<%BANNER%>

Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2014-12-31.

DARK ITEM
Permanent Link: http://ufdc.ufl.edu/UFE0043878/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2014-12-31.
Physical Description: Book
Language: english
Creator: Seltzer, Karl M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

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

Notes

Statement of Responsibility: by Karl M Seltzer.
Thesis: Thesis (M.E.)--University of Florida, 2011.
Local: Adviser: Sansalone, John.
Electronic Access: INACCESSIBLE UNTIL 2014-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043878:00001

Permanent Link: http://ufdc.ufl.edu/UFE0043878/00001

Material Information

Title: Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2014-12-31.
Physical Description: Book
Language: english
Creator: Seltzer, Karl M
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2011

Subjects

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

Notes

Statement of Responsibility: by Karl M Seltzer.
Thesis: Thesis (M.E.)--University of Florida, 2011.
Local: Adviser: Sansalone, John.
Electronic Access: INACCESSIBLE UNTIL 2014-12-31

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2011
System ID: UFE0043878:00001


This item has the following downloads:


Full Text

PAGE 1

1 PHOSPHOROUS KINETICS OF URBAN SOURCE AREA PARTITIONING AND VOLUMETRIC TREATMENT OF METALS SPECIES IN RAINFALL RUNOFF By KARL SELTZER A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ENGINEERING UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Karl Seltzer

PAGE 3

3 To my m other, my f ather, my sister, my girlfriend Lauren, my f amily and my f riend s, who have always supported and believed in my goal s and ambitions

PAGE 4

4 ACKNOWLEDGMENTS I would first like to thank Dr. John Sansalone, who guided me throughout the course of my graduate research and thesis writing. Dr. Ben Koopman and Dr. Mike Annable, who served in my committee and support ed my research. A mber Atteberry for helping me meet all of the deadlines. A nd lastly, I would like to thank all my co workers in the Stormwater Unit Operations and Processes Laboratory who helped with all the laboratory analysis and field collections; Hwan Chul Cho, Chri stina Herr, Hao Zhang, Adam Marquez, Josh Dickenson, Aniela Burant, Valerie Thorson, Greg Brenner, Julie Midgette and Moshik Doron.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 7 LIST OF FIGURES ................................ ................................ ................................ ......................... 8 LIST OF ABBREVIATIONS ................................ ................................ ................................ .......... 9 ABSTRACT ................................ ................................ ................................ ................................ ... 12 CHAPTER 1 GLOBAL INTRODUCTION ................................ ................................ ................................ 15 2 PHOSPHOROUS KINETICS IN RAINFALL RUNOFF FROM A BIOGENICALLY LOADED WATERSHED ................................ ................................ ................................ ...... 18 Objectives ................................ ................................ ................................ ............................... 19 Methodology ................................ ................................ ................................ ........................... 20 Source Area Catc hment ................................ ................................ ................................ ... 21 Sampling and Analytical Methods ................................ ................................ .................. 21 Phosphorous Partitioning ................................ ................................ ................................ 23 Characterization of Rainfall Runoff Washoff and Material Transport ........................... 24 Kinetic Equilibrium ................................ ................................ ................................ ......... 25 Statistical Analysis ................................ ................................ ................................ .......... 26 Results ................................ ................................ ................................ ................................ ..... 27 Chemical and Physical Characterization of Rainfall Runoff ................................ .......... 27 Phosphorous Loadings and Partitioning ................................ ................................ .......... 28 Time to Equilibrium ................................ ................................ ................................ ........ 30 Discussion ................................ ................................ ................................ ............................... 31 3 COMPARISON OF METALS REMOVAL IN RAINFALL RUNOFF BETWEEN VARYING UNIT OPERATIONS ................................ ................................ ......................... 44 Objectives ................................ ................................ ................................ ............................... 46 Methodology ................................ ................................ ................................ ........................... 47 Source Area Watershed and Hydrologic Data Collection ................................ ............... 47 Sampling Methods and Analytical Procedures ................................ ................................ 48 Metals Partitioning ................................ ................................ ................................ .......... 49 Event Mean Concentration and Removal Efficiencies ................................ .................... 50 Statistical Analysis ................................ ................................ ................................ .......... 51 Results ................................ ................................ ................................ ................................ ..... 53 Rainfall Runoff and Hydrologic Results ................................ ................................ ......... 53 Metals Loadings and Partitioning ................................ ................................ .................... 54

PAGE 6

6 Metals Removal Efficiency ................................ ................................ ............................. 56 Treatment of Metals by Size Gradation ................................ ................................ ........... 57 Statistical Comparison of Q p and its Relationship to Removal Efficiency ..................... 58 Discussion ................................ ................................ ................................ ............................... 59 4 GLOBAL CONCLUSIONS ................................ ................................ ................................ ... 74 LI ST OF REFERENCES ................................ ................................ ................................ ............... 77 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ......... 81

PAGE 7

7 LIST OF TABLES Table page 2 1 Event based hydrologic parameters for each storm event.. ................................ ............... 34 2 2 Chemical properties for each of the samples from the experimental procedure.. .............. 35 2 3 Physical properties for each of the samples f rom the experimental procedure. ............... 36 2 4 Phosphorous properties for each of the samples from the experimental procedure .......... 37 2 5 Kinetic results for ea ch of the experimental samples. ................................ ....................... 38 2 6 Statistical Variables and A nalysis of samples with a d 50 above 75 m and a d 50 below 75 m.. ................................ ................................ ................................ .................... 39 3 1 Hydrology data for all ten measured storms. ................................ ................................ ..... 61 3 2 Total event mean concentration (EMC) values for the influent and effluent of all ................................ ................................ ............ 62 3 3 10 d 50 and d 90 particle sizes ( m) for each of the 10 sampled storms. ................................ ................................ ................................ .................. 63 3 4 Mass for each PM gradation, PM total and metal. ................................ .............................. 64 3 5 Statistical comparison of particle bound metals species above and below designe d flow capacity for both units. ................................ ................................ .............................. 65

PAGE 8

8 LIST O F FIGURES Figure page 2 1 ....................... 40 2 2 Comparison of Time to Equilibrium and the Scale/Shape Factors of the Gamma Distribution between the two experimental populations. ................................ .................. 41 2 3 Dissolved and particulate bound phosphorous concentrations over time for samples with a d 50 greater than 75 m.. ................................ ................................ ........................... 42 2 4 Dissolved and particulate bound phosphorous concentrations over time for samples with a d 50 less than 75 m. ................................ ................................ ................................ 43 3 1 Time dependent fd values for Zn in runoff samples for 7 August 2010 and 13 August 2010 storm events and probability density functions of the equilibrium c oefficient, Kd, for zinc in the influent and effluent for both units. ................................ ..................... 66 3 2 Influent and Effluent PSDs for the August 1 st and July 31 st storms and August 23 rd and August 21 st ................................ ................................ ................................ ................. 67 3 3 Total mass based Mass for the three size gradations of particulate bound nickel and lead over the course of the study for the two stormwater units. ................................ ........ 68 3 4 Comparison of PM removal efficiency and particle bound metals removal efficiency for each of the storm events.. ................................ ................................ ............................. 69 3 5 Comparison of two storms with similar Q p values, both of which exceed the hydraulic design capacity for each of the respective units.. ................................ .............. 70 3 6 Comparison of two storms wit h similar Q p values, both of which were below the hydraulic design capacity for each of the respective units. ................................ ............... 71 3 7 Particle Bou nd removal efficiencies for each metals species as a function of the maximum flow rate for any given storm. ................................ ................................ ........... 72 3 8 Probability density functions of total particle bound copper total particle bound nickel and dissolved nickel in both the influent and effluent for each unit. ...................... 73

PAGE 9

9 LIST O F ABBREVIATIONS BHS baffled hydrodynamic separator C 0 initial concentration of phosphorous (mg/L) C e equilibrium concentration of phosphorous (mg/L) c(t) time variable of pollutant concentration d 50 particle size at 50% finer by mass DI de ionized D m dissolved metals concentration (mg/L) D p dissolved phosphorous concentration (mg/L) d rain duration of rainfall runoff event (minutes) EMC event mean concentration (mg/L) EMC IN event mean concentration of influent (mg/L) EMC OUT event mean concentration of e ffluent (mg/L) f d dissolved bound fraction of s ample (%) f p p article bound fraction of s ample (%) ICP AES inductively coupled plasma atomic emission spectroscopy IPRT initial pavement residence time (minutes) i rain max maximum rainfall intensity (in/hr) k equilibrium rate constant of second order kinetic model (g/mg min) shape factor of gamma distribution k d equilibrium partitioning coefficient (L/kg) M mass of sample (g) n 1 number of samples in population 1 n 2 number of samples in population 2 NRMSE normalized root mean squared error

PAGE 10

10 PDF probability density function PDH previous dry hours (hours) PM particulate matter PP polypropylene P m particle bound metals concentration (mg/L) P p particle bound phosphorous concentration (mg/L) PSD particle s ize dis tribution PVC polyvinyl chloride q(t) time variable of flow q e mass of phosphorous adsorbed/ mass of particulate matter (mg/g) Q p maximum flow rate (gpm) Q avg average flow rate (gpm) q t mass of phosphorous adsorbed/mass of particulate matter at t=t (mg/g) r 2 coefficient of determination RE(%) removal efficiency (%) S 1 standard deviation of population 1 (minutes) S 2 standard deviation of population 2 (minutes) SSC suspended sediment concentration (mg/L) S p 2 dimensionless statistical variable t 0 i nitial s ampling t ime (minutes) t o test statistic ( = 0.05) t* time to kinetic equilibrium (minutes) t rain event duration (minutes) TDP total dissolved phosphorous (mg/L) TDS total dissolved solids (mg/L)

PAGE 11

11 TM total percentage or concentration of a given metals species (% or mg/L) TP t otal p hosphorous (mg/L) TSS total suspended solids (mg/L) USEPA United States Environmental Protection Agency UOP unit operations and processes V volume of sample (L) V runoff volume of runoff (gal) VF volumetric filtration x 1 mean time to equilibrium for population 1 (minutes) x 2 mean time to equilibrium for population 2 (minutes) scale factor of gamma distribution

PAGE 12

12 Abstract of Thesis Presented to the Graduate School of the University of Florida in Parti al Fulfillment of the Requirements for the Degree of Master of Engineering PHOSPHOROUS KINETICS OF URBAN SOURCE AREA PARTITIONING AND VOLUMETRIC TREATMENT OF METALS SPECIES IN RAINFALL RUNOFF By Karl Seltzer December 2011 Chair: John J. Sansalone Major: Environmental Engineering Sciences The impairment of surface waters from rainfall runoff washoff has increasingly become a matter of concern due to excess loadings of pollutants such as nutrients and metals (Lai and Lam 2009; Vohla et al., 2011; M ahbub et al., 2011). These pollutants can eventually lead to water quality problems, eutrophication and the possible degradation of a water supply (Wang et al., 2003; Hatt et al., 2008; Ellis et al., 1987). Due to these harmful effects, the treatment of r unoff has progressively gained momentum. Various unit operations have been developed are currently being used to target pollutants before they become problematic (Dechesne et al., 2004; Stanley, 1996). Though, before a unit is installed for field use, its ability to adequately treat runoff must be verified. One of the main objectives of this study was to examine the kinetic s of partitioning of phosphorous in runoff samples between the dissolved and the particulate bound phase s Oftentimes, when runoff sam ples are being taken, an automated water quality sampling unit is used (Harmel et al., 2010). As a result, samples are left in the field and not immediately analyzed. In this study, runoff samples were sub sampled at recorded time intervals and the two par titions were divided to inhibit further kinetics. It was observed that over time, the dissolved

PAGE 13

13 phosphate ions adsorbed to the particulate matter present in the samples, ultimately increasing the overall fraction of particle bound phosphorous. This directi onal nature was seen for all the observed samples and the results from a second order kinetic model indicated that equilibrium was generally reached within 240 minutes. Furthermore, the particle size distribution (PSD) within a sample was statistically com pared to the time to equilibrium. It was observed that there is significance in the rate of phase equilibrium and that samples that are more heterodispersed with a larger mean particle size reach equilibrium in less time. The other focus of this study was to compare the removal efficiency of particle bound metals species between varying stormwater treatment un its. In this study, the two units that were analyzed were a baffled hydrodynamic separator and a membrane filtration unit. The baffled hydrodynamic se parator was able to target particulate matter and particle bound pollutants and remove them through hydrodynamic separation and sedimentation while the membrane filtration unit targeted the same type of pollutants but had the added benefit of a filtration mechanism. Both units received influent from the same biogenically loaded, impervious surface area over the course of two months. Both units targeted the removal of particulate matter through sedimentation and hydrodynamic separation but, one had the added element of a filtration mechanism in its treatment train. This filtration mechanism helped target a smaller particulate matter gradation while also providing a safety net lessening the harsh effects of phenomena such as scouring and re suspension. Results showed that both units were capable of removing a significant portion of the particulate bound metals from an influent stream. Though, the treatment unit with the added filtration mechanism generally performed better, especially in regards to the small pa rticle gradations. Furthermore, the influence of a changing flow rate on the removal of metals in runoff was a nalyzed. When comparing the two units against two high flow

PAGE 14

14 storms it was observed that the membrane filtration unit was able to provide better r emoval efficiencies. This was a result of its capability to contain much of the pollutants that were already removed from previous storms. In contrast, the hydrodynamic separator lost a portion of its already removed pollutants in some cases as a result of scour and resuspension. Without the added filtration mechanism, the smaller particles that were already retained by the unit were able to become resuspended and leade the unit through the outlet.

PAGE 15

15 CHAPTER 1 GLOBAL INTRODUCTION The omnipresent growth of impervious surfaces in our urban environment has led to a vast change in the hydrologic response in these said areas. Traffic, as well as agriculture and other industrial practices, add various pollutants to these surfaces that ev entually become washoff in rainfall runoff events (Mahbub et al., 2011, Soranno et al., 1996). In fact, the USEPA stated that runoff is the 4 th leading cause of water quality impairment in receiving water bodies (Tsihrintzis et al., 1997). Due to such harm ful loadings, excess nutrients, particulate matter and metals species find their way into the natural environment where their addition can cause severe harm to both the environment and a water supply source (Hatt et al., 2008, Ellis et al., 1987, Wang et a l., 2003). Nutrients such as phosphorous can dramatically change the landscape of a receiving water body by causing eutrophication when in high concentrations (Agudelo et al., 2010 ; Sharpley et al., 1994). Runoff, in particular, carries a large amount of biogenic mass that is rich in phosphorous. It has even been estimated that the amount of phosphorous storage in terrestrial and freshwater ecosystems are at least 75% greater than they were in the pre industrial times (Bennett et al., 2001). Additionally, metals in runoff can add to the degradation of a natural ecosystem or even a potable water supply (Hatt et al., 2008). This is especially evident in areas where high vehicular loadings are seen. Copper, lead, nickel and zinc, as well as a number of other m etals, come from the wear and tear of brakes, tires and the body of everyday vehicles (Sansalone and Buchberger, 1997). Due to these harmful effects, it is important to treat pollutants in runoff and ensure that the assessments associated with such treatm ents are quantified accurately.

PAGE 16

16 Both nutrients and metals in runoff can exist in different partitions and be associated with various gradations of particle sizes. In total, each pollutant can either be dissolved in the runoff or be adsorb ed to the surroun ding particulate matter. Adding to this complexity, pollutants oftentimes carry kinetic instability causing a portion of one partition to switch over to the other. Phosphorous, in particular, frequently exhibits this motion. When biogenic sediments come in to contact with water, phosphorous will exchange between the water and the sediments until equilibrium is reached (Zhou et al., 2005). This can become problematic if a treatment operation is being assessed and the samples are not immediately analyzed. When measuring the phosphorous concentrations of the various partitions in a runoff samples, what is seen in the laboratory might not be what was seen in the field by the treatment unit during the runoff event if kinetic instability was present. For that reaso n, it is important to separate the dissolved and particulate bound species in a sample before equilibrium is met if an accurate description of how the pollutants in the field behave is needed. In addition to the sampling and analytical techniques being sig nificant to the treatment of runoff, the type of unit and its associated treatment processes are highly important. Runoff treatment units come in various sizes and have different strategies for removing pollutants. Some units have added components to their treatment train, such as porous pavement, that allow them to target a smaller gradation of particles and add an extra layer of safety against phenomena such as scouring (Legret et al., 1999) while others may only have basic sedimentation techniques (Stanl it is important to estimate the added benefit of additional sediment removal and the protection of overflow and scour. Metals, much like nutrients, have various part itions and if a treatment process is able to remove smaller particles, it will have the added benefit of increasing its

PAGE 17

17 removal efficiency for the metals, especially if the metals species is largely in the particle bound phase. Therefore, it is critical to consider the various methods of pollutant removal in order to accurately achieve the desired results from a treatment unit.

PAGE 18

18 CHAPTER 2 PHOSPHOROUS KINETICS IN RAINFALL RUNOFF FROM A BIOGEN ICALLY LOADED WATERSHED While phosphorus plays an important role in the functional ability of a natural ecosystem, its excess loadings to receiving water bodies has increasingly become a concern due to eutrophication and the degradation of water quality (Lai and Lam, 2009; Vohla et al., 2011; Wang et al., 2003). Si nce the pre industrial times, the amount of phosphorus in terrestrial and freshwater systems has increased by at least 75% (Bennett et al., 2001). While its sources may be numerous and the concentrations of phosphorus in natural ecosystems can vary greatly its urban hydrologic transport has increasingly been studied and characterized. Many examinations correlate the transport of sediments and runoff from impervious watersheds to natural ecosystems to the overall buildup of phosphorus (Sharpley et al. 1994; Soranno et al., 1996). Additionally, as urban development and the amount of impervious surfaces continue to grow, this pollutant loading associated with the runoff leaving a particular watershed will persist (Tsihrintzis and Hamid, 1997). Adding to the co mplexity of phosphorous buildup is the fact that phosphorous exists in both the dissolved and particulate pollutant phases, making it difficult to pinpoint a solution for its harmful buildup (Sharpley et al. 1994). Additionally, phosphorus has a kinetic na ture in runoff which results in the affinity for dissolved phosphorus to attach itself to sediments (Haggard et al., 2004; Zhou et al., 2005). Overall, the partitioning of phosphorus between the dissolved and particulate bound phases is a complex phenomeno n that is a result of pH variances, flow velocity, contact time and other chemical, biological and physical processes (Agudelo et al., 2010; Bostrom et al., 1988; Zhou et al., 2005). All of these characteristics create variable and unique phosphorus partit ioning rates for all the individual flows that carry phosphorus to receiving water bodies.

PAGE 19

19 However, one major factor in the overall partitioning process for phosphorus is the retention capacity of dissolved phosphorus in sediments. Generally, dissolved ph osphorus adsorption capacity increases exponentially as the particle size distribution decreases, due in large part to the increased surface area that smaller particles carry (Wang and Li, 2010). Though, the capacity of a particular sample to adsorb phosph orous does not translate to the overall rate of adsorption. In a study by Kim and Sansalone (2008), the d 50 for event ranged from 20 to 300 m, which spans a variety of the PM size fractions, including the suspended, settleable and sm aller portion of the sediment size class (Sansalone and Kim, 2008). The d 50 is defined as the mid point of a particle size distribution where half of all the particles, by weight, are larger and smaller in size. Therefore, runoff covers a large range of si te characteristics for phosphorous adsorption. Additionally, in a study by Vaze and Chiew (2004), less than 15% of the total phosphorus content in stormwater is attached to particles over 300 m and nearly all of the particulate bound phosphorus in runoff was associated with particles ranging in the 11 to 150 m range. Similar results were found in other studies that illustrated suspended particles to have the highest concentration of mass based phosphorus, followed by settleable and finally sediment sized particles to have the smallest amount of phosphorus by mass (Ma et al., 2010). Building on this foundation of literature, this study first attempted to determine the direction of phosphorous kinetics in stormwater samples. Once the direction of partitioning was established, the intensity and duration of phosphorus partitioning in runoff samples was explored in an overall attempt to determine how long it took to reach equilibrium between the two phases. It was hypothesized that t he kinetics involved between the two phases would be rapid and that over time, the dissolved portion of phosphorus would decrease as phosphate ions adsorbed to PM

PAGE 20

20 in runoff, thus increasing the particulate bound portion of phosphorus. This concept would ca rry several implications. During runoff events, typically an automated water quality sampling unit is used to take samples (Harmel et al., 2010). As a result of this sampling procedure and coupled with the intensity of phosphorus partitioning, this study a lso hypothesizes that the analysis of dissolved and particulate bound forms of phosphorus may develop unforeseen amounts of error if the two phases are not quickly separated. Laboratory results describing the exact variance among the two partitions would n ot be precise when compared to what was established in the field. Objectives Based on the idea that phosphorous kinetics in runoff is a variable characteristic, it is hypothesized that physical indices, such as particle size, could influence the overall in tensity of this phenomena. In order to validate this hypothesis, a number of primary objectives were formulated. The first objective was to determine the direction of phosphorous kinetics in stormwater samples by physically sampling runoff from a biogenica lly loaded parking lot in Gainesville, FL and quickly fractionating a subsample from each primary sample. This will divide the dissolved and particulate bound variables and provide a benchmark of the two as close to the sampling time as possible. The secon d objective was to determine how long it took for each sample to reach equilibrium. This will eventually provide details on how the dissolved and particulate bound phosphorus concentrations interact over time. The third objective was to analyze the PSD of each sample identify whether or not the size of the particles in a stormwater sample has an effect on the rate of phosphorus kinetics.

PAGE 21

21 Methodology Source Area Catchment The watershed of interest was a public institutional parking lot located on the University of Florida in Gainesville, FL. This parking lot, known as the Reitz Union Parking Lot, is situated across from the campus student union and has a high volume of vehi cular, especially during weekday business hours. The influent runoff consisted of overland sheet flow from a catchment basin located in the southwest corner of the parking lot. In order to determine the total contribution area to this catchment basin, a su rvey was performed and data revealed that approximately 400 to 600 m2 of the parking lot contributed to this catchment, varying storm to storm due to differing wind patterns and intensity of rainfall. A majority of this contribution area consists of imperv ious asphalt surfaces while the other portion is composed of raised vegetative islands. These raised vegetative islands consist of grass and scattered trees, both of which contribute to nutrient and particulate matter loadings in stormwater runoff. During every studied rainfall event, total rainfall depth was measured using a tipping bucket rain gage manufactured by Texas Electronics Inc. with a rainfall depth capacity of 0.01 inch. Rainfall runoff collected by the stormwater catchment basin was re routed t hrough a butterfly value and into a 6 inch, PVC conduit system. Volume of runoff was then measured using a 1 inch parshall flume with depth measurements recorded every second by a 30 kHz ultrasonic sensor. The runoff was then transported to a sampling drop box, where samples were collected, before the stormwater once again entered another small 6 inch, PVC conduit system which took it straight to one of two stormwater treatment units. Sampling and Analytical Methods Each primary sample was collected using Nalgene polypropylene (PP) bottles that were sterilized in an acid bath for 24 hours and rinsed using deionized water. The samples were taken

PAGE 22

22 using the traditional grab method sampling technique following the parshall flume where the flow measurements were taken and before the stormwater made its way into the drop box. Once the primary samples were collected, they were immediately transported back to the lab and a homogenous, 60 mL sub sample was fractioned using a pressure pump and a 0.7 m glass fiber fil ter to separate the dissolved and particulate bound portions of the sub sample in accordance to Method D3977 97B (ASTM 2007). A filter size of 0.7 m was used in the fractionation process due to limiting choice of glass fiber filters. Glass fiber filters were chosen for this experimentation process to eliminate the possibility of leaching components that could alter the kinetic nature of the sample or add additional components. In addition, a number of other characteristics for each sample were determined. Separate sample bottles collected runoff and were used to solve for various other properties such as the PSD, concentration of sediment PM (> 75 m), settleable PM (25 75 m), suspended PM (< 25 m) and several other chemical properties. The PSD for each influent and effluent sample was measured using a laser diffraction particle analyzer (Malvern Instruments: Hydro 2000G) in a batch mode analysis. Sediment PM concentration was determined by passing the sample through a #200 sieve (> 75 m). Once the sampl e was passed through the sieve, it was then placed in an Imhoff cone for settleable particle separation. The supernatant leftover from the Imhoff cone was then fractionated to separate the dissolved constituents from the suspended PM. In order to test the kinetic changes in phosphorus, individual homogenous sub samples were taken from the primary 1.0 L sample at differing time intervals. Between each sub sample, the primary 1.0 L sample bottle rested on a countertop without agitation, allowing the contents inside to equilibrate naturally. Prior to taking the next sub sample, each primary sample bottle was lightly turned to ensure a homogenous mixture of particulate matter throughout the bottle so

PAGE 23

23 both partitions could be sub sampled. This process resulted i n an accumulation of sub samples, collected at differing time intervals, with their dissolved and particulate components divided. These sub samples were then preserved in a refrigeration unit until they were each tested for their individual TP concentratio ns. The analysis of total dissolved phosphorous (TDP) for each sub sample was performed using a HACH DR/5600 spectrophotometer in conjunction with the 4500 PE ascorbic acid method (APHA, 1995). In comparison, the PM based phosphorus sub samples followed t he 4500 PB persulfate digestion method. Following digestion, the ascorbic acid method was used to solve for TP in the sub sample. Phosphorous Partitioning As previously discussed, the influences effecting the partitioning of phosphorous are complex and nu merous. While the variables associated with phosphorous kinetics may be intricate, the two phases of phosphorous partitioning are quite simple. The total phosphorous concentration found in stormwater can be broken up into two categories; phosphorous dissol ved in the water or phosphorous attached to particles. This concept is described in the expression below. ( 2 1) In this expression, TP is total phosphorous, P P is the phosphorous concentration associated with particulate matter and D P is the dissolved phosphorous concentration. For any given sample, the TP value should not change throughout the partitioning process. Rather, the P P and D P values will change until equilibrium has been met. In order to provide quality assurance for the testing procedure throughout the partitioning experiment, the TP values for each time step were determined and compared to each of the other

PAGE 24

24 sample points. With a relatively stagnant TP value for all the time step samples, the fraction dissolved ( f d ) and the fraction particle bound were validated. T he fraction of phosphorous dissolved and the fraction of phosphorous particle bound were d etermined using the E quation 2 2 and Equation 2 3. ( 2 2) ( 2 3) The P P value includes each of the particle size gradations; sediment, settleable and suspended. Characterization of Rainfall Runoff Washoff and Material Transport An important characteristic dictating the type of particles entering a sample zone is the type and rate of pollutants washing off the impervious parking lot surface. Based on the hydrologic variations of a storm, two different categories can be discerned; a flow limited pollutant transport and a mass limited pollutant transport. During a flow limited transport event, the flow of water controls a majority of the pollutant transport and therefore, is mainly dictated by the dissolved pollutant concentrations. In contrast, during a mass limited transport event, the physical pollutant mass in the runoff carries a majority of the pollutants. This is highly characteristic of a first flush model. When the surface washoff from an impervious surface begins, the sedim ent pollution is generally at its highest and closely follows the hydrograph of the storm (Whipple, 1983). It is during this period when a majority of the particulate matter will be captured.

PAGE 25

25 Kinetic Equilibrium In order to quantify the time to equilibrium for each primary sample, a suitable kinetic model was needed. Though, it was necessary to first quantify the availability of the surface sites in relation to the amount of sorbates (Liu, 2005). This was determined by relating the total amount of phosphoro us adsorbed at equilibrium per unit mass of the available sediments in the sample bottle. The equation below outlines this process. ( 2 4) In the expression, C o and C e are the initial and equilibrium concentrations of phosphorous (mg/L) in the sample, V is the volume of the sample being tested, M is the total mass of sediment in the sample (g) and q e is the amount of phosphorous adsorbed onto the sediment per unit mass of sediment (mg/g) in the sample. For this experimentation process, all the samples were taken using 1 L sample bottles. It has previously been shown that the kinetic modeling of phosphorous adsorption onto particles closely relates to a second order kin etic model (Ozacar, 2003). This was also the same case for metals adsorption in urban runoff (Liu, 2005) and the adsorption of reactive dye 189 on cro ss linked chitosan beads (Chiou and Li, 2002). The differential equation used in this second o rder kinetic model is shown in E quation 2 5. ( 2 5) Integrating Equation 2 5 above with the boundary conditions, q t = 0 at t = 0 and q t = q t at t= t and rearranging the variables produces the simplified second order kinetic model show below.

PAGE 26

26 ( 2 6) In the expression, t is the time (minutes) at any interval after the initial sample t 0 q t is the amount of phosphorous adsorbed per unit mass of sediment at t ime = t, q e is the amount of phosphorous adsorbed per unit mass of sediment at equilibrium and k is the equilibrium rate constant of the second order kinetic model (g/mg min). The sample points for each sub sample in a given primary sample were plotted and Equation 2 6 was used to model the kinetic nature of phosphorous. By changing the values of the equilibrium dissolved phosphorous concentration and the equilibrium rate constant, k, the model was fit to each individual sample. As a quality check, both partitions of phosphorous were measured and the TP value was determined. With a constant value of TP at each sub sample point, the model was verified. Statistical Analysis To compute the influence of particle size on the time to equilibrium, the samples were split int o two separate populations: samples that contained PM with a d 50 greater than 75 m and samples that contained PM with a d 50 less than 75 m. The 75 m dividing line was chosen due to it being the dividing gradation line that separates settleable and sedim ent particles. With these two populations, an overall assessment on whether or not particle size is a variable that alters phosphorous kinetic equilibrium can be determined. A significance test between the two populations was performed, comparing the mean t ime to equilibrium values (McClave and Sincich 2005). Through this process, the significance of particle size on kinetics can be concluded. First, it was necessary to first construct a pooled sample estimator between the groups. This process is shown be low.

PAGE 27

27 ( 2 7) In the expression, S p 2 is a statistical variable, S 1 is the standard deviation of the time to equilibrium for the samples with a d 50 less than 75 m, S 2 is the standard deviation of the time to equilibrium for the samples with a d 50 greater than 75 m, n 1 is the number of samples with a d 50 less than 75 m and n 2 is the number of samples with a d 50 greater than 75 m. Using the S p 2 value found in Equati on 2 7, a test statistic is determined. This is shown in Equation 2 8 below. ( 2 8) For E quation 2 8, t o is the test statistic, x 1 is the mean time to equilibrium for the samples with a d 50 less than 75 m and x 2 is the mean time to equilibrium for the samples with a d 50 greater than 75 m. If the test statistic from this statistical analysis exceeded the t value obtained from the t table, then the two groups had a significant difference in the time to equilibrium Additionally, a Results Chemical and Physical Characterization of Rainfall Runoff Samples were taken from runoff events between August 2010 and November 2010 and varied between raw influ ent runoff samples from the impervious parking lot and effluent samples from two separate stormwater treatment units. The hydrologic characteristics for each sampled storm are summarized in Table 2 1. One unit was a hydrodynamic separator while the other w as a hydrodynamic separator with an added filtration mechanism. The two units added a dynamic to

PAGE 28

28 the amount of particulate matter in the effluent samples, which, in turn gave variance to the characteristics throughout all the primary samples. The chemical properties of the primary samples are summarized in Table 2 2. In contrast to the chemical properties which generally carried similar values, most of the physical characteristics of the samples had a large variability; shown in Table 2 3. The d 50 of the samples ranged from 2.0 m to 160.3 m with six of the samples having a d 50 below 75 m and four of the samples having a d 50 above 75 m. Additionally, the measured particle size distribution values were fit with a gamma distribution model and t he scale and shape factors, as well as the normalized root mean squared error (NRMSE) can be seen in Table 2 3. The PSD curves for each of the samples are divided between the two populations and shown in Figure 2 1. Additionally, a visual summary of the ga mma distribution for the two populations is shown in Figure 2 2. The runoff intensity and flow rate also played a large role in the amount of particulate matter that was in the samples. Varying flows throughout a given storm can change a portion of an eve nt from flow limited to mass limited, or vice versa. This created suspended sediment concentrations (SSC) ranging from 4.5 mg/L to 2301.7 mg/L, most of which was comprised of biogenic debris washoff from the raised vegetative islands. As expected, the two primary samples with the highest SSC concentrations were the first two influent samples taken from their respective storm. Correspondingly, the three primary samples with the smallest SSC concentrations were effluent samples that had already been treated b y one of the two stormwater treatment units.

PAGE 29

29 Phosphorous Loadings and Partitioning The TP concentrations measured from the ten primary samples had a range of 1.7 mg/L to 8.4 mg/L, which was strongly dictated by the amount of particulate matter captured. Th e highest TP measurements were found in the samples with the highest concentration of SSC while the lowest generally had the smallest mass of SSC in the sample bottle. The concentrations of TP for each of the samples, as well as the initial f d value and th e equilibrium f d value, are shown in Table 2 4. Out of the ten primary samples, eight had an initial f d value below 0.5, signifying a majority of the TP was associated with particles. Only two samples had an initial f d value above 0.5. These tendencies s how how much the SSC loadings in rainfall runoff dictate the amount of phosphorous leaving a watershed. This trend was also observed by Vaze and Chiew (2004). In an investigation on phosphorous in urban stormwater, it was found that most of the P was assoc iated with particles while only about 25% of the TP is dissolved. In this study, with the exception of two primary samples, the range of dissolved TP values was 18% 35%. The partitioning of phosphorous between the dissolved and particulate bound phases he ld the same trend throughout all the samples. Over time, the exchange of phosphorous within a primary sample always moved from the dissolved to the particulate bound phase. This trend held true in other experiments as well (Ozacar, 2003). Therefore, the in itial f d value for all samples was always higher than the equilibrium f d value. However, the extent, intensity and time of this exchange process varied between all primary samples. A majority of the primary samples had rather small kinetic changes with les s than a 10% change between partitions, while one sample had a change in f d values of 0.49.

PAGE 30

30 Time to Equilibrium The time to equilibrium for each sample was defined as the amount of time it took for the exchange of phosphorous between partitions within a p rimary sample to remain stable and all kinetic activity to stop. Using a pseudo second order kinetic model, the equilibrium dissolved phosphorous concentration was determined by fitting the model to the measured data and then finding the exact f d equilibri um value. With the dissolved phosphorous equilibrium value known, the time to equilibrium was in turn determined. These values, as well as the kinetic k value, the percentage of a samples kinetic exchange within the first 60 minutes and the goodness of fit between the experimental values and the model are presented in Table 2 5. Additionally, the time to equilibrium, change of dissolved and PM based phosphorous concentrations and the standard deviations for the two primary sample populations are illustrated in Figure 2 3 and Figure 2 4. Overall, the time to equilibrium values ranged from 146.1 minutes to 245.3 minutes and, as seen in Table 2 5, most of the kinetic exchange occurred within the first 60 minutes. In fact, two of the samples achieved over 95% eq uilibrium and nine of the ten samples achieved at least 70% equilibrium within the first hour. Overall, the time to equilibrium values are on the higher end of published results. In another experiment, it was found that the time it took for equilibrium to be reached between the two phosphorous phases was about 120 minutes (Ozacar, 2003). Though, this experiment was run with a controlled amount of sediment, type of sediment, pH and temperature. In this experiment, all of those values were different between each of the samples, as seen in Table 2 2 and Table 2 3. Particle Size Distribution Influence on Time to Equilibrium A comparison between the results from the pseudo second order kinetic model indicated that the size of the particles within a sample does have a significant effect on the time to

PAGE 31

31 equilibrium. As seen in Figure 2 2, the time to equilibrium can be related to the gamma distribution shape factor, which represents the size gradation parameter for that given distribution. For that reason, two statistical populations were formulated by separating all the samples along the 75 m line. Thus, one population encompasse d all the sample s that had a d 50 larger than 75 m and the other population included all the sample that had a d 50 smaller than 75 m. As a whole, results indicated that the population with the larger scale factor, and larger d 50 reached equilibrium in less time than the group of samples with a smaller d 50 value. In addition to the rate of phosphorous partitioning, the intensity of partitioning was also altered between the two groups. While there was an evident exchange of phosphorus between all the samples, the group with the smaller d 50 had a decrease in dissolved phosphorous concentration ranging between 1 6%. For the samples with the higher d 50 the decrease in dissolved phosphorous ranged from 5% to 10%. Though, there was one outlier sample with a 49% ch ange in dissolved phosphorous. Discussion The conclusions from this experiment explores the direction of phosphorous kinetics in stormwater samples and how the size of the particles within a given sample effects the amount of time a sample is needed before it reaches equilibrium. Samples were taken from an impervious parking lot in Gainesville, FL where most of the pollutants came from biogenically loaded surface washoff. Ten samples, spanning a three month time period, were analyzed to study the partitioni ng phosphorous and the intensity of its kinetic nature. The results generated changing dissolved and PM bound phosphorous concentrations over time and associated each primary sample with a certain particle size distribution.

PAGE 32

32 The amount of phosphorous and the fraction of f d and f p in each sample was relatively stable throughout the experiment. Most of the samples had TP values in the range of 1.7 5.2 mg P/L and these values were strongly dictated by the amount of particulate matter in each individual primar y sample. Additionally, the fraction of TP associated with the dissolved phase was generally between 0.18 and 0.35, which implies that phosphorous is largely associated with PM in runoff. All of the samples in this study showed an increasing concentration of PM bound phosphorous over time. While the direction of partitioning remained the same for all samples, there was a distinct difference in the time to equilibrium between the primary samples. To further analyze this trend, the total population of samples was broken up into two separate categories, samples with a d 50 greater than 75 m and samples with a d 50 less than 75 m. Samples with a d 50 greater than 75 m represented a group that had a majority of its particles in the sediment PM range and experienc ed a time to equilibrium generally below 3 hours. In contrast, samples with a d 50 less than 75 m represented a group that had a majority of its particles smaller than the sediment PM range and experienced a time to equilibrium in the 3.5 4 hour range. Tho ugh, a majority of the phosphorous kinetic partitioning occurred rapidly, as indicated by nearly all of the samples experiencing at least 70% equilibrium within the first hour. Many factors can play a part in determining how long it takes for a runoff sam ple to reach equilibrium with its phosphorous partitioning and it is well illustrated that the size of the particles within a given sample is one such element. While it would initially seem logical to assume that samples with a smaller scale factor would p roduce more rapid kinetics due to their increased surface area, it is proposed that the phosphorous associated with those particles are less labile and would not disassociate as much in the initial phase of runoff. Rather, the larger and

PAGE 33

33 more coarse partic les would be more susceptible to change and the phosphorous associated with those particles enter into the dissolved phase more easily. Then, once a sample is taken or the agitation associated with transport subsides, the dissolved phosphorous will then re attach itself to available surface sites on surrounding PM.

PAGE 34

34 Table 2 1 Event based hydrologic parameters for each storm event. Characteristics included are previous dry hours, event duration, rainfall depth, maximum rainfall intensity, initial pavement residence time, runoff volume and maximum flow rate. Event Date (2010) PDH d rain h rain i max IPRT V runoff Q p [hrs] [min] [mm] [mm/hr] [min] [L] [L/min] Aug 7 24 48 8.64 61 7 2623 795 Aug 23 48 42 2.79 5.1 20 310 76 Sept 26 40 78 3.56 5.1 11 1128 26 Nov 4 36 43 4.83 25.4 5 996 212 PDH Previous Dry Hours d rain Event Duration h rain Rainfall Depth i max Maximum Rainfall Intensity IPRT Initial Pavement Residence Time V runoff Runoff Volume Q p Maximum Flow Rate

PAGE 35

35 Table 2 2 Chemical properties for each of the samples from the experimental procedure. Characteristics included are pH, total dissolved solids concentration (TDS), redox, alkalinity, temperature and conductivity. PM Size Fraction Event Date (2010) Flow Type pH TDS Redox Alkalinity Temperature Conductivity [s.u.] [mg/L] [mV] [mg/L as CaCO 3 ] [ C] [ s/cm] d 50 > 75 m Aug 23 Influent 7.0 38.5 335.1 29.5 26.5 78.8 Aug 23 Effluent 6.7 90.0 321.7 88.4 25.9 161.5 Sept 26 Influent 6.6 77.0 204.3 75.8 23.5 157.0 Nov 4 Influent 7.1 63.0 374.3 34.7 22.1 128.5 d 50 < 75 m Aug 7 Effluent 6.6 57.5 360.6 65.5 25.5 117.9 Sept 26 Influent 7.2 42.5 493.7 30.9 23.8 86.9 Sept 26a Effluent 7.1 81.0 487.3 88.5 23.7 165.6 Sept 26b Effluent 6.9 31.0 584.3 79.0 24.0 63.2 Nov 4 Influent 6.8 105.0 351.1 169.7 21.5 214.4 Nov 4 Effluent 7.1 138.0 401.9 139.4 21.9 280.9 TDS Total Dissolved Solids

PAGE 36

36 Table 2 3 Physical properties for each of the samples from the experimental procedure. Characteristics included are sediment particle concentration, settleable particle concentration, suspended particle concentration (TSS), suspended sediment concentration (SSC) and d 50 value, value and NRMSE. PM Size Fraction Event Date (2010) Flow Type Sediment [mg/L] Settleable [mg/L] Suspended [mg/L] SSC [mg/L] d 50 [ m] NRMSE d 50 > 75 m Aug 23 Influent 39.7 2.6 1.0 45.8 147.2 371.6 0.6 .015 Aug 23 Effluent 3.3 2.5 2.0 5.5 106.7 147.9 1.0 .008 Sept 26 Influent 524.2 13.4 48.0 585.6 160.3 110.7 1.7 .021 Nov 4 Influent 50.0 4.4 41.0 246.9 98.3 333 0 0.5 .035 d 50 < 75 m Aug 7 Effluent 3.0 5.0 8.5 13.2 3.5 10.4 0.7 .041 Sept 26 Influent 7.4 1.6 4.0 38.3 20. 6 56.6 0.7 .021 Sept 26a Effluent 2.5 1.6 1.5 5.3 4.0 19.8 0.6 .062 Sept 26b Effluent 7.7 2.3 2.0 4.5 3.1 14.7 0.7 .069 Nov 4 Influent 1810.9 171.4 319.4 2301.7 73.4 288.5 0.6 .036 Nov 4 Effluent 10.3 3.2 6.0 104.0 2.0 0.7 3.3 .015 SSC Suspended Sediment Concentration scale factor of gamma distribution shape factor of gamma distribution NRMSE Normalized Root Mean Square Error of Modeled PSD to Experimental PSD

PAGE 37

37 Table 2 4 Phosphorous properties for each of the samples from the experimental procedure. Characteristics included are total phsphorous initial total dissolved phosphorous concentration, initial f d value, initial k d value, equilibrium total dissolved phosphorous concentration, equilibrium f d value and equilibrium k d value. PM Size Fraction Event Date (2010) Flow Type TP avg [mg/L] TP stdev [mg/L] Initial (t = 0) Equilibrium (t = t* ) TDP f d k d 10 4 TDP f d k d 10 4 [mg/L] [L/kg] [mg/L] [L/kg] d 50 > 75 m Aug 23 Influent 2.1 0.05 1.76 0.84 0.76 0.73 0.35 3.2 Aug 23 Effluent 1.7 0.02 1.31 0.75 5.57 1.14 0.65 7.8 Sept 26 Influent 3.2 0.02 0.82 0.26 0.41 0.66 0.21 0.4 Nov 4 Influent 5.2 0.01 1.80 0.34 3.59 1.66 0.32 3.7 d 50 < 75 m Aug 7 Effluent 2.5 0.04 0.78 0.31 10.68 0.67 0.26 11.4 Sept 26 Influent 2.6 0.02 0.67 0.26 14.58 0.47 0.18 16.1 Sept 26a Effluent 4.5 0.02 1.09 0.24 61.74 0.98 0.22 63.6 Sept 26b Effluent 3.1 0.02 1.07 0.35 16.75 0.93 0.30 17.9 Nov 4 Influent 8.4 0.01 1.59 0.19 0.30 1.38 0.16 0.3 Nov 4 Effluent 6.2 0.01 1.11 0.18 25.88 1.03 0.17 26.3 TP avg Average Total Phosphorous Concentration from each of the Sub Samples TP stdev Standard Deviation Between the Total Phosphorous Concentrations from each of the Sub Samples TDP Total Dissolved Phosphorous f d Fraction Dissolved k d Equilibrium Partitioning Coefficient

PAGE 38

38 Table 2 5 Kinetic results for each of the experimental samples. Characteristics included are time to equilibrium, the equilibrium rate constant of the second order kinetic model and the r 2 value comparing the experimental data to the model. PM Size Fraction Event Date (2010) Flow Type Equil 60 [%] t* [min] k [mg/mg*min] r 2 d 50 > 75 m Aug 23 Influent 96.5 148.9 20.2 .973 Aug 23 Effluent 95.3 146.1 15.0 .971 Sept 26 Influent 77.9 156.5 212.2 .945 Nov 4 Influent 77.6 153.4 38.5 .975 d 50 < 75 m Aug 7 Effluent 53.6 212.4 2.2 .871 Sept 26 Influent 74.9 225.0 3.2 .896 Sept 26a Effluent 76.2 245.3 2.7 .861 Sept 26b Effluent 78.7 227.2 2.0 .908 Nov 4 Influent 77.0 214.0 621.9 .900 Nov 4 Effluent 71.2 210.7 51.8 .899 Equil 60 % Equilibrium in 60 minutes t* Time to Equilibrium k Equilibrium Rate Constant of the Second Order Kinetic Model r 2 Goodness of fit between the experimental data and the second order kinetic model

PAGE 39

39 Table 2 6 Statistical Variables and Analysis of samples with a d 50 above 75 m and a d 50 below 75 m. Variables included are time to equilibrium mean, standard deviation and number of samples from each population. Additionally, statistical analysis variables includ e the test statistic (t o ) from the statistical analysis between the two means and the comparison of the t value obtained from the t table (t and the S p 2 statistical variable. Statistical Variable d 50 > 75 m d 50 < 75 m Mean t* 151.22 222.43 (minutes) Standard Deviation t* 4.63 13.13 (minutes) n 4 6 t o > t n 1 + n 2 2 10.47 > 2.31

PAGE 40

40 Figure 2 1 populations

PAGE 41

41 Figure 2 2 Comparison of Time to Equilibrium and the Scale/Shape Factors of the Gamma Distribution between the two experimental populations (greater than 75 m and less than 75 m).

PAGE 42

42 Figure 2 3 Dissolved and particulate bound phosphorous concentrations over time for samples with a d 50 greater than 75 m. The time to equilibrium is marked by the dashed vertical line. (a) Au gust 23 rd Influent (b) August 23 rd Effluent (c) September 26 th Influent (d) November 4 th Influent

PAGE 43

43 Figure 2 4 Dissolved and particulate bound phosphorous concentrations over time for samples with a d 50 less than 75 m. The time to equilibrium is marked by the dashed vertical line. (a) August 7 th Effluent (b) September 26 th Influent (c) September 26 th Effluent a (d ) September 26 th Effluent b (e) November 4 th Inf luent (f) November 4 th Effluent

PAGE 44

44 CHAPTER 3 COMPARISON OF METALS REMOVAL IN RAINFALL RUNOFF BETWEEN VARYI NG UNIT OPERATIONS As the expansion of urban impervious surfaces continues, the amount of pollutio n that collects on these areas will continue to grow. Urban traffic constantly adds various pollutants to roads and parking lots which eventually get washed off through rainfall runoff (Mahbub et al., 2011). This runoff then makes its way into the natural receiving water bodies where it can severely degrade an ecosystem or even pollute a potable water supply source (Hatt et al., 2008, Ellis et al., 1987). By some estimates, the metal pollution fraction in natural streams derived from highway runoff can be b etween 35 75% (Ellis et al., 1987). Even in low density residential areas, metals loadings from street surfaces can be problematic (Ellis et al., 1985). However, while the concentrations of metals in rainfall runoff may be quite high, the ability to pinpoi nt the exact culprits behind these nonpoint pollutants is challenging The rainfall that falls on these impervious surfaces can oftentimes be acidic and contain quantities of mercury and cadmium while automobile wear and tear can add lead, copper, nickel, cobalt, zinc, manganese, cadmium and other various metals (Whipple and Hunter, 1977, Pitcher et al., 2004, Granier et al., 1990). Additionally, wastewater and industrial leaks, septic tanks and solid waste disposal can add to the pollution stream (Whipple and Hunter, 1977). When a metals species makes its way into a runoff stream, its transport can occur in two separate manners. Similar to nutrients, metals can be in both the dissolved and the particulate bound forms (Yousef et al., 1984). Though, complicating the issue of metals treatment further is the notion that some metals tend to be mainly associated with the dissolved phase while others tend to favor being bound to particles. For example, when zinc is present in urban runoff, it tends to be more soluble while lead tends to adsorb to sediments and other particles while being transported (Reinelt and Horner, 1995, Mertz et al., 1999). Dissolved metals are of particular

PAGE 45

45 concern due to their ease of transport in a stream an d how they can be assimilated by aquatic plants and animals (Dean et al., 2005). Nonetheless, while dissolved pollutants add a more immediate threat due to their ease of transport, both partitions add a level of hazard that must be considered. Metals in s tormwater that are associated with particulate matter tend to be more heavily concentrated in the finer sized particles (Walker and Hurl, 2002, Rentz et al., 2011). These finer sized particles are more more easily transported and have a higher probability for re suspension. Additionally, if the rain in an area tends to be acidic, such as in cities with high vehicular traffic or industrial settings, the metals on particulate matter can leach into the dissolved phase (Dean et al., 2005). There have been seve ral treatment options tested to enhance the removal of metals in runoff. One of the most basic treatment tools used in stormwater management is the dry detention pond. Dry detention ponds primarily rely on the process of sedimentation to remove pollutants (Stanley, 1996) but, a lack of maintenance leads to large build ups of contaminated sediment that can eventually be re suspended or the metals associated with the sediment can leach into the water (Pitcher et al., 2004). Infiltration ponds are another opti on and achieve better removal of particle matter (Dechesne et al., 2004) but, also have trouble removing the soluble material unless it is aided by an engineered media or plant matter (Pitcher et al., 2004). To create a more natural looking treatment setti ng, wetlands are often used (Calijuri et al., 2011). In addition to wetlands being more attractive to the community, they can also be abundant in reeds and other plants that are highly efficient in removing soluble pollutants from water (Read et al., 2008) Though, once the soluble metals are removed by the aquatic biology, there is potential for it to make its way into the food web (Reinelt and Horner, 1995). Porous pavements provide onsite treatment for rainfall runoff and can additionally provide soluble metal removal if an

PAGE 46

46 admixture is added but generally only provide assistance at reducing the peak flow and strain on the downstream treatment options (Legret et al., 1999). Building on this foundation, this study attempted to evaluate the metals treatmen t efficiency of two separate stormwater unit operations and processes (UOPs) that received runoff from the same watershed. One unit was a baffled hydrodynamic separator (BHS) that specialized in the removal of particulate matter from an inflow stream. The other unit was also a baffled hydrodynamic separator but had an additional filtration mechanism installed into its treatment train. This filtration mechanism had a nominal pore size of 20 m and was the last step in the ll further be referred to as the volumetric filtration unit (VF) The benefit of the added membrane filter is the additional removal of the finer particulate matter. While neither treatment unit specifically targeted the dissolved portion of metals in runo ff, it was hypothesized that the added filtration mechanism in the volumetric filtration unit would provide significantly better results in removing the particulate bound metals species from runoff. Additionally, this mechanism would stabilize the treatmen t process and lessens the potential of negative removal efficiencies ( RE ) due to scour and re suspension. Objectives Building on the idea that the removal of metals in runoff could be more effective and consistent with a unit that has added components in its treatment train, a number of objectives were established. The first objective was to manually sample influent and effluent runoff from two varying types of stormwater treatment units that received flow from the same biogenically loaded, impervious parking lot located in Gainesville, FL. While the concentrations of a metals species in a particular gradation can varying drastically, by using the same watershed for both units, similar loadings should be somewhat expected. The second objective was to take those

PAGE 47

47 samples and separate all of the particulate matter from each of them by the varying types of size partitions (sed iment, settleable and suspended PM). In total, when combined with the aqueous portion of the sample, there were a total of 4 sub samples. The third and final objective was to analyze each of the 4 sub samples and measure the metals concentration in each of them. Using those results, the removal efficiencies between each of the two separate stormwater treatment units were compared, with a majority of the emphasis focusing on the removal of metals by PM size gradation. Additionally, the effect that the maximu m flow rate for each storm had on the total removal was observed. Methodology Source Area Watershed and Hydrologic Data Collection The watershed in this study was a public institutional parking lot, a little over 3 acres in size and located on the Universi ty of Florida campus in Gainesville, FL. Resting at the corner of Museum Road and Center Drive, this parking lot experiences a high density of car traffic due in large part to the Reitz Union, which is the student union, located across the street. From 7:3 0 AM until 4:30 PM, the Reitz Union parking lot mainly provides parking for faculty and staff while, outside those hours, it is open to the public. The runoff that made its way into the stormwater treatment units for this study came in large part from the southwest corner of the parking lot. Results from an onsite survey showed that approximately 400 to 600 m 2 of the parking lot actually contributed to the catchment basin that transported the water to the units. This area has a large variability due to the wind direction during a storm and the rainfall intensity. Additionally, there are two varying types of surfaces on the watershed site. The first is an impervious asphalt surface that comprises approximately 75% of the total area while the other portion is raised vegetative islands. These raised vegetative islands, which consist of grass and trees, are the main

PAGE 48

48 During rainfall runoff events, the rainfall intensity and depth was measured using a tipping bucket rain gage manufactured by Texas Electronics Inc. with a rainfall depth capacity of 0.01 inches while the duration of the storm was measured manually by onsite stopwatches. Runoff from the parking lot site entered into a stormwater catchmen t basin and was re routed through a 6 inch PVC conduit system by use of a butterfly valve. The PVC conduit system transported the runoff to a 1 inch parshall flume where the influent flow was recorded using a 30 kHz ultrasonic sensor. The influent was then sampled as it fell into a sampling dropbox before it made its way through additional PVC piping and into one of the two treatment units. Sampling Methods and Analytical Procedures Onsite sampling was carried out using the traditional grab method technique following the parshall flume and before the influent water reached the dropbox. The samples were collected in Nalgene polypropylene (PP) bottles that were cleaned and sterilized in an acid bath for 24 hours before a 24 hour immersion in a DI bath. Corresp onding influent and effluent samples were collected at approximate 2 10 minute intervals, depending on the rainfall intensity, throughout the storm to analyze the nature of the pollutant loading in the total stream over the length of the runoff period. Onc e the storm was completed, the samples were immediately taken to the laboratory for analysis. The analysis of the particulate matter loading began immediately following the storm to minimize the influence of coagulation and partitioning. The entire conten ts of a 1 L sample bottle, was passed through a 75 m sieve to remove any material falling into the sediment PM category. Following the sieve, the sample entered an imhoff cone for a 1 hour settling period to remove all the settleable sized PM (particulate matter larger than 25 m and smaller than 75 m), in accordance with S.M.2540 D. Once the hour passed, the supernatant was then

PAGE 49

49 fractionated through a 0.45 m filter, removing all the suspended PM left in the sample. Each of these partitions were then sto red in a sealed refrigerator until the metals analysis was performed. The remaining aqueous portion of the sample, which contained the dissolved partition of the metals, was acidified with 5% nitric acid and also stored in a sealed refrigerator in accordan ce to S.M.3030 D. Additionally, the particle size distribution (PSD) was analyzed for each of the samples using a laser diffraction particle analyzer (Malvern Instruments: Hydro 2000G) in a batch mode analysis. During sample analysis the particulate bound metals portion of the sample was digested using the nitric hydrochloric acid method, SW 846 Method 3015 (USEPA, 1990). Then, all of the metals partitions, including the dissolved portion, were analyzed in house with an Inductively Coupled Plasma Atomic Emission Spectroscopy (ICP AES). For this study, the ICP AES machine used was a Perkin Elmer Optima 3200 RL. Additionally, the analytical techniques used followed SW 846 Method 6020, as outlined by the USEPA (1996). Metals Partitioning As previously stated the existence of metals in stormwater can occur in two phases, metals that are associated with particulate matter and metals that are in the aqueous phase. While the cause and effect of metals partitioning may be difficult, the concept behind the two pha ses is rather simple. The equation below sums up the basic theory of partitioning. ( 3 1) In the expression above, TM is the total percentage or concentration of any given metals species in a sample, P m is the metals percentage or concentration of the given metals species associated with particulate matter and D m is the metals percentage or concentration of the given

PAGE 50

50 metals species that is dissolved in the sample. If a percentage value is used, TM will always be equal to 1.0 (10 0%). Building on this basic partitioning function, the fraction of particulate bound metals ( p ) and the fraction of dissolved metals ( d ) were determined. Between the two units in question, the main difference in the treatment trains was an added filtration mechanism, which targets particle based pollutants. Thus, varying values in p and d can significantly change the overall removal eff iciency. Equations 3 2 and 3 3 outline the two partitioning fractions. ( 3 2) ( 3 3) In the two equations above, D m is dissolved metals concentration and P m is the particulate bound metals concentration for a given sample or event mean concentration (EMC). It should be noted that the P m value in the equations above includes each of the particle size gradations; sediment, settleable and suspended. Event Mean Concentration and Removal Efficiencies To character ize the removal efficiency of a pollutant in a stormwater unit, the event mean concentration (EMC) was determined. This number is the average concentration of a pollutant throughout the course of the entire storm; whether it is for the influent or the effl uent. The EMC was the chosen variable of interest due to its stability and its ability to accurately quantify what is in the total flow stream in question. On the other hand, if the range of concentration values were used, phenomena such as the first flush can significantly skew the results. In order to find the EMC, two separate variables were needed; the total mass of the pollutant in the influent or

PAGE 51

51 effluent and the total volume of runoff that entered or left the unit. These variables were then entered i nto the Equation 3 4, as described by Sansalone and Buchberger (1997). ( 3 4) In Equation 3 4, M is the total mass of the metal species throughout the entire storm, V is the total volume of flow through the conduit system throughout the storm, c(t) is the time variable of the pollutant concentration, q(t) is the time variable of flow and t is the time. To quantify the difference in metals removal between the two units, the removal efficiency was determined. The removal efficiency computes the treatment ability of either unit for any given pollutant species at any size gradation. Equation 3 5 is a summary on finding the RE. ( 3 5) duration of a given storm, EMC IN and EMC OUT i s the EMC concentration of a metals species in the effluent. Statistical Analysis Each of the two treatment units have a maximum influent flow rate designated by their manufacturer. For the baffled hydrodynamic separator, the 100% flow rate capacity of the unit is 541 gpm. At this flow rate, bypass occurs and a portion of the raw runoff flows over an internal baffle and leaves the unit untreated. When this occurs, the RE of that discrete time period significantly decreases and the overall effluent EMC for t he event can be drastically increased. For the volumetric filtration unit, the 100% flow rate capacity is slightly higher and graded at a flow of 757 gpm. This flow capacity is greater due to it having a slightly larger body for additional storage.

PAGE 52

52 A statistical analysis was carried out, comparing the change in removal efficiency for individual treatment units maximum designated flow rate. By comparing the t wo populations, the the two units can be determined. In order to carry out this process, the construction of a pooled sample estimator between the groups must b e determined first. This process is shown in Equation 3 6. ( 3 6) In the above equation, S p 2 is a statistical variable, S 1 is the standard deviation of the removal efficiency for the baffled hydrodynamic separator, S 2 is the standar d deviation of the removal efficiency for the volumetric filtration unit, n 1 is the number of samples taken from the hydrodynamic population and n 2 is the number of samples taken from the volumetric filtration unit population. Using the statistical varia ble found in Equation 3 6, a test statistic is determined. This is shown in Equation 3 7 below. ( 3 7) In Equation 3 7, t o is the test statistic, x 1 is the mean removal efficiency for the baffled hydrodynamic separator and x 2 is the mean removal efficiency for the volumetric filtration unit. If the test statistic from this statistical analysis exceeded the t value obtained from the statistical t table, then the two groups had a significant difference in their removal efficienc ies. Additionally, for this statistical analysis, a value of 0.95 was always used.

PAGE 53

53 Results Rainfall Runoff and Hydrologic Results Data was collected from 10 different storm events between July 2010 and September 2010. In total, samples were taken from b oth the raw influent from the impervious parking lot surface and the effluent from each of the two units at their outlet point. Out of the ten observed storms, five of them were transported through the baffled hydrodynamic separator and the other five stor ms were treated by the volumetric filtration unit. The hydrology characteristics of all the sampled storms are described in Table 3 1. One of the main driving forces of influent pollution intensity into a stormwater unit is the number of previous dry hours since the last washoff. If many days pass without a runoff event, the pollution on the impervious parking lot surface accumulates and maintains its place until a driving force, such as stormwater, carries it away. The previous dry hours (PDH) observed thr oughout the ten storms ranged from 15 172 hours. Additionally, pollutant washoff can strongly be influenced by the intensity and duration of a storm event. The range of these results throughout the ten sampled events was 5.1 137.2 mm /hour for rainfall inte nsity and 25 104 minutes for rainfall duration. During the high intensity, long duration events, the potential for the biogenic material from the raised vegetative islands being washed off into the influent rises. The dissolved metals species oftentimes ad sorb to the biogenic material and are transported into one of the two units. For the baffled hydrodynamic separator, the main mechanism of pollution removal is through the settling of particulate matter. Though, when the flow through the unit is high, the re is potential for scour. Scour is the re suspension of the particulate matter and pollutants that have already been removed by the stormwater unit. When scour occurs, effluent pollutant concentrations oftentimes rise and, as a result, the removal efficie ncy lowers. This trend is far

PAGE 54

54 more prevalent in the baffled hydrodynamic separator than it is in the volumetric filtration unit. While re suspension may occur in the volumetric filtration unit, the added filtration mechanism helps prevent some of the small er sized particles from leaving the unit. When examining the potential for scour, one of the main factors to consider is the peak flow (Q p in Table 3 1). For the baffled hydrodynamic separator, the maximum influent flow rate ranged from 87.1 962 L/min and the Q p ranged from 75.7 802.5 L/min for the volumetric filtration. Overall, each of the units had two storms that had peak flows over their maximum graded capacity as defined by their manufacturer throughout this study. Metals Loadings and Partitionin g In total, four different metals species were analyzed throughout the ten storms: copper, nickel, lead and zinc. For each of the pollutants, laboratory analyses were carried out to determine their concentrations in the sediment, settleable and suspended P M phase; as well as their dissolved partition. The strongest contributing metals of the four analyzed to the pollution stream were nickel and zinc, which is generally a result of the wear of brakes, tires and the frame of vehicles (Sansalone and Buchberger 1997). The total EMC concentration for each metal species in the influent and the effluent for all of the analyzed storms are summarized in Table 3 2. The highest EMC influent total concentrations for the baffled hydrodynamic separator for each of the me tals pollutants were 165 g Cu/L, 4215 g Ni/L, 118 g Pb/L and 1562 g Zn/L. For the volumetric filtration unit, the highest EMC influent total concentrations for each of the metals pollutants were 412 g Cu/L, 894 g Ni/L, 107 g Pb/L and 1967 g Zn/L. F requently, the first flush had a strong influence on the EMC of a metals species and its partitioning for an entire storm. It was often observed that the total mass contributing to the metals EMC had little contributing mass from the samples taken later in a storm. This was

PAGE 55

55 strongly shaped by the first flush of particulate matter and pollutants off of the impervious surface, which carried a significant portion of the metals in the runoff. The results of this are displayed in Figure 3 1, which shows the chan ges of f d throughout the course of a storm. As shown, towards the beginning of the storm when the first flush occurs, a majority of the metals pollutants are associated with the particulate bound phase. In the case of Figure 3 1 the partitions of zinc are shown in two representative storms. The fraction of zinc in the effluent associated with the dissolved phase throughout the course of those two storms represents a majority of the pollutant and this is due to the removal of the particulate bound phase thro ugh treatment. The zinc equilibrium coefficient values also displayed in Figure 3 1 correspond to a consistently high f d value for effluent samples. As such, the fractional percentage of the EMC associated with the dissolved or particulate bound partition of a metals species and the particle size distribution of the influent PM largely 3 3 summarizes the PM concentrations for each gradation as well as the PM d 10 d 50 and d 90 for both the influent and effluent for each of the ten analyzed storms. As seen, the volumetric filtration unit was able to treat PM more effectively. The largest d 50 released by the volumetric filtration unit over the course of the study was 6 microns while the smallest d 50 r eleased by the baffled hydrodynamic separator was 25 microns. This is shown in greater detail in Figure 3 2, where the EMC PSD for four different storms is shown. Two storms (August 1 st and July 31 st ) experienced similar peak flow rates and both exceeded t he maximum graded capacity of their respective units. The other two storms (August 23 rd and August 21 st ) had similar peak flow rates that were well below the maximum graded capacity for their respective treatment units. In those results, both units treated the influent PM more effectively when the storm had lower peak flows but, the volumetric filtration

PAGE 56

56 un it performed considerably better, even during the high intensity flow rates. The importance of translates into a higher removal efficiency of metals. Metals Removal Efficiency Since both units targeted the removal of particulate matter from runoff, a majority of the metals treatment happened in the form of particulate bound metals removal. This gave a significant advantage to the volumetric filtration unit, wh ich was able to target considerably smaller PM size ranges and was less affected by high flow rates, as previously seen in Figure 3 2. Overall, the total removal efficiency for each metals species in this study is summarized in Table 3 4. Results indicate d that both units were able to successfully treat metals in runoff, with hydrodynamic separator. Those two storms, July 28 th and July 31 st both had significantly high peak flow rates and rainfall intensities, which resulted in unit bypass and a significant portion of untreated runoff. These hydrologic characteristics, coupled with the high runoff volume, contributed to these negative removals. Throughout the study, the volumetric filtration unit produced better results than the baffled hydrodynamic separator. In Figure 3 3, the change in total mass for nickel and lead are shown for the two units. For each of the size gradations, the volumetric filtration unit performed s ubstantially better than the baffled hydrodynamic separator. Additionally, the relationship between the removal of PM and particle bound metals between each of the units was explored. In Figure 3 4, the removal efficiency of PM was compared to the removal efficiency of particle bound metals for each of the storms. As seen, a higher treatment of particulate matter corresponds to a higher treatment of particle bound metals. This was also the case for another study done by Kim and Sansalone (2011), which utili zed a treatment unit similar

PAGE 57

57 to the BHS, defined as a secondary hydrodynamic separator. Though, as show in Figure 3 4, the treatment of PM oftentimes outperformed the treatment of particle bound metals. This is explained by the preferential adsorption of m etals to small PM in the runoff, which is more difficult to target in the treatment process. Treatment of Metals by Size Gradation When exploring the relationship between particle sizes and the removal of metals, it was observed that the metals associated with the larger sized particles were more easily removed. Figure 3 5 explores the nature of lead removal in each particle size gradation from two high intensity storms. On August 1 st the volumetric filtration unit treated a runoff event that had a peak f low rate of 802.5 L/min, which was over its maximum graded capacity. As seen, almost all of the Lead associated with the sediment size range was removed. The lead mass associated with the settleable and suspended size gradations had less favorable results, though both were still able to have positive removal efficiencies and the total lead removed from this storm exceeded 48%. In comparison, the baffled hydrodynamic separator treated a high intensity storm with a peak flow rate of 794.9 L/min on July 31 st The particle bound lead associated with the sediment size range did produce a positive removal efficiency. Though, it did not perform as well as the volumetric filtration unit, which had a similar peak flow but, a significantly higher maximum rainfall inte nsity, shorter rainfall duration and nearly 50% more runoff volume. Additionally, the lead mass associated with the settleable and suspended size ranges had negative removal efficiencies. This is due to the high flow rates causing turbulence in the unit an d re suspending some of the smaller material that had previously been removed. Overall, 26.10% of the lead in the runoff from that storm was removed by the baffled hydrodynamic separator. In comparison, Figure 3 6 displays the inter event removal of lead through two lower intensity storms. On August 23 rd the volumetric filtration unit treated a storm that had a peak

PAGE 58

58 flow rate of 75.7 L/min. and the baffled hydrodynamic separator treated a storm on August 21 st with a peak flow rate of 87.1 L/min. Largely due to the low intensity runoff between the storms, the two units removed nearly all of the lead associated with the sediment and settleable size gradations. The main difference between the two units was seen when comparing the smaller sized PM. While the VF had a small positive treatment for suspended bound lead, the BHS effluent nearly mirrored the mass in the influent and even produced slightly negative results. Statistical Comparison of Q p and its Relationship to Removal Efficiency Over the ten analyz ed storm events, each of the two units experienced two storms that had peak flows (Q p ) exceeding their maximum flow rate capacity. As previously seen, these high Q p values can cause a fluctuation in the RE of a metals species in a particular storm due to t he units resulting inability to settle material and phenomena such as scour. Figure 3 7 illustrates the effect Q p has on the particle bound RE for each of the analyzed metals species in both units throughout the study. As shown, there is oftentimes a clear difference between the RE of a pollutant in a low intensity event and the RE of a pollutant in a high intensity event. To further explore this trend, the particle bound RE for each metals species with a Q p above the design flow capacity (757 L/min.) and the particle bound RE for each metals species with a Q p below the design flow capacity (757 L/min) were statistically compared. For both units, there was not any statistical significant difference between the removal efficiency of particle bound metals whe n the Q p did not exceed the designated flow capacity of the units. However, copper, nickel did have significant differences in RE between the two units when the Q p exceeded capacity and this difference can be seen in Figure 3 8. As shown for both copper an d nickel, the particle bound concentrations were significantly different between the influent and effluent between the two units. Also, for comparison the probability density functions (PDF of the dissolved nickel between the two units is shown. As seen, the influent values nearly mirror

PAGE 59

59 located in Table 3 5. Discussion The conclusions of the study explore the removal efficiency of various metals between two di ffering stormwater treatment units that received raw runoff predominately from a biogenically loaded parking lot located in Gainesville, FL. Over the span of two months, ten different storms were analyzed with each of the two units in question receiving in fluent from five measured were through settling and hydrodynamic separation. The other unit was a volumetric filtration unit which had the same treatment capabil ities as the baffled hydrodynamic separator but, had the added benefit of a filtration device installed as the last process in the treatment train. The overall removal efficiency between each of the two units had many variables. There was added complexity with the metals concentrations throughout the study varying drastically, as well as the partitioning of each metal species. Coupling the stability of each of these partitions with the influent flow rates from a storm, the actual removal efficiency of a tre atment unit can be defined. With the addition of the filtration mechanism in the volumetric filtration unit, the removal of smaller particulate matter was apparent. For metals such as copper and nickel, the overall removal efficiency was consistently high er throughout all the measured storms. Even when the peak flows exceeded the maximum graded capacity of the two units, the volumetric filtration unit had statistically significant higher removal efficiencies for these metals. While the numbers were not sta tistically significant, the same trend was seen for both lead and zinc. The removal of lead and zinc through the volumetric filtration unit was relatively stable throughout the study and was generally higher than the baffled hydrodynamic separator. Overall many variables play a

PAGE 60

60 part in the removal of metals in stormwater runoff. In this study, it was observed that the volumetric filtration unit had a greater potential for the removal of metals and the ability to produce consistent results. This observation goes hand in hand with the added elements in its treatment train. To target and treat metals in stormwater to the fullest extent, it is important to characterize the pollutants and build an adequate treatment train around those characteristics that can ta rget each of the possible partitions.

PAGE 61

61 Table 3 1 Hydrology data for all ten measured storms. Unit Event date (2010) PDH d rain h rain I max IPRT t p V runoff Q p Q 50 [hr] [min] [mm] [mm/hr] [min] [min] [L] [L/min] [L/min] Baffled Hydrodynamic Separator (BHS) 28 July 36 50 19.1 137.2 1.0 26 8316.5 961.5 83.4 4.9 31 July 15 60 23.6 76.2 5.2 17 2386.0 794.9 163.8 3.8 13 August 112 25 2.8 55.9 2.0 6 83.0 193.1 3.6 1.4 14 August 28 31 2.5 22.9 1.6 12 157.0 147.6 10.8 1.4 21 August 83 31 2.8 45.7 2.1 5 79.0 87.1 2.1 1.6 Volumetric Filtration (VF) 1 August 24 36 30.0 127.0 5.0 17 3172.0 802.5 228.0 2.3 6 August 120 104 3.6 50.8 12.0 7 369.0 408.8 2.4 6.7 7 August 24 48 8.6 50.8 7.0 22 693.0 787.4 31.8 7.3 23 August 48 42 2.8 5.1 20.0 22 82.0 75.7 8.4 3.1 12 September 172 52 6.9 50.8 17.0 13 434.0 230.9 25.2 3.4 Statistic Median 42 45 5.1 50.8 5.1 15 401.5 319.9 18.0 3.3 Mean 66 48 10.2 63.5 7.3 15 965.2 448.9 56.0 3.6 Std. Dev. 53 23 10.2 40.6 6.8 7 1160.2 349.4 79.1 2.1 PDH Previous Dry Hours d rain Event Duration h rain Rainfall Depth i max Maximum Rainfall Intensity IPRT Initial Pavement Residence Time t p Time to Peak Flow V runoff Runoff Volume Q p Maximum Flow Rate Q p Median Flow Rate Unsteadiness Parameter

PAGE 62

62 Table 3 2 Total event mean concentration (EMC) values for the influent and effluent of all Unit Event date (2010) Flow Type Cu total Ni total Pb total Zn total [g/L] [g/L] [g/L] [g/L] Baffled Hydrodynamic Separator (BHS) 28 July Influent 23 50 23 317 Effluent 10 70 12 130 31 July Influent 90 178 89 677 Effluent 60 214 66 532 13 August Influent 165 59 107 1259 Effluent 30 11 26 949 14 August Influent 77 57 92 762 Effluent 26 7 51 276 21 August Influent 57 4215 118 1562 Effluent 11 282 27 744 Statistics Median Influent 77 59 92 762 Effluent 26 70 27 532 Mean Influent 83 912 86 916 Effluent 27 117 36 526 Std. Dev. Influent 52 1847 37 494 Effluent 21 125 21 334 Volumetric Filtration (VF) 1 August Influent 93 9 14 53 Effluent 13 2 7 30 6 August Influent 412 9 107 1451 Effluent 178 3 40 763 7 August Influent 118 13 41 725 Effluent 14 6 33 511 23 August Influent 16 894 27 1544 Effluent 2 16 16 637 12 September Influent 64 104 30 1967 Effluent 14 22 18 615 Statistics Median Influent 93 13 30 1451 Effluent 14 6 18 615 Mean Influent 141 206 44 1148 Effluent 44 10 23 511 Std. Dev. Influent 157 387 37 758 Effluent 75 9 13 283

PAGE 63

63 Table 3 3 10 d 50 and d 90 particle sizes ( m) for each of the 10 sampled storms. Unit Event date (2010) Flow Type Sediment PM Settleable PM Suspended PM d 10 d 50 d 90 Gamma Parameters [mg/L] [mg/L] [mg/L] [m] [m] [m] Baffled Hydrodynamic Separator (BHS) 28 July Influent 361.6 29.2 21.4 13 110 409 193.3 0.86 Effluent 90 24.2 23 7 48 206 101.6 0.84 31 July Influent 109.2 14.1 16.2 18 128 483 214.7 0.9 Effluent 45.8 15.9 19.6 5 32 129 53.5 0.92 13 August Influent 9573.2 41.5 2.1 8 147 673 403.9 0.63 Effluent 5.9 6 0.3 3 28 164 75.2 0.7 14 August Influent 243.1 36.5 4.1 18 182 799 414.3 0.74 Effluent 7.4 4.9 0.9 5 43 225 106 0.72 21 August Influent 301.1 36.8 2.2 6 86 610 329 0.56 Effluent 3.6 4.9 0.4 4 25 212 76.6 0.69 Statistics Median Influent 301.1 36.5 4.1 13 128 610 329.0 0.74 Effluent 7.4 6.0 0.9 5 32 206 76.6 0.72 Mean Influent 2117.6 31.6 9.2 13 131 595 311.0 0.74 Effluent 30.5 11.2 8.8 5 35 187 82.6 0.77 Std. Dev. Influent 4168.8 10.7 9.0 6 37 154 103.4 0.15 Effluent 37.5 8.6 11.4 1 10 40 21.5 0.10 Volumetric Filtration (VF) 1 August Influent 243 22.7 18.5 26 213 693 304.6 0.98 Effluent 4.8 8.4 6.9 2 6 17 4.6 1.57 6 August Influent 390.4 29.5 48 16 231 984 623.9 0.65 Effluent 13.2 2.9 12.1 1 3 18 5.2 1 7 August Influent 222.5 32.2 13.1 19 186 737 638.4 0.81 Effluent 1.6 5.2 7 1 4 12 3.4 1.39 23 August Influent 459.5 35.9 38.3 14 190 714 416.1 0.73 Effluent 2.9 3.2 5.1 2 4 40 15.9 0.75 12 Sept Influent 110.5 46 45.2 9 89 328 176.3 0.79 Effluent 2.7 4.1 11.6 1 2 8 1.5 1.98 Statistics Median Influent 243.0 32.2 38.3 16 190 714 416.1 0.79 Effluent 2.9 4.1 7.0 1 4 17 4.6 1.39 Mean Influent 285.2 33.3 32.6 17 182 691 431.9 0.79 Effluent 5.0 4.8 8.5 1 4 19 6.1 1.34 Std. Dev. Influent 139.4 8.6 15.9 6 55 235 200.8 0.12 Effluent 4.7 2.2 3.1 1 1 12 5.6 0.48 scale factor of gamma distribution shape factor of gamma distribution

PAGE 64

64 Table 3 4 M ass for each PM gradation, PM total and metal. Unit Event Date (2010) Particulate Matter (PM) PM total Cu total Ni total Pb total Zn total Sediment Settleable Suspended Baffled Hydrodynamic Separator (BHS) July 28 75.12% 17.11% 7.48% 64.71% 57.02% 41.94% 47.83% 58.95% July 31 58.06% 12.73% 20.44% 46.98% 33.21% 20.39% 26.10% 21.38% August 13 99.94% 85.48% 86.19% 97.51% 81.97% 82.33% 75.23% 24.61% August 14 96.96% 86.50% 77.74% 94.84% 66.01% 88.52% 44.49% 63.77% August 21 98.80% 86.76% 81.11% 97.74% 81.53% 93.31% 76.79% 52.39% Volumetric Filtration (VF) August 1 98.02% 63.00% 62.70% 96.86% 86.11% 80.62% 48.95% 42.72% August 6 96.62% 90.17% 74.79% 97.63% 56.96% 68.97% 62.36% 47.44% August 7 99.28% 83.85% 46.56% 88.13% 88.39% 55.32% 20.19% 29.48% August 23 99.37% 91.09% 86.68% 99.15% 87.73% 98.18% 39.55% 58.75% September 12 97.56% 91.09% 74.34% 97.78% 78.80% 78.47% 40.33% 68.75%

PAGE 65

65 Table 3 5 Statistical comparison of particle bound metals species above and below designed flow capacity for both units. Flow rates above 757 L/min represent flows above the maximum graded capacity as designated by their respective manufacturer. At those flows, the re is potential bypass. Flow Range Statistical Variable Cu particle Ni particle Pb particle Zn particle Q p > 757 L/min S p 2 161.98 378.16 256.23 535.40 |t| 3.58 5.32 1.38 0.71 Q p < 757 L/min S p 2 36.49 100.69 61.87 90.96 |t| 0.91 0.86 2.44 2.19

PAGE 66

66 Figure 3 1 Time dependent fd values for Zn in runoff samples for 7 August 2010 and 13 August 2010 storm events and probability density functions of the equilibrium coefficient, Kd, for zinc in the influent and effluent for both units.

PAGE 67

67 Figure 3 2 Influent and Effluent PSDs for the August 1 st and July 31 st storms (both of which had Q p values exceeding the hydraulic design capacity for the two respective units) and August 23 rd and August 21 st (both of which had Q p values below the hydraulic design capacity for the two respective units). Range bars represent the standard deviation of the measurements.

PAGE 68

68 Figure 3 3 Total mass based M ass for the three size gradations of particulate bound nickel and lead over the course of the study for the two stormwater units. Each bar represents the total volume from 5 separate events. Range bars represent potential mass based error.

PAGE 69

69 Figure 3 4 Comparison of PM removal efficiency and particle bound metals removal efficiency for each of the storm events. Solid diagonal lines represents even removal and dashed diagonal lines represent 10% difference in removal. SHS (secondary hydrodynamic separato r) data retrieved from Kim and Sansalone (2011).

PAGE 70

70 Figure 3 5 Comparison of two storms with similar Q p values, both of which exceed the hydraulic design capacity for each of the respective units. Results show total mass over total volume for each of the particle gradations.

PAGE 71

71 Figure 3 6 Comparison of two storms with similar Q p values, both of which were below the hydraulic design capacity for each of the respective units. Results show total mass over total volume for each of the particle gradations.

PAGE 72

72 Figure 3 7 Particle Bound removal efficiencies for each meta ls species as a function of the maximum flow rate for any given storm.

PAGE 73

73 Figure 3 8 Probability density functions of total particle bound copper total particle bound nickel and dissolved nickel in both the influent and effluent for each unit.

PAGE 74

74 CHAPTER 4 GLOBAL CONCLUSIONS This study explores the kinetics of phosphorous in stormwater samples and how the particulate matter within a sample can affect the intensity and rate of the kinetic motions. Rainfall runoff was manually collected from an impervious parking surface located across from the student union on the University of Florida campus in Gainesville, FL. Primary samples from various storm events were taken and sub samples were then fractionated over time to separate the dissolved and particulate bound pollutants. Each of these sub samples was analyzed to determine their TP concentrations and, in turn, their kinetic changes were observed. Additionally, this study assessed the removal efficiency of particle bound metals from runoff using various stormwater treatment units. Both units in this study utilized hydrodynamic separation and sedimentation in the removal of PM from runoff, with one unit having an added filtration mechanism in its treatment train. In total, ten storms were analyzed, with five storms going through each unit, and the resulting influent and effluent EMCs of four studied metals were determined. Copper, nickel, lead and zinc were the metals of concern for this study and their resulting concentrations associated with each PM size gradation were ascertained. Results from the experiment on phosphorous kinetics in stormwater samples indicated that there was a significantly positive correlation in the kinetic motion of phosphorous between its partitions and the particulate matter within a sample. For all the mea sured primary samples, the kinetic motion showed an increase in particulate bound phosphorous concentrations over time as the dissolved phosphate ions in a runoff sample adsorbed to the surrounding PM. This kinetic motion had varying intensities between ea ch of the samples. One sample was observed to have a fractional change in phosphorous associated with the dissolved phase of 1% while the most intense sample had a swing of 49%. Additionally, it was observed that the size of the PM in a

PAGE 75

75 sample had a direct relationship on the amount of time it took before a primary sample reached kinetic equilibrium. For heterodispersed runoff with a high variability in the size of the internal PM and larger particles, phosphorous kinetic equilibrium was achieved in statist ically less time. While larger biogenic particles, such as leaves, have less surface area than their smaller counterparts, it is assumed that they are also more labile and willing to disassociate in the initial phase of runoff. In comparison, the smaller particles would be less susceptible to release phosphorous into the surrounding water. O nce a sample is taken or the runoff slows the dissolved phosphorous will then reattach itself to available surface sites on surrounding PM. Results from the assessment on the treatment of particulate bound metals between two different stormwater units showed that the added filtration mechanism in the volumetric filtration device provided better treatment and less variability between storm even ts. Runoff can occur in various durations and have a large scale of flow rates that enter a treatment unit. If the flow rates abilities. When comparing the events w ith low maximum flow rates, both the volumetric filtration unit and the baffled hydrodynamic separator proved to not have significant differences in their removal efficiencies for copper, nickel, lead and zinc. Though, once the flow rate went above 757 L/m in and exceeded the maximum flow rated capacity for each unit as specified by the manufacturer, it was observed that the volumetric filtration unit performed better and had less variability as a result of the flow. Additionally, the amount of particle boun d copper and nickel removed had significant differences between the two units above 757 L/min. While the volumetric unit did perform better for both particle bound lead and zinc, the differences were not statistically significant. Reasoning for this can be explained by the higher concentrations of each those metals in the suspended PM gradation. Neither of the two studied units specifically targets

PAGE 76

76 particles throughout that size range and as a result, the pollutants associated with the gradation can enter t he effluent with more ease.

PAGE 77

77 LIST OF REFERENCES Agudelo, S.C., Nelson, N.O., Barnes, P.L., Keane, T.D., Pierzynski, G.M., 2010. Phosphorous Adsorption and Desorption Potential of Stream Sediments and Field Soils in Agricultural Watershe ds. Journal of Environmental Quality 40, 144 152. APHA American Public Health Association, 1995. In: Eation, A.D., Clesceri, L.S., Greenberg, A.E., (Eds.), Standard Methods for the Examination of Water and Wastewater, 19th ed. Washington, D.C. ASTM A merican Society for Testing and Materials, 2007. Standard Test Methods for Determining Sediment Concentration in Water Samples Method D3977 97B ASTM International, West Conshohocken, Pennsylvania. Bennett, E.M., Carpenter, S.R., Caraco, N.F., 2001. Hum an Impact on Erodable Phosphorous and Eutrophication: A Global Perspective. BioScience 51 (3), 227 234. Bostrom, B., Anderson, J.M., Fleischer, S., Jansson, M., 1988. Exchange of phosphorous across the sediment water interface. Hydrobiologia 170, 229 244. Calijuri, M.L., Santiago, A.F., Neto, R.F.M., Carvalho, I.C., 2011. Evaluation of the Ability of a Natural Wetland to Remove Heavy Metals Generated by Runways and Other Paved Areas from an Airport Complex in Brazil. Water Air Soil Pollution 219, 319 327. Chiou, M.S., Li, H.Y., 2002. Equilibrium and kinetic modeling of adsorption of reactive dye on cross linked chitosan beads. Journal of Hazardous Materials B93, 233 248. Dean, C.M., Sansalone, J.J., Cartledge, F.K., Pardue, J.H., 2005. Influence of Hydro logy on Rainfall Runoff Metal Element Speciation. Journal of Environmental Engineering 131 (4), 632 642. Dechesne, M., Barraud, S., Bardin, J.P., 2004. Spatial distribution of pollution in an urban stormwater infiltration basin. Journal of Contaminant Hydrology 72, 189 205. Ellis, J.B., Harrop, D.O., Revitt, D.M., 1985. Hydrological Controls of Pollutant Removal from Highway Surfaces. Water Resources 20 (5), 589 595. Ellis, J.B., Revitt, D.M., Harrop D.O., Beckwith, P.R., 1987. The Contribution of Highway Surfaces to Urban Stormwater Sediments and Metal Loadings. The Science of the Total Environment 59, 339 349. Granier, L., Chevreuil, M., Carru, A.M., Letolle, R., 1990. Urban Runoff Pollution by Or ganochlorines (Polychlorinated Biphenyls and Lindane) and Heavy Metals (Lead, Zinc and Chromium). Chemosphere 21 (9), 1101 1107. Gromaire Mertz, M.C., Garnaud, S., Gonzalez, A., Chebbo, G., 1999. Characterisation of Urban Runoff Pollution in Paris. Water Science Technology 39 (2), 1 8.

PAGE 78

78 Haggard, B.E., Ekka, S.A., Matlock, M.D., Chaubey, I., 2004. Phosphate Equilibrium Between Stream Sediments and Water: Potential Effect of Chemical Amendments. American Society of Agricultural Engineers 47 (4), 1113 1118. H armel, R.D., Slade, R.M., Haney, R.L., 2010. Impact of Sampling Techniques on Measured Stormwater Quality Data for Small Streams. Journal of Environmental Quality 39, 1734 1742. Hatt, B.E., Fletcher, T.D., Deletic, A., 2008. Hydraulic and Pollutant Remova l Performance of Fine Media Stormwater Filtration Systems. Environmental Science Technology 42, 2535 2541. Kim, J.Y., Sansalone, J.J., 2011. Representation of Rainfall Runoff Metal Loads, Partitioning, and Unit Operation Behavior through Differing Particu late Matter Indicies. Journal of Environmental Engineering (In Press). Lai, D., Lam, K.C., 2009. Phosphorous sorption by sediments in a subtropical constructed wetland receiving stormwater runoff. Ecological Engineering 35, 735 743. Liu, D., Sansalone, J .J., Cartledge, F.K., 2005. Adsorption Kinetics for Urban Rainfall Runoff Metals by Composite Oxide Coated Polymeric Media. Journal of Environmental Engineering 131 (8), 1168 1177. Legret, M., Nicollet, M., Miloda, P., Colandini, V., Raimbault, G., 1999. Simulation of Heavy Metal Pollution from Stormwater Infiltration Through a Porous Pavement with Reservoir Structure. Water Science Technology 39 (2), 119 125. Ma, J., Ying, G., Sansalone, J.J., 2010. Transport and Distribution of Particulate Matter Phosph orous Fractions in Rainfall Runoff from Roadway Source Areas. Journal of Environmental Engineering 136 (11), 1197 1205. Mahbub, P., Ayoko, G.A., Goonetilleke, A., Egodawatta, P., 2011. Analysis of the build up of semi and non volatile organic compounds on urban roads. Water Research 45, 2835 2844. McClave, J., Sincich, T., 2006. Statistics 10th Edition. Pearson Prentice Hall, p. 436. Ozacar, M., 2003. Equilibrium and Kinetic Modelling of Adsorption of Phosphorous on Calcined Alunite. Adsorption 9, 125 132. Pitcher, S.K., Slade, R.C.T., Ward, N.I., 2004. Heavy metal removal from motorway stormwater using zeolites. Science of the Total Environment 334 335, 161 166. Read, J., Wevill, T., Fletcher, T., Deletic, A., 2008. Variation among plant species in pollutant removal from stormwater in biofiltration systems. Water Research 42, 893 902.

PAGE 79

79 Reinelt, L.E., Horner, R.R., 1995. Pollutant removal from stormwater runoff by palustrine wetlands based on comprehensive budgets. Ecological Engineering 4, 77 97. Rentz, R., Widerlund, A., Viklander, M., Ohlander, B., 2011. Impact of Urban Stormwater on Sediment Quality in an Enclosed Bay of the Lule River, Norther Sweden. Water Air Soil Pollution 218, 651 666. Sansalone, J.J., Bucherger, S.G., 1997. Partitioning and First Flush of Metals in Urban Roadway Storm Water. Journal of Environmental Engineering 123 (2), 134 143. Sansalone J., Kim J.Y., 2008. Transport of Particulate Matter Fractions in Urban Source Area Pavement Surface Runoff. Journal of Environmental Q uality. 37 (5), 1883 1893. Sharpley, A.N., Chapra, S.C., Wedepohl, R., Sims, J.T., Daniel, T.C., Reddy, K.R., 1994. Managing Agricultural Phosphorous for Protection of Surface Waters: Issues and Options. Journal of Environmental Quality 23, 437 451. Sora nno, P.A., Hubler, S.L., Carpenter, S.R., 1996. Phosphorous Loads to Surface Waters: A Simple Model to Account for Spatial Pattern of Land Use. Ecological Applications 6 (3), 865 878. Stanley, D.W., 1996. Pollutant removal by a stormwater dry detention po nd. Water Environment Research 68 (6), 1076 1082. Tsihrintzis, V.A., Hamid, R., 1997. Modeling and Management of Urban Stormwater Runoff Quality: A Review. Water Resources Management 11, 137 164. USEPA United States Environmental Protection Agency, 199 2. Acid Digestion of Aqueous Samples and Extracts for Total Metals for Analysis by FLAA or ICP Spectroscopy. SW 846, Revision 1, Washington, D.C. USEPA United States Environmental Protection Agency, 1997. Inductively Coupled Plasma Mass Spectormetry. SW 846, Revision 1, Washington, D.C. Vaze, J., Chiew, F.H.S., 2004. Nutrient Loads Associated with Different Sediment Sizes in Urban Stormwater and Surface Pollutants. Journal of Environmental Engineering 130 (4), 391 396. Vohla, C., Koiv, M., Bavor H.J., Chazarenc, F., Mander, U., 2011. Filter materials for phosphorous removal from wastewater in treatment wetlands A Review. Ecological Engineering 37, 70 89. Walker, D.J., Hurl, S., 2002. The reduction of heavy metals in a stormwater wetland. Ecolo gical Engineering 18, 407 414.

PAGE 80

80 Wang, H., Appan, A., Gulliver, J.S., 2003. Modeling of phosphorous dynamics in aquatic sediments: I model development. Water Research 37, 3928 3938. Wang, Q., Li, Y., 2010. Phosphorous adsorption and desorption behavior on sediments of different origins. Journal of Soils Sediments 10, 1159 1173. Whipple, W., Hunter, J.V., 1977. Nonpoint sources and planning for water pollution control. Journal WPCF, 15 22. Whipple, W., Grigg, N., Grizzard, T., Randall, C., Shubinski R., Tucker, L., 1983. Stormwater Management in Urbanizing Areas. Prentice Hall, Inc., p. 68 69. Yousef, Y.A., Wanielista, M.P., Hvitved Jacobsen, T., Harper, H.H., 1984. Fate of Heavy Metals in Stormwater Runoff from Highway Bridges. The Science of the Total Environment 33, 233 244. Zhou, A., Tang, H., Wang, D., 2005. Phosphorous adsorption on natural sediments: Modeling and effects of pH and sediment composition. Water Research 39, 1245 1254.

PAGE 81

81 BIOGRAPHICAL SKETCH K arl Seltzer graduated with his bachel egree in e nvironmental e ngineering s ciences from the University of Florida in December 2009. Immediately after his BA, Karl began his Master of Engineering in e nvironmental e ngineering s ciences work at the University of Florida under the guidance of Dr. John J. Sansalone. His research primarily focused on the kinetic nature of phosphorous in stormwater samples as well as the treatment of metals species using various stormwater treatment techniques. Following graduation, Karl remained in Gainesville, F L and b egan work in the environmental e ngineering field.