PII: S0043-1354(00)00236-0

Wat. Res. Vol. 35, No. 1, pp. 151159, 2001

7 2000 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0043-1354/00/$ - see front matter

www.elsevier.com/locate/watres

CONTROLLED STRUVITE CRYSTALLISATION FOR

REMOVING PHOSPHORUS FROM ANAEROBIC DIGESTER
SIDESTREAMS
ELISABETH V. MUNCH* and KEITH BARRM
Brisbane Water, 240 Donaldson Rd, Rocklea, Brisbane, QLD 4106, Australia
(First received 20 January 2000; accepted in revised form 3 April 2000)
AbstractEnhanced biological phosphorus removal wastewater treatment plants that use anaerobic
digesters for sludge treatment, have high phosphorus concentrations in the sidestreams from their
sludge dewatering equipment. To remove phosphorus from such sidestreams controlled struvite
crystallisation can be used. Struvite (or MAP) is a naturally occurring crystal of magnesium,
ammonium and phosphate. We present operational results obtained with a continuously operated pilotscale MAP reactor. The pilot-scale reactor (143 l) was an air agitated column reactor with a reaction
and a settling zone, based on the Phosnix process of Unitika Ltd., Japan. The inuent to the MAP
reactor was centrate from the centrifuge that dewaters anaerobically digested sludge at the Oxley Creek
wastewater treatment plant in Brisbane. We used a 60% magnesium hydroxide slurry to add the
required magnesium to the process and to obtain the alkaline pH value required. The pilot-scale MAP
process achieved an ortho-P removal ratio of 94% from an average inuent ortho-P concentration of
61 mg/l. The reactor was operated at a pH of around 8.5. Insucient dosing of magnesium reduced the
P removal performance. There was no inuence of the hydraulic residence time on the process in the
range of 18 h. The dry MAP product had cadmium, lead and mercury concentrations well below the
legal limits for fertilisers in Queensland, Australia and can be reused as a valuable slow-release
fertiliser. 7 2000 Elsevier Science Ltd. All rights reserved
Key wordsstruvite, MAP, crystallisation, sidestream, return liquor, phosphorus removal

INTRODUCTION

Brisbane City Council strives to reduce the amount

of nutrients (nitrogen and phosphorus) entering the
Brisbane River and Moreton Bay from the 300 Ml/d
of treated euent discharged by Brisbane's ten
wastewater treatment plants (Abal et al., 1998).
Enhanced biological phosphorus removal (EBPR) is
the preferred technology to remove phosphorus from
wastewater in many instances. For EBPR plants that
use anaerobic digesters for sludge treatment, the diculty arises of how to deal with the sidestreams that
are high in phosphorus and contribute a signicant
phosphorus load to the plant.
Sidestreams, or ``return liquors'', are those ows
of wastewater that originate from digested sludge
dewatering equipment such as centrifuges or belt lter presses. The centrate, or ltrate from these facilities is typically very high in ammonia and
phosphorus because these compounds are released
in the anaerobic digestion process. Ammonia is produced from the break-down of proteins and bac*Author to whom all correspondence should be addressed.
Tel.: +61-7-3403-1523; fax: +61-7-3403-2603; e-mail:
ppoebw@brisbane.qld.gov.au
151

teria. The phosphorus that is incorporated into

bacterial biomass in the EBPR process is re-solubilised under the conditions of an anaerobic digester.
If the liquid from the sludge digestion process is
returned to the head of the plant, most of the phosphorus is only recirculated and not removed.
Controlled struvite crystallisation is a technology
that can be applied to remove phosphorus from
sidestreams, enabling EBPR plants to achieve very
low levels of euent P concentration. The three
molecules magnesium, ammonium and phosphate
are the ingredients for the struvite, or MAP, crystals. The process of MAP crystallisation removes
the phosphate and to a lesser extent ammonia from
the wastewater, and together with magnesium binds
them in a solid form, the struvite crystals. The recommended position of the MAP process within the
overall plant owsheet is shown in Fig. 1.
It has been estimated that the resource base for
phosphate rock which can economically be
exploited, will last, at best, little more than 100
years and could be depleted in as little as 50 years
(Driver et al., 1999). Human sewage and animal
manures and slurries are two major sources of
phosphorus from which phosphorus could be recycled. MAP is a valuable slow-release fertiliser and

152

Elisabeth V. Munch and Keith Barr

the MAP process hence allows recycling of the limited resource phosphorus.
The aim of this research project was to demonstrate the MAP process so that a process according
to Fig. 1(B) can be realised for example at the
Oxley Creek WWTP. We also investigated the
properties of the MAP product generated in order
to assess its value as a fertiliser.
LITERATURE REVIEW ON MAP PROCESS

Struvite has the following properties (Chirmuley,

1994):
. Struvite is magnesium ammonium phosphate hexahydrate
with
the
chemical
formula
MgNH4PO4.6H2O.
. White crystalline powder in its pure form but can
also occur either as large single crystals, very
small crystals, large curds or a gelatinous mass.
. Specic gravity is 1.7.
. Highly soluble at acidic pH and highly insoluble
at alkaline pH.
MINTEQA2 is an aquatic chemistry equilibrium
model (Buchanan et al., 1994). This model accepts
inputs of magnesium, ammonium, and phosphate
concentrations, pH, temperature, and ionic strength
of a solution and computes the degree of saturation
with respect to struvite. Key results from this computation are that:
. As the pH increases, the concentration of orthoP increases, while the concentrations of Mg2+
and NH4+ decrease. This phenomenon produces
an optimum pH for struvite formation.
. The pH range for which struvite can precipitate
was identied as approximately 711.
. The pH of minimum solubility for struvite is
around 9.0 for typical conditions in wastewater
treatment; minimum solubility for struvite implies
maximal potential for its formation.

Unintentional struvite formation has long been

known to be a nuisance in wastewater treatment
plants. Hard and thick struvite crusts can block up
valves, pipes, centrifuge bowls and pumps (for factors impacting on unintentional struvite formation
see for example Chirmuley, 1994).
Intentional struvite crystallisation, on the other
hand, can be harnessed to remove phosphorus from
wastewater. The process most widely described in
the relevant literature is the Phosnix process developed by Unitika Ltd., Japan (Brett et al., 1997;
Katsuura, 1998). This process has been operated at
full scale at several sites in Japan since 1987, with
plants ranging in capacity from 100 to 500 kl/d and
producing 100500 kg/d of MAP. A similar process
is used by Kurita Water Industries (Ishiduka et al.,
1998).
Despite the full-scale experience of these two
Japanese companies, very little scientic knowledge
has been published and little detailed information
about this process exists outside of Japan. It is the
purpose of this paper to describe the MAP process
in detail to a wider audience.
MAP process inuent
Most of the MAP processes described in the literature use sidestreams generated from anaerobic
digesters as the inuent. The supernatant from
digested sludge dewatering devices is remarkably
similar throughout the world (Table 1). Whilst the
ammonia content is generally in the range of 700
800 mg/l, the P content of the supernatant can vary
widely, depending on:
. Whether the activated sludge process incorporates enhanced biological phosphorus removal or
not.
. Whether MAP formation occurs in the anaerobic
digester itself, depending on the hardness of the
water (magnesium content).

Fig. 1. Process owsheet of EBPR wastewater treatment plant without sidestream treatment (A) and
with sidestream treatment by MAP process (B).

Struvite crystallisation for P removal

Operating parameters for MAP process

The MAP process depends mainly on two operating parameters:
1. Ratio of magnesium, ammonium and phosphate
in the reactor (in most cases, magnesium has to
be added to remove all phosphorus).
2. Reactor pH value (alkaline).
The two main types of magnesium used in the
MAP process are magnesium hydroxide (Mg(OH)2)
and magnesium chloride (MgCl2). The advantage of
using MgCl2 is that it dissociates faster than
Mg(OH)2 therefore resulting in shorter reaction
times. Magnesium hydroxide, however, is generally
cheaper, and has the advantage of also raising the
pH.
The required pH increase can be achieved in several ways. Research performed by Fujimoto et al.
(1991) indicated that addition of NaOH was more
eective than addition of lime or Mg(OH)2. Aeration itself can achieve the desired pH increase but it
requires longer residence times (Battistoni et al.,
1998).
Katsuura (1998) investigated the eect of ammonium concentration, Mg:P ratio and pH on the
P removal performance. He reported that:
. As the ammonia concentration increased from
100 to 500 mg/l, so did the P removal ratio.
. The P removal ratio increased sharply with the
pH being increased from 8.0 to 9.0, for all
ammonia concentrations tested. The increase in P
removal ratio was sharper, the higher the ammonia concentration.
. The P removal ratio increased with an increase in
Mg:P ratio. This increase was much more pronounced at low pH values (pH of 8.0) compared
to high pH values (pH of 9.0).
. At a pH of 9.0, the P removal ratio did not
increase any further even if the Mg:P ratio was
increased beyond about 1.3.
The value of struvite as a fertiliser
The fertiliser properties of struvite were already
demonstrated in the 1960s in Germany and the US
(as summarised by Schuiling and Andrade, 1998).
Struvite can be produced by chemical processes

153

from the raw materials magnesium oxide, phosphoric acid and ammonia, but this is not commonly
done as the costs are usually prohibitive. Bridger et
al. (1961) give a detailed description of the value of
MAP as a slow release fertiliser. MAP has been
used commercially for container plants and is also
appropriate for use on turf, tree seedlings, ornamentals, vegetables, and ower boards. The advantages of MAP as a fertiliser are:
. Because nutrients are released at a slower rate
throughout the season compared to soluble fertilisers, plants can take up most of the nutrients
without waste by leaching.
. Less frequent application is required.
. Fertiliser burn is not a problem even at high application rates.
Unitika Ltd. of Japan has been able to operate fullscale MAP reactors at a prot by successfully marketing their MAP product. The product, under the
name of ``Green MAP II'', has been promoted in
Japan as an ``environmentally friendly special fertiliser for forest regeneration of articially created
slopes'' as well as a ``plant friendly long lasting type
gardening fertiliser''.

PROCESS DESIGN

The pilot-scale MAP reactor used in this study

consists of a plexi-glass reactor, an inuent pump,
an air compressor, a magnesium/alkali dosing
pump, a pH probe and an air owmeter (Fig. 2).
The total reactor volume is 143 l and the reactor
has a reaction zone and a settling zone. The internal
diameters of the reaction zone and the settling zone
are 300 mm and 600 mm, respectively. The total
liquid height is 1365 mm.
The pilot-scale MAP reactor was based on the
same process principles as the Phosnix process
developed by Unitika Ltd. Brett et al. (1997)
describe the process principles: ``The Phosnix process is based on an air agitated column reactor,
with ancillary chemical dosing equipment. Digester
euent is fed into the base of the tower where it is
mixed with a magnesium solution to provide a
Mg:P molar ratio of 1:1.3. Alkali is dosed to ensure
a pH range of 8.59. Air that is sparged into the

Table 1. Characteristics of anaerobically digested sludge dewatering liquid (all in mg/l except pH); N/S=not specied
Ortho-P

TP

SS

NH4-N

Soluble Mg

pH

43.9

N/S

140

500

26

7.93

110
110250

270
11002200

750
820210

N/S
1322

207

N/S

N/S

756

293

N/S

N/S

775

16.2

80
7028

Plant location and reference

Seibu WWTP, Fukuoka City, Japan; only one train is operated as

EBPR (Abe, 1995)
N/S
Slough EBPR plant, UK (Williams, 1998)
7.6520.03 Oxley Creek WWTP (own data from 11 Nov. 98 to 11 Dec. 98;
average295% C.I.; N = 27)
7.37.5
Tobu sewage treatment plant, Fukuoka City, Japan (Fujimoto et
al., 1991)
N/S
Ballarat South EBPR plant, Australia (Webb et al., 1995)

154

Elisabeth V. Munch and Keith Barr

base of the column provides the agitation required

for complete mixing and suspension of the growing
particles.
The crystals grow in size until they sink to the
base of the tower where they are periodically
removed. The recovered product is then dewatered
from the small amount of solution which is lost
when the crystals are removed, either by a lter bag
system or natural drying.''
We used a magnesium hydroxide slurry (brand
name MHS-60, Orica, Australia) for both increase
of pH and magnesium addition. MHS-60 contains
55% w/w Mg(OH)2 and 42% w/w water, has a
bulk density of 1.5 t/m3 at 60% solids and a buer
pH of 9.5. Advantages of using this product in fullscale are that it does not have to be stirred and is

non-hazardous. The disadvantage of using MHS-60

in the MAP process is that the pH value and molar
ratio of Mg:P cannot be controlled independently
of each other.
MATERIALS AND METHODS

The MAP reactor inuent was centrate from the sludge

dewatering centrifuge at the Oxley Creek WWTP in Brisbane, Australia (Table 2). This plant is a conventional
BOD-removal activated sludge plant with an average dry
weather owrate of 55 Ml/d. The sludge that is being
dewatered in the centrifuge is anaerobically digested primary and thickened waste activated sludge. The centrate
was stored for several days in a wet well at the centrifuge,
from where it was continuously pumped to the MAP reactor. This centrate storage allowed some solids settling to
occur, resulting in lower suspended solids and TP concen-

Fig. 2. Diagram of pilot-scale MAP reactor with ancillary equipment.

Struvite crystallisation for P removal

155

Table 2. MAP reactor inuent characteristics (in mg/l except pH)

Average295% C.I.
Number of samples

pH

SS

NH4-N

PO4-P

TP

Sol. Mg

COD

Alkalinity as CaCO3

7.7120.02
22

160240
18

790230
26

6125
26

7824
23

1122
24

5002100
5

28002200
5

trations of the centrate compared to the results presented

in Table 1.
The centrate was continuously pumped from the centrifuge wet well to the MAP reactor for 5 days of the week.
During the weekend, the reactor was aerated but not fed.
The inuent owrate was varied over the course of the
reactor operation (owrate setpoints and measured values
are shown in Fig. 3; the inuent owrate was measured
using a measuring cylinder at the reactor outlet). As can
be seen from the graph, the actual owrate was dicult to
keep exactly at the setpoint, probably due to level variations in the centrate wet well.
The air owrate was adjusted to a level where the MAP
particles remained in suspension but no air bubbles
appeared at the top of the outer cylinder (the settling
zone). Foam formation on the top of the inner cylinder
also became excessive when the air owrate was too high.
The air owrate used was about 7 l/min.
At start-up, the reactor was seeded with 500 g of
crushed and sieved (1 mm) struvite from a deposit
removed from a pipe. Diculties with dosing of the MHS60 were experienced initially because the MHS-60 tended
to solidify and clog the small plastic tubes used for dosing.
As of 12 August 1999, a more reliable system for the
MHS-60 dosing was found. The MHS-60 was diluted
twenty times with tap water and continuously stirred to
prevent settling of solids. The stirring was performed with
an overhead stirrer in a 2 l plastic beaker. In the full-scale
application, such problems are not expected since larger
diameter pipes would be used.
The required MHS-60 dose rate was calculated from the
stoichiometric magnesium requirements to remove all
ortho-P from the centrate. The amount of MHS-60 solution required to remove a certain P load can be calculated as:
_ orthoP
QMHS60 M

MWMgOH 2
EF
1:5ton=kl 0:55
MWP

.
where QMHS60 is the owrate of MHS-60 required; MorthoP
is the P load to be removed; EF is the excess factor; a
Mg:P excess factor of 1.3 was chosen here as recommended by Katsuura (1998); MW is the molecular
weight (58.3 g/mol for Mg(OH)2 and 31 g/mol for P).
The factor 1.5 t/kl is the density of the MHS-60 slurry
and the factor 0.55 is the magnesium hydroxide content in
the MHS-60 slurry.
The MHS-60 dose rate setpoint chosen here was 12 ml/

h of pure MHS-60 solution, which would be sucient to

remove 61 mg/l of ortho-P at an inuent owrate of 65 l/
h. The resulting pH value was usually around 8.5, which is
suciently high for MAP crystallisation.
For the period where very high inuent owrates were
trialled (Fig. 3), the MHS-60 dose rate was increased to
match the higher P load to be removed. It was found that
in order to keep the pH value suciently high, the MHS60 dose rate had to be increased beyond the stoichiometric
requirements to a value of 21 ml/h (an average pH value
of 8.3 resulted during that period).
For the purpose of these experiments, overdosing with
magnesium was not considered a problem, because the
main aim was to demonstrate the highest possible P
removal from the centrate. In the full-scale process the
magnesium dose rate could be controlled using on-line
euent ortho-P measurements in order to avoid struvite
formation in the digester due to overdosing of magnesium.
It might also be more cost-eective to use a dierent alkali
so that the magnesium dose rate can be optimised independently from the pH value.
A fraction of the MAP product present in the reactor
was removed every ve days of operation by turning the
air o, letting the MAP particles settle and briey opening
a valve at the bottom of the reactor. The settling period
required only several minutes. The removed MAP slurry
dried easily in the air resulting in the nal dry MAP product.
Inuent and euent grab samples were taken at steady
state (after 23 hydraulic residence times). We found that
the euent samples for soluble compounds had to be preserved with a special preservation method. This had to be
done to avoid further crystallisation in the sample vials
while they were frozen which would have resulted in lower
ortho-P and ammonia-N concentrations. The euent
samples were therefore rst ltered through 0.45 mm lters
(to remove ne MAP particles) and then acidied (23
drops of concentrated sulfuric acid added to reduce pH to
around 6.5). At this pH value no further signicant MAP
crystallisation can occur in the sample vials which were
stored frozen until analysis.
Total P was analysed by ICP (inductively coupled
plasma) after digestion, soluble magnesium by ICP after
ltration, and ortho-P and ammonia were analysed by
FIA (ow injection analysis).
The dry, solid MAP product was analysed for TKN
using the standard digestion and distillation method
(APHA et al., 1992) with 0.1 g of solid sample. TKN
could be approximated to be total N, since there is unlikely to be any NOx in the solid MAP sample. The total
phosphorus content was analysed using digestion in aqua
regia and analysis by ICP. Aqua regia is a 3:1 volumetric
mixture of HCl and HNO3. We found that the initial
digestion method employed had not been strong enough
to digest the MAP solids fully, resulting in measured P
concentrations in the solid MAP samples that were an
order of magnitude too low.

RESULTS AND DISCUSSION

Fig. 3. Inuent owrates to the MAP reactor.

The operational results are shown in Figs 46

and additional average euent concentrations are
given in Table 3. The MAP reactor achieved excel-

156

Elisabeth V. Munch and Keith Barr

Table 3. Average euent concentrations from MAP reactor (all in mg/l)

Average295% C.I.
Number of samples

SS

NH4-N

TP

Organic N

BOD

COD

Alkalinity as CaCO3

290250
27

740230
37

2224
36

1325
4

50220
4

520270
5

28002200
5

lent P removal from the centrate sidestream (Fig. 4).

The average euent ortho-P concentration from 12
August 99 onwards (i.e. after the reliable MHS-60
dosing had been implemented) was 4 mg/l, corresponding to an ortho-P removal ratio of 94%.
The euent TP concentration after 12 August 99
was about 13 mg/l. Particulate phosphorus is not
available for the MAP crystallisation process.
Hence, only the ortho-P is removed in the MAP
reactor, unless the reactor is operated also for suspended solids removal. Here, the SS concentration
in the euent was higher than in the inuent
(Tables 2 and 3). Fine MAP particles suspended in
the euent also increase the measured euent TP
concentration.
The MAP process was operated at excessively
high MHS-60 dose rates and also without MHS-60
dosing to investigate the eect of magnesium concentration on the P removal performance (Fig. 5).

Fig. 4. Inuent and euent TP and ortho-P concentrations of MAP reactor.

The euent magnesium concentration was used

here as an indicator for the amount of magnesium
added to the reactor, because of diculties in determining the exact MHS-60 dose rate during some
time periods.
Figure 5 clearly shows that below a certain magnesium concentration, the P removal performance
rapidly decreases. This is partly, but probably not
only, due to the correlation between magnesium addition and pH increase. Because the pH was not
controlled separately from the magnesium dosing,
these two eects were impossible to delineate for
this study. For optimal P removal performance of
the process, MHS-60 should be dosed such that
there is at least 20 mg/l soluble magnesium in the
euent.
Figure 6 shows that in the range of about 18 h,
the hydraulic residence time has no eect on the
euent ortho-P concentration (at suciently high
magnesium dose rates). This is important as it
means small reactor volumes are sucient (12 h
HRT).
The MAP slurry usually occupied about 12 l of
reactor volume after settling. About every 5 days of
reactor operation, approximately 24 l of settled
MAP slurry were removed through the bottom
valve. The settled MAP slurry had a solids content
of about 16% which dried to the nal solids concentration (Table 4) in a few days. The drying properties of the MAP slurry are extremely favourable
and enable fast drying. The nal product has a ne,
powdery, non-odorous white/grey appearance.
The pilot-scale MAP reactor produced about
320 g of dry MAP product per day of reactor operation. This is in agreement with the value determined by stoichiometric calculations: if the average
inuent ortho-P concentration of 61 mg/l was
reduced to about 5 mg/l in the euent and the
average inuent owrate was about 42 l/h (Fig. 3),
then the expected MAP production is 327 g/d.
Some MAP was continuously lost as ne particles
in the euent (there was a thin lm of MAP in the
euent launder).
The ammonia-N removal ratio of the MAP process was about 6%, which was in agreement with
Table 4. Composition of MAP produced, theoretical and measured
weight fractions
Mg

Fig. 5. Eect of Mg concentration on euent ortho-P concentration.

Theoretical value from chemical formula (%) 9.9 5.7 12.6

Measured value, average of 5 samples (%)
9.1 5.1 12.4

H2O
44
39

Struvite crystallisation for P removal

157

Fig. 6. The euent ortho-P concentration is independent of the HRT in the range tested (for euent
soluble Mg higher than 20 mg/l).

the theoretically expected removal ratio based on

the N:P ratio of the inuent.
PROPERTIES OF MAP PRODUCT

Physical properties of MAP product

A sample was taken from the reaction zone of
the MAP reactor and analysed under the microscope (Fig. 7). The typical trapezoidal shape of
the MAP crystals is visible. Very few bacterial
ocs or other microorganisms existed in the reactor (this is also evident from the low VSS to SS
ratio of 19% for a sample taken from the wellmixed reaction zone during aeration, at a TS

content of about 5.4%). The small clumps of organic matter that were visible under the microscope were probably introduced through the
inuent.
The MAP product was analysed for struvite
using polycrystalline X-ray diraction analysis
with the X-ray Analysis Facility at Queensland
University of Technology, Brisbane. The majority
of the sample was struvite (>90 wt. %).
Another compound was present in minor to
trace quantities, which was possibly quartz.
To investigate the typical size of the MAP crystals we used a Malvern Mastersizer E (Department
of Chemical Engineering, The University of

Fig. 7. Photomicrograph of liquor from reaction zone of MAP reactor (100 times magnication).

158

Elisabeth V. Munch and Keith Barr

Queensland). Using a laser beam passing through a

suspension of the particles, the size distribution of
the suspended particles was determined. A sample
of settled MAP slurry taken on 13/8/99 was wet
sieved and more than 90% of the sample passed
through the 355 mm sieve. The smaller than 355 mm
fraction was then sized with the Malvern instrument. The volume median size of the sample was
110 mm and the 1090 vol% range was 28215 mm.
The distribution was found to have a slightly bimodal shape.
The size of the struvite crystals produced from
the pilot-scale MAP reactor was signicantly smaller than what has been achieved with full-scale
MAP reactors in Japan. The process used by Kurita
Water Industries is generally operated to achieve
grain sizes of 2.03.8 mm (Abe, 1995). The MAP
grain size is aected by the inuent phosphorus
concentration and MAP particle retention time in
the reactor. Abe (1995) reported that the MAP
growth rate was 0.173 mm/d with a high inuent P
concentration (>200 mg/l) and only 0.061 mm/d
for a low inuent P concentration (34100 mg/l).
Abe (1995) also reported that for high inuent
phosphorus concentrations, MAP was generated instantaneously near the inlet of the injection port and
did not grow larger. Therefore, a circulation line can
become necessary for dilution to obtain an appropriate phosphorus concentration (about 90 mg/l) in the
reactor. It was not part of our study to optimise the
operating conditions for larger MAP crystals.
Fertiliser qualities of MAP product generated
For a potential buyer of MAP fertiliser, the
nitrogen and phosphorus content of the product is
of importance. The magnesium content may also be
valuable, if the application is for magnesium decient soils (for example in forestry). The elemental
composition of the MAP product generated at the
Oxley Creek WWTP is shown in Table 4. The
measured composition of the MAP product is very
close to the theoretical values.
The main legal requirement for selling the MAP
product as a fertiliser in Queensland is that the cadmium, lead and mercury levels in the product are
lower than specied in the government regulation
(Table 5). These concentrations measured in the
MAP produced at the Oxley Creek WWTP were
mostly below the analytical detection limits and
well below the legal limits (Table 5). This conrms

the Japanese experience that the MAP product produced from wastewater by controlled MAP crystallisation contains very few contaminants.
BASIC ECONOMIC ANALYSIS OF MAP PROCESS

The capital costs for the MAP process are determined by the size of the MAP reactor required and
the costs for the associated equipment (e.g. pumps
and instrumentation). The total reactor volume for
the full-scale reactor is determined by multiplying
the inuent owrate with the hydraulic residence
time. From the results achieved with the pilot-scale
MAP reactor it appears that the hydraulic residence
time could be as short as 12 h (Fig. 6).
The inuent owrate equals the sidestream owrate to be treated. For example in the case of the
Oxley Creek WWTP which has an average dry
weather owrate of about 55 Ml/d, the sidestream
owrate is about 5 l/s (centrate owrate from the
sludge dewatering centrifuge). Therefore, the
required MAP reactor volume is about 27 kl.
The operational cost eectiveness of the MAP process compared to alternative sidestream treatment
processes depends mainly on the following factors:
. The cost of magnesium to be added to the process (MHS-60 is sold at 500 A$/ton (330 US$/
ton), or less depending on quantities used)
. The cost at which the MAP product can be sold,
which may also depend on the amount of MAP
produced. For an indicative price note that Unitika Ltd. sell their MAP product at about Y29
per kg (276 US$ per ton). Our own preliminary
market research indicated that in Australia the
price achievable may be in the range of 300500
A$/ton (198330 US$/t). The most likely use for
our MAP product appears to be as a component
of a potting mix or a commodity fertiliser to
replace other sources of the desired elements.
. The ortho-P concentration in the inuent to the
MAP reactor, since higher concentrations result
in more MAP production.
. The ortho-P removal performance in the MAP
reactor, since higher ortho-P removal implies
more MAP production.
Depending on these factors, the MAP process will
operate at a prot or a loss for a given site. For the
Oxley Creek WWTP, the operating prot based on
estimated magnesium and MAP prices, was calcu-

Table 5. Allowable limits for key contaminants in fertiliser (Queensland regulations) and results obtained for MAP produced in this study
Compound

Cadmium
Lead
Mercury

Legal limit

350
100
5

Unit

mg/kg phosphorus
mg/kg product
mg/kg product

Results (sample no.)

1

<4
<5
0.2

5.5
<5
< 0.1

<4
6.9
< 0.1

<4
5.2
< 0.1

Struvite crystallisation for P removal

lated to range from minus A$ 13,000 per year

(worst-case scenario) to plus A$ 149,000 per year
(best-case scenario). For the best-case scenario it
was assumed that the plant was upgraded to an
EBPR plant and the centrate ortho-P concentration
was consequently assumed to double.
The MAP process can also be used for nitrogen
removal from sidestreams. To achieve signicant
nitrogen removal, phosphorus has to be added in a
stoichiometric quantity. If the MAP product can be
sold at a favourable price, it may become economical
to operate the MAP process for N removal. The P
could be added in the form of technical grade phosphoric acid (H3PO4) or in the form of commercially
available fertilisers such as Superphosphate
(Ca3(PO4)2).
CONCLUSIONS

The MAP process (struvite crystallisation process) is a suitable technology to remove and reuse
phosphorus from wastewater treatment sidestreams.
It is a process ideally suited for enhanced biological
phosphorus removal plants with anaerobic digesters. The process was demonstrated and investigated
using a 143 l air agitated column reactor based on
the Phosnix process by Unitika, Japan. The inuent
to the MAP reactor was centrate from the sludge
dewatering centrifuge at the Oxley Creek WWTP in
Brisbane, Australia. Our study concluded that:
. The pilot-scale MAP process achieved an ortho-P
removal of 94% from an inuent ortho-P concentration of 61 mg/l (i.e. euent ortho-P concentration of 4 mg/l).
. The alkali and magnesium source used was a
60% magnesium hydroxide slurry (MHS-60). The
reactor was operated at a pH of around 8.5.
Insucient dosing of magnesium reduced the P
removal performance.
. The MAP process can be realised in a one-vessel
operation at short hydraulic residence times (1
2 h). For inuent owrates in the range of 20
120 l/h, the P removal performance was independent of the HRT.
. About 320 g of dry MAP product was produced
per day, at an average inuent owrate of 42 l/h.
. The MAP product fullled all the requirements
for its reuse as a slow-release fertiliser in Queensland, with cadmium, lead and mercury concentrations being well below the legal limits.
. The composition of the MAP product was 12.4%
P, 9.1% Mg, 5.1% N, and 39% crystalline water,
which was in accordance with the theoretically
expected values.
The MAP process enables enhanced biological
phosphorus removal plants that use anaerobic
digesters to achieve very low levels of euent
ortho-P concentrations. It also allows true recycling
of the limited resource phosphorus. How many

159

other domestic wastewater treatment processes have

the potential to generate a net prot by producing a
valuable by-product?
AcknowledgementsBrisbane City Council funded this
work. We would like to thank the operators at the Oxley
Creek WWTP. In addition we are grateful to our colleagues Julie McLellan and Mary Corbett, as well as
Justin Doyle (University of Queensland) and Mr Tamura
(Unitika Ltd.) for their friendly assistance and expert
advice.
REFERENCES

APHA AWWA WPCF (1992) Standard Methods for the

Examination of Water and Wastewater. Port City Press,
Baltimore.
Abal E. G., Holloway K. M. and Dennison W. C. (1998)
Interim Stage 2 Scientic Report. Brisbane River &
Moreton Bay Wastewater Management Study, Brisbane.
Abe S. (1995) Phosphate removal from dewatering ltrate
by MAP process at Seibu treatment plant in Fukuoka
City. Sewage Works in Japan, 5964.
Battistoni P., Pavan P., Cecchi F. and Mata-Alvarez J.
(1998) Phosphate removal in real anaerobic supernatants. Modelling and performance of a uidized bed
reactor. Wat. Sci. Tech. 38(1), 275283.
Brett S., Guy J., Morse G. K. and Lester J. N. (1997)
Phosphorus Removal and Recovery Technologies. SP
Publications Division Selper Limited, London.
Bridger G. L., Salutsky M. L. and Starostka R. W. (1961)
Metal ammonium phosphates as fertilisers. In 140th
meeting of the American Chemical Society, Chicago, Illinois.
Buchanan J. R., Mote C. R. and Robinson R. B. (1994)
Thermodynamics of struvite formation. Transactions of
the ASAE (American Society of Agricultural Engineers)
37(2), 617621.
Chirmuley D. G. (1994) Struvite precipitation in WWTPs:
Causes and solutions. Water (Journal of the Australian
Water Association) December, 2123.
Driver J., Lijmbach D. and Steen I. (1999) Why recover
phosphorus for recycling, and how? Environmental Technology 20(7), 651662.
Fujimoto N., Mizuochi T. and Togami Y. (1991) Phosphorus xation in the sludge treatment system of a biological phosphorus removal process. Wat. Sci. Tech. 23,
635640.
Ishiduka S., Sato S., Shibata M. and Togami Y. (1998)
Recovery of uoride and phosphate from wastewater.
In Symposium of the Society of Chemical Engineers in
Japan.
Katsuura H. (1998) Phosphate recovery from sewage by
granule forming process (full scale struvite recovery
from a sewage works at Shimane Prefecture, Japan). In
International conference on phosphorus recovery from
sewage and animal waste, Warwick University, UK.
Schuiling R. D. and Andrade A. (1998) Recovery of struvite from calf manure. In International conference on
phosphorus recovery from sewage and animal waste, Warwick University, UK.
Webb K. M., Bhargava S. K., Fraci A. J., Priestley A. J.,
Booker N. A. and Conney E. (1995) Struvite
(MgNH4PO4.6H2O) precipitation: potential for nutrient
removal and re-use from wastewaters. Chemistry in Australia October, 4244.
Williams S. (1998) Adapting strategies for avoiding struvite build-up in plants and transforming them into strategies for recovery. In International conference on
phosphorus recovery from sewage and animal waste, Warwick University, UK.