Wat. Res. Vol. 33, No. 17, pp.

3595±3602, 1999
# 1999 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
PII: S0043-1354(99)00082-2 0043-1354/99/$ - see front matter

www.elsevier.com/locate/watres

PHYSICO-CHEMICAL SCREENING OF PHOSPHATE-

REMOVING SUBSTRATES FOR USE IN CONSTRUCTED
WETLAND SYSTEMS
A. DRIZO1*, C. A. FROST2 M, J. GRACE1 and K. A. SMITH1
1
Institute of Ecology and Resource Management, University of Edinburgh, West Mains Road,
Edinburgh EH9 3JG, UK and 2The Environmental Protection Group, Scottish Agricultural College,
Bush Estate, Penicuik, Midlothian EH26 0PH, UK

(First received 1 June 1998; accepted in revised form 1 February 1999)

AbstractÐThe objective was to provide selection criteria for substrates that would enhance phosphate
removal from waste water in a constructed wetland system (subsurface horizontal ¯ow). Measured
properties of seven substrates (bauxite, shale, burnt oil shale, limestone, zeolite, light expanded clay
aggregates (LECA) and ¯y ash) were: pH, cation exchange capacity (CEC), hydraulic conductivity,
porosity, speci®c surface area, particle size distribution and phosphate (P) adsorption capacity. Fly ash
and shale had the highest P adsorption values, followed by bauxite, limestone and LECA. Longer-term
experiments in which synthetic waste water was passed over shale and bauxite gave maximum P uptake
values of 730 and 355 mg P kgÿ1, respectively. X-ray ¯uorescence measurements showed that
substantial precipitation of P had occurred on the shale surfaces. On the basis of these measurements it
was concluded that, of the seven materials examined, shale had the best combination of properties as a
substrate for constructed wetland systems (CWS). # 1999 Elsevier Science Ltd. All rights reserved

Key wordsÐwaste water, phosphorus, shale, bauxite, ¯y ash, light expanded clay aggregates, Phragmites
australis, removal eciency, life expectancy of a CWS

INTRODUCTION Results from several studies have shown that im-

Recent work has emphasised the potential import-
ance of natural and constructed wetland systems
(CWS) for purifying waste water. These systems are
useful for reducing biochemical oxygen demand
(BOD), suspended solids (SS), ammonium (N) and
phosphorus (P) in sewage (Moshiri, 1993; Kadlec
and Knight, 1996; Vymazal et al., 1998). Such sys-
tems are becoming widely accepted as a suitable
low-cost technology for small to medium-sized com-
munities (Moshiri, 1993; Vymazal et al., 1998).
Studies of the removal of P by wetland systems
have been carried out in the U.S.A. (Moshiri, 1993;
Kadlec and Knight, 1996), Australia (Mann and
Bavor, 1993) and several countries in Europe:
Denmark (Schierup et al., 1990), Norway (Zhu et
al., 1997), U.K. (Green, 1997), Czech Republic
(Vymazal, 1995) and more recently, Sweden
(Johansson, 1997) and the Netherlands (Schreijer et
al., 1997).
*Author to whom all correspndence should be addressed.
Present address: CGM Department, Ecole
Polytechnique de Montreal, P.O. Box 6079, Station
Centre-Ville, Montreal, Quebec, Canada H3C 3A7.
Tel.: +1-514-340-4711; e-mail: aleksandradrizo@
polymtl.ca
3595
mobilisation of P in CWS occurs through substra-
tum adsorption, chemical precipitation, bacterial
action, plant and algal uptake and incorporation
into organic matter (Kadlec and Knight, 1996;
Vymazal et al., 1998). Of these, the substrate may
play the greatest role, and could very well be the
factor that is most amenable to control.
Consequently, it is important to select those sub-
strates with the highest P adsorption capacity,
which is dependent upon chemical and physical
properties of the material (Zhu et al., 1997; Drizo,
1998). Such materials might include minerals with
reactive Fe or Al hydroxide or oxide groups on
their surfaces, or calcareous materials which can
promote Ca phosphate precipitation (Drizo et al.,
1997; Johansson, 1997; Zhu et al., 1997). Apart
from Fe, Al and Ca minerals, the rate of P adsorp-
tion is controlled by substrate Eh, pH and adsorp-
tive surface area (Vymazal et al., 1998). Fine
grained materials have large surface areas and
therefore the potential to enhance P adsorption ca-
pacity (Zhu et al., 1997). However, such materials
often have low hydraulic conductivity which leads
to the occurrence of overland ¯ows and insucient
contact between wastewater and substrate within
the CWS (Kadlec and Knight, 1996; Drizo, 1998).
3596 A. Drizo et al.
Therefore, the materials should be suciently per-
meable to prevent surface channelling of the CWS
(Kadlec and Knight, 1996). In addition, they should
be cheap and locally available in order to reduce
the costs of CWS construction.
This paper describes investigations of a range of
candidate substrates, all meeting the above criteria.
Comparison of seven materials is presented, based
on measurements of pH, cation exchange capacity
(CEC), hydraulic conductivity, porosity, speci®c
surface area and particle size distribution, and
phosphate adsorption capacity. These parameters
all in¯uence the e€ectiveness of a substrate for the
removal of phosphates, on a short-term basis. Also,
to estimate the capacity of substrates, laboratory in-
vestigations of the P saturation point are reported.
Finally, X-ray ¯uorescence spectrometric analysis
was used to demonstrate the presence of P on the
surface of the materials. The properties investigated
in the later two experiments in¯uence the e€ective-
ness of the substrate for the removal of phosphates,
on a long term basis.
MATERIALS AND METHODS
Materials
The seven materials chosen for investigation of their
physico-chemical properties were:
. Bauxite, a naturally occurring mixture of minerals
rich in hydrated aluminium oxides and ferric oxides
and low in alkali metals, alkaline earths and silicates,
obtained from Alcan Chemicals Europe, Burntisland,
Fife.
. Shale, an argillaceous rock, derived from the lower lime-
stone group of the carboniferous system, which is highly
®ssile and which splits readily into very thin laminae. It
is readily available material in Scotland (Drizo, 1998);
the one used for experiments was obtained from
Cultshill Quarry, Fife.
. Burnt oil shale (BOS), widely available in central
Scotland as the waste product from heating oil shale to
produce mineral oil.
. Limestone, composed largely of calcium carbonate
(CaCO3) in the form of calcite. The material used for
experiments was obtained from Middleton Lime
Quarry, Midlothian.
. Zeolite, a hydrated aluminium-silicate mineral with the
aluminium and silicon polyhedra linked by the sharing
of oxygen atoms.
. Lightweight expanded clay aggregates (LECA), formed
by expanding special clay minerals at high temperature
producing lightweight (300±400 kg mÿ3) ceramic peb-
bles. The material used for experiments was obtained
from the Clydesdale Trading Company, Lanark, who
had imported it from Norway.
. Fly ash, an inorganic waste product from coal combus-
tion, consisting mainly of spherical glassy particles of
silica (SiO2), alumina (Al2O3) and iron oxides. The ma-
terial used here was furnace bottom ash which has coar-
ser particle size and, hence, higher permeability than
®ne ¯y ash, and originated from Longannet Power
station, Fife.
All of these materials, except the limestone, were rich in
Fe and Al; the limestone was a material expected to pro-
mote precipitation of Ca phosphates.
Methods
Determination of pH, CEC, hydraulic conductivity, por-
osity, speci®c surface area and particle size distribution.
The pH of each substrate was determined with a pH meter
(model 701a, Orion, Cambridge, MA) after mixing 2 g of
milled (< 2 mm) material with 20 ml of 0.01 M CaCl2
(McLean, 1982). Cation exchange capacity was measured
using the colorimetric method with methylene blue (MB)
(Rhoades, 1982). From the number of ``equivalents'' of
MB adsorbed the concentration of cations exchanged in
the solution was calculated and CEC determined.
Hydraulic conductivity was measured using a constant
head permeameter (0.65 m height). Acrylic plastic tubes
with two di€erent bores (diameter of 0.1 and 0.07 m, re-
spectively) were used. The tubes were ®lled with the di€er-
ent substrates to a height of 0.65 m. Water was passed
through them, using an upward ¯ow to facilitate removal
of trapped air. The hydraulic conductivity was calculated
using Darcy's law:
Q ˆ kA DH=DL† 1†
where Q = volumetric ¯ow of the solution (m3 sÿ1);
k = saturated hydraulic conductivity (m sÿ1); A = cross
sectional area of the tube (m2); DH/DL = hydraulic gradi-
ent of the ¯ow system expressed as a decimal fraction
where DH = the hydraulic head (m) and DL = the length
of the column (m).
The porosity of substrates was determined using the
standard soil science procedure based on estimations of
bulk density and particle density (Klute, 1986). Speci®c
surface area was determined using the N2 gas adsorption
method (Davis and Kent, 1990). Particle size distribution
was measured by dry sieving each of the substrates
through four di€erent mesh sizes, ranging from 2.8 to
12.7 mm, and subsequent weighing.
Determination of maximum phosphate adsorption ca-
pacity. A slight modi®cation of a standard batch equili-
brium technique (Nair et al,, 1984) was used. Five
di€erent P solutions, ranging in concentration from 2.5 to
40 mg P lÿ1 (0.08 to 1.25 mM P as KH2PO4 in 0.01 M
CaCl2), were equilibrated with 20 g of the substrate for
24 h by continuous shaking on a rotating shaker (60 rpm)
at constant temperature (218C). Three drops of toluene
were added to each tube in order to inhibit microbial ac-
tivity. Suspensions were then centrifuged and the P
remaining in the supernatant was determined using a stan-
dard ammonium molybdate method (DoE, 1980). Phos-
phorus adsorption maxima were determined from the
Langmuir equation (Barrow, 1978):
abC
Pa ˆ 2†
1 ‡ aC †
where C is the concentration of P in the solution at equili-
brium (g mÿ3); Pa is the amount of P adsorbed per unit
weight of material (g kgÿ1); a is the constant related to the
binding strength of P on the material; b is the P adsorp-
tion maximum (g kgÿ1). a and b were determined using
the SigmaStat iterative curve ®tting package (Kuo et al.,
1992).
Determination of the saturation point of the substrate.
Twelve Perspex columns (inner diameter: 100 mm; height:
900 mm) were packed to a height of 600 mm with each
of four substrates selected as a result of the adsorption
capacity study (bauxite, shale, ¯y ash and LECA), giving
a replication of three per substrate. The columns were
fed from below with a synthetic waste water solution
devised by DoE (1981), from a common reservoir, using
a separate peristaltic pump for each column. The syn-
thetic waste water (made up by mixing 7 di€erent com-
ponents: bacteriological peptone, meat extract, urea,
sodium chloride, calcium chloride, magnesium sulphate
 CWS substrates for P removal
and dipotassium hydrogen phosphate) is widely used as a
substitute for real sewage in the Water Research Centre's
waste water treatment studies in Great Britain. Such syn-
thetic sewage after dilution normally contains approxi-
mately 106 mg lÿ1 organic C, 46 mg lÿ1 N and 5 mg lÿ1 P
(DoE, 1981). However, in an attempt to achieve a sub-
strate saturation by P, the K2HPO4-P concentration was
increased to 35±45 mg lÿ1 (about twice the average P con-
centration of domestic waste waters) and extremely high
loading rates (approximately 25 g P mÿ3 dÿ1) were used
because of the time limitations of the experiment. The
pumps operated for 10-min periods six times per day, sup-
plying a total of 3 l dÿ1. The average contact time (pore
volume retention time) between the solution and the sub-
strate was 12 h. Samples of the out¯ow from each column
were taken every second day and analysed for P content,
using a standard ammonium molybdate method (DoE,
1980), to determine when saturation of the substrates was
reached.
After 40 days, saturation had not been reached in any
of the substrates. As ¯y ash had already been used because
of its high P adsorption capacities in investigations carried
out by other authors (Mann and Bavor, 1993), it was
decided to continue the investigation using only the shale
and bauxite. To speed up the saturation of the substrates,
the P concentration of the feeding solution and the daily
¯ow through the system were doubled, thus increasing the
inlet loading rate 4-fold, to 100 g P mÿ3 dÿ1.
Examination of the substrate surfaces by X-ray ¯uor-
escence spectrometry (XRFS). In order to determine
whether phosphate precipitation had occurred during the
column experiment, individual samples of substrate ma-
terial were taken from the columns after the saturation
point had been reached, cut into 10±30 mm sections and
analysed using an X-ray ¯uorescence spectrometry
(XRFS) technique. The substrate fragments were bom-
barded by X-rays from a Rh-anode source and the inten-
sity of the characteristic ¯uorescent radiation from P
measured with a wavelength-dispersive spectrometer (Phi-
lips PW1480).
RESULTS
pH, CEC, hydraulic conductivity, porosity, speci®c
surface area and particle size distribution
The measured values of these parameters are
given in Table 1. In order to test whether there was
any signi®cant di€erence in these properties between
the substrates, ®ve one-way ANOVA analyses were
performed (one test for each property), using
SigmaStat statistical software (Kuo et al., 1992).
Table 1. pH, CEC, hydraulic conductivity, porosity and surface area of substrates. In any one column, values marked with di€erent
letters are signi®cantly di€erent from each other ( p < 0.05). Standard deviations are shown in parentheses (n = 3)
Substrate pH CEC (cmol kgÿ1) Hydraulic conductivity (m sÿ1  10ÿ4)
Bauxite 5.9b (0.1) 8.5a (1.3)
Shale 4.5a (0.1) 19.4b (2.8)
BOS 7.4d (0.2) 23.3b (1.0)
Limestone 7.8d (0.1) 22.8b (1.3)
Zeolite 6.5e (0.2) 23.0b (1.1)
LECA$ 8.2e (0.1) 9.5a (2.0)
Fly ash 8.3e (0.2) 9.3a (0.4)

BOS=Burnt oil shale.
$
LECA=Light expanded clay aggregates.
3597
The pH values covered a wide range. Shale was sig-
ni®cantly ( p < 0.05) the most acid, followed by
bauxite. Fly ash and LECA were the most alkaline
and not statistically di€erent from each other
(Table 1). CEC values fell into two groups, with
BOS, zeolite, limestone and shale having CEC
values very much higher than those for bauxite,
LECA and ¯y ash. Although hydraulic conduc-
tivities were generally between 10ÿ3 and 10ÿ4 m sÿ1,
except for that of zeolite (2.94  10ÿ3), they were
signi®cantly di€erent ( p < 0.05) from each other
(except for the two pairs, shale and BOS, and ¯y
ash and limestone). Zeolite had a porosity signi®-
cantly higher than those of the other substrates
( p < 0.05), whereas 5 of the others with lower
values did not di€er signi®cantly from each other.
The speci®c surface area for LECA and ¯y ash
could not be calculated because the ranges of the
relative pressures for these two substrates exceeded
the values for which the equation used in the N2
adsorption method (Davis and Kent, 1990) is valid
(between 0.05 and 0.30). The speci®c surface area of
the zeolite was higher than that of shale, which in
turn was signi®cantly greater ( p < 0.05) than the
other three materials (Table 1).
Particle size distribution varied considerably
among the substrates (Table 2), with LECA and
limestone containing by far the biggest proportion
of large particles >12.7 mm (70.2 and 61.7%, re-
spectively) and ¯y ash the largest proportion of
small particles < 2.8 mm (77.8%).
Maximum phosphate adsorption capacity
Fly ash and shale showed the greatest mean P
adsorption (21 S.D., n = 3) of the seven di€erent
substrates as a function of P concentration in sol-
ution, followed by bauxite, limestone and LECA
(Fig. 1). When P adsorption data were plotted
according to the Langmuir model of adsorption
(Fig. 2), a close ®t to Eq. (2) was observed for all
the substrates. Therefore, bonding capacities a and
theoretical P adsorption maxima b of the substrates
(equation 2) were determined. P adsorption maxima
varied between 0.42 and 0.86 g kgÿ1, while bonding
capacities varied between 0.07 and 0.6 (Table 3).
Porosity (%) Surface area (m2 gÿ1)
5.9b (0.3) 34.5a,b (2.8) 6.8a (0.4)
10.0d (0.7) 37.7a,b (1.9) 19.9c (1.6)
10.2d (0.3) 40.0a,b (2.0) 8.3b (0.3)
2.4a (0.1) 33.0a (1.6) 7.4a,b (0.5)
29.4e (0.5) 54.4c (1.0) 31.4d (0.2)
7.0c (0.2) 40.4a,b (2.9) ±
2.8a (0.3) 45.3b (3.0) ±
3598 A. Drizo et al.

Table 2. Particle size distribution of substrates

Substrate (% w/w)a
0.0±2.8 mm 2.9±4.0 mm 4.1±6.7 mm 6.8±12.6 mm > 12.70 mm

Bauxite 0.2 0.0 2.6 75.7 21.5

Shale 1.9 5.3 25.6 41.9 25.3
BOS 22.3 11.8 21.9 22.6 21.4
Limestone 1.2 2.2 8.6 26.3 61.7
Zeolite 0.0 0.0 4.2 83.6 12.2
LECA 0.4 0.1 4.3 25.0 70.2
Fly ash 77.8 3.5 6.5 12.2 0.0

a
WÐweight of size fraction as a percentage of total weight of sample.

Table 3. Bonding capacities, a, and theoretical P adsorption max-

ima, b, derived from the Langmuir equation (Eq. 2) (2=S.E.;
n = 7)

Substrate Bonding capacity, a P adsorption maximum, b (g kgÿ1)

Bauxite 0.26 20.01 0.61 20.08

Shale 0.61 20.02 0.65 20.07
BOS 0.06 20.01 0.58 20.09
Limestone 0.10 20.02 0.68 20.09
Zeolite 0.03 20.01 0.46 20.08
LECA 0.10 20.02 0.42 20.04
Fly ash 0.07 20.01 0.86 20.06

In order to ®nd the most signi®cant parameters

Fig. 1. Mean P adsorption (21 S.D., n = 3) for the seven
di€erent substrates as a function of P concentration in sol-
ution. Solution±substrate mixtures had been equilibrated
for 24 h at constant temperature (218C).
which in¯uenced the P adsorption capacity of the
substrates, regression analyses were made of the
theoretical maximum P adsorption values (obtained
from the Langmuir equation (Eq. 2) and Table 3)
plotted against pH, CEC, hydraulic conductivity,
porosity, surface area and bonding capacity of the
substrates. None of these properties had a signi®-
cant e€ect on P adsorption capacity.
The three substrates chosen for further study
were shale, bauxite and ¯y ash, which included
both the materials with the highest P adsorption ca-
pacities, and also represented a wide range in pH.
A fourth substrate, LECA, was also included,
despite its low P adsorption and bonding capacity
values, because experience in Scandinavian
countries suggested that it had promise as a sub-
strate (Brix, 1994).

Fig. 2. Langmuir plot of mean P adsorption data (21 S.E., n = 3) for all seven substrates. Bonding ca-
pacities and theoretical P adsorption maxima are presented in Table 3.
 CWS substrates for P removal 3599

Substrate saturation points could not be determined. The dark patches on

As indicated above, after 40 days of exposure of
the substrates to a P loading rate of 25 g
P mÿ3 dÿ1, shale had adsorbed the highest amount
of P (400 mg P kgÿ1), followed by ¯y ash (300 mg
P kgÿ1) and then bauxite (160 mg P kgÿ1), but sat-
uration had not been reached in any of the sub-
strates (Fig. 3a). The 4-fold increase in the inlet
loading rate to 100 g P mÿ3 dÿ1, at this point,
resulted in a sharp rise in the rate of P adsorption
by shale and bauxite over the following 20 days and
saturation was reached at about 70 d (Fig. 3b).
Overall, the total input of P to the substrates over
the whole period of investigation was 24.7 g, the
total output P from the shale column was 21.8 g
and the total adsorbed 2.9 g, giving a saturation
value of 730 mg P kgÿ1 (Fig. 3b). The correspond-
ing value for bauxite was ca. 350 mg P kgÿ1.
XRFS measurements
After the saturation point of P had been reached
(as described above), subsamples of the substrates
from each tube were examined; changes in colour
were observed on the surfaces of both bauxite and
shale, suggestive of precipitation of insoluble phos-
phates. Bauxite had dark patches and shale orange
ones. Investigation of these coatings by XRFS
showed that detectable P deposition had occurred
only on the shale particles. The concentration of P
on the shale had increased roughly 3-fold during
the period of exposure in the column, as indicated
by the X-ray count rate, which is directly pro-
portional to the P concentration (Brindley and
Brown, 1980); the absolute quantities deposited
bauxite did not show comparable P deposition;
their composition was not investigated further.
DISCUSSION
The measured hydraulic conductivities of 2±
29  10ÿ4 m sÿ1 (Table 1) were in the range of
values recommended for CWS (EC/EWPCA, 1990)
and it has been suggested that once developed, the
hydraulic conductivity of a CWS will stabilize and
maintain itself (Kickuth, 1977).
The physical and chemical properties listed in
Table 1 were not strongly linked with the theoreti-
cal P adsorption capacities. There is a considerable
controversy over the proper use of the Langmuir
equation for adsorption studies, arising from the
fact that it was developed from the assumption that
adsorbed gases form a monomolecular ®lm on uni-
form surfaces, whereas the adsorption of phosphate
on minerals involves adsorption onto non-uniform
surfaces (Barrow, 1978). The biggest advantage in
the application of the Langmuir model for adsorp-
tion studies is that it enables the calculations of
theoretical P adsorption maxima to be made.
Therefore, despite the limitations of the mechanistic
P adsorption equations, they are still widespread in
P adsorption studies, with Langmuir equations
being most frequently used (Stuanes, 1982).
The material with the smallest particle size distri-
bution will have the highest surface area, increasing
the potential for direct reaction with phosphates.
The measurements showed such a trend for most of
the substrates investigated, but zeolite was an
exception; it had the highest speci®c surface area

Fig. 3. Cumulative P adsorption capacities for shale, bauxite, LECA and ash over (a) 40 days and (b)
80 days. The columns (0.1 m diameter, 0.6 m height) were fed from below with 3 l dÿ1 of the synthetic
sewage solution containing 35±42 mg Pÿ1 (loading rate of 25 g P mÿ3 dÿ1) during the ®rst 40 days of
experimentation (a), followed by 70±90 mg P lÿ1 and 6 l dÿ1 of the solution, resulting in a 4-fold
increase of the inlet loading rate to 100 g Pmÿ3 dÿ1 (b) from a common reservoir, using a separate peri-
staltic pump for each column. The arrow in (b) indicates the place of the 4-fold increase in the inlet
loading concentration.
3600 A. Drizo et al.
(Table 1) but the lowest theoretical P adsorption ca-
pacity (Table 3 and Fig. 2c). The reason for this
may have been that P adsorption was reduced by
di€usion constraints: 96 % of the zeolite particles
were >6.8 mm in size (Table 2) and secondly, the
Langmuir equation may have been inappropriate in
this context.
The results of the P saturation point measure-
ments (Fig. 3) indicated that shale had the highest
P adsorption capacity, in excess of 650±
700 mg kgÿ1. The closest reported values for the
materials used in CWS are 430 mg P kgÿ1 for blast
furnace slag and 240 mg P kg ÿ1 for ¯y ash (Mann
and Bavor, 1993). Zhu et al. (1997) recently showed
that some types of Norwegian lightweight expanded
aggregates (LWA) have P adsorption capacity of
1390 mg P kgÿ1. On the other hand, Johansson
(1997) concluded that LWA produced in Sweden
was chemically non-reactive and therefore was not
suitable for P removal unless lime was added. As
the production of LWA requires substantial
amounts of energy, the cost e€ectiveness of this ma-
terial as a substrate for a CWS remains question-
able.
The presence of P deposition on shale surfaces
during exposure to the synthetic waste water sol-
ution was con®rmed by X-ray ¯uorescence analysis.
Such deposition may lead to enhanced removal of
P. This has particular implications for wastewater
treatment; Stuanes (1982) showed that P removal
by soils receiving long-term wastewater treatment
applications was much greater than that predicted
by adsorption maxima.
Life expectancy of shale based CWS
The results from this study are not sucient to
forecast accurately the lifetime of a CWS using
shale as a substrate. Full-scale and supplementary
greenhouse tests are necessary to allow such predic-
tion. However, data obtained from the batch exper-
iment using phosphate solution and the saturation
point measurements can be used to obtain an initial
estimate.
If the observed saturation capacity were to be
applied at a ®eld scale, it is expected that one tonne
of shale could absorb in excess of 0.7 kg P. The
quantity of phosphate (PO4) excreted per person
per day is 3 g PO4; in addition, another 4 g is dis-
charged from cleaning compounds (Laak, 1986),
giving a total of 7 g of PO4 (or 2.3 g of P) per per-
son per day. Most CWS use a septic tank or similar
settlement chamber as a primary treatment, to
remove half of the phosphorus in the accumulated
sludge (Frost, unpublished data), corresponding to
a discharge to the CWS of 1.2 g per person per day
(440 g P yÿ1). Assuming the land area required to
treat waste water from one person is 5 m2 and the
depth of the bed is 0.6 m (EC/EWPCA, 1990), 3 m3
or 4.5 tonnes of substrate would be needed to treat
the e‚uent from one person. If saturation is
reached at 0.7 g P kgÿ1 (Fig. 3b), this amount of
shale could adsorb 3.15 kg P, which in theory
suggests a lifetime of 7 years. Alternatively, adopt-
ing the design guidelines from Norway, with a land
area of 10 m2 per person and depth of 0.9 m
(Maehlum et al., 1995), the life of the system could
be increased to over 20 years. Such calculations
necessarily assume that ¯ow through the substrate
is well-distributed to avoid channelling, but never-
theless they indicate that considerable longevity
may be possible in a well-designed ¯ow regime.
A comparison between removal eciency and life
expectancy of the systems is dicult because of the
di€erences in the operational parameters such as
the loading rate, the age of the system, the hydrau-
lic design (subsurface, overland or vertical ¯ow),
the type of pretreatment unit used, the condition of
the vegetation and the physical and chemical prop-
erties of the substrate. However, one of the features
which is common to all CWS designed for P
removal is that P accumulates in the substrate and
plants until they are both saturated (Kadlec and
Knight, 1996; Vymazal et al., 1998).
Results from the literature indicate a typical life
span of CWS for P removal of only 2±5 years.
Pulverised fuel ash in the Valley®eld pilot plant in
south Scotland (Wolstenholme and Bayes, 1990), in-
itially produced a 60 % reduction in the applied
load, but this began to decline after 2 years of ser-
vice. Poor P removal eciency (20±40%) was also
observed in Danish CWS and it was concluded that
only systems with loading rates of less than
2 cm dÿ1 showed removals of over 50 % (Schierup
et al., 1990). Gravel-based systems in Richmond,
Australia showed a decline in removal eciency
after only 1±2 yr of operation (Mann and Bavor,
1993). Kadlec and Knight (1996) showed that initial
P removal rates from wetland systems in the U.S.A.
are often in excess of 90 % but decline sharply after
only 4±5 yr of cumulative P additions. Systems con-
structed in Norway, using LECA of extremely high
P adsorptive capacity (4 kg mÿ3) indicated promis-
ing P removal of r95% after 2±3 yr of operation;
suggestions have been made that in theory, the life
time of the systems could be more than two decades
(Maehlum et al., 1995), but that it will depend on
the plants' root development and the e€ects that
this could have on P removal. The longevity of a
CWS could be enhanced by designing a multistage
system where a gravel CWS (either vertical or hori-
zontal) is followed by a shale-based unit (Drizo,
1998). Such an arrangement would result in the sus-
pended solids being retained in the gravel and thus
avoid the blocking of the pathways of ¯ow through
the plate-like shale material. This substrate could be
excavated on an annual basis, which would result in
an increase of the P retention and prolong the life
expectancy of the main planted CWS.
 CWS substrates for P removal
CONCLUSIONS
The results showed that, of the seven substrates
examined, shale had the best combination of prop-
erties for use in CWS. It had the greatest potential
for P removal, with a P adsorption capacity as high
as 650±730 mg P kgÿ1. In addition, XRFS analysis
of the shale surface (after being exposed to high P
concentration for a period of 80 days) indicated
enhanced P removal via deposition on the material.
The physical and chemical properties examined
(pH, CEC, hydraulic conductivity, porosity, surface
area) were not strongly linked with the observed P
adsorption capacity and the di€erences between ma-
terials are better explored by direct observation of
P adsorption in laboratory experiments.
Shale shows promise as a substrate for con-
structed wetland systems, using low loading rates.
These results suggest it may have a potential life-
time of 20 years in such systems; however, data
from pilot-scale and full-scale CWS will be needed
to con®rm that the lifetime is of this order.
AcknowledgementsÐWe wish to thank Mr F. Daunt, Dr
P. Hiley (Yorkshire Water plc) and Professor Y. Comeau
(CGM Department, Ecole Polytechnique, Montreal,
Quebec) for their great help in the preparation of this
manuscript; Mr R. Speirs of Scottish Agricultural College,
Edinburgh (SAC) for his recommendation on the choice
of the substrates at the outset of the research; Mr J.
MacGowan of SAC for providing the equipment for P
analyses; Dr G. Angel and Mrs D. James, Department of
Geology, University of Edinburgh, for help with XRF
analyses; Mr J. Pearson of Wemyss Brick Co. Fife,
Scotland for the supply of shale and Mr T. Cole of Alcan
Chemicals Europe, Burntisland, Fife, Scotland for the
supply of bauxite. We acknowledge ®nancial support from
the Overseas Research Scholarship Fund of the
Committee of Vice-Chancellors and Principals of the
Universities of the United Kingdom (CVCP), the
Environmental Research Support Scheme (ERSS) of the
Soros Foundation (Central European University,
Budapest) and the Scottish Oce Agriculture and
Fisheries Department (SOAFD).
REFERENCES
Barrow N. J. (1978) The description of phosphate adsorp-
tion curves. J. Soil Sci. 29, 447±462.
Brindley G. W. and Brown G. (eds.) (1980) Crystal struc-
tures of clay minerals and their X-ray identi®cation.
Mineralogical Society Monograph No 5. Mineralogical
Society, London.
Brix H. (1994) Functions of macrophytes in constructed
wetlands. Water Sci. Technol. 29(4), 71±78.
Davis J. A. and Kent D. B. (1990) Surface complexation
modeling in aqueous geochemistry. In Mineral±Water
Interface Geochemistry, ed. P. H. Ribbe, Reviews in
Mineralogy, 23. Mineralogical Society of America,
Washington, DC USA.
Department of Environment (1980) Analysis of Raw,
Potable and Wastewaters. HMSO, London.
Department of Environment (1981) Assessment of
Biodegradability. Methods for the Examination of Waters
and Associated Materials. HMSO, London.
Drizo A., Frost C. A., Smith K. A. and Grace J. (1997)
Phosphate and ammonium removal by constructed
wetlands with horizontal subsurface ¯ow, using shale
3601
as a substrate. Water Sci. Technol. 35(5), 95±
102.
Drizo A. (1998) Phosphate and ammonium removal from
waste water, using constructed wetland systems. Ph.D.
thesis, University of Edinburgh.
EC/EWPCA (1990) European design and operations
guidelines for reed bed treatment systems. Presented to
the Conference ``Constructed Wetlands in Water
Pollution Control'', Cambridge, UK, September 1990.
Also available as WRc Report UI 17, WRc Swindon,
UK, December, 1990.
Green M. B. (1997). Experience with establishment and
operation of reed bed treatment for small communities
in the UK. In: Nutrient Cycling and Retention in
Wetlands, Vymazal J. ed. Wetlands Ecology and
Management Kluwer Academic Publishers, The
Netherlands. 4, 147±158.
Johansson L. (1997) The use of LECA (light expanded
clay aggregates) for the removal of phosphorus from
wastewater. Water Sci. Technol. 35(5), 87 ±93.
Kadlec R. H. and Knight R. L. (1996) Treatment
Wetlands. Lewis Publishers, Chelsea, MI.
Kickuth R. (1977) Degradation and incorporation of
nutrients from rural wastewaters by plant rhizosphere
under limnic conditions. In Utilization of Manure by
Land Spreading. Commission of the European
Communities, London, pp. 335±343, (EUR 5672e).
Klute A. (ed.) (1986) Methods of Soil Analysis. Part 1:
Physical and Mineralogical Methods, 2nd edn. American
Society of Agronomy, Madison, WI, USA.
Kuo J., Fox E. and McDonald S. (1992) SigmaStat User's
Manual. Jandel Scienti®c GmbH, Erkrath, Germany.
Laak R. (1986) Wastewater Engineering Design for
Unsewered Areas. Technomic Publishing, The Hague,
The Netherlands.
Maehlum T., Jenssen P. D. and Warner W. S. (1995)
Cold-climate constructed wetlands. Water Sci. Technol.
32(3), 95±101.
Mann R. A. and Bavor H. J. (1993) Phosphorus removal
in constructed wetlands using gravel and industrial
waste substrata. Water Sci. Technol. 27(5), 107±113.
McLean E. O. (1982) Soil pH and lime requirement. In
Methods of Soil Analysis. Part 2: Chemical and
Microbiological Properties, eds A. L. Page, R. H. Miller
and D. R. Keeney. American Society of Agronomy,
Madison, WI, USA.
Moshiri G. A. (ed.) (1993) Constructed Wetlands for
Water Quality Improvement. Lewis Publishers, Boca
Raton, FL, USA.
Nair P. S., Logan T. J., Sharpley A. N., Sommers L. E.,
Tabatabai M. A. and Yuan T. L. (1984) Interlaboratory
comparison of a standardized phosphorus adsorption
procedure. J. Environ. Qual. 13, 591±595.
Rhoades J. D. (1982) Cation exchange capacity. In
Methods of Soil Analysis. Part 2: Chemical and
Microbiological Properties, eds A. L. Page, R. H. Miller
and D. R. Keeney. American Society of Agronomy,
Madison, WI, USA.
Schierup H., Brix H. and Lorenzen B. (1990) Wastewater
treatment in constructed reed beds in Denmark: state of
the art. In Constructed Wetlands in Water Pollution
Control: Advances in Water Pollution Control No 11, eds
P. F. Cooper and B. C. Findlater, p. 495. Pergamon
Press, Oxford.
Schreijer M., Kampf R., Toet S., Verhoeven J. (1997) The
use of constructed wetlands to upgrade treated sewage
e‚uents before discharge to natural surface water in
Texel Island, The Netherlands: Pilot Study. Water Sci.
Technol. 35(5), 231±239.
Stuanes A. O. (1982) Phosphorus sorption by soils: a
review. In Alternative Waste Water Treatments. Low-
Cost Small Systems, Research and Development.
Proceedings of Conference, Oslo, Norway, September 7±
3602 A. Drizo et al.

10, 1981, eds A. S. Eikum and R. W. Seabloom, p. 145.
D. Reidel Publishing, Norway.
Vymazal J. (1995) Constructed wetlands for wastewater
treatment in the Czech Republic: state of the art. Water
Sci. Technol. 32(3), 357±365.
Vymazal J., Brix H., Cooper P. F., Green M. B. and
Zhu T., Jenssen P. D., Maehlum T. and Krogstad T.
Haberl R. (1998) Constructed Wetlands for Wastewater
Treatment in Europe. Backhuys Publishers, Leiden, The
Netherlands.
Wolstenholme R. and Bayes C. D. (1990) An evalu-
ation of nutrient removal by the reed bed treatment
system at Valley®eld, Fife, Scotland. In Constructed
Wetlands in Water Pollution Control: Advances in
Water Pollution Control No 11, eds P. F. Cooper and
B. C. Findlater, pp. 139±148. Pergamon Press,
Oxford.
(1997) Phosphorus sorption and chemical characteristics
of lightweight aggregates (LWA): potential ®lter media
in treatment wetlands. Water Sci. Technol. 35(5), 103±
108.