Water Practice and Technology sample issue

© IWA Publishing 2014 1

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.001

Methods to accompany and evaluate planning of combined sewer overflow treatment concepts for complex sewer systems K. Klepiszewski, S. Seiffert, M. Regneri and E. Henry Public Research Centre Henri Tudor, 29, avenue J.F. Kennedy, L-1855 Luxembourg. E-mail: kai.klepiszewski@tudor.lu

Abstract Simulation tools are in common use to evaluate combined sewer overflow (CSO) treatment concepts in complex sewer systems. However, the planning of CSO structures in a sewer system is a matter of local constraints, expert knowledge and trial and error. Common standards only provide general recommendations to plan CSO structures and work out management strategies. Additionally, modelling the emissions of complex sewer systems tends to result in comprehensive findings. Although, it is essential to understand local behaviour and interaction of CSO structures in a system to improve local and overall performance there is a lack of tools to illustrate comprehensive simulation results in a simple way. In this context the methods presented here are developed. These include clear illustrations of the as-is state in the catchment using Sankey diagrams to show relevant volume and pollutant flows. Furthermore, loading and treatment indicators are suggested to illustrate local loading conditions and treatment capabilities of CSO structures in relation to the overall system. Additional emission indicators provide information on local emissions and show interactions of CSO structures. The results indicate that the suggested methods contribute to an efficient evaluation of interactions and performances to improve treatment strategies in the planning phase.

Key words: performance indicators, planning complex sewer systems, substance flow analyses, visualization

INTRODUCTION Modelling volume and substance flows in sewer systems is widely applied to plan and evaluate combined sewer overflow (CSO) treatment concepts in complex sewer systems. Nevertheless, distribution of CSO structures in sewer systems, dimensioning of individual storage volume for CSO treatment and setting of CSO structures’ outflows to downstream sewer sections are a matter of local constraints, expert knowledge and trial and error.

Detailed planning of CSO structures is strongly linked to local constraints in catchments and sewer systems. Common standards only provide general guidelines and recommendations to support the planning of CSO structures and work out management strategies. This is reflected in several national guidelines at European level (Germany & Luxembourg: ATV 1992; Great Britain: FWR 1998, France: CERTU 2004, Fanders/Belgium CIW 2012, Austria: OEWAV 2007). Furthermore, none of the standards provide detailed information on how to include the interaction between CSO structures of complex systems into the planning procedure. The dimensioning procedure specified in the German guideline A 128 (ATV 1992) for instance asks for long-term simulation ( 10a rain data) of complex sewer systems to mimic system behaviour and CSO emissions. Further, it demands (i) not to exceed a threshold of the total chemical oxygen demand (COD) load emitted by the sum of all CSOs of the system and (ii) to meet specific mixing ratios of rain weather flow and dry weather flow at individual CSO structures.

Modelling volume and substance flows and resulting emissions of complex sewer systems tends to result in comprehensive findings. Since standard simulation tools for sewer systems are designed to

2

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.001

provide results as required by standards or approving authorities they rarely provide additional tools to assist more detailed analyses of the system behaviour. Due to this, there is a lack of tools to illustrate comprehensive simulation results in a simple way. It is essential to understand local behaviour and the interaction of individual CSO structures in a system to improve local and overall performance of the system.

Against this background the methods presented here were developed for the Luxembourgish water authority to enable an analysis of significant volume and substance flows in complex sewer systems.

Additionally, a set of three indicators for individual CSO structures in the system provide an overview on the local loading conditions, treatment capabilities and emissions with a link to the overall system.

The methods are used to accompany and evaluate the planning of CSO management concepts suggested by consultants which include results of hydraulic and pollution simulations. Due to this, these methods clearly illustrate relevant catchment characteristics and results of long-term simulations by showing volume and substance flows in drainage networks. Additionally, they suggest adapted dristributions of storage volumes as well as volume and substance flows in the network to improve the overall system performance.

METHODOLOGY Data on a sewer system were provided by a study dealing with development and implementation of an optimised predictive real-time control system for a rural sewer network in Luxembourg (Henry et al.

2007). The findings presented here focus on long-term flow and pollution load simulation results for the current system including 8 combined sewer overflow tanks (CSOT) and one CSO (Figure 1).

Figure 1 | Schema of the combined sewer system under investigation.

The simulation is carried out using InfoWorks ICM, 3 years continuous local rainfall data and a calibrated model of the sewer system. Previous extensive monitoring campaigns and model calibration focused on a sufficient modelling of three CSOT in the system to adapt catchment parameters like runoff coefficients, pollution build-up and wash off parameters. The findings were transferred to further sub-catchments with similar characteristics. The main objective of the

3

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.001

calibration was to get reliable data on total CSO volumes and loads. However, the main objective of the evaluation procedures presented here is to allow a relative comparison of simulation results for different scenarios which is required in most of the guidelines mentioned above. This does not necessarily call for calibrated models to provide input data for the evaluation procedures. In a first step, this information was visualized in Sankey diagrams, a special kind of flow diagram where the width of an arrow is proportional to its quantity (Figure 2). Sankey diagrams illustrate energy or material flows to identify inefficiency or saving potentials (Schmidt 2008). In a second step the indicator procedure was applied to evaluate results of a long-term flow and pollution load simulation for the sewer network shown in Figure 1. In this case study, adapted static outflows of individual CSO structures were suggested based on the evaluation of the indicator procedure. The objective of the adaptation was to reduce volume and pollutant emissions of individual CSO structures and the overall system.

Figure 2 | Sankey diagram showing structure of network, overflow volumes and loads, and connected area and PE for the adapted system.

Illustration of substance flows in sewer systems

To gain an overview on the volume and substance flows and the results of a long-term simulation which were distributed in various tables and plans, Sankey diagrams were used. They combine information on the layout of the sewer network, connected population equivalents (PE) and the area contributing to surface runoff with data on the CSO structures like storage volume and outflow to the downstream waste water treatment plant (WWTP). Additionally, they include information on substance flows within the sewer and of overflow volumes and associated COD loads calculated by the long-term simulation. Sankey diagrams have so far been used e.g. for visualization of substance flows in sewer/WWTP/river-systems (Benedetti 2006) or nutrient and energy fluxes in decentral sanitation systems (Campos et al. 2012). In this study, the determinative volume flows in rain weather conditions in the sewer system are represented by the sum of the connected area while the connected PE serve as a substitute parameter for the expected loads. All values (area, PE, overflow volume, overflow COD load, storage volume and throttled outflow) are given as a percentage of the respective totals to enable a quick overview on the fraction any given structure is contributing to the whole system.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.001

4

Indicators for local and global evaluation and improvement of treatment concepts

For this purpose a set of three indicators was calculated for each CSO structure: (i) a loading indicator, (ii) a treatment indicator and (iii) an emission indicators. For individual CSO structures the loading indicator represents the average of three parameters (s. Equation (1)) linked to the local percentage of: 1. Connected contributing area to the contributing area in the total catchment.

2. Connected PE to the PE in the total catchment.

3. Upstream CSO structures outflows to the design wet weather inflow of downstream WWTP. Equation (1) shows the possibility of a different weighting of the loading parameters included in the loading indicator. Due to this, specific boundary conditions weighting factors can be taken into account if necessary. In the case of the study presented here all weighting are set to one.

Iload,n

PEn Acon,n w1 þ w2 þ w3 PEtot Acon,tot ¼ 3 P wi

P

Qout,n,upstream QWWTP

100

(1)

i¼1

Iload,n: loading indicator for CSO structure n [%].

PEn: PE connected to structure n (PE connected to upstream CSO structures not included).

PEtot: PE in the total catchment.

Acon,n: contributing area upstream of CSO structure n (contributing area of upstream CSO structures not included).

Acon,tot: total contributing area in the catchment. ΣQout,n,upstream: sum of outflows of CSO structures upstream of structure n.

QWWTP: design inflow to WWT.P w1,w2,w3: optional weighting factors (in this case: w1 ¼ w2 ¼ w3 ¼ 1).

Subsequently, the normalised loading indicator is calculated to be compared with further indicators in the procedure (Equation (2)) Iload,n 100 IN,load,n ¼ Pm i¼1 Iload,i

(2)

IN,load,n: normalised loading indicator for CSO structure n [%].

Iload,n: loading indicator for CSO structure n. m: number of structures in the catchment.

An additional indicator is taken into account in the procedure to include the expected treatment capability of individual structures in the context of the total system. Similar to the loading indicator the treatment capability indicator is calculated from the average of the percentage of the: 1. Storage volume of an individual CSO structure to the total storage volume available in the sewer system.

2. Outflow to the downstream sewer system to the design wet weather inflow to WWTP. For further evaluation a normalised treatment capability indicator is calculated like for the loading indicator. Finally, three emission indicators are calculated based on the results of a long-term simulation for the sewer system indicating the treatment behaviour of individual structures in relation to the whole sewer system. The three emission indicators applied here represent the percentage of

5

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.001

1. CSO volume of an individual CSO structure to the total CSO volume of all structures in the system.

2. COD load in the CSO of an individual CSO structure to the total COD load emitted by all structures in the system.

3. Ammonia load in the CSO of an individual CSO structure to the total ammonia load emitted by all structures in the system.

Both quality-based emission indicators in addition to the CSO volume emission indicator provide a comprehensive overview of local CSO emissions in realtion to the behaviour of the total system. In the procedure the COD emission indicator represents a pollution parameter that originates from dry weather flow and surface runoff during rain events. In contrast to the COD emission indicator the ammonia emission indicator represents a substance almost exclusively originating from dry weather flow. Consequently the different emission indicators reflect specific catchment conditions (contributing area, number of inhabitants etc.) as well as treatment capability of local CSO structures.

Furthermore, all parameters included in the calculation indicators can be weighted as shown in Equation (1). This might be necessary to account for local conditions like sensitive surface waters that require a limitation of specific emissions, for example, peak overflow rates.

RESULTS AND DISCUSSION Illustration of substance flows in sewer systems

As can be seen in Figure 2, there are two groups of catchments contributing more or less equally to the hydraulic and substance load of the WWTP: the catchments of Eschdorf and Heiderscheid on the one hand, on the other hand the remaining six CSOT structures.

Eschdorf and Heiderscheid are responsible for 50.8% of overflow volume and 47.9% of overflow COD load while their share of the connected area and PE is only 44.2 and 31.9%, respectively.

The highest overflow volumes occur in Eschdorf/West and Kaundorf.

Indicators for local and global evaluation and improvement of treatment concepts

Figure 1 illustrates the indicators for the CSO structures in the as-is sewer system. All structures in the system except for the CSOT Eschdorf/East have a loading indicator of 10% or below. The high loading indicator of ca. 40% for Eschdorf/East is caused by the number of PE connected to the structure and the loading of the structure by the outflow of the upstream CSO Eschdorf/West.

Due to the high outflow of 110 l/s CSO Eschdorf/West has the highest treatment capability indicator representing about 45% of the total treatment capability of the system. This shows the significant contribution of the outflow components to the indicator calculation. The treatment capability indicators for other structures are in a range of 5 to 10%. The most relevant emissions of the sewer systems are caused by the CSO structures in Kaundorf, Eschdorf and Heiderscheid. More than 90% of the total CSO volume in the system is emitted there.

Since the structures in the system under observation are already built the only possibility to improve the overall system performance is to adapt the outflows of the individual structures. Due to this, the possibility to increase the treatment capability of the structures is quite limited. Furthermore, the maximum combined sewage inflow to the downstream WWTP of 57 l/s should not be exceeded. Consequently, the adapted system is based on a stepwise variation of the outflow of selected structures and an evaluation of the resulting indicators.

The outflow of CSO Eschdorf/West was reduced to 27 l/s to decrease the loading indicator of CSOT Eschdorf/East in an adapted model of the sewer system. Furthermore, the outflows of the CSOT Nocher-Route, CSOT Dahl and CSOT Nocher were reduced to be able to increase the outflows

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.001

6

Figure 3 | Normalised loading and treatment capability indicators as well as emission indicators of the CSO structures in the sewer system for the as-is system.

of CSOT Kaundorf, CSOT Büderscheid, CSOT Eschdorf/West and CSOT Heiderscheid. This would be possible since the outflows of the CSOT Nocher-Route, CSOT Dahl and CSOT Nocher is conveyed by pumps to the downstream sewer system. The outflows of CSO structures for the as-is system and for the adapted system are shown in Table 1. Figure 4 shows the outcome of the indicator procedure. It is based on the results of a long-term simulation for the adapted system. For the CSO structures in Kaundorf, Eschdorf/West, and Heiderscheid the comparison shows that the different indicators are much more balanced than for the as-is system. Although the treatment capability indicator could be increased for CSOT Eschdorf/East from 10 to 15% in the adapted system, the loading and the emission indicators still indicate a low performance of the structure. Another possibilty to increase the treatment capability indicator and consequently to reduce emission indicators of CSOT Eschdorf/ East would be to increase the storage volume.

The results of the long-term flow and pollution load simulation for the adapted system show that the total emissions of the system can be reduced significantly by applying the outflows shown in Table 1.

Hence, the yearly volume and ammonia load emitted by the total system could be reduced by 12%.

Due to a more efficient retention of first flushes the yearly COD load emissions could be reduced by about 15%.

Table 1 | Outflows of CSO structures in the as-is and the adapted system Outflow to downstream system (Qout) CSO structure

Kaundorf

System as-is

Adapted system

(l/s)

(l/s)

6.0

9.0

Nocher-Route

6.0

4.0

Nocher

6.0

2.0

Dahl

6.0

4.0

Büderscheid

4.0

6.0

Goesdorf

9.0

9.0

110.0a

27.0a

Eschdorf/Ost

11.0

12.0

Heiderscheid

9.0

11.0

57.0

57.0

Eschdorf/West

Sum a

Not included in sum since ouflow not directly discharged to downstream WWTP.

7

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.001

Figure 4 | Normalised loading and treatment capability indicators as well as emission indicators of the CSO structures in the sewer system for the adapted system.

CONCLUSION The application of the Sankey diagram gives a good overview on characteristics of the catchment and the CSO structures as well as on the substance flows in the system. Furthermore, the indicator procedure shows that loading, treatment capability and emission indicators calculated for all CSO structures within the system give a good overview on the overall treatment and emission characteristics of the system and indicate potentials or needs of improvement at individual CSO structures.

Besides, the indicators can also be calculated from detailed results of flow and pollutants modelling to support the evaluation of modelling results and improve the overall CSO treatment concept in complex sewer systems. It can be concluded that the procedure presented here as well as the suggestions to illustrate the results can significantly contribute to an improvement and a standardisation of the planning and evaluation of CSO treatment concepts in complex sewer systems. Especially, the indicator method enables planners to do system evaluations without sophisticated software tools.

ACKNOWLEDGEMENT This work was funded by the Luxembourgish Water Administration (Administration de la Gestion de l’Eau) and supported by the water board Syndicat Intercommunal de Dépollution des Eaux résiduaires du Nord (SIDEN). The authors would also like to thank Mr. Flies (Schroeder et Associés, Luxembourg) and Mr. Kroll (Aquafin, Belgium) for their advice.

REFERENCES ATV 1992 Richtlinie für die Bemessung und Gestaltung von Regenentlastungsanlagen in Mischwasserkanälen (Standards for the Dimensioning and Design of Stormwater Overflows in Combined Wastewater Sewers). ATV-Arbeitsblatt 128, German Water Association for Water, Wastewater and Waste, Hennef, Germany.

Benedetti, L. 2006 Probabilistic Design and Upgrade of Wastewater Treatment Plants in the EU Water Framework Directive Context. PhD Thesis, Ghent University, Belgium, pp. 304.

Campos, L. C., Jain, V. & Schuetze, M. 2012 Simulating Nutrient and Energy Fluxes in Non-networked Sanitation Systems. 2nd Faecal Sludge Management Conference, Durban, South Africa, 29.-31.10.2012.

CERTU 2004 La ville et son assainissement – Principes, méthodes et outils pour une meilleure intégration dans le cycle de l’eau.

(The City and its Sewerage – Principles, Methods and Tools for a Better Integration in the Water Cycle). Centre d’Etudes sur les Reseaux, les Transports, l’Urbanisme et les constructions publiques, Lyon, France.

8

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.001

CIW 2012 Code van goede praktijk voor het ontwerp, de aanleg en het onderhoud van rioleringssystemen (Code of good practice for the design, construction and maintenance of sewer systems). Werkgroep Waterzuivering van de Coördinatiecommissie Integraal Waterbeleid, Flanders, Belgium.

FWR 1998 Urban Pollution Management Manual (UPM). A Planning Guide for the Management of Urban Wastewater Discharges During Wet Weather. 2nd ed. Foundation for Water Research, Marlow, UK, report FR/CL0009.

Henry, E., Klepiszewski, K. & Schosseler, P. 2007 Modelling of a combined sewer system to support planning decisions and initialise the application of a RTC System. Proceedings of the 6th NOVATECH, Lyon, France.

OEWAV 2007 Richtlinien für die Bemessung von Mischwasserentlastungen (Guideline for the dimensioning of combined sewer overflows). Regelblatt 19. Österreichischer Wasser- und Abfallwirtschaftsverband, Vienna, Austria.

Schmidt, M. 2008 The sankey diagram in energy and material flow management. Journal of Industrial Ecology 12 (1), 82–94.

© IWA Publishing 2014 9

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

Revised design criteria for stormwater facilities to meet pollution reduction and flow control requirements, also considering predicted climate effects T. Larma and H. Almb a

Corresponding author. StormTac, Boo Strandv. 1J, S132 36 Saltsjö-Boo. E-mail: thomas.larm@stormtac.com

b

SWECO Environment, P.O. Box 34044 S100 26 Stockholm. E-mail: henrik.alm@sweco.se

Abstract There is a need to revise existing design methods for stormwater pollutant treatment, flow transport and detention facilities. The aim is to increase the accuracy in predicting the performance compared with design only based upon areal and volumetric functions and to optimize design by considering more site-specific data, receiving water quality criteria and forecasted climate effects. During the latest years, flow proportional concentration data from in- and outlets from wet ponds and constructed wetlands, have been compiled. Furthermore, other kind of data from the specific facilities have been compiled, such as areas, volumes, proportion of vegetation, outlet design details and length:width ratio. The parameters are used to revise design methods and are implemented in the operative stormwater and recipient software model StormTac. Design criteria and parameters for calculating design flow and sizing required detention volume are also presented. The climate effects on some of the studied parameters, e.g. design flow and inlet concentration, are discussed. The paper presents the climate factor based upon the hypothesis that it is a function of the design rain duration and reoccurrence time.

Key words: climate factor, design criteria, flow detention, StormTac, stormwater, treatment

INTRODUCTION The most commonly used design methods for stormwater treatment are based on volume or area related correlations (Larm & Hallberg 2008). Over the last years, the following-up of treatment systems for stormwater has been improved through consistent flow proportional sampling. This has highlighted the need for updated tools for designing ponds, wetlands and detention basins regarding site-specific conditions.

Larm & Hallberg (2008) evaluated these methods using data from 18 stormwater facilities in Norway and Sweden. One of the conclusions was that consideration of a minimum concentration at the inlet must be taken into consideration. Larm & Hallberg (2008) also discussed the uncertainty in the overall design and the impact of vegetation in stormwater systems. Also Persson & Pettersson (2009) and Pramsten (2010) considered these methods to be insufficient; the methods need to be complemented by taking into account more site-specific parameters.

The aim of this paper is, by evaluating new data, to develop site-specific parameters to improve the design of stormwater treatment facilities. In this paper, data are presented for suspended solids (SS), phosphorus (P), copper (Cu) and zinc (Zn). These 4 substances are selected since there are much available data and since these are generally of priority in different countries used as water quality criteria and as basis for designing stormwater treatment facilities.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

10

The updated design methods with these site-specific parameters can be used to design new facilities or to re-design existing facilities to meet requirements.

METHOD To develop site-specific parameters, data from total of 46 facilities (20 Scandinavian, one Irish and 25 American) have been processed in the compilation (Larm & Hallberg 2008; StormTac 2013). The following data is available for each facility: watershed area, runoff coefficient, permanent water area and water volume, permanent water area/reduced watershed area, permanent water volume/average runoff volume, surface loading, percentage of vegetation, bypass or not, treatment effects, the inlet and outlet concentrations.

The studied facilities are implemented in the watershed management model StormTac (Larm.

2000). The model is based on the five ‘boxes’ Runoff, Pollutant transport, Detention, Stormwater treatment and Receiving water, see Figure 1.

Figure 1 | Simplified flowchart of the watershed management model StormTac.

Revised design criteria ‘Runoff’ and climate change

The design flow Qdim is calculated in Equation (1), developed by and presented in Larm (2013). It is used for the design of storm sewer, concrete channels, ditches etc.

Qdim ¼ Qdimþ þ fc i ws As , where: Qdim Qdimþ fc i tr N w As The

(1)

Design flow (l/s).

Additional inflow i.e other constant, pumped inflow or base flow Qb. Climate factor.

Rain intensity (l/s/ha) for specific value of tr and N, tr . 10 min.

Design rain duration (min).

Reoccurrence time (years).

Specific runoff coefficient.

Specific watershed area (ha). design rain duration tr is calculated in Equation (2).

tr ¼ L=(60v) L Length, transport distance (m). v Mean water velocity (m/s).

(2)

11

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

If there are physical flow limitations in the transport system upstream the design point can be considered in StormTac. It is common in Sweden that 10 min rain duration always is used when calculating design flows. This often results in oversized transport systems and too small detention volumes. The design rain duration should be the calculated transport time (Equation (2)) if it is longer than, or at minimum 10 min (Swedish Water 2011). Chosen N depends on how often the transport system can be accepted to be flooded.

Climate factor fc

The climate factor fc is the ratio between the expected future and the present design rainfall intensities. In northern Europe more frequent and more severe storms are expected and the design intensities are expected to increase with a factor of 1.1–1.5 within the next 100 years (Arnbjerg-Nielsen 2008). According to the recommendations by Swedish Water Association, estimated short-term precipitation will increase by a factor of 1.05 to 1.3 in Sweden, while annual runoff volume is estimated to increase by a factor of 1.1– 1.2 (Swedish Water 2011). This is based on the climate scenario A2 and SMHI recommendations.

Based on extrapolation of annual precipitation increases from data from SMHI 1860–2012, the annual precipitation will increase from the period 1961–1990 to 2071–2100 corresponding the climatic factor 1.20, and from today (2013) 1.15.

The rain reoccurrence time and duration may affect the size of the climate factor (Arnbjerg-Nielsen 2008; Swedish Water 2011). In order to forecast the impact of climate effects on design flows and flow detention volumes, we need to calculate the climate factors on the basis of case specific design return periods and durations. We have, based on data from Swedish Water (2011), SMHI and data compiled

by Arnbjerg-Nielsen (2008), developed draft equations for calculating the climate factor based on these parameters. The equations show that the climate factor increases with increasing return period and decreases with increasing duration.

The Danish study recommends the use of different factors for different reoccurrence time, i.e. fc ¼ 1.2 for N ¼ 2 years, fc ¼ 1.3 for N ¼ 10 years and fc ¼ 1.4 for N ¼ 100 years, while Swedish recommendations suggest the use of fc ¼ 1.05 to 1.3 depending on regional climate conditions (Swedish Water 2011). Based on the Danish correlations between climate factor and return period a similar Swedish curve was adapted by a parallel shift downwards of the Danish curve. A trend line for a logarithmic function provided the best match to data, using the minimum value at N ¼ 1 is 1.05 (assumption) and the maximum value at N ¼ 100 is 1.3 (Larm 2013). Based on these Danish data, shares (Ktr) of the climate factor ( fc,tr) for different durations (1–12 h) are calculated, which assumed that the proportion of 1.0 is used for shorter durations than 1 hour, according to what the data shows. The latter means that no reduction factor is made for durations up to 1 hour. Over the duration of one hour is thus a reduced climate factor used. A trend line was created for data in scenario A (Arnbjerg-Nielsen 2008) and it is assumed that its function provides the effect of duration on the climate factor (Larm 2013).

We obtain the following relationship between climate factor, the return period and duration, valid for the Swedish climate conditions (Larm 2013), see Figure 2.

For the 10-year storm (which is common design in Sweden) and from a small area with a design flow time of 10 min, the climate factor 1.15 can be used for the design of e.g. stormwater sewers.

For a detention basin with the design duration of 5 hours, the climate factor of 1.10 can be used, see Figure 2.

‘Detention’

Facilities may need to be designed with a capacity for detention of intense flows to reduce flow due to capacity of the transport system downstream to prevent flooding. The detention volume (Vd) can be sized for a rain with a particular return time N (Larm 2013). Equation (3) and (4) were developed by

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

12

Figure 2 | The overall climate factor (fc) and its dependence on both the return period N (years) and duration tr (h). Larm (2013).

and presented in Larm (2013). The design discharge (outflow Qout) from a detention basin control required detention volume, which is the maximum volume between the design inflow (Qdim) and outflow at the rainfall duration (tr) that gives the maximum volume (Vd) at the design return period, Equation (3).

The design outflow, Equation (4), is not the maximum but the mean outflow (Swedish Water 2011).

Vdmax ¼ 60 tr Qdim Qout,ave , where: Qout,ave

Qout fQred , where

(3) (4)

Maximum required detention volume (m3).

Vdmax Qout,ave Design outflow, mean outflow (Swedish Water 2011) (l/s).

Qout Maximum outflow (l/s).

Factor for reducing the design outflow considering that the outflow is not at a maximum fQred other than at maximum detention level. Normally: 2/3, if flow regulator: 0.95, if pumped outflow: 1.0.

It is common in Sweden to generally design flow detention volumes using the rain duration 10 min.

This results in too small detention volumes with more frequent floodings. According to Swedish Water (2011), the rain duration together with the specific rain intensity for that duration that result in the largest detention volume Vdmax shall be used, see Equation (3). For small outflow values, very long rain durations shall be used, up to 24 hours (Swedish Water 2011).

‘Pollutant transport’

Stormwater discharge is identified as one of the major pollutant emissions in urban areas (Alm et al.

2010). There is a correlation between annual mean concentration and specific landuse in the catchments (Alm et al. 2010). The specific concentrations in stormwater water vary between different rain occasions (event mean concentration), during the specific rainfall occasions and between the different substances. StormTac uses annual mean data to cope with these variations. These ‘standard concentrations’ per land use are used to calculate specific concentrations at the inlet and pollution loads on recipient. With the expected climate change, the total precipitation will increase (in Sweden), maximum flows will increase and the period without precipitation will be longer. This is expected to result in higher peak concentrations and event mean concentrations, and an increased total load (Sharma et al. 2011) which may lead to that existing treatment facilities are undersized.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

13

Table 1 | Standard concentrations of stormwater and base flow for selected substances (total fractions) and land uses

Land use

Road 10,000 vehicles/day

Storm water P

Cu

Zn

SS

Base flow P

Cu

Zn

SS

mg/l

μg/l

μg/l

mg/l

mg/l

μg/l

μg/l

mg/l

0.18

38

164

87

0.052

13

77

25

Parking

0.10

40

140

140

0.029

11

47

35

Residential area

0.20

20

80

45

0.058

5.5

27

11

Terraced house area

0.25

25

85

45

0.073

6.9

28

11

Multi-family area

0.30

30

100

70

0.087

8.3

33

17

Downtown area

0.28

22

140

100

0.081

6,1

47

25

Industrial area

0,30

45

270

100

0.087

12

90

25

Park grounds

0.12

15

Forest

0.035

6.5

25

49

0.035

4.1

8,4

12

15

34

0.030

4.0

10

1.5

Agricultural property

0.22

14

20

100

0.22

14

20

100

Meadow

0.20

15

30

45

0.17

9.2

20

2,0

‘Storm water treatment’

With Equation (5) the permanent pool area for a wet pond and a constructed wetland is calculated (Larm 2000; Persson & Pettersson 2009) Ap ¼ wAKAw

(5)

Area of permanent pool [m2].

Ap KAw Regression constant, normally 150 (70–400) for wet ponds and 300 (100–800) for wetlands (StormTac, 2013).

Figure 3 presents the correlation between KAw (¼Ap/Ared ¼ Ap/Aw) and the reduction efficiency of studied substances from the studied treatments facilities, without using other site-specific parameters.

The fit of the data for the studied substances was less good (regression coefficients R 2 , 0.4), see Figure 3, indicating the influence of other parameters on treatment efficiency.

In Equation (6) the reduction efficiency is calculated based on the correlations in Figure 3, but complemented with different factors, representing included site-specific parameters (StormTac, 2013).

RE ¼ [k1 ln Ap =Aw þ k2 ] fCin fveg fbypass fVd fCirr ftemp fshape RE k1, k2 f Cin veg bypass Vd Cirr temp shape

(6)

Reduction efficiency for wet ponds and constructed wetlands [%].

Regression coefficients for each pollutant. factor. inlet concentration. vegetation. bypass. detention volume. irreducible concentration. temperature. shape.

The treatment effect can also be a function of permanent facility water volume (Vp) and the average runoff volume (Vr). Vp/Vr can in StormTac be used to replace Ap/Aw in Equation (6). However, generally the fit to data was better for Ap/Aw.

14

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

Figure 3 | Diagram of treatment effects based on sampling and trendline empirical relationships Ap/Ared, where Ared Âź Aw. Blue squares show data from the Scandinavian facilities, black dots show facilities from U.S. and one facility from Ireland.

Inlet concentration ( fCin)

The inlet concentrations Cin affect the relative treatment effect, see Figure 4. From the presented diagram of SS, one observation is that the reduction efďŹ ciency is over 80% for all facilities with higher inlet concentration than 100 mg/l.

15

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

Figure 4 | Diagram of treatment effects based on sampling and trendline empirical relationship of inlet concentrations Cin. Blue squares show data from the Scandinavian plants, black dots show plants from U.S. and one facility from Ireland.

16

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

Irreducible concentration ( fCirr)

The minimum outlet concentration, the ‘irreducible concentration’, is affected by incoming content and internal processes in plants (decomposition of plants, leakage from the bottom due to lack of oxygen, the exchange with sediment, stirring sediment because benthic animals, etc). In StormTac, the reduction efficiency is adjusted so that not less than its minimum concentration is obtained at the outlet. The model contains relationships for each subject between the percentage of vegetation and the irreducible concentrations. However, it is possible to unlock this restriction if it is believed possible to achieve lower levels by adapting the choice of plants or add filters or the like. The irreducible concentrations have been estimated from outflow concentration data from the studied facilities.

These are Cirr (P) ¼ 20–30 ug/l, Cirr (Cu) ¼ 6–7 ug/l, Cirr (Zn) ¼ 14–25 ug/l and Cirr (SS) ¼ 5–10 mg/l.

Vegetation ( fveg)

Increased proportion of vegetation of the facility area is expected to provide higher treatment effects (Larm & Hallberg 2008; Persson & Pettersson 2009). The vegetation reduces water velocity and stops particles and thus increases the effect of the sedimentation process. Vegetation also reduces resuspension of particles and provides an uptake of nutrients and metals that can be separated if the vegetation is harvested. Vegetation also provides large areas for microorganisms and contributes to the uptake of pollutants. The pollutants are mostly taken up in the roots where the contaminant concentration is highest (SMRC 2012). The roots oxygenate even surrounding sediments which increase microbial uptake of pollutants (Stottmeister et al. 2003). More data is needed and the relationship is uncertain, but an increase in treatment effect was observed for Zn, Cu and SS. The correlations indicate a negative effect of P, but this may be due to the influence of other parameters. An assessment is that vegetation has no effect on phosphorus. How the facilities are maintained is considered to play a major role, in some facilities have no harvest of vegetation occurred while harvest has taken place in others.

Bypass ( fbypass)

Pollutants in bypassed flows (overflow) are not treated within the facility. Bypass results in an increased pollutant reduction efficiency (%) within the facility since the flow to the facility is decreased. However, the total load to the recipient may increase (Vikström et al. 2004; Pramsten 2010). The part of the total yearly runoff volume that bypasses (untreated) the facility is calculated in StormTac. The developed flow model calculates the share of bypass of the total volume from this historical precipitation data, calculated time of concentration (minutes) and the bypassed flow divided by the reduced watershed area, i.e. the reduced flow Qred (l/s, hared).

Detention volume ( fdet)

There is often a need of both flow detention and pollutant treatment which can be achieved in the same facility (Whipple & Hunter 1981). By limiting the outflow an increased detention time and sedimentation time during the runoff events is generated, although it is generally accepted that most of the treatment occurs between the runoff events. This regulation implies that a greater portion of the facility is involved in the treatment; better mixing occurs and less risk of short circuit currents occurs. This might indicate a correlation with fshape.

A feature has been developed for use in StormTac based on data from the New Jersey Stormwater Best Management Practices Manual (2004), see Figure 5. The figure shows the treatment efficiency of SS as a function of Vp/Vr. There are curves for various detention times. The treatment efficiency is

17

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

Figure 5 | Reduction efficiency of SS as a function of Vp/Vr. New Jersey Stormwater Best management Practices Manual (2004).

then increased from the lower curve up to the percentage specified in the upper curve corresponding to a certain detention time.

Temperature ( ftemp)

To adapt the pollutant calculations for colder regions the runoff coefficients can be increased in StormTac to include the effects of decreased evaporation. Furthermore, the evaporation from upstream lakes and water courses in the runoff and the recipient model can be decreased. Also, the regression coefficient for calculation recipient lake pollutant concentrations using the OECDmodel can be changed (lower x and higher y). These adaptions shall be further studied by comparing data from colder and warmer regions.

From data from the studied facilities, the difference in reduction efficiency (%) per degree Celcius can be calculated in StormTac. The yearly average water temperature has been estimated for the case studies. For a new calculation the yearly average air temperature is input data and a factor for increased or decreased reduction efficiency is calculated from an empirical equation from the case studies.

Shape ( fshape)

From the case studies the mean form as length:width ratio was estimated. Input is an estimated ratio and output is a factor to increase the reduction efficiency for larger ratio or to decrease it for smaller ratio. The StormTac equations are based on data of hydraulic efficiency for different shapes from Vikström et al. (2004), assuming the same fraction for all parameters due to small differences between substances.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

18

RESULTS AND DISCUSSIONS The errors per parameter have been investigated and the parameters with most impacts are presented.

Table 2 presents compiled errors for the studied facilities per parameter and substance, and the average error per parameter. The data in Table 2 indicates that there is no sufficiently good match for any single parameter, therefore a combination of parameters need to be studied and used.

Table 2 | Correlation (R 2-valuee) between studied single parameters and measured reduction efficiency Parameter

P

Cu

Zn

SS

Mean

Ap/Ared

0.35

0.20

0.28

0.39

0.31

Cin

0.12

0.55

0.10

0.40

0.29

Vp/Vr

0.09

0.30

0.26

0.33

0.25

Vegetation

0.14

0.073

0.054

0.042

0.077

For detention volume, temperature, bypass and shape no regression coefficients could be developed because of the context and type of data used.

The strongest correlation was generally for the areametric correlation (Ap/Ared) for each substance.

Vegetation gave weak indications of increasing treatment efficiency with increasing percentage vegetation.

The data also shows that there are irreducible concentrations from stormwater ponds and wetlands.

These irreducible concentrations are interesting to compare with guidelines and environment quality standards deciding which treatment technique to be used. Furthermore, maximum treatment effects are identified and used.

Uncertainty remains high for individual parameters, so it is considered relevant to consider more parameters than not doing it. Relatively much data was used, but more data is needed in future to update the correlations.

The results can be used to design ponds and wetlands regarding a specific parameter or to identify which type of storm water treatment or combination of treatments facilities needed.

CONCLUSIONS The paper presents a selection of design criteria for flow transport, flow detention and pollutant treatment of stormwater. The model is updated with the optional climate factor. The relations need to be reviewed and revised regularly in accordance with current knowledge from SMHI and the Swedish Water’s forthcoming recommendations.

The following parameters have been selected, and can now be modelled in various combinations in StormTac for designing ponds and wetlands, and to compute their treatment effects.

• • • • • • • • •

Ap/Ared (area related).

Vp/Vr (volume related).

Cin (inlet concentration).

Cirr (irreducible concentration).

Detention volume.

Vegetation (wetland plants, plants in the water).

Bypass.

Temperature.

Shape.

19

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.002

The strongest influences on the reduction efficiency are the relationship between permanent pool area and the reduced sub-watershed area (Ap/Ared), and the inlet concentration (Cin).

The following parameters are to be investigated further because of high uncertainty: the effects of vegetation, the presence of a detentions volume and bypass.

There is also a need for a statistical analysis of the various parameters influence on the treatment efficiency, to quantify uncertainty. For the latter, it is planned some form of multi criteria analysis.

Although there is considerable uncertainty for each parameter separately studied, the reduction efficiency can be explained by the influence of site-specific conditions and the updated design methods with these site-specific parameters can now be used to design new facilities or to re-design existing facilities to meet requirements.

A comparison has been performed with data from the wet ponds Ladbrodammen and Tibbledammen (Alm et al. 2010) which showed a significantly better agreement between calculated and measured treatment effects when area and volume relationships were supplemented with the sitespecific parameters. Further evaluation will take place and data and the empirical equations will be continuously updated.

REFERENCES Alm, H., Banach, A. & och Larm, T. 2010 Occurrence and treatment of substances of priority, metals and some other substances in stormwater. Swedish Water Development, Report 2010-06. (In Swedish) Arnbjerg-Nielsen, K. 2008 Quantification of climate change impacts on extreme precipitation used for design of sewer systems.

Paper in 11th international conference on urban drainage, Edinburgh, Scotland, UK, 2008.

Larm, T. 2000 Watershed-based design of stormwater treatment facilities: model development and applications. PhD Thesis, Dep Civil & Environmental Engineering, KTH, Stockholm, Sweden.

Larm, T. 2013 Design of facilities for detention of stormwater flows. PM StormTac, v. 2013-02. (In Swedish) Larm, T. & Hallberg, M. 2008 Design methods for stormwater treatment – site-specific parameters. 11th International Conference on Urban Drainage, ICUD, Edinburgh, Scottland, UK, 2008.

New Jersey Stormwater Best management Practices Manual, February 2004, page 9.11-4.

Persson, J. & Pettersson, T. J. R. 2009 Monitoring, sizing and removal efficiency in stormwater ponds. E-water.

Pramsten, J. 2010 Reduction efficiency in stormwater ponds in relation to pond volume, overflow and inflow pollutant concentration. Vatten 66, 99–111, Lund, 2010. (In Swedish).

Sharma, A. K., Vezzaro, L., Birch, H., Arnbjerg-Nielsen, K. & Mikkelsen, P. S. 2011 Effect of climate change on stormwater characteristics and treatment efficiencies of stormwater retention ponds. Paper in 12th International Conference on Urban Drainage, Porto Alegre/Brazil, September 11–16, 2011 SMRC 2012 Technical Note #53 from Watershed Protection Techniques. 1 (4), 210–213. Pollutant Dynamics Within Stormwater Wetlands: Plant Uptake. The Stormwater Manager’s Resource Center (SMRC) Web site, http://www. stormwatercenter.net/. Article 92.

Stottmeister, U., Wiessner, A., Kuschk, P., Kappelmeyer, U., Kästner, M., Bederski, O., Müller, R. A. & Moormann, H. 2003 Effects of plants and microorganisms in constructed wetlands for wastewater treatment. Biotechnology Advances 22 (1–2), 93–117. Elsevier.

Swedish Water 2011 Precipitation data for the design and analysis of sewer systems. Swedish Water Report P104. (In Swedish) Vikström, M., Gustafsson, L.-G., German, J. & och Svensson, G. 2004 The reduction efficiencies of stormwater pondsinfluensing factors and evaluation methods. VA-FORSK rapport 2004–11.

Whipple, W. & Hunter, V. J. 1981 Settleability of urban runoff pollution. Water Pollution Control Federation Journal 53, 1726–1731.

Yu, S. L., Kuo, J.-T., Fassman, E. A. & Pan, H. 2001 Field test of grassed-swale performance in removing runoff pollution.

Journal of Water Resources Planning and Management, May/June 2001. 168–170.

© IWA Publishing 2014 20

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

French airport runoff pollution management (water and sludge): toward a new approach based on constructed wetlands? Case of Aéroports de Paris – Orly (France) Philippe Branchu1, Laetitia Gres2, Frederic Mougin3, Martin Le Blanc4, Emmanuelle Lucas4 and Benoit Mars5 1 Corresponding author. CETE Ile de France – ERA N°35 du LCPC – 12 rue Teisserenc de Bort – F 78190 Trappes – France.

E-mail: philippe.branchu@developpement-durable.gouv.fr 2

Laboratoire d’Aéroports de Paris - Bâtiment 631, Orly Sud 103, F-94396 ORLY Aérogares cedex, France.

E-mail: laetitia.gres@adp.fr

3

Aéroports de Paris – Aires Aéronautiques -Bâtiment 640, Orly Sud 103, F-94396 ORLY Aérogares cedex, France.

E-mail: frederic.mougin@adp.fr 4

Lyonnaise des Eaux – Centre Technique Usines Assainissement – Centre Régional Ile de France Sud – 51 avenue de Sénart, BP29, F-91230 MONTGERON, France. E-mail: emmanuelle.lucas@lyonnaise-des-eaux.fr

5 Service Technique de l’Aviation Civile (STAC) – Dept Aménagement, capacité, environnement – 9 av. du Docteur Maurice Gryfongel – BP 53735 – F 31037 Toulouse Cedex 1 – France. E-mail: benoit.mars@aviation-civile.gouv.fr

Abstract One of the environmental concerns for airport managers is stormwater management. Three kinds of contamination occur at airports: (1) chronic contamination, (2) seasonal contamination associated with de-icing procedures and (3) accidental pollution. At Orly Airport, a stormwater treatment plant (STEP) is devoted to removing chronic pollution. About 4 million m3 of water are processed yearly by the STEP, producing about 50–100 t of sludge. Mainly contaminated by petroleum hydrocarbons (Hc), it is removed as final wastes. During the winter, the STEP is unable to treat organic load associated with de-icers (formate and glycol). In order to improve onsite stormwater management, two experiments were carried out in real conditions between 2008 and 2010. The first one dealt with the ability of two vertical flow planted beds (VFPB) associated with two kinds of plant to treat waters contaminated by de-icers. Organic loads of up to 2,094 mg/L chemical oxygen demand (COD) were passed through sand filters in a closed-loop system. Organic degradation was characterized by a first-order kinetic constant driven by nutrient availability. Nutrient (N, P) addition was then necessary to reach treatment objectives (40 mg/L COD). The second system evaluated sludge Hc removal by four VFPB associated with several plant species. This experiment showed that the measured Hc removal rates of 70% for sludge and 95% for water did not depend on the choice of plant and that a rest time was necessary to allow Hc removal. Finally, a treatment system, based on VFPB, should be a good alternative which could optimize both sludge and contaminated water management.

Key words: airport, constructed wetland, de-icers, hydrocarbons, sludge treatment, vertical flow water treatment

INTRODUCTION Preservation of the air traffic in areas subjected to harsh winter conditions requires the implementation of specific measures to handle black ice and frost. One of these measures deals with the use of chemicals: acetates, formates and glycols. These runway and aircraft de-icing operations (Figure 1)

21

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

Figure 1 | De-icing operations of (A) runways and (B) planes.

have a significant environmental impact, in particular with relation to their high organic load, with biological oxygen demand (BOD) and chemical oxygen demand (COD) concentrations that can reach 3,500 mg/L in runoff water in winter. Impacts on receiving waters are thus temporary but can nevertheless be an obstacle to good ecological status achievement required by the European Water Framework Directive (2015) depending on the sensitivity of receiving waters (use or not, ecological).

Considering the risks caused to the environment, in particular to the water resource (de-oxygenation of surface water, problem of additives present in de-icing products), the implementation of de-icing operations has to be the object of particular environmental concerns. The French environmental regulation on water and aquatic environments stipulates that every airport should have adapted recovery and treatment equipment for de-icing products in order to respect the regulatory levels for water runoff discharges into receiving streams.

As far as water runoff treatment is concerned, the winter season is often problematic for airport managers, as glycol and formate/acetate deicers are not removed by common runoff water treatment systems, based on settlement and filtration processes. Moreover, the time needed to achieve natural attenuation in retention ponds may be very long at low temperatures (,10 °C). This situation is problematic with regard to the French regulatory requirements.

Specific treatment units are deployed all around the world to take this seasonal pollution into account (US-EPA 2000; Higgins & Maclean 2002); some of them are based on intensive treatments (bioreactors, ion exchange, reverse osmosis) and others on extensive treatments (soil degradation, constructed wetlands).

ORLY AIRPORT WATER TREATMENT FACILITY A stormwater treatment plant (STEP) was created at Orly Airport in 1996. This system, built on a single site, recovers all the runoff waters, generated on 520 ha impervious surfaces, in order to treat them. Various treatment paths were fitted out, according to the volumes and quality of waters (Figure 2). This complex plant, which treats about 4 million m3 of rainwater each year, allows to manage at a same time, through a deep (20 m depth) runoff sewer network, a pumping well, twelve pumps, a grit chamber, two lagoons of storage and settling and a filtration unit (eight sand filters), a maximum water volume of 65,000 m3 (runoff path in Figure 2). Treated water is then discharged to the Orge river. Due to groundwater inflow in the runoff sewer network a clear water path treats water trough a settling tank and then sand filters. When hydrocarbons are on line detected in the well, contaminated waters are sent to an oil/water separator and then to the filters (contaminated water path in Figure 2). In winter, when runoff waters are contaminated by de-icing products,

22

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

Figure 2 | STEP at Orly Airport.

STEP technology based on settlement and filtration is inefficient to remove this organic load characretrized by a high solubility. The de-icing products used at Orly Airport are potassium formates and propylene glycol. During the particularly harsh winter of 2009–2010, consumption of propylene glycol and potassium formate was about 720 m3 and 450,000 L, respectively. Figure 3 illustrates the total organic carbon (TOC) monitored online during the winter of 2009–2010 at the STEP inflow.

TOC concentration can thus reach up to 1,000 mg/L. On line TOC analyze controls the winter runoff path: TOC contaminated runoff water are stored in lagoons to promote biodegradation of de-icer (Figure 2). Low temperature, and restricted water storage capacity do not allow to reach quality objectives [daily average concentration: 10 mg/L TOC, 10 mg/L glycol, 40 mg/L COD, 30 mg/L suspended solids (SS)]. TOC contaminated runoff water is then sometimes discharged at low flow to the municipal wastewater sewer network (winter runoff path in Figure 2).

At the same time, linked to settling and filtration processes, this STEP annually generates between 50 and 100 tons (dry matter – DM) of sludge. Raw sludges extracted from dry weather settling tank, washing of sand filters and water/oil separator are then concentrated (sludge path in Figure 2). Sludge produced by the STEP has the usual features of stormwater runoff sediments:

• fine particles: more than 80% of particles are smaller than 80 μm, • relatively high levels of heavy metals associated to SS in runoff water (about 3,000 mg/kg DM – mainly zinc, copper and lead), but lower than French regulatory criterion for sewage sludge application.

The specificity of this sludge is its very high level of petroleum hydrocarbons (mean value of 40,000 mg/kg DM with a maximum concentration of 110,000 mg/kg DM) due to accidental releases

23

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

Figure 3 | Illustration of daily TOC concentration (mg/L) measured online at the STEP inflow (Orly Airport).

(for example, leakages in the hydrocarbon transport network). Once extracted, this concentrated sludge is managed offsite by a specialized firm.

OBJECTIVES As outflow water quality (realative to COD) at Orly airport is subject to temporal changes during winter time, winter runoff facilities need to be improved to respect quality objectives. Within the framework of their respective environmental missions, the Aéroports de Paris laboratory and the French civil technical center (French civil aviation authority/Ministry of Ecology) launched a program intended to experiment a vertical flow planted bed system in order to improve the degradation of the organic molecules formed by de-icing products in water runoff. The objectives of this pilot study introduced in 2008 with the cooperation of the CETE Ile de France (The Ministry of Ecology’s technical and scientific network) were to:

• test performances of an extensive treatment based on vertical flow planted beds (VFPB) for winter contaminated runoff treatment on an experimental scale but in real conditions (real runoff waters and environmental conditions), • identify and specify the role of influencing parameters, • propose the initial foundations for sizing an operational treatment plant.

To manage onsite sludges generated by the STEP Aéroports de Paris and Lyonnaise des Eaux (STEP operator), with the co-funding of the Agence de l’Eau Seine-Normandie, conducted a pilot experiment to test the possibility of treating this sludge onsite with a VFPB. The objectives were to estimate the performances of the system and to identify which plants are the most adapted to this purpose. Such a system is commonly used to achieve sewage sludge dewatering (Vincent et al. 2010); here, the specificity is relative to pollutant removal.

The final objective of both pilot experiments (water and sludge treatments) was to evaluate technical feasability of a complementary system to the present STEP based on the same technology, the VFPB.

METHOLOGY Water treatment

Experimental filters consisted of four cylinder-shaped lysimeters with a surface area of 0.75 m2 filled by granular materials to a depth of 0.75 cm. An experimental system was set up in order to test the influence of different types of vegetation, filtering material grain size, the number of daily feeding cycles (hydraulic

24

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

load, daily volume in Table 1) and availability of nutriments (Figure 4). Other tested parameters were environmental: temperature and runoff organic load and composition (Table 1).

The choice of vegetation was important due to its potential attractiveness for birds. Two kinds were tested:

• Scirpus lacustris, Juncus inflexus, Iris pseudocarus, • Phragmites australis.

Two washed silicate granular materials were tested, 0–4 sand and 0–2 crushed fine sand. A transition granular layer below the filter ended with a drain. A calibrated orifice regulated outflow to 4 L/min.m². Filters were fed by water pumped from a storage tank via a flow distributor (Figure 4).

Each lysimeter was then fed with a constant flow. Storage tank was filled with glycol contaminated water pumped from the STEP grit chamber (see Figure 2). Reference lysimeter was used to assess TOC oxydation without vegetation. The 0–2 fine sand was abandoned for the second winter, due to its sensitivity to frost.

Two experimental periods have been studied : winter 2008–2009 and winter 2009–2010. During the first winter (2008–2009), only one storage tank (1 m3) was used for the four lysimeters (Figure 4).

During the second one (2009–2010), each lysimeter was connected to a pump and a storage tank (Figure 4). As the experimental design was based on water recirculation (closed-loop system), the storage tank also acted as a feeding tank.

Figure 4 | Illustration of the water treatment experimental system.

Samples were taken on a daily basis, and concerned raw and treated waters. Analyses were performed in order to determine TOC, COD, BOD, glycol and formate concentrations. Chloride concentration was also analyzed.

Due to the relative duration of degradation during the first winter (2008–2009), nutrients (nitrogen and phosphorus) were added to one of the two lysimeters during the second winter (2009–2010). N and P addition enabled a BOD/N/P of 100/5/1 on a weight basis to be reached in the storage tank.

Sludge treatment

The experimental system is composed by 4 planted beds (Figure 5). Each 4 m2 cylinder-shaped cell is independant from the others. These vertical flow beds are filled up to 40 cm by sands with a grain size increasing downward to the drain.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

25

Figure 5 | Schematic illustration of the experimental pilot.

Plant choice has been done based on their potential to treat hydrocarbons

• beds 1 and 2: common reeds (phragmites australis), • bed 3: dwarf creeping bamboo (ploïoblastus pumilus), • bed 4: assemblage of iris (iris pseudocarus), carex (carex pendulat) and cattail (typha latifolia).

Before feeding the beds, sludges are diluted in a storage tank (from 150–200 g DM/L to 10–20 g DM/L – Figure 5). Two experimental periods have been studied, July to September 2008 and May to September 2009. For each period the feeding cycle was: two weeks with a daily feeding followed by a rest time of two weeks. Each cycle has led to a 150 kg DM application for each cell. An analytical program has been performed both on sludges (C10–C40 hydrocarbons) and percolating waters (COD, SS, C10–C40 Hydrocarbons).

RESULTS Water treatment

Two winters were studied: 2008–2009 and 2009–2010. Three treatment experiments were performed during the first winter, and seven during the second one. Table 1 presents features (weather and physical–chemical) of the 10 monitored events. The TOC concentration range in contaminated water during these experiments was between 71 and 2,094 mg/L corresponding to propylene glycol concentration range of 10 to 1,035 mg/L. Daily volume (Second column in Table 1) ranges from 300 L/day (0.4 m/day) to 3600 L/day (4.8 m/day) without dysfunction. Chloride concentrations range from 84 to

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

26

Table 1 | Summarized data on weather conditions, organic and chloride concentrations and degradation features during the 10 events monitored during the winters of 2008–2009 and 2009–2010 Time to reach Volume L/day

Propylene glycol (mg/L)

COD initial (mg/L)

COD final (mg/L)

80% removal rate (day)

Mean 4 Lys.

Mean 4 Lys.

Cl (mg/L)

T°C mean

T°c min

2008–2009 Evt 1

944

100

256

40

7

245

5.3

3.5

Evt 2

708

725

1470

234

18

951

3.4

1

Evt 3

708

528

867

330

. 18

504

5.6

2

Lys. 1

Lys. 3

Lys. 1

Lys. 3

345

622

9

174

7

.14

86.9

2.5

0.4

672

1276

18

25

5

5

445

3.5

1.5

10

71

14

24

2

.2

84

4.2

0.9

27

4

6

425

5.2

1.7

5

2

484

8.3

5.6

2009–2010 Lys. 1

Lys. 3

Evt 1

900

Evt 2

900

Evt 3

900

300

Evt 4

900

550

597

1005

15

Evt 5

900

1036

2094

271

Evt 6

3600

494

1435

684

Evt 7

3600

324

1435

548

17 26 16

.3

1

n.d.

5.7

.5

3

n.d.

2

2.2 1.5

Grey cases correspond to lysimeter with nutrient addition. n.d.: not determined.

Lys. 1: lysimeter with 0–4 sand and iris/scirpus/juncus assemblage.

Lys. 3: lysimeter with 0–4 sand and phragmites australis.

COD final: COD concentration when experiment is stopped.

Time to 80% is the expanded time to reach a 80% removal rate from the COD initial concentration. Estimation from lineary regression between measured values. ‘ . x’ indicates that 80% is not yet reached at the end of the experiment after x days.

951 mg/L. 2008–2009 and 2009–2010 winters were particularly harsh with recorded temperature lower than the normal ones. Mean event temperatures range from 2.5 to 8.3 °C and min event temperature from –1.5 to 5.6 °C. Physical and chemical data relating to concentration evolution and removal rate were processed.

The first experimental objective was to obtain a final removal rate of 80% of the organic load (expressed as COD or TOC). Figure 6 illustrates the cumulated removal rate computed during experiment 5 in the winter of 2009–2010.

Average COD degradation rate measured in reference experiment is 20% during 2008–2009 winter. The time needed to achieve a removal rate of 80% during 2008–2009 experiments varied from 1 to more than

Figure 6 | Removal rate of COD, TOC, propylene glycol and BOD determined during the fifth experiment of the winter of 2009– 2010 for lysimeters 1 (without nutrient addition) and 3 (with nutrient addition).

27

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

18 days and were lower than or equal to 7 days for 2009–2010 experiments with nutrient addition (Table 1).

Final COD concentrations were similar during 2008–2009 experiment between the four lysimeters, ranging from 40 to 330 mg/L. For 2009–2010 experiments final COD concentrations range from 9 to 27 mg/L for lysimeter with nutrient addition. Evt 7 was the unique exception with a final COD concentration of 548 mg/L. For lysimeter without nutrient addition concentrations range from 14 to 684 mg/L. Sludge treatment

Only bamboo development was affected by the sludge; for other plants, the growth was good in spite of the high level of metals and hydrocarbons. Aerial and root growth was better for reeds with a foliar cover allowing preservation of optimal degradation conditions.

Particle filtration was pretty good with suspended matter (and associated pollutants like COD) removal rate higher than 90% throughout the experiment. COD oncentration in percolating waters is however higher than the regulatory quality objective for discharge in river – (Figure 7). As for water treatment experiments, no difference in treatment performance was observed between the different planted beds. Diluting prior to application to beds is important in terms of clogging prevention. Clogging was indeed observed for sludge concentration higher than 40 g DM/L. Sludge analyses performed in surface two weeks after the end of the feeding showed a C10–C40 concentration decrease of about 70% (Figure 8 for the 2009 experiment). This removal rate was similar for the four beds. C10–C40 hydrocarbon concentrations in percolating waters were relatively variable between 0.08 and 0.46 mg/L which respect regulatory objective (1 mg/L). Considering water concentrations, the global hydrocarbon removal rate was then higher than 95%.

Figure 7 | COD logartitmic concentration evolution between raw and treated waters.

DISCUSSION Water treatment

During the first winter, it was thus not possible to determine the role of vegetation assemblage on degradation performance. It therefore seems that plants play an indirect role (Wymazal et al.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

28

Figure 8 | Concentration evolution (mg/kg DM) and computed removal rates for raw sludge and sludge sampled at bed surface 2 weeks after the end of feeding for the 2009 experiment. Data on percolating waters are also indicated.

1998). Degradation time does not seem to be linked to temperature (Table 1). Vegetation development at the end of the monitoring period was good, and seemed to be pollutant and salt tolerant (up to 484 mg/l of chloride). Absence of hydraulic dysfunction during all the experiences illustrates the ability of vertical reed bed to accept a large range of hydraulic load (Molle et al. 2004). The main differences between 2008–2009 and 2009–2010 experiments concern lower final COD concentration and time to reach 80% removal rate in 2009–2010 for experiments with nutrient addition.

Table 2 | k values computed for the 10 events monitored during the winters of 2008–2009 and 2009–2010 K 1 (day)

Volume L/day

2008–2009 Mean 4 Lys. Evt 1

944

0.3

Evt 2

708

0.09

Evt 3

708

0.06

2009–2010 Lys. 1

Lys. 3

Lys. 1

900

Evt 2

900

0.63

0.52

Evt 3

900

300

1.56

0.73

Evt 4

900

550

0.25

0.64

Evt 5

900

Evt 6

3600

Evt 7

3600

Grey cases correspond to lysimeter with nutrient addition.

Lys. 1: lysimeter with 0–4 sand and iris/scirpus/juncus assemblage.

Lys. 3: lysimeter with 0–4 sand and phragmites australis.

0.35

Lys. 3

Evt 1

0.3 0.2 0.13

0.08

1.81 1.43 0.78

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

29

The most important features for TOC removal process are then final concentration and time need to reach this concentration. From a chemical point of view these features are associated to kinetic.

Degradation kinetics

TOC degradation is assumed to follow first-order kinetic (Equation (1))

C0 Ln C

¼ kt

(1)

With C: TOC (or COD as COD ¼ 3.29 TOC þ 42.45/r 2 ¼ 0.91) concentration at time t, C0: initial TOC concentration (t ¼ 0) and k: degradation constant (day 1).

In such a degradation process, k (Table 2) is determined graphically as the slope of the linear regression between Ln (C0/C ) and t. The higher k is, the more rapid the degradation is, and shorter is the time to reach 80% of COD degradation (see Table 1).

It was not possible to establish a correlation between k values and degradation performance or grain size (2008–2009), nor was it possible with vegetation assemblage (2008–2009), pollutant initial concentration (2008–2009 and 2009–2010) and temperature. It has previously been shown that a temperature control of degradation kinetics below 10 °C exists. During this study, mean temperatures were systematically below 10 °C, so this parameter may not be determining.

Nutrient addition is the only determining parameter during experiments (see Figure 6 for nutrient control). Nutrient addition is thus a key parameter in achieving rapid degradation of the organic load at winter temperatures. Efficient degradation is probably associated with good biomass acclimatization and development that need time to become optimal (Evt 1 to Evt 5 for lysimeter 3, cf.

Figure 9). If nutrient load is stopped, treatment performance slowly decreases (Evt 5 to Evt 7, for Lysimeter 3, cf. Figure 9), probably due to nutrient reserve consumptions. Importance of daily feeding cycles (corresponding to daily volume in Table 2) was not clearly identified as a determining parameter. However lower k values determined during events 3 and 4 for lysimeter 3 with nutrient addition could be associated to relatively lower daily feeding volume (Table 2).

Figure 9 | k value evolution for lysimeter 3 during 2009–2010 Winter. Event n° refers to Tables 1 and 2.

Sizing considerations

One of the first approaches for sizing a full-scale wetland treatment system is based on first-order degradation kinetics associated with hydraulic considerations.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

30

For sizing considerations, several assumptions are made:

• • • • • •

an inflow COD concentration of 500 mg/L (coherent with Figure 3), a target COD concentration of 40 mg/L (regulatory objective for outflow), a daily treated flow of 10,000 m3 (estimation of the contaminated flow), 0.7 m-deep filtering sand (0–4), as for experiments, a regulated outflow (4 L/min.m²), as for experiments, k ¼ 1.5 day 1 with a feeding rate of 1.2 m day 1 and with k water-load dependent (assuming that hydraulic load is a determining parameter for degradation). A specific kinetic constant (k/q, with q being the daily hydraulic load in m.day 1) of 1.25 m 1 is then computed.

From a hydraulic approach, experiments have shown that filters can accept a water load of up to 4.8 m day 1 without malfunction. Common sizing considerations relative to vertical flow reed beds are based on an annual maximum load of up to 100 m.year 1 (Uhl & Dittmer 2005).

Considering a hydraulic load of 1 m day 1, a wetland surface of 1 ha should be able to pass 10,000 m3 a day.

From a chemical approach, and considering first-order kinetics, the target concentration would be reached after 2 days of water recirculation.

Some complementary design considerations should be integrated (Figure 10). The treatment unit should include a pond upstream in order to add and mix nutrients before feeding the constructed wetland. Nutrients may be added by a feed pump controlled by TOC online measurement. The storage basin should be equipped with an underflow baffle in order to retain floating pollution, and with a

Figure 10 | Schematic illustration of proposed facility for winter runoff treatment.

31

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

by-pass to prevent any hydraulic disturbance. In order to treat organic load which exceeds sizing assumptions, the hydraulic system should enable water recirculation based on TOC measurements at the outflow. In the same way, daily hydraulic load should be adjusted according to climatic conditions (rainfall, temperature) and to water features (volume, TOC concentration). Feeding should alternate with rest periods. The wetland system should be partitioned into two files in order to prevent disturbances (hydraulic, contamination) and it should be possible to partition each file into several cells in series in order to optimize performance. The wetland system should be equipped with a dead storage zone in order to withstand dry weather conditions. The unit should be equipped with a metrological system (flow and online TOC).

Sludge treatment

This work shows that VFPB are adapted to the removal of high suspended matter content and high hydrocarbon loads, and confirms that plant species are not determining for treatment performance and that they probably play an indirect role. Organic carbon concentration (expressed COD) in treated water is however too high to respect the quality objective for discharge in river. Removal rates for hydrocarbons are also insufficient, final sludge concentrations being higher than 15,000 mg/kg DM. VFPB can then not be proposed alone as a treatment facility. They could nevertheless be associated with an intensive system (such as a bioreactor, which was also tested during this work) as a finishing part (Figure 11). Sludges could be pumped from the STEP oil/water separator to an aerated bioreactor seeded with fungal cocktail (Figure 11).

Figure 11 | Schematic illustration of proposed facility for sludge treatment.

CONCLUSIONS AND PERSPECTIVES The experimental approach adopted by Aéroports de Paris illustrates that the use of extensive systems like VFPB may provide complementary facility to the present STEP which could resolve several problems for airport managers.

Vertical flow planted sand beds are then adapted to organic load removal from glycol contaminated runoff water if nutrients (nitrogen, phosphorus) are added in optimal quantity. Experiment have shown their ability to reduce high organic load (up to 2094 mg/L COD) to concentration in accordance with regulatory objectives (maximum of 40 mg/l COD) even at low temperature in an acceptable treatment time. This study allows to propose initial sizing foundations for operational equipment

32

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.003

adapted to the airport. This experimental study need however to be validated by scale study before a real scale treatment facility.

Sludge treatment experiments show that VFPB are able to reduce sludge hydrocarbon loads (up to 70%) but residual concentrations are still high (. 15,000 mg/gk DM). VFPB could nevertheless be associated with an intensive system (such as a bioreactor) as a finishing part. Pilot study with such an association (intensive þ extensive processes) should be performed in the future.

Results from both treatment experiments seem to indicate that plants only play an indirect role, as no difference in treatment performance was observed between the different planted beds.

At short term, 2014–2015 winter, Aeroports de Paris intends to complete the STEP facility for winter runoff treatment by a constructed wetland (vertical flow reed beds).

In perspective, it will be interesting to study the behavior of other organic substances in the treatment system (for example, phenols (Abira et al. 2004), benzotriazoles (Castro et al. 2001) and in particular, substances listed as priority substances in the European Water Framework Directive (alkylphenols (Chang et al. 2004; Langford et al. 2005).

REFERENCES Abira, M. A., Van Bruggen, J. J. A. & Demy, P. 2004 Potential of a tropical sub-surface constructed wetland to remove phenol from pre-treated pulp and papermill wastewater. 9th international conference on wetland systems for water pollution control. Avignon, France, pp. 353–358.

Castro, S., Davis, L. C. & Erickson, L. E. 2001 Plant-enhanced remediation of glycol-based aircraft deicing fluids. Practice Periodical of Hazardous, Toxic, and Radioactive Waste Management 5, 141–152.

Chang, B. V., Yu, C. H. & Yuan, S. Y. 2004 Degradation of nonylphenol by anaerobic microorganisms from river sediment.

Chemosphere 55, 493–500.

Higgins, J. & Maclean, M. 2002 The use of a very large constructed sub-surface flow wetland to treat Glycol contaminated stormwater from aircraft de-icing operations. Water Quality. Research Journal Canada 37 (4), 785–792.

Langford, K. H., Scrimshaw, M. D., Birkett, J. W. & Lester, J. N. 2005 Degradation of nonyphenolic surfactants in activated sludge batch tests. Water Research 39, 870–876.

Molle, P., Lienard, A., Grasmick, A. & Iwena, A. 2004 French vertical flow constructed wetlands: reed bed behavior and limits due to hydraulic overloading on first stage filters. Proceedings of the 9th IWA Conference Proceedings on Wetland systems in water pollution control, Avignon, France, September 2004.

Uhl, M. & Dittmer, U. 2005 Constructed wetlands for CSO treatment: an overview of practice and research in Germany. Water Science Technology 51 (9), 23–30.

US-EPA 2000 Constructed Wetlands treatment of municipal wastewaters. Manual. EPA/625/R-99/010. 165.

Vincent, J., Molle, P., Wisniewski, C. & Liénard, A. Sludge drying reed beds for septage treatment: towards design and operation recommendations. Proceedings of the 12th IWA Conference Proceedings on Wetland systems in water pollution control, Venice, Italy, October 2010.

Wymazal, J., Brix, H., Cooper, P. F., Haberl, R., Perfler, R. & Laber, J. 1998 Removal mechanisms and types of constructed wetlands. In: Constructed Wetlands for Wastewater Treatment in Europe (J. Wymazal, H. Brix, P. F. Cooper, M. B. Green & R. Haberl, eds). Backhuys Publishers, Leiden, (Nl), pp. 17–66.

© IWA Publishing 2014 33

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.004

Setting standards for the future based on evidence from the past – a UK perspective on the success of the approach S. Homewood and C. Snowdon WRc Plc, Frankland Road, Blagrove, Swindon, Wiltshire, SN5 8YF, United Kingdom. E-mail: sarah.homewood@wrcplc.co.uk, carmen.snowdon@wrcplc.co.uk

Abstract The Water Research Centre has collated, over a number of years, quantitative end-use (micro-component) water consumption information for over 700 properties in England and Wales. For a sample of this size, this is the most detailed set of data in the UK, and was used to inform some of the mandatory and voluntary standards that are in place in the UK today.

One such standard is in the new edition of Part G of the Building Regulations, published in 2010, which includes a water efficiency requirement for the first time. This is a whole-of-house water use standard set at a maximum of 125 litres per person per day. The voluntary Code for Sustainable Homes (the Code) also has a series of standards for water use in new homes varying from 90 to 120 litres per person per day. Both of these standards allow flexibility for people to choose what is installed in a new home (including new emerging technologies) provided they can achieve the specified water use figure. Whilst the Code is purely a design standard, the Building Regulation is a mandatory requirement and homes built to the building regulations should be inspected postinstallation to confirm the devices installed have specifications in-line with the design standard.

To help inform the UK Government on the success of the approach, and to aid UK water companies develop 25 year projections of demand as part of their water resources planning, research assessing the real consumption of homes designed to these water efficiency standards has been carried out. Meter readings from a sample of homes were analysed alongside property information to identify actual water use and trends amongst property cohorts.

Results from our statistical analysis show that occupancy is the single most significant factor influencing per capita consumption and whether a home meets a given water efficiency design standard.

Key words: household water use, UK standards, water efficiency, water use

INTRODUCTION Many areas of England and Wales are in a situation of water stress. The latest publication by the Environment Agency and Natural Resources Wales (Environment Agency and Natural Resources Wales 2013) shows, at a water body level, the current water stress situation. Under the Water Framework Directive the water body is the unit for which environmental objectives are set and may be entire lakes, parts of lakes, parts of rivers and streams, estuaries, coastal waters or groundwater.

Whilst many people living in the UK believe there is a water resources problem in the South and South East of England, this latest publication shows clearly that water bodies right across the UK are in a state of moderate or high stress.

Many parts of the UK experienced a period of drought between 2004 and 2006, which resulted in a focus by the Government on ways of managing supply and demand of water. Following consultation in 2005 on the introduction of a voluntary Code, which would assess homes against sustainability criteria including water efficiency, the Government considered if assessment of water efficiency should

34

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.004

be mandatory for all new homes as a way of developing infrastructure that helped drive down demand. In 2006 a consultation on options for water efficiency in new buildings was held (Department for Communities and Local Government [DCLG] and Department for the Environment Food and Rural Affairs [Defra] 2006).

In 2011, just over 11 billion m3 of water was abstracted from the non-tidal sources and almost 6 billion m3 was used for public water supply (Defra 2013). Approximately half of this is used to supply homes. Over 100,000 new homes are built each year in England and Wales (approximately 0.5% of the total housing stock). Therefore by making this new housing stock water-efficient over time, the total demand for water seen by homes across the UK should drop.

Since 1999, Water Research Centre (WRc) has worked with the UK water industry using bespoke hardware and software – the Identiflow® system – to monitor the end-uses (or micro-components) of water in homes. The evidence base, from over 700 homes, that this monitoring has produced allows us to understand how water is used in the home, and the trends in consumption associated with different appliances, as well as underpinning water company demand forecasts and policy relating to water use in the UK. Specifically ownership, volume per use and frequency of use information for individual appliances has been compiled (Snowdon et al. 2012). Evidence collected from this system, specifically frequency of use information, was used as a starting point for development of the water calculation methodology for new homes (DCLG 2009) referenced within Regulation 36 of the Building Regulations (Building Regulations 2010) which came into force in 2010.

The objective of the calculation methodology referred to in Regulation 36 was to maintain flexibility and customer choice in the selection of sanitary appliances used in the dwelling, whilst at the same time achieving a lower water usage with minimal impact on the inhabitants. It also opened the way to using alternative technologies such as harvested rainwater and grey water recycling for toilet flushing.

Regulation 36 of the Building Regulations (Building Regulations 2010) states: ‘The potential consumption of wholesome water by persons occupying a dwelling to which this regulation applies must not exceed 125 litres per person per day, calculated in accordance with the methodology set out in the document “The Water Efficiency Calculator for New Dwellings” … ’ The calculation methodology applies both to Regulation 36 of the Building Regulations, which is the minimum requirement for all new homes built, and also to homes built to the Code (DCLG 2010). The Code is the national standard for the sustainable design and construction of new homes, and sets minimum standard for energy and water use. Some local governments may require homes to be built to levels of the Code and all social housing is required to be built to Code level 3 as a minimum.

It is recognised by both the Water Industry and Government that Regulation 36 is a design standard and therefore may not reflect the actual water consumed within new homes for a number of reasons.

These include the different ways in which occupants use their domestic appliances (such as selecting different wash programmes on their washing machine) and the impact of product replacement over time. We also know from our research (Waylen et al. 2008) that households change how they use their water-using appliance over time. The single biggest change over the last ten years is the increase in duration per use and frequency of use of showers.

As water companies in England and Wales are required to forecast household water demand, including an allowance for new homes’ water use, as part of their Water Resources Management Plan, it was appropriate to identify an evidence-based estimate for this and explore reasons why new homes may not meet their design standard.

Between 2010 and 2013, on behalf of a group of UK Water Companies and Defra, WRc collated water meter readings from homes built to the new water efficiency standard (Regulation 36), or Code level 3 or 4.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.004

35

Table 1 | The sample structure of homes included within the analysis Occupancy

1

2

3

4

Flat

18

19

3

2

Bungalow

1

3

5

6

Total

Mean

42

1.7

4

1.8

23

2.3

68

2.6

Detached

4

12

2

5

Semi-detached

10

27

16

11

3

Terraced

12

13

8

5

2

40

2.3

Unknown

6

26

14

10

7

63

2.8

Total

51

100

43

33

12

240

2.4

1

1

METHODS The study set out to determine the actual water use of homes built to the Code and the new Regulation 36, and to specifically understand

• which factors influence per capita water use in new properties; • which factors influence the probability that a given household will exceed its water efficiency design standards; and

• which factors influence the amount by which a household exceeds its water efficiency design standard once that has happened?

To answer these questions data was collected from a sample of homes built to Regulation 36 or a Code Standard; whose occupancy was known. Regular meter readings were then collected for each of the properties identified in the sample and a questionnaire issued that sought information on a number of factors that were considered likely to be useful during data analysis. This collated information on tenure, property type, occupancy, and water butt and sprinkler ownership, as well as the types of devices installed at the property. Additionally, the questionnaire aimed to understand how satisfied consumers were with the appliances installed and if any of the appliances installed originally had been replaced.

The sample available for use in data analysis was 240 properties. Calculating an actual per capita consumption for comparison to the design standard meant that occupancy data was critical – this information is not routinely collected by water companies for all households. The data was therefore sought through either the limited existing records of the water companies or through the customer questionnaire. Of the sample of 240 properties; 52 were known to have been built to Code level 3 or 4 (per capita consumption of 105 litres per person per day, excluding outdoor water use). The design standard for the remaining 188 properties is unconfirmed but was assumed to comply with the minimum Building Regulations design standard of 125 litres per person per day, with a five litre allowance for outdoor water use.

To carry out statistical analysis to answer the research questions posed, the candidate factors that influence water consumption were selected. It was well known from previous research (Chambers et al. 2005) that the relationship between occupancy and household consumption is non-linear.

The design standards for new properties (whether from Part G Regulation 36 or from the Code) are measured in terms of water use per person (PCC). Hence the primary statistical model needed as a minimum to predict PCC and include occupancy and design standard as factors.

Factors influencing PCC

Data from the following property factors was assimilated from the questionnaire data and tested for their impact on PCC: dwelling type (e.g. flat, detached house etc.), presence of a water butt,

36

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.004

Figure 1 | Histogram of PCC values (litres/person/day) (n ¼ 240).

Figure 2 | Variation of PCC with occupancy.

replacement of any devices originally installed, water efficiency design standard, use of hosepipe/ sprinkler/external tap and tenure (owner occupied, rented, etc.) of the property.

A linear model was fitted to the dataset, however as the PCCs were skewed these were transformed to a log scale in order to satisfy the normality assumption of a linear model.

The general formula for a linear model is shown below yi ¼ b0 þ b1 x1i þ b2 x2i þ 1i where y is the response variable (e.g. PCC) for property i; b0 , b1 , b2 are the model coefficients; x1i , x2i are observations of factors 1 and 2 for property i; 1i is the random error associated with property i.

These error terms are assumed to be independent and normally distributed with a mean of zero.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.004

37

Figure 3 | Cumulative distribution of volume of exceedance of design standard for Code Level 3 or 4 properties (litres per person per day).

All model fitting was performed in R statistical software (R Development Core Team 2010).

A forward stepwise regression was performed for the dataset, retaining only those which were statistically significant (at a 5% significance level). When factors were only marginally significant the Akaike Information Criterion was used to compare the goodness of fit of the model before and after the factor was added (Burnham & Anderson 2002).

Probability of exceedance of design standard

The probability of a household exceeding the design standard was explored by setting up an indicator variable which took a value of 1 if a property exceeded its design standard and 0 if not. Such binary response variables are commonly modelled using logistic regression which is a form of generalised linear modelling (GLM). Hence, a binomial GLM with a logit1 link function was fitted to each of the dataset using forward stepwise regression.

The general formula for a binomial GLM with a logit link function is shown below yi Bin(1, pi )

pi log 1 pi

¼ b0 þ b1 x1i þ b2 x2i

where y is the response variable for property i which can takes values of 1 and 0 only; p is the probability of y taking a value of 1; β0, β1, β2 are the model coefficients; x1i , x2i are observations of factors 1 and 2 for property i; ð pi =1 pi Þ is the ‘odds’ of property i exceeding its design standard.

To explore design standard exceedance, a log transformation was applied to the highly skewed exceedance values and a forward stepwise linear regression performed. The accuracy of these models was lower than the results for all 240 properties, due to the smaller sample; 78 properties exceeded their design standard for water use.

1

The binomial GLM uses a logit function which maps numbers from ∞ to þ∞ onto the interval [0,1], and can therefore be used to model probability as a function of any set of predictor variables.

38

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.004

RESULTS Which factors influence PCC?

Occupancy was identified as a significant factor affecting PCC; homes with a lower occupancy had a higher PCC. This trend is in line with findings from previous studies (e.g. Chambers et al. 2005). For the 240 properties where occupancy information was available the average occupancy was 2.4 which was in line with the national average (DCLG 2012).

PCC was also found to vary with dwelling type and ownership of a water butt. However, smaller sample sizes meant the strength of the relationship between these is less than for occupancy, and further research would be required to strengthen the conclusions relating to the relationship of PCC with dwelling type and ownership of a water butt.

What proportion of homes exceed their design standard, and which factors influence the probability that a design standard will be exceeded?

Looking at the overall sample (240 homes) 70% of all homes used less than the maximum design standard of 125 litres per person per day. Homes built to Code level 3 or 4, with a design standard of 105 litres per person per day, appear, from the sample, to be more likely to exceed their design standard than the other homes in our sample which were built to an unknown design standard. It is inappropriate to draw any further conclusions from this result since the ‘Unknown’ group may have been built to more or less stringent standards than Code level 3 or 4, thus making direct comparison of the two samples unsound.

For homes built to Code levels 3 or 4, whilst the median exceedance volume is about 0 litres (i.e. the median PCC is around the expected design standard); the distribution is positively skewed. In other words, although 50% of homes use less and 50% use more that the design standard. The volume of exceedance at 25% of properties is greater than 20 litres (i.e. total water use is higher than even the Regulation 36 standard).

The factors that are most likely to influence whether or not a property exceeds its design standard are similar to those which have higher PCCs. Lower occupancy homes (especially one and two person households), homes without a water butt and those living in houses rather than flats were identified as being most likely to exceed the design standard.

Which factors influence the amount by which a household exceeds its water efficiency design standard once the design standard has been exceeded?

Only occupancy was found to be statistically significant when this question was explored; homes with a lower occupancy are likely to exceed their design standard by a larger amount than homes with a higher occupancy.

Study limitations

(1) The design standard to which many of the properties were built is unknown. Although all of the homes should have been built to a minimum design standard of 125 litres per person per day (in line with Regulation 36) it is possible that some of this sample were built to a more stringent standard e.g. Code level 3 or above. Therefore it has not been possible to draw conclusions in relation to the whole sample other than to explore the influences on PCC. (2) No information is available on the occupancy of new homes at a national level. Within the sample, flats, as a dwelling type, were under-represented and as these may be expected to have a lower occupancy level than houses, our sample mean occupancy may be high. As the results

39

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.004

indicate that a higher occupancy level results in a lower PCC this may mean that our study potentially under-estimates the number of properties that would exceed the design standard at a national level. In order to confirm this hypothesis, occupancy rates in new homes would need to be surveyed.

(3) Confounding factors – background changes. Water use in homes is inherently variable due to the necessary human interaction with water using devices to result in consumption. Whilst some uses of water – such as toilet flushing – are relatively functional, others have a number of behavioural and social elements associated with them (e.g. motivations for showering resulting in differing durations of shower event described in Pullinger et al. 2013). The human behaviour with regard to water use, and even choice over products to install and use in the home, are affected by a number of factors. Over the course of the study there were a number of initiatives and changes which may have influenced the households in the sample to change their water use or appliances within it. It is not possible to separate out these confounding factors from the analysis as there is, in effect, no control group who have not been exposed to these influences.

DISCUSSION When considering water efficiency and behaviour change, we believe it is appropriate not to consider policies or initiatives in isolation as the success of one is likely to depend upon implementation of another. A key example of this is the recent decision by UK retailers and builders merchants to display a water-use label on water-using products that they stock (Dickinson 2013). Water labelling has been a topic of discussion for over ten years and yet it has only been since the introduction of Regulation 36, which requires information on product water use performance, that there has been enough momentum to reach a widespread voluntary agreement on its use.

The question is then raised on how the effectiveness of the policy has been, or could be, further enhanced. For instance a number of water companies have embarked on public information and education campaigns around water efficiency, and drought. A government led campaign (Defra 2013) aimed to highlight the link between river health and water use, so that people understand and value water. These initiatives may have resulted in behaviour changes that will be attributed through a quantitative evidence-based study, such as this, to the performance of Policy in driving a change in consumption. However, the reality may be that these external influences are resulting in a culture change in how we use water.

The challenge for the future will be harnessing the power of policy in conjunction with these external influences to maximise the combined benefit. Without adequate policy and regulation in place, we consider it unlikely that the labelling debate would have gained traction – however it is also likely that without providing information on water efficiency performance for new water using products (e.g. the water label scheme) the ease of application of policy for developers may result in poor appliance choice and, hence, in more homes exceeding the design standard. This naturally raises the question of how much further might we be able to go and whether it would be more effective to implement more stringent standards, or to continue to implement other initiatives, perhaps focussing on the power of choice architecture, or customer engagement through gamification – the use of game theory in a non-game context – to keep changing customer behaviour as well as relying on technology.

Questions that remain unanswered

The results of this study have raised a number of new research questions. In particular, why new households consistently fail to achieve PCC of 105 litres per head per day against the relevant design standard. What consumer behaviour or appliance use is driving water use in one quarter of

40

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.004

new homes to exceed the minimum water efficiency standard of 125 litres per head per day? In addition, the longevity of the perceived effect of the design standard is of interest – will these homes still have the same level of water consumption in five or ten years’ time? This study is necessarily a snapshot of average water use over a two year monitoring period. Will the new design standard have a different, long-term effect on water use?

CONCLUSIONS The data collected and analysed for the study allows us to draw a number of conclusions, the main one being that there is a statistically significant variation in PCC with occupancy. The importance of this should be noted, particularly when considering future policy and when devising demand forecasts for the future. If average occupancies look set to fall over time as more people live alone or without children, then there is a risk that the design standard will be more regularly exceeded and by higher volumes resulting in no benefit to the water consumption levels across England and Wales. Water Companies will need to take care when forecasting consumption and pay particular attention to the anticipated occupancy levels in order to prevent an under-prediction of the demand from these homes.

ACKNOWLEDGEMENT The funders of this research are gratefully acknowledged: Bristol Water; Defra; Essex and Suffolk Water; Portsmouth Water; South East Water; South Staffordshire Water; Southern Water; Thames Water; Veolia Water and Yorkshire Water.

REFERENCES Burnham, K. P. & Anderson, D. R. 2002 Model Selection and Multimodel Inference: A Practical Information-Theoretic Approach (2nd ed.). Springer-Verlag.

Chambers, V., Creasey, J., Glennie, E., Kowalski, M. & Marshallsay, D. 2005 Increasing the Value of Domestic Water use Data for Demand Management – Summary Report. WRc report P6805. Project CP187.

DCLG 2012 Housing Statistics. [online] https://www.gov.uk/government/publications/house-building-in-england-july-toseptember-2012.

Department for Communities and Local Government and Department for the Environment, Food and Rural Affairs. 2006 Water Efficiency in New Buildings – A consultation document. Product Code: 06 BD04277. Crown Copyright, 2006.

Department for Communities and Local Government. 2009 The Water Efficiency Calculator for new dwellings. Crown Copyright. ISBN: 978-1-4098-1827-4.

Department for Communities and Local Government. 2010 Code for Sustainable Homes Technical Guide. November 2010.

Crown Copyright. ISBN 978 1 85946 331 4.

Department for the Environment, Food and Rural Affairs 2012 Water Abstraction from non-tidal surface water and groundwater in England and Wales, 2000 to 2011. DEFRA Statistics Release: 18th December 2012.

Department for the Environment, Food and Rural Affairs 2013 Love Your River website, http://www.defra.gov.uk/ loveyourriver/, visited 20 September 2013.

Dickinson, G. 2013 Bathroom retailers, merchants and manufacturers join forces to launch consumer water efficiency label website, http://www.wrap.org.uk/node/15323, visited 20 September 2013.

Environment Agency and Natural Resources Wales. 2013 Water stressed areas – final classification. Downloadable from https://www.gov.uk/government/publications/water-stressed-areas-2013-classification.

Pullinger, M., Browne, A., Anderson, B. & Medd, W. 2013 Patterns of water: The water related practices of households in southern England, and their influence on water consumption and demand management. Lancaster University, Lancaster, UK. Downloadable from https://www.escholar.manchester.ac.uk/uk-ac-man-scw:187780.

R Development Core Team 2010 R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna. Website: http://www.R-project.org.

41

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.004

Snowdon, C., Bewes, V., Bujnowicz, A., Doyle, T., Glennie, E., Kowalski, M. & Vigo, F. 2012 Compendium of Micro-component Statistics. WRc report P9193.03. Project CP499.

Statutory Instruments 2010 The Building Regulations 2010. S.I. 2010, No. 2214. Crown Copyright.

Waylen, C., Bujnowicz, A., Saddique, S. & Boyles, C. 2008 Water use in new dwellings. WRc report P7694. Project CP337.

© IWA Publishing 2014 42

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

Survey on full-scale drinking water treatment plants for arsenic removal in Italy S. Sorlinia,*, F. Gialdinia and M. C. Collivignarellib a Department of Civil Engineering, Architecture, Land, Environment and Mathematics, University of Brescia, Via Branze 43, Brescia, Italy b

Department of Hydraulic and Environmental Engineering, University of Pavia, Via Ferrata 1, 27100, Pavia, Italy

* Corresponding author. E-mail: sabrina.sorlini@ing.unibs.it

Abstract Arsenic in drinking water causes severe health effects and it is widely diffused in groundwater around the world.

This paper presents the results of a survey about the main arsenic removal technologies employed in Italy and the main features in the management of real treatment plants. 19 drinking water treatment plans were involved in this study. The specific aspects analysed in this survey were: type of technologies applied in the drinking water treatment plants (water characteristics, ionic form of As in raw water, etc.), technical aspects (chemical dosage, treatment steps, hydraulic load, retention time, etc.), operational aspects (backwashing, media regeneration, management of residues, etc.) and costs of these technologies. In Italy, the main technologies employed are chemical precipitation (10 plants) and adsorption with granular ferric hydroxide (GFH) (six plants). Two of these plants employ both chemical precipitation and GFH. Moreover, there are some applications of adsorption on titanium dioxide (two plants), reverse osmosis (two plants) and ionic exchange (two plants).

Key words: adsorption, arsenic, chemical precipitation, drinking water treatment plant, ion exchange, reverse osmosis

INTRODUCTION Arsenic is an ubiquitous element found in the atmosphere, soils and rocks, natural waters and organisms. It is mobilized through a combination of natural processes such as weathering reactions, biological activity and volcanic emissions as well as through a range of anthropogenic activities.

Among the various sources of As in the environment, drinking water probably poses the greatest threat to human health. Regulatory and recommended limits for arsenic in drinking water have been reduced in recent years following increased evidence of its toxic effects to humans. The World Health Organization (WHO) guideline value was reduced from 50 to 10 μg/L in 1993 (WHO 1993) although the recommendation is still provisional pending further scientific evidence.

In Italy, Legislative Decree 2001/31 reduced the limit of arsenic from 50 to 10 μg/L, in agreement with the European Directive 98/83/EC. Well-known critical areas for high-As concentration in groundwater have been found in many parts of the world such as Argentina, Bangladesh, Chile, China, Hungary and India (West Bengal) (Smedley & Kinniburg 2002). Also in many parts of Italy groundwater contains arsenic concentrations higher than the national regulatory standard of 10 μg/L. The most critical areas for groundwater contamination are located in the following regions: Lombardia, Piemonte, Veneto, Trentino Alto Adige, Emilia Romagna, Toscana, Umbria, Lazio and Campania where concentration reaches values up to 500 μg/L (Colombetti et al. 2011; Colombetti & Mantelli 2011).

43

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

Several treatment processes have been developed for arsenic removal from water, including precipitation, adsorption, ion exchange, membrane filtration and biological process. The most promising processes for arsenic removal from highly contaminated waters are chemical precipitation and adsorption because of their low cost and high efficiency (Song et al. 2006; Choong et al. 2007; Pallier et al. 2010). Chemical precipitation, using mainly iron or aluminium salts as coagulants, is a well-known method for arsenic removal that can convert soluble arsenic species into insoluble products. However, an oxidation step is required to better remove arsenite (As (III); Bissen & Frimmel 2003). Three main mechanisms have been suggested for arsenic removal: (1) adsorption involving the formation of surface complexes between soluble arsenic and active sites of formed hydroxides; (2) co-precipitation with incorporation of soluble arsenic species into a growing hydroxide phase by inclusion, occlusion, or adsorption (Lytle et al. 2005); (3) precipitation and formation of insoluble compounds (like FeAsO4) between As(V) and Fe(III) (Edwards 1994). Some authors (Jiang 2001; Thirunavukkarasu et al. 2005) showed that arsenic removal from water achieved by chemical precipitation process depends on initial arsenic concentration in water while other authors did not observe any influence (Cheng et al. 1994). The arsenic removal could reach 90% (Scott et al. 1995; USEPA 2000A; Gregor 2001). Iron coagulants are more effective than aluminium, titanium, and zirconium ones (Scott et al. 1995; Lakshmanan et al. 2008).

Silicate, phosphate and sulphate interfere with arsenic removal (Hering et al. 1996; Meng et al.

2000; 2002).

Zouboulis & Katsoyiannis (2002) studied arsenic removal by direct filtration with sand filters, instead of separation by sedimentation. Alum or ferric chloride were used as the coagulant agent and organic polymers were added to enhance the precipitation. The coagulants achieved up to 99% arsenic removal. Han et al. (2002) used ferric chloride and ferric sulphate as coagulants in arsenic removal. The results have shown a significant arsenic removal through adsorption mechanism onto the ferric complexes present. Chemical precipitation is an effective and widely used technology for treating drinking water to remove arsenic.

For small systems, however, adsorption onto media [granular ferric hydroxide (GFH) or granular ferric oxide (GFO)] is currently the process more adopted. GFH is effective in reducing both As (III) and As (V); removal efficiency and kinetics are slightly different (Thirunavukkarasu et al. 2003).

Optimal pH is included in the range of 6–8, with highest arsenic removal observed at pH 7.6 and arsenic removal yield of 95–97% (USEPA 2000a, b; Thirunavukkarasu et al. 2003). The recommended operating conditions include an EBCT of 3–6 min and a hydraulic loading rate between 10 and 20 m/h. The cost of this technology is high, but the advantages compared to other materials such as activated alumina are the following: adsorbent capacity (GFH: 4–13 g As/kg; activated alumina: 0.55 g As/kg), high values of Bed Volume (BV for GFH: 32,000–82,000; BV for activated alumina: 1,000–13,000) and arsenic pre-oxidation is not required (MWH 2005). Recently, regeneration techniques have been developed in real scale by flushing with a solution of sodium hydroxide with an efficiency of 95–97% (Streat et al. 2008).

Removal of arsenic in drinking water by granular titanium dioxide (TiO2) is a new technology that was developed in the USA. There are not many experimentations and bibliographic documentations about the adsorption capacity of TiO2, but studies show high efficiency yield for arsenic removal (about 95%). Adsorption results (Bang et al. 2005; Pena et al. 2005) indicate that As (V) removal efficiency is high at pH between 4 and 9, whereas As (III) removal efficiency is high at pH 7–8. This media does not require a pre-oxidation step, but the As (V) adsorption kinetics are faster than As (III) (Bang et al. 2005). Approximately 32,000–45,000 BV of water were treated by the TiO2 filter with an EBCT of 1–3 min and arsenic concentration in raw water of about 10–40 μg/L (Bang et al.

2005).

Ion exchange can be a viable process for the removal of arsenic from drinking water by means of strong-base anion exchange resin, that can remove only ionized compounds and consequently only

44

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

arsenate ion (As (III) is undissociated in natural water). The chloride-arsenate chemical reaction typically occurs close to the pH of neutrality:

2R-Cl þ HAsO2 4 $ R 2 -HAsO4 þ 2Cl ðR- indicates the exchange resin):

Design considerations for arsenic removal by ion exchange include: pre-treatment for the removal of undesirable compounds (for instance methane and hydrogen sulphide), oxidation state of arsenic, flushing by a bed of strong-base anion exchange resin. Process performance depends on: alkalinity, pH, type of competitive ions in water, affinity of arsenic with resin and other process items. The resins have a relatively high affinity for arsenic in the arsenate form (HAsO2 4 ), however, high sulfate and nitrate levels compete with arsenate and can reduce removal efficiency. The anionic exchange is scarcely effective when high concentration of TDS (.500 mg/L) and sulfate (.150 mg/L) are measured in water (AWWA 1999). Chloride-form resins are often used in arsenic removal. As a result, the potential increase in chlorides can greatly raise the corrosivity and reduce the pH. In situations where chlorides pose a problem, pH stabilization or/and post-treatment corrosion control, demineralization, blending, or alternative treatment techniques may be required to prevent disturbances in the distribution system. An advantage of using the ion exchange resin is that regeneration does not cause a reduction of ion exchange capacity, but a suitable management of spent regenerant has to be adopted.

Nanofiltration and reverse-osmosis are evaluated the most suitable membrane processes for arsenic removal and both inorganic and organic arsenic can be removed simultaneously with monovalent and bivalent ions. As (III) is more difficult to remove than As (V) (MWH 2005). Arsenic removal by reverse-osmosis is complicated because it involves: physical filtration, membrane pressure, electrical potential, polymer type, etc. Reverse-osmosis performance is adversely affected by the presence of turbidity, iron, manganese, silica, and other constituents. Reverse-osmosis requires pre-treatment for removal of particles and dissolved constituents (like iron, manganese and colloids) (USEPA 2000B). Reverse-osmosis is capable of achieving an arsenic removal yield close to 95% and finished water arsenic concentrations below 0.002 mg/L, when arsenic is present as As (V).

The concentrate flow is between 10 and 50 percent of the influent flow depending on raw water quality and membrane properties, with an arsenic removal from 50 to 99%. The permeate has typically low concentration of dissolved salts, so a by-pass of raw water can be applied to improve the quality of treated water. Alternatively post-treatments are required. The pH has an insignificant influence on the reverse-osmosis process. The rejection of As (V) decreases with the increase in temperature and decrease in pH (Bissen & Frimmel 2003). Various studies have reported arsenic removal ranging between 40–99% without specifying the arsenic species removed, but it would seem that reverse-osmosis removes As (V) to a greater degree than As (III) (Ning 2002; Bissen & Frimmel 2003; Shih 2005).

This paper presents the results of an investigation on the the main arsenic removal technologies applied in some Italian drinking water treatment plants (DWTPs) with a specific analysis of their operation and management criteria.

MATERIALS AND METHODS The investigation about the main arsenic removal technologies employed in Italy and the main features in the management of real treatment plants was performed on 19 drinking water treatment plans (Table 1).

The investigation was carried out using a questionnaire delivered to the DWTPs managers and operators for the collection of the following information:

• general aspects: water characteristics, ionic form of As in raw water, arsenic removal yield, etc.; • technical aspects: chemical dosage, treatment steps, hydraulic load, retention time, etc.;

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

45

Table 1 | Main data about the drinking water treatment plants involved in the survey Process

CP

CP þ Ads Ads

DWTP

Treatment chain

AsIN [μg/L]

QIN [L/s]

1

A þ BF þ S þ CP þ SF þ DIS

17

450

2

A þ BF þ S þ CP þ SF þ DIS

17

450

3

P þ SF þ S þ CP þ GAC þ DIS

20

13

4

CP þ P þ BF þ S þ CP þ SF þ GAC þ DIS

38

60

5

ST þ CP þ A þ BF þ CP þ SF þ DIS

50

9

6

P þ CP þ SF þ DIS

22

13

7

ST þ CP þ A þ BF þ S þ CP þ SF þ DIS

47

28

8

A þ BF þ DIS

20

20

9

A þ BF þ S þ CP þ SF þ GFH þ DIS

37

3

10

A þ BF þ CP þ SF þ GFH þ DIS

65

15

11

GFH

59

3

12

GFH

16

10

13

P þ GFH

61

27

14

T

32

0.6

15

P þ SF þ GAC þ T

48

nr

Ads þ M

16

P þ F þ GFH þ RO

50

34

M

17

F þ S þ P þ RO þ DIS

IE

18

ST þ A þ BF þ GAC þ IE þ DIS

30

33

19

IE þ DIS

30

2

100

0.03

A ¼ aeration; BF ¼ biological filtration; S ¼ softening; SF ¼ sand filtration; DIS ¼ disinfection P ¼ pre-oxidation; GAC ¼ granular activated carbon; ST ¼ stripping; F ¼ filtration; GFH ¼ granular ferric hydroxide; T ¼ titanium dioxide; RO ¼ reverse osmosis; IE ¼ ion exchange; CP ¼ chemical precipitation; Ads ¼ adsorption; M ¼ membrane filtration; nr ¼ not reported.

• operational aspects: filter backwashing, media regeneration, management of residues, etc.; • treatment of residues: residue characteristics, technical solution for their treatment and disposal, etc.; • costs of the technologies.

RESULTS Among the analysed DWTPs, the main technologies employed were chemical precipitation (10 plants) and adsorption with GFH (six plants). Two of these plants employed both chemical precipitation and GFH and other two plants employed GFH and reverse osmosis. Moreover, adsorption on titanium dioxide (two plants), reverse osmosis (two plants) and ionic exchange (two plants) processes were adopted.

DWTPs adopting chemical precipitation

Ten full-scale application plants applying chemical precipitation process were analyzed in the survey.

They are were located mainly in the North of Italy, particularly in the Po Basin (Mantua, Cremona, Brescia). Only one plant is located in the Centre-North of Italy (Pisa). The flow rates of treated water are extremely variable (450–3 L/s). The arsenic concentration in raw water is quite variable too, even if generally below 50 μg/L. When arsenic concentration in raw water was higher than 40 μg/L, a double stage of iron salt dosage was applied (plants 4, 5 and 7). Alternatively, the adsorption on GFH after chemical

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

46

precipitation was employed (plants 9 and 10). The purpose of this treatment was to increase the arsenic removal yields, to meet the regulatory limits.

The average arsenic removal yields were between 60 and 90%. The highest yields (higher than 90%) were obtained when a post-treatment of adsorption on GFH or a double stage of iron salts dosage were adopted. The reported yields refer to the arsenic removal in the oxidized form. In fact, before the precipitation treatment, the As(III) was always oxidized to As(V).

The main information related to DWTPs applying chemical precipitation are indicated in Table 2.

The drinking water treatment plants analyzed in the survey showed a highly complex sequence of treatments in addition to the simple precipitation treatment. Among the pre-treatments there were air stripping for the removal of volatile compounds, such as hydrogen sulphide and methane and aeration to increase the dissolved oxygen before biological filtration. The analysis of the drinking water treatment plant with chemical oxidation showed the presence of numerous biological oxidation treatments in addition to or in substitution of the chemical oxidation of arsenic. In particular, biological oxidation was adopted by eight plants out of ten and its application was mainly adopted in case of water contaminated by ammonium, in addition to iron, manganese and arsenic. These pollutants could be removed together with arsenic in the treatment plant. After ferric chloride dosage, the plant treatments were composed by filtration (on sand or, more rarely, on granular activated carbon) for the removal of insoluble compounds derived from chemical precipitation. Moreover, two plants employed, before the final disinfection, a post-treatment of filtration on GFH, to decrease the arsenic concentration below the regulation limit.

Table 2 | Main management data about the DWTPs with chemical precipitation (Number of plants: 10) Potassium permanganate, sodium hypochlorite, oxygen or hydrogen Pre-oxidation

Chemicals

peroxide

Chemical precipitation

Chemicals Dosage Contact time

Ferric chloride 3–7 mg/L as FeCl3 3–4.5 min

Biological filters

Media

Biolite, quartzite (either alone or mixed with manganese dioxide), sand and pyrolusite 6.5–11.8 min Air, water or air mixed with water 3 times per week - 3 times per day 10–44 min 3–10%

Contact time Mean for backwashing Backwashing frequency Backwashing duration Volume of water for backwashing/ volume treated water Secondary filters

Media Contact time Mean for backwashing Backwashing frequency Backwashing duration Volume of water for backwashing/ volume treated water

Sand, quartz and pyrolusite, granular activated carbon, anthracite, hydroanthracite, GFH 4.5–20 min Air, water or air mixed with water Sand filters: 3 times per week - 4 times per day; GAC: twice per day; GFH: once per month Sand filters: 14–52 min; GAC: 10 min; GFH: 25 min Sand filters: 7.8–15.5%; GAC: 5.6%; GFH: 0.065%

All the residuals generated by filter backwashing were liquid sludge. The amount of sludge produced and its quality were closely related to the characteristics of the raw water, the type of coagulant adopted (the use of lime will produce a higher amount of sludge compared to other coagulants) and the frequency of backwashing. These residuals could be discharged to sewerage or treated in a proper residual treatment line. When a specific line for wastewater treatment was available, it was composed by a thickening tank (in one case: plant number 9) or it could be more complex. In this case the treatment was arranged with a basin for wastewater storage, flocculation, thickening and

47

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

dewatering phases (in five case: plants number 1, 2, 4–6). The treatment of water from backwashing of filters in a dedicated line produces two flows: a liquid phase (called supernatant) that could be discharged into a sewerage or a water surface body and a solid phase (sludge) that was generally disposed of in landfill for not hazardous wastes (see Figure 1).

Figure 1 | Schematic diagram of treatment of wastewater from backwashing of filters in DWTPs adopting chemical precipitation.

DWTPs adopting adsorption

Five plants adopted GFH and two titanium dioxide as adsorption media. Plants 9 and 10 used adsorption treatment after a previous arsenic precipitation treatment by means of iron salts to guarantee the constant respect of the legal limits.

Moreover, one plant (number 16) treated water rich in arsenic (50 μg/L) and fluorides (2.8 mg/L) and had a combined treatment system: 50% of the flow was treated by reverse osmosis and the remaining 50% with GFH (about 15 L/s). The remaining two plants removed arsenic by adsorption on a granular material made of titanium dioxides.

The adsorption process management is not too difficult and the presence of specialized personnel is generally not required as the filter backwash was automated on the ground of load losses or timing.

The main information related to DWTPs applying chemical precipitation are indicated in Table 3.

As far as the GFH adsorption step is concerned, the residuals that could be produced are the filter backwashing waters, whose only aim is to move the filtering bed and remove the suspended solids that could have settled on the filter media. This kind of residual can be generally directly discharged into a surface water body, if the water treatment plant had only the adsorption step. If the water treatment scheme was composed of a mixed treatment of precipitation followed by adsorption, GFH filter backwash waters could be managed together with sand filter backwashing water. Some kinds of granular ferrous hydroxide were nonetheless subjected to exhaustion with time. When it was exhausted, the filtering material had to be removed and disposed of in appropriate facilities which were authorized to treat and dispose of this kind of waste. Exhausted GFH was collected and disposed of with the European Waste Code (EWC) 19 09 01.

The backwashing waters produced by titanium dioxide could be discharged in the sewerage. The material cannot be regenerated, so, when it is exhausted, it has to be substituted and disposed of in a dump. The material substitution frequency was around 12–24 months.

DWTPs adopting ion exchange

Table 4 shows the main process and plant characteristics concerning the two plants with ionic exchange treatment. The plant n. 18 applied a filter containing a type II anionic resin in chloride cycle while the plant n. 19 had a selective functionalized adsorbent resin with ferric hydroxide.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

48

Table 3 | Main management data about the DWTPs with adsorption process (Number of plants: 8) GFH

TiO2

Arsenic removal yields

75–99%

91–97%

Contact time

3–12 min

1.5–3 min

hydraulic load

6.5–12.8 m/h

n.a.

Filter backwashing mean

Water

Water

Backwashing frequency

Once a month – three times a week

Once a year

Backwashing duration

8–25 min

3–5 min

Backwashing flow rate

30–80 m3/h

n.a.

Volume of water for backwashing/volume treated water

0.02 to 0.9%

n.a.

Media life

9 months–3 years

100,000 BV

n.a.: not available data.

Table 4 | Main management data about the DWTPs with ion exchange process (Number of plants: 2) Type II anionic resin

FeOOH resin

Arsenic removal yields

.80%

.80%

Contact time

3–5 min

4–8 min

Hydraulic load

3.2–5.3 m/h

1.8–3.5 m/h

Backwashing mean

Water

Water

Backwashing frequency

Three times a month

Once a month

Backwashing duration

15 min

15 min

Backwashing flow rate

10 m3/h

7.5 m3/h

Volume of water for backwashing/volume treated water

0.05%

0.07%

Regeneration mean

NaCl and water

NaOH, NaCl and water

Regeneration frequency

Three times a month

Once a year

Regeneration duration

30 min

30 min

Regeneration flow rate

5 m3/h

n.d.

The inlet water in the first plant contained methane, ammonia, iron and manganese as well. These substances were removed by stripping and biological filters (two 4 meter-high sand filters in parallel, and two 2 meter-high active carbon filters always in parallel), which could also oxidize arsenic.

Finally, water was filtered through the three columns in parallel, filled with anionic resin, with 90 m3/h downflow. This kind of resin obtains high yields if there are few interfering elements, that is to say few sulfates (concentration ,50 μg/L), chlorides (concentration ,50 μg/L) and nitrates (concentration ,25 μg/L). The second plant, which treated 2.0 L/s flow, had a simpler treatment scheme (pumping station, two functionalized FeOOH resins in series, NaClO disinfection) as arsenic was the only critical pollutant.

This treatment produces the following residuals: backwashing water (derived from backwashing of resins) and spent brine (derived from resins regeneration).

If the regeneration occurs concomitantly with the backwashing, wastewater can be disposed of in a platform for hazardous waste or treated in a specific line for residue treatment. This treatment may consist of the following steps: storage of wastewater before the treatment; iron salt dosage for arsenic precipitation; settling; storage of spent brine.

Two streams arise from this type of treatment: the regenerated brine, which can be reused in subsequent resin regeneration treatment and sludge separated from the settling, which has to be disposed of in a landfill.

49

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

In the case of the plant n. 18, the sludge derived from brine regeneration and water from filter backwashing were treated in a special treatment plant within the company. The supernatant separated from the settling of the wastewater treatment plant was discharged into the surface water, while the sludge was treated with a dewatering centrifuge. The dried sludge was then collected by a company that makes compost. When the resins cannot be regenerated any more, they must be disposed of in landfills. After about 1,000 regeneration cycles, the resin was replaced and the exhausted resin was allocated to landfill with EWC 19 08 06.

DWTPs adopting membrane filtration

The following Table 5 shows the characteristics of reverse osmosis plants analyzed in the survey.

Table 5 | Main management data about the DWTPs with membrane process (Number of plants: 2)

Arsenic removal yields

. 95%

Modules

1–25

Pressure

11–14 bars

permeate flow/raw water flow

50–75%

concentrate flow/ raw water flow

25–50%

In these case studies high removal yields of both forms of arsenic were obtained, even though very different scales of application were applied (plant 17 was a point of use plant). The plant 17 treated the entire flow rate, while the plant 16 treated only 50% of raw water. In particular, in this plant, a 15 L/s water flow was filtered through the membrane, while the remaining 15 L/s were treated with GFH filtration, in order to remove fluoride as well as arsenic (fluoride was present in raw water at 2.8 mg/L). The permeate produced by plant 16 had an acid pH because of the presence of free CO2. In this case, the manager added sodium hydroxide to reduce the water aggressiveness in the distribution system.

Membrane processes (RO) do not require backwashing and regeneration operations, but generally a check of the proper operation of pre-treatment is needed, in particular for the softening, which requires periodic replacement of salt, microfiltration and disinfection.

If the membranes are protected from fouling, clogging and bacterial film formation, the management operations consist only of washing with cleaning and disinfectant materials with a frequency of once every few months. Membrane processes, however, produce a solution (concentrate) rich in contaminants, particularly arsenic, which can be discharged into the sewerage or be adequately treated. Other residues that arose from the process were the cartridge filters, used in pre-treatment to remove suspended solids, which must be replaced once every 2 months, and the UV lamps, which were adopted to reduce the bacterial load upstream of the membrane treatment. The used filters were disposed of in landfills for non-hazardous waste, as well as UV lamps. The non-hazardous nature of these compounds should be evaluated on a case by case basis. In the case of plant 16, however, the concentrate was discharged into the sewerage according to the regulation limits.

Cost of the technologies

In addition to very simple systems (for example, oxidation and filtration through sand and activated carbon), there are more complex treatments, designed also to eliminate other undesirable substances such as iron, manganese, ammonia, etc. Some treatments are time-proven (chemical precipitation) while other treatments are more innovative (selective resins, GFH or titanium dioxide). The

50

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

complexity and diversity of these issues lead to a variation in the treatment costs from one facility to another. The costs of the technologies adopted in the water treatment plants studied varied from 3.1 to 10.4 €cent/m3 of treated water for chemical precipitation, where the maximum cost is related to plants n. 1–2 with treatment of residues in addition to water treatment. For chemical precipitation combined with adsorption the cost varied from 12.3 to 32.6 €cent/m3 of treated water. For membrane filtration the indicated cost was 13 €cent/m3 of treated water while for ion exchange the cost ranged between 1.9 to 6.1 €cent/m3 of treated water.

CONCLUSIONS This survey involved 19 drinking water treatment plants for arsenic removal. The most spread technologies were chemical precipitation (10 plants) and adsorption with GFH (six plants). Two of these plants employed both chemical precipitation and GFH. Moreover, there were some applications of adsorption on titanium dioxide (two plants), reverse osmosis (two plants) and ionic exchange (two plants). The results of the survey show that arsenic in groundwater varied from 16 to 100 μg/L and flow rate was in the range 0.03–450 L/s. Generally, chemical precipitation was adopted in the case of high flow rate. Lower flow rates were associated to adsorption with titanium dioxide and reverse osmosis. Arsenic removal yields obtained with the different processes were 60–90% for chemical precipitation, 75–99% for adsorption, higher than 80% for ion exchange and higher than 95% for reverse osmosis.

The main residues derived from the studied processes were water from filter backwashing (chemical precipitation and adsorption), exhausted media (adsorption), retentate and spent brine (membrane filtration).

Finally, an operation cost up to 10.4 and to 32.6 €cent/m3 was declared respectively in plants with chemical precipitation alone or with chemical precipitation followed by adsorption on GFH.

ACKNOWLEDGEMENTS The authors gratefully acknowledge all the subjects who collaborated to the activity of the working group. A special acknowledgment is addressed to the companies that took part in the survey: A2A S.p.A. (Brescia), Acque S.p.A. (Ospedaletto, Pisa), Acquedotto del Fiora S.p.A. (Grosseto), AEM Gestioni S.r.l. (Cremona), AIMAG S.p.A. (Mirandola, Modena), Padania Acque Gestione S.p.A. (Cremona), SISAM S.p.A. (Castelgoffredo, Mantova), Culligan Italiana S.p.A., Gruppo Zilio S.p.A., TECAM S.r.l. and Tecnoimpianti Water Treatment S.r.l.

REFERENCES AWWA 1999 Water Quality & Treatment. A Handbook of Community Water Supplies. 5th edn, McGraw Hill.

Bang, S., Patel, M., Lippincott, L. & Meng, X. 2005 Removal of arsenic from groundwater by granula titanium dioxide adsorbent. Chemosphere 60, 389–397.

Bissen, M. & Frimmel, F. H. 2003 Arsenic- a review. Part II: oxidation of arsenic and its removal in water treatment. Acta Hydrochimica et Hydrobiologica 31 (2), 97–100.

Cheng, R. C., Liang, S., Wang, H. C. & Beuhler, J. 1994 Enhanced coagulation for arsenic removal. Journal of American Water Works Association 9, 79–90.

Choong, T. S. Y., Chuah, T. G., Robiah, Y., Koay, F. L. G. & Azni, I. 2007 Arsenic toxicity, health hazards and removal techniques from water: an overview. Desalination 217, 139–166.

Colombetti, A. & Mantelli, F. 2011 The presence of arsenic in water of Italy: a review. In: Arsenic in Water for Human Consumption. Technical Solutions for Arsenic Removal from Drinking Water (Sorlini, S. & Collivignarelli, C., ed). LAP Lambert Academic Publishing, Saarbrucken, Germany, pp. 14–17.

51

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.005

Colombetti, A., Mantelli, F., Ottaviani, M. & Veschetti, E. 2011 Panoramica sulla presenza di arsenico nelle acque italiane.

In: L’arsenico nelle acque destinate al consumo umano – Esperienze e applicazioni delle tecnologie di rimozione dell’arsenico e aspetti gestionali (Collivignarelli, C., Riganti, V. & Sorlini, S., ed). Dario Flaccovio editore, Palermo, Italy, pp. 9–19.

Edwards, M. 1994 Chemistry of arsenic removal during coagulation and Fe-Mn oxidation. Journal of American Water Works Association 86 (9), 64–78.

Gregor, J. 2001 Arsenic removal during conventional aluminium-based drinking-water treatment. Water Research 35 (7), 1659–1664.

Han, B., Runnells, T., Zimbron, J. & Wickramasinghe, R. 2002 Arsenic removal from drinking water by flocculation and microfiltration. Desalination 145, 293–298.

Hering, J. G., Chen, P. Y., Wilkie, J. A., Elimelech, M. & Liang, S. 1996 Arsenic removal by ferric chloride. Journal of American Water Works Association 88 (4), 155–167.

Jiang, J. Q. 2001 Removing arsenic from groundwater for the developing world – a review. Water Science and Technology 44 (6), 89–98.

Lakshmanan, D., Clifford, D. & Samanta, G. 2008 Arsenic removal by coagulation with aluminum, iron, titanium, and zirconium. Journal of American Water Works Association 100 (2), 76–88.

Legislative Decree n. 31 of 2 February 2001 Accomplishment of Directive 98/83/CE on the quality of water intended for human consumption.

Lytle, D. A., Sorg, T. J. & Snoeyink, V. L. 2005 Optimizing arsenic removal during iron removal: theoretical and practical considerations. Journal of Water Supply: Research and Technology – Aqua 54, 545–560.

Meng, X. G., Bang, S. & Korfiatis, G. P. 2000 Effects of silicate, sulfate, and carbonate on arsenic removal by ferric chloride.

Water Research 34, 1255–1261.

Meng, X. G., Korfiatis, G. P., Bang, S. & Bang, K. W. 2002 Combined effects of anions on arsenic removal by iron hydroxides.

Toxicology Letters 133, 103–111.

MWH 2005 Water Treatment Principles and Design. Wiley.

Ning, R. Y. 2002 Arsenic removal by reverse osmosis. Desalination 143, 237–441.

Pallier, V., Feuillade-Cathalifaud, G., Serpaud, B. & Bollinger, J. C. 2010 Effect of organic matter on arsenic removal during coagulation/flocculation treatment. Journal of Colloid And Interface Science 342 (1), 26–32.

Pena, M. E., Korfiatis, G. P., Patel, M., Lippincott, L. & Meng, X. 2005 Adsorption of As (V) and As (III) by nanocristalline titanium dioxide. Water Research 39, 2327–2337.

Scott, K. N., Green, J. F., Do, H. D. & McLean, S. J. 1995 Arsenic removal by coagulation. Journal of American Water Works Association 87 (4), 114–126.

Shih, M. C. 2005 An overview of arsenic removal by pressure-driven membrane processes. Desalination 172, 85–97.

Smedley, P. L. & Kinniburg, D. G. 2002 A review of the source, behaviour and distribution of arsenic in natural waters. Applied Geochemistry 17, 517–568.

Song, S., Lopez-Valdivieso, A., Hernandez-Campos, D. J., Peng, C., Monroy-Fernandez, M. G. & Razo-Soto, I. 2006 Arsenic removal from high-arsenic water by enhanced coagulation with ferric ions and coarse calcite. Water Research 40, 364–372.

Streat, M., Hellgardt, K. & Newton, N. L. R. 2008 Hydrous ferric oxide as an adsorbent in water treatment. Part 2. Adsorption studies. Process Safety and Environmental Protection 86, 11–20.

Thirunavukkarasu, O. S., Viraraghavan, T. & Subramanian, K. S. 2003 Arsenic removal from drinking water using granular ferric hydroxide. Water, Air and Soil pollution 29 (2), 161–170.

Thirunavukkarasu, O.S., Viraraghavan, T., Subramanian, K. S., Chaalal, O. & Islam, M. R. 2005 Arsenic removal in drinking water – impacts and novel removal technologies. Energy Source 27, 209–O219.

USEPA 2000A Arsenic Removal From Drinking Water by Coagulation/Filtration and Lime-Softening Plants, EPA/600/R-00/ 063.

USEPA 2000B Technologies and costs for removal of arsenic from drinking water, EPA 815-R-00-028.

WHO 1993 Guidelines for drinking water quality, 2nd edn, Volume 1, Recommendations, Geneva, Switzerland.

Zouboulis, A. & Katsoyiannis, I. 2002 Removal of arsenates from contaminated water by coagulation–direct filtration.

Separation Science and Technology 37, 2859–2873.

© IWA Publishing 2014 52

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

Lessons learned during startup, testing and optimization of membrane bioreactors systems for enhanced nutrient removal V. M. Maillard, K. L. Perri, J. M. VerNooy and T. A. Young 16701 Melford Blvd. Suite 330, Bowie, Maryland, 21401 USA. E-mail: vincent.maillard@ghd.com

Abstract Membrane bioreactors are known for producing high quality effluent from wastewater treatment facilities in order to meet stringent regulatory requirements (Fleischer et al. 2005), accommodate growth (Vadiveloo & Cisterna 2008), provide opportunities for water reuse (Schmidt et al. 2011), and achieve other operational goals for various municipalities, utilities and industries (Cummings & Frenkel 2008). The process of testing, starting up and optimizing an MBR process for enhanced nutrient removal at the end of a construction project is often overlooked. Even a well-designed MBR can fail to meet expectations if the system is not properly configured during the startup phase, making this a critical step in any successful implementation of membrane technology.

The startup phase of two municipal MBR plants were compared to demonstrate the importance of various strategies for initial process optimization, with a focus on lessons learned, techniques and performance expectations that can be applied to future projects.

Key words: MBR, nutrient removal, optimization, startup, testing

INTRODUCTION Membrane bioreactors (MBRs), which consist of a biological process and a membrane filtration system (MFS), offer several benefits which have led to their widespread application in the treatment of municipal and industrial wastewater (Cummings & Frenkel 2008). MBR technology allows for greater flexibility in both design and operation, as they are compatible with a broad range of process configurations, physical layouts, influent characteristics, operating conditions and control strategies (Pellegrin et al. 2012). MBRs are also ideal for expansion to accommodate growth or meet increasingly stringent effluent requirements within a relatively small footprint (Ferraris et al. 2009).

Some of the primary limitations to the application of MBR technology for wastewater treatment compared to conventional alternatives include membrane fouling caused by certain operating conditions or influent constituents, the capital cost of MBR equipment, the selection of process and equipment configurations for optimal performance, and increased energy consumption. Each of these items are currently the focus of on-going research, which continues to develop and improve new operation and control strategies (including membrane aeration and cleaning procedures) to reduce energy demand, pre-treatment techniques to reduce sources of fouling, membrane designs (such as arrangement, support, material, density and size) to extend equipment life and enhance performance for the given operating conditions, and treatment options (such as anaerobic and attached growth systems) (Pellegrin et al. 2012).

Perhaps one of the most significant advantages of MBR technology is its ability to consistently produce high quality effluent, including effluent total nitrogen (TN) concentrations of 3 mg/L and total phosphorus (TP) concentrations of 0.3 mg/L, with the possibility of even lower concentrations following process optimization (Fleischer et al. 2005). The selection of MBR technology is often based on

53

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

the benefits associated with the production of such high quality effluent, such as more efficient utilization of available discharge limits, a reduction in the facility’s environmental impact and the potential to implement water reuse. However, selection and design are only the first steps towards capitalizing on these benefits. In order to take advantage of the advanced treatment capabilities and maximize the value of the MBR system, it is also critical for each component to be properly installed and operated to avoid costly inefficiencies, underperformance or even process failures (Ellsworth & Riddell 2007).

The successful startup and optimization of MBR facilities requires an additional degree of planning and execution to address the unique challenges associated with MBR technology, including

• The large number of interconnected components operating as a single system.

• Reliance on complex automatic controls and interlocks with risk of costly failures.

• Use of proprietary terminology, configurations, operational schemes, testing protocols, etc., by different system suppliers.

• A relatively wide range of potential operating conditions, control strategies and performance targets.

Other project-specific factors, such as time or budget constraints and owner preferences, will determine the approach necessary to meet the goals established for each individual MBR installation.

To document the impact of different strategies for bringing an MBR plant online, two municipal MBR wastewater treatment plants (WWTP) located in Virginia (United States) were studied during startup, performance testing and initial operation. The first plant is a new 0.95 megalitre per day (ML/d) facility owned and operated by the small, seasonal town of Cape Charles, Virginia, which has about 1,000 permanent residents.

The Cape Charles WWTP was built on a greenfield site to replace the Town’s existing plant, which consisted of aging contact stabilization units. The new plant includes headworks, four-stage Bardenpho reactors, MFS, ultraviolet (UV) disinfection, and provisions for water reuse.

The second plant is a 2.65 ML/d facility owned and operated by the Town of Berryville, Virginia, which has about 4,000 residents. The MFS at the Berryville WWTP was added as part of a major upgrade project on the same site as the existing lagoon treatment plant.

The upgrade project also added new headworks, 2.65 ML of flow equalization, four-stage Bardenpho reactors, and UV disinfection.

Both plants were designed to meet enhanced nutrient removal (ENR) level effluent concentrations of 4.0 mg/L TN and 0.3 mg/L TP. A summary of each plant is shown in Table 1.

A basic flow schematic for both Cape Charles and Berryville is shown below in Figure 3.

METHODS Both plants went through a 90-day performance testing period for the new MBR systems after initiating wastewater treatment. The plants also went through multiple stages of preliminary testing with clean water prior to the introduction of wastewater. During the clean water testing phase, the installed equipment ran for multiple weeks and was inspected to verify proper function. In particular, preliminary clean water testing of the MBR system was utilized to check the automatic programming developed for each facility, including data transfer and recording, alarms, interlocks, alternate control modes and operator interface. Any discrepancies observed during the clean water testing period were tracked with project-specific checklists until they could be suitably resolved. This thorough testing procedure helped to identify and correct potential issues, such as missing steps in the membrane cleaning sequence and equipment failures after a power outage, to prevent delays or failures during wastewater treatment.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

54

Table 1 | Plant design criteria Parameter

Cape Charles WWTP

Berryville WWTP

Current Average Flow

0.4 ML/d

1.1 ML/d

Design Average Daily Flow

0.95 ML/d

2.65 ML/d

Design Peak Flow

2.84 ML/d

7.95 ML/d (with flow equalization)

Peaking Factor for MBR

3:1

3:1

BOD

206 mg/L

290 mg/L

TSS

226 mg/L

270 mg/L

TKN

39.7 mg/L

45 mg/L

Design Average Influent Concentrations

TP

6.1 mg/L

6.5 mg/L

Flow Equalization

No

Yes (2.65 ML)

Influent Pumping Operation

Constant Speed

Variable Speed

Biological HRT

14 hours

24 hours

Type of Membranes

Hollow Fiber (GE ZeeWeed 500D, 340 sf/module)

Hollow Fiber (GE Zeeweed 500D, 250 sf/module)

No. of Membrane Trains

2

4

Following the completion of clean water testing, each plant seeded the biological tanks with activated sludge from a nearby operating MBR facility. Previous studies with ultrafiltration membrane systems have shown that seeding with an activated sludge inoculum from an existing facility can reduce membrane fouling during initial operation and help to speed up the development of higher mixed liquor suspended solids (MLSS) concentrations (Di Bella et al. 2010). Seeding with activated sludge also helps the MBR system to achieve the maximum design solids concentrations during the specified testing period.

Cape Charles initially seeded the new MBR to obtain a starting MLSS concentration goal of over 2,000 mg/L, and utilized additional seed sludge to accelerate solids accumulation after the introduction of wastewater. Berryville seeded the new MBR for a starting MLSS concentration goal of about 900 mg/L, and did not utilize additional seed sludge following the introduction of wastewater. As a result, MLSS concentrations were significantly higher at Cape Charles WWTP throughout the startup period, as shown in Figure 1.

Raw wastewater flow from the collection system was diverted to each plant after initial activated sludge addition. Based on the construction schedule for the respective projects, the introduction of wastewater occurred in January 2012 for Cape Charles WWTP and in September 2012 for Berryville WWTP. The diversion of raw wastewater to the new MBR systems proceeded in accordance with the startup plan developed specifically for each facility. For Cape Charles, the startup plan recommended that the Town’s existing wastewater plant remain in operation until the new MBR facility was consistently producing high quality effluent. This allowed for a gradual increase in flow to the new MBR plant and a prolonged flow split between the old and new treatment facilities. The new MBR system at Cape Charles WWTP began receiving 100% of the Town’s raw wastewater flow two months after the initial startup. The MBR system at Berryville WWTP was constructed as part of an upgrade to the Town’s existing treatment plant, so all of the raw wastewater flow was immediately diverted from the existing treatment process to the new MBR facilities. Influent flow rates for each plant are shown in Figure 2.

During the performance testing period, operating conditions and effluent quality were closely monitored at each facility to assist with process optimization. A combination of on-site laboratory testing and off-site, third party laboratory analysis were used to monitor concentrations of nutrients and

55

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

Figure 1 | Cape Charles wastewater treatment plant.

Figure 2 | Berryville wastewater treatment plant.

Figure 3 | Flow schematic for Cape Charles and Berryville WWTP.

suspended solids in the raw influent, mixed liquor and plant effluent. At both plants, online instrumentation was utilized to continuously measure influent and effluent flow rates, as well as dissolved oxygen (DO) and nitrate (NO3) concentrations in the biological tanks. Instrumentation provided as part of the MFS for each facility was used to monitor plant feed (raw influent) flows, permeate (filtered effluent) flows, wastewater temperature, effluent turbidity and trans-membrane pressure (TMP).

56

Figure 4 | Startup phase MLSS concentrations for two municipal MBR plants.

Figure 5 | Startup phase influent flow rates for two municipal MBR plants.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

57

Both Cape Charles WWTP and Berryville WWTP were designed to use Micro C-G™ for supplemental carbon and aluminum sulphate (alum) as a metal coagulant. The startup plans for both plants called for the chemical feed systems, supplemental carbon and alum, to be started up sequentially to provide adequate time to establish dosing rates for the first chemical before starting addition of the second chemical. Full denitrification was listed as an initial milestone in the startup plan for Cape Charles, so the supplemental carbon feed system was brought online first. Supplemental carbon addition began 22 days after the introduction of wastewater at Cape Charles WWTP, followed by alum addition after 38 days of wastewater treatment. Berryville WWTP started with relatively high (.7 mg/L) effluent TP concentrations in the first month of performance testing, so the alum feed system was started first after 40 days of wastewater treatment and the supplemental carbon feed system followed at 54 days of wastewater treatment.

RESULTS AND DISCUSSION The raw influent wastewater characteristics during the startup phase of each plant are listed in Table 2. The influent characteristics at Berryville WWTP remained relatively consistent throughout Table 2 | Startup phase average influent wastewater characteristics for two municipal MBR plants Parameter

Cape Charles WWTP Average (Min-Max)

Berryville WWTP Average (Min-Max)

BOD

142 mg/L (64–223 mg/L)

362 mg/L (209–536 mg/L)

TSS

91 mg/L (30–256 mg/L)

125 mg/L (32–448 mg/L)

TKN

30 mg/L (14–44 mg/L)

55 mg/L (47–59 mg/L)

TP

3.7 mg/L (1.7–5.4 mg/L)

7.4 mg/L (1.7–5.4 mg/L)

Figure 6 | Startup phase effluent total nitrogen (TN) and total phosphorus (TP) concentrations for two municipal MBR plants.

58

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

the 90-day startup period. Influent wastewater characteristics at Cape Charles WWTP showed significant variation as a result of the Town’s sewage collection system. The collection system consists of two main raw wastewater pump stations: one serving a gravity sewer system in the Town’s historic district and the second connected to a vacuum sewer system serving primarily new development.

The influent wastewater from the gravity sewer system was relatively dilute due to generally older construction and higher inflow/infiltration (I&I) rates in the collection system. In contrast, the vacuum sewer system has relatively high strength wastewater coming from the area of new development where there are more low-flow plumbing fixtures and less I&I. Both pump stations cycled on and off throughout the day, producing significant variation in the influent wastewater characteristics.

This was particularly noticeable during the colder months of the year when Cape Charles WWTP began wastewater treatment (before the seasonal residents arrived in Town).

Full nitrification occurred rapidly at both MBR facilities (within 7 days for Cape Charles WWTP and 14 days for Berryville WWTP), and as expected the time required to achieve nitrification did not appear to be significantly impacted by the volume of activated sludge inoculum utilized (Di Bella et al. 2010). As shown in Figure 3, Cape Charles WWTP was able to achieve partial denitrification until the addition of supplemental carbon. Full denitrification was achieved shortly after startup of the supplemental carbon feed system. Berryville WWTP was able to achieve nearly complete denitrification without the addition of supplemental carbon addition; however, as shown in Figure 3, the startup of the supplemental carbon feed system provided stable and complete denitrification.

Cape Charles WWTP demonstrated significant phosphorus removal due to biological treatment after startup of the supplemental carbon system, as shown in Figure 3. Alum addition was required to maintain effluent TP concentrations below 0.3 mg/L as the plant flows and influent loads increased. Effluent TP concentrations at Berryville WWTP decreased below the target concentration of 0.3 mg/L after the addition of alum on day 40 of the performance testing period.

At Cape Charles WWTP, the supplemental carbon system started up on day 24 of the performance testing period, and the effluent TN concentrations quickly decreased from approximately 25 mg/L on day 22 to less than 3 mg/L by day 29. Supplemental carbon (Micro CG™) was initially dosed in manual control while the automatic control programming for the plant was finalized. While under manual control, the supplemental carbon was over-dosed at approximately 142 mg/L to ensure consistent denitrification, resulting in an average effluent TN concentration of 1.2 mg/L. After the automatic controls became available the dosing rate was flow-paced using the influent flow signal, allowing the feed rate to be adjusted to achieve the desired effluent quality. After a temporary increase in effluent TN concentration following the implementation of the automatic control programming, the effluent TN concentration was again maintained below 2 mg/L for the remainder of the performance testing period with typical supplemental carbon dosing rates of approximately 130 mg/L. The initial alum dose at Cape Charles WWTP was approximately 20 mg/L and was increased to 40 mg/L by the end of the performance testing period. During startup, Cape Charles tested for effluent TP using a 0.2 mg/L detection limit; however, more recent laboratory testing with a detection limit of 0.02 mg/L has shown that the MBR system is able to consistently produce at or below 0.02 mg/L with an alum dosing rate of approximately 30 mg/L. This effluent TP concentration is approaching the current limit of technology (Bott et al. 2009; deBarbadillo et al. 2010).

The alum feed system for Berryville WWTP was started on day 40 and resulted in a significant decrease in the effluent TP concentration from 7.7 mg/L on day 27 to 0.2 mg/L on day 48. Berryville continued to maintain alum feed rates for the remainder of the startup period to achieve an effluent TP concentration of 0.3 mg/L or less. Laboratory results for Berryville WWTP indicated TN concentrations as low as 4.3 mg/L without supplemental carbon addition, likely as a result of the favorable influent BOD/TKN ratio of about 6.6:1 (Fleischer et al. 2002). However, the supplemental carbon feed system for Berryville WWTP was started in automatic control on day 54 of the performance testing period to achieve the target effluent TN concentration of less than 4 mg/L.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

59

Both plants maintained average effluent turbidity levels below 0.1 NTU throughout the startup period, as shown in Table 3. The average pressure differential across the membranes (i.e. TMP) was minimal for Cape Charles WWTP and well below the maximum allowable TMP of 55 kPa for Berryville WWTP. Permeability data was also monitored for both plants as a relative indicator of membrane fouling and system efficiency, with higher permeability values generally representing decreased fouling and increased efficiency. The permeability of the membranes, reported in units of litres per square meter of membrane surface area per day per kPA of TMP (Lm2d/kPa), was approximately 40% higher for Cape Charles WWTP during the 90-day performance testing period.

The lower average permeability value for Berryville WWTP may have been a result of the limited volume of activated sludge inoculum used to seed the MBR system, as the biomass in the MBR system typically forms a protective layer around the surface of the membranes that can enhance filtering and reduce certain types of fouling (Di Bella et al. 2010).

Table 3 | Startup phase MFS performance data for two municipal MBR plants Parameter

Cape Charles WWTP Average (Min-Max)

Berryville WWTP Average (Min-Max)

Wastewater Temperature (°C)

15.5 (9.6–20.5)

12.4 (7.3–19.8)

Turbidity (NTU)

0.06 (0.04–0.80)

0.06 (0.00–0.76)

Trans-Membrane Pressure (kPa)

0.4 ( 3.5–9.8)

Permeability (Lm2d/kPa)

256 (5–296)

2.0 ( 7.2–2.1) 182 (6–295)

Table 4 | MBR effluent concentrations following plant startup Period

Cape Charles WWTP Average

Berryville WWTP Average

Startup (90 days)

7.1 mg/L TN 0.90 mg/L TP

6.3 mg/L TN 1.10 mg/L TP

0–6 months after Startup

1.7 mg/L TN 0.35 mg/L TP

2.3 mg/L TN 0.06 mg/L TP

6–12 months after Startup

2.0 mg/L TN 0.03 mg/L TP

Not available

System performance was also maintained during extreme conditions, including periods of low/ sporadic flow caused by cyclic operation of the raw wastewater pumps upstream for Cape Charles WWTP and high flow events such as Hurricane Sandy at Berryville WWTP.

CONCLUSIONS There are many different ways to start up an MBR, and each will yield different results, as evidenced herein. The milestones, durations, and sequencing associated with each phase of testing and optimization were especially important for the selected case studies.

In the case of Cape Charles, the Town expressed interest in maximizing nutrient removal through the new MBR facility to take advantage of available nutrient credits. In response to this objective, a significant volume of activated sludge inoculum was utilized for seeding to quickly increase the MLSS concentration in the biological tanks and achieve the desired performance. The chemical feed systems were also started up earlier to enhance nutrient removal during the preliminary stages of plant

60

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

operation. As a result, Cape Charles WWTP achieved ENR-level effluent concentrations of 4.0 mg/L TN and 0.3 mg/L TP approximately 2–3 weeks earlier in the startup process when compared to Berryville WWTP, likely as a result of the increased seeding volume and earlier chemical feed (Manninaa & Di Bella 2012). The larger volume of sludge inoculum also appears to have reduced fouling of the membranes at Cape Charles WWTP during the first 90 days of operation. By maintaining conservative dosing rates for supplemental carbon and alum Cape Charles WWTP has consistently achieved effluent concentrations below 2.0 mg/L TN and 0.05 mg/L TP following the completion of performance testing on the new MBR facilities.

Berryville WWTP took a different approach to startup of the new MBR facilities, choosing instead to develop MLSS concentrations gradually instead of relying on large volumes of activated sludge inoculum. This ultimately reduced the cost of the startup process, but delayed subsequent milestones related to chemical addition and solids processing that were dependent on reaching target solids concentrations in the MBR system. Berryville WWTP demonstrated the ability to meet ENR-level effluent concentrations within the 90-day performance testing period even with limited seeding. Berryville WWTP has averaged effluent concentrations of approximately 2.6 mg/L TN and 0.2 mg/L TP since the completion of startup testing.

Other lessons learned include:

• A project-specific startup plan is an important resource for coordinating between multiple parties, guiding on-site personnel, identifying and preparing for possible setbacks, and addressing variations in startup conditions. Establishing intermediate milestones, such as a target MLSS concentration, a certain number of consistent laboratory results, or a specific change in effluent quality, helped to track progress and focus efforts during startup.

• Point-by-point testing of the automatic controls for the MBR process during clean water testing can potentially resolve critical issues before wastewater is introduced and also serves as a valuable training exercise for plant operators. Thorough clean water testing procedures can take multiple weeks to complete, but can potentially eliminate costly delays or shutdowns during wastewater operation.

• Details of the MBR seeding plan, from the source and delivery method of the activated sludge inoculum to the amount used, can have a significant impact on startup in terms of process performance, scheduling, and potential risks to installed equipment. Screening of the activated sludge inoculum is recommended for protection of the MBR equipment, and the condition of the sludge (age, MLSS concentration, solids retention time of the seed plant, etc.) should be checked for each batch.

• Extensive laboratory testing and data collection is important for accurately estimating startup conditions, confirming influent characteristics, and monitoring performance of the new MBR for a successful startup. In the case of Cape Charles WWTP, the actual influent concentrations and raw wastewater flows at the time of plant startup were considerably different from the values estimated during the preliminary design phase of the project due to changes in growth projections and seasonal variations in the local population.

These and other valuable lessons from the case studies can be applied to the startup phase of future projects.

ACKNOWLEDGEMENTS The authors of this paper would like to thank Bob Panek, Dave Fauber and Patrick Christman (Town of Cape Charles); Dave Tyrrell (Town of Berryville); the staff of the Cape Charles and Berryville WWTPs; Shawn Addison and Jeny Chacko (GE Power and Water); Greg Jablonski and Sebastian Smoot (GHD).

61

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.006

REFERENCES Bott, C., Parker, D., Neethling, J. B., Pramanik, A. & Murphy, S. 2009 WEF/WERF Cooperative Study of BNR Plants Approaching the Limit of Technology: II. Statistical Evaluation of Process Reliability. WEF Proceedings, Nutrient Removal 2009.

Cummings, G. & Frenkel, V. 2008 Membranes for industrial water reuse – they’re not just for municipal applications anymore.

Water Environment Federation 77–91. deBarbadillo, C., Shellswell, G., Cyr, W., Edwards, B., Waite, R., Sabherwal, B., Mullan, J. & Mitchell, R. 2010 Development of Full-scale Sizing Criteria from Tertiary Pilot Testing Results to Achieve Ultra-low Phosphorus Limits at Innisfil, Ontario.

WEF Proceedings, WEFTEC 2010.

Di Bella, G., Durante, F., Torregrossa, M. & Viviani, G. 2010 Start-up with or without inoculum? Analysis of an SMBR pilot plant. Desalination 260, 79–90.

Ellsworth, S. & Riddell, K. 2007 Startup operation and maintenance of the Delphos MBR WWTP. WEF Proceedings WEFTEC 2007.

Ferraris, M., Innella, C. & Spagni, A. 2009 Start-up of a pilot-scale membrane bioreactor to treat municipal wastewater.

Desalination 237 (1–3), 190–200.

Fleischer, E. J., Broderick, T. A., Daigger, G. T., Fonseca, A. D. & Holbrook, R. D. 2002 Membrane bioreactor pilot facility achieves level-of-technology effluent limits. WEF Proceedings, WEFTEC 2002.

Fleischer, E. J., Broderick, T. A., Daigger, G. T., Fonseca, A. D., Holbrook, R. D. & Murthy, S. N. 2005 Evaluation of membrane bioreactor process capabilities to meet stringent effluent nutrient discharge requirements. Water Environment Research 77 (2), 162–178.

Manninaa, G. & Di Bella, G. 2012 Comparing two start-up strategies for MBRs: experimental study and mathematical modelling. Biochemical Engineering Journal 68, 91–103.

Pellegrin, M., Greiner, A. D., Aguinaldo, J., Diamond, J., Gluck, S., Burbano, M. S., Arabi, S., Wert, J., McCandless, R., Padhye, L. P. & Shoaf, R. 2012 Membrane processes. Water Environment Research 84 (10), 1114–1216mi.

Schmidt, H., Glenny, Y., Hirani, Z., Casado, L., Arrebola, V. & Ferguson, J. 2011 Miami-Dade’s reuse pilot program – investigates the limits of technologies to meet ultra-low levels of nutrients and micropollutants. Water Environment Federation 6301–6313.

Vadiveloo, E. & Cisterna, R. 2008 Wastewater reuse: South Florida’s answer to a sustainable 21st century. Water Environment Federation 726–739.

© IWA Publishing 2014 62

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.007

Solar sludge drying in Pafos wastewater treatment plant: operational experiences I. Oikonomidis and C. Marinos ENVITEC AE, 12-14 Ag. Ioannou, Chalandri 15233, Athens, Greece. E-mail: oikonomidis@envitec.gr; marinos@envitec.gr

Abstract In this study, one year of operating experience with a solar sludge drying plant in Pafos wastewater treatment plant is discussed. The plant had a total area of 3,853 m2 and consisted of four parallel lines. Two types of dewatered sludge with considerably different dry matter content were used. 5,678 tn of dewatered sludge were introduced in the plant and the dried sludge had an average dried solids content of 80.9%, which corresponded to a surface annual evaporation rate of 1.14 tn H2O m 2 a 1. The results indicated that solar drying may be a particularly favorable option for sewage sludge drying at climatic conditions such as those prevailing in the East Mediterranean area.

Key words: energy, evaporation, recycling, sewage, sludge, solar

INTRODUCTION Sewage sludge originates from the process of treatment of wastewater and is source of nutrients, mainly phosphorus and nitrogen, organic matter, and energy. Its total quantities throughout Europe are rising due to the progressive implementation of the EU Directive for Urban Waste Water Treatment (UWWT) by the Member States. In 2012, sewage sludge production in EU was nearly 11 million tons (dry matter) and will increase by at least 10% until 2020 (EUREAU 2012).

Sewage sludge disposal routes include landfilling, incineration and recycling, which includes all the processes that result in reuse of the sludge.

In Cyprus, full implementation of the requirements of the EU UWWT Directive for wastewater connection and treatment was expected by the end of 2012. The production of sewage sludge in 2006 was 7,586 tons of dry solids (DS), which gives a specific sludge production of 10 kg DS/capita.a, which is about half the average in EU27. The production is estimated to reach 17,620 tons by 2020 (EC 2008). In Cyprus, recycling of sewage sludge in agriculture is promoted by incorporating the Sewage Sludge Directive 86/278/ EEC into national law (Water Pollution Control Laws 2002–2009) and the Water Pollution Control Regulations 2002 (No. 517/2002). In 2006, 41% of the sludge was used in the agriculture (EC 2008).

Sewage sludge drying

Drying is an important step in all disposal routes: First, it diminishes the bound water present in the sludge, thus resulting in volume and mass decrease, which leads to a cost reduction in transport, handling and storage. Second, drying at high temperatures can kill pathogens and stabilize the sludge to some extent. Finally, the removal of water can increase the calorific value of the sludge, which can be used as acceptable combustible source (Chai 2007). Digested and dried sludges (90% DS and 50% mineral fraction in DS) have a calorific value of 10–15 MJ kg 1 DS, which is equivalent to that of brown coal (Flaga 2009).

63

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.007

Depending on the type of energy delivered in the sludge in order to evaporate water, sewage sludge drying systems can be either thermal or solar. In the thermal systems, exchange of heat and mass between dried sludge and air is achieved through convection, conduction and radiation, while in the solar systems, the solar energy is provided for evaporation (Flaga 2009). The main advantage of solar drying over thermal is the use of free solar energy. The drying effect is based on the vapor pressure difference between the warm sludge, which absorbs the sunlight, and the air in the drying hall (Luboschik 1999). The main design components of solar sludge drying plants (SSDP) are: (1) Sludge heating through the effective usage of the solar energy provided by the greenhouse effect obtained in a greenhouse chamber where the sewage sludge is spread in layers in the floor. (2) Aeration of the surface of the sludge. This is achieved by means of indoor ventilators, which remove the saturated layer formed on the surface of the sludge by providing air turbulence and exhaust fans, which discharge the saturated air accumulated in the plant and provide air renewal. (3) Sludge renewal, e.g. mixing of the sludge for renewal purposes using robots, drums with rotating scrapers etc. (Salihoglu et al. 2007). Several components can be added to increase the efficiency of the SSDP, such as heat pump, infrared lamps, floor heating or rock-bed as thermal energy storage system (Bennamoun 2012). If a second source of energy beside the free solar energy is used, then these are called hybrid SSDP, in contrast with pure SSDP, and may utilize waste heat from combined heat and power unit (CHP).

Solar drying technology has been developed substantially in recent years and economical, automatic systems with additional options like extra heating have been designed by several suppliers such as IST, Thermo System, Huber etc. Of particular interest is the continuous and automatic filling Bilfinger Passavant Water Technologies’ PTS Passavant Trocknungssystem, of which the innovative turning technology is based on continuous sludge distribution/turning with chain scrapper and harrows, thus allowing 19% higher surface area.

This paper discusses one year of operational experience with the Pafos SSDP, from the beginning of June 2012 until the end of May 2013. Performance, energy saving and disposal issues are discussed.

MATERIAL AND METHODS Site description

The Pafos WWTP is located approximately 10 km east of Pafos and serves the city of Pafos and its touristic area. The works was first established in December 2002 with a design flow of 8,100 m3 d 1.

In 2008, the company ENVITEC AE took over the operation of the plant and was responsible for the design and construction of extension works to accommodate a Phase 2 design flow of 19,700 m3 d 1 and thereafter operation and maintenance of the site for a period of 10 years. The extended plant was completed in 2010 as a plant with two new extended aeration activated sludge (AS) lines, operating in parallel with the existing AS plant and a new sludge treatment plant consisting of a mesophilic anaerobic digester, a CHP, a centrifuge and a SSDP. Some part of the stabilized surplus activated sludge (SAS) was homogenized and dewatered in belt filter presses, which were part of the existing plant, while primary sludge was homogenized and fed to the digester. The primary anaerobically digested sludge was then mixed with the remaining quantity of SAS and were dewatered in the centrifuge. The dewatered sludge was then fed into the SSDP. Pafos solar sludge drying plant

The plant was constructed and operated by ENVITEC AE, based on Bilfinger Passavant Water Technologies’ PTS solar PTS Passavant Trocknungssystem and was designed to achieve 70% DS annual

64

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.007

average in the outlet. A four chamber greenhouse mounted on concrete floor was constructed, with total area of 3,853 m2 (104 37.04 m). The process diagram of the plant is shown in Figure 1. It was covered with transparent polycarbonate walls and roof permeable to solar radiation (SR). For each chamber, the mechanically dewatered sludge was dropped from container skips into an intermediate storage bunker, separated by its adjacent by lateral walls. Sludge was fed continuously into the drying plant by four hydraulic pushing rods. Inside the structure, sludge distribution/turning to ensure maximum surface for evaporation took place continuously by chain scrapers with harrows and as sludge dried, it was transported gradually by the devices to the outlet. Across the plant, the sludge bed was no higher than 15 cm. The completely closed cover was fitted with air flaps to allow air circulation. At the inlet, the plant was equipped with exhaust ventilators (four in each chamber) of total capacity of 400,000 m3 h 1, which provided air exchange when the flaps were close as a result of precipitation, high humidity or strong wind. Moreover, vertical ventilators were installed along the roof (two in each chamber, total capacity of 120,000 m3 h 1). At the outlet, dried sludge was withdrawn into a container skip through a belt conveyor. The plant had a design evaporation capacity of 5,760 tn H2O annually, or 1.5 tn H2O m 2 a 1. The plant is presented in Figure 2.

Figure 1 | Process diagram of Pafos SSDP (one of the four identical chambers is depicted together with the outlet of the plant): (a) intermediate storage bunker and pushing rods; (b) chain scrapper with harrows; (c) air flaps; (d) exhaust ventilators; (e) vertical ventilators; (f) sludge belt conveyor; (g) sludge screw conveyor.

During the process, a meteorological station provided continuous monitoring of ambient temperature (T), relative humidity (RH), wind speed and direction, precipitation, global SR and atmospheric pressure. Data were received every 2 seconds by the MKIII-LR sensor assembly (RainWise, Inc.).

Indoor T and RH was also measured. Three PLCs controlled all the electromechanical components of the plant automatically, while a SCADA system allowed process control and data logging.

There were two basic operation modes of the turning device; the fully automatic and the semi-automatic mode. In the former, data from the weather station were used to control the movement of the turning device automatically. In the latter, five different options were available, depending on the desired speed of sludge.

The total solids (DS) of the sludge and the fecal coliform content were measured in accordance with the Standard Methods for the Examination of Water and Wastewater (1998).

65

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.007

Figure 2 | Pafos SSDP: (A) overview; (B) intermediate storage bunkers and exhaust ventilator outlets; (C) sludge transportation by chain scrappers with harrows; (D) chain scrapper drive station and dried sludge belt conveyor at the outlet of the plant.

SSDP operation

During the first six months of operation, three chambers of the plant (chambers one to three) were fed with sludge dewatered in the belt filter presses (with DS concentration of 14.6% on average) and one (chamber four) was fed with centrifuged sludge (average DS concentration was 23.7%). During the remaining six months, chambers one to three were fed with both centrifuged and belt filter press sludge, depending on the relative quantities produced by each of the dewatering devices, while chamber four was fed only with centrifuged sludge, as before. The turning device operated semi-automatically. Basic maintenance of the SSDP included some harrows replacement and in some cases manual labor inside the chambers to assist sludge transport.

RESULTS AND DISCUSSION Meteorological parameters

The average values of the main climatic parameters throughout the year are shown in Figure 3. As expected, there was a gradual decrease in SR, ambient and indoor T during the winter months. As

66

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.007

Figure 3 | Variations in the main meteorological parameters in Pafos SSDP.

a result of the greenhouse effect, the average T inside the plant was consistently higher than that outside by 2.86 + 1.09 K. This difference was similar to that reported in Bux et al. (2001), who carried out trials in commercial full-scale SSDP for municipal sludge. Similarly, the ambient RH was higher than the indoor RH by 4.44 + 3.22%. The differences between ambient and indoor T and ambient and indoor RH were decreasing as weather was becoming cooler. The average daily SR over the year was 5.21 kWh m 2 d 1 which is typical of the climatic conditions in Cyprus (Kalogirou 2006).

Figures 4 and 5 shows the variations in the main meteorological parameters on a typical summer day and on a sunny fall day (15/06/2012 and 20/11/2012). Ambient and indoor T and ambient and indoor RH were nearly the same at night, when cooling occurred due to SR absence. The greenhouse effect took place during the day and there was a larger increase in indoor T than in ambient T, with the former reaching a maximum value of 45 째C during the summer period. Similar trends were observed for RH.

Figure 4 | Variations in the main meteorological parameters on 15/06/2012.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.007

67

Figure 5 | Variations in the main meteorological parameters on 20/11/2012.

SSDP performance

The process parameters of the plant for the one year period are shown in Table 1.

Table 1 | Process parameters in Pafos SSDP Chamber Units 1 1

Moist sludge loading

tn a

Moist sludge DS content

%

Moist sludge DS loading

tn DS a

Dried sludge discharge

tn a

2

1195 14.9 1

1

Dried sludge DS content

%

Evaporation rate

tn H2O a

Annual surface evaporation rate

tn H2O m

3

1334 16.5

4

1205 17.6

Total

1944

5678

23.7

173

217

209

456

1055

214

268

259

564

1305

954

1059

939

1380

4332

80.9 1 2

a

1

1.03

1.10

0.98

1.43

1.14

Over the year, 1.14 tn H2O per m2 were removed. This value was quite higher than the rate of 0.75 tn H2O m 2 a 1 reported in Germany (Luboschik 1999). Due to sticking problems and harrows strain potentially occurring during winter for moist sludge with DS content less than 16%, mostly lower loading rates were applied in chambers one to three than in chamber four, in which moist sludge had a higher DS content (23.7% on average).

The performance of the plant in terms of monthly average dried sludge DS content over the year is shown in Figure 6. Overall, the average dried sludge DS content was 80.9 + 11.9%, which exceeded the 70% DS design minimum annual average. The highest average value was observed in June (91.3%) and the lowest in December (55.7%). The performance of the plant in terms of dried sludge DS content was clearly related to the seasonal variations in indoor T. As a result, the lowest DS contents were observed between November 2012 and March 2013, when the daily average indoor T remained mostly below 20 째C.

68

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.007

Figure 6 | Monthly average DS content of the dried sludge in Pafos SSDP.

The variations in the surface evaporation rate in all chambers are depicted in Figure 7 (based on a year). The rates varied according to the quantities of moist sludge entering the chambers (which resulted from the primary and secondary sludge production in the WWTP) and the DS content of the moist and dried sludge. The low evaporation rates in December were due to limited moist sludge loading. As suggested by the evaporation rates in chamber four, more water than that deďŹ ned by the design evaporation capacity could be removed in the SSDP, both in warm and cool months, especially when the moist sludge DS content was higher than 20%. In particular, the evaporation capacity was exceeded by 38% in June in this chamber.

Figure 7 | Variations in evaporation rates in Pafos SSDP.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.007

69

Substantial changes in the physical properties of the sludge were evident during drying. As sludge dried and transported gradually to the outlet, there was a decrease in the mean particle size, with particles with sizes less than 1 cm predominating at the outlet. Moreover, the surface of the particles became rough and amorphous as a result of extensive shrinkage and crack development. Those are two well-established phenomena during wastewater sludge drying (Bennamoun 2012).

Energy consumption

The daily thermal energy demand for water evaporation was 11,393 kWh d 1, assuming that 1 ton of sludge requires 960 kWh of thermal energy. Compared to the average daily SR over year, it was concluded that approximately 56% of it was effective for evaporation. This was partly because of transmission losses. The specific electrical energy consumption was 77.3 KWh tn 1 H2O. This consumption was only 8.0% of the free thermal energy consumption and also much less than the total energy demand of thermal driers as reported in the literature (800–1000 kWh tn 1 H2O) (Bux et al.

2001). Similar electrical energy requirements have been reported elsewhere (Mathioudakis et al.

2010). On average, 59% of the electrical energy demand was consumed by the exhaust ventilators, 28% by the turning devices and the remaining 13% by the hydraulic device, the indoor ventilators and the conveyors. The climatic conditions, i.e. wind velocity, precipitation and humidity strongly influenced the electrical energy consumption by controlling the operation of the exhaust ventilators.

Disposal options

Currently, the Sewage Board of Pafos is responsible for the monitoring and management of the dried sludge from Pafos SSDP, according to the provisions of the 70/2012 Wastewater Management Permit.

The sludge is used as a fertilizer and soil conditioner in selected farms. The heavy metal concentrations in the final sludge product were well within the limits imposed by the Wastewater Management Permit, as shown in Table 2 (data given by the Sewerage Board of Pafos).

Table 2 | Heavy metals content in dried sludge from Pafos SSDP (sample collected on 21/2/2013) Metal

Units

Value

Management permit limit

mg kg

1

DS

Chromium

mg kg

1

DS

30.54

1,000

Copper

mg kg

1

DS

184.21

1,000

Lead

mg kg

1

DS

20.55

750

Zinc

mg kg

1

DS

439.28

2,500

Nickel

mg kg

1

DS

14.54

300

Mercury

mg kg

1

DS

0.75

10

Cadmium

,0.002

10

The fecal coliforms in dried sludge were measured on 11/07/2012 and on 21/02/2013 and were 2 103 and 1.2 103 CFU gr 1 DS, respectively (latter data given by the Sewerage Board of Pafos).

Measurements in the dewatered sludge showed that fecal coliforms were reduced by approximately two orders of magnitude (data not shown). Moreover, no eggs of intestinal parasites were found (measured on 21/02/2013). Despite the considerable pathogen reduction, the EPA Class A pathogen reduction requirement of 103 CFU gr 1 DS was not met. This was in agreement with previous research on sanitization of sludge subjected to solar drying, which showed that even though pathogen numbers were of the same order of magnitude as the EPA Class A limits, those were not satisfied (Bux et al. 2001, Salihoglu et al. 2007, Ogleni & Ozdemir 2010, Mathioudakis et al. 2010), It has been

70

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.007

reported that limiting liming (15%–20% of the DS before solar drying) was effective in achieving the EPA Class A pathogen reduction requirement (Bux et al. 2001, Salihoglu et al. 2007).

CONCLUSIONS In this study, one year of operational experience with the solar sewage drying plant in Pafos WWTP is presented. The plant discharged dried sludge with DS content of 80.9% on average and this was higher than the design 70%. It was shown that at cooler weather there was a decrease in plant performance in terms of DS content. The annual surface evaporation rate was 1.14 tn H2O m 2 a 1.

Evaporation rates higher than the design ones occurred, especially when moist sludge DS content was higher than 20%. The daily thermal energy consumption of 11,393 kWh d 1 was covered exclusively by solar energy free of charge. The electrical energy demand was only 8% of the thermal energy one. With more than 300 days of the year considered as having sunny weather, it is proved that solar resource is a suitable candidate to be used as a renewable energy source for sludge drying in Cyprus.

The dried product meets the requirements for agricultural reuse in terms of heavy metals content.

Moreover, there was a considerable pathogen reduction along the process, even though the sludge did not meet the EPA Class A pathogen reduction requirements. Use as a renewable biofuel could also be an alternative disposal option.

REFERENCES Bennamoun, L. 2012 Solar drying of wastewater sludge: a review. Renewable and Sustainable Energy Reviews 16, 1061–1073.

Bux, M., Baumann, R., Philipp, W., Conrad, T. & Muhlbauer, W. 2001 Class-A by solar drying recent experiences in Europe. In: Proc. WEFTEC, Atlanta, USA, 2001.

Chai, L. H. 2007 Statistical dynamic features of sludge drying systems. International Journal of Thermal Sciences 46, 802–811.

EC 2008 Environmental, economic and social impacts of the use of sewage sludge on land, Final Report. http://ec.europa.eu/ environment/waste/sludge/index.htm (accessed 9 May 2013).

EUREAU 2012 Positions paper on how the revision of the Fertiliser Regulation should promote sustainable use of sludge in agriculture. http://eureau.org/sites/eureau.org/files/documents/2012.03.29_EUREAU_PP_Sust%20use_Sludge_in_agri. pdf (accessed 10 May 2013).

Flaga, A. 2009 Sludge drying. In: Proc. Polish-Swedish-Ukrainian Seminar Research and Application of new Technologies in Wastewater Treatment and Municipal Solid Waste Disposal in Ukraine, Sweden and Poland (E. Plaza & E. Levlin, eds).

Lviv, Ukraine, 26–28 October, 2006.

Kalogirou, S. A. 2006 Environmental friendly energy sources in Cyprus. In: Proc. 2nd Eco Forum, on CD-ROM, Nicosia, Cyprus, 2006.

Luboschik, U. 1999 Solar sludge drying-based on the IST process. Renewable Energy 16, 785–788.

Mathioudakis, V. L., Kapagiannidis, A. G., Athanasoulia, E., Diamantis, V. I., Melidis, P. & Aivasidis, A. 2009 Extended dewatering of sewage sludge in solar drying plants. Desalination 248, 733–739.

Mathioudakis, V. L., Kapagiannidis, A. G., Athanasoulia, E., Paltzoglou, A. D., Melidis, P. & Aivasidis, A. 2010 Sewage sludge solar drying in a pilot scale greenhouse. In: Proc. Protection and Restoration of the Environment X, Corfu, Greece, 5–9 July 2010.

Ogleni, N. & Ozdemir, S. 2010 Pathogen reduction effects of solar drying and soil application in sewage sludge. Turkish Journal of Agriculture and Forestry 34, 509–515.

Salihoglu, N. K., Pinarli, V. & Salihoglu, G. 2007 Solar drying is sludge management in Turkey. Renewable Energy 32, 1661– 1675.

Standard Methods for the Examination of Water and Wastewater 1998 20th edn, American Public Health Association/ American Water Works Association/Water Environment Federation, Washington DC, USA.

© IWA Publishing 2014 71

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.008

A study of the digestion process of sewage sludge from a dairy wastewater treatment plant to determine the composition and load of reject water W. Da˛ browski Faculty of Civil and Environmental Engineering, Bialystok University of Technology, 45A Wiejska St., 15-355 Bialystok, Poland E-mail: dabrow@pb.edu.pl

Abstract The issue of reject water has to be considered in almost every biological municipal or industrial wastewater treatment plant (WWTP) that applies aerobic or anaerobic digestion of sewage sludge. Reject water is usually returned to the beginning of the treatment process, which results in periodical disturbances in stable and efficient sewage treatment. Due to planned modernization of one of the biggest dairy WWTPs in Poland, a laboratory scale research has been carried out to determine quality characteristics of reject water. Aerobic and anaerobic digestion was applied to a mixture of two kinds of sludge: excessive and flotation. According to research performed by the author results (range value) of reject water were: 7.3 to 12.9 mg N-NH4/L after aerobic and 460.0 to 574.0 mg N-NH4/L after anaerobic digestion. The study has confirmed a higher value of organic substances in reject water after anaerobic digestion in comparison with aerobic. Due to high concentration of ammonia nitrogen in reject water obtained during co-digestion of excessive and flotation sludge, a separated system for its treatment should be applied. The results of research work presented in this paper provided a base for the project of the pilot installation with constructed wetland.

Key words: aerobic and anaerobic digestion, dairy WWTP, reject water, sewage sludge

INTRODUCTION The problem of reject water from sewage sludge treatment refers mainly to the wastewater treatment plants (WWTPs) that are characterized by personal equivalent (PE) of over 100,000 and apply anaerobic digestion of the excess sludge. Smaller WWTPs generally use aerobic digestion of the excessive sludge. This process takes place in separate chambers or simultaneously in the aeration tanks. In these WWTPs the amount of sludge is too low. The anaerobic digestion along with electricity and heat, as well as biogas production, is not profitable. The load of reject water in municipal treatment plants which apply aerobic sludge digestion is low in relation to the raw sewage load, thus there is no need for the separate treatment. Studies carried out in two such WWTPs having similar PE showed that the share of reject water load in raw sewage load ranged from 1.1 to 1.7% for five-day biochemical oxygen demand (BOD5) and from 2.8 to 4.9% for N-NH4. The quantity of reject water ranged from 2.1 to 2.9% of the raw sewage stream (Da˛ browski 2010). In the case of a municipal WWTP which uses anaerobic sludge digestion, the high concentration of ammonia nitrogen in reject water is the biggest problem. Reject water affects significantly the treatment process and the quality of treated sewage discharged into the receiver. According to a study conducted by Ryziń ska in Polish WWTPs, the quantity of reject water ranged from 2.7 to 7% of the amount of raw sewage, while the share of ammonia nitrogen load in the raw sewage ranged from 15.6 to 47% (Ryziń ska 2006). According to a study conducted

by Fux as well as Gajewska and Obarska-Pempkowiak, the share of ammonia nitrogen load in raw

72

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.008

sewage load ranged from 2 to 20% (Fux et al. 2006; Gajewska & Obarska-Pempkowiak 2011). In turn, research carried out in Dutch WWTPs showed a 25% share of the total nitrogen load in relation to the load in raw sewage. The quantity of reject water less than 2% of raw sewage flow (Janus & van der Roest 1997). A similar situation was observed during research conducted in Germany when 20 WWTPs were examined. The load of ammonia nitrogen was up to 25% of the load observed in raw sewage (Meyer & Wilderer 2004).

In dairy WWTPs that use aerobic digestion of sewage sludge, the problem of reject water from the sludge treatment did not result from a high concentration of ammonia nitrogen. In the author’s study on the characteristics of reject water conducted in 2008 in Mlekovita WWTP in Wysokie Mazowieckie, which applied the excessive sludge aerobic digestion, it was indicated that the concentration of ammonia nitrogen in reject water ranged from 4.9 to 26.4 mg N-NH4/L, while in the raw dairy wastewater it ranged from 1.1 to 8.3 mg N-NH4/L (Da˛ browski 2009). Given that the amount of reject water in the dairy wastewater stream was at approximately the 10% level, a large share of the nitrogen load from the reject water was returned to the biological treatment (Da˛ browski 2007, 2010). Till the end of 2012, anaerobic sludge digestion was not used in dairy WWTPs in Podlasie (east part of Poland). Modernization of the WWTPs was based on keeping the current method of sewage treatment. The flotation process was commonly introduced, which contributed to the reduction in the load within the aeration chambers, and on the other hand, led to the production of flotation sludge. Application of anaerobic sewage sludge stabilization has lots of advantages – the most important being biogas production. The aerobic stabilization process is less expensive (building and exploitation) but there is no possibility of energy production.

Modernization of the largest dairy WWTP in Poland located in Wysokie Mazowieckie started in 2012. According to the project, it will be treating about 7,500 m3 sewage daily, while PE will reach 350,000.

After the modernization, the process of aerobic excess sludge digestion will be replaced by anaerobic digestion of mixed sludge: excessive after thickening and flotation. The modernization project does not include separate treatment of reject water.

The aim of this study was to determine the characteristics of reject water produced after the sludge digestion process in the dairy WWTP and the possibility of its separate treatment. Laboratory studies of the digestion process were carried out due to lack of experiments associated with both the aerobic and anaerobic co-digestion of excessive and flotation sludge. Results from the study will be used to optimize the treatment process at the end of the plant modernization in 2013. Eventually, most of the dairy WWTPs in Poland will be operating with anaerobic sewage sludge digestion. Designers should consider the necessity of separate reject water treatment. The results will also be applied in the design and building of the pilot installation for reject water treatment. Research results have been used to determine the load of reject water after application of anaerobic excessive and flotation sludge digestion.

METHODS The subject of research conducted in 2012 was the digestion process of mixed excessive and flotation sludge from a dairy WWTP. The sludge samples were prepared according to the modernization project of Mlekovita in Wysokie Mazowieckie WWTP. The excess sludge made up 70% of the load, while flotation sludge was 30% of the load when recalculated as a dry mass (DM). The load of the chamber for anaerobic digestion was approximately 0.7 kg DM/m3 per day, which is a typical value for sludge stabilization (Bień 2002). Anaerobic and aerobic digestion of mixed sludge samples were carried out simultaneously. This allowed the comparison of the characteristics of reject water produced during the digestion processes.

73

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.008

Figure 1 | Research installation for anaerobic sewage sludge digestion.

The research installation for anaerobic digestion consisted of three closed chambers equipped with stirring and heating devices (Figure 1). Anaerobic digestion process parameters were adopted from the project of the WWTP modernization. The process was carried out at 35 °C (mesophilic digestion) with a solid retention time (SRT) of 20 days.

The aerobic digestion was conducted using three aeration chambers equipped with fine bubble diffusers. The SRT of the aerobic digestion process was the same as for the anaerobic. During the tests, the constant content of dissolved oxygen was kept at the level of 1.0 mg/L.

Sampling and scope of the determination

In order to obtain reproducible results, nine series were performed. The start up of anaerobic stabilization chambers was done with inoculums taken from the municipal WWTP digestion chamber. In every series the decrease of organic matter and the production of biogas containing more than 60% methane were the indicators of stabilization. Upon the completion of each series, samples of sludge from anaerobic and aerobic digestion chambers were collected. Sludge was dehydrated using a special cloth used in the bag filter press, then physical and chemical analyses were performed. Values and concentrations of the following pollutants were measured:

• Organics (BOD5, chemical oxygen demand [COD], total organic carbon [TOC]), • Total suspended solids (TSS), • Nitrogen forms: total Kjeldahl nitrogen (TKN), ammonia nitrogen NH4-N, nitrate nitrogen N-NO3 and nitrite nitrogen N-NO2.

• Total phosphorus.

Determinations were conducted in a certified laboratory in accordance with the procedures set out in the Regulation of the Environmental Protection Minister from 24 July 2006 and in accordance with the American Public Health Association APHA (2005). Tests of the analysis of COD, TOC, Kjeldahl nitrogen, NH4-N, NO3-N, N-NO2, and total phosphorus recommended by Merck were applied. BOD was determined using OXI-TOP. In order to assess the degree of sludge digestion, besides studies on reject water, organic matter content in anaerobic and aerobic digested sludge was determined as well.

The results were statistically evaluated using the StatSoft STATISTICA 8.0.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.008

74

RESULTS AND DISCUSSION The achieved results are shown in Tables 1–3. Median, minimum and maximum values, as well as the average values, are presented for the statistical evaluation of the process. Table 1 shows the results from studies on the characteristics of the sludge subjected to the digestion process. The effect of the digestion process was similar in both cases. After 20 days, the content of organic matter decreased to an average of 54.1% for the anaerobic and to 58.7% DM for the aerobic digestion. The initial content of organic matter in mixed sludge subjected to anaerobic and aerobic digestion ranged from 78.0 to 84% DM, (80.1%, on average).

The project of modernization of the WWTP in Wysokie Mazowieckie was assumed to reduce the content of organic matter to about 50% DM. This value is often quoted in literature when describing the digestion of sludge under mesophilic conditions (Bień 2002; Bordeleau & Droste 2011). During the study, a major decrease in organic matter content was observed at the initial phases of both anaerobic and aerobic digestion.

Table 2 illustrates the characteristics of reject water after the digestion process. Substantial differences in its composition were observed, both for organic matter content measured by the BOD5, COD and TOC as well as the concentrations of total nitrogen (TN) and ammonia. The average value of BOD5 in reject water after anaerobic digestion was more than five times higher as compared Table 1 | The characteristics of sewage sludge Type of sludge

Organic substance (%)

Sludge before digestion

80.1/79.0 (78.0–84.0)

After anaerobic digestion

54.1/52.0 (49.0–60.0)

After aerobic digestion

58.7/55.0 (53.0–66.0)

Median/mean (minimum – maximum).

Table 2 | The characteristics of reject water Parameter

Anaerobic digestion

BOD5

468.3/420.0 (310.0–660.0)

COD

952.3/912.0 (730.0–1,250.0)

TOC

98.2/96.0 (90.0–112.0)

TSS

231.7/234.0 (205.0–260.0)

TN

554.8/548.3 (510.3–621.6)

NH4-N

527.8/530.0 (465.0–574.0)

Aerobic digestion

87.3/90.0 (75.0–95.0) 179.9/185.0 (160.0–190.0) 38.2/38.0 (35.0–42.0) 196.8/190.0 (180.0–217.0) 40.2/40.2 (37.1–43.6) 9.8/9.1 (7.3–12.9)

Org-N

26.4/29.0 (5.0–47.0)

NO3-N

0.4/0.4 (0.2–0.6)

1.2/1.0 (0.9–1.5)

NO2-N

0.1/0.1 (0.1–0.2)

0.2/0.2 (0.1–0.2)

Ptot

36.3/36.9 (32.0–39.7)

29.0/28.7 (28.1–30.3)

19.6/19.1 (17.0–22.4)

Median/mean (minimum – maximum).

Table 3 | The characteristics of bioavailability of reject water Parameter

Anaerobic digestion

Aerobic digestion

BOD/COD

0.49

0.48

BOD/TN

0.84

2.17

75

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.008

with the value obtained after the aerobic digestion. A similar relationship was observed in the case of COD and TOC. BOD5 and COD values were similar to those reported by Gajewska & ObarskaPempkowiak (2008, 2011), who conducted research related to the treatment of reject water after anaerobic digestion with the constructed wetlands (CW) method. The amount of total suspended solids was similar after completing both processes. During decomposition of organic matter in anaerobic conditions, CH4, CO2 and H2O are produced. The analysis of nutrient concentration in reject water indicates that nitrogen from anaerobic digested sludge, as expected, occurs mainly in the form of ammonia ions. The concentration of total nitrogen is more than 13 times higher in the case of reject water from the anaerobic versus aerobic digestion process. This disproportion is even greater when compared to ammonia nitrogen concentration, which was 527.8 mg N-NH4/L after anaerobic and 9.8 mg N-NH4/L after aerobic digestion. A low concentration of ammonia nitrogen was caused by nitrification and denitrification processes. Aerobic stabilization resulted in a low quantity of organic substances in reject water. Organic nitrogen, nitrate, and nitrite concentrations were similar after both the aerobic and anaerobic digestion processes. In the case of phosphorus, a slightly higher concentration was observed after the anaerobic digestion. The concentration of ammonia nitrogen in reject water from the anaerobic digestion process is similar to that reported in the literature by Bień (2002), Fux et al. (2006), Ryziń ska (2006), Janus & van der Roest (1997), as well as Gajewska & Obarska-Pempkowiak (2008, 2011). The results from studies on reject water after aerobic digestion are similar to those achieved by the author during the research of the characteristics of reject water from treatment of sewage sludge in the dairy WWTPs in Podlasie province (Da˛ browski 2007, 2009, 2011). The studies included nine individual dairy wastewater treatment plants that used simultaneous or separate aerobic digestion of excessive sludge. According to a study performed in 2008, the average value of BOD5 in reject water amounted to 114 mg O2/L, ammonium nitrogen 18.2 mg NNH4/L, and total phosphorus 7.3 mg P/L (Da˛ browski 2009).

To determine the susceptibility of reject water to biological degradation, values of BOD5 to COD and BOD5 to total nitrogen ratios were calculated and presented in Table 3. Any biological process requires BOD5/COD .0.45, and BOD5/N .4 (Heidrich et al. 2008). Results achieved during laboratory tests referring to reject water after both the aerobic and anaerobic digestion demonstrate its low susceptibility to biological decomposition during its biological treatment.

The results are similar to those reported by Gajewska & Obarska (2011). Similar values of BOD5/ COD occur in the leachate from municipal landfills after many years of exploitation (Wojciechowska et al. 2010). The BOD5/TN ratio was 0.84 for reject water after anaerobic digestion, while 2.17 after aerobic digestion. The analysis reveals that reject water after aerobic digestion is much more susceptible to biodegradation.

Determining the load of reject water from digestion process and possibilities for its treatment in dairy WWTP

Returning the reject water to the beginning of the WWTP increases the load of biological treatment carried out in sludge activated chambers and, as a consequence, the efficiency of organic substances and the removal of nutrients. In the case of the dairy WWTP belonging to the Mlekovita Company in Wysokie Mazowieckie, a change in the sewage sludge digestion would result in a significant increase in the sewage load due to the reject water being returned to the main stream of the treatment line.

According to the exploiter’s data and own research conducted from January to September 2012, the average sample of raw sewage which included reject water in this period was approximately 6,000 m3/d, while the average value of BOD5 amounted to 2,800 mg O2/L, ammonia nitrogen to 2.5 mg N-NH4/L, and total nitrogen to 170 mg N/L. The sewage treatment facility still exclusively digests the excessive sludge under aerobic conditions during the WWTP modernization, and the amount of reject water is about 400 m3/d, on average. The average content of BOD5 in reject

76

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.008

water before modernization was 100 mg O2/L, concentration of ammonia nitrogen 12.0 mg N-NH4/ L, and total nitrogen 36 mg N/L. As a result, the load due to reject water in the WWTP is minimal and amounts to about 0.2% for BOD5. In the case of ammonia, the load of reject water in relation to raw wastewater load was as high as 32%.

It was assumed that the load of raw dairy sewage and the amount of reject water after starting the anaerobic digestion process would not change. Taking into consideration the parameters for reject water obtained during this study, the results of which are shown in Table 2, the reject water load in reference to the raw sewage load is expected to be 1.1% for BOD5 and 21% for total nitrogen.

The ammonia nitrogen load in the reject water would be approximately 14 times higher than the load specified for the raw dairy sewage. The share of sewage load (ammonia nitrogen) in the reject water load from anaerobic digestion should be about 7.1%.

Such a large estimated load of the WWTP with ammonia nitrogen contained in the reject water that would be produced after starting the anaerobic digestion process under mesophilic conditions will require the use of separate treatment. Advanced technologies applied for the treatment of reject water are known nowadays. The most common is the SHARON method (Kempen et al. 2001; Meyer & Wilderer 2004; Jenicek et al. 2006). However, they are expensive and complicated. Constructed wetlands may be an alternative to them. CW provide a long-term removal of pollutants along with stability and high efficiency at very low operating costs (Gajewska et al. 2004; Bulc 2006; Kinsley et al. 2006; Obarska-Pempkowiak et al. 2010; Wojciechowska et al. 2010). The author’s own research (Da˛ browski 2009) of the treatment effectiveness of reject water from aerobic excessive sludge digestion in dairy WWTPs was the basis for the design and construction of the implementation system at the dairy WWTP in Bielsk Podlaski. Also, a pilot study conducted in Gdansk municipal WWTP confirmed the possibility of applying a hybrid system based on a vertical and horizontalflow CW for the treatment of reject water in a municipal sewage treatment plant which applies anaerobic digestion (Gajewska & Obarska-Pempkowiak 2010).

Achieved results were used to design the pilot installation, which was built in autumn 2013 in the dairy WWTP Mlekovita in Wysokie Mazowieckie. A similar installation operates in Bielsk Podlaski dairy WWTP (Da˛ browski 2013). It consists of two constructed wetlands (vertical and horizontal) that can be used in different configurations and will be supplied with reject water from the anaerobic co-digestion of excessive and flotation sludge. The study will include the specification of the parameters for such reject water treatment and guidelines for designing a system for its treatment in full scale in order to reduce the load of the dairy WWTP.

CONCLUSIONS The study showed that the use of anaerobic sludge digestion for a mixture of excessive and flotation sludge in a dairy WWTP will result in a significant increase in the load of pollution in reject water.

Estimated load of pollutants in reject water measured by means of BOD5 can be increased up to five times compared to that observed in the current process of separate or simultaneous aerobic digestion of excessive sludge. The main problem will result from an increase in ammonium nitrogen load and total nitrogen in reject water. According to the research and data on treatment parameters applied in Wysokie Mazowieckie in 2012, the load of ammonia nitrogen in reject water from the anaerobic digestion treatment of excessive and flotation sludge will be much higher than the load contained in the raw dairy sewage. It is necessary to apply a separate reject water treatment to reduce the load of the biological part of the dairy WWTP. The analysis of reject water composition confirms the desirability of the use of a constructed wetland system for its treatment. The pilot testing system

77

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.008

designed on the basis of the survey results was installed in 2013 along with the start of exploiting the anaerobic digestion in the dairy WWTP.

ACKNOWLEDGEMENTS The research presented in the paper has been done within the frames of the statute activity S/WBiIŚ 4/ 2011 implemented at the Department of Engineering Technology and Environmental Protection, Bialystok University of Technology (BUT), as well as with technical and scientific cooperation between BUT and Mlekovita in Wysokie Mazowieckie. Special thanks go to Marek Kajurek, the dairy WWTP manager.

REFERENCES American Public Health Association (APHA) 2005 Standard Methods for Examination of Water and Wastewater 21st edn, American Public Health Association, Washington, DC. Bień , J. 2002 Osady sć iekowe: teoria i praktyka (Sewage sludge: theory and practice). Cze˛stochowa University of Technology.

Bordeleau, E. L. & Droste, R. L. 2011 Comprehensive review and compilation of pre-treatments for mesophilic and thermophilic anaerobic digestion. Water Science & Technology 63 (2), 291–296.

Bulc, T. G. 2006 Long term performance of a constructed wetland for landfill leachate treatment. Ecological Engineering 26, 365–374.

Da˛ browski, W. 2007 Study upon the quality of reject water arising from aerobic treatment of sewage sludge in dairy wastewater treatment plants. Polish Journal of Environmental Studies 16 (2A), 31–35.

Da˛ browski, W. 2009 Metoda hydrofitowa do oczyszczania odcieków z tlenowej stabilizacji osadów w oczyszczalni sć ieków mleczarskich, Monografia P.A.N. Gospodarka odpadami komunalnymi, (Constructed wetlands for reject water treatment in dairy wastewater treatment plant, Monograph Municipal waste disposal in 2009). Polish Academy of Science, Koszalin 2009 V, 205–214.

Da˛ browski, W. 2010 Charakterystyka odcieków z tlenowej przeróbki osadów w komunalnych i przemysłowych oczyszczalniach województwa podlaskiego (Characteristic of reject water from aerobic digestion in municipal and industrial WWTPs of podlaskie province in 2010). Engineering and Protection of the Environment 13 (1), 43–51.

Da˛ browski, W. 2011 Removal of organic and biogenic compounds from reject water with constructed wetlands. Ecological Chemistry and Engineering 18 (9–10), 1203–1214.

Da˛ browski, W. 2013 Okresĺ enie moz˙ liwosć i zmniejszenia obcia˛ z˙ enia oczyszczalni sć ieków mleczarskich przez zastosowanie wydzielonego oczyszczania odcieków z przeróbki osadów Rocznik Ochrona Ś rodowiska (The possibility of dairy WWTP load decreasing using constructed wetland for reject water treatment). Annual Set of Environment Protection 15, 901–913.

Fux, C. h., Valten, S., Carozzi, V., Solley, D. & Keller, J. 2006 Efficient and stable nitrification and denitrification of ammoniumrich sludge dewatering liquor using SBR with continuous loading. Water Research 40 (14), 2765–2775.

Gajewska, M. & Obarska-Pempkowiak, H. 2008 Effect of recycling of reject waters from dewatering the sludges on the operations efficiency of a wastewater treatment plant. Przemysł Chemiczny 87 (5), 448–451.

Gajewska, M. & Obarska-Pempkowiak, H. 2011 The role of SSVF and SSHF beds in concentrated wastewater treatment, design recommendation. Water Science & Technology 64 (28), 431–439.

Heidrich, Z., Kalenik, M., Podedworna, J. & Stań ko, G. 2008 Sanitacja wsi. Rural Areas Sanitation. Seidel-Przywecki, Warszawa.

Janus, H. M. & van der Roest, H. F. 1997 Do not reject the idea of treating reject water. Water Science & Technology 35 (10), 27– 34.

Jenicek, P., Svehla, P., Zabranska, J., Dohanyos, M. & Vondrysova, J. 2006 Denitrification of reject water using primary sludge as organic substrate. Treatment of reject waters after sludge dewatering, IWA congress, Moscow, pp. 538–543.

Kempen, R., Mulder, J. W., Uijetrlinde, C. A. & Loosdrecht, M. C. M. 2001 Overview: full scale experience of the SHARON process for treatment of rejection water of digested sludge dewatering. Water Science and Technology 44 (1), 145–152.

Kinsley, C. B., Crolla, A. M., Kuyucak, N., Zimmer, M. & Lafléche, A. 2006 Nitrogen dynamics in a constructed wetland system treating landfill leachate. In: Proc. of 10th International Conference on Wetland Systems for Water Pollution Control, Portugal, pp. 295–305.

Meyer, S. S. & Wilderer, A. 2004 Reject water: treating of process water in large wastewater treatment plants in Germany-A case study. Journal of Environmental Science and Health 39 (7), 1645–1654.

Nosrat, M., Amani, T. & Sreekrishnan, T. R. 2011 Thermophilic anaerobic digestion of waste activated sludge versus mesophilic anaerobic digestion. International Conference on Advances in Biotechnology and Pharmaceutical Sciences (ICABPS’2011), Bangkok, pp. 226–229.

78

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.008

Obarska-Pempkowiak, H., Gajewska, M. & Wojciechowska, E. 2010 Application of vertical flow constructed wetland for highly contaminated wastewater treatment: preliminary results. In: Water and Nutrient Management in Natural and Constructed Wetlands (J. Vymazal, ed). Springer Science þ Busines Media B.V., pp. 37–51.

Rozporza˛ dzenie Ministra Ś rodowiska z dnia 24 lipca 2006 roku w sprawie warunków jakie nalez˙ y spełnić przy wprowadzaniu sć ieków do wód lub ziemi, oraz w sprawie substancji szczególnie szkodliwych dla sŕ odowiska wodnego Regulation of the Minister of Environment of 24th of July 2006 on conditions to be met for disposal of treated sewage into the water and soil and due to substances harmful to the environment (official Journal no. 137, 987, Poland, pp. 9787–9820 (in Polish).

Ryziń ska, J. 2006 Problem wód osadowych i moz˙ liwosć i ich oczyszczania w Polsce, Problems of reject water in Poland. Gaz Woda i Technika Sanitarna 7–8, 58–62. (In Polish).

Song, Y. C., Kwon, S. J. & Woo, J. H. 2006 Mesophilic and thermophilic temperature co-phase anaerobic digestion compared with single-stage mesophilic- and thermophilic digestion of sewage sludge. Water Research 38 (7), 1653–1662.

Wojciechowska, E., Gajewska, M. & Obarska-Pempkowiak, H. 2010 Treatment of landfill leachate by constructed wetlands: three case studies. Polish Journal of Environmental Studies 19 (3), 643–650.

© IWA Publishing 2014 79

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.009

Heavy metals in groundwater from the Wolonghu coal mine, northern Anhui Province, China and their hydrological implications Linhua Sun*, Herong Gui and Weihua Peng School of Earth Science and Engineering, Suzhou University, Suzhou, Anhui 234000, China *Corresponding author. E-mail: sunlinh@126.com

Abstract Groundwater is important for either resource usage or safety of coal mining in north China. In this study, concentrations of eight kinds of heavy metals (Pb, Cd, Cu, Ni, Cr, Zn, Fe and Mn) in groundwater from three deep aquifers in the Wolonghu coal mine, northern Anhui Province, China have been analyzed for water quality assessment and water source identification. The results suggest that the groundwater have different concentrations of heavy metals among aquifers, which might be the results of different occurrence forms of heavy metals (e.g. Pb, Cd, Cu and Cr were adsorbed by iron hydroxides) and different kinds of water rock interactions (e.g. Mn and Zn originated from carbonate rocks). In comparison with the groundwater quality standard of China and WHO, most of them cannot be used for drinking directly but must be treated before drinking, especially the Pb, Cd and Fe contents. Moreover, hydraulic connection between aquifers has been identified by plots of factor scores and cluster analysis, which is similar to the results obtained by previous studies. Moreover, discriminant analysis demonstrated that heavy metals can be used for identification the source of inrush water in coal mines.

Key words: coal mine, groundwater, heavy metals, hydraulic connection, quality

INTRODUCTION To be one of the most important fresh water storage, the groundwater system plays important role in the current world. Their formation and evolution have long been affected by various physical, chemical and biological processes in the atmosphere, hydrosphere, biosphere and lithosphere, and thus participate in the global water cycle, which can lead to the recycling and exchange of materials among all of the spheres [1]. Moreover, they are also important strategic resources. Taking for instance, more than 50% of drinking water was supplied by groundwater in America, whereas more than 400 cities in China use groundwater for domestic use.

Excessive exploitation and utilization of groundwater resources has led to many environmental problems, especially the groundwater contamination. The survey of China Environmental Protection Administration indicated that more than 64% of the groundwater in the cities has been seriously polluted and 33% of them have been lightly polluted [2]. Therefore, groundwater contamination has long been concerned by governments, scientists and other humans [3–6]. The environmental problems were key issues in the 2004 International Geological Congress, 33rd and 34th International Hydrogeological Congress.

Heavy metals, especially the toxic metals (e.g. Hg, Cd, Pb, Zn et al.), have long been concerned by scientists. For example, their toxicological studies [7, 8], concentrations, distribution and forms in the water systems [9, 10], sources [11], migration, enrichment and transformation [12]. However, the similar work related to groundwater is limited and most of them are focused on the assessment of potential threaten [13–15], whereas other aspects have not been well determined.

80

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.009

Groundwater is important in northern Anhui Province, China. Firstly, it is the main water supplied for either domestic use or irrigation, but also a potential threaten for the safety of coal mining in the area (e.g. water inrush), especially in the deep underground system. Human in the mining area need groundwater for surviving, whereas the safety of coal mining needs to the discharging of water. Therefore, the rational use of mine water has become an important issue currently and some studies have been carried out [16, 17].

In this study, concentrations of eight kinds of heavy metals in the groundwater from the Wolonghu coal mine, northern Anhui Province have been analyzed, the goals of this study include: (1) environmental purpose: concentrations and quality and (2) engineering purpose: water source identification.

HYDROLOGICAL BACKGROUND The Wolonghu coal mine, which belongs to the Wanbei Coal-Electricity Group, is located 25 km southwest to the Suixi County, northern Anhui Province, China (Figure 1), the length is 8–9 km from south to north, and the width is 3–4 km from east to west, a total area is 28.9 km2. Previous investigations revealed that the groundwater system in the mine can be divided into three major aquifer systems from up to down: loose layer aquifer system (LA), coal bearing sandstone aquifer system (CA) and the underlying limestone aquifer system (TA). Moreover, each major aquifer system can be subdivided into some small secondary aquifers and their characteristics are as follows: LA: it can be subdivided into four aquifers, including the first, second, third and fourth aquifers from shallow to deep. The first and second aquifers are buried in shallow environment and they are used for water supply in the area (e.g. wells for drinking and irrigation). The depth of the third aquifer is between 138 and 179 m and the main rock types in the aquifer are fine and medium sandstones and the water storage in it is rich. The depth of the fourth aquifer is up to 234 m and the host rocks are mainly composed of clay, sandstone and conglomerate. CA: it can also be subdivided into four secondary aquifers: the 5, 6, 7 and 8th aquifers, they are composed of grey-white sandstones with a large number of fractures. The total thickness of CA is approximately 240 m and the water storage is medium in the 5 and 7th aquifers. TA: it is mainly composed of limestone and also rich with water because of its karstic composition.

Figure 1 | Location of the Wolonghu coal mine. Lines and circles in the map are major roads and cities in the area, respectively. (after ref [18])

81

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.009

In brief, the first threat for coal mining is the CA because it is facing the coal mining directly. However, LA and CA cannot be ignored as a potential source of inflow because the water storage in these two aquifers is also rich and they can recharge CA by faults or other channels. More information can be archived from ref [18].

SAMPLING AND ANALYSIS Twenty-five groundwater samples from LA (10), CA (10) and TA (5) have been collected from the alley in the Wolonghu coal mine between June and August, 2012. Water samples were filtered through 0.45 μm pore-size membrane and collected into a 2.0 L polyethylene bottles that had been cleaned in the laboratory, and immediately acidified to pH ,2 by HNO3 for prevention of element precipitation and/or adsorption by the bottle. Then the samples were sent to the laboratory for analysis in 24 hours.

Analytical processes were taken place in the Engineering and Technology Research Center of Coal Exploration in Anhui Province. Atomic absorption spectrometer have been applied for analysis of eight kinds of heavy metals (Pb, Cd, Cu, Ni, Cr, Zn, Fe and Mn), and the standard solutions of China after dilution have been applied for calibration, and the relative standard deviation is limited to less than 3%.

Three kinds of statistical analysis have been applied for the dataset, including factor analysis, cluster analysis and discriminant analysis. Factor analysis was used for identifying the correlation between heavy metals, and then for understanding their sources. Plots of factor scores were used for understanding the hydraulic connection between aquifers in combination with cluster analysis.

Moreover, discriminant analysis was used for water source identification. All of the analysis was processed by SPSS (version 16).

RESULTS AND DISCUSSIONS Heavy metal concentrations

All of the analytical results are listed in Table 1. As can be seen from the table, the concentrations of Pb, Cd, Cu, Ni, Cr and Zn are 6.32–31.8, 2.88–8.90, 2.75–28.0, 19.0–187, 0.067–6.40 and 30.3–66.5 ug/ l, respectively. Their mean concentrations are 16.1, 5.07, 10.7, 70.0, 1.88 and 51.2 ug/l, respectively.

Following the order of Ni . Zn . Pb . Cu . Cd . Cr. As to the Fe and Mn, they have much higher concentrations than other six kinds of heavy metals. Their concentrations are 524–614 and 2– 882 ug/l, respectively, and their mean concentrations are 562 and 191 ug/l, respectively.

It is also noticed that the heavy metal concentrations of groundwater samples from the LA, CA and TA are different (Figure 2): TA samples have the highest Pb, Cd, Cu, Cr and Zn concentrations than other two aquifers, whereas CA samples possess the highest Ni and Mn concentrations. Moreover, LA samples have the lowest concentrations for most of the heavy metals, such as Pb, Cd, Ni, Cr and Mn. These differences are considered to be related to different kinds of water rock interactions in different aquifers because of their different compositions of wall rocks [16, 17] and/or, different degrees of water rock interactions. And therefore, provide the probability that they can be used for water source identification.

Quality assessment

To better manage groundwater resources, the Chinese government have subdivided groundwater into five classes based on the concentrations of pollutants (Table 2): Class I–III can be used for drinking, irrigation and industry, and class IV can be used for irrigation and industry directly, but must be

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.009

82

Table 1 | Concentrations of heavy metals ID

Source

Pb (ug/l)

WLH-7

LA

8.71

WLH-8

LA

6.89

WLH-9

LA

WLH-10

LA

WLH-13

LA

WLH-25

LA

WLH-26

LA

WLH-27

LA

WLH-28

LA

WLH-29

LA

10.3

3.77

WLH-1

CA

11.7

4.34

3.74

WLH-2

CA

11.8

4.09

2.75

WLH-4

CA

21.3

6.57

11.8

WLH-5

CA

26.0

5.56

11.6

WLH-6

CA

11.4

WLH-12

CA

WLH-14

CA

WLH-15

CA

WLH-20 WLH-21

10.7

Cd (ug/l)

Cu (ug/l)

Ni (ug/l)

Cr (ug/l)

Zn (ug/l)

Fe (mg/l)

Mn (mg/l)

3.16

6.25

3.12

5.86

29.1

0.067

53.8

0.528

0.134

24.2

0.372

50.1

0.538

0.127

3.97

7.37

40.7

0.692

54.4

0.536

0.230

9.61

3.51

7.61

40.8

0.513

50.4

0.524

0.147

6.42

2.96

4.94

25.9

0.893

53.7

0.577

0.211

3.25

5.40

24.2

0.386

48.3

0.528

0.120

3.32

5.93

32.9

0.639

45.2

0.536

0.051

3.93

9.17

57.8

0.940

47.7

0.568

0.116

8.15

32.8

0.744

50.1

0.573

0.247

45.9

0.449

44.9

0.545

0.033

99.9

1.07

50.5

0.554

0.005

3.33

46.5

0.575

0.002

38.1

2.44

61.6

0.582

0.860

37.6

2.27

63.6

0.577

0.882

10.0 6.62 10.5 9.39

4.28

12.4

4.24

3.67

2.88

5.46

18.0

5.64

8.69

17.8

5.77

9.39

CA

14.8

5.23

5.75

CA

14.9

5.42

5.30

WLH-3

TA

29.7

7.72

WLH-11

TA

30.5

6.72

WLH-22

TA

31.3

WLH-23

TA

31.8

WLH-24

TA

29.9

6.32

109

187

1.16

42.9

0.546

0.014

0.475

57.8

0.574

0.013

164

0.697

32.4

0.56

0.024

166

0.162

30.3

0.557

0.027

92.0

2.74

42.0

0.604

0.029

87.1

0.294

42.8

0.575

0.028

18.7

61.9

4.76

66.5

0.577

0.216

16.2

69.5

6.40

62.9

0.575

0.089

7.97

28.0

67.9

3.99

61.4

0.554

0.188

8.72

27.4

64.0

4.95

62.5

0.567

0.239

8.90

26.1

67.1

3.78

62.9

0.614

0.241

19.0

Figure 2 | Concentration variations of heavy metals in groundwater from different aquifers.

treated before drinking, whereas class V and the worse cannot be used for drinking, and either irrigation or industrial use must be carefully selected.

Based on these standards, all of the samples can be subdivided to class III or better according to the concentrations of Pb, Cd and Cu. As to the Cr concentrations, twenty-four samples can pass the class III criterion and one sample can pass the class IV criterion. However, only twelve, ten and eleven samples can pass the class III criterion according to their Ni, Zn and Mn concentrations, respectively (Figure 3). Moreover, no sample has Fe concentration lower than 300 ug/l, but all of them have Fe

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.009

83

Table 2 | Quality standards for groundwater in China (GB/T 14848-9) Class I

Class II

Class III

Class IV

Class V

Pb (μg/l)

5

10

50

100

. 100

Cd (μg/l)

0.1

1

10

10

. 10

Cu (μg/l)

10

50

1,000

1,500

. 1,500

Ni (μg/l)

5

10

50

100

. 100

Cr (μg/l)

5

10

50

100

. 100

Zn (μg/l)

50

500

1,000

5,000

. 5,000

Fe (mg/l)

0.1

0.2

0.3

1.5

. 1.5

Mn (mg/l)

0.05

0.05

0.1

1.0

. 1.0

Figure 3 | Sample distribution according to the groundwater quality classification.

concentrations lower than 1,500 ug/l, and therefore all of them can only be classified to be class IV according to their Fe concentrations. Such result suggests that these groundwater samples can be used for irrigation and industry purpose. However, if they are used for drinking, their Ni, Zn, Fe and Mn concentrations must be treated before application.

Moreover, comparisons of heavy metal concentrations in the groundwater with the World Health Organization guidelines for drinking water quality (WHO, 2008) suggest that no sample can meet the requirement of Fe (300 ug/l) and only seven and two samples can meet the requirements of Pb (10 ug/l) and Cd (3 ug/l), respectively. However, only seven and two samples cannot meet the requirements of Ni (70 ug/l) and Mn (400 ug/l). Therefore, different with those of the Chinese standards, Pb and Cd should be firstly noticed before the application of drinking.

Statistical analysis

Statistical analyses, including factor, cluster and correlation analysis, are useful tools for understanding the environmental data set (e.g. hydro-chemical data). They can be used for identification of the source of heavy metals in soils, the major ions in groundwater and the hydraulic connection between aquifers et al. [20, 21]. In this study, factor analysis has been firstly processed by using the heavy metal concentrations, and the result is listed in Table 3.

Two factors with CA (EV) higher than one after vari-max rotation have been obtained (Table 3), the total variance explanation is 76.7%. The first factor (FA1), which accounts for 51.0% information (Var %), is dominated by Pb, Cd, Cu, Cr and Fe, whereas the second factor (FA2) accounts for 25.7% information, and is dominated by Zn and Mn. Moreover, Ni is participated in FA2 with high negative loading.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.009

84

Table 3 | Results of factor analysis (after vari-max rotation) Pb

Cd

Cu

Ni

Cr

Zn

Fe

Mn

EV

Var%

FA1

0.96

0.97

0.84

0.28

0.87

0.46

0.62

0.25

4.1

51.0

FA2

0.13

0.02

0.27

0.89

0.18

0.82

0.02

0.69

2.1

25.7

There is no direct method for determine the potential meanings of the factors because the lack of heavy metal concentrations in the wall rocks. However, as can be seen from the results, Pb, Cd, Cu, Cr and Fe are considered to have similar sources or, they are bounded together during variation, whereas Zn and Mn are considered to be originated from a similar source or under the condition of similar forms. Previous studies revealed that clay minerals and/or iron hydroxides are good carrier for heavy metals [22]. Therefore, the close correlations between Pb, Cd, Cu, Cr and Fe suggest that these four kinds of heavy metals are adsorbed by iron hydroxides. Moreover, it is also found by previous studies that Mn concentrations are increasing from clastic rocks (e.g. sandstone and shale) to carbonate rocks (e.g. limestone). Therefore, Mn should be higher in TA groundwater if its concentrations are mainly controlled by water rock interactions. This consideration is demonstrated by the Mn concentrations in the groundwater samples. Except for two samples (WLH-4 and 5) from CA, the samples from TA have the highest Mn concentrations (Table 1 and Figure 2). Another explanation is that Mn tends to be more soluble than Fe in the water [22]. It means that Mn and Zn in the groundwater samples are more likely to be dissolved form rather than adsorbed by clay minerals or iron hydroxides. Moreover, Ni is considered to be enriched in clastic rocks rather than carbonate rocks. In conclusion, FA1 and FA2 are interpreted to be iron hydroxides and carbonate factors, respectively.

In the plots of factor scores (Figure 4), all of the samples can be subdivided into four groups: TA group located in the right with high factor score 1, CA group in the lower with medium factor score 1 and lowest factor score 2, LA group in the left with lowest factor score 1. There are only two samples (WLH-4 and 5) with highest factor score 2 in the upper, which is because of their highest concentrations of Mn (Table 1). Moreover, it can also be found in the plot that one sample (WLH-12) from CA is plotted in the ďŹ eld of LA, this might be an indication of hydraulic connection between these two aquifers [18]. The conclusion can also be demonstrated by Q-mode cluster analysis that four CA samples are classiďŹ ed into a same group with LA samples (Figure 5).

Figure 4 | Plots of factor scores.

85

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.009

Figure 5 | Dendrogram of Q-mode cluster analysis (Sample 1–10, 11–20 and 21–25 are from LA, CA and TA, respectively).

Water source identification by using heavy metals

Water source identification is an important work after water inrush, and hydro-chemistry had played irreplaceable role. Previously, major ion concentrations, trace elements (including rare earth elements) and stable isotopes have long been used for water source identification [22–24]. However, heavy metals, as reported in this study, have not been applied in the issue.

As mentioned above, the groundwater samples from LA, CA and TA show different concentrations of heavy metals, which provided the possibility for water source identification by using them. Moreover, their factor score plots (Figure 4) and Q-mode cluster analysis (Figure 5) revealed that the groundwater samples from different aquifers are distinguishable. Therefore, we use the concentrations of heavy metals in this study, following the method reported by ref [24], to establish the models for water source identification. Because some of the samples (WLH-4, 5 and 12) are not ‘pure samples’ as defined by ref [24], before processing the discriminant analysis, they have been assigned as unknown source, and the rest of the sample have been chosen as ‘pure samples’ for model establishment.

The following equations and discriminant diagram (Figure 6) have been obtained, and they can be used for tracing the source of groundwater. Take samples (WLH-4, 5 and 12) as examples, their sources are classified to be LA, similar to the results obtained by Q-mode cluster analysis (Figure 5).

F1 ¼ 0.431 Pb þ 1.73 Cd-0.028 Cu þ 0.024 Ni þ 0.505 Cr þ 0.336 Zn þ 0.584 Fe-15.2 Mn-33.1; F2 ¼ 0.265 Pb-2.89 Cd þ 0.430 Cu-0.030 Ni þ 0.050 Cr-0.027 Zn-2.47 Fe þ 15.7 Mn þ 8.86.

The unit for Pb, Cd, Cu, Ni, Cr and Zn are ug/l, for Fe and Mn are mg/l.

CONCLUSIONS Based on the heavy metal concentrations in the groundwater from three different aquifers in the Wolonghu coal mine, northern Anhui Province, China, the following conclusions have been made: (1) The groundwater samples from different aquifers show variable heavy metal concentrations.

(2) Most of the groundwater samples cannot be used for drinking directly, but almost all of they can be used for drinking, irrigation and industry after treatment.

86

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.009

Figure 6 | Diagram of discriminant analysis.

(3) Adsorption by iron hydroxides is responsible for the concentrations of Pb, Cd, Cu, Cr and Fe in the groundwater, whereas dissolution from carbonate rocks is responsible for the concentrations of Mn and Zn. (4) Plots of factor scores and Q-mode cluster analysis, as well as discriminant analysis of heavy metals concentrations indicate that they can be used for water source identification.

ACKNOWLEDGEMENT This work was financially supported by National Natural Science Foundation of China (41302274 and 41173106), the Natural Science Foundation of Anhui Province (1308085QE77) and the Program for Innovative Research Team in Suzhou University (2013kytd01).

REFERENCES [1] Wang, Y. X. 2007 Groundwater Contamination. Higher Education Press, Beijing.

[2] Yin, G. X. & Li, Z. G. 2005 Groundwater Pollution and Prevention. China Environmental Science Press, Beijing.

[3] Al-Bassam, A. M. & Al-Rumikhani, Y. A. 2003 Integrated hydrochemical method of water quality assessment for irrigation in arid areas: application to the Jilh aquifer, Saudi Arabia. Journal of African Earth Sciences 36, 345–356.

[4] Jeevanandam, M., Kannan, R., Srinivasalu, S. & Rammohan, V. 2006 Hydrogeochemistry and groundwater quality assessment of lower part of the Ponnaiyar River basin, Cuddalore district, south Inida. Environmental Monitoring and Assessment 132 (1–3), 263–274.

[5] Ma, J., Ding, Z., Wei, G., Zhao, H. & Huang, T. 2009 Sources of water pollution and evolution of water quality in the Wuwei basin of Shiyang River, Northwest China. Journal of Environmental Management 90, 1168–1177.

[6] Pritchard, M., Mkandawire, T. & O’Neill, J. G. 2008 Assessment of groundwater quality in shallow wells within the southern district of Malawi. Physics and Chemistry of the Earth 33, 812–823.

[7] Forstner, U. 1982 Accumulative phases for heavy metals in limnic sediments. Hydrobiologia 91, 269–284.

[8] Naseem, R. & Tahir, S. S. 2001 Removal of Pb (II) from aqueous-acidic solutions by using bentonite as an adsorbent. Water Research 35, 3982–3986.

[9] Wu, X. R., Yin, P. H., Zhao, L., Li, S. T. & Yang, Y. F. 2010 Health risk assessment of heavy metals in the water of surface and subsurface microlayers from Guangzhou section of Pearl River. Journal of Jinan university (Natural Science) 31 (1), 84–88.

[10] Duan, X. C., Wang, W. J., Dang, Z., Zhou, J. M. & Feng, X. D. 2007 Distribution of heavy metals in water around the Dabaoshan mine. Earth and Environment 35 (3), 255–260.

[11] Zhang, Z. Y., Jilili, A., Feng, J. Q., Muysal, T. & Wang, S. P. 2012 Contents and sources of heavy metals in surface water in the Tianshan Mountain. Chinese Environmental Science 32 (10), 1799–1806.

87

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.009

[12] Cheng, J. P., Liu, C. E. & Wang, W. H. 2005 Transportation and accumulation of mercury and other heavy metals in waterbiofilm-sediment system of Huangpu River. Advances in Water Science 16 (6), 767–772.

[13] Bakis, R. & Tuncan, A. 2011 An investigation of heavy metal and migration through groundwater from the landfill area of Eskisehir in Turkey. Environmental Monitoring and Assessment 176 (1–4), 87–98.

[14] El Khalil, H., El Hamiani, O., Bitton, G., Ouazzani, N. & Boularbah, A. 2008 Heavy metal contamination from mining sites in South Morocco: monitoring metal content and toxicity of soil runoff and groundwater. Environmental Monitoring and Assessment 136 (1–3), 147–160.

[15] Lee, J. Y., Choi, J. C. & Lee, K. K. 2005 Variations in heavy metal contamination of stream water and groundwater affected by an abandoned lead–zinc mine in Korea. Environmental geochemistry and health 27 (3), 237–257.

[16] Sun, L. H., Gui, H. R. & Lin, M. L. 2013 Major ion chemistry of groundwater from limestone aquifer in Taoyuan coal mine, northern Anhui Province, China. Fresenius Environmental Bulletin 22 (2a), 537–543.

[17] Sun, L. H. & Gui, H. R. 2013 Groundwater from deep limestone aquifer in Linhuan coalfield, northern Anhui Province, China: quality and controlling factor analysis. International Journal of Applied Environmental Sciences 8 (2), 167–176.

[18] Sun, L. H. In press Statistical analysis of hydrochemistry of groundwater and its implications for water source identification: a case study. Arabian Journal of Geosciences (doi: 10.1007/s12517-013-1061-8).

[19] World Health Organization (WHO) 2008 Guidelines for Drinking Water Quality. 3rd edn, Incorporating the first and second addenda. Volume 1 Recommendations. Geneva. Available at: http://www.who.int/water_sanitation_health/dwq/ fulltext.pdf.

[20] Sun, L. H. 2013 Hydrochemical variation during groundwater mixing: a case study with multivariate statistical approach.

Water Practice and Technology 8 (3), 399–408.

[21] Rapantova, N., Krzeszowski, S., Grmela, A. & Wolkersdorfer, C. 2012 Quantitative assessment of mine water sources based on the general mixing equation and multivariate statistics. Mine Water and the Environment. 31 (4), 252–265.

[22] Gui, H. R. & Chen, L. W. 2007 Hydrogeochemistric Evolution and Discrimination of Groundwater in Mining District.

Geological Publishing House, Beijing.

[23] Veselič, M. & Mali, N. 1991 Reliability of hydrogeochemical methods as water inrush risk prediction tools. Proceedings, 4th International Mine Water Association Congress, 1, pp. 313–326.

[24] Sun, L. H. & Gui, H. R. 2012 Establishment of water source discrimination model in coal mine by using hydrogeochemistry and statistical analysis: a case study from Renlou Coal Mine in northern Anhui Province, China.

Journal of Coal Science and Engineering (China) 18 (4), 385–389.

© IWA Publishing 2014 88

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.010

Innovative sewer inspection as a basis for an optimised condition-based maintenance strategy H. Plihal*, F. Kretschmer, D. Schwarz and Th. Ertl Institute of Sanitary Engineering and Water Pollution Control, University of Natural Resources and Life Sciences, Vienna, Muthgasse 18, 1190 Vienna, Austria. E-mail: florian.kretschmer@boku.ac.at; dominik.schwarz@boku.ac.at; Thomas.ertl@boku.ac.at *Corresponding author. E-mail: hanns.plihal@boku.ac.at

Abstract Currently, around 100,000 km of public sewers are operated in Austria, with cleaning as one of the main tasks for a wastewater utility. Due to the precarious financial situation of many cities and municipalities sewer cleaning represents a considerable financial burden, resulting in the extension of cleaning intervals and the delay of required cleaning activities. Different approaches to sewer management can be distinguished. If a selective strategy is adopted, only pipe sections with deposits are cleaned. Thus, detailed information regarding the amount of deposits is required. A simple and quick method to inspect the sewer system and assess the degree of sediments is by means of a manhole-zoom camera. The current research project INNOKANIS investigates the operational condition assessment of sewers by means of different manhole-zoom cameras. So far the data suggest that the majority of investigated pipe sections belongs to the self-cleaning category. Only 1% of the pipe sections of combined sewers and 11% of sanitary sewers require additional cleaning. The example of the city of Salzburg illustrates the potential savings in connection with sewer cleaning if a selective approach is adopted. Following a strategic change, the sewer cleaning expenses decreased by 60% within a year.

Key words: cleaning performance, INNOKANIS, manhole-zoom camera, monitoring, selective strategy

INTRODUCTION At present around 100,000 km of public sewers are in operation in Austria, with cleaning as one of the main tasks for wastewater utilities. For many Austrian cities and municipalities sewer cleaning represents a considerable financial burden due to their precarious financial situation, which results in the extension of cleaning intervals and the delay of required cleaning measures. The Austrian technical guideline for sewer operation (‘OEWAV RB 22’ – draft version 2012) recommends a forward-looking sewer management and maintenance strategy to reduce costs and to minimize the environmental impact.

In a survey preceding the OEWAV-KAN convention (national board of operators of sewer systems and wastewater treatment plants) in 2010, the basic and operational data of the participating sewer operators were collected by means of a questionnaire. Among other aspects, the survey also covered the topic of sewer cleaning. Figure 1 illustrates the surveyed quantities of disposed waste (t/km) for the year 2009. It was found that the data ranged considerably from below 1 to 100 t/km, with the main part of disposed waste quantities between 1–10 t/km.

Figure 1 further suggests that participants with significantly more than 10 t/km of disposed waste probably cleaned some pipe sections with blockages. Participants with less than 1 t/km of disposed waste are supposed to have cleaned sewer sections with hardly any deposits at all (possibly within the framework of sewer mapping).

The cleaning strategy adopted by the participants of this survey can only be assumed, but plays an essential role if sewer management is to be optimized.

89

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.010

Figure 1 | Surveyed quantities of disposed waste for the year 2009.

MAINTENANCE STRATEGIES Inspection and maintenance represent essential aspects of sewer management. Different approaches to sewer management can be adopted: A reactive strategy (emergency strategy), a proactive strategy (interval-based maintenance strategy) and a selective strategy (condition-based maintenance strategy).

Following a reactive strategy, the sewer is not cleaned on a regular basis, but only as required, for example, in case of failure due to a blockage caused by increased deposits.

If a proactive strategy is adopted, sewer cleaning is conducted based on predefined intervals. To this end, the sewer system is often divided into areas resp. districts which are assigned to an annual sewer cleaning cycle. Once all sewer areas have been cleaned, the cycle starts again. Since the entire sewer system is covered, pipe sections with both high deposits and few or no deposits are cleaned.

The selective strategy, however, only considers sewer pipes with high amounts of deposits for cleaning. Thus, the cleaning occurs on a demand-oriented basis. However, such a selective approach to sewer management requires accurate knowledge of the sewer system. Prior to cleaning, the degree of sediments has to be determined. A simple and quick method to inspect the sewer system and to assess the amount of deposits in the pipe sections is by means of a manhole-zoom camera (‘electronic sewer mirror’). A selective approach to sewer management has the following advantages: sewer cleaning based on the amount of deposited material, less incidents due to blockages and finally an optimization of the management process. The disadvantage of this strategy, however, is the additional work caused by inspection in order to assess the degree of sediments.

SEWER INSPECTION BY MEANS OF A MANHOLE-ZOOM CAMERA (‘ELECTRONIC SEWER MIRROR’) As mentioned above, detailed information regarding the amount of deposits in each pipe section is required in order to adopt a selective approach to sewer management.

90

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.010

Among others, the current research project INNOKANIS conducted by the University of Natural Resources and Life Sciences, Vienna and several partners investigates the operational condition assessment of sewers applying different manhole-zoom cameras (see Figure 2). So far, 4 sewer catchment areas (342 sewer sections totalling 9.4 km) have been inspected by means of two different camera types. Both combined sewer systems (242 sections) and sanitary sewer systems (100 sections) were considered. Pipe dimensions (from DN 150 to egg-shaped 1300/900) as well as pipe materials (asbestos cement, clay, concrete, PVC) varied in the investigated sewer sections. With the aid of the manhole-zoom camera videos the deposited material in each section was estimated in relation to the pipe dimension and averaged over the pipe length. In a next step, the resulting quantities of deposited material were allocated to the following categories: cross-section reduction 2.5%, .2.5%– 15% and .15%. Based on a study by the Ruhr University Bochum (2008) a maximum threshold value of .15% has been specified for required sewer cleaning. In contrast to this study, the threshold value for only slightly deposited sewer sections has been set at 2.5% (instead of 10%). The reason for this modification was that in most pipe sections of the current investigation the deposited material did not exceed the threshold value of 2.5%.

Figure 2 | Sewer inspection by means of a manhole-zoom camera. Left: Manhole-zoom camera used in an egg-shaped profile.

Right: Assessing the degree of sediments (Plihalet al. 2013).

Figure 3 summarizes the quantities of deposited material found in these 4 catchment areas for combined and sanitary sewers respectively. As can be seen, for 63% of the investigated pipe sections in combined sewer systems and for 73% of the pipe sections in sanitary sewer systems the rate of deposited material in relation to the pipe dimension (i.e. the cross-section reduction) was 2.5%. For 36% of the inspected combined sewers and 16% of the sanitary sewers the cross-section reduction ranged between .2.5% and 15%. These findings suggest that no additional cleaning measures are required for the majority of the investigated pipe sections due to the intact self-cleaning effect. Only for 1% of the pipe sections in combined sewers and for 11% of the pipe sections in sanitary sewers a cross-section reduction of .15% was found, implying that the self-cleaning effect is not sufficient in these cases so that additional cleaning measures have to be taken.

MONITORING OF THE CLEANING PERFORMANCE In addition to determining the degree of sediments in the pipe sections a further important aspect of a selective approach to sewer management is the assessment of the cleaning performance. The resulting information is crucial to determine in how far the cleaning targets have been met and if some of the sewer sections have to be cleaned again. Figure 4 compares selected sewer sections prior to and after cleaning. (The differing image qualities are the result of using different types of manhole-zoom cameras.)

91

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.010

Figure 3 | Rates of deposited material in combined sewer systems (left) and sanitary sewer systems (right).

Figure 4 | Monitoring of the cleaning performance [modified 6].

Example 1 shows a sewer section managed by means of a reactive strategy. Prior to cleaning the manhole-zoom camera did not identify any deposited material in this sewer section, only spider webs throughout the entire pipe length. In contrast, following cleaning the pipe material (clay) could be correctly identified and the spider webs, which would have interfered with the conventional closed circuit TV (CCTV) system for structural condition assessment, had been removed.

Example 2 shows a sewer section with deposited material prior to sewer cleaning. After the sewer has been cleaned, no deposits remain. Thus, the cleaning performance was optimal and the cleaning target has been fulfilled.

Example 3 illustrates a case in which the sewer cleaning performance was not optimal and the cleaning target has not been fulfilled. There was more deposited material found in this sewer section after than prior to cleaning. Thus, this particular pipe section has to be recleaned.

EXAMPLE OF A SELECTIVE CLEANING STRATEGY: CITY OF SALZBURG Until 2010, the entire sewer system of the city of Salzburg (totalling 375 km) was cleaned by means of their own 3 sewer cleaning vehicles. Due to optimisation processes, the approach to sewer management

92

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.010

was changed from a proactive to a selective strategy in the year 2010. For this purpose a manhole-zoom camera was purchased in order to determine the amount of deposited material in the sewer system and to identify pipe sections requiring intensive cleaning. Based on a study conducted by the Ruhr University Bochum (2008) the degree of pollution in the pipe sections was classified according to ‘amount of deposited material’, ‘flow and stream rating of the wastewater’ and ‘consistency of deposited materials’.

The main finding of this optical inspection was that for 98% of the investigated sewer sections of the city of Salzburg the amount of deposited material ranged between 0–10%. Only for 1% of the sewer sections a rate of deposited material between either 10–15% or .15% was found (see Figure 5). Since an intact self-cleaning effect could be confirmed for the majority of sewer sections, today only those pipe sections have to be considered for additional cleaning for which self-cleaning is not effective (e.g. due to insufficient pipe slopes, structural or operational defects).

Figure 5 | Degree of pollution of pipe sections – city of Salzburg [modified 6].

Figure 6 shows an analysis of the quantities of deposited material removed by the city of Salzburg between 2006 and 2012. Following the change of the cleaning strategy in 2010, an increase in the quantity of removed deposits (in tons per cleaned km) is clearly discernible. In 2009, prior to changing the strategy, the deposited material removed from the sewer system ranged between 0.5–0.7 t/km. Following the strategic change, this rate increased up to 2.1 t/km in 2010, further rising to 1.3 t/km in 2011 and 1.5 t/km in 2012 respectively. According to these findings, the quantities of deposited material (t/km) removed from the sewer system have multiplied considerably since 2009.

Figure 6 | Quantities of disposed waste (t/km) – city of Salzburg.

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.010

93

However, contrary to expectations predicting much higher amounts of deposited material, the average quantity of removed deposits (t/km) only doubled. This unexpected result may be explained by the fact that in addition to the selective cleaning of pipe sections with a high degree of pollution also sections with little or no deposits were cleaned due to an upcoming CCTV inspection for sewer mapping. However, differentiating the reason for cleaning in terms of either a high degree of pollution or the upcoming CCTV inspection was impossible based on the available data.

The previous findings regarding the quantities of deposited material in particular sewer sections were used to calculate future cleaning intervals. In addition to the degree of pollution, the gradient of each pipe section, the sewer type (combined or sanitary system) and the pipe material were also considered. This analysis resulted in a preliminary cleaning interval for each particular sewer section, ranging from annual cleaning to a 5-year interval. However, cleaning based on this preliminary schedule is preceded by a sewer inspection using a manhole-zoom camera to ensure that the pipe sections in question are as highly polluted as expected. The results of these inspections are considered in the continuous calculation of future cleaning intervals in order to optimize the cleaning schedule.

An important outcome of changing the cleaning strategy from proactive to selective was that the city of Salzburg could eliminate 2 of their 3 sewer cleaning vehicles which were not required any more. Since the self-cleaning effect of most sewer sections was found to be sufficient to ensure that the wastewater is forwarded to the wastewater treatment plant, the sewer cleaning expenses of the city of Salzburg decreased by 60% within a year following the strategic change.

CONCLUSION A selective approach to sewer cleaning requires accurate knowledge of the behaviour of a sewer system regarding the site and amount of deposited material. This can be achieved by means of a manhole-zoom camera which represents a simple, quick and cost-efficient tool for sewer inspection.

Recent findings based on manhole-zoom camera investigations within the framework of the INNOKANIS project suggest the following:

• degree of pollution (cross-section reduction) for approximately 2/3 of all pipe sections in sanitary and combined sewer systems 2.5%,

• thus self-cleaning effect is sufficient for majority of sewer sections, • degree of pollution (cross-section reduction) for only 1% of combined sewers and 11% of sanitary sewers .15%.

These results are confirmed by the example of the city of Salzburg (mainly combined sewer system)

• degree of pollution (cross-section reduction) for 98% of all pipe sections in Salzburg ranging from 0– 10%, • thus no cleaning required for majority of sewer sections, • sewer cleaning expenses decreased by 60% within a year.

Concluding, it can be summarised that the majority of pipe sections is self-cleaning. Thus, a proactive strategy, with predefined cleaning intervals for all pipe sections regardless of their degree of pollution, represents an inefficient approach. In contrast, a selective cleaning strategy decreases expenses and optimises sewer management.

94

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.010

ACKNOWLEDGEMENT The current paper is based on investigations within the framework of the research project INNOKANIS, funded by the Federal Ministry of Agriculture, Forestry, Environment and Water Management.

The authors are grateful for this support.

REFERENCES Ertl, T. h. 2011 Ergebnisse der KAN-Umfrage ‘Stamm- und Betriebsdaten Kanalisation 2010’ (Results of the KAN survey ‘Basic and operational sewer data 2010)’, Presentation within the framework of the KAN Conference, 7–8 September 2011, Pregarten, Austria.

Innokanis – Innovative Inspection Methods For Optimisation Of Selective Sewer Operation Strategies, research project, Institute of Sanitary Engineering and Water Pollution Control, University of Natural Resources and Life Sciences, Vienna, Austria.

Oewav Regelblatt 22 (Entwurf 11.09.2012) ‘Betrieb, Wartung und Überprüfung von Kanalanlagen’ (Sewer Operation and Maintenance), Guideline 22, draft version 11.09.2012, Austrian Water and Waste Management Association, Vienna, Austria.

Plihal, H. 2012 Ablagerungen in Mischwasserkanälen (Deposits in combined sewer systems), Presentation, KAN Conference, 5–6 September 2012, Pregarten, Austria.

Plihal, H., Kretschmer, F. & Ertl, T. h. 2013 Innovative Kanalinspektion als Grundlage für die bedarfsorientierte Kanalreinigung (Innovative sewer inspection methods as the basis for needs-oriented sewer cleaning), Austrian Water and Waste Management OEWAW 3-4/2013; doi: 10.1007/s00506-013-0062-x; published by Springer.

Ruhr-University Bochum 2008 Zustands-, Prozess- und Wirkungsanalyse zur Entwicklung einer bedarfsorientierten Reinigungsstrategie für Kanalnetze (Analysing condition, process and effect to develop a condition-based sewer cleaning strategy), Germany.

© IWA Publishing 2014 95

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.011

Geochemical characteristics of deep groundwater from sandstone aquifer in Qianyingzi mine, northern Anhui province, China He Rong Gui School of Earth Sciences and Engineering, Suzhou University, Suzhou, Anhui 234000, China. E-mail: hrgui@163.com

Abstract Major ions, trace elements and isotope concentrations for eight groundwater samples were tested, which collected from sandstone aquifer in Qianyingzi mine, northern Anhui province, China. The Geochemical characteristic of groundwater samples were studied based on the conventional graphical and multivariate statistical approach, and the resulted showed: two types of groundwater could be identified through the Piper diagram, which have high concentrations total dissolved solids (1,164–5,165 mg/L), with alkaline environment (pH ¼ 8.02–8.90) in nature; the rare earth element of groundwater samples are characterized by enrichment of HREEs compared to LREEs when normalized to PAAS, which presented from the NdSN/YbSN ratios ranging from 0.042 to 0.121, with an average 0.075; groundwater characterized by negative Ce anomalies and positive Eu anomalies, what could be caused by the Ph conditions and exchange reaction between Eu2þ and Sr2þ, respectively; δ18O and δ2H of groundwater varied from 8.78 to 8.36‰ and 68.5 to 59.5‰, respectively.

The detritus and the exchange reaction between groundwater and alkyl could be the reason of obviously drift of δ2H. Key words: deep groundwater, geochemical characteristic, mining area, Northern Anhui province

INTRODUCTION Many studies, focusing on the hydro-geochemical process, groundwater quality and water rock interaction, have great significance for the exploitation of groundwater (Kumar et al. 2009; Ramkumar et al. 2013). Especially for these days, with the development of society and economy, deep groundwater plays a key role for the natively character without any anthropogenic influences. However, the studies which devoted to the hydro-geochemical process in deep groundwater (depth . 400 m) are limited, for the sample collection is difficult.

Deep coal mining provides opportunities for the sample collection of deep groundwater. Deep groundwater in mining area has two typical characters for the specific district. First, it is the nicer and abundant water resource; in addition, it is the hidden danger for coal mine safety. Thus, deep groundwater investigations, such as major ions, trace elements, as well as isotopic studies, have been carried out in mining area (Chen et al. 2011; Gui et al. 2011; Sun et al. 2011). Especially with the evolution of analytical techniques, the rare earth elements (REEs) concentrations in groundwater have been studied for many scientists. REEs have the unique characteristics in diverse geological processes, what could be as the indicator in the water source discrimination (Chen et al.

2011; Sun et al. 2011).

Qianyingzi mine is located in northern Anhui Province, China, where has abundant coal resources.

For the coal mine safe exploitation, water source discrimination is of great interest for the scientists and producer. Thus, so many studies what focused on the hydro-geochemical character, and then as basis to discuss the discrimination of groundwater have been carried out (Sun & Gui 2012). However,

96

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.011

these studies frequently to discuss only one or two type chemical elements, such as major elements, trace elements or isotope, the studies about the hydro-geochemical process combining the major ions and trace element with isotopic composition are limited. Especially in Qianyingzi mining, the information about hydro-geochemical has not been printed. The purpose of the study is to discuss the geochemistry character of groundwater from sandstone aquifer, using the major, trace element and isotopic data. The major targets are to (1) define the geochemical evolution of sandstone aquifer system in coal bearing; (2) understand the REEs characteristic of groundwater and its source; (3) identify the origin of isotope in groundwater samples. These results could be used as a basis for making sustainable groundwater development schemes and tracing the origin of deep groundwater.

MATERIALS AND METHODS The Qianyingzi Coal Mine is located in the southern part of the Huaibei coalfield, which constituted by 23 active underground coal mines. Huaibei coalfield is one of the major coalfields in China, being located in the northern Anhui province, China. The basement of coal mine in the district is composed

by Archean and Proterozoic metamorphic rock, with cover strata are stable sedimentation between late-Proterozoic and Permian (Gui et al. 2011). Groundwater system in the district from shallow to deep could be subdivided into four aquifers: the Quaternary aquifer, the Permian aquifer (Coal bearing aquifer), the Carboniferous aquifer (Taiyuan Formation limestone aquifer), and the Ordovician aquifer.

The groundwater samples were collected from sandstone aquifer in coal bearing of Qianyingzi mine, northern Anhui province, China (Figure 1). A total eight samples were collected from the Qianyingzi mining area. Water samples were collected via drainage holes in alleys, and then filtered through 0.45 μm pore-size membrane and collected into polyethylene bottles that had been cleaned using trace element clean procedures. All the eight samples were analyzed for major ions, trace elements and isotope.

Figure 1 | Location of study area in northern Anhui Province, China.

Major ions were analyzed in the analysis testing center of department of coal geology of Anhui pro vince, China. The Kþ and Naþ were analyzed by atomic absorption spectrometry, SO2 4 and Cl by ion chromatography, Ca2þ and Mg2þ by EDTA titration and alkaline by acid-based titration. The trace element and isotopic compositions were analyzed in the laboratory of Institute of Karst Geology, Chinese Academy of Geological Sciences. Trace element concentrations were determined after

97

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.011

pre-concentration by liquid-liquid extraction and analyzed by inductively coupled plasma mass spectrometry (ICP-MS, POEMS III) in the State Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences (Wuhan). The analytical precision for all trace elements was 10% relative standard deviation or better. The isotopic compositions were analyzed in the laboratory of Institute of Karst Geology, Chinese Academy of Geological Sciences. The isotopic data are reported with respect to standard mean ocean water, and δ18O and δD have an overall precision of 0.2 and 2‰, respectively.

The piper diagram and calculates about carbonate equilibrium, total dissolved solids (TDS), density, conductivity and hardness were accomplished by software Aqqa, the statistical of samples data were completed by Excel (version 2007) and the SPSS (version 17).

RESULTS AND DISCUSSION Major ions

The chemical data of groundwater samples are listed in Table 1, which used to discuss the geochemical characteristics of groundwater. The Naþ þ Kþ are dominant in the cation, whereas the SO2 4 and

HCO3 are mainly anion in the samples, with the low degree content of Cl (Table 1, Figure 2). In general, the pH values of groundwater varied from 8.02 to 8.90, with an average value 8.42, which indicates water is alkaline in nature. The amount of TDS of groundwater ranges from 1,164 to 5,165 mg/L, with an average of 2,428.8 mg/L. The geochemical data of groundwater were plotted on a Piper diagram (Figure 2), it can be seen almost all the ground water could be described the Na · K-SO4 · HCO3 type. Eight groundwater samples could be subdivided into two types in detail, the Na · K-HCO3 and Na · K-SO4 types.

If the dissolution of calcite, dolomite and gypsum are dominant reaction in hydro-geochemical pro cess, the ratio between Ca2þ þ Mg2þ and SO2 4 þ HCO3 will be close to the 1:1. Ion exchange tends to 2 shift the points to right due to an excess of SO4 þ HCO 3 (Fisher & Mulican 1997). The plot of 2þ 2þ 2

Ca þ Mg versus SO4 þ HCO3 (Figure 3(a)) shows that almost all the groundwater sample below the 1:1 line which indicate ion exchange is obviously. In addition, all the sample points are placed below the 1:1 line, what combined with the high concentrations of Naþ þ Kþ, indicating the silicate weathering in the hydro-geochemical process.

The plot of Ca2þ versus Mg2þ of the groundwater suggests the dominance of the dissolution of calcite and dolomite that present in the sandstone aquifer with some Coal bearing strata (Figure 3(b)). If the points near to the 1:1 line, dissolution of dolomite should occur, whereas a high ratio is indicative of great calcite contribution. Most of the samples near to the line 1:1, which two samples above the line 1:1, what should be indicated that the dissolution of dolomite is exited, whereas the calcite weathering is dominant.

Rare earth element

The REE concentrations of groundwater samples are listed in Table 1 and the PAAS (Post Archean Average Shale) normalized REE patterns are presented in Figure 4 (Taylor & McLennan 1985).

The total REE concentrations (∑REE) are varied from 0.047 to 0.086 mg/L, with an average 0.066 mg/L. The normalized NdSN/YbSN ratios (SN means PAAS normalization) varied from 0.042 to 0.121, with an average 0.075, what combined with the feature presented from Figure 4 revealed that all of the groundwater samples are characterized by enrichment of HREEs compared to LREEs when normalized to PAAS. In addition, the groundwater samples have weak negative PAAS normalized Ce anomalies, with the Ce/Ce* values rang from 0.71 to 0.002, with an average

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.011

98

Table 1 | Major (mg/L), trace element (μg/L) and isotopic composition and other relations parameter of groundwater samples from sandstone aquifer in Qianyingzi mine, northern Anhui province, China Parameter

QYZ1

QYZ2

QYZ3

QYZ4

QYZ5

QYZ6

QYZ7

QYZ8

Kþ þ Naþ

652.33

582.45

1178.38

1718.35

543.67

471.82

524.29

949.88

Ca

4.75

3.17

24.54

26.13

6.33

5.54

16.63

102.13

Mg2þ

3.84

3.36

4.8

11.53

3.84

3.84

19.2

87.37

Cl

181.67

282.21

199.31

149.93

188.73

146.4

407.44

287.5

SO2 4

375.79

4.53

1687.97

2837.57

125.13

21.4

8.64

1899.53

HCO 3

738.59

817.06

736.28

817.06

290.82

CO2 3

115.78

129.4

Hardness

27.68

21.75

Alkalinity

798.75

885.81

PH

8.54

TDS

1714

La Ce

2þ

687.81

853.99

860.91

70.38

70.38

72.65

81.05

112.68

31.63

29.65

120.59

614.81

603.79

681.4

817.67

827.14

670.04

296.27

8.9

8.05

8.55

8.31

8.65

8.02

8.36

1425

3675

5165

1377

1164

1394

3516

0.0031

0.0041

0.0016

0.0029

0.0026

0.0033

0.0067

0.0032

0.0086

0.012

0.0032

0.004

0.0022

0.006

0.0036

0.0047

Pr

0.0013

0.0017

0.0008

0.0009

0.0019

0.0009

0.0017

0.0008

Nd

0.0032

0.0079

0.0061

0.01

0.0072

0.0079

0.0053

0.0059

Sm

0.0054

0.0072

0.012

0.01

0.008

0.013

0.0048

0.0071

Eu

0.0025

0.0037

0.0034

0.0069

0.0044

0.0082

0.03

0.0027

34.05

Gd

0.0028

0.0063

0.0035

0.0057

0.0067

0.0044

0.0049

0.0049

Tb

0.0009

0.0009

0.0013

0.0022

0.0038

0.0014

0.001

0.0018

Dy

0.0034

0.005

0.0046

0.0069

0.0012

0.0045

0.0076

0.002

Ho

0.0016

0.0014

0.0046

0.01

0.0011

0.002

0.0008

0.0025

Er

0.0095

0.0079

0.0062

0.0099

0.0079

0.0079

0.0042

0.0045

Tm

0.001

0.0012

0.0013

0.002

0.0017

0.0017

0.0006

0.0009

Yb

0.0022

0.0091

0.0059

0.012

0.0143

0.01

0.0052

0.0086

Lu

0.0015

0.0011

0.0013

0.0029

0.0028

0.0014

0.0018

0.0009

Zr

0.26

0.13

0.1

0.24

0.2

0.14

0.18

0.15

∑REE

0.047

0.070

0.056

0.086

0.066

0.073

0.078

0.051

NdSN/YbSN

0.121

0.072

0.086

0.069

0.042

0.066

0.085

0.057

Ce/Ce*

0.024

0.002

0.217

0.248

0.710

0.097

0.609

0.169

Eu/Eu*

0.47

0.41

0.33

0.63

0.45

0.66

1.46

0.33

δD

58.5

64.8

68.5

64.8

64

67.8

64.2

66.9

18

8.58

8.59

8.59

8.58

8.36

8.75

8.7

8.78

δ O

0.26, what calculated by Ce/Ce* ¼ log (2*CeSN/(LaSN þ PrSN)). Whereas the Eu presented positive anomalies, with the Eu/Eu* values varied from 0.33 to 1.46, with an average 0.59, what calculated by Eu/Eu* ¼ log (2*EuSN/(SmSN þ GdSN)).

The negative Ce anomalies may reflect oxidative conditions of the aquifer, for Ce3þ is oxidized to the Ce4þ, reducing the concentrations of soluble Ce (Leybourne et al. 2000). However, there are no obviously proof supporting this view, for the groundwater collected from sandstone aquifer in coal bearing, where are regard as deoxidize environment for abundant coal resource. Another possible explanation is that the Ce anomalies of groundwater reflect differences in solubility of Ce redox species related to pH conditions in the aquifer system. Previous studies showed that degree of negative Ce anomalies varied with increasing pH, for Ce3þ is more stable in low pH conditions (Johannesson et al. 2005). The plot of Ce anomalies between pH showed that the degree of Ce anomalies varied

99

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.011

Figure 2 | Piper diagram of groundwater from sandstone aquifer in Qianyingzi mining, northern Anhui province, China.

2þ Figure 3 | The scatter diagrams of Ca2þ þ Mg2þ versus SO24 þ HCO versus Mg2þ of groundwater from sandstone 3 , Ca aquifer in Qianyingzi mine, northern Anhui province, China.

100

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.011

Figure 4 | PAAS normalized REE patterns of groundwater from sandstone aquifer in Qianyingzi mine, northern Anhui province, China.

along with the pH values (Figure 5), indicating that such a mechanism could be main reason for Ce depletion. The positive Eu anomalies in groundwater were related to the redox conditions (Jeong 2001). The positive Eu anomalies in deep groundwater possibly due to the exchange reaction between Eu2þ and Sr2þ in reducing conditions, for the chemical properties of Eu2þ and Sr2þ are similar (Yan et al. 2012). Therefore, if the concentration of Sr2þ is high in groundwater, the exchange reactions between Eu2þ and Sr2þ might occur. Consequently, Sr2þ is readily precipitated, whereas the concentration of Eu2þ is still high in groundwater. This mechanism possibly contributes to the positive Eu anomalies in groundwater in this study.

Figure 5 | Scatter diagram between pH versus Ce/Ce* of groundwater from sandstone aquifer in Qianyingzi mine, northern Anhui province, China.

Zr and Sr are always considered to be representative of terrigenous and marine material in marineterrigenous facies (e.g. the sandstone in the coal bearing strata), although Sr can also represent plagioclase in terrigenous detritus. Previous study showed that the REEs concentrations in sandstone are

101

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.011

dominated by the Zr (Gui et al. 2011). The REEs concentrations and fractionation in groundwater from sandstone aquifer are also expected be controlled by Zr, and the positive correlation between REEs and Zr are expected exiting too. This view was supported from the Figure 6, what presented positive correlation between Zr and REEs with the correlation coefficient 0.83. And the low positive coefficient (0.31) between Sr and REEs revealed that the contribution of REEs from marine material is limited.

Figure 6 | Scatter diagram between REE versus Sr, Zr of groundwater from sandstone aquifer in Qianyingzi mine, northern Anhui province, China.

Isotope character

The results of the δD and δ18O analysis for groundwater from sandstone aquifer are plotted in Figure 7 (a). The stable isotope values for groundwater samples were found be varied between 8.78 to 8.36‰ in δ18O with an average of 8.62‰ and from 68.5 to 59.5‰ in δ2H with an average of 64.94‰, respectively. In order to identify more information about the isotopic feature of groundwater samples, the other data are required. Thus, global meteoric water line (GMWL), local meteoric water line (LMWL) and local surface water line (LSWL) about the δD and δ18O were gathered. The GMWL was described by δD ¼ 8*δ18O þ 10.56, which defined by Craig (1961); the LMWL was characterized as δD ¼ 7.9*δ18O þ 8.2, which summarized from the measured data of the stable

Figure 7 | Diagram between δ18O and δD of groundwater from sandstone aquifer in Qianyingzi mine, northern Anhui province, China.

102

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.011

isotopes (Zhang 1989) and the LSWL is printed as the formula δD ¼ 6.74*δ18O 3.33(Gui et al.

2005). All the lines and the isotopic message of the groundwater samples are plotted in Figure 7.

The LSWL are below the LMWL and GMWL, which are presented from Figure 7, indicating that the surface water was shaped through the evaporation (Barth 2000).

Seven groundwater samples are plotted below the LSWL, LMWL and GMWL, except one groundwater sample be on the GMWL, what combined with the feature one sample among the LMWL and LSWL, indicating that the groundwater from sandstone aquifer be supplied from the meteoric water, with varied degree evaporation or not. In addition, the δ18O and δ2H are slightly variety, and the relationship between δ18O and δ2H in the groundwater form sandstone aquifer could be defined as δD ¼ 7.54*δ18O þ 0.0005. Further more, the values of δ2H are variety obviously, whereas the δ18O are stabilization reversely. The plots upward movement with a direction vertical or approximate to vertical, and all the plots are not exceed the LMWL, what revealed that the δ2H have been drifted obviously.

Generally speaking, the variety of δ18O and δ2H could be caused by such factors as evaporation, reservoir temperature, residence time and water-rock interaction (Truesdell & Hulston 1980). The sight presented in Figure 7(a) could be interpreted by the sandstone aquifer being reducing environment, where exchange reaction of δ2H between groundwater and alkyl is more markedly, whereas the exchanging of δ18O has been equilibrium. And the deduction could be supported by the previous study (Gui et al. 2005). However, the positive correlation between δ2H and Zr (Figure 7(b)), with the correlation coefficient 0.74, what implying that the variety of δ2H are relate to the concentrations of Zr. In another word, the character of δ2H could be caused by the detritus which have high degree concentration of Zr. However, the concretely process and influencing mechanism need the further work.

CONCLUSIONS The concentration of major ions, trace elements and isotopic in groundwater samples collected from sandstone aquifer in Qianyingzi mine, northern Anhui province, China had been tested, the geochemistry characteristic of groundwater were analyzed, a series of conclusion could be obtained: Groundwater from the sandstone aquifer could be subdivided into Na · K-HCO3 and Na · K-SO4 types, what have high concentrations TDS (1,164–5,165 mg/L), with alkaline environment (Ph ¼ 8.02–8.90) in nature. The silicate weathering is dominated in the hydro-geochemical process, and the dissolution of calcite and dolomite were also exited, what be obtained from the relation between 2þ

and Mg2þ, and the high concentration of Naþ þ Kþ.

Ca2þ þ Mg2þ and SO2 4 þ HCO3 , Ca ∑REE concentrations are varied from 0.047 to 0.086 mg/L, with an average 0.066 mg/L, what could be contributed from the detritus with high concentrations of Zr. The normalized NdSN/ YbSN ratios varied from 0.042 to 0.121, with an average 0.075, revealing all of the groundwater samples are characterized by enrichment of HREEs compared to LREEs when normalized to PAAS. Slight negative Ce anomalies and marked positive Eu anomalies are exited in all samples, what could be caused by the Ph conditions and exchange reaction between Eu2þ and Sr2þ, respectively.

The value of δ18O are varied between 8.78 to 8.36‰, with an average of 8.62‰, whereas the 2 δ H ranging from 68.5 to 59.5‰, with an average of 64.94‰. In addition, the values of δ2H are variety obviously, otherwise, the δ18O are stabilization reversely. Groundwater from sandstone aquifer could be supplied from the meteoric water, with varied degree evaporation or not, what deduced from the feature of the relation between samples plots and LSWL, LMWL and GMWL. The drift of δ2H could be caused by the detritus and the exchange reaction between groundwater and alkyl.

103

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.011

ACKNOWLEDGEMENT Foundation item: Project supported by the National Nature Science Foundation of China (41173106, 41373095), the Anhui province Department of education Natural Science Foundation (KJ2013A249, KJ2013B289) and the Program for Innovative Research Team in Suzhou University (2013kjtd01).

REFERENCES Barth, S. R. 2000 Stable isotope geochemistry of sediment hosted groundwater from a Late Paleozoic-Early Mesozoic section in central Europe. Journal of Hydrology 235, 8–80.

Chen, L. W., Gui, H. R. & Yin, X. X. 2011 Monitoring of flow field based on stable isotope geochemical characteristics in deep groundwater. Environ Monit Assess 179, 487–498.

Chen, S., Gui, H. R., Sun, L. H. & Liu, X. H. 2011 Rare earth element fractionation between groundwater and wall rock in limestone aquifer: sample from Taiyuan formation limestone aquifer in Renlou coal mine, northern Anhui province.

Geoscience 25 (4), 802–807. (in Chinese) Craig, H. 1961 Isotopic variation in meteoric water. Science 133, 1702–1703.

Fisher, R. S. & Mulican, W. F. 1997 Hydrochemical evolution of sodium-sulfate and sodium-chloride groundwater beneath the Northern Chihuahuan desert, Trans-Pecos, Rexas, USA. Hydrogeol J 10 (4), 455–474.

Gui, H. R., Sun, L. H., Chen, L. W. & Chen, S. 2011 Rare earth element geochemistry of ground water from a deep seated sandstone aquifer, northern Anhui province, China. Mining Science and Technology 21, 477–482.

Gui, H. R., Chen, L. W. & Song, X. M. 2005 Drift features of oxygen and hydrogen stable isotopes in deep groundwater in mining area of northern Anhui. Journal of Harbin Institute of Technology 37, 111–114. (in Chinese) Jeong, C. H. 2001 Mineral-water interaction and hydro-geochemistry in the Samkwang mine area, Korea. Geochemical Journal 35, 1–12.

Johannesson, K. H., Cortes, A., Leal, J. A. R., Ramirez, A. G. & Durazo, J. 2005 Rare Earth Elements in Groundwater Flow Systems. Springer, Netherlands, pp. 188–222.

Kumar, M., Kumari, K., Singh, U. K. & Ramanathan, A. L. 2009 Hydrogeochemical processes in the groundwater environment of Muktsar, Punjab: conventional graphical and multivariate statistical approach. Environ Geol 57, 873–884.

Leybourne, M. I., Goodfellow, W. D., Boyle, D. R. & Hall, G. M. 2000 Rapid development of negative Ce anomalies in surface waters and contrasting REE patterns in groundwaters associated with Zn-Pb massive sulphide deposits. Appl Geochem 15, 695–723.

Ramkumar, T., Venkatramanan, S., Anithamary, I. & Ibrahim, S. M. S. 2013 Evaluation of hydrogeochemical parameters and quality assessment of the groundwater in Kottur blocks, Tiruvarur district, Tamilnadu, India. Arab J Geosci 6, 101–108.

Sun, L. H., Gui, H. R. & Chen, S. 2011 Rare earth element geochemistry of groundwater from coal bearing aquifer in Renlou coal mine, northern Anhui Province, China. Journal of Rare Earths 29, 185–192.

Sun, L. H. & Gui, H. R. 2012 Establishment of water source discrimination model in coal mine by using hydrogeochemistry and statistical analysis: a case study from Renlou Coal Mine in northern Anhui Province, China. Journal of Coal Science and Engineering (China) 18 (4), 385–389.

Taylor, S. R. & McLennan, S. M. 1985 The continental crust: its composition and evolution. Blackwell Scientific Publications, Oxford.

Truesdell, A. H. & Hulston, J. R. 1980 Isotopic evidence of environments of geothermal systems. In: Handbook of environmental Isotope geochemistry, 1: The terrestrial Environment (Fritz, P. & Fontes, JCh., eds). Elsevier, Amsterdam, pp. 179–126.

Yan, Z. C., Liu, G. J., Sun, R. Y., Tang, Q., Wu, D., Wu, B. & Zhou, C. C. 2012 Geochemistry of rare earth element in groundwater from the Taiyuan Formation limestone aquifer in the Wolonghu Coal Mine, Anhui province, China. Journal of Geochemical Exploration 135, 54–62.

Zhang, H. P. 1989 Study on the background value of stable isotopes in precipitations of China. Site Investigation Science and Technology 1, 6–13. (in Chinese)

© IWA Publishing 2014 104

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

Transforming ‘value engineering’ from an art form into a science – process resilience modelling J. Curriea, N. Wraggb, C. Robertsa, J. Tattersallc and G. Leslied a

Black & Veatch Australia Pty Ltd, 492 St Kilda Road (6th Floor) Melbourne, VIC 3004, Australia. E-mail: curriejd@bv.com

b

DNV GL Ltd

c

Black & Veatch Ltd, 69 London Road, Redhill, UK

d

University of New South Wales, Sydney, Australia

Abstract The resilience of a treatment facility should be an important part of its design and operation throughout its service life to ensure it meets compliance and production expectations. This has traditionally been difficult to assess and quantify, and as a consequence its management has largely been ignored, or has been reduced to a function of how many treatment stages are provided with redundancy and/ or backup ‘stand-by’ facilities. Without proper resilience assessment there will always be a tendency to undertake ‘gold-plate’ engineering producing specifications much higher than the business need. This consequently leads to higher capital and operational expenditure over the life of a treatment asset. Value engineering then ends up an art form, where negotiating the line between risk and cost is often more to do with good luck than judgement. Resilience assessment makes value engineering a science rather than an art, as well as providing a critical means of influencing and assessing investment decisions and operational and maintenance planning to minimise the overall cost of compliance.

Asset resilience assessment techniques have been developed in other industries over the last 15 years. Recently the authors have applied these tried and tested approaches to water and wastewater treatment assets.

Key words: asset management, reliability, resilience, risk based planning, value engineering

INTRODUCTION For as long as engineers have been designing major water and wastewater treatment facilities they have had to decide not only on the treatment techniques to be used but also the degree of ‘robustness’ or redundancy to be provided in their designs. In making these decisions they have had to balance the competing requirements of providing additional or spare equipment to allow the facility to continue to perform in the event of equipment failure, and providing a facility at a cost the end-user can afford.

In recent years, as the cost and complexity of facilities has increased, the techniques of value engineering have been more commonly applied in the water industry. These techniques have provided a structured approach to determine the balance between robustness and cost, however this still relies on the judgment of engineers and operators to determine what will be acceptable for long-term operation. While there have been many successes using this approach there have also been a number of occasions where value engineering has resulted in plants with poor operability or insufficient capacity to maintain the required quality and quantity criteria. In these cases additional equipment has had to be added. As this has been done as a retrofit it has cost significantly more than if it had been done as part of the original project. One good example of this is a wastewater treatment plant that removed two final clarifiers during value engineering then had to build them a few years later when the plant was unable to meet solids consent during peak flow events.

105

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

Various forms of Reliability Assessment, or Resilience Assessment have been long used in other industrial sectors such as oil and gas, and Nuclear (OREDA 2009). This paper describes the application of these tools in the Water industry in concert with engineering judgement to provide a more consistent and defined output from ‘value engineering’ in the design and construction phase, as well as providing a tool for sustainable operation and maintenance of the asset to minimise both whole of life cost and environmental footprint. Resilience assessment tools are also of great benefit in this context as they can be used in a business-as-usual fashion to investigate poor performance, identify critical equipment, quantify risk to public health, optimise capital investment and guide maintenance regimes. The proposed approach goes beyond previous reliability studies of water and wastewater systems that have considered failure rates or critical instrumentation (Corominas et al. 2011) or simulated plant performance under a range of conditions by applying statistical methods to vary output of steady state process models [3 & 4]. In particular, this approach uses actual failure rates for all components across a complete asset register for both conventional and advance water treatment process, including membrane systems. The work is timely because of the growth in the use of membrane systems, particularly in wastewater recycling (Seah et al. 2003) and that again, previous work on membrane systems has been narrowly focused on the failure of single components (Childress et al. 2005).

RELIABILITY ANALYSIS AND RESILIENCE Reliability analysis is used to predict the likely performance of a system (OREDA 2009) Reliability analysis involves building a model to represent the impact of equipment failures on the performance metric of interest, typically plant availability (i.e. the proportion of time the system is operational and capable of performing as intended).

A reliability block diagram is typically used to represent the impact of failures with blocks arranged in series and in parallel. If a simple two component system is considered, a series arrangement indicates that the failure of either results in system unavailability whereas a parallel arrangement indicates that both items have to fail for the system to become unavailable. By using multiple nested series and parallel groups the reliability of complex systems can be represented.

Figure 1 | Components in series and parallel.

When considering the performance of production systems (typically in the oil and gas industry) there are often complexities that require a different approach to traditional reliability techniques.

Such complexities are non-simple failure and repair processes, partial levels of operation, deferred impact of failures, system configuration changes and capacity variations. Monte Carlo techniques

106

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

can be applied to capture the impact of these random events. The approach consists of developing a model of the system (normally in the form of a reliability block type representation) and subjecting it to events (typically failures and repairs) that can occur during the lifetime of the system. Such events are generated stochastically, through the use of random numbers. The nature of the model means that it is possible to include a wide variety of complex component and system behaviours. By using this technique a simulation of a system’s lifetime can be undertaken by stepping through events as they occur. However, an individual simulation is not necessarily a reasonable indication of average performance as it might have been subjected to events that were more, or less, favourable than the average. To obtain an indication of average performance and the likely range of performance, it is necessary to undertake multiple simulations (as many as 10,000 simulations are commonly used).

The Optagon software from DNV GL has been developed over a number of years to utilise this approach; moreover the software extends this concept by associating a capacity, or flow, with each component. Key inputs to such modelling are the impact of equipment failures on system performance and reliability data (in terms of mean time between failures and mean down-time) for the failures and repairs. Logistic delays to undertake repairs can also be included as part of the downtime. An example of this use in the gas industry was reported by Rogers (Rogers 2000).

By using a Monte Carlo package a range of performance statistics can be obtained about the reliability and availability of a system. These include

• shortfall (the proportion of capacity that is not supplied); • unavailability (the proportion of time when output is below the required system capacity level); and • component criticality (contribution to loss of output).

In applying this approach in the water industry a further step has been taken beyond reliability analysis for production systems and this is termed ‘Resilience’. This additional step involves taking into account compliance. The resilience of a wastewater treatment facility, as an example, is a fundamental factor in maintaining continuous compliance with its environmental discharge consent.

Resilience is defined as the ability of a system to perform and maintain its function in routine, as well as unexpected circumstances. The overall resilience of a given facility combines the performance of the treatment process with the availability of the associated critical equipment. Process performance relates to the dynamics of the facility – tools such as BioWin can be used to define how many equipment items need to fail before an impact on quality might be realised as well as how long until a breach in consent is expected (this is also referred to as ‘deferred effect’). The availability of equipment items is dependent on the number of critical failures that occur and how long it takes to address those failures.

USE IN OTHER INDUSTRIES Reliability analysis in the form of reliability, availability and maintainability (RAM) studies has been used in the Aeronautical and Defence industries for many years. More recently, it has also been used in the oil and gas industries.

Different industries have different drivers. The aeronautical and defence industries are typically driven by availability of equipment (eg maximising the amount of time a military or commercial aircraft is in a serviceable condition and available for operations). In the oil and gas industry, key reasons for carrying out RAM studies are to maximise revenues through increased production, benchmark performance and quantify ‘lost’ production potential, reduce CAPEX and OPEX expenditure, optimise design and operation, target investment and maintenance activity, and reduce contractual penalties by optimising commercial strategies. RAM studies are applicable through

107

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

the full lifecycle of a facility, from concept to front end engineering design, Detailed Design through to Maintenance and Operation, and can provide information for key decisions at all of these different stages.

Such use of RAM studies in these industries is very much accepted practice as it is seen to add significant value during all stages of the lifecycle of a system or facility. Recent examples include

• $435M CAPEX þ $25M/annum OPEX savings through optimisation of storage and shipping requirements for a LNG supply chain.

• $300M CAPEX saving for a major oil and gas operator through optimisation of installed equipment redundancy.

• 10% increase in throughput through optimised maintenance for an offshore gas platform.

APPLYING RESILIENCE MODELLING TO THE WATER INDUSTRY Resilience of a desalination plant – throughput based

The Authors have recently applied the techniques of resilience modelling to a desalination plant.

There were significant concerns on the initial reliability of the facility and the owner wished to establish the reliable output of the plant and to determine if cost effective options were available to improve this.

The first part of the study consisted of establishing a model of the plant that identified the various plant components, their interconnections and the reduction in output that occurred if one or more of each component was unavailable. This information together with reliability data was used to create the resilience model. The model was loaded into the Optagon software to allow analysis to be performed As mean-time-to-failure (MTTF) and mean-time-to-repair (MTTR) is not readily available for the Water industry the – Offshore Reliability Data Handbook (OREDA) (OREDA 2009) data from the offshore oil and gas industry was used as a basis. Examination of this data suggested that the repair times given were unrealistic in an environment without 24-hour working by maintenance personnel, and so this data was multiplied by a factor of 3 and then rounded up to the nearest day to give a more representative value for this application.

This model was then run as described above. A typical throughput simulation at the 50th Percentile (P50) condition is shown in Figure 2 below.

The Monte Carlo Analysis described above looked at many thousands of flow profiles, similar to that shown in Figure 2, to determine the probability of production availability. These profiles were randomly generated largely using the MTTF and MTTR data. This is shown in graphical form in Figure 3 below and the P5, P50 and P95 values are given in Table 1. P50 represents the availability which 50% of all simulations exceed. A production availability of 100% represents the case with no shortfall expected and therefore would require everything to work perfectly – in reality, this is not achievable. Over the life of a facility there will be periods where production is operating at 100% expected throughput level, however there will be certain periods of time where reduced or no throughput is experienced.

Further analysis of this data allowed an evaluation of the contribution of the different system components to the production shortfall. The results of this analysis grouped by type are shown in Figure 4.

One surprising outcome from this analysis was the contribution of the motor control centres (MCCs) to the reduction in plant throughput. The plant uses active front ends (AFE) for the large drives on the plant. Each of these are water cooled with primary and secondary cooling loops and heat exchangers. In the secondary loops there is a single pump for each AFE with no redundancy

108

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

Figure 2 | 50th percentile throughput simulation of a desalination plant.

Figure 3 | Probability of exceeding a speciďŹ ed production availability at a desalination plant.

in the system. This cooling system introduced a reduction in reliability into the design that had not been appreciated until this analysis was done.

Following the identiďŹ cation of the Base Case condition, a number of different scenarios were investigated to identify their potential impact on the Plant. These ranged from relatively low cost options such as increasing the redundancy of the MCC cooling to the high cost option of adding an additional RO train. The model was rerun for each of these scenarios to establish the impact on production availability.

As can be seen in Figure 5 the largest impact on the production availability was not enabled by the addition of additional plant but by the change in the time to repairs. While 24 hour maintenance is not a practical option in most cases this information provides information to target maintenance contracts

109

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

Table 1 | Percentile production availability for a desalination plant Base case result

Production availability (%)

P5 (5th percentile)

91

P50 (50th percentile)

86

P95 (95th percentile)

79

Figure 4 | System component contribution to production shortfall.

Figure 5 | Probability of production availability scenarios.

with guaranteed response times and repair times for the equipment which is identified as critical. It also highlights the importance of generating water industry specific information on MTTR. Resilience of a wastewater treatment works facility – quality based

A decision support tool was developed to assess the resilience of a wastewater treatment plant. Performance was measured by non-compliant events based on breaches in environmental discharge

110

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

Figure 6 | Resilience sensitivity analysis for a wastewater treatment works.

consent of the final effluent. The tool considered historical equipment reliability data, equipment configuration and criticality and insights from site personnel including aspects relating to maintenance, spares, equipment reliability and impacts of equipment failure. In order to define criticality Biowin modelling of the system was carried out with equipment in a failed condition. This allowed input to the model of data on the plant capacity with equipment failed and the time delays between equipment failure and deterioration of plant effluent.

Key results included a prediction of the number of non-compliant events and identification of the key contributors to non-compliant events. A number of sensitivity cases were developed to highlight ways that would improve or reduce plant resilience including consideration of critical spares availability, equipment failure frequency, additional impacts identified by the project team and equipment criticality.

The flexible tool enables the operator to update the model and use it to support cost effective asset management decisions in the following areas:

• Installing spare equipment: By highlighting key contributors to non-compliant events, the benefit to plant resilience can be quantified by considering additional equipment redundancy.

• Maintenance and Repair Strategy: By changing either the frequency of equipment failures and/or how long repairs take, the most sensitive equipment to plant performance can be identified.

• Spares Holding: Having boxed and stored spares available on-site could significantly reduce the number of non-compliant events.

• Plant Load: Impacts can be changed depending on the flow through the facilities. For example, plant load may need to increase due to population growth. The model will predict the expected increase in non-compliant events.

• Additional Impacts: Any known or expected issues (eg extreme weather events, blockages during winter) can be included to quantify their impact on plant resilience.

• Removal of old and installation of new equipment: comparative assessments between different options can be performed by changing the model parameters.

Resilience research project of a typical water re-use facility

The approach has also been taken a step further in the Australian Water Recycling Centre of Excellence National Demonstration Education and Engagement Program project. The project seeks to address concerns regarding the use of recycled water as a potential source of potable water. The project is multi-faceted having many streams of work. One sub-stream is focused on trying to quantify and provide credible evidence that the mechanical reliability of water recycling plants are robust and pose no more

111

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

risk to public health than those of a more typical water treatment process – in essence determining the resilience of the water recycling process. The project is ongoing, however the approach being taken is to create a representative Reference Plant (with several configuration options) to represent a typical water recycling process, and then to use industry data to inform the model from plants around the world that have been operating for a long period. The requirements around the data are complex owing to the multiplicitous nature of the data collected from the different operators. It has been vital to develop a robust data standard to streamline the collection, analysis and storage of the supplied data. The focus of the resilience modelling for this project is the quality of its performance as opposed to its production throughput. The results aim to quantify the risk to public health by predicting likely frequency of non-compliance over a set time period. The vital data for the modelling, in addition to the reliability data, is the equipment set redundancy and its critical number, and the likely deferred impact time.

The critical number refers to the point at which the number of equipment items would lead to a quality breach. It is expected that the Reference Plant and the supporting data will be a great source of information and support for asset management planning in the future (eg optimizing maintenance strategies in plants, better targeting of critical equipment for renewal planning purposes).

THE CHALLENGES OF AVAILABILITY DATA The need for good quality data that is representative of critical outages of equipment is important for ensuring meaningful results, so that any outputs can be used with an acceptable level of confidence.

Ideally equipment performance should be based on actual operating records where these are available. However, historical operational data is usually limited in its use since the main purpose of these records is typically for detailing personnel working hours and providing an overview of maintenance tasks performed and not primarily for the purpose of recording critical equipment outages that would feed into resilience modelling. It is therefore important that if equipment performance is to be based on historical records there are good maintenance management systems (eg Maximo, SAP) in place so that relevant information is recorded.

Where historical records are deemed to be unsuitable, data from generic industry sources may be used. For example, the oil and gas industry has a commonly used data source, OREDA, that has been developed through contributions from operators across the entire industry – this handbook includes failure frequencies and repair times for different equipment types at both the equipment and the component level. Such data is very useful for representing facilities that are yet to be built, but it is not currently widely available in the water industry. Vendor data may be used but normally is not representative of expected performance since equipment behaviour will be process and operation specific, and vendors may present optimistic views of the reliability of their equipment. As noted above, a project is currently in progress that assesses the resilience of a typical water re-use facility with equipment reliability data based on historical data from several sites from around the world.

In terms of data handling, general rules should be applied to the data set, where possible, in order to filter out any events which are not relevant for modelling purposes (i.e. events that have not resulted in a critical outage of equipment should be excluded). For example, work order priority may be an indicator of event criticality. Depending on the data recorded, a line-by-line review may be the only option in order to determine whether an event is deemed to be critical and how long the critical outage lasted – however, this may be a time consuming exercise that requires significant further investigation.

Previous resilience modelling experience highlighted the following challenges with using historical data:

• Recording of information may be inconsistent. For example, a higher work order priority may be assigned that would indicate that a critical outage has occurred, whereas in reality it was not a critical event.

112

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

• Outage durations may not necessarily indicate how long equipment has been unavailable for – typically an overestimation is provided since the work order may include diagnosis of problems, delays in administrative procedures and times awaiting a spare component. In reality, during such activities the equipment may actually be operating. There is also the possibility for double-counting when more than one person has worked on the maintenance activity – for example, 30 maintenance hours might be recorded when in reality 2 maintenance crew attended to the repair and therefore the equipment was unavailable for 15 hours.

• The level of detail provided in the outage durations is typically limited. The resilience model can represent a breakdown of times including logistic delays, spares lead times, actual repair duration and restart times – however, this level of information is not typically recorded hence a single overall outage duration is usually considered.

Key benefits in using historical operating data include:

• Not having to rely on expert knowledge or judgement – hard recorded data that can be audited will provide a consistent approach in data handling. Benchmarking is also possible against actual performance experienced.

• Provided that the level of data recording allows it, a detailed representation beneath the equipment level (i.e. component level) can be represented including consideration of critical spares holding arrangements.

• Regular updates of models are possible to account for newly recorded events.

One area of interest would be to determine the impact on facilities performance with different maintenance strategies. It would be expected that with increased and improved preventative (planned) maintenance, equipment failures would be less frequent. This could be achieved by performing sensitivity analysis and comparing predicted performance with various changes to equipment reliability and planned maintenance activities. It should be noted that resilience modelling will not directly quantify how much equipment reliability changes if a certain maintenance regime is applied – expert input will be required to allow sensitivity analysis to be undertaken. Improved data gathering on the benefits of preventative maintenance would therefore be valuable.

In order to take into account uncertainty in equipment performance, sensitivity cases can be developed to quantify the change in resilience if equipment reliability varied (i.e. increased or reduced outage times, more or less frequent critical failures).

The following improvements to collecting historical performance data could be made:

• Clear distinguishing of critical from non-critical outages in order to take into account the correct events.

• A more detailed breakdown of event durations could be provided in order to separate critical outage durations from other activities when equipment may still be operational.

• As more resilience analysis is undertaken it would be beneficial to compile data from various assets to create a water industry ‘typical’ performance data set that is similar to the oil and gas industry generic data.

A VISION OF THE FUTURE Since the financial downturn of 2008, utilities have found themselves in a situation where, despite the need to invest in rehabilitation or replacement of critical infrastructure to maintain serviceability, the regulatory appetite to raise prices to fund this investment has not been favourable. The result has been an increasing focus on extending the useful life of the asset base, deferring investment and ‘sweating’

113

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

the assets. Good practice asset management approaches are being adopted to better target the funding that is available. Associated with these approaches, utilities are seeking to better understand and manage the asset lifecycle, increasingly using risk-based approaches that balance the criticality of the assets and their likelihood of failure. This is a trend that is likely to increase in the future.

Adopting a lifecycle asset management approach will necessitate a greater understanding of asset resilience as utilities use risk to target investment. At each stage of the lifecyle, from planning, through design, construction, operations, maintenance and replacement, the effects of asset resilience on funding decision making will need to be considered. As an example, resilience will likely become a key design factor, being built into rehabilitation and replacement projects, as well as new-build and expansion projects.

As well as being optimised for performance and cost, the designs of the future will be required to perform in terms of resilience having to meet set targets. Life cycle costing, using resilience as a key determinant will become the norm. Resilience will be an equally important consideration from the viewpoint of operations and maintenance optimization, and the types of models described above will need to be enhanced to enable scenario testing as resilience becomes a core of lifecycle optimization. In the future, design approaches and their supporting tools within the utilities corporate environment will need to integrate or incorporate resilience functionality, making it a business-as-usual process.

As utilities install enterprise asset management decision support systems, especially computerised maintenance management systems, the improvement in data, analysis and decision making around maintenance optimization will change exponentially. The financial drivers mentioned above will encourage the optimization of maintenance, which in turn will likely reduce the margins for error.

Understanding and modelling the effects of the resilience of the asset base on maintenance expenditure will also become business-as-usual.

The lack of good quality data is likely to be a key issue in the short to medium timeframe. Water operators will need to share their data, like the oil and gas industry has in creating and maintaining OREDA, to make available a common industry source of good quality equipment performance related data. This data will be very useful for operators during design (i.e. for equipment not yet built) and generally where past practices have meant historic performance data at equipment level is limited or of poor quality.

As described above, the key to understanding resilience is to evaluate whether the asset base can perform and maintain its desired function under both routine and unexpected circumstances. As part of vulnerability assessments, utilities need to consider a broader range of potential threats than ever, including cyber security and terrorism. The resilience modelling of the future will need to enable utilities to build such considerations into their assessments, considering the broader range of factors that can cause the asset system to be performing under ‘unexpected circumstances’. The mitigation measures for these new threats will involve new assets, processes and procedures and the effectiveness of these measures will need to be assessed in cost benefit terms. This is the key challenge for the resilience models of the future.

ACKNOWLEDGEMENTS This work was supported in part by a grant from the Australian Water Recycling Centre of Excellence, Brisbane Australia.

REFERENCES Aguado, Rosen 2008 Multivariate statistical monitoring of continuous wastewater treatment plants. Engineering Applications of Artificial Intelligence 21, 1080–1091.

114

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.012

Childress, Le-Clech et al. 2005 Mechanical analysis of hollow fiber membrane integrity in water reuse applications.

Desalination 180 (1–3), 5–14.

Corominas, Villez et al. 2011 Performance evaluation of fault detection methods for wastewater treatment processes.

Biotechnology and Bioengineering 108 (2), 333–344.

OREDA Project Participants, SINTEF Industrial Management 2009 OREDA Reliability Data Handbook, Norway.

Rogers 2000 Use of the OPTAGON package to predict the reliability and availability of a gas facility, determine component criticality and assist in gas contract decisions proc. ESREL Conference, Edinburgh 2000.

Rosen, Rieger et al. 2008 Adding realism to simulated sensors and actuators. Water Science and Technology 57 (3), 337–344.

Seah, Poon et al. 2003 Singapore’s NEWater demonstration project: another milestone in indirect potable reuse. Water 30 (4), 74–77.

© IWA Publishing 2014 115

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.014

Twenty-two percent less water on maize with a new subscriber decision support tool Bertrand Pinela and Marianne Moalicb a

Terrena Innovation, France. E-mail: bpinel@terrena.fr

b

Terrena Agronomic Department, France

Abstract As a case study, this paper shows how a French cooperative seized on a new technology (capacitance sensors), tested it in real conditions (irrigation of maize in several Agronomic Department trial fields) and is now able to offer to its farmer members a subscriber decision support tool (DST) to increase water use efficiency.

Thus, technical and scientific trials have been conducted with fourteen farmers over the last four years. Managing irrigation with capacitance probes has resulted in an increase in water efficiency: most of the time, there is a reduction in water supply (the average is 22% (range 4–44%) less water use than without the probes). Sometimes, only a yield increase is observed with the same water quantity (see 2010 trial – farmer 2).

In 2012 Terrena conducted market research on 28 farmer members and 20 crop advisors. This showed that most farmers who irrigate do not use irrigation management tools. However with a tightening of regulations and an increase in irrigation costs, attitudes are changing. Thus, 12 farmers questioned, said they were ready to pay for an irrigation management service based on capacitance probes (continuous real time monitoring of soil moisture). It also showed that irrigation advice needs to take into account global irrigation system constraints (number of rollers, pump speed and the like). Farmers are ready to pay for an efficient irrigation management DST. So as to be ready for a 2013 launch, a DST soft-launch was set-up in 2012 on ten experimental farms. The usual working practices of irrigators cannot be expected to change overnight. However, this study shows how important is the farmers’ involvement in the building of this service, coupled with that of their crop advisors. Because farmers were involved at the beginning of the DST creation process, it made it possible to convince all of them to use this tool in a more sustainable manner.

PART 1 – TERRENA, ECOLOGICALLY INTENSIVE FARMING (EIF) AND WATER MANAGEMENT Terrena (www.terrena.fr) is one of the biggest French agricultural cooperatives and is located in the West of France. Its turnover was € 4.6 billion in 2012. It groups together 22,000 farmers with approximately 2 billion hectares of Utilised Agricultural Area and 12,000 employees. As other French cooperatives, there are three main different jobs in Terrena

• Upstream, the commercialisation of farming inputs (pesticides, fertilisers, seeds, irrigation, animal feed and care, etc., and technical advice to help members to correctly use these inputs (around 300 advisors). In France, a large part of the private extension services (advice, knowledge transfer, etc) is carried out by agricultural cooperatives.

• Downstream, Terrena collects its farmer members’ production (cereals, wine, milk, cattle, pigs, poultry, etc) to the value of € 2.4 billion, approximately 10% of the agricultural production of the West of France, a key European region).

Also downstream, using its industrial equipment, Terrena processes and commercialises the equiv• alent of € 2.9 billion of products (agro-food industry) through major retail outlets to the general public and final consumers.

116

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.014

In 2008, in a general context of increasing staple food demand, energy price rises and increasing social demand for more sustainable farming, the Advisory Board of the cooperative decided to focus its strategy on EIF. EIF, also called Sustainable Crop Production Intensification by FAO, could be one of the present paradigms for world agriculture, aiming at ‘producing more and better with lower inputs’, for example, less irrigation water.

On the one hand, food production needs to be increased to feed a world population of 9 billion inhabitants expected in 2050 whose eating habits will certainly evolve. Indeed, FAO says it will be necessary to increase farming production by 70% by 2050 (FAO 2011). Clearly, it is on the African and Asian continents that the problem will be most keenly felt. But in a world market, commercial flows are interdependent and this issue notably involves the equilibrium of European markets.

In Europe, on the other hand, farmers and society want to reduce the use of artificial and nonrenewable inputs (pesticides, fertilisers, antibiotics, energy, water, and so on). In fact, these inputs will become rarer and rarer and more expensive as they are related to the price of oil) and they possibly result in a negative impact on ecosystems (water pollution, air pollution, etc) that society is no longer prepared to accept.

Quantitative water management: a major social issue

Water resources in France are not restricted as in some countries, but periods of summer drought are becoming more and more pronounced. Arable crop irrigation is a relatively recent practice. In fact, it started in the 1970s. Surface areas irrigated have trebled since then but have stagnated since 2000.

‘With our present dry summers and their potential to create problems for the environment and leisure activities, the use of water by and for farming can be a source of social conflict between rural and urban populations. There is therefore a real need for irrigation management’ (Bergez et al. 2008).

In France, 50% of irrigated surfaces are covered by maize (forage, grain or seed production) (RGA: 2010), namely, around 750,000 ha annually. Depending on French regions and their different soil and climate conditions, the average water needs for maize irrigation vary between 150 and 300 mm (1,500–3,000 m3/ha), in other words a 25–30 mm water supply either every 5 or 10 days.

Farm level irrigation management implies a two-stage thought process for the farmer (Bergez et al.

2008)

• strategic considerations before the irrigation campaign; • tactical considerations during the irrigation campaign.

It is important to note that farm level quantitative water management is not only a question of irrigation. Farm level irrigation management initially involves strategic considerations regarding

• choice of crops and varieties; • choice of crop management (technical and cultural procedure); • choice of material.

Strategic considerations entail long term decisions concerning the organisation of the farm (choice of material, water access rights, resource creation, etc) and short term decisions (crop rotation choices, planning a projected irrigation schedule, and so on).

For these strategic considerations, technical advisors, like Terrena’s, already operate and, generally, farmers have access to efficient advice.

Afterwards, tactical considerations involve the farmer’s action plan regarding irrigation management, in other words, how to adapt the supply to the situation and the system as it is at the moment of action. Quantitative water management involves irrigation management and in particular irrigation scheduling during the campaign: the farmer must implement the projected irrigation

117

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.014

schedule and adapt it to soil and climate conditions at the moment of action. There are numerous irrigation DSTs available today. The best known are the use of a tensiometric probe and the calculation of crop water balance. But these tools are little used for arable crops (maize).

A few years ago, capacitance probes appeared in the labs and then on the market (Paltineanu & Starr 1997). They represent a real innovation and seem to offer a good quality-price ratio, mainly due to the fact that soil moisture levels can be monitored in real time. Even though prices have fallen, they are still relatively expensive, while becoming more robust and their readings more reliable.

They therefore seem to be a good tool with which to read soil moisture levels with a potential mass market in the farming world when compared with other tools on the market.

For the record, the operating principle of these probes is to feed an electric current to each sensor (one sensor for one depth, four sensors on one probe). The surrounding soil is covered by a 5 cm radius electric field. In this way the soil dielectric is read at each depth and as this depends essentially on the volume of water present, soil moisture at each level can be calculated.

The probe readings are taken every thirty minutes and recorded by a data logger. They are then sent automatically to an Internet server every six hours. Consequently, soil moisture level changes can be monitored in real time from our computer through specific software.

Taking all of these elements into consideration, Terrena cooperative has decided to offer its members an irrigation management service for maize, based on capacitance probes.

The initial idea is to create a turnkey service for farmers who irrigate, including

• Material: installation/ extraction of the capacitance probes in the fields.

• Advice and technical assistance: deliverable, commitment service.

• After sales/ repair service.

To ensure that advisors and farmers gave the best reception possible to a new DST, technical and commercial analysis was required.

PART 2 – 4 YEARS’ TECHNICAL TRIALS In order to help farmers improve their water supply management, Terrena has set up many trials on irrigation management. The cooperative has focussed on the main irrigated crop: maize (silage, grain and seed production). The goal: how to reduce water use without reducing yields.

Between 2009 and 2011, the Agronomic Department of Terrena decided to test 4 different kinds of capacitance probes, with respect not only to the strength and viability of soil moisture readings, but also the price and the economic return for the farmers. In 2012, Terrena decided to use exclusively one of these kinds of probe, which the Agronomic Department considered the most robust and convenient (Figure 1).

The probes used by Terrena over this 3-year trial period were

• A capacitance probe containing 4 sensors at 10, 20, 30 and 50 cm respectively, allowing for readings at these different depths. It is possible to connect 4 capacitance probes to one data logger.

• A data logger, supplying the probe with electricity, and recording and sending the readings via a GPRS (data transmission service) link. This data logger runs off a battery linked to a solar panel.

The first three years of trials (2009–2011) took place in fields made available by volunteer members of the cooperative. The experimental objectives and protocols were different from one year to another and are summarised in the following table (Figure 2).

118

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.014

Figure 1 | The more recent probe, used in 2012 and 2013 by Terrena’s Agronomic Department.

Figure 2 | 2009–2011 trials.

119

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.014

The main results: improved water irrigation efficiency

Water efficiency was measured by comparing the DST managed strips and those irrigated in the farmer’s traditional manner. The two managements use the same machines (rollers or drippers), and the same water. The differences between the two systems are based on the different periods and quantities of water applied.

These three experimental years highlighted an improvement in water efficiency on all the trial farms (Figure 3). The results for 2010, for example, can be seen in the following figure:

Figure 3 | The 2010 results for 4 trials.

In 2010, seven trials were conducted. But technical problems made it difficult to get useable results from three of them, so only four results are available for use.

For farmer 1, using the probe resulted in a 32% water saving, namely 2 water rounds of 30 mm, while maintaining yield.

For farmer 2, there was no water saving but the probe managed water supply was better positioned eliminating water stress. This resulted in a 1.9 T/ha increase in relation to the usual practice.

For farmer 3, he could have saved a water round of 25 mm on the managed plot but he did not follow the advice. This fact is also a kind of result: embrace change, accept less irrigation!

For farmer 4, probe management linked to surface sprinklers resulted in a 41% water saving while maintaining yield.

It is worth noting that the probe management often influenced the farmer’s ‘traditional’ practise.

The true difference in water efficiency between the two methods is therefore often underestimated.

2012 : A soft-launch year

In July and August 2012, in addition to the trials presented above, the Agronomic Department of Terrena decided to ‘soft-launch’ its future management advice service. Within the 10 farms setting up comparative monitoring of managed and non-managed irrigation amounts, the Agronomic Department simulated a weekly advice fax based on the monitoring of soil moisture graphs, weather forecasts and the water needs of crops. Apart from the technical results, this ‘soft-launch’ year helped to validate the ‘logistic’ and commercial feasibility (Cf. following section) of the future DST. The new DST has been called ‘Niléa’.

Figures 4–6 present one of the ten 2012 trials: on P2, with Niléa advice, the farmer decided to begin the irrigation only on week 29. Without Niléa, on the control plots, the farmer had begun to irrigate much earlier: on week 27 (Figure 6). The probes installed in these control plots showed the saturation of the soil at 50 cm deep (Figure 6). Rooting was penalised, especially as the beginning of July was cold and wet. The soil moisture curves did not flex: the crop’s water demand was very low.

In this example, water efficiency has been doubled, from 11 to 22 kg/mm of water applied.

120

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.014

Figure 4 | ‘Triganne trial’: an example of one of the 2012 trial fields with maize seed production.

Figure 5 | ‘Triganne trial’ results.

Figure 6 | ‘Triganne trial’ soil moisture curves in the area without capacitance probe management (Niléa).

These 4 experimental years on farms resulted in a positive change in their methods. They have adopted a more ‘reasoned’ irrigation approach. They all agree on the fact that the tool indicated when they should start irrigating, when they should stop, when they should begin again after rainfall and how much they should apply. All of this is done in a pedagogical and visual way through the use of graphs.

These helped farmers to manage irrigation but also to get to know their soil, rooting depth and crop water requirements better. They could see in real time the effect of their practises and of the climate.

121

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.014

In total, after the experimental years (2009–2012), the average water saving, only through changes in irrigation thanks to soil moisture monitoring, is estimated at 1 to 2 water rounds, namely 25–50 mm (250–500 m3/ha). This represents an average saving of 22%, with large variations depending on situation, from a saving of around 0% to almost 44%.

Technical problems, breakdowns

Robustness: despite these positive results in terms of farming practices and the interesting practicalities of the tool, some of the capacitance probes we have tested were very fragile. Numerous breakdowns occurred during the trials: cables cut between the probe and the data logger, due to tractors or animals on the field, sensor electric welds breaking, and problems with the printed circuit boards or with the GPRS connection.

Installation: putting the probes in the ground is also a very delicate process, and requires technical skill to have a good connection between the soil and the probe. If this connection is weak, the readings are incorrect. Because the reading radius is very small, there should be no stone or air bubble between the soil and the probe Viability: in certain extreme types of soil (very sandy, very stony, etc), moisture measurement does not always seem to be viable. There is often a variation between the quantity of water measured with the captors and the real quantity of water on the field (irrigation or rainfall). This can be explained by a bad soil/probe connection or by a ‘temperature’ effect, which influences the dielectric reading. The salinity and organic matter content of the soil both also influence dielectric monitoring.

The probe reading values must therefore be treated with caution. Experience in using the tool and the presence of a rainfall gauge next to the probe can indicate the need to lower or raise probe reading values.

Above all, the probes are not sufficient, on their own, to form the basis of good advice for the farmers. The 2012 soft-launch also showed that the soil moisture has to be analysed weekly by taking into consideration many other aspects, for example: the crop stage, past irrigation rounds, the weather (past and forecast) and irrigation system constraints. Only if the DST is able to take all of these aspects into account will it be interesting or of value for the irrigators. Thus, the 2012 survey (Cf. next section) showed that the farmers were expecting a turnkey service.

PART 3 – MARKET RESEARCH Beside the technical and scientifical trials, a qualitative market survey was conducted by the Marketing Department of Terrena to validate the interest to farmers of this kind of new technology. In 2012, 28 surveys were carried out.

The initial hypothesis was that

• There would be a development potential for this irrigation management DST over the cooperative region and mainly in three sectors, according to the crop advisors’ experience ○ The Vallée de l’Auhion: an emblematic seed production area conditioned by irrigation: 850 irrigators present in the area of which 100 are maize seed producers for Terrena Semences. ○ The Pays de Retz: maize grain silage production area but also field grown vegetables with an 80strong irrigators association. ○ The Sud Vienne: arable crop area, irrigation is essential to guarantee yields: 855 irrigators present in the administrative area organised into the Vienne Irrigators Association (ADIV).

• To begin with this service could be developed for maize (grain, seed, and silage): the main irrigated crops on the farms. The maize seed producers, a high added value crop, would invest more easily in an irrigation management subscriber service.

• Partnerships could be envisaged with certain water management players in the targeted areas.

122

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.014

Objectives

To validate all of these hypotheses, the market study relied in particular on surveys carried out on farmers who irrigate (commercial target). The main objectives of the surveys were

• To have an overall approach to irrigation management on the farm: to understand the cooperative members’ practices and identify the ways in which they regard irrigation on their farms (advantages and disadvantages).

• To identify their expectations with regard to irrigation and more particularly the management tool.

• To present the capacitance probes and collect opinions on this new tool: level of interest, checks on the use of such a tool, expectations of a possible service provided by the cooperative.

• Also, to collect the feedback from the six 2011 cooperative capacitance probe trial farmers: their expectations, level of satisfaction, software ergonomics … with the aim of having a more detailed vision of tool use, the software, and the sort of advice and support that members wanted, if the service was set-up.

The representative sample was organised jointly by the Marketing Department and the crop advisors according to classical marketing methods (Churchill & Iacobucci 2009). For each of the 3 areas identified the advisors suggested the names of members they knew. To be representative, each advisor provided the names of farmers who, according to them, were potentially interested in irrigation management and water saving, as well as others not necessarily interested in the subject.

This second farmer group was, in effect, essential to discovering why the majority of irrigators do not use irrigation management tools today: what stops them?

To obtain useable trends, the objectives were fixed for 10 surveyed farmers per target area including the 2011 capacitance probe trial farmers according to the following distribution: 5 farmers interested in irrigation management and 5 farmers having no prior interest in irrigation management.

The survey was based on a questionnaire designed jointly by the Agronomic and Marketing departments, and the EIF R&D Team of the cooperative. To meet the requirements, the questionnaire was divided into three main sections

• General information about the farmer and the farm.

• Irrigation on the farm: surface of agricultural land irrigated, irrigation equipment, irrigation water sources, irrigation scheduling.

• The market approach: economic interest of irrigation for the farm, current sources of information and advice on irrigation management, presentation of the capacitance probes and the farmer’s immediate reaction to the interest of the tool.

For the 2011 probe trial farmers, the ‘market approach’ section of the questionnaire was modified to collect feedback about their experiences and their expectations.

Regarding irrigation scheduling on the farm, the questionnaire was targeted mainly at maize growing: the main irrigated crop on the farms and the first crop targeted if the service was to be developed.

Interviews took place on the farms in a semi-directive way to allow the farmers to develop each question and to get the maximum number of elements.

At the end of the first five interviews, the questionnaire underwent some changes mainly in changing the wording of the questions.

The answers to the questionnaire were analysed through the development of an Excel database and the summaries of surveyed farmers, it being essentially a qualitative and not quantitative approach.

Without entering into the details of all the results of this survey, we can attempt a summary of the knowledge obtained:

• 19 of the 28 farmers agreed that irrigating was a constraint in terms of working time especially during busy periods such as harvest time. Nevertheless, for the 9 other farmers, it was not

123

Water Practice & Technology Vol 9 No 1 doi: 10.2166/wpt.2014.014

considered a constraint as it was part of their job and the final result was worth it: irrigation equals a certainty of harvest and income and/or business existence for specialised arable farms (seed production, vegetables, flower production, etc). Working time constraint should therefore be taken into consideration when developing an irrigation DST. It is obvious that the use of capacitance probes should not increase the farmer’s workload.

• To be useful for the farmers, the DST should be offered as a turnkey service. That means that during the summer campaign (July-August), irrigators consider they don’t have enough time left to take care of the probes (installation, repairs, and so on), or go onto the Internet to download the soil moisture data, and they have even less time to analyse the curves. They are willing to pay for this kind of turnkey DST service because it will bring them information that they couldn’t get readily on their own.

Even though water shortage has to be put into perspective on the surveyed farms at present, the set• ting-up of quotas in the Authion River, their reduction in the Vienne, and an ever increasing tightening of regulations regarding the building of hill reservoirs results in a major water economy challenge. This makes for development potential in ways to optimise water management within farms and so, the use of irrigation management tools.

• The fact that farmers do not know the price of water or energy implies that as yet it is not sufficiently expensive for them to turn to the use of irrigation management enabling them to make efficiency savings.

Finally, it was a surprise to learn that 66% of the farmers who had been surveyed were interested in the technology and the potential launch of a DST tool by the cooperative.

CONCLUSION In 2013, Terrena reached its objective: the cooperative launched its new DST service. The price of this service has been calculated with all the costs that are needed to produce irrigation advice (probes, advisors, weather forecast, after-sales service, etc). The price of this service for the next (2013) irrigation campaign is 1,800 €/farm. Terrena decided to sell this DST to only 30 farmers, even if many more farmers wished to try it for the campaign. Once more, this shows the great interest that farmers have in improving their irrigation practices, if such improvement does not mean ‘more complications, more working time …’. It will also give the cooperative the opportunity to adjust its offer for next year.

Terrena really considers the capacitance probes as a great new technology. They are potentially very useful for optimising irrigation practises on maize. The probes are still fragile and expensive, however.

For the next few years, one topic of R&D projects could be to try to extrapolate the soil moisture data on a larger areal scale, by analysing and modelling the moisture readings in a network of capacitance probes.

REFERENCES Bergez, J.-E., Garcia, F., Leenhardt, D. & Maton, L. 2008 Optimising irrigation management at the plot scale to participate at the regional scale water resource management. In: Topics on System Analysis and Integrated Water Resource Management (Castelletti, A. & Soncini Sessa, R., eds), pp. 141–160.

Churchill, G. A. & Iacobucci, D. 2009 Marketing Research: Methodological Foundations. Cengage Learning, pp. 625.

FAO 2011 Save and Grow. FAO, Rome June 2011.

Paltineanu, I. C. & Starr, J. L. 1997 Real-time soil water dynamics using multisensory capacitance probes: laboratory calibration.

In Soil Science Society of America Journal 61 (6), 1676–1585.