One of the options to recycle phosphorus (P) in the wastewater sector is to recover it as struvite crystals from digested sludge. Measurements on a full-scale demonstration plant in Leuven, Belgium, yielded a first indication of the profitability of struvite recovery, in function of different variables such as incoming PO43− concentration, MgCl2 dosing, improved dewaterability, etc. An uncertainty and sensitivity analysis was carried out. Although possible improvement in sludge dewaterability when recovering struvite from digested sludge has a positive economic amortization effect, it is at the same time the largest source of financial risk. A theoretical exercise showed that for struvite recovery from centrate, uncertainty would be lower, and the largest sensitivity would be attributed to ingoing PO43− concentration. Although struvite recovery from digested sludge is riskier, it is an investment with potentially a higher return than investment in struvite recovery from centrate. The article provides information for possible financial incentive schemes to support P-recovery.

INTRODUCTION

Global phosphorus (P) reserves are gradually declining and are limited to a few exporting countries. Hence, nutrient recovery comes into the picture in different sectors of the P-cycle such as wastewater treatment (Cordell et al. 2009). Orthophosphate concentrations are generally rather low throughout a wastewater treatment plant (WWTP). In that way, P can only reasonably be recovered in sections of high (free) P concentration, namely from ashes of mono-incinerated sludge, from sludge waters of dewatered digested sludge or from digested sludge itself (Cornel & Schaum 2009).

This paper looks at P-recovery as struvite crystals (MgNH4PO4.6H2O) from digested sludge. In the past, struvite formation in WWTPs was first reported as a problem of natural scaling, and only later on as a means of recovery of phosphorus (Doyle & Parsons 2002; Liu et al. 2013). Controlled struvite crystallization requires an excess concentration of , together with a sufficiently high pH (around 7.5) and a source of magnesium, often added as MgCl2 or Mg(OH)2 (Hanhoun 2011).

P-recovery by means of struvite precipitation can typically only retrieve around 20% of the total P entering a WWTP. However, struvite recovery is a rather cheap technology as compared with other P-recovery techniques (Egle 2014), and a relatively clean end product is produced (Moerman et al. 2012; Muster et al. 2013). When alternative sources of Mg2+ ions could be used in the future as described by Lahav et al. (2013) and Kruk et al. (2014), this could still improve the financial viability of this recovery technique. If struvite is recovered from digested sludge, there is an additional positive financial impact of improved dewaterability of the sludge (STOWA 2012).

The European Commission (2013) states, however, that there is a clear need for context-specific life cycle and cost analysis of these recovery techniques, case by case. The WWTP of Leuven (120,000 people equivalent), operated by Aquafin nv, was found to be a suitable location for having a full-scale demonstration plant of struvite recovery from digested sludge. The recovery technique that was used was developed and patented by NuReSys® and was in Leuven for the first time used on sludge instead of on water. This demonstration plant has been in operation since April 2013 and intensive measurement campaigns have been carried out since then.

In this paper, a cost–benefit analysis, including uncertainty and sensitivity analysis, is carried out on the basis of actual full-scale results from this demonstration plant. To our knowledge, this is one of the first papers presenting an operational and financial balance for this relatively new technique applied at full scale.

MATERIAL AND METHODS

Model, inputs, and outputs

The operation of enhanced biological P-removal in the water line of WWTP Leuven and the presence of an anaerobic digester in the sludge line result in sufficient concentrations of and after digestion. The full-scale installation for struvite recovery from digested sludge (DIG) is fully described in Marchi et al. (2015) and a flowchart is presented in Figure 1.

Figure 1

Flowchart of the full-scale installation for struvite recovery from digested wastewater sludge in Leuven, Belgium.

Figure 1

Flowchart of the full-scale installation for struvite recovery from digested wastewater sludge in Leuven, Belgium.

The digested sludge goes through a cutter toward the stripper for natural pH increase by means of CO2-stripping. In a crystallizer, MgCl2 is added and the sludge moves by gravity to the harvester. The MgCl2 dosing is set at a molar ratio of 1.75 Mg:P at the end of the test year. The installation operates at a pH set point of 7.8 in the stripper and 7.5 in the reactor. In the latter, NaOH is added if needed to correct pH against the acidifying effect of MgCl2 addition and struvite formation. The installation chemically removes free from 220 down to 40 mg/L on average. The operational settings and results of the test year were used for calculating the cost–benefit. Full details on the chemical and mass balance results of the monitoring campaigns in Leuven can be found in Marchi et al. (2015). For the cost–benefit analysis, a financial model was built that takes into account the most important revenues and expenditures (Table 1).

Table 1

Positive and negative cash flows taken into account for the financial analysis of the struvite installation in Leuven, Belgium

Positive flows 
 Sales price of struvite 
 Reduced aeration cost for N-removal (reduction via feedback flow of the sludge waters) 
 Reduced cost for carbon source for P-removal (reduction via feedback flow of the sludge waters) 
 Reduced maintenance cost for clogged pipes (industrial cutter) 
 Lower sludge disposal cost due to improved dewaterability 
 Reduced polymer use in the dewatering process 
 Rest value metal 
Negative flows 
 Investment main installation, 8 m³/h sludge flow 
 Investment in chemical storage and dosing (to allow for bulk delivery and thus reduce chemical-dosing costs) 
 Investment struvite washing (due to working on digested sludge) 
 Operation man-hours 
 Operation electrical power 
 Operation MgCl2 consumption 
 Operation NaOH consumption 
 Maintenance 
Not considered 
 Research costs 
 Fixed costs for digestion, dewatering, drying, … 
 Possible benefits attributable to a lower Cd and U content in the struvite fertilizer as compared with natural P ore 
 Claimed benefits related to a lower P content in dried sludge pellets that go to the cement industry (better drying properties of the cement) 
Positive flows 
 Sales price of struvite 
 Reduced aeration cost for N-removal (reduction via feedback flow of the sludge waters) 
 Reduced cost for carbon source for P-removal (reduction via feedback flow of the sludge waters) 
 Reduced maintenance cost for clogged pipes (industrial cutter) 
 Lower sludge disposal cost due to improved dewaterability 
 Reduced polymer use in the dewatering process 
 Rest value metal 
Negative flows 
 Investment main installation, 8 m³/h sludge flow 
 Investment in chemical storage and dosing (to allow for bulk delivery and thus reduce chemical-dosing costs) 
 Investment struvite washing (due to working on digested sludge) 
 Operation man-hours 
 Operation electrical power 
 Operation MgCl2 consumption 
 Operation NaOH consumption 
 Maintenance 
Not considered 
 Research costs 
 Fixed costs for digestion, dewatering, drying, … 
 Possible benefits attributable to a lower Cd and U content in the struvite fertilizer as compared with natural P ore 
 Claimed benefits related to a lower P content in dried sludge pellets that go to the cement industry (better drying properties of the cement) 

The relation between Mg:P molar ratio and removal from the digested sludge is based on current experiences from the full-scale installation (see also Marchi et al. 2015). These curves are the basis to calculate the chemical removal percentage of in function of Mg dosing at a given fixed pH in the model. Other operational costs and benefits depend on the chosen set points of the installation and on ingoing flow and concentration.

From the model, the price per ton of struvite recovered necessary to have a discounted payback time for the installation of 10 years (interest rate 5%, inflation 3%, and technical depreciation period 20 years) is iteratively calculated. To do so, the actual struvite market price was set at zero, and it was calculated what the price should necessarily become to make the installation economically viable. The model also allows calculating discrete discounted payback times at given struvite prices.

Default values used to produce the currently most plausible results are presented in Table 2. The defaults are based on current operational settings and/or the experience of 1-year operation and/or the warranties of the provider. For example, the physical recovery percentage of 25% is currently reached, but in the uncertainty section, also scenarios with lower and higher values of recovery are entered in the model. Physical recovery percentage is the weight percentage of the chemically formed struvite that has actually been harvested. The incoming was kept variable between 150 and 500 mg P as /L, being a typical range for digested Bio-P sludge according to the authors’ experience.

Table 2

Default values and ranges around default values for different variables of the financial model of struvite recovery from digested sludge in WWTP Leuven, Belgium

Variable Unit Lower limit Upper limit Default 
Ingoing orthophosphate mg/L 150 500 220a 
Improved dewaterability Absolute % of dry matter 1.5b 
Mg dosing (32% concentration) Mg:P 1.1 2.2c 1.75a 
NaOH dosing (29% concentration) L/m³ DIG 0.5 0.0a 
Decreased polymer-use Relative weight % 16 8b 
Recovery percentage Weight % 15 65 25b,c 
Electricity price €/kWh 0.08 0.12 0.1a 
Man-hours h/week 0.5 1b 
MgCl2 price (32% concentration) €/kg product 0.05 0.09 0.07b 
NaOH price (29% concentration) €/kg product 0.128 0.208 0.168b 
Ingoing flow m³/h 8a 
Meters of piping still scaled with the installation on 55 4b 
Investment cost € d − 25% d + 25% d 
Polymer (PE) price €/ton active PE 260 460 360 
Man-hour cost € d − 20% d + 20% d 
Variable Unit Lower limit Upper limit Default 
Ingoing orthophosphate mg/L 150 500 220a 
Improved dewaterability Absolute % of dry matter 1.5b 
Mg dosing (32% concentration) Mg:P 1.1 2.2c 1.75a 
NaOH dosing (29% concentration) L/m³ DIG 0.5 0.0a 
Decreased polymer-use Relative weight % 16 8b 
Recovery percentage Weight % 15 65 25b,c 
Electricity price €/kWh 0.08 0.12 0.1a 
Man-hours h/week 0.5 1b 
MgCl2 price (32% concentration) €/kg product 0.05 0.09 0.07b 
NaOH price (29% concentration) €/kg product 0.128 0.208 0.168b 
Ingoing flow m³/h 8a 
Meters of piping still scaled with the installation on 55 4b 
Investment cost € d − 25% d + 25% d 
Polymer (PE) price €/ton active PE 260 460 360 
Man-hour cost € d − 20% d + 20% d 

aCurrent setting.

bAccording to operational experience.

cGamma distribution around current performance.

dConfidential. Variations take into account price offers of different providers.

Uncertainty and sensitivity

A sensitivity and uncertainty analysis is carried out by varying inputs in different Monte Carlo runs (Sin et al. 2011). Indeed, in order not to provide absolute figures, ranges are defined for the positive and negative financial flows. These ranges are chosen according to experience, literature, and/or the provider's information.

One thousand, one hundred Monte Carlo simulations led to different results of financial feasibility of struvite recovery in Leuven. The variation ranges of the input variables for the Monte Carlo runs are presented in Table 2.

Uniform distributions between the limit values of the inputs were considered, except for dewaterability, MgCl2 dosing, NaOH dosing, polymer price, man-hour cost, and investment for which normal distributions were selected. The reason for this is that for these variables, there is more certainty about the mean value after 1-year testing. Standard deviations for these normal distributions are chosen in order to have 99% of the randomly generated values fall within the chosen range (Table 2). For the recovery percentage, a gamma distribution was selected with a non-centred peak at 25% recovery. With enough Monte Carlo runs, this centres model results around the most plausible operation range of the installation (Sin et al. 2011).

Inter-parameter dependencies were taken into account for the dewaterability–Mg dosing relation and for the dewaterability–NaOH dosing relation. This is because Marchi et al. (2015) found that the improvement in sludge dewaterability largely depends on the ratio of bivalent over monovalent ions in the sludge. This ratio is clearly influenced by Mg2+ and Na+ dosing. By limiting the random search range for dewaterability improvement to plausible ranges in the case of higher or lower than average Mg or NaOH dosing, the theoretically impossible scenarios are excluded from the Monte Carlo.

Standardized regression coefficients (SRCs) are calculated as a measure for the sensitivity of the result in terms of changing variables. The sensitivity measures in this methodology are derived from linear regression of the outputs from the Monte Carlo runs. The necessary price per ton of struvite recovered to have a discounted payback of 10 years is selected as main objective value. A first-order linear, multivariate model is fitted to relate this price to all (changing) variables. Then, the regression coefficients of each variable are scaled by multiplying them with the ratio of standard deviations of the model inputs over standard deviation of the output from the Monte Carlo simulations. The obtained sensitivity indices are only valid if the fitted model is sufficiently linear, i.e. having an R² larger than 0.7 (Saltelli et al. 2006). The sensitivity indices obtained can have a value between −1 and +1. Coefficients close to zero mean that the output is not sensitive to that parameter. To reduce the uncertainty of a certain model or scenario, one should focus on finding a more reliable measure for the most uncertain factor. The square of the sensitivity indices is a measure for the percentage of the total variation that is explained by a specific variable.

Theoretical comparison with the same installation working on centrate waters in Leuven

To assess whether this DIG installation is optimally positioned within the water line, a theoretical comparison was made with an installation that would recover struvite from the sludge waters (centrate, ‘CEN’). These sludge waters leaving the dewatering equipment tend to be considerably loaded with and , because digested sludge is dewatered. For that theoretical CEN model, the financial flows of Tables 1 and 2 were taken into account, with the following differences:

  • The investment is lower because the investment for a struvite washing entity is not needed.

  • Improved dewaterability and decreased polymer-use most probably do not apply. One could argue that due to the feedback flow of CEN to the WWTP, part of the beneficial operational effects of struvite removal would be visible in the sludge at steady state. Because this feedback flow only represents a maximum of 15–20% of the incoming P-load in the WWTP, this effect is believed to be minimal (Marchi et al. 2015).

  • The DIG installation has a design capacity of 8 m³/h. With the same size of installation, the double ingoing flow of CEN can be treated. So the given investment cost of a DIG installation of 8 m³/h provides an installation of 16 m³/h hydraulic capacity when used for CEN (provider, personal communication).

  • Optimum set points for MgCl2 and NaOH dosing are lower for CEN. For MgCl2, the optimum Mg:P molar ratio is 1:1 with a narrower uncertainty range. Mg consumption for CEN is markedly lower than for DIG due to the low suspended solids content. Average NaOH consumption for CEN is at 0.1 L/m³ CEN (provider, personal communication based on full-scale recovery plants running on water fraction).

It should be noted that the above settings are derived from secondary information in the case of CEN, whereas for DIG, they relate to the full-scale installation.

RESULTS AND DISCUSSION

Payback and uncertainty

Figure 2 presents the results of the cost–benefit analysis including the uncertainty cloud. This analysis provides a relation between incoming concentration in the DIG and the necessary selling price per ton of struvite recovered to have an economically viable recovery technique. The curves could be the basis for an incentive subsidiary scheme to overcome the unprofitability of such a project, taking into account different ranges of P-price volatility. Based on an incoming concentration, necessary subsidies can be derived by subtracting the actual struvite market price from the presented curve.

Figure 2

Necessary selling price per ton of recovered struvite (€/ton) from digested sludge to have a discounted payback time of 10 years, in function of varying incoming orthophosphate concentration. Dots are the result of the uncertainty analysis, the full line represents the result for the default inputs. The colour codes represent varying improvement in dewaterability in absolute percentages of dry matter (the full colour version of this figure is available online at http://www.iwaponline.com/wst/toc.htm).

Figure 2

Necessary selling price per ton of recovered struvite (€/ton) from digested sludge to have a discounted payback time of 10 years, in function of varying incoming orthophosphate concentration. Dots are the result of the uncertainty analysis, the full line represents the result for the default inputs. The colour codes represent varying improvement in dewaterability in absolute percentages of dry matter (the full colour version of this figure is available online at http://www.iwaponline.com/wst/toc.htm).

The required value (whether or not subsidized) per ton of struvite recovered should be between €590 and €440/ton at incoming concentrations between 150 and 450 mg/L, respectively, according to the current default values (full line with equation in Figure 2). Yet, the risk of the operator deciding to invest in struvite recovery is still considerable (depicted by the point cloud around the cost function). With a guarantee of high ingoing concentration, the uncertainty range becomes somewhat smaller.

Given the current operational settings and results in Leuven, a struvite subsidy or selling price of €530/ton struvite recovered is enough to guarantee a discounted payback of 10 years in the installation of Leuven. Such a price is not impossible to be reached one day. During the 2008 food crisis, international P-rock prices were already much closer to this required selling price.

It is, however, also clear that within the current uncertainty cloud there is also a fraction of the 1,100 scenarios that support struvite recovery as an economically viable operation already at present. Therefore, the reader should look for scenarios (points) that lie below the current struvite selling price of around €50/ton in Figure 2.

The following items would logically make the recovery of struvite from digested wastewater sludge more beneficial:

  • Lower investment due to higher number of produced recovery units.

  • Higher international P-price.

  • Higher concentrations, e.g. by pretreatments before digestion causing increased hydrolysis.

  • Increased certainty about the improved dewaterability.

For CEN, the situation is different (Figure 3). Profitability is very low at lower P concentrations (<150 mg P/L), but on the other hand, the uncertainty on this investment is also far less than for recovery from digested sludge. This is mainly because the possibility of improved dewaterability is a highly sensitive parameter for the financial model of DIG, but it is not an influencing factor for CEN. The payback of the CEN installation is more dependent on the sales of struvite themselves and therefore on the recovery percentage and on the incoming concentration. Comparing Figures 2 and 3, one can see that investment in struvite recovery from discounted payback time (DIG) is riskier, but potentially with a higher return than investment in struvite recovery from CEN. For CEN, there are indeed no chances (none out of the 1,100 Monte Carlo runs) to have, for example, a discounted payback time (PBTdisc) 10 years for ingoing concentrations of 200 mg P/L at current struvite prices around €50/ton. In Figure 2 for DIG, this is still possible, with a considerable chance.

Figure 3

Cost per ton of recovered struvite (€/ton) from centrate water at a struvite market price of €0 to have a discounted payback time of 10 years, in function of varying incoming orthophosphate concentration. Dots are the result of the uncertainty analysis, the full line represents the result for the default inputs. The colour codes represent varying improvement in dewaterability in absolute percentages of dry matter. Note that the same investment cost as for digested sludge can treat a double hydraulic flow (the full colour version of this figure is available online at http://www.iwaponline.com/wst/toc.htm).

Figure 3

Cost per ton of recovered struvite (€/ton) from centrate water at a struvite market price of €0 to have a discounted payback time of 10 years, in function of varying incoming orthophosphate concentration. Dots are the result of the uncertainty analysis, the full line represents the result for the default inputs. The colour codes represent varying improvement in dewaterability in absolute percentages of dry matter. Note that the same investment cost as for digested sludge can treat a double hydraulic flow (the full colour version of this figure is available online at http://www.iwaponline.com/wst/toc.htm).

Due to the dilution of sludge with a water–polymer-mixture during dewatering, there is also a considerable dilution in concentration ingoing in the struvite recovery installation on CEN. This means that very high concentrations will not easily be reached in reality at WWTPs, even if the WWTP dewaters digested sludge. To have such concentrations in the DIG situations is more plausible.

Sensitivity analysis

The following Figure 4 presents the SRC of the financial model of DIG and CEN, respectively.

Figure 4

Results of the sensitivity analysis (SRC) for 17 model variables for struvite recovery from digested sludge at the demonstration plant of Leuven, Belgium (a) and for struvite recovery from centrate (b). R² of the analyses was 0.85 and 0.79, respectively, values that according to Sin et al. (2011) provide enough certainty about a linear approximation for the uncertainty of this model. Negative SRCs mean that increases in those values decrease the payback time; positive SRCs mean that increases in those values increase the payback time.

Figure 4

Results of the sensitivity analysis (SRC) for 17 model variables for struvite recovery from digested sludge at the demonstration plant of Leuven, Belgium (a) and for struvite recovery from centrate (b). R² of the analyses was 0.85 and 0.79, respectively, values that according to Sin et al. (2011) provide enough certainty about a linear approximation for the uncertainty of this model. Negative SRCs mean that increases in those values decrease the payback time; positive SRCs mean that increases in those values increase the payback time.

The square of the individual SRCs in Figure 4 is a measure of the percentage of the total variation in profitability that is explained by that variable. The profitability of DIG is most sensitive to changes in dewaterability of the treated sludge (45% of the total variability in cost), to the recovery percentage (20% of the variability) (Figure 4). To a smaller extent, incoming (6% of the total variation), MgCl2 price (6%), and operational man-hours (4%) cause uncertainty.

Incoming orthophosphate concentration is the primal source of variability of the financial model for CEN. It explains 70% of the variability. Recovery percentage is obvious too, but the variation limits for this variable were kept narrow between 55 and 75% expected performance in the CEN model.

Discussion on the cost–benefit

Expressed in cost per kilogram of P recovered and at the average ingoing concentration of 220 mg P as /L ingoing, the cost of struvite recovery from digested sludge is €4.40/kg P with current specs and for a discounted payback of 10 years, and between <€0 and €17.5/kg P recovered, according to the uncertainty analysis. For CEN with an ingoing concentration of 220 mg P as /L, recovery cost is €3.93/kg P recovered, but with a smaller uncertainty range of €2.54–5.55/kg P recovered. For CEN with an ingoing concentration of 110 mg P as /L, recovery cost rises sharply. This should be kept in mind when comparing the two techniques (CEN and DIG) for the same WWTP due to dilution of sludge waters after polymer addition.

Expressed in terms of people equivalents in Leuven, the currently estimated cost per inhabitant is around €0.32/year/people equivalent to have P recovered from digested sludge as struvite. Indicatively, adding this information to the comparative study by Egle (2014) suggests that the presented case of Leuven is a relatively cheap recovery technique.

It should be clear that this tentative subsidy curve is only based on one installation and only valid in the presented range of the curve. Although no investment costs for larger installations are available, it is believed that scale benefits, proportional to incoming flow, will decrease the subsidy curves for larger flows. Uncertainty around the curves is, however, not positively influenced by larger scale.

Discussion on struvite recovery opportunities

For Aquafin nv, which operates all municipal WWTPs in the northern part of Belgium, the technique of P-recovery as struvite can only be applied on around 20% of its total wastewater stream, because it operates enhanced biological P-removal on one-fifth of its stream. Recovery of P as struvite from digested sludge can only be achieved with a Bio-P installation equipped with a digester (possibly after transport) or any other sludge treatment provoking release of and from the cells (e.g. electroporation and thermal hydrolysis). From Marchi et al. (2015), it is concluded that maximum 20% of Ptotal entering the WWTP can be recovered. Therefore, the theoretical recoverable P via this technique accounts for less than 4% of Aquafin's total phosphorus in the influent. For special streams such as separately collected urine, in which absolute concentration is high and in which represents a very high proportion of Ptotal, this recovery technique has additional potential.

CONCLUSION

The full-scale installation for struvite recovery from digested wastewater sludge in Leuven, Belgium has been operational for over a year now. This experience allowed to propose a first tentative subsidy scheme for the recycling of as struvite. It should be considered whether a subsidy scheme, with the presented financial risk, could be the way forward at a moment of low yet volatile international P-price.

Although possible improvement in sludge dewaterability when recovering struvite from digested sludge has a positive economic amortization effect, it is at the same time the largest source of financial risk. With low recovery and mainly on the basis of operational benefits, the installation can be paid back within a reasonable time of 10 years if there is certainty about the possible dewaterability improvement being at the maximum presented in this study. This article shows that investment in struvite recovery from digested sludge is riskier, but potentially with a higher return than investment in struvite recovery from CEN.

Should a certain ecological role for this relatively clean product be recognized financially, it would support the presented P-recovery route.

ACKNOWLEDGEMENTS

We are grateful to VLAKWA, VITO, VIA-Flanders in action, NuReSys, and Akwadok for supporting this research and demonstration project.

REFERENCES

REFERENCES
Cordell
D.
Drangert
J.-O.
White
S.
2009
The story of phosphorus: global food security and food for thought
.
Global Environ. Change
19
,
292
305
.
Cornel
P.
Schaum
C.
2009
Phosphorus recovery from wastewaters: needs, technologies and costs
.
Water Sci. Technol.
59
(
6
),
1069
1076
.
Doyle
J. D.
Parsons
S. A.
2002
Struvite formation, control and recovery
.
Water Res.
36
,
3925
3940
.
Egle
L.
2014
(Comparative)
Technical, Environmental and Economic Assessment of Phosphorus Recycling Technologies from Wastewater
.
Institute for Water Quality, Resource and Waste Management. Vienna University of Technology, Karlsplatz 13/226-1, 1040 Vienna, Austria
.
European Commission
2013
Science for Environment Policy In-depth Report: Sustainable Phosphorus Use
.
Report produced by Science Communication Unit, University of the West of England, Bristol for the European Commission DG Environment, October 2013. http://ec.europa.eu/science-environment-policy
.
Hanhoun
M.
2011
Analysis and Modelling of Struvite Precipitation: Towards the Treatment of Industrial Waste-water Discharges
.
PhD Thesis
,
Université de Toulouse
.
Kruk
D. J.
Elektorowicz
M.
Oleszkiewicz
J. A.
2014
Struvite precipitation and phosphorus removal using magnesium sacrificial anode
.
Chemosphere
101
,
28
33
.
Lahav
O.
Telzhensky
M.
Zewuhn
A.
Gendel
Y.
Gerth
J.
Calmano
W.
Birnhack
L.
2013
Struvite recovery from municipal-wastewater sludge centrifuge supernatant using seawater NF concentrate as a cheap Mg(II) source
.
Sep. Purif. Technol.
108
,
103
110
.
Marchi
A.
Geerts
S.
Weemaes
M.
Schiettecatte
W.
Vanhoof
C. M.
2015
Full-scale phosphorus recovery from digested wastewater sludge in Belgium – part I: technical achievements and challenges
.
Water Sci. Technol. – Special Issue on Full Scale Investigations
71
(
4
),
487
494
.
Moerman
W. H. M.
de Danschutter
J.
de Ru
B.
Veltman
A.
Verwoert
G.
Reitsma
B. A. H.
van den Bulk
J.
2012
Phosphate recovery as nuisance control for struvite clogging sensitive wastewater or sludge
. In:
IWA Nutrient Removal and Recovery 2012, September 23–25
,
Harbin, China
.
Muster
T. H.
Douglas
G. B.
Sherman
N.
Seeber
A.
Wright
N.
Guezuekara
Y.
2013
Towards effective phosphorus recycling from wastewater: quantity and quality
.
Chemosphere
91
,
676
684
.
Saltelli
A.
Ratto
M.
Tarantola
S.
Campolongo
F.
2006
Sensitivity analysis practices: strategies for model-based inference
.
Reliab. Eng. Syst. Saf.
91
,
1109e1125
.
STOWA
2012
Struviet productie door middel van het Airprex proces
.
Rapport 2012–27. STOWA, Amersfoort, The Netherlands
.