Abstract

On-site wastewater treatment systems are approved by the French regulation based on the results of platform tests following the European standard NF EN 12566-3. In addition to this approval for the treatment system, at least 90% of outlet concentrations have to be below 30 mg L−1 for total suspended solids (TSS) and 35 mg L−1 for biochemical oxygen demand. The aim of this study is to evaluate the effluent quality of these treatment systems on site, i.e. under real operating conditions, and to assess their performances. Between 2011 and 2016, 1,286 treated wastewater samples were taken from 231 on-site sanitation facilities in France. Data collected are heterogeneous and a robust statistical methodology (using a generalized log-linear model) was used to study the effects of four explanatory variables (treatment systems, loading rate, aging and sampling methods) on the distribution of treated wastewater concentrations. The model calculates median outlet concentrations depending on the effects identified. Its application allowed studying and comparing the outlet median concentrations of 21 on-site sanitation systems classified into nine categories and three groups. Four treatment systems out of the 21 monitored showed TSS median outlet concentrations below 10 mg L−1 and four treatment systems have TSS medians higher than the regulatory threshold of 30 mg L−1.

INTRODUCTION

In the 2000s in France, on-site wastewaters were generally treated using a septic tank combined with (i) on-site infiltration based on soil purification capacities or (ii) filters filled with sand or zeolite. These treatment systems also called ‘traditional systems’ are still allowed by the current regulation (Decree of 7 September 2009 (Arrêté du 7 septembre 2009)).

From 2009, changes in the French regulations (Decree of 7 September 2009) have enabled the marketing of new treatment systems approved by the French government. The approval procedure is based on a complementary analysis of the results of platform tests performed using the EU procedure NF EN 12566-3. This standard does not refer to any thresholds to qualify the treated wastewater quality (NF EN 12566-3+ A2 2013). Therefore, France has defined regulatory thresholds as part of the approval procedure. A treatment system is approved and marketed in France if at least 90% of the effluent measured concentrations comply with the maximum thresholds of 30 mg L−1 for total suspended solids (TSS) and 35 mg L−1 for biochemical oxygen demand (BOD5) and if none of the effluent concentrations exceeds the thresholds of 85 mg L−1 for TSS and 50 mg L−1 for BOD5. The number of samples (20 to 26 during the standard procedure) is imposed by the EU standards.

Following this procedure, 650 approvals have been delivered to 63 manufacturers from 2009 to 2016. These approved systems thus increased the technical possibilities offered to any owner who needs to purchase or rehabilitate their on-site wastewater treatment facility.

In this context, this study aimed to evaluate the effluent quality of on-site sanitation systems (i.e. domestic facilities of less than 20 population equivalent (PE)) in France and to assess their performances under real operating conditions.

From 2011 to 2016, 1,448 samples were collected from 246 different facilities. All treatment systems studied are based on biological processes that have proven to be efficient in the domain of domestic wastewater treatment. The facilities are divided into three groups of wastewater treatment processes: (i) attached growth systems on fine media (AGSFM), (ii) submerged attached growth processes (SAGP) and (iii) activated sludge processes (ASP) and into 13 categories and 33 systems as described in Table 1.

Table 1

Description of the three groups of biological processes subdivided into 13 categories and 33 systems

Groups Description of the process Categories Systems 
Attached growth systems on fine media (AGSFM) Septic tank + filter filled with fine materials Suspended solids are retained at the surface by mechanical filtration and the dissolved pollution is degraded by fixed-film bacteria (Metcalf & Eddy, Inc. 2003). Sand (S)
Constructed wetland (CW)
Zeolite (Z)
Coconut shavings (CS)
Rock wool (RW)
Pine barks (PB) 
S1, S2, S3
CW1
Z1, Z2
CS1, CS2
RW1, RW2
PB1 
Submerged attached growth processes (SAGP) Primary settlement tank + bioreactor filled with media (fixed or fluidized) + clarifier The dissolved pollution is degraded by fixed-film bacteria and the sludge in the clarifier is recirculated into the primary settlement tank by a pump. Fixed media (Fx)
Fluidized media (Fl)
Rotating biological contactors (RBC)a 
Fx1 to Fx9
Fl1
RBC1 
Activated sludge processes (ASP) Classical: primary settlement tank + bioreactor + clarifier The dissolved pollution is degraded by suspended bacteria in the bioreactor (Dubois & Boutin 2017). Sludge in the clarifier is recirculated into the primary settlement tank by a pump. Classical (C)
Without primary settlement tank (WoPST)
With additional treatment by filtration (WAT)
Sequencing batch reactor (SBR) 
C1, C2, C3
WoPST1, WoPST2
WAT1 SBR1 to SBR5 
Groups Description of the process Categories Systems 
Attached growth systems on fine media (AGSFM) Septic tank + filter filled with fine materials Suspended solids are retained at the surface by mechanical filtration and the dissolved pollution is degraded by fixed-film bacteria (Metcalf & Eddy, Inc. 2003). Sand (S)
Constructed wetland (CW)
Zeolite (Z)
Coconut shavings (CS)
Rock wool (RW)
Pine barks (PB) 
S1, S2, S3
CW1
Z1, Z2
CS1, CS2
RW1, RW2
PB1 
Submerged attached growth processes (SAGP) Primary settlement tank + bioreactor filled with media (fixed or fluidized) + clarifier The dissolved pollution is degraded by fixed-film bacteria and the sludge in the clarifier is recirculated into the primary settlement tank by a pump. Fixed media (Fx)
Fluidized media (Fl)
Rotating biological contactors (RBC)a 
Fx1 to Fx9
Fl1
RBC1 
Activated sludge processes (ASP) Classical: primary settlement tank + bioreactor + clarifier The dissolved pollution is degraded by suspended bacteria in the bioreactor (Dubois & Boutin 2017). Sludge in the clarifier is recirculated into the primary settlement tank by a pump. Classical (C)
Without primary settlement tank (WoPST)
With additional treatment by filtration (WAT)
Sequencing batch reactor (SBR) 
C1, C2, C3
WoPST1, WoPST2
WAT1 SBR1 to SBR5 

aRBC has been attributed to the SAGP category although not exactly meeting the criteria.

Among the 33 systems studied, 32 are ‘approved’ systems and one is a ‘traditional’ system, composed of a vertical-flow sand filter equipped with drains.

Data collected are heterogeneous and may be below the quantification limit for some the six physico-chemical parameters analyzed. The latter have been left-censored. One of the recommended statistical methods to consider censored data is the analysis of a generalized log-linear model (Kalbeisch & Prentice 2002; Olivier et al. 2018a).

Such a model has been developed to compare distributions and explain the variations of outlet concentrations by analyzing the impact of four explanatory variables: (i) the treatment group/system, (ii) the estimated loading rate, (iii) the age of the facility at the time of sampling and (iv) the sampling method.

MATERIAL AND METHODS

Data collection and validation

Facilities were selected based on five criteria: (i) the facility has been qualified as compliant with the regulations by a public on-site sanitation service, (ii) it must be installed in a principal residence, (iii) only domestic wastewater is treated, (iv) the sampling point of the facility must be accessible and (v) the owner must be willing to participate (Olivier et al. 2018a).

A total of 1,448 data samples were therefore collected on 246 facilities according to a pre-defined protocol which includes the sampling conditions, the technical characteristics of the facility, the maintenance operations carried out and the possession or non-possession of a maintenance contract (Olivier et al. 2018b). Many practitioners were mobilized during 6 years to carry out the sampling in compliance with the established protocol (Boutin et al. 2017); the number of samples taken per system was liable to vary (actually between 1 and 41) according to the local conditions. Parameters analyzed at the outlet of the treatment system are the following: TSS, chemical oxygen demand (COD), BOD5, ammonium nitrogen (NH4+-N), nitrate nitrogen (NO3-N) and Kjeldahl nitrogen (TKN). They have been analyzed using different French standards methods (Boutin et al. 2017).

Data were compiled and validated following the methodology described in Olivier et al. (2018a). Due to either problems with the sampling conditions (no outflow, improper sampling point, outflow diluted with storm water, sample contaminated with deposits) or with the analysis (late analysis due to transport or laboratory delays, analytical error, inconsistency in analysis results), 11% of data was considered as non-validated and therefore removed from the dataset.

After data validation, the analysis was made on 1,286 data issued from 231 facilities. Thirty percent of those facilities have a maintenance contract.

As mentioned above, data were divided into three groups of biological processes, 13 categories and 33 systems (Table 1).

Among the three groups, 13 different categories were identified to subdivide systems based on common design characteristics.

  • AGSFMs are classified according to the filter media, as its intrinsic properties, although not precisely known for each system, are likely to condition the performance.The biological degradation within the filter bed is performed under aerobic conditions. Hence, conversion of nitrogen forms stops at the nitrification stage except for the constructed wetland (CW) system. Indeed, the CW system is composed of two stages: the first stage is a vertical-flow filter which allows the nitrification to take place and the second stage is a horizontal-flow filter with anoxic conditions allowing the denitrification (Vymazal & Kröpfelová 2008).

  • SAGPs are classified according to the type of media in the bioreactor: fixed (Fx) or fluidized (Fl). Rotating batch contactors (RBCs) are also part of this group. All categories within this group show conditions allowing nitrification and denitrification.

  • ASPs are classified in reference to the classical configuration (three tanks). Then, three other categories are identified: classical without primary settlement tank (WoPST), classical with additional treatment by filtration (WAT), sequencing batch reactor (SBR). All categories within this group show conditions allowing nitrification and denitrification.

The distinction between a treatment system and a specific facility is important. A treatment system refers to a product marketed by manufacturers, except for the vertical-flow constructed wetland. Each system is characterized by technical aspects specific to its category: design, aeration system, number of tanks, depth of the filter, etc. A facility is a specific treatment plant installed to treat wastewater from a given individual house. For a given treatment system, several facilities have thus been sampled.

Figure 1 shows the percentage of samples validated per category and the number of collected samples per system.

Figure 1

Number of validated samples per system for each category (meaning of abbreviations of categories is available in Table 1).

Figure 1

Number of validated samples per system for each category (meaning of abbreviations of categories is available in Table 1).

Figure 1 also shows the heterogeneity of samples by groups and categories. AGSFMs represent 47.5% of the total dataset, SAGPs 30.2% and the ASPs 22.3%.

The number of samples per treatment system is variable: from 1 (S2) to 126 (CW1). Considering the dataset, differences in the results were supposed to be due to a too small sample size when it was less than 13. Eleven systems were therefore removed from the analysis (S2, PB1, Fx2, Fx5, Fx8, RBC1, WoPST2, C2, C3, SBR4 and SBR5). The S3 system has also been removed because only one facility was sampled. The systems' analysis was made on the 21 remaining systems (1,194 samples) and the analysis of the complete dataset (three groups) was made on the 33 systems (1,286 samples).

Data analysis using a generalized log-linear model

The impact of four explanatory variables on the outlet concentrations was investigated considering six physico-chemical parameters (TSS, COD, BOD5, NH4+-N, NO3-N and TKN).

The four explanatory variables are as follows:

  • The treatment groups/systems: 21 systems gathered into three groups. This variable can take three modalities when comparing the three groups and 21 modalities for the systems' comparison.

  • The estimated loading rate: the actual organic load was not measured during the sampling. Therefore, a daily load (in COD) was estimated based on the number of inhabitants, their activities and the theoretical daily load of pollution per person per day (Jönsson et al. 2005). Three classes (= three modalities) of estimated loading rates were created: low (<30%), intermediate (30–70%) and high (>70%).

  • The aging: the age of the studied facilities was between 1 and 12 years. Since 80% of the facilities are under 4 years old, three modalities were used as follows: young (<2 yr), intermediate (2–4 yr) and old (>4 yr).

  • The sampling method: two different sampling methods (= two modalities) were used: ‘grab’ samples and time-proportional 24 h composite samples. The dataset is composed of 778 grab samples and 508 composite samples.

The obtained dataset is heterogeneous (Figure 2 and Table 2) and required robust data processing. The analysis of the dataset distribution for each parameter showed that data are not normally distributed. Therefore, due to the heterogeneity of the dataset, non-Gaussian distribution and censored data (concentration lower than quantification limit), a statistical methodology has been developed based on generalized linear models introduced by Nelder & Wedderburn (1972). A generalized log-linear model was created and has been calibrated from observed data using the maximum likelihood estimation theory that allows taking into account censored data. This model enabled identification of significant effects which have an impact on the distribution of outlet concentrations for the six physico-chemical parameters. Results are analyzed by comparison of median values to the one obtained with the reference modality, the latter being the modality comprising the most data within the data subsets.

Table 2

Concentration ranges of treated wastewater for the 21 systems

 Concentration (mg L1)
 
 TSS COD BOD5 NH4+-N NO3-N TKN 
Average 57 141 23 24 37 30 
Median 17 85 11 28 15 
Minimum 30 
Maximum 4,500 3,900 350 223 199 341 
Number of values 1,187 1,189 654 1,107 1,177 950 
 Concentration (mg L1)
 
 TSS COD BOD5 NH4+-N NO3-N TKN 
Average 57 141 23 24 37 30 
Median 17 85 11 28 15 
Minimum 30 
Maximum 4,500 3,900 350 223 199 341 
Number of values 1,187 1,189 654 1,107 1,177 950 
Figure 2

TSS outlet concentrations (zoom from 0 to 500 mg L−1) observed at the outlet of the 21 systems.

Figure 2

TSS outlet concentrations (zoom from 0 to 500 mg L−1) observed at the outlet of the 21 systems.

The model enabled to determine if modalities are significantly different from the reference using p-value (i.e. probability of wrongly rejecting the null hypothesis) calculation. The null hypothesis H0 is stated as ‘There is no difference between modalities’ and the significance threshold was generally set to 0.1% in order to ensure only predominant effects were retained, except where otherwise mentioned in the results when such a low value was too restrictive.

The model allowed quantification of the effect of the modality on outlet concentrations by providing theoretical medians that were compared to empirical medians (field data).

RESULTS AND DISCUSSION

Descriptive statistical analysis

A significant variation of the concentrations was observed for the six parameters, with maximum concentrations of treated wastewater higher than 3,800 mg L−1 for TSS and COD (Table 2).

Average concentrations are strongly influenced by high values and are up to 3.3 times higher than median concentrations. Therefore, median outlet concentrations have been used to compare datasets.

A strong heterogeneity was observed between the 21 systems due to significant differences between the three treatment groups. Ranges of TSS outlet concentrations for example are important for the 21 treatment systems (Figure 2), with median values varying from 5 mg L−1 (CW1 and Fx1 systems) to 120 mg L−1 (WoPST1 system). However, due to the wide ranges within each system distributions, the statistical analysis was not able to highlight any significant differences in the water quality observed at the outlet of the 21 systems. Therefore, the performances of the treatment systems were analyzed at the group scale (i.e. comparing systems belonging to one group).

Statistical analysis with generalized log-linear model

The statistical analysis was first implemented on the complete dataset to compare the three groups. Then, the model was used to compare systems within each treatment group. Effects identified as significant by the model for all physico-chemical parameters are presented in Table 3. Nitrogen pollution parameters were not compared at the three groups scale because they are not designed to reach the same nitrogen removal performances.

Table 3

Identification of significant effects of different explanatory variables

Group Effect identified by the model Physico-chemical parameters 
Three groups Group TSS COD BOD5 – – – 
Loading rate TSS COD BOD5 – – – 
Sampling method  COD  – – – 
AGSFM System TSS COD BOD5 TKN NH4+-N NO3-N 
Loading rate   BOD5   NO3-N 
Aging      NO3-N 
Sampling method   BOD5    
SAGP System TSS COD BOD5 TKN NH4+-N NO3-N 
Loading rate    TKN NH4+-N NO3-N 
ASP System TSS COD BOD5 TKN NH4+-N NO3-N 
Loading rate    TKN   
Aging     NH4+-N  
Group Effect identified by the model Physico-chemical parameters 
Three groups Group TSS COD BOD5 – – – 
Loading rate TSS COD BOD5 – – – 
Sampling method  COD  – – – 
AGSFM System TSS COD BOD5 TKN NH4+-N NO3-N 
Loading rate   BOD5   NO3-N 
Aging      NO3-N 
Sampling method   BOD5    
SAGP System TSS COD BOD5 TKN NH4+-N NO3-N 
Loading rate    TKN NH4+-N NO3-N 
ASP System TSS COD BOD5 TKN NH4+-N NO3-N 
Loading rate    TKN   
Aging     NH4+-N  

For the complete dataset, three variables (group, loading rate and sampling method) were identified by the model to have a significant effect on COD parameter. However, the number of data available for this parameter was not sufficient to analyze the three effects. Therefore, only combined effects on TSS and BOD5 parameters were studied in this paper.

For the analysis of the 21 systems, two effects (system and loading rate) were frequently identified by the model as variables which have an impact on outlet concentrations. Aging was identified as an explanatory variable for treatment systems of AGSFMs and ASPs.

Several combined effects were identified for a parameter; for example, in the AGSFM group, three effects (system, loading rate and sampling method) had an impact on BOD5 outlet concentration. However, the number of data available for this parameter was not sufficient to analyze combined effects. Therefore, only parameters with a single effect are presented.

The statistical model provided theoretical medians of physico-chemical parameters according to the effects identified. These medians were compared to empirical medians and are presented in Tables 4 and 5. Results obtained with a p-value higher than 0.1% are given for information.

Table 4

TSS and BOD5 theoretical and empirical medians for the three groups: AGSFM, SAGP and ASP

   AGSFM SAGP ASP 
TSS (mg L−1Loading rate <70% 9* (10) 14 (17) 40 (40) 
>70% 15 (16) 23 (24) 64 (61) 
BOD5 (mg L−1Loading rate <70% 3* (5) 6 (8) 17 (14) 
>70% 8 (12) 14 (10) 38 (43) 
   AGSFM SAGP ASP 
TSS (mg L−1Loading rate <70% 9* (10) 14 (17) 40 (40) 
>70% 15 (16) 23 (24) 64 (61) 
BOD5 (mg L−1Loading rate <70% 3* (5) 6 (8) 17 (14) 
>70% 8 (12) 14 (10) 38 (43) 

X (X) = theoretical median (empirical median).

*Reference modality.

Table 5

Theoretical and empirical median concentrations (mg L1) for each treatment system of each group

AGSFM systems S1 CW1 CS1 CS2 RW1 RW2 Z1 Z2 
TSS 8 (9) 4* (5) 23 (26) 10 (12) 31 (33) 13 (13) 36 (26) 26 (21) 
COD 37 (35) 54* (50) 83 (89)
p-value = 0.5% 
54 (65) 130 (140) 108 (103) 100 (87) 104 (94) 
TKN 5* (8) – 27 (29) 5 (6) 62 (79) 52 (48) 20 (23) 31 (39) 
NH4+-N 3* (4) – 23 (25) 5 (6)
p-value = 3% 
57 (71) 44 (45) 17 (24) 27 (32) 
SAGP systems Fx1 Fx3 Fx4 Fx6 Fx7 Fx9 Fl1  
TSS 4 (5) 12 (10) 24* (25) 24 (24) 24 (25) 16 (17)
p-value = 5% 
24 (29)  
COD 61 (57) 111 (97) 111* (108) 111 (103) 111 (101) 111 (90) 111 (112)  
BOD5 2 (5) 13 (13) 13 (10) 13* (13) ** 7 (6)
p-value = 0.5% 
**  
ASP systems SBR1 SBR2 SBR3 WAT1 C1 WoPST1   
TSS 19 (23) 19* (20) 19 (18) 19 (19) 62 (88) 133 (120)   
COD 81 (75) 81* (78) 63 (58)
p-value = 5% 
81 (97) 180 (199) 245 (204)   
BOD5 6 (6) 6* (7) ** ** 37 (60) 28 (22)   
AGSFM systems S1 CW1 CS1 CS2 RW1 RW2 Z1 Z2 
TSS 8 (9) 4* (5) 23 (26) 10 (12) 31 (33) 13 (13) 36 (26) 26 (21) 
COD 37 (35) 54* (50) 83 (89)
p-value = 0.5% 
54 (65) 130 (140) 108 (103) 100 (87) 104 (94) 
TKN 5* (8) – 27 (29) 5 (6) 62 (79) 52 (48) 20 (23) 31 (39) 
NH4+-N 3* (4) – 23 (25) 5 (6)
p-value = 3% 
57 (71) 44 (45) 17 (24) 27 (32) 
SAGP systems Fx1 Fx3 Fx4 Fx6 Fx7 Fx9 Fl1  
TSS 4 (5) 12 (10) 24* (25) 24 (24) 24 (25) 16 (17)
p-value = 5% 
24 (29)  
COD 61 (57) 111 (97) 111* (108) 111 (103) 111 (101) 111 (90) 111 (112)  
BOD5 2 (5) 13 (13) 13 (10) 13* (13) ** 7 (6)
p-value = 0.5% 
**  
ASP systems SBR1 SBR2 SBR3 WAT1 C1 WoPST1   
TSS 19 (23) 19* (20) 19 (18) 19 (19) 62 (88) 133 (120)   
COD 81 (75) 81* (78) 63 (58)
p-value = 5% 
81 (97) 180 (199) 245 (204)   
BOD5 6 (6) 6* (7) ** ** 37 (60) 28 (22)   

X (X) = theoretical median (empirical median) in mg L−1.

*Reference modality.

**Insufficient data – no possible comparison.

Three groups

AGSFMs deliver better outlet quality for the same loading rate, compared to the other groups (Table 4). Indeed, medians calculated by the model are 15 mg L−1 for TSS and 8 mg L−1 for BOD5 for a loading rate close to the nominal load.

ASPs have the highest median concentrations especially when the loading rate is high (64 mg L−1 and 38 mg L−1 for TSS and BOD5 parameters).

Loading rate effect is clearly visible since effluent median concentrations are 1.6 to 2.3 times higher for high loading rate.

For TSS, ASPs medians for both loading rates are higher than the threshold of 30 mg L−1 allowed by the French regulation. For BOD5, ASPs median concentration with a high loading rate is higher than the threshold of 35 mg L−1.

AGSFM systems

The model was used to compare eight different treatment systems (S1, CW1, CS1, CS2, RW1, RW2, Z1 and Z2) according to the explanatory variables. The model identified a single effect ‘system’ for TSS, COD, TKN and NH4+-N parameters (Table 3).

The CW system was not studied for nitrogen pollution parameters (TKN and NH4+-N) because this system allows nitrification and denitrification to take place whereas other systems are only nitrifying.

The TSS parameter has permitted all systems of AGSFMs to be discriminated since they are all different from the reference (CW1). The system with the lowest TSS median outlet concentration is CW1 (4 mg L−1) followed by S1 (8 mg L−1), CS2 (10 mg L−1), RW2 (13 mg L−1), CS1 (23 mg L−1), Z2 (26 mg L−1), RW1 (33 mg L−1) and Z1 (36 mg L−1).

RW1 and Z1 have the lowest removal performances and 50% of their data are higher than the regulatory threshold of 30 mg L−1.

For COD, the model did not identify significant differences between CW1 and CS2 systems. The three systems with the best treated wastewater quality are the same as for TSS (S1, CW1 and CS2), then CS1 and Z1, and the three treatment systems with the lowest removal performances are Z2, RW2 and RW1.

Analysis of nitrification parameters has given the same classification for TKN and NH4+-N: treatment systems with the highest nitrification rate are S1 and CS2, followed by Z1, CS1, Z2, RW2 and RW1. These three last systems have TKN median concentrations higher than 31 mg L−1 meaning that the nitrification is incomplete due to lack of oxygen. Nitrogen parameters are a good indicator of filter health (Liénard et al. 1990): they show the sensitivity of the filter regarding the lack of oxygen and indicate a risk of clogging by the development of bacteria characteristic of anaerobic systems.

The multi-parameter analysis leads to the following classification: S1, CW1, CS2, CS1 = Z1, Z2, RW2, RW1. This may be explained by design differences. Indeed, the traditional sand filter (MEDDE 2014) and the CW system have the largest filter surfaces of, respectively, 5 m2 PE−1 and 4 m2 PE−1. Both coconut shavings systems show different quality and nitrification rates. Indeed, CS2 has better removal performances than CS1, which can be explained by design differences since CS1 is more compact (0.65 m2 PE−1) compared to CS2 (0.85 m2 PE−1) and then receives higher organic loading rate. Rock wool filters have lower removal performances compared to other treatment systems studied. Their surface is the smallest of all studied systems of the AGSFM group (0.26 m2 PE−1). Therefore, filters are highly solicited and generate very poor quality of treated wastewater (Dubois & Boutin 2017). The reduction of surfaces leads to intensive use of the filters (McKinley & Siegrist 2011).

SAGP systems

The model was used to compare seven different treatment systems (Fx1, Fx3, Fx4, Fx6, Fx7, Fx9, Fl1) according to explanatory variables.

The model identified the ‘system’ as a single effect for TSS, COD and BOD5 parameters (Table 3).

The model does not differentiate Fx4, Fx6, Fx7 and Fl1 treatment systems whatever the parameter analyzed. These four systems have the highest median outlet concentrations of TSS, COD and BOD5 (24 mg L−1, 111 mg L−1 and 13 mg L−1 respectively).

For the COD parameter, only the Fx1 system is different from the reference and has the lowest median concentration (61 mg L−1).

All parameters combined, Fx1 has the lowest median concentrations of the SAGP group (Table 5). This treatment system shows the best treated wastewater quality compared to other systems with medians of TSS, COD and BOD5 of 4 mg L−1, 61 mg L−1 and 2 mg L−1 respectively. Then, Fx9 and Fx3 have lower medians than other treatment systems for the TSS parameter. Fx9 has better removal performances than Fx3 for the BOD5 parameter (7 mg L−1 and 13 mg L−1 respectively) with a p-value of 0.5%.

Differences between Fx1, Fx9, Fx3 and the other treatment systems could be explained by the volume of the primary settlement tank and for Fx1 the nature of the fixed media (ribbon) in its clarifier that could prevent resuspension of sludge flocs by nitrogen production and could explain the better performances of Fx1 regarding the other systems.

The multi-parameter analysis leads to the following classification: Fx1, Fx9, Fx3, Fx4 = Fx6 = Fx7 = Fl1.

ASP systems

The model was used to compare six different treatment systems (WoPST1, C1, WAT1, SBR1, SBR2 and SBR3) according to explanatory variables.

The model identified the ‘system’ as a single effect for TSS, COD, BOD5 and NO3-N parameters (Table 3). The interpretation of NO3-N results is difficult without NH4+-N. Therefore, only carbon pollution parameters were analyzed.

For the TSS parameter, only C1 and WoPST1 are different from the reference and have the highest median concentrations (62 and 133 mg L−1 respectively). The analysis of COD and BOD5 parameters leads to the same conclusions.

From Table 5, SBR1, SBR2, SBR3 and WAT1 appear to be the best treatment systems of the ASP group for all parameters. They are all from the SBR and ‘with additional treatment’ categories. It can be explained by the sequential batch, which has a buffering effect on hydraulic loads. The WAT1 system has an additional treatment (sand filter) that can retain more TSS. Then the classification is: SBR3 = SBR1 = SBR2 = WAT1 better than C1 and then WoPST1. The WoPST1 system has the highest medians of TSS, COD and BOD5 compared to all systems studied and this could be explained by the absence of primary settlement tank which has a double function: i) buffering effect on hydraulic loads and ii) storage of biological sludge (Canler et al. 2005; Tscherter et al. 2015).

Overall results

TSS and COD parameters were chosen to compare graphically the 21 treatment systems using median concentrations calculated by the model (Figure 3). This figure summarizes the results of the three analyses of treatment systems done at the group scale. The comparison of all treatment systems with each other was made possible since only one significant effect (treatment system) has been identified by the model for TSS and COD (Table 3).

Figure 3

Theoretical median TSS /COD effluent concentrations for the 21 treatment systems.

Figure 3

Theoretical median TSS /COD effluent concentrations for the 21 treatment systems.

The treated wastewater quality varies among systems. Although all systems studied comply with the French regulation, they deliver variable treated wastewater quality with medians from 4 to 133 mg L−1 for TSS and 36 to 244 mg L−1 for COD.

A group of four treatment systems show good removal performances (median TSS between 4 and 10 mg L−1, median COD between 36 and 61 mg L−1). Three are from the AGSFM group but three different categories (sand, constructed wetland and one system of coconut shavings), and one treatment system is from the SAGP group (fixed bed category).

For TSS, four treatment systems have medians higher than the regulatory threshold of 30 mg L−1: RW1 and Z1 from the AGSFM group and C1 and WoPST1 from the ASP group. For the 13 other systems, the medians are not sufficient to compare them to regulatory thresholds or to draw any conclusions.

CONCLUSION

The statistical model has permitted comparison of median outlet concentrations of carbon and nitrogen pollution parameters taking into account four effects: treatment groups/systems, loading rate, aging, and sampling method.

The analysis of the three groups has highlighted different performances within the groups. The AGSFM group has better treated wastewater quality for TSS and BOD5 compared to the other groups (SAGP and ASP) and ASPs have the highest outlet median concentrations. For the three groups, the effluent quality is degrading as the loading rate increase. For high loading rates (>70%), TSS and BOD5 median concentrations of the ASP group (respectively 64 and 38 mg L−1) exceed the regulatory thresholds.

However, the analysis at the group scale hides heterogeneity that the systems' analysis has been able to identify.

Based on TSS and COD medians comparison, a group of four treatment systems has the lowest median outlet concentrations (S1, CW and CS2 from AGSFM and Fx1 from SAGP). All of them, excepting CW (receiving raw wastewaters), have a large primary settlement tank. Moreover, the three systems of AGSFMs have the largest filter surfaces and hence receive lower organic loading rate. The nature of the fixed media (ribbon) in the Fx1 clarifier, which prevents resuspension of sludge flocs, could explain these better performances. Four systems have outlet median concentrations (WoPST1 and C1 from ASP, and RW1 and Z1 from AGSFM) higher than the TSS regulatory threshold of 30 mg L−1. Between those two groups, 13 systems show variable removal performances with medians between 12 and 26 mg L−1 for TSS and 63 and 125 mg L−1 for COD. Since only four systems out of 21 have good performances and considering that 80% of the studied facilities are less than 4 years old, the situation is worrying.

All these processes have proven their efficiency in collective sanitation. However, this study shows that the treated wastewater quality is highly variable regarding the different on-site treatment systems. In collective sanitation, weekly monitoring and maintenance operations are a guarantee of the systems' performances (Tscherter et al. 2015). In on-site sanitation, even with a maintenance contract, operating activities are rarely performed, which may have a negative impact on the system performances.

The development of new on-site wastewater treatment systems is increasing in France because of their compactness and easy implementation. These new systems are approved by the French regulation based on platform test results of the European Standard NF EN 12566-3. Discussions are underway to lower regulatory thresholds of the French approval procedure based on those results.

ACKNOWLEDGEMENTS

The authors warmly thank about 150 people involved in this study: public on-site sanitation services, French departments, water agencies and French authorities as well as the French Agency for the Biodiversity for their financial support.

REFERENCES

REFERENCES
Arrêté du 7 septembre 2009 fixant les prescriptions techniques applicables aux installations d'assainissement non collectif recevant une charge brute de pollution organique inférieure ou égale à 1.2 kg/j de DBO5 paru au JO du 25 avril 2012 (Decree of 7 September 2009 on technical requirements for individual wastewater facilities responding to a load of organic pollution less or or equal to 1.2 kg/J of BOD5). Journal officiel “Lois et Décrets” – JORF. No. 0098, 25 April 2012
.
Boutin
C.
,
Olivier
L.
,
Agenet
P.
,
Parisi
S.
,
Artuit
P.
,
Branchu
P.
,
Decout
A.
,
Dubois
V.
,
Dubourg
L.
,
Dhumeaux
D.
,
Jousse
S.
,
Leval
C.
,
Mouline
B.
,
Portier
N.
,
Rambert
C.
,
Souliac
L.
&
Szabo
C.
2017
On-site Wastewater Treatment: in situ Assessment of Wastewater Treatment Plants from 2011 to 2016
. .
Canler
J. P.
,
Cauchi
A.
,
Cotteux
E.
,
Graveleau
L.
,
Hyvrard
N.
,
Larrigauderie
A.
,
Meinhold
H.
&
Pujol
R.
2005
Dysfonctionnements biologiques des stations d’épuration : origines et solutions (Biological malfunctions of treatment plants : origins and solutions)
.
Document Technique FNDAE
33
,
99
.
Dubois
V.
&
Boutin
C.
2017
Comparison of the design criteria of 141 onsite wastewater treatment systems available on the French market
.
Journal of Environmental Management
http://dx.doi.org/10.1016/j.jenvman.2017.07.063
.
Jönsson
H.
,
Baky
A.
,
Jeppsson
U.
,
Hellström
D.
&
Karrman
E.
2005
Composition of Urine, Faeces, Greywater and Biowaste for Utilization in the URWARE Model. s.l
.
Chalmers University of Technology
,
Sweden
, p.
49
.
Kalbeisch
J. D.
&
Prentice
R. L.
2002
The Statistical Analysis of Failure Time 415 Data
,
2nd edn
.
Wiley Interscience
,
Hoboken, NJ, USA
.
Liénard
A.
,
Boutin
C.
,
Esser
D.
1990
Domestic Wastewater Treatment with Emergent Hydrophyte Beds in France. Constructed Wetland in Water Pollution Control (Adv. Wat. Pollut. Control N°11)
(
Cooper
P. F.
&
Findlater
B. C.
, eds).
Pergamon Press
,
UK
, pp.
183
192
.
McKinley
J. W.
&
Siegrist
R. L.
2011
Soil clogging genesis in soil treatment units used for onsite wastewater reclamation: a review
.
Critical Reviews in Environmental Science and Technology
41
(
24
),
2186
2209
.
MEDDE
2014
Agréments des Dispositifs de Traitement en France (Approvals of On-site Sanitation Systems in France). http://www.assainissement-non-collectif.developpement-durable.gouv.fr/agrement-des-dispositifs-de-traitement-r92.html (accessed 6 July 2017)
.
Metcalf & Eddy, Inc
.
2003
Wastewater Engineering. Treatment, Disposal and Reuse
,
4th edn
.
McGraw-Hill
,
New York, USA
.
Nelder
J. A.
&
Wedderburn
R. W. M.
1972
Generalized linear models
.
Journal of the Royal Statistical Society
135
,
370
384
.
NF EN 12566-3 + A2
2013
Small Wastewater Treatment Systems for up to 50 PTE. AFNOR publication
.
Olivier
L.
,
Artuit
P.
,
Branchu
P.
,
Decout
A.
,
Dubois
V.
,
Dubourg
L.
,
Dhumeaux
D.
,
Jousse
S.
,
Leval
C.
,
Mouline
B.
,
Portier
N.
,
Souliac
L.
,
Szabo
C.
,
Parisi
S.
&
Boutin
C.
2018b
Assainissement non collectif en France : synthèse du suivi in situ des installations réalisé de 2011 à 2016 (On-site wastewater treatment in France : summary of the in situ assessment carried out from 2011 to 2016)
.
TSM
7–8
,
83
96
.
Tscherter
C.
,
Boutin
C.
,
Caquel
O.
,
Dimastromatteo
N.
,
Dumaine
J.
,
Fernandes
G.
,
Gervasi
C.
,
Parotin
S.
&
Prost Boucle
S.
2015
Ouvrages de Traitement par Boues Actives – Guide D'exploitation (Activated Sludge Processes – Operating Guide)
. .
Vymazal
J.
&
Kröpfelová
L.
2008
Wastewater Treatment in Constructed Wetlands with Horizontal Sub-Surface Flow
.
Environmental pollution series
, Volume
14
.
Springer, Dordrecht
,
The Netherlands
.