3

Stormwater Sampling and Data Collection

Sampling is required in the Multi-Sector General Permit (MSGP) because it provides information on the quality of stormwater leaving an industrial site and on the performance of stormwater control measures (SCMs) in reducing pollutant burden. However, stormwater sampling is complicated by the dynamic characteristics of stormwater flow, the diffuse nature of many stormwater flows, and the myriad potential pollutants and pollutant characteristics that may exist on an industrial property. This chapter discusses the many challenges in quantifying pollutant discharges and includes recommendations to enhance the reliability and consistency of stormwater monitoring, laboratory analysis, and data management to improve industrial stormwater management under the MSGP.

CHALLENGES OF QUANTIFYING STORMWATER POLLTUANT DISCHARGE

Quantifying stormwater pollutant concentrations, loads, and subsequent environmental impacts is a challenge due to the variability in activities taking place on the land, storm occurrences, stormwater flows, and pollutant concentrations (Breault and Granato, 2000; Bent et al., 2001). Variations in water quality parameters can occur within a single storm, between storms, seasonally, and annually. Stormwater composition shows great temporal variation, especially in the early stages of runoff, for many reasons. Storms of different intensity (rainfall energy) mobilize and transport pollutants at different times after runoff begins. Runoff from different parts of a facility reaches the sample point at different times. Additionally, different pollutants mobilize after different periods in contact with flowing water.

Generally, stormwater pollutant concentrations will follow a first-flush pattern, with the highest concentrations occurring early in the storm. During the start of a storm, the rainfall is washing the drainage area at its most polluted state. As the duration of the storm continues, the concentrations of pollutants generally fall (e.g., Sansalone and Cristina, 2004; Han et al., 2006c). The difference between concentrations in first-flush runoff and later runoff can be an order of magnitude or more for some pollutants. This is not always the case, however, because changing rainfall intensity during a storm can provide energy mid-storm that may scour the drainage areas and produce high concentrations after the first flush. Concentrations in the effluent of treatment SCMs typically do not vary as much as in untreated stormwater. Treatment SCMs generally will reduce concentration and buffer high- and low-concentration excursions. For SCMs that have stormwater storage, sampling of initial discharge at the outfall may consist mostly of (treated) water that has been stored from the previous storm. Different sampling approaches, therefore, can lead to different results.

Effect of Sampling Methodology

The volume-weighted (or flow-weighted) pollutant concentration, also called the event mean concentration (EMC), provides the most consistent and comprehensive assessment of stormwater pollutant discharges and

loads. Pollutant loads are important for understanding longer-term water-body impairments and toxicity concerns. The EMC is defined as the total pollutant mass discharged in the stormwater divided by the total runoff volume for the storm event, as measured at a specific outfall or measurement point:

where C is the pollutant concentration, Q is the flow rate, t is time, and T_d is the storm duration. Determinations of pollutant mass load or the EMC requires comprehensive understanding of the flows and concentrations occurring over the entire storm event, which can be measurement intensive. A pollutant concentration measured at a single time during a stormwater event cannot be considered to be representative of the EMC.

Different types of sampling schemes can be used to quantify stormwater pollutant discharges, ranging from simple grab samples to volume-weighted automatic composite sampling that supports the calculation of the EMC. Stormwater sampling can be resource intensive, and sampling plan decisions need to balance the benefits of the information obtained against the costs and labor requirements.

Grab Sampling

A grab sample will always be a snapshot of a rapidly changing situation. Trying to infer an EMC from a grab sample is not scientifically justifiable. However, the more controlled and consistent the collection of the grab sample(s), the more valuable and comparable the information becomes. Comparing grab samples that come from stormwater collected at different parts of the respective storm hydrographs will not have meaning. However, if samples are collected at the same (or near-same) sampling time during each storm, grab samples can be more reliable measures of stormwater pollution, subject to the limitations described about differences in rainfall energy in separate storms.

The current MSGP requires benchmark grab sampling to occur within 30 minutes of the start of runoff at the discharge point (see Table 1-1). The concentrations in a sample taken in the first 30 minutes of a storm are likely to be higher than the event mean concentration due to the effect of the first flush (unless the discharge is coming out of a treatment SCM or other device that stores water from the prior storm). Thus, current MSGP monitoring provides a low-cost, coarse indicator of the effectiveness of nonstructural and structural SCMs, and potential stormwater discharge pollution concerns. Carefully collected and analyzed grab samples, as part of benchmark monitoring, have value in this regard.

Sampling the first runoff could add further consistency and comparability to the grab sample data set and reduce monitoring variability. Inexpensive passive first-flush samplers are currently available that automatically capture the initial runoff from a storm. These samplers hold approximately 1 liter and are placed in the field before a storm event. They will fill with the first-flush runoff flow. A float (plastic ball) or other mechanism blocks the collector input once the vessel is full. Commercial first-flush samplers appear to provide useful, reproducible information on runoff water quality (Landsman and Davis, 2018). However, these samplers collect the first flow reaching the collection vessel. This flow could be highly contaminated if it is the first wash of the drainage area, or, conversely, it could be the first flow from stored water in an SCM and be relatively unpolluted. The use of first-flush samplers may eliminate some of the variation associated with direct human collection of samples, such as inconsistent placement of the sample bottle in the stormwater stream and variable time of collection. This type of sampling can also reduce the burden of sampling of remote sites.

An additional problem of grab sampling is lack of mixing of solids and the associated pollutants in the water column. Grab sampling often consists of inserting a bottle into the flow at the end of an outfall, and it is important to realize that the location of sampling within the stormwater flow can introduce variability, particularly when sampling runoff has not been treated in a structural SCM to remove particulates. As a result of poor mixing, sampling near the bottom of the pipe can result in higher total suspended solids (TSS) concentrations than samples collected at the water surface. To address this, bedload samplers have been developed and tested that can be installed in stormwater pipes to capture the solid material that will not be collected in traditional grab sampling or even by automatic samplers (Burton and Pitt, 2002).

Composite Sampling

Composite sampling involves taking multiple samples and combining them to obtain a measure of the overall stormwater condition. Several different techniques of sample compositing are possible, depending on the number of samples collected, if they are weighted by time or stormwater volume, and if they are collected manually or automatically. Generally, the greater the number of samples used in the compositing, the closer the resulting sample is to the EMC of the pollutants in the stormwater (Ma et al., 2009).

Manual Composite Sampling. A reasonably complete picture of pollutant concentrations in stormwater can be attained by collecting a composite sample over some period of time, although not necessarily the entire discharge from a storm. Taking multiple grab samples during a storm event, especially at equally timed intervals, and combining them into a single sample for analysis will provide a closer approximation of the true volume-weighted average concentrations of pollutants in the runoff than single grab samples (Burton and Pitt, 2002; Ma et al., 2009). Extended composite sampling also would likely reduce the likelihood of exceeding a benchmark compared to first-flush sampling because of the expected high first-flush concentration. However, this will come at the expense of increased sampling time and complexity.

Automated Composite Sampling. The most comprehensive approach for assessing pollutant discharges is to install automated composite sampling equipment, connected to stormwater flow measurement devices (calibrated flumes or weirs), both of which are widely available. Composite volume-weighted sampling can be employed to reliably quantify stormwater EMCs and pollutant loads for most pollutants. Protocols for full-storm monitoring are now well established for volume-weighted composite sampling and quality assurance and quality control procedures¹ (Water Quality Program, 2011). Volume-weighted automated composite sampling has multiple advantages. It can reliably collect sample aliquots after passage of specific stormwater volumes to obtain pollutant EMCs. Automated sampling also reduces the labor costs of sampling, although the samplers must be carefully set up and maintained before runoff begins. Additionally, automated samplers can be set to collect samples during a storm that does not occur during normal working hours. Finally, the EMCs resulting from composite sampling are likely to be lower than pollutant concentrations in a grab sample collected during the first flush and therefore may reduce the likelihood of an exceedance that could trigger additional compliance requirements for the discharger. Collection of composite volume-weighted samples over multiple storm events can provide a comprehensive understanding of annual pollutant discharge loads from a site and/or SCM performance over storms of different size.

Nonetheless, automated sampling may not be appropriate for some water quality parameters. For example, samples for bacteria (not an MSGP parameter) are usually not collected with an automatic sampler because of holding time and contamination concerns. Most automatic samplers cannot sample bedload material (e.g., larger solids that are moving in the flow but near the bottom of the pipe or channel and below the sampler intake tube). Therefore, automated sampling may miss sediment and associated pollutants that are traveling very near the pipe bottom. Because they settle quickly, bedload-sized particles are rarely seen in the effluent of structural SCMs. Also, the MSGP uses TSS as the measure of particulate matter, which does not generally include bedload material. The goal, whether in grab or composite sampling, is to find a location where the flow and solids are well mixed (Fischer et al., 1979; Saunders et al., 1983).

Composite volume-weighted sampling can be resource intensive, with initial costs of equipment and installation of the flume/weir, and operational costs with the time and expertise needed to maintain the sampling site, program the samplers, and check equipment calibrations. For a facility, unless required, this level of accuracy may not merit the additional costs, depending on the goals of the sampling scheme.

Other Sources of Error and Variability

Additional choices regarding sampling and analysis, such as sampling design, equipment, analytical protocols, and operator skill, will affect the results. Example sources of sampling and analysis error capable

___________________

¹ See www.bmpdatabase.org/monitoring-guidance.html.

TABLE 3-1
Example Sources of Sampling Error and Variability in MSGP Stormwater Monitoring Results

of causing a difference between true water quality at the desired measuring point and water quality measurements are highlighted in Table 3-1.

Sampling variability can result from inconsistencies in selecting the monitoring point. Because of the diffuse nature of stormwater, isolating a specific discharge point may be challenging. Because of the highly variable conditions at industrial facilities, stormwater sampling points can be difficult to define. Facilities regulated under the current MSGP that have multiple discharges may collect their stormwater samples at one discharge point and list it as being representative of all discharge points. This is a valid procedure if the activities taking place on each drainage area are adequately similar. If the land uses are not similar, such an approach can provide an inaccurate assessment of the pollutant discharges from the site. Where industrial activity is not equally distributed across various discharge points, the MSGP requires that multiple sampling points be included. Complex facilities with a large footprint will need to sample multiple discharge points to represent the myriad activities taking place at the facility.

Once the outfalls are selected, locating an appropriate point in the flow path also is required, as discussed in the previous section on the effects of sampling approaches. The type of equipment used and its installation location also will impact the results (Winterstein and Stefan, 1983; Graczyk et al., 2000; Cristina et al., 2002; Clark et al., 2009). In discharges from large pipes, a single grab sample may not reflect the true volume-averaged or depth-averaged concentration at that sampling time. Collecting samples at different flow depths (depth-integrated sampling) can provide concentrations more representative of the true values (Selbig et al., 2012). Finally, collecting from definable channels is much more repeatable than sheet flow sampling. Sheet flow can be difficult to monitor and multiple samples may have to be collected at different points spatially in order to generate a clear picture of the pollutant concentrations.

Variation in sample processing and analysis protocols can also affect results. The time prior to sample processing in the field or laboratory has been noted as an important factor affecting solids and metals measurements due to the creation and dissolution of flocs in the stormwater over time (Furumai et al., 2002; Kayhanian et al., 2005; Li et al., 2005). The analytical method chosen also can be a factor in variability of results (Gray et al., 2000; Clark and Siu, 2008). Variation in technique and skill among analysts within a single laboratory also has an effect (Clark and Pitt, 2008). In addition to the sources of variability discussed above, other factors that affect sample results could include sample contamination and improper data handling. In all stormwater monitoring situations, it is important to minimize the error and understand and manage the variability.

RECOMMENDED IMPROVEMENTS TO SAMPLING AND ANALYSIS PROTOCOLS

Several opportunities exist to reduce error in industrial stormwater monitoring. In this section, the committee recommends improvements to the MSGP sampling requirements and training for sampling and laboratory personnel. The committee also discusses novel technologies that in the years ahead may offer additional efficiencies and improved accuracy for MSGP monitoring.

Sampling Type

The committee recommends that the Environmental Protection Agency (EPA) allow and promote the use of composite sampling for benchmark monitoring for all pollutants except those that transform or degrade rapidly or are especially time sensitive (e.g., pH). As discussed previously, composite sampling provides more consistent and reliable data and resolves or reduces the problem of intrastorm variability in pollutant concentrations. For sites with treatment SCMs that store water from the previous storm, composite sampling would provide more accurate stormwater discharge results. The committee recognizes that the cost and complexity of managing composite samplers may be more than some permittees want to bear, and therefore, composite sampling should not be required. However, the advantages of composite sampling (including the increased likelihood of meeting the benchmark with an EMC as compared to a grab) may encourage the investment in higher-quality data collection.

The current MSGP requirement for grab samples during the first 30 minutes to capture the first flush is inconsistent with the methods used to derive benchmark thresholds. Technology-based MSGP benchmark thresholds are derived from Nationwide Urban Runoff Program (NURP) values and secondary treatment requirements. NURP values are calculated from event mean concentrations in stormwater discharges that were characterized using composite sampling techniques (EPA, 1983). Secondary treatment requirements are derived from measured performance of publicly owned treatment works over longer time periods, and the values that form a basis for MSGP benchmark thresholds are based on 30-day averaging periods (EPA, 1984, p. 37006). Composite sampling over a period up to 24 hours also better aligns with the time period used to express water quality-based effluent limits in permits. Although EPA recommends that acute aquatic life criteria be expressed as maximum 1-hour average concentrations (EPA, 1985), effluent limits in NPDES permits that are derived from acute aquatic life criteria are expressed as daily average limits (see 40 CFR § 122.45(d)). EPA explains in its Technical Support Document for Water Quality-Based Toxics Control (EPA, 1991) that the 1-hour averaging period was derived primarily from data on response time for ammonia, a fast-acting toxicant. EPA (1991) states that there are scientifically justifiable alternative averaging periods and that, in practice, 1-day periods are the shortest periods for which wasteload allocation modelers and enforcement personnel have adequate data.

Multiple types of composite sampling techniques could be used, including more simplistic manual composite methods that characterize the first few hours of discharge and more complex automated methods that composite samples over a discharge period up to 24 hours. EPA originally required both grab and composite sampling in the 1992 baseline permits and as part of the 1992 group application process. At that time EPA stated that it was necessary to provide information on the first-flush concentration as well as the average concentration of pollutants discharged during an event (EPA, 1992a, p. 41296). In issuing the 1995 MSGP, EPA allowed the use of grab sampling for the vast majority of permit-required sampling, presumably as a burden reduction (EPA, 1995, p. 50828). In issuing the 2008 MSGP, EPA discontinued all remaining composite sampling and essentially disallowed the use of composite sampling in favor of grab sampling to characterize the high pollutant concentration that would occur during a first-flush effect (EPA, 2008b). The limitations of grab samples, including the high variability, are now well understood. Composite sampling technology, which is widely available and relatively inexpensive, could significantly improve the consistency and reliability of stormwater data and should be encouraged.

Monitoring Frequency

The current MSGP allows benchmark monitoring to be stopped for that permit term if the mean concen-

tration of the four previous quarterly measurements is below the benchmark threshold. Given the pronounced variability in the flow and quality of stormwater and the potential for major changes in site management over time, four quarterly samples are insufficient to assess the adequacy of stormwater management at a facility over the course of a permit term of 5 years.

The first concern relates to the number of storm event samples that is sufficient to determine a benchmark exceedance, given the inherent variability in stormwater runoff. Stormwater pollutant concentrations will vary with antecedent dry conditions, stormwater flow rates, industrial activity on the site, and many other factors.

The number of samples required to be statistically confident that the sample mean is less than a specific value, such as a benchmark, is dependent on the acceptable error and the coefficient of variation (COV, standard deviation of the samples divided by the mean) (see Burton and Pitt, 2002; Ott and Longnecker, 2015). Figure 3-1 displays the acceptable difference between the stormwater discharge sample mean and the respective benchmark based on the number of samples and COV of the data set (α = 0.05; power = 0.80).² In an analysis of the National Stormwater Quality Database,³Burton and Pitt (2002) determined that coefficients of variation for stormwater runoff across multiple sites when appropriately categorized by land use, region, and sometimes seasons were 0.5 to 1.0 (measured using composite sampling). Use of grab samples and other sources of error and variability (see Table 3-1) will increase the COV.

For example, with a COV of 1.0 (optimistic for grab samples), with collection of only four samples, the acceptable error is 125 percent, as the difference between the measured mean value and the benchmark. Therefore, for a TSS benchmark of 100 mg/L, any quarterly average concentration from 0 to 225 mg/L is statistically indistinguishable from the benchmark. COV greater than 1.0 would produce a larger range. Reducing this range to a scientifically preferred value, such as 20 percent (80 to 120 mg/L TSS) would require 150 samples at 1.0 COV.

Obviously, this level of sampling is unrealistic. Collection of more samples increases the confidence that a site is complying with the requirements by reducing the acceptable error. Ultimately the decision on the number of samples to require is based on what amount of error is acceptable, relative to the cost of the increased monitoring. Technology verification for SCMs used in municipal stormwater requires monitoring of a minimum of 12 storm events (composite sampling) over a range of storm intensities (with other constraints on type of storm, etc. [Water Quality Program, 2011]).

In addition to the drawbacks of a limited sampling population, the MSGP sampling waiver after four samples poses additional temporal concerns, because a facility would then not be required to monitor stormwater for up to 4 years (or more) until the next permit term begins. Various modifications to the facility, changes in activities at the facility, and turnover in site personnel could take place during the period of monitoring relief, all of which could impact the stormwater discharge characteristics. Structural SCM performance can also degrade over time, if not maintained, usually due to clogging in a media filter or sediment buildup that reduces the treatment volume and increases scour in sedimentation devices. Sustained monitoring can help ensure that permittees continue to implement and maintain SCMs consistently during the entire permit period. More frequent continual sampling allows a consistent representation of stormwater discharge as operations and personnel change over the duration of a permit term. Additional sampling throughout the permit term also helps reduce the uncertainty associated with natural variability among storms and wet versus dry years.

Some states have acted to increase the frequency of chemical monitoring beyond that specified in the MSGP (see Appendix A). Washington allows monitoring relief only after having eight consecutive quarterly samples with concentrations less than the benchmark. California allows a reduction in sampling frequency to

___________________

² The Type 1 error rate (α) is the risk of a false positive, where something is assumed to be true when it is actually false. According to Burton and Pitt (2002), “an example would be concluding that a tested water was adversely contaminated, when it actually was clean. The most common value of α is 0.05 (accepting a 5 percent risk of having a Type 1 error).” Power is 1 – β. Type 2 error rate (β) is the risk of “a false negative, or assuming something is false when it is actually true. An example would be concluding that a tested water was clean when it actually was contaminated. If this was an effluent, it would therefore be an illegal discharge with the possible imposition of severe penalties from the regulatory agency” (Burton and Pitt, 2002).

³ See http://www.bmpdatabase.org/nsqd.html.

once per 6 months after four consecutive samples with concentrations less than the benchmarks.

The MSGP should include a minimum of annual sampling for those that qualify for monitoring relief to ensure that appropriate stormwater management continues throughout the permit term and to provide additional data to indicate the effectiveness of SCMs. Furthermore, EPA should also analyze COVs for sector- and site-specific industrial stormwater data to evaluate the benefits of additional increases in sampling frequencies (such as 2 years of quarterly monitoring, or twice annual monitoring for those with monitoring relief) in terms of reductions in acceptable error.

Role of Training for Sampling and Laboratory Personnel

Data and field experience show substantial differences in the reliability of samples collected by facility personnel as compared to trained (watershed agency)

personnel (K. Schiff, Southern California Coastal Water Research Project, personal communication, 2018). This difference is attributed to the fact that agency personnel are trained in water, wastewater, and stormwater sampling procedures and pollutant transport concepts and have experience with multiple stormwater situations. In contrast, industrial facility staff may not be trained in stormwater concepts or procedures. Inconsistent sampling at a given facility across multiple storms may result from using untrained personnel or by employing different personnel who implement procedures differently. The committee recommends training and guidance, including the possibility of a training/certificate program in stormwater collection and monitoring, to reduce the variation in sampling design and sample collection.

Water quality analysis of stormwater samples is most often performed by private contract laboratories, but may also be done in house, particularly for facilities of large corporations. Some variation in measured pollutant concentrations can routinely be expected due to variability in laboratory methods, individual behaviors, reporting levels, and degree of quality control, which affect the accuracy and precision of measurements (Clark and Pitt, 2008). For example, for TSS, three methods are approved in 40 CFR § 163.3,⁴ which yield different results for known concentrations of stormwater solids even when a well-trained analyst is performing the analysis, with results varying by up to 25 percent (Gray et al., 2000; Clark and Siu, 2008). There was added variability when different well-trained analysts measured TSS using one of the methods. Much of this variability was attributed to the methods employed to obtain the aliquot used in the analytical method and the difficulty in capturing larger, heavier particles in the subsampling of the initial sample.

The industrial stormwater matrix poses particular challenges in analysis because many of the pollutants of concerns (e.g., metals, organics) are likely to sorb to solids. For some organics, the difficulty of extracting the pollutant into an aqueous or solvent phase for measurement can result in matrix interferences that reduce the accuracy of the analytical method (EPA, 1986).

Laboratory certification programs evaluate and certify the technical competence of laboratories to perform specific types of testing and measurements. While the Safe Drinking Water Act contains laboratory certification requirements, other major federal environmental statutes including the Clean Water Act do not contain similar requirements. Some states have independently established certification programs for additional environmental media, including wastewater, solid and hazardous wastes, and air samples, and require the use of certified laboratories through their own statutes and regulations. The National Environmental Laboratory Accreditation Program (NELAP) exists to promote technical competence of environmental laboratories and develops nationally recognized standards for accreditation.⁵ For stormwater, EPA encourages the use of laboratories certified by agencies accredited by NELAP (EPA, 2009a). Minnesota is an example of a state that has a Clean Water Act laboratory accreditation program and requires use of an accredited laboratory in its MSGP (MPCA, 2015). Because stormwater is distinct from wastewater, it is important to understand to what extent the laboratory certification program includes evaluation of technical competence with the stormwater matrix.

Periodic interlaboratory calibration programs represent another approach that has been used to promote the comparability of testing results among laboratories. These programs bring standardization to analytical procedures for stormwater samples. Such programs facilitate communication among laboratory personnel, help set performance-based criteria to measure success, and utilize a locally derived stormwater matrix while improving comparability and reliability of stormwater sampling results used to determine compliance with water quality benchmarks. After interlaboratory coefficients of variation were noted in California that were greater than 40 percent for some stormwater pollutants, the Southern California Stormwater Monitoring Coalition (SMC) developed guidelines and protocols to ensure comparability of laboratory results for stormwater samples (Gossett and Schiff, 2010). The SMC also instituted interlaboratory calibration exercises⁶ that have reduced variability

___________________

⁴ Standard Methods SM 2540D-2011, USGS I-3765, and ASTM D5709-18.

⁵ See https://cfpub.epa.gov/si/si_public_record_report.cfm?dirEntryId=56216.

⁶ See http://socalsmc.org/smc-regional-stormwater-monitoringcomparison-and-evaluation.

(K. Schiff, SCCWRP, personal communication, 2018), but such regional efforts are rare.

To reduce analytical variability and improve the utility of monitoring results, the committee recommends that EPA encourage state adoption of national laboratory accreditation programs for the Clean Water Act with a focus on the stormwater matrix and on reducing the variabilities associated with stormwater pollutants that have been noted above. EPA should initiate and encourage interlaboratory calibration efforts, including the establishment of performance quantification levels developed from samples with a stormwater matrix. EPA should publish guidance and case studies on interlaboratory calibration, with specific focus on known challenges to stormwater analysis (e.g., solids capture, matrix interference). These efforts would promote the comparability and reliability of test results reported to permitting authorities.

New Methodologies or Technologies for Industrial Stormwater Monitoring

In this section, a few examples are offered of potential improvements in monitoring technology for industrial stormwater discharge; some of these are available today, and others may be reasonably expected in a not-too-distant future. Monitoring technology is considered here in the broadest sense and includes hardware, software, sensors, sampling techniques and timing, mobile technologies, and apps.

Visual monitoring information could be addressed with the future development of mobile apps that may be useful in identifying stormwater clarity, sheens, or other visual water quality indicators via still imaging or video. Drone imaging may be useful in visual stormwater discharge monitoring and delineation of drainage areas and covered/exposed areas.

Sensors and real-time control are ubiquitous in process control, water quality measurements, and documenting water quality in drinking water and wastewater treatment facilities of all sizes. Stormwater applications are obviously complicated by the episodic and dynamic nature of stormwater flows and quality. Advances in sensors can lead to improved monitoring of stormwater discharges and SCM performance. Field-employable flow/moisture sensors are available now, as are turbidity, pH, dissolved oxygen, and temperature sensors. Data can be collected in the field in real time over wireless networks. Reliable field turbidity sensors now available could open up the possibility of using turbidity as a surrogate for TSS in future permit terms. For construction site erosion, turbidity, instead of TSS, is used to document performance of erosion control practices in several states, such as Vermont (State of Vermont, 2008).

Field sensors that accurately and reliably measure other water quality parameters are currently available, although they are somewhat costly. A wider range of sensors at lower cost is anticipated to be available in the near future. Newly developed sensors may be able to directly measure concentrations of water quality parameters of interest, or may measure useful surrogates of water quality. The overall result will be more reliable monitoring of industrial stormwater discharges. The use of real-time control of SCM performance using water level/storage information and weather predictions is also possible (Kerkez et al., 2016).

Future revisions of the MSGP monitoring requirements should consider advances in sensor technology and reductions in costs since the previous permit release. By considering the latest technology, EPA can take advantage of opportunities to improve the value of the MSGP monitoring program, providing more or better-quality information for similar or reduced costs.

TIERED APPROACH TO MONITORING

The current MSGP monitoring approach could be substantially improved to provide more useful information on the quality of stormwater discharges and their impact on receiving waters while balancing the net burden to industry and limited agency resources. In this section, a tiered approach, with different levels of inspection or monitoring according to risk, complexity, and past performance of stormwater management, is recommended. This approach is expected to provide better overall protection of the environment and public health.

Proposed Categories of Monitoring

The committee envisions a framework (see Table 3-2) where the most complex, high-risk facilities or those with recurring exceedances would be

TABLE 3-2
Table of Criteria and Implications for Proposed Monitoring Tiers

NOTE: In all categories, no exposure exemptions may be granted. Additional monitoring may be required for facilities with effluent limitation guidelines (see Appendix B).

required to conduct more-sophisticated monitoring to assess their impact to receiving waters and target future improvements to stormwater control measures, consistent with the recommendations of the National Research Council (NRC, 2009). Those facilities required to conduct benchmark monitoring would continue to do so, but a much larger number of facilities that currently conduct only visual monitoring would also monitor for TSS, pH, and chemical oxygen demand (COD) as part of industry-wide monitoring (see Chapter 2). Facilities with low risk and low likelihood of substantial pollutant discharges could replace water quality monitoring with rigorous inspections and reporting. Details of the monitoring categories proposed by the committee are provided below.

Inspection Only (Category 1)

The committee recommends that EPA define and create a category for facilities that could rely on inspection by a permitting authority or certified inspector as a complete alternative to chemical discharge monitoring. Providing an option for inspection in lieu of monitoring can reduce the burden on small, low-risk

facilities while improving stormwater management. This recommendation is distinct from how the current MSGP operates. Facilities would be allowed to rely on inspections as an alternative to chemical discharge monitoring if they exhibit a low risk of contributing to water quality problems via stormwater discharge. This determination would be facility based rather than sector based and would be verified by a certified inspector. Given the limited capacity for permitting authorities to perform compliance inspections, the committee envisions that certified inspectors would be contracted by many of the permittees to conduct the inspection. The certified inspector could be an employee of a municipal separate storm sewer system (MS4), a private third-party company, or a parent corporation, as long as the inspector is not directly involved in the day-to-day operation or oversight of the facility being inspected.

The value of facility inspections to evaluate site conditions and pollutant discharge potential was recognized early in the implementation of industrial stormwater permitting (see Box 3-1). In-person education was highlighted by several state permitting agencies in meetings with the committee as an effective means to improve stormwater management and permit compliance at small industrial sites. Small businesses may also have difficulty collecting reliable monitoring data, because they may have limited financial resources or limited staff to develop monitoring expertise. Many of those businesses, and some larger businesses, operate small facilities that qualify as low risk. For those facilities, both the permittee and the permitting authority could potentially be better served with a rigorous inspection instead of monitoring. A site inspection by a certified inspector would increase the reliability of the results, and the business may welcome the exemption from chemical monitoring.

The committee recognizes the difficulty in defining the characteristics of a low-risk facility. For a site to be considered at low risk of impacting water quality, it should have a low likelihood of discharging toxic substances in toxic amounts, generally have a small area of exposed industrial activity, and be well managed. A simple approach would be to base low pollutant risk on the amount of surface area that contains industrial activities exposed to stormwater. Many very small facilities (e.g., less than 0.5 to 1 acre of industrial activities) have relatively few activities exposed to stormwater that pose a pollution risk to water quality, assuming that the facility is not part of larger network of integrated operations at multiple facilities. Additionally, a site that has little area of exposed industrial activities reasonably may be expected to discharge lower volumes of industrial runoff (and corresponding lower mass load of pollutants) for a given size of precipitation event compared to facilities with larger industrial operations. A 1-acre threshold is used to distinguish the relative risk of water quality impacts associated with discharges of

stormwater associated with construction activity (EPA, 1999). However, several states established a smaller area threshold for exempting certain erosion controls, especially in high-value watersheds.

A criterion of facility area, although simple to implement, nonetheless is not a robust indicator of risk. A small industrial facility may or may not store hazardous materials, handle materials in large volume, or rely on outdoor equipment or operations that release stormwater pollutants during operation. Small facilities in some industry categories have the potential to generate substantial amounts of pollutants capable of causing harm in receiving waters. In urban areas, clusters of small facilities in aggregate may generate substantial discharges. At the same time, many midsized industrial facilities conduct limited activities exposed to stormwater, for which effective management strategies are relatively easy to implement and maintain. Industrial stormwater discharges from these mid-sized facilities would be expected to produce much lower pollutant mass loadings compared to smaller facilities with more active operations. Research has documented substantial variation among facilities of a given sector in the type, extent, and intensity of industrial activities they conduct that may be expected to govern a facility’s risk of discharging pollutants (Swamikannu et al., 2000; Cross and Duke, 2008). Because industrial facilities are so highly variable, classifying facility risk is most accurately based on a characterization of the intensities and types of industrial activities conducted at each facility.

Specific criteria could be developed that characterize the presence or absence of activities considered likely to generate stormwater pollutants that could cause water quality problems. The criteria could be similar to those developed for no exposure exemptions, which describe in narrative form all activities that, if present, would preclude a facility from qualifying for the exemption. The criteria envisioned here are admittedly more complex, because they are intended to assess the expected magnitude or scale of pollutants from activities rather than simply the presence or absence of an activity. However, the process of establishing criteria is the same, and EPA can rely on its experience and the experience of the states to define activities that may reasonably be expected to discharge toxic pollutants in toxic amounts during routine operation. Examples of possible criteria for low pollutant discharge risk are presented in Table 3-3. Conformance to the criteria would be verified by an inspection.

Also important in the determination of a low-risk facility is certification that the site is well managed. To assess this the committee recommends that the facility inspection be conducted at least once per permit term and include the elements of a stormwater compliance inspection, such as

• Reviewing the permit and the stormwater pollution prevention plan (SWPPP) and determining whether the SWPPP meets the requirements set forth in the permit;
• Reviewing records, including self-inspection reports, to verify that the permittee is complying with the permit and the SWPPP;
• Walking the site and verifying that the SWPPP is accurate and that the SCMs are in place and functioning; and
• Identifying actions that need to be taken to effectively manage stormwater pollution.

In addition, inspections can provide opportunities to educate facility operators on the most effective steps to improve stormwater management.

A publicly accessible report filed with the permitting authority would document the findings of the inspection and any specific concerns and recommendations for additional SCMs. Current facility conditions would be compared to previous conditions documented in prior inspection reports. If the inspection indicates substantial concerns, recurrent problems that have remained unaddressed, or a lapse in inspections, the permitting authority or inspector could recommend the facility be placed in another category that would include required chemical monitoring. Local entities such as MS4 permittees, agencies responsible for total maximum daily load (TMDL) implementation, or those responsible for other watershed protection programs could also petition the permitting authority to exclude from Category 1 industry types or individual facilities that are found to be potentially discharging pollutants that are causing or contributing to the impairment of receiving waters.

Because inspection would serve as an alternative to chemical benchmark monitoring, an inspector certification program (see Box 3-2) is recommended to promote

TABLE3-3
Example Criteria for Determining Low-Risk Facilities (Category 1)

NOTE: These criteria are intended to lead to a determination that the type, intensity, and extent of industrial activities are unlikely to generate discharges of pollutants of a kind and a quantity that may cause or contribute to water quality problems in receiving waters. The intent is to create a category of facilities that do not meet the rigorous criteria of “no exposure” but encompass facilities with activities that are small but nonzero in spatial extent, frequency, intensity, and/or presence of residuals. These are committee suggestions, but EPA should develop concrete and implementable criteria conditions.

confidence in the thoroughness and reliability of results. The certified inspector would evaluate the facility’s SCMs and conformance to the criteria for this low-risk category. An inspector certification program would provide a means to certify and track the credentials of the inspector, promote inspector accountability, and help inspectors stay current with the latest developments, skills, and technologies available to promote MSGP permit compliance.

Industry-wide Monitoring (Category 2)

The committee recommends that the MSGP continue to have a category of facilities that are not subject to sector-specific benchmark monitoring based on a determination that they do not have the potential to discharge sector-specific pollutants at a level of concern. However, the committee recommends that all facilities without sector-specific benchmark monitoring conduct industry-wide monitoring for pH, TSS, and COD, as discussed in Chapter 2, in addition to the currently required visual monitoring of stormwater discharge and routine site inspections by facility staff. As also discussed in Chapter 2, the committee recommends that EPA conduct data and literature reviews prior to the next permit renewal and, as a part of each permit renewal, determine whether benchmark monitoring should be added for some industries currently exempted from benchmark monitoring, and whether the pollutant-specific benchmark monitoring requirements for each sector should be revised.

Benchmark Monitoring (Category 3)

Chemical-specific benchmark monitoring in the current MSGP applies to 55 percent of industrial permittees⁷ (R. Marcus, EPA, personal communication, 2018). These permittees are classified within sectors for which it has been determined that potential pollutant levels warrant such monitoring and SCMs are reasonably available for additional pollutant reduction (see Table 1-1). The committee recommends ongoing use of this category. Nevertheless, the specific benchmark monitoring requirements should be updated based on recent data and literature to reflect recent knowledge of pollutant toxicity, sector risks, and stormwater management capabilities, as discussed in Chapter 2. All facilities with benchmark monitoring should also conduct industry-wide monitoring for pH, TSS, and COD, in addition to the currently required visual monitoring of stormwater discharge and routine site inspections by facility staff.

Enhanced Monitoring (Category 4)

A fourth category of enhanced monitoring is envisioned for industrial facilities with the highest risk for discharging pollutants that may adversely impact surface waters. This designation should be based on past repeated exceedances of benchmarks (e.g., AIM Tier 3; see Box 1-3), severe concerns raised upon site inspections of Category 1 facilities, or recommendations of the permitting authority for sites that are large and complex with high pollutant discharge potential or where TMDL development and implementation merits additional monitoring. The largest facilities will typically produce the greatest volume of runoff, leading to high risk from high pollutant mass loads. Complex sites could include those with multiple outfalls and varying land uses throughout the industrial site or high-risk chemicals used in exposed areas.

Monitoring plans would be developed as appropriate for the site and the site issues that need to be addressed. For sites with repeated exceedances, facilities may need to monitor at multiple outfalls and to implement volume-weighted composite monitoring to calculate stormwater discharge event mean concentrations to help to determine whether they are causing or contributing to violations of water quality criteria. For the largest and most complex sites, facilities would be expected to develop and implement a sampling program that is spatially and temporally representative of stormwater discharges from all parts of the facility where industrial activities are conducted. This information may be needed to determine whether and where additional stormwater control measures are warranted, including moving exposed industrial activities under cover or enhanced treatment. Should monitoring and subsequent actions be implemented that bring the site into compliance, the permitting authority could evaluate whether the facility can return to Category 1, 2, or 3.

___________________

⁷ Data applicable to the states and territories permitted by the federal MSGP and does not include data from states with delegated regulatory authority.

For facilities that use event mean concentrations to determine compliance, some consideration is needed regarding extreme storms. EPA should establish a “nonrepresentative storm” criterion that would exclude event mean concentration data for extreme events that are expected to exceed SCM design criteria. Under extreme conditions, SCM performance will be compromised and stormwater bypass will occur. It is reasonable to expect that the discharge of stormwater pollutants associated with industrial activity and the effectiveness of stormwater control measures implemented are most representative for water quality purposes when the sampling is conducted on discharges resulting from frequent storm events and not large extreme events. This event size may be based on a statistical review of long-term rainfall records to establish wet weather precipitation conditions when they become less relevant for water quality. This criterion may be a storm of a certain return frequency such as a 10-year storm, or a multiple of the 90th percentile rainfall depth, or a multiple of the long-term average rainfall depth for the area. Using nonrepresentative storm criteria, a permittee would either not submit EMC data from storms that exceed the criterion or these data would not be evaluated against the benchmarks.

Enhanced stormwater monitoring is considered to be within the financial resources and/or expertise of a major industrial facility and may prove beneficial to the industry by more accurately characterizing the stormwater discharge than by using grab-sample first-flush benchmark monitoring. Full-storm data can provide a much more complete picture of the industrial stormwater discharge from a site. Additionally, when faced with designing treatment SCMs for a high-risk and/or complex site, the flow and water quality data collected by composite sampling are critical to ensuring the sizing and design are appropriate. The MS4 entity could be an active participant with Category 4 facilities, potentially reimbursed to conduct the enhanced monitoring on behalf of the larger facilities in the watershed, so that the data are consistent and useful both at a site level and on a watershed basis.

Benefits of Tiered Monitoring Requirements

The current MSGP includes several levels of monitoring based on expected sector-specific stormwater pollutant discharge. The committee encourages EPA to add both enhanced and reduced levels of monitoring to the existing program. The elimination of benchmark monitoring by low-risk facilities would provide a nonmonitoring option for oversight of these facilities and eliminate some of the most suspect, unreliable monitoring data. This approach also ensures that high-risk industries that are more likely to be significant sources of stormwater pollution invest in the necessary monitoring to confirm that SCMs are effective in reducing pollutants and risks to receiving waters. In total, this proposed framework is expected to reduce the monitoring burden on the lowest-risk facilities while increasing the quality of the data available on the overall population of industrial facilities including the largest, highest-risk facilities. Combined with suggested improvements to monitoring protocols, training, and data management discussed in this chapter, the tiered approach is also expected to increase the usefulness of the data collected toward improving the management of industrial stormwater.

Exemptions, Additions, and Other Permitting Alternatives

Within the tiered framework envisioned by the committee, there are exceptions, additional monitoring, and other permitting options as are currently applicable to the current MSGP.

No Exposure

No-exposure certification is allowed under the current MSGP for sites, regardless of size or complexity, at which “all industrial materials and operations are protected by a storm resistant shelter to prevent exposure to rain, snow, snowmelt, and/or runoff” (EPA, 2015d). With no-exposure certification, required once every 5 years, facilities are exempt from the requirements of the MSGP, including monitoring. Certification requires facility owners to confirm no-exposure conditions by answering specific questions about industrial materials or activities exposed to precipitation and to allow the permitting authority to inspect the property, although such inspections are rarely conducted.

The committee agrees that monitoring is not needed at facilities with no exposure but recommends

verification of no exposure by a certified inspector or the permitting authority. Maryland is an example of a jurisdiction that currently requires third-party verification of no exposure.

Effluent Limitation Guidelines

As discussed in Chapter 1, EPA has established effluent limitation guidelines (ELGs) for 10 subsectors of industrial facilities (see Appendix B), with required monitoring at least once per year at each outfall. This ELG monitoring, required by law, would supplement the MSGP monitoring envisioned in Table 3-2.

Individual Stormwater Permit Monitoring

In its original regulatory strategy for industrial stormwater (EPA, 1990, p. 48002), EPA identified an individual permit category for situations where the MSGP benchmark monitoring requirements, SWPPPs, and SCMs may be inadequate to address pollution from stormwater discharges associated with industrial activity. Federal regulations empower the permitting authority to exclude facilities from the MSGP and require individual NPDES permits when special considerations such as a large quantity of pollutant discharge, proximity to receiving waters, and the characteristics of pollutants are at issue (40 CFR § 122.28(b)(3)). Extensive stormwater discharge characterization for conventional and nonconventional pollutants, toxic pollutants, hazardous substances, and treatment units must be submitted with the permit application. Based on this information, an individual stormwater permit can require more extensive monitoring and/or a greater number of pollutants compared to the MSGP, where benchmark monitoring is determined by standard industrial classification code. Individual permits can also be structured with enforceable discharge criteria expressed as numerical effluent limits, which trigger a permit violation if exceeded. This stricter enforcement of pollutant exceedances can be helpful for sites that represent a high public concern or that raise environmental justice issues. Federal law authorizes any “interested person” to petition the permitting authority to require an individual permit.

Advanced Analyses Possible Under Enhanced Monitoring

Under the AIM process (still to be developed) and the enhanced monitoring category envisioned within the tiered framework for large, complex sites with repeated benchmark exceedances, there are opportunities to use advanced tools and analyses to better understand water quality impacts from individual facilities. These tools, such as wet weather dilution or the biotic ligand model, may require monitoring of receiving water flows or quality, more complex sampling techniques, and modeling, so they are not viewed as tools that should be required of all permittees. Nevertheless, for facilities struggling with repeated exceedances, these advanced tools and analyses can clarify where further SCMs are necessary to protect receiving water quality.

Wet Weather Dilution and Mixing Zones

Many MSGP benchmarks are based on water quality criteria (see Table 1-3) and all MSGP benchmarks are applied at the point of discharge without dilution. By its very nature, industrial stormwater discharges occur during wet weather conditions when the receiving stream is expected to be flowing at some reasonable capacity above base flow, which could provide dilution of stormwater discharges. NPDES regulations allow for municipal and industrial process wastewater discharges to incorporate dilution and an impacted mixing zone when evaluating instream toxicity. According to EPA (2014), a mixing zone is “a limited area of volume of water where initial dilution of a discharge takes place and where certain numeric water quality criteria may be exceeded.” State regulations generally limit these areas based on widths or cross-sectional areas and lengths on a case-by-case basis, and the use of mixing zones is at the discretion of the permitting authority.

Explicit inclusion of a dilution allowance in deriving benchmark thresholds for the MSGP has not been done by EPA and would be challenging, given the state-to-state variability in how mixing-zone allowances are included as part of state water quality standards and the site-specific analysis normally conducted to implement the allowance for a discharge. However, facilities that repeatedly fail to reach benchmarks and are elevated to the upper tiers of the AIM process should be permitted

a mixing-zone allowance, as is allowed with municipal and industrial process wastewater dischargers, after the facility has applied all reasonable SCMs. A mixing-zone allowance would allow facility operators to set site-specific criteria that are protective of ambient water quality in the receiving waters.

Calculating a stormwater mixing zone based on best available science may require the use of data sets characterizing upstream flow and water quality conditions and dynamic water quality models to understand the impact of stormwater runoff on receiving waters. These water quality models are typically calibrated with site-specific water quality and hydrology data. Applicable water quality model types may be “far field,” where water quality is influenced by the hydrodynamics of the receiving water, or “near field,” where pollutant concentrations at the discharge location are determined from plumes at the facility outfalls (Gawad et al., 1996; Jirka et al., 1996; Davis, 2018). EPA should develop guidance for using water quality models for calculating stormwater mixing zones.

Alternative Metals Benchmarks

The 2015 MSGP requires total metals analyses (rather than dissolved), but questions have emerged from industry about whether dissolved metal analyses or the biotic ligand model would provide a more accurate assessment of stormwater pollution. Both approaches require more rigorous monitoring that may be a burden if applied uniformly to all permittees. However, if permittees have repeated exceedances of metals benchmarks (see Appendix D), they may benefit from enhanced monitoring of dissolved metals or in support of the biotic ligand model.

Dissolved Metals. Dissolved metals are more biologically available than particulate-bound metals and are more important in assessing pollutant risk. According to EPA (1996b),

Dissolved metals are used to determine acute and chronic aquatic life criteria. According to EPA (1996b), dissolved metals are operationally defined as “that which passes through a 0.45 µm or a 0.40 µm filter.” With this operational definition, a fraction of the metals measured as dissolved consists of small particulate or colloidal metals that are able to pass through the filter or metals that are complexed with organic ligands, which may not be biologically available. Dissolved metals require field or laboratory filtration within 15 minutes of sample collection (40 CFR § 136.3) because metal species continue to change between dissolved, precipitated, and sediment-sorbed forms after the sample is collected.

Several studies have been conducted to characterize metal concentrations in urban stormwater based on total metals (Pitt et al., 2004a,b; Shaver et al., 2007). In a number of stormwater studies, a significant fraction (approximately 30 to 70 percent) of copper, cadmium, and zinc was found in the dissolved form (Pitt et al., 1995; Crunkilton et al., 1996; Sansalone and Buchberger, 1997; Pitt and Clark, 2010). Differences in stormwater chemistry, receiving water chemistry, temperature, and sediment composition will affect the fraction of metals that are bound or dissolved (Weiner, 2008). Runoff that is collected from receiving waters will often have higher amounts of metals in particulate form while stormwater collected from pipes will have a higher dissolved fraction (Clary et al., 2011).

Because dissolved metal concentrations provide a more accurate measure of potential toxicity, it would be reasonable for the MSGP to allow industries that have had repeated exceedances of benchmark levels for total metals to sample for dissolved metals and compare this quantity against the existing benchmark. However, sampling for dissolved metals requires more complex sampling methodology, including filtering within 15 minutes of sampling. Because rapid filtering for dissolved metals puts an additional burden on industry, the committee does not recommend that dissolved metals analyses be required for all permittees covered by the MSGP, but should be an option if all proper sampling procedures are followed.

Biotic Ligand Model. As discussed in Chapter 2, the Biotic Ligand Model (BLM) is an aquatic toxicology tool that is used to determine the bioavailability

of metals in aquatic ecosystems. Lethal accumulation values of metals on the gill surface, when fish toxicity is being considered, are used to predict lethal metals concentration values with the BLM (Niogi and Wood, 2004). EPA already uses the BLM as a tool in the Ambient Water Criteria in surface waters (Jarvis and Wisniewksi, 2006), but to develop a BLM, site-specific water quality parameters, including hardness, pH, and dissolved organic carbon, need to be measured. As with dissolved metals discussed in the previous section, the MSGP should allow those who exceed total metals benchmarks to analyze receiving waters to calculate pollutant toxicity associated with a facility’s stormwater discharge. However, the facility would need to do additional sampling beyond the current MSGP requirements to acquire the data needed by the BLM.

Watershed-based collaborative relationships among industries, municipalities, and other dischargers could help facilitate the characterization of receiving water chemistry, as required for use of the BLM, at reduced cost. Multiple dischargers could combine resources to appropriately characterize the necessary water quality parameters over a range of flows, seasonal variations, and other important conditions. With these data, BLM modeling could be completed to establish watershed-specific benchmark concentrations for copper for all dischargers to the receiving water. This characterization procedure for copper and the BLM has been done in Oregon (OR DEQ, 2018). Collaborative monitoring could be expanded to other pollutants that need receiving water quality information to determine discharge concentrations.

UPDATING AND UPGRADING CURRENT METHODS OF DATA MANAGEMENT

Submitting, managing, and reviewing data collected under the MSGP has been challenging. In 1995 when the first MSGP was issued, EPA’s national data system for the NDPES program did not accommodate stormwater permits. It could not address the need to enter the type and numbers of sources and reporting against benchmarks instead of enforceable numeric limits. The data system had been in place since 1982, 5 years prior to Congress’s action to expand stormwater permitting (EPA, 2013, p. 46011). The transition to an updated national data system, the inclusion of stormwater permits in the system, and the eventuality of self-reporting monitoring data into the system has taken years.

With the 2015 MSGP, EPA required as of December 2016 that permittees submit their discharge monitoring reports (DMRs) electronically, including those operating under EPA or a state MSGP, into the national eDMR data system, unless a waiver is obtained (EPA, 2015a, p. 64066). Prior to 2015, monitoring was often submitted in paper format, making review of these data and permit compliance cumbersome and staff intensive. When EPA reviewed benchmark monitoring data for development of the 2015 MSGP, only 485 of the 1,200 covered facilities required to perform benchmark monitoring submitted their results electronically and many of the records were unusable (EPA, 2012). Data collected outside of the MSGP have no single or linked repository for storage and public access.

As part of the information gathering conducted for this study, states acknowledged that they have been limited in their ability to receive, review, and respond to MSGP monitoring due to staffing shortfalls. However, many states reported that they do have the capacity to review data electronically and that digital reporting improves the effectiveness of staff oversight, particularly in states with limited staffing. Automated searchable data systems streamline environmental compliance. States using these systems can autogenerate reminders and compliance advisories. The level of electronic reporting is increasing as permittees become aware of and adept at electronic reporting, state data systems capabilities grow, and compliance rates increase. States are required to share MSGP monitoring information with EPA via the national data system, and data sharing is increasing. Two particular advantages arising from improving data management tools are an improved capacity to screen data automatically for outliers and errors and to analyze large data sets using data visualization software.

Screening for Errors and Omissions

The new era of automated data systems and electronic self-reporting offers many opportunities for improving data quality. Illegible discharge reports are eliminated. Permittees enter results into screens prepopulated with information on outfalls, sampling fre-

quencies, parameters, and units. If consistent units for benchmarks are used based on the value from which the benchmark was derived, as recommended in Chapter 2, unit errors could be substantially reduced. The systems have the capability of providing permittees immediate electronic feedback, such as by alerting or requiring facilities to check and correct decimal point placement and verify results that exceed the benchmark threshold, helping to reduce transcription errors. Several entries among the 2015 MSGP appeared erroneously high or low, suggesting the data management system could improve its alerts to permittees of outlier data or “less than” values that exceed the benchmark, thereby further reducing errors (see Appendix D).

Electronic data can also improve agency oversight. EPA and states can generate automated reports, which streamlines the identification of omissions and exceedances. More complete information regarding whether a facility did in fact report a discharge during the monitoring period is becoming available.

Data Analysis and Visualization

With improved data quality and more data becoming available through application programming interfaces and web services, the ability to evaluate the data for patterns, trends, and correlations is expected to increase. For example, California Water Boards’ data center contains industrial stormwater effluent water quality data and an assessment tool that can be used for quick illustration of query results. These visualizations can then develop “data stories” that help to understand industrial stormwater effluent quality and progress made under the MSGP.

A simple visualization example using California Water Boards’ tool compiles the most recent 5 years of data by facility and compares the median value (using a minimum of three samples) to the benchmark. Facilities where the median is below the benchmark are shown on the left side of Figures 3-2 and 3-3 for lead and TSS, respectively, and facilities where the median is above the benchmark are shown on the right side of each figure (note that the scales are different). A comparative analysis by pollutant indicates that TSS is a greater water quality challenge compared to lead for facilities covered by the California equivalent of the MSGP.

A quick temporal analysis can be made using this visualization for lead. Given that EPA only included benchmarks in the 1995 MSGP in cases where the median of all samples for a given sector exceeded the potential benchmark value, and that the benchmark threshold for lead has not changed since the 1995 MSGP, the fact that relatively few facilities currently exceed the benchmark indicates that lead pollutant levels have significantly improved during the time period the MSGP has been in place. This may be explained by improvements in pollution prevention measures that have been implemented in this time period, including the removal of lead from gasoline and

paints and improved housekeeping at sites. For TSS, the data visualization could be used to target compliance assistance efforts; for example, it could be used to find opportunities where treatment SCMs could supplement existing site measures to reduce pollution discharge levels.

The committee recommends that EPA continue to compile data for facilities operating under the EPA MSGP and state MSGPs nationally and make these data publicly available in a timely manner. The committee also recommends that EPA develop visualization tools that can be used by others to easily examine data for patterns, trends, and correlations.

CONCLUSIONS AND RECOMMENDATIONS

The current MSGP benchmark monitoring requirement focuses on low-cost, coarse indicators of site problems, and the usefulness of the data can frequently be hampered by its variability. Stormwater monitoring data display variability that originates from many different sources, including the variability of precipitation within and among storms and changes in operations over the course of time. In this chapter, the committee recommends improvements in sampling design and procedures, laboratory analysis protocols, and data management to reduce error and improve the reliability of monitoring results to support improved stormwater management.

EPA should update and strengthen industrial stormwater monitoring, sampling, and analysis protocols and training to improve the quality of monitoring data. Specifically, EPA should

• Consider a training or certificate program in stormwater collection and monitoring to ensure that required sampling and data collection are representative of stormwater leaving the site to the greatest extent possible.
• Stay abreast of advancements in monitoring, sampling, and analysis technology that can provide more or better-quality information for similar or reduced costs and consider these in future revisions of the MSGP.

EPA should allow and promote the use of composite sampling for benchmark monitoring for all pollutants except those affected by storage time. EPA’s disallowance of composite sampling and reliance on grab sampling in the interest of discrete characterization of the highest pollutant concentration is not warranted based on the methods used to derive benchmark thresholds. Multiple composite sampling techniques are available that provide more consistent and reliable quantification of stormwater pollutant discharges compared to a single grab sample. Composite samplers have become common in stormwater monitoring as experience with this approach has increased and costs have declined, and the EMCs that result from composite sampling may reduce the likelihood of exceeding the benchmark compared to first-flush grab sampling. Composite sampling is not appropriate for pollutants for which the results may vary over time with

storage, such as those that transform or degrade rapidly or interact with the atmosphere (e.g., pH).

Quarterly stormwater event samples collected over 1 year are inadequate to characterize industrial stormwater discharge or describe industrial SCM performance over the permit term. Under the MSGP, if a permittee’s average of four consecutive quarterly samples meets the benchmark, a waiver is granted for the remainder of the permit term. For permittees with average results that meet the benchmark, the MSGP should require a minimum of continued annual sampling, to ensure appropriate stormwater management throughout the remainder of the permit term. Extended sampling over the course of the permit would provide greater assurance of continued effective stormwater management and help identify adverse effects from modifications in facility operation and personnel over time. Given the natural variability and the limitations of grab samples, substantial uncertainty is associated with using the average of only four stormwater samples. EPA should analyze industrial stormwater data and sector-specific coefficients of variation to recommend additional increases in sampling frequency, consistent with EPA’s determination of an acceptable level of error for this indicator of SCM performance. Additional continued monitoring at a lower intensity throughout the permit would also increase the overall sample size and thereby reduce the uncertainty in the monitoring results.

State adoption of national laboratory accreditation programs for the Clean Water Act with a focus on the stormwater matrix and interlaboratory calibration efforts would improve data quality and reduce error. NPDES laboratory accreditation programs and stormwater interlaboratory calibration efforts would improve the comparability and reliability of monitoring data. To support these efforts, EPA should publish guidance and case studies on interlaboratory calibration specifically focused on the stormwater matrix, including the establishment of performance quantification levels for stormwater samples. These efforts would promote similar procedures at a national level to ensure the comparability and reliability of test results reported to permitting authorities.

To improve stormwater data quality while balancing the burden of monitoring, EPA should expand its tiered approach to monitoring within the MSGP, based on facility risk, complexity, and past performance. The committee proposes four categories:

1. Inspection only. Low-risk facilities could opt for permit-term inspection by a certified inspector or the permitting authority in lieu of monitoring. Facilities could be classified as low risk based on facility size (e.g., less than 0.5 or 1 acre of industrial activity), recognizing that size may not fully represent the risk profile, or more accurately based on a detailed assessment of the type and intensity of industrial activities conducted on site, or a hybrid approach.
2. Industry-wide monitoring only. All facilities in sectors that do not merit additional pollutant monitoring would conduct industry-wide monitoring for pH, TSS, and COD. These data would provide broad, low-cost indicators of the effectiveness of stormwater control measures on site.
3. Benchmark monitoring. Sectors that merit additional pollutant monitoring, based on the most recent data and industry literature review, would conduct sector-specific benchmark monitoring in addition to pH, TSS, and COD, which would be collected by all facilities with chemical monitoring.
4. Enhanced monitoring. Facilities with repeated benchmark exceedances or those characterized by the permitting authority as large complex sites with high pollutant discharge potential would conduct more rigorous monitoring, in consultation with the permitting authority. These facilities could collect volume-weighted composite samples at multiple outfalls if appropriate. Additional tools and monitoring strategies could be used to assess the water quality impact to receiving waters from stormwater discharge, including wet weather mixing zones, dissolved metal sampling, and site-specific interpretation of water quality criteria, with additional guidance from EPA. EPA should develop “nonrepresentative storm” criteria to exclude monitoring for events that would not be representative of facility stormwater discharge.

This tiered system would improve the overall quality of monitoring data to inform future iterations of the MSGP while balancing the overall burden to industry and permitting agencies.

To improve the ability to analyze data nationally and the efficiency and capability of oversight by permitting agencies, EPA should enhance electronic data reporting and develop data management and visualization tools. Electronic reporting has only been required of permittees since 2016, and the data management capabilities are still developing to make the most use of this information at the national and state levels. Automated compliance reminders, improved checks on missing or unusual data, and data analysis and visualization capabilities would improve the effectiveness of staff oversight and provide new opportunities to analyze trends. EPA should develop national visualization tools that can be used to easily examine data for patterns, trends, and correlations.