F A C U L T Y O F S C I E N C E

U N I V E R S I T Y O F C O P E N H A G E N

PhD Thesis Gry Lyngsie

Sorbents for phosphate removal from agricultural

drainage water

Academic advisor: Hans Christian Bruun Hansen, University of Copenhagen

Co-advisors: Ole K. Borggaard, University of Copenhagen

Chad J. Penn, Oklahoma State University

Submitted: October 2013

ii

Name of department: Department of Plant and Environmental Sciences

Faculty of Science

University of Copenhagen

Author: Gry Lyngsie

Titel: Sorbents for phosphate removal from agricultural drainage water

Academic advisor: Hans Christian Bruun Hansen, University of Copenhagen

Co-advisors: Ole K. Borggaard, University of Copenhagen

Chad J. Penn, Oklahoma State University

Submitted: October 2013

Front-page picture by the author

iii

Preface

This PhD thesis presents a research study carried out as a collaboration between University of

Copenhagen and Oklahoma State University, between July 2010 and September 2013. The study

was conducted within the research project SupremeTech funded by The Danish Council for

Strategic Research (grant no. 09-067280).

During my study, I was enrolled at Department of Plant and Environmental Sciences, Faculty of

Science, University of Copenhagen (KU), which was my main workplace. My advisors were

Professor Hans Christian Bruun Hansen and Professor Emeritus Ole K. Borggaard. Part of the

experimental work was carried out at KU and part of it was done at Oklahoma State University

(OSU), where I worked with Associate Professor Chad Penn at Department of Plant and Soil

Science for six months. I returned to OSU for five weeks in spring 2013 to write up some of the

experimental work carried out the year before.

This thesis comprises introductory chapters about the focus of the thesis (Chapter 1), phosphorus

(P) in soil and transport paths to the aquatic environment (Chapter 2) and technologies for removing

non-point P pollution (Chapter 2), followed by a chapter on phosphate sorbing materials (Chapter

4). Three research papers, which form the core of the thesis, are presented in a results section

(Chapter 5-7), and the thesis ends with a general discussion of the main findings and their

perspectives (Chapter 8-10).

Gry Lyngsie

Copenhagen, October, 2013

iv

Acknowledgments

First I would like to acknowledge my two supervises Hans Christian Bruun Hansen and Ole K.

Borggaard for support, guidance and valuable discussions. A special thanks to Ole who, despite of

his retirement, is a rock in the sea; always answering mails, taking time to explain and discuss my

many questions and never letting a thing rest until he is satisfied. For all his efforts I am truly

grateful.

I would also like thank my external adviser Chad Penn for giving me the opportunity to visit

Oklahoma twice, for his guidance and discussions and for giving me completely free reign in the

lab - your enthusiasm, energy and confidence always provide a much appreciated boost. I would

also like to thank Research Specialist Stuart Wilson at OSU, and his wonderful family, for taking

such good care of me from the moment I arrived in Oklahoma. You guys made it easy to be many

miles away from home. I would also like to thank the boys in Soil Chemistry lab at OSU for vivid

discussions about science and culture, and for making me laugh several times a day.

Next, I would like to thank the technical staff at the Section for Environmental Chemistry and

Physics, KU, for their help, guidance and patience; Bente Postvang for helping with the screening

experiment; Section colleagues, especially the other PhD students and Post docs who have become

dear friends.

A special thanks to my former advisor Professor Henrik Breuning-Madsen who is never letting me

forget why I love science so much and for always having an idea for a new project. Finally, I would

like to thank my family and friends for their endless support and for reminding me of the fun world

outside science.

v

Abstract

Eutrophication of lakes and streams, due to phosphorus (P) coming from drained farmlands, is a

serious problem in areas with intensive agriculture as in Denmark. Installation of P sorbing filters at

drain outlets has been proposed as a solution. Efficient sorbents to be used as such filters must

possess high affinity to retain orthophosphate (phosphate) at low concentrations. In addition, high

phosphate sorption capacity, fast bonding and low desorption are necessary. This PhD study

therefore seeks to identify phosphate sorbing materials (PSMs) that are capable of removing and

retaining phosphate at low concentrations and with short reaction times. Additionally, the aim of

this thesis is to get a better understanding of the sorption reactions, e.g. surface complexation and/or

precipitation, related to different types of commercially available PSMs.

Five potential filter materials (Filtralite®P, limestone, calcinated diatomaceous earth, shell-sand and

iron-oxide based CFH) fractionated in four particle size intervals were investigated under field

relevant phosphate concentrations (0-161 µM) and retention times of 0-24 min. For the same

particle size sorption decreased on the order: CFH ≈ Filtralite®P > limestone > shell-sand ≈

calcinated diatomaceous earth. The finest CFH and Filtralite®P fractions (0.05-0.5 mm) were found

to be the best sorbents with a phosphate retention of ≥ 90% from an initial concentration of 161 µM

corresponding to 14.5 µmol/kg sorbed within 24 min. Both PSMs were also capable of retaining

≥90% of phosphate from a 16 µM solution within 1.5 min. However, only the finest CFH fraction

was also able to retain ≥90% of phosphate sorbed from the 16 µM solution when subjected to 4

times desorption sequences with 6 mM KNO3.

Filtralite®P and CFH was also tested in a flow-through setting. The flow-through investigation

showed that retention times between 0.5-9 min and inlet phosphate concentrations between 1.6 and

32 µM influenced the materials phosphate sorption capacity and affinity. CFH shows up to 10 times

higher phosphate sorption capacity and affinity compared Filtralite®P. The difference is especially

pronounced at the low phosphate concentrations (1.6 and 3.2 µM). CFH released less than 10% of

previously sorbed phosphate compared to Filtralite®P, which released ≥ 35% in the desorption

experiments. Both materials’ phosphate sorption capacity and affinity are highly dependent on the

phosphate inlet concentration, which illustrates how important it is to test a PSM using field

relevant concentrations. Furthermore, phosphate sorption by Filtralite®P was also positively

correlated to retention time. This was not the case for CFH, indicating that CFH is capable of

removing phosphate even at high flow rates.

In order to improve the understanding of the phosphate sorption reactions and kinetics for different

types of commercial available PSMs, CFH, Filtralite®P and Limestone were studied by means of

isothermal titration calorimetry (ITC), sorption isotherms, sequential extractions and SEM-EDS.

For CFH, the results indicate formation of strong Fe-P bonding under formation of surface

vi

complexes on outer surfaces followed by slow migration into interior sorption sites. For Filtralite-P

and Limestone, the phosphate sorption strongly depended on reaction time and pH, probably

because the phosphate retention was due to formation of Ca-phosphate precipitates.

In conclusion, of the five materials investigated the results from phosphate sorption and desorption

studies clearly demonstrate that regarding phosphate sorption affinity, capacity and reactivity, CFH

is superior as a filter material compared to the other tested materials.

vii

Resumé (Danish abstract)

Eutrofiering af søer og vandløb på grund af fosfor (P), der stammer fra drænede landbrugsområder,

er et alvorligt problem i områder med intensivt landbrug. Installering af P sorberende filtre for

enden af hoveddræn kan være en løsning. Den anvendte sorbent i sådanne filtre skal have en høj P

bindingsaffinitet for at fastholde ortho-fosfat (fosfat) ved relativt lave koncentrationer. Desuden er

en stor fosfat sorptionkapacitet og lav desorption nødvendig. Dette ph.d.-studie søger derfor at

identificere fosfat sorberende materialer (PSM), der er i stand til at fjerne og fastholde fosfat fra

drænvand ved lave koncentrationer og ved hurtige retentionstider. Derudover er formålet med

studiet at få en bedre forståelse af sorptions reaktioner og kinetikken relateret til forskellige typer af

kommercielle tilgængelige PSM.

Fem mulige filtermaterialer (Filtralite®P, kalksten, kalcineret moler, knuste muslinge skaller og

jern-oxid baseret CFH) fraktioneret i fire partikelstørrelsesintervaller er blevet undersøgt ved

relevante fosfat koncentrationer (0-161 µM) og retentionstider på 0-24 min. For samme

partikelstørrelse sorption faldt på ordren: CFH ≈ Filtralite®P> kalksten> kalcineret diatomejord ≈

shell-sand De bedste sorbenter var de fineste CFH og Filtralite®P fraktioner (0.05-0.5 mm) med en

fosfat tilbageholdelse kapacitet på ≥ 90 % ved en initial koncentration på 161 µM, svarende til en

sorption på 14.5 µmol /kg inden for 24 min. Begge sorbenter var yderligere i stand til at fastholde ≥

90 % af fosfat i en 16 µM opløsning inden for 1 ½ min. Men kun de fineste CFH fraktioner var også

i stand til at tilbageholde ≥ 90 % af det sorberede fosfat fra 16 µM opløsningen når de udsattes for 4

gange desorptions sekvenser med 6 mM KNO3. De rasterende tre testede materialer

Filtralite®P og CFH blev yderligere undersøgelse i en flow-through opstilling. Undersøgelsen viste

at retentionstid mellem ½ -9 min og fosfat indløbskoncentration på 1.6, 3.2, 16 og 32 µM influerede

på materialernes fosfat sorptionskapacitet og affinitet. CFH udviste op til 10 gange højere fosfat

sorptionskapacitet og affinitet sammenlignet med Filtralite®P. Forskellen var især markant ved de

lave baggrunds indløbskoncentrationer (1.6 og 3.2 µM ). CFH desorberer endvidere < 10 % af

tidligere sorberet fosfat i forhold til Filtralite®P, som frigiver ≥ 35% i desorptions eksperimenterne.

Begge materialers fosfat sorptionskapacitet og affinitet er meget afhængig af indløbskoncentration,

dette illustrerer, hvor vigtigt det er at teste en PSM under felt relevante fosfatkoncentrationer.

Endvidere var fosfat sorption af Filtralite®P også positivt korreleret til retentionstiden. Dette var

ikke tilfældet for CFH hvilket indikerer at CFH vil kunne fjerne fosfat selv ved høje

strømningshastigheder.

For at fremme forståelsen af sorptions reaktioner og kinetik, i forbindelse med forskellige typer af

kommercielt tilgængelige PSM, er reaktioner involveret i fosfat sorption til Ca/Mg-oxid baserede

Filtralite®P, CFH og kalksten undersøgt, ved hjælp af isotermtitreringskalorimetri (ITC), sorptions

viii

isotermer, sekventielle ekstraktioner og SEM-EDS. For CFH viser resultaterne dannelse af en stærk

Fe-P binding under dannelse af overflade kompleksation på udvendige overflader, efterfulgt af

langsom migration til indvendige sorptionspladser. For Filtralite®P og kalksten, er fosfat sorption

stærkt afhængig af reaktionstiden og pH, formentlig fordi fosfat tilbageholdelse er betinget af

dannelse af Ca-phosphat udfældninger.

På baggrund af de opnåede resultater i dette studie af fosfat sorption og desorption, kan det

konkluderes at, af de fem undersøgte materialer, er den jernoxid baserede CFH overlegen som

filtermateriale i forhold til de andre testede materialer.

ix

List of Publications (peer-reviewed papers)

Lyngsie, G., Borggaard, O.K., Hansen, H.C.B. Testing phosphate sorption efficiency of five

potential drainage water filter materials under field-relevant conditions (In print Water Research).

Lyngsie, G., Penn, C.J., Hansen, H.C.B., Borggaard, O.K. Phosphate sorption by three potential

filter materials as assessed by isothermal titration calorimetry (Submitted to Journal of

Environmental Management).

Lyngsie, G., Penn, C.J., Pedersen, H.L., Borggaard, O.K., Hansen, H.C.B. Modelling of phosphate

retention by two promising drainage filter materials calcium-based Filtralite-P and an iron-rich CFH

under flow-through conditions (Draft manuscript).

Presentations

2013 Lyngsie, G., Penn, C., Borggaard, O.K., Hansen, H.C.B. Sorbents for phosphate

removal from agricultural drainage water. 7th International Phosphorus Workshop,

Uppsala, Sweden

Sorbents for phosphate retention in landscape filters – chemical (WP1). SupremeTech

PhD workshop, Copenhagen.

Sorbents for phosphate retention in landscape filters. PLEN PhD workshop,

Copenhagen.

2012 Lyngsie, G., Penn, C., Borggaard, O.K., Hansen, H.C.B. ‘Reactions and mechanisms

of phosphate sorption by two potential filter materials assessed by isothermal titration

calorimetry’. 13th International Conference on Wetland Systems for Water Pollution

Control, Perth, Australia

Sorption properties for phosphate sorbing materials. SupremeTech workshop, Århus.

2010 Phosphate sorption properties of filter materials (WP1) SupremeTech workshop,

Århus.

x

Abbreviations

ACP Amorphous calcium phosphates

CDE Calcinated diatomaceous earth

ITC Isothermal titration calorimetry

P Phosphorus

Phosphate Otho-phosphate

Pmax Phosphate sorption maximum

PSM Phosphate sorbing material

RT Retention time

SSA Specific surface area

xi

Table of Contents

Preface ................................................................................................................................................. iii

Acknowledgments ............................................................................................................................... iv

Abstract ................................................................................................................................................ v

Resumé (Danish abstract) .................................................................................................................. vii

List of Publications (peer-reviewed papers) ....................................................................................... ix

Presentations ....................................................................................................................................... ix

Abbreviations ....................................................................................................................................... x

Introduction

Chapter 1 - Focus of thesis ................................................................................................................... 2

Chapter 2- P in soil and transport paths to aquatic environment ......................................................... 4

2.1 Agricultural application of manure and fertilizer ....................................................................... 4

2.2 Phosphorus forms in soil ............................................................................................................ 5

2.2.1 Availability of phosphate ..................................................................................................... 6

2.3 Transport of phosphate to the aquatic environment ................................................................... 6

2.3.1 Distribution of phosphate in drainage water through-out the year ...................................... 7

Chapter 3 – Technologies for the removal of non-point P pollution ................................................... 9

Chapter 4 – Phosphate sorbing materials (PSM) ............................................................................... 12

4.1 Sorption of phosphate to PSMs ................................................................................................ 12

4.1.1 Phosphate sorption to metal oxides ................................................................................... 13

4.1.2 Phosphate sorption to CaCO3 ............................................................................................ 14

4.2 Different test approached of PSM ............................................................................................ 14

4.2.1 Batch tests .......................................................................................................................... 15

4.2.2 Flow-through tests ............................................................................................................. 15

4.2.2 Desorption tests ................................................................................................................. 16

4.2.4 Field tests ........................................................................................................................... 16

4.3 Choosing the right material ...................................................................................................... 16

Results

Chapter 5 – Paper I............................................................................................................................. 20

Chapter 6 – Paper II ........................................................................................................................... 43

Chapter 7 – Paper III .......................................................................................................................... 77

xii

Main findings

Chapter 8 – Combined discussion .................................................................................................... 106

8.1 Influence of retention time on PSM sorption ......................................................................... 106

8.2 Influence of pH on PSM sorption........................................................................................... 107

8.3 Influence of the background electrolyte on PSM sorption ..................................................... 107

8.4 Influence of phosphate concentration on PSM sorption ........................................................ 108

8.5 Retention of sorbed phosphate by the PSMs .......................................................................... 109

8.6 Influence of test method on PSM sorption ............................................................................. 109

Chapter 9 - Conclusion .................................................................................................................... 111

Chapter 10 – Perspective.................................................................................................................. 112

References ........................................................................................................................................ 113

1

Introduction

2

Chapter 1 - Focus of thesis Phosphorus (P) is an essential element for life and a macronutrient for plants, microorganisms and

animals. In order to maximize crop yields, agricultural soils in industrialized countries have

received massive inputs of P during the last century. High inputs have resulted in P accumulation in

soil and potential P transport to the aquatic environment. Excess P in aquatic environments can lead

to excess algal growth, resulting in oxygen depletion and fish death. Thus, P concentrations in lakes

higher than ~50 µg/L (1.6 µM) may cause eutrophication (Smil, 2000). In Denmark, P loss

originating from agriculture is a main contributor of P in surfaces water bodies (Kronvang et al.,

2009a). On average, 1400 kg P/ha has accumulated in arable Danish soils during the last century

(Poulsen and Rubaek, 2005), and P is continuously accumulating in areas of high livestock density

due to application of animal manure (Kronvang et al., 2009a). Manure application is regulated by

the content of nitrogen (N) and may therefore result in over-fertilization with P (Tarkalson and

Leytem, 2009). However, a P-enriched soil only becomes an environmental problem when

connected to the aquatic environment by an effective transport pathway such as artificial drains

(Heathwaite et al., 2003). In Denmark, more than half of the farmland is artificially drained

(Breuning-Madsen et al., 1992) and one-third of P losses occur via leaching through the soil profile

to subsurface drains (Kronvang et al., 2009a). Despite substantial efforts to reduce the transport of

P from agricultural land to the aquatic environment, it is still a serious and costly problem in

Denmark and elsewhere (Ballantine and Tanner, 2010; Buda et al., 2012; Delgado and Scalenghe,

2008). Thus, reaching the goal of good water quality as stated in the EU Water Framework

Directive (WFD) requires a substantial reduction of P loss from farmland in Europe and elsewhere

(Kaasik et al., 2008; Søndergaard et al., 2005). Hence, this thesis focuses on reducing P,

specifically otho-phosphate, losses leached through subsurface drains in agricultural land.

Several types of technology addressing reduction of nonpoint phosphate losses are currently

established e.g. the end-of-pipe approach carried out in connection with a small-scale constructed

wetland (Koskiaho et al., 2003; Reinhardt et al., 2005), the use of flow-through filter structures in

ditches (Penn et al., 2007) and the diffuse approach where a phosphate reactive material is used as a

reactive barrier placed in a ditch or around drain pipe (Buda et al., 2012; Groenenberg et al., 2013).

The common denominator for these different approaches is that they all need a phosphate removing

agent in order to be effective. Hence, a great variety of different types of phosphate sorbing

materials (PSMs) have been tested (Ballantine and Tanner, 2010; Cucarella and Renman, 2009;

Johansson Westholm, 2006; Penn et al., 2007; Vohla et al., 2011). Most studies on these materials

have been carried out in relation to high phosphate concentrations in waste-waters (Cucarella and

Renman, 2009; Johansson Westholm, 2006; Penn et al., 2007; Vohla et al., 2011). However, a

direct transfer of the PSM data obtained from these high phosphate concentration studies to the low

concentrations in drainage water is not meaningful. Hence, the main aim of this thesis is to

3

identify PSMs that are capable of removing and retaining phosphate both at low concentrations and

during fast contact times in order to remove it from drainage water. This thesis will focus solely on

the phosphate sorbing aspect of PSMs. The hydraulic conductivity of a PSM has proven to be very

important in a phosphate removal structure (Penn et al., 2012) but test of the PSMs hydraulic

properties is beyond the scope of this thesis although the behavior of different particle sizes was

investigated. In addition, the aim of the study described in this thesis is to further understand the

sorption reactions relating to different types of PSMs and to improve the general understanding of

PSM behavior. These aims were investigated through laboratory experiments. The thesis is built on

three submitted manuscripts based on the following three objectives:

The first objective was to screen five selected commercial available PSMs at four specific particle

size fractions for their ability to sorb and retain phosphate, at low concentration with short contact

time in a batch set-up. The hypothesis was that some of the tested PSM would be capable of

removing and retaining phosphate at low concentrations and that particle size would have an

influence on the sorption capacity, affinity and reactivity (Paper I).

The second objective was to study how PSMs behaved in a flow-through setting, regarding

affinity, reactivity, capacity and desorption. The two best performing PSMs from the screening

study in three particle size fractions were treated with varied phosphate inlet concentrations (< 32

µM) and different retention times (< 10 min). The hypothesis was that the retention time and

phosphate inlet concentration influenced the sorption capacity and affinity of the PSMs (Paper II).

The third objective was to study the sorption reaction of three different PSMs in order to better

understand their different behavior in regard to sorption and desorption. The hypothesis was that the

sorption reactions varied among Ca-based and Fe-based PSMs but also between different types of

Ca-based PSMs (Paper III).

4

Chapter 2- P in soil and transport paths to aquatic environment

This chapter introduces the general concepts of P behavior in soil and links it to the transport of P to

the aquatic environment. Fig. 2.1 shows different sources and potential pathways for P to the

aquatic environment. Focus in this thesis is on agricultural P losses, identified by the red circle.

Fig. 2-1 Modified schematic diagram showing different point and non-point P sources to the aquatic environment and

different P storages in the landscape (Kronvang et al. 2009a). The red circle indicates the agricultural contributions, the

focus area of this thesis.

2.1 Agricultural application of manure and fertilizer

In many catchments around the world, agriculture is the major contributor of P to surface waters

and P in agricultural soils (a non-point or diffuse source) is considered to be a main source of P

losses to the aquatic environment (Kronvang et al., 2009b). Inputs of nutrients, such as P, are

necessary in farming systems as nutrients are critical in maintaining soil fertility and increasing crop

and forage productivity. Where nutrients are in deficit soil fertility can decline, while with an excess

of nutrients necessary for plant growth there is a risk of nutrient transportation to water, which is the

case in OECD countries (OECD, 2013a).

5

Overall, OECD agricultural P surpluses have in fact shown a continuous downward trend from

1990 to 2009, both in absolute amount and in terms of P surplus per hectare agricultural land. The

average OECD P application surplus to agriculture land decreased from 13 kg P/ha in 1990-92 to 9

kg P/ha in 1998-2000 and to 7 kg P/ha in 2007-09, giving an average annual decrease of 4% in the

first period and 5% in the second. The reduction in Denmark was approximately the same as the

average in OECD (OECD, 2013a). These numbers reflects an overall improvement in nutrient use

efficiency by farmers and slower growth in agriculture production for many countries in the 2000s.

To a large extent, especially over the past decade, the OECD reduction in P surplus is a

consequence of the higher fertilizer prices and farmers realizing that their soil had high levels of

accumulated P. Crops and pasture can draw from these reserves without further application, at least

for a number of years. Despite the overall improvement in lowering P surpluses, intensity levels per

hectare in arable land remains at very high levels in terms of their potential to cause environmental

damage. In most OECD member states agriculture is responsible for more than a third of the total

nutrient discharge to surface and coastal waters (OECD, 2013b), where background (or natural) loss

of P is around 0.1 kg/ha depending on the concentration in the sediment and rocks (OECD, 2013a).

Hence, it is likely that there will be a considerable time lag for many countries between reductions

in P surpluses leading to lower P concentrations in water bodies (OECD, 2013b). Indeed, P

concentrations in rivers and lakes could continue to rise over decades also in the future.

2.2 Phosphorus forms in soil

P can be leached from the agricultural land to the aquatic environment through drain-pipes, ditches

and the upper sections of groundwater. P can be leached either as particulate, organic and inorganic

dissolved P and can exist in drainage water in these various forms.

Organic P content of soils mainly consists of inositol phosphates, nucleic acids, phospholipids, and

sugar-based phosphates such as glycerophosphates, mentioned in order of abundance. Dissolved

organic material with incorporated P is produced via microbial decomposing of organic material

(organic manure, plant residuals and dead roots). The decomposition is controlled by the pH,

nutrients, soil temperature, oxygen and water content and the composition of the organic material.

Dissolved organic material is transported by the percolating water but can also be adsorbed by the

soil’s humus fraction and be a part of an equilibrium controlled by the concentration of dissolved

organic material in the soil solution. Dissolved organic material can further be sorbed to the soil’s

mineral fraction (clay particles) in a reaction with iron and aluminum and other polyvalent cations.

(Andersen et al., 2006)

Dissolved inorganic P (Pi), which is the focus of this thesis, can be transported from the top soil to

the drains through either the macropores or through the soil matrix. Pi in the natural environment

consists of free orthophosphate compounds. These species are mainly H2PO4- and HPO4

2- but can

6

occur as H3PO4 and PO43-

in strongly acid and strongly alkaline environments, respectively. Here,

orthophosphate species are simply referred to as phosphate.

2.2.1 Availability of phosphate

Phosphate is in equilibrium between solid and solution phases by sorption and desorption processes.

Desorption of phosphate occurs by diffusion from the solid phase to the solution in response to

changes in the soil solution, e.g. decreased phosphate concentration due to plant uptake or leaching.

Soil phosphate is often divided into three pools; phosphate in solution (dissolved phosphate),

phosphate loosely adsorbed and readily exchangeable (labile phosphate), and phosphate strongly

sorbed to soil particles (non-labile phosphate). Phosphate in solution and labile phosphate is

available for plant and organism uptake, whereas non-labile phosphate is seen as a sink, which only

becomes available to plants and organisms by depletion of the two other pools. The dissolved

phosphate pool is generally very low in soil, constituting ≤0.01% of total P. The non-labile pool is

by far the largest form of total P (Mengel et al., 2001). The non-labile pool increases as time

proceeds due to continuous slow migration of phosphate within the soil particles (Barrow, 1983).

Hence, inputs of P are needed in cultivated soils as most soil P becomes unavailable to plants over

time (Barrow, 1983).

2.3 Transport of phosphate to the aquatic environment

In order for the soil P to possess a risk for the aquatic environment there needs to be a connection,

or pathway, between the soil and the aquatic environment. Tile drains and ditches connect fields to

recipient waters and act as highways for both soluble and particulate P, thereby reducing the time in

which the soil water is in contact with the soil matrix compared to undrained systems.

Phosphate can be leached from the topsoil to the drain system via the pathways created by

macropores and through the soil matrix. For macropores it is the topsoil’s water soluble phosphate

that gives an indication of the leaching potential. For transportation through the matrix it is the

soil’s phosphate binding capacity, from the surface to the drain-pipes, which matters. Factors

promoting phosphate leaching via subsurface pathways are dependent on soil characteristics (e.g.

low phosphate sorption capacity, soil conducive to promoting preferential flow pathways, reducing

conditions etc.), agronomic management practices (e.g. amount and form of P added and

application method including liming strategies) and climatic conditions (Gburek et al., 2005).

As clay minerals exhibit strong sorption with phosphate, leaching from fine-textured soils should

therefore seem less pronounced than in less adsorptive sandy soils (McDowell et al., 2001). Fine-

textured soils, however, have an uneven distribution of different pore size classes and preferential

flow paths, in contrast with uniform flow through a larger volume of the matrix in sandy soils.

Preferential flow paths in fine-textured soils have been shown to permit direct transport of

phosphate placed on the surface through the soil profile, bypassing the soil matrix and soil

7

adsorption sites (Kleinman et al., 2009). Preferential flow leads to chemical non-equilibrium. This

is due to insufficient reaction time of phosphate with soil sorption sites either due to low chemical

kinetics or slow diffusion when water is rapidly flowing by. Physical heterogeneity factors such as

the degree of preferential flow pathways, as well as the distribution of Fe and Al oxides and clay

minerals in the soil, lead to layers of different reactivity within soil (Hansen et al., 1999). It should

be kept in mind that it is in fact the fine-textured soils that are drained.

2.3.1 Distribution of phosphate in drainage water through-out the year

There is a large temporal variation in P concentration coupled with an even larger temporal

variation in instantaneous flow in streams (Gburek et al., 2005). Andersen et al. (2006) found that

the median phosphate concentration in five drains in Denmark was around 1.6 µM. This is in line

with other measurement of phosphates in drainage water in Denmark (Poulsen and Rubaek, 2005)

and is at the same level as the phosphate concentration in streams (Andersen et al., 2006). 10-20%

of the Danish drains have much higher concentrations (Andersen et al., 2006; Poulsen and Rubaek,

2005) and a few drains have very high phosphate concentration up to 42 µM (Andersen et al., 2006;

Jørgensen et al., 2001). However, Penn et al. (2007) found more than four times this concentrations

in Maryland, US.

The seasonal variation in phosphate concentrations in stream flow can be large and is dependent on

catchment size and climatic conditions. The highest concentration of phosphate is often found

during the driest part of the year due to less dilution of phosphate delivered from point sources.

However, a tendency for higher phosphate concentration during winter time than during summer

time on fine-textured soils and the opposite for coarser soils has been noted (Gburek et al., 2005). A

high hourly variation in both phosphate concentration and flow rate is seen during rainstorms (Fig.

2-2) and fast frost-thaw transitions (Cade-Menun et al., 2013; Grant et al., 1996; Johnes, 2007; Penn

et al., 2012). In fact Penn et al. (2012) found that 75% of the phosphate delivered from a 320 ha

suburban watershed was delivered during the six largest rainfall events over a 5 month period.

8

Traveling time associated with water moving from the soil surface to water bodies vary according

to the related flow processes. Fast flow processes that by-pass most of the phosphate sorbing soil

matrix are characterized by a relatively short travel time, typically in the order of minutes to hours.

Travel time from the soil surface to drains may be only a few minutes (Gburek et al., 2005). In

cases such as this, the water only interacts with a thin soil layer close to the surface and most likely

it has only minimal contact with the surface (matrix) material of the macropores through which it

flows. This results in the chemical non-equilibrium as described in the previous section.

In summary the drainage system has low base-flow phosphate concentrations when the flow rate is

low and contact time thereby high, compared with rainfall events with high phosphate concentration

and a fast water flow resulting in short contact time.

Fig. 2-2 Concentration of different P fractions and runoff (Q) during two single storm events at different

times of the year in a Danish lowland stream draning a 10 km2 catchment. DRP: Dissolved Reactive P; PIP:

Particulate Inorganic P; POP:Particulate Organic P; PP: Particulate P; DOP: Dissolved Organic P. Figure

modified from Kronvang et al., 2003

Q

Q

9

Chapter 3 – Technologies for the removal of non-point P pollution

Despite the widespread implementation of conservation and nutrient management strategies to

control agricultural P losses, a lack of water quality improvement in inland waters and estuaries is

still seen (Buda et al., 2012). This can be explained by the lag phase described in section 2.1.

A wide variety of phosphate sorbing technologies have been conceived through the last two

decades, ranging from preventive measures that focus on P immobilization in soils to remedial

measures that utilize filters to remove phosphate from runoff water. In contrast to the soil amending

approaches that only immobilize the phosphate but do not remove it from the system, the remedial

Fig. 3-1 Example of end-of-pip remedial strategy targeting condense P source.

Constructed wetland with different matrix flows. 1a) vertical downward flow b) vertical

upward flow, c) horizontal flow, d) snapshot from the field after implementation of

different matrix structures. Figure modified form www.supremetech.dk , d) obtained from

Folkebladet, onsdag 15. Maj 2013.

10

technologies, which will be focused on here, address the phosphate after it has been mobilized to

the agriculture drainage water.

There are two overall strategies that are typically used as remedial technologies. The first strategy is

a so call end-of-pipe approach targeting condense agricultural phosphate sources such as main tile

drains, barnyard and golf course runoff. This end-of-pipe method consists of a variety of filter

technologies such as constructed wetland with matrix flow (Fig 3-1), closed well filters (Fig. 3-2),

box ditch filets (Penn et al., 2012), and large dam ditches (Bryant et al., 2012), amongst others. The

second remediation strategy focusses on the filtration of phosphate from diffuse flows such as soil

water and shallow lateral groundwater flows on-site. Examples of potential on-site remediation

techniques are introducing a permeable reactive barrier (Fig. 3-3) or enveloping a tile drain with

PSM (Fig 3-3) (Groenenberg et al., 2013). The basic idea of both types of remedial technologies is

to control phosphate losses by intercepting the flow path with a PSM and thereby removing it from

the agricultural runoff before it enters the wider water environment.

Fig 3-2 Example of end-of-pipe remedial strategies targeting a condense P source. Two different

types of drainage well filters, red arrows indicate inflow, green arrows outflow. a) from the

impetration of the wells, photo from www.supremetech.dk, b) after implementation in the

landscape, c) cylinder filter cassette filter filled with crushed shell-sand. The water penetrates the

filter cylinder from the outer surface and exit through the bottom in the middle, d) combined

lamellar and rectangular cassette filter filled with crushed shell-sand. The lamellae are thought to

make the suspended particles settle on the bottom of the well to avoid clogging of the phosphate

sorbing filter. b-d) Photos taken by the author.

11

There are advantages and drawbacks to each of these different remedial strategies: It can been

argued that constructed wetland follows a wide range of recreational benefits such as more wild life,

which will increase the hunting prospect of an area, more insects, which would lead to more bird

life, whereas as on the disadvantage side it could be argued that constructed wetlands take up too

much space of agricultural land and they are expensive to construct. When discussing box ditches

and wells it can be argued that they are easy to maintain and do not take up much space. However,

it could be claimed that because of their smaller size they need more maintenance and can handle

less water than wetlands. Regarding the on-site remediation strategies an advantage of the

“enveloping” drainage pipes and the permeable reactive barriers is that they do not take up space in

the landscape whereas one drawback is that maintenance, if needed, is not easy. Hence, when

choosing one or the other remediation technology such pro and cons should indeed be discussed and

the choice should reflect the system regarding water regime, phosphate concentration and location

in the landscape at each specific site.

As indicated above, the common denominator for all of P sorbing technologies is that in order to be

efficient they need to contain a PSM that can chemically retain the phosphate (Ballantine and

Tanner, 2010; Buda et al., 2012; Johansson Westholm, 2006; Vohla et al., 2011). A number of

PSMs have been evaluated in the literature and their basic characteristics will be described in

Chapter 4.

Fig 3-3 Example of remedial

strategies addressing diffuse

phosphate from soil and shallow

groundwater. Implementing a

permeable reactive barrier with iron

coated sand (left) and enveloping a

tile drain with iron coated sand

(middle and right). Figure modified

form Buda et al. 2012.

12

Chapter 4 – Phosphate sorbing materials (PSM)

The overall phosphate removal capacity of a landscape P filter is strongly dependent on the sorption

characteristic of the construction e.g. size, type, and PSM used. Considerable attention must

therefore be given to the choice and composition of a suitable PSM in order to ensure optimal

phosphate sorption and retention over the long term. It is also important to understand the

underlying mechanism behind phosphate sorption to PSM in order to choose the right one for

various systems, as pointed out in a review by Klimeski et al. (2012). In the following chapter a

short introduction to the different types of PSM is given, and basic sorption reactions and typical

test approaches will be explained.

Since the 1960s scientists worldwide have carried out research on PSM (Klimeski et al., 2012).

According to their origin, PSMs are often classified into three groups: natural material, industrial

by-products, and manufacture materials (Cucarella and Renman, 2009). A wide range of natural

materials have been tested as PSM, including apatite, bauxite, dolomite, calcite/aragonite

(limestone), gravels, sand, clays, sub-soil, opoka, peat, shellsand, wollastonite, zeolites and tree

bark. Industrial by-products tested include various forms of coal ash, fly ash, water treatment

residuals, ocher, slag, blast furnace and mine residuals. Manufacture materials such as Light-weight

aggregates (LWA) or light expanded clay aggregates (LECA), Filtralite–P (LECA modified

material), iron oxides and concrete have also been tested in both laboratory and field studies

(Ballantine and Tanner, 2010; Cucarella and Renman, 2009; Johansson Westholm, 2006; Klimeski

et al., 2012; Vohla et al., 2011). The PSMs tested are often related to the materials available in the

geographical area of the investigation.

PSMs can also be classified according to their chemical composition; metals, mainly Fe/Al-

containing materials, materials containing soluble divalent alkaline earth metals (Ca and Mg) or a

mixture of the two (Klimeski et al., 2012). In this thesis the PSM will be described according to

their active sorbent, which controls the phosphate sorption reactions.

4.1 Sorption of phosphate to PSMs

Physical properties that affect the phosphate sorption to a particle include the shape, size and

porosity of the particles, all of which affect their reactive surface and hydraulic conductivity.

Materials that consist of fine particles usually have a large specific surface area (SSA) and may

have high phosphate sorption potential. Together with the pH of the solid-solution system, the PSM

chemical composition and crystallographic properties are the determining factors for the PSM

phosphate affinity and capacity. The most efficient PSMs tend to be those that contain Fe/Al

hydroxides or easily soluble Ca/Mg compounds. Their generic sorption reaction will be explained in

the following section.

13

4.1.1 Phosphate sorption to metal oxides

Adsorption describes sorption reactions taking place at the surface of particles. This can occur by

electron donor and electron acceptor interaction, electrostatic attraction or covalent bonding.

Phosphate adsorbs to Fe and Al oxides by surface complexation involving covalent bonds and

thereby forming strong complexes which are not readily susceptible to desorption (Borggaard et al.,

2005; Chardon et al., 2012; Heal et al., 2003; Strauss et al., 1997; Chapter 5-7). This adsorption is

site-specific; one or two hydroxyl groups of the metal oxide are exchanged with one or two of the

oxygen atoms of the phosphate iron resulting in a monodenate/mononuclear or bidentate/binuclear

surface complex (Fig. 4-1). The reaction is considered reversible although the release of absorbed P

may in practice be minimal (Klimeski et al., 2012). For adsorption on a solid surface, such as metal

oxide, the maximum phosphate retention is achieved when all sorption sites become occupied.

Surface complexation between Al/Fe metal oxides and phosphate is pH and concentration

dependent. Al/Fe metals oxides are positively charged at low pH and negatively charged at high pH

values. As phosphate is an anion, adsorption is greatest at low pH and decreases with increasing pH

(Klimeski et al., 2012). Still, complexes do form at high pH values because the adsorption is site-

specific. Increasing phosphate concentration in solution increases adsorption onto metal oxides until

the adsorption maximum capacity has been reached (Hiemstra and Van Riemsdijk, 1996).

Phosphate adsorption onto Fe oxides is rapid, but equilibrium is not rapidly achieved (Chardon et

al., 2012, Chapters 5-7). Adsorption can be followed by slow migration of adsorbed P via solid-

state diffusion of P towards less accessible interior sorption sites, forming inner P complexes (van

Riemsdijk et al., 1984).

Fig. 4-1 Specific

adsorption of

phosphate onto

an iron oxide

surface.

Modified from

Borggaard and

Elberling (2004)

14

Water saturated conditions, when the oxygen concentration become very low, can cause reducing

conditions where Fe(III) is reduced to Fe(II). Adsorption of phosphate is much higher to Fe(III)

oxides than to Fe(II), which tends to be soluble or sorbed onto catirons exchange sites. Reducing

conditions will therefore lead to the release of phosphate into solution and in turn, increased

solubility and leaching of phosphate upon continued reductive dissolution of Fe(III) oxides

(Heiberg et al., 2010).

4.1.2 Phosphate sorption to CaCO3

At high pH, phosphate adsorbs to surfaces of CaCO3. This is followed by precipitation of secondary

Ca-phosphates which range from amorphous calcium phosphates (Ca3(PO4)2*nH2O) to the well

crystalline hydroxyapatite (Ca5(PO4)3OH). The product precipitating is highly dependent on pH,

time of reaction and competitive ions present, both cat- and anions (Cao et al., 2007; Cao and

Harris, 2008). Phosphate adsorption onto Ca is fairly rapid (Sø et al., 2011; Chapters 5- 6) and

completed within 2-3 h (Sø et al., 2011). Precipitation of Ca-phosphates occurs in phosphate

solutions when the solubility product of the mineral is exceeded. With high amounts of solution

phosphate (e.g. when using high phosphate concentration) more Ca-phosphate can precipitate and

as the Ca concentration in solution decreases upon precipitation, more Ca will dissolve from

CaCO3. This continues as long as soluble Ca and phosphate are present in in concentration

exceeding the solubility product of the precipitate. Precipitation of Ca phosphate can be divided into

three phases; induction, rapid crystal growth and maturation in crystal growth (Grossl and Inskeep,

1991). Cao et al. (2007) and Cao and Harris (2008) found that precipitation of Ca-phosphates in a

pure system occurred within the first hour as the concentration of Ca and phosphate rapidly

declined within the first 10 minutes, followed by a slow decline until 60 min and leveling off

beyond 60 min. For poorly crystalline Ca-phosphates, the reaction may be fully reversible and a

decrease in pH, phosphate or Ca concentrations may result in the dissolution of the precipitates

(Ádám et al., 2007; Diaz et al., 1994).

The Mg-phosphate system is similar to that of Ca-phosphate. However, one of the two will

normally dominate over the other and the present of both ions in solution may have an inhibiting

effect on the Ca/Mg- phosphate precipitation (Cao et al., 2007; Cao and Harris, 2008; Ferguson and

McCarty, 1971).

4.2 Different test approached of PSM

In the design of a phosphate retaining filter bed, important parameters assessed in the laboratory

include the phosphate sorption capacity and affinity of the PSM as well as the kinetics of the

reaction. The capacity is of course linked to the amount of phosphate a PSM can retain and thus to

the effective lifespan of the PSM. Reaction kinetics relate to the effects of contact time of retained

phosphate. For applications that aim to remediate large water volumes in short contacts periods, fast

15

reaction is clearly desirable. In the first phase of material characterization, laboratory experiments

often serve to estimate retention capacity and reaction rate. Laboratory tests are commonly

performed either in a closed system as batch experiments or in an open system as in flow-through

columns (Klimeski et al., 2012). It has been emphasized by several authors that regardless of the

great amount of existing published data, the potential and efficiencies of different PSMs are very

difficult to compare directly because of the different test conditions (Johansson Westholm, 2006;

Klimeski et al., 2012; Vohla et al., 2011).

4.2.1 Batch tests

Closed batch experiments are rather easy and fast to conduct. They can be used to estimate the first

predictions of phosphate retention. In a typical batch experiment, a fixed amount of material is

shaken with a fixed amount of a phosphate solution at varying phosphate concentrations. The

estimate maximum retention can then be calculated or modeled via the phosphate sorption isotherm.

However, batch studies commonly overestimate the performance of the PSM, particularly for

precipitation-controlled P retention (Klimeski et al., 2012).

4.2.2 Flow-through tests

The flow-through set-up is more time consuming. A flow-through cell is an open system with no

accumulation of dissolved species in the reaction vessel and the test may provide a more realistic

picture of what happens in the field (Klimeski et al., 2012; Penn et al., 2011). In an open system

flowing water will remove part of the dissolved Ca2+

or Mg2+

, and the yield of the precipitates is

then no longer proportional to the mass of soluble Ca or Mg compounds initially present. Factors

that may substantially influence the outcome of flow-through tests include particle size, the PSM’s

active sorbent, initial phosphate concentration and flow rate (contact time) (Stoner et al., 2012;

Chapter 6). Time needed to achieve phosphate saturation in flow-through test may vary

considerably depending on the amount of material, its physical and chemical properties, the applied

flow rate and the initial phosphate concentration. A high initial phosphate concentration will shorten

the time needed to obtain phosphate saturation. However, as phosphate retention by sorption

mechanism is an equilibrium reaction more phosphate is retained when using high initial

concentration, i.e. the PSM’s phosphate sorption capacity may be overestimated, as mentioned

previously. A small amount of material (e.g. mixing the PSM with quartz sand) or an increase in

the flow rate are two other options for shortening the time needed to reach the PSM saturation.

However, as flow rate influence on the sorption capacity especially for Ca based PSMs (Chapter 6;

Penn et al., 2011; Stoner et al., 2012) this is not a good approach. If the phosphate saturation is not

achieved it may be possible to estimate it through modeling (Chapter 6; Penn et al., 2011; Stoner et

al., 2012).

16

4.2.2 Desorption tests

PSM desorption characteristics should be taken into consideration when estimating the PSM

phosphate sorption behavior. In landscape P filters, ambient conditions vary as described in section

2.4, and any available P sink would be affected by this. For Ca-phosphate precipitates, change in pH

and solution and phosphate concentration may dissolve the precipitates (Chapters 5-6). For metal

oxides, phosphate desorption may occur as result of changes in the phosphate concentration, the

degree of ironic saturation and ionic strength. The redox state of the system may affect the

sorption/desorption reaction for Fe-based PSMs, as described in 4.1. Desorption tests can be

conducted in both batch and flow-through settings and are usually conducted in pure water or in

phosphate free electrolyte solutions. In dissolution tests, more aggressive solvents or reducing

agents may be used.

4.2.4 Field tests

PSMs have been applied as a reactive media in landscape filter beds in both small and large scales.

The phosphate sorption capacity of the PSM usually proves to be lower in larger scale applications

than in laboratory tests (Klimeski et al., 2012). As the scale increase, obstacles to remove phosphate

are often connected to hydraulic conductivity and high flows. Precipitates, suspended solids and

organic material may clog the filter. The incoming flow also tends to form preferential pathways in

the material, resulting in a short retention time, limited contact between the phosphate and the

material and consequently lower than expected removal efficiency of the filter system (section 2.3).

4.3 Choosing the right material

In order to choose the right PSM for a system, it is important to understand the system you are

dealing with, in regard to hydraulic load, flow rate, phosphate concentration, pH etc. For instance is

it, as previously mentioned, the fine grained particles which normally have the highest P sorption

potential. However, these materials normally have a low hydraulic conductivity resulting in

overland flow and incomplete contact with the drainage water, if the filters hydraulic conductivity

does not match the system. Therefore, an ideal material should possess both high phosphate

sorption capacity and affinity as well as a percolation rate that is suitable for the given system (Cui

et al., 2008).

The lifetime or effective capacity of a PSM will, besides the sorption characteristic of the material,

also depend on hydraulic loading rate, phosphate concentration and flow dynamics (Ballantine and

Tanner, 2010; Chapter 6). Irrespective of the lifetime or phosphate sorption capacity of a PSM,

sorption of phosphate is a finite process and the recycling of phosphate saturated PSM could be a

sustainable way of dealing with the consumption of world´s limited P reserves and rising costs of P

fertilizers (Ballantine and Tanner, 2010). Phosphate-saturated PSM can be used directly as a P-rich

soil amendment or alternatively, the phosphate may be reclaimed from the material and used as a

17

nutrient specific fertilizer (Sibrell et al., 2009). However, if PSMs are being used directly as a soil

amendment, the possible toxicity of the PSM should be taken into consideration in order to avoid

contamination of the soil. For example, by-products such as steel slags can contain high

concentrations of heavy metals that can exceeds the levels permitted for disposal on land

(Ballantine and Tanner, 2010) and could therefore not be redisposed on the field.

From a practical and partly a sustainable perspective, the PSM chosen for a landscape filter should

ideally be low-cost (e.g. industrial by-products), generated locally (to reduce transport costs),

widely available in large quantities (Chardon et al., 2012), non-toxic and either be reusable once

saturated with P or able to be readily and safely disposed of (Ballantine and Tanner, 2010). In

countries such as Denmark, with limited access to relevant by-products and local reactive products

in general, it is not easy to fulfill these demands, which is why the PSMs investigated in this thesis

are mainly manufactured, or partly manufactured, products and not of all a national origin.

18

19

Results

20

Chapter 5 – Paper I

A three-step test of phosphate sorption efficiency of potential agricultural drainage filter

materials

G. Lyngsie*, O.K. Borggaard, H.C.B. Hansen

University of Copenhagen, Department of Plant and Environmental Sciences, Thorvaldsensvej 40,

DK-1871 Frederiksberg C, Denmark

*Corresponding author: chj177@alumni.ku.dk

In print Water Research

Keywords: Water quality; eutrophication; iron oxides; carbonates; farmland drainage

Abstract

Phosphorus (P) eutrophication of lakes and streams, coming from drained farmlands, is a serious

problem in areas with intensive agriculture. Installation of P sorbing filters at drain outlets may be a

solution. Efficient sorbents to be used for such filters must possess high P bonding affinity to retain

ortho-phosphate (Pi) at low concentrations. In addition high P sorption capacity, fast bonding and

low desorption is necessary. In this study five potential filter materials (Filtralite-P®, limestone,

calcinated diatomaceous earth, shell-sand and iron-oxide based CFH) in four particle size intervals

were investigated under field relevant P concentrations (0-161 µM) and retentions times of 0-24

min. Of the five materials examined, the results from P sorption and desorption studies clearly

demonstrate that the iron based CFH is superior as a filter material compared to calcium based

materials when tested again criteria for sorption affinity, capacity and stability. The finest CFH and

Filtralite-P® fractions (0.05-0.5 mm) were best with a P retention of ≥90% of Pi from an initial

concentration of 161 µM corresponding to 14.5 mmol/kg sorbed within 24 min. They were further

capable to retain ≥90% of Pi from an initially 16 µM solution within 1½ min. However, only the

finest CFH fraction was also able to retain ≥90% of Pi sorbed from the 16 µM solution against 4

times desorption sequences with 6 mM KNO3. Among the materials investigated, the finest CFH

fraction is therefore the only suitable filter material, when very fast and strong bonding of high Pi

concentrations is needed, e.g. in drains under P rich soils during extreme weather conditions.

21

1 Introduction

Soils in intensively farmed countries may be sources of phosphorus (P) due to decades of surplus

application of P in organic and inorganic fertilisers (Heal et al., 2005; Delgado and Scalenghe 2008;

Buda et al., 2012). However, a P-enriched soil only becomes an environmental problem when

connected to the aquatic environment by an effective transport pathway such as artificial drains

(Heathwaite et al., 2003). Tile drains and ditches connect fields to recipient waters and act as

highways for both soluble and particulate P (Ulén et al., 2007). Despite substantial efforts over

many years to reduce this transport, leaching of P, from agricultural land to the aquatic environment

is still a serious and costly problem in many parts of Europe and elsewhere (Delgado and

Scalenghe, 2008; Ballantine and Tanner, 2010; Buda et al., 2012). Thus, to reach the goal of good

water quality as stated in the EU Water Framework Directive (WFD) requires a substantial

reduction of the diffuse P loss from farmland in many parts of Europe (Søndergaard et al., 2005;

Kaasik et al., 2008).

While a range of P mitigation options have been tested for surface transport (Hoffmann et al.,

2009), this is not the case for subsurface transport (Kröger et al., 2008). However, by installing a

filter construction at the end of a drainage pipe, Penn et al. (2007) demonstrated an immediate

reduction of drainage P leaching. This end-of-pipe approach could be carried out in connection with

a small-scale constructed wetland (Reinhardt et al., 2005) or with flow-through filter structures in

ditches (Penn et al., 2007). A great variety of different types of filters and filter materials for P

retention have been described for retention of high P concentrations in waste waters (Johansson

Westholm, 2006; Penn et al., 2007; Cucarella and Renman, 2009; Vohla et al., 2011). However, it is

questionable to directly transfer the experience obtained from these high P concentration studies to

the low P concentrations in drainage water as the filter materials may behave differently at high and

low solution concentrations (Agyei et al., 2002; Ádám et al., 2007). Therefore, the filter materials

need to be tested at low P concentrations and short reaction times relevant for cleaning P

contaminated drainage water.

P in drainage water may consist of P in dissolved organic matter, particulate P and dissolved ortho-

phosphate (Pi). The focus in this investigation will be on Pi. In natural, unpolluted areas in

Denmark, the Pi concentration in base flow drainage water is typically ≤1.6 µM but the

concentration can be up to 42 µM in farmland drains (Andersen et al., 2006) and Penn et al. (2007)

22

found more than four times this concentration in Maryland, US. The high water flow during

rainstorms and fast frost-thaw transitions is critical because of high or very high P leaching during

such peak flows (Grant et al., 1996; Johnes, 2007). To effectively remove the Pi during peak flows

with high Pi concentrations, the filter material must react fast and possess high Pi sorption stability

and capacity. The Pi removal efficiency of a filter material is closely related to the content of

various Al, Ca, Fe and Mg (hydr)oxides and carbonates (Johansson Westholm, 2006; Ballantine and

Tanner, 2010; Vohla et al., 2011). In addition to the elemental composition, the specific surface area

(SSA) is important as sorption normally increases at increasing SSA. Although it is outside the

scope of this study, high hydraulic conductivity of the filter material is also essential for use in high-

flow drainage filters.

As a practical test of high Pi removal efficiency also under extreme conditions, we suggest the

following three sorption/desorption criteria: (i) A capacity to retain ≥90% Pi from 161 µM solution

within 24 min; (ii) A reactivity resulting in retention of ≥90% Pi from 16 µM solution within 1½

min; (iii) A stability resulting in dissolution of <10% of this retained Pi after four desorptions with

artificial drainage water (6 mM KNO3). A detailed discussion of these test criteria is given later in

the paper.

Accordingly, the aim of the present investigation is to assess the sorption parameters of five

manufactured filter materials (so-called Filtralite-P®, limestone, calcinated diatomaceous earth

CDE, shell-sand and iron oxide based CFH). The assessment includes a batch-mode testing of the Pi

removal efficiency according to above-mentioned criteria of the materials in four particle sizes at

low concentrations of Pi in artificial and natural drainage waters. To the best of our knowledge, this

is the first study where various potential phosphorus filter materials are tested under the same field-

relevant conditions, i.e. both at low to rather low Pi concentrations and very short to semi-short

sorption (reaction) times.

2 Materials and methods

2.1 Filter material characterization

Filtralite-P® is a Light Expanded Clay Aggregates (LECA) resembling material calcinated at 1200

˚C that was provided by Weber, Norway. The porous material contains granules of Ca/Mg oxides,

which is the active sorbent. Limestone consists of a mixture of bryozo and coral chalk from the

23

Danian formation at Faxe. The dried product was provided by Faxe Kalk A/S, Denmark. Calcinated

diatomaceous earth (CDE) from the Fur formation calcinated at 750 ˚C was provided by Damolin

A/S, Denmark. Shell-sand consisting of crushed sea shells was provided by DanShells Aps,

Denmark. CFH-12 (CFH) consists of dried iron oxides, which was provided by Kemira Oyj,

Finland.

All analyses were carried out on ball-milled bulk materials. pH was measured potentiometrically in

0.01 M CaCl2 using a solid:solution ratio of 1:2.5. The mineralogy of the materials was assessed by

X-ray diffraction analysis on unoriented samples using a Siemens 5000 instrument equipped with

Co-Kα radiation and a diffracted beam monochromator. Diffractograms were recorded from 10 to

90° 2θ using 0.03° 2θ steps and a step speed of 2 s. Diffraction peak positions were used to

calculate d-values for mineral identification. The carbonate content was determined volumetrically

by a calcimeter (Allison and Moodie, 1965). Oxalate-extractable aluminium (Alox) and iron (Feox)

were determined by extraction with 0.2 M ammonium oxalate for 2 h at pH 3 in the dark

(Schwertmann, 1964). Citrate-bicarbonate-dithionite-extractable Al and Fe (AlCBD, FeCBD) were

determined by three sequential extractions for 15 min at 70 °C as described by (Mehra and Jackson,

1960). Total Al, Fe, Ca and Mg (Altotal, Fetotal, Catotal, Mgtotal) were determined after dissolution of

the materials in a mixture of concentrated nitric acid, hydrogen peroxide, hydrochloric acid and

hydrofluoric acid (EPA 3052). Al and Fe in oxalate and CBD extracts were determined by atomic

absorption spectroscopy (AAS) using a Perkin Elmer 3300 and the concentrations of total Al, Fe,

Ca and Mg were determined by inductive coupled plasma mass spectroscopy (ICP-MS) on an

Agilent 7500C instrument. Total (inorganic) P (Ptotal) was determined by extracting the material

with 6 M H2SO4 for 10 min at 70 °C (Mehta et al., 1954). The concentration of P in the extract was

determined by the molybdenum-blue method (Murphy and Riley, 1962).

To assess the influence of particle size on Pi sorption, the filter materials were fractionated into

particle size intervals of 2-4 mm, 1-2 mm, 0.5-1 mm and 0.05–0.5 mm by sieving. The (external)

specific surface area (SSA) of the different fractions was determined by applying the BET equation

(Brunauer et al., 1938) to N2 adsorption data obtained by means of a Micromeritic Gemini VII

2390a instrument. Total element composition of the CFH and Filtralite-P® fractions was determined

by X-ray fluorescence at Actlabs, Canada.

24

2.2 Phosphate sorption

2.2.1 Influence of Pi concentration and background electrolyte

The filter materials were soaked in a background electrolyte consisting of 6 mM potassium nitrate

(KNO3) for 24 h before performing phosphate sorption. This solution called artificial drainage water

(ADW) was used as a proxy of natural drainage water (NDW) to avoid interfering reactions such as

formation of Ca precipitates if using NDW, and 6 mM KNO3 has the same electric conductivity

(ionic strength) as was found in the NDW sample described below. The sorption tests were carried

out as follows: Nine 1 g samples of each of the different particle sizes of the five filter materials

were shaken (175 rpm) in open beakers for 24 min with 100 mL 6 mM KNO3 and 7 initial KH2PO4

concentrations ranging between 0 and 161 µM adjusted to pH 6 with 0.1 M NaOH. After shaking,

pH was measured and an aliquot of the solution was withdrawn with a syringe and filtered through

0.2 µm Millipore syringe-filter. The filtrate was added 20 µL 2 M sulphuric acid per mL for

preservation, and stored (< 1 week) at 5 °C until P determination.

The sorption behaviour was also tested with NDW collected from a drainage well at Tåstrup,

Denmark under a cultivated soil (Typic Argiudoll) developed on calcareous morainic material from

the Weichelian Glaciation. The pH of this NDW was 6.9, it had an electric conductivity of 0.6 dS/m

and an alkalinity (HCO3-) of 4.3 mM. The NDW contained 2.5 mM Ca

2+, 0.3 mM Mg

2+, 0.4 mM

Na+, 1.6 µM K

+ and 1.6 µM Pi; Cl

- and NO3

- were identified but not quantified. Sorption by the 1-2

mm fractions of CFH, Filtralite-P® and limestone was tested with NDW as background electrolyte

using the same Pi concentrations and procedure as described above.

2.2.2 Pi sorption and desorption kinetics

The importance of contact time for short term Pi sorption was determined by shaking (175 rpm) 0.7

g samples of the different filter materials with 70 mL 6 mM KNO3 containing 16 µM KH2PO4

adjusted to pH 6 for 90 s, 180 s, 360 s, 720 s and 1440 s (24 min). The importance of contact time

for semi long-term Pi sorption was performed with CFH, Filtralite-P® and limestone (1-2 mm) as

sorbents in a similar way but the shaking times were extended to 45 min, 1.5 h, 3 h, 6 h, 12 h, 24 h

and 48 h. After shaking, the extracts were treated as described in section 2.2.1.

Immediately after the 24-min sorption from 16 µM Pi solution, desorption was carried out on all

samples by means of four successive extractions, each by shaking the sample (175 rpm) for 15 min

25

with 50 mL of 6 mM KNO3, centrifugation of the suspension and replacement of the extract with a

fresh portion of 6 mM KNO3. After extraction the extracts were acidified as described above.

The phosphate concentrations in the filtrates and extracts were determined by the molybdenum blue

method using flow injection analysis on a FIAstar 5000 instrument (Ruzicka and Hansen, 1988).

Sorbed Pi in µmol/kg was calculated from the difference between the Pi concentrations before and

after shaking with the filter materials. In the desorption investigation, desorbed Pi in each

desorption step was calculated from the Pi concentration in the extract corrected for left-over from

the previous desorption.

All experiments were carried out in triplicate. The glass- and plastic wares were acid-washed, the

chemicals were pro analysis or of better quality and triple deionized water was used throughout.

2.3 Data handling

Sorbed Pi was plotted against the solution concentration resulting in the isotherms. Sorbed Pi

corresponding to initial concentrations of 16 µM and 161 µM were read on the isotherms for the

various materials and particle sizes (S16, S161).

The sorption versus time data from the short-term (90-1440 s) and semi long-term (0.75-48 h)

sorption investigations were fitted to the hyperbola equation:

tK

tSSK

max

(Eq. 1)

SK is the time-dependent amount of sorbed Pi (µmol/kg) at time t (min). Smax is the maximum Pi

sorption (µmol/kg) and K (min) is a fitting parameter determined by the shape of the sorption curve.

The curves were fitted with SigmaPlot (v. 12.0, Systat Software, Inc.).

3 Results and discussion

3.1 Filter material characteristics

The filters comprise three kinds of minerals including carbonates (limestone, shell-sand), iron

oxides (CFH) and mixed minerals (CDE, Filtralite-P®) (Table 1). According to XRD, the limestone

consists of pure calcite (CaCO3), while shell-sand contains two calcium carbonates, i.e. aragonite

and calcite, together with dolomite (CaMg(CO3)2). Both materials contain small amounts of Al and

26

Fe oxides. CFH consists of poorly ordered Fe oxides (Table 1) as also indicated by XRD. In

addition to Fe, CFH contains Ca and Mg carbonates. CDE and Filtralite-P® are calcinated at high

temperature, which makes the clay mineral structures collapse and turn into amorphous solids.

Together with the elemental analyses it seems, however, rather clear that they consist of mixtures of

poorly ordered Al, Ca, Fe and Mg silicates. In addition, Filtralite-P® undoubtedly also contains

Ca/Mg oxides as indicated by the very alkaline pH (Table 1). Important is to note that all materials

contain minor to small but significant amounts of Pi, especially the CDE and CFH materials. The

contents of Al and Fe (hydr)oxides are very important in relation to Pi sorption, as Al and Fe oxides

(AlCBD, FeCBD), especially the poorly ordered, oxalate extractable forms (AlOx, FeOx), are well

known effective Pi sorbents (Borggaard et al., 2005; Johansson Westholm, 2006; Cucarella and

Renman, 2009; Vohla et al., 2011; Penn et al., 2011). Accordingly, CFH with very high FeCBD and

FeOx contents is expected to be a very strong Pi sorbent. As Ca and Mg carbonates and oxides are

important Pi sorbents, Filtralite-P®, limestone and shell-sand may also be considered to act a Pi

sorbents. On the other hand, CDE will probably be a rather poor sorbent as most silicates even tiny

clay silicates possess limited Pi sorption capacity (Ballantine and Tanner, 2010).

Table 1. Chemical composition of the five potential filter materials (CDE, CFH, limestone, Filtralite-P® and shell-sand)

based on analyses of bulk samples.

Element/

characteristic Form Unit CDE CFH Limestone Filtralite-P

® Shell-sand

Fe

Oxalate mmol/kg 26.0±0.4 7310±90 5.0±0.1 22.0±0.2 5.0±0.1

CBD mmol/kg 218±11 7790±150 5.0±0.0 24.0±0.3 4.0±0.1

Total mmol/kg 815±30 7360±420 8.0±0.1 737±30 11±1

Al

Oxalate mmol/kg 26±3 15.0±0.2 15.0±0.1 31.0±0.4 15±1

CBD mmol/kg 31±1 6.0±2.8 nd2

21.0±1.6 nd

Total mmol/kg 629±27 32±2 nd 2350±160 23±2

Ca Total mmol/kg 381±10 658±23 9590±50 1720±60 9280±120

Mg Total mmol/kg 100±18 1390±100 103±1 1560±110 565±26

Pi Total mmol/kg 12.0±0.5 8.0±0.3 1.0±0.1 3.0±0.1 4.0±0.1

Carbonate1

% 0.0±0.0 5.0±0.1 101±3 6.0±0.2 102±2

pH 4.7±0.1 8.2±0.1 7.9±0.1 11.9±0.2 8.1±0.1

Minerals

Mixed

calcined

clay

silicates

Two line Fe

oxides Calcite

Calcite, Ca/Mg

oxides, clay

silicates

Calcite,

aragonite,

dolomite

1Carbonate expressed as CaCO3.

2Not determined

27

The different particle size fractions had somewhat different specific surface areas, SSAs (Table 2).

Except for CDE, the smallest particles had – as expected - the highest SSAs. Consequently, the

order of decreasing Pi sorption per mass unit is expected to be 0.05-0.5 mm > 0.5-1 mm > 1-2 mm >

2-4 mm as Pi sorption is normally found to increase with increasing SSA for the same material

(Ballantine and Tanner, 2010; Penn et al., 2011). SSA also differs among the materials with CFH

and CDE having much larger SSAs than the other materials. The poorly ordered Fe oxides are

responsible for the substantial CFH SSAs, whereas CDE has high SSA due to larger internal (N2

accessible) pores generated by the diatomite framework. The total element composition of the

Filtralite-P® and CFH fractions (Supplemental data, Table S1) showed the finer Filtralite-P

®

fractions were somewhat enriched in Ca and Mg but not in Al and Fe, whereas the three CFH

fractions had the same composition.

3.2 Phosphate retention

3.2.1 Phosphate sorption capacity

The 24-min Pi sorption isotherms for the five materials are very different with limited sorption to

CDE, limestone and shell-sand; in fact CDE releases Pi at low concentration (Fig. 1). In contrast,

CFH and Filtralite-P® possess much higher high Pi sorption capacity. Except CFH, the shape of the

isotherms comprises two steps.

For soils and various materials such as Al and Fe oxides, Pi sorption isotherms can often be fitted

with the Langmuir equation resulting in sorption maximum and stability e.g. (Borggaard et al.,

2005; Johansson Westholm, 2006; Vohla et al., 2011). This is not the case for the 2-steps isotherms

in Fig. 1, i.e. Langmuirian sorption capacity and stability cannot be determined. Similar non-

Langmuirian shaped sorption isotherms have also been reported for similar and other materials

tested for their suitability as P filters and the shape of the curves has been ascribed to a combination

of adsorption and precipitation reactions at low and high Pi concentrations, respectively (Ádám et

al., 2007; Kaasik et al., 2008; Cucarella and Renman, 2009). Therefore, it was decided to

characterize the Pi retention capacity of all materials by means of Pi sorbed at initial Pi

concentrations of 16 µM (S16) and 161 µM (S161).

28

Fig. 1. Pi sorption isotherms for the 3-4 particle size fractions of the five potential drainage water filter materials. The

initial Pi concentrations covered the range 0 to161 µM, the solution:solid ratio was 100, the sorption time was 24 min

and the background electrolyte was 6 mM KNO3. The values are averages of the triplicate and error bars represent

standard deviation. Red arrows indicate possible change in sorption mechanisms.

Sorption (S161) at the highest Pi concentration (161 µM) was chosen to show sorption at a high filter

inlet concentration comparable to Pi concentrations reported for high Pi drainage waters (Penn et al.,

2007; Chardon et al., 2012). The first test criterion was the material ability to remove at least 90%

of Pi from the highest initial concentration within 24 min corresponding to sorption of ≥14.5

mmol/kg. As indicated in Fig. 1, S161 strongly depends on material and particle size (Table 2). Apart

29

from shell-sand, sorbed Pi generally increases with decreasing particle size or increasing SSA in

agreement with most other studies (Cucarella and Renman, 2009; Ballantine and Tanner, 2010;

Vohla et al., 2011). For the same particle size, S161 decreases in the order: CFH ≈ Filtralite-P® >

limestone ≈ CDE > shell-sand. In fact, the 0.05-0.5 mm and 0.5-1 mm Filtralite-P® fractions and the

smallest CFH fraction sorbs all or almost all added Pi, i.e. these samples fulfil the test criterion of

≥14.5 mmol/kg sorption within 24 min.

Table 2. Phosphate sorption (mean and SD) by the particle size fractions of the five filter materials from solutions with

initially 16 µM Pi (S16) and 161 µM Pi (S161) together with pH and the specific surface area (SSA) of the various size

fractions. Pi sorption expressed as µmol/kg and percent of added Pi. S16 and S161 were taken from isotherms in Fig. 1;

sorption time 24 min and solution:solid ratio 100.

Material Particle

size

S16 S161 pH 1SSA

mm µmol/kg (%) µmol/kg (%) m2/g

Artificial drainage water (ADW) as background electrolyte

CDE

0.05-0.5 142±15 (9) 4020±325 (25) 5.3 26±1

0.5-1 102±30 (6) 3870±222 (24) 5.2 26±1

1-2 -21±60 (-1) 2150±952 (13) 5.5 32±4

2-4 -157±32 (-10) 2440±141 (15) 5.5 30±2

CFH

0.05-0.5 1580±51 (98) 16100±50 (100) 8.2 44±7

0.5-1 1570±32 (97) 11000±195 (68) 6.9 32±10

1-2 1100±51 (68) 7930±140 (49) 6.7 32±2

Limestone

0.05-0.5 1320±208 (82) 4340±1860 (68) 8.9 1.3±0.2

0.5-1 258±34 (16) 2030±1410 (13) 8.8 <0.1

1-2 188±19 (12) 2670±251 (17) 8.8 <0.1

2-4 195±76 (12) 2230±276 (14) 8.6 <0.1

Filtralite-P®

0.05-0.5 1590±52 (99) 16200±12 (100) 11.7 3.2±0.2

0.5-1 1130±365 (70) 14900±839 (92) 10.8 1.2±0.0

1-2 571±136 (35) 10200±2100 (63) 10.1 0.9±0.2

2-4 524±213 (33) 9430±21400 (59) 9.8 0.9±0.1

Shell-sand

0.05-0.5 479±28 (30) 3330±461 (21) 9.1 <0.1

05-1 358±12 (22) 2800±45 (17) 8.6 <0.1

1-2 283±12 (18) 2560±65 (16) 8.6 <0.1

2-4 250±11 (16) 2450±60 (15) 8.3 <0.1

Natural drainage water (NDW) as background electrolyte

CFH 1-2 894±133 (47) 5550±273 (37) 8.2 32±2

Limestone 1-2 299±85 (16) 1970±85 (12) 8.3 <0.1

Filtralite-P®

1-2 418±33 (22) 1980±96 (12) 8.3 0.9±0.2 1Limit of detection was 0.1 m

2/g.

However, SSA cannot adequately explain the big differences between the various fractions of, in

particular Filtralite-P® and CFH. The differences can neither be explained by the similar (CFH) or

rather similar (Filtralite-P®) total element compositions (Table S1) indicating that other material

characteristics are responsible for the differences. The much higher sorption by the finer particle

30

size fractions (Fig. 1, Table 2) may be due to more internal hydrophilic sorption sites in the finer

than coarser fractions, which are accessible to Pi but not to the hydrophobic N2, and hence not

accounted for by the N2 determined SSA (Makris et al., 2004). Comparison with previously

published data for Pi sorbed to similar materials such as iron oxide sludge, Filtralite-P®, limestone

and shell-sand, is difficult as in general much higher P concentrations and reaction times have been

used. However, for comparable solution Pi concentrations our short term S161 for Filtralite-P® is in

the same range as sorbed amounts after reaction for 24 h found by Ádám et al. (2007), while for

iron oxides our S161 is more than 10 times lower than the sorbed amounts after 1 – 21 d (Chardon et

al., 2012). It demonstrates that iron oxides may have very high sorption capacities that increases

strongly with time (Makris et al., 2004).

In the field, however, efficient filters must react fast to retain drainage water P during fast leaching

events such as rainstorms or when the transition between frost and thaw is very short but they must

also be able to reduce Pi to very low concentrations as Pi concentrations in lakes higher than ~50

µg/L (1.6 µM) may cause eutrophication (Smil, 2000). Consequently, the Pi threshold at the filter

outlet is set at 1.6 µM. The determination of sorbed P (S16) at an initial Pi concentration of 16 µM

was used to test the capacity of the materials to substantially sorb Pi even at low Pi concentration. In

fact, the criterion was 90% retention (1440 µmol/kg) from the 16 µM Pi solution. The sorption

criterion of 1440 µmol/kg (S16) is fulfilled by the two finest CFH particle size fractions and the

0.05-0.5 mm Filtralite-P® fraction (Table 2), whereas the other materials and size fractions are less

effective sorbents; although the 1-2 mm CFH, the finest limestone and the 0.5-1 mm Filtralite-P®

fractions are rather close to the threshold.

As also seen in Table 2, sorbed Pi depends on the background electrolyte even though the ADW and

the NDW have similar electric conductivities, i.e. similar ionic strengths. However, the ion

composition of the two electrolytes is different as NDW is dominated by Ca2+

and HCO3-, whereas

K+ and NO3

- are the only ions in ADW. The influence of the background electrolyte on sorbed Pi

(S16 and S161) differs for the three materials tested. Thus, CFH seems more effective with ADW as

background electrolyte (sorbs 20-30% less with NDW), while limestone sorbs more (60% increase)

from the initial 16 µM Pi solution but less (25%) from the 161 µM solution with NDW as

background electrolyte instead of ADW. The biggest difference is seen for Filtralite-P®, where

especially S161 is substantially reduced (80%) with NDW as background electrolyte. This difference

31

may be explained by a substantial drop in pH from 10.1 in ADW to 8.3 in NDW because of HCO3-

in NDW, which forms CaCO3 by reaction with Ca2+

in Filtralite-P® changing the sorption S161 to be

similar to limestone (Table 2). The observed influence of the background electrolyte on Pi sorption

is in accordance with previous studies showing that the composition and concentration of the

various background electrolytes may affect the sorption (Johansson Westholm, 2006; Ádám et al.,

2007; Sø et al., 2011).

3.2.2 Phosphate sorption kinetics

Even though the kinetics of Pi sorption by soils and similar materials have been investigated for

more than half a century (Chardon et al., 2012), short-time kinetics is less studied. An exception is

the work by Heal et al. (2003), who reported fast sorption of Pi by ochre from mine water treatment

plants able to reduce the Pi concentration from 161 µM to <0.3 µM within 8 min. However,

normally sorption times of several hours and days are used in an attempt to reach equilibrium as

indicated above.

32

Fig. 2. Pi sorption kinetics for the 1-2 mm size fractions of CFH, Filtralite-P® and Limestone. The initial Pi

concentration was 16 µM, the solution:solid ratio was 100 and the background electrolyte was 6 mM KNO3. The data

were fitted by the Eq. 1 and shown as curves; the dashed line corresponds to sorption of 1440 µmol/kg. The values are

averages of triplicates and error bars represent standard deviation.

The sorption kinetic data were fitted by Eq. 1 (Fig. 2) and Table 3 shows the sorption maxima

(Smax), the time parameter (K), and sorption after 1½ min (S1½) calculated by Eq. 1. Smax is the

maximum sorption under the current experimental conditions, in casu sorption from a 16 µM Pi

solution with a solution:solid ratio of 100, i.e. Smax in Table 3 cannot exceed 1600 µmol/kg. The

above-mentioned threshold sorption of 1440 µmol/kg corresponding to 90% of the initially added Pi

is only fulfilled by the two finest CFH size fractions and the finest Filtralite-P® size fraction. The

fitting parameter K is determined by the shape of sorption vs time curve and increases as the curve

becomes more straight-lined, i.e. a low K value corresponds to a strongly bending curve as seen for

the CFH sample in Fig. 2 with K = 5.8 min compared to the more smoothly bending Filtralite-P®

curve with K = 160 min (Table 3). In other words, K is a measure of the sorption rate, the smaller

the K, the faster the sorption. Fast reaction of CFH is in line with reported initial fast sorption of Pi

33

by various pure (synthetic) Fe oxides (Strauss et al., 1997; Chardon et al., 2012). The fast sorption

by Filtralite-P® is assumed due to rapid reaction between Pi and hydrated Ca/Mg oxides as well as

between Pi and reactive Al and Fe compounds in the Filtralite-P® sorbent (Table 1). Even more

important is to note that after as little as 1½ min, the finest CFH and Filtralite-P® size fractions have

passed this threshold showing that these samples are fast reacting even at this low Pi concentration

(Table 3). Extending sorption time up to 48 h, the 1-2 mm CFH and Filtralite-P® samples can sorb

all added Pi, whereas the limestone sorbs less than half of the added Pi.

Table 3. Influence of time on phosphate sorption by the particle size fractions of the five filter materials divided in

short-term sorption, 1½-24 min, and semi long-term sorption, 3/4-48 h. Sorption data were fitted to Eq. 1 resulting in

sorption maximum (Smax) and sorption curve shape factor (K). SK1½ is sorbed Pi after 1½ min calculated by Eq. 1.

Initial Pi concentration 16 µM, material:solid ratio 100 and background electrolyte ADW (6 mM KNO3).

Material Particle size Smax K SK1½

mm µmol/kg min µmol/kg

Short-term kinetic series 1½-24 min

CDE

0.05-0.5 8±25 1nd nd

0.5-1 9±36 nd nd

1-2 -84±9 nd nd

2-4 -134±14 nd nd

CFH

0.05-0.5 1610±10 0.02±0.01 1590±10

0.5-1 1550±8 0.66±0.01 1100±20

1-2 1140±30 1.68±0.20 590±45

Limestone

0.05-0.5 1170±10 0.13±0.01 1080±20

0.5-1 343±26 0.56±0.01 252±7

1-2 219±7 0.12±0.01 201±10

2-4 99±10 2.3±0.9 44±7

Filtralite-P®

0.05-0.5 1570±10 0.00±0.00 1570±10

0.5-1 913±50 0.8±0.3 620±95

1-2 412±10 2.0±0.2 195±17

2-4 476±81 1.5±0.1 257±108

Shell-sand

0.05-0.5 569±46 2.3±0.7 283±64

0.5-1 208±24 0.4±0.5 142±36

1-2 237±9 1.3±0.2 149±15

2-4 233±11 2.0±0.4 120±13

Semi long-term kinetic series 3/4-48 h

CFH 1-2 1620±30 5.8±0.1 -

Limestone 1-2 630±41 320±66 -

Filtralite-P®

1-2 1610±50 160±18 - 1Not determined as the CDE graphs could not be fitted by Eq. 1.

3.2.3 Phosphate sorption stability

Even though fast and substantial Pi sorption is mandatory for the practical use of the filter materials,

it is also important with a stable bonding of sorbed Pi to ensure that Pi is not desorbed when the

sorption condition changes, e.g. because of decreasing Pi concentration in the drainage water (Grant

34

et al., 1996). As a test criterion for the stability of Pi retention, desorption of no more than 10% of

the Pi sorbed during 4 successive desorption steps from the initial 16 µM solution was adopted.

After four desorptions by ADW, less than 10% was desorbed from the three CFH size fractions

indicating high stability of the bonding of Pi by this material (Fig. 3). Considering the high content

of Fe oxides in this material (Table 1), strong Pi retention is expected as Fe oxides are known to

form strong bonds with Pi (Strauss et al., 1997; Heal et al., 2003; Borggaard et al., 2005; Chardon et

al., 2012). All the other samples lost more than 10% in the desorption test. In fact, more than 10%

was desorbed already after the first desorption (Fig. 3), and for CDE (data not shown) more Pi was

desorbed than sorbed from the initial 16 µM Pi solution, which may be ascribed to the original Pi

content of this material (Table 1, Fig. 1). Apart from CFH, which showed very little Pi release, the

relative desorption seen in this study (Fig. 3) is high compared to the relative desorption of 10-32%

of sorbed Pi seen for six industrial by-products (Penn et al., 2011), which may be attributed to

different reaction times.

Fig. 3. Pi desorption by means of 6 mM KNO3 from the 3-4 particle size fractions of the five potential drainage water

filter materials loaded with Pi from 16 µM solution. The P-loaded material was shaken 4 times, each time for 15 min

using 6 mM KNO3 (ADW).

35

3.3 Suitability of the materials as phosphate filters

The five materials in 3-4 particle size fractions have been tested with respect to Pi sorption

efficiency, i.e. sorption capacity, sorption reactivity and sorption stability. Emphasis has been put

on test conditions considered important for possible use of the materials as filters to reduce Pi in

drainage water to a non-eutrophication level before it enters open water bodies, especially streams

and lakes, also under extreme drainage water flow conditions. Consequently, high, fast and strong

Pi retention was the test criteria in this investigation. Obviously, only the 0.05-0.5 mm CFH fraction

fulfils all three criteria. Filtralite-P® has a rather high sorption capacity and the finest fraction (0.05-

0.5 mm) sorbs >90% Pi from the initially 16 µM solution but all Filtralite-P® fractions fail to retain

at >90% of the sorbed Pi in the desorption test. In the test, CDE came out as the least suitable

material as it has a low sorption capacity and more Pi was released during desorption than originally

sorbed (data not shown). The carbonate-based materials, limestone and shell-sand possess

characteristics between those of Filtralite-P® and CDE and they are all not suitable as fast reacting

filter materials. The remarkable behaviour of Filtralite-P®

with high capacity at high pH but limited

capacity at neutral pH, when NDW was used as background electrolyte (Table 2), limits the

practical suitability of this material.

The criteria set in this study may be considered very rigorous, and possibly less strict criteria may

occasionally be justified when for example drainage water from an agricultural field is diluted with

water from an unpolluted area, e.g. old forests, resulting in ecosystem-tolerable Pi concentrations in

receiving open water bodies. Furthermore, the result does not exclude that some of the other tested

materials can be used (and are maybe superior to CFH) in remediation of certain wastewaters, as

landscape barriers, as amendments to control Pi release in very P-rich soils and other uses than as

filter material for drainage water. Finally it must be emphasized that despite the favorable

phosphate sorption characteristics of CFH, assessment of the suitability of this material in drainage

water filters requires consideration of additional factors such as hydraulic properties, costs,

availability, toxicity, long-term performance and use of the P-saturated (used) filter material as P

fertilizer (Ballantine and Tanner, 2010; Vohla et al., 2011). However, assessment of these additional

and important factors is outside the scope of the present study.

36

4 Conclusions

Materials proposed as efficient drainage water filters must possess high capacity, reactivity and

stability to retain Pi at rather low and very low concentrations. To address these requirements also

under extreme subsurface runoff conditions, a 3-step test was proposed:

1. Capacity: Retention of ≥90% of Pi from 161 µM solution corresponding to 14.5

mmol/kg within 24 min with solution:material ratio = 100.

2. Reactivity: Retention of ≥90% of Pi from 16 µM

3. solution decreasing solution Pi to ≤1.6 µM (50 µg/L) within 1½ min (representing high

flow situations) and 24 min (representing low flow situations).

4. Stability: Retention of ≥90% of Pi sorbed from 16 µM solution within 24 min against 4

times desorption with 6 mM KNO3.

Application of the test to sorption and desorption results obtained for the four particle size fractions

of the five potential filter materials showed that criteria 1 and 2 were passed by the finest CFH and

Filtralite-P® fractions, while all CFH samples fulfilled the third criterion. Consequently, only the

finest (0.05-0.5 mm) CFH sample fulfilled all criteria of this very strict test. The choice of

background electrolyte, whether ADW or NDW, affected the sorption, in particular for Filtralite-P®.

5 Supplemental data

Table S1. The main element composition of the 3 CFH and 4 Filtralite-P® fractions as determined

by XRF.

6 Acknowledgement

The project was carried out in the frames of SupremeTech project funded by The Danish Council

for Strategic Research (grant no. 09-067280). Special thanks to Bente Postvang for laboratory

assistance.

37

7 References

Ádám, K., Krogstad, T., Vråle, L., Søvik, A. K., Jenssen, P. D. 2007. Phosphorus retention in the

filter materials shellsand and Filtralite P®—Batch and column experiment with synthetic P solution

and secondary wastewater. Ecological Engineering 29 (2), 200-208.

Agyei, N. M., Strydom, C. A., Potgieter, J. H. 2002. The removal of phosphate ions from aqueous

solution by fly ash, slag, ordinary Portland cement and related blends. Cement and Concrete

Research 32 (12), 1889-1897.

Allison, L. E. and Moodie, C. D. (1965) Carbonate. In: Black, C.A., Evans, D.D., White, J.L.,

Ensminger, L.E., Clark, F.E. (Eds.), Methods of Soil Analysis: Part 2. Chemical and

Microbiological Properties, 1392-1395, Soil Science Society of America, Madison, Wisconsin.

Andersen, H. E., Larsen, S. E., Kronvang, B., Hansen, K. M., Laubel, A., Windolf, J., Muus, K.

2006. Fosfat i drænvand. Vand og Jord 13 (4), 152-156. (In Danish)

Ballantine, D. J. and Tanner, C. C. 2010. Substrate and filter materials to enhance phosphorus

removal in constructed wetlands treating diffuse farm runoff: a review. New Zealand Journal of

Agricultural Research 53 (1), 71-95.

Borggaard, O. K., Raben-Lange, B., Gimsing, A. L., Strobel, B. W. 2005. Influence of humic

substances on phosphate adsorption by aluminium and iron oxides. Geoderma 127 (3–4), 270-279.

Brunauer, S., Emmett, P. H., Teller, E. 1938. Adsorption of gases in multimolecular layers. Journal

of the American Chemical Society 60 (2), 309-319.

Buda, A. R., Koopmans, G. F., Bryant, R. B., Chardon, W. J. 2012. Emerging technologies for

removing nonpoint phosphorus from surface water and groundwater: Introduction. Journal of

Environmental Quality 41 (3), 621-627.

Chardon, W. J., Groenenberg, J. E., Temminghoff, E. J. M., Koopmans, G. F. 2012. Use of reactive

materials to bind Phosphorus. Journal of Environmental Quality 41 (3), 636-646.

38

Cucarella, V., Renman, G. 2009. Phosphorus sorption capacity of filter materials used for on-site

wastewater treatment determined in batch experiments – A Comparative Study. Journal of

Environmental Quality 38 (2), 381-392.

Delgado, A., Scalenghe, R. 2008. Aspects of phosphorus transfer from soils in Europe. Journal of

Plant Nutrition and Soil Science 171 (4), 552-575.

Grant, R., Laubel, A., Kronvang, B., Andersen, H. E., Svendsen, L. M., Fuglsang, A. 1996. Loss of

dissolved and particulate phosphorus from arable catchments by subsurface drainage. Water

Research 30 (11), 2633-2642.

Heal, K., Younger, P. L., Smith, K., Glendinning, S., Quinn, P., Dobbie, K. 2003. Novel use of

ochre from mine water treatment plants to reduce point and diffuse phosphorus pollution. Land

Contamination and Reclamation 11 (2), 145-152.

Heal, K., Dobbie, K., Bozika, E., McHaffie, H., Simpson, A., Smith, K. 2005. Enhancing

phosphorus removal in constructed wetlands with ochre from mine drainage treatment. Water

Science and Technology 51 (9), 275-282.

Heathwaite, L., Sharpley, A., Bechmann, M. 2003. The conceptual basis for a decision support

framework to assess the risk of phosphorus loss at the field scale across Europe. Journal of Plant

Nutrition and Soil Science 166 (4), 447-458.

Hoffmann, C. C., Kjaergaard, C., Uusi-Kämppä, J., Hansen, H. C. B., Kronvang, B. 2009.

Phosphorus Retention in Riparian Buffers: Review of Their Efficiency. Journal of Environmental

Quality 38 (5), 1942-1955.

Johansson Westholm, L. 2006. Substrates for phosphorus removal—Potential benefits for on-site

wastewater treatment? Water Research 40 (1), 23-36.

Johnes, P. J. 2007. Uncertainties in annual riverine phosphorus load estimation: Impact of load

estimation methodology, sampling frequency, baseflow index and catchment population density.

Journal of Hydrology 332 (1-2), 241-258.

39

Kaasik, A., Vohla, C., Mõtlep, R., Mander, Ü, Kirsimäe, K. 2008. Hydrated calcareous oil-shale ash

as potential filter media for phosphorus removal in constructed wetlands. Water Research 42 (4–5),

1315-1323.

Kröger, R., Holland, M. M., Moore, M. T., Cooper, C. M. 2008. Agricultural drainage ditches

mitigate phosphorus loads as a function of hydrological variability. Journal of Environmental

Quality 37, 107-113.

Makris, K. C., Harris, W. G., O'Connor, G. A., Obreza, T. A. 2004. Phosphorus immobilization in

micropores of drinking-water treatment residuals: Implications for long-term stability.

Environmental Science & Technology 38 (24), 6590-6596.

Mehra, O. P. and Jackson, M. L. (1960) Iron oxide removal from soils and clays by dithionite-

citrate system buffered with sodium bicarbonate. In: Proceedings of 7th

National Conference of

Clay and Clay Minerals. Washington 1958, 317-327.

Mehta, N., Legg, J., Goring, C., Black, C. 1954. Determination of organic phosphorus in soils: I.

Extraction method. Soil Science Society of America Journal 18 (4), 443-449.

Murphy, J. and Riley, J. P. 1962. A modified single solution method for the determination of

phosphate in natural waters. Analytica Chimica Acta 27 (0), 31-36.

Penn, C. J., Bryant, R. B., Kleinman, P. J. A., Allen, A. L. 2007. Removing dissolved phosphorus

from drainage ditch water with phosphorus sorbing materials. Journal of Soil and Water

Conservation 62 (4), 269-276.

Penn, C. J., Bryant, R. B., Callahan, M. P., McGrath, J. M. 2011. Use of industrial by-products to

sorb and retain phosphorus. Communications in Soil Science and Plant Analysis 42 (6), 633-644.

Reinhardt, M., Gächter, R., Wehrli, B., Müler, B. 2005. Phosphorus retention in small constructed

wetlands treating agricultural drainage water. Journal of Environmental Quality 34 (4), 1251-1259.

Ruzicka, J. and Hansen, E. H. 1988. Homogeneous and heterogeneous systems: Flow injection

analysis today and tomorrow. Analytica Chimica Acta 214 (0), 1-27.

40

Schwertmann, U. 1964. Differenzierung der Eisenoxide des Bodens durch Extraction mit

Ammoniumoxalat-Lösung. Zietschrift für Pflanzenernährung, Düngung, Bodenkunde 105, 194-202.

(in German)

Smil, V. 2000. Phosphorus is the environment: Natural flows and human interferences. Annual

Review of Energy and the Environment 25 (1), 53-88.

Sø, H. U., Postma, D., Jakobsen, R., Larsen, F. 2011. Sorption of phosphate onto calcite; results

from batch experiments and surface complexation modeling. Geochimica et Cosmochimica Acta 75

(10), 2911-2923.

Søndergaard, M., Jeppesen, E., Jensen, J.P., Amsinck, S.L. 2005. Water Framework Directive:

ecological classification of Danish lakes. Journal of Applied Ecology 42 (4), 616-629.

Strauss, R., Brümmer, G. W., Barrow, N. J. 1997. Effects of crystallinity of goethite: II. Rates of

sorption and desorption of phosphate. European Journal of Soil Science 48 (1), 101-114.

Ulén, B., Bechmann, M., Folster, J., Jarvie, H. P., Tunney, H. 2007. Agriculture as a phosphorus

source for eutrophication in the north-west European countries, Norway, Sweden, United Kingdom

and Ireland: a review. Soil Use and Management 23, 5-15.

Vohla, C., Kõiv, M., Bavor, H. J., Chazarenc, F., Mander, Ü. 2011. Filter materials for phosphorus

removal from wastewater in treatment wetlands—A review. Ecological Engineering 37 (1), 70-89.

41

Supplemental Information

A three-step test of phosphate sorption efficiency of potential agricultural drainage filter materials

G. Lyngsie*, O.K. Borggaard, H.C.B. Hansen

University of Copenhagen, Department of Plant and Environmental Sciences, Thorvaldsensvej 40, DK-1871 Frederiksberg C, Denmark *Corresponding author: chj177@alumni.ku.dk

1 Page

1 Table

Table S1 Main elements obtained by XRF on fractionated CFH and Filtralite-P® given in % oxides.

LOI - Loss on Ignition. Detection limit 0.01%

Material Fraction SiO2 Al2O3 Fe2O3 MnO MgO CaO Na2O K2O TiO2 P2O5 Cr2O3 V2O5 LOI Total

CFH

0.5-0.05 1.77 0.46 56.32 0.208 5.98 3.34 0.07 0.16 0.21 0.15 < 0.01 0.004 27.15 95.82

1-0.5 1.5 0.47 58.36 0.21 5.97 2.62 0.04 0.13 0.21 0.13 < 0.01 0.003 26.99 96.63

2-1 1.34 0.26 58.74 0.216 6.01 2.54 0.05 0.18 0.21 0.13 < 0.01 0.005 27.05 96.73

Filtralite®

P

0.5-0.05 51.16 15.13 7.04 0.101 6.74 8.03 1.78 3.5 0.72 0.21 0.01 0.016 5.49 99.92

1-0.5 54.66 16.2 7.05 0.107 5.32 6.08 1.87 3.76 0.77 0.22 0.01 0.018 3.13 99.19

2-1 56.81 16.85 7.39 0.107 4.85 5.16 1.92 3.9 0.8 0.23 0.01 0.02 2.06 100.1

4-2 56.15 16.78 7.15 0.109 4.97 5.36 1.91 3.88 0.8 0.23 0.01 0.019 2.22 99.58

42

43

Chapter 6 – Paper II

Modelling of phosphate retention by two promising drainage filter materials calcium-based

Filtralite®

P and an iron-rich CFH under flow-through conditions.

Lyngsie, Gry1*

, Penn, Chad J.2, Pedersen, H.L.

3, Borggaard, Ole K.

1, Hansen, Hans C. B.

1

1University of Copenhagen, Department of Plant and Environmental Sciences, Thorvaldsensvej 40,

DK-1871 Frederiksberg C, Denmark. 2Oklahoma State University, Department of Plant and Soil

Sciences, 368 Agricultural Hall, Stillwater, OK 74078-6028, USA. 3

University of Copenhagen,

Department of Mathematical Sciences, Universitetsparken 5, DK-2100 København Ø, Denmark. *Corresponding author: chj177@ku.dk

Draft manuscript

Abstract

Subsurface transport of phosphate (Pi) from fertilized agricultural fields to freshwaters may lead to

eutrophication, reduced biodiversity and fish kills in lakes. Reduction of Pi transport by means of

filters in drains with Pi sorbing materials (PSM) may be one way to improve water quality. The

aims of the study were to model the effect of retention time (RT) and inlet Pi concentration on Pi

sorption and desorption by two PSMs, Ca-based Filtralite®P and Fe-based CFH-12, in a flow-

through setting. Four inlet Pi concentrations (1.6, 3.2, 16 and 32 µM) were all tested with 6

different retention times (0.5-9 min). Both materials’ Pi sorption capacity and affinity is highly

dependent on the Pi inlet concentration. The Fe oxide-based CFH displayed a P sorption capacity

and affinity ten times higher than Ca-based Filtralite®P. CFH released less than 10% of previously

sorbed P compared to Filtralite®P, which released ≥35% during desorption. Furthermore, Pi

sorption by Filtralite®P was also positively correlated to RT. This was not the case for CFH

indicating that CFH will be capable of removing Pi even at high flow rates (i.e. low RT). Therefore,

CFH would be the preferred material for a P removal structure with fluctuating inlet Pi

concentrations and RT. A PSM specific prediction model was produced using flow-through

sorption data, which predicts how much Pi can be removed and how long a specific material will

last until Pi saturation for a given set of site conditions.

1 Introduction

Phosphorus (P) contributes a major role in regard to eutrophication of lakes and inland water, and in

many countries P export from agriculture is now the main source of P to surface waters (Kronvang

44

et al., 2009). To improve quality of the aquatic environment as stated by the European Water

Framework Directive (WFD), substantial reduction of diffuse P loss from farmlands in Europe is

required (Kaasik et al., 2008). P in drainage water exists in different forms including organic P,

particulate P, and dissolved orthophosphate (Pi). The focus in this investigation will be on Pi. In

Denmark, drainage water Pi concentrations in base flow are typically 1.6 µM or lower in natural

(non-polluted) areas but can be up to 42 µM in farmlands (Andersen et al., 2006). Penn et al. (2007)

found more than four times this concentration in Maryland, US and Groenenberg et al., (2013)

reported average Pi concentration in the Netherlands of ≈100 µM but with large spatial variation.

Most P is desorbed and dissolved from soils in high runoff water flows during rainstorms,

especially within the first flush, where high peak concentrations often are observed (Grant et al.,

1996; Johnes, 2007; Kronvang et al., 2003; Penn et al., 2012; Ulén, 1995).

Various Pi sorbing technologies have been conceived during the last two decades, ranging from Pi

immobilization (stabilization) in the soil to different remediation measures that utilize filters to

remove Pi from runoff water (Buda et al., 2012). In contrast to soil immobilization which does not

allow for P removal, the use of filters, which will be focused on here, can ultimately remove Pi

from drainage water.

Different Pi sorbing materials (PSMs) have been tested but mainly with high Pi concentrations as in

wastewaters (Ballantine and Tanner, 2010; Cucarella and Renman, 2009; Johansson Westholm,

2006; Penn et al., 2007). However, it is not appropriate to directly transfer the results obtained from

such studies to the much lower Pi concentrations in most drainage waters (Johansson Westholm,

2006; Klimeski et al., 2012; Vohla et al., 2011).

In order to estimate the effectiveness and longevity of PSMs in drainage water filters, the impact of

the solution Pi concentration and the retention time (RT) on Pi sorption must be known; RT is the

solution-to-PSM contact time. Many studies on Pi sorption to PSMs have used batch experiments

but the results of such experiments may overestimate the performance of the PSM, particularly the

precipitation-controlled Pi retention (Klimeski et al., 2012; Stoner et al., 2012; Penn and McGrath,

2011). A flow-through cell is an open system resembling a field-installed drainage water filter that

is considered to provide a more realistic picture of what happens in the field (Klimeski et al., 2012;

Penn and McGrath, 2011). Variables that may influence the outcome of flow-through tests include

PSM particle size and composition, pH, Pi concentration and flow rate (and hence RT) (Stoner et

al., 2012).

Due to the high number of variables affecting Pi sorption and the expense and time consumption of

flow-through cell experimentation, estimation of PSM performance through modeling is attractive.

45

Thus Penn et al. (2012) got a fair estimate of the “lifetime” of a PSM filter material (steel slag) by

means of an empirical exponential model. Based on a screening of five potential PSMs in four

different particle sizes assessed through a batch study conducted by Lyngsie et al. (2013a), the two

best performing PSMs were chosen for further investigation in a flow-through setting. These were

Fe-based CFH and Ca-based Filtralite®P. The specific aims of the study were (i) to test the

performance of the two materials in a flow-through setting, (ii) to use an empirical mechanistic

model to describe the effect of retention time (RT) and inlet Pi concentration on Pi sorption by the

materials in a flow-through setting, and (iii) to model the effect of RT on desorption of Pi from the

P-treated materials, also in a flow-through setting.

2 Materials and methods

2.1 Materials

Filtralite®P, provided by Weber, Norway, is a Light Expanded Clay Aggregates (LECA)-

resembling material calcinated at 1200 ˚C, containing Ca/Mg oxide granules. CFH-12 (CFH)

consists of dried iron oxide, which was provided by Kemira Oyj, Finland.

Prior to the flow-through Pi sorption experiments, the filter materials were fractionated into particle

size intervals of 4-2 mm, 2-1 mm and 1-0.5 mm by sieving. As CFH had no particles >2 mm the

coarser fraction does not exist for this product.

2.2 General characterization of sorption materials

The determination of the mineralogy and specific surface area of the PSMs as well as their chemical

characterization are described in Supplemental Information.

2.3 Flow-through experiments

The flow-through cells were constructed from high-density polyethylene as described in DeSutter et

al. (2006). A diagram of the setup is found in Supplemental Information, Fig. S1. The PSM was

mixed with acid-washed sand (pure Si sand, 14808–60–7; Acros Organics, Morris Plains, NJ) and

total of 5 g of mixed material was placed in the cell. The proportion of PSM to sand depended on

the Pi sorption capacity of the PSM and varied from 0.1 to 3 g (Tables S2-4, S5-6). A suitable

amount resulting in sorption >0 but <100% of the added Pi was determined by trial and error. A

0.22 μm filter was placed beneath the samples and the bottom of the cell was connected to a

peristaltic pump (VWR, variable rate “low flow” 61161–354) using plastic tubing. The desired RT

(RT (min) = pore volume (mL)/flow rate (mL min-1

)), i.e. the time needed for the pore volume to

pass through the cell, was achieved by varying the pump flow rate. Flow rates giving the desired

RTs of 0.5, 1, 1.5, 3, 6 and 9 min were 2.5, 1.26, 0.84, 0.42, 0.21 and 0.14 mL min-1

, respectively.

These RTs represent reasonable times for runoff water to pass through a PSM filter structure in the

46

field (Penn et al., 2012). A Mariotte bottle was used to maintain a constant solution flow. Sorption

experiments were run for 5 h, during which the outlet solution from the cells was sampled at 0, 30,

60, 90, 120, 150, 180, 210, 240, 270 and 300 min, except for the 9 min RT, which ran for 8 h with

hourly sampling. Four different Pi concentrations were tested (1.6, 3.2, 16, and 32 µM) using

solutions made from KH2PO4 in 6 mM KNO3 as background electrolyte. These Pi concentrations

correspond to Pi concentrations found in various drainage waters (Andersen et al., 2006; Jørgensen

et al., 2001; Poulsen and Rubaek, 2005). Immediately after the sorption experiment with RT 0.5, 1,

1.5, 3 and 6 min, the materials were exposed to desorption with 6 mM KNO3. Desorption flow-

through was conducted for two hours with outflow samples taken at 30 min intervals. All solutions

were analysed for Pi by the Murphy-Riley molybdate blue method (Murphy and Riley, 1962) on a

Spec-20 with a limit of detection of 0.05 µM and a limit of quantification of 0.17µM. pH was

measured in outlet samples from experiments with RT of 0.5, 1 and 1.5 min during both sorption

and desorption. Due to small sample volumes, pH was not measured in outlet samples obtained

from RT 3, 6 and 9 min. All flow-through RT and P inlet concentration combinations were

triplicated for each material, resulting in a total of 360 experimental units.

Pi sorption by the materials was calculated using the following equation, where t is sampling time:

[

]

( (

)[

]) ( [

] ( [ ]))

[ ] (Eq. 1)

The Pi desorption was calculated using Eq. 2:

[

]

(

)[

] ( [

] ( [ ]))

[ ] (Eq. 2)

3 Data analysis and model development

3.1 Modelling of Pi sorption

Cumulated sorbed Pi (µmol kg-1

) under flow-through was described as a function of Pi load on the

material (µmol kg-1

) using a hyperbola model:

( ) (Eq. 3)

Where LSS is the Local Sorption Saturation (µmol kg-1

), c is Pi inlet concentration (µmol L-1

), f is

flow rate (L min-1

), t is time (min), m is mass (kg) and K is a coefficient determined from the slope

of the relationship between cumulative sorbed Pi and cumulative Pi load. The quantity f∙c∙t/m is the

amount of Pi added per kg sorbent (µmol kg-1

) at any given time, t. The fitting parameter LSS is an

47

estimate of the maximum Pi sorption capacity under the specific RT and Pi inlet concentration

conditions and corresponds to an outlet Pi concentration equal to the inlet concentration. The fitting

parameter K is a measure of the affinity between sorbent and sorbate.

The model was run in R (version 2.15.0), using the command nls (non-linear least square) and using

individual data points. The output from this analysis gave K and LSS, significant levels for each

parameter and the residual standard error for the model. To test if RT and Pi inlet concentration had

a significant influence on the estimated LSS and K values, a one-way analysis of variance

(ANOVA) was conducted using SigmaPlot ANOVA function.

3.2 Modeling of P desorption

Relative cumulative desorbed Pi (%) under flow-through conditions was described as a function of

the amount of 6 mM KNO3 added to the material (L kg-1

) using a log modified logistic model:

( ( ⁄ ) ) (Eq. 4)

Where RDM is the Relative local Desorption Maximum (% of the previously sorbed Pi), D is a

measure of the lag phase and d is the slope of the curve. RDM, which logically cannot exceed

100%, estimates the relative maximum desorption corresponding to a zero outlet Pi concentration.

The more negative the slope parameter d, the faster RDM is reached while a larger D corresponds to

retardation of RDM.

Eq. 4 was also analyzed in R (version 2.15.0), using the command nls (non-linear least square)

using individual data points. The output from R provided P values for RDM, D and d, in addition to

residual standard error for the model. If the model estimated a RDM value > 100%, RDM was set to

equal 100 (the highest possible value) and the D and d parameters were estimated again. To test if

RT and Pi inlet concentration had a significant influence on the estimated RDM values, a one-way

ANOVA was conducted as described above.

3.3 Prediction model

As LSS and K in Eq. 3 are linearly related to RT and the Pi inlet concentration (c), two multiple

linear regression (MLR) models were constructed to predict the LSS and K obtained by Eq. 3 for

each Pi removal curve as a function of RT and Pi inlet concentration:

LSS=( ) ( ) (Eq. 5a)

K= ( ) ( ) (Eq. 5b)

48

Where α and ε are slope coefficients related to RT, β and ϑ are coefficients related to Pi inlet

concentration, and γ and δ are the intercepts. If one of the independent variables was not significant

the variable was removed and a new reduced MLR was fitted. Inserting Eq. 5a and 5b in Eq. 3 gives

the final sorption model:

(( ) ( ) )

(( ) ( ) ) (Eq. 6)

A similar modeling approach was successfully utilized by Stoner et al. (2012), except that an

exponential function was used to describe P sorption.

4 Results and discussion

4.1 Pi removal and retention by Filtralite®P

An example of a Filtralite®P Pi sorption curve can be seen in Fig. 1a. The fitting parameters

obtained from Eq. 3 can be found in Tables S2-S4 and a summary of LSS and K is given in Fig. 2.

The model parameters are highly significant and the residual standard error of the model is low.

Fig. 1. a) Examples of experimentally determined flow-through P removal curves for CFH and Filtralite®P (2-1 mm

size fraction). Inlet Pi concentration was 16 µM and RT was 1½ min. b) Corresponding experimentally determined

flow-through P relative desorption curve conducted with P-free solution. Desorbed P is presented as a % of previously

sorbed P.

There is no significant difference in LSS and K between the 4-2 mm and 2-1 mm particle size

fractions indicating that they have the similar affinity and capacity for Pi (Fig. 2). Further, there is

no significant difference in LSS and K values between the two inlet Pi concentrations representing

base-flow (1.6 and 3.2 µM). LSS and K values increase significantly as the inlet Pi concentration

increases. RT has a statistically significant positive effect on LSS (Fig. 2) even though there are no

significant differences between adjacent values.

49

Fig. 2. Filtralite®P LSS (a, b, and c) and K values (d, e, and f) estimated for five different RTs and four inlet Pi

concentrations using Eq. 3. among the 4-2 mm (a, d), 2-1 mm (b,e), and 1-0.5 mm (c, f) sized fractions. Error bars

indicate standard error of estimate.

The estimated LSS values were comparable to what was sorbed to Filtralite®P in a batch experiment

by Lyngsie et al. (2013a) when the equilibrium solution Pi concentration reached 1.6, 3.2, 16 and 32

µM (Table S5a). LSS values obtained from low flow-through input Pi concentrations were actually

greater than the sorption values obtained in the batch study (Table S5; 24 min RT). The reason for

the good agreement between the batch and the flow-through is most likely due the short retention

time and the low initial concentration in the batch study. Compared to a classic long-term batch

experiment (Fig. S2) where the RT was 7 days and the initial concentration between 0-5000 µM,

50

the flow-through LSS at the same Pi concentration was between 5 and 32 times less than Langmuir

sorption maxima (Pmax) estimated by the batch experiment (Table S5). The LSS values estimated in

this study when Pi inlet concentration was 32 µM (950-3750 µmol/kg) were also low compared

with the results of Ádám et al. (2007), who found that Filtralite®P sorbed between 4480 and 6770

µmol kg-1

when the outlet concentration reached 32 µM in a flow study. However, this is to be

expected since the study by Ádám et al. (2007) utilized a higher Pi inlet concentration (320 µM)

and longer RT (3.5 to 4 d) compared to the current study. The Pmax obtained for Filtralite-P over a 7

day batch sorption experiment (Fig S2 and Table S5b) are in the same range as Pmax findings of

Ádám et al. (2007) and the variation is most likely due to the different particle size and the low pH

in this study. The Pmax values are, however at least 20 times higher than the highest LSS value for

Filtralite®P (Tables S2-4).

The RT also has a significant influence on the K values, particularly at a RT of 6 and 9 min which

are significantly larger than the K values measured at lower RT (Fig. 2); in general, K, typically

associated with affinity, increases when RT increases. The LSS and K dependency on P inlet

concentration indicates that Filtralite®P sorbs less Pi at lower inlet concentration, meaning that

Filtralite®P will not be as efficient in removing Pi at low concentration of Pi as it is at higher

concentrations. The LSS and K dependency on RT further indicate that Filtralite®

P might have

problems sorbing Pi in peak flow situations.

The lack of significant difference between the 4-2 and 2-1 mm fractions is expected as they have

similar contents of active Ca, i.e. 321±8 mmol kg-1

for the 4-2 mm fraction, 289±14 mmol kg-1

for

the 2-1 mm fraction while the active Ca was 460±8 mmol kg-1

in the 1-0.5 fraction (Table S1). This

may explain why the 1-0.5 mm fraction has greater K and LSS values than the two coarser

fractions.

An example of a Filtralite®P relative desorption curve may be seen in Fig. 1b. The fitting

parameters (D and d) obtained from Eq. 4 can be found in Tables S2-S4 and the relative desorption

maximum (RDM) is shown in Fig. 3. All three Filtralite®P fractions have a minimum RDM of

approximately 35% of the sorbed Pi. The general but not statistically significant tendency is that Pi

sorbed at lower concentration is more prone to desorb than Pi sorbed at higher concentrations. The

difference in RDM between low and high Pi inlet concentrations may be attributed to different

sorption processes such as adsorption at the low Pi concentration and precipitation at the higher

concentration. Further, RDM seems to decrease with increasing RT, i.e. at short RT (fast flow rate)

more Pi is desorbed. This further suggests the dominance of Ca phosphate precipitation mechanism

of previously sorbed P. The dissolution of Ca phosphate is not only enhanced in the flow-through

process by a constant removal of reaction products, but also through a decrease in pH. In the

51

current study, the initial pH was ~10 and decreased (Tables S2-S4) over the 5 h experiment. The pH

further decreased during the 2 h desorption experiment by 0.6 units or less. Lyngsie et al. (2013b)

determined that Pi retention by Filtralite®P is very pH dependent as also found by Herrmann et al.

(2013) and Karabelnik et al. (2012). Thus, when pH decreased below 8.2, Filtralite®P nearly ceased

to sorb Pi and even began to desorb indigenous Pi (Lyngsie et al., 2013b). However, none of the

samples in the current study decreased to a pH below the critical value of 8.2 in the desorption

experiment, which indicates that the desorption was also driven by the Pi concentration gradient

(Tables S2-S4). Similarly, Stoner et al. (2012) found that the ability of a by-product to buffer the

pH > 6.5 was significantly related to P removal in a flow-through setting, specifically for Ca

dominated materials.

Fig. 3. Filtralite

®P relative desorption maximum (RDM) values tested under five different RTs and estimated using Eq.

4. a) 4-2 mm, b) 2-1 mm and c) 1-0.5 mm size fraction from samples previously sorbed with an inlet Pi concentration of

1.6, 3.2, 16, and 32 µM as indicated on the x-axis. Error bars indicate standard error of estimate.

52

These results suggest that even though Filtralite®

P has a relatively high Pi sorption efficiency (Figs.

1a and 2; Lyngsie et al. 2013a), it might have problems retaining sorbed Pi in a system with

fluctuating inlet Pi concentrations where high inlet Pi concentrations are followed by low Pi inlet

concentrations as might be the case when a high Pi concentration in drainage water during peak

flow is followed by low Pi concentrations during base flow conditions (Kronvang et al., 2003; Penn

et al., 2012).

4.2 Pi removal and retention by CFH

An example of a CFH Pi removal curve is seen in Fig. 1a. The fitting parameters obtained from Eq.

3 can be found in Tables S6 and S7. A summary is given in Fig. 4. The model parameters are highly

significant but the standard errors of the model parameters and the residual standard errors of the

model are high. The main reason for the high standard errors of fitting parameters for CFH is that

CFH was never saturated with Pi, but reached a steady state where the outflow Pi concentration

became nearly constant but lower than the inlet concentration, which resulted in a continued and

linear P removal, as exemplified in Fig. 1a.

Fig. 4. CFH LSS (a, b) and K values (c, d) estimated for five different RTs and four inlet Pi concentrations using Eq. 3.

among the 2-1 mm (a, c) and 1-0.5 mm (b, d) sized fractions. Error bars indicate standard error of estimate.

53

Overall LSS and K increased with increasing Pi concentration but decreased with increasing

particle size (Fig. 4). Compared with Filtralite®P (Fig. 2), LSS and K for CFH are at least 10 times

higher. The influence of RT on LSS and K is strongly dependent of the inlet Pi concentration. At

the lowest inlet Pi concentration (1.6 µM), LSS and K tended to increase with increasing RT,

whereas an inverse relationship is occasionally observed at high Pi concentrations (32 µM). While

somewhat counter intuitive, a decrease in P removal with increasing RT has been observed in other

studies (Penn et al., 2014; Stoner et al., 2012), and indicates extremely fast chemical kinetics with

the overall kinetics limited by physical transport. Penn et al (2014) made the same observation on a

soil sample and further explored this phenomenon using calorimetry/kinetic analysis. They found

that the increase in P removal with flow rate was due to increased pore flow turbulence, which

would reduce the thickness of the low velocity boundary layer of fluid near the mineral surface and

result in a smaller boundary layer. The higher flow rates then resulted in faster transport of

reactants from the bulk flow to the mineral surface, and/or faster transport of reaction products

away from the mineral surface into the bulk flow which prevents the complete reaction from

coming to a ”pre-mature” chemical equilibrium.

LSS values estimated for the four inlet concentrations are up to 5 times higher compared to low

flow concentrations and in the same range for the high concentrations as the short-term (24 min)

batch isotherm values from Lyngsie et al. (2013a), Table S5. Consequently, it appears that short-

term batch experiments can serve as an indicator of the sorption capacity at given concentration,

even though it underestimates the sorption capacity at the low concentrations. The LSS values

obtained by flow-through are between one and three magnitude lower than the values obtained by

the long-term batch experiment (Fig. S3, Table S5a) at equilibrium Pi concentrations of 1.6, 3.2, 16

and 32 µM. Hence, the long-term batch experiment overestimates the flow-through sorption

capacity of CFH at a given concentration.

Steady state P removal observed in this study (Fig. 1a) was also observed in a sorption batch

experiment where CFH was titrated with a 10 mM Pi solution (Lyngsie et al., 2013b) and by

Chardon et al. (2012) who conducted a Pi leaching column study with Fe-oxide coated sand and

different amounts of iron sludge. The steady state P removal observed for CFH and other Fe-oxides

may be explained by the fast and strong sorption onto surface sites, which occur at a greater rate

than the Pi migration to the surface of interior sorption sites. The steady outlet P concentration or

level of the steady state seems to be dependent on the amount of Fe oxide in the column i.e. the

sorbent to sorbate ratio (Chardon et al., 2012).

The LSS and K parameters obtained indicate that the 1-0.5 mm fraction has a much higher P

sorption capacity and affinity than the 2-1 mm fraction (Fig 4). Theoretically, this would be

54

expected due to the increased surface area and therefore site density compared to larger particles

(Nair et al., 1984). However, SSA, oxalate-extractable Fe, and Pmax from long term batch

experiments are approximately the same for the two fractions (Tables S1, S5, Fig. S3), suggesting a

different explanation for the different behavior of the two fractions, such as a better accessibility to

the surface of the smaller sized fraction or less limiting physical kinetics as previously described

above.

Fig. 5. CFH relative desorption maximum (RDM) values tested under five different RTs and estimated using Eq. 4. a)

2-1 mm and b) 1-0.5 mm size fraction from samples previously sorbed with an inlet Pi concentration of 1.6, 3.2, 16, and

32 µM as indicated on the x-axis. Error bars indicate standard error of estimate.

An example of a relative desorption curve for CFH is seen in Fig. 1b. The fitting parameters

obtained from Eq. 4 can be found in Tables S6 and S7 and RMD are summarized in Fig. 5. In

general, CFH had RDM values < 10% of the previously sorbed Pi. In contrast to the RDM

dependencies on inlet Pi concentration and RT for Filtralite®P (Fig. 3), RDM for CFH seems

unaffected by inlet Pi concentration as well as RT. This may be due to the high stability of the Pi-

CFH sorption complex as also observed in previous batch experiments conducted on Fe oxide-

based materials (Lyngsie et al., 2013a,b; Chardon et al., 2012; Penn et al., 2011; Harvey and Rhue,

2008; Klimeski et al., 2012; Zeng et al., 2004).

55

The high LSS and K values and the low RDM values obtained for CFH indicates that this material

has a high capacity and affinity for Pi and that P sorption to CFH seems to be a mainly irreversible

reaction. Thus, CFH is more capable of retaining P in systems with varying P inlet contractions

compared to Filtralite®P.

4.3 Prediction models

The MLR analyses show that for CFH neither LSS nor K are significantly dependent on RT (Table

1) as also shown in Fig. 4, which indicate fast sorption of Pi by CFH. RT was therefore excluded

and a new reduced MLR was executed with Pi inlet concentration as the only independent variable.

The high coefficients (β = 188 and 290, and ϑ = 236 and 437 for CFH 2-1 mm and 1-0.5 mm,

respectively) indicate that both K and LSS are strongly dependent on the inlet Pi concentration. The

intercept should ideally be zero or close to zero. However, the high intercepts show that even at low

Pi concentration, the LSS value will be high. For Filtralite®P, the MLR shows that LSS and K

values are significant and positively correlated with both RT and inlet Pi concentration as also

shown in Fig. 2. The fact that Filtralite®P has a lower Pi sorption capacity and affinity compared to

CFH is reflected in observation of the respective coefficients obtained with Eqs. 5a and 5b.

Table 1. Model coefficients obtained from Eq. 5a and 5b for prediction LSS and K respectively for each material.

NS – not significant. Significant level codes; 0 = ‘***’, 0.001 <‘**’, 0.01< ‘*’, and 0.05< ‘.’

56

The dependency of LSS and K on RT and P inlet concentrations observed for Filtralite®P was also

observed for other Ca-based PSMs with low pH buffer capacity such as gypsum (Stoner et al.,

2012). However, Stoner et al. (2012) also found that Pi sorption by Fe/Al oxide-based PSMs or Ca-

based PSMs with high pH buffer capacity such as fly-ash did not show a dependency on RT.

Based on the description given for a Pi removal structure in Penn et al. (2012) a calculation example

of the capacity and efficiency of CFH and Filtralite®P 0.5-1 mm was generated. Approximately

90% of Pi added to the systems comes in rainfall events (Penn et al. 2012). A typical example of a

rainfall event could produce a flow rate of 130 L min-1

and we then assumed an average Pi inlet of

20 µM. The RT for the six largest rainfall events that delivered 75% of the Pi to the P removal

structure was 8.9 min (Penn et al. 2012). If the flow of 130 L min-1

lasted for 7 h (420 min), 1.1 mol

Pi is added to the structure in the rainfall event. The removal structure holds approximately 2.7 ton

PSM (Penn et al. 2012) i.e. approximately 400 µmol Pi kg-1

is added under these conditions.

According to Penn et al. (2012), 7.7 mol Pi is added to this system over a five month monitoring

period, which gives an average loading of nearly 600 µmol Pi kg-1

month-1

to the structure.

Continuing with this scenario of RT of 9 min and Pi inlet concentration of 20 µM, LSS values for

CFH and Filtralite®P 1-0.5 mm fraction can be calculated by Eq. 5a to yield approximately 13000

µmol P kg-1

and 2300 µmol P kg-1

, respectively. Because LSS is asymptotic there is no meaning in

calculating a filter’s lifetime based on LSS alone as the associated P loading is impossible to

calculate. When the Pi removal curve (Figure 6) approaches LSS, the filters’ affinity towards Pi

becomes very low (i.e. the slope of the curve is less steep) and the filter removes very little.

Therefore a practical cutoff Pi removal value needs to be set, which can be estimated by

normalizing the slope of the Pi removal curve relative to the initial slope. For this example, the

cutoff value will be the point at which the slope equals 1% of the initial slope, which corresponds to

an added Pi of 9×K. Due to the nature of the model, this 1% slope cutoff value is equal to 90% of

the filter’s full LSS. If the material is P loaded to 90% of the full LSS, then Filtralite®P and CFH

would have a lifetime of about three and half and 20 years, respectively, for the described scenario

assuming an average Pi loading of 600 µmol kg-1

month-1

. However, at the end of that Pi removal

lifetime, Filtralite®P and CFH would both only remove 8% of the added Pi.

57

Fig. 6 Examples of P removals curves calculated via the prediction model (Eq. 6) and coefficients obtained via Eq. 5a

and b (Table 1) for a) Filtralite®P and b) CFH. Inlet P concentration is set to 20 µM and RT to 9 min. The green solid

line indicates the LSS cutoff value and the blue dotted line indicate 50% of the LSS value of respective material. Note

the different scale y-axis.

From the curve in Fig. 6 generated by Eq. 6 it is clear that the affinity for Pi and P removal

efficiency decreases as the material becomes saturated with Pi. For the following example, we

calculated the hypothetical replacement time based on the assumption of material replacement after

it reaches 50% (instead of 90%) of the maximum Pi sorbed (i.e. LSS). This is indicated by the blue

58

dotted line in the inset of Fig. 6. This means that CFH could retain 6500 µmol Pi kg-1

and

Filtralite®P 1150 µmol Pi kg

-1 before it would need to be replaced. Further, assuming that this

hypothetical Pi removal structure also hold 2.7 ton Filtralite®P and the average loading rate was 600

µmol µmol Pi kg-1

month-1

, it would have a removal capacity of 3.1 mol Pi structure-1

and an

estimated lifetime of five month before the material needed to be changed. For the same material

mass and scenario, CFH would have a Pi removal capacity of 18 mol structure-1

, leading to a

lifetime of 27 months before the material is P saturated to 50% of the LSS. When the material

reaches 50% of LSS, CFH would retain 41% of the cumulative added Pi and Filtralite®P would

retain ~38%, meaning that even though Pi is retained in the system, an appreciable amount of Pi

escaped i.e. 59 and 62% of P added to CFH and Filtralite®

P, respectively. If a higher P removal or a

longer lifetime is desired, a larger structure with more PSM is needed. Furthermore, it should be

kept in mind that the desorption of the particle is not accounted for in this example and that

Filtralite®P desorbed more than 35% but CFH less than 10% of the previously retained Pi, when Pi

free solution was added during the flow-through experiments. In order to validate the new model

approach, field experiments need to be conducted.

Conclusion

The flow-through approach described in this paper illustrates the influence of RT and inlet Pi

concentration on the Pi sorption capacity and affinity for the PSMs tested. The Fe oxide-based CFH

has a Pi sorption capacity and affinity up to 10 times higher than the Ca-based Filtralite®P. The

difference is especially pronounced at the low base flow concentrations (1.6 and 3.2 µM). CFH

released less than 10% of previously sorbed P compared to Filtralite®P, which released ≥35%

during the desorption test. Both materials’ Pi sorption capacity and affinity is highly dependent on

the Pi inlet concentration which illustrates how important it is to test a PSM using anticipated field

concentrations. Furthermore, Pi sorption by Filtralite®P was also positively correlated to RT. This

was not the case for CFH indicating that CFH will be capable of removing Pi even at high flow

rates. Therefore, CFH would be the preferred material for a P removal structure with fluctuating

inlet Pi concentrations and retention times. From the flow-through approach a PSM specific

prediction model aids in predicting how much Pi can be removed and how long a specific material

will last until Pi saturation if the Pi loading rate for a specific site is known.

Supplemental information (SI)

Fig. S1Diagram of the flow-through cell used to evaluate the effect of retention time and Pi

concentration on Pi sorption. Figure from Penn and McGrath. 2011.

Fig. S2 Filtralite®P isotherm obtained from log-term sorption experiment (7d) and high initial Pi

concentrations (0-5000µM). Error bars are standard deviations. Pmax and K values are obtained by

59

the Langmuir equation (Sorbed P = Pmax∙ Equilibrium P con. ∙ KL/(1+KL∙ Equilibrium P con. ) on

non-averaged data.

Fig. S3. CFH isotherm obtained from log-term sorption experiment (7d) and high initial Pi

concentrations (0-20000µM). Error bars are standard deviations. Pmax and K values are obtained

by the Langmuir equation (Sorbed P = Pmax∙ Equilibrium P con. ∙ KL/(1+KL∙ Equilibrium P con. )

on non-averaged data.

Table S1. General characterization of the materials used in this study for parameters that may

impact phosphorus (P) sorption.

Table S2. Model parameters and pH for Filtralite®P 4-2 mm.

Table S3. Model parameters and pH for Filtralite®P 2-1 mm.

Table S4. Model parameters and pH for Filtralite®P 1-0.5 mm fraction.

Table S5a Sorbed Pi (µmol/kg) when equilibrium Pi concentrations are 1.6, 3.2, 16 and 32 µM

Table S5b Pmax values for selected studies

Table S6. Model parameters and pH for CFH 2-1 mm.

Table S7. Model parameters and pH for CFH 1-0.5 mm

Acknowledgement

The project was carried out as part of the SupremeTech project funded by The Danish Council for

Strategic Research (grant no. 09-067280). Special thanks to Anita Schjødt Sandager for laboratory

assistance.

60

References

Ádám, K., Krogstad, T., Vråle, L., Søvik, A.K., Jenssen, P.D., 2007. Phosphorus retention in the

filter materials shellsand and Filtralite P®—Batch and column experiment with synthetic P solution

and secondary wastewater, Ecol. Eng. 29, 200-208.

Andersen, H.E., Larsen, S.E., Kronvang, B., Hansen, K.M., Laubel, A., Windolf, J., Muus, K.,

2006. Fosfat i drænvand, Vand og jord 13, 152-156. (In Danish)

Ballantine, D.J., Tanner, C.C., 2010. Substrate and filter materials to enhance phosphorus removal

in constructed wetlands treating diffuse farm runoff: a review, New Zealand J. Agric. Res. 53, 71-

95.

Buda, A.R., Koopmans, G.F., Bryant, R.B., Chardon, W.J., 2012. Emerging technologies for

removing nonpoint phosphorus from surface water and groundwater: Introduction, J. Environ. Qual.

41, 621-627.

Chardon, W.J., Groenenberg, J.E., Temminghoff, E.J.M., Koopmans, G.F., 2012. Use of reactive

materials to bind phosphorus, J. Environ. Qual. 41, 636-646.

Cucarella, V., Renman, G., 2009. Phosphorus sorption capacity of filter materials used for on-site

wastewater treatment determined in batch experiments - A comparative study, J. Environ. Qual. 38,

381-392.

DeSutter, T.M., Pierzynski, G.M., Baker, L.R., 2006. Flow-through and batch methods for

determining calcium-magnesium and magnesium-calcium selectivity, Soil Sci. Soc. Am. J. 70, 550-

554.

Grant, R., Laubel, A., Kronvang, B., Andersen, H.E., Svendsen, L.M., Fuglsang, A., 1996. Loss of

dissolved and particulate phosphorus from arable catchments by subsurface drainage, Water Res.

30, 2633-2642.

Groenenberg, J.E., Chardon, W.J., Koopmans, G.F., 2013. - Reducing phosphorus loading of

surface water using iron-coated sand, J. Environ. Qual. 42, 250-259.

61

Harvey, O., Rhue, R., 2008. Kinetics and energetics of phosphate sorption in a multi-component Al

(III)–Fe (III) hydr (oxide) sorbent system, J. Colloid Interface Sci. 322, 384-393.

Herrmann, I., Jourak, A., Hedström, A., Lundström, T.S., Viklander, M., 2013. The effect of

hydraulic loading rate and influent source on the binding capacity of phosphorus filters, PloS one 8,

e69017.

Johansson Westholm, L., 2006. Substrates for phosphorus removal—Potential benefits for on-site

wastewater treatment? Water Res. 40, 23-36.

Johnes, P.J., 2007. Uncertainties in annual riverine phosphorus load estimation: Impact of load

estimation methodology, sampling frequency, baseflow index and catchment population density, J.

Hydro. 332, 241-258.

Jørgensen, J.O., Kronvang, B., Paulsen, I., 2001. Fosfor i jord og vand – udvikling, status og

perspektiver, 380, 45-54. (In Danish)

Kaasik, A., Vohla, C., Mõtlep, R., Mander, Ü, Kirsimäe, K., 2008. Hydrated calcareous oil-shale

ash as potential filter media for phosphorus removal in constructed wetlands, Water Res. 42, 1315-

1323.

Karabelnik, K., Kõiv, M., Kasak, K., Jenssen, P.D., Mander, Ü, 2012. High-strength greywater

treatment in compact hybrid filter systems with alternative substrates, Ecol. Eng. 49, 84-92.

Klimeski, A., Chardon, W.J., Turtola, E., Uusitalo, R., 2012. Potential and limitations of phosphate

retention media in water protection: A process-based review of laboratory and field-scale tests,

Agric. Food Sci. 21, 206-223.

Kronvang, B., Bechmann, M., Pedersen, M.L., Flynn, N., 2003. Phosphorus dynamics and export in

streams draining micro‐catchments: Development of empirical models, J. Plant Nutr. Soil Sci. 166,

469-474.

Kronvang, B., Behrendt, H., Andersen, H.E., Arheimer, B., Barr, A., Borgvang, S.A., Bouraoui, F.,

Granlund, K., Grizzetti, B., Groenendijk, P., Schwaiger, E., Hejzlar, J., Hoffmann, L., Johnsson, H.,

Panagopoulos, Y., Lo Porto, A., Reisser, H., Schoumans, O., Anthony, S., Silgram, M., Venohr, M.,

62

Larsen, S.E., 2009. Ensemble modelling of nutrient loads and nutrient load partitioning in 17

European catchments, J. Environ. Monit. 11, 572-583.

Lyngsie, G., Borggaard, O.K., Hansen H.C.B., 2013a. Testing phosphate sorption efficiency of five

potential drainage water filter materials under field-relevant conditions, Water Res. In Print.

http://dx.doi.org/10.1016/j.watres.2013.10.061

Lyngsie, G., Penn, C.J., Hansen H.C.B., Borggaard, O.K., 2013b. Phosphate sorption by three

potential filter materials as assessed by isothermal titration calorimetry, Submitted to J. Environ.

Manag.

Murphy J., Riley J.P., 1962. A modified single solution method for the determination of phosphate

in natural waters, Anal. Chim. Acta. 27, 31-6.

Nair, P., Logan, T., Sharpley, A., Sommers, L., Tabatabai, M., Yuan, T., 1984. Interlaboratory

comparison of a standardized phosphorus adsorption procedure, J. Environ. Qual. 13, 591-595.

Penn, C.J., Heeren, D., Fox, G., Kumar, A.J. 2014. Application of isothermal calorimetry to

phosphorus sorption onto soils in a flow-through system, Soil Sci. Soc. Am. J. In Print.

doi:10.2136/sssaj2013.06.0239

Penn, C.J., McGrath, J.M., Rounds, E., Fox, G., Heeren, D., 2012. Trapping phosphorus in runoff

with a phosphorus removal structure, J. Environ. Qual. 41, 672-679.

Penn, C.J. and McGrath, J.M. 2011. Predicting phosphorus sorption onto normal and modified slag

using a flow-through approach, J. Wat. Res. Protec. 3:235-244.

Penn, C.J., Bryant, R.B., Callahan, M.P., McGrath, J.M., 2011. Use of industrial by-products to

sorb and retain phosphorus, Commun. Soil Sci. Plant Anal. 42, 633-644.Penn, C.J., Bryant, R.B.,

Kleinman, P.J.A., Allen, A.L., 2007. Removing dissolved phosphorus from drainage ditch water

with phosphorus sorbing materials, J. Soil Water Cons. 62, 269-276.

Poulsen, H.D., Rubaek, G.H., 2005. Fosfat i dansk landbrug - Omsætning, tab og virkemidler mod

tab. DJF rapport Husdyrbrug, 93-121. (in Danish)

63

Stoner, D., Penn, C., McGrath, J., Warren, J., 2012. Phosphorus removal with by-products in a

flow-through setting, J. Environ. Qual. 41, 654-663.

Ulén, B., 1995. Episodic precipitation and discharge events and their influence on losses of

phosphorus and nitrogen from tile-drained arable fields, Swedish J. Agric. Res. 35, 25-31.

Vohla, C., Kõiv, M., Bavor, H.J., Chazarenc, F., Mander, Ü, 2011. Filter materials for phosphorus

removal from wastewater in treatment wetlands—A review, Ecol. Eng. 37, 70-89. Zeng, L., Li, X.,

Liu, J., 2004. Adsorptive removal of phosphate from aqueous solutions using iron oxide tailings,

Water Res. 38, 1318-1326.

64

65

Supplemental Information

Phosphorus sorption and desorption on Filtralite®P and an iron-rich material under flow-through conditions:

Influence of retention time and phosphate concentration

Lyngsie, Gry1*

, Penn, Chad J.2, Pedersen, H.L.

3, Borggaard, Ole K.

1, Hansen, Hans C. B.

1

1University of Copenhagen, Department of Plant and Environmental Sciences, Thorvaldsensvej 40, DK-1871 Frederiksberg C.

2Oklahoma State University, Department of Plant and Soil Sciences, 368 Agricultural Hall, Stillwater, OK 74078-6028.

3University of Copenhagen, Department of Mathematical Sciences, Universitetsparken 5, DK-2100 København Ø

*Corresponding author: chj177@ku.dk

11 Pages

3 Figures

7 Tables

66

Fig. S1Diagram of the flow-through cell used to evaluate the effect of retention time and Pi concentration on Pi sorption. Figure from Penn and McGrath. 2011.

67

Fig. S2 Filtralite®P isotherm obtained from log-term sorption experiment (7d) and high initial Pi concentrations (0-5000µM). Error bars are standard deviations. Pmax and K values are

obtained by the Langmuir equation (Sorbed P = Pmax∙ Equilibrium P con. ∙ KL/(1+KL∙ Equilibrium P con. ) on non-averaged data.

68

Fig. S3. CFH isotherm obtained from log-term sorption experiment (7d) and high initial Pi concentrations (0-20000µM). Error bars are standard deviations. Pmax and K values are obtained by

the Langmuir equation (Sorbed P = Pmax∙ Equilibrium P con. ∙ KL/(1+KL∙ Equilibrium P con. ) on non-averaged data.

69

General characterization

The mineralogy of the materials was assessed by X-ray diffraction analysis

on unoriented samples using a Siemens D5000 instrument equipped with

Co-Kα radiation and a diffracted beam monochromator. Diffractograms

were recorded from 5 to 80o 2θ using 0.02

o 2θ steps and a step speed of 10

s. Diffraction peak positions were used to calculate d-values for mineral

identification. pH was measured potentiometrically in 0.01 M CaCl2 using a

material:solution ratio of 1:2.5. Total Al, Fe, Ca and Mg (AlTotal, FeTotal,

CaTotal, MgTotal) were determined after dissolution of the materials in a

mixture of concentrated nitric acid, hydrogen peroxide, hydrochloric acid

and hydrofluoric acid (EPA 3052). Oxalate-extractable aluminium (Alox)

and iron (Feox) were determined by extraction with 0.2 M ammonium

oxalate for 2 h at pH 3 in the dark (Schwertmann. 1964) and determined by

AAS. To determine maximum amount of active Ca, each size fraction of the

Filtralite-P was extracted for 1 h with 1 M hydrochloric acid using a

material:solution ratio of 1:12.5 and the contents of Ca in the extracts were

determined by AAS. Total inorganic P (Pinorg) was determined by extracting

bulk material with 6 M H2SO4 for 10 min at 70 °C. The concentration of P

in the extract was determined by the molybdenum-blue method (Murphy

and Riley. 1962). All analyses were carried out in triplicates. The specific

surface area (SSA) of the different fractions was determined by applying the

BET equation to N2 adsorption data obtained by means of a Micromeritic

Gemini VII 2390a instrument (Brunauer et al. 1938) after the samples have

been under 12 h of vacuum at room temperature. The reported SSAs are the

average of five separate measurements.

Table S1. General characterization of the materials used in this study for parameters that

may impact phosphorus (P) sorption.

CFH Filtralite®P

Bulk 2-1 mm 1-0.5 mm Bulk 2-4 mm 2-1 mm 1-0.5 mm

AlTotal†

mmol kg-1

32±2 ND ND 2345±161 ND ND ND

FeTotal† 7360±420 ND ND 737±30 ND ND ND

FeOx‡ 7310±90 7380±91 7440±230 22±0.2 ND ND ND

CaTotal 658±23 ND ND 1720±56 ND ND ND

CaHCl§ ND ND ND ND 321±8 289±14 460±8

MgTotal 1390±97 ND ND 1562±110 ND ND ND

MgHCl§ ND ND ND ND 493±5 344±8 346±12

Pinorg.¶ 8±0.3 ND ND 3±0.1 ND ND ND

SSA†† m2 g-1 ND 32±2 32±11 ND 0.9±0.1 0.9±0.2 1.2±0.0

pH¤ 8.2±0.1 ND ND 11.9±0.2 ND ND ND

† Total digestion: EPA 3052, ‡ Extraction with ammonium oxalate, § 1 M HCl, ¶ 6 M H2SO4, †† Specific surface

area, ¤ pH measured in 0.01 M CaCl2, ND: Not determined.

70

Table S2. Model parameters and pH for Filtralite®P 4-2 mm.

Sorption† Desorption‡ pH

Inlet P conc RT Mass

LSS Std. Err. l.s K Std. Err. l.s R std err

RDM Std. Err. l.s D Std. Err. l.s d Std. Err. l.s R std err

Start sorp. SD End sorp. SD End desorp. SD

µM Min g

µmol/kg

%

Fil

tral

ite®

P 4

-2 m

m

1.6

½ 2.0

135 18 *** 177 67 * 19

100 §

866 312 * -1.34 0.1 *** 1.64

9.5 0.49 9.4 0.43 8.9 0.01

1 2.0

158 3 *** 231 9 *** 1

49 12.2 * 821 456 0.13 -1.67 0.3 ** 0.85

9.8 0.30 9.6 0.40 9.6 0.03

1½ 1.5

164 5 *** 245 12 *** 1

56 18.1 * 790 894 0.42 -1.74 0.5 * 2.11

9.4 0.36 9.2 0.18 9.4 0.04

3 1.0

103 12 *** 160 34 *** 5

60 18.6 * 404 451 0.41 -1.73 0.5 * 2.59

6 0.75

469 144 ** 924 316 ** 2

18 5.1 * 721 2790 0.81 -2.56 1.7 0.20 3.24

9 0.50

225 41 *** 319 87 ** 6

3.2

½ 2.0

182 4 *** 192 16 *** 7

56 12.0 ** 408 216 . -1.4 0.2 *** 1.30

10.1 0.07 9.7 0.06 9.3 0.06

1 2.0

146 3 *** 185 11 *** 3

51 7.1 ** 948 603 0.18 -1.9 0.2 *** 1.11

9.7 0.46 9.6 0.18 9.1 0.15

1½ 1.5

217 9 *** 285 24 *** 5

49 11.2 ** 698 282 0.28 -1.1 0.4 ** 1.37

9.9 0.05 9.8 0.03 9.5 0.10

3 1.0

249 17 *** 379 44 *** 5

29 0.6 ** 307 110 * -1.5 0.2 *** 0.53

6 0.75

422 72 *** 667 148 *** 5

100 §

9240 45000 0.84 -2.5 1.5 0.14 15.15

9 0.51

382 44 *** 510 94 *** 9

16

½ 2.0

782 10 *** 845 48 *** 19

33 2.5 *** 362 142 * -1.5 0.1 *** 0.62

9.9 0.11 9.5 0.03 9.4 0.09

1 2.0

549 22 *** 628 86 *** 31

39 7.1 ** 704 829 0.434 -1.9 0.4 ** 1.73

9.8 0.52 9.7 0.06 9.8 0.01

1½ 1.5

715 16 *** 972 57 *** 14

41 5.1 *** 807 694 0.30 -1.9 0.3 ** 1.28

9.8 0.03 9.6 0.02 9.4 0.08

3 1.0

877 60 *** 1320 177 *** 28

26 2.1 *** 375 153 . -1.8 0.2 *** 0.44

6 0.75

987 66 *** 1640 172 *** 14

26 2.8 *** 208 125 0.16 -1.9 0.3 ** 0.77

9 0.50

1260 114 *** 2010 318 *** 33

32

½ 3.0

1160 29 *** 1410 131 *** 45

38 6.8 *** 218 195 0.30 -1.6 0.4 ** 1.92

9.5 0.15 8.4 0.01 8.5 0.13

1 3.0

949 64 *** 1260 237 *** 65

45 4.8 *** 357 187 0.12 -1.9 0.1 *** 1.02

9.9 0.07 9.8 0.03 9.9 0.04

1½ 2.5

1390 27 *** 1980 79 *** 14

37 2.8 *** 324 87 * -1.8 0.1 *** 0.42

10.0 0.09 9.4 0.46 9.5 0.14

3 1.5

1480 104 *** 2270 284 *** 36

34 9.4 * 200 199 0.36 -1.8 0.5 * 1.64

6 1.0

2060 288 *** 3140 637 *** 43

27 8.1 * 108 137 0.47 -1.9 0.7 . 2.00

9 0.75 2160 349 *** 3440 887 *** 73

Standard error ‘Std. err.’, level of significant ‘l.s’, residual standard error ‘R std. err’ and standard error ‘SD’. Significant level codes; 0 = ‘***’, 0.001 <‘**’, 0.01< ‘*’, and 0.05< ‘.’

† Parameters obtained via Eq. 3 on non-averaged data

‡ Parameters obtained via Eq. 4 on non-averaged data

§ D parameter set to 100%

71

Table S3. Model parameters and pH for Filtralite®P 2-1 mm.

Sorption† Desorption‡ pH

Inlet P conc RT Mass

LSS Std. Err. l.s K Std. Err. l.s R std err

RDM Std. Err. l.s D Std. Err. l.s d Std. Err. l.s R std err

Start sorp. SD End sorp. SD End desorp. SD

µM min g

µmol/kg

%

Fil

tral

ite®

P 2

-1 m

m

1.6

½ 2.0

107 2 *** 80 8 *** 4

100 §

667 294 * -1.5 0.1 *** 2.94

9.7 0.34 9.5 0.18 9.4 0.53

1 2.0

127 3 *** 140 7 *** 2

75 8 *** 776 176 ** -1.7 0.1 *** 0.55

9.8 0.26 9.3 0.03 9.5 0.00

1½ 1.5

161 13 *** 194 29 *** 5

78 7 *** 792 178 ** -1.7 0.1 *** 0.56

9.7 0.09 9.4 0.11 9.4 0.08

3 1.0

213 8 *** 266 16 *** 2

54 11 ** 545 219 . -1.7 0.2 *** 0.77

6 0.75

253 59 *** 409 119 ** 3

39 8 ** 240 91 * -1.7 0.2 *** 0.66

9 0.50

284 47 *** 365 87 *** 5

3.2

½ 2.0

138 4 *** 90 15 *** 9

96 7 *** 670 152 ** -1.6 0.1 ** 0.81

10.1 0.05 9.7 0.08 9.5 0.00

1 2.0

156 3 *** 165 10 *** 4

73 14 ** 666 387 0.15 -1.7 0.2 *** 1.43

9.8 0.13 9.6 0.12 9.4 0.10

1½ 1.5

209 14 *** 244 37 *** 10

54 11 ** 773 559 0.23 -1.7 0.3 ** 1.30

10.0 0.02 9.5 0.54 8.9 0.14

3 1.0

301 34 *** 358 71 *** 11

37 20 0.13 363 434 0.44 -1.6 0.7 * 1.70

6 0.75

401 17 *** 528 31 *** 2

53 32 0.15 270 174 0.18 -1.6 0.5 * 1.55

9 0.51

362 37 *** 395 72 *** 11

16

½ 2.0

889 12 *** 845 49 *** 22

36 4 *** 467 219 . -1.6 0.2 *** 0.69

9.9 0.05 9.5 0.10 9.2 0.48

1 2.0

736 24 *** 766 76 *** 28

41 6 ** 699 589 0.29 -1.9 0.3 ** 1.27

10.1 0.01 9.9 0.03 9.8 0.07

1½ 1.5

717 10 *** 767 30 *** 11

42 2 *** 883 228 * -1.9 0.1 *** 0.39

9.8 0.06 9.1 0.44 9.1 0.63

3 1.0

865 40 *** 1040 104 *** 25

35 13 * 366 634 0.59 -1.8 0.8 . 2.44

6 0.75

1020 120 *** 1150 234 *** 41

26 12 . 211 337 0.56 -1.8 0.8 . 1.99

9 0.51

1340 45 *** 1650 105 *** 17

32

½ 3.0

1220 100 *** 1300 417 *** 164

36 5 *** 291 199 0.19 -1.7 0.3 *** 1.78

10.0 0.05 9.4 0.21 9.1 0.01

1 3.0

984 14 *** 1120 47 *** 16

46 2 *** 358 85 ** -1.9 0.1 *** 0.46

9.9 0.02 9.8 0.00 9.5 0.17

1½ 2.5

1200 23 *** 1330 61 *** 18

37 5 *** 306 208 0.20 -1.9 0.3 ** 1.11

9.9 0.02 9.5 0.29 9.1 0.20

3 1.5

1550 92 *** 1950 220 *** 39

34 7 ** 202 175 0.30 -1.8 0.4 ** 1.41

6 1.0

1930 169 *** 2580 350 *** 34

27 6 ** 131 107 0.27 -1.9 0.5 * 1.24

9 0.75 2000 108 *** 2340 232 *** 37

Standard error ‘Std. err.’, level of significant ‘l.s’, residual standard error ‘R std. err’ and standard error ‘SD’. Significant level codes; 0 = ‘***’, 0.001 <‘**’, 0.01< ‘*’, and 0.05< ‘.’

† Parameters obtained via Eq. 3 on non-averaged data

‡ Parameters obtained via Eq. 4 on non-averaged data

§ D parameter set to 100%

72

Table S4. Model parameters and pH for Filtralite®P 1-0.5 mm fraction.

Sorption† Desorption‡ pH

Inlet P conc RT Mass

LSS Std. Err. l.s K Std. Err. l.s R std err

RDM Std. Err. l.s D Std. Err. l.s d Std. Err. l.s R std err

Start sorp. SD End sorp. SD End desorp. SD

µM min g

µmol/kg

%

Fil

tral

ite®

P 1

-0.5

mm

1.6

½ 2.0

182 13 *** 129 31 *** 18

94 35 * 628 351 0.11 -1.4 0.2 *** 2.09

10.2 0.05 9.7 0.08 9.6 0.28

1 2.0

170 11 *** 168 23 *** 6

76 52 0.20 1560 5465 0.79 -2.0 1.3 0.19 8.46

10.4 0.05 10.1 0.03 9.9 0.03

1½ 1.0

195 8 *** 180 17 *** 6

100 §

2430 738 * -1.7 0.1 *** 1.42

10.0 0.06 9.6 0.03 9.4 0.05

3 0.75

244 26 *** 317 54 *** 6

60 73 0.45 875 2297 0.72 -1.7 1.2 0.22 4.82

6 0.50

379 34 *** 546 63 *** 2

40 10 * 1170 1611 0.50 -2.1 0.6 * 1.90

9 0.30

416 48 *** 551 94 *** 7

3.2

½ 2.0

281 8 *** 215 23 *** 12

63 13 ** 635 328 . -1.6 0.2 *** 1.20

10.4 0.19 9.6 0.49 9.4 0.25

1 2.0

351 26 *** 353 54 *** 14

48 12 * 887 669 0.24 -1.8 0.3 ** 1.16

10.3 0.12 10.1 0.04 9.5 0.04

1½ 1.0

291 6 *** 268 16 *** 6

60 6 *** 1720 802 . -1.8 0.2 *** 0.81

9.9 0.04 9.4 0.53 8.8 0.06

3 0.75

462 11 *** 549 22 *** 3

48 18 * 739 777 0.38 -1.7 0.4 * 1.66

6 0.50

836 90 *** 1160 157 *** 4

50 63 0.47 425 418 0.36 -1.5 0.7 0.10 1.96

9 0.30

681 63 *** 712 112 *** 17

16

½ 2.0

1080 32 *** 923 113 *** 55

37 6 *** 441 361 0.26 -1.6 0.3 ** 1.29

10.2 0.04 9.8 0.05 9.5 0.09

1 2.0

1290 23 *** 1220 54 *** 18

35 6 ** 726 661 0.32 -1.9 0.4 ** 1.15

10.4 0.02 10.1 0.03 9.9 0.09

1½ 1.0

1260 22 *** 1300 62 *** 21

42 2 *** 1600 592 * -1.9 0.1 *** 0.49

9.4 0.07 9.3 0.31 9.2 0.01

3 0.75

1440 41 *** 1660 95 *** 21

31 2 *** 644 230 * -1.8 0.1 *** 0.40

6 0.50

1500 107 *** 1670 212 *** 38

32 12 * 465 848 0.61 -1.8 0.8 . 2.29

9 0.30

1840 129 *** 2420 337 *** 56

32

½ 3.0

2360 415 *** 2650 1260 * 399

35 2 *** 276 65 ** -1.7 0.1 *** 0.39

10.1 0.15 9.5 0.11 9.1 0.41

1 2.6

2230 140 *** 2510 338 *** 83

37 3 *** 726 162 * -1.8 0.2 *** 0.57

10.1 0.28 10.0 0.04 9.5 0.31

1½ 1.5

2290 83 *** 2370 203 *** 61

39 6 ** 743 584 0.26 -1.8 0.3 ** 1.08

10.1 0.05 9.0 0.61 9.3 0.45

3 1.0

2690 110 *** 3140 237 *** 42

37 22 0.15 434 1360 0.76 -1.9 1.3 0.22 4.77

6 0.75

3760 384 *** 4740 679 *** 47

29 3 *** 234 91 . -1.8 0.2 *** 0.53

9 0.50 3540 214 *** 4310 443 *** 57

Standard error ‘Std. err.’, level of significant ‘l.s’, residual standard error ‘R std. err’ and standard error ‘SD’. Significant level codes; 0 = ‘***’, 0.001 <‘**’, 0.01< ‘*’, and 0.05< ‘.’

† Parameters obtained via Eq. 3 on non-averaged data

‡ Parameters obtained via Eq. 4 on non-averaged data

§ D parameter set to 100%

73

Long-term batch sorption experiments The sorption isotherms was carried out as follows: Nine 1 g samples of each of the

different particle sizes of the two filter materials were gently shaken horizontally in blue-

cap bottles for 7 d with 100 mL 6 mM KNO3 and nine initial KH2PO4 concentration

adjusted to pH 6 with 0.1 M NaOH. Initial KH2PO4 concentrations ranging between 0 -

5000 µM for Filtralite-P and 0-20000 µM for CFH. After shaking, pH was measured and

an aliquot of the solution was withdrawn with a syringe and filtered through 0.2 µm

Millipore syringe-filter. The phosphate concentrations in the filtrates were determined by

the molybdenum blue method using flow injection analysis on a FIAstar 5000 instrument

(Ruzicka and Hansen. 1988). Sorbed Pi in µmol/kg was calculated from the difference

between the Pi concentrations before and after shaking with the filter materials. All

experiments were carried out in triplicate. The glass- and plastic wares were acid-washed,

the chemicals were pro analysis or of better quality and triple deionized water was used

throughout.

Table S5a Sorbed Pi (µmol/kg) when equilibrium Pi concentrations are 1.6, 3.2, 16 and 32 µM

Filtralite®P

CFH

Equi. P

concentration µM

Sorbed Pi (µmol/kg)

24 min§ 7 d#

24 min§ 7 d¤

1.6

0.5-1 mm 170 2000

2400 3.5 *10^4

2-1 mm 70 1000

560 3.5 *10^4

4-2 mm 60 4500

3.2

0.5-1 mm 750 4400

3450 4.8*10^4

2-1 mm 170 1300

800 5.4*10^4

4-2 mm 120 6000

16

0.5-1 mm 17000 14900

6700 20*10^4

2-1 mm 600 6000

2700 12*10^4

4-2 mm 700 10100

32

0.5-1 mm nd 21700

8600 28*10^4

2-1 mm 2300 9300

4600 19*10^4

4-2 mm 3800 16000

§From Lyngsie et al. 2013. Initial P conc. 0-161 µM # Initial P conc. 0-5000µM, ¤ Initial P conc. 0-20000 µM

Table S5b Pmax values for selected studies

Particle

size

pH

Reaction

time

Pmax

µmol/kg

Filtralite®P 0.5-1 mm 8.1 7 d (6.55±0.62)*104 This study

Filtralite®P 2-1 mm 7.8 7 d (4.46±0.18)*104 This study

Filtralite®P 4-2 mm 7.8 7 d (7.20±0.34)*104 This study

Filtralite®P ?

7.75*104 Ádám et al. (2007)

CFH 0.5-1 mm 7.4 7 d (1.24±0.04)*106 This study

CFH 2-1 mm 7.3 7 d (1.20±0.04)*106 This study

Amorphous Hydrouse

iron oxide (FeOOH)

3.5 24 h 9.15*105 Parfitt et al. (1975)

Akaganeite (β-FeOOH) 0.32-2 mm 5.5 4 d 7.22*105 Genz et al. (2004)

Akaganeite (β-FeOOH) 0.32-2 mm 8.2 4 d 5.23*105 Genz et al. 2004

Akaganeite (β-FeOOH)

3.5 24 h 8.28*105 Parfitt et al. 1975

Lepidocrocite (γ-FeOOH)

3.5 24 h 5.18*105 Parfitt et al. 1975

Goethite (α-FeOOH)

3.5 24 h 2.10*105 Parfitt et al. 1975

Hematite (α-Fe2O3)

3.5 24 h 1.64*105 Parfitt et al. 1975

Dry Tailings

3.5 24 h 2.48*105 (Zeng et al. 2004)

Polkemmet ochre

8.06*105 (Heal et al. 2005)

74

Table S6. Model parameters and pH for CFH 2-1 mm.

Sorption† Desorption‡ pH

Inlet Pi conc RT Mass

LSS Std. Err. l.s K Std. Err. l.s R std err

RDM Std. Err. l.s D Std. Err. l.s d Std. Err. l.s R std err

Start sorp. SD End sorp. SD End desorp. SD

µM min g µmol/kg %

CF

H 2

-1 m

m

1.6

½ 0.50

1520 42 *** 1390 81 *** 23

2.3 0.9 * 7,41*10^5 1,56*10^7 0.96 -2.88 4.76 0.55 1.36

6.8 0.02 6.4 0.01 6.4 0.08

1 0.40

830 119 *** 710 233 ** 82

3.6 1.0 * 2,68*10^6 5,68*10^7 0.96 -3.30 4.86 0.53 1.25

7.1 0.18 6.7 0.06 6.6 0.32

1½ 0.30

1340 372 ** 1350 630 * 104

12.1 6.4 0.12 3,26*10^4 2,77*10^5 0.91 -2.09 1.91 0.32 2.42

6.7 1.00 6.6 1.12 6.5 0.11

3 0.21

2000 349 *** 2120 498 *** 33

11.7 3.0 * 4,75*10^4 1,7*10^5 0.79 -2.22 0.85 * 0.99

6 0.15

3340 517 *** 36600 6370 *** 8

7.0 0.6 *** 1,27*10^4 1,88*10^4 0.53 -2.23 0.39 ** 0.29

9 0.10

9720 2686 ** 9310 2810 ** 21

3.2

½ 0.50

1850 91 *** 2050 241 *** 69

4.3 0.2 *** 1,03*10^4 1,05*10^4 0.36 -1.98 0.24 *** 0.13

7.7 0.52 7.0 0.14 7.0 0.26

1 0.41

1360 66 *** 1250 147 *** 51

7.5 5.0 0.19 5,68*10^3 2,90*10^4 0.85 -1.68 1.26 0.24 0.88

7.2 0.80 6.9 0.16 6.6 0.02

1½ 0.30

1320 58 *** 1200 123 *** 42

9.2 1.9 ** 7,20*10^3 1,27*10^4 0.60 -1.73 0.42 ** 0.40

7.0 0.30 6.8 0.06 6.5 0.11

3 0.20

1460 39 *** 1310 69 *** 19

13.2 9.1 0.21 6,73*10^4 5,12*10^5 0.90 -2.22 1.80 0.27 2.09

6 0.15

1710 188 *** 1220 212 *** 40

8.1 0.4 *** 5,93*10^3 3,67*10^3 0.17 -2.01 0.17 *** 0.14

9 0.10

3320 607 *** 4090 1040 *** 64

16

½ 0.50

5400 394 *** 10100 1780 *** 303

7.4 0.8 *** 2,27*10^3 2,03*10^3 0.30 -1.54 0.01 *** 0.19

6.7 0.05 6.5 0.05 6.8 0.18

1 0.50

3260 192 *** 3860 626 *** 190

5.9 1.2 ** 1,36*10^4 4,95*10^4 0.80 -2.08 0.89 . 0.54

7.0 0.42 6.7 0.14 6.3 0.58

1½ 0.40

3510 332 *** 4650 1020 *** 233

6.6 1.7 * 1,96*10^4 1,02*10^5 0.86 -2.19 1.26 0.14 0.87

6.7 0.10 6.5 0.10 6.8 0.13

3 0.30

3490 440 *** 4890 1170 *** 185

7.4 1.4 ** 7,27*10^3 2,27*10^4 0.76 -2.15 0.85 . 0.66

6 0.15

3560 310 *** 3650 603 *** 128

9.3 1.4 ** 2,49*10^4 1,06*10^5 0.82 -2.53 1.14 . 1.04

9 0.10

5080 320 *** 7880 951 *** 126

32

½ 1.0

14900 2080 *** 25800 590 *** 534

8.8 3.4 * 4,16*10^2 244 0.12 -1.16 0.25 *** 0.20

7.3 0.23 7.0 0.36 7.1 0.22

1 0.75

8840 931 *** 12600 2370 *** 332

4.3 0.6 *** 6,07*10^3 1,32*10^4 0.67 -2.09 0.59 * 0.26

7.5 0.90 6.9 0.05 6.8 0.09

1½ 0.65

7570 893 *** 11700 2300 *** 247

5.2 0.5 *** 3,51*10^5 1,26*10^5 0.79 -2.71 0.98 * 0.47

7.6 0.52 6.8 0.07 6.8 0.04

3 0.50

5160 845 *** 6330 1900 ** 323

4.9 0.5 *** 1,88*10^3 2,44*10^3 0.48 -2.05 411.00 ** 0.21

6 0.30

4900 411 *** 7030 973 *** 108

7.0 0.3 *** 4,89*10^3 5,87*10^3 0.44 -2.58 0.39 ** 0.27

9 0.20 8260 754 *** 14000 1990 *** 140

Standard error ‘Std. err.’, level of significant ‘l.s’, residual standard error ‘R std. err’ and standard error ‘SD’. Significant level codes; 0 = ‘***’, 0.001 <‘**’, 0.01< ‘*’, and 0.05< ‘.’

† Parameters obtained via Eq. 3 on non-averaged data

‡ Parameters obtained via Eq. 4 on non-averaged data

75

Table S7. Model parameters and pH for CFH 1-0.5 mm.

Sorption† Desorption‡ pH

Inlet P conc RT Mass

LSS Std. Err. l.s K Std. Err. l.s R std err

RDM Std. Err. l.s D Std. Err. l.s d Std. Err. l.s R std err

Start sorp. SD End sorp. SD End desorp. SD

µM min g

µmol/kg

%

CF

H 1

- 0

.5 m

m

1.6

½ 0.50

3680 589 *** 3570 844 *** 95

1.1 0.1 *** 1,12*10^5 9,67*10^5 0.91 -2.72 2.04 0.22 0.23

7.0 0.12 6.5 0.49 6.7 0.08

1 0.41

5880 570 *** 6030 693 *** 18

27 101 0.80 8,18*10^3 1,32*10^4 0.56 -1.38 1.02 0.23 0.85

7.3 0.86 7.0 0.19 7.2 0.26

1½ 0.22

6350 1480 *** 7280 2050 ** 57

8.4 3.7 . 5,49*10^4 4,17*10^4 0.90 -2.11 1.65 0.26 1.44

6.8 0.21 6.5 0.24 6.7 0.14

3 0.15

6920 1260 *** 7880 1690 *** 43

12 5.2 . 2,71*10^4 8,29*10^4 0.76 -1.89 0.73 * 0.73

6 0.10

11000 2020 *** 12200 2420 *** 18

1.6 2.7 0.59 2,7*10^5 3,15*10^6 0.94 -2.34 2.80 0.44 0.26

9 0.10

13300 7490 . 13900 8500 0.11 56

3.2

½ 0.50

6670 477 *** 6030 670 *** 98

3.3 1.5 . 1,28*10^4 6,67*10^4 0.85 -1.89 1.22 0.17 0.48

6.9 0.06 6.4 0.25 6.6 0.16

1 0.30

5560 955 *** 5280 1390 *** 184

7.5 1.2 ** 3,07*10^4 4,18*10^4 0.49 -1.88 0.31 ** 0.21

6.6 0.12 6.9 0.05 6.9 0.01

1½ 0.20

5380 220 *** 4520 298 *** 53

8.9 12 0.77 9,81*10^3 7,60*10^4 0.90 -1.59 1.79 0.42 1.30

7.0 0.08 7.0 0.05 6.7 0.04

3 0.15

6550 604 *** 6140 747 *** 50

100 §

1,04*10^4 7,04*10^3 0.19 -0.95 0.12 *** 0.19

6 0.10

18300 2000 *** 19300 2280 *** 11

9.5 9.7 0.37 4,05*10^6 1.18*10^6 0.97 -3.12 6.43 0.65 4.83

9 0.10

11200 3020 ** 11800 3640 ** 54

16

½ 0.50

12000 986 *** 18200 2800 *** 395

7.1 1.3 ** 3,88*10^3 5,38*10^3 0.09 -1.61 0.35 ** 0.28

7.1 0.04 6.8 0.05 6.8 0.01

1 0.30

549 22 *** 4180 575 *** 31

6.5 2.3 * 3,28*10^4 1,65*10^5 0.85 -1.99 1.10 0.13 0.74

6.9 0.21 6.6 0.05 6.5 0.17

1½ 0.21

9240 796 *** 13300 2250 *** 378

6.5 2.8 . 3,93*10^5 6,34*10^6 0.95 -2.53 3.34 0.48 2.13

7.0 0.10 6.6 0.00 6.5 0.02

3 0.15

7800 349 *** 8530 762 *** 172

9.8 2.5 * 3,69*10^4 2,10*10^5 0.87 -2.20 1.30 0.15 1.32

6 0.10

7580 583 *** 8690 1180 *** 198

9.5 9.7 0.37 4,05*10^6 1,19*10^8 0.97 -3.12 6.43 0.65 4.83

9 0.10

11000 1240 *** 11500 2160 *** 304

32

½ 1.0

28700 1100 *** 35800 2010 *** 181

19 35 0.61 808 635 0.24 -1.04 0.34 * 0.26

7.2 0.14 6.9 0.06 6.7 0.22

1 0.50

18500 3300 *** 24000 7190 ** 934

4.9 2.8 0.74 3,56*10^4 3,56*10^4 0.84 -1.82 1.23 0.20 0.58

7.0 0.08 7.0 0.09 7.0 0.05

1½ 0.51

15500 1080 *** 22700 2560 *** 260

4.4 2.4 0.12 1,39*10^4 1,00*10^5 0.90 -2.03 1.78 0.31 0.81

7.3 0.06 6.8 0.03 7.0 0.00

3 0.30

13900 1340 *** 16800 2520 *** 275

4.9 0.6 *** 1,88*10^3 2,40*10^3 0.48 -2.05 0.41 *** 0.21

6 0.20

11200 1110 *** 12800 1950 *** 223

5.9 1.0 ** 5,59*10^3 1,64*10^4 0.75 -2.25 0.86 * 0.53

9 0.20 16200 2370 *** 20000 4120 *** 259

Standard error ‘Std. err.’, level of significant ‘l.s’, residual standard error ‘R std. err’ and standard error ‘SD’. Significant level codes; 0 = ‘***’, 0.001 <‘**’, 0.01< ‘*’, and 0.05< ‘.’

† Parameters obtained via Eq. 3 on non-averaged data

‡ Parameters obtained via Eq. 4 on non-averaged data

§ D parameter set to 100%

76

References

Ádám K, Sovik A, Krogstad T, Heistad A. Phosphorous removal by the filter materials light-weight aggregates and shellsand-a review of processes and

experimental set-ups for improved design of filter systems for wastewater treatment. Vatten 2007;63:245.

Brunauer S, Emmett PH, Teller E. Adsorption of Gases in Multimolecular Layers. J Am Chem Soc 1938;60:309-19.

Genz A, Kornmüller A, Jekel M. Advanced phosphorus removal from membrane filtrates by adsorption on activated aluminium oxide and granulated ferric

hydroxide. Water Res 2004;38:3523-30.

Heal K, Dobbie K, Bozika E, McHaffie H, Simpson A, Smith K. Enhancing phosphorus removal in constructed wetlands with ochre from mine drainage

treatment. Water Science and Technology 2005;51:275-82.

Lyngsie, G., Borggaard, O.K., Hansen H.C.B., 2013a. Testing phosphate sorption efficiency of five potential drainage water filter materials under field-relevant

conditions, Submitted to Water Res.

Murphy J, Riley JP. A modified single solution method for the determination of phosphate in natural waters. Anal Chim Acta 1962;27:31-6.

Parfitt RL, Atkinson RJ, Smart RSC. The mechanism of phosphate fixation by iron oxides. Soil Sci Soc Am J 1975;39:837-41.

Penn CJ, McGrath JM. Predicting phosphorus sorption onto steel slag using a flow-through approach with application to a pilot scale system. J.Water

Resour.Prot 2011;3:235-44.

Ruzicka J, Hansen EH. Homogeneous and heterogeneous systems : Flow injection analysis today and tomorrow. Anal Chim Acta 1988;214:1-27.

Schwertmann U. Differenzierung der Eisenoxide des Bodens durch Extraction mit Ammoniumoxalat-Lösung. Zietschrift für Pflanzenernährung, Düngung,

Bodenkunde 1964;105:194-202.

Zeng L, Li X, Liu J. Adsorptive removal of phosphate from aqueous solutions using iron oxide tailings. Water Res 2004;38:1318-26.

77

Chapter 7 – Paper III

Phosphate sorption by three potential filter materials as assessed by isothermal titration

calorimetry

Lyngsie, Gry1*

, Penn, Chad J.2, Hansen, Hans C. B.

1, Borggaard, Ole K.

1

1University of Copenhagen, Department of Plant and Environmental Sciences, Thorvaldsensvej 40,

DK-1871 Frederiksberg C. 2Oklahoma State University, Department of Plant and Soil Sciences, 368

Agricultural Hall, Stillwater, OK 74078-6028.

Submitted to Journal of Environmental Management

*Corresponding author:chj177@ku.dk

Abstract

Phosphorus eutrophication of lakes and streams, coming from drained farmlands, is a serious

problem in areas with intensive agriculture. Installation of phosphate (P) sorbing filters at drain

outlets may be a solution. The aim of this study was to improve the understanding of reactions

involved in P sorption by three commercial P sorbing materials, i.e. Ca/Mg oxide-based Filtralite®P,

Fe oxide-based CFH-12 and Limestone in two particle sizes (2-1 mm and 1-0.5 mm), by means of

isothermal titration calorimetry (ITC), sorption isotherms, sequential extractions and SEM-EDS.

The results indicate that P retention by CFH is due to surface complexation by rapid formation of

strong Fe-P bonds. In contrast, retention of P by Filtralite®P and Limstone strongly depends on pH

and time and is interpreted due to formation of calcium phosphate precipitate(s). Consequently,

CFH can unambiguously be recommended as P retention filter material in drain outlets, whereas the

use of Filtralite®P and Limestone has certain (serious) limitations. Thus, Filtralite

®P has high

capacity to retain P but only at alkaline pH (pH ≥ 10) and P retention by Limestone requires long-

time contact and a high ratio between sorbent and sorbate.

1 Introduction

Surface and subsurface transport of phosphate (P) from fertilized agricultural fields to open waters

may lead to eutrophication, reduced biodiversity and fish kills in lakes and streams (Ballantine and

78

Tanner, 2010; Delgado and Scalenghe, 2008). To mitigate eutrophic waters is difficult (Sharpley et

al., 2003) but reduction of agricultural P leaching by means of filter structures intercepting drains

and ditches with P sorbing materials may be one way to improve the water quality (Ballantine and

Tanner, 2010; Penn et al., 2007). Several materials have been proposed as P sorbing materials

(PSM) (Ballantine and Tanner, 2010; Cucarella and Renman, 2009; Vohla et al., 2011; Westholm,

2006) for use in landscape P filters (Penn et al., 2007; Reinhardt et al., 2005).

To handle the high water flow and relatively high P concentrations during peak flows found in

farmland drains and ditches, PSMs must react fast and have a high affinity for P in order to

effectively remove it. The P removal efficiency of most PSMs is typically related to pH and various

Al, Ca and Fe compounds (Ballantine and Tanner, 2010; Cucarella and Renman, 2009; Vohla et al.,

2011; Westholm, 2006). Additionally, the particle size and hence specific surface area (SSA) of the

sorbent particles increases reactivity and sorption capacity (Nair et al., 1984). Besides high P

affinity and fast kinetics, PSM filters must also have good hydraulic conductivity to handle the high

water flow seen in connection with rain-storms but at the same time also allow P-rich water to come

into contact with the materials. Despite the abundance of literature on sorption to various PSMs

such as Al and Fe oxides, limestone, shellsand and various by-products from industry (Ballantine

and Tanner, 2010; Cucarella and Renman, 2009; Vohla et al., 2011; Westholm, 2006), doubt still

exists about detailed sorption reactions of these materials, and hence safe use and management of

these filters is not yet ensured.

Sorption isotherms are commonly used in P sorption studies of PSMs (Ballantine and Tanner, 2010;

Klimeski et al., 2012; Westholm, 2006). Although useful for assessment of P sorption capacity and

affinity, isotherms are not suited for determination of the precise sorption reactions (Veith and

Sposito, 1977). Isothermal titration calorimetry (ITC) provides a sensitive and direct quantitative

measure of heat of a reaction and can be used as a complementary technique for establishing P

sorption reactions. For instance, Penn and Zhang (2010) found that FeCl3 titrated with NaH2PO4

changed from exothermic to endothermic as the titration proceeded and interpreted this to a change

of the P sorption process from adsorption to precipitation. ITC can be used for estimation of

enthalpy for mineral-solution interactions in connection with traditional sorption measurements as

the reaction proceeds (Appel et al., 2013; Kabengi et al., 2006; Penn and Warren, 2009; Penn and

Zhang, 2010; Rhue et al., 2002). Briefly, ITC measures changes in heat emitted (exotherm) or

79

absorbed (endotherm) along product formation when a solute is added stepwise to a solution or

solid suspension. For each solute addition (injection), the heat q (J s-1

) released or absorbed is given

by Eq. 1:

pnHVq (1)

Where Δ[np] (mol L-1

s-1

) is the change in product concentration, ΔH is the enthalpy of the reaction

(J mol-1

product), and V (L) is the volume of reaction mixture. Because q is directly proportional to

the increase in mass of product formed at each injection, its magnitude will gradually decrease as

the reaction approaches saturation (equilibrium) of the system. The time integrated heat Q (J)

released or absorbed is directly proportional to the energy of interaction:

pnHVQ (2)

For a full introduction to this method see Steinberg (1981) and Freire et al. (1990). ITC is not a

“stand alone” technique, as correct interpretation of the data requires additional knowledge of the

system being studied (Penn and Zhang, 2010; Rhue et al., 2002). Furthermore, the calculated

thermodynamic properties (q and Q) can only be meaningfully interpreted for pure systems with

one (or very few) reactions that can be identified.

In order to know which type of PSM to choose for landscape filters, it is central to understand the

particular type of sorption reaction, e.g. whether it is a precipitation, surface complexation or ion

exchange reaction. Thus, by means of ITC, sorption isotherms, sequential extractions and PSM

characteristics, the objective of this study was to improve understanding of the reactions involved in

P sorption by three potential commercial PSMs including CaCO3-based Limestone, Ca/Mg oxide-

based Filtralite®P and Fe oxide-based CFH-12 at two different particle sizes. These three materials

were choosen based on a prior screening (Lyngsie et al., 2013).

2 MATERIALS AND METHODS

2.1 Materials

Limestone consists of a mixture of bryozo and coral chalk from the Danian formation at Faxe. The

dried product was provided by Faxe Kalk A/S, Denmark. Filtralite®P produced by Weber, Norway

80

is a Light Expanded Clay Aggregate (LECA)-resembling material calcinated at 1200 ˚C consisting

of granules of Ca/Mg oxides in a collapsed clay matrix. CFH-12 (CFH) produced by Kemira Oyj,

Finland, consists of dried iron oxide. The particles of the three PSMs were separated by sieving into

a 2-1 mm fraction and a 1-0.5 mm fraction.

2.3 Methods

All chemicals were of pro analysis or better quality and the water was double deionized (DI). All

glass and plastic wares were acid-washed prior to use.

2.3.1 PSM characteristics

The mineralogy of the materials was assessed by powder X-ray diffraction analysis (XRD) on

unoriented specimens using a Siemens D5000 instrument equipped with Co-Kα radiation and a

diffracted beam monochromator. Diffractograms were recorded from 5 to 80o 2θ using 0.02

o 2θ

steps and a step speed of 10 s. Diffraction peak positions were used to calculate d-values for mineral

identification.

pH of the PSMs was measured potentiometrically in 0.01 M CaCl2 using a material:solution ratio of

1:2.5. Total contents of Al, Fe, Ca and Mg (AlTotal, FeTotal, CaTotal, MgTotal) were determined after

dissolution of the materials in a mixture of concentrated nitric acid, hydrogen peroxide,

hydrochloric acid and hydrofluoric acid (EPA 3052) and were measured by inductive coupled

plasma mass spectroscopy (ICP-MS) on an Agilent 7500C instrument. Oxalate-extractable iron

(FeOx) was determined by extraction with 0.2 M ammonium oxalate for 2 h at pH 3 in the dark

(Schwertmann, 1964) with quantification of Fe by inductively coupled plasma atomic emission

spectroscopy (ICP-AES) using a Spectro Ciros CCD instrument. Reactive Ca and Mg in Filtralite®P

and Limestone were determined by titration of 0.25 g PSM suspended in 7.5 mL water with 0.005

M HNO3 while stirring to the same final pH as after the 25 titrations with 0.01 M NaH2PO4

described below. The suspension was filtered using a 0.22 µm Millipore filter and the Mg and Ca

concentrations were determined in the clear filtrate by atomic absorption spectroscopy (AAS) using

a Perkin Elmer 3300 instrument. Total inorganic P (Ptotal) was determined by extraction with 6 M

H2SO4 for 10 min at 70 °C (Mehta et al., 1954). Plant available phosphorus (POlsen) was determined

through 3 combined sequential extractions with 0.5 M NaHCO3 adjusted to pH 8.5 (Olsen and

Sommers, 1982). The concentration of P in the extracts was determined by the molybdenum-blue

81

method (Murphy and Riley, 1962). All analyses were carried out in triplicates. The specific surface

area (SSA) of the different fractions was determined by applying the BET equation to N2 adsorption

data obtained by means of a Micromeritic Gemini VII 2390a instrument (Brunauer et al., 1938)

after the sample has been outgassed for 12 h at room temperature. The reported SSAs are the

average of five separate measurements.

2.3.2 Isothermal Titration Calorimetry

All ITC experiments were conducted on a CSC 4200 Isothermal Titration Calorimeter (CSC Inc.,

Lindon, UT) at 25°C. The ITC has a sensitivity of 0.418 μJ detectable heat effect and a “noise

level” of ±0.0418 μJ s-1

(deconvoluted signal). The ITC investigations were carried in two ways, i.e.

as single-point and as multiple-points titrations. For single-point P sorption titrations, 100 mg of the

material was placed in a 1.3 mL reaction vessel and suspended in 750 µL of DI water, and 250 µL

0.01 M NaH2PO4 was added in one injection and the heat production monitored for the following 5

h. For the mulitiple-points P sorption titrations, 25 mg of the material was placed in the reaction

vessel and suspended in 750 µL of water. Under continuous stirring, the suspension was titrated

with 0.01 M NaH2PO4 by adding a total of 250 µL in 25 increments of 10 µL with 5 min between

each injection. In order to compensate for heat of dilution, a blank was run where P solution was

stepwise added in the same manner to water as to the PSM suspension. In addition, the heat of

neutralization of Filtralite®P was determined by titration of 25 mg Filtralite

®P in 750 µL of water

with 0.005 M HNO3 to pH 7.4, which was the final pH after titration with 0.01 M NaH2PO4. All

ITC experiments were run as duplicates.

2.3.3 P sorption isotherms

Supplementary batch sorption isotherms were conducted in the same manner and at the same

solid:solution ratio as the corresponding ITC titration, but scaled up 10 times. Accordingly, 0.25 g

of PSM was suspended in 7.5 mL water and stepwise added 0.1 mL of 0.01 M NaH2PO4 with 5 min

between each addition. Only the first 8 injections and the 25th

were recorded. P sorption

experiments resembling the single-point ITC titrations were also conducted, where 1 g PSM was

suspended in 7.5 mL water and 2.5 mL of 0.01 M NaH2PO4 was added and allowed to react for 5 h.

The suspensions were filtered using 0.22 µm Millipore filters. In the clear filtrate, pH was measured

by a combination electrode and the P concentration determined spectrophotometrically by the

82

molybdenum blue method (Murphy and Riley, 1962). All sorption experiments were done in

triplicates

2.3.4 Sequential P fractionation

Materials from the single-point P sorption experiment were sequentially extracted with a modified

Hedley sequence (Hedley et al., 1982). 0.5 g of previously P-treated material from the single-point

P sorption experiments was extracted with 30 mL water by overnight shaking, centrifugation at

3000 g for 15 min, and filtered through a 0.22 µm Millipore filter. The sample was further

sequentially extracted with 0.5 M NaHCO3, 0.1 M NaOH and 1 M HCl in the same manner as

described for the water extraction. The P concentrations in the extracts were determined as

described above and referred to as PH2O, PNaHCO3, PNaOH and PHCl pools depending on the treatment.

2.3.5 Surface examination

Samples from the multiple-points titration experiments were examined by scanning electron

microscopy (SEM, FEI Quanta 600FE) and energy dispersive X-ray spectroscopy (EDS, EVEX

Nanoanalysis Quantum Dot Detector using EVEX Nanoalysis Software (EVEX, Princeton, New

Jersey)). Samples were prepared for SEM-EDS by depositing on a C stub mount and coating with C

to prevent charging. The samples were also examined by XRD in the same manner as described in

section 2.3.1.

3 RESULTS AND DISCUSSION

3.1 Filter characteristics

CFH consists of amorphous (oxalate-extractable) Fe oxides, 2-line ferrihydrite according to XRD,

and minor contents of Al, Ca and Mg compounds (Supported Information (SI), Table S1). SSAs of

the two particle size fractions are the same, 32 m2 g

-1. This is rather low compared to SSA of other

Fe oxide materials used for P sorption (Chardon et al., 2012; Willett et al., 1988). The very high

content of amorphous Fe oxides strongly suggests CFH as a strong P bonding sorbent (Lyngsie et

al., 2013; Klimeski et al., 2012). Accordingly, although CFH contained substantial amount of P

(indigenous P), it is unavailable as shown by undetectable POlsen.

Filtralite®P contains little Fe and it is dominated by Al, Ca and Mg compounds (SI, Table S1).

According to XRD, Filtralite®

P consists of silicates, calcite and Ca/Mg oxides. No Al oxides were

83

detected by XRD and the content of oxalate-extractable Al was negligible. Al oxides, which have

good P retention properties (Vohla et al., 2011), may therefore be considered of very limited

importance as P sorbents in Filtralite®P. Due to the high content of Ca/Mg oxides, an aqueous

suspension of Filtralite®

P has a pH of nearly 11. Under aquatic conditions the Ca/Mg oxides will

hydrate and form Ca/Mg hydroxides, which together with the calcite may cause P sorption.

Indigenous P amounted to 3 mmol kg-1

that was partly available (POlsen = 0.2 mmol kg-1

). The SSA

of both fractions are considerable higher than expected from their particle sizes, indicating that both

fractions have active/exposed sites on external as well as internal surfaces of the particles.

Limestone is dominated by Ca, with minor contents of Mg, Al and Fe (SI, Table S1). XRD showed

the material to be strongly dominated by crystalline calcite, but also aragonite and dolomite were

identified. Limestone contained 1 mmol kg-1

of indigenous P, of which one fifth was available as

POlsen. CaCO3 in different forms, e.g. sea-shells, egg-shell and limestone is often used as a P

removal material because it is relatively cheap and available at many places (Ballantine and Tanner,

2010).

3.2 P sorption to CFH

Both CFH particle size fractions readily sorbed added P (Fig. 1a), which was expected due to the

high content of amorphous Fe oxides (Klimeski et al., 2012). After addition of 25 injections of 0.01

mL of 10 mM P per 0.025 g CFH corresponding to 100 mmol kg-1

in the multiple-points titration, a

maximum of 55 mmol kg-1

or about half of the added P was sorbed. However, as indicated by the

almost linear sorption curve, none of the particle size fractions reached the maximum sorption

capacity (Pmax). Using the pedotransfer function by (Borggaard et al., 2004) and assuming that Feox

is the only P sorbing agent, a rough estimate of Pmax can be calculated as 0.12×FeOx resulting in Pmax

= 890 mmol kg-1

for both CFH size fractions. This confirms that the Pmax for both CFH size

fractions is far from reached. In the single-point P sorption, CFH sorbed all (25 µmol kg-1

) of the

added P (Table 1). The sequential extraction (Fig. 2b) showed that less than 5% of the extracted P

was in the labile pool (PH2O + PNaHCO3). The PNaOH, which is the pool associated to P bound to Al

and Fe, amounted between 14 and 18% (5-7 mmol kg-1

) of the sorbed P. PHCl is by far the largest

pool (28-29 mmol kg-1

) and is usually considered to comprise P associated with Ca as in Ca

phosphates. Considering the large PHCl and the low Ca content in CFH (SI, Table S1) together with

the modest PNaOH pool it is likely that the PHCl pool also includes Fe bound P. Nonetheless both the

84

PNaOH and PHCl pools are considered non-labile emphasizing that the P is strongly sorbed to CFH

and little sorbed P is expected to be released to the aqueous phase from the CFH. The EDS (SI, Fig.

S1a) showed that P is present on the surface (6%) and randomly distributed on the surface of the

particle (overlay image not showed). The SEM image (SI, Fig. S1a) did not show formation of any

secondary minerals nor did the XRD analysis.

Fig. 1. Sorption of P by the two particle size fractions of CFH. a) Showing the amount of sorbed P as a function of

injection number for direct comparison with the multi-points P thermogram; b) pH resulting from the P additions shown

in a), c) thermograms for multi-points titration of CFH with 0.01 M NaH2PO4, d) corresponding integration of the

thermograms. Error bars are standard deviations. Exothermic upwards.

The sorption of P by CFH is an exothermic reaction as shown by the ITC thermograms (Figs. 1c

and 1d). Evolved heat decreases gradually with successive titration and after around 10 injections a

steady-state heat evolution at each increment seems achieved. However, heat evolution does not

cease even after addition of 100 mmol P kg-1

(all 25 injections). This might be attributed to sorbed P

being considerably less than the estimated Pmax (see above) corresponding to low P saturation and

hence almost constant bonding energies. The rapid rise of the evolved heat peaks after each

85

injection (Fig. 1c) indicates fast reaction between sorbent and sorbate (Penn and Warren, 2009;

Steinberg, 1981). After this rapid heat evolution, the peaks for the first 6 injections (Fig. 1c) became

somewhat broadened (tailing). This is even more clearly seen on the single-point thermogram (Fig.

2a), where the fast initial heat evolution is followed by a slower heat release shown by the rather

long tail. The peak morphology indicates P sorption in two steps including a fast initial sorption

followed by a slower reaction. This may be attributed to fast bonding of P to Fe oxide sorption sites

at outer particles surfaces followed by diffusion into interior sorption sites as demonstrated for P

sorption by ferrihydrite (Willett et al., 1988).

Table 1. Final pH for the single-point P addition experiment, total amount of P sorbed for the single-point (Psingle) and

the multiple P injection experiment (Pmulti), corresponding heat released for the single-point (Q) and multiple (∑Q)

experiment, heat normalized to sorbent (Q norm), and heat released per mol P sorbed (∆H).

CFH Filtralite

®P

Limestone

2-1 mm 1-0.5 mm 2-1 mm 1-0.5 mm 2-1 mm 1-0.5 mm

pHsingle 7.7±(0.04) 7.8±(0.02) 8.1±(0.26) 8.4±(0.06) 8.3±(0.09) 8.3±(0.24)

Psingle mmol kg-1 24±(0.5) 25±(0.2) 10±(0.7) 11±(0.4) 20±(0.1) 21±(0.1)

Pmulti mmol kg-1 47±(4.5) 55±(7.0) 2.5±(1.9) 5.8±(1.8) 3.1±(1.8) 6.6±(1.9)

Qsingle mJ -170 -175 -139 -148 -33 -32

∑Qmulti mJ -29 -41 -48 -44 -17 -12

Qsingle norm kJ kg-1 solid -1.7 -1.7 -1.4 -1.5 -0.3 -0.3

∑Qmulti norm kJ kg-1 solid -1.2 -1.6 -1.9 -1.8 -0.7 -0.5

∆Hsingle kJ mol-1 P -71 -70 -139 -135 -16 -15

∆Hmulti kJ mol-1 P -25 -30 -774 -303 -222 -73

The CFH single-point thermogram (Fig. 2a) indicates that exothermic heat peaked 210 s after P

injection, and then the heat of reaction gradually decreased, and returns to baseline after approx. 1

h. Harvey and Rhue (2008) and Appel et al. (2013) examined P sorption by amorphous Fe and Al

oxides using flow calorimetry and also found that the reaction between the oxides and P were

exothermic and proceeded within about 1 h. The molar reaction heat (∆Hr) for the P sorption in this

study was ~ -70 kJ mol-1

and -25 to -30 kJ mol-1

sorbed P for the single-point and multiple-points P

titrations, respectively (Table 1). The difference in ∆Hr can be attributed to about twice as much P

was sorbed in the multi-points than in the single-point experiments (Table 1), as Q expressed per kg

CFH (denoted Qnorm in Table 1) gave values not too different (-1.2 to -1.7 kJ kg-1

solid). The ∆Hr

86

obtained in the multi-points titration is considered the most reliable as in the single-point

experiment all added P was sorbed. In good agreement with ∆Hr obtained from the multi-points

titration, Harvey and Rhue (2008) found ∆Hr in the range -25 to -39 kJ mol-1

for P sorption by a

mixed Al/Fe oxide sorbent.

Fig 2. a) Thermograms from single-point P injection with 250 µL of 0.01 M NaH2PO4. Shown is the heat formation

from 1-0.5 mm size fraction of CFH, Filtralite®

P and Limestone; exothermic upwards. b) Sequentially extracted P on

previously P-treated material from the single-point P titration experiment. The P was extracted with H2O, NaHCO3, HCl

and NaOH. The amount of P extracted is given in the table in mmol kg-1

.

Miltenburg and Golterman (1997) titrated Fe oxide with a P solution, resulting in a thermogram

resembling those in Fig. 1c. Further, an initial exothermic reaction found by ITC of FeCl3 titrated

87

with P was also interpreted as phosphate complexation to Fe(III), while further titration resulting in

an endothermic reaction was considered due to an unspecified precipitation (Penn and Zhang,

2010). However, Fe-P precipitates such as strengite (FePO4) only forms under acidic conditions

(and elevated P concentrations) but not at neutral to weakly alkaline pH, where Fe will precipitate

as Fe oxides (Lindsay, 1979) that will sorb P through surface complexation. This also explains why

in the present study at neutral pH, no Fe(III)-phosphate minerals were observed in the SEM images

or in XRD despite detection of P on the CFH surface (SI, Fig. S1a).

The results of the present study and numerous other investigations (Harvey and Rhue, 2008; Kwon

and Kubicki, 2004; Penn and Zhang, 2010) have shown that P sorption by Fe oxides can be

explained by surface complexation but doubt still exists about the precise surface complex(es)

formed (Dideriksen and Stipp, 2003; Kwon and Kubicki, 2004). Thus, diprotonated,

monoprotonated and deprotonated forms of bidentate, binuclear and monodentate, mononuclear

complexes have been suggested (Arai and Sparks, 2001; Kwon and Kubicki, 2004; Persson et al.,

1996). Probably the complex formed depends on various external conditions such as pH and P

concentration. Thus, based on comprehensive quantum mechanical calculations, Kwon and

Kubicki, 2004 concluded that bidentate, binuclear complexes formed at acidic pH (4-6), whereas at

neutral pH as in the present study monodentate (mononuclear) complexes are the most stable form,

although occurrence of a deprotonated bidentate, binuclear complex cannot be completely ruled out.

The results of the present investigation may therefore be interpreted as fast and strong P sorption by

CFH under formation of probably monodentate complexes on outer surfaces followed by slow

migration into interior sorption sites.

3.3 P sorption by Filtralite®

P

P sorption by the two Filtralite®P particle size fractions was very different (Fig. 3a). While the 2-1

mm fraction did not significantly sorb P during the first 8 injections, the 1-0.5 mm fraction sorbed

almost the same amount of P as CFH up to the fifth injection, which is equivalent to about half (~10

mmol kg-1

) of the added P (20 mmol kg-1

). However, after further P addition, P sorption completely

ceased and previously sorbed P was desorbed. This decrease in P sorption coincides with a sharp

drop in pH from about 10 to nearly 8 at the 6th

P injection (Fig. 3b). For the 2-1 mm fraction, the pH

immediately dropped to about 8 after the first injection and no P sorbed. Evidently, P retention by

Filtralite®P is strongly pH dependent. Previous studies have shown that P sorption by Ca-rich

88

materials is dependent on both the Ca concentration and elevated pH (Claveau-Mallet et al., 2012;

Klimeski et al., 2012; Stoner et al., 2012; Vohla et al., 2011). However, after 25 injections in the

multi-points experiment, the 1-0.5 mm fraction sorbed about 6 mmol kg-1

and the 2-1 mm fraction

about 2.5 mmol kg-1

(Table 1; Fig. 3a), while both size fractions sorbed about 40% (10 mmol kg-1

)

in the single-point ITC investigation.

Fig. 3. Sorption of P by the two particle size fractions of Filtralite®

P. a) Showing the amount of sorbed P as a function

of injection number for direct comparison with the multi-points P thermogram; b) pH resulting from the P additions

shown in a), c) thermogram for multi-points titration of Filtralite®

P with 0.01 M NaH2PO4, d) corresponding integration

of the thermograms. Error bars are standard deviations. Exothermic upwards

The sequential extraction of Filtralite®P showed that 3-4 mmol P kg

-1 was extractable by water and

that the labile P pool (PH2O + PNaHCO3) accounted for about 65% of the extracted P (Fig. 2b),

demonstrating that the sorbed P is very reactive. PHCl accounted for ~30% (~5 mmol kg-1

) of the

desorbed P, but about half of this most likely originated from indigenous P present in the

Filtralite®P (SI, Table S1). SEM (SI, Fig. S1b) did not show formation of new minerals on the

surface nor was any detected by XRD. The overlay EDS indicates that Ca is connected to the light

89

(white) spots on the images of the surface and that no P was detected. EDS close-up on light spots

showed that P is present on the surface but in small amounts < 1% (SI, Fig. S1d).

Obviously, P sorption by Filtralite®P is very pH sensitive with high sorption at pH higher than ~10

but little or no sorption at neutral pH (Fig. 3a-b). Therefore, the P sorbed during the first five

injections must be retained in a sorbent-sorbate interaction, possibly a Ca phosphate, which readily

dissolves P when pH is lowered. Thus, a precipitate of amorphous calcium phosphate (ACP),

Ca3(PO4)2*nH2O, is readily formed under alkaline conditions (Cao and Harris, 2008; Lindsay,

1979). When pH decreases, the stability of ACP decreases eventually resulting in dissolution of this

precipitate (Fig. 3a). However, although likely, formation of ACP during the initial P injections

(Fig. 3a) is so far a postulate that needs be verified by future studies.

The dominance of exothermic peaks that gradually decrease during multiple-points P titrations (Fig.

3c) and the corresponding accumulated heat curves (Fig. 3d) for the reaction between P and

Filtralite®P resemble those for P sorption by CFH (Figs. 1c and 1d). However, closer inspection of

Fig. 3c shows the presence of small endothermic contributions during the first four injections. These

endothermic humps appear at the foot of the exothermic peaks, i.e. they correspond to a slower

reaction than the reaction causing the exothermic peaks. Formation of ACP or another sparingly

soluble calcium phosphate cannot explain the strong exothermic peaks in Fig. 3a as formation of

these precipitates leads to limited heat exchange (SI, Table S2). However, in addition to P sorption,

the neutralization of OH- (Fig. 3b) is an exothermic process: H

+ + OH

- → H2O with a reaction heat

(∆Hr) of -56 kJ mol-1

. To estimate the heat of neutralization, 25 mg Filtralite®P was titrated with

0.005 M HNO3 to pH 7 in the same way as for the multi-points P sorption (SI Fig. S3b). According

to the result of the neutralization titration ∑Qmulti = -45 mJ, which is very close to ∑Qmulti for both

single-point and multiple-points P sorption titrations (Table 1). This is in line with formation of

ACP or another sparingly soluble calcium phosphate with limited heat production (SI, Table S2). In

fact, the estimated ∑Qmulti = -45 mJ corresponds to -36 kJ mol OH-1

, which is not very different

from strong acid/base neutralization heat of -56 kJ mol-1

(SI, Table S2). The small endothermic

humps after the first four P injections (Fig. 3c) may be ascribed to dissolution of small amounts of

the precipitate due to small pH decreases after each titration increment (Fig. 3b). The significant but

limited sorption after 25 injections, i.e. after 2 h (Fig. 3a) may be attributed to sorption of P by

CaCO3 formed by reaction of Ca(OH)2 with CO2 as shown subsequently for the P-Limestone

system.

90

3.4 P sorption by Limestone

None of the two Limestone fractions sorbed any P in the first 8 injections (Fig. 4a). On the contrary,

indigenous P in the Limestone (SI, Table S1) was desorbed. After the 25th

P injections, the two

Limestone size fractions had sorbed about 3 and 7 mmol kg-1

corresponding to 3 and 7% of the

added P (Table 1). However, in the single-point sorption both Limestone size fractions sorbed

approximately 80% of the added P. About 1% of the P sorbed in the single-point experiment was

extractable by water, whereas the PNaHCO3 accounted for 75-80% of the sorbed P (Fig 2b). Thus,

Limestone sorbed more P than Filtralite®P in the single point P addition experiment, and the P

sorbed by Limestone seems slightly more available than P sorbed by Filtralite®P but much more

available than P sorbed to CFH (Fig. 2b). No secondary phases were observed in the SEM images

(SI, Fig. S1d) or by XRD. EDS (SI, Fig. S1d) shows that P is randomly distributed on the surface of

the particle (overlay image not shown). The pH in the multiple-points titration started at 9.3 and

decreased to about 8 after the first injection and hereafter it gradually declined to 7.5 (Fig. 4b). The

alkaline initial pH and the rapid pH decline indicates occurrence of small amounts of CaO

(Ca(OH)2) in the Limestone samples that seems turned into CaCO3 after the first injection.

The observed sorption of P by the Limestone samples (and other CaCO3 products) is considered due

to precipitation of Ca phosphates but precipitation of Ca phosphates is a rather slow process that

needs an hour or more to be completed especially in the presence of carbonate ions (Cao et al.,

2007; Grossl and Inskeep, 1991). This reluctance may explain why Limestone sorbed no P during

the first multi-points injections with only 5 min between each injection. However, after 25

injections, i.e. after 2 h, significant amounts were sorbed and even more in the single-point

experiment lasting for 5 h (Table 1).

91

Fig. 4. Sorption of P by the two particle size fractions of Limestone. a) Showing the amount of sorbed P as a function of

injection number for direct comparison with the multi-points P thermogram; b) pH resulting from the P additions shown

in a), c) thermogram for multi-points titration of Limestone with 0.01 M NaH2PO4, d) corresponding integration of the

thermograms. Error bars are standard deviations. Exothermic upwards

The Limestone thermograms are dominated by exothermic peaks that gradually decrease during the

titration (Fig. 4c). The presence of an endothermic contribution during the first injection is probably

related to the pH drop and the possible CaCO3 precipitation (SI, Table S2) mentioned above. For

the multiple-points titration, ∑Qnorm = ~ -0.6 kJ kg-1

resulting in ∆Hr of -222 and -73 kJ mol-1

for the

2-1 mm and the 1-0.5 mm fractions, respectively. These enthalpies seem unrealistic high as for

example formation of brushite (CaHPO4 ∙ 2H2O) from the reaction of HPO42-

with CaCO3 only

results in ∆Hr of -8 kJ mol-1

(SI, Table S2). Furthermore, as shown in Table S2, ∆Hr for formation

of other sparing soluble Ca phosphates are positive (endothermic) and close to zero. Even formation

of hydroxyapatite (Ca5(PO4)3OH) from the reaction between CaCO3 and HPO42-

results in ∆Hr of

zero (SI, Table S2). The single-point thermograms (Fig. 2a) show a net exothermic reaction with

Qnorm values of -0.3 kJ kg-1

of solid resulting in ∆Hr= -15 kJ mol-1

P for both factions (Table 1),

92

which seems more sound. However, whether formation of brushite, apatite or another sparingly

soluble Ca phosphate is responsible for the P sorbed by Limestone (Table 1) cannot be determined

from these ITC data but the high pool of readily available P (PH2O + PNaHCO3) indicates that the

sorbed P did not occur in apatite, which needs HCl for dissolution (Hedley et al., 1982), but as a

more reactive Ca phosphate such as ACP, brushite, monetite or tricalcium phosphate (SI, Table S2).

3.5 CFH, Filtralite®

P and Limestone as P sorbing filters

In regard to the possible use as PSM in filter drains and ditches to reduce or stop the P transport

from agricultural fields to streams and lakes, CFH seems very promising because it is fast reacting

and possesses high P sorption capacity and binds P strongly in forms that are not easily redissolved.

Furthermore, the similarity between sorption isotherms and thermograms of the two particle size

fractions is important. The similar behavior of the two particle size fractions including sorption

strength, capacity and reaction rate shows that either fraction may be used. Probably the coarser (2-

1 mm) fraction will be preferred due to higher hydraulic conductivity.

Although the present investigation was not able to unambiguously show the precise mechanisms for

the reaction between P and Filtralite®P, the results obtained demonstrate that Filtralite

®P probably

has limited interests as a PSM for trapping P in agricultural drainage water, as it only exhibited fast

reaction and high P sorbing capacity at alkaline conditions (pH ≥~10). Hence, for fast P sorption at

neutral to acidic pH as found in most drainage and ditch waters, Filtralite®P appeared unsuited.

However, if pH in the P contaminated water is augmented to ~10 by base addition before the water

enters the filter, Filtralite®P might be an effective P sorbing filter material. In addition, it would be

an attractive sorbent because after being used, the P loaded Filtralite®P would be a useful P

fertilizer slowly releasing the sorbed P along with decreasing pH when spread on the field, where

the soil pH is neutral to acidic. On the other hand, P removal at such high pH might present a

problem if the treated water also becomes very alkaline. Then the outlet water leaving the

Filtralite®P filter would need neutralization before entering the recipient water body to avoid

harmful effect on the biota. Furthermore, the practical management of such a system with pH

regulations of the water before and after passing the filter might be rather costly reducing the

competitiveness of Filtralite®P as PSM.

93

The results obtained for Limestone demonstrated that it is a slowly reacting sorbent with limited

sorption capacity. It is therefore considered unsuited to trap P under extreme conditions with sudden

high water and P flows in drains as can be seen in drainage water from P rich fields during

rainstorms (Grant et al., 1996; Penn et al., 2012). However, Limestone and probably also other

CaCO3-based materials (as well as Filtralite®

P) may be used as PSM under condition with

prolonged contact between the P contaminated water and the filter material because CaCO3

materials are comparatively cheap and naturally occurring at many places.

4 Conclusion

Sorption isotherm and ITC results have clearly demonstrated that CFH possesses high efficiency as

P sorbent, whereas Filtralite®P only sorbs P at high pH while Limestone requires a high

sorbent:sorbate ratio and a long contact time for P removal. Therefore, CFH can unambiguously be

recommended in drainage filters while the use of Filtralite®P and Limestone has certain limitations.

For CFH, the results indicate formation of strong Fe-P bonding under formation of surface

complexation on outer surfaces followed by slow migration into interior sorption sites. For the Ca-

based materials, this study indicates that kinetics and pH are the most important factors in

determining formation of Ca phosphate precipitates. P retention by Filtralite®P under alkaline

conditions is speculated due to formation of amorphous calcium phosphate precipitate, which may

readily dissolve in water at neutral to acidic pH. The slow P retention by Limestone is considered

due to precipitation of Ca-phosphate(s), soluble in bicarbonate but not in water.

Supplemental information (SI)

Fig. S1a SEM image of CFH after sequental P titration with 0.01 M NaH2PO4 and corresponding

EDS analysis.

Fig. S1b SEM image of Filtralite®P after sequental P titration with 0.01 M NaH2PO4 and

corresponding EDS analysis.

Fig. S1c SEM image of Filtralite®P white spots after sequental P titration with 0.01 M NaH2PO4 and

corresponding EDS analysis.

94

Fig. S1d SEM image of Limestone after sequental P titration with 0.01 M NaH2PO4 and

corresponding EDS analysis.

Fig. S2. Neutralization thermogram (a) and integration of thermogram (b) from Filtralite®P (1-0.5

mm) titrated with 0.005 M HNO3 to pH ~7. Exothermic is upwards.

Table S1. General characterization of unfractionated (bulk) and the two particle size fractions of

CFH, Filtralite®P and Limestone.

Table S2. Equilibrium reactions for some of the possible reactions occurring, their corresponding

∆H values for reaction, product name, Ca:P ratio and solubility product (Ksp).

Acknowledgement

The project was carried out as part of the SupremeTech project funded by The Danish Council for

Strategic Research (grant no. 09-067280). Special thanks to Stuart Wilson for laboratory assistance.

95

References

Appel, C., Rhue, D., Kabengi, N., Harris, W., 2013. Calorimetric investigation of the nature of

sulfate and phosphate sorption on amorphous aluminum hydroxide, Soil Sci. 178, 180-188.

Arai, Y., Sparks, D., 2001. ATR–FTIR spectroscopic investigation on phosphate adsorption

mechanisms at the ferrihydrite–water interface, J. Colloid Interface Sci. 241, 317-326.

Ballantine, D.J., Tanner, C.C., 2010. Substrate and filter materials to enhance phosphorus removal

in constructed wetlands treating diffuse farm runoff: a review, New Zealand J. Agric. Res. 53, 71-

95. doi: 10.1080/00288231003685843.

Borggaard, O.K., Szilas, C., Gimsing, A.L., Rasmussen, L.H., 2004. Estimation of soil phosphate

adsorption capacity by means of a pedotransfer function, Geoderma 118, 55-61. doi:

10.1016/s0016-7061(03)00183-6.

Brunauer, S., Emmett, P.H., Teller, E., 1938. Adsorption of gases in multimolecular layers, J. Am.

Chem. Soc. 60, 309-319. doi: 10.1021/ja01269a023.

Cao, X., Harris, W., 2008. Carbonate and magnesium interactive effect on calcium phosphate

precipitation, Environ. Sci. Technol. 42, 436-442. doi: 10.1021/es0716709.

Cao, X., Harris, W.G., Josan, M.S., Nair, V.D., 2007. Inhibition of calcium phosphate precipitation

under environmentally-relevant conditions, Sci. Total Environ. 383, 205-215. doi:

10.1016/j.scitotenv.2007.05.012.

Chardon, W.J., Groenenberg, J.E., Temminghoff, E.J.M., Koopmans, G.F., 2012. Use of reactive

materials to bind phosphorus, J. Environ. Qual. 41, 636-646. doi:10.2134/jeq2011.0055

Claveau-Mallet, D., Wallace, S., Comeau, Y., 2012. Model of phosphorus precipitation and crystal

formation in electric arc furnace steel slag filters, Environ. Sci. Technol. 46, 1465-1470. doi:

10.1021/es2024884

Cucarella, V., Renman, G., 2009. Phosphorus sorption capacity of filter materials used for on-site

wastewater treatment determined in batch experiments – A comparative study, J. Environ. Qual. 38,

381-392. doi:10.2134/jeq2008.0192

96

Delgado, A., Scalenghe, R., 2008. Aspects of phosphorus transfer from soils in Europe, J. Plant

Nutr. Soil Sci. 171, 552-575. doi: 10.1002/jpln.200625052.

Dideriksen, K., Stipp, S., 2003. The adsorption of glyphosate and phosphate to goethite: a

molecular-scale atomic force microscopy study, Geochim. Cosmochim. Acta 67, 3313-3327.

Freire, E., Mayorga, O.L., Straume, M., 1990. Isothermal titration calorimetry, Anal. Chem. 62,

950A-959A.

Grant, R., Laubel, A., Kronvang, B., Andersen, H.E., Svendsen, L.M., Fuglsang, A., 1996. Loss of

dissolved and particulate phosphorus from arable catchments by subsurface drainage, Water Res.

30, 2633-2642.

Grossl, P.R., Inskeep, W.P., 1991. Precipitation of dicalcium phosphate dihydrate in the presence of

organic acids, Soil Sci. Soc. Am. J. 55, 670-675.

Harvey, O., Rhue, R., 2008. Kinetics and energetics of phosphate sorption in a multi-component Al

(III)–Fe (III) hydr (oxide) sorbent system, J. Colloid Interface Sci. 322, 384-393.

Hedley, M., Stewart, J., Chauhan, B., 1982. Changes in inorganic and organic soil phosphorus

fractions induced by cultivation practices and by laboratory incubations, Soil Sci. Soc. Am. J. 46,

970-976.

Kabengi, N.J., Daroub, S.H., Rhue, R.D., 2006. Energetics of arsenate sorption on amorphous

aluminum hydroxides studied using flow adsorption calorimetry, J. Colloid Interface Sci. 297, 86-

94. doi: 10.1016/j.jcis.2005.10.050.

Klimeski, A., Chardon, W.J., Turtola, E., Uusitalo, R., 2012. Potential and limitations of phosphate

retention media in water protection: A process-based review of laboratory and field-scale tests,

Agric. Food Sci. 21, 206-223.

Kwon, K.D., Kubicki, J.D., 2004. Molecular orbital theory study on surface complex structures of

phosphates to iron hydroxides: Calculation of vibrational frequencies and adsorption energies,

Langmuir 20, 9249-9254. doi:10.1021/la0487444

Lindsay, W.L., 1979. Chemical equilibria in soils. John Wiley & Sons, Inc., Canada.

97

Lyngsie, G., Borggaard, O.K., Hansen H.C.B., 2013. Testing phosphate sorption efficiency of five

potential drainage water filter materials under field-relevant conditions, Submitted to Water Res.

Mehta, N., Legg, J., Goring, C., Black, C., 1954. Determination of organic phosphorus in soils: I.

Extraction method, Soil Sci. Soc. Am. J. 18, 443-449.

Miltenburg, J., Golterman, H., 1997. The energy of the adsorption of o-phosphate onto ferric

hydroxide, Hydrobiologia 364, 93-97. doi: 10.1023/A:1003107907214.

Murphy, J., Riley, J.P., 1962. A modified single solution method for the determination of phosphate

in natural waters, Anal. Chim. Acta 27, 31-36. doi: 10.1016/S0003-2670(00)88444-5.

Nair, P., Logan, T., Sharpley, A., Sommers, L., Tabatabai, M., Yuan, T., 1984. Interlaboratory

comparison of a standardized phosphorus adsorption procedure, J. Environ. Qual. 13, 591-595.

Olsen, S.R., Sommers, L.E., 1982. Phosphorus, A.L. Page, R.H. Miller, D.R. Keeney (Eds.),

Methods of Soil Analyses, Part 2, Chemical and Microbiological Properties (2nd ed.), American

Society of Agronomy, Madison, WI, USA, 403–430.

Penn, C.J., McGrath, J.M., Rounds, E., Fox, G., Heeren, D., 2012. Trapping phosphorus in runoff

with a phosphorus removal structure, J. Environ. Qual. 41, 672-679. doi:10.2134/jeq2011.0045

Penn, C.J., Warren, J.G., 2009. Investigating phosphorus sorption onto kaolinite using isothermal

titration calorimetry, Soil Sci. Soc. Am. J. 73, 560-568. doi: 10.2136/sssaj2008.0198.

Penn, C.J., Bryant, R.B., Kleinman, P.J.A., Allen, A.L., 2007. Removing dissolved phosphorus

from drainage ditch water with phosphorus sorbing materials, J. Soil Water Conser. 62, 269-276.

Penn, C.J., Zhang, H., 2010. Isothermal titration calorimetry as an indicator of phosphorus sorption

behavior, Soil Sci. Soc. Am. J. 74, 502-511.

Persson, P., Nilsson, N., Sjöberg, S., 1996. Structure and bonding of orthophosphate ions at the iron

oxide–aqueous interface, J. Colloid Interface Sci. 177, 263-275.

Reinhardt, M., Gächter, R., Wehrli, B., Müler, B., 2005. Phosphorus retention in small constructed

wetlands treating agricultural drainage water, J. Environ. Qual. 34, 1251-1259.

98

Rhue, R.D., Appel, C., Kabengi, N., 2002. Measuring surface chemical properties of soil using

flow calorimetry 1, Soil Sci. 167, 782-790.

Schwertmann, U., 1964. Differenzierung der Eisenoxide des Bodens durch Extraction mit

Ammoniumoxalat-Lösung, Zietschrift für Pflanzenernährung, Düngung, Bodenkunde 105, 194-202.

Sharpley, A.N., Weld, J.L., Beegle, D.B., Kleinman, P.J.A., Gburek, W.J., Moore, P.A., Mullins,

G., 2003. Development of phosphorus indices for nutrient management planning strategies in the

United States, J. Soil Water Conser.58, 137-152.

Steinberg, G., 1981. What you can do with surface calorimetry, CHEMTECH December, 730e737 .

Stoner, D., Penn, C., McGrath, J., Warren, J., 2012. Phosphorus removal with by-products in a

flow-through setting, J. Environ. Qual. 41, 654-663. doi:10.2134/jeq2011.0049

Veith, J., Sposito, G., 1977. On the use of the Langmuir equation in the interpretation of

“adsorption” phenomena, Soil Sci. Soc. Am. J. 41, 697-702.

Vohla, C., Kõiv, M., Bavor, H.J., Chazarenc, F., Mander, Ü, 2011. Filter materials for phosphorus

removal from wastewater in treatment wetlands—A review, Eco. Eng. 37, 70-89. doi:

10.1016/j.ecoleng.2009.08.003.

Westholm, L.J., 2006. Substrates for phosphorus removal – Potential benefits for on-site wastewater

treatment? Water Res. 40, 23-36. doi.org/10.1016/j.watres.2005.11.006

Willett, I., Chartres, C., Nguyen, T., 1988. Migration of phosphate into aggregated particles of

ferrihydrite, J. Soil Sci. 39, 275-282.

99

Supplemental Information

Phosphate sorption by three potential filter materials as assessed by isothermal titration

calorimetry

Lyngsie, Gry1*

, Penn, Chad J.2, Hansen, Hans C. B.

1, Borggaard, Ole K.

1

1University of Copenhagen, Department of Plant and Environmental Sciences, Thorvaldsensvej 40,

DK-1871 Frederiksberg C. 2Oklahoma State University, Department of Plant and Soil Sciences, 368

Agricultural Hall, Stillwater, OK 74078-6028.

*Corresponding author: gryly@life.ku.dk

6 pages

2 figures

2 table

100

Fig. S1a SEM image of CFH after multiple P titration with 0.01 M NaH2PO4 and corresponding

EDS analysis.

Fig. S1b SEM image of Filtralite®P after multiple P titration with 0.01 M NaH2PO4 and

corresponding EDS analysis.

101

Fig. S1c SEM image of Filtralite®P white spots after multiple P titration with 0.01 M NaH2PO4 and

corresponding EDS analysis.

Fig. S1d SEM image of Limestone after multiple P titration with 0.01 M NaH2PO4 and

corresponding EDS analysis.

102

Fig. S2. Neutralization thermogram (a) and integration of thermogram (b) from Filtralite®P (1-0.5

mm) titrated with 0.005 M HNO3 to pH ~7. Exothermic is upwards.

103

Table S1. General characterization of unfractionated (bulk) and the two particle size fractions of

CFH, Filtralite®P and Limestone.

CFH Filtralite®

P Limestone

Bulk 2-1 mm 1-0.5 mm Bulk 2-1 mm 1-0.5 mm Bulk 2-1 mm 1-0.5 mm

AlTotal† mmol kg-1 32±2 ND ND 2345±161 ND ND 2.4±1.7 ND ND

FeTotal† mmol kg-1 7360±420 ND ND 737±30 ND ND 8.0±0.1 ND ND

FeOx‡ mmol kg-1 7310±90 7380±91 7440±230 22±0.2 ND ND 5.0±0.1 ND ND

CaTotal mmol kg-1 658±23 ND ND 1720±56 ND ND 9585±45 ND ND

CaHNO3§ mmol kg-1 ND ND ND ND 13.9±2.8 18.4±3.5 ND 10.8±1.4 16.3±6.1

MgTotal mmol kg-1 1390±97 ND ND 1562±110 ND ND 103±1 ND ND

MgHNO3 mmol kg-1 ND ND ND ND 0.3±0.1 0.4±0.1 ND 0.7±0.1 0.5±0.1

Pinorg.¶ mmol kg-1 8.0±0.3 ND ND 3.0±0.1 ND ND 1.0±0.1 ND ND

Polsen# mmol kg-1 0.0±0.0 ND ND 0.2±0.0 ND ND 0.2±0.0 ND ND

SSA†† m2 g-1 ND 32±2 32±11 ND 0.9±0.1 1.2±0.0 BDL BDL BDL

pH¤ 8.2±0.1 ND ND 11.9±0.2 ND ND 7.9±0.1 ND ND

† Total digestion: EPA 3052, ‡ Extraction with ammonium oxalate, §Extraction with 0.05 M HNO3, ¶Extraction with 6 M H2SO4, # Extraction with Olsen solution, †† Specific surface

area, ¤ pH measured in 0.01 M CaCl2, ND: Not determined, BDL below detection limit.

104

Table S2. Equilibrium reactions for some of the possible reactions occurring, their corresponding

∆H values for reaction, product name, Ca:P ratio and solubility product (Ksp). Reaction

∆Hr‡ Mineral phase Ca:P¤ Ksp (of product)¤

kJ/mol

Ca2+ + CO32- CaCO3 12 Calcite

H+ + OH- H2O -56

3Ca2+ + 2HPO42- + nH2O Ca3(PO4)2*nH2O + 2H+ endo Amophous calcium phosphate (ACP) ~1.5

Variable, more soluble than

crystalline phosphate

Ca2+ + HPO42- + 2H2O CaHPO4*2H2O 3 Brushite 1.0 2.49*10-7

CaCO3 + HPO42- + 2H2O CaHPO4*2H2O + CO3

2- -9 Brushite 1.0 2.49*10-7

Ca2+ + HPO42- CaHPO4 20 Monetite 1.0 1.26*10-7

3Ca2+ + 2HPO42- β-Ca3(PO4)2 + 2H+ 89 β-Tricalcium phosphate 1.5 1.20*10-29

3Ca2+ + 2OH- + 2HPO42- β-Ca3(PO4)2 + 2H2O 263 β-Tricalcium phosphate 1.5 1.20*10-29

8Ca2+ + 6HPO42- + 5H2O Ca8H2(PO4)6*5H2O + 4H+ 36 Octacalcium phosphate 1.3 1.25*10-47

5Ca2+ + 3HPO42- + H2O Ca5(PO4)3OH + 4H+ -2250 Hydroxyapatite 1.67 4.70*10-59

5CaCO3 + 3HPO42- + H+ + H2O Ca5(PO4)3OH + 5HCO3

- -2 Hydroxyapatite 1.67 4.70*10-59

Fe3+ + HPO42- + 2H2O FePO4 * 2H2O + H+ 25 Strengite 1.0

¤From Valsami-Jones E. (2001), ‡ values calculated from Woods and Garrels (1987).

References

Valsami-Jones, E., 2001. Mineralogical controls on phosphorus recovery from wastewaters, Mineralogical

Magazine 65, 611-620.

Woods, T.L., Garrels, R.M., 1987. Thermodynamic Values at Low Temperature for Natural Inorganic

Materials: An Uncritical Summary. Oxford University Press New York, NY.

105

Main findings

106

Chapter 8 – Combined discussion

In order to remove phosphate during both peak and base flows a PSM needs to react fast and to

strongly retain high amounts of phosphate even at relative low phosphate concentrations. The

phosphate removal efficiency of PSM is closely related to the content of sparingly soluble Al/Fe

oxides and/or easy soluble Ca/Mg compounds as well as the crystallographic properties of these

compounds including shape, size and porosity. Furthermore, other factors affecting PSM efficiency

include phosphate concentration, time of reaction, pH, electrolyte composition, solid:solution ratio

and whether a batch or flow-through method is used (Ballantine and Tanner, 2010; Johansson

Westholm, 2006; Klimeski et al., 2012; Vohla et al., 2011). Although outside the scope of this

study, the hydraulic conductivity of the PSM is also essential for use in high–flow drainage

systems. In this study the primary focus has been on testing PSMs at low phosphate concentrations

and short retention times.

8.1 Influence of retention time on PSM sorption

This PhD study clearly shows that retention time has an influence on PSMs’ phosphate sorption,

even among the relative short reaction times used (Papers I and II). Phosphate retention by the Fe-

based CFH and the Ca-based Filtralite®P is rapid compared with the other PSMs tested (Paper I).

Limestone needs more time to retain phosphate as seen both in the kinetic study in Paper I but in

particular in Paper III where limestone sorbed significantly higher amounts in the 5 h sorption

experiment than in the 2 h experiment.

The flow-through study, which simulates constant application of phosphate as may be seen in field

filter beds, clearly showed that sorption by Filtralite®P is positively and significantly dependent of

the reaction time (Paper II), when retention times were less than 10 min. In other words, an increase

in reaction time both increases the material’s capacity and affinity for phosphate sorption.

Increased sorption capacity with increasing retention times was also seen with Filtralite®P in a

flow-through column study by Herrmann et al. (2013b) and by other Ca-based PSMs with low

buffer capacity, such as gypsum, in flow-through study by Stoner et al. (2012). The batch kinetic

study (Paper I) also showed a clear positive correlation between reaction time and sorbed phosphate

was seen within both the 24 min and 48 h experiments.

For the Fe-based CFH system, however, the time dependency is not straight forward (Paper II). At

low concentrations a positive but non-significant trend is seen meaning that the sorption capacity

increases with increasing retention time, in line with the kinetic study in Paper I. For the highest

peak concentration (32 µM) an inverse relationship between reaction time and capacity was

observed (Paper II). This indicates that at reaction times between 0.5 and 6 min the limiting factor

of the capacity is not the kinetics of the reaction, but the speed in which phosphate is added to the

107

particle. The inverse relationship between retention time and sorption capacity was also seen on a

floodplain soil from the Illinois River watershed (Penn et al., unpublished results). At the longest

retention time (9 min) an increase in phosphate sorption capacity is seen (Paper II). This could be

interpreted as diffusion of phosphate from the surface to less accessible interior sorption sites. This

is supported by ITC results for the reaction between phosphate and CFH, where phosphate sorption

seems to occur in two steps including a fast initial sorption followed by slower reaction (Paper III).

As CFH consists of so-called 2-line ferrihydrite (Paper I) this interpretation is in line with the time-

dependent sorption of phosphate by synthetic ferrihydrite described by (Willett et al., 1988).

8.2 Influence of pH on PSM sorption

The Fe-based CFH does not seem to be significantly dependent on pH at the neutral to slightly

alkaline conditions used in this study. In contrast, the phosphate-Filtralite®P system is very pH

dependent (Papers I-III). Filtralite®P possesses high capacity to fast sorption of phosphate at

alkaline pH (pH ≥ 10) but loses this capacity at neutral pH (pH < 8.2) (Paper III). However, after

longer reaction time this PSM regains some phosphate sorption capacity probably because the CaO

by reaction with CO2 is turned into CaCO3, which has some long-term (hours, days) phosphate

sorption capacity (Papers I and III; Vohla et al., 2011).

8.3 Influence of the background electrolyte on PSM sorption

The composition of the background electrolyte affects phosphate sorption as shown by comparing

isotherms made with the artificial electrolyte (6 mM KNO3) with isotherms made with natural

drainage water (Paper I). These solutions had the same ionic strength but natural drainage water was

dominated by Ca2+

and HCO3-. In addition, the influence of the composition of the background

electrolyte differed among the three materials tested including CFH, Filtralite®P and limestone,

although the sorption capacity for all three materials decreased when natural drainage water was

used instead of 6 mM KNO3 (Paper I). A reduction in sorption capacity when using a more

buffered solution that also contained competitive anions is in agreement with observations on the

Filtralite®P capacity to sorb phosphate from wastewater and from an artificial phosphate solution

(Ádám et al., 2007; Herrmann et al., 2013b).

The ideal composition of an electrolyte resembling Danish drainage water from agricultural areas

should be dominated by Ca2+

and HCO3-, but collection of natural drainage water in the quantities

needed for the studies described in Papers I and II was considered too time consuming (Sø et al.,

2011). Furthermore, the composition of natural drainage water exhibits great spatial and temporal

variation. A rather comprehensive attempt to produce artificial drainage water containing

Ca(HCO3)2 with the same ionic strength as found on average for natural drainage water samples

(and 6 mM KNO3) failed due to precipitation of CaCO3 and a test of phosphate sorption from a

Ca(NO3)2 solution resulted in higher sorption using this background electrolyte the using KNO3,

108

probably due to precipitation of Ca-phosphate (Lyngsie, unpublished results). It was therefore

decided to test the PSMs’ ability to remove phosphate from solutions with KNO3 as background

electrolyte.

8.4 Influence of phosphate concentration on PSM sorption

Overall, this study has shown that Fe-based and Ca-based PSMs possess different sorption

properties when exposed to high and low phosphate concentrations. A higher sorption capacity is,

as expected, seen with an increase in the phosphate concentration (Papers I and II).

CFH has higher sorption capasity at low phosphate concentrations than all the other tested materials

(Papers I and II). Furthermore, CFH exhibited a steady-state sorption behavior where effluent

solution concentration stayed at a constant level, which was less than the inlet concentration, i.e.

CFH continued to sorb phosphate at a constant rate (Paper II and III). The amount of phosphate

removed at this steady-state increased with increasing phosphate concentration (Papers II and III)

indicating an equilibrium, where the phosphate uptake is proportional with the application rate. A

steady-state behavior was also seen by Chardon et al. (2012) who did a phosphate column study

with Fe oxide-coated sand. Based on mainly the ITC data (Paper III), the phosphate removal

mechanism by CFH is considered due to surface complexation, in line with phosphate sorption by

Fe oxides and Fe-based PSMs (Harvey and Rhue, 2008; Kwon and Kubicki, 2004; Penn and Zhang,

2010).

The Ca-based PSMs (limestone, shell-sand and Filtralite®P) show different sorption behavior

compared to CFH (Papers I and III). The non-Langmuir but s-shaped sorption isotherms seen in

Paper I indicate a change in the concentration dependency in these PSM-phosphate systems, which

has also been observed by other studies of similar systems e.g. Ádám et al. (2007), Cucarella and

Renman (2009), Freeman and Rowell (1981) Kaasik et al. (2008). The s-shaped appearance of the

isotherms is interpreted as a change in sorption reactions from adsorption at the low concentrations

(inlets < ~30 µM) to precipitation at higher concentrations (Paper I). The ITC results (Paper III)

indicates that amorphous calcium phosphate (ACP) is the precipitating Ca-phosphate in the

Filtralite®P system, at high P concentrations and high pH. This may explain both the high

desorption tendency and the dissolution of the product at pH < 8.2 (Paper III). The precipitation of

ACP is supported by FT-IR studies of the phosphate-Filtralite®P systems (Herrmann et al., 2013a).

Even though CFH and Filtralite®P exhibit high phosphate sorption capacities (Paper II), these

capacities are much lower than the Pmax values obtained by applying the Langmuir equation to the

results obtained by long-time (7 days) sorption at high phosphate concentrations (Paper II).

However, Pmax corresponds to an infinitely high phosphate concentration, which is irrelevant for

PSMs, where fast sorption at relatively low concentrations is mandatory.

109

8.5 Retention of sorbed phosphate by the PSMs

Even though fast and substantial phosphate sorption is mandatory for the practical use of PSMs it is

equally important, but often neglected, that PSMs are able to retain the sorbed phosphate when

sorption condition change i.e. at fluctuating phosphate concentration seen in drainage water. CFH is

the best retainer of phosphate of the tested PSMs (Papers I-III). CFH desorbs more in absolute

amount when it has been exposed to higher inlet phosphate concentrations, but in relative amounts

desorption from CFH does not exceed 10% of the retained phosphate (Papers I and II). This strong

retention was also confirmed by sequential extraction where less that 5% of the sorbed phosphate

was in the labile pool (Paper III). Strong phosphate retention by Fe oxides is also seen studies of

Chardon et al. (2012), Harvey and Rhue (2008), Zeng et al. (2004). Furthermore, the phosphate

retention by CFH was not dependent of the reaction time (Paper II). This is a very important

property for a PSM.

In general, the Ca-based systems are more prone to release of retained phosphate compared to CFH

(Papers I-III). In the batch desorption study these PSMs desorbed 25-100% with a trend that the

small particle size fractions desorbed the least. A decrease in desorption is seen with increasing

retention time indicating that the retained phosphate (Ca-phosphate precipitate) becomes less

soluble when it has had longer time to react but further investigations are needed to confirm this

(Paper II). The relative desorption seen for the PSMs tested in this study is rather high compared to

the relative (10-32%) desorptions seen for six industrial by-products (Penn et al., 2011) which may

be attributed to the different reaction times used in the two studies.

Sequential extraction (Paper III), which was conducted on samples that had reacted with phosphate

for 5 h, showed that phosphate retained by Filtralite®P was more labile than that retained by

limestone even though limestone desorbed more than Filtrate®P in the batch desorption, where

reaction was 24 min (Paper I). This indicates that desorption is related to both sorption reaction time

and concentration and that phosphate retained by CaO is more labile than CaCO3. The higher

degree of stability seen at higher phosphate concentration and longer reaction time is likely due to

rather slow formation of more stable Ca-phosphate(s) (Grossl and Inskeep, 1991; Chapter 4).

8.6 Influence of test method on PSM sorption

Batch studies work well as a quick and rough screening tool but do not resemble field conditions

and often have a tendency to overestimate the sorption capacity of the solid if long retention times

are used (Papers I and II; Klimeski et al., 2012; Vohla et al., 2011). Flow-through studies are more

time consuming and therefore more expensive than batch methods, but they better at mimic field

conditions and the extra information that can be obtained by this method makes it worthwhile. In

addition, there is the prospect of making a PSM-specific prediction model based on flow-through

110

data which can be used to predict the lifetime of the PSM and help predict size of the removal

structure to be installed in the field (Paper II). Nevertheless more and better model work is needed

and especially a model that can take the PSM desorption into consideration as well.

The time and concentration dependency found in this study makes it clear how important it is to test

PSMs with reaction times and concentrations that may be expected in a phosphate removal

structure. Furthermore, the dependency of phosphate sorption on parameters such as electrolyte

composition (Paper I; Ádám et al., 2007; Herrmann et al., 2013b) and pH (Paper III; Genz et al.,

2004; Herrmann et al., 2013a; Herrmann et al., 2013b; Karabelnik et al., 2012; Zeng et al., 2004)

shows that the PSM test conditions can have great influence on the results and thereby making

comparison between different studies difficult. Hence, this study supports the recommendation

proposed by Klimeski et al. (2012), who recommended standardized laboratory test protocols for

PSM testing in order to ensure comparability of the results.

111

Chapter 9 - Conclusion

The primary aim of the PhD study was to identify PSMs that are capable of removing and retaining

phosphate from drainage water at low concentrations and with fast retention times. Additionally,

this thesis aimed to get a better understanding of the sorption reactions and kinetics related to

different types of commercial available PSMs in order to get a general understanding of the PSM

behavior. The major conclusions are:

Of the five tested commercial available materials the Fe-based CFH showed the best

sorption properties: CFH is capable of sorbing phosphate at both base flow and high peak

concentrations and retaining >90% of the sorbed phosphate regardless of the time of

reaction and concentration. Sorption by CFH is a fast reaction although CFH also sorbs

more when reaction time becomes longer. No secondary iron phosphate is detected on the

surface of CFH and the exothermic heat of reaction and pH range suggest that adsorption by

surface complexation is the sorption mechanism.

Filtralite®P lacks the necessary properties of an effective PSM in a drainage system both

because it is not capable of retaining the sorbed phosphate, and because it only works at

alkaline pH. However, Filtralite® P has a high affinity towards phosphate at high phosphate

concentrations and the kinetics of the reaction is fairly rapid if pH is high (>~10). It could

therefore be used in systems with consistent high phosphate concentrations at the inlet and

where pH can be adjusted so the effluent does not harm the environment. The fact that

Filtralite®P is prone to desorb at neutral to acid pH, the sorption is considered due to

precipitation forming of precipitant e.g. amorphous calcium phosphate that dissolves at

neutral to acid pH.

Limestone has proven to be unsuited as PSM in drainage water systems as it does not sorb

phosphate at low concentration nor does it retain it at short reaction times. However,

limestone can sorb phosphate under production of a fairly stable product if the phosphate

concentration is high, retention time long and solid:solution ratio low.

Both Fe- and Ca-based PSM’ sorption capacity is dependent on the inlet phosphate

concentration which makes it evident that PSM need to be tested at field relevant

concentrations in order to evaluate their sorption properties in relation to what is needed as

efficient drainages water filters.

Flow-through studies are superior to batch studies in regard to evaluation of PSM because it

better mimics field conditions. Furthermore, flow-through data can be used to estimate a

realistic lifetime of a material and therefore is useful when designing a phosphate removal

structure.

112

Chapter 10 – Perspective

The focus of the thesis was solely on P sorbing aspects of the PSM, despite the hydraulic

conductivity of PSM is equally important in a P removal structure. Even though this study points to

CFH as being the best PSM, the size of the particle and particle sorting of the raw material has

proven to cause problems for the hydraulic conductivity (Canga et al. 2012). Hence, may CFH not

the ideal PSM. The Ultrasorb project, which is a collaboration between University of Copenhagen

and two industrial partners, tries to address this challenge.

Another issue, which has not been addressed in this thesis, is the cost effectiveness of the PSM. It is

it evident that a PSM needs to be cost effective in order to be applied in as a drainage water filter.

Commercially available PSM may be too expensive, which is one of the several reasons why

industrial by-products have been tested in several studies. In line with using a by-product is the

reuse of used PSM either as a direct soil amendment or in the fertilizer industry as a raw material.

Both the use of by-product and reuse of PSM will contribute to a more sustainable system; but they

need to be investigated further. By-products may release pollutants to the environment, along with

the phosphate because they may contain, also have retained other compounds e.g. heavy metals,

antibiotics, growth hormones, and pesticides that may threaten the environment.

113

References

Ádám, K., Sovik, A., Krogstad, T., Heistad, A., 2007. Phosphorous removal by the filter materials

light-weight aggregates and shellsand-a review of processes and experimental set-ups for improved

design of filter systems for wastewater treatment, Vatten, 63, 245.

Andersen, H.E., Larsen, S.E., Kronvang, B., Hansen, K.M., Laubel, A., Windolf, J., Muus, K.,

2006. Fosfat i drænvand, Vand og jord, 13, 152-156, (in Danish).

Ballantine, D.J., Tanner, C.C., 2010. Substrate and filter materials to enhance phosphorus removal

in constructed wetlands treating diffuse farm runoff: a review, New Zealand J. Agr. Res. 53, 71-95.

doi: 10.1080/00288231003685843.

Barrow, N., 1983. A mechanistic model for describing the sorption and desorption of phosphate by

soil, J. Soil Sci. 34, 733-750.

Borggaard, O.K., Elberling, B., 2004. Pedological Biogeochemistry, Institute of Geography,

University of Copenhagen, Copenhagen

Borggaard, O.K., Raben-Lange, B., Gimsing, A.L., Strobel, B.W., 2005. Influence of humic

substances on phosphate adsorption by aluminium and iron oxides, Geoderma 127, 270-279. doi:

10.1016/j.geoderma.2004.12.011.

Breuning-Madsen, H., Noerr, A.H., Holst, K.A., 1992. The Danish Soil Classification. The Royal

Danish Geographical Society, Copenhagen.

Bryant, R.B., Buda, A.R., Kleinman, P.J., Church, C.D., Saporito, L.S., Folmar, G.J., Bose, S.,

Allen, A.L., 2012. Using flue gas desulfurization gypsum to remove dissolved phosphorus from

agricultural drainage waters, J. Environ. Qual. 41, 664-671.

Buda, A.R., Koopmans, G.F., Bryant, R.B., Chardon, W.J., 2012. Emerging technologies for

removing nonpoint phosphorus from surface water and groundwater: Introduction, J. Environ. Qual.

41, 621-627.

Cade-Menun, B.J., Bell, G., Baker-Ismail, S., Fouli, Y., Hodder, K., McMartin, D.W., Perez-

Valdivia, C., Wu, K., 2013. Nutrient loss from Saskatchewan cropland and pasture in spring

snowmelt runoff, Can. J. Soil Sci. 93, 1-14.

Canga, E., Iversen, B. V., Kjaergaard, C., 2012. A simplified transfer function for estimating

saturated hydraulic conductivity of porous drainage filters. Water, Air Soil Pol. (under review)

114

Cao, X., Harris, W., 2008. Carbonate and magnesium interactive effect on calcium phosphate

precipitation, Environ. Sci. Technol. 42, 436-442. doi: 10.1021/es0716709.

Cao, X., Harris, W.G., Josan, M.S., Nair, V.D., 2007. Inhibition of calcium phosphate precipitation

under environmentally-relevant conditions, Sci. Total Environ. 383, 205-215. doi:

10.1016/j.scitotenv.2007.05.012.

Chardon, W.J., Groenenberg, J.E., Temminghoff, E.J.M., Koopmans, G.F., 2012. Use of reactive

materials to bind phosphorus, J. Environ. Qual. 41, 636-646.

Cucarella, V., Renman, G., 2009. Phosphorus sorption capacity of filter materials used for on-site

wastewater treatment determined in batch experiments - A comparative study, J. Environ. Qual. 38,

381-392.

Cui, L., Zhu, X., Ma, M., Ouyang, Y., Dong, M., Zhu, W., Luo, S., 2008. Phosphorus sorption

capacities and physicochemical properties of nine substrate materials for constructed wetland, Arch.

Environ. Contam. Toxicol. 55, 210-217.

Delgado, A., Scalenghe, R., 2008. Aspects of phosphorus transfer from soils in Europe, J. Plant

Nutr. Soil Sci. 171, 552-575. doi: 10.1002/jpln.200625052.

Diaz, O., Reddy, K., Moore, P., 1994. Solubility of inorganic phosphorus in stream water as

influenced by pH and calcium concentration, Water Res. 28, 1755-1763.

Ferguson, J.F., McCarty, P.L., 1971. Effects of carbonate and magnesium on calcium phosphate

precipitation, Environ. Sci. Technol. 5, 534-540. doi: 10.1021/es60053a005.

Freeman, J.S., Rowell, D.L., 1981. The adsorption and precipitation of phosphate onto calcite,

J.Soil Sci. 32, 75-84.

Gburek, W.J., Barberis, E., Haygarth, P.M., Kronvang, B. & Stamm, C. 2005. Phosphorus mobility

in the landscape. In: Phosphorus: agriculture and the environment (eds Sims, J.T., Sharpley, A.N.),

941–979. Agronomy Monographs No. 46, American Society of Agronomy, Inc., Madison,

Wisconsin, USA.

Genz, A., Kornmüller, A., Jekel, M., 2004. Advanced phosphorus removal from membrane filtrates

by adsorption on activated aluminium oxide and granulated ferric hydroxide, Water Res. 38, 3523-

3530.

115

Grant, R., Laubel, A., Kronvang, B., Andersen, H.E., Svendsen, L.M., Fuglsang, A., 1996. Loss of

dissolved and particulate phosphorus from arable catchments by subsurface drainage, Water Res.

30, 2633-2642.

Groenenberg, J.E., Chardon, W.J., Koopmans, G.F., 2013. - Reducing Phosphorus Loading of

Surface Water Using Iron-Coated Sand, J. Environ. Qual. 42, 250-259. doi:

10.2134/jeq2012.0344.

Grossl, P.R., Inskeep, W.P., 1991. Precipitation of Dicalcium Phosphate Dihydrate in the Presence

of Organic Acids, Soil Sci. Soc. Am. J. 55, 670-675.

Hansen, H.C.B., Bergen Jensen, M., Magid, J., 1999. Phosphate sorption to matrix and fracture wall

materials in a Glossaqualf, Geoderma 90, 243-261.

Harvey, O., Rhue, R., 2008. Kinetics and energetics of phosphate sorption in a multi-component Al

(III)–Fe (III) hydr (oxide) sorbent system, J. Colloid Interface Sci. 322, 384-393.

Heal, K., Younger, P.L., Smith, K., Glendinning, S., Quinn, P., Dobbie, K., 2003. Novel use of

ochre from mine water treatment plants to reduce point and diffuse phosphorus pollution. Land

Contam. Reclam. 11, 145-152.

Heathwaite, L., Sharpley, A., Bechmann, M., 2003. The conceptual basis for a decision support

framework to assess the risk of phosphorus loss at the field scale across Europe, J. Plant Nutr. Soil

Sci.166, 447-458. doi: 10.1002/jpln.200321154.

Heiberg, L., Pedersen, T.V., Jensen, H.S., Kjaergaard, C., Hansen, H.C.B., 2010. A comparative

study of phosphate sorption in lowland soils under oxic and anoxic conditions, J. Environ. Qual. 39,

734-743.

Herrmann, I., Jourak, A., Gustafsson, J.P., Hedström, A., Lundström, T.S., Viklander, M., 2013a.

Modeling phosphate transport and removal in a compact bed filled with a mineral-based sorbent for

domestic wastewater treatment, J. Contam. Hydrol.

Herrmann, I., Jourak, A., Hedström, A., Lundström, T.S., Viklander, M., 2013b. The effect of

hydraulic loading rate and influent source on the binding capacity of phosphorus filters, PloS one 8,

e69017.

116

Hiemstra, T., Van Riemsdijk, W.H., 1996. A surface structural approach to ion adsorption: The

charge distribution (CD) model, J. Colloid Interface Sci. 179, 488-508. doi:

http://dx.doi.org/10.1006/jcis.1996.0242.

Johansson Westholm, L., 2006. Substrates for phosphorus removal - Potential benefits for on-site

wastewater treatment? Water Res. 40, 23-36. doi: 10.1016/j.watres.2005.11.006.

Johnes, P.J., 2007. Uncertainties in annual riverine phosphorus load estimation: Impact of load

estimation methodology, sampling frequency, baseflow index and catchment population density, J.

Hydrol. 332, 241-258. doi: 10.1016/j.jhydrol.2006.07.006.

Jørgensen, J.O., Kronvang, B., Paulsen, I., 2001. Fosfor i jord og vand – udvikling, status og

perspektiver, 380, 45-54, (in Danish).

Kaasik, A., Vohla, C., Mõtlep, R., Mander, Ü, Kirsimäe, K., 2008. Hydrated calcareous oil-shale

ash as potential filter media for phosphorus removal in constructed wetlands, Water Res. 42, 1315-

1323. doi: 10.1016/j.watres.2007.10.002.

Karabelnik, K., Kõiv, M., Kasak, K., Jenssen, P.D., Mander, Ü, 2012. High-strength greywater

treatment in compact hybrid filter systems with alternative substrates, Ecol. Eng. 49, 84-92.

Kleinman, P.J., Sharpley, A.N., Saporito, L.S., Buda, A.R., Bryant, R.B., 2009. Application of

manure to no-till soils: phosphorus losses by sub-surface and surface pathways, Nutr. Cycling

Agroecosyst. 84, 215-227.

Klimeski, A., Chardon, W.J., Turtola, E., Uusitalo, R., 2012. Potential and limitations of phosphate

retention media in water protection: A process-based review of laboratory and field-scale tests, Agr.

Food Sci. 21, 206-223.

Koskiaho, J., Ekholm, P., Räty, M., Riihimäki, J., Puustinen, M., 2003. Retaining agricultural

nutrients in constructed wetlands—experiences under boreal conditions, Ecol. Eng. 20, 89-103. doi:

10.1016/S0925-8574(03)00006-5.

Kronvang, B., Behrendt, H., Andersen, H.E., Arheimer, B., Barr, A., Borgvang, S.A., Bouraoui, F.,

Granlund, K., Grizzetti, B., Groenendijk, P., Schwaiger, E., Hejzlar, J., Hoffmann, L., Johnsson, H.,

Panagopoulos, Y., Lo Porto, A., Reisser, H., Schoumans, O., Anthony, S., Silgram, M., Venohr, M.,

Larsen, S.E., 2009a. Ensemble modelling of nutrient loads and nutrient load partitioning in 17

European catchments, J. Environ. Monit. 11, 572-583.

117

Kronvang, B., Rubæk, G.H., Heckrath, G., 2009b. International phosphorus workshop: Diffuse

phosphorus loss to surface water bodies - Risk assessment, mitigation options, and ecological

effects in river basins all rights reserved, J. Environ. Qual. 38, 1924-1929.

Kwon, K.D., Kubicki, J.D., 2004. Molecular orbital theory study on surface complex structures of

phosphates to iron hydroxides: Calculation of vibrational frequencies and adsorption energies,

Langmuir 20, 9249-9254.

McDowell, R.W., Sharpley, A.N., Condron, L.M., Haygarth, P.M., Brookes, P.C., 2001. Processes

controlling soil phosphorus release to runoff and implications for agricultural management, Nutr.

Cycl. Agroecosyst. 59, 269-284.

Mengel, K., Kosegarten, H., Kirkby, E.A., Appel, T., 2001. Principles of plant nutrition. Springer.

OECD, 2013a. Nutrients: Nitrogen and phosphorus balances, in: Anonymous . Organisation for

Economic Co-operation and Development, 67-78.

OECD, 2013b. Water quality: Nitrates, phosphorus and pesticides, in: Anonymous . Organisation

for Economic Co-operation and Development, 117-130.

Penn, C.J., McGrath, J.M., Rounds, E., Fox, G., Heeren, D., 2012. Trapping phosphorus in runoff

with a phosphorus removal structure, J. Environ. Qual. 41, 672-679.

Penn, C.J., Bryant, R.B., Callahan, M.P., McGrath, J.M., 2011. Use of industrial by-products to

sorb and retain phosphorus, Commun. Soil Sci. Plant Anal. 42, 633-644. doi:

10.1080/00103624.2011.550374.

Penn, C.J., Bryant, R.B., Kleinman, P.J.A., Allen, A.L., 2007. Removing dissolved phosphorus

from drainage ditch water with phosphorus sorbing materials, J. Soil Water Conserv. 62, 269-276.

Penn, C.J., Zhang, H., 2010. Isothermal titration calorimetry as an indicator of phosphorus sorption

behavior, Soil Sci. Soc. Am. J. 74, 502-511.

Poulsen, H.D., Rubaek, G.H., 2005. Fosfat i dansk landbrug - Omsætning, tab og virkemidler mod

tab. DJF rappot Husdyrbrug , 93-121, (in Danish).

Reinhardt, M., Gächter, R., Wehrli, B., Müler, B., 2005. Phosphorus retention in small constructed

wetlands treating agricultural drainage water, J. Environ. Qual. 34, 1251-1259.

118

Sibrell, P.L., Montgomery, G.A., Ritenour, K.L., Tucker, T.W., 2009. Removal of phosphorus from

agricultural wastewaters using adsorption media prepared from acid mine drainage sludge, Water

Res. 43, 2240-2250.

Smil, V., 2000. Phosphorus is the Environment: Natural flows and human interferences, Annu. Rev.

Energy Environ. 25, 53-88. doi: 10.1146/annurev.energy.25.1.53.

Sø, H.U., Postma, D., Jakobsen, R., Larsen, F., 2011. Sorption of phosphate onto calcite; results

from batch experiments and surface complexation modeling, Geochim. Cosmochim. Acta 75, 2911-

2923. doi: 10.1016/j.gca.2011.02.031.

Søndergaard, M., Jeppesen, E., Peder Jensen, J., Lildal Amsinck, S., 2005. Water framework

directive: ecological classification of Danish lakes, J. Appl. Ecol. 42, 616-629. doi: 10.1111/j.1365-

2664.2005.01040.x.

Stoner, D., Penn, C., McGrath, J., Warren, J., 2012. Phosphorus removal with by-products in a

flow-through setting, J. Environ. Qual. 41, 654-663.

Strauss, R., Brümmer, G.W., Barrow, N.J., 1997. Effects of crystallinity of goethite: II. Rates of

sorption and desorption of phosphate, Eur. J. Soil Sci. 48, 101-114. doi: 10.1111/j.1365-

2389.1997.tb00189.x.

Tarkalson, D.D., Leytem, A.B., 2009. Phosphorus mobility in soil columns treated with dairy

manures and commercial fertilizer, Soil Sci. 174, 73-80.

van Riemsdijk, W.H., Boumans, L.J.M., de Haan, F.A.M., 1984. Phosphate sorption by soils: I. A

model for phosphate reaction with metal-oxides in soil1, Soil Sci. Soc. Am. J. 48, 537.

Vohla, C., Kõiv, M., Bavor, H.J., Chazarenc, F., Mander, Ü, 2011. Filter materials for phosphorus

removal from wastewater in treatment wetlands - A review, Ecol. Eng. 37, 70-89. doi:

10.1016/j.ecoleng.2009.08.003.

Willett, I., Chartres, C., Nguyen, T., 1988. Migration of phosphate into aggregated particles of

ferrihydrite, J. Soil Sci. 39, 275-282.

Zeng, L., Li, X., Liu, J., 2004. Adsorptive removal of phosphate from aqueous solutions using iron

oxide tailings, Water Res. 38, 1318-1326.