimentation obtained by other studies at a small scale of time and space are validated by our result at the scale of 1 month and over the entire network, although the methodology used is different.
Continuous monitoring is an interesting tool for better understanding the sewer system and to optimize wastewater management. This use of continuous monitoring allows better comprehension of the erosion/sedimentation dynamic in sewers, which is one of the most important parameters in discharge loads by CSOs. So, this tool could be used to limit raw wastewater discharge into the receiving streams.
Financial support for this project by the city of Pau (Communauté d'Agglomération de Pau Pyrénées) and the Adour Garonne Water Agency (Agence de l'Eau Adour Garonne) is gratefully acknowledged.