Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2013-08-31.

MISSING IMAGE

Material Information

Title:
Record for a UF thesis. Title & abstract won't display until thesis is accessible after 2013-08-31.
Physical Description:
Book
Language:
english
Creator:
Herr,Christina Beth
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.E.)
Degree Grantor:
University of Florida
Degree Disciplines:
Environmental Engineering Sciences
Committee Chair:
Sansalone, John
Committee Members:
Chadik, Paul A
Koopman, Ben L

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 Christina Beth Herr.
Thesis:
Thesis (M.E.)--University of Florida, 2011.
Local:
Adviser: Sansalone, John.
Electronic Access:
INACCESSIBLE UNTIL 2013-08-31

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2011
System ID:
UFE0043266:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 EXPERIMENTAL ANALYSIS OF RAINFALL RUNOFF SAMPLING TECHNIQUE S AND VOLUMETRIC CLARIFYING FILTER TREATMENT PROCESSES By CHRISTINA HERR A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILL MENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ENGINEERING UNIVERSITY OF FLORIDA 2011

PAGE 2

2 2011 Christina Herr

PAGE 3

3 To my wonderful fianc Matt, my parents, my sister Jenny, and my grandma, who have always believed in me

PAGE 4

4 ACKNOWLEDGMENTS I w ould like to thank Dr. John J. Sansalone, who mentored me throughout the course of my graduate research. I would also like to thank Dr. Paul A. Chadik and Dr. Ben Koopman, who supported me by serving as members of my committee. Lastly I would like to thank Hwan Chul Cho, Josh Dickenson, Karl Seltzer, Hao Zhang, Greg Brenner, Aniela Burant, Moshik Doron, Jacqueline Martin, Adam Marquez, Julie Midgette, and Valerie Thorsen, who collaborated on fieldwork and laboratory analysis.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................. 7 LIST OF FIGURES ........................................................................................................... 8 LIST OF ABBREVIATIONS ............................................................................................. 9 ABSTRACT ................................................................................................................... 13 CHAPTER 1 GLOBAL INTRODUCTION ..................................................................................... 15 2 THE ROLE OF SAMPLING IN REPRESENTING RAINFALL RUNOFF PARTICLE MATTER GRANULOMETRY .................................................................................. 17 Introduction .............................................................................................................. 17 Objectives ................................................................................................................ 19 Methodology ............................................................................................................ 20 Unit Operations .................................................................................................. 20 Urban Source A rea Watershed ............................................................................. 21 Rainfall Runoff Event based Data Collection ........................................................ 21 Laboratory Analysis ............................................................................................ 22 Data Analysis and Elaboration .............................................................................. 23 Event mean concentration .............................................................................. 23 Removal efficiency ....................................................................................... 24 Particle size distribution (PSD) ...................................................................... 24 Cumulative gamma distribution ..................................................................... 25 Particle number d ensity (PND) ...................................................................... 26 Kruskal Wallis H test for significant difference ............................................... 27 Normalized root mean square error ................................................................. 28 Newtons law ............................................................................................... 28 Results and Discussion............................................................................................... 29 Conclusions .............................................................................................................. 36 3 PARTICULATE MATTER AND PHOSPHORUS REMOVAL IN RAINFALL RUNOFF TREATED WITH VOLUMETRIC FILTRATION ........................................ 52 Introduction .............................................................................................................. 52 The Role of Urban Development .......................................................................... 52 History of Stormwater Regulations ....................................................................... 53 Current Research ................................................................................................ 54 Objectives ................................................................................................................ 55

PAGE 6

6 Methodology ............................................................................................................ 55 Watershed Configuration ..................................................................................... 55 Volumetric Clarifying Filter and Monitoring System .............................................. 55 Calibration Procedures ........................................................................................ 58 Calibra tion of MJK ultrasonic sensor for flow depth measurements ................... 58 Calibration of pressure transducers ................................................................. 58 Preliminary hydraulic testin g ......................................................................... 59 Monitoring Methodology ..................................................................................... 59 Assessing PM Loads and Transport in Urban Rainfall Runoff ................................. 60 Data Analysis and Elaboration .............................................................................. 62 Volumetric rainfallrunoff coefficients ............................................................ 62 Event mean value ......................................................................................... 62 Removal efficiency ....................................................................................... 63 Particle size distribution (PSD) ...................................................................... 63 Mass balan ce error ........................................................................................ 64 Mann Whitney U test for significant difference ............................................... 64 Newtons law ............................................................................................... 65 Surface overflow rate .................................................................................... 65 Surface loading rate ...................................................................................... 66 Cumulative gamma distribution (CGD) function .............................................. 66 Results and Discussion............................................................................................... 67 Event Hydrology ................................................................................................. 67 Particulate Matter (PM) ....................................................................................... 68 Particle Size Distribution (PSD) ........................................................................... 70 Phosphorus ......................................................................................................... 73 Conclusion ............................................................................................................... 74 4 GLOBAL SUMMARY AND CONCLUSIONS ........................................................... 92 LIST OF REFERENCES .................................................................................................. 95 BIOGRAPHICAL SKETCH ........................................................................................... 100

PAGE 7

7 LIST OF TABLES Table page 21 Monitored rainfall runoff event characteristics ......................................................... 38 22 Summary of event based PM fractions .................................................................... 39 23 Summary of event based PM indices of d10, d50, and d90. ........................................... 40 24 NRMSE between the incremental mass of automated samples and manual samples at each particle size across the PSD ............................................................................ 41 25 Summary of event based cumulative gamma distribution parameters for PSD. ............ 42 26 Summary of event based particle number density (PND) indices. ............................... 43 27 Summary of event based power law model parameters for PND ................................ 44 31 Monitored rainfall runoff event characteristics ......................................................... 76 32 Summary of event based values for PM fractions ..................................................... 77 33 Comparison of event based PM indices of d10, d50, and d90 for influent, primary effluent, and secondary effluent .............................................................................. 78 34 Comparison of event based SSC for influent, primary effluent, and secondary effluent ................................................................................................................ 79 35 Comparison of event based gamma parameters for influent, primary effluent, and secondary effluent ................................................................................................. 80 36 Summary of event based values for dissolved and particulate bound phosphorus ......... 81

PAGE 8

8 LIST OF FIGURES Figure page 21 Elevation view of monitoring system design ............................................................ 45 22 Example of manual and automated influent PSD (14 August 2010 rainfall runoff event) .................................................................................................................. 46 23 Manual and automated influent and effluent gamma distribution parameters ............... 47 24 Manual and automated influent and effluent power law parameters for particles less than 75 m ........................................................................................................... 48 25 Relationship between turbidity and PND ................................................................. 49 26 Examples of manual and automated influent PSD vs. PND (13 August 2010 and 23 August 2010 rainfall runoff events) ......................................................................... 50 27 Kruskal Wallis H test results for PSD ..................................................................... 51 31 Volumetric clarifying filter treatment unit design ..................................................... 82 32 Cumulative frequency distribution of rainfall depth for Gainesville, FL (GNV) based on 30 years of rainfall data (19802010) .................................................................. 83 33 Suspended (< 25 m), settleable (25 75 m), sediment (> 75 m) PM fractions and SSC (all PM fractions) for influent and effluent ........................................................ 84 34 Probability density functions (p = 0.05) ................................................................................................................. 85 35 Median and range of variation for the influent and effluent PSD from the entire monitoring campaign (n=34) .................................................................................. 86 36 Measured influent and secondary effluent PSD and modeled primary effluent PSD for the 1 August 2010 event ................................................................................... 87 37 Probability density functions (pdfs) for surface overflow rate, SOR and surface ............................................................................. 88 38 Event based cumulative gamma distribution parameters for measured influent, modeled primary eff luent, and measured secondary effluent PSD ............................... 89 39 Event based influent and effluent dissolved phosphorus and PM based phosphorus ..... 90 310 Probability density functions (pdfs) for flow rate and dissolved and particulate bound ...................................................................................... 91

PAGE 9

9 LIST OF ABBREVIATION S Afilter, s surface area of filter (m2) As surface area (m2) Aw contributing area of the watershed (L2) b cumulative power law exponent based on particle diameter, dimensionless BHS baffled hydrodynamic separator c runoff coefficient (dimensionless) flow weighted mean concentration of the constituent of interest for an e vent (g/L) Cd frictional drag coefficient (dimensionless) CGD cumulative gamma distribution Ci average concentration associated with period i (mg/L) Ci IN mean influent concentration associated with sampling period i (mg/L) Cj EFF mean effluent concentration associated with sampling period j (mg/L) CWA clean water act dp diameter of a particle (m) EMC event mean concentration (mg/L) EM V event mean value EPA E nvironmental P rotection A gency g gravitational acceleration (m/s2) h rainfall depth (cm ) Hcalc calculated test statistic of the Kruskal Wallis H test for significant difference In rainfall intensity of sample period n (L/s) IPRT initial pavement residence time (minutes) K total number of treatments in a Kruskal Wallis H test

PAGE 10

10 l particle size parameter for the cumulative power law model m total number of effluent measurements taken during an event M total mass of the constituent of interest ( g ) MBE mass balance error (%) eff cumulative mass load of the effluent across an entire event (g) inf cumulative mass load of the influent across an entire event (g) rec total mass of PM recovered after an event (g) MS4 municipal separate storm sewer system MT FR maximum treatment flow rate ( L/s ) n total number of influent measurements tak en during an event N particle number density for the cumulative power law model ( count/m3) NPDES N ational Pollutant Discharge Elimination S ystem NPS nonpoint source NRMSE normalized root mean square error NWS N ational W eather S ervice PDH previous dry hours ( hr ) PLM power law model PM particulate matter PND particle number d istribution ( count/L ) PSD particle size distribution ( mg/L ) Q volumetric flow rate (m3/s) QA/QC quality assurance / quality control Qmax maximum volumetric flow rate (L/s)

PAGE 11

11 Qmea n mean volumetric flow rate (L/s) Qmed median volumetric flow rate (L/s) Qn volumetric flow rate of sample period n (L/s) Rd particle Reynolds number (dimensionless) ri the rank sum for treatment i RMSE root mean square error ( mm) RPD ratio of predi ction to deviation (dimensionless) s total number of data points in a Kruskal Wallis H test SD standard deviation SDV standard deviation of the measured values in the prediction set si the number of samples in treatment i SLR surface loading rate (L/m in/m2) SSC suspended sediment concentration ( mg/L ) SSE sum of squared error SOR surface overflow rate (m/s) TSS total suspended solids ( mg/L ) Ui Mann Whitney test statistic for treatment i UOP unit operations and processes V tot al volume of runoff i n an event ( L ) Vi volume of flow during period i (L) Vi IN volume of influent flow for sampling period i ( L ) Vj EFF volume of effluent flow for sampling period j ( L ) VCF volumetric clarifying filter

PAGE 12

12 Vs particle settling velocity ( m/s ) Vs,crit critica l settling velocity (m/s) VSS volatile suspended solids (mg/L) X1 j jth measurement of treatment 1 mean of the measurements of treatment 1 X2 j jth measurement of treatment 2 mean of the measurements of treatment 2 xmax ma ximum measured value of the manual sample PSD power law index based on particle diameter (count/L3) cumulative power law exponent based on particle diameter (dimensionless) k shape factor of cumulative gamma distribution function scale factor of cumulative gamma distribution function f fl uid viscosity (g/ms) f mass density of a fluid (g/m3) p mass density of a particle (g/m3)

PAGE 13

13 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree o f Master of Engineering EXPERIMENTAL ANALYSIS OF RAINFALL RUNOFF SAMPLING TECHNIQUES AND VOLUMETRIC CLARIFYING FILTER TREATMENT PROCESSES By Christina Herr August 2011 Chair: John J. Sansalone Major: Environmental Enginee r ing Sciences Rainfall runoff carries particulate matter (PM) from urban and agricultural land to surface waters and is the leading cause of impairment across all types of waterbodies. This impairment is partly due to the ecological impacts of PM on receiving waters and partly because many constituents of concern, like nutrients and metals, partition into these particles and are transported to surface waters in urban rainfall runoff. It is crucial to accurately assess PM granulometry, mass load, and transport in order to select an eff ective unit operation to protect receiving waters from impairment. One main objective of this study i s to examine the hypothesis that a PM analysis based on manual sampling provides a more representative assessment of PM load and transport in urban rainfal l runoff for a small watershed than one based on peristaltic pumpdriven automated sampling. The other main objective i s to examine the in situ testing of a volumetric clarifying filter ( VCF ) treatment unit for the removal of PM and phosphorus from urban rainfall runoff for a small watershed. This study use s a series of paired manual grab samples and automated samples analyzed for suspended sediment concentration (SSC) particle size distribution (PSD) particle number distribution (PND), and PM fractions to compare the two sampling techniques. Tests for

PAGE 14

14 significant difference (Kruskal Wallis H t est) are conducted on PSDs and PNDs. The cumulative gamma distribution (CGD) is used to model PSDs. I nfluent CGD parameters are significantly different (p < 0.05) between sampling methods. PND results are modeled by the power law model (PLM) Neither influent nor effluent PLM parameters are significantly different between sampling methods. Results indicate manual sampling is more appropriate f or size heterodisperse PSDs. However, for size monodisperse effluent, as generated through primary clarification with or without filtration, with a relatively small median particle size, d50, properly deployed automatic sampling can represent PSD and PND. In order to assess the treatment potential of a VCF unit, field testing including insitu intra event data collection and analysis of 25 a ctual rainfall runoff events, is conducted on the unit. PM i s separated into size fractions (sediment, settleable, and suspended PM; SSC) and quantified for both pre and post treatmen t samples. Laser diffraction i s conducted to determine PSD. PM fractions a re digested and analyzed for soluble reactive phosphorus (SRP) concentration. Results indicate that the VCF shows significant removal capabilities for both PM and particulate bound P. The volume weighted mean removal efficiency for total s olids (in the form of SSC) is 98% with the greatest removal coming from coarse particles (> 75 m). Sediment bound phosphorus is removed with the highest volume weighted mean efficiency of 99% while the volume weighted mean removal of total phosphorus is 68%. PM separation is divided into the two removal mechanisms utilized by the VCF, sedimentation and filtration to determ ine the relative contributions of each mechanism. Results indicate that removal of PM is significantly increased when filtration is utilized in addition to sedimentation.

PAGE 15

15 CHAPTER 1 GLOBAL INTRODUCTION P articulate matter (PM) from urban and agricultural land is conveyed to surface waters in rainfall runoff. This is the leading cause of surface water impairment across all types of w aterbodies (EPA, 2000). Impairment is due partly to the ecological impacts of PM on receiving waters and partly because man y constituents of concern, like nutrients and metals, partition into these particles and are transported to surface waters in urban rainfall runoff (Legret and Pagotto, 1999; Sansalone and Kim, 2008; Liu et al., 2010). It is crucial to accurately assess P M granulometry, mass load, and transport both because of impacts of the solids and particulate bound constituents on receiving waters and because most stormwater treatment devices are assessed by their ability to remove these solids (Greb and Bannerman, 1997; Cristina et al., 2002; Sansalone and Kim, 2008). One difficulty associated with stormwater sampling is choosing an appropriate sampling technique. The two most common techniques are manual grab sampling and automated sampling. Automated sampling is commonly used to sample rainfall runoff for nonpoint source pollutants in small watersheds whereas manual sampling is more often used to sample larger streams and rivers. Automated samplers are often preferred in sampling projects because manual sampling can be more challenging. Manual grab sampling requires personnel to anticipate rainfall events, travel to the sampling site, and work in adverse weather conditions, all of which can be costly, timeconsuming, and inconvenient (Harmel et al., 2003). However, studies have indicated that automated samplers used in rainfall runoff monitoring are not capable of repeatedly capturing particles greater than 250 Clark et al., 2007). Using automated samplers in rainfal l runoff monitoring has the effect of misrepresenting PSDs, PM loads, and particulate bound phosphorus loads and can potentially

PAGE 16

16 lead to the improper selection and design of unit processes. Sampling technique is also critical in evaluating the treatment p otential of stormwater unit operations and processes (UOPs). Only with an appropriate sampling technique can rainfall runoff treatment technologies be assessed accurately A major challenge in stormwater management is selecting an appropriate rainfallr unoff treatment strategy. In many urban areas where there is limited area and high land cost, combined unit operations are preferred to individual treatment processes and large settling tanks (Grizzard et al., 1986). Baffled hydrodynamic separator s ( BHS s ) have been favored in recent years for their small footprint (Wong et al., 1995; Walker et al., 1999) ; h owever, filtration unit operations have become more common for PM separation in rainfall runoff (Liu et al., 2010). Selection of filtration unit opera tions depends on knowledge of PM load and granulometry, along with headloss development and desired effluent water indices such as suspended sediment concentration ( SSC ) total suspended solids ( TSS ) and particle size distribution ( PSD ) (Clark, 2000). I t is therefore critical to quantify and qualify the inputs to the unit and the outputs to receiving waters under a variety of rainfall intensities, runoff durations, and PM loads when evaluating the performance of a volumetric clarifying filter ( VCF ) unit.

PAGE 17

17 CHAPTER 2 THE ROLE OF SAMPLING IN REPRESENTING RAIN FALL RUNOFF PARTICLE MATTER GRANULOMETRY Introduction Rainfall runoff carries particulate matter (PM) from urban and agricultural land to lakes, streams, and rivers and is the leading cause of impairment across all types of waterbodies. In surface waters, the PM contained in surface runoff is a major constituent contributing to poor water quality (EPA, 2000). This is partly because particles can have ecological impacts on receiving waters and partly b ecause many constituents of concern, like nutrients and metals, partition into the solid phase and are transported to surface waters in urban rainfall runoff (Legret and Pagotto, 1999; Sansalone and Kim, 2008; Liu et al., 2010). The partitioning of these constituents into PM in rainfallrunoff is highly affected by land use (e.g. transportation, recreation, and construction), particle size, and pollutant age (Liebens, 2001). It is therefore crucial to accurately assess PM granulometry as well as loads and transport in rainfall runoff in order to select an effective unit operation or source control plan and to protect receiving surface waters from impairment (Greb and Bannerman, 1997; Cristina et al., 2002; Sansalone and Kim, 2008). Currently, there are sev eral difficulties associated with accurately assessing PM loads and transport in rainfall runoff. One issue is choosing the PM index to be studied. There are two common gravimetric methods for quantifying PM: total suspended solids (TSS), Standard Methods 2540D (Standard Methods, 1995), and suspended sediment concentration (SSC), ASTM D397797 (Gray et al., 2000). Unfortunately both methods fail to give information about particle size distribution (PSD) and particle number distribution (PND). They only y ield total mass load and mass concentration, which are insufficient for evaluating the behavior and maintenance of unit operations (Greg and Bannerman, 1997; Cristina et al., 2002). Additionally,

PAGE 18

18 with the TSS method, a subsample is used for the analysis. Studies have shown that, for a sample containing a significant portion of coarse particles, defined as PM greater than 75 m in diameter (ASTM, 2002; Rushton et al., 2007), the SSC method depicts a more accurate assessment of the total PM mass present in a sample (Grey et al., 2000; Clark and Siu, 2007). The aliquot taken for TSS is often not representative of the whole sample, which tends to underestimate the PM load of the whole sample and therefore underestimate the actual PM load of the runoff. On t he contrary, the SSC method overcomes the challenge of taking a representative subsample by analyzing the entire volume. Therefore, if the sample is truly representative of the runoff, the SSC method is not affected by heterodispersity or the presence of sediment size particles ( > 75 m ) (Sansalone and Kim, 2008). Another difficulty associated with accurately assessing PM loads is the selection of a sampling technique. There are two common sampling techniques: manual grab sampling and automated sampling When sampling rainfall runoff for nonpoint source pollutants in small watersheds, automated sampling is most commonly used, whereas manual grab sampling is more commonly used for larger streams and rivers (Harmel et al., 2003). Although automated sampl ers were originally developed for use in wastewater treatment, they are often preferred in stormwater sampling projects for small watersheds because the alternative, manual sampling, can be much more difficult. Manual grab sampling requires personnel to a nticipate rainfall events, travel to the sampling site, and work in adverse weather conditions. This can be costly, time consuming, and inconvenient (Harmel et al., 2003). However, historically, stormwater monitoring studies utilizing automated samplers produce results that differ from those of studies using manual sampling (Sansalone et al., 1998; Furamai et al., 2002; Westerlund and Viklander, 2006). Currently, automated samplers used in

PAGE 19

19 rainfall runoff monitoring are not capable of repeatedly capturing particles greater than 250 m This is mainly due to lack of sufficient turbulence in pipes leading to a poorly mixed water column (Bent et al., 2001; Burton and Pitt, 2001; Clark et al., 2007). U sing automated sampling methods requires several assumpti ons to be made and, therefore, has several constraints. Automated sampling strategies typically assume that the entire watershed of interest is homogeneous and that water quality can be sampled at one intake point and be valid across the entire stream (Harmel et al., 2003). While this is usually the case with wastewater, it is often not with urban rainfall runoff. Studies show that there are major differences between the PM of wastewater and that of urban rainfall runoff (Furamai et al., 2002; Li et al., 2005). The PSD of typical wastewater is fine and size monodisperse. However, the PSD of typical urban rainfall runoff is coarser and size heterodisperse with particles ranging from 1 to 10,000 microns in size (Li et al., 2005; Sansalone and Kim, 2008). Although studies show that coarse PM is more easily removed from runoff than the suspended fraction, accurate information about this part of the gradation is important (Andral et al., 1999). A significant proportion of the total mass of constituents par titioning into the PM are associated with the sediment size fraction. Studies show that more than 60 percent of particulate bound metal mass (including Cd, Cu, Pb, and Zn) is associated with particles greater than 250 m in diameter (Sansalone and Cristi na, 2004). Furthermore, coarser fractions can significantly impact drainage systems conveyance capacity and runoff leaching in conveyance systems Objectives The primary goal of this study is to examine the hypothesis that a PM analysis, based on manual grab sampling paired with the SSC method and PSD characterization, can provide a more representative assessment of PM load and transport in rainfall runoff for a small watershed

PAGE 20

20 than a PM analysis, based on peristaltic pumpdriven automated sampling paired with the SSC method and PSD characterization. The second objective is to identify for which particle sizes if any, the results of the two sampling methods are significantly different. The third objective is to examine the hypothesis that although autom ated sampling may not provide representative results for PM load in size heterodisperse rainfall runoff, representative results are produced for the size monodisperse treated effluent from a volumetric clarifying filter (VCF) This is based on a VCF which implements hydrodynamic separation and filtration and therefore produces effluent that does not contain any sediment size particles. Methodology Unit Operations This study utilizes two separate unit operations, a baffled hydrodynamic separator (BHS) and a VCF. The BHS unit used in this study is 1.2 meters in diameter and holds a volume of 1703 liters. It has a maximum treatment flow rate (MTFR) of 9.1 L/s The system components include an inlet, an inlet drop pipe surrounded by a weir, an oil port, an outlet riser pipe, and an outlet. The vertical distance between the outlet pipe invert and the bottom of the base slab is 173 centimeters. The VCF unit is also 1.2 meters in diameter and is configured with two standard cartridges, each with a 70 millimete r orifice cartridge lid, and one draindown cartridge with a 35millimeter orifice cartridge lid. Each cartridge contains eleven 20 micrometer filter tentacles made up of two 69centimeter long segments. The system components include a 46centimeter diame ter maintenance access pipe, a pressure relief pipe, a cartridge deck, a backwash pool weir located around the two standard cartridges, and a separator skirt. The unit has a MTFR of 12.6 L/s and is configured to begin bypassing flow when head loss reaches 46 centimeters. This is equivalent to a water elevation of 46 centimeters above the cartridge deck level.

PAGE 21

21 Urban Source Area Watershed The Reitz Union surface parking facility located on the University of Florida campus serves as the watershed in this stu dy. Rainfall runoff is collected from an existing stormdrain inlet at the southwest parking lot pavement edge as shown in Figure 21. The watershed is divided into two main subcatchments, a parking area on the southeast side and an access road on the northwest side. Depending on the storm event, the watershed area ranges from 400 to 600 square meters with a median area of approximately 500 square meters. A butterfly valve is used at the inlet to prevent unwanted flow from entering the system. A 15 cen timeter diameter PVC pipe carries runoff at a 6 percent slope from the inlet to a Parshall flume located 10 meters downstream. An ultrasonic sensor is fixed in the Parshall flume to record hydrology data. A 61 cm x 46 cm x 46 cm influent sampling drop box is located immediately downstream of the Parshall flume. A 15centimeter diameter PVC tee section carries the runoff from the outfall of the dropbox to the two treatment units. Diversion g ate valves are used on each side of the tee to direct all of the influent to one of the two units. The VCF and the BHS are installed in parallel and a 61 cm x 46 cm x 46 cm effluent sampling drop box is located at each outfall. There is an additional bypass pipe network for the VCF. It consists of a 15 centimeter di ameter PVC pipe elevated to 122 centimeters above the units internal cartridge deck level, with a 5 centimeter diameter siphon break 30 centimeters above that. In the event that bypass occurs during an event treated by the VCF, untreated bypass effluent and treated effluent are sampled either as a composite or as two individual streams in the effluent sampling drop box. Bypass flow rate is measured with a 1 psi pressure transducer located in the bypass pipe. None of the events monitored for this study g enerated bypass. Rainfall Runoff Event based Data Collection Hydrology and PM data are collected for 8 events. Runoff samples are collected

PAGE 22

22 using both manual grab sampling and automated sampling on a time basis using radar and observations made by physic a l inspection. The automated samplers used in this study are ISCO 3700 peristaltic pumpdriven automated samplers. Hydrology data are measured from an ultrasonic sensor located in the Parshall flume. Synchronized stop watches are used to record all samp ling times and samples are volume weighted based on these sampling times and paired flow rate measurements Manual influent samples are taken in the influent sampling drop box. Manual effluent samples are taken in the effluent sampling drop box of the appropriate unit. Automated influent samples are taken in one of two ways; for four of the monitored events, the influent automated sampling intake tube is located directly downstream of the inlet at the parking lot pavement edge. The intake tube is placed at the bottom of the pipe and oriented to face the flow head on. For the other four monitored events, the influent automated sampling intake tube is located in the outfall of the Parshall flume. The intake tube is placed in the vertical and horizontal c enter of flow, facing the flow. Automated effluent samples for all events are taken the same way. An effluent automated sampling intake tube is fed into the outlet pipe of each unit the VCF and the BHS, and fixed along the bottom of the outlet pipe so a s to face flow. Placing the automated sampler intake tube facing the flow and either locating it at the bottom of the flow or locating it at the center of flow are the two most common automated sampling collection methods recommended by manufacturers (Tel edyne Isco, Inc., Lincoln, NE; American Sigma, Inc., Loveland, CO), required by regulators (EPA, 1992; Colorado Department of Public Health and Environment, 1996; State of Maine Bureau of Land and Water Quality Management, 2007), and ultimately selected fo r implementation (de Leon and Lowe, 2009; Reed, 1981). Laboratory Analysis After samples are collected, samples are transported directly back to the laboratory to be analyzed. Within the first 2 hours after the end of an event, turbidity measurements are taken on

PAGE 23

23 each sample using a t urbidimeter ( Hach 2100AN Turbidimeter, ISO Method 7027). A 10 mL subsample is taken from the well shaken 0.5 liter PSD samples for this turbidity measurement. Within the first 6 hours, laser diffraction is conducted using a laser diffraction particle analyzer (Malvern Instruments: Hydro 2000G) to determine PSDs from the 0.5 liter PSD samples. Prior to being poured into the analyzer, each sample is pre screened with a #4 2000m sieve. This is done so that PM or other debris larger than 2000 m is not allowed into the laser chamber. For each sample, the material collected in the #4 sieve is placed in a preweighed pan, dried in the oven at 105C for 24 hours, and a final weight is recorded. Once samples are added to the wet chamber of the analyzer, multiple repetitions are required to reach stability across the entire gradation. The output from this instrument is in terms of percent finer by volume for each particle size ranging from 0.02 to 2000 m. With paired SSC data ( concentration and sample volume) from the remaining 0.5 liter samples, this output is converted to percent finer by mass. SSC measurements are considered to be less timesensitive than the PSD analysis and are conducted within 24 hours of the end of an e vent. 0.5 liter SSC samples are filtered through a 1 m glass fiber filter with a vacuum pump. Filters are pre washed with DI water, dried in a furnace at 550C for 30 minutes, and weighed prior to use. The precise volume of each sample being filtered is also recorded. After filtering the samples, the filters and PM material captured by the filters are dried at 105C. The final weights are recorded and the difference in dry weights is used to calculate the mass of PM and the SSC. Data Analysis and Elab oration Event mean concentration While the inter event distr ibution of PM is log normal, a volume weighted event mean concentration (EMC) is utilized as a comparative index for characterizing PM concentrations that

PAGE 24

24 represent a flow weighted PM concentratio n for a rainfall runoff event (Sansalone and Kim, 2008). EMC is calculated using equation 21. C V M EMC ( 21) In this expression, M represents the total mass of the constituent of interest (for example, SSC) over an entire event, V r epresents the total volume of runoff over that same event, and C represents the flow weighted average concentration of the constituent of interest for an entire event. Removal efficiency While effluent concentrations provide a more phys ically discernable measurement of treatment performance, a common comparative index for evaluating PM separation is percent removal. Percent removal is calculated using equation 22. 100 ) ( ) ( ) ( (%)1 1 1 IN i n i IN i EFF j m j EFF j IN i n i IN iC V C V C V PR (2 2) In this expression, Vi IN is the volume of influent flow for sampling period i; Vj EFF is the volume of effluent flow for sampling period j; Ci IN is the mean influent concentration associated with period i; Cj EFF is the mean effluent concentration associated with period j; n is the total number of influent measurements taken during an event; m is the total number of effluent measurements taken during an event. Particle size distribution (PSD) PSD is a primary measurement of PM granulometry. The output of the laser diffraction instrument is in percent finer by volume. A more common index is percent finer by mass, and

PAGE 25

25 the commensurate indices of d10, d50, and d90 by mass In order to make this conversion, the measured gravimetric mass (from SSC) is distributed over the entire PSD for each samp le. Massbased cumulative PSDs (i.e. percent finer by mass) for the samples are computed by integrating the gravimetric concentration of individual particle sizes for samples, corresponding flow rates, and sam pling time intervals. With these paired data, PM total volume concentration for each particle size is converted to a gravimetric concentration in mg/L. Cumulative gamma distribution The PSDs are modeled gravimetrically using a cumulative gamma distribution (CGD). In order to represent any PSD both a dispersivity parameter and a representative parameter of PM size for the PSD are needed. The dispersivity parameter can be thought of as a characteristic of gradation uniformity or for example as a PM size sorting coefficient (Dickenson and Sansalone 2009). Therefore a series of PSDs with the same representative PM size parameter but differing shapes can be generated. The PM size parameter is a representation of the fineness or coa rseness of a PSD, for example a phi scale ( n) particle size gradation parameter or more commonly a d50m, where m indicates on a gravimetric basis (Dickenson and Sansalone, 2009). Modeling PSDs of differing uniformity and PM size representation is required for influent and effluent urban runoff. Previous studies (Sansalone and Ying 2008; Lin et al. 2009) have utilized a two parameter CGD to model size heterodisperse PSDs. The shape and scale factor parameters in the epresentative coarseness or fineness of the PSD. E quations 23 and 24 represent the gamma distribution ( f(x) ) and CGD ( F(x) ), respectively ) ( ) / ( ) () / ( 1k e x x fx k (2 3)

PAGE 26

26 xdx x f x F0) ( ) ( (2 4) In this expression x represents the particl e diameter, d (m), k represents a shape the gamma function. Particle number density (PND) While PM granulometry is most commonly measured and modeled based on common gravimetric size distributions (P SD and CGD), PM granulometry can also be examined based on particle number distribution (PND), the number of particles of a discrete size increment per a given volume of liquid. PND results are commonly modeled based on the assumption that PND can be rep resented by a truncated hyperbolic function, as a power law model (PLM) across a particle size range with details provided elsewhere (Bader 1970; Cristina et al. 2002; Sansalone and Cristina, 2004). A similar presentation to PSD results can be made for PND, as percent finer by particle number. Following the determination of the cumulative PND, the distribution is modeled using a two parameter PLM. The basic form of the PLM is summarized in the following expression and is plotted in a log log domain to yield a linear representation of PND as a function of PM size across a truncated PSD. ) log( ) log( ) ( / log(p pd d d dN (2 5) In this expression N is the PND, dp is t are physically p versus dp is plotted with the intercept representing and the slope representing As a gravimetric index of granulometry, PSD is dominated by the PM fraction of the gradation with the most mass, commonly the coarser PM, whereas PND as a number based index of granulometry is dominated by the PM fraction of the gradation with the highest PM number, commonly the finest PM.

PAGE 27

27 Kruskal Wallis H test for significant difference When comparing two or more sets of results that are not normally distributed, a useful test for significant difference is the Kruskal Wallis H test. For this test, an H value is calculated and compared to a critical H value from the Chi Square Distribution. An H value is calculated by combining all the samples in the k treatments and ordering them from smallest to largest. Each sample is assigned a rank, with the smallest sample value being assigned first rank and the largest s ample value being assigned nth rank. If any 2 or more data points are equal, each is assigned the average of the ranks. Next, the samples are divided into their original treatment groups and the ranks for each treatment are summed. An H value is then ca lculated using equation 26. ) 1 ( 3 ) 1 ( 122 s s r s s Hi i calc ( 26) In this expression, s is the total number of data points, ri is the rank sum for treatment i, si is the number of samples in treatment i. For the Kruskal Wallis H tests conducted in this study, influent results are separated into two treatments, manual samples and automated samples. Effluent results are separated into four treatments, manual samples from the VCF, automated samples from the VCF, manual samples from the BHS, and automated samples from the BHS For all tests, the null hypothesis (ho) is hypothesis (ha) is that at least two of the treatments being tested a re significantly different (p < = 0.05). When the calculated H value exceeds the critical H value, 2 ho is rejected and the conclusion is that at least two of the treatments are significantly different

PAGE 28

28 Normalized root mean square error Normalized r oot mean square error (NRMSE) is another useful parameter in quantifying the difference between predicted values and measured values. In order to calculate NRMSE, it is necessary to first calculate the sum of squared error (SSE). SSE is calculated using equation 27. ... ) ( ) (2 2 2 2 1 1 x x x x SSEj j ( 27) In this expression, x1j is the jth measurement of treatment 1, 1x is the mean of the measurements of treatment 1, x2j is the jth measurement of treatment 2, 2x is the mean of the measurements of treatment 2. RMSE is then calculated using equation 28 k n SSE RMSE ( 28) In this expression, n is the total number of samples and k is the number of treatments. RMSE is then normalized using equation 29. maxx RMSE NRMSE (2 9) In this expression xmax is the maximum measured value in the manual PSD distribution. Newtons law Settling velocity of a spherical particle in a fluid is calculated using Newtons Law provided in equation 210 p f f p d sd C g V 3 4 ( 210) In this expression, Vs p is the f is the fluid density, Cd is the frictional drag coefficient, and dp is the particle diameter. Cd is calculated using equation 211.

PAGE 29

29 44 0 3 24 d d dR R C ( 211) In this expression, Rd is the particle Reynolds number. Results and Discussion The eight storms monitored in this study cover a wide range of rainfall intensities, runoff dura tions, and flow rates. Event duration ranges from 25 to 78 minutes. Rainfall depth ranges from 2.5 to 23.6 mm. Runoff volume ranges from 283 to 12010 liters. Peak flow rate ranges from 0.5 to 13 L/s and median flow rate ranges from 0.1 to 5 L/s. Event characteristics are shown in Table 21. Measured SSC and PSD results for the eight events are included in Tables 2 2 and 23, respectively. PSD results of the 8 storms are pooled and Kruskal Wallis H tests for significant difference are conducted on each PM size ranging from 0.02 to 2000 m in diameter using the measured values of percent finer by mass for all samples. For the influent samples, there is no significant difference smaller than 159 m in diameter. However, f or all particles larger than or equal to 159 m in diameter, there is a significant difference (p < nual samples, indicating that although automated samplers accurately represent the fine fraction of particles, they fail to representatively capture particles larger than 159 m in diameter. For the effluent samples, there is no significant difference (p manual samples from the same unit operation, but there is a significant difference (p < between the PM separation provided by both the VCF and the BHS This indicates that the automated sampler represen tatively captures particles across the entire PSD when it is size monodisperse and contains mostly fine particles. The significant differences between automated and manual PSDs for the 14 August event are illustrated in Figure 2 2. The influent and effluent

PAGE 30

30 uncertainty bars in Figure 22 represent the standard deviation between all of the influent and effluent samples taken across the event. In order to validate the manual sampling results and give meaning to the significant difference tests discussed abo ve, the mass balance for PM of the BHS is determined using manual sampling results for influent and effluent as well as manual recovery of all PM separated by the BHS. The mass balance recovery is 95.1% with a mass balance error of 4.9% across the entire monitoring campaign. Therefore, results of manual sampling are taken as the representative influent and effluent values to which automated sampling results are compared. A statistically significant difference between results of the two sampling methods th erefore indicates that results of automated sampling are significantly different from the actual representative PM results. N RMSE is calculated for automated and manual PSDs. This is done for the diffe rence in mass of each particle size in the PSD for automated and manual samples of all monitored events. NRMSE for i nfluent PSD ranges from 0.11 to 0.30 with a volume weighted mean of 0.13. NRMSE for effluent PSD from the BHS ranges from 0.09 to 0.60 with a volume weighted mean of 0.50. NRMSE for efflu ent PSD from the VCF ranges from 0.09 to 0.99 with a volume weighted mean of 0.28. NRMSEs are tabulated in Table 24. NRMSEs are relatively large for both influent and effluent PSDs, which indicates significant error between manual and automated PSDs. T hese results are consistent with those of the Kruskal Wallis H test discussed previously. In addition to NRMSE, PSD is modeled using a CGD. For this distribution, the PSD for each sample is modeled an CGD, the shape parameter represents the uniformity of the PM size gradation and the scale parameter represents the coarseness of the PM. Influent k for manual samples ranges from 0.61

PAGE 31

31 to 1.09 with a volume weighted mean of 1.01. Influent k for automated samples ranges from 0.75 to 1.12 with a volume weighted mean of 1.08. The higher volume weighted k value for influent automated samples indicates that the distribution provided by auto mated samples is more uniform than that of manual samples and therefore contains a narrower range of particle sizes. This means that in failing to capture the large particles in a sample, the automated sampler is also failing to accurately assess the unif ormity of particle sizes. Effluent k for manual samples from the BHS ranges from 0.86 to 1.25 with a volume weighted mean of 1.21. Effluent k for automated samples from the BHS ranges from 0. 83 to 1.15 with a volume weighted mean of 1.12. Effluent k for manual samples from the VCF ranges from 0.66 to 1.3 5 with a volume weighted mean of 1.25. Effluent k for automated samples from the VCF ranges from 0. 55 to 1.34 with a volume weighted mean of 0.99. Effluent automated and manual sample k values cover the same range of values and therefore do not appear to be different. This is because there are few to no coarse particles in the treated effluent of either unit and the automated sampler r manual samples ranges from 99.4 to 390 with a volume weighted mean of 284 146 to 268 with a volume weighted mean of 229. The lower volume weighted automated samples indicates that the overall coarseness of particles is underrepresented by the from the BHS ranges from 58.4 to 88.9 with a volume weighted mean of 61.2 from the BHS ranges from 83.8 to 133.5 with a volume weighted mean of 92.8. from the VCF ranges from 4.1 to 25.4 with a volume weighted mean of 18.6 samples from the VCF ranges from 5.0 to 36.8 with a volume weighted mean of 28.6. As with the effluent automated and manual samples cover the same range of

PAGE 32

32 values and therefore do not appear to be significantly different. CGD parameters for all events are shown in Table 25. From these CGD results, Kruskal Wallis H tests for significant difference are conducted on < is size heter odisperse and contains coarse particles, and the automated sampler fails to accurately represent both the uniformity and coarseness of the PSD. For the effluent samples, there is no between any of the f our treatments. This is because there are few coarse particles in the effluent and the auto mated sampler is able to accurately represent the uniformity and coarseness of the PSD. The significant differences between automated and manual k s are illustrated in Figure 2 3. In order to evaluate the role of the automated sampler intake tubes inner diameter in the significant difference between automated and manual samples at larger particle sizes, settling velocity is calculated for spherical particles ranging from diameter 0.02 to 2000 m and compared to the uptake velocity of the automated sampler used in this study Particles are assumed to be spherical in shape and a median particle density of 2.45 g/cm3 is applied from previously measured data. In addition, the uptake velocity is calculated using an average measured flow rate and a 1 cm inner tubing diameter. The uptake velocity of the automated sampler tubing is 468 mm/s. The settling velocity of a spherical particle with diameter 159 m is approximately 26 mm/s. The uptake velocity of the automated sampler is an order of magnitude higher than the settling velocity of a 159m particle but the sampler still fails to repeatedly capture particles greater than or equal to this diameter. This demonstrates that the

PAGE 33

33 significant difference between automated and manual samples containing particles at or above 159 m in diameter cannot be attributed to intake tube selection alone. In addition to PSD, PND is an important granulometric index of PM. Particle number indices d10, d50, and d90 (by number) for the eight events monitored in this study are shown in Table 2 6. From these results, Kruskal Wallis H tests for significant difference are conducted on each particle size ranging from 0.02 to 2000 m in diameter using the measured values of percent finer by number for all samples. For both influent and effluent PNDs, there is no significant difference number of particles of each particle size between 0.02 and 2000 m in diameter. This is because although coarse particles contain more mass than fine particles and dominate the PSD, fine particles are more plentiful than coarse particles and dominate the PND. Therefore, si nce the automated sampler accurately represents the fine fraction, it also accurately represents the PND. The PND for each sample is modeled using a cumulative PLM. The PLM is applied only to particles ranging from 0.02 to 75 m in diameter (fine fraction ). This range is chosen because the fine fraction dominates the PND and for all storms, at least 99.98% of all particles by number are finer than 75 m. For the PLM, the PND for each sample is linearly regressed with a slope of and a y intercept of The slope represents the dominant particle fraction, where a steeper slope indicates a preference for large particles and a flatter slope indicates a preference for small particles. The y intercept represents the variability of particle concentration. For the comparisons between automated and manual samples for both influent and effluent, all of the PLM parameters are fairly consistent and generally have the same range of values for t he two sampling methods. PLM parameter results for all events are s hown in Table 27.

PAGE 34

34 From these PLM results, Kruskal Wallis H tests for significant difference are conducted on parameters for all samples. For the influent samples, there is no significant difference (p This indicates that the automated sampler accurately represents the total number of particles in a sample and the dominance of fine particles in that count. For the effluent samples, there is also no significant 0.05) between automated and manual samples from the same unit operation PLM parameters are illustrated in Figure 2 4. A Note about the Relationship between PND and Turbidity Turbidity is a measure of the scattering of li ght by suspended particles (<75 m). It significantly impacts the filtration and disinfection of surface waters. Therefore, the turbidity of a water sample is an important water quality characteristic to determine. Since both turbidity and PND are heavi ly influenced by fine particles, the measured values of percent finer by number for a variety of particle sizes between 0.02 and 75 m for a set of samples are plotted with paired turbidity measurements, taken with a turbidimeter. Linear regressions for t urbidity as a function of PND, in the form of percent finer by number, showed that the highest correlation between turbidity and PND occurs at a nominal 1m particle size. This indicates that not only is 1 m the average d50 by number, but it is also the particle size that is most highly correlated to turbidity. The relationship between turbidity and the percent of particles finer than 1 m by number is illustrated in Figure 2 5. Additionally, the coefficient of determination, r2, between percent finer by number and turbidity increases with increasing particle size from 0.02 to 1 m, peaks at about 1 m, and then decreases with increasing particle size until it levels off from 1 to 75 m. This is also shown in Figure 25.

PAGE 35

35 PND and PSD both play an integr al part in characterizing rainfall runoff. The PSD provides information about how the mass of PM is distributed which is important for the fate and transport of nutrients and metals. The PND provides information about how the number of particles is distr ibuted which is important for the quantification of cyst organisms. Sediment size particles represent a large amount of the total mass and therefore tend to largely affect the PSD. However, they do not represent a large number of the total count and ther efore do not tend to largely affect the PND. Fine particles represent a large number of the total count and therefore tend to largely affect PND. However, they do not represent a large amount of the total mass and therefore do not tend to largely affect the PSD. Hence, it is expected that PSD is more affected by automated sampling than PND, since automated sampling mainly impacts the collection of larger particles. This is illustrated in Figure 2 6, in which manual and automated influent PSDs and PNDs f or the 13 August 2010 event and the 23 August 2010 event are plotted together. For the 13 August 2010 event, the manual influent d50 by mass is 182 m and the automated influent d50 by mass is 117 m. However, the manual influent d50 by number is 0.9 m and the automated influent d50 by number is 0.5 m. For the 23 August 2010 event, the manual influent d50 by mass is 190 m and the automated influent d50 by mass is 120 m. However, the manual influent d50 by number is 0.9 m and the automated influent d50 by number is 0.7 m. There are significant differences (p < between samples taken manually and samples taken with an automated sampler. The finest particles have the smallest difference in percent finer by mass, but as particle size increases, so does this difference. A turning point occurs in the size gradation at 159m, when this difference becomes statistically significantly different. Figure 2 7 illustrates the increasing difference between the critical H value2 and the calculated H value of the Kruskal Wallis H test for PSD with increasing par ticle size. These

PAGE 36

36 results indicate that the critical particle diameter at which there is a significant difference between manual sampling and automated sampling is 159 m. This implies that for any size monodisperse treated effluent, like that of the VCF with a d50 below 159 m, there is no significant difference (p > between paired automated and manual samples. However, for any size heterodisperse influent, like that of typical urban rainfall runoff there is a significant difference (p < between paired automated and manual samples. Conclusions This study examines manual and automated sampling results generated through paired sampling and analysis for PSD, PND, SSC and PM fractions. Results are based on eight runoff events captured from an urban source area in Gainesville, Florida. Whole sample fra ctionation is utilized for SSC and PM fractions, and laser diffraction for PSD and PND. Influent and effluent sampling for four events is examined for a VCF (primary sedimentation and filtration) and the additional four events for a BHS providing only pri mary sedimentation. Non parametric testing is utilized to compare sampling methods for influent and effluent. Based on this framework of testing and analysis results illustrate that there are statistically significant differences between event based manua l grab sampling and automated sampling for influent SSC and PSD While there is no significant difference between the two PND characterizations for size heterodisperse runoff with a d50 (by mass) at or above 159 m, there is a significant difference betwe en the two PSD characterizations. This indicates that manual grab sampling paired with the SSC method and PSD characterization is a more accurate PM analysis for most untreated urban rainfall runoff from small watersheds. However, for runoff containing mostly fine particles, this is not the case. For size monodisperse runoff with a d50 (by mass) below 159 m, there is no significant difference between the two PM analyses. This indicates that either technique, if used appropriately,

PAGE 37

37 provides a representat ive PM characterization of treated effluent from a unit implementing primary sedimentation as well as filtration. A preliminary PM analysis is required in this case to determine if the effluent is truly size monodisperse and composed of fine particles sma ller than 159 m. There are two main constraints of automated samplers that are responsible for the difficulties associated with stormwater monitoring. The first is that they do not sample the entire depth of flow. The location of the intake tube within the depth of flow greatly impacts sampling, since higher concentrations of particles are found at lower depths The second is that the uptake velocity of the automated sampler is usually not equal to the localized streamflow velocity. Runoff flow rates a re not constant; however intake tube uptake velocities are fixed. This leads to mostly non isokinetic sampling and substantial error due to inertial effects of the particles.

PAGE 38

38 Table 2 1. Monitored rainfall runoff event characteristics The line s eparating the first set of events and the second set of events represents the breakpoint between the BHS events (top) and the V C F events (bottom). Event D ate (2010) PDH (hr) Event duration (min) Rainfall depth ( mm ) Runoff volume (L) Peak flow (L/s) Median flow (L/s) Mean flow (L/s) 31 July 72 60 23.6 9019 13.26 0.60 2.50 13 August 144 25 2.79 309 3.21 0.06 0.21 14 August 28 31 2. 54 594 2.47 0.25 0.38 21 August 168 31 2.79 283 1.46 0.03 0.16 1 August 24 36 30.0 12007 13.26 4.74 5.49 7 August 24 48 8.64 2757 13.26 0.23 0.73 23 August 48 42 2.79 312 1.25 0.01 0.12 26 September 40 78 3.56 1129 0.45 0.26 0.24

PAGE 39

39 Table 2 2. Summary of event based PM fractions Influent and effluent refer to untreated and treated rainfall runoff, respectively. T he line separating the first set of events and the second set of events represents the breakpoint between the BHS events (top) and the V C F events (bottom). Event Date (2010) PM < 75 m [mg/L] PM < 1000 m [mg/L] PM < 2000 m [mg/L] PM > 2000 m [mg/L] SSC [mg/L] Manual Auto Manual Auto Manual Auto Manual Auto Manual Auto 31July Influent 9.2 14.5 61.0 93.9 81.0 121.6 11.6 7.0 92.7 128.6 Effluent 9.7 14.6 38.8 67.9 47.0 83.9 0.0 0.0 47.0 83.9 13Aug ust Influent 33.3 52.9 190.6 311.6 258.2 408.9 356.2 61.3 614.4 470.2 Effluent 4.0 4.0 12.8 14.8 15.3 17.8 0.0 0.0 15.3 17.8 14August Influent 24.2 19.0 198.6 102.1 282.7 131.0 52.4 10.3 335.1 141.3 Effluent 4.1 4.1 14.4 14.9 17.3 18.1 0.0 0.0 17.3 1 8.1 21August Influent 69.2 84.5 368.2 414.5 477.6 539.9 57.3 77.4 534.9 617.3 Effluent 3.3 3.3 10.0 13.9 12.0 17.0 0.0 0.0 12.0 17.0 1August Influent 32.4 23.9 183.6 129.7 245.1 168.6 0.0 0.0 245.1 168.7 Effluent 2.8 3.5 6.6 8.6 7.6 10 .0 0.0 0.0 7.6 10.0 7August Influent 8.7 29.1 60.7 185.7 82.7 242.6 34.4 27.1 117.1 269.7 Effluent 6.5 6.2 12.4 11.8 13.9 13.2 0.0 0.0 13.9 13.2 23August Influent 24.7 40.2 155.1 224.5 210.5 292.3 345.3 37.6 555.8 329.9 Effluent 1.9 8.5 4.1 15.1 4. 7 16.9 0.0 0.0 4.7 16.9 26September Influent 17.4 9.1 89.9 36.1 114.4 44.4 3.5 0.0 117.9 44.4 Effluent 1.3 5.2 2.6 9.9 2.9 11.2 2.1 0.0 5.0 11.2

PAGE 40

40 Table 2 3. Summary of event based PM indices of d10, d50, and d90. Influent and effluent refer to untreated and treated rainfall runoff, respectively. T he line separating the first set of events and the second set of events represents the breakpoint between the BHS events (top) and the V C F events (bottom). Event Date (2010) PSD (m) d 10 d 50 d 90 Manual Auto Manual Auto Manual Auto 31July Influent 23.0 16.3 150.1 110.1 442.8 180.1 Effluent 6.5 8.8 40.4 69.4 124.3 205.3 13August Influent 16.3 16.6 181.5 117.2 764.5 407.2 Effluent 4.4 5.7 31.9 41.3 145.5 174.6 14August Inf luent 24.5 12.3 177.4 101.2 727.4 409.8 Effluent 5.7 6.0 46.0 46.3 233.2 199.7 21August Influent 5.9 8.7 80.4 88.5 552.7 300.9 Effluent 4.6 8.6 33.5 77.7 224.3 203.6 1August Influent 23.9 26.1 199.2 185.7 531.0 523.9 Effluent 1.5 1.5 5.5 8.3 17.1 39.3 7August Influent 18.6 19.3 186.2 147.7 737.3 407.2 Effluent 0.9 0.9 3.5 3.7 11.8 13.2 23August Influent 13.7 13.5 190.4 119.7 713.7 464.3 Effluent 1.5 1.4 3.9 2.4 40.4 4.3 26September Influent 3.6 4.7 34.6 76.9 173.2 230.6 Efflue nt 1.0 1.1 3.1 4.3 55.5 13.3

PAGE 41

41 Table 2 4. NRMSE between the incremental mass of automated samples and manual samples at each particle size across the PSD Influent and effluent refer to untreated and treated rainfall runoff, respectively. T he line separating the firs t set of events and the second set of events represents the breakpoint between the BHS events (top) and the V C F events (bottom). Event Date (2010) PSD NRMSE Influent Effluent 31 July 0.135 0.534 13 August 0.194 0.093 14 August 0.301 0.091 21 August 0.300 0.599 1 August 0.113 0.286 7 August 0.114 0.091 23 August 0.274 0.990 26 September 0.191 0.498

PAGE 42

42 Table 2 5. Summary of event based cumulative gamma distribution parameters for PSD. k Shape factor of gamma distribution function Scale factor of gamma distribution function Influent and effluent refer to untreated and treated rainfall runoff, respectively. T he line separating the first set of events and the second set of events represents the breakpoint between the BHS events (to p) and the V C F events (bottom). Event Date (2010) k Manual Auto Manual Auto 31July Influent 1.08 1.12 189.93 220.10 Effluent 1.25 1.15 58.42 91.08 13August Influent 1.09 0.84 389.68 267.79 Effluent 0.86 0.83 82.67 133.49 14August Influent 0.80 0.85 353.70 201.76 Effluent 0.91 0.90 88.94 83. 76 21August Influent 0.61 0.75 385.93 267.03 Effluent 0.91 0.94 66.43 120.79 1August Influent 1.05 1.12 337.71 252.97 Effluent 1.30 0.93 22.04 36.77 7August Influent 0.79 1.06 384.97 182.18 Effluent 1.35 1.14 4.11 5.55 23August Influen t 0.78 0.87 365.21 244.67 Effluent 0.83 0.55 25.44 5.01 26September Influent 0.86 0.77 99.40 145.77 Effluent 0.66 1.34 15.79 4.98

PAGE 43

43 Table 2 6. Summary of event based particle number density (PND) indices Influent and effluent refer to untreated and treated rainfall runoff, respectively. T he line separating the first set of events and the second set of events represents the bre akpoint between the BHS events (top) and the V C F events (bottom). Event Date (2010) PND (m) d 10 d 50 d 90 Manual Auto Manual Auto Manual Auto 31July Influent 1.4 1.5 2.0 2.1 4.3 4.3 Efflue nt 1.4 0.9 2.0 1.4 4.1 3.1 13August Influent 0.5 0.4 0.9 0.8 1.9 1.8 Effluent 0.4 0.4 0.5 0.5 1.1 1.1 14August Influent 0.8 0.5 1.2 0.8 2.6 1.9 Effluent 0.4 0.4 0.5 0.5 1.1 1.1 21August Influent 0.4 0.4 0.5 0.6 1.2 1.3 Effluent 0.3 0.3 0.5 0.5 1.0 1.0 1August Influent 1.0 1.1 1.8 1.8 3.6 3.6 Effluent 1.0 1.0 1.2 1.2 2.5 2.4 7August Influent 0.5 0.5 0.9 0.9 2.1 1.7 Effluent 0.3 0.3 0.5 0.5 1.0 1.0 23August Influent 0.5 0.4 0.9 0.7 2.0 1.6 Effluent 0.8 0.9 1.3 1.5 2.2 2.5 26September Influent 0.4 0.3 0.6 0.5 1.3 1.1 Effluent 0.5 0.4 0.9 0.7 1.5 1.5

PAGE 44

44 Table 2 7. Summary of event based power law model parameters for PND coefficient of the power law model (PLM), physical index for number of particles exponent of PLM, physical index for slope on log log plot and size distribution of PM Influent and effluent refer to untreated and treated rainfall runoff, respectively. T he line separating the first set of events and the second set of events represents the bre akpoint between the BHS events (top) and the V C F events (bottom). Event Date (2010) Manual Auto Manual Auto 31July Influent 4.44x10 7 5.40x10 7 2.03 2.09 Effluent 6.89x10 7 4.79x10 7 2.38 2.12 13August Influent 2.25x10 8 1.53x10 7 2.40 2.13 Effluent 4.61x10 7 5.79x10 7 2.37 2.37 14August Influent 6.53x10 7 4.37x10 7 2.09 2.31 Effluent 4.97x10 7 3.33x10 7 2.35 2.30 21August Influent 2.84x10 7 7.52x10 7 2.47 2. 29 Effluent 3.72x10 7 2.70x10 7 2.45 2.25 1August Influent 3.56x10 6 1.15x10 8 2.18 2.17 Effluent 1.52x10 8 8.91x10 7 3.45 2.92 7August Influent 5.16x10 7 2.38x10 8 2.31 2.41 Effluent 8.63x10 8 4.65x10 8 3.89 3.67 23August Influent 2.84x10 8 1.12x1 0 8 2.26 2.31 Effluent 8.13x10 7 3.87x10 8 3.08 4.13 26September Influent 8.51x10 7 9.57x10 7 2.49 2.50 Effluent 1.26x10 8 3.94x10 8 3.26 3.71

PAGE 45

45 Figure 2 1. Elevation view of monitoring system design VCF

PAGE 46

46 Particle diameter ( m) 1 10 100 1000 Percent finer by mass (%) 0 20 40 60 80 100 Manual Influent Auto Influent Manual d50 = 177.5 m Auto d50 = 101.2 m 14 August 2010 Figure 2 2. E xample of manual and automated influent PSD (14 August 2010 rainfall runoff event) Influent and effluent uncertainty bars represent the standard deviation in PSDs between all of the influent and effluent samples taken during a given event, respectively.

PAGE 47

47 k 0.0 0.4 0.8 1.2 1.6 2.0 2.4 5th percentile 10th percentile 25th percentile 50th percentile 75th percentile 90th percentile 95th percentile 0 x 1 10 100 1000 Manual Influent Auto Influent Manual VCF Effluent Auto VCF Effluent Manual BHS Effluent Auto BHS Effluent Figure 2 3. Manual and automated influent and effluent gamma distribution parameters

PAGE 48

48 0 1 2 3 4 5 6 1e+6 1e+7 1e+8 1e+9 1e+10 Manual Influent Auto Influent Manual VCF Effluent Auto HDS Effluent Auto VCF Effluent Manual HDS Effluent Figure 2 4. Manual and automated influent and effluent power law parameters for particles less than 75 m

PAGE 49

49 Percent Finer By Number 0 20 40 60 80 100 Turbidity (NTU) 0 20 40 60 80 100 August 7, 2010 August 13, 2010 August 14, 2010 Particle Size (m) 1 10 r-squared 0.0 0.1 0.2 0.3 0.4 0.5 0.6 August 14, 2010 Figure 2 5. Relationship between turbidity and PND. The coefficient of determination between turbidity and percent finer by number is shown as a function of particle size (d = 0.02 25 m ).

PAGE 50

50 PSD: Percent finer by mass (%) 0 20 40 60 80 100 0 20 40 60 80 100 Manual Influent PSD Auto Influent PSD Manual Influent PND Auto Influent PND 13 August 2010 runoff eventPND: Percent finer by number (%) Particle diameter ( m) 1 10 100 1000 PND: Percent finer by number (%) 0 20 40 60 80 100 PSD: Percent finer by mass (%) 0 20 40 60 80 100 23 August 2010 runoff event Figure 2 6. Examples of manual and automate d influent PSD vs. PND (13 August 2010 and 23 August 2010 rainfall runoff events). PSD method results are significantly different (p <

PAGE 51

51 Particle Diameter (m) 0.01 0.1 1 10 100 1000 H-Test Differential -8 -6 -4 -2 0 2 4 6 8 10 159 m Fi gure 2 7. Kruskal Wallis H test results for PSD. The difference between the critical H value and the measured H value increases with increasing particle size from 0.02 to 2000 m. At a particle diameter of 159 m the difference between the two values becomes so great that there is a statistically significant difference (p < automated and manual samples. For all particle diameters greater than or equal to 159 m, there is a significant difference between the two sampling techniques.

PAGE 52

52 CHAPTER 3 PARTICULATE MATTER A ND PHOSPHORUS REMOVA L IN RAINFALL RU NOFF TREATED WITH VOLUMETRIC FILTRATION Introduction Rainfall runoff from impervious surfaces and urban land use carries particulate matter (PM), nutrients (Marsalek, 1999; Vaze and Chiew, 2003), heavy metals (Liebens, 2001; Sansalone and Cristina, 2004; L au and Stenstrom, 2005), and other constituents (Van Dolah et al., 2005; Khan et al., 2006) to lakes, streams, and rivers. Studies have identified nonpoint source (NPS) pollution from urban rainfall runoff as the leading cause of impairment across all ty pes of waterbodies (EPA, 2000; Lee and Bang, 2000; Gnecco et al., 2005). This is partly due to the ecological impacts of particles on receiving waters and partly because many constituents of concern, including metals and nutrients, partition into the PM a nd are transported to surface waters in urban rainfall runoff (Legret and Pagotto, 1999; Sansalone and Kim, 2008; Liu et al., 2010). An excessive concentration of phosphorus is believed to be the leading cause of eutrophication in freshwaters (Corell, 1998). In order to reduce the adverse effects of rainfall runoff on surface waters, the water resources community has been forced to undertake the challenging task of monitoring stormwater quality (Harmel et al., 2003). The Role of Urban Development The PM found in rainfall runoff originates from many sources in the urban environment (Sansalone and Kim, 2008). Infrastructure such as roads, sidewalks, and residential structures that are added during land development and urbanization significantly increase the runoff volumes and mass loads discharged to receiving waters (Glen et al., 2004; Gobel et al. 2007). Urban development decreases the amount of permeable surface in a watershed (Roesner, 1999). Additionally, with urbanization comes increased population density leading to contributions of dirt, contaminants from vehicles, debris from streets, and publicly used chemicals to urban

PAGE 53

53 rainfall runoff from these impervious surfaces (Sartor and Boyd, 1972; Grottker, 1987; Pratt et al., 1995). Activities such as transportation and urban infrastructure maintenance generate significant PM ranging in size anywhere from less than 1 m to greater than 10,000 m (Sansalone and Kim, 2008). A study by Sartor and Boyd showed that concentrations of suspended solids, nutrie nts, and other constituents are much higher in urban rainfall runoff than in runoff from rural areas (1972). This impact greatly depends on traffic density, wind drift, duration and intensity of rainfall events, and duration of dry weather periods between events (Gabel et al., 2007). History of Stormwater Regulations Stormwater runoff (rainfall runoff and snowmelt) was first recognized as a potential source of surface water impairment in the 1960s when government agencies first began to research the issue At this time it was determined that stormwater runoff was responsible for almost half of all pollutant load discharged to surface waters (Pitt et al., 1999). In response to growing concern, the Clean Water Act (CWA) was amended in 1987 to prohibit the discharge of pollutants from stormwater runoff to receiving waters unless it was in compliance with a National Pollutant Discharge Elimination System (NPDES) permit. This included industrial and construction stormwater discharges along with municipal separ ate storm sewer systems (MS4s). In 1990, federal regulations (Phase I) were enacted to guide municipalities in preparing NPDES permit applications. In order to comply, municipalities had to collect existing information about the quantity and quality of st ormwater discharge, receiving waters, management programs, and financial resources. Then, they had to create a program to reduce NPS pollution to the maximum achievable extent. Phase I applied to all medium and large

PAGE 54

54 MS4s, cities with populations greater than 100,000, operators of construction activities affecting at least 5 acres of land, and certain industrial activities (EPA, 2000). In 1999, the final rule (Phase II) was published by the Environmental Protection Agency (EPA). None of the permit requi rements changed in Phase II, but they now additionally applied to all small MS4s located within an urban area with a population of at least 50,000 and a population density of at least 1,000 people per square mile along with operators of construction activi ties affecting at least 1 acre of land. Phase II permit coverage became mandatory in 2003 (EPA, 2000). Current Research In response to the NPDES amendments of the CWA, unit operations are being explored for the separation of PM in stormwater treatment (Nix et al., 1988; James, 1999). In many urban areas where there is limited area and high land cost, combined unit operations are preferred to extended treatment processes and large settling basins (Grizzard et al., 1986). Preliminary unit operations such a s hydrodynamic separators were initially favored for their small footprint (Wong et al., 1995; Walker et al., 1999). However, since hydrodynamic separator s primarily remove gross solids and require frequent maintenance, over the last decade filtration uni t operations have become more common for PM separation in runoff (Liu et al., 2010). Such filtration unit operations can be implemented in a variety of configurations, including replaceable radial cartridge filters (Pathapati and Sansalone, 2009) and axial filters. The selection and design of filtration unit operations depends on knowledge of PM load and granulometry (Greb and Bannerman, 1997; Cristina et al., 2002; Sansalone and Kim, 2008), unit filter run volume, headloss development, and desired efflu ent water indices such as suspended sediment concentration (SSC), total suspended solids (TSS), turbidity, and particle size distribution (PSD) (Clark, 2000). It is therefore critical, when assessing the performance of a stormwater treatment unit, to

PAGE 55

55 quantify and qualify the inputs to the unit and the outputs to receiving waters under a variety of event durations, flow intensities, and loads (Liu et al., 2010). Objectives The primary objective of this study is to evaluate the removal of particulate matter using an in situ volumetric clarifying filter ( VCF ) treatment unit. Variable flow rates, storm durations, and pollutant loads are tested and subsequent event mean values (EMVs) are calculated in order to evaluate the performance of a VCF during actual rai nfall runoff events. Performance, in terms of total mass load reduction and effluent particle size distribution, is evaluated for each of the two main mechanisms of removal in the unit, sedimentation and filtration. An additional objective of this study is to evaluate the removal of dissolved and particulate bound phosphorus using the same treatment device. Methodology Watershed Configuration The Reitz Union paved surface parking facility located on the University of Florida campus serves as the watershed in this study. Rainfall runoff is collected from an existing stormdrain inlet at the southwest parking lot pavement edge. The watershed is divided into two main subcatchments, a parking area on the southeast side and an access road on the northwest sid e. Due to the parking lots low slope in the direction of drainage, the contributing watershed area varies with rainfall intensity and wind direction. Depending on the storm event, the watershed area varies from 400 to 600 square meters with a median area of approximately 500 square meters. Volumetric Clarifying F ilter and Monitoring System This study utilizes a volumetric clarifying filter ( VCF ) for runoff treatment as shown in Figure 3 1. The VCF unit is 1.2 meters in diameter and is configured with two standard

PAGE 56

56 cartridges, each with a 70 millimeter orifice cartridge lid, and one draindown cartridge with a 35millimeter orifice cartridge lid. Each cartridge contains eleven 20 micrometer filter tentacles made up of two 69centimeter long segments. The s ystem components include a 46 centimeter diameter maintenance access pipe, a pressure relief pipe, a cartridge deck, a backwash pool weir located around the two standard cartridges, a separator skirt, and a settling zone. The unit has a m aximum t reatment f low r ate (MTFR) of 12.6 liters per second and is configured to begin bypassing flow when head loss reaches 46 centimeters. This is equivalent to a water elevation of 46 centimeters above the cartridge deck level. The monitoring station design i s driven by the monitoring procedures related to the physical chemical processes being investigated ; PM and phosphorus removal. Rainfall runoff chemistry is monitored across the treatment system during rainfall events. Runoff is transported to the treatment syst em after collection by catch basin Inlet A as shown in Figure 21. The cylindrical manhole is characterized by a 9 inch high inlet at the top and a 6 inch diameter outlet pipe at the bottom. A butterfly valve is used at the inlet to prevent unwanted flow from entering the system. A 15 centimeter diameter PVC pipe carries runoff at a 6 percent slope from the inlet to a Parshall flume located 10 meters downstream. A 61 cm x 46 cm x 46 cm influent sampling drop box is located immediately downstream of the Parshall flume. A 15centimeter diameter PVC teesection carries the runoff from the outfall of the drop box to the treatment unit. Diversion gate valves are used on each side of the tee to direct all of the influent runoff either to the VCF or to a BHS that is installed in parallel for a separate study. A 61 cm x 46 cm x 46 cm effluent sampling drop box is located at the outfall. Additionally there is a bypass pipe network for the VCF It consists of a 15 centimeter diameter PVC pipe elevated to 122 centimeters above the units internal cartridge deck level,

PAGE 57

57 with a 5 centimeter diameter siphon break 30 centimeters above that. In the event that bypass occurs during an event treated by the VCF untreated bypass effluent and treated effluent are sampled either as a composite or as two individual streams in the effluent sampling drop box. None of the events monitored for this study generated bypass. In order to investigate the head loss caused by the 20micrometer filter tentacles during rainfall runoff ev ents, continuous head measurements are taken both inside and outside of the filter cartridges. Two 1 psi pressure transducers (model PDCR 1830 1 psig) are used for these measurements (DRUCK Inc.). Measured differences in water depth from the two sensors provided the head loss caused by the filters. Additionally, bypass flow rate is measured with a 1psi pressure transducer located in the bypass pipe (model PDCR 1830 1 psig). The CR1000 model data logger manufactured by Campbell Scientific Inc. is used f or the transducers. In order to measure flow depth, the Parshall flume is equipped with a Shuttle Level Transmitter model ultrasonic s ensor which i s calibrated at the site to measure water depth. An MJK 4 20mA model data logger is used for the sensor. The site is equipped with a tipping bucket rain gauge manufactured by ISCO Inc. (0.01 inch bucket capacity). It is located on the roof of the Stormwater Unit Operations and Processes (UOP) laboratory, which is about 180 meters southwest of the watershed. Additionally, two rain gauges located at the University of Floridas Physics Department (west side of catchment) and at the University of Floridas Dental Tower (southeast side of catchment) provide a validation of rainfall measurements. The roof of the UOP laboratory has also been equipped for rainfall chemistry monitoring. In addition to the manually sampled rainfall runoff, rainfall is collected on the roof prior to

PAGE 58

58 reaching the ground. Four DI rinsed Pyrex dishes are used for rainfall collection during the events. Calibration Procedures Calibration of MJK ultrasonic sensor for flow depth measurements A Shuttle Ultrasonic sensor manufactured by MJK is mounted 76 cm above the invert surface of the Parshall flume. The sensor has an accuracy of + 0.10 cm The sensor is connected to a logic unit equipped with a display that converts the voltage response to water depth once the sensor has been properly installed and leveled and the system has been calibrated. The initial setup consisted of selecting the out put unit (inches) and adjusting the zero level to an absolute reference value. For this the invert surface of the flume is cleaned and dried and a sensor measurement is taken. This value is then set as the control reading. Then a series of discrete de pths are generated for the expected range of flow depths. Measurements are recorded and a calibration curve is developed. Calibration of pressure transducers The three 1 psi pressure transducers (DRUCK Inc.) are connected to the CR1000 data logger (Campbe ll Scientific Inc.). During the calibration phase the real time data are acquired and monitored by a field laptop connected to the data logger. Pressure data are reported as mV and converted to water depth (feet) through a calibration curve unique to eac h pressure transducer. The calibration phase consisted of placing each transducer in a water column under static conditions. Various water depths are measured manually and the corresponding voltage measurements are recorded. A linear regression between voltage measures and water depth provided the multiplier and offset parameters. These parameters are then input into the sensors data acquisition program.

PAGE 59

59 Preliminary hydraulic testing Hydraulic testing is performed to quantify the cleanbed (time 0) hea d loss as a function of steady flow across the VCF Clean bed, potable water (TSS < 0.1 mg/L) head loss measurements are taken by two, 1psi pressure transducers at steady increments of flow rates. The VCF s internal deck is used as the datum for all head measurements. During the testing, the VCF is configured to have two standard cartridges with 70 mm orifice lids and one draindown cartridge with a 35 mm orifice lid. This is the same configuration that is used for actual rainfallrunoff events. Water i s pumped through the system at 20 gpm increments starting at 40 gpm, peaking at 260 gpm, and returning to 40 gpm. Flow rate is regulated with a flow meter. Monitoring Methodology A threshold rainfall depth of 2.5 mm and a minimum dry period of 6 hours are selected for rainfall runoff events. Prior to an event, radar tools such as Wunderground ( www.wunderground.com ) and National Weather Service (NWS) are utilized to determine if an approaching rainfall event is o f sufficient intensity, spatial coverage, and trajectory. Although the monitoring station is permanently assembled, some setup prior to an event is required. The YSI Multi Parameter sensor, the MJK Ultrasonic level sensor, and the pressure transducers ar e connected to the CR1000 data logger. Two synchronized stop watches are used for sampling, one each for influent and effluent. Runoff samples are taken manually at two sampling points, one upstream of the treatment system and one downstream of the treatm ent system, at regular intervals in the range of two to five minutes. Frequency is determined based on radar and rainfall intensity/flow inspections. Manual samples are taken across the full cross section of flow, with mathematical compositing based on m easured flow volume and time.

PAGE 60

60 For both influent and effluent sampling, one set of 1.1L replicate samples and two sets of 0.5L replicate samples are taken. One of the sets of 0.5 L samples is used for a particle size distribution (PSD) analysis. The other set is used to determine SSC. Corresponding PSD and SSC results are paired so that the PSD results (volume %) are converted to a gravimetric concentration (mg/L). The set of 1.1 L samples is used to determine turbidity, sediment PM, settleable PM, TSS and dissolved and particulate bound phosphorus. Assessing PM Loads and Transport in Urban Rainfall Runoff At the end of a monitored event, samples are taken back to the laboratory for analysis. The rainfall is also collected from the Pyrex troughs for immediate analysis. The first analysis in the lab uses laser diffraction on a set of 0.5 L samples to determine PSD. This is done for runoff samples only. The laser diffraction particle analyzer (Malvern Instruments: Hydro 2000G) is operated in batch mode. Prior to being poured into the analyzer, each sample is prescreened with a #4 2000m sieve. This is done so that PM or other debris larger than 2000 m is not allowed to enter the laser chamber. Once samples are added to the wet chamber of the analy zer, multiple repetitions (3 9) are required until the entire gradation reaches stability. A paired SSC analysis is conducted within the first 24 hours following an event. This is done for runoff samples only. For the SSC analysis, t he other set of 0.5L samples is filtered through a 1m glass fiber filter using a vacuum pump. Filters are washed, dried, and weighed prior to use. The sample volumes are recorded and filters are dried in an oven at 105C. Final filter weights are then recorded. The diff erence between final and initial filter weights provides the mass of PM, which can then be converted to a concentration (as SSC) with sample volume. The set of 1.1L samples is analyzed in parallel with the PSD and SSC analyses. These samples are first us ed for turbidity measurements. Then, samples are screened through a #200 75m sieve in order to separate and collect sediment size particles. The sample volume is

PAGE 61

61 measured and recorded prior to sieving. The filtrate is then poured into an Imhoff cone a nd allowed to settle for one hour under quiescent conditions. At the end of the hour, the settleable PM is carefully recovered from the Imhoff cone by decanting the supernatant from the cone and collecting the PM from the bottom. The settleable PM is pla ced in a pre weighed pan, dried at 105C, and a final weight is recorded. A volumetric fraction of the Imhoff cone supernatant (60 mL) is then passed through a pressurized stainless steel fractionation column with a 0.45m membrane using a vacuum pump. T he fractionated filtrate is used immediately for total dissolved phosphorus (TDP), measured with a HACH DR 5600 Spectrophotometer. The 0.45 m filters used for fractionation are preserved at 4C for total suspended phosphorus analysis. The remainder of th e Imhoff cone supernatant is used to determine the suspended PM fraction (as TSS). About 100 mL is taken from the well mixed supernatant and filtered through a 1 m glass fiber filter using a vacuum pump. Filters are washed, dried, and weighed prior to use. The sample volumes are recorded and filters are dried in an oven at 105C. Final filter weights are then recorded. The difference between final and initial filter weights provides the mass of PM, which can then be converted to a concentration (as TS S) with sample volume. The following fractions of phosphorus are measured by the spectrophotometer (HACH DR 5600): TDP, sediment bound total phosphorus, settleable bound total phosphorus, and suspendedbound total phosphorus. The method used for the measurement of phosphate (soluble reactive phosphorous) (orthophosphate) is Standard Method 4500P E Ascorbic Acid method. A HACH DR 5600 spectrophotometer is used to determine phosphate concentration. To determine sediment bound and settleable bound total phosphorus, small amounts (0.5 g) of sediment PM and settleablePM are acid digested. To determine suspendedbound total phosphorus, the 0.45-

PAGE 62

62 m membranes from the fractionation process are acid digested. To determine the TDP, fractionated filtrate is analy zed in the spectrophotometer. As part of the quality assurance/ quality control (QA/QC) procedure, all probes are calibrated with known standards. For the ions analyzed by the spectrophotometer, calibration curves for reagents are developed with at least 5 data points. The coefficient of determination R2 must be greater than or equal to 0.99. A new 5point calibration curve is created for each new reagent lot number Data Analysis and Elaboration Volumetric rainfallrunoff coefficients The runoff coeffici ent in this study is a volumetric based coefficient ranging from 0 to 1. It represents the fraction of runoff resulting from a known volume of rainfall across an entire event. The volumetric runoff coefficient is expressed as equation 31. n w n nA I Q c (3 1) In this expression, c represents the runoff coefficient, Qn represents the volumetric flow rate of sample period n observed at the outlet (L3/s), In represents the rainfall intensity of the nth interval (L/s), and Aw represents the m edian contributing area of the watershed (L2). Event mean value While the inter event distr ibutions of PM and phosphorus are log normal, a volume weighted event mean value (EM V ) is utilized as a comparative index for characterizing PM and phosphorus concen trations that represent a flow weighted PM and phosphorus concentration for a rainfall runoff event (Sansalone and Kim, 2008). EMV is calculated using equation 32.

PAGE 63

63 C V V C EMVn i i n i i i 1 1 ( 32) In this expression, Ci represents the average value associated with period i ( often mg/L), Vi represents the volume of flow during period i (L), and C repres ents the flow weighted average value for an entire event ( often mg/L). Removal efficien cy Treatment capacity is often quantified as a percent removal. This is calculated for the constituent of interest using the inflow and outflow loads and equation 33. 100 ) ( ) ( ) ( (%)1 1 1 IN i n i IN i EFF j m j EFF j IN i n i IN iC V C V C V PR (3 3) In this expression, Vi IN is the volume of influent flow for sampling period i (L); Vj EFF is the volume of effluent flow for sampling period j (L); Ci IN is the mean influent concentration associated with period i (mg/L); Cj EFF is the mean effluent concentration associated with period j (mg/L); n is the total number of influent measurements taken during an event; m is the total number of effluent measurements taken during an event. Particle size distribution (PSD) PSD is useful in characterizing PM. The output of the laser diffraction instrument is in percent finer by volume. A more useful unit is percent finer by mass, which can be used to find a d10, a d50, and a d90. In order to make this conversion, the measured gravimetric mass (from SSC) is distributed over the entire PSD for each sample. Mass based cumulative PSDs (i.e. percent finer by mass) for the samples are computed by integrating the gravimetric concentration

PAGE 64

64 of individual particle sizes for samples, corresponding flow rates, and sampling time intervals. With this paired data, PM total volume c oncentration for each particle size is converted to a gravimetric concentration in mg/L. Mass balance error A mass balance is often conducted after all of the events are captured in order to ensure mass conservation based on influent mass, effluent mass, a nd recovered PM mass trapped in the treatment unit due to sedimentation and filtration. A mass balance error (MBE) of 10% is chosen as the maximum allowable for this study. MBE is calculated using equation 34. 100 (%)inf inf M M M M MBErec eff ( 34) In this equa inf eff are the cumulative mass load of influent and effluent rec is the total mass of PM recovered after an event, including PM trapped by sedimentation and filtration. Mann Whitney U test for signifi cant difference When comparing two sets of data, a useful test for significant difference in data that is not normally distributed is the MannWhitney U test. For this test, a U value is calculated for each set of data, or treatment, and the lowest U is c ompared to a critical U value from a U table. U values are calculated by combining all the samples in the k treatments and ordering them from smallest to largest. Each sample is assigned a rank, with the smallest sample value being assigned first rank an d the largest sample value being assigned nth rank. If any 2 or more data points are equal, each is assigned the average of the ranks. Next, the samples are divided into their original treatment groups and the ranks for each treatment are summed. A U va lue for each treatment is then calculated using equation 35.

PAGE 65

65 2 ) 1 ( i i i is s r U ( 35) In this expression, ri is the rank sum for treatment i, si is the number of samples in treatment i. When the smallest of the calculated U values exceeds the critical U value, the conclusion is that the two treatment means are significantly different. Newtons law Settling velocity of a spherical particle in a fluid is calculated using Newtons Law provided in equation 36. p f f p d sd C g V 3 4 ( 36) In this expression, Vs is the settling velocit p is the f is the fluid density, Cd is the frictional drag coefficient, and dp is the particle diameter. Cd is calculated using equation 37. 44 0 3 24 d d dR R C ( 37) In this expression, Rd is the particle Reynolds number, which can be calculated for the transitional settling region using equation 38. f s p f dV d R ( 38) f is again the fluid density, dp is again the particle d iameter, Vs is again the settling velocity, and f is the fluid viscosity. Surface overflow rate The critical settling velocity is equal to the surface overflow rate which is calculated using equation 39.

PAGE 66

66 s crit sA Q V SOR ( 39) In this expres sion, SOR is surface loading rate, Vs,crit is the critical settling velocity, Q is the flow rate, and As is the surface area. Surface loading rate The surface loading rate is calculated using equation 10. s filterA Q SLR, (3 10) In this expres sion, SLR is surface loading rate, Q is the volumetric flow rate, and Afilter, s is the entire surface area of the filter. Cumulative gamma distribution (CGD) function The PSDs are modeled gravimetrically using a cumulative gamma distribution (CGD). In or der to represent any PSD both a dispersivity parameter and a representative parameter of PM size for the PSD are needed. The dispersivity parameter can be thought of as a characteristic of gradation uniformity or for example as a PM size sorting coefficient (Dickenson and Sansalone 2009). Therefore a series of PSDs with the same representative PM size parameter but differing shapes can be generated. The PM size parameter is a representation of the fineness or coarseness of a PSD, for example a phi scal e ( n) particle size gradation parameter or more commonly a d50m, where m indicates on a gravimetric basis (Dickenson and Sansalone, 2009). To model PSDs of differing uniformity and PM size representation is required for influent and effluent urban runoff. Previous studies (Sansalone and Ying 2008; Lin et al. 2009) have utilized a two parameter CGD to model size heterodisperse PSDs. The shape and scale factor parameters in the presentative

PAGE 67

67 coarseness or fineness of the PSD. Equations 311 and 312 represent the gamma distribution ( f(x) ) and CGD ( F(x)), respectively. ) ( ) / ( ) () / ( 1k e x x fx k (3 11) xdx x f x F0) ( ) ( (3 12) In this expression x represents the part icle diameter, d (m), k represents a shape Results and Discussion Event Hydrology Event based hydrologic indices including previous dry hours (PDH), event duration (drain) peak flo w rate (Qmax) median flow rate (Qmed) mean flow rate (Qmean), total influent volume (Vi), total effluent volume (Ve) rainfall depth (hrain) initial pavement residence time ( IPRT), and runoff coefficient (c) for a total of 25 events occurring between Ma y 28, 2010 and June 27, 2011 a re shown in Table 31. For the 25 storms monitored in this study, event duration ranges from 26 to 691 minutes. IPRT ranges from 3 to 33 minutes. Runoff volume ranges from 205 to 13,229 liters. Peak rainfall intensity ranges from 5.1 to 137.2 mm/hr. Peak runoff flow rate ranges from 0.5 to 14.3 L/s, median flow rate ranges from 0.1 to 5.5 L/s, and mean flow rate ranges from 0.1 to 5.5 L/s. Rainfall depth ranges from 0.3 to 5.0 cm. Figure 32 illustrates the cumulative fr equency distribution of event rainfall depth for Gainesville, FL based on historical data (NWS). The storms monitored in this study and their respective rainfall depths are also shown. For the 25 storms monitored in this study PDH range from 10 to 438 hours with a median of 75 hours. This median is for an entire year of study covering both the wet season and the dry

PAGE 68

68 season. This result is consistent with previous years in Gainesville, Florida. Based on historical data from the last ten years (NWS), the overall average dry period between rainfall events is 82 hours. The dry period between rainfall events varies greatly with seasonality. For this study, the median dry period during the wet season and during the dry season is 66 hours and 164 hours, resp ectively. Based on historical data for Gainesville over the past ten years, the average dry period during the wet season and during the dry season is 69 hours and 96 hours. These results indicate that while the wet season monitored in this study is typic al of Gainesville, the dry season is uncharacteristically dry. The runoff coefficient for the site ranges from 0.12 to 0.83. The low est runoff coefficient (0.12) i s observed on the 20 April 2011 event characterized by low rainfall depth and intensity and the lowest runoff volume. The highest runoff coefficient (0.83) i s observed on the 30 June 2010 event characterized by high rainfall intensity and a large runoff volume. These results for runoff coefficient are typical of urban catchments. A study by Cr istina and Sansalone found similar results for a small urban catchment in Cincinnati, Ohio with ranges of 0.2 to 0.4 and 0.6 to 0.8 for low and high intensity events, respectively (2003). Particulate Matter (PM) PM based studies are difficult to compare du e to differences in methodologies. For example, sampling can be done manually or with automated samplers. However, this study has concluded that automated samplers are not designed to sample coarse size heterodisperse PM representatively, but instead are designed to sample size monodisperse suspended wastewater flocs. In addition to differences in sampling technique, the analyses performed affect the comparability of studies. TSS is often used in studies as a gravimetric index for PM. However, studies have shown that the required subsampling step of the TSS procedure can introduce errors due to the difficulty of taking a representative subsample. Furthermore, TSS only provides

PAGE 69

69 information about fine particles (< 75 m). Although coarse particles (> 75 m) have been found to have lower pollutant concentrations (Sansalone and Buchberger, 1997; Roger et al., 1998; Lau and Stenstrom, 2005), they can be retained in treatment systems and result in clogging and a need for increased maintenance. An SSC analys is paired with PSD provides information about the entire gradation. Event based PM fractions including sediment PM, settleable PM, suspended PM, turbidity, SSC, and volatile suspended solids (VSS) for the 25 events are shown in Table 32. There i s signif icant variability in the ratios of suspended, settleable, and suspended PM to total PM between the 25 events. PM fractions for both influent and effluent runoff are illustrated in Figure 33. The influent suspended fraction of PM ranges from 1 to 29% wit h a volume weighted mean of 6 %. The effluent suspended fraction of PM ranges from 17 to 69% with a volume weighted mean of 5 4%. The influent settleable fraction of PM ranges from 0 to 30% with a volume weighted mean of 9%. The effluent settleable fracti on of PM ranges from 7 to 44% with a volume weighted mean of 21%. The influent sediment fraction of PM ranges from 55 to 99% with a volume weighted mean of 84%. The effluent sediment fraction of PM ranges from 12 to 65% with a volume weighted mean of 24% Figure 3 3 illustrates the dominance of the sediment PM fraction in the influent and the dominance of the suspended PM fraction in the effluent. This is due to the high removal of sediment and settleable PM fractions by the sedimentation and filtration processes utilized in the VCF unit. The percent removal of suspended PM ranges from 48 to 96% with a volume weighted mean of 77 %. The percent removal of settleable PM ranges from 65 to 99% with a volume weighted mean of 92%. The percent removal of sedi ment PM ranges from 97 to 100% with a volume weighted mean of

PAGE 70

70 100%. The percent removal of total PM (as SSC) ranges from 89 to 100% with a volume weighted mean of 98 %. Probability density functions (pdfs) for runoff flow rate, turbidity, suspended PM, set tle able PM, sediment PM, and SSC a re also developed for influent and effluent samples based on all 25 monitored events. The measured distributions are illustrated in Figure 3 4. All of the measured results for both influent and effluent samples are log normally distributed with an R squared greater than 0.90. MannWhitney rank sum tes ts a re conducted on turbidity and PM fractions to compare influent results and effluent results. Results of the tests indicate that, for all PM fractions and for turbidity, the effluent is significantly lower (p < than the influent. Particle Size Distribution (PSD) The percentages finer by mass, d10, d50, and d90, are shown in Table 33. The d50 represents the particle diameter for which 50 percent of the particles by mass are smaller than or the same siz e as that diameter. Similarly, the d10 and the d90 represent the particle diameters for which 10 and 90 percent of the particles by mass are smaller than or the same size as those diameters. For the 25 events monitored in this study, influent runoff d10 ranges from 2 to 54 m with a median of 9 m. Effluent runoff d10 ranges from 0 to 2 m with a median of 1 m. Influent runoff d50 ranges from 22 to 263 m with a median of 82 m. Effluent runoff d50 ranges from 1 to 11 m with a median of 3 m. Influe nt runoff d90 ranges from 173 to 1016 m with a median of 401 m. Effluent runoff d90 ranges from 2 to 52 m with a median of 12 m. The spread of the influent PSD range indicates that PM in the raw influent runoff is size heterodisperse and varies great ly with changes in rainfall intensity, traffic, and previous dry hours. This heterodispersity is illustrated in Figure 35 which shows the median PSD curve for the 25 monitored events influent runoff. The solid curve represents the median PSD, the upper and lower dotted curves represent the standard deviation of PSD across all events. The spread of the effluent PSD range is mainly

PAGE 71

71 attributed to filter ripening. Filter ripening is the complex process by which a colmation layer, composed of protozoa, bact eria, and algae, develops on the filter and contributes to PM removal (Dizer et al., 2004). PSD is highly dependent on hydrology. The most intense rainfall runoff events occur on 21 June 2010, 1 August 2010, and 14 May 2011 with peak rainfall intensities of 121.9, 127.0, and 137.2 mm/hr, res pectively. The 21 June event has the highest influent d10 and d50 values of 54 and 263 m, respectively. T he least intense events occur on 23 August, and 26 September 2010 and 20 April 2011 with peak rainfall intensiti es of 5.1, 5.1, and 15.2 mm/hr respe ctively. The 20 April event has the lowest d10 and d50 vely. The 26 September event has the second lowest influent d10 and d50 values of 4 and 35 m, respectively. A comparison of PM removal and PSD between primary effluent (sedimentation only) and secondary effluent (sedimentat ion + filtration) is also conducted for the 25 storms For this comparison, measured i nfluent PM, PSD and flow data a re used to model primary effluent PM and PSD for the VCF after sedimentation alone. For particles ranging from 0.02 to 2000 m, settling velocity is calculated using Newtons l aw. Newtons law is widely used to model Type I (discrete) settling because it is applicable for all settling regions (laminar, transitional, and turbulent) (Metcalf and Eddy, 2003). An iterative com putation of sett ling velocity is used with drag coefficient and Reynolds number. The flow rate at each sample time i s used to calculate surface overflow rate, or the critical settling velocity at that sample time. Particle sizes with settling velocities greater than the critical settling velocity a re assumed to be fully removed, whereas particle sizes with settling velocities less than the critical settling velocity a re assumed to be removed at a fraction equal to the ratio of particle settling velocity to critical settling velocity (Lawson, 1994). For each particle size of a given sample, removal rates are applied to

PAGE 72

72 the influent mass load and a new effluent SSC and PSD is generated from the result. The total mass load reduction during the sedimentation proce ss ranges from 65 to 99% with a volume weighted mean of 8 7 % and effluent d50 ranges from 3 to 25 m with a median of 10 m. Mann Whitney rank sum tests a re conducted on primary and secondary effluent SSC results as well as primary and secondary effluent PSDs. Although both PM removal and effluent PSD of primary and secondary effluent are significantly different, the comparison indicates that PM is substantially removed by hydrodynamic separation alone. This is attributed to the large surface area of the treatment unit and the low median flow rate of rainfall runoff during the 25 events monitored in this study which generate s relatively low surface overflow rates within the unit. Results of the PSD comparison are shown in Table 33 and results of the SSC comparison are shown in Table 34. The most significant deviations between primary and secondary effluent occur during high flow events. The three events wit h the highest flow rates occur on 21 June, 30 June, and 1 August 2010 with median flow rates of 5.5, 3.3, and 4.7 L/s and average flow rates of 5.1, 4.0, and 5.5 L/s, respectively. These events have among the highest modeled event based effluent d50 results of 21, 14, and 25 m and the highest modeled event based effluent d90 results of 44, 54, and 67 m, res pectively. The measured influent and effluent PSDs are shown with the modeled effluent PSD for the 1 August event in Figure 36. Probability density functions (pdfs) for surface overflow rate (SOR) and surface loading rate (SLR) are developed from the 25 monitored events. The measured distributions are illustrated in Figure 3 7. Both SOR and SLR are log normally distributed with an R squared greater than 0.83. The distribution of SOR is indicative of the VCF units settling capacity as a function of inf luent flow rate. The median SOR across all events is 0.20 mm/s. This

PAGE 73

73 corresponds to an equivalent particle diameter of 15 m. The distribution of SLR is indicative of the typical load on the filters for a small watershed during rainfall runoff events of varying intensity and duration. The median SLR across all events is 0.13 L/min/m2. The maximum SLR during this study is 8.05 L/min/m2, which is an order of magnitude below the optimum loading rate range for a typical granular filter (82 to 204 L/min/m2) (Metcalf and Eddy, 2003). Cumulative mass distributions of PM acr oss the entire size gradation are modeled with a CGD. The modeling process fit parameters k and in the CGD function. The parameters a re optimized by minimizing the sum of squared error ( SSE) and maximizing the correlation coefficient such that the pvalue i s greater than 0.05 between modeled and measured data. Cumulative trends are well behaved and consistently fit a CGD For measured influent PSD, k range s from 0.45 to 1.29 with a vol ume weighted mean of 0.77 ranges from 131 to 579 with a volume weighted mean of 302. For modeled primary effluent PSD, k ranges from 0.79 to 2.00 with a volume weighted mean of 1.28 and ranges from 2.0 to 27.5 with a volume weighted mean of 14.3. For measured secondary ef fluent PSD, k ranges from 0.20 to 2.01 with a volume weighted mean of 1.15 and ranges from 1.6 to 46.3 with a volume weighted mean of 10.0. Event based k and as a function of cumulative treated volume are illustrated in Figure 3 8. Gamma parameters f or all events are shown in Table 3 5. Phosphorus Event based phosphorus fractions including sediment P, settleable P, suspended P, dissolved P, and total P for the 25 events are shown in Table 36. Studies estimate that the EMV for total P resulting from urban rainfall runoff is around 300 g/L (Brown et al., 2003), considering both new and old urban development. T he values found in this study a re significantly higher. The volume weighted influent total P of 3283 g/L is an order of magnitude higher tha n concentrations reported in literature. Influent and effluent event based

PAGE 74

74 concentrations for dissolved and PM based phosphorus are shown in Figure 39. The major source of phosphorus in the watershed is biogenic material, from leaf fall and grass cuttin g maintenance, deposited on the pavement and mobilized during rainfall runoff events. Probability density functions (pdfs) for TD P, suspended P, sett leable P, sediment P, and TP are provided for influent and effluent samples based on all 25 monitored events. The measured distributions are illustrated in Figure 3 10. All of the measured results for both influent and effluent samples are log normally distributed at a 0.05 significance level with an R squared greater than 0.85. MannWhitney rank sum tests are conducted on dissolved and particulate bound fractions to compare influent and effluent results. Results of the tests indicate that, for all particulate bound phosphorus fractions as well as TP, the effluent concentrations are significantly lower (p < than the influent. However, there is no significant difference (p > between influent and effluent dissolved phosphorus. This is partly because there is no mechanism in the treatment unit to remove dissolved nutrients, but it is also because, during the time between treated storm events, phosphorus bound to the PM retained in the unit is resuspending into the dissolved phase and leaving the unit as part of the effluent during events. This indicates that regular maintenance for the uni t is required to periodically collect PM trapped in the unit during treatment. Conclusion The VCF unit evaluated in this study shows significant removal capabilities for both PM and particulate bound phosphorus. Coarse particles (sediment PM) are removed with the highest volume weighted mean efficiency of 100%. Fine particles (settleable and suspended PM) are al so significantly removed with a volume weighted mean removal efficiency of 8 7 %. The volume weighted mean removal efficiency of total PM (as SSC) is 9 8%. The median influent d50 for this study is 8 2 m and the median effluent d50 is 3 m. A comparison of PM removal

PAGE 75

75 and PSD between primary effluent (sedimentation only) and secondary effluent (sedimentation + filtration) demonstrated that filtration significantly increases the PM removal efficiency of the unit. The volume weighted mean removal of SSC without filtration for the 25 monitored events is 87% and the median effluent d50 is 10 m. Similar trends in removal of phosphorus are seen, with the highest removal efficiency occurring for sediment bound P. The volume weighted mean removal efficiencies for sediment P, settleable P, suspended P, and dissolved P are 99%, 94%, 40%, and 0%, respectively. The volume weighted mean removal efficiency for total P is 68%.

PAGE 76

76 Table 3 1. Monitored rainfall runoff event characteristics Event Date PDH (hr) IPRT (min) d rain (min) h rain (mm) i rain max (mm/hr) c V i (L) V e (L) Q max (L/s) Q med (L/s) Q mean (L/s) 28 May (2010) 96 10 112 20.6 76.2 0.73 7465 1849 4.3 1.0 1.1 16 June 20 18 61 16.0 61.0 0.63 5006 4760 5.4 0.7 2.2 21 June 42 6 43 23.4 121.9 0.74 8695 8517 7.5 5.5 5.1 30 June 24 8 50 13.2 76.2 0.83 5459 5342 9.1 3.3 4.0 15 July 75 8 28 9.7 91.4 0.75 3608 3310 13.3 1.4 3.1 1 August 24 5 36 30.0 127.0 0.80 11973 11616 14.3 4.7 5.5 6 August 120 12 104 3.6 50.8 0.78 1357 1101 6.8 0.1 0.3 7 August 24 7 48 8.6 50.8 0.61 2622 2513 13.3 0.4 0.9 23 August 48 20 42 2.8 5.1 0.22 312 177 1.3 0.1 0.1 12 September 141 17 52 6.9 50.8 0.48 1643 1503 3.9 0.1 0.5 26 September 40 11 78 3.6 5.1 0.63 1129 701 0.5 0.3 0.2 27 September 10 20 388 15.2 101.6 0.50 3841 3729 10.9 0.1 0.2 4 November 36 5 43 4.8 25.4 0.41 994 490 3.5 0.1 0.4 16 November 286 8 34 3.3 25.4 0.18 305 155 1.8 0.1 0.1 5 January 72 3 125 21.3 10 6.7 0.54 5800 4866 7.4 0.2 1.1 10 January 106 3 26 5.1 91.4 0.44 1129 1098 3.3 0.1 0.4 25 January 365 7 389 38.9 17.0 0.56 12406 12289 3.0 0.6 0.7 7 February 12 7 306 32.8 30.5 0.81 13229 13100 2.2 0.8 0.7 9 March 79 10 691 29.2 16.2 0.69 10551 9744 3. 1 0.1 0.2 28 March 438 7 66 2.5 33.9 0.41 522 497 1.0 0.1 0.1 30 March 48 33 179 15.5 76.2 0.49 3766 3732 5.6 0.1 0.3 20 April 196 9 61 3.6 15.2 0.12 205 116 3.3 0.1 0.1 14 May 188 5 295 50.3 137.2 0.43 10802 10596 11.5 0.1 0.6 6 June 541 19 69 4.1 22 .9 0.47 960 712 1.6 0.1 0.2 27 June (2011) 88 3 50 11.4 42.9 0.59 3383 2341 3.4 0.1 2.0 Mean 125 10 135 15.1 58.4 0.55 4686 4194 5.7 0.8 1.2 Median 75 8 61 11.4 50.8 0.56 3608 2513 3.9 0.1 0.5 SD 141 7 160 12.9 40.1 0.20 4286 4246 4.2 1.5 1.6

PAGE 77

77 Table 3 2. Summary of event based values for PM fractions Event Date Turbidity Suspended PM ( < 25 m) Settleable PM Sediment PM ( >75 m) SSC ( all PM fractions) % Volatile EMV i (NTU) EMV e (NTU) EMV i [mg/L] EMV e [mg/L] (%) EMV i [mg/L] EMV e [mg/L] (%) EMV i [mg/L] EMV e [mg/L] (%) EMV i [mg/L] EMV e [mg/L] (%) EMV i (%) EMV e (%) 28 May (2010) 35.6 14.1 43.7 11.9 87 45.4 6.9 93 435.9 6.2 99 532.2 15.4 99 51.0 40.2 16 June 32.7 10.7 40.2 19.7 53 39.6 2.0 95 78 8.4 7.0 99 1401.7 18.1 99 65.1 26.4 21 June 4.7 3.0 18.4 9.9 48 29.7 1.8 94 2328.4 5.6 100 1162.9 7.4 99 78.7 28.4 30 June 9.8 6.5 12.2 5.8 53 13.0 1.6 88 318.3 8.0 98 444.5 5.4 99 84.1 33.1 15 July 31.2 7.1 23.7 6.9 73 68.1 1.4 98 937.8 5.1 100 812.2 8 .4 99 74.7 65.9 1 August 14.8 3.9 18.5 6.9 64 111.7 8.7 93 243.0 4.8 98 245.1 7.7 97 70.5 52.7 6 August 51.9 1.4 48.0 12.1 82 29.5 2.9 93 390.4 13.2 98 308.4 7.3 98 51.3 3.0 7 August 15.6 3.8 13.1 7.0 49 32.2 5.2 85 222.5 1.6 99 115.8 13.9 89 42.3 31.0 23 August 46.6 5.3 38.3 5.0 92 35.9 3.2 95 459.5 2.9 100 555.8 4.7 100 69.1 46.9 12 September 27.9 3.6 45.2 11.6 76 46.0 4.1 92 110.5 2.7 98 261.5 5.8 98 56.3 40.7 26 September 21.4 3.3 11.2 2.2 85 5.1 2.4 65 1262.4 8.7 100 117.9 5.0 97 58.5 80.0 27 Se ptember 14.1 5.1 44.5 5.0 89 50.0 2.1 96 874.1 3.2 100 765.1 6.8 99 55.1 37.9 4 November 82.5 5.5 93.6 6.7 96 51.5 3.0 97 486.6 4.2 100 477.1 9.6 99 46.2 53.0 16 November 171.0 10.8 123.2 9.4 96 137.8 2.4 99 332.2 19.7 97 543.6 12.2 99 43.5 10.7 5 Janua ry 65.7 10.1 68.6 13.0 84 83.6 2.9 97 1411.7 3.2 100 693.2 8.7 99 69.4 52.2 10 January 38.0 3.3 20.8 3.1 86 60.0 3.6 95 453.0 1.4 100 211.1 3.0 99 68.0 24.8 25 January 28.2 6.8 32.3 3.5 89 37.5 3.6 91 390.9 2.0 100 105.8 4.2 96 68.1 34.8 7 February 30.0 5.9 20.4 4.4 79 14.4 0.9 94 639.7 2.5 100 438.3 7.6 98 75.8 54.5 9 March 19.4 2.4 21.9 8.8 62 10.6 1.3 88 69.6 1.9 97 78.2 2.8 97 57.8 31.2 28 March 61.8 3.5 48.9 9.8 84 19.5 2.9 88 101.1 2.9 98 102.8 5.6 96 54.6 19.7 30 March 70.7 4.6 44.9 5.1 89 59.6 2.2 96 667.5 1.0 100 443.7 7.3 98 60.2 5.6 20 April 112.2 2.4 65.7 7.9 93 78.0 3.5 98 254.6 2.3 100 921.7 6.1 100 44.7 22.8 14 May 19.9 5.6 33.9 11.3 67 43.2 1.2 97 602.5 0.7 100 487.3 5.3 99 66.0 10.2 6 June 38.4 3.7 54.2 10.6 85 31.4 2.6 94 313.1 1.1 100 237.5 9.0 97 54.9 25.4 27 June (2011) 63.8 3.4 54.3 10.1 83 77.1 2.6 97 586.7 5.2 99 591.7 9.8 98 81.3 3.8 Mean 44.3 5.4 41.6 8.3 78 48.4 3.0 93 587.2 4.7 99 482.2 7.9 98 61.9 33.4 Median 32.7 4.6 40.2 7.9 84 43.2 2.6 94 453.0 3.2 100 444.5 7.3 99 60.2 31.2 SD 36.7 3.1 26.3 3.9 15 31.5 1.8 7 494.7 4.3 1 338.4 3.7 2 12.1 19.6

PAGE 78

78 Table 3 3. Comparison of event based PM indices of d10, d50, and d90 for influent, primary effluent, and secondary effluent Event Date Measured influent PSD (m) Primary effluent PSD (after settling) (m) Measured secondary effluent PSD (settling + filtration) (m) d 10 d 50 d 90 d 10 d 50 d 90 d 10 d 50 d 90 28 May (2010) 7 69 915 5 16 34 2 11 34 16 June 28 242 1016 4 17 38 1 6 16 21 June 54 263 769 6 21 44 1 6 34 30 June 8 75 271 2 14 54 1 5 17 15 July 40 225 628 5 23 53 2 6 17 1 August 26 213 693 4 25 67 2 6 17 6 August 16 231 984 3 12 40 1 3 18 7 August 19 186 737 2 9 37 1 4 12 23 August 14 190 714 2 6 17 2 4 40 12 September 9 89 328 3 11 25 1 2 8 26 September 4 35 173 1 4 11 1 3 52 27 September 15 136 723 2 14 47 1 3 11 4 November 3 68 401 1 6 21 1 2 9 16 November 5 51 610 1 6 18 1 2 12 5 January 15 110 794 2 12 36 1 3 12 10 January 8 117 227 1 6 23 1 2 6 25 January 7 63 308 1 8 27 0 1 2 7 February 7 68 36 9 2 8 23 1 3 18 9 March 6 57 278 2 5 14 1 3 7 28 March 4 32 200 1 3 8 1 3 8 30 March 6 44 176 2 10 35 1 3 7 20 April 2 22 310 1 4 14 0 1 8 14 May 10 80 705 2 12 36 1 3 8 6 June 10 99 345 2 8 19 1 2 7 27 June (2011) 10 82 310 2 11 29 1 6 14 Mean 13 114 519 2 11 31 1 4 16 Median 9 82 401 2 10 29 1 3 12 SD 12 74 270 1 6 15 0 2 12

PAGE 79

79 Table 3 4. Comparison of event based SSC for influent, primary effluent, and secondary effluent Event Date Influent Primary e ffluent (after settling) Measured secondary effluent (settling + filtration) SSC [ mg/L ] SSC [ mg/L ] (%) SSC [ mg/L ] (%) 28 May (2010) 532.2 51.8 90 15.4 99 16 June 1401.7 68.3 88 18.1 99 21 June 1162.9 12.6 99 7.4 99 30 June 444.5 158.5 65 5.4 99 15 July 812.2 147.0 82 8.4 99 1 August 245.1 42.2 83 7.7 97 6 August 308.4 49.7 84 7.3 98 7 August 115.8 14.3 88 13.9 89 23 August 555.8 39.7 93 4.7 100 12 September 261.5 27.5 89 5.8 98 26 September 117.9 9.9 92 5.0 97 27 September 765.1 72.0 91 6.8 99 4 November 477.1 127.4 7 5 9.6 99 16 November 543.6 90.4 84 12.2 99 5 January 693.2 99.6 86 8.7 99 10 January 211.1 12.4 76 3.0 99 25 January 105.8 28.7 88 4.2 96 7 February 438.3 38.0 92 7.6 98 9 March 78.2 29.9 83 2.8 97 28 March 102.8 16.0 84 5.6 96 30 March 443.7 12.6 98 7.3 98 20 April 921.7 100.4 90 6.1 100 14 May 487.3 70.9 85 5.3 99 6 June 237.5 43.2 82 9.0 97 27 June (2011) 471.3 103.7 78 9.8 98 Mean 477.4 58.7 86 7.9 98 Median 444.5 43.2 86 7.3 99 SD 337.6 43.8 7 3.7 2

PAGE 80

80 Table 3 5. Comparison of event based gamma parameters for influent, primary effluent, and secondary effluent Event Date Influent PSD Primary effluent PSD (after settling) Measured secondary effluent PSD (settling + filtration) k k k 28 May (2010) 0.45 295.2 2.00 9.2 1.19 12.9 16 June 0.86 427.9 1.58 12.9 1.44 5.3 21 June 1.29 291.3 1.83 13.0 1.21 7.6 30 June 0.77 155.0 0.82 26.8 1.08 8.1 15 July 1.02 240.6 1.54 17.5 0.91 14.7 1 August 1.01 337.8 1.17 27.5 1.29 6.4 6 August 0.60 579.0 1.17 14.2 0.83 5.8 7 August 0.74 404.3 1.10 12.2 1.31 3.7 23 August 0.71 399.1 1.39 5.5 0.68 23.0 12 September 0.65 401.7 1.58 8.3 1.81 2.1 26 September 0.66 130.6 1.57 3.3 0.65 15.9 27 September 0.73 414.3 0.79 26.6 1.24 4.4 4 Nove mber 0.51 212.2 1.15 7.5 1.94 1.6 16 November 0.53 461.6 1.27 6.6 0.47 46.3 5 January 0.56 487.7 1.13 14.9 1.39 3.7 10 January 0.80 146.1 1.08 8.4 1.99 1.6 25 January 0.85 174.5 1.00 11.4 1.29 6.1 7 February 0.61 234.0 1.37 7.5 0.20 34.8 9 March 0.64 370.7 1.68 3.8 1.39 2.8 28 March 0.62 155.9 1.74 2.0 2.01 1.7 30 March 0.59 453.8 0.95 16.2 0.83 4.2 20 April 0.45 493.0 1.31 4.5 0.52 6.9 14 May 0.63 300.5 0.99 16.5 1.68 2.0 6 June 0.71 294.0 1.51 6.2 0.23 14.9 27 June (2011) 0.88 140.3 1.26 11.4 0.62 33.7 Mean 0.71 320.0 1.32 11.8 1.13 10.8 Median 0.66 300.5 1.27 11.4 1.21 6.1 SD 0.19 130.4 0.32 7.2 0.53 11.8

PAGE 81

81 Table 3 6. Summary of event based value s for dissolved and particulate bound phosphorus Event Date Dissolved P (DP) fraction Suspended P ( 1 to 25 m) Settleable P Sediment P ( >75 m) Total P EMV i [g/L] EMV e [g/L] (%) EMV i [g/L] EMV e [g/L] (%) EMV i [g/L] EMV e [g/L] (%) EMV i [g/L] EMV e [g/L] (%) EMV i [g/L] EMV e [g/L] (%) 28 May (2010) 498 362 64 714 359 75 175 26 93 1025 15 99 2405 762 84 16 June 328 318 8 444 402 32 266 13 95 2218 20 99 3256 666 81 21 June 175 229 0 116 112 6 103 6 94 5489 13 100 5883 360 94 30 June 216 249 0 265 172 37 76 9 88 660 17 98 1216 447 64 15 July 638 453 35 319 258 26 357 7 98 2234 12 99 3548 731 81 1 August 193 402 0 566 468 20 407 31 93 1175 23 98 2342 920 62 6 August 665 521 43 343 363 23 11 1 93 1021 35 98 2040 920 67 7 August 199 318 0 588 607 1 166 27 84 4 54 3 99 1407 955 35 23 August 527 655 23 472 232 69 19 2 95 592 4 100 1570 883 65 12 September 690 1045 0 753 237 71 21 2 92 671 17 98 2135 1300 44 26 September 669 854 6 562 591 23 11 5 65 1825 13 99 3068 1463 65 27 September 496 714 0 1218 1008 19 70 3 96 1279 5 100 3063 1730 45 4 November 1437 1030 64 1979 1347 66 355 21 97 1240 11 100 5011 2409 76 16 November 1921 1598 55 2619 787 84 1470 25 99 2891 165 97 8793 2574 84 5 January 1059 1360 0 820 733 23 171 5 98 1944 6 100 3947 2104 54 10 January 1036 1387 0 1106 1085 8 334 20 94 1378 4 100 3853 2496 39 25 January 450 507 0 1802 488 73 125 12 90 2120 11 99 4497 1146 75 7 February 684 626 9 706 546 23 13 1 94 1552 6 100 3407 1179 66 9 March 394 628 0 470 309 37 44 6 88 176 5 97 887 806 11 28 Mar ch 4670 2557 56 1780 1162 38 113 18 87 231 6 98 5718 3132 56 30 March 878 1111 0 1382 1196 14 255 9 96 1874 3 100 4364 2474 44 20 April 2045 4002 0 790 756 47 164 7 98 3505 3 100 6504 4769 59 14 May 740 790 0 618 685 0 128 3 98 1507 2 100 2994 1480 51 6 June 1177 1881 0 735 471 51 175 14 94 669 2 100 2769 2368 35 27 June (2011) 791 2182 0 830 557 37 370 13 97 1237 7 99 3228 2758 20 Mean 903 1031 15 880 597 36 216 11 93 1559 16 99 3516 1633 58 Median 669 714 0 714 546 32 164 9 94 1279 7 99 3228 1300 6 2 SD 925 876 23 608 342 25 290 9 7 1151 32 1 1825 1039 21

PAGE 82

82 Figure 3 1. Volumetric clarifying filter treatment unit design

PAGE 83

83 Rainfall depth, mm 0.01 0.1 1 10 100 1000 Cumulative frequency distribution, % 0 20 40 60 80 100 GNV Measured 28 March 23 Aug. 16 Nov. 6 Aug.; 26 Sep.; 20 April 4 Nov. 10 Jan. 12 Sep. 7 Aug. 15 July 30 June 27 Sep.; 30 March 16 June 28 May (2010) 5 Jan. 21 June 1 Aug.; 9 March 7 Feb. 25 Jan. 14 May 6 June 27 June (2011) Figure 3 2. Cumulative frequency distribution of rainfall depth for Gainesv ille, FL (GNV) based on 30 years of rainfall data (19802010)

PAGE 84

84 PM fraction [mg/L] 1 10 100 1000 SSC [mg/L] 1 10 100 1000 Influent Effluent Suspended Settleable Sediment SSC Figure 3 3. Suspended (< 25 m) settleable (25 75 m) sediment (> 75 m) PM fractions and SSC (all PM fractions) for influent and effluent

PAGE 85

85 Flow rate (L/s) 0.0001 0.001 0.01 0.1 1 10 100 pdf 0.00 0.02 0.04 0.06 0.08 0.10 0.12 Measured Flow Rate Modeled Flow Rate Flow rate = 1.3 L/s = 1.7 L/s Turbidity (NTU) 0.01 0.1 1 10 100 1000 10000 pdf 0.0 0.1 0.2 0.3 0.4 Effluent Measured Influent Measured Effluent Modeled Influent Modeled Turbidity Suspended PM (mg/L) 0.01 0.1 1 10 100 1000 10000 pdf 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Effluent Measured Influent Measured Effluent Modeled Influent Modeled Suspended PM Settleable PM (mg/L) 0.01 0.1 1 10 100 1000 10000 pdf 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Effluent Measured Influent Measured Effluent Modeled Influent Modeled Settleable PM Sediment PM (mg/L) 0.01 0.1 1 10 100 1000 10000 pdf 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Effluent Measured Influent Measured Effluent Modeled Influent Modeled Sediment PM SSC (mg/L) 0.01 0.1 1 10 100 1000 10000 pdf 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 Effluent Measured Influent Measured Effluent Modeled Influent Modeled SSC Figure 3 4. Probability d ensity functions (pdfs) for flow rate, turbi dity, and PM fractions (p > = 0.05)

PAGE 86

86 Particle Diameter (m) 0.1 1 10 100 1000 Percent Finer by Mass 0 20 40 60 80 100 Mean Effluent PSD Mean Influent PSD Range of Variation Figure 3 5. Median and range of variation for the influent and effluent PSD from the entire monitoring campaign (n=35)

PAGE 87

87 Particle Diameter (m) 0.1 1 10 100 1000 Percent Finer by Mass 0 20 40 60 80 100 Secondary Effluent Influent Primary Effluent 1 August 2010 Measured Influent d50 = 213 m Primary Effluent d50 = 25 m Secondary Effluent d50 = 6 m Effluent Mass = 416 g Treated Volume = 12 m3 Figure 3 6. Measured influent and secondary effluent PSD and modeled prima ry effluent PSD for the 1 August 2010 event

PAGE 88

88 Surface overflow rate, SOR (mm/s) 0.001 0.01 0.1 1 10 100 1000 pdf 0.00 0.02 0.04 0.06 0.08 0.10 0.12 Measured SOR Modeled SOR = 0.66 mm/s = 1.36 mm/s1.1 3.3 10.6 33.7 113.2 578.2 Equivalent particle diameter ( m)s = 2.65 g/m3T = 20oC Surface loading rate, SLR (L/min/m2) 0.001 0.01 0.1 1 10 100 1000 pdf 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16 Measured SLR Modeled SLR Granular Filter Loading Rate Figure 3 7. Probability density functions (pdfs) for surface overflow rate SOR and surface loading rate, SLR (p > = 0.05)

PAGE 89

89 k 0.0 0.7 1.4 2.1 2.8 Influent Primary Effluent 0.0 0.7 1.4 2.1 2.8 Influent Primary Effluent k 0.0 0.7 1.4 2.1 2.8 Secondary Effluent Primary Effluent 0.0 0.7 1.4 2.1 2.8 Primary Effluent Secondary Effluent 1 10 100 1000 Influent Primary Effluent l 1 10 100 1000 Influent Primary Effluent Cumulative treated volume (m3) 0 30 60 90 120 l 0.1 1 10 100 1000 Secondary Effluent Primary Effluent 1 10 100 Primary Effluent Secondary Effluent Figure 3 8. Event based cumulative gamma distribution parameters for measured influent modeled primary effluent, and measured secondary effluent PSD

PAGE 90

90 Dissolved phosphorus [mg/L] 0.01 0.1 1 10 Influent Dissolved P Effluent Dissolved P Dissolved phosphorus [mg/L] 0.1 1 10 Influent Effluent Cumulative treated volume (m 3 ) 0 30 60 90 120 PM-based phosphorus [mg/L] 0.01 0.1 1 10 100 Influent PM-based P Effluent PM-based P PM-based phosphorus [mg/L] 0.1 1 10 Influent Effluent Figure 3 9. Event based influent and effluent dissolved phosphorus and PM based phosphorus

PAGE 91

91 Flow rate (L/s) 0.0001 0.001 0.01 0.1 1 10 100 pdf 0.00 0.02 0.04 0.06 0.08 0.10 0.12 Measured Flow Rate Modeled Flow Rate Flow rate = 1.3 L/s = 1.7 L/s Dissolved P [ g/L] 0.001 0.01 0.1 1 10 100 1000 pdf 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 Measured Effluent Measured Influent Modeled Influent Modeled Effluent Dissolved P inf = 849 g/L inf = 976 g/L eff = 806 g/L eff = 566 g/L Suspended P [ g/L] 0.001 0.01 0.1 1 10 100 1000 pdf 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Measured Effluent Measured Influent Modeled Effluent Modeled Influent Suspended P inf = 906 g/L inf = 661 g/L eff = 593 g/L eff = 371 g/L Settleable P [ g/L] 0.0001 0.001 0.01 0.1 1 10 100 pdf 0.00 0.04 0.08 0.12 0.16 0.20 Measured Effluent Measured Influent Modeled Effluent Modeled Influent Settleable P inf = 217 g/L inf = 314 g/L eff = 12 g/L eff = 10 g/L Sediment P [ g/L] 0.0001 0.001 0.01 0.1 1 10 100 1000 pdf 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Measured Effluent Measured Influent Modeled Influent Modeled Effluent Sediment P inf = 1526 g/L inf = 1164 g/L eff = 19 g/L eff = 34 g/L Total P [mg/L] 0.01 0.1 1 10 100 1000 pdf 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Measured Effluent Measured Influent Modeled Influent Modeled Effluent Total Pm inf = 3.5 mg/L s inf = 1.9 mg/L m eff = 1.4 mg/L s eff = 0.8 mg/L Figure 3 10. Probability density functions (pdfs) for flow rate and dissolved and particulate bound phosphorus (p >

PAGE 92

92 CHAPTER 4 GLOBAL SUMMARY AND CONCLUSIONS This study evaluates the two primary sampling techniques for rainfall runoff monitoring to determine which provide s a more representat ive PM analysis under various rainfall intensities, runoff durations, and PM loads. Monitoring i s conducted on the Reitz Union parking facility located on the University of Florida campus in Gainesville, FL. PM from eight wet weather events was sampled i n parallel using manual and a utomated sampling techniques and samples are analyzed for PSD, PND, SSC, and PM fractions. Also in this study, the treatment efficiency of a volumetric clarifying filter ( VCF ) unit i s assessed under various rainfall intensitie s, runoff durations, and P M loads. Manual grab samples a re taken during 25 w et weather events and samples are analyzed for PM fractions, like sediment PM, settleable PM, TSS, and SSC, PSD, and dissolved and particulate bound phosphorus. Influent and effl uent event mean concentrations a re calculated along with removal efficiencies to assess treatment capabilities. This chapter summarizes the conclusions reached in this study. Results of the analyses on sampling techniques indicate that there are statistic ally significant differences between a PM analysis using manual sampling and one using automated sampling. For a size heterodisperse runoff with a median particle diameter (by mass) above 159 m, samples taken with an automated sampler significantly under estimate the PM load. This indicates that manual grab sampling paired with the SSC method and PSD characterization is a more accurate PM analysis for most untreated urban rainfall runoff from small watersheds. On the contrary, for runoff containing mostl y fine particles (< 75 m), this is not the case. For size monodisperse runoff with a d50 (by mass) below 159 m, there is no significant difference between the two PM analyses. This indicates that either sampling technique representatively characterizes the treated effluent from a rainfall runoff treatment unit implementing

PAGE 93

93 sedimentation with or without filtration. A preliminary PM analysis is required in this case to determine if the effluent is truly size monodisperse and composed of fine particles sm aller than 159 m. There are two main constraints of automated samplers that are responsible for the difficulties associated with stormwater monitoring. The first is that they do not sample the entire depth of flow. The location of the intake tube within the depth of flow greatly impacts sampling, since higher concentrations of particles are found at lower depths. The second is that the uptake velocity of the automated sampler is usually not equal to the localized streamflow velocity. Runoff flow rates are not constant; however intake tube uptake velocities are fixed. This leads to mostly non isokinetic sampling and substantial error due to inertial effects of the particles. Results of the VCF treatment assessment indicate that this unit shows significa nt removal capabilities for both PM and particulate bound phosphorus. Coarse particles (> 75 m) are removed with the highest volume weighted mean efficiency of 100%. Fine particles (< 75 m) are also significantly removed (p < with a volume we ighted mean removal efficiency of 87%. The volume weighted mean removal efficiency of total PM (as SSC) is 98%. The median influent particle diameter for this study is 82 m and the median effluent particle diameter is 3 m. A comparison between both me asured mass load reduction and effluent PSD for the VCF and a modeled mass load reduction and effluent PSD for a treatment unit of the same size without filtration demonstrated that filtration significantly increases the PM removal efficiency of the unit. The volume weighted mean removal of SSC by a unit with hydrodynamic separation alone for the 25 monitored events is 87% and the median effluent particle diameter is 10 m. Similar trends in removal of phosphorus are seen, with the highest volume weighted mean removal efficiency of 99% occurring for sediment bound P. All other particulate bound

PAGE 94

94 phosphorus fractions are significantly lower (p < in the effluent than in the influent. The volume weighted mean remo val efficiency for total P is 68%. However, there is no significant decrease in dissolved phosphorus concentration in the effluent as compared to the influent.

PAGE 95

95 LIST OF REFERENCES Andral, M.C., Roger, S., MontrjaudVignoles, M., Herremans, L., 1999. Particle size di stribution and hydrodynamic characteristics of solid matter carried by runoff from motorways. Water Env ironment Research 71 (4), 398407. American Society for Testing and Materials (ASTM), 1997. Standard methods for determining sediment concentration in w ater. Annual Book of ASTM Standards, Designation D397797(B). ASTM, West Conshohocken, PA. Bader, H., 1970. The hyperbolic distribution of particle sizes. Journal of Geoph ysical Research 75 (15), 28222830. Barrett, M.E., Irish, L.B., Jr., Malina, J.F., Jr., Charbeneau, R.J., 1998. Characterization of highway runoff in Austin, Texas area. Journal of Environme ntal Engineering 124 (2), 131137. Bent, G.R., Gray, J.R., Smith, K.P., Glysson, G.D., 2001. A synopsis of technical issues for monitoring sediment in highway and urban runoff U.S. Geolog ical Survey OpenFile Report 00497. U.S. Geological Survey. Brown, P.A., Gill, S.A., Allen, S.J., 2000. Metal removal from wastewater using peat. Water Research 16, 39073916. Burton, G.A., Pitt, R., 2001. Stormwater Effects Handbook: A Toolbox for Watershed Managers, Scientists, and Engineers Lewis Publishers, Boca Raton, FL. 928 pages. Clark, S.E., 2000. Urban stormwater filtration: Optimization of design parameters and a pilot scale evaluation. PhD dissertation, University of Alabama at Birmingham, Birmingham, Alabama. Clark S.E ., Siu, C.Y.S., 2007. Measuring solids concentration in stormwater runoff: comparison of analytical methods Environmental Scie nce and Technology 42(2), 511516. Clark, S. E., Siu, C. Y. S., Roenning, C. D., Treese, D. P., Pitt. R., Bathi, J. R., 2007. Automatic sampler efficiency for stormwater solids. Presented at the 2007 Pennsylvania Stormwater Management Sympo sium. Villanova, PA. October 18, 2007. Colorado Department of Public Health and Environment, 1996. CDPS Stormwater Sampling Guidance Document. Corell, D. L., 1998. The role of phosphorus in the eutrophication of receiving waters: a review. Journal of Environmental Quality 27, 261266. Cristina, C., Tramonte J., Sansalone, J.J., 2002. A granulometry based selection methodology for separation of traffic generated particles in urban highway snowmelt runoff. Water, Ai r, and Soil Pollution 136, 3353.

PAGE 96

96 Cristina, C., Tramonte, J.J., Cartledge, F.K., Pardue, J.H., 2005. Influence of hydrology on rainfall runoff metal element speciation. Journal of Environmental Engineering ASCE 131(4), 632642. De Leon, D.B., Lowe, J., 2009. Washington State Department of Ecology Standard Operating Procedure for Automatic Sa mpling for Stormwater Monitoring. Dickenson, J.A., Sansalone, J.J., 2009. Discrete phase model representation of particulate matter (PM) for simulating PM separation by hydrodynamic unit operations. Environmental Science and Techn ology 43 (21), 82208226. Dizer, H., Grutzmacher, G., Bartel, H., Wiese, H.B., Szewzyk, R., and Lopez Pila, J.M., 2004. Contribution of the colmation layer to the elimination of coliphages by slow sand filtration. Water Science and Technology 50 (2) 210 214. EPA, 1992. NPDES Storm Water Sampling Guidance Document. EPA, 2000. 2000 National Water Quality Inventory. National Water Quality Report to Congress under Clean Water Act Section 305(b) http://www.epa.gov/305b/2000report/ (last accessed January 11, 2011). Furamai, H., Balmer, H., Boller, M., 2002. Dynamic behavior of suspended pollutants and particle size distribution in highway runoff. Water Science and Technology 46(1112), 413418. Glenn, T., Tribouillard, S., Mishra, S., Singh, V., Sansalone, J., 2001. Pollutant mass and physical particulate pollution as a function of particle size and traffic for urban snow runoff, NOVATECH Conference Proceedings Lyon, Frankreich, 945952. Gnecco, I., Berretta, C., La nza, L.G., La Barbera, P., 2005. Storm water pollution in the urban environment of Genoa, Italy. Atmospheric Research 77 (14), 60 73. Gobel, P., Dierkes, C., Coldeway, W.G., 2007. Storm water runoff concentration matrix for urban areas. Journal of Conta minant Hydrology 91 (12), 2642. Gray, J.R., Glysson, G.D., Turcios, L.M., Schwartz, G.E., 2000. Comparability of suspended sediment concentration and total suspended solids data. U.S. Dept. of the Interior, USGS Information Services, Reston, Vir ginia. Greb, S.R., Bannerman, R.T., 1997. Influence of particle size on wet pond effectiveness. Water Envi ronment Research 69 (6), 11341138. Grizzard, T.L., Randall, C.W., Weand, B.L., Ellis, K.L., 1986. Effectiveness of extended detention ponds, pp. 323 337. In Urbonas, B., and Roesner, L.A., Urban runoff quality Impact of quality enhancement technology. ASCE, New York.

PAGE 97

97 Grottker, M., 1987. Runoff quality from a street with medium traffic loading. Science of the Total Environment 59, 457466. Harmel, R.D., King, K.W., Slade, R.M., 2003. Automated storm water sampling on small watersheds. Applied Engineerin g in Agriculture, 19 (6), 667674. James, R.B., 2003. TSS: A viable neasure of storm water pollutants? Proceedings of the 1st Annua l North American Surface Water Quality Conference and Exposition, StormCon 2003, San Antonia, TX. Avaliable at http://204.202.251.206/assets/55TSS.pdf (accessed on March 14, 2011). Khan, S., Lau, S. L., Kayhanian, M., Stenstrom, M.K., 2006. Oil and grease measurement in highway runoff: sampling time and event mean concentrations. Journal of Envir onmental Engineering 132, 415422. Lau, S.L., Stenstrom, M.K., 2005. Metals and PAHs adsorbed to stree t particles. Water Research 39, 40834092. Lawson, T.B., 1994. Fundamentals of aquacultural engineering. Kluwer Academic Publishers, Norwell. Legret, M., Pagotto, C., 1999. Evaluation of pollutant loadings in the runoff waters from a major rural hi ghway. Science of the Tot al Environment 235 (13), 143150. Lee, J.H., Bang, K.W., 2000. Characterization of urban stormwater runoff. Water Resources 34, 17731780. Li, Y.X., Lau, S.L., Kayhanian, M., and Stenstrom, M.K., (2005). Particle size distribution in highway runoff. Journal of Environmental Engineering ASCE 131 (9), 2671276. Liebens, J., 2001. Heavy metal contamination of sediments in stormwater management systems: The effect of land use, particle size, and age. Environmental Geology 41, 341351. Lin, H., Ying, G., Sansalone, J., 2009. Granulometry of noncolloidal particulate matter transported by urban runoff. Water Air Soil Pollution 198, 269284. Liu, B., Sansalone, J., Ying, G., 2010. Volumetric filtration of rainfall runoff: (I ) separation of particulate matter and filter forensics. Journal of Environment al Engineering 136 (12), 13211330. Marsalek, J., Rochfurt, Q., Brownlee, B., Mayor, T., Servos, M., 1999. An exploratory study of urban runoff toxicity. Water Science an d Technology 39 (12), 3339. Metcalf, L., Eddy, H., 2003. Wastewater Engineering: Treatment and Reuse. McGraw Hill, Boston, MA.

PAGE 98

98 Nix, S.J., Heaney, J.P., Huber, W.C., 1988. Suspended solids removal in detention basins. Journal of Environmental Engin eering 114 (6), 13311341. Pathapati, S., Sansalone, J.J., 2009. Combining particle analyses with computation fluid dynamics modeling to predict hetero disperse particulate matter fate and pressure drop in a passive runoff radial filter. Journal of Environm ental Engineering 135 (2), 7785. Pitt, R., S. Clark, P.D. Johnson, R. Morquecho, S. Gill, and Pratap, M., 2004. High level treatment of stormwater heavy metals. Proceedings of the 2004 World Water and Environmental Resources Congress: Critical Transitions in Water and Environmental Resources Management Jun 27Jul 1 2004, Salt Lake City, UT, 917926. Pratt, C.J., Mantle, J.D.G., Schofield, P.A., 1995. UK research into the performance of permeable pavement, reservoir structures in controlling stormwater discharge quantity and quality. Water Science and Technology 32 (1), 63 69. Reed, G.D., 1981. Evaluation of Automatic Suspended Solids Sampling Procedures. Water Pollution Control F ederation 53(10), 1481149. Roesner, L.A., 1999. Urban runoff pollution summary thoughts the state of practice today and for the 21st century. Water Science and Technology 39 (12), 353 360. Roger, S., MontrejaudVignoles, M., Andral, M.C., Herremans, L., Fortune, J.P., 1998. Mineral, physical and chemical analysis of the solid matter carried by motorway runoff water. Water Research 32 (4), 11191125. Rushton, B., England, G., Smith, D., 2007. ASCE guidelines for monitoring stormwater gross pollutants. The 9th biennial conference on stormwater research and watershed management. Orlando, Florida. Sansalone, J.J., Buchberger, S.G., 1997. Characterization of solid and metal element distributions in urban highway stormwater. Water Science and Technology 36 (89), 155160. Sansalone, J.J., Buchberger, S.G ., 1997. Partitioning and first flush of metals in urban roadway storm water. Journal of Environmental Engineering ASCE 123(2), 134143. Sansalone, J.J., Cristina, C.M., 2004. Prediction of gradationbased heavy metal mass using granulometric indices of snowmelt particles. Journal of Environmental Engineering 130, 14881497. Sansalone, J.J., Kim, J.Y., 2008. Transport of particulate matter fractions in urban source area pavement surface runoff. Journal of Environmental Quality 37, 18831893. Sans alone, J.J., Koran, J.M., Smithson, J.A., Buchberger, S.G., 1998. Physical characteristics of urban roadway solids transported during rain events. Journal of Environmental Engineer ing ASCE 124(5), 427440.

PAGE 99

99 Sartor, J.D., Boyd, G.B., Agardy, F.J., 1974. Wa ter pollution aspects of street surface contaminan ts. Journal WPCF 46 (3), 458467. Standard Methods for the Examination of Water and Wastewater, 1995. American Public Health Association, Washington, DC. State of Maine Bureau of Land and Water Qualit y Management, 2007, Standard Operating Procedure for Effluent Monitoring of Stormwater Discharges for Facilities Covered Under the MPDES Multi Sector General Permit for Stormwater Discharges Associated Industrial Activities. Doc No. DEPLW0859. Stenstrom R.S., Clausen, J.C., Askew, D.R., 2002. Treatment of parking lot stormwater using a StormTreat system. Environmental Science and Technology 36 (20), 44414446. Taylor, G.D., Fletcher, T.D., Wong, T.H.F., Breen, P.F., Duncan H.P., 2005. Nitrogen com position in urban runoff implications for stormwater management. Water Research 39 (10), 19821989. Van Dolah, R.F., Riekirk, G.H.M., Levisen, M.V., Scott, G.I., Fulton, M.H., Bearden, D., Sivertsen, S., Chung, K.W., Sanger, D.M., 2005. An evaluati on of polycyclic aromatic hydrocarbon (PAH) runoff from highways into estuarine wetlands of South Carolina. Archives of Environmental Contamination and Toxicology 49, 391396. Vaze, J., Chiew, F.H.S., 2003. Study of pollutant washoff from small impervious experimental plots. Water Re sources Research 39 (6), 11601170. Westerlund, C., Viklander, M., 2006. Particles and associated metals in road runoff during snowmelt and rainfall. Science of the Total Environment 362(13), 143156. Williams, P.C., 2001. Implementation of near infrared technology. Near Infrared Technology in the Agricul ture and Food Industries, 145 171. Wu, J.S., Homan, R.E., Dorney, J.R., 1996. Systematic evaluation of pollutant removal by urban wet detention ponds. Journal of Environmental Engineering, ASCE 122 (11), 983988. Ying, G., Sansalone, J.J., 2008. Partitioning and granulometric distribution of metal leachate from urban traffic dry deposition particulate matter subject to acidic rainfall and runoff retention. Water Research 42 (15), 21462162.

PAGE 100

100 BIOGRAPHICAL SKETCH Christina Herr graduated magna cum laude with her bachelors d egree in e nvironmental e ngineering s ciences with a minor in s ustainability s tudies from the University of F lorida in 2010. Christina receiv e d the degree of Master of Engineering in e nvironmental e ngineering s ciences from the University of Florida in August 2011. Her masters research was focused on experimentation of urban stormwater treatment unit operations and processes. She worked under the guidance of Dr. John J. Sansalone in the department of e nvironmental e ngineering s ciences. After graduation, Christina moved to Atlanta, Georgia to work in the Water Resources field.