ARTICLE IN PRESS

Available at www.sciencedirect.com

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 1 1 4 3 – 1 1 5 4

0043-1354/$ - see frodoi:10.1016/j.watres

�Corresponding aE-mail addresses:

journal homepage: www.elsevier.com/locate/watres

TM

Sorption of phosphorous to Filtralite-P —The effect ofdifferent scales

Kinga Adama,�, Anne Kristine Søvikb, Tore Krogstadc

aDepartment of Mathematical Sciences and Technology, Norwegian University of Life Sciences, P.O. Box 5003, 1432 As, NorwaybCentre for Soil and Environmental Research (Bioforsk), Fredrik A. Dahls vei 20, 1432 As, NorwaycDepartment of Plant and Environmental Sciences, Norwegian University of Life Sciences, P.O. Box 5003, 1432 As, Norway

a r t i c l e i n f o

Article history:

Received 26 April 2005

Received in revised form

7 January 2006

Accepted 11 January 2006

Keywords:

Constructed wetland

Filtralite-PTM

P fractionation

P sorption

SEM

nt matter & 2006 Elsevie.2006.01.009

uthor. Tel.: +47 97778106kinga.adam@umb.no (K.

A B S T R A C T

Sorption of P to the filter material Filtralite-PTM was examined at a small, medium and large

scale. In the small- and meso-scale laboratory models, the sorbed amount of total

phosphorus (P) was heterogeneously distributed with more P sorbed in the inlet zone and

the bottom layers. The full-scale system had, on the other hand, the highest sorbed

concentration in the outlet region. The overall P sorption capacity of the material was 8030,

4990 and 521 mg P kg�1 Filtralite-PTM for Box 1, Box 2 and meso scale, respectively. This

equals 4.4, 2.8 and 0.29 kg P m�3 material, respectively. However, the maximum sorption

capacities found were 2500, 3887 and 4500 mg P kg�1 Filtralite-PTM for the two small-scale

box systems and the meso-scale container, respectively. In the full-scale system the overall

P sorption capacity of the material was 52 mg P kg�1 Filtralite-PTM (0.029 kg P m�3 Filtralite-

PTM) with a maximum sorbed amount of P of 249 mg P kg�1. Results from both the small-

and meso-scale system show that when a constructed wetland (CW) is saturated, i.e. when

the outlet concentration has reached its maximum allowed concentration of 1.0 mg P l�1,

only parts of the filter material will have reached the sorption capacity. Sequential

extractions of Filtralite-PTM showed that the loosely bound P, Ca–P and Al–P were the

primary P sorption pools both in the small-scale models and in the full-scale CW. However,

the proportion of these three fractions varied with time and change in pH. A white product

precipitated in the outlet zone of both the small-scale box models as well as the onsite CW.

The surface of these precipitation particles was identified by X-ray diffraction and SEM

method as CaCO3 and precipitated Ca- and Mg-phosphates.

& 2006 Elsevier Ltd. All rights reserved.

1. Introduction

Constructed wetlands (CWs) have shown their ability to

remove large amounts of phosphorous (P) (Kadlec and Knight,

1996; Jenssen et al., 2005; Vymazal, 2004) from wastewater

using special filter materials with an enhanced P sorption

capacity. The physico-chemical characteristics of different

filter material have been investigated by several scientists in

order to evaluate which parameters are the most important

r Ltd. All rights reserved.

; fax: +47 64948810.Adam), anne.sovik@biofo

for the P sorption (Drizo et al., 1997, 1999, 2002; Johansson,

1998; Arias et al., 2001; Arias and Brix 2005; Jang and Kang,

2001; Khadhraoui et al., 2002; Molle et al., 2005). Their

conclusions show that grain size distribution, pH, specific

surface area and the content of Al, Fe or Ca ions are

particularly important material properties for P sorption.

These filter substrates have to be available in large quantities

at low cost, but with long lasting P sorption capacity

(Johansson, 1998). As potential fertilizers, these filter materi-

rsk.no (A.K. Kristine Søvik), tore.krogstad@umb.no (T. Krogstad).

ARTICLE IN PRESS

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 1 1 4 3 – 1 1 5 41144

als must remove P from wastewater in such a way that the P

may be recovered and available for plant uptake. Kvarnstrom

et al. (2004) showed that in Sweden 15% of the commercial

fertilizer P could be replaced by P from such filter substrates.

Filtralite-PTM, the new generation of Norwegian produced

light weight aggregate (LWA) is especially developed for P

sorption. Filtralite-PTM is an illite based clay mineral with a

high pH and high Ca and Mg content. Previous batch

experiments have shown high P sorption capacity of the

material up to 12 g P kg�1 Filtralite-PTM (6.6 kg P m�3 Filtralite-

PTM) (Jenssen and Krogstad, 2003). Drizo et al. (2002) and

Adam et al. (2005) showed that batch experiments can lead to

misinterpretation and overestimation of the P retention

capacity of the material. Cheung and Venkitachalam (2000)

suggested that the maximum sorption capacity of the

materials estimated by Langmuir isotherms may not be

accurate but it could still be useful in comparing alternative

materials. Drizo et al. (2002) also recommended that batch

experiments should be coupled with long-term investigations

of the material’s performance. The question is whether these

long-term laboratory experiments may mimic the processes

of full-scale CWs? Can we use them as representative models

of the P removal processes of full-scale systems?

In this paper experiments were conducted to determine the

long-term P sorption capacity of Filtralite-PTM in small- and

meso-scale laboratory filter systems. The results were com-

pared with data from existing full-scale CW systems. The

spatial distribution of sorbed P was measured at all scales.

Fractionation of P from the small- and the full-scale systems

was also conducted to find potential patterns and pools of P

removal. The mineralogical composition of Filtralite-PTM

before and after exposure to P (small-scale system) as well

as precipitates from the small- and full-scale systems was

examined by scanning electron microscope (SEM).

2. Material and methods

2.1. Filter medium

A filter material called Filtralite-PTM has been used in the

experiments at all three scales. The material is a commer-

cially available expanded clay product with a high pH (410)

and a high Ca and Mg content. It has a grain size in the range

0–4 mm, effective porosity of 40% (Optiroc, 2003) and particle

porosity of 68% (Suliman et al., 2005a) with a saturated

hydraulic conductivity of 100 m day�1. The uniformity coeffi-

cient (d60/d10) of the material has an average value of 2.6. Arias

et al. (2001) suggested that the uniformity coefficient should

be less than 4 in order to secure an adequate hydraulic

conductivity. Filtralite-PTM is fulfilling all the requirements to

be a good P removal agent in CW systems.

2.2. Study sites—sampling

2.2.1. Small scaleTwo boxes with dimensions 26 cm long, 7 cm wide and 12 cm

high from an experiment described by Adam et al. (2005) were

used representing the small-scale system (Fig. 1a). The boxes

were filled with material up to 9 cm height but the top 2 cm of

the material was used only as an insulation layer. Each box

contained 953g of material but only about 700g was saturated

with P solution. One of the boxes was fed with 15ppm P

solution with a hydraulic loading rate (HLR) of 5 l day�1 (Box 1).

The other one was fed with the same P solution but with a HLR

of 1.25 l day�1 (Box 2). The theoretical residence time in the

boxes was 5 and 20 h in Box 1 and Box 2, respectively (Suliman

et al., 2005a). Phosphate concentrations and pH in the inlet and

outlet water were analyzed. The inlet concentration for both

boxes was kept constant at 15 ppm P for the entire experiment.

The inflow to the boxes was stopped when the boxes were

saturated with phosphorous, i.e. when the inlet concentration

was equal to the outlet concentration. When the experiment

was ended the filter material was sampled at three different

depths (top layer: 2–3 cm, middle layer: 4–5 cm and bottom

layer: 6-7 cm below the soil surface) at nine different locations

(Fig. 1) in order to measure the accumulated content of total P

(TP). Sequential extractions (Table 1) were carried out on a

composite sample from each box. The composite samples were

obtained by mixing the entire content of each box and a

representative sample was collected. During the first 2–3

months of the experiment associated with the very high pH

(11–12), white precipitate developed in the outlet draining tubes

of the boxes. The amount of precipitate was proportional to the

inlet P concentration and to the loading rate. The development

of the precipitates stopped when large amounts of Ca2+ and

Mg2+ were washed out from the system lowering the pH to

about 8–9.

2.2.2. Meso scaleThe meso-scale filter system consists of a container of

Plexiglass with the dimensions 3 m long, 0.29 m wide and

0.8 m high (Fig. 1b). The height of the filter material was 0.75 m.

The total amount of the material in the container was 653 l or

359 kg. Vertical PVC pipes were inserted at 12 locations in the

filter material (Fig. 1b) in order to examine solute concentra-

tions throughout the volume of the filter material. Solutes and

water were distributed evenly through a sponge replacing the

entire width of the top 0.14 m of the filter material. A P-solution

with a concentration of 6 ppm was entering the system as

pulse injections; every 2.75 min approximately 135 ml of

solution was applied to the inlet area. A constant head of

0.70 m was defined by the height of the outlet tube and the

head at the inlet was 0.72 m. The theoretical and measured

residence time in the system were approximately 86 and 103 h,

respectively (Suliman et al., 2005b).

The container was fed with an average concentration of

6 ppm P solution for about 1.5 years (17th March 2003–18th

October 2004). Water samples were taken from the observa-

tion wells as well as from the outlet and analyzed for PO43�–P.

Soil samples were collected after the experiment was finished

to measure the total sorbed P in the material. The samples

were taken from 13 locations at three different depths (10, 40

and 60 cm below the soil surface) marked with black crosses

in Fig. 1b. Sequential extraction of the material was not

conducted on this scale.

2.2.3. Large scaleA subsurface flow wetland system at Dal, Norway (CW1)

(Suliman et al., 2005b) has been operating since autumn 2000

ARTICLE IN PRESS

Table 1 – Chemical extraction scheme for 1 g of filter material

P-fraction Extractants Treatment

Loosely bond P 1 M NH4Cl 50 ml extractant, shaken for 1 h

Al–bond P 0.5 M NH4F 50 ml extractant, shaken for 1 h

Fe–bond P 0.1 M NaOH 50 ml extractant, shaken for 18 h

Ca–bond P 1 M H2SO4 50 ml extractant, shaken for 1 h

Occluded P KCl–C6H8O6–EDTA 50 ml extractant, shaken for 1 h

The soil was washed with 2� 25 ml saturated NaCl-solution between the different extractions.

0.3 m 0.5 m 0.9 m 1.7 3 m

0.19 m

0.38 m

0.57 m

0.75 m

3.0 m

Inlet

Outlet

Plexiglass tank filled with Filtralite PTM.

Wells for sampling water at different depths.

Sponge

TP soil samples

9 cm

2 cm

26 cm

7 cm

7 cm

Inlet

Outlet

Saturatedlayer

Insulation layer

Filtralite-P

(a)

(b)

Fig. 1 – (a) Illustrative sketch of the boxes used in the small-scale box experiment. Black crosses mark material sampling

points (modified from Suliman et al., 2005a,b). (b) Illustrative sketch of the middle scale box.

WAT E R R E S E A R C H 40 (2006) 1143– 1154 1145

and is treating wastewater from an elementary school. The

average hydraulic load is 1.6 m3 day�1. The system consists of

three parts: a 16 m3 septic tank, a vertical 0.7 m deep, 8 m wide

and 11 m long ðV ¼ 61:6 m3; M ¼ 33 800 kgÞ pre-treatment filter

with 10 domes and a horizontal flow filterbed with dimen-

sions of 25 m long, 8 m wide and 0.9 m high

ðV ¼ 180 m3; M ¼ 99 000 kgÞ. The horizontal filterbed is filled

with Filtralite-PTM (0–4 mm) except the last 2 m, where

Filtralite-PTM (2–4 mm) is used for better drainage. The 11 m

long pre-treatment system is located on top of the first part of

the horizontal filterbed and is filled with Filtralite-PTM

(2–4 mm) (Fig. 2). The wastewater is pumped in pulses from

the septic tank to the pre-treatment filter and is evenly

distributed on the surface of the filter by 10 special spraying

nozzles developed by Heistad et al. (2001). The pre-treated

wastewater is collected at the bottom of the pre-treatment

filter and then fed to the horizontal filterbed. The wastewater

is collected at the bottom of the outlet zone by a perforated

manifold, led through an outlet pipe with a level control and

further discharged to the local brook. The theoretical and

ARTICLE IN PRESS

Fig. 2 – Illustrative sketch of CW1.

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 1 1 4 3 – 1 1 5 41146

measured retention time of the system were 800 and 425 h,

respectively (Suliman et al., 2005b). The system has an

estimated lifetime of 15 years.

Soil and water samples from the horizontal filterbed were

collected in the autumn of 2003 when the system was about 3

years old. Water samples were taken at the inlet and outlet

zones and from the nine observation wells (Fig. 2) and

analyzed for PO4–P, while the soil samples were taken close

to these observation wells. Soil samples were collected from

nine spots (Fig. 2) at three different depths (40–50, 60–70 and

70–80 cm below the surface) to measure the content of TP.

Sequential extractions (Table 1) were carried out on samples

from two locations marked as number 2 and 5 in Fig. 2. One

soil sample was divided into two fractions (0–2 and 2–4 mm)

to examine the grain size influence on the P sorption. All

measured values of P in filter material from all three scales

were corrected for the background P concentration in

Filtralite-PTM (0.038 mg P kg�1).

Precipitates in the filter can clog the system and reduce the

hydraulic capacity of the media. For this purpose precipitate

samples were taken from a recently established full-scale

wetland system at Norderas, Norway (CW2) (Heistad et al.,

2004) that still has a high content of soluble Mg and Ca, and

fresh precipitates at the outlet zone. This system has been

operating since the winter of 2002 and receives ordinary

domestic wastewater from three family flats. Precipitate was

collected and analyzed as described below.

2.3. Chemical analysis

The soil samples were extracted for TP according to Møberg

and Pettersen (1982). Sequential extraction was conducted

according to Chang and Jackson (1957), modified by Hartikai-

nen (1979) (Table 1). Extractants were filtered through a

0.45mm membrane filter and determined by the ascorbic

acid—ammonium molybdate method of Murphy and Riley

(1962). SEM analyses were conducted in order to characterize

the microstructure and composition of the Filtralite-PTM

(0–4 mm) material before and after use, as well as the white

precipitates from the small boxes. The chemical composition

of the precipitate from CW2 was analyzed by using X-ray

diffraction (model: Philips, PW1730/10) analysis.

2.4. Statistical analyses

Statistical analyses (SAS) were run to find significant (Po0.05)

differences in the amount of sorbed P between the different

scales as well as compare each individual scale by width,

length and depth.

3. Results

3.1. Spatial distribution of P in water and soil

3.1.1. Small scaleThe inlet P concentration to both boxes was 15 ppm for

the entire experimental period. The outlet concen-

trations increased gradually in both boxes and the experi-

ment was stopped when the outlet concentrations approxi-

mately exceeded the inlet concentration. It took 150 and

538 days for Box 1 and 2, respectively, to become saturated

with P.

The total amount of P added to each box for the entire

experimental period was 7.4 and 4.8 g for Box 1 and 2,

respectively. Box 1 sorbed 76% and Box 2 sorbed 73% of the

added P. Thus the filter material in both boxes was approxi-

mately saturated at the end of the experiment. Mass balance

calculations show that the average amount sorbed was 5.6 g P/

700 g material and 3.5 g P/700 g material, resulting in overall P

sorption of 8.0 g P kg�1 (4.4 kg P m�3 material) and 5.0 g P kg�1

(2.8 kg P m�3 material) for Box 1 and 2, respectively. The

ARTICLE IN PRESS

WAT E R R E S E A R C H 40 (2006) 1143– 1154 1147

samples from the filter material were analyzed for TP, and the

results show a very heterogeneous pattern (Fig. 3a and b). The

maximum amount of P extracted from the material was 2500

and 3887 mg P kg�1 material in Box 1 and Box 2, respectively

(equal to 1.4 and 2.1 kg P m�3 material in the two boxes). In

Box 2 the amount of sorbed P decreased towards the outlet

(Fig. 3b), while in Box 1 there seemed to be an increase in the

sorbed concentration towards the outlet (Fig. 3a). In both

cases the sorbed P accumulated at the bottom of the boxes.

Still the amount of total P sorbed in the two boxes was not

significantly different (P40.05). There was also no signifi-

cantly difference between the amounts of total P in the

different layers within each box. The maximum sorption

capacity estimated by extraction of total P (2500 mg P kg�1)

and the average sorption capacity based on mass balance

calculations (8030 mg P kg�1) differed significantly in the case

of Box 1. One possible explanation could be that P was

precipitated along with Ca and Mg, followed by leaching of

Top layer

600

600

600

1600

1600

1600

1400

1400

1400

1200

1200

1200

1200

1000

1000

1000

1000

800

800

800

800

800

600

600

600

600

Width (cm)

1 3 5 6 7

Leng

th (

cm)

Leng

th (

cm)

5

10

15

20

25

Midd

800

800

800

1600

1400

1200

1200

101000

1000

1000

6

600

1400

1200

1200

1000

10001000

1000

1000

800

800800

800

800

600

600600

600

600

1000

8001600

1600

140014001400

1200

12001200

1200 1800

5

10

15

20

25

1200

1200

1000

1000

800

800

600

600

2 4

Wid

1 32

Wid

1 32

Width (cm)

1 3 5 6 72 4

1000

1000800600

1000800600

(a)

(b)

Fig. 3 – (a) Distribution of P (mg P kg�1 Filtralite-PTM) in different d

concentration of 15 ppm P). (b) Distribution of P (mg P kg�1 Filtra

loading rate of 1.25 l day�1 and an inlet concentration of 15 ppm

the boxes.

precipitates out from the system and accumulation in the

bottom of the collecting containers (Adam et al., 2005).

3.1.2. Meso scaleDuring the experiment, the P concentration in the outlet

reached 0.2 ppm. The average concentration in the deeper

wells (0.57 cm below the surface) was lower than the average

concentration in the upper wells (0.38 and 0.19 cm below the

surface) (Fig. 4a). Similar results were found for a Br-tracer

experiment in the same container (French et al., 2004). Thus

there seems to be a preferential flow in the upper parts of the

container. The water samples collected from the observation

wells also showed that the P concentration in the pore-water

decreased towards the outlet (Fig. 4b).

The total amount of P added to the container was 189 g for

the entire period and the total amount sorbed to the filter was

187 g. Thus according to mass balance calculations the

material sorbed 99% of the added P, resulting in overall P

le layer

800

800

800

16001400

14001200

12001000

1000

00

600

600

00

Bottom layer

2500

2500

2000

2000

2000

15002000

1500

1500

1500

1500

1500

1000

1000

1000

1000

1000

500

1600

1

1200

1200

1000

1000

1000

800

800

800

800

600600

600

600

600

1500

1000

1000

1000

1000

500

500 1000

1000

1000

1000

500

30002500

200015001500

1500

th (cm)

5 6 74

Width (cm)

1 3 5 6 72 4

Width (cm)

1 3 5 6 72 4

th (cm)

5 6 74

400

1400

2000

epths in Box 1 (hydraulic loading rate of 5 l day�1 and an inlet

lite-PTM) in different depths in Box 2 (hydraulic

P). The arrows represent the direction of the flow in

ARTICLE IN PRESS

0

2

4

6

80.3 m

0.5 m

0.9 m

1.7 m

outlet

Date

01.04.03 01.08.03 01.12.03 01.04.04 01.08.04

P C

on

cen

trat

ion

(m

g l-1

)P

Co

nce

ntr

atio

n (

mg

l-1)

0

2

4

6

80.19 m

0.38 m

0.57 m

outlet

(a)

(b)

Fig. 4 – (a) P concentration in the pore-water at the depth of 0.19, 0.38 and 0.57 m as well as in the outlet as a function of time.

(Each point on the figure at each depth represents an average value of four different distances from the inlet.). (b) P

concentrations in the pore-water (depth average) at different distances from the inlet as a function of time.

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 1 1 4 3 – 1 1 5 41148

sorption of 521 mg P kg�1 material (0.29 kg m�3 material).

However, in areas along the right side of the container P

sorption as high as 4500 mg P kg�1 Filtralite-PTM was mea-

sured (extraction of total P), indicating that the material could

sorb much higher amounts of P. The content of total P in the

filter material showed the same pattern as the water samples,

i.e. the sorbed P accumulated in the inlet section of the

container and decreased towards the outlet as well as towards

the bottom (Fig. 5a). In the inlet section more P was also found

in the material from the center of the container than in the

material close to the walls. With increasing depth and

increasing distance from the inlet the sorbed content was

higher in the material along the wall than in the material

from the center. The right side of the container also tended to

sorb more P in all three depths. The SAS showed that the inlet

section sorbed significantly (Po0.0001) higher amounts of P

than the rest of the container, there was, however, no

significant difference in P sorption by depth and width.

3.1.3. Large scaleThe inlet P concentration to the pre-treatment system at CW1

was 2.9370.68 ppm while the average outlet P concentration

was about 0.083 ppm. The cumulative amount of P added to

the system has been about 5.3 kg while the cumulative

amount of P in the outlet has been only about 150 g over the

3 years period. The media thus sorbed 97% of the added P,

resulting in overall P sorption of 52 mg P kg �1 material. This

shows that the media is far from being saturated, and this is

anticipated since CW1 was designed to last for 15 years and it

has only been working for 3 years. The amount of TP in the

material sampled from CW1 varied from 106 to 249 mg P kg�1

Filtralite-PTM and was highest at location 7 (249 mg P kg�1

Filtralite-PTM, which equals 0.14 kg P m�3 material) (Figs. 2 and

5b). There was no big difference between the different layers

at each location (Fig. 5b). The SAS showed no significant

difference (P40.05) regarding P sorption between the different

layers, however, the right side and the outlet zone had

significantly higher amounts of sorbed P than the left side and

the inlet zone. The concentrations in the water samples were

extremely low between 0 and 0.035 mg l�1 (results are not

shown). The highest PO4–P concentration (0.035 mg l�1) in the

water samples was measured in well 7 (Fig. 2). These results

were consistent with the findings in the soil samples.

During the 3 years period, the large-scale system had

sorbed significantly lower (Po0.0001) amounts of P on the

weight basis than the small- and medium-scale systems. The

ARTICLE IN PRESS

Top layer

500

500

500500

500

150015001500

100010001000

40003500 30003000

2500

2500

2500

2000

2000

2000

Width (cm)

5 10 15 20 25

Width (cm)

5 10 15 20 25

Width (cm)

5 10 15 20 25

Width (cm)

5 10 15 20 25

Width (cm)

5 10 15 20 25

Width (cm)

5 10 15 20 25

Leng

th (

cm)

Leng

th (

cm)

50

100

150

200

250

300Middle layer

25002500

20002000

15001500

10001000

500

500 500

500

45004000

35003000

3000

Bottom layer

500

500

500

500

500

500

500

1000

10003000

25002000

1500

1500

140140

140 160

160

150

150

150

130130130

130130

130

140140

190180

170

170

2

4

6

8

10

12

14

140

140

140

140

140

140140

150

150

150 180170

160

160

120

120

120

220200

180

180

160

160

140

140

140

40003500

(a)

(b)

Fig. 5 – (a) Distribution of P in the meso-scale container at different depths. The values are given as mg P kg�1 Filtralite-PTM. (b)

Distribution of P in the full-scale CW system. The values are given as mg P kg�1 Filtralite-PTM. The arrows represent

the direction of the flow in the boxes.

WAT E R R E S E A R C H 40 (2006) 1143– 1154 1149

average sorbed P concentrations based on mass balance

calculations were 8.03, 4.99, 0.52 and 0.052 g P kg�1 Filtralite-

PTM for Box 1, Box 2, meso-scale container and CW1,

respectively.

3.2. Sequential extraction

It was a priori assumed that Ca–P would be the main P pool

because of the high pH and Ca content of the media. The

results from the onsite system justified this hypothesis since

the Ca–P pool contributed 50–54% of the total P removal. In

the small boxes, on the other hand, only 29–41% of the P

removed was found as Ca–P (Table 2). The results for loosely

bound P were opposite, a high percentage (30–53%) in the

small boxes and a very low percentage (4–5%) in the onsite

system. The percentage of Al bound P was relatively high in

the onsite system (40–43%) and lower (16–27%) in the small

boxes. The difference in the proportion of loosely bound Ca–

and Al–P in the small boxes and the onsite system can be

caused by different operating times (Table 2). The sorbed P

was thus transformed from loosely bound P into stronger

bound Ca– and Al–P by time. Occluded P and Fe bound P were

not important P pools, this is due to the high pH and the

anaerobic conditions in the systems. These results are

consistent to the findings of Zhu (1998). Ca-bound P and

loosely bound P are the most available P sources for plant

uptake, especially in acidic soils where the solubility of Ca–P

compounds increases as the pH drops below 6 (Tisdale et al.,

1993). It was a priori expected that the smaller particles would

sorb more P based on their higher specific surface area,

however, there was no significant difference between the two

Filtralite-PTM fractions when it comes to P removal capacity

and sequential extraction. These results are consistent with

the ones found by Johansson (1998).

The results from the sequential extractions were supported

by SEM analysis of virgin and saturated Filtralite-PTM from the

small boxes. The results showed that there was only a

negligible amount of P in the virgin material, which is

consistent with the analytical laboratory measurements

(0.038 mg P kg�1 material). On the other hand, the P saturated

Filtralite-PTM had a high content of P connected to surfaces

where both Ca and Mg or Al were present (Fig. 6). Fe did not

seem to be important in the P accumulation process in the

system.

ARTICLE IN PRESS

Table 2 – P fractions in the boxes (Box 1 and Box 2) and in different locations (2, 5) of the CW1

Location Looselybound (%)

Ca-bound(%)

Al-bound(%)

Fe-bound(%)

Occluded(%)

PP

extracted(mg P kg�1

FL-PTM)

Operationtime

Box 1a 53 29 16 1 1 1186.7 150 days

Box 2b 30 41 27 1 1 878.7 538 days

CW1 (2) 40–50 cm

(0–2 mm)

5 50 43 0 2 267.4 3 years

CW1 (2) 40–50 cm

(2–4 mm)

4 54 40 0 2 254.5 3 years

CW1 (2) 40–50 cm 5 51 42 0 2 250.5 3 years

CW1 (2) 60–70 cm 5 50 43 0 2 259.8 3 years

CW1 (2) 70–80 cm 5 50 43 0 2 252.6 3 years

CW1 (5) 70–80 cm 4 53 40 1 2 272.3 3 years

a Hydraulic loading rate of 5 l day�1 and inlet P concentration of 15 ppm.b Hydraulic loading rate of 1.25 l day�1 and inlet P concentration of 15 ppm.

Fig. 6 – Surface of Filtralite-PTM particle after 1.5 years of P solution application in Box 1.

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 1 1 4 3 – 1 1 5 41150

3.3. Precipitates

With X-ray diffraction analysis the principal peaks of CaCO3

were observed from the samples of precipitate taken from the

outlet of CW2. Since the media is very rich in Ca2+ and has

high pH (10–12) the Ca2+ leaching out of the system

precipitates as CaCO3 in the presence of CO2 and oxygen in

the outlet zone. With SEM the precipitates from the small

boxes could be identified and an interesting structure was

observed. The inner core (Fig. 7b) contained Ca- and Mg-

oxides and -carbonates and was surrounded by Ca- and Mg-

phosphates (Fig. 7a). In the first phase of the experiments the

Ca and Mg leached out of the system and precipitated as Ca-

and Mg-oxides and -carbonates. In the second phase, P

appeared in the outlet zone together with Ca and Mg and

precipitated as Ca- and Mg-phosphate.

4. General discussion

The meso-scale system in the laboratory was designed based

on data from a small onsite system in Norway treating

wastewater from one household. We expected to get the same

loading rate and P inlet concentration at the Dal school as for

this small onsite system. Unfortunately the school has

produced less and more diluted wastewater than what the

system was designed for. This shows the possible variations

in full-scale systems, which makes them difficult to model.

For the small-scale boxes, higher loading rates and inlet P

concentrations were chosen in order to shorten the duration

of the experiment. If we had chosen the same loading rates

and P concentrations as used in the onsite system the

experiment would have taken at least 3 years to accomplish.

Different inlet P concentrations and loading rates do not

influence our results and conclusions regarding the P

distribution and fractionation on each scale.

4.1. Spatial variation at the microscale

The SEM results showed that the P was not evenly distributed

on the surface of the Filtralite-PTM particles (Fig. 6). One

reason could be that the material is rather heterogeneous on

this scale. The most obvious reason is, however, that calcium

phosphate crystals have a tendency to grow from relatively

few spots on the surfaces as shown by Griffin and Jurinak

(1973). According to Cole et al. (1953) the reaction of

phosphate with calcite surfaces involves adsorption of small

ARTICLE IN PRESS

Fig. 7 – SEM pictures of the same precipitate particle from the outlet tubes of the boxes. (a) The outside cover of the particle

with clear peaks of O, Mg, P and Ca. (b) The core of the particle with peaks of O and Mg.

WAT E R R E S E A R C H 40 (2006) 1143– 1154 1151

amounts of phosphate followed by precipitation of calcium

phosphate.

4.2. Spatial variation at the small, meso and large scale

The spatial distribution of P in the meso-scale system gives

valuable information regarding the sorption pattern of CWs

using Filtralite-PTM. As Fig. 4a and b as well as Fig. 5a shows

the material in the container is successively saturated with P,

i.e. there is like a P plume moving from the inlet to the outlet

and at the center of the plume the concentration in the water

is equal to the inlet concentration and the material is

saturated with P. This pattern was also observed for a similar

container using shell sand as filter material (Søvik and Kløve,

2005) as well as in a wastewater treatment system at Tveter

(south-eastern Norway) (Zhu et al., 2002). The small-scale

boxes did not show a similar pattern, which could be due to a

length of only 26 cm (Fig. 1a) while the plume in the container

is developing over a distance of 1 m from the inlet (Fig. 5a). In

CW1 this TP pattern is, however, not observed as statistical

results showed that there were no significant differences

between the various locations within the filter system. The

highest concentrations were also observed close to the outlet

where Filtralite-PTM (2–4 mm) material was used. This sug-

gests that short cuts or preferential pathways play an

important role in this particular wetland. Either chemical or

physical processes may cause the relatively high concentra-

tion of P in soil and water in the outlet zone. Ca2+ and Mg2+

ions, which are continuously leached out of the system, may

deposit in the outlet zone and thus induce P precipitation. On

the other hand the variable distribution of TP might be due to

irregularity in hydraulic parameters such as the hydraulic

conductivity, the grain size distribution or the porosity

(Suliman el al., 2005b).

4.3. Binding capacities

It is well known that most batch experiments overestimate

the P retention capacity of the material (Drizo et al., 2002).

Thus one of the questions addressed in this paper is whether

small and meso-scale flow experiments in the laboratory are

better tools for giving more reliable P sorption capacities,

which may be used for predicting the long-term behavior and

lifetime of a CW with Filtralite-PTM. Experiments with the first

generation of Filtralite showed that the maximum sorbed

concentration in the field was about half of the maximum

sorbed concentration found in batch experiments. Previous

batch experiments with Filtralite-PTM have shown a sorption

capacity of 12 g P kg�1 material (6.6 kg P m�3) (Jenssen and

Krogstad, 2003), and thus the authors assumed that the field

capacity of Filtralite-PTM would be 6 g P kg�1 material

(3.3 kg P m�3). In our experiments the maximum amount of

extracted TP obtained in the P saturated inlet section of the

meso-scale container was 4.5 g P kg�1 Filtralite-PTM with an

overall P sorption of 0.521 g P kg�1 material (Table 3). In the

small boxes the maximum amount of extracted TP was 2.5

and 3.9 g P kg�1 Filtralite-PTM with an overall P sorption of 8.03

and 4.99 g P kg�1 material in Box 1 and Box 2, respectively

(Table 3). In CW1 the maximum extracted TP concentration

was significantly lower than in the laboratory models, only

0.25 g P kg�1 with an overall P sorption of 0.052 g P kg�1

Filtralite-PTM. Thus the field sorption capacity of 6 g P kg�1

material estimated by Jenssen and Krogstad (2003) was found

in the box experiments and the inlet section of the meso-

scale experiment, where the media was almost completely

saturated with P.

By plotting the added P (g P kg�1 material) versus sorbed P

(g P kg�1 material), it is clearly seen that the total amount of P

added to the meso- and full-scale systems were lower than

the total amount added to the small-scale boxes (Fig. 8). This

gives lower overall sorbed P concentrations in the meso- and

full-scale systems. When it comes to the maximum sorbed

concentration, the small-scale boxes and the inlet zone of the

meso-scale container were close to P saturation, while the

inlet part of the full-scale system was still very far from P

saturation. The most probable reason for the low maximum

sorbed concentration in the full-scale system compared to

the laboratory system was a lower inlet concentration of P. As

already stated the school produced less and more diluted

wastewater than what was expected. Thus the intention of

comparing the laboratory experiments with the full-scale

system was not feasible. Other reasons for lower sorbed

concentrations in the full-scale system may be the loading

with real wastewater and thus the potential development of

biofilm, something which may reduce the sorption capacity of

the material compared to the lab experiments. A minimum

50% reduction in P sorption capacity of full-scale system-

s—due to the presence of solids and organic matter—was

suggested by Drizo et al. (2002). CW1 is also provided with a

pre-filter where substantial amounts (6–12 months of P

ARTICLE IN PRESS

Ta

ble

3–

Ad

ded

,m

ea

sure

da

nd

calc

ula

ted

Pin

the

dif

fere

nt

sca

les

Ma

sso

fFi

ltra

lite

-P

TM

(kg)

Vo

lum

eo

fFi

ltra

lite

-PT

M(m

3)

HR

T(h

)M

ass

of

Pa

dd

ed

(gP

kg�

1Fi

ltra

lite

-P

TM

)

Ma

sso

fP

ad

ded

(gP

m�

3Fi

ltra

lite

-P

TM

)

Mea

sure

dm

ax

imu

mso

rbed

con

cen

tra

tio

n(g

Pk

g�

1Fi

ltra

lite

-P

TM

)

Ca

lcu

late

davera

ge

sorb

ed

con

cen

tra

tio

n(g

Pk

g�1

Filt

rali

te-

PT

M)

Sm

all

-sca

leB

ox

10.7

00

1.2

7�

10�

35

10.6

5826

2.5

8.0

3

Sm

all

-sca

leB

ox

20.7

00

1.2

7�

10�

320

6.9

3779

3.8

87

4.9

9

Meso

sca

le359

0.6

53

86

0.5

26

289

4.5

0.5

21

La

rge

sca

le99

000

180

800

0.0

535

28

0.2

49

0.0

52

Large scale

Meso scale

Box 2

Box 1

0

2

4

6

8

10

12

0 4 6 8 10 12

Added P (g kg-1)

Sorb

ed P

(g

kg-1

)

2

Fig. 8 – P efficiency curve with the overall P sorption

calculated from the mass balance.

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 1 1 4 3 – 1 1 5 41152

sorption based on previous experience) of P may be retained,

further reducing the concentration in the water going into the

horizontal filterbed. On the other hand, the presence of plants

in CW1 may have contributed to an enhance P removal, but

might also have created cracks and pores in the material,

which could have led to preferential pathways. To sum up,

our results show that laboratory models give more reliable

estimates of the sorption capacity of the material than what

batch experiments might do and are in accordance with the

findings reported by Drizo et al. (2002). On the other hand,

even with laboratory models of this size, it may be difficult to

extrapolate the results to full-scale systems. Full-scale CWs

are more complex systems with varying inlet concentrations

as well as preferential pathways.

4.4. Estimates of lifetimes for a CW system withFiltralite P

By Norwegian guidelines a household of five persons will for a

period of 15 years need 40 m3 of filter material in order to have

an effluent concentration that does not exceed 1 mg l�1 P. This

means that there will be about 8 m3 person�1 for 15 years.

Jenssen and Krogstad (2003) have previously calculated the

lifetime of a CW by multiplying the maximum field sorption

capacity with the total mass of filter material. As previously

mentioned, the field sorption capacity was estimated to be

50% of the capacity measured with batch experiments using a

360 ppm phosphate solution, and equals 6 g P kg�1 Filtralite-

PTM (or 3.3 kg P m�3). Knowing that one person produces about

1.5 g P a day (equals 547.5 g a year) (Holtan et al., 1988), about

2.5 m3 of material is needed for one person over 15 years.

In Norway the outlet concentration from CWs treating

wastewater is not supposed to exceed 1.0 mg P l�1. The meso-

scale container was stopped before this concentration was

reached in the outlet (Fig. 4). However, some months before

the end of the experiment (middle of April, 2004) the depth

average concentration at 0.9 m from the inlet was at about

1.0 mg P l�1 (Fig. 4b). Still at the end of the experiment only

part of the material within the first 90 cm of the container had

attained the maximum sorption capacity (Fig. 5a). This shows

that not only a sorption maximum criterion is needed, but

also the average sorption concentration has to be taken into

account in the calculation of lifetime for a CW system. To

estimate the volume of the material needed for one person

ARTICLE IN PRESS

WAT E R R E S E A R C H 40 (2006) 1143– 1154 1153

over 15 years based on the meso-scale experiment’s P

extraction data we consider the first 0.9 m of the container

as a separate system. If we roughly assume that half of this

volume has attained the maximum sorption capacity of

4500 mg P kg�1 (2.5 kg P m�3), while the remaining half has

just sorbed about 1700 mg P kg�1 (equal to 0.94 kg P m�3) (Fig.

5a), then the average sorption capacity would be 1.7 kg P m�3.

Using this average P sorption capacity about 4.8 m3 of filter

material would be needed for 1 person over 15 years. This

amount is the double of the calculated amount based on the

capacity of Jenssen and Krogstad (2003), but still well within

the limits of the Norwegian guidelines.

No CWs are run to full saturation yet and no data are

available on maximum accumulation of these systems using

Filtralite-PTM. Ongoing CWs have to be followed over years to

better understand the processes going on for P accumulation

and to find laboratory systems that can be used to calculate

lifetime more accurately.

Acknowledgment

The work has been partly funded by the PRIMROSE project on

constructed wetlands (http://primrose.jordforsk.no) financed

by the European Commission contract CVK1-2000-00065.

R E F E R E N C E S

Adam, K., Krogstad, T., Suliman, F.R.D., Jenssen, P.D., 2005.Phosphorous sorption by Filtralite-PTM—small-scale box ex-periment. J. Environ. Sci. Health 40 (6–7), 1239–1250.

Arias, C.A., Brix, H., 2005. Phosphorous removal in constructedwetlands: can suitable alternative media be identified? WaterSci. Technol. 51 (9), 267–273.

Arias, C.A., Del Bubba, M., Brix, H., 2001. Phosphorous removal bysands for use as media in subsurface flow constructed reedbeds. Water Res. 35 (5), 1159–1168.

Chang, S.C., Jackson, M.L., 1957. Fractionation of soil phosphorus.Soil Sci. 84, 133–144.

Cheung, K.C., Venkitachalam, T.H., 2000. Improving phosphateremoval of sand infiltration system using alkaline fly ash.Chemosphere 41, 243–249.

Cole, C.V., Sterling, R., Scott, C.O., 1953. The nature of phosphatesorption by calcium carbonate. Soil Science Society Proceed-ings, pp. 352–356.

Drizo, A., Frost, C.A., Smith, K.A., Grace, J., 1997. Phosphate andammonium removal by constructed wetlands with horizontalsubsurface flow, using shale as a substrate. Water Sci. Technol.35 (5), 95–102.

Drizo, A., Frost, C.A., Grace, J., Smith, K.A., 1999. Physico-chemical screening of phosphate-removing substratesfor use in constructed wetland systems. Water Res. 33 (17),3595–3602.

Drizo, A., Comeau, Y., Forget, C., Chapuis, R.P., 2002. Phosphoroussaturation potential—parameters for estimating the longevityof constructed wetland systems. Environ. Sci. Technol. 36 (21),4642–4648.

French, H.K., Kløve, B., Søvik, A.K., 2004. Flow and transportpathways in small scale experimental subsurface flow con-structed wetlands. In: Proceedings to the Ninth (IWA) Inter-national Conference on Wetland Systems for WastewaterPollution Control, vol. 2. Avignon, France, 26–30 September,2004.

Griffin, R.A., Jurinak, J.J., 1973. Reaction of phosphate with calcite.Soil Sci. Soc. Am. Proc. 37, 847–850.

Hartikainen, H., 1979. Phosphorus and its reactions interrestrial soils and lake sediments. J. Sci. Agric. Soc. Finland51, 537–624.

Heistad, A., Jenssen, P.D., Frydenlund, A.S., 2001. A new combineddistribution and pretreatment unit for wastewater soil infil-tration systems. In: Manel, K. (Ed.), Onsite WastewaterTreatment, Proceedings of Ninth International Conference onIndividual and Small Community Sewage Systems. ASAE, pp.200–206.

Heistad, A., Vrale, L., Paruch, A.M., Adam, K., Jenssen, P.D., 2004.From constructed wetlands to compact filter systems. In:Proceedings to the Ninth (IWA) International Conference onWetland Systems for Wastewater Pollution Control, vol. 2.Avignon, France, 26–30 September, 2004, pp. 759–766.

Holtan, H., Kamp-Nielsen, L., Stuanes, A.O., 1988. Phosphorous insoil, water and sediment: an overview. Hydrobiologia 170,19–34.

Jang, H., Kang, S.-H., 2001. Phosphorous removal using cowbone in hydroxyapatite crystallization. Water Res. 36,1324–1330.

Jenssen, P.D., Krogstad, T., 2003. Design of constructed wetlandsusing phosphorus sorbing lightweight aggregate (LWA). In:Mander, U., Jenssen, P.D. (Eds.), Constructed Wetlands forWastewater Treatment in Cold Climates. International Serieson Advances in Ecological Sciences. WIT Press, Great Britain.

Jenssen, P.D., Mæhlum, T., Krogstad, T., Vrale, L., 2005. Highperformance constructed wetlands for cold climates. J. En-viron. Sci. Health 40 (6–7), 1343–1353.

Johansson, L., 1998. Phosphorous sorption to filter substrate-s—potential benefits for on-site wastewater treatment. De-partment of Civil and Environmental Engineering, RoyanInstitute of Technology. Kungl Tekniska Hogskolan, Stock-holm.

Kadlec, R.H., Knight, R.L., 1996. Treatment Wetlands. CRC Press,Lewis Publishers, Boca Raton, 893pp (ISBN:0-87371-930-1).

Khadhraoui, M., Watanabe, T., Kuroda, M., 2002. The effect of thephysical structure of a porous Ca-based sorbent on itsphosphorous removal capacity. Water Res. 36, 3711–3718.

Kvarnstrom, M.E., Morel, C.A.L., Krogstad, T., 2004. Plant-availability of phosphorous in filter substrates derivedfrom small-scale wastewater treatment systems. Ecol. Eng. 22(1), 1–15.

Molle, P., Lienard, A., Iwema, A., Kabbabi, A., 2005. Apatiteas an interesting seed to remove phosphorous from waste-water in constructed wetlands. Water Sci. Technol. 51 (9),193–203.

Møberg, J.P., Petersen, L., 1982. Øvelsesvejledning til geologi ogjordbundslære II (Practical lessons for geology and soilscience). Den kgl. Veterinær- og Landbbohøgskole, Købehavn.,136pp (in Danish).

Murphy, J., Riley, J.P., 1962. A modified single solution method forthe determination of phosphate in natural waters. Anal. Chim.Acta 26, 31–36.

Optiroc Group AB, 2003. Product specification of Filtralite-PTM,Oslo, Norway.

Suliman, F., Haugen, L.E., Jenssen, P.D., 2005a. Salinity effect onthe characteristics of saturated horizontal flow in small boxescontaining Filtralite-P. J. Environ. Sci. Health 40 (6–7),1403–1413.

Suliman, F., French, H., Haugen, L.E., Jenssen, P.D., Kløve, B.,2005b. The effect of scale of horizontal subsurface flowconstructed wetlands on flow and transport parameters.Water Sci. Technol. 51 (9), 259–266.

Søvik, A.K., Kløve, B., 2005. Phosphorus retention processes inshell sand filter systems treating municipal wastewater. Ecol.Eng. 25 (2), 168–182.

ARTICLE IN PRESS

WAT E R R E S E A R C H 4 0 ( 2 0 0 6 ) 1 1 4 3 – 1 1 5 41154

Tisdale, S.L., Nelson, W.L., Beaton, J.D., Havlin, J.L., 1993. SoilFertility and Fertilizers, fifth ed. Macmillan Coll Div. 634pp.

Vymazal, J., 2004. Removal of phosphorus via harvesting ofemergent vegetation in constructed wetlands for wastewatertreatment. In: Proceedings to the Ninth (IWA) InternationalConference on Wetland Systems for Wastewater PollutionControl, vol. 2. Avignon, France, 26–30 September, 2004, pp.415–422.

Zhu, T., 1998. Fate of phosphorous in a light-weight aggregate(LWA) wastewater treatment wetland. Ph.D. Thesis, NLH,Norway.

Zhu, T., Mæhlum, T., Jenssen, P., Krogstad, T., 2002. Fate ofphosphorus in a light-weight aggregate wastewater treatmentwetland. In: Proceedings to the Eighth (IWA) InternationalConference on Wetland Systems for Water Pollution Control,University of Dar Es Salaam, 16–19 September 2002.