Measuring and Removing Dissolved Metals from Stormwater in Highly Urbanized Areas (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

A-1 A P P E N D I X A Practical Guidance for Measurement of Dissolved Metals in Stormwater Runoff

A-2 T A B L E O F C O N T E N T S A-3 Chapter 1 Introduction A-4 1.1 The Issues A-8 Chapter 2 Preparation and Planning A-10 Chapter 3 Constituent Selection A-10 3.1 Conventional Parameters A-11 3.2 Major Ions A-11 3.3 Organic Composition A-12 Chapter 4 Equipment A-14 4.1 Sampler Intake Strainer, Intake Tubing and Flexible Pump Tubing A-15 4.2 Composite Containers A-16 4.3 Flow Monitoring A-17 4.4 Rainfall Gauges A-17 4.5 In Situ Sensors A-17 4.6 Power A-17 4.7 Telecommunication for System Command/Control and Data Access A-18 Chapter 5 Sample Collection and Handling Procedures A-19 5.1 Pacing the Sampler A-20 5.2 Percent Capture A-21 5.3 Peak Capture A-21 5.4 Evaluation of Representative Sampling A-22 5.5 Sample Splitting A-24 5.6 Filtration Issues A-25 5.6.1 Filtration Methods A-27 5.6.2 Real-time Filtration A-29 Chapter 6 Analytical Methods and Quality Assurance/Quality Control A-29 6.1 Detection Limits, Data Quality Objectives, and Measurement Quality Objectives A-30 6.2 Precision, Bias, Accuracy, Representativeness, Completeness, and Comparability A-30 6.2.1 Precision A-30 6.2.2 Bias A-31 6.2.3 Accuracy A-32 6.2.4 Representativeness, Comparability, and Completeness A-34 6.3 Laboratory Quality Assurance/Quality Control A-34 6.4 Field QA/QC A-34 6.4.1 Blanks A-35 6.4.2 Field Duplicates A-36 Chapter 7 Equipment Cleaning, Blanking, and Tracking A-37 7.1 Laboratory Sample Bottles A-38 Chapter 8 Reporting and Documentation A-40 Bibliography

A-3 One major challenge associated with characterization of receiving waters and stormwater runoff, as well as assessment of BMP effectiveness are the low concentrations that must be achieved to meet water quality standards (Table 1). When hardness is depressed in a freshwater receiving water body, the water quality standards for dissolved copper, lead and cad- mium can approach the detection limits that can reasonably be achieved by most laboratories. The difficulties in measur- ing trace metals in ambient waters became readily apparent in the late 80s and early 90s when surveys using ultraclean, reliable sampling and analytical methods (Shiller 1987; Horowitz et al. 1994; Windom et al. 1991) first demonstrated the lack of reliability of previous metals data. Contamination resulting from cumulative effects of sampling equipment and processing techniques used prior to 1990 were found to bias the concentrations of cadmium, copper, lead, and zinc that had been reported in surface waters. Each aspect of sampling, processing, and analyzing water quality becomes increasing critical as lower project report- ing limits are required to evaluate concentrations of dissolved metal in environmental samples. Methods and materials used in sample collection, processing, and analysis may introduce contamination or artificially reduce measured concentra- tions of dissolved metals in the samples. An extensive suite of quality control measures is necessary to address potential contamination and adsorptive losses. This starts with assur- ing that all parts of the sampling system that come into con- tact with the water are comprised of appropriate materials, proper cleaning procedures are employed, and systems are in place to document cleanliness of all system components. Over the past 15 years, a number of documents have been developed to provide guidance for monitoring of stormwater discharges. Three of these documents are comprehensive in scope and should be considered when developing any storm- water monitoring effort. The Caltrans Stormwater Moni- toring Protocols (Caltrans 2003) were initially developed in 2000 and later updated and expanded in 2003. This set of protocols was developed to provide comprehensive guidance to Caltrans management, staff, and contractors for use in the planning and implementation of stormwater monitoring programs. Step-by-step descriptions are provided to guide the user through the processes used to plan and implement a successful water quality monitoring program specific to run- off from transportation-related facilities. Soon after the first edition of the Caltrans manual, another manual was devel- oped for the FHWA (Strecker, Mayo et al. 2001). The FHWA stormwater monitoring manual included monitoring strate- gies for five different objectives common to stormwater pro- grams, equipment selection, installation and applicable QA/ QC requirements. In the same time frame, U.S. Geological Survey and the FHWA (Granato, Zenone and Cazenos, eds, 2003) compiled another significant resource consisting of a series of chapters written by experts in each field. The indi- vidual chapters address a wide range of topics directly related to monitoring stormwater. These chapters included basic information requirements and data quality; quality assurance and quality control practices; measurement of precipitation and runoff flow; the geochemistry of runoff; measurement of sediments, trace elements, and organic chemicals in runoff; assessment of the potential ecological effects of runoff; moni- toring atmospheric deposition; and interpreting runoff data using appropriate statistical techniques. Many additional resources are available that provide spe- cific information on cleaning procedures, sampling methods, general water quality testing, and documentation necessary to support and validate monitoring data. Horowitz et al. (1994), Keith (1996) and the most current version of the USGS National Field Manual (NFM) for Collection of Water Quality Data (various dates) are essential sources of information on cleaning, sampling and processing water samples. The USGS NFM is maintained online at http://pubs.water.usgs.gov/ twri9A/ in order to provide easy access to periodic modifica- tions and improvements. These documents should be consid- ered central to most any water quality monitoring effort. A C h A P T E r 1 Introduction

A-4 number of other resources have been developed in recent years that contain useful information for planning and implement- ing stormwater monitoring efforts that focus on dissolved met- als at concentrations relevant to water quality criteria. Many of the more valuable guidance documents and areas addressed by these documents are summarized in Table 2. This manual is intended to provide specific monitoring guidance for sampling, processing, and analyzing dissolved metals in highway and urban runoff. Clean sampling tech- niques are required whenever low-level analytical reporting limits are necessary for analysis of metals. It is well recognized that measurement of dissolved metals at levels relevant to water quality criteria can be extremely challenging. Extreme care is required to avoid contamination during the collection, processing, transport, and analysis of samples. The validity of the data is dependent upon implementation of quality assurance measures to demonstrate that all materials contacting the samples are free of contamination and that sampling handling processes do not introduce contamination. Monitoring guidelines emphasize the steps necessary to collect, process, and analyze stormwater samples while providing the documentation necessary to demonstrate that the data are reli- able and repeatable. Access to supporting documentation allows other researchers to independently validate the monitoring data and to make informed decisions regarding appropriate use of the data for other purposes. Although this manual focuses on assessment of dissolved metals, monitoring protocol are intended to provide repre- sentative and accurate measurement for all phases of metals associated with runoff from small, highly impervious drain- ages as well as supplementary information necessary to evalu- ate the metals data. Recommended protocols are provided for development of a monitoring program designed for assess- ment of loads and concentrations of dissolved trace metals. This requires representative rainfall, flow, and water quality information. These are intended to provide insight as the strategies, approaches, and techniques useful for monitoring highway or urban stormwater runoff. Specific objectives and site con- ditions must be considered when establishing a sampling program. The overall objectives of monitoring programs for dissolved metals will be quite varied and may necessitate different monitoring strategies. This manual is intended to address all considerations necessary to assure that sampling and analytical processes are well documented and provide the necessary sensitivity to address water quality objectives. The manual was designed to be used by both field and laboratory staff. The intent was to produce a document that serves as a practical working document. Protocols are pre- sented to assist staff in all phases of the process starting with designing a successful study and the preparations necessary for successful implementation. 1.1 The Issues The National Highway Runoff Water-Quality Data and Methodology Synthesis (Granato, Zenone et al. 2003) pro- vided a comprehensive look at technical issues associated Freshwater1 Marine2 Total Dissolved Total Total Metal Hardness CMC CCC CMC CCC Acute Chronic As 25 340 150 340 150 69 36 Cd 25 0.52 0.10 0.52 0.09 40 8.8 Cr(III) 25 580 28 180 24 Cr(VI) 25 16 11 16 11 1100 50 Cu 25 3.8 2.9 3.63 2.73 4.8 3.1 Pb 25 14 0.5 14 0.5 210 8.1 Ni 25 140 16 140 16 74 8.2 Ag 25 0.4 0.3 0.954 Zn 25 37 37 36 36 90 81 1 Freshwater criteria are based upon hardness with the exception of arsenic (As) and trivalent chromium (Cr+3). CMC (Criteria Maximum Concentration) is the highest level for a 1-hour average exposure not to be exceeded more than once every three years. CCC (Criteria Continuous Concentration) is for a 4-day exposure not to be exceeded more than once every three years. 2 Criteria for marine waters based upon unfiltered samples. 3 Based upon CTR (2000) hardness-based calculation. Lower criteria can result from use of the EPA (2007) biotic ligand model when pH is depressed (6.0-6.5) and dissolved organic carbon content is low (<2 mg/L). 4 Criterion is halved for consistency with 1985 guideline derivation. Table 1. Water quality criteria for metals.

A-5 Table 2. Summary of key resources for the design and implementation of a sampling program for dissolved metals at low levels. Reference P re pa ra tio n an d Pl an ni ng Si te a nd E qu ip m en t Se le ct io n Cl ea ni ng a nd B la nk in g Pr oc ed ur es Fi el d Sa m pl in g – Eq ui pm en t & S tra te gi es Fi el d Sa m pl in g – Sa m pl e Ha nd lin g A na ly tic al M et ho ds a nd Qu al ity A ss ur an ce D at a M an ag em en t Caltrans (2003). Caltrans Comprehensive Protocols Guidance Manual. Granato, G. E., C. Zenone (2003). National Highway Runoff Water-Quality Data and Methodology Synthesis, Volume I: Technical Issues for Monitoring Highway Runoff and Urban Stormwater. Strecker, E., Mayo, J. (2001). Guidance Manual for Monitoring Highway Runoff Water Quality. Keith, L. H., ed (1996). Principles of Environmental Sampling, Second Edition. Horowitz, A.J., et al. (1994). U.S. Geological Survey Protocol for the Collection and Processing of Surface Water Samples for the Subsequent Determination of Inorganic Constituents in Filtered Water. U.S. EPA (1996). Method 1669: Sampling Ambient Water for Trace Metals at EPA Water Quality Criteria Levels. U.S. EPA (1996). Guidance on the Documentation and Evaluation of Trace Metals Data Collected for Clean Water Act Compliance Monitoring Burton, G. A. and R. Pitt (2002) Stormwater effects handbook: a toolbox for watershed managers, scientists, and engineers. U.S. Department of Agriculture-National Resources Conservation Services. (2003). National Handbook of Water Quality Monitoring. Shaver, E. R., R. Horner, J. Skupien, C. May, and G. Ridley (2007). Fundamentals of Urban Runoff Management: Technical and Institutional Issues Law, Neely L., L. Fraley-McNeal, K. Cappiella, and R. Pitt (2008). Monitoring to Demonstrate Environmental Results: Guidance to Develop Local Stormwater Monitoring Studies Using Six Example Study Designs Florida Department of Environmental Protection 2008. Standard Operating Procedures for Field Activities, Surface Water Sampling and General Sampling U.S. Geological Survey (Various Dates). Techniques of Water Resources Investigations, Book 9, Handbook for Water Resources Investigations, National Field Manual for Collection of Water Quality Data. http://pubs.water.usgs.gov/twri9A/ Geosyntec Consultants and Wright Water Engineers, Inc. (2009). Urban Stormwater BMP Performance Monitoring.

A-6 with sampling, processing, and analyzing dissolved metals in highway and urban runoff. Runoff from highly impervious pavements and roadways transports dissolved, colloidal, and suspended solids in a heterogeneous mixture and, along with pH, alkalinity, traffic levels, and residence time, influences the partitioning of heavy metals. Water quality characteris- tics such as low alkalinity, low hardness, and short pavement residence times can cause a majority of the heavy-metal mass to remain in solution at the edge of the pavement, with par- titioning coefficients approaching equilibrium conditions only toward the end of the event, as heavy metals partition to entrained solids (Sansalone and Buchberger 1997; Sansalone and Glenn 2000; Glenn, Liu et al. 2001; Magnuson, Kelty et al. 2001; Dean, Sansalone et al. 2005; Barber, Brown et al. 2006). The rate at which these geochemical processes occur may be an important consideration when considering the applica- tion of typical best management practices (BMPs) for high- way runoff within the right-of-way. BMPs such as a detention basin or roadside swale may be necessary to detain runoff and produce sufficient residence time so that partitioning to the entrained solids occurs. This calls for an improved understand- ing of the concentrations of dissolved metals in stormwater at the point of entry into a BMP or receiving water body as well as better geochemical and hydrological information to assess factors that will contribute to modification of the dissolved fraction (fd) of trace metals. The numerous hydrologic and geochemical factors con- trolling speciation of metals and the varying rates at which metals partition between dissolved and particulate forms presents a significant problem with respect to the quanti- fication of the dissolved fraction. All research conducted to investigate the transformational processes that occur in stormwater runoff from highways has been based upon grab samples that received filtration within one to six hours. The current criteria used for flow-weighted composites necessi- tates filtration after completion of all flow from a storm event, a process that can extend from less than one hour to days. The EPA method for analysis of dissolved metals specifies that samples should be filtered within 15 minutes or as soon as possible after completion of sampling. Caltrans guidelines (Caltrans 2003) called for filtration to occur within 48 hours of the time the last aliquot is sampled, but did identify the basis for this specification. The time that geochemical and hydrologic factors are expected to result in the partitioning process to near equilibrium is estimated to take 2–12 hours. Given the uncertainties in the duration of storm events and the many factors that could influence partitioning, none of the existing guidelines for holding times would be expected to provide consistent data that would be considered useful or even comparable among storm events at a given location. Current stormwater practices that delay filtration until sam- ples reach the laboratory are likely to be more representative of equilibrium conditions than conditions during the storm at the monitoring point. Without modification of current commercial storm- water monitoring systems to enable real-time filtration, it is unlikely that concentrations of dissolved metals during the storm event can be adequately characterized at the sampling point with flow-weighted composites. EPA’s Engineering and Analysis Division (U.S. EPA 1998) experimented with adap- tation of in-line filtration procedures to automated com- positing devices at a wastewater treatment facility with the objective of assembling a system from commercially available parts in a cost-effective manner capable of obtaining clean field blanks and uncontaminated samples. Using an equip- ment configuration similar to that proposed in this guidance document, they were able to achieve the objective of clean field blanks and uncontaminated samples but were not suc- cessful in assembling a system that provided immediate filtra- tion. Development of an automated sampling system with the capabilities of providing immediate, real-time filtration for dissolved metals should remain an important goal for programs using flow-weighted composite methods to char- acterize dissolved metals. For purposes of this document, we utilize the commonly accepted operational definition for “dissolved” metals that is based upon the concentration of metals in whole-water samples after passing through a 0.45 µm membrane filter. Depending upon the particular metal of concern and geo- chemical factors associated with the sample, it is recognized that “dissolved” fraction is not comprised solely of ionic species but can contain a substantial proportion of colloi- dal metals (Horowitz 1994; Bricker 1999; Tucillo 2006). In actuality, only a small proportion of some metals in a filtered water sample may be in the dissolved state as the majority are adsorbed to colloids or bound to humic or fulvic compounds. The importance of colloids will depend upon specific site conditions and project objectives. Studies suggest that trans- port of metals associated with colloids can be enhanced in cold climates where roads are salted with NaCl (Amrhein, Mosher and Strong 1993; Amrhein et. al 1994). Tucillo (2006) found that copper and zinc were primarily in the dissolved concentrations (less than 10Kda or 0.003 micron) or frac- tions greater than 20 microns. This was thought to explain high variability in performance of BMPs that relied upon settling. Artifacts of the filtration process influence whether col- loidal metals are captured in the filtrate or on the filter. The volume filtered, amounts of sediment present, the effective surface area of the filter, and the type of membrane material used to manufacture the filter can all influence what passes through the filter (Horowitz, Lum et al. 1996). As such, this simple definition for dissolved chemical constituents must be applied with some caution. Despite the known flaws in

A-7 using filtration pore size to distinguish between particulate and dissolved metal species, monitoring typically requires assessment of dissolved metals since they serve as the basis for both interpretation of biological effects and current water quality criteria. It becomes more important to standardize and document the actual filtration techniques used as well as control factors such as filter packing, which is known to increase variability. This guidance manual provides guidance for the design and implementation of stormwater monitoring for dissolved metals. These protocols include (1) batch cleaning of sam- pling equipment, bottles, hoses; (2) quality checks necessary to verify that each batch of components subjected to clean- ing meet project objectives; (3) tracking the use of all sys- tem components by batch so that they can be linked to the field samples; (4) establish requirements and guidelines for representative samples; (5) detailed sample handling and transportation procedures; (6) sample processing in the laboratory; (7) laboratory accreditation and participation in inter-laboratory calibration studies; and (8) the basic level of documentation necessary for final reports to allow others to assess the validity of the data set. This guidance also includes important ancillary data and standardized explanatory vari- ables recommended to both provide a common basis for comparison among studies and to account for some of the variability in the data.

A-8 Preparation and Planning Developing a stormwater monitoring program that is capa- ble of obtaining environmental data that are representative, accurate, and provide the supporting information needed to be used by other programs requires careful preparation and plan- ning. A monitoring program designed to obtain environmen- tal data for decision making purposes should be conducted in three phases: planning, implementation, and assessment. Jones (1999) provides a detailed outline of the necessary elements to be addressed as part of the project planning process. In addi- tion, EPA provides a set of extensive guidelines (http://www. epa.gov/quality/qa_docs.html) for the design and implementa- tion of monitoring programs to enable validated data with min- imized total error that can be used to support decision-making processes. These guidelines provide a basis for development of the major components of a SAP with emphasis on use of clean sampling methods and providing the documentation necessary to support use of the data by others. The preparation and planning stage includes development of a SAP that will serve as the roadmap for all phases of the program. An example of the contents and structure of a typi- cal SAP for stormwater monitoring is shown in Figure 1. The SAP includes an overview of the monitoring objectives, project organization, sampling site locations, monitoring constituents and detection limits, analytical methods, sampling frequency, field equipment operation, sample collection methods, QA/ QC protocols, and logistical details for the project. The SAP incorporates Data Quality Objectives (DQOs) developed to meet the goals of the program and expected use of the data. The SAP should provide information on data manage- ment and reporting which is an important element of assur- ing that basic data generated by the program are (1) available for public use as original data, (2) supported by sufficient quality assurance information (to indicate the validity, reli- ability, and compatibility of data from different sources), and (3) available in electronic files in database formats that are universally available. These factors were among the impor- tant issues cited by Granato, Bank et al. (2003) as primary requirements for assuring data that are both accessible and useful for incorporation into larger data sets for larger scale comparison. The SAP should provide details on the data validation pro- cess to describe how the data will be validated to determine if DQOs were met. The data quality assessment report should identify any data quality issues that were encountered, correc- tion action taken and any necessary qualifications applied to the data. This report should provide an overall assessment of the uncertainty of data based upon project QA/QC activities and any limitations on interpretation of the study results that might result from data quality issues. Preparation and planning is especially critical to measure- ment of dissolved metals. Collection of high quality samples that are free of external contaminants and representative of actual field conditions will require an extensive expenditure of effort in the planning stages. A large proportion of the effort necessary to obtain representative stormwater samples is expended well before the actual sampling events. C h A P T E r 2

A-9 SAMPLING AND ANALYSIS PLAN 1.0 PROJECT OVERVIEW/DESCRIPTION 1.1 Description of why the project is being conducted 1.2 Description of who is conducting the project 1.3 General scope of monitoring activities 1.4 Project organization/roles and responsibilities 2.0 MONITORING SITE(S) 2.1 Site location (map) 2.2 Written driving directions 2.3 Site access instructions (gates, locks, keys, combinations) 2.4 Notification procedures 3.0 ANALYTICAL CONSTITUENTS List of constituents for sampling and analysis (including sample collection method, bottle type, volume required, preservation, and laboratory performing analysis) 4.0 DATA QUALITY OBJECTIVES (DQOs) 4.1 Analytical reporting limits 4.2 Analytical precision, accuracy, and completeness 5.0 FIELD EQUIPMENT MAINTENANCE 5.1 Equipment calibration 5.2 Equipment maintenance 5.3 Equipment cleaning (bottles/lids/tubing) 6.0 MONITORING PREPARATION AND LOGISTICS 6.1 Weather tracking 6.2 Storm selection criteria 6.3 Storm action levels 6.4 Communications/notification procedures 6.5 Sample bottle ordering 6.6 Sample bottle labeling 6.7 Field equipment preparation 7.0 SAMPLE COLLECTION, PRESERVATION, AND DELIVERY 7.1 Sample collection methods 7.2 Field measurement methods 7.3 Field equipment list Figure 1. Outline of typical sampling and analysis plan for stormwater monitoring.

A-10 Constituent Selection A substantial amount of research has been conducted on the characterization, speciation, and removal of dissolved metals in highway runoff. This research has demonstrated that metal partitioning between the dissolved and particulate-bound fractions in stormwater is a dynamic process that is influ- enced by a number of different factors (Doner 1978; Bauske and Goetz 1993; Granato, Church, and Stone 1995; Breault et al. 1996; Sansalone and Buchberger 1997). Monitoring of dissolved metals in runoff should incorporate measurement of the major constituents known to influence partitioning. This includes geochemical data such as sediment, pH, alkalin- ity, major ions, hardness, and dissolved organic compounds (DOC). Use of geochemical or biotic ligand models to examine conditions that may mobilize metals and make them more available has been suggested as an alternative to direct mea- surement of dissolved metals (Geosyntec Consultants and Wright Water Engineers 2009). 3.1 Conventional Parameters Conventional parameters should include total dissolved solids (TDS), total suspended solids (TSS) or suspended sol- ids concentration (SSC), total organic carbon (TOC), and DOC. Specific conductance and pH should be measured during both the subsampling process and immediately upon arrival at the laboratory since values can be subject to change over relatively brief periods. Field values should be used for purposes of a geochemical model. Laboratory data should be considered as backup if field measurements experience QA problems. Accurate measurement of pH will require use of a probe designed specifically to measure pH in waters with low ionic strength (less than 200 µmho conductivity). The USGS NFM (Chapter 6, Section 6.4 pH) provides specific guidance for measurement of waters with low ionic strength. Waters with low ionic strength tend to cause sluggish, drifting responses that fail to stabilize. A variety of manufacturers make pH probes that are specifically designed for low-ionic strength waters. These sensors normally provide higher bleed rates at the reference junction that result in faster stabilization and increase accuracy and reproducibility. Low ionic strength buffers are also available that help response time and accu- racy even when used with standard pH probes. Temperature will also be an issue when measuring pH in stormwater sam- ples. Stormwater composite samples should be maintained at temperatures of 4–6°C due to refrigeration of icing of the samplers. The temperature of the subsample taken for pH measurements should be maintained at the composite sample temperature and the temperature recorded with the measurement. Measurement of solids in stormwater or receiving water has received a substantial amount of discussion with respect to both sampling and analytical methods. Peristaltic pumps have been shown to be limited sediment size fractions of 250 µm or less (Clark, Siu et al. 2009). Furthermore, the SSC method is generally recommended for more accurate quan- tification of suspended particulates but method is not strictly suited for composite sampling. The SSC method requires the whole sample be processed. Although the method is used, it is based upon processing a complete subsample from the composite. Selection of a method of measuring solids should consider overall objectives of the program and the known limitations of peristaltic sampling methods. Additional mea- sures such as particle size distributions may be necessary if the objectives include the need to understand partitioning among available sediment particles. Total alkalinity should be measured along with the three major components: carbonate alkalinity, bicarbonate alkalin- ity, and hydroxide alkalinity. Hardness as CaCO3 is another critical parameter since water quality criteria for many of the dissolved metals are a function of hardness. The most accurate measurement of hardness is based upon individual analysis of calcium (Ca) and magnesium (Mg), the two major cations comprising hardness, and then calculating total hardness. C h A P T E r 3

A-11 Temperature can be an important factor especially in regions that receive summer storm events and where ther- mal effects are important to measure. Temperature can also be important when using a geochemical model to determine speciation of dissolved metals. This would normally be mea- sured with use of a thermistor attached near the intake. 3.2 Major Ions As a minimum, selected constituents should include the major cations and anions. Recommended cations include calcium (Ca) and magnesium (Mg), typically measured to calculate hardness, and both sodium (Na) and potassium (K). In addition, concentrations of iron (Fe) and aluminum (Al) will be useful to support geochemical speciation and adsorp- tion modeling. Major anions should include sulfate (SO4) and chloride (Cl). The quality of the chemical analyses should be checked by calculation of the ion balance. Ion balance is calculated as the total cation concentration minus the total anion concentration divided by the total concentration of ions in the solution. Con- centrations are based upon the milliequivalents of each ion. 3.3 Organic Composition TOC measurements provide a surrogate measure of the organic matter content in water; however, its utility for esti- mating the extent of metal binding to solid organic matter or dissolved organic matter is questionable. Geochemical computer models such as model VI (Tipping, 1998) and the NICA-Donnan model (Kinniburgh et al., 1999) require esti- mation of the concentration of binding sites and the metal- ligand equilibrium constant. Ideally, these model constants are obtained by calibrating the model to data obtained by titrating concentrated organic matter into a solution containing the metal ion and measuring free metal concentration as a function of organic matter concentration DOC. However, these types of measurements are not routinely conducted. As a result, modelers often assume a range of assump- tions regarding the fraction of humic and fulvic acids and the fraction of inert organic matter present in a water. Recent data (Amery et al. 2008) suggests that one easily measurable parameter, SUVA, correlates well with the available fraction of organic matter. Therefore, we suggest that specific UV at 254 nm be measured to provide an assessment of the active fraction of fulvic acid in the sample.

A-12 C h A P T E r 4 This guidance is based on use of automated stormwater monitoring equipment to collect flow-weighted compos- ite samples of runoff for analysis of dissolved metals at lev- els relevant to receiving water quality criteria. Stormwater monitoring systems are commercially available from several sources. The two most common systems are manufactured by American Sigma and Isco. These systems were originally designed for waste water treatment plants where conditions are relatively constant. Manufacturers of these systems started modifying the design of the equipment in the 90s to provide functionality for delivering fixed volumes of water even when water levels were changing quickly. They then were integrated with a variety of datalogger/flow modules that provided bet- ter flow quantification and the ability to pace the samplers in response to changing flows. Monitoring equipment has continued to be improved by improved software, integration with rain gauges, and in situ water quality instrumentation and telecommunications. Providing reliable telecommunica- tions for real-time access to data and to provide command and control functionality has greatly improved efficiency and contributed to improved stormwater data. Most automated stormwater sampling stations incorporate peristaltic pumping systems for sample collection. Alternative systems (e.g. Manning Environmental, Sirco, etc.) are also avail- able that utilize vacuum pumps to draw water into a measuring chamber. These vacuum systems have some distinct advan- tages when it comes to sampling water containing high loads of suspended particulates and sampling situations requiring high vertical lift. Although the measuring chambers allow collection of very repeatable sample volumes, they are generally consid- ered difficult to configure for ultraclean sampling and have not been extensively used for stormwater monitoring applications. The two most common stormwater monitoring systems both use peristaltic pumping systems. When appropriate measures are taken, it has been demonstrated that these systems are capable of collecting blanks that are uncontaminated and high quality, reproducible data using detection limits appropriate to water quality criteria. In order to accomplish this, extreme care must be taken to avoid introduction of contaminants. Requirements include: • Assurance that all materials coming into contact with the samples are intrinsically low in trace metals and do not adsorb/absorb metals (Table 3); • Material coming into contact with the sample water are subjected to intensive cleaning using standardized proto- col and subjected to systematic blanking to demonstrate and document that blanking standards are met; • All cleaned sampling equipment and bottles are appropri- ately tracked so that blanking data can be associated with all components deployed in the field; • Samples are collected, processed, and transported, taking extraordinary care to avoid contamination from field per- sonnel or their gear; and • Laboratory analysis is conducted in a filtered air environ- ment using ultrapure reagents. Although the technical limitations of autosamplers are often cited, they still provide the most practical method for collecting representative samples of stormwater runoff for characterization of water quality and have been heavily uti- lized for this purpose for the past 20 years. The alternative, manual sampling, is generally not practical for collection of flow-weighted composite samples from a large number of sites or for sampling events that occur over an extended period of time. Despite the known drawbacks, autosamplers combined with accurate flow metering remain the most com- mon and appropriate tool for monitoring stormwater run- off. Understanding the limitations of autosamplers allows the user to minimize and compensate for those limitations. Among the most common limitations cited are the following: • Lack of ability to collect isokinetic samples • Poor representation of suspended particles and particle size distributions Equipment

A-13 Construction material for sampling equipment (does not apply to well casing) Target analyte(s) Material Description Inorganic Organic Plastics1 Fluorocarbon polymers2 (other varieties available for differing applications) Chemically inert for most analytes. (potential source of Fl) (Sorption of some organics.) Polypropylene Relatively inert for inorganic analytes. * Do not use. Polyethylene (linear) Relatively inert for inorganic analytes. * Do not use. Polyvinyl chloride (PVC) Relatively inert for inorganic analytes. * Do not use. Silicone Very porous. Relatively inert for most inorganic analytes. (potential source of Si) Do not use. Nylon Relatively inert for inorganic analytes. * Do not use. Exception: CFCs Metals Stainless steel, 316- grade (SS 316) SS 316—metal having the greatest corrosion resistance. Comes in various grades. Used for submersible pump casing.3 * (potential source of Cr, Ni, Fe, and possibly Mn and Mo, if corroded4) Do not use for surface-water sampling: equipment must have a plastic coating (does not apply to submersible pumps). * Do not use if corroded.4 Stainless steel, 304- grade (SS 304) Similar to SS 316, but less corrosion resistant. Do not use. * Do not use if corroded.4 Other metals: brass, iron, copper, aluminum, and galvanized and carbon steels Refrigeration-grade copper or aluminum tubing are used routinely for collection of3H/3He, CFC, and SF-6 samples. Do not use (except as noted for isotopes). * Routinely used for CFCs. Do not use if corroded. Glass Glass, borosilicate (laboratory grade) Ceramic Relatively inert. Potential sorption of analytes. (glass is potential source of B and Si) * * Generally appropriate for use shown; Fl, fluoride; Si, silica; Cr, chromium; Ni, nickel; Fe, iron; Mn, manganese; Mo, molybdenum; 3H/3He, tritium/helium-3; CFC, chlorofluorocarbon; SF-6, sulfur hexafluoride; B, boron] 1 Plastics used in connection with inorganic trace-element sampling must be uncolored or white (Horowitz and others, 1994). 2 Fluorocarbon polymers include materials such as Teflon®, Kynar®, and Tefzel® that are relatively inert for sampling inorganic or organic analytes. 3 Most submersible sampling pumps have stainless steel components. One can minimize effects on inorganics samples by using fluorocarbon polymers in construction of sample-wetted components (for example, for a bladder, stator, impeller) to the extent possible. 4 Corroded/weathered surfaces are active sorption sites for organic compounds and can leach trace elements. Reproduced from USGS NFM -Table 2-1. General guidelines for selecting equipment on the basis of construction material and target analyte(s) (From USGS NFM, Table 2-1) 3 Table 3. Appropriate materials for environmental samples.

A-14 • Sample line velocities that decrease with head differentials (ability to meet minimum recommended line velocity of 2.0 ft/s diminishes) without the use of auxiliary pumps • 100-foot line limitation • Sample volume repeatability Most of these limitations can be minimized by incorpora- tion of these considerations in the site selection process and when installing the equipment. In small drainages, a primary control device such as a flume or weir is necessary to obtain accurate measurement of flow. Flumes are generally pre- ferred since they are not as prone to trapping sediments as weirs. Conditions necessary for collecting accurate flow mea- surements are not conducive to collection of representative water samplers. When designing the installation, consider- ation should be given to providing a hydraulic jump near the primary control device so that turbulent mixing is encour- aged. This can be accomplished by installing a sampling port downstream of the flume. The intake should be set on the bottom of the sampling port. The sampling port should pro- vide a minimal retention volume (depth of approximately one inch) to allow for the intake to be covered during very low flow conditions but minimize potential trapping of par- ticulates. In most cases, the intake can and should be placed on the bottom of the sampling point. The sampling equipment should be placed as close as pos- sible to the sampling point without creating a safety issue. Safety should always be the primary consideration but sites are preferred that provide safe access combined with the abil- ity to minimize lengths of the intake hose and head differen- tials. Whenever possible, intake hose should be limited to 10 to 20 feet in total length with a head differential of no more than 4 to 6 feet. Laboratory tests with common autosamplers (Clark, Siu et al. 2009) concluded that the peristaltic pumps used in these systems can provide accurate measures of sol- ids that are less than 250 microns in diameter with a specific gravity of 2.65. The ability to effectively represent particles less than 250 microns started to degrade when head differen- tials exceeded 2.5 meters (~8.3 feet). 4.1 Sampler Intake Strainer, Intake Tubing and Flexible Pump Tubing Intake strainers are used to prevent small rocks and debris from being drawn into the intake tubing and causing block- ages or damage to the pump and peristaltic pump tubing. Strainers typically come in a limited number of sizes and are constructed of a combination of Teflon® and 316 stain- less or simply stainless steel. The low profile version is typi- cally preferred to provide greater ability to sample shallow flows. Intake strainers constructed of stainless steel should be coated with a fluoropolymer coating if they are to be used for sampling dissolved metals. If the stainless steel intake is not coated, the strainer should never be subjected to cleaning with acids. Cleaning should be limited to warm tap water, laboratory detergents and MilliQ water rinses. Tubing comprised of 100% fluorinated ethylene propylene (FEP) should be used for the intake tubing. Several alternative fluoropolymer products are available but 3/8″ ID solid FEP tubing has the chemical characteristics suitable for sampling metals (and organics) at low levels and appropriate physical characteristics. The rigidity of FEP tubing provides resistance to collapse at high head differentials but still is manageable for tight configurations. The rigidity of the tubing can also be a problem if not installed correctly. The length of the tubing should be tai- lored to the site to minimize the overall length and provide a smooth path to the pump. Anti-kink should be used whenever the tubing passes through a point where it would be suscep- tible to bend. If the tubing develops a kink, the entire intake line will require replacement (at a substantial cost!). Tubing constructed of polyethylene with a Teflon® liner is available and was at one time recommended by compa- nies supplying stormwater monitoring equipment, but these products should strictly be avoided. Zinc contamination can occur if water gets between the liner and the polyethylene tubing. The liner also can be uneven leading to collapse of the lightweight liner and will often compress and distort when attaching fittings. The peristaltic hose used in autosamplers is a medical- grade silicon product. The specifications for the peristaltic pump hoses used in these samplers are unique to the samplers. It is very important that the hose specified and provided by the manufacturers of the autosamplers be used. Minor differ- ences in the peristaltic hose can cause major deterioration in performance of the samplers. Use of generic peristaltic pump hose from other sources can lead to problems with the abil- ity to calibrate the samplers and maintain intake velocities of greater than 2.5 feet per second with higher lift requirements. The peristaltic hose is connected to the FEP tubing and fed through the pump head leaving the minimum amount necessary to feed the peristaltic pump hose into the top of the composite bottle. The composite container should always have a lid to prevent dust from settling in the container. A number of issues must be considered when installing the intake strainer and hose. Many of the key issues are addressed in the Caltrans Stormwater Monitoring Guidance and the National Highway Runoff Data and Methodology Synthe- sis (Breault and Granato 2000; Granato, Zenone et al. 2003). Use of autosamplers with a relatively constant intake flow rate and a fixed input point have inherent problems with regards to collecting samples with particle size distributions that are truly representative of the discharge (Edwards and Glysson

A-15 1999; Smith and Granato 2010). A number of approaches have been suggested for minimizing the bias introduced by autosamplers but each requires special consideration of site specific conditions. Edwards and Glysson (1999; p. 27) provide a list of important factors that should be addressed when positioning an intake when using autosamplers. Intake strainer are often located near the invert of small pipes or channels but this can often lead to biasing the samples towards coarser sediment, burial of the intake within bedload material, or clogging by large particles. In general, intakes should be oriented in a downstream direction since field testing (Edwards and Glysson 1999) sug- gests that this allows for more representative collection of coarser particles and limits the tendency for debris to foul the intake. Whenever possible, the intake should be located in a position where active mixing occurs to enhance suspension of the full range of particles being transported. Other options have been proposed for improving representativeness of sam- ples collected with autosamplers. These include installation of static mixers (Figure 2) such as developed by USGS (Smith 2002; Smith and Granato 2010), mechanical float devices to allow sampling at a depth of roughly 0.6 times the depth of the flow (Eads and Thomas 1983; Gettel et al. 2011), or the Depth-Integrated Sample Arm (DISA) recently tested by Selbig and Bannerman (2011). The extensive efforts made to develop intake systems that provide more representative samples illus- trate the importance of this issue. Although some of these pro- posed solutions are complex, they all share a common goal of seeking to improve the ability of autosamplers to obtain rep- resentative samples. The ultimate approach to mounting the intake should be based upon site conditions and the simplest, most reliable method to optimize representative sampling. If trash and debris are common at the site, a stainless steel cage may need to be manufactured to deflect plastic bags or heavy objects moving through the channel. If such protection is required, the cage should be constructed with 316 grade stainless steel (Table 3) and preferably coated with fluoropoly- mer coating. The intake hose and strainer must be attached in a manner that provides a high probability of remaining in place during high flows. Use of conduit to protect the intake hose and allow it to be securely attached to the channel is recom- mended for most sites. The intake hose also must be installed without low points or loops so that water is not retained. 4.2 Composite Containers A composite container should be used that can be dem- onstrated to be free of metals of interest at the desired levels (U.S. EPA 1996). Containers constructed of fluoropolymers (FEP, PTFE), conventional or linear polyethylene, polycar- bonate, polysulfone, polypropylene, or ultrapure quartz are considered optimal but borosilicate glass has been shown to meet sampling objectives for trace metals measured at lim- its appropriate to EPA water quality criteria. High capacity borosilicate media bottles (20 liters or 5 gallons) are preferred for storm monitoring since they can be cleaned and suitably blanked for analysis of both metals and organic compounds. The transparency of the bottles is also a useful feature when subsampling and cleaning the containers for reuse. These large media bottles are designed for stoppers and thus do not come with lids. Suitable closure mechanisms must be fabricated for use during sampling, transport, and storage of clean bottles. The preferred closure mechanism is a Teflon® stopper fitted with a Viton O-ring (23/8″ - ID × 2¾″- OD) that seals the lid against the media bottle. A polypropylene clamp (Figure 3) is used to seal the Teflon® stopper and O-ring to the rim of the composite sample bottle. Two polypropylene bolts with wing-nuts are used to maintain pressure on the seal or to assist in removal of the lid. Every composite bottle requires one solid lid for use in pro- tecting the bottle during storage and transport. A minimum Figure 2. Static mixer (from Smith and Granato (2010). Figure 3. Composite bottle with label and installed tubing.

A-16 of one Teflon® stopper should be available for each monitor- ing site during storm events. Each field sampling crew should have additional stoppers with holes (“sampling stopper”) that would be available if a sampling stopper is accidentally contaminated during bottle changes or original installations. The holes in the sampling stoppers should be minimally larger than the external diameter of the peristaltic hose. If a tight fit exists, the pressure created when water is pumped into the bottle will cause the hose to be ejected and the sam- pling event will need to be abandoned. Transporting composite bottles is best accomplished by use of 10-gallon Brute® containers to both protect them from breakage and simplify handling. They also provide additional capacity for ice while transporting full bottles to the labora- tory or subsampling site. Bottle bags (Figure 4) are also useful in allowing full bottles to be handled easier and reduce the need to contact the bottles near the neck. They are important for both minimizing the need to handle the neck of the bottle and are also an important health and safety issue. The empty bottles weigh 15 pounds and they hold another 40 pounds of water when full. These can be very slippery and difficult to handle when removing them from the autosamplers. Bags can be easily fabricated out of square- mesh nylon netting with nylon straps for handles. Use of bottle bags allows two people to lift a full bottle out of the ice in the autosampler and place it in a Brute® container. Whether empty or full, suitable restraints should be provided whenever the 20-L composite bottles and Brute® containers are being transported. 4.3 Flow Monitoring Collection of flow-weighted stormwater samplers requires the ability to accurately measure flow over the full range of conditions that occur at the monitoring site. The ability to accurately measure flow should be an important aspect of site selection. Hydraulic characteristics necessary to allow for accurate flow measurement include a relatively straight and uniform length of pipe or channel without major confluences or other features that would disrupt establishment of uniform flow conditions. The actual measurement site should be located sufficiently downstream from inflows to the drainage system to achieve well-mixed conditions across the channel. The flow sensor and sample collection inlet should be placed a minimum of five pipe diameters upstream and 10 pipe diameters down- stream of any confluence to minimize turbulence and ensure well-mixed flow. The Isco Open Channel Flow Measurement Handbook (Walkowiak 2008) is an invaluable resource to assist in selection of the most appropriate approach for flow measure- ments and information on the constraints of each method. Use of a primary control device is recommended for most sites. Flumes are generally preferable to weirs as they tend to be self-cleaning and are less subject to the errors due to approach velocities. When properly sized and installed, they can provide accuracy of +/-3%. Flumes generally require a still- ing well and communication port connecting to the flume at the measurement point. Bubblers are recommended for mea- surement of water levels within the stilling wells as they provide stable, accurate stage measurements for calculation of flow. Calibration checks are essential to assure that flow mea- surements are accurate (Maheepala et al. 2001; Church, Granato and Owens 2003; Smith and Granato 2010). This is particularly important if flow measurements do not utilize a primary control device. Calibration of flow meters in small pipes (less than 12-16 inches) can typically be performed by using a portable water tank and then pumping water through the system with centrifugal pumps over a range of conditions expected. Flow data from the meter being calibrated are com- pared with paired flow measurements taken at a downstream location using a volumetric weir or secondary flow meter. For larger pipes, a nearby fire hydrant fitted with a certified flow meter and totalizer may be used as a water source, preferably in association with routine servicing of the hydrant. Before conducting any calibration checks involving discharges to the storm drain, one should consult with local authorities. If discharges are not permitted, calibrations may need to be performed during an actual storm event using dilute solutions of fluorescent dyes such as Rhodamine WT and a fluorometer. Figure 4. Composite bottle showing bottle bag used for transport and lifting.

A-17 4.4 Rainfall Gauges Installation of electronic tipping bucket rain gauges is highly recommended for monitoring of smaller drainages. Use of a localized rain gauge provides better representation of conditions at the site. This is less important at larger water- sheds best characterized by a network of rain gauges. The basic requirements for collecting, documenting, and report- ing precipitation data are summarized in work conducted under the National Highway Runoff Data and Methodology Synthesis (Church, Granato et al. 2003). A variety of quality instruments are available but all require substantial maintenance to ensure maintenance of high data quality. Typical gauges will have 8-inch diameter cones, pro- vide 1 tip per 0.01″ of rain and have an accuracy of ± 2% up to 2″/hr. The accuracy of tipping bucket rain gauges can be impacted by very intense rainfall events but errors are more commonly due to poor installations. Proper installation and maintenance of the rain gauge is as important as gauge prox- imity to the monitoring site. Proper installation requires that the equipment is level, clear of trees and other major obstruc- tions, and mounted firmly to avoid introduction of artifacts due to vibration. Monthly maintenance should be conducted to assure that no obstructions have developed due to birds, insects and spiders, or other debris. Supporting data are necessary to document the accuracy and representativeness of the precipitation data. Basic information needs include: • Location relative to the monitored drainage area • Length of record • Recording frequency • Equipment (model, specifications, and measuring methods) • Records of maintenance visits and calibration records • A data quality assessment • Availability and access to final data records Continuous data records should be maintained through- out the wet season with data being output and recorded for each tip of the bucket. Data for storm events should include total precipitation, hyetographs, maximum 5-minute inten- sity, the length of the antecedent dry period, and total pre- cipitation associated with the antecedent event. 4.5 In Situ Sensors Both standard commercial stormwater monitoring equip- ment and custom systems designed around datalogger/control modules such as those made by Campbell Scientific are now capable of incorporating an array of additional water quality sensors using an SDI-12 interface. The wide variety of SDI-12 compatible stage, flow, and water quality probes that are available can allow collection of important ancillary data such as temperature and conductivity. Continuous records of temperature and conductivity can provide important infor- mation for general characterization or as input for modeling equilibrium conditions between particulate and dissolved forms. Most monitoring equipment has the ability to trigger sampling based upon specified water quality conditions. 4.6 Power Stormwater monitoring equipment can generally be pow- ered by battery or standard 120VAC. Standard systems come with built-in batteries but use of external, sealed deep-cycle marine batteries are recommended if the equipment must run under battery power. Even systems with access to 120VAC should be equipped with batteries that can provide backup power in case of power outages during an event. All batteries should be placed in plastic marine battery cases to isolate the terminals and wiring. Systems equipped with telecommuni- cation packages should have a second battery that is obligated to telecommunications equipment and, if conditions are suit- able, equipped with a solar panel to assure that a full charge is available when needed for a storm event. 4.7 Telecommunication for System Command/Control and Data Access The ability to remotely communicate with the monitoring equipment allows for more efficient and representative sam- pling. Remote communication facilitates preparation of sta- tions for storm events and making last minute adjustments to sampling criteria based upon the most recent forecasts. Com- munication with the sites also reduces the number of field visits by monitoring personnel. Remote two-way communica- tion with monitoring sites allows the project manager (storm control) to make informed decisions during the storm as to the best allocations of human resources among sampling sites. By remotely monitoring the status of each monitoring site, the manager can more accurately estimate when composite bottles will fill and direct field crews to the site to avoid disruptions in the sampling. Real time access to flow, sampling, and rain- fall data also provides important information for determin- ing when sampling should be terminated and crews directed to collect and process the samples. Increases in both efficiency and sample quality make two-way communication with moni- toring stations a necessity for most monitoring programs.

A-18 Sample Collection and Handling Procedures Monitoring stations typically include a flow monitoring component that incorporates instrumentation and software for measuring and recording flow. The flow meters are com- prised of a datalogger/control module to allow storage and processing of the flow data and communication with a com- posite sampling device. When collecting flow-weighted composite samples, the flow meter is programmed to send a pulse to the sampler each time a specified volume of water has passed the site. Upon receiving a pulse from the flow meter, the autosampler initi- ates a process designed to collect a fixed volume of water (ali- quot) and discharge it into a composite bottle. Autosamplers are often programmed to go through a purge cycle before and after collecting a sample aliquot. This helps assure that the line remains clear and water does not get trapped in the line and incorporated into the sample aliquot. It also assists in clearing debris that might collect around or within the intake strainer. The time required to complete this cycle under worst case conditions (maximum head differentials) plays a critical role in determining sampling rates. Elimination of the post purge process is recommended to reduce cycling time and improve the overall ability to obtain representative samples of runoff from smaller drainages. Cycling times will increase with increasing lengths of intake hose, higher head differentials, and as sampling vol- umes are increased for the sampling aliquot. The cycling time defines the minimum time interval that can be maintained between sampling pulses and still maintain a flow-weighted composite sample. If pulses exceed this rate for a substantial period of time, the sampling pulses become “stacked” in the autosampler’s memory and sampling will continue at this maximum rate until such time that the sampler catches up with declining storm volumes. Unfortunately, stacking of sampling pulses is a common but rarely recognized occurrence in many stormwater moni- toring studies that utilize flow-weighted composited sam- pling strategies. Furthermore, the impacts of stacking on the data are often impossible to assess. Brief occurrences of stack- ing that recover within a few cycles would not be expected to bias the data but extended periods of stacked sampling com- mands that continue to execute well after major flows have subsided most certainly impact the data. This latter condi- tion is most common when multiple (four bottle) composite bottles are used to sample runoff in small drainages. This sampling configuration has been used as a strategy to provide field duplicates and/or additional water samples for QA/QC. Cycling times associated with this sampling configu- ration can easily run up to five minutes when the full purge- sample-purge cycle is performed for each bottle. Although samplers may be programmed to eliminate purges between bottles, this modification further impairs the effectiveness in sampling suspended sediments and flow conditions may still change radically by completion of the sampling routine. It is extremely common for sites using this configuration to exhibit “stacking” and miss aliquots intended for the third or fourth bottles due to rapidly changing water levels. Small, highly impervious drainages associated with high- way and ultra-urban areas are particularly challenging. These types of sites tend to have “flashy” flows that are highly respon- sive to changes in rainfall intensity. Monitoring strategies for these types of sites must take into consideration the range of flow conditions and the rates at which changes occur at these types of sites. Monitoring under these types of conditions can often benefit from use of smaller, more frequent sample aliquots. This improves the chances of sample aliquots being taken near the time and under the flow conditions respon- sible for initiating the sampling sequence. However, there is a point of diminishing returns as the size of the aliquot decreases; decreases in the system cycling time become less significant and controlled primarily by the length of the intake hose and the head differential. With the exception of extremely small drainages, it is generally recommended that 200 ml be used as the lower limit for aliquot volumes. A volume of 1,000 ml is recommended as the upper limit for aliquot size in most situations. Sampling of discharges from a pump is an exception since water is often discharged C h A P T E r 5

A-19 at very high rates for brief periods of time. More frequently, aliquot volumes are set around 500 ml. With aliquot volumes set to 500 ml, a 20-liter composite bottle provides capacity for 40 aliquots without changing bottles. 5.1 Pacing the Sampler Designing a sampling strategy for obtaining a representa- tive composite sample from any given storm event requires consideration of the following key factors: • Total sample volume required - fixed • Capacity of the composite bottle - fixed • Volume of individual aliquots - fixed • Quantitative precipitation forecast (QPF) for the event • Maximum sustained rainfall intensity - fixed • Drainage area - fixed • Cycling Time - fixed A sampling program designed to focus on assessment of metals and ancillary parameters will require less than four liters of water. The composite containers provide a total capacity of 20 liters. Given these parameters, the sampling strategy would be based upon providing twice the required sample volume for the forecasted storm event and would still provide capacity for effectively sampling runoff that was up to 2.5 times greater than expected. The flow volume per sample (the amount of flow that passes the sampling point between each aliquot collected) must be programmed into the flow meter in proportion to the predicted rainfall amount for each storm event, to set the sample pacing so as to fill the composite bottle(s) at an appropriate rate. Calculation of the flow volume per sample (Vs) is per- formed using the predicted rainfall amount (QPF), the known or estimated drainage area (A), and the compos- ite runoff coefficient (C) for the area monitored to calcu- late the expected runoff flow volume for the storm event. The runoff coefficient for a specific drainage area is defined as the fraction of total precipitation volume delivered to the area that ends up as runoff at the point of discharge. Flow volume per sample can be determined using the formulas presented below: ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) = × × × × = = × × × × inches 1 ft 12 inches acres 43,560 sq. ft 1 acre min inches hr 1 hr 60 min 1 ft 12 inches acres 43,560 sq. ft 1 acre Vr cf QPF A C Vs cf Vr cf CSA Qm cf PIm A Vr = estimated runoff volume in cubic feet QPF = Quantitative Precipitation Forecast in inches A = drainage area in acres C = runoff coefficient Vs = flow volume per aliquot in cubic feet CSA = number of composite sample aliquots required for complete composite PIm = maximum 5-minute precipitation intensity Qm = maximum flow rate in cubic feet per minute Successful collection of representative samples from a small, highly impervious drainage requires a sampling strat- egy that optimizes the probability of obtaining the necessary sample volumes and aliquot counts even when actual rainfall is found to substantially differ from predictions. Use of high capacity composite containers is an important part of the sampling strategy that enables effective sampling even when monitoring events that significantly differ from weather fore- casts. Determination of appropriate settings requires knowl- edge of the maximum time for the autosampler to go through a full sampling cycle (Ts), the minimum volumes necessary to complete all analyses (Va), and the volume of each aliquot produced by the aliquot. The following calculations illustrate an iterative approach used to determine initial setting. Example: QPF = 0.55 inches = QPF (quantity of precipitation forecast) A (drainage area) = 4.5 acres C (runoff coefficient) = 0.9 CSA (target number of composite sample aliquots for sample composite) = 20 Va = 4,000 ml = minimum volume nec- essary for all analyses Vc = 20,000 ml = composite bottle capacity Ts = 0.36 minutes based upon a Vq of 400 ml PIm = 2.2 inches/hour (based upon 5-minute rainfall intensity) ( ) ( ) = × × × × = = × = = = = × × × × × = 0.55 inches 1ft 12 inches 4.5 acres 0.9 43,560 sq. ft 1 acre 8086 cf 4,000 2 ml 20 400 ml per aliquot 8086 cf 20 aliquots 404 cf per aliquot 2.2 inches 1 hr 1 hr 60 min 1 ft 12 inches 4.5 acres 43,560 sq. ft 1 acre 0.9 540 cf per min Vr Vq Vs Qm

A-20 Ts = Maximum time (minutes) for sampler to go through the full sampling cycle. Va = Minimum volume (milliliters) necessary to complete all analyses. Vq = Aliquot volumes (milliliters), 200 to 1000 ml In this example, we have applied a margin of safety by dou- bling the total volume required for conducting all analyses. Doubling of the aliquot volume (Vq) provides the ability to successfully capture storm events ranging from ½ to nearly 2½ times the predicted rainfall volumes without a bottle change. A margin of safety can be applied in several different ways but one must always consider impacts on other sam- pling constraints. Any adjustments in the volume used for sampling aliquots (Vq) will require examination of impacts on cycling time (Ts) for the autosampler. A rough estimate of the potential maximum sustained flow rate (Qm) from the drainage area will help to reduce the pos- sibility of setting sampling rates (Vs) that commonly exceed the time required for the monitoring system to go through the full sampling process (Ts) before receiving a pulse to trig- ger initiation of another sample aliquot. This example uses a maximum sustained rainfall intensity of 2.2 inches/hour. This was based upon rainfall records from Long Beach, California. In this region, 97.5% of the storm events monitored over a 10-year period had maximum 5-minute rainfall intensities that were less than this level. If it is assumed that the site is estab- lished with an 8-foot length of FEP tubing and the required head is 40 inches, a sampling cycle (Ts) could be completed every 0.36 minutes (22 seconds) if a post purge is not used (Figure 5). The sampling flow volume (Vs) used for this hypothetical site should always be greater than 200 cubic feet in order to handle the estimated maximum sustained flow (Qm) of 540 cubic feet per minute. This is roughly half of the proposed flow volume (Vs) of 404 cf and thus would provide ample cushion to increase the CSA without the risk of stacking sample pulses. Based upon these initial calculations, the total number of aliquots (CSA) could be increased to 30. Adjusting the vol- ume of each aliquot (Vq) to 300 ml would maintain the same projected sample volumes. This would have a minor effect of reducing cycling time (Ts) from 22 to 20 seconds and the flow volume per aliquot (Vs) would 270 cubic feet per aliquot. Reducing the volume of each aliquot to 300 ml would reduce the lower limit for Vs to 180 cubic feet. This adjustment would still allow a comfortable margin of safety to handle unexpected high intensity rainfall and provide for approximately 15 ali- quots if the storm yielded half of the expected runoff. Once a sampling strategy is established, most parameters such as the volumes used for each aliquot (Vq) and target number of aliquots (CSA) should remain constant. The only adjustments would be based upon predicted rainfall (QPF) for each event. After monitoring the first few storm events, the calculations for predicted runoff volume should be reviewed and adjusted as necessary to provide the optimal settings for assuring representative samples. This review should include an assessment of measured rainfall intensity and resulting flows to verify that settings used for the monitoring station are within the capacity of the equipment to maintain flow- weighted pacing. 5.2 Percent Capture Percent storm capture (Percent Capture) is the percentage of the total event flow effectively represented in a composite sample. This is determined by the following formula: = × ×Percent Capture total # of aliquots Total Run of Volume 100 Vr f This calculation assumes that the storm volume repre- sented by each aliquot is constant and not impacted by stack- ing of sampling pulses. The actual data should be reviewed to verify that this assumption is correct. Percent Capture will rarely equal 100%. A number of fac- tors contribute to small reductions in Percent Capture. Fre- quent periods of low flows can lead to reductions in percent capture since autosamplers require suspension of sampling when water levels go below a minimum level. Flow mea- surements with primary control devices allow flow to be monitored below typical sampling limits. With small, highly responsive watersheds, water levels may frequently drop below levels that can be sampled. This condition also occurs at the end of the event as flows subside. The most significant losses in percent capture tend to occur when sampling requires that composite bottles be replaced during a storm event. This can happen due to the need for high sample volumes for other analytical or toxi- cological purposes or as the result of substantial deviations Figure 5. Example of autosampler cycling time with and without a post purge.

A-21 from predicted rainfall. Minimizing these losses requires the ability to remotely monitor progress at the site so that sam- pling crews can be alerted and directed to the site before the last aliquot is pumped. Crews should remove the composite bottle immediately after the last sample is taken and place a new bottle in the sampler before another sample is initiated. At high flows, this can be a matter of just a few minutes. In the real world, factors that contribute to the need for composite bottles to be changed at one site typically have comparable impacts at other monitoring sites. Intense shower activity also limits the ability of field crews to safely get through traffic to service the site. If a program typically requires use of multiple composite bottles, one should plan on using more field crews to avoid this problem. 5.3 Peak Capture Successful sampling conducted throughout periods of high flows is considered important for sampling to be considered representative. This is generally reported as a “yes” or “no” situation, but minor loss of sampling efficacy around periods of peak flow should not be considered important if all other quality measures are met. Periods of high flows cause ele- vated sampling frequency, which may cause composite bottle capacities to be exceeded faster than anticipated. Maintaining effective sampling through these time periods will generally result in the capture of high percentages of the total storm flows. If the monitoring station was not effectively sampling during peak flow, the sampling data should be examined and professional judgment applied to determine whether data should be qualified. 5.4 Evaluation of Representative Sampling A sample is considered representative if (1) the sampling aliquots were taken in direct proportion to the flow, (2) a min- imum number of aliquots are obtained, and (3) the sample represents runoff for the entire duration of the event. Each of these issues requires careful examination and documentation in order to assess representativeness of samples collected by the autosampler. Although some guidelines (Table 4) have been suggested for addressing representativeness (Strecker, Mayo et al. 2001; Caltrans 2003), standards have not been clearly established to determine acceptance or rejection of the data set. Documentation included with the reports should include the standards used to assess representativeness of the samples and include a discussion of the potential impacts of any excursions on use of the data. Recommended criteria are discussed below. The representativeness of sampling tends to improve with increased sampling density. Using overly aggressive sampler settings with the intent of providing high sampling density also increases the chances of sample pulses becoming backed up. Evidence of possible stacking of sampling pulses is evalu- ated by examining the times that each aliquot is taken, the volume represented by that aliquot, and the total volume of stormwater runoff that has occurred up to the point that the sample was taken. Stacking becomes evident when the timing between samples becomes constant despite changing flow rates. This will occur if flows exceed the ability of the equipment to complete the sampling cycle before receiving a pulse indicating that another sampling sequence should be initiated. If flow exceeds this limit for an extended time, the sampler will just continue to take samples until the system catches up with the total volume of stormwater runoff. Brief and limited excursions do not typically have a severe impact on representativeness. When three to five sampling pulses become backed up on the autosampler, there is a substantial increase in the potential to introduce bias. Such cases must be critically examined and evaluated to determine whether the data warrant rejection or qualification based upon the magnitude of these excursions. Overly conservative setting can result in an insufficient number of aliquots being collected over the duration of the storm event for the composite sample to be considered rep- resentative of the storm event. Guidance as to the minimum number of aliquots needed to be considered representative Total Event Precipitation Strecker et al. (2001)1 Caltrans (2003)2 Minimum Acceptable Number of Aliquots Percent Capture Requirement Minimum Acceptable Number of Aliquots Percent Capture Requirement 0-0.25” 12-16 (10-20% of the mean) >60% (note events with <80%) 6 85 0.25-0.5” 8 80 0.5-1” 10 80 >1” 12 75 1 Strecker et al. (2001). Guidance Manual for Monitoring Highway Runoff Water Quality. FHWA-EP-01-022 2 Caltrans (2003). Comprehensive Protocols Guidance Manual: Stormwater Quality Monitoring Protocols. CTSW-RT-03-105.51.42 Table 4. Criteria used to evaluate representativeness of stormwater runoff samples.

A-22 is limited. Strecker, Mayo et al. (2001) conducted a sampling exercise that suggested that 12–16 aliquots were necessary for samples to be considered representative. Twelve aliquots provided resulted in estimates within 20% of the mean. Increas- ing the aliquot count to 16 samples provided estimates within 10% of the mean. Caltrans (2003) took a different approach suggesting that the minimum acceptable number of aliquots should be considered to vary with the magnitude of the storm event. Acceptable aliquot counts ranged from six for an event with under 0.25 inches of rainfall to 12 for an event exceeding 1 inch of rain. Given the variability that can exist within a single storm event, we recommend using a lower limit of 12 aliquots. Unusual circumstances such as a monitoring event associated with extremely small rainfall events (lower than 0.15 inches) may warrant acceptance of lower numbers of aliquots if all other criteria are acceptable. A Percent Capture of greater than 70 to 80% is generally considered the minimum to be considered as an acceptable sample (Table 4). Opinions differ as to what levels represent a significant loss of storm coverage. If a sampling event results in a Percent Capture of less than 70%, the specific circum- stances causing the reduced coverage should be examined to determine the cause of the reduction. Professional judgment should be used to determine whether the data from the event can be used to address the project objectives. Data should be appropriately qualified in cases where storm coverage is less than this limit and it is determined that data are still useful for the purposes of the program. The average Percent Cap- ture for events at most sites should be at least 85–90%. 5.5 Sample Splitting The process of sample splitting can be a major source of error. Errors are more likely to be introduced when whole stormwater samples contain suspended sediments with par- ticles larger than 250 µm. In addition it is unlikely that particle sizes greater than 250 µm will be effectively represented in samples taken with autosamplers (Clark et al. 2009). The two methods recommended in the USGS National Field Manual for Collection of Water-Quality Data (http:// pubs.water.usgs.gov/twri9A) include 14-L churns (plastic or fluoropolymer) and Dekaport cone splitters (Figure 6 and Figure 7). Each method has advantages and disadvantages as noted in Table 5. Many of the most significant disadvantages can be mitigated by having pre-cleaned and packaged split- ting equipment available for each site. This may be practical for some programs if metals are the only key analytes and the 14-L plastic USGS modified churns (~$600–$800) can be uti- lized; however, few programs can bear the cost of 14-L fluoro- polymer churns, which currently cost approximately $6000 each with disposable supplies. The Dekaport fluoropolymer cone splitter can be used to (1) split samples that are to be analyzed for both organic and inorganic compounds and (2) is not constrained by an upper limit on volume. The price point of the Dekaport splitter (~$1200) also makes it more reasonable to have precleaned equipment for each site, but field cleaning procedures specified by USGS (1997) have been shown to be effective in producing blanks at appropriate levels. Proper use (From USGS NFM, Chapter A2. Selection of Sampling Equipment) A. 14 L Plastic Churn B. 14 L Fluoropolymer Churn Figure 6. Churns used for splitting whole stormwater samples. (From USGS NFM and Capel and Larsen 1996) Figure 7. Cone splitter.

A-23 (Based upon and modified from USGS NFM Table 2-6) Splitter Advantages Limitations Fluoropolymer churn splitter Suitable for both inorganic and non-volatile organic analyses. Simple to operate. Easy to clean. No modification of the splitter design is necessary. Can be used to split samples with particle sizes 250 µm and suspended-sediment concentrations 1,000 mg/L, but splitting accuracy becomes unacceptable for particle sizes >250 µm and suspended-sediment concentrations >1,000 mg/L. Sample volumes less than 4 L or greater than 13 L cannot be split for whole-water subsamples from this 14-L churn. If pre-cleaned equipment is not available for each site, the churn requires field decontamination and additional blanking. Plastic (non- fluorocarbon polymer) churn splitter Simple to operate. Easy to clean. Plastic (nonfluorocarbon polymer) churn splitters must not be used to composite samples for determination of organic compounds. Can be used to split samples with particle sizes 250 µm and suspended-sediment concentrations 1,000 mg/L, but splitting accuracy becomes unacceptable for particle sizes >250 µm and suspended-sediment concentrations >1,000 mg/L. When using the 14-L churn, sample volumes that total less than 4 L or greater than 13 L cannot be split for whole-water subsamples. Requires a modified spigot and construction of a funnel assembly. If pre-cleaned equipment is not available for each site, the churn requires field decontamination and additional blanking. Fluorocarbon- polymer cone splitter Can be used on samples with suspended- sediment concentrations from 0 to 10,000 mg/L. Samples containing sediment particles ranging in size from very fine clay and silt (1 to 10 µm) to sand-size particles (250 µm) can be split. Samples as small as 250 mL can be split into 10 equal subsamples. Samples greater than 13 L can be processed. Suitable for samples to be analyzed for both inorganic and non-volatile organic analyses. The FEP distributor tubes can be pre-cleaned and replaced for each sample to reduce field decontamination efforts. Accuracy of the volume equivalents must be verified before using a new or modified cone splitter. Splitter is awkward to operate and clean in the field. If pre-cleaned equipment is not available for each site, the cone splitter requires field decontamination and additional blanking. Sample is vulnerable to contamination from atmospheric sources or from improper operation. Splitting accuracy for sediment particles >250 µm or sediment concentrations >10,000 mg/L must be quantified by the user. The cone splitter must be level for proper operation. Magnetic Stir Bar and Autosampler Pump All equipment can be pre-cleaned, no field decontamination needed. Minimizes contact with additional surfaces and secondary containers. Very adaptable to processing large numbers of samples in limited time. Wet sieving of whole storm samples through 64 µm prior to splitting would make reproducibility acceptable. Ineffective with coarse sediments (>150-200 µm). Accuracy and reproducibility with significant sediment loads must be verified. Accuracy and reproducibility when performed by different personnel must be verified. [L, liter; mg/L, μm, micrometer; milligrams per liter; mL, milliliters; >, greater than; ≤, less than or equal to.] Table 5. Advantages and disadvantages of splitting equipment and procedures.

A-24 of churns and cone splitters are well described in the USGS National Field Manual (various dates). Selbig, Bannerman, and Bowman (2007) tested an alterna- tive approach to processing samples. Whole storm samples were initially filtered through 63 µm nylon filters to remove all particles in the sand fraction. The remaining storm sam- ples were then split and subsampled into standard 250 ml containers for analysis of trace metals. They further modified the procedure by digesting the entire 250 ml sample rather than using the standard procedure of withdrawing an aliquot of the 250 ml sample for digestion. With only the silt and clay fraction remaining in the storm- water composite sample, alternative splitting methods such as used by Gossett and Schiff (2010) for developing subsamples for a large interlaboratory comparison study. Sample split- ting was accomplished by the use of a large-capacity stirrer and a 2- to 3-inch, pre-cleaned, Teflon-coated stir bar (larger stirbars can be used for larger volume containers) placed in the bottom of 20-L borosilicate media bottle containing the stormwater composite. Stirring speed was controlled to avoid creation of a large vortex within the composite bottle that can actually concentrate heavier particles. The final speed setting was based on a visual assessment of the most even mixing throughout the composite bottle. Sub-sampling from the homogenized composite bottle was accomplished using a peristaltic pump and pre-cleaned (inside and outside) sub- sampling hose. The sub-sampling hoses consisted of a short piece of silicone hose with FEP tubing attached to both ends. The rigid FEP tubing was long enough to reach the bottom of the 20-L composite bottle and extend at least 6 inches to allow space for manual manipulation of the tube. Filling sample containers by pumping from the composite bottle was performed by two people. One person was respon- sible for filling individual sample containers and one person was responsible for constantly moving the intake tubing up and down in the water column of the composite sample. Based on experimental evidence, this up and down movement of the intake is a procedure that helps obtain a more repre- sentative sub-sample since some stratification of heavier par- ticles in the composite sample despite the mixing created by the stirrer. (Note that these samples were whole storm com- posites that included suspended material greater than 63 µm in diameter.) They recommend limiting the up and down movement of the intake tubing should to approximately 80 to 90% of the depth of the water column and never touch the bottom of the composite bottle. Regardless of the methods and approach selected to split large composite samples, initial testing is essential to demonstrate that the process can be performed without introducing contamination and assure that subsamples can be obtained that consistently meet project requirements for precision, accuracy, and representativeness. 5.6 Filtration Issues A number of important questions remain unanswered with respect to the magnitude and importance of changes in the par- titioning of dissolved and particulate metals during the process of collecting and handling the composite samples. The sam- pling process may occur over the period of just a couple of hours to a couple of days. Once storm discharges are complete, the composite sample is removed and subjected to strong mix- ing in order to obtain uniformly mixed subsamples for analy- sis. The federal guidelines indicate that filtration should occur within 15 minutes of splitting the composite sample. These guidelines are not considered practical for most stormwater programs. The Caltrans stormwater monitoring guidelines (Caltrans 2003) established a holding time of 48 hours for fil- tration of flow-weighted composite samples to allow samples to be transported and filtered in the laboratory. The time required to attain an equilibrium distribution of constituents between the dissolved and particulate fractions in stormwater merits further research investigation. If there is a significant trend toward increased particulate phase met- als during storage, then this trend could be useful in BMP selection. BMPs that can store the runoff can provide time for increased particulate phase concentrations and therefore greater metals removal, because metal removal by BMPs is primarily due to suspended solids removal. The hydraulic detention time for many BMPs, such as detention basins, is quite small compared to the time typically allowed for equi- librium to be reached in most samples collected as part of monitoring programs. Holding times for both filtration and analyses are measured from the time the last aliquot is taken to the time when samples are filtered in the laboratory. By any measure, this is a highly arbitrary starting point. The extent to which geochemical and physical partitioning processes may have progressed by the end of the storm event cannot be established. Ultimately, the issue becomes a question of how rapidly these partitioning processes occur and whether exposure times in a BMP are sufficient to accelerate precipitate and sequester metals. It is estimated that partitioning processes can approach equilibrium within 1 to 12 hours depending largely on the sediment content and size composition. The National Highway Runoff Water-Quality Data and Methodology Synthesis (Breault and Granato 2000) iden- tified the existence of dissolved matrix sampling artifacts associated with historical water quality data that could be classified as either environmental or procedural. Environ- mental factors were defined as geochemical composition, sus- pended solids, concentrations of colloids, and the amount of organic material in the samples. All of these factors varied sub- stantially between storm events and contributed to variability. Procedural artifacts were identified as those introduced by use

A-25 of different filter types, pore sizes, filter surface areas, and methods of filtration. They also demonstrated how filter pack- ing could alter the size classes of sediment that pass through the filter. Even though the filter was designed to let size classes of 0.45 µm or less to pass through the filter, it is evident that the filter packing set up a finer filter that limited many smaller particles from crossing the filter. In an attempt to overcome some of these issues, Breault and Granato (2000) suggested three potential options: • Large surface area stainless steel, Teflon filter holders with 0.45 µm filters; or • Use of pre-centrifugation to remove sediment and colloid that might clog the membrane. This eluate would then be filtered through a standard 0.45 µm; or • “Exhaustive filtration” with the more complex techniques that are designed to remove colloids and thus leave only metals that are truly “dissolved.” Since these volumes were published in 2003, few changes have been made in the manner in which stormwater samples have been collected and tested for dissolved and total trace metals. The majority of stormwater monitoring has been conducted with autosamplers taking flow-weighted com- posites for the full duration of an event. Two recent studies based upon intensive grab sampling (Dean, Sansalone et al. 2005; Lau, Han et al. 2009) and filtration within 2–4 hours of collection both illustrate a substantial degree of variability in the dissolved fraction (fd) of metals and generally higher proportions of dissolved metals compared to flow-weighted composite samples filtered up to 48 hours after the end of an event. Work conducted in Baton Rouge, Louisiana (Dean, Sansalone et al. 2005) showed that metal partitioning and speciation was influenced by hydrologic conditions. High intensity rainfall was noted to result in increased fd values that was attributed to shortened residence times. During high flow rates, cadmium, copper and zinc tended to exhibit low fd values due to a more pronounced diluting effect on dis- solved metal concentrations and a more pronounced delivery of particulate-bound metals. Another study (Lau, Han et al. 2009) conducted in southern California also primarily used grab sampling to characterize highway runoff at three sites with average daily traffic (ADT) volume greater than 260,000. The study was conducted over a period of three storm sea- sons from 2000 to 2003. Samples were transported to the nearby UCLA campus laboratory and filtered within a few hours of the collection. Lau, Han et al. (2009) provided tables comparing calculated dissolved and total recoverable metals EMC results for the Los Angeles highway sites with storm- water composite data from six years of monitoring at Caltrans sites throughout the state. These data were restructured by Minervini (2010) to compare the average percentages of met- als present in the dissolved form (Table 6). The UCLA study found that cadmium, copper, nickel and zinc were predomi- nantly in the dissolved fractions (71 to 82% dissolved). It was suggested that the much shorter holding times used for com- pleting the filtration process was a likely factor in the higher percentages of the metals in the dissolved fraction. Another contributing factor may have been the low concentrations of suspended solids available in the samples relative to those reported in the flow-weighted composite samples from other Caltrans monitoring efforts. 5.6.1 Filtration Methods Filtration is best conducted in the laboratory where the environment can be better controlled. Given the lack of com- parability and potential for introduction of contaminants, it is recommended that samples be transported to the laboratory as soon as practical and filtered immediately. Filtration meth- ods should be consistent with those outlined in EPA Method 1669. When necessary, these filtration methods can be uti- lized in a range of environments. If filtration is required to be conducted in a field situation, it would best be performed in Parameter Lau et al. 2009 Caltrans (1997-2003) Los Angeles (>260,000 ADT) Statewide Nonurban (<30,000 ADT) Urban (>30,000 ADT) TSS (mg/L) 68 175 232 152 Cd 72 38 40 33 Cr 28 32 29 33 Cu 71 34 35 33 Ni 79 40 34 42 Pb 15 10 12 10 Zn 82 37 35 37 (Lau, Han et al. 2009; Minervini 2010.) Table 6. Average percent dissolved metals.

A-26 a mobile laboratory or walk-in shelter that provides a clean, dry, and well lit location for handling the samples. Use of a glovebox is recommended for maintaining clean conditions. The glovebox should be set up in a protected area that has been set up with the intent of filtering water. Capsule filters are recommended for use in filtering storm- water samples since they minimize potential for introduction of contaminants and studies have demonstrated that they are capable of producing suitable blanks at levels appropri- ate to EPA water quality criteria. Pall AquaPrep 600 Capsules 0.45 µm tortuous-path capsule filters (Catalog #12175) pro- vide a 600 cm2 effective filtration area (EFT) that can be suf- ficient for filtration of less than 1 liter of stormwater with relatively low quantities of suspended solids. This capsule fil- ter uses Pall Supor membrane made of polyester-reinforced polysulfone (PES). The Pall GWV High Capacity Ground- water Sampling Capsules (Catalog #12178) are often a better filtration capsule for stormwater due to their ability to handle larger sediment loads before clogging. The GWV capsule pro- vides 0.45 µm filtration but has an EFT of 700 cm2 and uses Pall’s Versapor® membrane (acrylic copolymer on a non- woven support). Comparable alternative capsule filters are available with similar specifications. Regardless of the selected brand and membrane type, blanking studies are necessary to assure that the filters can produce blanks that are below target detection limits for the program. The Pall capsule filters are intended for pressure filtration. A small, portable peristaltic pump is used to pump the sample water through the capsule filters. Cole Parmer sells a Master- flex L/S drive (Catalog No. H-07570-10) equipped with a quick load pump head that is commonly used for this purpose due to the range of pumping rates, working pressure, and flexibility to operate off AC or DC power. C-Flex™ peristaltic pump tub- ing (3/8 in. ID by approximately 3 feet) is the preferred tubing for use with this pump. C-Flex ™ is a thermoplastic elastomer made from styrene-ethylene, butylene-styrene (SEBS) resins. SEBS resintubing can be subjected to more aggressive cleaning with acids and does not tend to sorb metals. Unfortunately, this material will not hold up to the forces of the high-volume peristaltic pumps used for wastewater and stormwater sam- pling. The C-Flex™ peristaltic hose is connected to short 6-8 inch lengths of FEP tubing at both ends. Appropriately sized PVC, clear polyethylene, or fluoropolymer “barbed” straight connectors that have been subject to protocol clean- ing procedures can be used to connect segments of tubing. Tubing is cleaned by soaking in a nitric acid solution for 8–24 hours, rinsed with reagent water and air-dried. After drying, the ends are capped, tubing is bagged in clear polyeth- ylene bags, serialized with a unique number associated with that cleaning batch, and stored until use. The capsule filters and tubing should be pre-rinsed and conditioned by pumping approximately 1 liter of Type 1 laboratory water (Milli-Q or equivalent) through the cap- sule filters and discarded. This filter rinse is recommended even though Pall now indicates that each capsule filter is now certified to have been rinsed in this manner. Water from a minimum of 1 out of every 20 filters or one in every batch should be analyzed for metals to verify that quality standards are being maintained. After the pre-rinse is complete, the intake is placed in the 1-liter HDPE bottle that was designated for dissolved metals and an initial portion of the sample (~50–100 ml) is pumped into a waste container. The remainder of the filtered water is directed into the final sample container and fixed with an ultrapure nitric acid such as Seastar Chemicals BASELINE nitric acid. Five milliliters of 10% Seastar Chemicals BASE- LINE nitric acid is typically sufficient to fix a 1-liter sample. If samples are fixed in the field, polypropylene or Teflon® vials of 50% Ultrex-grade nitric acid should be used (USGS 1998). A minimum of one (2 ml each) vial should be used for each 250 ml of sample in order to adjust pH to less than 2. The filtration process is best performed with two person- nel with one assigned to be “clean hands (CH)” and the other “dirty hands (DH)”. The following is a summary of the pro- cess used to obtain a filtered composite sample for analysis of dissolved metals. 1. Set up the filtration system and supplies in a comfortable, convenient, and well-lit working location. DH opens the plastic bag and CH removes a section of C-Flex™/FEP tubing. 2. DH closes the bag and then assists CH in loading the peristaltic hose into the pump. 3. DH opens the storage bag containing the sample to be filtered and CH removes the water sample from the inner storage bag and places the bottle on the counter. 4. CH also places two clean empty sample bottles, a bottle containing reagent water, and a bottle for waste in the glove bag. 5. CH removes the lid of the reagent water bottle and places the end of the pump tubing in the bottle. 6. DH starts the pump and passes approximately 200 mL of reagent water through the tubing and filter into the waste bottle. 7. CH then moves the outlet tubing to a clean bottle and collects the remaining reagent water as a blank. DH stops the pump. 8. CH removes the lid of the sample bottle and places the intake end of the tubing in the bottle. 9. DH starts the pump and passes approximately 50 mL through the tubing and filter into the remaining clean sample bottle and then stops the pump. CH uses the fil- trate to rinse the bottle, discards the waste sample, and returns the outlet tube to the sample bottle.

A-27 10. DH starts the pump and the remaining sample is pro- cessed through the filter and collected in the sample bot- tle (final volume 500 mL). If preservation is required, the sample is acidified at this point. 11. CH replaces the lid on the bottle, returns the bottle to the inside bag, and zips the bag. 12. CH then places the zipped bag into the outer bag held by DH. 13. DH zips the outer bag and places the double-bagged sample bottle into a clean, ice-filled cooler for immediate shipment to the laboratory. 5.6.2 Real-time Filtration A conceptual approach has been developed for perform- ing real-time filtration (Figure 8) that would enable collec- tion of flow-weighted composite samples for dissolved metals that represent actual conditions at the sampling point. This approach also avoids the additional handling of the composite sample to obtain subsamples for filtration thus providing less potential for introduction of contaminants. This approach would require some experience with electronics and the knowledge of the autosamplers. Minor fabrication would also AUTOSAMPLER PUMP HEAD CAPSULE FILTER 1-LITER HDPE 20-LITER BOROSILICATE GLASS MASTERFLEX PUMP PUMP/CR 600 COMMUNICATIONS BATTERY PURGE PUMP PURGE COLLECTION CHAMBER/CELL/TUBING (LEVEL) ~10-15 ml + + VENT ON OFF CR 600 - - Figure 8. Diagram showing conceptual real-time filtration system for dissolved metals.

A-28 be required to construct the sampling capsule from appropri- ate materials. A small, inline capsule could be used to trap a small quan- tity (~20ml) of water from each aliquot. Water would enter and exit near the top of the capsule. The bulk of the aliquot sample (typically 200 to 500 ml) would pass through and get discharged into a large, standard composite container that would be used for measurement to total metals, sediment, major ions, etc. A third port would be located at the bottom of the capsule. The filtration system would be connected at this point. The capsule would need a mounted level such that a consistent quantity of water would be subsampled during the pumping process, and the volume remaining in the cell would not be impacted by the purge cycle. Most modern autosamplers have a range of potential for outputting information to a port indicating that a successful aliquot has been taken. The signal can be read by any basic data logger/control module such as a Campbell Scientific CR200X. The logger would be connected to a control switch or relay that allows switching a connection between two terminals (3 and 4) that control whether the Masterflex Pump is run- ning or not. A closure of the remote switch contacts will turn the pump system on. Opening the contact will turn the pump system off. The datalogger would be programmed to run for a fixed time of roughly 10-20 seconds at 200/ml per minute to pump all of the water out of the bottom of the capsule and through a 0.45 micron capsule filter into a container. Tubing coming from the exit of the capsule filter would enter through the lid of the sample container and a second small hose would exit the lid and be vented through another small hose fitted with a PTFE vent filter. Use of small subsamples on the order of 20 ml would allow 50 aliquots to be sampled before filling a 1-liter bottle. If the high volume capsule filters are used, filter clogging should not be an issue for samples of this size.

A-29 The sampling and analysis plan (SAP) should incorporate a quality assurance and quality control (QA/QC) plan. This plan should detail activities that will be conducted to assure that both chemical and physical measurements meet the stan- dard of quality needed to evaluate dissolved and particulate metals at levels relevant to EPA water quality criteria. Data comparability is an important consideration when devel- oping these project plans. Intergovernmental Task Force on Monitoring Water Quality (ITFM 1995) defines comparabil- ity as the “characteristics that allow information from many sources to be of definable or equivalent quality so that it can be used to address program objectives not necessarily related to those for which the data were collected.” One important aspect of comparability is the use of ana- lytical laboratories that are accredited under a program such as the National Environmental Laboratory Accreditation Program (NELAP) or a well-qualified research laboratory. In addition, the laboratory should be a participant in a labora- tory proficiency and intercalibration program. The online National Environment Methods Index (2011; http://acwi.gov/methods/index.html) is a key resource for selection of analytical methods that meet detection limits needed for the project and help insure that data are com- parable. The SAP should incorporate all protocols that will be used to validate the analytical data. Protocols provided in the National Functional Guidelines for Inorganic Super- fund Data Review (USEPA540-R-10-011) (U.S. EPA 2010) and the Guidance on the Documentation and Evaluation of Trace Metals Data Collected for Clean Water Act Compli- ance Monitoring (EPA/821/B/95/002) (U.S. EPA 1996) are recommended. The sections that follow address activities associated with both field sampling and laboratory analyses. Quality assur- ance activities start with procedures designed to assure that errors introduced in the field sampling and subsampling processes are minimized. Field QA/QC samples are collected and used to evaluate potential contamination and sampling error introduced into a sample prior to its submittal to the analytical laboratory. Laboratory QA/QC activities are used to provide information needed to assess potential labora- tory contamination, analytical precision and accuracy, and representativeness. 6.1 Detection Limits, Data Quality Objectives, and Measurement Quality Objectives There are a great number of definitions related to detec- tion and quantification limits. A basic understanding of the derivation of these various limits is essential when data from the monitoring effort are being reported and assessed. The Method Detection Limit (MDL) for any test is cur- rently defined in Title 40 of the Code of Federal Regulations Part 136 as “the minimum concentration of a substance that can be measured and reported with 99% confidence that the analyte concentration is greater than zero, and is determined from analysis of a sample in a given matrix containing the analyte.” (40 CFR 136, Appendix B, revision 1.11). The MDL is derived from seven analyses conducted on a sample con- taining concentrations of the analyte of concern. The MDL is calculated by multiplying the standard deviation of the seven replicates by the appropriate 1-tailed Student’s T value for 6 degrees of freedom (n-1) and a probability of 0.01 (3.143). Therefore the MDL is actually a distribution with the limit established at the mean or 50% percentile value. The Limit of Quantification (LOQ) is defined as the lowest concentration of an analyte in a sample that can be determined with accept- able precision and accuracy under the stated operational con- ditions of the method. The LOQ, which here considered the Reporting Limit, is approximately 10 times the standard devi- ation of the seven replicates analyzed in the MDL study. A reporting limit (or LOQ) is a “threshold based on a measure of the variability or noise inherent in the laboratory process, while a detection limit is a threshold below which measured C h A P T E r 6 Analytical Methods and Quality Assurance/ Quality Control

A-30 values are not considered statistically different from a blank signal” (Helsel 2005). Reporting limits for the metals and many of the other pro- posed constituents are often near the practical limits of modern instrumentation with use of implementation of methods that require extraction and concentration of the metals. These sug- gested reporting limits were established based upon the need to evaluate the data at levels appropriate to receiving water quality standards. The relative accuracy of most laboratory methods decreases as concentrations approach the detection limits. For this reason it is recommended that, where possible, reporting limits should be established at levels that are 0.1 to 0.5 times the level of concern. These lower reporting limits are necessary in order to minimize analytical variability so that data can be used to effectively assess water quality relative to the established standards. Data that are below reporting limits should be identified as <RL. For purposes of calculating descriptive statistics, one of three methods for handling left-censored data with single or multiple detection limits (Helsel 2005, 2009). These three methods [maximum likelihood estimation (MLE); regres- sion on order statistics (ROS); and Kaplan-Meier (K-M)] are summarized and compared in Geosyntec and Wright Water Engineers (2009; http://bmpdatabase.org). The ROS method is recommended by the International Stormwater Best Man- agement Practices (BMP) Database project and is currently used by Caltrans studies in California. A Microsoft Excel add-in macro and instructions for the ROS method can be accessed from the Caltrans Stormwater website (http://www. dot.ca.gov/hq/env/stormwater/pdf/DAT.zip). Performance acceptance criteria, also known as Data Qual- ity Objectives (DQOs), should be specified for the project, to help ensure that the data produced by the monitoring program will be useful in addressing program objectives. DQOs are intended to limit the probability of making deci- sion errors by considering the purpose for collecting the data, defining the appropriate data needed, and defining the tol- erable probabilities of making decision errors. Measurement Quality Objectives (MQOs) provide specific acceptance cri- teria (constituent reporting limits, analytical precision, accu- racy, completeness, and representativeness) for each analytical method. MQOs, as defined for this guidance document, specify the quality of data required to support the specific objec- tives of the monitoring program. Table 7 and Table 8 pro- vide suggested MQOs for measurement of dissolved metals at levels of ecological concern and provide the ancillary data that are useful in the interpretation of the data. The MQOs for ancillary data were selected to provide the abil- ity to run various geochemical models that can improve assessment of bioavailability and mobility of metals in stormwater runoff. 6.2 Precision, Bias, Accuracy, Representativeness, Completeness, and Comparability The overall quality of analytical measurements is assessed through evaluation of precision, accuracy/bias, representative- ness, comparability, and completeness. Precision and accuracy/ bias are measured quantitatively. Representativeness and com- parability are both assessed qualitatively. Completeness is assessed in both quantitative and qualitative terms. The follow- ing sections examine how these measures are typically applied. 6.2.1 Precision Precision provides an assessment of mutual agreement between repeated measurements. These measurements apply to field duplicates, laboratory duplicates, matrix spike dupli- cates, and laboratory control sample duplicates. Monitoring of precision through the process allows for the evaluation of the consistency of field sampling and laboratory analyses. The Relative Percent Difference (RPD) will be used to evaluate precision based upon duplicate samples. The RPD is calculated for each pair of data and is calculated as: [ ] [ )( ) ( )= − ∗ +RPD x x 100 x x 21 2 1 2 Where: x1 = concentration or value of sample 1 of the pair x2 = concentration or value of sample 2 of the pair In the case of matrix spike/spike duplicate, RPDs are com- pared with MQOs established for the program. Suggested MQOs are provided in Table 8 but final objectives should be established based upon the objectives of the monitoring program and in consultation with the laboratory. In the case of laboratory or field duplicates, values can often be near or below the established reporting limits. Calculation of RPDs is not appropriate if one or both values are below reporting lim- its. When both values are near the detection limits RPDs may exceed established limits but the absolute difference between the two samples should not exceed the reporting limit. 6.2.2 Bias Bias is the systematic inherent in a method or caused by some artifact or idiosyncrasy of the measurement system. Bias may be either positive or negative and can emanate from a number of different points in the process. Although both positive and negative biases may exist concurrently in the same sample, the net bias is all that can be reason- ably addressed in this project. Bias is preferably measured through analysis of spiked samples so that matrix effects are incorporated.

A-31 6.2.3 Accuracy Accuracy is a measure of the closeness of a measurement or the average of a number of measurements to the true value. Accuracy includes a combination of random error as mea- sured by precision and systematic error as measured by bias. An assessment of the accuracy of measurements is based on determining the percent difference between measured val- ues and known or “true” values applied to surrogates, matrix spikes (MS), laboratory control samples (LCS) and standard reference materials (SRM). Surrogates and matrix spikes evaluate matrix interferences on analytical performance, while laboratory control samples, standard reference materi- als, and blank spikes (BS) evaluate analytical performance in the absence of matrix effects. Assessment of the accuracy of measurements is based upon determining the difference between measured values and the true value. This is assessed primarily through analysis of spike recoveries or certified value ranges for SRMs. Spike recoveries are calculated as percent recovery according to the following formula: [ ]( )= − α ∗Percent Recovery t x 100% Analyte Method Type EPA Method Number Holding Time Container Type Preservation Detection1,2 Limit/ Resolution Units Conventionals pH electrometric 150.1 15 minutes glass or PE none +/- 0.1 std. units Alkalinity titrimetric; SM 2320B Filter ASAP, 14 days PE none 1 mg/L Carbonate titrimetric; SM 2320B Filter ASAP, 14 days PE none 1 mg/L Bicarbonate titrimetric; SM 2320B Filter ASAP, 14 days PE none 1 mg/L Hydroxide titrimetric; SM 2320B Filter ASAP, 14 days PE none 1 mg/L Conductivity conductivity meter SM 2510 28 days glass or PE 4°C; filter if hold time >24 hours +/- 1 µmho/cm Hardness titrimetric; colorimetric; calculation; ICP-AES 130.2; 130.1; SM 2340B; 200.7 6 months glass or PE 4°C and HNO3 or H2SO4 to pH<2 1 mg/L Calcium ICP-AES SM 2340B; 200.7 6 months glass or PE 4°C and HNO3 or H2SO4 to pH<2 0.01 mg/L Magnesium ICP-AES SM 2340B; 200.7 6 months glass or PE 4°C and HNO3 or H2SO4 to pH<2 0.02 mg/L TOC; DOC oxidation or combustion 415.1 28 days glass or PE 4°C and HCl or H2SO4 to pH<2 1 mg/L TDS dried filtrate weight 160.1 7 days glass or PE 4°C 1 mg/L TSS dried filter weight 160.2 7 days glass or PE 4°C 4 mg/L SSC dried filter weight ASTMD3977B 7 days glass or PE 4°C 4 mg/L Ca, Mg, Na, K ICP-AES 200.7; 6010B 6 months glass or PE 4°C and HNO3 or H2SO4 to pH<2 0.01-0.3 mg/L SO4, Cl- Ion chromatography 300 28 days glass or PE 4°C 0.02 mg/L Table 7. Recommended constituents, analytical methods, and specifications for sample handling. (continued on next page)

A-32 Where: t = total concentration found in the spiked sample x = original concentration in sample prior to spiking, and a = actual spike concentration added to the sample 6.2.4 Representativeness, Comparability, and Completeness Representativeness is the degree to which data accurately and precisely represent the natural environment. For storm- water runoff, representativeness is first evaluated based upon the automated flow-composite sample and the associated hydrograph. To be considered as representative, the autosam- pler must have effectively triggered to capture initial runoff from the pavement and the composite sample should (1) be comprised of a minimum number of aliquots over the course of the storm event, (2) effectively represent the period of peak flow, (3) contain flow-weighted aliquots from over 80% of the total runoff volume, and (4) demonstrate little or no evidence of “stacking.” Stacking occurs when the sampling volume is set too low and commands back up in the memory of an auto- sampler causing it to continuously cycle until it catches up with the accumulation of total flow measured by the stormwater monitoring station. Representativeness is also assessed through the process of splitting or subsampling 20 L composite bottles into indi- vidual sample containers being sent to the laboratory. The first subsamples removed from the composite bottle should have the same composition as the last. Subsampling stan- dard operation procedures (SOPs) should be established and closely followed to minimize variability in this process. As noted in Section 5.5, sample splitting can introduce a sub- stantial amount of error especially if significant quantities of coarse sediments (greater than 250 µm) represent a signifi- cant fraction of the suspended sediments. Wet sieving of the whole storm sample with a 63 µm mesh prior to splitting the sample (Selbig and Bannerman 2008) and subsequent diges- tion of the entire unfiltered and filtered subsample containers has shown some promise of providing better assessments of the particulate and dissolved loads of metals in stormwater. The sand fraction can be analyzed independently as a single sample or further fractionated to assess partitioning within the sand fractions. Removal of the sand fraction from the whole stormwater material greatly reduces variability among split samples regardless of the method used. Comparability is the measure of confidence with which one dataset can be compared to another. The use of standard- ized methods of chemical analysis and field sampling and processing are ways of insuring comparability. Application of Analyte Method Type EPA Method Number Holding Time Container Type Preservation Detection1,2 Limit/ Resolution Units Metals Aluminum GF-AA; ICP-MS 200.9; 200.8 filter for dissolved fraction and preserve within 24 hours; 6 months to analysis Teflon, PE, or borosilicate glass 4°C and HNO3 to pH<2 1 µg/L Arsenic ICP-MS 200.8; 1632 1.4 µg/L Cadmium ICP-MS 200.8; 1638 0.5 µg/L Chromium GF-AA; ICP-MS 200.8 0.9 µg/L Copper GF-AA; ICP-MS 200.8; 1638 0.5 µg/L Iron GF-AA; colorimetric 200.7; 200.9; SM 3500-Fe B 10 µg/L Lead GF-AA; ICP-MS 200.8;1638 0.6 µg/L Nickel GF-AA; ICP-MS 200.8; 1638 0.5 µg/L Selenium GF-AA; GH-AA; ICP-MS 200.9; 200.8; 1638; 1639 0.6 µg/L Silver GF-AA; ICP-MS 200.8; 1638 01 µg/L Zinc GF-AA; ICP-MS 200.8; 1638 1.8 µg/L 1 When multiple analytical methods are stated, the analytical method associated with the detection limit was bolded. 2 Suggested detection limits are based upon the National Environmental Methods Index (https://www.nemi.gov/apex/f?p=237:1:175271955775976), accessed September 30, 2011. Table 7. (Continued).

A-33 Target5 Detection Limit Holding Times Accuracy Precision Completeness Analyte Spike Recovery SRM2 Recovery Matrix Spike RPDs Laboratory Duplicate RPDs Conventionals pH 0.1 units Immediately1 — — — 20% 95% Alkalinity 1 mg/L Filter ASAP, 14 days — — — 20% 95% Carbonate 1 mg/L Filter ASAP, 14 days — — — 20% 95% Bicarbonate 1 mg/L Filter ASAP, 14 days — — — 20% 95% Hydroxide Alkalinity 1 mg/L Filter ASAP, 14 days — — — 20% 95% Spec. Conductance 1 mhos/cm 28 days 1 — 80-120% — 25% 95% Hardness as CaCO3 1 mg/L 6 months 80-120% 80-120% 25% 20% 95% Calcium 0.01 mg/L 6 months 80-120% 80-120% 25% 20% 95% Magnesium 0.02 mg/L 6 months 80-120% 80-120% 25% 20% 95% TDS 1 mg/L 7 days3 80-120% 80-120% 25% 20% 95% TSS 1 mg/L 7 days3 80-120% 80-120% 25% 20% 95% SSC 1 mg/L 7 days3 80-120% 80-120% 25% 20% 95% TOC 1 mg/L 28 days 85-115% 85-115% 25% 15% 95% DOC 1 mg/L 28 days 85-115% 85-115% 25% 15% 95% Major Ions Sulfate 0.02 mg/L 28 days 80-120% 80-120% 25% 20% 95% Chloride 0.02 mg/L 28 days 80-120% 80-120% 25% 20% 95% Calcium 0.01 mg/L 6 months 80-120% 80-120% 25% 20% 95% Magnesium 0.02 mg/L 6 months 80-120% 80-120% 25% 20% 95% Potassium 0.30 mg/L 6 months 80-120% 80-120% 25% 20% 95% Sodium 0.03 mg/L 6 months 80-120% 80-120% 25% 20% 95% Metals (dissolved4 and total) Aluminum 1.0 6 months 75-125% 80-120% 20% 20% 95% Arsenic 1.4 g/L 6 months 75-125% 80-120% 20% 20% 95% Cadmium 0.5 g/L 6 months 75-125% 80-120% 20% 20% 95% Chromium 0.9 g/L 6 months 75-125% 80-120% 20% 20% 95% Copper 0.5 g/L 6 months 75-125% 80-120% 20% 20% 95% Iron 10 g/L 6 months 75-125% 80-120% 20% 20% 95% Lead 0.6 g/L 6 months 75-125% 80-120% 20% 20% 95% Nickel 0.5 g/L 6 months 75-125% 80-120% 20% 20% 95% Zinc 1.8 g/L 6 months 75-125% 80-120% 20% 20% 95% 1 Performed in field if possible 2 SRM recovery values based upon values provided with each specific SRM 3 7 days based upon limit for measuring TSS/no regulatory limit. 4 Filter within 24 hours of end of storm 5 Suggested detection limits are based upon the National Environmental Methods Index (https://www.nemi.gov/apex/f?p=237:1:175271955775976), accessed September 30, 2011. Table 8. Suggested measurement quality objectives.

A-34 consistent sampling and processing procedures is necessary for assuring comparability among data sets. Thorough docu- mentation of these procedures, quality assurance activities, and a written assessment of data validation and quality are necessary to provide others with the basic elements to evalu- ate comparability. Completeness is a measure of the percentage of the data judged valid after comparison with specific validation cri- teria. This includes data lost through accidental breakage of sample containers or other activities that result in irreparable loss of samples. Implementation of standardized chain-of- custody procedures which track samples as they are trans- ferred between custodians is one method of maintaining a high level of completeness. A high level of completeness is essential to all phases of this study due to the limited number of samples. Of course, the overall goal is to obtain completeness of 100%; however, a realistic data quality indicator of 95% insures an adequate level of data return. 6.3 Laboratory Quality Assurance/ Quality Control The quality of analytical data is dependent on the ways in which samples are collected, handled, and analyzed. DQOs provide the standards against which the data are compared to determine if they meet the quality necessary to be used to address program objectives. Data should be subjected to a thorough verification and validation process designed to eval- uate project data quality and determine whether data require qualification. The three major categories of QA/QC checks, accuracy, precision, and contamination, were discussed in the previ- ous section. As a minimum, the laboratory should incorpo- rate analysis of duplicate samples, method blanks, and matrix spike/spike duplicates with each analytical batch. Use of a certi- fied reference material (CRM) or standard reference material (SRM) is also recommended as these allow assessment of long term performance of the analytical methods so that represen- tativeness can be assessed. Laboratories often use an internal CRM that is analyzed with each batch to evaluate any potential long-term shift in performance of the analytical procedures. Recommended minimum quality control samples are pro- vided in Table 9. 6.4 Field QA/QC 6.4.1 Blanks A thorough system of blanking is an essential element of monitoring. Much of the blanking processes are performed well in advance of the actual monitoring in order to dem- onstrate that all equipment expected to contact water is free of contaminants at the detection limits established for the program. Equipment components are cleaned in batches. Analyte Equipment/ Container Blanks Filter Blanks Method Blanks Field and Laboratory Duplicates MS/ MSDs1 LCSs2 or SRMs3 pH — — — — — Alkalinity — — — Carbonate — — — Bicarbonate — — — Hydroxide — — — Spec. Conductance — — — — Hardness as CaCO3 — — TDS — — TSS — — — TOC DOC Cations (Ca, Mg, K…) — — Anions (Cl, SO4…) — Metals 1 Matrix Spike and Matrix Spike Duplicate 2 Laboratory Control Samples 3 Standard Reference Materials Table 9. Laboratory quality control samples by analyte.

A-35 Subsamples from each cleaning batch are rinsed with Type 1 laboratory blank water and submitted to the laboratory for analysis. If hits are encountered in any cleaning batch, the entire batch is put back through the cleaning and blanking process until satisfactory results are obtained. If contaminants are measured in the blanks, it is often prudent to reexamine the cleaning processes and equipment or materials used in the cleaning process. Equipment requiring blanks and the fre- quency of blanks is summarized below and in Table 10. 6.4.2 Field Duplicates Composite subsampling duplicates associated with flow- weighted composite samples are often referred to as field duplicates but, in fact, they are subsampling replicates. These replicates help assess combined variability associated with subsampling from the composite container and variability associated with the analytical process. They are evaluated against the same criteria as used for laboratory duplicates. Intake Hose One per batch System Component Blanking Frequency Peristaltic Pump Hose One per batch1 or 10% for batches greater than 10 Composite Bottles One per batch or 10% for batches greater than 10 Subsampling Pump Hose One per batch or 10% for batches greater than 10 Laboratory Sample Containers 2% of the lot2 or batch, minimum of one Capsule Filter Blank3 One per batch or 10% for batches greater than 10 Churn/Cone Splitter When field cleaning is performed, process one blank per session 1 A batch is a group of samples that are cleaned at the same time and in the same manner. 2 If decontaminated bottles are sent directly from the manufacturer, the batch would be the lot designated by the manufacturer in their testing of the bottles. 3 If filtration is performed in the laboratory, the capsule filter blanks would be considered part of laboratory QA/QC. Table 10. Summary of blanking requirements for field equipment.

A-36 C h A P T E r 7 Sample collection, handling, and processing materials can contribute and/or sorb trace elements within the time scales typical for collection, processing, and analysis of run- off samples. The relative effect of potential contamination and/or attenuation of trace elements in runoff samples is a function of the concentrations of major and trace ele- ments, organic chemicals, and sediment in solution. Sam- pling artifacts are especially important when measured concentrations that are at or near analytical detection limits (Horowitz 1997). Therefore, great care is required to collect and process samples in a manner that will minimize poten- tial contamination and variability in the sampling process (Breault and Granato 2000). Sampling conducted to measure dissolved metals at levels relevant to EPA water quality criteria requires documenta- tion that all sampling equipment is free of contamination and that the processes used to obtain and handle samples do not introduce contamination. This requires documentation that methods used to collect, process, and analyze the samples do not introduce contamination. Documentation should include establishing detailed written procedures for cleaning all com- ponents of the sampling system, blanking processes necessary to verify that system components and sample handling are not introducing contamination, and a system of tracking deploy- ment of protocol-cleaned equipment in the field. All sample containers and equipment used for sample col- lection in the field and/or sample storage in the laboratory must be decontaminated and cleaned prior to use. These include the FEP tubing, Teflon® lids, strainers, and hoses/ fittings that are used in the subsampling process (USGS 1993). It is important that personnel assigned to clean and handle the equipment are thoroughly trained and familiar with the cleaning, blanking, and tracking procedures. All field sampling staff should be familiar with these processes so that they have a better understanding of the importance of using clean sampling procedures and the effort required to elimi- nate sources of contamination. Cleaning procedures require use of correct equipment, protective gear, and proper handling of waste products. Work should be conducted in a well-ventilated area using appropriate protective garments and goggles for eye protection whenever cleaning the sampling equipment with the acid. Cleaning protocols should also be consistent with ASTM (2008) standard D5088–02 that covers cleaning of sampling equipment and sample bottles. The generalized cleaning pro- cess is based upon a series of washings that typically start with tap water with a phosphate-free detergent, a tap water rinse, soaking in a 10% solution of reagent grade nitric acid, and a final series of rinses with laboratory Type 1 water. SOPs for cleaning sampling equipment are provided in Appendix B. The Florida State Department of Environmental Protection (DEP) also maintains a list of SOPs that contain detailed cleaning procedures for equipment to be used for ultra-trace metals sampling and general water quality sam- pling. These SOPs are available on-line at http://www.dep. state.fl.us/water/sas/sop/sops.htm. The SOPs include guidelines for packaging and labeling cleaned equipment by batches, blanking each batch, and tracking deployment of each system component in order to link site data with blanking data for all bottles and tubing. These are intended as general guidelines that can be modi- fied to fit specific project requirements. Ultimately, appli- cation of the final cleaning protocols should be tested to verify that blanks can consistently be obtained that meet the project objectives before implementation of the field sam- pling program. If contamination is routinely encountered, the procedures and materials used in the cleaning process should be reexamined and modified as necessary to elimi- nate possible sources of contamination. Sample contamination has long been considered one of the most significant problems associated with measure- ment of dissolved metals. One of the major elements of QA/QC documentation is establishing that clean sampling procedures are used throughout the process and that all Equipment Cleaning, Blanking, and Tracking

A-37 equipment used to collect and process the water samples are free of contamination. 7.1 Laboratory Sample Bottles In order to account for any contamination introduced by sampling containers, blanks must be collected for compos- ite bottles and laboratory bottles used for sample storage for metals analysis. A sampling container blank is prepared by filling a clean container with blank water and measuring the concentrations of selected constituents (typically metals and other trace contaminants for composite bottles and metals analysis only for metals storage bottles). These blanks may be submitted “blind” to the laboratory by field personnel or prepared internally by the laboratory. Certified pre-cleaned QC-grade bottles can be used. These bottles are cleaned using acceptable protocol for analysis of dissolved metals and tracked by lots. They come with stan- dard certification forms that document the concentration to which the bottles are considered “contaminant-free,” but these concentrations are not typically suitable for program reporting limits required for measurement of dissolved met- als. Manufacturers may provide an option of certification to specific limits required by a project but it is preferable to purchase the QC bottles that are tracked by lot and conduct internal blanking studies. Lots not meeting project require- ments should be returned to the manufacturer and exchanged for containers from another lot. At least 2% of the bottles in any lot or batch should be blanked at the program detection limits with a minimum frequency of one bottle per batch. A batch is a group of samples that are cleaned at the same time and in the same manner; or, if decontaminated bottles are sent directly from the manufacturer, the batch would be the lot designated by the manufacturer in their testing of the bottles. After cleaning, sample bottles and laboratory-cleaned sam- pling equipment are handled only while wearing clean, powder- free nitrile gloves. All laboratory-cleaned sampling equipment and metals analysis storage bottles are double bagged in clean zip-lock plastic bags for storage or shipment. Clean bottles are stored in a clean area with lids properly secured. Immediately prior to the filling of laboratory sample bot- tles, the bottle labels should be checked, and date and time added using a waterproof pen. Attempting to label grab sam- ple bottles after sample collection may be difficult because of wet labels. The remainder of the information on the labels is prepared in advance to minimize handling requirements during the field effort.

A-38 Reporting and Documentation The National Highway Runoff Data and Methodology Synthesis project (Granato, Zenone et al. 2003) involved an extensive review of available monitoring data and provided a number of key recommendations for improving the infor- mation and data generated by runoff monitoring programs would be valuable for local, regional, and national needs. In an effort to improve data comparability, increase data shar- ing, and allow use of data for secondary purposes, the Meth- ods and Data Comparability Board of the National Water Monitoring Council (2006) developed a set of standardized data elements that should be incorporated in any program. Additional information is provided by Geosyntec Consultants and Wright Water Engineers (2009). This document provides specific guidance on data structures and metadata necessary to submit data to the International Stormwater BMP data- base. A data entry file is provided on the www.bmpdatabase. org web site that identifies all required and recommended data elements needed for submission of stormwater BMP moni- toring data. Among the most important issues cited was a lack of basic information necessary to determine the uncertainty of data sets and documentation of QA/QC practices. This informa- tion was considered critical for evaluating the potential utility of available water quality information for any given purpose (Church, Granato et al. 2003). They indicated that all studies should provide certain basic data that would eventually lead to the ability of other researchers to independently review the information and make informed decisions as to the utility of the data for other purposes. Without the ready availability of the basic data and the necessary information to evaluate the supporting quality assurance information, studies are often considered to be of limited value. Ideally, all data and reports should be archived on an inter- net site where the information can be readily accessed. Con- sideration should be given to making data available through regional data centers that are becoming more common for consolidating and accessing research and monitoring infor- mation to improve distribution and documentation. These centers may have specific requirements for data formats which should be a consideration when designing the moni- toring program. Full access to data along with appropriate documentation generated from these types of investigations will promote better science and help speed progress towards more effective BMPs. This allows data to be used with con- fidence in other studies, encourages improved sampling designs, and will contribute to an improved understanding of the complex processes involved as metals are mobilized from impervious surfaces and transported to the storm drains and receiving waters. Understanding major factors controlling the rates of these processes and the state of metals in the run- off at different locations along the discharge pathway will lead to better BMPs. Table 11 provides a summary of major elements needed to improve acceptance and usability of stormwater runoff data. The Methods and Data Comparability Board (2006) also pro- vides a summary and structure for water quality data that is considered the minimum for assuring that appropriate infor- mation is provided to assure that data can be shared and used by others. Copies of the SAP, Data Assessment and Validation Report, and any final reports summarizing and interpreting the results should be available as PDF documents. The Data Assessment and Validation Report should provide a thorough review of QA/QC activities and summarize the results. This report should provide a systematic assessment of the data and quality assurance measurements against the MQOs estab- lished in the SAP. This report is often included as an appendix to main body of the report. Information should include: • Tabular summaries of all quality assurance measures, • Identify cases where MQOs may not have been met, • Discuss the potential cause of these excursions, • Identify corrective actions taken, • Discuss impacts of any data quality issues, and • Apply qualifiers based upon criteria established in the SAP. C h A P T E r 8

A-39 Electronic data files should be included that contain infor- mation on the monitoring sites; rainfall, runoff, sampling data and water quality data associated with each monitoring event; and time series data for seasonal flow and precipitation information. If possible, this information should be provided in a standardized format such as that used by the Caltrans stormwater monitoring program in California (Caltrans 2003). If data are not provided in one of several established data base formats, metadata should be included with the files to describe the overall structure and contents of each field. Table 11. Summary of recommended information and data needed to improve use of the data in addressing regional and national issues. Written Documents SAP Project Report(s) Data Assessment and Validation Report ELECTRONIC DATA FILES (Excel, Access or Other Database Formats) SITE DATA Monitoring Site Information Location Size of drainage area Average slope Soil type Percent impervious Estimate runoff coefficient Average Annual Daily Traffic (AADT) Type of pavement or land use BMPs applicable to site EVENT DATA (Include metadata identifying database structure and fields) Rainfall Total rainfall Seasonal rainfall to date Antecedent rainfall (inches) Antecedent dry weather (days) Maximum rainfall intensity Date/time of start and end of rainfall Duration of rainfall Runoff Total runoff volume Peak flow rate Date/time of start and end of runoff Duration of runoff Sampling Data # of aliquots Time of each aliquot % storm capture Peak flow capture? Water Quality Data Validated field and laboratory measurements TIME SERIES DATA (Seasonal or Event Data) Flow Site data only Precipitation Site or regional data If regional, include location, site ID and agency responsible for site

A-40 Alpers, C. N., H. E. Taylor, and J. L. Domagalski., eds. (2000). Metals Transport in the Sacramento River, California, 1996-1997. Water-Resources Investigations Report 99-4286. American Public Health Association (1995). Standard Methods for the Examination of Water and Wastewater, 19th Edition. Amery, F., F. Degryse, K. Cheyns, I. De Troyer, J. Mertens, R. Merckx and E. Smolders (2008). “The UV-Absorbance of Dissolved Organic Matter Predicts the Fivefold Variation in its Affinity for Mobilizing Cu in an Agricultural Soil Horizon.” European Journal of Soil Sci- ence, 59, December: 1087–1095. Amrhein, C., P.A. Mosher, and J. E. Strong (1993). “Colloid-Assisted Transport of Trace Metals in Roadside Soils Receiving Deicing Salts.” Soil Science Society of America Journal, 57(5): 1212–1217. Amrhein, C., P.A. Mosher, J. E. Strong, and P. G. Pacheco. (1994). “Trace Metal Solubility in Soils And Waters Receiving Deicing Salts.” Journal of Environmental Quality, 23 (2): 219–227. ASTM D5088–02. (2008). Standard Practice for Decontamination of Field Equipment Used at Nonradioactive Waste Sites. Barber, M. E., M. G. Brown, et al. (2006). Phase I: Preliminary Environ- mental Investigation of Heavy Metals in Highway Runoff. Washington State Department of Transportation. Bauske, B. and Goetz, D. (1993). “Effects of Deicing Salts on Heavy Metal Mobility.” Acta hydrochimica et hydrobiologica, 21 (1): 38–42. Bent, G. C., Gray, J. R., Smith, K. P., and Glysson, G. D. (2001). A Synopsis of Technical Issues for Monitoring Sediment in Highway and Urban Runoff. U.S. Geological Survey Open File Report 00-497. Breault, R. F., Colman, J. A., Aiken, G. R., and McKnight, D. (1996). “Copper Speciation and Binding By Organic Matter in Copper- Contaminated Streamwater.” Environmental Science and Technology, 30 (12): 3477–3486. Breault, R. F. and G. E. Granato. (2000). “A Synopsis of Technical Issues of Concern for Monitoring Trace Elements in Highway and Urban Runoff.” National Highway Runoff Data and Methodology Synthesis, prepared in cooperation with the Federal Highway Administration. Bricker, O. P. (1999). An Overview of the Factors Involved in Evaluat- ing the Geochemical Effects of Highway Runoff on the Environment. U.S. Geological Survey Open File Report 98-630, http://pubs.water. usgs.gov/ofr98-630/. Capel, P. D. and Larson, S. J. (1996). “Evaluation Of Selected Informa- tion on Splitting Devices for Water Samples.” U.S. Geological Survey Water-Resources Investigations Report, 95–4141. Capel, P. D., Nacionales, F. C., and Larson, S. J. (1995). Precision of a Splitting Device for Water Samples. U.S. Geological Survey Open File Report 95–293. Caltrans. (2003). Caltrans Comprehensive Protocols Guidance Manual. CTSW-RT-03-105.51.42. Church, P. E., G. E. Granato, et al. (2003). Basic Requirements for Col- lecting, Documenting, and Reporting Precipitation and Stormwater- Flow Measurements. Clark, S. E., C. Y. S. Siu, et al. (2009). “Peristaltic Pump Autosamplers for Solids Measurement in Stormwater Runoff.” Water Environ- ment Research, 81: 192–200. CRT (2000). http://www.epa.gov/region9/water/ctr/ Dean, C. M., J. J. Sansalone, et al. (2005). “Influence of Hydrology on Rainfall-Runoff Metal Element Speciation.” Journal of Environmen- tal Engineering, 131(4): 632–642. Doner, H. E. (1978). “Chloride as a Factor in Mobilities of Ni(II), Cu(II), and Cd(II) in Soil.” Soil Science Society of America Journal, 42 (6): 882–885. Eads, R. and Thomas, R. (1983). Evaluation of a Depth Proportional Intake Device for Automatic Pumping Samplers. Journal of the American Water Resources Assoc., 19 (2): 289–292. Edwards, T., and Glysson, G. (1999). Field Methods for Measurement of Fluvial Sediment. Techniques of Water-Resources Investigations, Book 3, Applications of Hydraulics, Chapter C2, U.S. Geological Survey. EPA (2007). http://water.epa.gov/learn/training/standardsacademy/ blm_index.cfm Geosyntec Consultants and Wright Water Engineers, I. (2009). Urban Stormwater BMP Performance Monitoring. Gettel, M., Gulliver, J. S., Kayhanian, M., Degroot, G., Brand, J., Mohseni, O., Erickson, A. J. (2011). “Improving Suspended Sediment Mea- surements by Automatic Samplers.” J Environ Monit., Sept. 5. Glenn, D. W., D. Liu, et al. (2001). “Influence of Highway Runoff Chemistry, Hydrology, and Residence Time on Nonequilibrium Partitioning of Heavy Metals: Implications for Treatment at the Highway Shoulder.” In Transportation Research Record: Journal of the Transportation Research Board, No. 1755, TRB, National Research Council, Washington, DC: 129–140. Gossett, R. and K. Schiff. (2010). Stormwater Monitoring Coalition Laboratory Guidance Document, Third Edition. Southern Califor- nia Coastal Water Research Project Technical Report 615, West- minister, CA. Granato G. E., Church, P. E., and Stone V. J. (1995). “Mobilization of Major and Trace Constituents of Highway Runoff in Groundwa- ter Potentially Caused by Deicing-Chemical Migration.” In Trans- portation Research Record 1483, TRB, National Research Council, Washington, DC: 92–104. Bibliography

A-41 Granato, G. E., F. G. Bank, et al. (2003). Data Quality Objectives and Criteria for Basic Information, Acceptable Uncertainty, and Quality- Assurance and Quality-Control Documentation. Granato, G. E., C. Zenone, et al. (2003). National Highway Runoff Water- Quality Data and Methodology Synthesis, Volume I:Technical Issues for Monitoring Highway Runoff and Urban Stormwater. Helsel, D. R. (2004). Nondetects and Data Analysis—Statistics for Censored Environmental Data. Hoboken, N. J., John Wiley & Sons Publishing Inc. Helsel, D. R. (1990). “Less Than Obvious: Statistical Treatment of Data Below the Detection Limit.” Environmental Science and Technology, 24(12): 1766–1774. Kinniburgh, D, van Riemsdijk, W., Koopal, L., Borkovec, M., Benedetti, M., and Avena, M. (1999). Ion Binding to Natural Organic Matter: Competition, Heterogeneity, Stoichiometry and Thermodynamic Consistency. Colloids and Surfaces A: Physicochemical and Engineer- ing Aspects, 151 (1-2): 147–166. Horowitz, A. J., Demas, C. R., Fitzgerald, K. K., Miller, T. L., and Rickert, D. R. (1994). U.S. Geological Survey Protocol of the Collection and Processing of Surface-Water Samples for the Subsequent Deter- mination Of Inorganic Constituents in Filtered Water. Open-File Report 94–539. Horowitz, A. J. (1997). “Some Thoughts on Problems Associated With Various Sampling Media Used for Environmental Monitoring.” Analyst, 122: 1193–1200. Horowitz, A. J., Smith, J. J., and Elrick, K. A. (2001). Selected Labora- tory Evaluations of the Whole-Water Sample-Splitting Capabilities of a Prototype Fourteen-Liter Teflon Churn Splitter. U.S. Geological Survey Open File Report 2001–386. Horowitz, A. J., K. R. Lum, et al. (1996). “The Effect of Membrane Fil- tration on Dissolved Trace Element Concentrations.” Water, Air, and Soil Pollution, 90(1): 281-294. ITFM (1995). The Strategy For Improving Water-Quality Monitor- ing in the United States: Final Report of the Intergovernmental Task Force on Monitoring Water Quality. U.S. Geological Survey, Reston, VA. Jones, B. E. (1999). Principles and Practices for Quality Assurance and Quality Control. Open-File Report 98-636. U. S. Geological Survey. Keith L. H. (1996). Principles of Environmental Sampling, second edi- tion, American Chemical Society, Washington, DC. Lau, S. L., Y. Han, et al. (2009). “Characteristics of Highway Stormwater Runoff in Los Angeles: Metals and Polycyclic Aromatic Hydro- carbons.” Water Environ. Res., 81(3): 308–318. Magnuson, M. L., C. A. Kelty, et al. (2001). “Trace Metal Loading on Water-Borne Soil and Dust Particles Characterized Through the Use of Split-Flow Thin-Cell Fractionation.” Anal. Chem., 73(14): 3492–3496. Maheepala, U.K, A.K Takyi, and B. J.C. Perera (2001). “Hydrological Data Monitoring for Urban Stormwater Drainage Systems.” Journal of Hydrology, 245 (1-4): 32-47. Methods and Data Comparability Board (2006). Data Elements for Reporting Water Quality Monitoring Results for Chemical, Biologi- cal, Toxicological, and Microbiological Analytes. NWQMC Technical Report No. 3, April. Methods and Data Comparability Board (2011). National Environmental Methods Index, https://www.nemi.gov/apex/f? p=237:1:4348976520358323. Minervini, W. (2010). “Of: Characteristics of Highway Stormwater Runoff in Los Angeles: Metals and Polycyclic Aromatic Hydro- carbons.” Water Environment Research, 82(9): 861-861. National Water Monitoring Council (2011). Publications of the Meth- ods and Data Comparability Board. http://acwi.gov/methods/pubs/ index.html. Sansalone, J. J. and S. G. Buchberger (1997). “Partitioning and First Flush of Metals in Urban Roadway Storm Water.” ASCE Journal of Environmental Engineering, 123: 134–143. Sansalone, J. J. and D. W. Glenn (2000). “Temporal Variations in Heavy Metal Partitioning and Loading in Urban Highway Pave- ment Sheet Flow: Implications for In Situ Treatment Design.” In Transportation Research Record: Journal of the Transporta- tion Research Board, No. 1720, TRB, National Research Council, Washington, DC: 100-111. Selbig, W. R. and Bannerman, R. T. (2008). A Comparison of Runoff Quantity and Quality from Two Small Basins Undergoing Imple- mentation of Conventional- and Low-Impact Development (LID) Strategies: Cross Plains, Wisconsin, Water Years 1999–2005. U.S. Geological Survey Scientific Investigations Report 2008–5008. Selbig, W. R. and Bannerman, R. T. (2011). “Development of a Depth- Integrated Sample Arm (DISA) to Reduce Solids Stratification Bias in Stormwater Sampling.” Water Environment Research. Selbig, W. R., Bannerman, R., and Bowman, G. (2006). “Use of Wet Sieving to Improve the Accuracy of Sediment and Sediment- Associated Constituent Concentrations in Whole-Water Samples.” Proceedings of the 8th Federal Interagency Sedimentation and Hydro- logic Modeling Conference, Reno, NV, April. Selbig,W. R., R. Bannerman, and G. Bowman (2007). “Improving the Accuracy of Sediment-Associated Constituent Concentrations in Whole Storm Water Samples by Wet-Sieving.” J. Environ. Qual., 36:226–232 Shiller, A. M., and Boyle, E. A. (1987). “Variability of Dissolved Trace Metals in the Mississippi River.” Geochimica et Cosmochimica Acta, 51: 3273–3277. Smith, K. P. (2002). Effectiveness of Three Best Management Practices for Highway-Runoff Quality Along the Southeast Expressway, Boston, Massachusetts. U.S. Geological Survey Water-Resources Investiga- tions Report 02-4059. http://pubs.usgs.gov/wri/wri024059/. Smith, K. P., and Granato, G. E. (2010). Quality of stormwater runoff discharged from Massachusetts highways, 2005–07. U.S. Geological Survey Scientific Investigations Report 2009–5269. http://pubs. usgs.gov/sir/2009/5269/. Strecker, E., L. Mayo, et al. (2001). Guidance Manual for Monitor- ing Highway Runoff Water Quality. Publication No. FHWA- EP-01-022. Tuccillo, M. E. (2006). “Size Fractionation of Metals in Runoff from Residential and Highway Storm Sewers.” Sci. Total Environ., 355: pp. 288–300. U.S. Geological Survey (1993). Trace-Element Contamination—Findings of Study on the Cleaning of Sampler Caps, Nozzles, Bottles, and Bags for Trace-Element Work at the Part-Per-Billion Level. Office of Water Quality Technical Memorandum 93.06. http://water.usgs.gov/admin/ memo/QW/qw93.06.html. U.S. Geological Survey. (1997). Protocols for Cleaning a Teflon Cone Splitter to Produce Contaminant-Free Subsamples for Subsequent Determinations of Trace Elements. Office of Water Quality Techni- cal Memorandum 97.03. http://water.usgs.gov/admin/memo/QW/ qw97.03.html. U.S. Geological Survey. (1998). Change in Nitric Acid Preservative for Trace Element Samples. Office of Water Quality Technical Memorandum 98.06. http://water.usgs.gov/admin/memo/QW/ qw98.06.html.

A-42 U.S. EPA. (1996). Guidance on the Documentation and Evaluation of Trace Metals Data Collected for Clean Water Act Compliance Moni- toring - Draft. Office of Water, Office of Science and Technology, E. A. D. (4303), M. S. SW and D. Washington. U.S. EPA. (1996). Method 1669 Sampling Ambient Water for Trace Metals at EPA Water Quality Criteria Levels. U.S. EPA. (2010). National Functional Guidelines for Inorganic Super- fund Data Review. U. C. L. Program. U.S. EPA, E. A. D. (1998). Evaluating Field Techniques for Collecting Effluent Samples for Trace Metals Analysis. Walkowiak, D. K., ed. (2008). Isco Open Channel Flow Measurement Handbook. Windom, H. L., Byrd, J. T., Smith, R. G., Jr., and Huan, F. (1991). “Inadequacy of NASQAN Data for Assessing Metal Trends in the Nation’s Rivers:” Environmental Science and Technology, 25(6): 1137–1142.