<%BANNER%>

Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2012-12-31.

Permanent Link: http://ufdc.ufl.edu/UFE0042481/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2012-12-31.
Physical Description: Book
Language: english
Creator: Wu, Tingting
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: Environmental Engineering Sciences -- Dissertations, Academic -- UF
Genre: Environmental Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by Tingting Wu.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Sansalone, John.
Electronic Access: INACCESSIBLE UNTIL 2012-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042481:00001

Permanent Link: http://ufdc.ufl.edu/UFE0042481/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2012-12-31.
Physical Description: Book
Language: english
Creator: Wu, Tingting
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: Environmental Engineering Sciences -- Dissertations, Academic -- UF
Genre: Environmental Engineering Sciences thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Statement of Responsibility: by Tingting Wu.
Thesis: Thesis (Ph.D.)--University of Florida, 2010.
Local: Adviser: Sansalone, John.
Electronic Access: INACCESSIBLE UNTIL 2012-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0042481:00001


This item has the following downloads:


Full Text

PAGE 1

1 MASS TRANSFER EQUILIBRIA AND KINET ICS OF GRANULAR FILT ER MATERIALS FOR PHOSPHORUS IN URBAN WATER By TINGTING WU A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Tingting Wu

PAGE 3

3 This work is dedicated to my father Changzhi Wu; my mother Yuqing Ma ; and my brother Yue Wu.

PAGE 4

4 ACKNOWLEDGMENTS I wish to express my sincerest and deepest appreciation and gratitude to my advisor, Dr. John J. Sansalone, for his guidance and encouragement throughout my research. His patience and support helped me overcome difficult situations and finish this dissert ation. His words of urgings and careful reading and reviewing of all my writing will never be forgotten. I came to UF to gain knowledge and capability to take me to a higher professional level, what I come away with is so much more. In th e same vein, I want to thank my doctoral committee---Dr. James P. Heaney, Dr. Joseph J. Delfino and Dr. Mark W. Clark for their valuable support and advices on my research. My thanks also go to the past and present members of my group, for many fond memories and insightful discussions involving our research. I am grateful to Mr. Gary Scheiffele and Mr. Bill Brubaker for helping me with the analysis conducted in PERC. Many friends have helped me overcome setbacks and stay focused on my graduate study. I greatly value their f riendship and deeply appreciate their belief in me. Special thanks go to Fangyuan Hua, my previous roommate, for being such a great companion and helping each other stay affirmative through these years. Finally, I would like to express my heart felt grati tude to my parents and brother. They are always behind me and I could hardly accomplish anything without the love and support of my family.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 8 LIST OF FIGURES ........................................................................................................ 10 A BSTRACT ................................................................................................................... 12 CHAPTER 1 GLOBAL INTRODUCTION ..................................................................................... 14 2 BATCH EQUILIBRIA BETWEEN AQUEOUS PHOSPHORUS SPECIES AND ALUMINUM OXIDE COATED GRANULAR SUBSTRATES ................................... 21 Background ............................................................................................................. 21 Objectives ............................................................................................................... 23 Materials and Methodology ..................................................................................... 23 Sorbents and Media Preparation ...................................................................... 23 Ra w substrate (UCM) preparation ............................................................. 24 Aluminum oxide coated media (AOCM) preparation .................................. 24 Chemicals ......................................................................................................... 25 s) ......................................................... 25 Powder X ray Diffraction (XRD) ..................................................................... 25 Fourier Transform Infrared Spectrometry (FTIR) .............................................. 26 Experimental Matrices and Adsorption Equilibria (I sotherm) Study .................. 26 Equilibria Model ...................................................................................................... 27 Results and discussion ........................................................................................... 30 Physicoche mical Characteristics of Sorbents ................................................... 30 Studies in Synthetic Runoff Matrix (Hypothesis (1) and (2)) ............................. 33 Studies in Rainfall runoff and Wastewater Matrices (Hypothesis (3)) ............... 38 Summary ................................................................................................................ 41 Nomenclature ......................................................................................................... 43 3 BATCH EQUILIBRIA B ETWEEN AQUEOUS PHOSPHORUS AND GRANULAR FILTER MATERIALS .............................................................................................. 60 Background ............................................................................................................. 60 Objectives ............................................................................................................... 62 Materials and Methods ............................................................................................ 63 Media ................................................................................................................ 63 Physical Characteristics ................................................................................... 65 Chemicals ......................................................................................................... 65 Batch Adsorption Study .................................................................................... 65

PAGE 6

6 Phosphorus Analysis ........................................................................................ 66 Metal Leachi ng Tests ....................................................................................... 66 The Freundlich Isotherm and Unified Sorption Variable, Ku ............................. 67 Results .................................................................................................................... 69 Physical Characteristics ................................................................................... 69 P Removal Efficiency and Sorption Isotherm ................................................... 70 Metal Leaching from Media .............................................................................. 72 Discussions and Implications ........................................................................... 73 Summary ................................................................................................................ 77 Nomenclature ......................................................................................................... 78 4 MASS TRANSFER KINETICS OF GRANULAR FILTER MATERIALS FOR PHOSPHORUS IN AQUEOUS MATRICES ........................................................... 89 Background ............................................................................................................. 89 Objectives ............................................................................................................... 91 Materials and Methods ............................................................................................ 92 Media ................................................................................................................ 92 Chemicals ......................................................................................................... 93 Experimental Matrices and Strategy ................................................................. 93 Batch Kinetic Study .......................................................................................... 94 Overall Mass Transfer Rate Modeling .............................................................. 95 Results and Discussions ......................................................................................... 99 Tests in Synthetic Runoff (C0 = 1.0 mg/L) ........................................................ 99 Intra media Diffusion ...................................................................................... 101 Kinetics in RainfallRunoff and Wastewater Matrices ..................................... 103 Summary .............................................................................................................. 106 Nomenclature ....................................................................................................... 108 5 BREAKTHROUGH OF PHOSPHORUS FROM GRANULAR FILTER METERIALS ......................................................................................................... 125 Background ........................................................................................................... 125 Materials and Methods .......................................................................................... 128 Results and Discussions ....................................................................................... 132 Phosphorus Breakthrough Profiles in Synthetic Runoff Matrix for Each Media System ............................................................................................. 132 Desorption Experiments ................................................................................. 134 Phosphorus Breakthrough Profiles in Actual Matrix for AOCMc and AOCMpcc ...................................................................................................... 134 Effect of Flow Rate/Surface Loading Rate on Media Breakthrough Curves ... 136 Correlations between SLR and Thomas Model Parameters .......................... 137 Summary .............................................................................................................. 139 Nomenclature ....................................................................................................... 141 6 CONCLUSIONS ................................................................................................... 155

PAGE 7

7 LIST OF REFERENCES ............................................................................................. 162 BIOGRAPHICAL SKETCH .......................................................................................... 177

PAGE 8

8 LIST OF TABLES Table page 1 1 Water quality standards for the state of Floridas lakes and flowing waters (The final rule signature: Nov. 14, 2010) ............................................................ 20 2 1 Wate r chemistry of experimental matrix ............................................................. 45 2 2 Major composition of raw substrates .................................................................. 46 2 3 Media characteristics .......................................................................................... 47 2 4 Freundlich model parameters in synthetic runoff matrix ..................................... 48 2 5 Freundlich model parameters in rainfall runoff and wastewater matrix ............... 49 2 6 Predicted Saturation Index (SI) for AOCMpcc in different matrices ...................... 50 3 1 Summary of sorption media used in short term isotherm batch experiments to r emove P ........................................................................................................ 80 3 2 Media characteristics and cost ........................................................................... 81 3 3 Freundlich isotherm parameters ......................................................................... 82 3 4 Summary of metals leaching capacity of media.................................................. 83 4 1 Media characteristics ........................................................................................ 111 4 2 Summary of standard method utilized in the experiments ................................ 112 4 3 Water chemistry of experimental matrix ........................................................... 113 4 4 Parameters of 2nd order potential driving model in synthetic runoff (C0 = 1.0 0.05 mg/L) ........................................................................................................ 114 4 5 The sorption kinetics rate parameters in synthetic runoff (C0 = 1.0 0.05 mg/L) with water chemistry summarized in Table 42. ...................................... 115 4 6 The sorption kinetics parameters for AOCM and UCM and water chemistry matrices shown in Table 42. ............................................................................ 116 4 7 Parameters of 2nd order potential driving model in synthetic runoff, rainfall runoff and wastewater with the same C0 of 2.7 mg/L ....................................... 117 5 1 Media characteristics ........................................................................................ 143 5 2 Water chemistry of experimental matrix ........................................................... 144

PAGE 9

9 5 3 Loadresponse parameters for a suite of media utilizing water chemistry matrices shown in Table 52 and an axial reactor ............................................ 145 5 4 Thomas model parameters for AOCM forms utilizing water chemistry matrices shown in Table 52 and goodness of fit indices between modeled and monitored results ....................................................................................... 146 5 5 Relationships between Thomas model parameters and surface loading rate .. 147

PAGE 10

10 LIST OF FIGURES Figure page 2 1 FTIR spectra of media coating. ........................................................................... 51 2 2 Comparison of AOCM and UCM in synthetic runoff matrix. ................................ 52 2 3 Comparison of different media size in synthetic runoff matrix ............................ 53 2 4 Comparison of different AO CM and UCM in synthetic runoff matrix. .................. 54 2 5 Comparison of AOCM in rainfall runoff matrix. ................................................... 55 2 6 Comparison of AOCM in wastewater matrix. ...................................................... 56 2 7 Comparison of AOCMc in synthetic runoff, rainfall runoff and wastewater matrices. ............................................................................................................. 57 2 8 Compar ison of AOCMpcc in synthetic runoff, rainfall runoff and wastewater matrices. ............................................................................................................. 58 2 9 Comparison of AOCMp in synthetic runoff, rainfall runoff and wastewater matrices. ............................................................................................................. 59 3 1 Comparison of P reduction by media of granular filter materials. ....................... 84 3 2 Measured and Freundlich modeled isotherms for P removed by Tire crumb B&G and Expanded shale. ................................................................................. 85 3 3 Measured and Freundlich modeled isotherms for P removed by each granular filter material. ........................................................................................ 86 3 4 Comparison of Ku distribution for media of granular filt er materials. ................... 87 3 5 Comparison of Ku values at Ce = 0.1 mg/L (USEPA, 1986) and Ce = 0.03 mg/L (Genz et al., 2004; Oguz, 2005) for granular filter materials. ..................... 88 4 1 Schematic experimental configuration of kinetic tests on media. ...................... 118 4 2 A) E): Kinetics of P mass transfer with AOCM and UCM in synthetic runoff, C0 = 1.0 mg/L, modeled utilizing a 2nd order potential driving model; F): K inetics of P mass transfer for different AOCM in synthetic runoff, C0 = 1.0 mg/L. ................................................................................................................ 119 4 3 Kinetics of P mass transfer for suite of media tested in a synthetic runoff matrix (C0 = 1.0 mg/L). All kinetics data are modeled using a 2nd order potential driving model; with exception of zeoliteperlite GAC media and tire

PAGE 11

11 crumb (Black & Gold) where net rate of P mass transfer is not signific antly different than 0 (P 0.05) .............................................................................. 120 4 4 Root time plot for P transfer for suite of media tested in a synthetic runoff matrix ( C0 = 1.0 mg/L), with exception of zeoliteperlite GAC media and tire crumb (Black & Gold) where net rate of P mass transfer i s not significantly different than 0 (P 0.05) .............................................................................. 121 4 5 Mass transfer kinetics of P onto AOCM and UCM in a Rainfall runoff matrix with C0 = 2.7 mg/L utilizing a 2nd order potential driving model. ........................ 122 4 7 Mass transfer kinetics for P onto AOCM comparing kinetics between aqueous matrices. Data are modeled utilizing a 2nd order potential driving model, C0 = 2.7 mg/L for all aqueous matrices. ................................................ 124 5 1 Schematic experimental configuration of column breakthrough experiments. .. 148 5 2 Breakthrough curves for AOCM (2 4.75 mm) with an influent pH of 7.0 and C0 of 0.5 mg/L at a surface loading rate of 40 L/minm2. Range bar represent standard deviation between duplicate samples. ............................................... 149 5 3 Breakthrough curves for commercially available media wit h an influent pH of 7.0 and C0 of 0.5 mg/L at a surface loading rate of 40 L/minm2. Range bar represent standard deviation between duplicate samples. ............................... 150 5 4 Phosphorus desorption from media with D.I. water. Range bar represent standard deviation between duplicate samples. ............................................... 151 5 5 A): Breakthrough curves for AOCMc with rainfallrunoff as influent at a surface loading rate of 40 L/minm2 (C0 = 0.5mg/L); B): Breakthrough curves for AOCMpcc with wastewater as influent at a surface loading rate of 40 L/min m2 (C0 = 0.5 mg/L). Range bar represent standard deviation between duplicate samples. ............................................................................................ 152 5 6 Effect of surface loading rate on breakthr ough curves of AOCMc and AOCMp.in synthetic runoff at C0 = 0.5 mg/L. ..................................................... 153 5 7 Correlations of surface loading rates and Thomas model parameters for AOCMc and AOCMp.at C0 = 0.5 mg/L. Range bar represent standard c and AOCMp are significantly different at P = 0.05. ................................................................................................................. 154

PAGE 12

12 Abstract Of Dissertation Presented To The Graduate School Of The University Of Florida In Partial Fulfillment Of The Requirements For The Degree Of Doctor Of Philosophy MASS TRANSFER EQUILIBRIA AND KINET ICS OF GRANULAR FILT ER MATERIALS FOR PHOSPHORUS IN URBAN WATER By Tingting Wu December 2010 Chair: John J. Sansalone Major: Environmental Engineering Sciences This study examined a suite of granular filter substrates as potential adsorptivefiltration media for P control in urban water. T he batch equilibria between dissolved phosphorus (DP) and three engineered AlOx coated media ( soil based, AOCMc; pumice based, AOCMp; Portland cement concretebased, AOCMpcc) were investigated. Media isotherms are evaluated with surrogate matrix in deioni zed (DI) water, municipal wet weather flow (rainfall runoff) and dry weather (wastewater) matrices Compared to the uncoated media substrates, AOCM provides statistically significantly higher equilibra capacity. Wet and dry weather flow matrices had lower DP adsorption compared to the DI matrix due to competitive effects, in particular SO4 2 -. As compared to AOCMc and AOCMp isotherms AOCMpcc exhibited an increasing mass transfer gradient with increasing DP concentration; driven by surface precipitation. A se ries of granular filter materials were also tested in batch equilibrium experiments using the equivalent surrogate matrix of DI water Results were modeled using Freundlich isotherm equation Another unified sorption variable, Ku was calculated and served as a more appropriate parameter for media evaluation and

PAGE 13

13 comparison. At Ce = 0.03 mg/L, Ku values ranged from 25.68 to 0.09 L/g for the media tested in this study, indicating a significant variability of the media capacity. This study also examined the ki netics of DP mass transfer from the aqueous matrix to the media. A 2nd order potential driving model is utilized for the overall rate process Comparing synthetic and actual rainfall runoff matrices demonstrated that actual runoff produced slower overall rate kinetics for AOCMc and AOCMp primarily due to the presence of competitive ions (SO4 2 -) while elevated Ca2+ at the alkaline surface of AOCMpcc offset the effect of SO4 2 -, through surface precipitation. C ommercially available media exhibited a large var iance of capacities Column breakthrough experiments of DP from granular media were also conducted Thomas model was utilized to simulate the breakthrough curves of AOCM T he effect of surface loading rate on breakthrough behavior was investigated and nove l mathematical correlations were developed to relate Thomas model parameters (KT and q0) with surface loading rates Additionally, AOCMc and AOCMpcc were examined in rainfall runoff and wastewater, respectively.

PAGE 14

14 CHAPTER 1 GLOBAL INTRODUCTION Phosphorus (P) has been recognized as a primary nutrient discharged in excess to surface waters and ground water; and such enrichment by phosphorus is regarded as a main reason for eutrophication (U.S. Geological Survey, 1999; Markris et al., 2004; Yoshida et al., 2004; Oleszkiewicz and Barnard, 2006; Endards and Withers, 2007). P can be generated from point sources, for example domestic and industrial wastewater; and nonpoint sources such as agricultural and urban runoff (Namasivayam and Sangeetha, 2004). The built environs as the interface that transforms rainfall into runoff, is a major nonpoint source and pathway of nutrients including P and urban runoff has been noted to contain high phosphorus concentrations that can be accelerating the eutrophication resulting in aesthetic, chemical and ecological degradation of receiving waters (USEPA 1993, Bowes et al. 2005; Kim et al. 2008). While WEF/ASCE (1998) documented that the event mean concentration (EMC) for total phosphorus (TP) 0.33 mg/L and dissolved phosphorus is 0.12 mg/L; it has been reported that EMC of TP in urban runoff ranged from 0.03 to 5.29 mg/L (Davis, 2007) and for highway runoff the median TP ranged from 0.19 to 1.8 mg/L (Drapper et al., 2000). Dean et al. (2005) reported that urban runoff tr ansported DP from 0.3 to 1.4 mg/L (EMC) in an urban pavement catchment. A desired goal for prevention of plant nuisances in flowing waters not discharging directly to lakes or impoundments is 0.1 mg/L (USEPA, 1986). (2005) reported that TP concentrati on over 0.03 mg/L will likely trigger algal blooms in confined water bodies. Especially, EPA recently proposed new numeric nutrients criteria for Flo ridas lakes and flowing waters, which are summarized in Table 11 the specific values for different water bodies.

PAGE 15

15 However, while the technology exists and is relatively mature for treat ing point sources, control of nonpoint sources has not kept pace with point source control, in part because of the much greater challenges associated with diffuse, stochastic and variable loadings and with technology that is still developing. Many physical, chemical and biological control technologies applied to nonpoint loads can be disadvantaged by unfavorable cost/benefit relationships, low or variable effectiveness, or res trained by operational instability (for example desorption from media systems or lack of maintenance) and lack of sustainability (Karaca, 2004; Sansalone, 2005; Huang et al., 2008). P occurs in the forms of both dissolved phosphorus (DP) and particulate phosphorus (PP) and mainly exists as phosphate in aquasystems (Wei et al., 2008; Chitrakar et al., 2006). Regarding the control of P loadings in urban rainfall runoff, many phosphorus treatment methods have been investigated, such as constructed wetland system s (Gervin and Brix 2001; Seo et al. 2005), filtration, infiltration and detention/retention basins (Dechesne et al. 2005; Hunt et al. 2008; Hsieh et al, 2007), in situ filtration (Sansalone and Teng 2005, Teng and Sansalone 2004), storm treat systems (Sonstrom et al. 2002), urban wet detention ponds (Wu et al. 1996; Comings et al. 2000; Wang et al. 2004), and vegetative controls (Barrett et al. 1998). While sedimentation occurs in all unit operations, whether by design or otherwise, designs based stri ctly on sedimentation through overflow rate concepts are limited in populated urban area since there is generally a requirement of large operational area and volume, if suspended particulate matter (PM) requires separation. Additionally, sedimentation is

PAGE 16

16 a mechanism to separate PM, not soluble phosphorus, which, unfortunately, is very biologically active, and imparts the most negative influence on our surface waters. For over a decade adsorptivefiltration has attracted interest; for example through deploym ent of economical filter substrates for DP adsorption. A category of the adsorptive media incarnations have included filter substrates that range from common nonengineered materials (Sakadevan and Bavor, 1998; Johansson, 1999; Bubba et al., 2003; Hsieh an d Davis, 2005; Davis et al., 2006; Hsieh et al., 2007) to industrial by products or wastes (Kostura et al., 2005; Drizo et al., 2006; Lu et al., 2008) in order to reduce the initial cost of media systems. These studies have demonstrated that many of these media do not have sufficient capacity for successful deployment to control DP (Bubba et al. 2003; Forbes et al. 2004, Kim et al. 2008, Sansalone and Ma 2009). Recently engineered media with Fe or Al coating has been reported to exhibit significantly impro ved adsorptive capacity as compared to nonengineered media (Ayoub et al., 2001; Arias et al., 2006; Boujelben et al., 2008; Sansalone and Ma 2009). However, while these recent studies have pointed to the role of metal oxides for phosphate adsorption, Liu et al 2005 (a, b) reported that media function and deployment is also influenced by the properties of the media substrate itself such as porosity, pore size distribution, strength and substrate chemistry. To date, research has focused on comparisons of ec onomical substrates such as reuse or waste substrates (Agyei et al., compared to substrate combined with an oxide coating for adsorption of phosphate. Additionally, if med ia and media substrates require comparison, testing that allows comparison between media for a given media phenomena under equivalent conditions

PAGE 17

17 are required. The most common controlled surrogate aqueous matrix is deionized (D.I.) water which provides a c onsistent and reproducible matrix for media testing and facilitating media comparisons subject to different controlled loadings (Namasivayam and Sangeetha, 2004; Arias et al., 2006; Georgantas and Grigoopoulou, 2007; Boujelben et al., 2008). However, whil e such a consistent and reproducible matrix provides a vehicle for media comparability, studies have also shown that the equilibrium capacity of DP can be nonrepresentative when comparing results from a DI matrix and an actual aqueous matrix subject to si milar pH and DP concentrations (Millero et al., 2001; Chen et al., 2002; Shilton et al., 2005; dm et al., 2007, Sansalone and Ma 2009). To date little quantitative evaluations of adsorptivefilter media for DP adsorption have been conducted, especially fixed bed/column tests. However, there continues to be limited information available and large performance claims for phosphorus removal for both manufactured and natural filtration systems using unquantified media. Media applications that do exist with reported information tend to experience high performance variability (Davis et al. 2006, Hsieh et al. 2007) or result in exporting phosphorus, or leaching of other pollutants such as metals or pH. Again, these challenges are due to the limited quantitative information of the behavior of the filter material. Successful deployment of any media system requires the comparison of media behavior on a rigorous and equal basis. In this study, three Al oxide coated substrates: clayey soil based (AOCMc, 2 gradations) pumice based (AOCMp), and Portland cement concretebased (AOCMpcc, 2 gradations), along with their uncoated media substrates (UCM) served as controls, and a suite of nonengineered commercially available media,

PAGE 18

18 including activated alumina, bioretention soil media, expanded shale (2 gradations), iron coated perlite, tire crumb B&G (2 gradations) and Zeoliteperlite GAC are evaluated through three quantitative descriptions: equilibrium isotherms, kinetics and column breakthrough using controlled surrogate aqueous matrix of deionized (D.I.) water. Additionally, metal leaching and desorption tests are carried out to determine if metals leach out from the media materials during adsorption tests or the immobilized P can be easily sloughed off after adsorption. Furthermore, AOCM are examined utilizing aqueous matrices of synthetic runoff, actual rainfall runoff and wastewater under three hypotheses: First, a mediabased AlOx reactant, as a coating or admixture can significantly increase the capacity of a given s ubstrate as compared to the substrate alone. Second, the physical and chemical indices of the substrate impact the capability of the resulting media with AlOx. Third, the chemistry of an actual runoff or wastewater matrix with competing solutes significan tly affects the overall P mass transfer rates and media capacity. Moreover, the application of batch experimental data is often difficult to apply to fixed bed adsor ption system and inferences from such data may lead to false direction without proper interpretation. Beyond equilibria, media capacity for a solute can be represented through breakthrough modeling under a known set of loading conditions (Kse and ztrk, 2008). The time (or more appropriately volume treated) until a set breakthrough or exhaust ion level are commonly reported characteristics of a given media, system and set of loading conditions; while there has been more limited analytical model development to describe breakthrough in time and throughout the media profile (Aksu et al., 2007) unl ess carried out with an advectivedispersive

PAGE 19

19 numerical model (Liu et al. 2005). However, introduction of lumped system parameters and constrained correlations have provided tractable analytical models for onedimensional flow systems (Lee et al., 2000; Ko et al., 2001; Aksu et al., 2007). One such bed depth service time (BDST) model is the Thomas model that has been widely used in the design of fixed bed adsorption systems (Rao and Viraraghavan, 2002; Thirunavukkarasu et al., 2003; Aksu and Gnen, 2004; Ak su et al., 2007; Kse and ztrk, 2008; Pokhrel and Viraraghavan, 2008; Sivakumar and Palanisamy, 2009). The main advantages of the Thomas model are simplicity and reasonable accuracy in predicting media breakthrough behavior for adsorptive media. Additionally the kinetics coefficient and adsorption capacity from the model can be determined from monitored data. Typically examined for a single SLR, examination across a range of SLRs can extend the use of the Thomas model to incorporate behavior subject to unsteady phenomena, for example for a fixed bed media absorber utilized after an equalizing or primary treatment operation of runoff (Erickson et al., 2007). While there have been many studies regarding the use of such models (Koh et al., 1998; Aksu and Gnen, 2004; Mantovaneli et al., 2004; Aksu et al., 2007; Kse and ztrk, 2008), illustrating the relationship between these Thomas model parameters and SLR has not been demonstrated through a physical model. Given the variability in actual application, esp ecially the stochastic loading of rainfall runoff, such relations will be of great practical significance and immediate benefits.

PAGE 20

20 Table 11. Water quality standards for the state of Floridas lakes and flowing w aters (The final rule signature: Nov. 14, 2010) Lake c lass TP [mg/L] Colored l akes > 40 PCU 0.050 0.157 Clear lakes, a lkaline 3 0.030 0.087 Clear lakes, a cidic 3 0.010 0.030 Watershed Regions TP [ mg/L] Panhandle 0.043 Bone v alley 0.739 Peninsula 0.107 North c entral 0.359 Waterway TP [mg/L] Canals 0.042

PAGE 21

21 CHAPTER 2 BATCH EQUILIBRIA BETWEEN AQUEOUS PHOSPHORUS SPECIES AND ALUMINUM OXIDE COATED GRANULAR SUBSTRATES Background Phosphorus (P) can be generated from point sources, for example domestic and municipal wastewater; and nonpoint sources such as agricultural and urban runoff (Namasivayam and Sangeetha, 2004). The build environs of urban land uses represent an interface that transforms rainfall into runoff, w hile mobilizing and transporting nutrients including P. Urban runoff has been shown to contain high P concentrations, resulting in an adverse impact on receiving water bodies (Dean et al. 2005; Kim et al., 2008). As a primary limiting nutrient in surface w aters a range of limits for discharge of have been identified; for example 0.03 mg/L of P was suggested as a criteria level with advances in treatment of point source P, for example at a municipal wastewater treatment plant (WWTP) is economical and efficient compared to treatment of diffuse and stochastic loadings of nonpoint P (Shin and Han, 2004, Duda, 1993; UEPA 2003; Sansalone 2005). For over a decade adsorptivefiltratio n has attracted interest; for example through deployment of economical filter substrates for phosphate adsorption (Gray et al., 2000; Liu and Yang, 2002; Mortula et al., 2007, Sansalone and Ma 2009). There is also a significant body of research that has d emonstrated that P in runoff can be adsorbed by soils (Edzwald et al., 1976; Wang and Tzou, 1995; Fontes and Weed, 1996). In part, this adsorption can be attributed to Al and Fe oxides. For example, Al oxides can range from 9.4 to 12.3 g/kg; while Fe oxides can range from 10.8 to 23.3 g/kg in silt loam and sandy loam soil (Morillo et al., 2004). In many soils Al and Feoxides are present

PAGE 22

22 making soil a reactive substrate, specifically for phosphate species (Fontes and Weed, 1996; Nooney et al., 1998; Arias et al., 2006; Boujelben et al., 2008). Parfitt (1978) indicated that phosphate, within a broad concentration range was adsorbed on the oxide surfaces of clayey soils as binuclear bridging complexes by replacing two singly coordinated hydroxyl groups on the oxide surface per molecule of phosphate. However, van der Zee and van Riemsdijk (1988) suggested that phosphate adsorption consisted of two distinct mechanisms: one relatively fast, reversible adsorption process and a relatively slow, largely irreversibl e precipitation mechanism; for example in soils classified texturally as sandy. Goldberg and Sposito (1985) reported that the mechanism for phosphate adsorption is ligand exchange by hydrous Al/Fe oxides of soil minerals. More recently, metal oxides have been investigated to improve phosphate adsorption (Ayoub et al., 2001; Arias et al., 2006; Boujelben et al., 2008). However, while these recent studies have pointed to the role of metal oxides for phosphate adsorption, Liu et al 2005 (a, b) reported that media function and deployment is also influenced by the properties of the media substrate itself such as porosity, pore size distribution, strength and substrate chemistry. To date, research has focused on comparisons of economical substrates such as reus e or waste substrates (Agyei et al., compared to substrate combined with an oxide coating for adsorption of phosphate. Additionally, in many of these studies a controlled surrogate aqueous matrix such as the common use of deionized water has provided a consistent and reproducible matrix for testing media and facilitating media comparison (Fontes and Weed, 1996; Namasivayam and Sangeetha, 2004; Arias et al., 2006; Georgantas and Grigoopoulou,

PAGE 23

23 2007; Boujelben et al., 2008). However; actual rainfall runoff or wastewater matrices are much more complex with competitive interactions; and rainfall runoff is highly variable spatially and temporally (Dean et al. 2005, Sansalone et al. 2005). As a result, studies have either utilized a reproducible surrogate aqueous matrix or separately, an actual environmental matrix without compari son to the surrogate matrix Objectives This study focuses on phosphate (as DP) adsorption equilibria by AlOx as a mediabased reactant. Comparisons are made as function of media substrates, and between aqueous matrices of (DI water as a surrogate), wastewater and rainf allrunoff at equal DP concentrations. This study is driven by the following hypothes es. First, a mediabased AlOx reactant, as a coating or admixture can significantly increase the adsorption equilibria of a given substrate as compared to the substrate alone. Second, the physical and chemical indices of the substrate impact the adsorption equilibria of the resulting media with AlOx. Third, at the same pH, redox and DP concentration, the chemistry of a actual runoff or wastewater matrix with competing solutes with result in a statistically significant decrease in adsorption equilibria for a given media. Materials and Methodology Sorbents and Media Preparation Clays, pumice and Portland cement concrete are all reported to be effective for removing phosphate ions from water (Fontes and Weed, 1996; Agyei et al., 2002; Ayari et al., 2005; Akb al 2005; Liu et al., 2005). In this study these three materials were used to prepare raw substrates and then coated with Aluminum Oxide. In addition, the raw substrates were tested as control in the experiments. The effective size of the sorbents ranged fr om 0.85 4.75 mm for clay and concrete; 2 4.75 mm for pumice.

PAGE 24

24 Raw substrate (UCM) preparation UCMc: Bentonite with mineralogical composition of sodium hydrated aluminosilicate was purchased from Halliburton Energy Services, Inc. The clay was first mixe d with explosive reagents and water, and then sintered in kiln to remove the moisture and be solidized, following by crushing and sieving to recover the desired gradation. This porous substrate was then soaked in an acid solution (H2SO4, 10 15%) for 24 h to removal the phosphorus residual. After that, the raw substrate (UCMc) was rinsed with D.I. water until the pH was higher than 4.5 and then dried at 105 oC. UCMp: The pumice (2 4.75 mm) was purchased as commercials from Hess Pumice Products. The same acid soaking pretreatment was applied to pumice to produce pumice raw substrate (UCMp). UCMpcc: The concrete raw substrate was produced using a mass ratio 1:1 (water to cement). This cementitious slurry was then cured and size graded. Aluminum oxide c oated media (AOCM) preparation AOCMc and AOCMp: AOCMc and AOCMp were prepared by applying aluminum as surface coating. Stock solution of 1M Al (III) was prepared by dissolving Al(NO3)3 9H2O in De ionized (D.I.) water. Al (III) solution was added to the raw substrate contained in a Pyrex tray so that the substrate was just submerged. This Pyrex tray was then put in the oven and temperature was kept at around 200 oC. At least 24 hours were needed to fully dry and coat the substrate. Later, the aluminum coated sorbents was washed with D.I. water to remove the surficial fine materials. Then the sorbent was dried at 105 oC and finally stored in polypropylene bottles for further use. AOCMpcc: AOCMpcc was prepared by integrating aluminum solution into the cementitious matrix as the aqueous admixture. The mass ratio of aqueous Al(NO3)3

PAGE 25

25 solution (0.5 M) to cement is 1:1. This cementitious slurry was then cured and size graded in the same way as raw substrate preparation. Chemicals A phosphorus stock solution was prepared by dissolving anhydrous potassium dihydrogen orthophosphate (KH2PO4) powder (Analytical reagent grade, Fisher Scientific) into D.I. water. The stock s olution was calibrated to 100 mg/L and 1000 mg/L concentrations, which were utilized as standard solutions as necessary. Phosphorus concentration was measured by HACH DR/5000 Spectrophotometer using PhosVer 3 Ascorbic Acid Method (Standard Method 1998). A ll measurements were duplicated and high repeatability was assured by controlling experimental error within 5%. Measurements of pH were carried out using a multi meter (Orion 290A combination electrode), which was preliminary calibrated using a 3point calibration curve with reference units at a pH of 4, 7, and 10. Surface C s) Surface charge for the tested media was measured using a potentiometric titration methodology described by Van Raij and Peech (1972). pHpzc was then c alculated by linearly interpolating between the last positive and first negative surface charge s) measurement utilized a helium gas pycnometer. Powder X ray Diffraction (XRD) XRD analysis was conducted on oxide coating materials and raw clay substrate. The instrument used was XRD Philips APD 3720. Cu Kwas employed, with a tube current of 20 mA and a tube voltage of 40 kV. Scans were conducted from 10 to 120

PAGE 26

26 Fourier Transform Infrared Spe ctrometry (FTIR) FTIR measurement was performed with a Nicolet Magna IR 760 system. Coating materials were removed from the coated media surface for FTIR analysis. Reynolds Alpha Alumina and Inframat Gamma Alumina were also analyzed as references. 3.5 milligrams of the prepared material were pressed with KBr (1:100) to a pellet. The measurements were carried out in the midinfrared range from 4000 to 400 cm1. Experimental Matrices and Adsorption Equilibria (Isotherm) Study In order to estimate static or dynamic adsorption capacity, it is essential to have information on adsorption equilibrium to quantify how much phosphorous can be adsorbed. A common way to represent the adsorption equilibrium is the equilibrium isotherm fro m which the amount of phosphorus species adsorbed under a given set of chemistry conditions can be determined for a given media. Comparison and evaluation of the engineered media as well as the raw substrates were carried out in three different experiment al matrices, including synthetic runoff, filtered rainfallrunoff and wastewater matrix. The chemistry of experimental matrices was summarized in Table 2 1. In present study, the specific working phosphorus concentrations utilized to obtain isotherm data w ere: 0, 0.05, 0.5, 1.0, 2.5, 10, 25 and 50 mg/L. For the synthetic runoff matrix, the above initial TDP concentrations were obtained by diluting stock phosphorus solution and ionic strength was fixed at 0.01 M using KCl. Solution pH was adjusted to 7 at the beginning of experiments by dropwise addition of HCl or NaOH. For rainfallrunoff and wastewater matrix, the initial concentration of the rainfall runoff or wastewater was kept to replace one of the eight working phosphorus concentrations, whichever was the closest. Then the others were obtained by diluting the raw sample with D.I. water or adding high concentration

PAGE 27

27 phosphorus stock solution, in such a way that the original chemistry of the matrix was best maintained. In addition, the real rainfall runoff sample was filtered through 0.7 m paper filter to remove the coarse solids in the feeding solution before dilution or phosphatespike. 0.5 g adsorbent and 40 ml of working phosphate solution were added into a 50 ml polyethylene centrifuge tube. The mixture was thoroughly mixed on a horizontal bench shaker at a rate of 100 rpm allowing adsorption to take place at 20C until equilibrium reached, usually 24 hr. The content in the centrifuge tube was then filtered through 0.45m syringe filter and the total dissolved phosphate concentration in the filtrate was measured using spectrophotometer (HACH, DR5000). The amount of adsorbate that was adsorbed on the adsorbent was then calculated from the difference between the initial and equilibrium solute concentrations. Equilibria Model E quilibria represent one of four fundamental categories of adsorption models; models that quantitatively describe media or soil behavior (Liu et al. 2005b). Although an equilibrium isotherm does not account for temporal mass transfer effects of most transport phenomena; the operating curve of an equilibrium isotherm model represents an asymptotic limit as a function of concentration for adsorption capacity of a given soil or media system At each unique set of aqueous solid adsorbent phase conditions there is a unique media or soil capacity qe [mg/g] W V C C qe e) (0 (1) C0 and Ce are the initial and equilibrium solut e concentrations [mg/L], respectively; V the volume of the solution [L] and W the mass of the adsorbent [g ]. The relationship

PAGE 28

28 between solute adsorbed, q, as a function of the solute concentration in the fluid phase, C, at temperature T is called the adsorp tion isotherm at T. A singular point for a unique set of aqueous solid phase conditions cannot represent an isotherm (Sansalone and Ma 2009). ) ( c q q at T (2) An isotherm model can be represented by a number of mathematical forms that can be combined kinetics and breakthrough representations (Suzuki, 1990). A common isotherm is the Langmuir model. bC C bq qm 1 (3) q is the solute adsorbed per unit mass of adsorbent and qm is the maximum adsorbed phase concentration where all the adsorption sites are occupied. As ident ified by Tien,( 1994) the assumptions include (1) the solute molecules are adsorbed at localized states ; (2) e ach adsorption site accommodates one solute molecule and all the sites are energetically identical ; and (3) there are no lateral interactions betwe en sites These assumptions result in an idealized adsorbent surface; in contrast to the heterogeneity of actual surfaces. Even across the surface of a perfect crystal, Everett ( 1993) indicated that the energy of interaction with a solute varies across the surface and such heterogeneity cannot be neglected However, the surface heterogeneity can be modeled through an adsorption energy distribution, and m athematically, a heterogeneous surface may be considered to be composed of many homogeneous sites (Ti en, 1994; Rocha et al., 1997). Therefore, a heterogeneous adsorption surface can be considered as a summation of simplified homogenous Langmuir surfaces.

PAGE 29

29 i i i m m mC b q C C b C q b C b C q b q 1 ) ( 1 12 2 2 1 1 1 (4) qmi and bi are values of qm and b for the ith site. With ib b 1 eq (4) can be rewritten. i i mC b q C q ) ( (5) Each adsorption site is characterized by specific adsorption energy. By assuming a continuous distribution of the site energy, equation (5) can be written as: 0) ( b d b N C b C q (6) The site frequency distribution N(b') discretizes sites having adsorption energies between b' and b' + db'. T ien (1994) modeled the distribution of N(b') ; with A as a constant. 1 1) ( ) ( pb A b N (7) Substitute equation (7) with equation (6) and simplying : 0 1 1 1 1 01 ) ( ) ( b d C b b A b d C b b C A qp p (8) When p > 1, solution of the above equation leads to the Freundlich model, where ] ) 1 1 sin[( ) 1 (1p C A qp (9)

PAGE 30

30 Assuming that 1 1 1 p equation (9) can be approximated as: p pAC C A q1 1sin ) 1 ( ( sin A A ) (10) This has the form of the non linear Freundlich isotherm equation, commonly used to model heterogeneous adsorbent surface s. The heterogeneous composition and structure of the adsorbent substances and the potential of multiple sorption mechanism contributes to nonlinear isotherms. The commonly utilized form of Freundlich, as used herein is: n FC K q (11) KF and n are constants incorporating all parameters affecting the adsorption process, i.e. adsorption capacity and intensity respectively. Ce is t he concentration of solute in solution at equilibrium [mg/L]. The Freundlich model and experimental data were fit using nonlinear regression analysis. Results and discussion Physicochemical Characteristics of Sorbents The major components of pumice are am orphous silicon dioxide (SiO2) bo nded m etal oxides such as Al, Ca and Mg etc as reported by Panuccio et al. (2009). The concrete substrate of Type II cement, with major components of C S H (Calcium Silicate Hydrate, 50 60%), Ca(OH)2 (20 25%)and CSA ( ca lcium sulfoaluminates, 1520%). XRD analysis was carried out to determine the mineral composition of raw quartz (SiO2), Goethite (FeOOH) and Calcite

PAGE 31

31 (CaCO3) were identified. The major composition of the three substrates is summarized in Table 2 2. FTIR combined with XRD was used to analyze the composition and structure of media coating. No aluminum related crystals were found during the XRD analysis, which may lead to the conclusion that the coating was amorphous not crystalline. Figure 2 1 illustrates the FTIR spectrum of the media coating The broad band between 2800 ~ 3700 cm1 is associated with OH stretch (Klute, 1986; Liu, 1999; Chandrasekharan et al., 2008). More accurately, 3700 cm1 corresponds to free O H stretch, 3675 ~ 3540 cm1 to OH stretch and 3390 ~ 2800 cm1 to bonded O H stretch (Klute, 1986). Bands below 1000 cm1 represent Al O bonds (Costa et al., 2007). The peak at 1384 cm1 alumina is associated with nitrate bond (Smidt E et al., 2005; Garai J. et al., 2006), which originates from the precursor during aluminum oxide preparation. The transformation between different Al2O3 phases strongly depends on the precursors and the heat treatment. Usually the spectrum of l ow temperature phase alumina shows a broader band of OH group. With increase of treatment temperature, the bands tend to decrease the intensity due to the removal of water molecules. In the mean time, the intensity corresponding to Al O bond increases. Th ese phenomena were observed, as compared the three spectra shown in Figure 2 1. Since the media coating in this study was prepared with Al(NO)3 and only calcined at 200 C, broader band of OH stretch and lower Al O intensity were identified in FTIR test, and also the nitrate coming from the Al(NO)3 cannot be eliminated. Table 2 3 summarizes the physical characteristics of the sorbents. Surface area has been regarded as an important indicator of the media capacity in some

PAGE 32

32 circumstances (Colombo et al., 1993; Liu et al., 2001; Arias et al., 2006). Compared with UCM, the SSAs of both AOCMc and AOCMpcc are higher, while it is opposite for AOCMp and UCMp. This may be due to the highly porous nature of pumice substrate, which can be blocked or filled during the coating process. In contrast, for clay and concrete based media, aluminum oxide contributes to the surface roughness and results in a higher SSA. The literature has reported a wide range of specific surface area of aluminum oxide, from < 15 m2/g to > 200 m2/g (Goldberg et al., 2001; Genz et al., 2004). Surface area was found to be affected by various factors, such as concentration of the starting reagents, aging, heating and reaction in aqueous solution. Hietela et al. (1989, 1990) interpreted the increas e in heating temperature can result in the increase in surface area of low surface material. In their study of surface characterization of amorphous aluminum oxides, Goldberg and her coworkers (2001) found aluminum oxide minerals of widely differing initial surface areas measured in the dry state had the surface areas of comparable magnitude upon reaction in aqueous. Therefore, initial surface area may be a conditional indicator of chemical reactivity of synthetic amorphous aluminum oxide. For all of the t hree substrates based media (clay, pumice and concrete), there is not much difference between coated media and corresponding raw substrates in pHpzc. In addition, pHpzc of activated alumina has been reported to be 8.1 ~ 9.1 (Genz et al., 2004), which is much higher than 4 ~ 5 as shown by AOCMc and AOCMp. This implies that the electrical properties of the underlying material was not masked by the surface aluminum coating, which is also exemplified by the coating efficiency of AOCMc (33.13 18.32 mg aluminum coating /m2 UCMc) and AOCMp (7.16 0.87 mg aluminum coating

PAGE 33

33 /m2 UCMp). Unlike clay and pumice based AOCM, concrete based media has a much higher pHpzc. The high pH value of point of zero charge of AOCMpcc and UCMpcc is due to the presence of large amount of CaO/Ca(OH)2 in the substrate, which can definitely overwhelm the aluminum coating especially when aluminum was integrated into the cementitious matrix and not applied as a surface coating. Studies in Synthetic Runoff Matrix (Hypothesis (1) and (2)) Hypothesis (1) Comparison of the performance of AOCM and UCM in synthetic runoff matrix is shown in Figure 2 2. The isotherms rise sharply in the initial stages due to plenty of readily accessible sites and eventually reach a plateau indicating that the adsorbent is saturated at this level. In all three cases (clay based, pumice based and cement concrete based), the coating media showed much better capacity. The Freudlich isotherm was used to describe the sorption equilibria. The model parameters are summarized in Table 2 4. All R2s are no less than 0.96, indicating a good simulation of experimental data. The values of KF and n determine the steepness and curvature of the isotherm. Usually n values between 0 and 1 represent benefici al adsorption and indicate that significant adsorption takes place at low concentration but the increase in the amount adsorbed with concentration becomes less significant at higher concentration. Also, the higher the KF value, the greater the adsorption intensity. Therefore, the higher KF values for the coated media confirm that the adsorption capacity of AOCM was greater than that of UCM. Since for each pair of media, AOCM and UCM were in the same gradation and both experiments were carried out under th e same conditions, this difference can be reasonably ascribed to aluminum coating. Two gradations (0.85 2 mm) and (2 4.75 mm) were tested for AOCMc/UCMc and

PAGE 34

34 AOCMpcc/UCMpcc in this study. Figure 2 3 depicts the media size effects in synthetic runoff mat rix. Not surprisingly, small media size exhibited greater sorption capacity. Since the adsorption experiments were carried out with the same amount of adsorbents (0.5 g), this phenomenon can be easily related to the specific surface area. Although in the earlier discussion the initial SSA of amorphous alumina as a reliable indicator of potential reactivity was questioned, that was stated for the absolute SSA value of one media. Here, SSA of two adsorbents is compared and the trend that higher SSA better rea ctivity still stays true. The SSA ratio of AOCMc (0.85 2 mm) to AOCMc (2 4.75 mm) is about 8 and that of AOCMpcc (0.85 2 mm) to AOCMpcc (2 4.75 mm) is about 9. Larger SSA means more active sites could be available for phosphate to attach with and in return higher removal efficiency will be achieved. However, although there was notable increase in percentage of phosphate removal as the adsorbent particle size decreased, the increase was not exactly proportionate to the increased surface area. This indicates the phosphate is removed from the liquid phase mainly by chemisorptions rather than physical adsorption, especially at high P concentration. The higher removal efficiency of coated media comes from surface reactions between the media coating and aqueous phosphate. Although there are rare direct evidences, it is generally accepted that phosphates are adsorbed on hydrous aluminous oxide through ligand exchange between phosphates and the surface hydroxyl groups (Goldberg and Sposito, 1985; Genz et al. 2004; Kim and Kirkpatrick, 2004; Arias et al. 2006; Yang et al., 2006; Georgants and Grigoropoulou, 2007). When anions have a specific affinity for the metal atoms in the hydrated oxide, ligand exchange can result in formation of surface complexes contai ning no water molecules between the surface

PAGE 35

35 Lewis acid site and the adsorbed ion, which are referred to as inner sphere. Huang (1975) clearly demonstrated the significance of specific chemical adsorption by calculating free energy for the adsorption of pho Al2O3. A generalized ligand exchange reaction for phosphate ions was summarized by Goldberg and Sposito (1985): aSOH (s) + HbPO4 b 3 (aq) + cH+ (aq) aHcPO4 (s) + bH2O (l) + (a b)OH(aq) Where S refers to a metal ion, OH to a reactive surface hydroxyl and b degree of protonation of the phosphate ion. In this study metal ion is Al and at pH of 7, the majority of phosphate ions exist as the form of H2PO4 Therefore, the reaction can be rewritten as: 2 OH + H2PO4 2HPO4 + H2O + OHIn spite of controversy arising from the problem of what stereochemical configuration a phosphate takes on when it is complexed by a surface Lewis acid site due to the difficulty of probing the structure of the phosphate surface complex in an aqueous environment (Rajan, 1975; Guan et al., 2005), an increase of pH usually accompanies the adsorption reaction, as demonstrated in the above equation (Chang et al., 1997; Kitis et al., 2007). In this study, it was noticed that pH decreased as the wor king solution got contacted with AOCMc or AOCMp and then increased after certain time of reaction. The initial drop of pH may be due to the redundant H+ from pretreatment of acid soaking. Then following pH increase can be ascribed to the surface reaction, which released OHas a result. For AOCMpcc, an enhancement of solution pH was always observed and this should be correlated with the dominant influence of alkalinity caused by the concrete media. However, although the ligand exchange is

PAGE 36

36 favored as a major mechanism of phosphatemetal oxide adsorption, many researchers have noted that the whole reaction is a complicated process and more than one mechanism could contribute to the phosphate removal (Johnson et al., 2002; Kim and Kirpatrick, 2004). Additional ly, Arias et al. (2006) and Yang et al. (2006) pointed out that phosphorus sorption is a continuous sequence of adsorption and precipitation and is very difficult to make a distinctive quantitative analysis of the amount of adsorption and chemical copreci pitation. The role of each mechanism largely depends on the chemical nature of media and working solution, as discussed later. Hypothesis (2) Figure 2 4 compares the isotherm profiles of three AOCMs as well as UCMs in synthetic runoff matrix. In order to exclude particle size effect, all the results were obtained from the same gradation (2 4.75 mm). For UCMs, it can be easily noticed that UCMc and UCMp almost performed identically while there was a significant difference between those two and UCMpcc. It i s believed that the capacity of UCMc and UCMp for phosphorus removal from aqueous phase is usually related with the Al or Fe content in the substrate, which is consistent with the literature (Fontes and Weed, 1996; Nooney et al., 1998; Arias et al., 2006; Boujelben et al., 2008). For UCMpcc, positive charge on the adsorbent surface (at pH 7) increased its affinity for the negatively charged phosphate ions and the phosphate was removed from solution via deprotonation and subsequent precipitation as Ca3(PO4)2. Due to the high content of Calcium, UCMpcc exhibited a much higher removal capacity compared with UCMc and UCMp.

PAGE 37

37 The isotherm profile of AOCMs in synthetic runoff matrix is very instructive especially in contrast with that of UCMs. Unlike the uncoated c ounterparts, AOCMc and AOCMpcc showed similar higher capacity, while AOCMps efficiency was relatively low. On the contrary, the SSA of AOCMp (2 4.75 mm) is the highest among the three of them, which should apparently promise a better performance. One reason for this observation is the physical nature of pumice. Pumice is a light, porous igneous volcanic rock. In spite of the highly porous structure of AOCMp, quite percent of the porosity may come from micro pores which could decrease penetrability and hi nder the ingress of phosphate ions to active sites. As discussed earlier, aluminum coating on the raw substrates resulted in a higher capacity and for AOCMpcc surface precipitation can also be an important mechanism. Therefore, investigation of the different performance between AOCMp and AOCMc logically lead to examination of coating efficiency. As reported earlier, less than aluminum coating (in mass) was contained on AOCMp compared with AOCMc, which can be the direct cause for the low capacity of AOCMp. Even though exactly same methods were employed for AOCMc and AOCMp, the coating efficiency can be varied due to different physical and chemical nature of the raw substrates. For example, it is very possible that not all of the aluminum on AOCMp serves as a specific surface functional group since part of them may just penetrate and fill into the pumice pores. This speculation is supported by the surface area change of pumice media after aluminum coating, as shown in 2 3. The raw substrate surface chemistry can also impact the interactions among surfaces and aluminum oxide during binding. Bentonite has a 2:1 layer clay mineral with two silica tetrahedral sheets bonded to a central alumina octahedral sheet The isomorphic substitution of Al3+ with Fe2+ and

PAGE 38

38 Mg2 + and Si4+ with Al3+ results in net negative charge, which is balanced by the cations such as Na+ and Ca2+ located between the layers and surrounding edges ((nal and bentonite) with Na+ being the exchangeable cation. During the pretreatment of acid soaking, quite amount of the alkali metal ions (Na+) contained in the clay are exchanged for the hydrogen and create Brnsted acid centres. Furthermore, these H+ protons can migrate, disrupt t he bonds between Si, O and Al, and lead to the formation of new groups (Christidis et al., 1997). These new groups may be bound in a less cohesive manner and thus leave a lower energy barrier for the interaction of aluminum coating with the substrate surface, resulting in a higher coating efficiency. On the other hand, pumice is mainly formed by SiO2 4 tetrahedral networks, where approximately one in ten Si atoms is randomly replaced by Al atoms, maintaining the same coordination (Floriano, 1994). The subst itution of Al3+ for Si4+ also involves the addition of monovalent cations, usually sodium and potassium to balance the net charge. It can be assumed that the chargebalanced cations in pumice is mainly located within the network and not as easily and large ly exchanged with H+ as for bentonite, thus the coating efficiency for pumice substrate is lower. Studies in Rainfall runoff and Wastewater Matrices (Hypothesis (3)) As has been proved a better performance in synthetic runoff matrix, AOCM was further test ed in rainfall runoff and wastewater (after advanced treatment) matrices. The results are illustrated in Figure 2 5 and 2 6. The model parameters were summarized in Table 2 5. Figures 2 7 to 2 9 elucidate effect of matrices on the performance of AOCMc, AO CMpcc and AOCMp respectively.

PAGE 39

39 Although capacity of the three AOCMs followed the same order as in synthetic runoff matrix (Figure 2 4), much more significant difference was observed between AOCMpcc and the other two media. Meanwhile, less divergence between the isotherm profiles was shown between AOCMc and AOCMp. In fact, it can be seen from Figure 2 5 and Figure 2 6 that the isotherm curvature for AOCMpcc has shifted from convex to concave as the matrix changed from synthetic runoff to rainfall runoff/waste water matrix. This noticeable change along with more than 99% phosphate removal efficiency even under the condition of 50 mg/L initial concentration clearly indicates an inherent mechanism transit. On the other hand, the anomalies in the results can also be perceived through Freundlich model parameter evaluation. In Table 2 5, abnormal large KF are reported and the values of n are greater than 1 which has actually violated the monolayer surface coverage assumption (n < 1). In this case, Freundlich model c an only be regarded as an empirical model mathematically simulating the experimental results, but not an adsorption model with the key parameter of the maximum surface coverage. All these evidences have shown that the major mechanism of phosphorus removal for AOCMpcc transited from adsorption to surface precipitation as the working matrix changed. Agyei et al. (2002) studied the removal of phosphate ions from aqueous solution by fly ash, slag, ordinary Portland cement and found that removal efficiency of PO4 3 increased in such an order that mimicking the order of increasing percent CaO in the adsorbents, clearly indicating the affinity of phosphate for CaO. It has been a consensus that the formation of insoluble calcium phosphate and consequently the CaP p recipitation is a significant mechanism for P removal using furnace slag or cement (Agyei et al., 2002; Lu et al. 2008).

PAGE 40

40 In this study, the water chemistry of rainfall runoff and wastewater has promoted the precipitation of phosphate ion greatly and changed surface precipitation to the dominant mechanism. This transition should be attributed to the high content of CaO/Ca(OH)2 in concrete substrate as well as the water chemistry of rainfall runoff and wastewater matrices. Some suspended and settleable soli ds may be present in runoff and wastewater and it has been recognized that phosphorus has an affinity to react with solid particles (Appan and Wang, 2000; Georgantas and Grigoropoulou, 2007). Therefore, the suspended/settleable solids in samples may serve as nuclei and promote the aggregation and precipitation of phosphorus on AOCMpcc. It is also believed the high concentration of Ca2+ in runoff (2 5.45 mg/L) and wastewater matrix (51.32 mg/L) could also contribute to the enhancement of phosphate surface precipitation in addition to those coming from dissolution of the solid AOCMpcc. Visual MINTEQ ver. 2.61 has been used to predict the potential P precipitates for AOCMpcc in all of the three matrices under both natural and phosphate spiked conditions. Solution conditions were based on the results from the final equilibrium state where 0.5 g of AOCMpcc were added to 40 mL working solutions of synthetic runoff, rainfall runo ff and wastewater, respectively. The saturation index (SI) representing the tendency of a mineral will be dissolved or precipitated. It can be seen from Table 2 6 that most of the SIs are well above 0, which reinforce the occurrence of surface precipitati on. Contrary to AOCMpcc, from Figure 2 7 and 2 9 it can be found that adsorption capacity of AOCMc and AOCMp severely decreased in rainfall runoff and wastewater matrices as no major removal mechanism transition occurred. In their study of rainfall runoff mental element speciation, Dean et al. (2005) detected that different anions and

PAGE 41

41 cations were transported into the dissolved phase in urban rainfall runoff. Besides PO4 3 -, the anions included NO3 -, SO4 2 and CO3 2 -. In this study, similar results were obt ained as shown in Table 2 1. Since the presence of multi anions, simultaneous adsorption may occur and cause direct competition for adsorption sites. Ma (2005) found SO4 2 showed apparent competition on P adsorption. Here, SO4 2 concentration in rainfall r unoff was found to be 10 mg/L, about 40 times more than that of synthetic runoff and an even higher SO4 2 concentration (10 6 mg/L) was detected in wastewater samples. The result of competitive effects could be a dramatic decrease in the adsorption capacity of PO4 3 In addition, although the presence of Ca2+ could increase the phosphorus adsorption through formation of ternary complex ( Sansalone and Ma, 200 9 ), it was found the effect can only be significant in base conditions. Therefore, SO4 2 competition should be a primary reason for weaker A OCMc and AOCMp isotherms in runoff and wastewater matrices. Summary This research investigated equilibria between aqueous phosphorus species (as total dissolved phosphorus, TDP) and aluminum oxidebased (AlOx) media consisting of three separate substrates ; clay ey soil pumice and Portland cement concrete. Equilibrium was modeled as Freundlich isotherms for three matrices: rainfallrunoff wastewater and deionized (DI) water as a surrogate runoff matrix. The pH of each aqueous matrix was 7 .0 When comparing the suite of media; results demonstrated that each AOCM exhibited significant ly higher capacities than the corresponding UCM substrate i n synthetic runoff matrix With the exception of the concretebased substrate the uncoated clay or pumice had relatively small equilibria capacity across the entire TDP

PAGE 42

42 range from 0.05 to 50 mg/L in comparison to AOCM of the same substrate. This result supports the studies hypothesis that the inclusion of AlOx either as an admixture or coating will statistically significantly increase adsorption capacity P revious studies have demonstrated ligand exchange as a mechanism for phosphate adsorption onto hydrous metal oxide surfaces In addition to ligand exchange, results of this study indicate that additional mechanisms, for example surface precipitation for the concretebased substrate (AOCMpcc), occur. AOCMc and AOCMpcc both provided similar high equilibrium capacity as compared to AOCMp. In constrast, the equilibrium capacity of the UCMpcc dominated the uncoated clay and pumice substrates due to surface precipitation. The smaller media size range of 0.85 to 2.0 mm generated higher equilibrium capacity as compar ed to the 2.0 to 4.75 mm size range on a dry media mass basis; but the increase was not as quantitatively large as inclusion of AlOx to each substrate. While a controlled reproducible DI surrogate matrix was utilized to compare isotherms for each AOCM and companion substrate; isotherm results from this surrogate matrix were compared to uncontrolled environmental rainfall runoff and wastewater matrices. While the DI surrogate matrix generated the highest equilibrium capacity across the range of TDP concent rations examined for AOCMc and AOCMp, this surrogate matrix generated the lowest equilibrium capacity for AOCMpcc as compared to the actual environmental matrices. Examining each AOCM type separately, the rainfall runoff and wastewater matrices generated isotherms that were similar for each matrix across the range of concentrations examined. With ligand exchange as the primary adsorption mechanisms, the decreased equilibrium capacity of AOCMc and AOCMp

PAGE 43

43 loaded by rainfall runoff and wastewater matrices can be attributed to multiple anion competition, such as SO4 2 competing with phosphate for media exchange sites. In contrast, AOCMpcc capacity combines the dominant mechanism of surface precipitation, ternary complexation with Ca2+ (Sansalone and Ma 2009) and ligand exchange to generate nearly identical equilibrium capacity across the entire TDP concentration range which was significantly greater than the surrogate DI matrix. Nomenclature AOCM: Alumina oxide coated media AOCMc: A lumina oxide coated media with clay as raw substrate AOCMp: Alumina oxide coated media with pumice as raw substrate AOCMpcc: Alumina oxide media with Portland cement concrete as raw substrate AlOx: Aluminum oxide b: The Langmuir constant (L/mg) C: The conc entration in the fluid phase (mg/L) C0: The initial phosphate concentration (mg/L) Ce: The phosphate concentration at equilibrium (mg/L) DI water: De ionized water DP: Dissolved phosphorus KF: The Freundlich constant N(b): The site energy distribution n: The Freundlich exponent pHpzc: pH of point of zero charge q: The amount adsorbed (mg/g) qe: The amount of adsorption at equilibrium (mg/g)

PAGE 44

44 qm: The maximum adsorbed phase concentration (mg/g) R2: Coefficient of determination SI: Saturation index SSA: Specific surface area (m2/g) T Temperature V: The volume of the solution (L); W: The weight of adsorbent used (g); UCM: Uncoated media UCMc: Uncoated media with clay as raw substrate UCMp: Uncoated media with pumice as raw substrate UCMpcc: Uncoated media with Portland cement concrete as raw substrate s: Specific gravity

PAGE 45

45 Table 21 W ater chemistry of experimental m atrix Matrix Synthetic r unoff Filtered rainfall r unoff (Event on 05/16/08) Wastewater (Secondary e ffluent) pH 7.00 0.05 7.11 0.07 7.14 0.10 Conductivity [s/cm] 1525 37 218 12 780 14 Turbidity [NTU] 2.75 0.56 2.80 1.02 4.40 0.64 Alkalinity [mg/L] as CaCO 3 7 1 74 2 77 3 Ionic Strength [M] 0.010 0.002 0.004 0.002 0.012 0.003 DOC [mg/L] 0.3 0.1 28.9 3.6 10.3 2.2 Cation [mg/L] K + a 388 2 4.68 0.07 12.83 0.14 Ca 2+ 0.021 0 .003 25.45 1.26 51.32 1.90 Mg 2+ BMLb 3.04 0.12 30.56 1.07 Pb 2+ 0.018 0.036 0.098 0.044 Cu 2+ 0.049 0.02 1.16 0.03 Zn 2+ 0.107 0.15 14.89 0.33 Cr 2+ 0.673 0.013 4.01 0.04 Mn 2+ 0.127 0.001 0.734 0.015 Al 3+ 0.212 0.032 5.56 0.090 Fe 3+ 0.128 0.038 6.34 0.11 Sum of cation c harge [mmol] 9.95 1.71 7.10 Anion [mg/L] NO 3 N 0.81 0.13 3.08 0.15 3.14 0.17 SO 4 2 0.23 0.09 10.06 1.08 106 1 Cl a 363 13 8.05 0.15 146 1 HCO 3 4.2 0.6 45.1 1.2 46.9 1.7 PO 4 3 0 ~ 50 (n=8) 1.13 0.17 9.15 0.43 Sum of anion c harge b [mmol] 10.36 1.43 7.46 Data are mean standard deviation a 0.01M KCl was used to adjust the ionic strength of synthetic runoff. b In synthetic runoff, PO4 3 was added as KH2PO4 and not included in the charge summation.

PAGE 46

46 Table 22 Major composition of raw s ubstrates Raw s ubstrate Major c ompositions Clay SiO 2 (48 70%); FeOOH (9 17 %); CaCO3 (3 6%) (Measured by XRD analysis) Portland cement c oncrete C S H ( 50 60 %); Ca(OH) 2 (20 25%) CSA (Calcium sulfoaluminates, 1520%) (Mehta, 1987) Pumice SiO 2 (72%); Al 2 O 3 (12%); K2O (5%); Fe 2 O 3 (2%) (Panuccio et al., J. Environ. Manage., 2009)

PAGE 47

47 T able 2 3 Media characteristics Media Substrate Size [mm] Specific g ravity pHpzc AOCMc Clay 0.85 2.0 2.29 0.04 4.61 0.06 2.0 4.75 UCMc Clay 0.85 2.0 2.31 0.05 4.20 0.08 2.0 4.75 AOCMpcc Concrete 0.85 2.0 2.67 0.04 11.84 0.16 2.0 4.75 UCMpcc Concrete 0.85 2.0 2.67 0.03 11.98 0.12 2.0 4.75 AOCM p Pumice 2.0 4.75 2.27 0.01 4.59 0.10 UCM p Pumice 2.0 4.75 2.29 0.02 4.61 0.10 pHpzc = pH of point of zero charge Data here are mean standard deviation (n = 3)

PAGE 48

48 Table 2 4 Freundlich model parameters in synthetic r unoff m atrix Media Size [mm] KF n AOCMc 0.85 2.0 1.35 0.33 2.0 4.75 1.05 0.36 AOCMpcc 0.85 2.0 1.32 0.58 2.0 4.75 0.89 0.39 AOCM p 2.0 4.75 0.39 0.26 UCMc 0.85 2.0 0.018 0.79 2.0 4.75 0.012 0.86 UCMpcc 0.85 2.0 0.52 0.53 2.0 4.75 0.49 0.52 UCM p 2.0 4.75 0.012 0.84 *All R2 values between measured and modeled exceed 0.96.

PAGE 49

49 Table 2 5 Freundli ch m odel parameters in rainfall runoff and wastewater m atrix Media [mm] Rainfall runoff Wastewater K F n K F n AOCM c (0.85 2.0) 0.32 0.43 0.35 0.31 AOCM c (2.0 4.75) 0.25 0.41 0.33 0.23 AOCM p (2.0 4.75) 0.16 0.33 0.17 0.20 AOCM pcc (0.85 2) 998 2.51 995 1.84 AOCM pcc (2 4.75) 312 2.29 109 1.93 *All R2 values between measured and modeled exceed 0.95.

PAGE 50

50 Table 2 6 Predicted Saturation I ndex (SI) for AOCMpcc in d ifferent m atrices Potential precipitates Actual aqueous m at rices* and phosphate spiked m atrices [mg/L] Synthetic r unoff Rainfall r unoff Wastewater 0* 25 50 1.13* 25 50 9.15* 25 50 Ca 3 (PO 4 ) 2 (am1) NA 2.16 2.72 0.45 2.19 2.73 1.49 2.33 2.93 Ca 3 (PO 4 ) 2 (am2) 4.93 5.49 2.33 4.96 5.50 4.26 5.10 5.68 Ca 3 (PO 4 ) 2 (beta) 6.02 6.58 3.42 6.05 6.59 5.35 6.19 6.78 Ca 4 (HPO 4 ) 3 :3H 2 O 3.20 4.07 0.32 3.67 4.51 2.80 4.07 4.94 Hydroxyapatite [HAP, Ca 5 (PO 4 ) 3 OH] 19.5 9 20.3 9 15.2 9 19.2 2 20.0 0 17.9 8 19.2 3 20.0 9 SI = logIAP logKsp: 1 < SI < 0, saturated; SI< 1, undersaturated; SI > 0 supersaturated.

PAGE 51

51 Wavenumber [cm-1] 500 1000 1500 2000 2500 3000 3500 4000 Absorbance 0.00 0.05 0.10 0.15 0.20 0.25 0.30 OH-NO3 -Al O Figure 2 1 FTIR spectra of m edia c oati ng

PAGE 52

52 AOCMc (0.85 -2.0 mm) UCMc (0.85 -2.0 mm) Freundlich Model Qe [mg/g] 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 AOCMc (2.0 4.75 mm) UCMc (2.0 4.75 mm) Freundlich Model Ce [mg/L] 0 10 20 30 40 50 AOCMpcc (0.85 2.0 mm) UCMpcc (0.85 2.0 mm) Freundlich Model Qe [mg/g] 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 AOCMpcc (2.0 4.75 mm) UCMpcc (2.0 4.75 mm) Freundlich Model Ce [mg/L] 0 10 20 30 40 50 Qe [mg/g] 0.0 0.2 0.4 0.6 0.8 1.0 1.2 AOCMp (2.0 4.75 mm) UCMp (2.0 4.75 mm) Freundlich Model Clay substrate Portland cement substrate Pumice substrate Clay substrate Portland cement substrate F igure 22 Comparison of AOCM and UCM in synthetic runoff m atrix

PAGE 53

53 Qe [(mg/g] 0.0 1.0 2.0 3.0 4.0 AOCMpcc (0.85 2 mm) AOCMpcc (2 4.75 mm) Freundlich Model Qe [mg/g] 0.0 1.0 2.0 3.0 4.0 AOCMc (0.85 2 mm) AOCMc (2 4.75 mm) Freundlich Model Qe [mg/g] 0.0 0.1 0.2 0.3 0.4 UCMc (0.85 2 mm) UCMc (2 4.75 mm) Freundlich Model Ce [mg/L] 0 10 20 30 40 50 Qe [mg/g] 0.0 0.5 1.0 1.5 2.0 2.5 3.0 UCMpcc (0.85 2 mm) UCMpcc (2 4.75 mm) Freundlich Model Figure 2 3 Comparison of d ifferent m edia size in synthetic r unoff m atrix

PAGE 54

54 Ce [mg/L] 0 9 18 27 36 45 AOCMc (Clay) AOCMpcc (Portland cement) AOCMp (Pumice) Ce [mg/L] 0 10 20 30 40 50 Qe [mg/g] 0 1 2 3 4 UCMc (Clay) UCMpcc (Portland cement) UCMp (Pumice) 2.0 4.75 mm 2.0 4.75 mm Figure 2 4 Comparison of d ifferent AOCM and UCM in synthetic runoff m atrix

PAGE 55

55 0 5 10 15 20 25 30 Qe [mg/g] 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 AOCMc(0.85 2 mm) AOCMc(2.0 4.75 mm) Freundlich Model 0.00 0.04 0.08 0.12 0.16 0 1 2 3 4 5 AOCMpcc(0.85 2 mm) AOCMpcc(2.0 4.75 mm) Freundlich Model Ce [mg/L] 0 10 20 30 40 50 Qe [mg/g] 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 AOCMp(2.0 4.75 mm) Freundlich Model Ce [mg/L] 0 10 20 30 40 50 0 1 2 3 4 5 AOCMc (Clay) AOCMpcc (Portland cement) AOCMp (Pumice) 2.0 4.75 mm Figure 2 5 Comparison of AOCM in rainfallrunoff m atrix

PAGE 56

56 0 10 20 30 40 50 Qe [mg/g] 0.0 0.2 0.4 0.6 0.8 1.0 1.2 AOCMc (0.85 2 mm) AOCMc (2 4.75 mm) Freundlich 0.00 0.03 0.06 0.09 0.12 0.15 0 1 2 3 4 5 AOCMpcc (0.85 2 mm) AOCMpcc (2 4.75 mm) Freundlich Model Ce [mg/L] 0 10 20 30 40 50 Qe [mg/g] 0.0 0.1 0.2 0.3 0.4 AOCMp (2 4.75 mm) Freundlich Model Ce [mg/L] 0 10 20 30 40 50 0 1 2 3 4 5 AOCMc (Clay) AOCMpcc (Portland cement) AOCMp (Pumice) 2.0 4.75 mm Figure 2 6 Comparison of AOCM in wastewater m atrix

PAGE 57

57 Qe [mg/g] 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Synthetic Runoff Rainfall-runoff Wastewater Freundlich Model AOCMc (0.85 2.0 mm) Ce [mg/L] 0 10 20 30 40 50 Qe [mg/g] 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Synthetic Runoff Rainfall-runoff Wastewater Freundlich Model AOCMc (2.0 4.75 mm) Figure 2 7 Comparison of AOCMc in s ynthetic runoff, r ainf all r unoff and w astewater m atrices

PAGE 58

58 0 1 2 3 4 5 6 Qe [mg/g] 0 1 2 3 4 5 Synthetic Runoff Rainfall-runoff Wastewater Freundlich Model Ce [mg/L] 0 5 10 15 20 Qe [mg/g] 0 1 2 3 4 5 Synthetic Runoff Rainfall-runoff Wastewater Freundlich Model AOCMpcc (0.85 2.0 mm) AOCMpcc (2.0 4.75 mm) Figure 2 8 Comparison of AOCMpc c in synthetic runoff, rainfall runoff and wastewater matrices

PAGE 59

59 Ce [mg/L] 0 10 20 30 40 50 Qe [mg/g] 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Synthetic Runoff Rainfall-runoff Wastewater Freundlich Model AOCMp (2.0 4.75 mm) Figure 2 9 Comparison of AOCMp in synthetic runoff, r ainfall r unoff and w astewater m atrices

PAGE 60

60 CHAPTER 3 BATCH EQUILIBRIA BETWEEN AQUEOUS PHOSPHORUS AND GRANULAR FILTER MATERIALS Background It has been well documented that increased phosphorus (P) concentrations in water bodies is a serious pollution concern since P is one of the major limiting factors responsible for eutrophication (U.S. Geological Survey, 1999; Markris et al., 2004; Oleszkiewicz and Barnard, 2006; Endards and Withers, 2007). Consequently, there is increasing pressure and changes in water discharge regulation and legislation for lowering phosphorus concentration in surface water as well as phosphorus total maximum daily load (TMDL) (Read and Fernandes, 2003; USEPA, 2000a; Havens and Schelske, 2001; Walker, 2003). For instance, regulation for P discharge can be as low as 50 g/L for wastewater and concentratio ns in excess of 30 g/L have been established as an indication of excessive nutrients enrichment in lakes and other confined water bodies (Genz et al., 2004; Oguz, 2005). Industrial and municipal wastewater, agriculture and animal feeding operations are wellknown phosphorus sources. In recent years, it has also been recognized that urban rainfall runoff can transport a large amount of nutrients, including phosphorus and nitrogen and severely impair the receiving waters (USEPA, 2002). Phosphorus occurs in the forms of both dissolved P (DP) and particulate P (PP) and mainly exists as phosphate in aquasystems and secondary wastewater effluents (Wei et al., 2008; Ma, 2005; Chitrakar et al., 2006). P can be generated from both point sources like domestic and i ndustrial discharge and diffuse source such as urban rainfall runoff. P can be effectively treated via settling and filtration, chemical precipitation, biological method and constructed wetland. However, these technologies are either

PAGE 61

61 more effective for PP removal or constrained from diffuse sources control. For example, the traditional chemical coagulants and biological methods suffer the drawbacks of considerable capital investment, maintenance cost for infrastructure, reagents and sludge disposal. A feas ible process for P control from diffuse sources is adsorptive filtration with employment of low cost filter media. Diverse solids with regarding to P removal capacity have been tested and evaluated, including natural materials, industrial by product/waste, active filter substrates, engineered media etc. To exempt form a tedious enumeration here, the author summarizes the recent studies in Table 3 1. Generally P is bound to the media as a result of adsorption and precipitation reactions with Ca, Al and Fe in the media substrate and the P adsorption capacity of a material represents a central parameter for comparing and selecting candidate materials to be used as P removal media (Seo et al., 2005). However, although lots of materials have been tested, there ar e still some media applications without proper evaluation and suffering a low efficiency. Moreover, although lot of quantitative information of various adsorbents is available in literature, comparisons and material selections can still be problematic. One reason is that usually these media were tested under different experimental conditions and thus lost the equal base. Even within comparable laboratory control, selections are still not straightforward due to the lack of proper quantitative parameters. The P adsorption capacity is frequently estimated using Langmuir or Freundlich isotherm equations because of the simple form and satisfactory goodness of fit with the experimental data. However, it has been demonstrated that the Langmuir equation can lead to biased and unrealistic estimates of the adsorption

PAGE 62

62 parameters (Seo et al., 2005). As a nonlinear adsorption isotherm, Freundlich model has been favored due to the mathematical simplicity as well as the characteristic of heterogeneous surface of the media materials and the diversity of sorption mechanisms. But some researchers have noticed that neither of the two Freundlich parameters can independently represent the media capacity and comparisons among different adsorbents must be taken with caution due to the hidden complexities of KF (Bowman, 1981, 1982; Chen et al., 1999). Furthermore, the P adsorption capacity as the first and most important criteria in media selection is intuitive and indubitable; but it is not the only requirement that needs to be sati sfied. In addition to P removal capability, the candidate media should also be low cost with steady supply and not contain or release hazardous pollutants into environment. While economical efficiency has drawn enough attention since a variety of low cost materials and industrial by product/waste are investigated as P adsorbents (Table 3 1), potential environment contaminations caused by the media materials are frequently overlooked. Objectives In this paper, seven granular filter materials, including Tire crumb B&G (2 gradations), Expanded shale (2 gradations), Bioretention soil media, ZeolitePerlite GAC, Activated alumina, Fe coated perlite and clay based Aluminum Oxide Coated Media (AOCMc) were chosen and tested through equilibrium isotherm approach. Stu dies are carried out to validate the following hypotheses: Under consistent surrogate matrix and identical experimental conditions, the suite of media examined can exhibit large variance in aspect of P removal capacity, which can be quantitatively characterized by utilization of Freunlich model. Furthermore, media comparison can be

PAGE 63

63 facilitated by a newly defined unified sorption variable Ku developed by Chen et al. (1999). Finally, deployment of some high capacity media may be constrained by unfavorable met al leaching. Materials and Methods Media AOCMc (Aluminum Oxide Coated Media with expanded clay as substrate) An engineered media, AOCMc was developed and tested in this study. Bentonite with mineralogical composition of sodium hydrated aluminosilicate was purchased from Halliburton Energy Services, Inc. The raw substrate was then prepared through procedures of mixing with explosive reagents and water, sintering in kiln, crushing and sieving, acid washing, DI rinsing and drying. AOCMc was prepared by applyi ng aluminum oxide as surface coating. Stock solution of 1M Al (III) was prepared by dissolving Al(NO3)3 9H2O in DI water. Al (III) solution was added to the raw substrate and the mixture was then put in the oven at a temperature around 200 oC for at least 24 hours. Later, the aluminum coated sorbents was washed with DI water to remove the surficial fine materials. Then the sorbent was dried at 105 oC and stored in polypropylene bottles for further use. The effective gradation utilized here are 0.85 to 2 mm Activated a lumina Activated alumina was purchased as a commercial filter media from Schoofs Inc. CA, U. S. A. The alumina is primarily composed of Al2O3 (91.5%), SiO2 (0.02%), Fe2O3 (0.03%) and Na2O (0.35%). The particle size is within the range of 1.4 2.41 mm. Bioretention soil media

PAGE 64

64 Bioretention soil media tested in this study is commercial mixtures for Prince Georges County, Maryland with P index of 25. Soil mixtures include sand, loamy sand, and sandy loam Expanded shale Expanded shale is another commercial filter media. It is a porous material and usually manufactured by firing shale at a high temperature ( 2000 F Forbes et al., 2004). Little research has been undertaken to examine the applicability of expanded shale for phosphorus removal. In this study, the expanded shale was sorted by two size gradation: 2 10 mm and 0.45 4.5 mm. Fe coated perlite Fe coated perlite used in this research was collected from Imbrium Systems, Inc. Fer ric oxide was coated on perlite to develop a light weighted engineered media. The size ranges from 0.3 to 0.85 mm. Tire crumb B&G Tire crumb is a waste material and has been historically stockpiled due to limited economic and environmental feasibility of tire regeneration. As part of resources recovery with sustainable implication tire crumb B&G has been used as construction materials and roadbed support and fills (Yoon et al., 2006; Liu et al., 2000). Lisi et al. (2004) also studied the use of granulated tires as a subsurface drainage layer to mitigate nutrient leaching. However, more information and tests are needed to evaluate the potential of tire crumb as a phosphorus adsorptive material. In this study, green roof mixture of mainly tire crumb B&G was tested to assess its phosphorus removal ability with two gradations: 0.5 4.75 mm and 0.85 10 mm.

PAGE 65

65 Zeolite Perlite G AC Zeolite Perlite GAC is a commercial mixture of zeolite, perlite and granulated activated carbon with the volume ratio of and size range of 2.29 9.65 mm. A brief description of the media utilized in this study is summarized in Table 3 2. Physical Characteristics Specific surface area measurements were conducted using a modified ethylene glycol monoethyl ether method described by Sansalone et al. (1998). Surface charge for the tested media was measured using a potentiometric titration methodology described by Van Raij and Peech (1972). pHpzc was then calculated by linearly interpolating between the last positive and first negative sus) measurement utilized a helium gas pycnometer. Chemicals A phosphorus stock solution was prepared by dissolving anhydrous potassium dihydrogen orthophosphate (KH2PO4) powder (Analytical reagent grade, Fisher Scientific) into D.I. water. The stock solution was calibrated to 100 mg/L and 1000 mg/L concentrations which were utilized as standard solutions as necessary. Batch Adsorption Study Batch studies were performe d to examine phosphate adsorption onto the different media. 40 mL of synthetic runoff dosed with predetermined amount of phosphate (0, 0.05, 0.5, 1.0, 2.5, 10, 25 and 50 mg/L) and 0.5 g of the tested media were agitated on a horizontal bench shaker at a rate of 100 rpm allowing adsorption to take place at 20C until equilibrium reached, usually 24 hr. I onic strength was fixed at 0.01 M using KCl and s olution pH was adjusted to 7 at the beginning of experiments by dropwise addition of HCl or NaOH. Measurements of pH were carried out using a multi meter (Orion 290A

PAGE 66

66 combination electrode) which was preliminary calibrated using a 3 point calibration curve with reference units at a pH of 4, 7, and 10 Phosphorus Analysis After adsorption equilibrium reached, sam ples were taken and filtered through 0.45m syringe filter T he total dissolved phosphorus concentration in the filtrate was measured by HACH DR/5000 Spectrophotometer using PhosVer 3 Ascorbic Acid Method (Standard Method 1998). All measurements were dupli cated and high repeatability was assured by controlling experimental error within 5%. The amount of adsorbate that was adsorbed on the adsorbent qe [mg/g], was then calculated from the difference between the initial and equilibrium solute concentrations. W VC C qe e) (0 (1) C0 and Ce are the initial and equilibrium solute concentrations [mg/L], respectively; V the volume of the solution [L] and W the dry mass of the adsorbent [g]. Me tal Leaching Tests Metal leaching experiments were carried out to verify if the treated water quality may be contaminated with soluble metals leaching from the treatment media materials during adsorption. The possibility of metal leaching was examined by mixing deionized water samples with each of the media tested, mostly mimicking the experimental conditi ons in batch equilibrium study, followed by analysis of the water. Specifically, 0.5 g of each media was added to 40 mL of deionized water in a 50 mL centrifuge tube and then shaken for 24 hours at 100 rpm at 20 C Ionic strength was maintained at 0.01 M with KCl. Water samples were then filtered into a sample vial and acidified prior to analysis. The samples were analyzed for dissolved metal content using ICP

PAGE 67

67 instrumentation ( Perkin Elmer Plasma 3200). Similar experimental approach was reported in the lit erature by Sibrell et al. (2009). The Freundlich Isotherm and Unified Sorption Variable, Ku Equilibrium isotherm is a graphical representation showing the relationship between the amount adsorbed by a unit weight of adsorbent and the amount of adsorbate remaining at equilibrium. The operating curve of an equilibrium isotherm model represents an asymptotic limit as a function of concentration for adsorption capacity of a given soil or media system. An isotherm model can be represented by a number of mathem atical forms that can be combined kinetics and breakthrough representations As the isotherm maps the distribution of adsorbate between the liquid and solid phases, a critical parameter is the concentration ratio of solid phase to aqueous phase. Usually, t his ratio is described by partitioning coefficient, Kd. Specifically to this study, Kd can be represented as: e e dC q K (2) where Kd is the partitioning coefficient [L/g]; qe is the concentration of solids adsorbed on the solid phase [mg/g] and Ce is the concentration of P in the aqueous phase at equilibrium [mg/L]. The partitioning process can be fundamentally understood from a thermodynamics point of view, such as free energy and chemical potential, and be expressed on a mole fraction basis as following ( Schwarzenbach, 2002); ] exp[ T R G Kd (3)

PAGE 68

68 T is the absolute temperature. When a linear isotherm is assumed, Kd is conastant. However, in most of the media adsorption system, t he heterogeneou s composition and structure of the media substances and the potential of multiple sorption mechanism contributes to the variation and the nonlinear Freundlich isotherm equation is commonl y used to model heterogeneous adsorbent surfaces (Coles and Yong, 2006). The equation for Freundlich isotherm is: n e F eC K q ( 4 ) KF is the Freundlich constant and n is the adsorption intensity and depends on the linearity of the isotherm, usually between 0 and 1. Thus KF represents the sorption capacity of the media when the equilibrium solution concentration is 1.0 mg/L. When n = 1, the isotherm becomes linear and KF = Kd. The nonlinearity of Freundlich isotherm accommodates the heterogeneous composition and structure of different media materials and diverse sorption mechanisms, resulting in a widely application in many sorption systems (Chitrakar et al., 2006; Kim et al., 2006; Mortula et al., 2007; Wei, et al., 2008; Sibrell et al., 2009). Usually the mathematical modeling facilitates quantitative comparison among different media with model parameters and such comparison and evaluat ion are often imperative when a variety of media are available. However, in Freundlich isotherm the units of KF depend on the value of n, whenever n values are different among media, the KF values are not directly comparable. To address this limitation and make the model parameter

PAGE 69

69 more amenable to media comparison, Cheng et al (1999) proposed a unified sorption variable, Ku, which has the same units as Kd. The expression of Ku was derived as following: e u n e FC K C K ( 5 ) ) 1 (/ n e F e n e F uC K C C K K ( 6 ) I f Ce has the unit of mg/L and qe of mg/g as in this study, then the unit of Ku is g/L. Actually in contrast to linear isotherm system where the partitioning coefficient of sorbate between the aqueous and solid phase is a constant, Ku represents the distrib ution of the coefficients in nonlinear system. Plot of Ku vs C0 (or Ce) can be a instructive illustration of the dynamic mass transfer of the sorbate between the two phases and provide a more reliable and predictive estimation of media capacity. Additional ly, single point adsorption isotherm approach is frequently employed in researches (Gao et al., 2004; Penn et al., 2005; Raposo et al., 2009; Karathanasis Shumaker, 2009). However, Eq ( 6 ) clearly demonstrates that partitioning coefficient is a function of equilibrium concentration and a singlepoint isotherm can only represent one point of the whole distribution. Moreover, real situations such as rainfall runoff events are dynamic processes that exhibit changes not only in constituent concentration (can var y by 1 order of magnitude) but also the speciation and partitioning. In that case, partitioning coefficient is also a function of the immediate environment chemistry. Results Physical Characteristics Table 32 summarizes the physical properties of all the media utilized in this study. Two size gradations of expanded shale and tire crumb B&G were tested. Bioretention

PAGE 70

70 soil media has the widest size range, partially due to the miscellaneous constitutes of soil samples. The high specific surface area of activated alumina and AOCMc should be ascribed to the highly porous structure of alumina and the presence of GAC contributes to the high SSA of Zeolite Perlite GAC. The third highest SSA of Fe coated perlite mostly stems from the fine particle sizes (0.35 0.85). The variety of specific gravity reflects the diverse materials that constitute the media. Also, pHpzc ranges widely from 4.61 to 9.84. P Removal Efficiency and Sorption Isotherm The percent of P rem oved based on different initial P concentrations is depicted in Figure 3 1. The initial DP (dissolved phosphorus) concentration varied from 0.05 to 10 mg/L. It can be seen generally the percent removed decreased with the increase of initial P concentrations, indicating saturation of the media sorption capacity. However, activated alumina and Fecoated perlite maintained a removal efficiency of more than 90% within the tested concentration range and AOCMc also achieved more than 70% removal at C0 of 50 mg/L, exhibiting high capability with respect to P removal. Futhermore, P removal efficiency was presented in the form of mass of P removed per media mass [mg/g] as a function of the equilibrium concentration of P in the aqueous phase and the Freundlich equati on was applied to model the sorption isotherms. Figure 3 2 compares the performance of two size gradations of tire crumb B&G and expanded shale. It can be found that tire crumb B&G b and expanded shale b showed better capacities, corresponding to a smaller particle size and higher SSA. Therefore, only tire crumb B&G b and expanded shale b were plotted in Figure 3 3, which illustrates the comparison of sorption capacities among all the media tested in this study. It is easily noticed that in Figure 3 3, the eight filter media can be divided into

PAGE 71

71 two groups according to the P removal capacity. Activated alumina, Fecoated perlite and AOCMc constituted one group whose capability of P sorption was far superior to those of the rest. Especially for activated alumi na and Fecoated perlite, the isotherm profiles are almost parallel with the Qe axis within the concentration range used in this study. On the contrast, tire crumb B&G b exhibited very limited capacity with an isotherm curve nearly parallel with Ce axis. The Freundlich model parameters are summarized in Table 33 and the high value of correlation coefficient indicates a good fitting between the experimental and calculated data. The heterogeneity of media materials and variety of sorption mechanisms make Fr eundlich isotherm an appropriated equation to describe the process involving P mass transfer between aqueous and solid phase. Although the nonlinear Freundlich isotherm is widely used, comparison of model parameters KF of different media is constrained by the unit. The unit of KF is determined by n and cannot be compared directly between different sorbents even when the experiments are carried out under identical condtions. For this reason, Chen et al. (1999) proposed a unified sorption variable Ku with the same unit as distribution coefficient, Kd. In this study, the unit of mg/L was assigned to Ce and mg/g to qe, therefore, Ku has a unit of g/L. Ku was calculated from equation ( 6 ) and Figure 3 4 plotted Ku as function of Ce An approximately Lshape curve was obtained for all the media tested, showing an initially rapid decrease in Ku followed by a more moderate decrease. This result is consistent with those of Chen et al. (1999) and Marchi et al. (2006). Here the values of Ku ranged from >20 L/g to < 0.2 L/g. Two orders of magnitude variation indicates significant differences of the media capacity, therefore in Figure 3 4 a log scale of Ku was

PAGE 72

72 employed for a better demonstration of the comparison. One reason that Freundlich e quation has been widely favored is its simple mathematical form with only two parameters. However, as was pointed out by Bowman (1981), the hidden complexities in KF provide no rational basis for comparing sorption capacities among different sorbents. By employing the unified sorption variable of Ku, the media capacity can be reasonably quantitatively evaluated and compared. Metal Leaching from Media Besides a satisfactory efficiency, i t is very important that the water treatment process must not impair the treated water quality by contamination of the water with soluble salts or metals from the treatment media material (Sibrell et al., 2009) 13 metals in the filtrate from metal leaching tests, including Al, As, Ba, Ca, Cd, Cr, Cu, Fe, Mg, Mn, Ni, Pb and Zn were measured as described previously. All the results were summarized in Table 3 4, expressed as milligram of metal leached per gram of media. It can be noted that the concentration of As was below Method Detection Limit (0.05 mg/L) for all the media and so were the concentrations of Cr (MDL: 0.002 mg/L) and Ni (MDL: 0.005 mg/L) for most media. Generally, these media contained Al, Ca, Fe and Mg and other metals measured in this study were mostly in trace level. Previous studies have shown that p hosphorus sorption is often positively related to media properties of Al, Fe, and Ca content (Kang et al. 2003; Karageorgiou et al., 2007; Mortuna et al., 2007). Figure 3 5 illustrates the amount of Al, Fe and Ca leached out of each media in contrast with the unifi ed sorption coefficient Ku at Ce of 0.03 mg/L which was generally accepted as an indication of excessive nutrient enrichment in confined water bodies (Genz et al., 2004; Oguz, 2005) and at Ce of 0.1 mg/L (USEPA, 1986). It can be seen from Figure 3 5 that F e coated perlite leached significant amount of Fe as compared

PAGE 73

73 with the other media. Activated alumina, and bioretention soil media were the two adsorbents that released most Al among all the media. Most media leached certain amount of Ca. Comparing expanded shale a and expanded shale b, it can be found that the smaller gradation resulted in a higher metal concentrations in the filtrate. The same result was obtained for tire crumb B&G a and b, indicating that particle size impacted on metal leaching quantiti es as well. In addition, if the amount of metal leached could be positively related to the content of metal in each media, Figure 3 5 implies it is not necessarily true that higher content of Al/Fe/Ca results in higher removal efficiency. The media capacit y is determined by many factors, such as physical characteristics, chemical properties and the form and activity of metal content. Discussions and Implications As summarized in Table 31, alumina related residuals, by products and sludge have been extens ively studied as potential adsorbent for P removal In many soils Al and Feoxides are present making soil a reactive substrate, specifically for phosphate species (Fontes and Weed, 1996; Nooney et al., 1998; Arias et al., 2006; Boujelben et al., 2008). It has been reported that aluminum hydroxide is formed on the surface of activated aluminum oxide in water and then reacts with phosphate ions through ligand exchange as shown in the following equations (Urano and Tachikawa, 1991): Al2O3 + 3H2O 3 Al (OH)3 + H2PO4 4 + OH+ 2H2O Analogous mechanism accounted for the reaction between phosphate ion and surface alumina coating on AOCMc. Similarly, Fe oxide tailings, blast furnace slag, steel slag and granulated ferric hydroxide were tested for removal of P due to the richness of Fe in the materials (Genz

PAGE 74

74 et al., 2004; Zeng et al., 2004; Drizo et al., 2006; Mortula et al., 2007.) It is generally agreed that the mechanism of phosphate removal involved the formation of inner sphere complexes on Feoxi de surfaces (Genz et al., 2004). Hence, along with the fine particle size (therefore, relatively higher SSA), high P removal efficiency was achieved by Fe coated perlite in this study. Additionally, here the batch equilibrium experiments were carried out at pH of 7, which was lower than pHpzc (7.91) of Fecoated perlite and two mechanisms including electrostatic attraction and ligand exchange (inner sphere complexes) could be involved for P removal(Chitrakar et al., 2006). The moderate P uptake capability of expanded shale was possibly ascribed to the major content of calcite. It has been proved that the P removal of Cacontaining materials was through the formation of insoluble CaP compounds and consequently eorgiou et al., 2007). Zeolite perlite GAC is a mixture of three filtration media: zeolite, perlite and GAC with volume ratio of 5:4:1. The major component zeolite is a family of natural and synthetic aluminosilicate minerals. It is the main absorptive material used in agricultural and industrial situati ons. Perlite is a naturally occurring volcanic product characterizing with lightweight, chemically inert, coarse, and granular. It is an effective physical filtration media. GAC is a highly porous adsorbent material, but not usually used for P removal (Mortula et al., 2002). The zeoliteperlite GAC mixture may be more effective as filtration media and the limited capacity of DP adsorption can be ascribed to the Al and Ca oxides contained (Sakadevan andBavor, 1998). As a relatively novel best management practice for urban storm water, bioretention has been studied and employed to minimize the impact of urban runoff during storm

PAGE 75

75 events (Hsieh and Davis, 2005; Li and Davis, 2008). Generally, bioretention consist s of porous media layers of mixes of soil, sand and organic matters and utilizes the chemical, biological and physical properties of plants, microbes and soils for removal of pollutants from storm water runoff. I t appeared that phosphorus removal was the m ost variable of six pollutants examined in column and grab filed studies ( 240% ~ over 80%, Davis et al., 2009) The bioretention soil media tested in this study is commercial mixtures for Prince Georges County, Maryland with P index of 25. As a result, v ery low DP uptake capacity was achieved. In one aspect soil sorption of P could be long term process, which may not take place and finish within the time scale of this study; in another aspect, the complexity of the chemistry of these species as well as t he soil characteristic accounts for the capacity variety. I t has been recognized that treatment of dissolved phosphorus species is much more sensitive to media characteristics and proper media selection is critical for effective DP removal (Davis et al., 2 009; Hunt et al., 2006). Another media tested in this study which was also known to be utilized in BMPs for stormwater treatment was tire crumb B&G the main composition of a commercial green roof mixture. Green roofs offer an aesthetically pleasing treatment solution that utilizes unused space to treat and store stormwater runoff and pollutants may be removed via several processes, such as sedimentation, adsorption, filtration, bioremediation etc. The efficiency of the system will be determined from the total precipitation and the total overflow. Actually, it has been reported that t he green roof should be irrigated with recycled stormwater runoff, which will contain suspended and dissolved nutrients that are repeatedly recycled in order to maximize t he natural nutrient utilization capabilities

PAGE 76

76 of the biological processes taking place (Hardin, 2006). In this study, very limited DP uptake capacity was achieved for these media. As it has been shown the media chemical characteristics were correlated with pollutant removals (Hsieh and Davis, 2005), results obtained here indicate that further modification of the filter media should be conducted, more effectively designed process should be applied or other practices/techniques should be conjunct with to achieve favorable DP removal efficiency. It has been reported that p hosphorus sorption is often positively related to media properties such as Al, Fe, and Ca content (Kang et al. 2003; Karageorgiou et al., 2007; Mortuna et al., 2007). In order to improve the media capacity, efforts have been put to modify the natural adsorbents and develop engineered media, usually by applying Al or Fe coating (Ayoub et al., 2001; Arias et al., 2006; Boujelben et al., 2008). On one hand higher content of Al/Fe in media should be desirable from the efficiency point of view; on the other hand, unfavorable metal leaching could be an environmental concern associated with the application of such engineered media. These two apparently contradictions not only result in more challenges i n media preparation but also require more discretion in media evaluation. Among the media tested in this study, Fecoated perlite and activated alumina showed best P removal capacity. However, at the same time 3.49 0.05 mg Fe/ g Fecoated perlite and 1.0 9 0.02 mg Al / g activated alumina were released, respectively. Moreover, it should be noticed that these results were obtained at pH of 7. Much more acidic condition could be encountered in the actual environment and may result in higher amount of metal leaching. Therefore, albeit the high sorption capacity of activated alumina and Fecoated perlite, caution should be

PAGE 77

77 taken with the selection of these media since quite amount of Al and/or Fe leaches into the treated water. Comparatively, a satisfactory P removal efficiency can be achieved with AOCMc but without the compromise of significant metal leaching, which indicates AOCMc can be a promising adsorptivefiltration media for P control. Summary In this study, batch equilibria between dissolved P (DP) and 9 adsorptive media were tested and evaluated. Results demonstrated a significant capacity variation among different media. The unified sorption coefficient Ku developed by Chen et al. (1999), which illustrates the distribution of partitioning coefficient of sorbate between the aqueous and solid phase at various equilibrium concentrations, was adopted as a quantitative index to facilitate the comparison of media capacity. Results indicated that activated alumina, Fecoated perlite and AOCMc were the three media with good equilibrium capacity while the rest including tire crumb B&G, expanded shale, ZeolitePerlite GAC and bioretention soil media exhibited very limited capabilities. The possible treated water contamination due to media application was exami ned through metal leaching tests. Al, Ca, Fe and Mg were found to be the major metals released into the water sample. It was also found the high P removal efficiency of activated alumina and Fe coated perlite was accompanied by a significant amount of Al and Fe leaching, imposing an environmental concern of these media application. On the contrary, AOCMc seemed to be able to satisfy both P removal and minimal contamination requirement and thus it is considered as a promising cost effective adsorptive filtration media for P treatment.

PAGE 78

78 Nomenclature AOCMc: Alumina oxide coated media with clay as raw substrate T Temperature C0: The initial phosphate concentration (mg/L) Ce: The phosphate concentration at equilibrium [mg/L] DP: Dissolved phosphorus DI water: D e ionized water Free energy transfer between two phases Kd, Partitioning coefficient [L/g]. KF: The Freundlich constant Ku Unified sorption variable [L/g] n: The Freundlich exponent pHpzc: pH of point of zero charge PP Particulate phosphorus q: The am ount adsorbed (mg/g) qe: The amount of adsorption at equilibrium [mg/g] qm: The maximum adsorbed phase concentration [mg/g] R The general gas constant R2: Coefficient of determination SSA: Specific surface area [m2/g] T The absolute temperature TMDL Total maximum daily load V: The volume of the solution (L);

PAGE 79

79 W: The weight of adsorbent used (g); s: Specific gravity

PAGE 80

80 Table 31 Summary of sorption media used in s hort ter m isotherm batch experiments to r emove P Media Aqueous Matrix Matrix P [mg/L] pH (s.u.) Capacity [mg/g] Reference Alum sludge DI 2.5 7 0.674 Mortula et al., 2007 wastewater 3.7 6 .5 7 0.185 Acid mine drain sludge DI 20 7 9.89 ~ 31.97 Wei et al., 2008 Acid mine drain sludge DI 1.0 --1.82 23.9 Sibrell et al., 2009 AlOx MBR Filtrate 4 7.9 8.7 12.0 13.8 Genz et al., 2004 AlOx media DI 0 8.16 7 2.1 Sansalone & Ma, 2009 Al residual DI 0 100 --1.33 48.7 Dayton et al., 2003 Alum sludge DI 5 4.3 9 0.7 3.5 Yang et al., 2006 Bioretention media DI 3 7 0.081 Hsieh and Davis, 2005 Blast furnace slag DI 16.3 163 9.8 11.7 3.2 18.9 Kostura et al., 2005 Blast furnace slag DI 326.3 6.7 6.8 46.5 Lu et al., 2008 Calcite Tap water 32.63 7.6 12 3.1 Karageorgiou et al 2007 Coal bottom ash Landfill leachate 0.05 0.09 7 0.008 0.034 Lin and Yang, 2002 Double hydroxides DI 16.3 6 44.3 Das et al., 2006 Expanded shale DI 0.5 3.0 7 0.97 Forbes et al., 2004 Fe5HO84H2O WWTP effluent 2.7 7 5 Kang et al., 2003 FeOx MBR Filtrate 4 7.9 8.7 16.9 23.3 Genz et al., 2004 FeOOH seawater 0.3 1.0 7.8 10 Chitrakar et al., 2006 Fly ash DI 26.1 11.5 71.87 Oguz 2005 Iron oxide tailings DI 20 6.7 6.8 7 Zeng et al., 2004 KAl 3 (SO 4 ) 2 (OH) 6 DI 62 5 145.6 zacar, 2003 Lignocellulose media DI 50 1000 --16.4 80.9 Kim et al., 2006 Masonry sand DI 0.5 3.0 7 0.59 Forbes et al., 2004 Mineral based sorbent DI 5 8 1.9 19 Gustafsson et al., 2008 Powered alum sludge DI 4.8 4.2 0.98 Fu et al., 2008 Recovered ochre DI 100 --2.5 Heal et al., 2004 Sands DI 2.5 320 7.6 9.6 0.004 1.19 Bubba et al., 2003 Steel furnace slag DI 326.3 6.7 6.8 33.3 Lu et al., 2008 Synthetic hydrotalcite DI 200 8.6 47.3 Kuzawa et al.,2006 Zirconium hydroxide DI 1 10 5.0 30 Chitrakar et al., 2006 Seawater 0.30 7.7 10 wastewater 2 8.8 15 ZnCl 2 activated coir pith carbon DI 3.3 13.01 4 0.47 1.35 Namasivayam and Sangeetha, 2004 Suspended PM a Rainfall runoff 0.082 7 8 3.27 Sansalone et al., 2010 Settleable PM 0.42 Sediment PM 0.22 a PM = Particulate matter

PAGE 81

81 Table 32 Media c haracteristics and c ost Media Size [mm] SSA [m 2 /g] Specific Gravity pHpzc Cost (2010 $/lb) Activated alumina (commercial alumina) 1.4 2.41 571 4 2.67 0.01 8.83 0.04 2 4 AOCM c (clay based aluminum oxide coated media) 0.85 2 282 22 2.29 0.04 4.61 0.06 3.5 5.5 Bioretention soil media (commercial mixture for Prince Geoges County, Maryland, P index = 25) 0.001 10 22.2 4.5 2.69 0.03 6.35 0.03 0.10 0.25 Expanded shale (commercial expanded shale) 2 10 3.90 1.03 1.91 0.02 9.84 0.10 0.75 1.00 0.45 4.5 4.71 1.04 Iron coated perlite (iron oxide coated perlite) 0.3 0.85 41.8 2.2 1.24 0.03 7.91 0.08 1.50 2.25 Tire crumb B&G (green roof mixture of mainly tire crumb) 0.85 10 30.5 6.2 1.19 0.04 8.58 0.08 0.12 1.30 0.5 4.75 35.5 9.2 Zeolite Perlite GAC (commercial mixture) 2.29 9.65 93.7 16.9 2.02 0.01 7.51 0.04 0.65 0.95

PAGE 82

82 Table 3 3 Freundlich isotherm p arameters All R2 in this study are greater than 0.96. Media Size [mm] Matrix KF n Reference Activated alumina 1.4 2.41 DI 5.49 0.56 This study AOCM c 0.85 2 1.35 0.33 Bioretention soil media 0.001 10 0.18 0.39 Expanded shale 2 10 0.06 0.53 0.45 4.5 0.15 0.54 Fe coated perlite 0.3 0.85 6.72 0.64 Tire crumb 0.85 10 0.017 0.51 0.5 4.75 0.033 0.44 Zeolite Perlite GAC 2.29 9.65 0.051 0.59 Acid mine drain sludge --DI 17.0 0.25 Wei et al., 2008 Acid mine drain sludge 0.15 4 DI 3.26 4.38 0.24 0.50 Sibrell et al., 2009 Alum sludge 1.25 (d60) DI 1.03 0.48 Mortula et al., 2007 Blast furnace slag 0 0.18 DI 0.29 11.11 0.06 0.47 Kostura et al., 2005 Blast furnace slag 0.01 15 DI 6.37 1.74 Lu et al., 2008 Fe5HO84H2O 0.0035 WWTP effluent 19.4 0.05 Kang et al., 2003 KAl 3 (SO 4 ) 2 (OH) 6 0.09 0.15 DI 1.60 0.83 zacar, 2003 Iron oxide tailings --DI 3.59 0.19 Zeng et al., 2004 Lignocellulose media --DI 8.34 0.20 Kim et al., 2006 Recovered ochre --DI 3.5 0.48 Heal et al., 2004 Steel furnace slg 0.01 15 DI 1.71 1.72 Lu et al., 2008 Zirconium hydroxide 0.15 0.23 DI 17.4 0.33 Chitrakar et al., 2006 Seawater 43.5 1 wastewater 14.9 0.09 ZnCl2 activated coir pith carbon 0.25 0.5 DI 1.61 0.33 Namasivayam & Sangeetha, 2004

PAGE 83

83 Table 3 4 Summary of metals leaching capacity of m edia Metal Activated alumina Bioretention soil media Expanded shale a Expanded shale b Fe coated perlite Tire crumb a Tire crumb b ZPG AOCMc Al 1090 (20) a 972 (23) 67 (2) 120 (2) 27 (3) 156 (6) 260 (6) 137 (5) 285 (4) Ba 0.8 (0.07) 14 (0.1) 2 (0.1) 6 (0.1) 1 (0.1) 7 (0.1) 3 (0.1) 20 (0.1) 4 (0.1) Ca 22 (1) 456 (7) 1180 (20) 1600 (20) 599 (8) 576 (10) 876 (10) 2790 (20) 78 (0.6) Cd 0.4 (0.07) 0.4 (0.06) 0.3 (0.08) 0.4 (0.04) 0.6 (0.08) 0.4 (0.01) 0.4 (0.06) 0.4 (0.09) 0.3 (0.06) Cu 1 (0.1) 0.9 (0.08) BDL 1 (0.1) 3 (0.1) 0.6 (0.06) 0.6 (0.08) 0.9 (0.08) 1 (0.2) Fe BDL 657 (22) 15 (1) 24 (1) 3490 (50) 125 (6) 215 (5) 24 (2) 0.2 (0.1) Mg 17 (1) 97 (1) 116 (1) 119 (1) 78 (1) 88 (0.5) 258 (4) 328 (4) 6 (0.1) Mn 0.08 (0.008) 40 (0.4) 0.4 (0.02) 0.8 (0.01) 15 (0.3) 5 (0.1) 3 (0.1) 1 (0.1) 0.5 (0.01) Pb 10 (2) 9 (2) 6 (3) 3 (2) 12 (5) 5 (0.5) 7 (3) 11 (3) 6 (2) Zn BDL 2 (0.2) 7 (0.1) 4 (0.2) 1 (0.1) 17 (0.3) 53 (0.2) 3 (0.1) 13 (0.4) a: Values in parentheses are standard deviations; b: BDL Below Detection Limit; As, Cr, Ni are BDL.

PAGE 84

84 Initial DP Concentration [mg/L] 0 3 6 9 12 15 DP Removed (%) 0 20 40 60 80 100 AOCM c Activated alumina Iron coated perlite Expanded shale b Expanded shale a Tire crumb a Tire crumb b Zeolite-perlite-GAC Bioretention media Figure 3 1 Comparison of P reduction by media of granular filter m aterials

PAGE 85

85 Ce [mg/L] 0 10 20 30 40 50 Qe [mg/g] 0.00 0.05 0.10 0.15 0.20 0.25 Tire crumb a (0.85-10 mm) Tire crumb b ( 0.5-4.75 mm) Freundlich Model Ce [mg/L] 0 10 20 30 40 50 0.0 0.3 0.6 0.9 1.2 1.5 Expanded shale a (2-10 mm) Expanded shale b (0.45-4.5 mm) Freundlich Model F igure 3 2 M easured and Freundlich modeled isotherms for P r emoved by Tire crumb B&G and Expanded s hale

PAGE 86

86 Ce [mg/L] 0 10 20 30 40 50 60 Qe [mg/g] 0.0 0.9 1.8 2.7 3.6 4.5 AOCMcExpanded shale b Biorentention soil media Zeolite-perlite-GAC Tire crumb b Activated alumina Fe-coated perlite Figure 33 Measured and Freundlich m odeled isotherms for P removed by each g ranular filter m aterial

PAGE 87

87 Ce [mg/L] 0.1 1 10 100 Ku [L/g] 0.001 0.01 0.1 1 10 100 AOCMcFe-coated perlite Activated alumina Bioretention soil media Zeolite-perlite-GAC Expanded shale a Expanded shale b Tire crumb a Tire crumb b Figure 3 4 Comparison of Ku d istribution for m edia of g ranular filter m aterials

PAGE 88

88 Al, FE, Ca Content [mg/g] 0 1 2 3 4 Ku [L/g] at Ce = 0.1 mg/L 0.01 0.1 1 10 100 Al Fe Ca Ku AOCM c Activated alumina Bioretention soil media Fe-coated perlite Zeolite-perlite-GAC Expanded shale a Expanded shale b Tire crumb B&G a Tire crumb B&G b Al, Fe, Ca Content [mg/g] 0 1 2 3 4 Ku [L/g] at Ce = 0.03 mg/L 0.01 0.1 1 10 100 Al Fe Ca Ku Figure 3 5 Comparison of Ku v alues at Ce = 0.1 mg/L (USEPA, 1986) and Ce = 0.03 mg/L (Genz et al., 2004; Oguz, 2005) for granular filter m aterials

PAGE 89

89 CHAPTER 4 MASS TRANSFER KINETICS O F GRANULAR FILTER MATERIALS FOR PHOSPHORUS IN AQUEOUS MATRICES Background Recent studies of P loadings to aquatic systems focused on anthropogenically induced fluxes to aquatic systems demonstrate that such loadings accelerate eutrophication (Karaca et al., 2004; dm et al., 2007; Yu et al., 2008). Point sources such as conventional municipal wastewater treatment plants (WWTP) and nonpoint sources such as urban runoff along with sources from agricultural land uses are sources of nutrients that result in receiving water im pairments (USEPA, 1996). While the technology exists (although not always implemented) to identify and treat point sources such as discharges from WWTP, control of nonpoint sources has not kept pace with point source control, in part because of the much g reater challenges associated with diffuse, stochastic and variable loadings and with technology that is still developing. Many physical, chemical and biological control technologies applied to nonpoint loads can be disadvantaged by unfavorable cost/benefi t relationships, low or variable effectiveness, or restrained by operational instability (for example desorption from media systems or lack of maintenance) and lack of sustainability (Karaca, 2004; Sansalone, 2005; Huang et al., 2008). One technology that has been deployed to control P in runoff and wastewater is granular filtration media systems for PP; with recent incarnations as adsorptive media to facilitate DP control. A category of the adsorptive media incarnations have included filter substrates that range from common non engineered materials (Sakadevan and Bavor, 1998; Johansson, 1999; Bubba et al., 2003; Hsieh and Davis, 2005; Davis et al.,

PAGE 90

90 2006; Hsieh et al., 2007) to industrial by products or wastes (Kostura et al., 2005; Drizo et al., 2006; Lu et al., 2008) in order to provide reduce the initial cost of media systems. These studies have demonstrated that many of these media do not have sufficient capacity for successful deployment to control DP (Bubba et al. 2003; Forbes et al. 2004, Kim et al. 2008, Sansalone and Ma 2009). Recent studies have also demonstrated that engineered coatings and media substrates can significantly improve the adsorptive capacity of granular media as compared to nonengineered media (Ayoub et al., 2001; Arias et al., Li u et al. 2005a; Ko et al., 2007; Boujelben et al., 2008; Sansalone and Ma 2009). While recent studies have illustrated the equilibrium capacity of oxides of Fe and Al for phosphate (Arias et al., 2006; Boujelben et al., 2008; Sansalone and Ma 2009), little research has focused on examining the role of the substrate in combination with the oxide for mass transfer of DP. Furthermore, if media and media substrates require comparison, testing that allows comparison between media for a given media phenomena a re required; specifically overall mass transfer rate for DP. When comparing media phenomena under equivalent conditions a controlled surrogate matrix is required. The most common controlled surrogate aqueous matrix is deionized (D.I.) water which provides a consistent and reproducible matrix for media testing and facilitating media comparisons subject to different controlled loadings (Namasivayam and Sangeetha, 2004; Liu et al. 2005b, Arias et al., 2006; Georgantas and Grigoopoulou, 2007; Boujelben et al ., 2008). However, while such a consistent and reproducible matrix provides a vehicle for media comparability, studies have also shown that the equilibrium capacity of DP can be nonrepresentative when comparing results from a DI matrix and an actual aque ous matrix

PAGE 91

91 subject to similar pH and DP concentrations (Millero et al., 2001; Chen et al., 2002; Shilton et al., 2005; dm et al., 2007, Sansalone and Ma 2009). Specific to this study, a wastewater matrix is more complex than a DI water matrix with compe ting and synergistic ions. Actual rainfall runoff subject to advective, diffusive and reactive multi phase phenomena at much shorter interaction times than wastewater matrices in a WWTP illustrate a higher level of complexity as a result of competitive interactions, variable species and a highly heterodisperse particulate matter (PM) phase (Dean et al., 2005; Sansalone, 2005). Monitoring and modeling of media phenomena with a surrogate matrix such as DI water can provide a comparative basis for media subj ect to similar actual matrix chemistry such as pH, redox, and DP concentrations. However, resulting media phenomena and model parameters such as overall mass transfer rates from a DI surrogate matrix may be less than representative of actual aqueous matri ces such as rainfall runoff and wastewater for a given media. O bjectives Mass transfer processes between the aqueous and solid media phase can be described and quantified macroscopically through equilibria, kinetics, breakthrough and leaching. Equilibria approaches provide distribution coefficients and provide information on the final state of a reaction. However, in order to evaluate the timedependent profile of mass transfer phenomena, examine mechanisms, and develop design and model parameters for media based technologies, knowledge of kinetics is needed. There is a significant body of literature on equilibrium capacity for DP and metals for nonengineered and engineered media. However, kinetics and overall mass transfer rates for DP are needed to compare media deployed in control technologies (unit operations and processes) for DP; and are compared herein. Additionally, it is hypothesized that

PAGE 92

92 the role of media substrates for media engineered with oxides for overall mass transfer can be significant yet has not been illustrated in the literature for DP; and this role is examined herein. It is hypothesized that media testing using actual uncontrolled and complex runoff or wastewater matrix, as compared to the potential nonrepresentativeness of a cont rolled and reproducible DI surrogate matrix, produces disparate overall mass transfer rate parameters. Therefore in addition to comparing media (three of AOCM, three of UCM and six commercially available media) the role that an aqueous matrix imparts on overall mass transfer rates and capacity is examined. Materials and Method s Media Three aluminum oxide coated media substrates : AOCMc ( fired clay substrate), AOCMp (pumice substrate), AOCMpcc (Portland cement concrete substrate ) and their uncoated media (U CM) substrates are monitored and modeled. Clay and concrete substrates are crushed into two selected size gradations. A ctivated alumina; bioretention soil media; expanded shale; iron coated perlite; tire crumb (B&G) and z eolite p erlite GAC (volume ratio: 5:4:1) are also monitored and modeled in this study as supplied or deployed in unit operations and processes Table 4 1 summarizes the media characteristics. AOCMc and AOCMp were prepared by applying aluminum as surface coating. Al (III) solution was added to the raw substrate contained in a Pyrex tray so that the substrate was just submerged. This Pyrex tray was then put in the oven and temperature was kept at around 200 oC. At least 24 hours were needed to fully dry and coat the substrate. Later, the aluminum coated sorbents was washed with D.I. water to remove the surficial fine materials. AOCMpcc was prepared by integrating aluminum

PAGE 93

93 solution into the cementitious matrix as the aqueous admixture. The mass ratio of aqueous Al( NO3)3 solution (0.5 M) to cement is 1:1. This cementitious slurry was then cured and size graded in the same way as raw substrate preparation. Chemicals A 1000 mg/L phosphorus stock solution utilized is prepared by dissolving anhydrous potassium dihydrogen orthophosphate (KH2PO4) powder (Analytical reagent grade, Fisher Scientific) into DI water for the synthetic runoff. The stock solution is diluted as required for the synthetic runoff. Dilute HCl and NaOH solution (0.02N) were used to adjust the pH. Ioni c background was fixed as 0.01M as KCl. All measurements ( Eaton et al., 1998, summarized in Table 4 2) are duplicated and repeatability is assured by controlling experimental error within 5%. Experimental Matrices and Strategy Batch kinetic are carried out utilizing three different aqueous matrices: synthetic runoff, actual rainfall runoff and secondary wastewater as upflow sand filter effluent from the University of Florida Advanced Wastewater Treatment (AWT) facility. The chemistry of each aqueous ma trix is summarized in Table 4 3. Given that a DI aqueous matrix is consistent and reproducible surrogate the entire suite of media is monitored in synthetic runoff, with a DP concentration of 1.0 mg/L. This DP concentration is chosen based on the event based concentrations of DP during actual rainfall runoff events (Sansalone and Ma 2009). The AOCM and UCM forms are further examined in actual rainfall runoff and wastewater with their measured DP concentrations (2.7 mg/L and 11.1 mg/L, respectively). It is noted that the University of Florida catchment (Lake Alice watershed) where field monitoring is conducted is an asphalt paved surface parking facility with raised vegetated islands and therefore high biogenic loadings including

PAGE 94

94 nutrients. While the wastew ater is treated to AWT levels and effluent is reclaimed for campus wide irrigation, N and P concentrations remain elevated in the effluent. Finally, the role of each aqueous matrix on media mass transfer is investigated. In this objective, the DP concentr ation of synthetic runoff and wastewater is adjusted to the same level as that of runoff (2.7 mg/L) to eliminate the interference of different driving force. Specifically, a synthetic runoff with initial concentration of 2.7 mg/L was prepared following the same procedure as described above for 1.0 mg/L; wastewater sample with initial P concentration of 11.1 mg/L was pretreated with AOCMc to remove the excessive phosphorus. It has also been validated that no major chemistry of the wastewater sample which can influence media performances was altered after the pretreatment 2+4 2 -] < 8%). Batch Kinetic Study Batch kinetics is an experimental technique which is considered representative of more realistic conditions and can be convenientl y employed to determine the sorption parameters (Liu et al., 2005). In this study, m ass transfer rates investigated by the batch kinetic s experiments are facilitated using a recirculating flow through reactor Here, a short column (40 mm i.d., 90 mm length, Teflon PFA column) packed with test media served as the reactor. The sorbent solution ratio was fixed at 20 g to 2 L for all the tests and comparisons were made on equal mass base. The experimental set up is illustrated in Fig ure 4 1. The ionic strength is set at 0.01 M with KCl and pH was adjusted to 7 in synthetic runoff For the axial filter column t he surface loading rate ranges within 20 to 60 L/minm2. In this study, surface loading rate of 40 L/minm2 are chosen and kept throughout the experiments, based on the database of peak flow of 1year return event ( Sansalone and Teng, 2005).

PAGE 95

95 Duplicate 10 mL samples are collected from the sampling port shown in Figure 4 1 The time intervals of sampling were predetermined according to the preliminary tests of media kinetics. The sampled solution is filtered through 0.45 m filter The DP (as phosphate) concentration is determined using a spectrophotometer (HACH, DR5000). DP mass transfer to the media from the aqueous matr ix at each time intervals, qt [mg/g], and at equilibrium qe [mg/g] is computed. W C C V qt t) (0 (1) W C C V qe e) (0 (2) In these expressions C0 is the initial concentration [mg/L] ; Ct and Ce are the solution concentrations at time t and equilibrium [mg/L], respectively; V the volum e of the solution [L] and W the dry weight of media used [g]. Overall Mass Transfer Rate Modeling Modeling of kinetics can be carried out by empirical or fundamental models. In this case a 2nd order potential driving model is illustrated (Liu et al. 200 5a). The elementary secondorder reaction with different reactants can be generalized. P B Ak (3 ) The differential equation and analytical solutio n with initial condition of [A] = [A]0 and [B] = [B]0 at t = 0 are as follows. ] ][ [ ] [ B A k dt A d (4) ) ( exp0 0 0 0 0 0A B kt B A B A A (5 )

PAGE 96

96 In these expressions k is the reaction rate constant (M1s1), t is the reaction time (s), and [A]0 and [B]0 are initial concentrations of reactants A and B at t = 0, r espectively. For a heterogeneous solution (DP concentration at the near surface differs from bulk solution), the surface reaction mechanism with a hydrous oxide and may be depicted as surface complexation reactions. In developing the mathematical descript ion of the overall mass transfer process, certain assumptions are made. There can be various rate limiting steps, such as chemical reaction, film diffusion and intraparticle diffusion. For the kinetic model of this study, an overall rate covering all the processes is utilized to simplify the presentation of kinetics from aqueous solution to media. This overall rate is considered to be related to the sorption potentials (chemical or electrical or other energy forces of the diffusion processes) of surface si tes and sorbates. Since only differences in potential are ever physically meaningful, and the differences of potential can be from the variation of space or time, the potential at equilibrium ( t = ) can be set to zero or as a standard potential. Thus the potential for any time, t, is t ( t is the absolute potential at time t, and is the absolute potential at t = ). Concentration differences between the bulk solution and media surface sites are utilized at time, t, and at equilibrium for model com putations. Potential differences are assumed to be correlated to the difference in bulk solution concentration of DP and number of surface sites. Moreover, equilibria follows the Freundlich isotherm and this has been confirmed previously (Sansalone and Ma 2009). With these assumptio ns the second order rate, eq. (5) can be correlated to differences in DP concentrations and available surface site number between time t and equilibrium. Using a similar hypothesis, kinetics of divalent metal adsorption onto oxi de-

PAGE 97

97 coated media is successfully modeled (Liu et al, 2005a). A potential driving rate differential equation based on equation ( 5 ) can be represented as follows. e t t e tC C S S k dt dC ( 6 ) Here St is the number of active sites occupied on the media at time t and Se is the number of active sites occupied at equilibrium on the media. St and Se can be represented as follows. a C C St t 0 ( 7 ) a C C Se e 0 ( 8 ) Here C0 is the initial DP concent ration, Ct is the DP concentration at time t Ce is the DP concentration at equilibrium, and a is the sorbent/solute ratio. It has been shown that the integrated potential driving model rate equation can be linearized. 2 0 0 0 e e tC C k a C C t C C t (9 ) The overall mass transfer rate constant k and equilibrium concentration Ce can be determined by plotting t/(C0Ct) against t (Liu et al. 2005a) Additionally, substitute V W a and Eq. (1) and (2) into Eq. (9), we can get: 21e e tkq q t q t (1 0 ) qe and qt [mg/g] is the amount of phosphorus transferred at equilibrium and time t respectively.

PAGE 98

98 To compare monitored and modeled result, RMSE (Root Mean Square Error) and chi square tests are utilized and calculated for each kinetics run. n q E q Q RMSEn i i i 1 2)] ( ) ( [ (1 1 ) In this expression Q(qi) is the observed (monitored) value and E(qi) is the expected (modeled) value. The fit between monitored and modeled is also evaluated with a chi 22 measures the difference between expected and observed values and is a 2 is defined in the following manner and a 2 indicates a better fit: n i i i iq E q E q O1 2 2) ( )] ( ) ( [ (1 2 ) Intra media diffusion model While the 2nd order potential driving model is intended to simulate the overall mass transfer rate the model does not separately identify mechanisms. For media that has internal porosity there is intramedia diffusion that occurs during the overall mass transfer process. From a system design and modeling point of view, a lumped or overall mass transfer rate is appropriate (zacar, 2003) but may not clearly illustrate the mechani sm of diffusion that occurs as part of the overall mass transfer. Hence, a intra media model is also utilized for the purpose of identifying diffusion. A functional relationship common to most intramedia diffusion (qt) phenomena is that solute mass transfer rate varies proportionately with the squareroot of time (Alvin et al, 2010). ) (5 0t f qt (1 3 )

PAGE 99

99 The rate parameter for intraparticle diffusion (kp) can be determined 5 0t k qp t (1 4 ) kp is the intra media diffusion rate constant (mg/g 0.5). A plot of qt versus t0.5 may illustrate multilinearity, indicating multiple rate controlling steps occur during the overall mass transfer Results and Discussions Tests in Synthetic Runoff (C0 = 1.0 mg/L) Kinetic s are determined for three AOCM forms and their UCM substrates. AOCM and UCM media parameters and synthetic aqueous parameters are summarized in Table 4 1 and 4 3 respectively. Overall mass transfer kinetics of DP from synthetic runoff to the media are shown in Figure 4 2 illustrating that equilibrium is approached in approximately 90 minutes for AOCM forms and 150 minutes for UCM forms without Al Ox. The kinetic data are simulated with the 2nd order potential driving model and Table 4 4 lists the model parameters 2 and RMSE for each media form Results demonstrate that compared with UCM, the mass transfer kinetics for all AOCM forms were much greater. With the exception of the PCC substrate the equilibrium capacity of the Al Ox media was significantly higher than the UCM substrate indicating that the Al Ox improves the media capacity The mass transfer of phosphates to hydrous aluminous oxide is through ligand exchange to surface hydroxyl groups (Goldberg and Sposito, 1985; Genz et al., 2004; Kim and Kirkpatrick, 2004; Arias et al. 2006; Yang et al., 2006; Georgants and Grigoropoulou, 2007). Therefore it is hypothesized that an increase of pH should be observed during mass transfer given the release of surface hydroxyl groups for all AOCM forms The higher kinetics and capacity of the AOCM as compared

PAGE 100

100 to the UCM substrate is a result of the Al Ox ligand exchange and the number of surface sites providing such exchange, and therefore higher equilibrium capacity as compared to UCM of the same substrate. While AOCMpcc kinetics and capacity were not markedly higher tha n other AOCM forms, the permeable concrete media substrate (UCMpcc) exhibited much higher kinetics and capacity as compared to the permeable clay substrate (UCMc) and permeable pumice substrate (UCMp). This is attributed to the high content of CaO in concr ete substrate; as shown for increasing CaO content in fly ash, slag and conventional Portland cement (Agyei et al., 2002). The affinity of phosphate for CaO, and the resulting formation of insoluble CaP precipitation indicate that surface precipitation a nd ligand exchange combine to provide higher kinetics and equilibrium capacity However, the relative significance of each mechanism is a function of the media composition combined with the aqueous matrix ; which will be discussed later in the results Fig ure 4 2 and Table 4 4 illustrate that there is not a marked difference between the two size gradations of clay and PCC substrates Usually, if the mass transfer is chemically rate controlled, the rate constant is independent of particle diameter (Ho and Mc Kay, 1999), suggesting that chemisorption is the dominant mechanism with surface precipitation as a co mechanism. Since media size (a surrogate for SSA) does not provide a marked difference for kinetics or capacity for AOCMc and AOCMpcc the balance of this study utilized only the 2 ~ 4.75 mm gradation. Plot f of Figure 4 2 superimposes AOCM (2 ~ 4.75 mm) results from plots a, b and c to illustrate kinetics and equilibrium capacity. Although AOCMp displayed slower kinetics and lower capacity no marked vari ance among AOCMc, AOCMp and AOCMpcc is observed, in contrast to a UCM comparison

PAGE 101

101 from plots a, b and c. Results with the synthetic matrix illustrate that the Al Ox coating promoting ligand exchange provides the dominant mass transfer mechanism generating comparable kinetics and capacity results; despite differences illustrated for mass transfer by each individual substrate. In addition to AOCM, media utilized in stormwater BMP deployments are monitored and modeled for overall mass transfer kinetics using the synthetic runoff matrix. These media as summarized in Table 4 1 include activated alumina, bioretention soil media, expanded shale (a, b) iron coated perlite, tire crumb (a, b) (Black&Gold) and z eolitep erlite GAC The kinetics of each media including AOCM (2 4.75 mm) is shown in Figure 4 3. Results illustrate a wide range of kinetics; ranging from the highest for AOCM forms to essentiall y no different from 0 ( p ) for tire crumb (B&G) and zeoliteperlite GAC. While the r eactiv ity of the activated alumina and iron contributes to their relatively high kinetics and equilibrium capacity the fine gradation of the iron perlite limits the range of hydraulic applicability and the potential for ferric ferrous reduction under anaerobic conditions limits applicability to aerobic environments, while the potential for leaching is potentially limiting for both media. Table 4 4 summarizes the 2nd order model parameters for the conditions of the synthetic matrix. The model did not fit the tire crumb or zeoliteperlite GAC media since the mass transfer rate to, and from, these media were essentially equal resulting in media capacity that is not diff erent from 0 (p ) Intra media Diffusion M ultiple transport processes or rate limiting steps can exist with the overall mass transfer process. The overall mass transport process generally involves: (i) mass transfer of sorbate from the bulk phase to media surface; (ii) a reaction mechanism at

PAGE 102

102 the surface site or sites ; (iii) intra media diffusion, which includes diffusion of sorbate occluded in micropores (pore diffusion) and along porewall surfaces (surface diffusion). The combination of pore and surface diffusion within the immediate region are referred to as intramedia diffusion. The reaction at the surface (step ii) is usually rapid and not rate limiting; while usaually step (i) is rate limit ing for a system with poor mixing, low sorbate con centration, high affinity to the adsorbent and small particles (Choy et al., 2004; Alvin et al., 2010). Therefore, the intramedia diffusion is hypothesized as the rate determining step. According to Eq. ( 14) if qt as a function of t1/2 is linear then intra media diffusion is rate limiting Previous studies have shown that this function will not have a 0intercept if an external surface adsorption o r instantaneous adsorption stage exists (Hameed and Daud, 2008; Khaled et al., 2009). Plotting such a function can therefore illustrate multi linearity and the mass transfer process can be broken down into two or more steps: the first is the external surface adsorption rate ; the second rate is the slower mass transfer stage with intramedia diffusion control ling the rate; and the third stage is the final equilibrium stage. The slope of the second rate is called as intra media diffusion rate constant, kp (zacar, 2003; Karaca et al., 2004; Kostura et al., 2005). Figure 4 4 illustrates the root time plot s for DP mass transfer for the media exhibiting overall kinetics that are not significantly different from zero and tested in synthetic runoff (C0 = 1.0 mg/L) As can be seen, the external surface adsorption (first step) is completed in less than 5 min utes (when the first sample was taken). The other media of essentially zero capacity did not exhibit this behavior After 5 minutes t he intra-

PAGE 103

103 media diffusion control is attained and continued for times that ranged from 20 to 40 min utes depending on the media after which the final equilibrium stage is started. Table 4 5 summarizes the rate parameters of kp1 and kp2, with kp1 as the intra media diffusion rate constant and kp2 as the rate constant of the equilibrium stage. The 2nd order model and intramedia model simulate the monitored data. While the diffusion model illustrates discrete mass transfer stages and mechanisms t he 2nd order model simulates the overall mass transfer rate continuum over the entire range of times with a single set of parameters applied to the entire mass transfer process. Since the focus is to illustrate the overall mass transfer rate and rate parameters for modeling the 2nd order potential driving model is utilized for actual rainfall runoff and wastewater matrices Kinetics in RainfallR unoff and Wastewater Matrices Studies have illustrated that water chemistry influences media capacity (Shilton et al., 2005; dm et al., 2007, Sansalone and Ma 2009). Hypothesizing that the selection of the aqueous matrix also influences the kinetics as well as equilibrium capacity for a media, AOCM is also tested in actual rainfall runoff (C0 = 2.7 mg/L for DP ) and secondary effluent from an AWT (C0 = 11. 1 mg/L for DP ) with the water chemistry of these matrices listed in Table 4 3. The experimental resul ts are presented in Figure 4 5 and Figure 4 6 respectively. Kinetic s data are simulated with the 2nd order model and the rate constants for each media for each aqueous matrix are summarized in Table 4 6 AOCM is compared with UCM in plots a, b, c in Figure 4 5 and 6. While comparative trends between AOCM and the UCM substrate are similar to trends obtained in the synthetic runoff matrix under the same monitored conditions the magnitudes of the model parameters are different Giv en that the actual rainfall runoff had a higher DP concentration (2.7 mg/L) as compared to the synthetic runoff (1.0 mg/L) the equilibrium

PAGE 104

104 capacity is comparably higher, but lower than the initial concentration ratio of 2.7 suggesting the role of competiti ve interactions on equilibrium capacity. In contrast, when comparing the rate constants in Table 4 4 and 4 6 the results clearly indicate a significant decrease in the overall mass transfer rate constant for actual rainfall runoff when compared to synthet ic runoff despite the DP enrichment in actual rainfall runoff. Despite a clear influence of the aqueous matrices, Al Ox coating exhibited a significant influence on media capacity in particular for clay and pumice substrates with ligand exchange as dominant mechanism and surface precipitation a negligible mechanism as compared to AOCMpcc. Similar results are shown for wastewater as summarized in Figure 4 6. Figure 4 5 d and 4 6 d overlay each AOCM kinetics profile thereby illustrating the role of the subst rate While the kinetics of AOCMc and AOCMpcc are still similar to each other, albeit with commensurately different equilibrium capacities and kinetics as compared to the synthetic runoff, there is a marked and significant difference for AOCMp, especially in the wastewater matrix. AOCMp exhibited slower kinetics and lower equilibrium capacity. These results indicate that although an effective coating is critical to enhance substrate behavior coupled physic o chemical characteristics of the substrate and imposed water chemistry also impact media capability, as illustrated with these actual and more complex aqueous matrices. In order to further compare and investigate the influence of the aqueous matrix on overall mass transfer kinetics, synthetic runoff, ac tual ra infallrunoff and wastewater are batched at the same initial DP concentrati on of 2.7 mg/L. Thus, the effect of different driving forces for reaction and diffusion caused by different C0 levels is

PAGE 105

105 excluded. Figure 4 7 illustrates the comparison among various water matrices as well as pH variation along with the contact time for AOCMc, AOCMp and AOCMpcc, respectively. Table 4 7 summarizes the 2nd order model parameters for all aqueous matrices with an equal DP (C0 = 2.7 mg/L) As identified earlier, reactions between aqueous phosphate and Al Ox are ligand exchange between phosphate ions (H2PO4 or HPO4 2 -) and the surface hydroxyl group (OH-), resulting in OHrelease into s olution. As shown in Table 4 7 the change in pH ( pH) increases across the period of mass transfer yet is different between synthetic and actual matrices (rainfall runoff and wastewater) with an initial pH of all the matrices in the range of 7 .0 as shown in Table 4 3 Th e differences in the pH be tween synthetic and actual rainfall runoff and wastewater matrices is in part a result of higher buffer capacity in these latter two matrices due to the presence of primarily HCO3 -. Figure 4 7a overlays the kinetics of AOCMc in each matrix Kinetic rates follow the order of synthetic runoff, rainfall runoff and wastewater The actual matrices introduce the presence of competitive ions such as SO4 2 which compete for the active sites and decrease the media kinetics rates and capacity. However the impact i s insignificant (p 0.05) for AOCMc kinetics and capacity. Similar increases in pH (as pH) are obtained for AOCMp, as presented in Table 4 7 In contrast to Figure 4 7a, 7b illustrates significantly lower (p and apparently capaci ty when comparing synthetic and actual matrices. The time to equilibrium is nearly doubled in the actual matrices compared to AOCMc. The kinetic rates of AOCMp in these complex matrices are significantly less than AOCMc.

PAGE 106

106 D ue to the high content of CaO in cement concrete substrate higher increases in pH (as pH) are obtained for AOCMp cc, as presented in Table 4 7 In contrast to Figure 4 7a, and 7b; 7c illustrates lower kinetic rates for AOCMpcc in the actual rainfall runoff matrix. In contrast to AOCMc and AOCMp; AOCMpcc exhibit s higher kinetic rates in synthetic runoff and wastewater than in rainfall runoff. It has been reported that both SO4 2 and Ca2+ can influence the P removal by AOCM (Sansalone and Ma, 2009). SO4 2 decreases P sorption due to the competitive effect on ligand exchange while Ca2+ increases P sorption through promotion of surface precipitation. Thus, t he presence of both SO4 2 and Ca2+ in one matrix can exert contrary effects on the kinetic rates The relative significance of eac h effect is determined by the dominant mechanism involved, i.e., if ligand exchange is the major mechanism, competitive effect will be more evident; if surface precipitation is dominant, promotion effect will be observed. Furthermore, mechanism transition is a function of media composition combined with the aqueous matrix H igh concentration of Ca2+ in wastewater matrix ( 56 mg/L) high content of CaO, alkaline condition in the solution all facilitate Ca3(PO4)2 precipitation and overwhelm the SO4 2 competiti on effect. As a result the surface precipitation of Ca3(PO4)2 is enhanced in wastewater as compared to rainfall runoff. Summary Overall mass transfer rates of DP from a series of aqueous solutions by a suite of granular filter media in a re circulating reactor is monitored and modeled. The m edia include three Al Ox coated media: AOCMc (clay based), AOCMp (pumice based), AOCMpcc (Portland cement concrete based) and their uncoated media (UCM) substrates ; and a series of commercially avail able media: activated alumina;

PAGE 107

107 bioretention soil media; expanded shale (a, b) ; iron coated perlite; tire crumb (a, b) (B&G) and z eolite p erlite GAC Kinetic rates are examined in a synthetic runoff matrix, an actual rainfall runoff matrix, and a secondar y effluent wastewater matrix. Kinetics parameters are developed based on simulation of monitored data with a 2nd order potential driving model for overall mass transfer and an empirical model to identify intramedia diffusion as a rate limiting mechanism. Compared to the UCM controls for each AOCM, the Al Ox coated media achieved significantly higher kinetic rates and equilibrium capacity as irrespective of the aqueous matrix The role of the AOCM substrates did not illustrate a significant effect in a DI matrix While the kinetic rates for the commercially available media were all below that of AOCM forms in a DI matrix, these media exhibited varied capacity and decreased in the order from iron coated perlite, ac tivated alumina, expanded shale and bioretention soil media. Tire crumb (B&G) media used for green roof applications and z eolite p erlite GAC mass transfer rates to and from were equal and therefore net equilibrium capacity was not significantly greater than 0. For the DI matrix the 2nd order potential driving model simulated the overall mass transfer rate and the intra media diffusion model identified the rate limiting diffusion of the permeable AOCM and other media. The DI matrix is a controlled and reproducible matrix that allows comparison of media on an equivalent basis but a basis that may not represent the complex interactions of actual matrices such as rainfall runoff and wastewater. In order to examine the k inetic rate s of D P mass transfer for actual aqueous matrices, municipal wet weather flow (rainfall runoff, C0 = 2.7 mg/L) and dry weather ( secondary wastewater, C0 = 11. 1 mg/L) matrices are tested with AOCM and UCM

PAGE 108

108 Again each AOCM out performed the UCM substrate. However, while the type of AOCM substrate had an insignificant effect in the synthetic matrix, AOCMp kinetics rates and equilibrium capacity are significantly lower than AOCMc and AOCMpcc in both rainfall runoff and wastewater matrices. The choice of an aqueous matrix impacts resulting media parameters as developed from the 2nd order model. Therefore to investigate the effect of the aqueous matrix the DP concentration in each matrix (synthetic and actual) are adjusted to that of actual rainfall runoff (2.7 mg/L) to eliminate the interference of different driving forces. The presence of SO4 2 and Ca2+ in the actual matrices can exert contrary effects on phosphate mass transfer as SO4 2 competes for the active sites with phosphate ions through ligand exchange and Ca2+ enhances the surface precipitation of phosphate as Ca3( PO4)2. Results demonstrated that the competition effects of SO4 2 dominates in the rainfall runoff matrix and decreases the AOCM overall mass transfer rate. While this competitive effect is observed for AOCMc and AOCMp in wastewater, surface precipitation overweighed the competition effect for AOCMpcc, due to the high concentration of Ca2+ (56 mg/L) in the aqueous matrix as well as the alkaline condition in aqueous phase created through contact with concretebased media, resulting in a faster kinetics. Nomenclature [A]0: Boundary condition of reactant A at t = 0 a: Sorbent solution ratio (g/L) AOCM: Aluminum oxide coated media AOCMc: Aluminum oxide coated media with clay as raw substrate AOCMp: Aluminum oxide coated media with pumice as raw substrate

PAGE 109

109 AOCMpcc: Aluminum oxide media with Portland cement concrete as raw substrate AlOx: Aluminum oxide [B]0: Boundary condition of reactants B at t = 0 B&G: Black & gold C0: The initial DP concentration (mg/L) Ct: DP concentration at time t (mg/L) Ce: DP concentration at equilibrium (mg/L) DI water: De ionized water DP: Dissolved phosphorus E(qi) Expected value k: Rate constant of 2nd order mass transfer ( g/mg ) kp: The intra media diffusion rate constant (mg/g 0.5) pHpzc: pH of point of zero charge PP: Particulate phosphorus qe: The amount of mass transfer at equilibrium (mg/g) qt: The amount of mass transfer at time t (mg/g) Q(qi): Observed value RMSE : Root Mean Square Error Se: Number of active sites occupied at equilibrium St: Number of act ive sites occupied on media at time t SSA: Specific surface area (m2/g) UCM: U ncoated media UCMc: Uncoated media with clay as raw substrate

PAGE 110

110 UCMp: Uncoated media with pumice as raw substrate UCMpcc: Uncoated media with Portland cement concrete as raw substr ate V: The volume of the solution (L) W: The dry mass of media used (g) WWTP: Wastewater treatment plants t: Absolute potential at time t : Absolute potential at t = 2: Chisquare

PAGE 111

111 Table 41. Media c haracteristics Media s ubstrate Size [mm] Specific g ravity pHpzc AOCM and UCM AOCMc 0.85 2.0 2.290.04 4.61 0.06 2.0 4.75 UCMc 0.85 2.0 2.310.05 4.20 0.08 2.0 4.75 AOCMpcc 0.85 2.0 2.670.04 11.84 0.16 2.0 4.75 UCMpcc 0.85 2.0 2.670.03 11.98 0.12 2.0 4.75 AOCM p 0.85 2.0 2.270.01 4.59 0.10 UCM p 2.0 4.75 2.290.02 4.61 0.10 Activated alumina 1.4 2.41 2.67 0.01 8.83 0.04 Bioretention soil media 0.001 10 2.69 0.03 6.35 0.03 Fe coated perlite 0.3 0.85 1.24 0.03 7.91 0.08 Zeolite Perlite GAC 2.29 9.65 2.02 0.01 7.51 0.04 Expanded shale a 2 10 1.91 0.02 9.84 0.10 b 0.45 4.5 Tire Crumb (Black & Gold) a 0.85 10 1.19 0.04 8.58 0.08 b 0.5 4.75 AOCM = Aluminum Oxide Coated Media UCM = Un Coated Media c = clay based PCC = Portland cement concrete based p = pumice based pHpzc = pH at point of zero charge Entries are mean standard deviation (n = 3)

PAGE 112

112 Table 4 2 Summary of standard method utilized in the experiments Measurement Method pH S.M. 4500 H + B Conductivity S.M.2510 Turbidity S.M.2130 Alkalinity S.M.2320 DP S.M. 4500 P E:Ascorbic Acid Method SO 4 2 SulfaVer 4 Method NO 3 Cadmium Reduction Method Cl Mercuric Thiocyanate Method Dissolved m etal S.M. 3030 B S.M.: Standard Method; DP: dissolved P

PAGE 113

113 Table 4 3 Water chemistry of e xperimental m atrix Matrix Synthetic r unoff Rainfall r unoff (Event on 01/21/10) Wastewater (Secondary e ffluent) pH 7.00 0.05 6.98 0.23 7.16 0.10 Conductivity [s/cm] 1025 56 224 8 782 12 Turbidity [NTU] 0.83 0.46 24.7 3.8 1.25 0.76 Alkalinity [mg/L] as CaCO 3 5.4 0.8 75 1 48 2 Ionic Strength [M] 0.010 0.002 0.003 0.002 0.010 0.004 DOC [mg/L] 0.3 0.1 37.8 4.1 12.4 2.8 Ca 2+ BDL c 28.67 1.34 55.98 2.10 Sum of Cation Charge [mmol] 10.05 1.83 5.49 Anion [mg/L] NO 3 N 0.45 0.08 3.25 0.10 1.45 0.05 SO 4 2 0 19.0 6.0 95 5 Cl a 360 10 9.63 0.54 108 3 HCO 3 3.3 0.7 45.8 1.8 29.3 2.1 PO 4 3 b 1.00 0.05 2.71 0.12 11.08 0.21 Sum of Anion Charge b [mmol] 10.33 1.73 5.78 Data are mean standard deviation a 0.01M KCl was used to adjust the ionic strength of synthetic runoff. b In synthetic runoff, PO4 3 was added as KH2PO4. c BDL: Below Detection Limit

PAGE 114

114 Table 4 4 Parameters of 2nd order potential driving model in synthetic runoff (C0 = 1.0 0.05 mg/L) Media Substrate Size [mm] q e [mg/g] k [g/mgmin] 2 RMSE AOCM and UCM AOCMc 0.85 2.0 0.13 0.76 0.036 0.009 2.0 4.75 0.12 0.62 0.012 0.009 UCMc 0.85 2.0 0.03 5.64 0.001 0.001 2.0 4.75 0.03 4.80 0.002 0.002 AOCMpcc 0.85 2.0 0.13 0.51 0.006 0.007 2.0 4.75 0.14 0.35 0.016 0.009 UCMpcc 0.85 2.0 0.13 0.26 0.027 0.010 2.0 4.75 0.14 0.21 0.026 0.009 AOCM p 0.85 2.0 0.12 0.45 0.018 0.010 UCM p 2.0 4.75 0.03 0.70 0.008 0.002 Activated alumina 1.4 2.41 0.12 0.40 0.013 0.008 Bioretention soil media 0.001 10 0.02 5.08 0.005 0.010 Iron coated perlite 0.3 0.85 0.13 0.40 0.005 0.052 Expanded shale a 2 10 0.03 0.85 0.007 0.002 b 0.45 4.5 0.09 0.48 0.016 0.006 2: Chisquare RMSE: Root Mean Square Error Insignificant net capacity for Tire crumb (a, b) and ZeolitePerlite GAC, for which rate of adsorption/desorption were equal (P ).

PAGE 115

115 Table 4 5 The sorption kinetics rate parameters in synthetic runoff (C0 = 1.0 0.05 mg/L) with water chemistry summarized in Table 4 2. Media Substrate Size [mm] k p1 [mggmin 0.5 ] RMSE1 k p2 [mggmin 0.5 ] RMSE2 AOCM c 2 4.75 0.034 0.0027 0.005 0.0021 AOCM p 2 4.75 0.034 0.0025 0.006 0.0040 AOCM pcc 2 4.75 0.027 0.0056 0.003 0.0031 Activated alumina 1.4 2.41 0.025 0.0043 0.004 0.0024 Bioretention soil media 0.001 10 0.002 0.0003 0.0002 0.0002 Iron coated perlite 0.3 0.85 0.025 0.0027 0.008 0.0032 Expanded shale a 2 10 0.005 0.0013 0.0004 0.0003 b 0.45 4.5 0.016 0.0054 0.0009 0.0006 AOCM in the size range of 2 4.75 mm Insignificant net capacity for Tire crumb (a, b) and ZeolitePerlite GAC, for which rate of adsorption/desorption were equal (P ).

PAGE 116

116 Table 4 6 The sorption kinetics parameters for AOCM and UCM and water chemistry matrices shown in Table 4 2. Media Rainfall Runoff Wastewater q e [mg/g] k [g/mgmin] 2 RMSE q e [mg/g] k [g/mgmin] 2 RMSE AOCM c 0.28 0.063 0.002 0.003 1.25 0.009 0.221 0.100 UCM c 0.04 3.56 0.003 0.001 0.05 0.30 0.004 0.003 AOCM p 0.27 0.011 0.013 0.006 0.55 0.007 0.109 0.022 UCM p 0.024 0.85 0.006 0.002 0.077 0.074 0.009 0.004 AOCM pcc 0.30 0.036 0.039 0.007 1.29 0.022 0.200 0.106 UCM pcc 0.24 0.036 0.014 0.009 1.35 0.011 0.122 0.071 AOCM in the size range of 2 4.75 mm; Runoff: C0 = 2.7 0.1 mg/L; Wastewater: C0 = 11.1 0.2 mg/L.

PAGE 117

117 Table 4 7 Parameters of 2nd order potential driving model in synthetic runoff, rainfall runoff and wastewater with the same C0 of 2.7 mg/L Media AOCMc AOCMp AOCMpcc Synthetic runoff q e [mg/g] 0.26 0.30 0.31 k [g/mgmin] 0.18 0.046 0.12 2 0.010 0.004 0.045 RMSE 0.009 0.004 0.025 0.56 0.49 4.18 Rainfall Runoff qe [mg/g] 0.28 (108%) 0.27 (90%) 0.30 (97%) k [g/mgmin] 0.063 (35%) 0.011 (24%) 0.036 (30%) 2 0.002 0.013 0.039 RMSE 0.003 0.006 0.007 0.42 0.40 4.29 Wastewater qe [mg/g] 0.25 (96%) 0.26 (87%) 0.37 (119%) k [g/mgmin] 0.083 (46%) 0.011 (24%) 0.041 (34%) 2 0.030 0.027 0.071 RMSE 0.012 0.009 0.027 0.49 0.35 3.17 % as compared to that of synthetic runoff; AOCM in the size range of 2 4.75 mm; C0 of synthetic runoff and wastewater was adjusted to the same level as that of rainfall runoff, i.e. 2.7 mg/L.

PAGE 118

118 Figure 4 1 Schematic experimental configuration of kinetic t ests on media

PAGE 119

119 AOCMpcc UCMpcc Model qe [mg/g] 0.00 0.02 0.04 0.06 0.08 0.10 0.12 AOCMc UCMc Model 2 4.75 mm AOCMc UCMc Model 0.85 2 mm qe [mg/g] 0.00 0.02 0.04 0.06 0.08 0.10 0.12 AOCMpcc UCMpcc Model 2 4.75 mm 0.85 2 mm t (min.) 0 30 60 90 120 150 qe [mg/g] 0.00 0.02 0.04 0.06 0.08 0.10 0.12 AOCMp UCMp Model 2 4.75 mm t (min.) 0 30 60 90 120 150 AOCMc AOCMp AOCMpcc Model 2 4.75 mmA B C D E F Figure 4 2 A) E ): Kinetics of P mass transfer with AOCM and UCM in synthetic runoff, C0 = 1.0 mg/L, modeled utilizing a 2nd order potent ial driving model; F ): Kinetics of P mass transfer for different AOCM in synthetic runoff, C0 = 1.0 mg/L

PAGE 120

120 t (min.) 0 30 60 90 120 150 180 210 240 qe [mg/g] -0.02 0.00 0.02 0.04 0.06 0.08 0.10 0.12 AOCMc AOCMp AOCMpcc Activated alumina Iron coated perlite Biofilter soil media zeolite-perlite-GAC Tire crumb a Tire crumb b Expanded shale a Expanded shale b Figure 4 3 Kinetics of P mass transfer for suite of media tested in a synthetic runoff matrix (C0 = 1.0 mg/L). All kinetics data are modeled using a 2nd order potential driving model; with exception of zeoliteperlite GAC media and tire crumb (Black & Gold) where net rate of P mass transfer is not significantly different than 0 (P

PAGE 121

121 t0.5 (min0.5) 0 2 4 6 8 10 12 14 qe [mg/g] 0.00 0.02 0.04 0.06 0.08 0.10 0.12 AOCMc AOCMp AOCMpcc t0.5 (min0.5) 0 2 4 6 8 10 12 14 Biofilter soil media Expanded shale a Expanded shale b Activated alumina Iron coated perlite A B Figure 4 4 Root time plot for P transfer for suite of media tested in a synt hetic runoff matrix (C0 = 1.0 mg/L), with exception of zeoliteperlite GAC media and tire crumb (Black & Gold) where net rate of P mass transfer is not significantly different than 0 (P

PAGE 122

122 qe [mg/g] 0.00 0.05 0.10 0.15 0.20 0.25 0.30 AOCMc UCMc AOCMp UCMp t (min.) 0 200 400 600 800 1000 1200 qe [mg/g] 0.00 0.05 0.10 0.15 0.20 0.25 0.30 AOCMpcc UCMpcc t (min.) 0 400 800 1200 1600 2000 AOCMc AOCMp AOCMpcc Model A B C D Figure 4 5 Mass transfer kinetics of P onto AOCM and UCM in a Rainfallr unoff matrix with C0 = 2.7 mg/L utilizing a 2nd order potential driving model

PAGE 123

123 t (min.) 0 400 800 1200 1600 2000 AOCMc AOCMp AOCMpcc Model AOCMp UCMp qe [mg/g] 0.0 0.2 0.4 0.6 0.8 1.0 1.2 AOCMc UCMc t (min.) 0 100 200 300 400 500 qe [mg/g] 0.0 0.2 0.4 0.6 0.8 1.0 1.2 AOCMpcc UCMpcc A B C D Figure 4 6 Mass transfer kinetics of P onto AOCM and UCM in a wastewater matrix with C0 = 11.1 mg/L utilizing a 2nd order potential driving model

PAGE 124

124 q e [mg/g] 0.00 0.06 0.12 0.18 0.24 0.30 0.36 Synthetic runoff Rainfall-runoff Wastewater Model AOCMcA q e [mg/g] 0.00 0.06 0.12 0.18 0.24 0.30 0.36 Synthetic runoff Rainfall-runoff Wastewater Model B AOCMp t (min.) 0 400 800 1200 1600 2000 q e [mg/g] 0.00 0.06 0.12 0.18 0.24 0.30 0.36 Synthetic runoff Rainfall-runoff Wastewater Model C AOCMpcc Figure 47 Mass transfer kinetics for P onto AOCM comparing kinetics between aqueous matrices. Data are modeled utilizing a 2nd order potential driving model, C0 = 2.7 mg/L for all aqueous matrices

PAGE 125

125 CHAPTER 5 BREAKTHROUGH OF PHOSPHORUS FROM GRANULAR FILTER METERIALS Background Phosphorus is often identified as the most limiting element and a primary cause of eutrophication (Gomez et al., 1999; Aldridge and Ganf, 2003; Yoshida et al., 2004; Bowes et al. 2005). As an major nonpoint source and pathway of nutrients, high phosphorus loadings of urban runoff can elevate the tropic level in receiving water (USEPA 1993, Kim et al. 2008). While WEF/ASCE (1998) documented that the event mean concentration (EMC) for total phosphorus (TP) 0.33 mg/L and dissolved phosphorus is 0.12 mg/L; it has been reported that EMC of TP in urban runoff ranged from 0.03 to 5.29 mg/L (Davis, 2007) and for highway runoff the median TP ranged from 0.19 to 1.8 mg/L (Drapper et al., 2000). Dean et al. (2005) reported that urban runoff transported DP from 0.3 to 1.4 mg/L (EMC) in an urban pavement catchment. A desired goal for prevention of plant nuisances in flowing waters not discharging directly to lakes or impoundments is 0.1 mg/L (USEPA, 1986). concentration over 0.03 mg/L will likely trigger algal blooms in confined water bodies Given that runoff discharges of TP from urban land uses can be an order of magnitude higher than levels required to trigger algal blooms, there is a focus on control strategies (after source controls) to separate TP and DP from runoff discharges to receiving waters. The utilization of filter media as a control strategy for DP has been a long term interest for wastewater (Pollio and Kumin, 1968; Boari et al., 1976; Yoshida, 1983; Zhao and Sengupta, 1997) and more recently for urban runoff (Gray et al., 2000; Lin and Yang, 2002; Mortula et al., 2007, Sansalone and Ma 2009). While many of these media

PAGE 126

126 have been nonengineered substrates such as sand or perlite with filtration potential but little adsorptive capacity ( Bubba et al., 2003; Forbes et al., 2004; Wu et al., 2008); more recent engineered media with Fe or Al coating have shown significantly improved adsorptive capacity in comparison (Ayoub et al., 2001; Arias et al., 2006; Boujelben et al., 2008; Sansal one and Ma 2009). While the most common metric for adsorptive media capacity is the equilibria isotherm (Lee et al., 2003), successful employment of adsorption for phosphorus removal requires breakthrough analysis ( Metcalf and Eddy 2003). Application of b atch equilibria is often difficult to apply to flowthrough mediabased adsorption system without kinetics data and application of so called single point isotherms represent a further mis representation of such mediabased adsorption (Sperlich et al. 2005, Sansalone and Ma 2009). Breakthrough testing of media utilizing axial benchscale physical models permits a comparison of media and an economical and controlled method to determine media parameters at a given surface loading rate (SLR). While it is reco gnized that the most common surrogate aqueous matrix, deionized (D.I.) water, does not reproduce the complex interactions of rainfall runoff, this matrix is a consistent and reproducible matrix for media testing and facilitating media comparisons subject to controlled variations in loading conditions (Namasivayam and Sangeetha, 2004; Arias et al., 2006; Georgantas and Grigoopoulou, 2007; Boujelben et al., 2008). Hence, although studies with a surrogate matrix such as DI water can provide a comparative bas is for media, media phenomena and model parameters such as media breakthrough, a comparative assessment utilizing the actual aqueous matrices such as rainfall runoff and wastewater is needed.

PAGE 127

127 Beyond equilibria, media capacity for a solute can be represented through breakthrough modeling under a known set of loading conditions (Kse and ztrk, 2008). The time (or more appropriately volume treated) until a set breakthrough or exhaustion level are commonly reported characteristics of a given media, system a nd set of loading conditions; while there has been more limited analytical model development to describe breakthrough in time and throughout the media profile (Aksu et al., 2007) unless carried out with an advectivedispersive numerical model (Liu et al. 2 005). However, introduction of lumped system parameters and constrained correlations have provided tractable analytical models for onedimensional flow systems (Lee et al., 2000; Ko et al., 2001; Aksu et al., 2007). One such bed depth service time (BDST) model is the Thomas model that has been widely used in the design of fixed bed adsorption systems ( Rao and Viraraghavan, 2002; Thirunavukkarasu et al., 2003; Aksu and Gnen, 2004; Aksu et al., 2007; Kse and ztrk, 2008; Pokhrel and Viraraghavan, 2008; Si vakumar and Palanisamy, 2009). The main advantages of the Thomas model are simplicity and reasonable accuracy in predicting media breakthrough behavior for adsorptive media. Additionally the kinetics coefficient and adsorption capacity from the model can be determined from monitored data. Typically examined for a single SLR, examination across a range of SLRs can extend the use of the Thomas model to incorporate behavior subject to unsteady phenomena, for example for a fixed bed media absorber utilized af ter an equalizing or primary treatment operation of runoff ( Erickson et al., 2007). While there have been many studies regarding the use of such models (Koh et al., 1998; Aksu and Gnen, 2004; Mantovaneli et al., 2004; Aksu

PAGE 128

128 et al., 2007; Kse and ztrk, 2008), illustrating the relationship between these Thomas model parameters and SLR has not been demonstrated through a physical model. In this study, an axial media reactor is utilized for treatment of DP in aqueous matrices: surrogate runoff, rainfall runoff and wastewater. The study hypothesizes that filter media adsorption behavior is significantly different subject to the same controlled loadings. It is further hypothesized that for the AlOx coated media that media breakthrough behavior is a function of the substrates for the same diameter substrates. Given that a media intended for DP control should not act as a source of DP, desorption behavior is also quantified for each media. For the media tested that illustrate adsorptive capacity, SLR is hypothes ized to have a quantifiable impact on breakthrough behavior. Furthermore it is hypothesized that correlations between SLR and Thomas model parameters can be developed for adsorptive media and a comparison of these correlations can illustrate the differing influence of the media substrate. Materials and Methods This study examines AlOx coated engineered media: AOCMc (clay based), AOCMp (pumice based), AOCMpcc (Portland cement concrete based) as well as nonengineered media: tire crumb (a, b), expanded shale (a, b), bioretention soil media and Zeolite perlite GAC (volumetric ratio: 5:4:1). AOCMc and AOCMp is prepared by applying aluminum as surface coating (Sansalone and Ma 2009). AOCMpcc is prepared by adding aqueous Al(NO3)3 solution (0.5 M) to produce a water:cement ratio of 0.5. This slurry is cured, crused and sizegraded. Specific gravity analysis is conducted with a pycnometer (Quantachrome Ultrapycnomet er 1000). Bulk density is determined by measuring the dry mass and corresponding volume of the media bed in the column. Surface charge for the media is measured using a potentiometric titration methodology

PAGE 129

129 of Van Raij and Peech (1972). A constant head test is used to determine the hydraulic conductivity as described by Braja (2001). Media characteristics are summarized in Table 5 1. DP adsorption subject to continuous flow through loading is investigated in an axial flow with length of 267 mm and internal diameter of 45 mm. With a fixed bed volume of 350 mL, the ratio of packed media length to the reactor internal diameter (L/D) is fixed at 5. The mass of media utilized is dependent on the media. The constant influent flow is supplied by a peristaltic pump. Samples are collected at regular time intervals and effluent DP concentration (Ce) is determined. The axial column configuration is shown in Figure 5 1. In this study the breakthrough level (Vb) is set to Ce/C0 = 0.1 (effluent Ce = 0.05 mg/L) and exhausti on level (Vexh) is set to Ce/C0 = 0.9. B reakthrough curve are presented as a function of the number of bed volumes (NBV) treated. Three aqueous matrices are utilized: (1) synthetic runoff to compare media and develop Thomas model parameters as a function of SLR, (2) primary rainfall runoff influent, and (3) secondary wastewater sand filter effluent from a local Advanced Wastewater Treatment (AWT) facility. The chemistry of each aqueous matrix is summarized in Table 5 2. All media are tested in synthetic runoff with a constant influent DP (C0) of 0.5 mg/L and pH of 7 (controlled throughout the experiment) and a SLR fixed at 40 L/minm2 based on the SLR data for a paved urban Cincinnati catchment (Sansalone and Teng, 2005). The breakthrough behavior of AOC Mc and AOCMpcc are further examined in a rainfall runoff and wastewater matrix, respectively.

PAGE 130

130 A phosphorus stock solution is prepared by dissolving anhydrous potassium dihydrogen orthophosphate (KH2PO4) powder (Analytical reagent grade, Fisher Scientific ) into D.I. water. The stock solution is calibrated to 1000 mg/L concentration, and utilized as standard solution as necessary. Phosphate concentration is measured by a HACH DR/5000 Spectrophotometer using PhosVer 3 Ascorbic Acid Method ( Eaton et al., 199 8). All measurements are duplicated and repeatability is within 5%. Measurements of pH, conductivity and redox are carried out using a calibrated Thermo Scientific Orion 5 Star multi meter. The influence of SLR is varied from 10 to 160 L/minm2 for the synthetic runoff matrix. This loading range covers the typical SLR of source areas where unit operation and process are applied for in situ treatment (Sansalone and Buchberger, 1997; Liu et al., 2005). Since hydrology and SLR are variable under event base d conditions, a range of EBCT is experienced for the media. EBCT is the entire volume of packed media bed divided by Q, the volumetric flow rate [L/min]. The flow is volumetrically calibrated periodically and any measured variance is less than 2%. The dry media mass utilized, SLR, and EBCT for each run are summarized in Table 5 3. After the media reaches exhaustion in axial column adsorption experiments, the solution in the reactor is drained and desorption study is conducted after 24 hours of gravitati onal drainage so that only residual moisture content remains in the media bed. In desorption tests, the DP solution is replaced with D.I. water and the media loaded at the 40 L/minm2. Media bed effluent samples are filtered and analyzed for DP leached from the media bed.

PAGE 131

131 The Thomas model widely used for fixed bed adsorption systems (Viraraghavan and Rao, 2002; Thirunavukkarasu, 2003; Sivakumar and Palanisamy, 2009) has the following form. ] )( exp[ 1 10 0 0Q V C m q K C CT e (1) Ce is the effluent DP concentration [mg/L], C0 the influent DP concentration [mg/L], KT the Thomas rate constant [L/minmg], q0 media phase c oncentration of the solute [mg/g], m the mass of the adsorbent [g], V the throughput volume [L], and Q is the volumetric flow rate [L/min]. Thomas model results are compared to monitored data utilizing the Root Mean Square Error (RMSE). In this expression Q(qi) is the monitored value and E(qi) is the modeled value. n q E q Q RMSEn i i i 1 2)] ( ) ( [ ( 2) The fit is also evaluated with a chi 22 measures differences betwe2 the better the fit between monitored and modeled. n i i i iq E q E q O1 2 2) ( )] ( ) ( [ (3) The relationship between the model parameters, KT (rate parameter) and q0 (capacity parameter) are examined as a function of SLR for two adsorptive media (media for which adsorption >> desorption). AOCMC and AOCMP are chosen since the mass transfer phenomena are primary ligand exchange and do not include surface precipitation that may dominate AOCMpcc

PAGE 132

132 Results and Discussions Phosphorus Breakthrough Profiles in Synthetic Runoff Matrix for Each Media System Breakthrough profiles are presented by plotting the ratio of effluent P concentration (Ce) to the initial concentration (C0), Ce/C0, against time elapsed or NBV treated. Here, 1 BV is fixed at 350 mL for all media tested. Therefore: ) 350 ( ) )( ( L) / 1000 ( mL t Q mL NBV (4) In this expression Q [L/min] is the volumetric flow rate and t is elapsed time (min.). The breakthrough curves for AOCM forms and common nonengineered media are shown in Figure 5 2 and 5 3 respectively. Breakthrough curves shown in Figure 5 2 and 5 3 are conducted in synthetic runoff matrix at surfac e loading rate = 40 L/minm2 (EBCT = 5.7 min) with C0 = 0.5 mg/L. From Figure 5 2, results indicate that AOCMc breakthrough occurred at 900 BV and exhaustion is at greater than 4000 BV. In contrast, breakthrough for AOCMp and AOCMpcc are approximately 90 and 120 BV, respectively. The NBV of exhaustion for AOCMp is greater than 2000 and 800 for AOCMpcc. These results indicate that besides the active AlOx coating the physiochemical properties of substrates can also impact the engineer ed media capacity. The effectiveness of AlOx coating for DP removal promotion has been reported in literature and mass transfer of phosphates to hydrous aluminous oxide is through ligand exchange to surface hydroxyl groups, resulting in OHrelease into solution (Genz et al., 2004; Kim and Kirkpatrick, 2004; Arias et al. 2006; Yang et al., 2006; Georgants and Grigoropoulou, 2007; Sansalone and Ma, 2009). In this study, pH of the feeding P solution is fixed at 7 and ionic strength is adjusted to

PAGE 133

133 0.01M with 2 mol/L KCl solution. An increase of effluent pH is observed for AOCMc and AOCMp as a result of ligand exchange between surface hydroxyl groups and phosphate ions. A greater pH increase from AOCMpcc is difficult to discern for ligand exchange due to the CaO in the cement concrete substrate. The conductivity of the first 2 or 3 samples were usually higher in the effluent than in the feed solution; and became relatively constant and similar to the influent. The initial enhancement of conductivity can be attri buted to the minerals dissolved from the media surface. The amount of such minerals varied with substrates and was more evident for concrete based media (AOCMpcc). The redox ranged between 200 to 300 mV a nd decreased to below 100 mV at the end of each run. Similar redox decline was reported in literature (Onar et al., 1996). For media with adsorptive capacity (rate of adsorption >> desorption) the Thomas model is used to simulate adsorption breakthrough behavior as shown in Figure 5 2. The model parameter s q0 and KT are estimated and summarized in Table 5 4. The 2 and RMSE indicate that Thomas model predicts the entire breakthrough behavior. In addition to the engineered media of AOCM, the commonly available media utilized in unit operation and process media deployments, including bioretention soil media, expanded shale (a, b), tire crumb (a, b) and zeoliteperlite GAC are also tested in the f low through column reactor using the synthetic runoff matrix. The resulting breakthrough curves are shown in Figure 5 3. While variability does exist; tire crumb, expanded shale, bioretention soil media and zeoliteperlite GAC are exhausted within a small number of bed volumes, indicating very limited media capacity. With adsorption

PAGE 134

134 desorption such media could not be modeled with the Thomas model. The media breakthrough and exhaustion capacity are summarized in Table 5 3 for each run. Desorption Experi ments Desorption study with D.I. water are shown in Figure 5 4. Figure 5 4a and 5 4b summarize the DP effluent profiles. Figure 5 4c and 5 4d compares the total amount of DP adsorbed and desorbed per gram of media used. Results indicate that for AOCMc and AOCMp little adsorbed phosphorus is released. After the first one to two samples the DP concentration in the effluent is below the detection limit (0.02 mg/L), suggesting strong adsorption bond between adsorbate and adsorbents. However, AOCMpcc did release phosphorus. Due to the high surface pH of AOCMpcc, phosphorus is removed through surface precipitation (Agyei et al., 2002) and adsorption. The results obtained from this study suggested the temporary physical capture and sloughing of the precipitate d eposits from the media. Studies have demonstrated that the the age of calcium precipitate is the factor that affects the P mobility. The aging process may take months to form the stable CaP (Gomez et al., 1999). Given that the contact time of P with media is less than 6 minutes; and the whole adsorptiondesorption test is conducted in 48 hours the precipitated phosphorus can potentially be weakly desorbed as compared to clay and pumice substrates. Figure 5 4b and 5 4d indicates that the very limited DP ads orbed by the commercially available media is a result of weak physical bonds media such as Tire crumb a and Zeoliteperlite GAC; the amount of DP desorbed is more than adsorbed. Phosphorus Breakthrough Profiles in Actual Matrix for AOCMc and AOCMpcc Although tests with a surrogate matrix of D.I. water provide reproducible comparative data on media capacity, water chemistry influences media adsorption

PAGE 135

135 (Shilton et al., 2005; dm et al., 2007, Sansalone and Ma 2009). In order to evaluate media behavior in ac tual aqueous matrices, AOCMc and AOCMpcc are also tested in rainfall runoff and wastewater at a SLR of 40 L/minm2.. Results are presented in Figure 5 2 and RMSE are summarized in Table 5 4. As shown in Figure 5 5, breakthrough initiated at approximately 390 BVs for AOCMc in rainfallrunoff and 1200 BVs for AOCMpcc in wastewater. When comparing the media capacity at breakthrough (X/Mb) of AOCMc in Table 5 3, the results indicate that media capacity when loaded by rainfall runoff decreased more than 50% with an X/Mb (0.41 mg/g) as compared to the synthetic matrix (0.94 mg/g), while the inlet DP concentration were the same as 0.5 mg/L. Results suggest the role of competitive interactions, prim arily SO4 2 (Sansalone and Ma, 2009), have a significant effect on AOCMc breakthrough behavior. In contrast, wastewater chemistry and a Carich substrate increased both breakthrough (X/Mb, 0.79 mg/g) and exhaustion (X/Mexh, 2.35 mg/g) capacity of AOCMpcc a s compared to synthetic runoff matrix (X/Mb, 0.06 mg/g; X/Mexh, 0.23 mg/g). Additionally, AOCMpcc did not reach breakthrough until 1200 BVs and exhaustion occurred at greater than 7000 BVs. This breakthrough phenomena exhibited by AOCMpcc in wastewater, al ong with the desorption results; strongly indicate the occurrence of surface precipitation. Unlike clay and pumice substrates with ligand exchange as dominant mechanism, surface precipitation is important for AOCMpcc due to the high content of CaO. Further more, surface precipitation is promoted by the high concentration of Ca2+ in wastewater, resulting in the high capacity as shown in Figure 5 5b.

PAGE 136

136 Effect of Flow Rate/Surface Loading Rate on Media Breakthrough Curves The effect of SLR on DP adsorption of AO CMc and AOCMp is examined by varying the surface loading rate from 10 to 160 L/min m2, while the influent DP concentration is held constant at 0.5 mg/L. The plots of comparative normalized DP concentration versus number of bed volume treated at different S LR are given in Figure 5 6 SLR has a direct impact on contact time with media and hence influences removal efficiency, especially when the adsorption is diffusion controlled (Gen Fuhrman et al., 2005). The EBCT ranged from 22.6 to 1.4 min as SLR is varied from 10 to 160 L/minm2. As shown in Figure 56 the breakthrough curves became steeper and shifted towards the origin with increasing SLR while both the media breakthrough and exhaustion time decreased. As a result, the solute does not have enough time to interact with the sorbate and diffuse into the pores. For example, at a surface loading rate of 10 L/minm2, more than 2000 bed volumes of solution can be treated before AOCMc breakthrough. When the surface loading rate increased to 160 L/minm2, the br eakthrough occurred at less than 500 bed volumes. Table 5 3 summarizes breakthrough and exhaustion capacity of AOCMc and AOCMp at different SLRs and both X/Mb and X/Mexh decreased with increasing loading rates. Similar results were obtained from other adsorption studies (Aksu and Gnen, 2004; Gen Fuhrman et al., 2005; Malkoc and Nuhoglu, 2006; Aksu et al., 2007). The data at various SLRs are also fit to the Thomas model to determine the Thomas rate constant (KT) and maximum media concentration (q0). All t he parameter values are listed in Table 5 2 and RMSE. Breakthrough curves modeled at all SLRs reproduce monitored data. As SLR increases,

PAGE 137

137 the value of KT increases and q0 decreases. These results are in agreement w ith the literature (Mantovaneli et al., 2004; Malkoc and Nuhoglu, 2006; Aksu et al., 2007). Correlations between SLR and Thomas Model Parameters DP mass transfer processes from an aqueous solution to media is dynamic involving external mass transfer, film diffusion, and intraparticle diffusion. Given that these multiple processes occur simultaneously, simulating every process will require a numerical solution unless these processes can be lumped as a function of a primary system parameter for a given medi a. It is hypothesized that volumetric flow rate (as SLR) is the primary system parameter that influences external mass transfer mechanisms, such as axial dispersion and film diffusion, but may be less significant for intraparticle (media) mass transport at small contact times; for example intraevent runoff loadings times as compared to longer inter event times (Koh et la., 1998; Lee et al., 2003, Liu et al. 2005). The intraevent significant of intraparticle diffusion at a given SLR is also dependent on the pore size distribution in the media. In the Thomas model, the rate constant (KT), characterizing the rate of solute transfer from liquid to media phase, increases with SLR, indicating the influence of external mass transfer on system kinetics. It is further hypothesized that this relationship is a function of the adsorptive media. Results indicate that in a 1dimensional flow system with adsorptive media (AOCMC or AOCMP) that KT is proportional to SLR; and the following relationship is proposed across the flow rate range. SLR KT* (5)

PAGE 138

138 In this expression SLR is the surface loading rate [L/minm2 proportionality coefficient [m2/mg] which is a function of the system and the adsorptive T values at different SLRs using Equation 5 and is presented in Table 5 5. Results in Figure 5 7 illustrate the proportionality of the relationship between KT and SLR as described in Equation 7. Results also indicate that p and AOCMc are statistically significantly different (at p = 0.05) The second parameter examined in the Thomas model, q0 [mg/g], is an index for the media phase capacity for a solute, in this case DP. Ideally, in a 1dimensional flow system of fixed depth the SLR and contact time with the media are directly related, hence it is hypothesized that not only is the mass transfer rate a function of SLR but so is the media capacity. Both intraparticle and external film resistance play an important role in the adsorption process and the surface reaction (ligand exchange) is not the rate l imiting step. Intra particle (media) mass transfer rate is positively related with flow contact time (Liu et al. 2005) and there should be an inverse relationship, albeit nonlinear, between intraparticle mass transfer rate and SLR. In contrast, a higher SLR will decrease the film resistance and result in an increase in the external mass transfer (Ko et al., 2000). Given the separate and divergent influence of these two mechanisms in response to SLR the following equation is proposed. n mSLRq q ) (0 (6) Here, qm [mg/g] can be regarded as the theoretical maximum media capacity with infinite contact time, while n represents the influence of SLR. Higher values of n represent a greater influence of SLR on q0. Both qm and n are hypothesized to be a

PAGE 139

139 function of the adsorptive media. Therefore an adsorptive media with a high mass transfer rate and equilibrium capacity is expected to have a lower nexponent as compared to another adsorptive media of a lower mass transfer rate and capacity. The q0 (determined from monitored data) as a function of SLR is fit with Equation 6 and qm and n are determined. The results are summarized in Table 5 5. Figure 5 7 illustrates that the relationship provided by Equation 6 predicts q0. Results also indicate that the n exponent of each relationship (plotted as log transformed axes) for AOCMp and AOCMc are statistically significantly different (at p = 0.05) The relationships of Equation 5 and 6 as a function of the primary operating parameter, SLR can represent the breakthrough of AOCMc and AOCMp, both adsorptive media with a primary mechanism of ligand exchange. Summary The breakthrough of influent DP feed at 0.5 mg/ L from engineered granular filter media deployed in axial reactors (1 BV = 350 mL and L/D of 5.0), adsorptive media represented by three substrates supporting aluminum oxide, is examined as a function of SLR ranging from 10 to 160 L/minm2 with SLR as a surrogate for flow rate. These adsorptive media are compared to common nonengineered media at a fixed SLR of 40 L/min m2 with respect to breakthrough and exhaustion capacity and desorption. In these breakthrough and desorption examinations the aqueous mat rix utilized is a controlled synthetic runoff matrix of D.I. water, recognizing that while this matrix does not provide the complexity of rainfall runoff or wastewater, the matrix is consistently reproducible for comparative or parameterization studies.

PAGE 140

140 Results indicate that AOCMc had higher breakthrough (X/Mb) and exhaustion (X/Mexh) capacity as compared to AOCMpcc and AOCMp; AOCMc largely as a function of the substrate providing higher specific surface area for the AlOx. While AOCMpcc has a higher breakthrough capacity than AOCMp the exhaustion capacity is lower; both of these results a function of surface precipitates and the instability thereof at the AOCMpcc surface, in addition to ligand exchange generated by the AlOx. In contrast the primary adsorption mechanism of AOCMc and AOCMp is ligand exchange. Results indicate the nonengineered media has very limited DP breakthrough capacity (adsorption rate Leaching results using D.I. water at a pH of 7.0 and an SLR of 40 L/minm2 indicate that a significant, in some cases a predominant fraction of DP adsorbed, is subsequently leached despite a 24 hour inter event time providing for intraparticle mass transfer. In contrast, all of the three AOCM exhibit low desorption with AOCMc providing the lowest leaching concentrations and AOCMpcc providing a higher but nominal level of leaching from surface precipitates. Comparing breakthrough results between the synthetic matrix and rainfall runoff or wastew ater matrices for AOCMc and AOCMpcc respectively; results illustrate differing behavior. Competing solutes reduce the breakthrough capacity of AOCMc primarily functioning through ligand exchange. In contrast, the additional mechanism of calcium based precipitation by AOCMpcc generated by the CaO on the hydrated cementitious surface and the higher Ca concentrations in solution results in an increase in breakthrough capacity for AOCMpcc in wastewater. Results indicate that the surface properties of the medi a whether providing ligand exchange or a separate

PAGE 141

141 mechanism such as chemical precipitation must be considered in conjunction with the aqueous matrix. SLR ranging from 10 to 160 L/minm2, typical of SLRs to runoff filter systems, is a primary explanatory variable for DP breakthrough from adsorptive media. Increasing SLR results in consistently lower BVs at breakthrough and exhaustion levels. The Thomas model reproduced physical model data across all SLRs. The Thomas model rate constant (KT) and mediaphase capacity (q0) are a function of SLR for a given adsorptive media; media where the adsorption rate >> desorption rate or media that did not include a significant surface precipitation mechanism in addition to an exchange reaction. Mathematical relations hips (linear and power law) to SLR are demonstrated for KT and q0, respectively. The proportionality constants relating KT and SLR are significantly different (p = 0.05) as are the power law exponent relating q0 and SLR for the adsorptive media tested. N omenclature AOCM: Aluminum oxide coated media AOCMc: Aluminum oxide coated media with clay as raw substrate AOCMp: Aluminum oxide coated media with pumice as raw substrate AOCMpcc: Aluminum oxide media with Portland cement concrete as raw substrate B&G: Black & gold BV Bed volume (mL) C0: The initial DP concentration (mg/L) Ce: DP concentration at equilibrium (mg/L) DI water: De ionized water DP: Dissolved phosphorus

PAGE 142

142 EBCT Empty Bed Contact Time (min) E(qi) Expected value KT Thomas rate constant (L/minmg) m : The dry mass of media used (g) n C apacity coefficient of surface loading rate NBV Number of bed volume pHpzc: pH of point of zero charge q0 S olid media phase concentration of DP (mg/g) qm M aximum media capacity with infinite contact time (mg/g) Q Volumetric flow rate (L/min) Q(qi): Observed value RMSE : Root Mean Square Error SLR Surface Loading Rate (L/minm2) V Throughput volume (L) V0: The packed media volume (L) Vb number of bed volume (NBV) treated at breakthrough level of C / C0 = 0.1 Vexh number of bed volume (NBV) treated at exhaustion level of C / C0 = 0.9 X/Mb Breakthrough capacity at C/C0 = 0.1 X/Mexh Exhaustion capacity at C/C0 = 0.9 2: Chisquare P roportionality coefficient [m2/mg] representing conversion constant between surface loading rate and Thomas rate constant

PAGE 143

143 Table 51 Media characteristics Media Size [mm] Specific g ravity (at 20) Bulk d ensity [g/cm 3 ] Hydraulic conductivity [cm/s] pHpzc AOCM c 2.0 4.75 2.29 0.04 0.43 0.04 882 24 4.61 0.06 AOCM p 2.0 4.75 2.27 0.01 0.66 0.06 936 18 4.59 0.10 AOCMpcc 2.0 4.75 2.67 0.04 0.89 0.03 912 14 11.84 0.16 Bioretention soil media 0.001 10 2.69 0.03 1.35 0.12 6 1 6.35 0.03 Zeolite Perlite GAC 2.29 9.65 2.02 0.01 0.71 0.08 864 12 7.51 0.04 Expanded shale a 2 10 1.91 0.02 0.81 0.06 1098 12 9.84 0.10 b 0.45 4.5 0.86 0.05 384 6 Tire Crumb (B & G) a 0.85 10 1.19 0.04 0.53 0.07 138 3 8.58 0.08 b 0.5 4.75 0.65 0.06 30 6 AOCMc = Aluminum Oxide Coated Media, clay based AOCMp = Aluminum Oxide Coated Media, pumice based AOCMpcc = Aluminum Oxide Coated Media, Portland cement concrete based pHpzc = pH at point of zero charge Entries are mean standard deviation (n = 3)

PAGE 144

144 Table 5 2 Water chemistry of experimental m atrix Matrix Synthetic r unoff Rainfallr unoff (Event on 05/16/08) Wastewater (Secondary e ffluent) pH 7.00 0.05 7.11 0.07 7.13 0.08 Conductivity [s/cm] 1025 56 218 12 812 6 Turbidity [NTU] 0.83 0.46 2.80 1.02 3.24 0.22 Alkalinity [mg/L] as CaCO 3 5.4 0.8 74 2 66 1 Ionic Strength [M] 0.010 0.002 0.004 0.002 0.013 0.001 DOC [mg/L] 0.3 0.1 28.9 3.6 12.4 2.8 Ca 2+ BDL a 25.45 1.26 60.52 7.36 NO 3 N 0.45 0.08 3.08 0.15 1.52 0.08 SO 4 2 0 10.06 1.08 118 2 Cl b 360 10 8.05 0.15 92 7 HCO 3 3.3 0.7 45.1 1.2 29.3 2.1 Data are mean standard deviation a BDL: Below Detection Limit () b 0.01M KCl was used to adjust the ionic strength of synthetic runoff

PAGE 145

145 Table 5 3 Loadresponse parameters for a suite of media utilizing water chemistry matrices shown in Table 5 2 and an axial reactor Media Matrix Mass (g) SLR [L/min m 2 ] EBCT (min.) X/M b [mg/g] X/M exh [mg/g] V b V exh AOCMc Synthetic runoff 160 10 22.6 2.38 >4.17 2300 >5800 20 11.3 1.24 >2.83 1200 >4000 40 5.7 0.94 >2.45 900 >3500 80 2.8 0.85 >2.33 860 >3500 160 1.4 0.53 >1.93 400 >3200 AOCMp 240 10 22.6 0.58 1.71 830 4600 20 11.3 0.32 >0.96 480 >2200 40 5.7 0.06 >0.69 90 >1200 80 2.8 0.03 >0.52 60 >1600 160 1.4 0.005 >0.32 5 >1000 AOCM pcc 313 40 5.7 0.06 0.23 120 800 Bioretention soil media 470 <0.001 0.0031 <1 25 ZPG 200 <0.001 0.0013 <1 5 Tire Crumb (B & G) a 186 <0.001 0.0012 <1 5 Tire Crumb (B & G) b 227 <0.001 0.0024 <1 10 Expanded Shale a 284 <0.001 0.0021 <1 8 Expanded Shale b 302 <0.001 0.0073 1 35 AOCMc Rainfall runoff 160 0.41 1.09 390 1800 AOCM pcc Wastewater 313 0.79 > 2.35 1200 > 7300 Axial reactor: L/D = 5; D (internal diameter of column) = 4.5 cm; 1 bed volume = 350 mL (pore volume equivalent varies with media); pHinfluent = 7.0 0.2; C = Effluent DP concentration, [mg/L] C0 = Influent DP concentration, 0.5 mg/L SLR = Surface Loading Rate [L/minm2] EBCT = Empty Bed Contact Time; X/Mb = breakthrough capacity at C/C0 = 0.1; X/Mexh = exhaustion capacity at C/C0 = 0.9; Vb = number of bed volume (NBV) treated at breakthrough level of C / C0 = 0.1; Vexh = number of bed volume (NBV) treated at exhaustion level of C / C0 = 0.9; > breakthrough capacity is larger than the stated values(the experiment terminated before C / C0 = 0.9)

PAGE 146

146 Table 5 4 T homas m odel parameters for AOCM forms utilizing water ch emistry matrices shown in Table 52 and goodness of fit indices between modeled and monitored results Media Matrix SLR [L/min m 2 ] q 0 [mg/g] K T (0 3 ) [L/min mg] 2 RMSE AOCMc Synthetic runoff 10 4.21 0.107 0.497 0.026 20 2.93 0.236 0.169 0.021 40 2.59 0.480 0.168 0.030 80 2.37 0.939 0.479 0.058 160 1.95 1.850 0.192 0.041 AOCMp 10 1.66 0.108 0.729 0.043 20 0.97 0.365 0.408 0.049 40 0.69 0.763 0.391 0.049 80 0.51 1.956 0.243 0.054 160 0.33 4.555 0.073 0.049 AOCM pcc 40 0.24 2.293 0.048 0.035 AOCMc Rainfall runoff 40 1.08 1.174 0.063 0.017 AOCM pcc Wastewater 40 2. 88 0. 184 0. 318 0.0 53 RMSE = Root Mean Square Error; 2 = chi square.

PAGE 147

147 Table 5 5 Relationships between Thomas model parameters and surface loading rate Media q 0 [mg/g] ( n mSLR q q ) (0 ) K T [L/min mg] (SLR KT*) qm [mg/g] n 2 RMSE (0 5 ) 2 (0 6 ) RMSE (0 5 ) AOCM c 7.55 0.278 0.072 0.211 1.16 1.46 1.02 AOCM p 6.64 0.614 0.021 0.053 2.71 272 22.1 SLR = surface loading rate, [L/minm2]; q0 = solid media phase concentration of DP, [mg/g]; qm = maximum media capacity with infinite contact time ; n = capacity coefficient of surface loading rate; KT = Thomas rate constant [L/min mg]; = proportion ality coefficient [m2/mg] representing conversion constant between surface loading rate and Thomas rate constant

PAGE 148

148 Figure 5 1 Schematic experimental configuration of column breakthrough experiments

PAGE 149

149 NBV 0 800 1600 2400 3200 4000 C/C0 0.0 0.2 0.4 0.6 0.8 1.0 AOCMC AOCMp AOCMpcc Model Figure 5 2 Breakthrough curves for AOCM (2 4.75 mm) with an influent pH of 7.0 and C0 of 0.5 mg/L at a surface loading rate of 40 L/minm2. Range bar represent standard deviation between duplicate samples

PAGE 150

150 NBV 0 10 20 30 40 50 60 70 C/C0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Tire crumb a Tire crumb b Expaded shale a Expanded shale b Bioretention soil media Zeolite-perlite-GAC Figure 5 3 Breakthrough curves for commercially available media with an influent pH of 7.0 and C0 of 0.5 mg/L at a surface loading rate of 40 L/minm2. Range bar represent standard deviation between duplicate samples

PAGE 151

151 NBV 0 20 40 60 80 DP Desorbed [mg/L] 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 AOCMc AOCMp AOCMpcc DP Adsorbed/Desorbed [mg/g] 10-410-310-210-1100101 Adsorption Desorption AOCMcAOCMpAOCMpccTire Crumb a Tire Crumb b Expanded shale a Expanded shale b Bioretention soil media NBV 0 20 40 60 80 Tire crumb a Tire crumb b Expanded shale a Expanded shale b Bioretention soil media Zeolite-perlite-GAC Adsorption Desorption Zeolite-perlite-GACA B C D Figure 54 Phosphorus desorption from media with D.I. water. Range bar represent standard deviation between duplicate samples

PAGE 152

152 NBV 0 2000 4000 6000 8000 Wastewater Synthetic matrix Model NBV 0 2000 4000 6000 8000 C/C0 0.0 0.2 0.4 0.6 0.8 1.0 Rainfall-runoff Synthetic matrix Model AOCMcAOCMpccA B Figure 55 A) : Breakthrough curves for AOCMc with rainfall runoff as influent at a surface loading rate of 40 L/minm2 (C0 = 0.5 mg/L); B) : Breakthrough curves for AOCMpcc with wastewater as influent at a surface loadin g rate of 40 L/min m2 (C0 = 0.5 mg/L). Range bar represent standard deviation between duplicate samples.

PAGE 153

153 NBV 0 1000 2000 3000 4000 5000 6000 7000 C/C0 0.0 0.2 0.4 0.6 0.8 1.0 10 20 40 80 160 Model L/min-m2A. AOCMc NBV 0 1000 2000 3000 4000 5000 6000 7000 10 20 40 80 160 Model B. AOCMpL/min-m2 Figure 56 Effect of surface loading rate on breakthrough curves of AOCMc and AOCMp. in synthetic runoff at C0 = 0.5 mg/L

PAGE 154

154 0 30 60 90 120 150 180 KT (*103) [L/min-mg] 0.0 1.5 3.0 4.5 6.0 AOCMc AOCMp Model SLR KT 510 16 1 510 71 2 SLR [L/min-m2] 10 100 q0 [mg/g] 0.1 1 10 AOCMc AOCMp Model n mSLR q q ) (0n = 0.278 n = 0.614 A B Figure 57 Correlations of surface loading rates and Thomas model parameters for AOCMc and AOCMp.at C0 = 0.5 mg/L. Range bar represent standard c and AOCMp are significantly different at P = 0.05

PAGE 155

155 CHAPTER 6 CONCLUSIONS This research examined a suite of granular filter substrates as potential adsorptivefiltration media for phosphorus control in urban drainage. The media tested were aluminum oxide (AlOx) coated media (AOCM) consisting of three separate substrates : clayey soil (AOCMc) pumice (AOCMp) and Portland cement concrete (AOCMpcc) along with their uncoated media substrate (UCM); a set of commercially available media including activated alumina, bioretention soil media, expanded shale (2 gradations), Fecoated perlite, tire crumb B&G (2 gradations) and Zeoliteperlite GAC. The media were evaluated through 3 qualitative descriptions of equilibrium isotherm, kinetics and column breakthrough under equivalent conditions of surrogate matrix using DI water. Additionally, metal leaching during adsorption and P desorption after media exhausti on were also monitored. Furthermore, media phenomena of AOCM were investigated as functions of media coating and substrates as well as synthetic runoff, actual rainfall runoff and wastewater matrices. Batch equilibrium studies were firstly conducted betwe en the three substrates based AOCM&UCM and aqueous phosphorus species (as dissolved phosphorus, DP) Equilibrium was modeled as Freundlich isotherms for three matrices: rainfall runoff, wastewater and deionized (DI) water as a surrogate runoff matrix. When comparing the suite of media; results demonstrated that each AOCM exhibited significantly higher capacities than the corresponding UCM substrate in surrogate matrix. With the exception of the concretebased substrate the uncoated clay or pumice had relatively small equilibria capacity across the entire DP range from 0.05 to 50 mg/L in comparison to AOCM of the same substrate. This result supports the studys hypothesis that the

PAGE 156

156 inclusion of AlOx either as an admixture (a chemical addition to the water c ement ratio) or coating will statistically significantly increase adsorption capacity. Previous studies have demonstrated ligand exchange as a mechanism for phosphate adsorption onto hydrous metal oxide surfaces. In addition to ligand exchange, results of this study indicate that additional mechanisms, for example surface precipitation for the concretebased substrate (AOCMpcc), occur. AOCMc and AOCMpcc both provided similar high equilibrium capacity as compared to AOCMp. In contrast the equilibrium cap acity of the UCMpcc dominated the uncoated clay and pumice substrates due to surface precipitation. While a controlled reproducible DI surrogate matrix was utilized to compare isotherms for each AOCM and companion substrate; isotherm results from this surrogate matrix were compared to uncontrolled environmental rainfall runoff and wastewater matrices. While the DI surrogate matrix generated the highest equilibrium capacity across the range of DP concentrations examined for AOCMc and AOCMp, this surrogate m atrix generated the lowest equilibrium capacity for AOCMpcc as compared to the actual environmental matrices. Examining each AOCM type separately, the rainfall runoff and wastewater matrices generated isotherms that were similar for each matrix across the range of concentrations examined. With ligand exchange as the primary adsorption mechanisms, the decreased equilibrium capacity of AOCMc and AOCMp loaded by rainfall runoff and wastewater matrices can be attributed to multiple anion competition, such as SO4 2 competing with phosphate for media exchange sites. In contrast, AOCMpcc capacity combines the dominant mechanism of surface precipitation, ternary complexation with Ca2+ and ligand exchange to generate nearly identical

PAGE 157

157 equilibrium capacity across the entire DP concentration range which was significantly greater than the surrogate DI matrix. Results from b atch equilibrium experiments of the suite of commercially available media exhibited significantly varied capacities. The unified sorption coeffici ent Ku demonstrated the distribution of partitioning coefficient of sorbate between the aqueous and solid phase at various equilibrium concentrations and provided a quantitative index of media capacity, which can be evaluated and compared among different m edia. Results indicated that activated alumina, Fe coated perlite and AOCMc were three media with good P removal potential while the others such as tire crumb, expanded shale, ZPG and bioretention soil media exhibited limited capacities. The possible treat ed water contamination due to media application was examined through metal leaching tests. Al, Ca, Fe and Mg were found to be the major metals released into the water sample. It was also found the high P removal efficiency of activated alumina and Fe coate d perlite were accompanied by a significant amount of Al and Fe leaching, imposing an environmental concern of these media application. On the contrary, AOCMc seemed to be able to satisfy both P removal and minimal contamination requirement and thus it is considered as a promising cost effective adsorptive media for P treatment. Overall mass transfer rates of DP from a series of aqueous solutions by the suite of granular filter media in a recirculating reactor was also monitored and modeled. Kinetic rates are examined in a synthetic runoff matrix, an actual rainfall runoff matrix, and a secondary effluent wastewater matrix. Kinetics parameters are developed based on simulation of monitored data with a 2nd order potential driving model for overall mass

PAGE 158

158 tra nsfer and an empirical model to identify intramedia diffusion as a rate limiting mechanism. Compared to the UCM controls for each AOCM, the Al Ox coated media achieved significantly higher kinetic rates and equilibrium capacity as irrespective of the aqueous matrix. The role of the AOCM substrates did not illustrate a significant effect in a DI matrix. While the kinetic rates for the commercially available media were all below that of AOCM forms in a DI matrix, these media exhibited varied capacity and decreased in the order from Fe coated perlite, activated alumina, expanded shale and bioretention soil media. Tire crumb (B&G) media used for green roof applications and zeolite perlite GAC mass transfer rates to and from were equal and therefore net equilibrium capacity was not significantly greater than 0. For the DI matrix the 2nd order potential driving model simulated the overall mass transfer rate and the intramedia diffusion model identified the rate limiting diffusion of the permeable AOCM and other media. The DI matrix is a controlled and reproducible matrix that allows comparison of media on an equivalent basis but a basis that may not represent the complex interactions of actual matrices such as rainfall runoff and wastewater. In order to exam ine the kinetic rates of DP mass transfer for actual aqueous matrices, municipal wet weather flow (rainfall runoff, C0 = 2.7 mg/L) and dry weather (secondary wastewater, C0 = 11.1 mg/L) matrices are tested with AOCM and UCM. Again each AOCM out performed t he UCM substrate. However, while the type of AOCM substrate had an insignificant effect in the synthetic matrix, AOCMp kinetics rates and equilibrium capacity are significantly lower than AOCMc and AOCMpcc in both rainfall runoff and wastewater matrices.

PAGE 159

159 The choice of an aqueous matrix impacts resulting media parameters as developed from the 2nd order model. Therefore to investigate the effect of the aqueous matrix the DP concentration in each matrix (synthetic and actual) are adjusted to that of actual r ainfall runoff (2.7 mg/L) to eliminate the interference of different driving forces. The presence of SO4 2 and Ca2+ in the actual matrices can exert contrary effects on phosphate mass transfer, as SO4 2 competes for the active sites with phosphate ions thr ough ligand exchange and Ca2+ enhances the surface precipitation of phosphate as Ca3(PO4)2. Results demonstrated that the competition effects of SO4 2 dominates in the rainfall runoff matrix and decreases the AOCM overall mass transfer rate. While this com petitive effect is observed for AOCMc and AOCMp in wastewater, surface precipitation overweighed the competition effect for AOCMpcc, due to the high concentration of Ca2+ (56 mg/L) in the aqueous matrix as well as the alkaline condition in aqueous phase cr eated through contact with concretebased media, resulting in a faster kinetics. Finally, c olumn adsorption breakthrough experiments were conducted at a surface loading rate of 40 L/minm2 and initial P concentration of 0.5 mg/L in a fixed bed reactor with L/D =5 for all of the media studied, followed by desorption tests 24 hrs later. In these breakthrough and desorption examinations the aqueous matrix utilized is a controlled synthetic runoff matrix of D.I. water, recognizing that while this matrix does not provide the complexity of rainfall runoff or wastewater, the matrix is consistently reproducible for comparative or parameterization studies. Results indicate that AOCMc had higher breakthrough (X/Mb) and exhaustion (X/Mexh) capacity as compared to AOCMpcc and AOCMp; AOCMc largely as a function of

PAGE 160

160 the substrate providing higher specific surface area for the AlOx. While AOCMpcc has a higher breakthrough capacity than AOCMp the exhaustion capacity is lower; both of these results a function of surface precipitates and the instability thereof at the AOCMpcc surface, in addition to ligand exchange generated by the AlOx. In contrast the primary adsorption mechanism of AOCMc and AOCMp is ligand exchange. Results also indicate the nonengineered media has very limited DP breakthrough capacity (adsorption rate desorption rate) compared to the AlOx adsorptive media. Leaching results using D.I. water at a pH of 7.0 and an SLR of 4 0 L/min m2 indicate that a significant, in some cases a predominant fraction of DP adsorbed, is subsequently leached despite a 24 hour inter event time providing for intraparticle mass transfer. In contrast, all of the three AOCM exhibit low desorption wi th AOCMc providing the lowest leaching concentrations and AOCMpcc providing a higher but nominal level of leaching from surface precipitates. Comparing breakthrough results between the synthetic matrix and rainfall runoff or wastewater matrices for AOCMc a nd AOCMpcc respectively; results illustrate differing behavior. Competing solutes reduce the breakthrough capacity of AOCMc primarily functioning through ligand exchange. In contrast, the additional mechanism of calcium based precipitation by AOCMpcc gen erated by the CaO on the hydrated cementitious surface and the higher Ca concentrations in solution results in an increase in breakthrough capacity for AOCMpcc in wastewater. Results indicate that the surface properties of the media whether providing ligand exchange or a separate mechanism such as chemical precipitation must be considered in conjunction with the aqueous matrix.

PAGE 161

161 SLR ranging from 10 to 160 L/minm2, typical of SLRs to runoff filter systems, is a primary explanatory variable for DP breakthro ugh from adsorptive media. Increasing SLR results in consistently lower BVs at breakthrough and exhaustion levels. The Thomas model reproduced physical model data across all SLRs. The Thomas model rate constant (KT) and mediaphase capacity (q0) are a f unction of SLR for a given adsorptive media; media where the adsorption rate >> desorption rate or media that did not include a significant surface precipitation mechanism in addition to an exchange reaction. Mathematical relationships (linear and power law) to SLR are demonstrated for KT and q0, respectively. The proportionality constants relating KT and SLR are significantly different (p = 0.05) as are the power law exponent relating q0 and SLR for the adsorptive media tested.

PAGE 162

162 LIST OF REFERE NCES dm K., Krogstad T., Vrle L., Svik A. K. and Jenssen P. D. (2007) Phosphorus retention in the filter materials shellsand and Filtrate P Batch and column experiment with synthetic P solution and secondary wastewater. Ecol. Eng., 29, 200 208 Agyei N. M., Strydom C. A. and Potgieter J. H. (2002) The removal of phosphate ions from aqueous solution by fly ash, slag, ordinary Portland cement and related blends. Cem. Concr. Res, 32, 1889 1897 Akbal F. (2005) Adsorption of basic dyes from aqueous solution onto pumice powder. J. Colloid Interface Sci., 286, 455 458 Aksu Z., Continuous fixed bed biosorption of reactive dyes by dried Rhizopus arrhizus: determination of c olumn capacity. 143, 362 371 Aksu Z. and Gnen F. (2004) biosorption of phenol by immobilized activated sludge in a continuous packed bed: prediction of breakthrough curves. Process Biochem., 39, 599 613 Aldridge K. T. and Ganf G. G. (2003) Mod ification of sediment redox potential by three contrasting macrophytes: implications for phosphorus adsorption/desorption. Mar. Freshwater Res., 54, 87 94 Alvin W. M. Ip, Barford J. P. and McKay G. (2010) A comparative study on the kinetics and mechanisms of removal of Reactive Black 5 by adsorption onto activated carbons and bone char. Chem. Eng. J., 157, 434 442 Appan A. and Wang H. (2000) Sorption isotherms and kinetics of sediment phosphorus in a tropical reservoir. J. Environ. Eng. 126 (11), 993 998 Arias M., Da Silva Carballal J., GarcaRo L., Mejuto J. and Nez A. (2006) Retention of phosphorus by iron and aluminum oxidecoated quartz particles. J. Colloid Interface Sci., 295, 65 70 Ayari F., Srasra E., and Trabelsi Ayadi M. (2005) Characterization of bentonitic clays and their use as adsorbent. Desalination, 185, 391 397 Ayoub G. M., Koopman B. and Pandy N. (2001) Coated filter media for low concentration of phosphorus removal. Water Environ. Res., 73, 478 485 Barrett M. E., Walsh P. M., Malina J. F., and Charbeneau R. J. (1998) Performance of vegetative controls for treating highway runoff. J. Environ. Eng., 124 (11), 11211128

PAGE 163

163 Boari G., Liberti L. and Passino R. (1976) Selective renovation of eutrophic wastes: phosphate removal. Water Res., 10, 421 428 Boujelben N., Bouzid J., Elouear Z., Feki M., Jamoussi F. and Montiel A.(2008) Phosphorus removal from aqueous solution using iron coated natural and eng ineered sorbents. J. Hazard. Mater., 151, 103 110 Bowes M.J., House W.A., Hadgkinson R.A. and Leach D.V. (2005) Phosphorus discharge hysteresis during storm events along a rivers catchment: the river swale,UK. Water Res., 29, 751762 Bowman B. T. ( 1981) Anomalies in the log Freundlich equation resulting in deviations in adsorption K values of pesticides and other organic compounds when the system of units is changed. J. Environ. Sci. Health B 16 (2), 113 123 Bowman B. T. (1982) Conversion o f Freundlich adsorption K values to the mole fraction format and the use of Sy values to express relative adsorption of pesticides. Soil Sci. Soc. Am. J., 46, 740 743 Braja M. D. (2001) Principles of Geotechnical Engineering 5th edition, Thomson Learning Bubba M. D., Arias C. A. and Brix H. (2003) Phosphorus adsorption maximum of sandsfor use as media in subsurface flow constructed reed beds as measured by the Langmuir isotherm. Water Res., 37, 3390 3400 Chandrasekharan R., Zhang L., Ostroverkhov V., Prakash S., Wu Y., Shen Y. and Shannon M. A.(2008) High temperature hydroxylation of alumina crystalline surfaces. Surf. Sci., 602, 1466 1474 Chang Y., Li C.W. and Benjamin M. M. (1997) Iron oxidecoated media for NOM sorption and particulate filtration. J. Am. Water Works Assoc., 89, 100 113 Chen J. P., Chua M L and Zhang B. (2002) Effects of competitive ions, humic acid and pH on removal of ammonium and phosphorous from the synthetic industrial effluent by ion exchange resins. Waste Manage., 22, 711 719 Chen Z., Xing B. and McGill W. B. (1999) A unified sorption variable for environmental applications of the Freundlich Equation. J. Environ. Qual., 28, 1422 1428 Christidis G. E., Scott P. W. and Dunham A. C. (1997) Acid activation and bleaching capacity of bentonites from the islands of Milos and Chios, Aegean, Greece. Appl. Clay Sci., 12, 329 347

PAGE 164

164 Chitrakar R., Tezuka S., Sonoda A., Sakane K., Ooi K. and Hirotsu T. (2006) Selective adsorption of phosphate fr om seawater and wastewater by amorphous zirconium hydroxide. J. Colloid Interface Sci., 297, 426 433 Chitrakar R., Tezuka S., Sonoda A., Sakane K., Ooi K. and Takahiro H. (2006) Phosphate adsorption on synthetic goethite and akaganeite. J. Colloid I nterface Sci., 298, 602 608 Choy K. K. H., Porter J. F. and McKay G. (2004) Intraparticle diffusion in single and multi component acid dye adsorption from wastewater onto carbon. Chem. Eng. J., 103, 133 145 Coles C. A. and Yong R. N. (2006) Use of equilibrium and initial metal concentrations in determining Freundlich isotherms for soils and sediments. Eng. Geol., 85, 19 25 Phosphate adsorption and desorption in relation to morphology and crystal properties of synthetic hematites. Geochim. Cosmochim Acta, 58 (4), 1261 1269 Comings K. J., Booth D. B., and Horner R. R. (2000) Storm water pollutant removal by two wet ponds in Bellevue, Washington. J. Environ. Eng., 126 (4), 321330 Costa H. S., Pereira M. M. and Andrade G. I. (2007) Characterization of calcium phosphate coating and zinc incorporation on the porous alumina scaffolds. Mat. Res., 10 (1), 27 29 Danny C. K., Ko K., Porter J. F. and Mckay G.(2000) Optimized correlations for the fixed bed adsorption metal ions on bone char. 55, 5819 5829 Danny C. K., Ko K., Porter J. F. and Mckay G. (2001) Determination of solidphase loading for the removal of metal ion form effluents using fixedbed adsorbers. Envion. Sci. Technol., 35, 2797 2803 Das J., Patra B. S., Baliarsingh N. and Parida K. M. (2006) Adsorption of phosphate by layered double hydroxides in aqueous solutions. Appl. Clay Sci., 32, 252 260 Davis A. P. (2007) Field performance of bioretention: water qualit y. Environ. Eng. Sci., 24 (8) 1048 1064 Davis A. P., Hunt W. F., Traver R. G. and Clar M. (2009) Bioretention technology: overview of current practice and future needs. J. Envi r on. Eng., 135 (3), 109 117

PAGE 165

165 Davis P., Shokouhian M., Sharma H., and Minami C. (2006) Water qualityimprovement through bioretention media: nitrogen and phosphorus removal. Water Environ. Res. 78 (3), 284 293 Dayton E. A., Basta N. T., Jakober C. A. and Hattey J. A. (2003) Using treatment residuals to reduce phosphorus in agriculture runoff. J. Am. Water Works Assoc., 95 (4), 151 158 Dean C. M., Sansalone J. J., Cartleedge F. K. and Pardue J. H. (2005) Influence of hydrology on rainfall runoff metal element speciation. J. Environ. Eng., 131 (4), 632 642 Dech esne M., Barraud S., and Bardin J. P. (2005) Experimental assessment of stormwater infiltration basin evolution. J. Environ. Eng., 131 (7), 10901098 Drapper D., Tomlinson R. and Williams P. (2000) Pollutant concentrations in road runoff: Southeast Queensland case study. J. Environ. Eng. 124 313 320 Drizo A., Forget C., Chapuis R. P. and Comeau Y. (2006) Phosphorus removal by electric arc furnace steel slag and serpentinite. Water Res., 40 (8), 1547 1554 Duda A.M. (1993) Addressing nonpoint sources of water pollution must become an international priority. Water Sci. Technol., 28, 111 Eaton A. D., Clesceri L. S. and Greenberg A. E. (1998) Standard Methods for the Examination of Water and Wastewater. 20th Edition. American Public Health Association, American Water Works Association and Water Environment Federation. Edwards A. C. and Withers P. J. A. (2007) Linking phosphorus sources to impacts in different types of water body. Soil Use Manage., 23, 133 143 Edzwald J. K., Toensing D. C. and Leung M. C Y (1976) Phosphate adsorption reactions with clay minerals. Environ. Sci. Technol., 10 (5), 485 490 Everett D. H. (1993) Some problems in the study of the heterogeneity of solid surfaces. Langmuir, 9, 2586 2592 Erickson A. J., Gulliver J. S. and Weiss P. T. (2007) Enhanced sand filtration for storm water phosphorus removal. J. Environ. Eng., 133 (5), 485 497 Floriano M. A. (1994) The structure of pumice by neutron diffraction. J. Appl. Cryst., 27, 271 277

PAGE 166

166 Fontes M. P. F. and Weed S. B. (1996) Phosphate adsorption by clays from Brazilian Oxisols: relationships with specific surface area and mineralogy. Geoderma, 72, 37 51 Forbes M. G., Dickson K. R., Golden T. D., Hudak P. and Doyle R. D. (2004) Dissolved phosphorus retention of light weight expanded shale and masonry sand used in subsurface flow treatment wetlands. Environ. Sci. Technol., 38, 892 898 F u J. F., Zhao Y. Q., Razali M. and Bruen M. (2008) Response surface optimization of phosphorus species adsorption onto powdered alum sludge. J. Environ. Sci. Health, Part A, 43, 1100 1107 Gao Y., Wahi R., Kan A. T., Falkner J. C., Colvin V. L. and To mson M. B. (2004) Adsorption of cadmium on anatase nanoparticles s e ffect of crystal size and pH. Langmuir, 20, 9585 9593 Gen Fuhrman H., Bregnhj H., and McConchie D. (2005) Arsenate removal from water using sandred mud columns. Water Res., 39, 2944 2954 Genz A., Kornmller A. and Jekel M. (2004) Advanced phosphorus removal from membrane filtrates by adsorption on activated aluminum oxide and granulated ferric hydroxide. Water Res., 38, 3523 3530 Georgantas D. A. and Grigoropoulou H. P. (2007) Orthophosphate and metaphosphateion removal from aqueous solution using alum and aluminum hydroxide. J. Colloid Interface Sci., 315, 70 79 Gervin L., and Brix H. (2001) Removal of nutrients from combined sewer overflows and lake water in a vertical flow constructed wetland system. Water Sci. Technol., 44(1112), 171 176 Goldberg S., Lebron I., Suarez D. L. and Hinedi Z. R. (2001) Surface characterization of amorphous aluminum oxides. Soil Sci. Soc. Am. J., 65, 78 86 Goldberg S. and S posito G. (1985) On the mechanism of specific phosphate adsorption by hydroxylated mineral surfaces: A review. Commun. Soil Sci. Plant Anal., 16, 801 821 Gomez E., Durillon C., Rofes G. and Picot B. (1999) Phosphate adsorption and release from sedi ments of brackish lagoons: pH, O2 and loading influence. Water Res., 33, 2437 2447 Gray S., Kinross J., Read P. and Marland A. (2000) The nutrient assimilative capacity of maerl as a substrate in constructed wetland systems for waste treatment. Wate r Res., 34, 2183 2190

PAGE 167

167 Guan X H., Liu Q., Chen GH., Shang C. (2005) Surface complexation of condensed phosphate to aluminum hydroxide: An ATR FTIR spectroscopic investigation. J. Colloid Interface Sci., 289, 319 327 Gustafsson J. P., Renman A., Renman G. and Poll K. (2008) Phosphate removal by mineral based sorbents used in filters for small scale wastewater treatment. Water Res., 42, 189 197 Hameed B. H. and Daud F. B. M. (2008) Adsorption studies of basic dye on activated carbon derived from agriculture waste: Hevea brasiliensis seed coat. Chem. Eng. J., 139, 48 55 Hardin M. E.I. (2006) The effectiveness of a specifically designed green roof stormwater treatment system irrigated with recycled stormwater runoff to achieve pollutant removal and stormwater volume reduction. Master thesis, University of Central Florida Havens K. E. and Schelske C. L. (2004) The importance of considering biological processes when setting total maximum daily loads (TMDL) for phosphorus in shallow lak es and reservoirs. Environ. Pollut., 113 (1), 1 9 Heal K. V., Smith K. A., Younger P. L. and McHaffie H. and Batty L. C. (2004) Removing P from sewage effluent and agriculture runoff using ochre recovered from mine eater treatment In: Valsami Jones E. (Ed.), Phosphorus in Environmental Technology: Removal, Recovery and Applications. IWA Publishing, London, 320 334 Hietala S. L., Smith D. M., Golden J. L. and Brinker C. J. (1989) Anomalously low surface area and density in the silicaalumina gel system. J. Am. Ceram. Soc. 72, 2354 2358 Hietala S. L., Smith D. M., Brinker C. J., Hurd A. J., Carim A. H. and Dando N. (1990) Structural studies of anomalous behavior in the silicaalumina gel system. J. Am Ceram. Soc., 73, 2815 2821 Hsieh C. h and Davis A.P. (2005) Evaluation and optimization of bioretention media for treatment of urban storm water runoff, J. Environ. Eng., 131 (11), 15211531 Hsieh C. h., Davis A.P. and Needelman B.A. (2007) Bioretention column studies of phosphorus removal from urban stormwater runoff, Water Environ. Res. 79 (2), 177184 Ho Y. and McKay G. (1999) Pseudosecond order model for sorption processes. Process Biochem., 34, 451 465

PAGE 168

168 Huang C. (1975) Al2O3 el ectrolyte interface. J. Colloid Interface Sci., 53, 178 186 Huang W., Wang S., Zhu Z., Li L., Yao X., Rudolph V. and Haghseresht F. (2008) Phosphate removal from wastewater using red mud. J. Hazard. Mater., 158, 35 42 Hunt W. F., Jarrett A. R., S mith J. T., and Sharkey L. J. (2006) Evaluating bioretention hydrology and nutrient removal at three field sites in North Carolina. J. Irrig. Drain. Eng 132 (6), 600 608 Hunt W. F., Smith J. T., Jadlock S. J., Hathaway J. M. and Eubanks P. R. (2008) Pollutant removal and peak flow mitigation by a bioretention cell in urban charlotte, N.C. J. of Environ. Eng., 134 (5), 403 408 Johnson B. B., Ivanov A. V., Antzutkin O. N. and Forsling W. (2002) 31P nuclear magnetic resonance study of the adsorption of phosphate and phenyl phosphate Al2O3. Langmuir, 18, 1104 1111 Johansson L. (1999) Industrial by products and natural substrata as phosphorus sorbents. Environ. Technol., 20, 309 316 Kang S K, Choo K H and Lim K H (2003) Use o f iron oxide particles as adsorbents to enhance phosphorus removal form secondary wastewater effluent. Separ. Sci. Technol., 38 (15)., 3853 3874 Kinetic modeling of liquid phase adsorption of phosphate on dolomite. J. Colloid Interface Sci., 277, 257 263 Karageorgiou K., Paschalis M. and Anastassakis G. N. (2007) Removal of phosphate species from solution by adsorption onto calcite used as natural adsorbent. J. Hazard. Mater., 39, 447 452 Karathanasis A. D. and Shumaker P. D. (2009) Organic and inorganic phosphate interactions with soil hydroxyl interlayered minerals. J. Soils Sediments, 9, 501 510 Khaled A., Nemr A. E., Amany ES and Abdelwahab O. (2009) Treatment of artificial t extile dye effluent containing Direct Yellow 12 by orange peel carbon. Desalination, 238, 210 232 Kim J., Mann J. D. and Kwon S. (2006) Enhanced adsorption and regeneration with lignocelluloses based phosphorus removal media using molecular coating nanotechnology. J. Environ. Sci. Health, Part A, 41, 87 100

PAGE 169

169 Kim J Y ., Ma J., Howerter K., Garofalo G., and S ansalone J.J (2008) Interactions Of Phosphorus With Anthropogenic And Engineered Particulate Matter As A Function Of Mass, Number And Surface Area, Urban Water Systems, Monograph 11, Edited by James W. CHI Publications, Guelph, Ontario, 20 pp. February. Kim Y. and Kirkpat rick R. J. (2004) An investigation of phosphate adsorbed on aluminum oxyhydroxide and oxide phases by nuclear magnetic resonance. Eur. J. Soil Sci., 55, 243 251 Kitis M., Kaplan S. S., Karakaya E., Yigit N. O. and Civelekoglu G. (2007) Adsorption o f natural organic matter from waters by iorn coated pumice. Chemosphere, 66, 130 138 Klute A. (1986) Methods of soil analysis Part 1: Physical and mineralogical methods. American Society of Agronomy, Inc., Soil Science Society of America, Inc., Madi son, Wisconsin, USA Ko H., Davis A. P., Kim J Y, and Kim K W (2007) Arsenic removal by a colloidal iron oxide coated sand. J. Environ.Eng., 133 (9), 891898 Koh J H, Wankat P. C. and Wang N. H. L. (1998) Pore and surface diffusion and bulk phase mas s transfer in packed and fluidized beds. Ind. Eng. Chem. Res., 37, 228 239 Kse T. E. and ztrk N. (2008) Boron removal from aqueous solutions by ionexchange resin: column sorptionelution studies. J. Hazard. Mater., 152, 744 749 Kostura B., Ku lveitov H. and Leko J. (2005) Blast furnace slags as sorbents of phosphate form water solutions. Water Res., 39, 1795 1802 Kuzawa K., Jung Y J., Kiso Y., Yamada T., Nagai M. and Lee T G (2006) Phosphate removal and recovery with a synthetic hydrotalcite as an adsorbent. Chemosphere., 62, 45 52 Lee V. K. C., Porter J. F. and McKay G. (2000) Development of fixedbed adsorber correlation models. Ind. Eng. Chem. Res., 39 (7), 2427 2433 Lee V. K. C, Porter J. F., and Mckay G. (2003) Fixed bed modeling for acid dye adsorption onto activated carbon. J. Chem. Techol. Biotechnol., 78, 1281 1289 Li H. and Davis A. P. (2008) Urban particle capture in bioretention media I: Laboratory and field studies. J. Environ. Eng., 134 (6), 409 418

PAGE 170

170 Lin C. Y. and Yang D. H. (2002) Removal of pollutants from wastewater by coal bottom ash. J. Environ. Sci. Health, Part A: Toxic/Hazard. Subst. Environ., 37, 1509 1522 Lisi R. D., Park J. K. and Stier J. C. (2004) Mitigating nutrient leaching with a sub surface drainage layer of granulated tires. Waste Mange., 24, 831 839 Liu H. S., Mead J. L. and Stacer R. G. (2000) Environmental effects of recycled rubber in light fill applications. Rubber Chem. Technol., 73, 551 564 Liu D., Sansalone J J., and Cartledge F. K. (2005a) Adsorption kinetics for urban rainfall runoff metals by composite oxidecoated polymeric media. J. Environ. Eng.,131 (8), 1168 1177 Liu D., Sansalone J. J., and Cartledge F. K. (2005b) Comparison of sorptive filter media for treatment of metals in runoff. J. Environ. Eng., 131 (8), 1178 1186 Liu D., Teng Z., Sansalone J. J. and Cartledge F. K. (2001) Surface characteristics of sorptivefiltration storm water media. I: Low s < 1.0) oxidecoatd buoyant media. J. Environ. Eng., 127 (10), 868 878 Liu X. (1999) alumina with FCC catalysts Via steaming. J. Phys. Chem. B, 103, 2647 2652 Lu S G, Bai S Q and Shan H D (2008) Mechanisms of phosphate removal from aqueous solution by blast furnace slag and steel furnace slag. J. Zhejiang Univ. Sci. A, 9 (1), 125 132 Ma J. (2005) Speciation, equilibrium distribution and kinetics of rainfall runoff phosphorus in adsorptivefiltration unit opera tion and process. PhD dissertation, Louisiana State University Makris K. C., Hassan E H, Harris W. G., OConnor G. A. and Obreza T. A. (2004) Intraparticle phosphorus diffusion in a drinking water treatment residual at room temperature. J. Colloid Int erface Sci., 277, 417 423 Malkoc E. and Nuboglu Y. (2006) Fixed bed studies for the sorption of chromium (VI) onto tea factory waste. Chem. Eng. Sci., 61, 4363 4372 Mantovaneli I. C. C., Ferretti E. C., Simes M. R. and Ferreira da Silva C. (2004) The effect of temperature and flow rate on the clarification of the aqueous stevia extract in a fixedbed column with zeolites. Braz. J. Chem. Eng., 21 (3), 449 458 Marchi G., Guilherme L. R. G. and Lima J. M. (2006) Adsorption/Desorption of organic anions in Brazilian oxisols. Commun. Soil Sci. Plant Anal., 37, 1367 1379

PAGE 171

171 Mehta P.K., (1987) Concrete Structure, Properties, and Materials. PrenticeHall, Inc. New Jersey, pp. 453. Metcal and Eddy (2003) Wastewater engineering: treatment and reuse. McGraw Hill. Millero F., Huang F., Zhu X., Liu X. and Zhang, J (2001) Adsorption and desorption of phosphate oncalcite and aragonite in seawater. Aqua. Geochem., 7, 33 56 Morillo E., Undabeytia T., Cabrera A., Villaverde J. and Maqueda C. (2004) Effect of soil type on adsorption desorption, mobility, and activity of the herbicide norflurazon. J. Agric. Food Chem., 52, 884 890 Mortula M., Gibbons M. and Gagnon G. A. (2007) Phosphorus adsorption by naturally occurring materials and industrial by products. J. Environ. Eng. Sci., 6, 157 164 Namasivayam C. and Sangeetha D. (2004) Equilibrium and kinetic studies of adsorption of phosphate onto ZnCl2 activated coir pith carbon. J. Colloid Interface Sci., 280, 359 65 Nooney M. G., Campbell A., Murrell T. S., Lin X. F., Hossner L. R., Chusuei C. C. and Goodman D. W. (1998) Nucleation and growth of phosphate on metal oxide thin films. Langmuir, 14, 2750 2755 Sorption of phosphate from solid/liquid interface by fly ash. Colloids and Surfaces A: physiochem. Eng. Aspects, 262, 113 117 Removal of phosphate form wastewater. Cem. Concr. Res., 33, 1109 1112 Oleszkiewicz J. A. a nd Barnard J. L. (2006) Nutrient removal technology in North America and the European Union: a review. Water Qual. Res. J. Can., 41 (4), 449 462 Preparation and characterization of acidactivated bentonite powders. P owder Technol., 172, 14 18 Onar A. N., Balkaya N. and Akyz T. (1996) Phosphate removal by adsorption. Environ. Tech., 17 (2), 207 213 zacar M. (2003a) Equilibrium and kinetic modeling of adsorption of phosphorus on calcined alunite. Adsorption, 9, 125 132 zacar M. (2003b) Phosphate adsorption characteristics of alunite to be used as a cement additive. Cem. Concr. Res., 33, 1583 1587

PAGE 172

172 Rao J. R. and Viraraghavan T. (2007) Biosorption of phenol from an aqueous solution by Aspergillus niger biomass. 85, 165 171 Raposo F., De La Rubia M. A. and Borja R. (2009) Methylene blue number as useful indicator to evaluate the adsorptive capacity of granular activated carbon in batch Read P. and Fernandes T. (2003) Management of environmen tal impacts of marine aquaculture in Europe. Aquaculture, 226 (14), 139 163 Pafitt R. L. (1978) Anion adsorption by soils and soil materials. Adv. Agron., 30, 1 50 Panuccio M. R., Sorgon A., Rizzo M. and Cacco G. (2009) Cadmium adsorption on vermiculite, zeolite and pumice: Batch experimental studies. J. Environ Manage. 90, 364 374 Penn C. J., Mullins G. L. and Zelazny L. W. (2005) Mineralogy in relation to phosphorus sorption and dissolved phosphorus losses in runoff. Soil Sci. Soc. Am. J., 69, 1532 1540 Pokhrel D. and Viraraghavan T. (2008) Arsenic removal in an iron oxidecoated fungal biomass column: analysis of breakthrough curves. Bioresour. Technol., 99, 2067 2071 Pollio F. X. and Kumin R. (1968) Tertiary treatment of municipal sewage effluents. Environ.Sci. Technol., 2 (1), 54 61 Rajan S. S.S. (1975) Adsorption of divalent phosphate on hydrous aluminum oxide. Nature, 253, 434 436 Rocha M. S. D., Iha K., Faleiros A. C., Corat E. J., Surez Iha M. E. V. (1997) Freundlichs isotherm extended by statistical mechanics. J. Colloid Interface Sci., 185, 493 496 Sakadevan K. and B avor H. J. (1998) Phosphate adsorption characteristics of soils, slags and zeolite to be used as substrates in constructed wetland s ystems. Water Res., 32, 393 399 Sansalone J. J. (2005) Perspective on the synthesis of unit operations and process (UOP) concepts with hydrologic controls for rainfall runoff, J. Environ. Eng, 131 (7), 995 997. Sansalone J. J. and Buchberger S. G. (1997) Partitioning and first fluch of metals in urban roadway storm water. J. Environ. Eng., 123 (2), 134 143

PAGE 173

173 Sansalone J. J., Koran J. M., Smithson J A. and Buchberger S. G. (1998) Physical characteristics of urban roadway solids transported during rain. J. Environ. Eng., 124 (5), 427 440 Sansalone J. J., Liu B. and Ying G. (2010) Volumetic Filtration of rainfallrunoff. II: event based and interevent nutrient fate. J. Environ. Eng., 136 (12) Sansalone J. J. and Ma J. (2009) Parametric evaluation of batch equilibria for stormwater phosphorus adsorption on aluminum oxide media, J. Envion. Eng, 135 (9), 737 746 Sansalone J.J., and Teng Z. (2005a). Transient rainfall runoff loadings to a partial exfiltration system: implic ations for urban water quantity and quality J. Environ. Eng., 130 (9), 9901007 Sansalone J.J. and Teng Z. (2005b) Transient rainfall runoff loadings to a partial exfiltration system: implications for urban water quantity and quality, J. Environ. Eng ., 131 (8), 11551167 Schwarzenbach R. P., Gschwend P. M., Imboden D. M. Environmental Organic Chemistry 2nd ed. (2002) Wiley: Hoboken, NJ Seo D. C., Cho J. S., Lee H. J., and Heo J. S. (2005) Phosphorus retention capacity of filter media for stima ting the longevity of constructed wetland. Water Res., 39(11), 24452457 Shilton A., Pratt S., Drizo A., Mahmood, B., Banker S., Billings L., Glenny S. and Luo D. (2005) Active filters for upgrading phosphorus removal from pond systems. Water Sci.T echnol., 51 (12), 111 116 Shin E. W. and Han J. S. (2004) Phosphate adsorption on aluminum impregnated mesoporous silicates: Surface structure and behavior of adsorbents. Environ.Sci T echnol., 38, 912 917 Sibrell P. L., Montgomery G. A., Ritenour K. L. and Tucker T. W. (2009) Removal of phosphorus from agricultural wastewaters using adsorption media prepared from acid mine drainage sludge. Water Res., 43, 2240 2250 Sivakumar P. and Palanisamy P. N. (2009) Adsorptive removal of reactive and direct dyes using nonconventional adsorbent column studies. J. Sci. Ind. Res., 894 899 Smidt E., Eckhardt K., Lechner P., Schulten H. and Leinweber P. (2005) Characterization of different decomposition stages of biowaste using FT IR spectroscopy and pyrolysis field ionization mass spectrometry. Biodegradation, 16, 67 79

PAGE 174

174 Sperlich A., Werner A., Benz A., Amy G., Worch E. and Jekel M. (2005) Breakthrough Behavior of Granular Ferric Hydroxide (GFH) FixedBed Adsorption Filters: Modeling and Experimental Approaches.., Water Res., 39, 11901198. Suzuki M. (1990) Adsorption engineering, Elsevier Science, Chemical engineering monographs; vol. 25 Teng Z.and Sansalone J.J. (2004). In situ partial exfiltration of rainfall runoff. II: particle separation. J. Environ. Eng., 130 (9), 10081020 Thirunavukkarasu O.S., Viraraghavan T. and Subramanian K.S. (2003) Arsenic removal from drinking water using iron oxidecoated Sand. Water Air, Soil Pollut., 142, 95 11 1 Tien C. (1994) Adsorption calculation and modeling, ButterworthHeinemann series in chemical engineering, Newton, MA Urano K. and Tachikawa H. (1991) Process development for removal and recovery of phosphorus from wastewater by a new adsorbent. 1. Preparation method and adsorption capability of a new adsorbent. Ind. Eng. Chem. Res., 30, 1893 1896 U. S. Environmental Protection Agency (USEPA). (1986) Quality criteria for water. Office of Water Regulation and Standards, US Government Printing Office (PB81 226759), Washington, DC 20460 U. S. Environmental Protection Agency (USEPA) (1993). Handbook: Urban Runoff Pollution Prevention and Control Planning. EPA l625/R 93/004 Office of Research and Development, Washington, DC U. S. Environmental Protection Agency (USEPA) (1996) National Water Quality Inventory, Reports Congress, U. S. Environmental Protection Agency, Office of Water: Washington, D. C., Sec. 1. U S. Environmental Protection Agency (USEPA). (2002a) Environmental assessment for the final effluent limitations guidelines, pre treatment standards for new and existing sources and new source performance standards for the centralized waste treatment. U. S. Environmental Protection Agency (USEPA). (2002b) National water quality inventory: 2000 Report. EPA841 R 02001 USEPA, Washington, DC. U S Environmental Protection Agency (USEPA) (2003) Decentralized approaches to wastewater treatment [online]. available from http://www.epa.gov/owm/septic/pubs/septic_presentation.pdf

PAGE 175

175 U. S. Geological Survey (1999) The quality of our nations waters: nutrients andpesticides. U.S. Geological Survey Circular 1225. USGS Information Services, Denver, CO, 88 pp. Van der Zee S. E. A. T. M. and van Riemsdijk (1988) Model for long term phosphate reaction kinetics in soil. J. Environ. Qual., 17, 35 41 Van Raij B., and Peech M. (1972) Electrochemical properties of some oxides and alfisols of the tropics. Soil Sc i. Soc. Am. Proc., 36, 587 593 Walker Jr. and William W. (2003) Consideration of variability and uncertainty on phosphorus total maximum daily loads for lakes. J. Water Resour. Plann. Mange., 129 (4), 337 344 Wang G. T., Chen, S., Barber, M. E., a nd Yonge, D. R. (2004). Modeling flow and pollutant removal of wet detention pond treating stormwater runoff. J. Environ. Eng., 130 (11), 13151321 Wang M. K. and Tzou Y. M. (1995) Phosphate sorption by calcite, and ironrich calcareous soils. Geoder ma, 65, 249 261 Water Environment Federation and ASCE (WEF/ASCE) (1998) Urban runoff quality management. WEF Manual of Practice No. 23, ASCE Manual and Rep on Engineering Practice No. 87, WEF, Alexandria, Va., ASCE, Reston, Va Wei X., Viadero Jr. R. C. and Bhojappa S. (2008) Phosphorus removal by acid mine drainage sludge from secondary effluents of municipal wastewater treatment plants. Water Res., 42, 3275 3284 Wu J. S., Holman, R. E., and Dorney, J. R. (1996). Systematic evaluation of pollution removal by urban wet detention ponds. J. Environ. Eng., 122 (11), 983988 Wu T Gnecco, I., Berretta, C., Ma, J. and Sansalone, J. J.(2008) Stormwater phosphorus adsorption on oxidecoated media, Proceedings of the Water Environment Federation, Session 91 through Session 100 pp. 6945 6957(13), Chicago, IL Yang Y., Tomlinson D., Kennedy S. and Zhao Y. Q. (2006) Dewatered alum sludge: a potential adsorbent for phosphorus removal. Water Sci. Tech., 54 (5), 207 213 Yang Y., Zhao Y. Q., Babatunde A. O., Wang L., Ren Y. X., and Han Y. (2006) Characteristics and mechanisms of phosphate adsorption o n dewatered alum sludge. Sep. Purif. Technol., 51, 193 200

PAGE 176

176 Yoon S., Prezzi M., Siddiki N. Z. and Kim B. (2006) Construction of a test embankment using a sandtire shred mixture as fill material. Waste Mange., 26, 1033 1044 Yoshida H., Jitsukawa H. and Galinada W. A. (2004) Breakthrough and elution curves for adsorption of phosphates on an OH type strongly basic ion exchanger. Ind. Eng. Chem. Res., 43, 3394 3402 Yoshida I. (1983) Studies on the selective adsorption of anion by metal ion lo aded ionexchanges resin. Sep. Sci. Technol., 18 (1), 73 82 Yu M J, Li X. and Ahn W S (2008) Adsorptive removal of arsenate and orthophosphate anions by mesoporous alumina. Microporous Mesoporous Mater., 113, 197 203 Zeng L., Li X. and Liu J. (2004) Adsorptive removal of phosphate from aqueous solutions using iron oxide tailings. Water Res., 38, 1318 1326 Zhao D. and Sengupta A. K. (1997) Ultimate removal of phosphorus from wastewater using a new class of polymetric ion exchangers. Water Res., 32 (5), 1613 1625

PAGE 177

177 BIOGRAPHICAL SKETCH Tingting Wu received her b achelor s degree in e nvironmental engineering from Tongji University, Peoples Republic of China, in 2001 She got her m asters degree in chemical and biomolecular engineering from National University of Singapore, Singapore, in 2005. She started her doctoral degree in the D epartment of Environmental E ngineering and S cience at University of Florida in August 2006. Her graduate research was focused on equilibria and kinetics of granular filter materials for mass transfer of phosphorus in urban water. She worked under the guidance of Dr. John J. Sansalone. Tingting Wu received the degree of Doctor of Philosophy from University of Florida in December 2010