Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff (2019)

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.

12 This chapter describes the basis for the pollutant categories included in this report and pro- vides guidance on how to determine the significance of stormwater POCs for roadways in rela- tion to TMDLs and impaired water bodies. In the preliminary phases of TMDL development, an understanding of the watershed’s POCs and their associated land use contributions can assist state DOTs in determining the appropriate engagement and implementation strategy. The objective of this section is to identify data sources and describe the procedures to assist state DOTs with identifying watershed conditions, quantifying the impact of background concentrations, and comparing land use loading rates. State DOT practitioners can use this information to determine the anticipated roadway contributions and facilitate their engage- ment in the development of and compliance with TMDLs that are discussed in subsequent sections. Finally, typical runoff concentrations for pollutants discharged from highways and other land uses are provided, along with a methodology to assist with evaluating relative load contributions by land use within a watershed. TMDL Pollutants of Concern for Roadways The number of TMDLs approved each year has increased over time, with a record of more than 15,000 TMDLs approved in 2013 (Figure 5) (EPA 2017B). Concurrently, the number of new impaired water bodies listed on state-reported 305(b)-designated-use lists has shown a decrease over time (Figure 5) (EPA 2017B). As an increasing number of impaired water bodies has been identified, regulatory action has shifted from identifying impairments to creating TMDLs. The purpose of this guidance document is to provide state DOT practitioners with available resources and tools to develop engagement strategies to address the increasing number of TMDL implementations. To focus on state DOT involvement in TMDL development and to provide specific guid- ance for compliance, it is critical to understand what pollutants are anticipated in TMDLs that identify roadways as a source. EPA’s Assessment, Total Maximum Daily Load Tracking and Implementation System (ATTAINS) Ask WATERS query tool was used to gather information relating to impaired water bodies and TMDLs (EPA 2017D). The percentage of urban runoff impaired water bodies and TMDLs that listed state transportation agencies is summarized in Table 1, based on pollutant groups and constituents: • Urban Runoff Impaired Water Bodies: In compliance with Section 305(b) of the Clean Water Act, EPA compiles annual state monitoring data regarding the intended use, the cause of impairment, and the source of impairment for assessed waters. Roadways are listed as a source subcategory of urban runoff or grouped with generic source categories, such as urban non-point source pollution or MS4s. The impaired water body list was used to characterize C H A P T E R 3 Significance of Stormwater Pollutants

Significance of Stormwater Pollutants 13 the types of pollutants that would be expected to contain roadway WLAs, since a similar approach is used for both current and future TMDL development. The 305(b) list provides a representative data set on the types of impairments associated with urban and roadway runoff. However, it only includes data for assessed water bodies that states have reported to EPA. Of the 93,877 water body segments in the database through the 2016 reporting cycle, 14,449 (15 percent) list urban runoff as a source of impairment (EPA 2017B). • TMDLs Listing State Transportation Agencies: To contrast impairment status with actual TMDL development, the ATTAINS application query tool was used to catalog TMDLs that listed state transportation agencies as a source. The search terms that were queried to cover all transportation agency names included “department of transportation,” “transportation cabi- net” (Kentucky), “highway department” (Massachusetts), “department of roads” (Nebraska), “highway and transportation department” (New Mexico), and “agency of transportation” (Vermont). The summary in Table 1 indicates those TMDLs for which the transportation agency names were identified in the queries performed. The presence of a TMDL is not nec- essarily indicative of an assigned WLA. Also, state DOTs may need to comply with a TMDL, even if the TMDL is not specifically 303(d)-listed. However, it is helpful to identify the TMDLs for which transportation agencies may be more directly involved. Of the 27,036 developed TMDLs cataloged in the ATTAINS document database through 2017, 5,340 (20 percent) include state transportation agencies listed as a source in the TMDL document (EPA 2017A). This analysis represents a subset of the total approved TMDLs, as only 27,036 of the 73,951 (37 percent) approved TMDLs are cataloged in the ATTAINS document database. This analysis assumes that this subset is a representative sample of existing TMDLs. Appendix A identifies all 50 state DOTs that are assigned WLAs for various POCs, based on the ATTAINS database. The variety of impairments caused by urban runoff is extensive and includes more than 250 classifications. This extensive variety of impairments presents a challenge to developing generalized TMDL guidance because numerous scenarios are possible. However, the primary classifications that are shown in Table 1 could be used to determine the general impairment cause groups that are most common nationally. Several cause groups had a relatively high percentage of water body impairments but low per- centages of implemented TMDLs, including categories such as Unknown/Other, Flow/Habitat Alterations, Impaired Biota, and Pesticides/Emergent Contaminants/Toxins. The Unknown/ Other, Flow/Habitat Alterations, and Impaired Biota categories were representative of water N um be r o f A pp ro ve d TM D Ls o r Im pa ire d W at er B od y Li st in gs Year 20,000 15,000 10,000 5,000 0 Approved TMDLs New Impaired Water Body Listings Figure 5. Comparison of approved TMDLs to new impaired water body listings over time.

14 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff Cause Group Water Bodies, Urban Runoff Impairment TMDLs, Transportation Agencies Listed Percentage of 27,036 TMDLsa Primary Classification (proportion of group total) b Percentage of 5,340 TMDLsa Primary Classification (proportion of group total) b Pathogens 21.5 Fecal Coliform (0.42) E. coli (0.34) Enterococcus (0.08) 29.0 Fecal Coliform (0.62) E. coli (0.28) Enterococcus (0.05) Unknown/Other 17.6 Unknown/Placeholder (0.96) Debris/Trash (<0.01) Taste and Odor (<0.01) 0.4 Debris/Trash (0.48) Urban Stormwater Pollutants (0.20) Unknown (0.12) Sediment 13.1 Sedimentation/Siltation (0.76) Turbidity (0.12) Total Suspended Sediment (0.09) 9.0 Sediment (0.56) Total Suspended Sediment (0.18) Sedimentation/Siltation (0.15) Nutrients 10.0 Nutrients/Eutrophication (0.43) Total Phosphorus (0.36) Total Nitrogen (0.10) 15.4 Total Phosphorus (0.63) Total Nitrogen (0.31) Nitrate/Nitrite (0.04) Flow/Habitat Alterations 8.6 Flow Alteration (0.70) Habitat Alteration (0.19) Vegetative Cover Alteration (0.06) 0.1 Habitat Alteration (1.00) Organic Enrichment/ Oxygen Depletion 7.9 Dissolved Oxygen (0.87) Biological Indicators (0.08) Biochemical Oxygen Demand (BOD) (0.05) 3.9 Biochemical Oxygen Demand (0.43) Dissolved Oxygen (0.31) Oxygen Demand (0.14) Impaired Biota 6.4 Biological Impairment (0.42) Algae/Weeds (0.41) Exotic/Invasive Species (0.04) 0.3 Biological Impairment (0.80) Chlorophyll A (0.12) Algae/Weeds (0.10) Pesticides/Emergent Contaminants/Toxins 6.2 Polychlorinated Biphenyls (0.38) Dioxin (0.10) Toxicity Assessment (0.05) 1.8 Polychlorinated Biphenyls (0.15) Atrazine (0.15) Chlordane (0.10) Metals 4.1 Arsenic (0.17) Mercury (0.17) Copper (0.13) 33.0 Iron (0.58) Aluminum (0.13) Manganese (0.09) Salinity/Dissolved Constituents 2.6 Total Dissolved Solids (0.39) Chlorides (0.35) Sulfates (0.15) 1.0 Chlorides (0.50) Total Dissolved Solids (0.25) Conductivity (0.13) pH/Acidity/Caustic Conditions 1.0 pH (1.00) 2.5 pH (0.99) Alkalinity (0.01) Temperature 0.6 Temperature/Thermal Modification (1.00) 2.5 Temperature/Thermal Modification (0.99) Riparian Shade (<0.01) Oil and Grease 0.5 Oil and Grease (0.93) Petroleum Hydrocarbons (0.07) 0.03 Oil and Grease (1.00) Total 100 99 aFor water body segments and TMDLs listing multiple impairments, each impairment was included; 6,990 pollutants are listed in 5,340 TMDLs listing state DOTs. bPrimary classifications are the top three pollutants by percentage listed within the cause group as reported in the ATTAINS database (EPA 2017A). Table 1. Cause groups and primary classifications for urban runoff impaired water bodies and TMDLs listing state transportation agencies.

Significance of Stormwater Pollutants 15 body measurements that were difficult to translate to land use WLAs. This limitation can result in the selection of alternative cause classifications that are more specific or delay the TMDL development until quantifiable impairment causes are identified. The Pesticides/Emergent Contaminants/Toxins category includes measurable constituents; however, this cause group is a minority within urban runoff or state DOT-source TMDLs. This could be due to limited monitoring data for these constituents, both with regard to total number of samples analyzed and the number of samples analyzed using laboratory methods sensitive enough to detect pol- lutants at the levels of concern. Because of the variety of contaminant types, missing data sets for the Pesticides/Emergent Contaminants/Toxins category and limited TMDL development for the Unknown/Other, Flow/Habitat Alterations, and Impaired Biota categories, these classifications were not included in this analysis. Priority pollutants (Table 2) were selected based on state DOT TMDL WLAs that had the majority of classifications (Table 1). While other pollutants are likely of interest in certain situ- ations, the pollutants listed in Table 2 are assumed to represent the primary constituents for TMDL development for state DOTs nationally. Since the focus of this evaluation is on state DOT preparation and compliance with TMDLs, it only considers constituents that can be directly monitored in runoff. In addition, the developed guidance is based on land use loading rates, performance monitoring data, and existing modeling tools. The most common POCs for many state DOTs with regard to TMDL development are Pathogens, Sediment, Nutrients, Metals, Organic Enrichment/Oxygen Depletion, and Salinity/Dissolved Constituents. These pollutant categories represent 90 percent of the TMDLs catalogued in the ATTAINS database with transportation agencies listed (5,340). The Pathogens, Sediment, and Nutrients cause groups account for the majority of both impaired water bodies and developed TMDLs (Table 1). The Metals cause group has a distinctly greater percentage of developed TMDLs compared to water body impairments. The Organic Enrichment category/Oxygen Depletion cause group is predominantly based on dissolved oxygen for impairments, but BOD is the dominant pol- lutant for developed TMDLs. The Salinity/Dissolved Constituents group is a minority per- centage for both impairments and developed TMDLs; however, it is included because of its correlation with roadway deicing agents in cold weather climates. The three lowest minority cause groups (pH, Temperature, and Oil and Grease) are not included as priority constituents nationally because of infrequent occurrence. Identifying Impaired Water Bodies State DOT personnel likely have access to local information pertaining to impaired water bodies and developed TMDLs, with the most common source being the state regulatory author- ity responsible for water quality assessments under Clean Water Act Sections 303(d) and 305(b). However, it may be useful to identify additional information in the planning phase to support Pathogens E. coli Fecal Coliform Metals Aluminum (Al) Arsenic (As) Cadmium (Cd) Copper (Cu) Iron (Fe) Organic Enrichment / Oxygen Depletion Biochemical Oxygen Demand (BOD) Dissolved Oxygen (DO) Nutrients Nitrogen, Total Nitrogen (TN), Nitrate/Nitrite (NO2,3 as N) Salinity / Dissolved Constituents Chloride (Cl-) Total Dissolved Solids (TDS) Nitrogen, Total Kjeldahl Nitrogen (TKN) Phosphorus, Total Phosphorus (TP), Dissolved Phosphorus (DP) Lead (Pb) Manganese (Mn) Mercury (Hg) Zinc (Zn) Sediment Total Suspended Sediment (TSS) Table 2. TMDL pollutant categories and pollutants targeted for analysis.

16 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff TMDL development and to research how other jurisdictions are structuring TMDLs for the same constituents or constituent categories. The following data sources may be useful to state DOT practitioners to identify impaired water bodies and developed TMDLs, as well as for other types of watershed information: 1. ATTAINS Database: EPA’s ATTAINS database includes a collection of reports and tools for determining water body impairment nationally and locally (EPA 2017A). The database con- tains information that was reported by state agencies to EPA during 2-year reporting cycles in compliance with the Clean Water Act. Two pertinent tools to TMDL development that are available include the following: a. ATTAINS Web Reports: Reports can be generated by state and watershed boundary. Information is provided on water body characteristics, cause of impairment, and number of TMDLs by pollutant type. b. Ask WATERS Query Tool: This query tool allows for searching 303(d) state lists and TMDL documents in tabular format based on location, cause of impairment, source of the impairment, or keywords. 2. WATERS GeoViewer: This tool can create custom surface water condition maps because it includes the National Map (hydrography and elevation layers) and the National Surface Water Network monitoring data. It can also spatially portray water body impairments, show designated uses, display water quality monitoring data, and site the location of discharges. 3. U.S. Geological Survey (USGS) StreamStats Tool: The StreamStats tool automates the delin- eation of watersheds using digital elevation models and a user-provided delineation point. It identifies the basin characteristics, such as land use, elevation, slope, total area, and impervi- ous surface area (USGS 2017B). Flow statistics are also available for mapped stream gauges. 4. USGS Streamer Tool: This tool can trace upstream and downstream connected water bod- ies based on the selection of a point in a water body (USGS 2017A). The available data from counties and states that traced their streams can be used to identify the highway areas that contribute to 303(d)-listed water bodies at their intersections. Land Use and Background Runoff Concentrations Roadway runoff pollutant concentrations are influenced by contributions from run-on, vehi- cle deposition, atmospheric fallout, and highway maintenance (Figure 6). An understanding of the magnitude and controllability of these sources is an essential step in determining equitable WLAs and planning management actions. This section presents data sets that state DOT prac- titioners can use to assess the impacts of land use, soil type, and atmospheric deposition on downstream water quality. Land Use The objective of the land use assessment is to compile data from monitoring databases to compare runoff concentrations from various land uses. This information is useful to state DOT practitioners interested in determining the relative pollutant contributions of roadways com- pared to adjacent land uses. The pertinent data were identified in four runoff databases and compiled into a master data set to generate an assessment of optimal runoff concentration dis- tributions from various land uses. The data sources included the following: • Highway Runoff Database (HRDB): The HRDB includes data from 10 studies that docu- ment 119 highway-runoff monitoring sites based on 4,210 storm events in the conterminous United States. The data set is the predominant repository of highway-runoff monitoring data (Granato 2010A).

Significance of Stormwater Pollutants 17 • National Stormwater Quality Database (NSQD): The NSQD is a compilation of monitoring data collected by NPDES permit holders. The data set contains 8,602 sampled events from 104 cities around the United States and identifies the predominant site land use (Pitt and Maestre 2015). • International Stormwater Best Management Practices Database (BMPDB): The BMPDB is a comprehensive international repository for stormwater BMP monitoring projects. The database includes influent and effluent BMP hydrology and water quality data, as well as metadata on watershed and site characteristics. It also contains data sets from nearly 650 BMP performance studies. In addition, the influent concentration data that had a predominant land use description was used in the characterization of land use runoff concentrations (Geo- syntec Consultants and Wright Water Engineers 2014). • Agricultural Best Management Practices Database (AgBMPDB): The AgBMPDB is a grow- ing repository of performance data for agricultural study sites with and without implemented management practices (Wright Water Engineers and Geosyntec Consultants 2017A). The Version 2 data set includes monitoring results from 345 field sites with predominantly corn and soybean crops. Data were aggregated into seven land use categories based on the categorical descriptions provided in each database. The following land uses were used: • Highways: Surface street, highway, and interstate monitoring sites. Data collected from the HRDB is organized by annual average daily traffic (AADT). • Commercial: Office, retail, and mixed-use urban sites. • Industrial: Manufacturing and processing sites that were classified from light to heavy industry. • Institutional: School, medical, government, or business campuses classified as institutional. Figure 6. Conceptual roadway pollutant load mass balance (Harned 1988).

18 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff • Residential: Housing areas of low-to-high density and mixed-use areas that are predomi- nantly residential. • Open Space: Undeveloped or preserved areas, including grasslands, forests, and parks. • Agricultural: Crop production land uses (predominantly corn and soybean) exclusively from the AgBMPDB. The fields that were considered included variations in nutrient management, tillage, and crop rotation practices. The data were aggregated based on the presence of edge- of-field (EOF) practices, including buffers, water quality basins, and controlled drainage man- agement. Agriculture data were only available for nutrient and sediment constituents and are only representative of crop production areas. Land use runoff concentration summary statistics were calculated for each targeted pollutant (Table 2). Furthermore, the data from the HRDB, NSQD, and BMPDB were aggregated and used to characterize nonagricultural land uses. The AgBMPDB was used exclusively to characterize the agricultural land use due to differences in reporting statistics and limited agricultural data in the HRDB, NSQD, and BMPDB. The AgBMPDB reports values that are composite statistics for a site because it does not currently have individual sample statistics. This results in a single value per site for each of the concentration statistics reported. For nonagricultural land uses, the number of sites, samples, and percentage of non-detect values are reported. The summary statistics that report the median and interquartile range (25th percentile to 75th percentile) are reported for each land use and constituent (Table 3 to Table 8). The data summaries allow state DOT practitioners to compare median highway runoff concentrations for common TMDL pol- lutants to other land uses, as well as input them into the land use load comparison protocol (see Chapter 5). Data availability can be highly variable, with certain land uses and pollutants having limited sample collection or a significant number of non-detect samples (Table 3 to Table 8). Land Use Category Quantity of Sites/Samples (percentage of samples non-detect) TSS TN TKN NO2-3 as N TP DP Highway (AADT 0–25K) 32 / 665(1%) No Data 28 / 492 (16%) 13 / 207 (2%) 28 / 433 (16%) 4 / 46 (7%) Highway (AADT 25–50K) 45 / 873(<1%) No Data 38 / 567 (2%) 22 / 290 (<1%) 43 / 559 (2%) 8 / 54 (2%) Highway (AADT 50–100K) 15 / 266(<1%) No Data 13 / 164 (1%) 12 / 123 (<1%) 18 / 266 (7%) 4 / 41 (22%) Highway (AADT 100K+) 27 / 598(<1%) No Data 21 / 335 (<1%) 11 / 120 (0%) 20 / 329 (6%) 3 / 28 (32%) Highway (all AADT combined)a 193 / 4,160(<1%) 12 / 389 (0%) 164 / 3,052 (4%) 123 / 2,265 (3%) 184 / 3,325 (6%) 71 / 1,361 (21%) Commercial 124 / 2,155(<1%) 43 / 769 (0%) 104 / 1,608 (3%) 110 / 1,787 (2%) 140 / 2,200 (3%) 33 / 704 (18%) Industrial 97 / 1,225 (<1%) 13 / 144 (3%) 88 / 973 (6%) 76 / 856 (5%) 104 / 1,203 (4%) 16 / 301 (24%) Institutional 21 / 287(0%) 1 / 3 (0%) 6 / 73 (0%) 21 / 275 (<1%) 21 / 191 (1%) 2 / 7 (57%) Residential 250 / 4,011(<1%) 10 / 156 (3%) 36 / 428 (2%) 36 / 469 (2%) 44 / 607 (2%) 9 / 228 (5%) Open Space 42 / 589 (<1%) 44 / 766 (19%) 224 / 3,084 (1%) 213 / 3,112 (5%) 271 / 4,146 (2%) 39 / 1,192 (11%) Agriculture (EOF)b 7 / x(NA) 4 / x (NA) 3 / x (NA) 12 / x (NA) 4 / x (NA) 1 / x (NA) Agriculture (No EOF) 16 / x(NA) 11 / x (NA) 3 / x (NA) 24 / x (NA) 14 / x (NA) 8 / x (NA) Agriculture (All) 23 / x (NA) 15 / x (NA) 6 / x (NA) 36 / x (NA) 18 / x (NA) 9 / x (NA) Note: K = thousand; x = individual samples that were not reported in the AgBMPDB. Composite value for the site from all monitored events are used; NA = not available. aHighway (all AADTcombined) includes sites for which AADT was not classified. bEOF = edge-of-field practice and includes buffers, water quality basins, and controlled drainage management. Table 3. Sediment and nutrient land use site and sample quantities.

Significance of Stormwater Pollutants 19 Land Use Category Median Concentrations of All Data from All Sites Combined (25th to 75th percentiles of all data from all sites combined) TSS (mg/L) TN (mg/L) TKN (mg/L) NO2-3 as N (mg/L) TP (mg/L) DP (mg/L) Highway (AADT 0–25K) 43.0(16.0–120) No Data 0.82 (0.34–1.7) 0.43 (0.25–0.72) 0.12 (0.04–0.26) 0.07 (0.05–0.12) Highway (AADT 25–50K) 96.0(51.0–179) No Data 1.8 (1.08–3.00) 0.84 (0.5–1.38) 0.23 (0.14–0.38) 0.16 (0.06–0.27) Highway (AADT 50–100K) 55.0 (24.5–144) No Data 1.78 (1.13–3.2) 0.80 (0.46–1.25) 0.15 (0.07–0.27) 0.10 (0.04–0.21) Highway (AADT 100K+) 93.0(45.3–187) No Data 1.68 (1.09–2.41) 0.95 (0.40–1.67) 0.19 (0.1–0.32) 0.07 (0.03–0.2) Highway (all AADT combined)a 67.5(29.0–137) 1.36 (0.84–2.18) 1.40 (0.77–2.3) 0.56 (0.30–1.05) 0.18 (0.08-0.32) 0.05 (0.02–0.12) Commercial 52.2 (24.0–122) 1.20 (0.80–2.10) 1.10 (0.68–1.88) 0.40 (0.21–0.71) 0.20 (0.11–0.37) 0.04 (0.01–0.11) Industrial 61.0(21.0–148) 2.08 (0.95–3.24) 1.40 (0.84–2.45) 0.54 (0.30–1.02) 0.22 (0.12–0.44) 0.08 (0.03–0.3) Institutional 36.6(16.5–65.1) 1.81 (1.47–2.00) 1.20 (0.90–1.65) 0.37 (0.23–0.73) 0.12 (0.07–0.22) 0.05 (0.04–0.07) Residential 60.0 (22.0–145) 1.69 (1.00–2.67) 1.40 (0.81–2.33) 0.58 (0.31–0.96) 0.25 (0.12–0.46) 0.09 (0.03–0.18) Open Space 80.0(28.7–181) 1.73 (1.19–2.67) 1.40 (0.87–2.46) 0.64 (0.31–1.14) 0.33 (0.18–0.59) 0.12 (0.03–0.26) Agriculture (EOF)b 69.3(51.1–155) 5.78 (5.25–6.10) 4.81 (4.28–4.95) 2.13 (1.52–3.04) 0.66 (0.63–0.71) 0.41 (0.22–0.76) Agriculture (No EOF) 363 (224–825) 5.40 (4.79–7.08) 3.55 (3.49–3.61) 2.81 (1.88–3.94) 0.86 (0.79–2.11) 0.33 (0.16–0.58) Agriculture (All) 248(151–419) 5.55 (4.79–6.36) 3.71 (3.58–4.55) 2.73 (1.65–3.72) 0.81 (0.74–1.40) 0.16 (0.16–0.16) Note: All non-detect values assumed to be half the reporting limit. aHighway (all AADTcombined) includes sites for which AADT was not classified. bEOF = edge-of-field practice and includes buffers, water quality basins, and controlled drainage management. Table 4. Sediment and nutrient land use runoff concentrations. Land Use Category Quantity of Sites/Samples (percentage of samples non-detect) Aluminum Arsenic Cadmium Copper Iron Lead Manganese Mercury Zinc Highway (AADT 0–25K) No Data 19 / 339 (56%) 27 / 446 (51%) 35 / 651 (10%) 13 / 241 (3%) 33 / 621 (21%) 3 / 32 (0%) No Data 35 / 639 (2%) Highway (AADT 25–50K) No Data 10 / 151 (17%) 17 / 226 (13%) 21 / 301 (10%) 12 / 193 (0%) 21 / 301 (4%) 6 / 70 (0%) No Data 21 / 313 (<1%) Highway (AADT 50–100K) No Data 9 / 152 (32%) 23 / 334 (13%) 27 / 561 (2%) 9 / 203 (0%) 27 / 550 (4%) 2 / 28 (0%) No Data 27 / 501 (<1%) Highway (AADT 100K+) No Data 23 / 325 (23%) 46 / 586 (6%) 48 / 779 (6%) 16 / 150 (0%) 49 / 717 (1%) 9 / 55 (0%) 4 / 15 (100%) 49 / 796 (0%) Highway (all AADT combined)a 1 / 16 (0%) 100 / 1,764 (28%) 162 / 2,613 (23%) 196 / 3,692 (5%) 61 / 1,028 (2%) 187 / 3,411 (9%) 20 / 185 (0%) 7 / 39 (100%) 200 / 3,722 (1%) Commercial No Data 25 / 262(57%) 61 / 851 (57%) 98 / 1,396 (8%) 4 / 109 (0%) 86 / 1,101 (18%) 1 / 20 (10%) 20 / 155 (95%) 110 / 1,643 (2%) Industrial 3 / 24 (4%) 27 / 179 (73%) 58 / 692 (49%) 91 / 1,109 (10%) 2 / 18 (0%) 80 / 875 (21%) No Data 23 / 186 (95%) 97 / 1,166 (2%) Institutional No Data 3 / 8(13%) 4 / 24 (17%) 6 / 39 (0%) No Data 4 / 26 (4%) No Data No Data 21 / 248 (0%) Residential No Data No Data 125 / 1,658(55%) 216 / 2,792 (10%) 14 / 347 (4%) 184 / 2,373 (19%) 5 / 225 (7%) 44 / 360 (80%) 230 / 3,213 (3%) Open Space No Data 15 / 174 (56%) 22 / 256 (53%) 30 / 388 (10%) No Data 30 / 365 (15%) No Data 9 / 97 (91%) 35 / 431 (0%) aHighway (all AADTcombined) includes sites for which AADT was not classified. Table 5. Metals land use site and sample quantities.

Land Use Category Median Concentrations of All Data from All Sites Combined (25th to 75th percentiles of all data from all sites combined) Aluminum µg/L Arsenic µg/L Cadmium µg/L Copper µg/L Iron µg/L Lead µg/L Manganese µg/L Mercury µg/L Zinc µg/L Highway (AADT 0–25K) No Data 0.66 (0.24–2.00) 0.13 (0.03–0.50) 9.40 (4.00–20.0) 870 (228–5,240) 7.00 (1.80–47.9) 22.1 (13.5–88.6) No Data 60.0 (28.0–130) Highway (AADT 25–50K) No Data 1.20 (0.59–2.45) 0.35 (0.16–0.80) 20.0 (10.0–37.0) 1,900 (841–3,620) 12.0 (4.80–48.0) 35.6 (19.2–67.3) No Data 91.0 (48.0–195) Highway (AADT 50–100K) No Data 1.10 (0.60–1.70) 0.70 (0.30–10.0) 31.1 (15.0–65.0) 5,000 (2,300–9,000) 71.0 (6.93–500) 41.1 (23.2–129) No Data 180 (90.0–310) Highway (AADT 100K+) No Data 1.80(0.95–3.00) 0.83 (0.50–1.80) 42.0 (20.5–71.0) 4,990 (1,730–10,900) 37.0 (14.0–200) 152 (59.0–258) Excessive Non-Detects 230 (120–380) Highway (all AADT combined)a 4,810 (3,180–9,610) 1.23 (0.54–2.40) 0.50 (0.20–1.00) 22.0 (9.71–48.0) 1,900 (590–5,820) 17.0 (4.85–75.0) 57.3 (21.4–152) Excessive Non-Detects 130 (54.0–274) Commercial No Data 1.29(0.65–2.97) 0.31 (0.11–0.90) 11.0 (6.00–20.6) 374 (171–642) 8.97 (3.09–20.0) 46.0 (29.0–68.0) Excessive Non-Detects 70.0 (34.0–140) Industrial 2,640(491–4,540) 0.31 (0.10–1.71) 0.37 (0.10–1.11) 13.0 (4.30–30.0) 5,190 (1,400–8,530) 10.0 (2.29–31.6) No Data Excessive Non-Detects 87.7 (30.0–190) Institutional No Data 5.28(3.72–9.75) 0.41 (0.19–0.52) 9.00 (4.40–16.7) No Data 5.00 (4.00–9.23) No Data No Data 51.0 (34.0–81.0) Residential No Data No Data 0.28(0.07-0.85) 12.0 (4.99–30.0) 1,040 (396–3,520) 9.62 (2.40–28.6) 56.0 (30.0–120) Excessive Non-Detects 80.0 (31.0–175) Open Space No Data 1.32(0.65–4.00) 0.38 (0.14–1.00) 12.2 (6.95–27.2) No Data 15.0 (5.00–42.0) No Data Excessive Non-Detects 99.0 (45.8–190) Note: Robust regression-on-order statistics method, as described by Helsel and Cohn (1988), was used for all non-detect values. If non-detects exceeded 80 percent, then the data set was insufficient and no statistics were generated. aHighway (all AADTcombined) includes sites for which AADT was not classified. Table 6. Metals land use runoff concentrations.

Land Use Category Quantity of Sites/Samples (percentage of samples non-detect) E. coli Fecal Coliform BOD DO CL- TDS Highway (AADT 0–25K) No Data 2 / 4(0%) 6 / 66 (8%) No Data 12 / 423 (4%) 1 / 4 (0%) Highway (AADT 25–50K) No Data 1 / 9(0%) 5 / 39 (3%) No Data 10 / 605 (0%) No Data Highway (AADT 50–100K) No Data 3 / 15 (0%) 8 / 112 (1%) No Data 10 / 568 (<1%) No Data Highway (AADT 100K+) No Data 6 / 56(0%) 12 / 118 (2%) No Data 13 / 583 (<1%) 1 / 4 (0%) Highway (all AADT combined)a 1 / 22(0%) 50 / 404 (5%) 50 / 602 (4%) 4 / 18 (0%) 51 / 2,249 (1%) 47 / 918 (4%) Commercial 10 / 166(4%) 39 / 526 (4%) 65 / 1,099 (7%) 3 / 43 (0%) 17 / 271 (11%) 55 / 887 (1%) Industrial 7 / 61(5%) 42 / 344 (8%) 67 / 742 (6%) 9 / 51 (0%) 11 / 165 (12%) 49 / 619 (7%) Institutional No Data 2 / 34(0%) 4 / 66 (0%) No Data 17 / 165 (4%) 3 / 49 (0%) Residential 15 / 150(4%) 82 / 851 (4%) 161 / 1,887 (5%) 16 / 104 (0%) 28 / 464 (15%) 115 / 1,566 (<1%) Open Space 2 / 14(0%) 14 / 181 (3%) 24 / 244 (7%) 3 / 12 (0%) 3 / 108 (7%) 24 / 336 (<1%) aHighway (all AADTcombined) includes sites for which AADT was not classified. Table 7. Pathogens, organic enrichment, and dissolved constituents land use site and sample quantities.

Land Use Category Median Concentrations of All Data from All Sites Combined (25th to 75th percentiles of all data from all sites combined) E. coli (MPN/100 mL) Fecal Coliform (MPN/100 mL) BOD (mg/L) DO (mg/L) Cl- (mg/L) TDS (mg/L) Highway (AADT 0–25K) No Data 25,500(12,300–376,000) 5.00 (3.06–7.00) No Data 16.8 (5.07–109) 48.0 (41.5–50.8) Highway (AADT 25–50K) No Data 901(841–1,600) 4.60 (3.00–7.00) No Data 27.4 (10.7–183) No Data Highway (AADT 50–100K) No Data 1,640 (771–6,580) 12.0 (8.00–23.5) No Data 24.9 (8.99–104) No Data Highway (AADT 100K+) No Data 1,600 (807–13,800) 5.62 (3.51–9.75) No Data 48 (19.4–271) 230 (54.8–446) Highway (all AADT combined)a 1,250 (388–3,630) 1,600 (400–11,000) 6.83 (4.00–13.9) 8.80 (5.90–9.68) 27.9 (10.0–147) 58.0 (32.0–98.0) Commercial 1,520 (135–8,260) 5.370 (977–21,400) 8.00 (4.90–14.0) 8.30 (7.12–9.57) 7.60 (3.15–21.6) 75.0 (49.5–122) Industrial 1,400 (368–4,400) 2,150 (500–11,100) 9.00 (5.00–15.2) 8.00 (6.90–8.93) 8.96 (2.10–53.0) 101 (53.0–204) Institutional No Data 4,060 (769–13,900) 7.00 (4.00–12.0) No Data 43.3 (7.44–73.9) 88.0 (58.0–132) Residential 2,090 (452–6,230) 3,300 (403–22,000) 8.00 (5.00–14.5) 7.85 (6.48–9.03) 7.12 (2.00–36.3) 78.0 (50.0–132) Open Space 1,580 (703–3,380) 5,600 (1,500–33,000) 10.2 (6.00–18.5) 8.65 (7.82–9.20) 4.05 (1.15–11.8) 77.9 (46.8–130) Note: MPN = most probable number. aRobust regression-on-order statistics method, as described by Helsel and Cohn (1988), was used for all non-detect values. Table 8. Pathogens, organic enrichment, and dissolved constituents land use runoff concentrations.

Significance of Stormwater Pollutants 23 The size of the data set should be considered when evaluating its applicability as a calculation input, including the following: • Sediment and Nutrients (Tables 3 and 4): The nutrient and sediment data sets are relatively robust for all constituents and land uses. The institutional land use and dissolved phosphorus analytes are limited for certain land use–pollutant pairs. However, the agricultural land use concentrations reported in the AgBMPDB are generally greater than all other land uses for sediment, nitrogen, and phosphorus analytes. The nutrient runoff concentrations are similar between roadway, commercial, industrial, and residential land uses. Also, the sediment and TP analytes show a correlation with AADT, with increasing concentrations at greater traffic volumes. • Metals (Tables 5 and 6): The metal constituents have high quantities of sample values for copper, lead, and zinc constituents. However, the other analytes are limited, with too few sample values for aluminum and manganese to conduct land use comparisons. The mercury samples are predominantly below method detection limits, with greater than 80 percent non- detect for all land-uses. The arsenic and cadmium samples were also frequently below the detection limits, with more than 20 percent of roadway and 40 percent of other land uses as non-detects. For copper, lead, and zinc, the median roadway concentration—all AADT—is frequently greater than all other land uses. The comparisons between the land uses for other constituents are more variable because of small data sets, and the non-detects limit the con- clusions that can be drawn. • Pathogens, Organic Enrichment, and Dissolved Constituents (Tables 7 and 8): The data set size for these analytes is variable and dependent on land use and constituent. Moreover, the data sets are especially limited for E. coli and dissolved oxygen. The median fecal coliform, BOD, and TDS concentrations for roadways were generally lower than all other land uses. The median chloride concentration for roadways was greater than all other land uses, with the exception of institutional. Background Soil Concentrations The background concentrations of constituents in soils can also influence runoff loading rates. The chemical and physical properties of the soil affect the rate and composition of runoff due to erodibility, cohesion, hydraulic conductivity, and soil chemistry (Geosyntec Consultants et al. 2015). Pollutants can also be transported in both the dissolved and particulate forms, with runoff characteristics primarily determined by the pollutants’ solubility properties. The percent- age of a pollutant in stormwater being transported in particulate form (attached to sediment) is shown in Table 9. The built-up inorganic nitrogen typically dissolves into a soluble form when it encounters rainfall (Miguntanna et al. 2013). However, phosphorus and certain metals are pre- dominantly particulate-bound and can be mobilized with sediment (Helmreich et al. 2010). This has important implications for both the source of these pollutants in runoff and their removal through stormwater BMP implementation. Constituent Particulate-Bound (%) Dissolved Constituents Source Total Nitrogen 24 DON, NH3-N, NO2,3-N (Taylor et al. 2005) Total Phosphorus 71 O-PO4 3- (Miguntanna et al. 2013) Zinc 73 Dissolved Zn (Helmreich et al. 2010) Copper 79 Dissolved Cu (Helmreich et al. 2010) Lead 100 na (Helmreich et al. 2010) Note: DON = dissolved organic nitrogen; na = not applicable. Table 9. Nutrient and metal particulate and dissolved constituent proportions in stormwater runoff.

24 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff Presumably, the predominant source of sediment from roadways and other land uses is the adjacent in situ soil or fill material. An understanding of the background pollutant soil con- centrations and the particulate-bound washoff characteristics is used to assess the propor- tion of pollutant runoff concentrations that are potentially attributed to soil concentrations. The objective of this analysis is to compare background soil concentrations to roadway runoff concentrations and to determine the relative contribution potentially attributable to soils. This information can help state DOTs evaluate whether background pollutant concentrations in soils are a potentially significant source of highway runoff loads. The USGS North American Soil Geochemical Landscapes Project (NASGLP) has spatially cataloged soil concentrations for phosphorus and metal constituents within the conterminous United States (Smith et al. 2014). Figure 7 displays the NASGLP data set for soil phosphorus concentrations by percentile for the top 5 cm of the soil profile based on an interpolation of 4,841 soil samples. The top 5 cm data set is used because the majority of sediment in storm water runoff would likely be generated from the soil surface, except for during large mass wasting events or bank erosion instances where deeper soil horizons may contribute. The background soil concentrations for target pollutants can vary widely based on geographic location (Smith et al. 2014). The map shown in Figure 7 can be downloaded as a Google Earth KML file for Figure 7. Phosphorus soil concentrations in the top 5 cm (Smith et al. 2014).

Significance of Stormwater Pollutants 25 phosphorus and metal analytes (https://mrdata.usgs.gov/soilgeochemistry/#/summary). These files can also be used to estimate the soil concentrations for a specific site. Using the surface soil (top 5 cm) statistics from the NASGLP, one can calculate the potential concentration that is associated with soil erosion or washoff. By monitoring or estimating TSS concentrations and knowing the soil concentration percentile based on site location, one can estimate an associated concentration for phosphorus and metal analytes. The concentrations presented in Table 10 are per 100 mg/L TSS. These concentrations can also be adjusted by mul- tiplying the site or receiving water TSS concentration and then dividing by 100. This process can be used by state DOT practitioners as a planning-level source identification tool when prepar- ing for TMDL development or estimating compliance. It can also be used to accomplish two objectives: 1. Background Pollutant Load: In the planning phase of TMDL development, it may be ben- eficial to identify the background pollutant loading rate for an undeveloped land use. The concentrations identified in Table 10 can be used to estimate the pollutant concentration dis- tributions using TSS as a surrogate and its associated data sets to calculate pollutant loading rates. An alternative method for determining background soil concentrations using the Sto- chastic Empirical Loading and Dilution Model (SELDM) for specific watersheds is described in the next section, Simulating Receiving Water Quality Using SELDM. Using SELDM to calculate pollutant loads based on land use and impervious cover is described in the Land Use Load Comparison Protocol section. This information could also be used to advocate for equitable WLAs based on an understanding of the estimated predeveloped loading caused by soil washoff. 2. Prioritization of Erosion-Prone Areas: Pollutant loading from bank erosion or unstable landscapes could be substantial sources for TSS, TP, and identified metal constituents. Fraley et al. (2009) estimated that bank erosion accounted for 43 percent of TSS loading in an urban Pennsylvania watershed. If erosional areas are identified within the watershed, the concentrations presented in Table 10 could be used to attribute a load reduction associated with stabilization. This could be beneficial to both the WLA and compliance strategy iden- tification phases. For analytes with sufficient roadway runoff concentration data sets (more than 100 samples), the concentrations associated with soil washoff were compared to the measured roadway runoff concentrations (Table 11). This analysis shows the proportion of roadway runoff concentrations that are attributed to soil washoff. Phosphorus, aluminum, arsenic, and iron have relatively high proportions of runoff concentration attributed to background soils (Table 11). This indi- cates that loading from erosion-prone areas and background sources is especially relevant for these analytes, particularly in areas with high background-soil percentiles. The iron percentiles greater than 100 percent of roadway runoff concentrations indicate higher soil concentrations Soil Percentilea Constituent Runoff Concentrations (µg/L) Corresponding with 100 mg/L TSS TP Al As Cd Cu Fe Pb Mn Hg Zn 5% 14.0 920 0.14 0.005 0.38 380 0.76 7.90 0.0005 1.20 25% 36.0 3,200 0.31 0.01 0.88 1,280 1.40 29.0 0.001 3.60 50% 58.0 4,670 0.52 0.02 1.40 1,950 1.80 49.0 0.002 5.80 75% 84.0 6,000 0.76 0.03 2.10 2,660 2.40 79.0 0.004 8.00 95% 139 7,980 1.30 0.07 4.30 4,560 4.50 150 0.01 13.0 aSoil percentile refers to the nationwide surface soil (top 5 cm) statistics from the NASGLP. Table 10. Constituent runoff concentration (lg/L) attributed to surface soil washoff based on soil concentration percentiles.

26 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff than roadway runoff concentrations. This indicates that the source of iron in highway runoff may often be primarily background soils. Similarly, aluminum in highway runoff could also potentially be attributed to high concentrations of background soils. Cadmium, copper, lead, and zinc each have relatively low concentrations of roadway runoff attributed to soil washoff and likely have anthropogenic sources (vehicular and road construction materials). Simulating Receiving Water Quality Using SELDM SELDM uses Monte Carlo methods to estimate streamflow concentrations and loads for a site of interest and upstream basin (Granato 2013). This tool was collaboratively developed by the USGS and FHWA. This tool provides a valuable resource to state DOT practitioners interested in assessing highway loading contributions to a catchment and the effect of manage- ment actions on receiving water concentrations. When a POC of interest lacks receiving water monitoring data, the SELDM model can be used to estimate concentrations based on a sedi- ment transport curve. The following steps provide an overview on using the SELDM Lake–Basin analysis package to define background receiving water concentrations. This approach is further elaborated in the case study analysis conducted in Granato and Jones (2016). 1. Define Site Location and Upstream Basin: Site characteristic are defined to represent the physical parameters of a site and upstream basin. The USGS StreamStats Tool (USGS 2017B) can be used to determine estimates for watershed area and slope. 2. Simulate Hydrology: Precipitation statistics, pre-stream flows, impervious cover, and exist- ing BMPs are defined and used in the Monte Carlo assessment to determine the distribution of daily stream flow rates and annual volumes. 3. Simulate Upstream Water Quality: A sediment transport curve is used to define the TSS concentration based on stream flow rate for the upstream basin. This can be user defined if data are available or set to regional values defined based on ecoregion in SELDM (Granato 2013). A water quality transport curve can be defined based on background soil or bed sedi- ment concentrations for a POC. Values from the NASGLP (Table 11)—or alternative local data sets—can be used to relate suspended sediment concentrations to POC concentrations. The variation in the water quality transport curve is defined in SELDM using a standard- normal variate (Krandom). A Monte Carlo analysis is conducted to determine the population distribution of daily stream flow concentrations and annual loads. 4. Simulate Site Runoff Water Quality: The event mean concentrations (EMCs) for the site are defined using statistical distributions. EMC distributions can be defined based on local Soil Percentilea Proportion of median roadway runoff concentration (µg/L) attributed to surface soil washoffb TP Al As Cd Cu Fe Pb Zn (180)c (4,810)c,d (1.23)c (0.50)c (22.0)c (1,900)c (1,700)c (130) c 5% 0.05 0.13 0.08 0.01 0.01 0.14 0.03 0.01 25% 0.14 0.45 0.17 0.01 0.03 0.46 0.05 0.02 50% 0.22 0.66 0.29 0.03 0.04 0.70 0.07 0.03 75% 0.32 0.84 0.42 0.04 0.06 0.95 0.09 0.04 95% 0.52 1.12 0.72 0.09 0.13 1.63 0.18 0.06 aSoil percentile refers to the nationwide surface soil (top 5 cm) statistics from the NASGLP. bSoil runoff concentrations were calculated at the median roadway TSS concentration of 70 mg/L (Table 4). cMedian roadway runoff concentrations, shown in parentheses, are as reported in Tables 4 and 6 for all AADT highways. dThere are limited highway runoff data points for aluminum (Table 5). Table 11. Proportion of constituent runoff concentrations attributed to surface soil washoff—assuming all sediment in runoff is background soil— at median roadway runoff concentrations.

Significance of Stormwater Pollutants 27 monitoring data or land use summary statistics, as described in Tables 14 to 20 of this chapter. Storm event runoff loads are computed from the EMCs and hydrologic parameters to esti- mate instream concentrations upstream and downstream of a site. An assessment of background receiving water concentrations can be used by state DOT prac- titioners for several aspects of TMDL assessment and implementation planning. By defining background receiving water body conditions, state DOT practitioners may be better positioned to advocate for equitable and scientifically based WLAs. For pollutants with substantial back- ground concentrations, defining anticipated natural sources and inputs may be critical for assessing anticipated relative contributions of roadways to total loadings. Using SELDM to compare anticipated loading rates for land uses present in the watershed is further described in the Land Use Load Comparison Protocol section. Once a TMDL is implemented, an assessment of modeled receiving water body concentrations can be used to determine the effectiveness of management action. BMP performance modeling is further discussed in Chapter 6. Atmospheric Deposition Atmospheric data are limited for the majority of state DOT TMDL POCs (Table 12). How- ever, deposition data for mercury (Hg), sulfate (SO4 2-), nitrate (NO3 -), ammonium (NH4 +), and chloride (Cl-) have been extensively cataloged by the National Atmospheric Deposition Program (NADP) (2017). The NADP provides access to National Trends Network (NTN) and Mercury Deposition Network (MDN) data. Moreover, the grid data representing 2015 annual average concentrations of the pollutants mentioned were obtained from the NTN and MDN. The latest year for which published data were available at the time of this writing is 2015. A spatial assess- ment was conducted using the NTN and MDN grid data to compute grids that show the percent contribution of atmospheric deposition to the average national highway runoff concentrations of the five pollutants listed above. The average national highway runoff concentrations for the five atmospheric deposition pollutants are shown in Table 12. This analysis was limited to certain areas of the country because of limited data availability within the NADP database for certain regions. The following procedure was used to conduct the spatial assessment: 1. Identified the 25th and 75th percentile runoff concentrations (as available) for highway land uses (all traffic patterns) from the HRDB for the five atmospheric deposition pollutants (Hg, SO4 2-, NO3, NH4 +, and Cl-). 2. Downloaded the annual precipitation-weighted wet deposition concentration spatial grid layers (mg/L) from the NADP for the identified pollutants. 3. Computed within each grid cell the ratio of the atmospheric deposition constituent concen- tration at the cell to the national average 25th percentile runoff concentration. The computa- tion was repeated for the 75th percentile runoff concentration. Pollutant HRDB Runoff Concentrations (all traffic types) 25th Percentile (mg/L) Median (mg/L) 75th Percentile (mg/L) Ammonia (NH4 +) 0.20 0.31 0.45 Nitrate (NO3) 0.30 .56 1.05 Chloride (Cl-) 10.0 27.9 147 Mercury (Hg) BDL BDL BDL Sulfate (SO4 2-) NA NA NA Note: BDL = below detection limit; NA = not available. Table 12. Highway runoff concentrations for atmospheric deposition constituents (25th and 75th percentile concentrations from HRDB).

28 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff 4. Repeated Step 3 for each of the atmospheric deposition pollutants to determine the propor- tion of highway runoff contribution to each pollutant attributed to deposition. The results of the atmospheric deposition analysis are shown in Figures 8 to 18. For each analyte, maps depicting the atmospheric deposition concentrations from the NTN or MDN and calculated ratios of atmospheric deposition to the 25th and 75th percentile runoff concentra- tions are provided. • Ammonia (Figures 8–10): With the exception of a few areas in the Pacific Northwest and the East Coast, atmospheric deposition appears to contribute greater than 80 percent of the national average 25th percentile ammonium ion concentrations in highway runoff (Figure 9). Ammonium as NH4+ (mg/L) 0 0.20 0.40 0.60 0.80 ≥ 1.00 Figure 8. Estimated ammonium ion concentrations attributed to atmospheric deposition from 2016 NTN data. Proportion of Ammonium Potentially Attributed to Atmospheric Deposition >0.8 0.6–0.8 0.4–0.6 0.2–0.4 0–0.2 Figure 9. Estimated proportion of 25th percentile highway ammonium ion concentration attributed to atmospheric deposition from 2016 NTN data.

Significance of Stormwater Pollutants 29 State DOTs across the nation whose sampling data indicate ammonium ion concentrations close to the national 25th percentile concentration should evaluate atmospheric deposition as a possible significant contributor. Similarly, Figure 10 indicates that, for state DOTs located in certain parts of the country, sample concentrations close to the 75th percentile national average concentration could still attribute a significant portion of their loads to atmospheric deposition. • Nitrate (Figures 11–13): The nitrate deposition contributes to roadway runoff concentra- tions, as shown by its similar trend to ammonia. Nitrate depositions contribute greater than 80 percent of the 25th percentile nitrate highway runoff concentration for a majority of the Proportion of Ammonium Potentially Attributed to Atmospheric Deposition >0.8 0.6–0.8 0.4–0.6 0.2–0.4 0–0.2 Figure 10. Estimated proportion of 75th percentile highway ammonium ion concentration attributed to atmospheric deposition from 2016 NTN data. 0.8 0.4 1.6 1.2 0 Nitrate as NO3- (mg/L) ≥ 2.0 Figure 11. Estimated nitrate ion concentrations attributed to atmospheric deposition from 2016 NTN data.

30 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff country (Figure 12). A substantial portion of the 75th percentile highway runoff concentra- tion is also observed for large portions of the country (Figure 13). • Chloride (Figures 14–16): Atmospheric deposition does not appear to be a significant con- tributor for state DOTs with chloride concentrations that are close to the national average 25th percentile chloride ion concentrations except for a few areas on the West Coast and the East Coast and around the Great Salt Lake, which are affected by deposition from adjacent salt water sources (Figure 15). Similarly, atmospheric deposition is not a significant contributor for nearly all parts of the country with concentrations close to the 75th percentile national average concentration (Figure 16). • Mercury (Figure 17): The HRDB does not currently have sufficient data to characterize mer- cury. More than 80 percent of the samples collected are below the detection limit (ranging Proportion of Nitrate Potentially Attributed to Atmospheric Deposition >0.8 0.6–0.8 0.4–0.6 0.2–0.4 0–0.2 Figure 12. Estimated proportion of 25th percentile highway nitrate ion concentration attributed to atmospheric deposition from 2016 NTN data. Proportion of Nitrate Potentially Attributed to Atmospheric Deposition >0.8 0.6–0.8 0.4–0.6 0.2–0.4 0–0.2 Figure 13. Estimated proportion of 75th percentile highway nitrate ion concentration attributed to atmospheric deposition from 2016 NTN data.

Significance of Stormwater Pollutants 31 from 0.1 µg/L to 1 µg/L). To adequately conduct the runoff and deposition comparison analy- sis, more robust sampling standards and sample results are needed. However, atmospheric deposition data do indicate that it could be a significant contributor for mercury to large portions of the country (Figure 17). • Sulfate (Figure 18): The HRDB does not currently have enough data for computing national average sulfate ion concentrations. Therefore, the atmospheric deposition analysis to esti- mate the contribution of atmospheric deposition to sulfate concentrations in highway runoff was not performed. Figure 18 shows the average sulfate ion concentrations using 2016 data from the NTN. State DOTs with measured sulfate concentrations that are significantly CI- (mg/L) 0 0.20 0.40 0.60 0.80 ≥ 1.00 Figure 14. Estimated chloride ion concentrations attributed to atmospheric deposition from 2016 NDN data. Proportion of Chloride Potentially Attributed to Atmospheric Deposition >0.30 0.17–0.30 0.10–0.17 0.05–0.10 0.0–0.05 Figure 15. Estimated proportion of 25th percentile highway chloride ion concentration attributed to atmospheric deposition from 2016 NTN data.

32 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff Proportion of Chloride Potentially Attributed to Atmospheric Deposition >0.30 0.17–0.30 0.10–0.17 0.05–0.10 0.0–0.05 Figure 16. Estimated proportion of 75th percentile highway chloride ion concentration attributed to atmospheric deposition from 2016 NTN data. Figure 17. Estimated total mercury concentrations from 2015 MDN data.

Significance of Stormwater Pollutants 33 lower than the concentrations shown in the figure can conclude that atmospheric deposition is not a significant contributor. However, those with measured concentrations close to the values shown can justify further investigation to quantify the contribution of atmospheric deposition. The deposition of nitrogen analytes appears to be a substantial contributor to roadway runoff concentrations. In areas where measured roadway runoff concentrations are near the 25th per- centile national average, atmospheric deposition appears to be the primary source of ammonia and nitrate (Figure 9 and Figure 12). The maps shown in this analysis only include wet depo- sition; however, the wet deposition concentrations are consistent with literature findings on nitrogen deposition sources. Dry deposition—the settling of high-density dust particles—can also contribute to substantial pollutant loads. In a study measuring atmospheric deposition at a highway bridge deck in the North Carolina Piedmont, Wu et al. (1998) found that between 70 percent and 90 percent of TN in highway runoff can be attributed to deposition. Further investigation into the proportion of load attributed to atmospheric deposition may be warranted by state DOTs in regions with nitrogen WLAs and low expected or measured roadway runoff concentrations. For other analytes, deposition appears to generate a small portion of the overall pollutant load; however, it may be significant in certain situations. The California Department of Transporta- tion (Caltrans) (2003B) evaluated the contribution of wet deposition to highway runoff in the Lake Tahoe area and found that the contribution was relatively small for most constituents. For chloride, deposition was only substantial for certain coastal areas with low roadway runoff con- centrations (Figure 15). Phosphorus and metal constituents may occasionally be contributed from deposition sources. Phosphorus was not evaluated in the atmospheric deposition analysis because of the lack of data. According to Colman et al. (2001), phosphorus is not reported by NADP because concentrations in wet deposition are typically too low to be measured accurately. However, dissolved phosphorus in rainfall was attributed to 72 percent of the runoff concen- tration in the Caltrans (2003B) study. Dry deposition of phosphorus and metals can further be a substantial contributor to highway runoff that enters surface waters (Dudley et al. 1997). Wu et al. (1998) attributed between 30 percent and 50 percent of highway copper and lead loads Sulfate as SO42- (mg/L) 0 0.5 1.0 1.5 2.0 ≥ 2.5 Figure 18. Estimated sulfate ion concentrations from 2016 NDN data.

34 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff to combined wet and dry deposition. In urban settings, much of this deposition may be from local sources, such as emissions and airborne dust from vehicles and industry. Depending on the location, phosphorus, metals, chloride, and sulfide constituents may be substantial contributors to roadway pollutant loads; however, these constituents do not appear to be a primary pollutant source nationally. Roadway Use Inputs In addition to background soil concentrations and deposition, anthropogenic inputs can have a substantial effect on roadway runoff concentrations. Deicing chemicals, traction sand, vehicle use, litter, and adjacent land uses can all contribute pollutants to highway runoff. These sources are discussed in detail based on POCs in Chapter 4. Land Use Load Comparison Protocol State DOTs engaged in the beginning phases of TMDL development may be interested in comparing unit area loading rates for highways to adjacent land uses to advocate for equitable WLAs. The objective of this section is to explain the approach for using SELDM to compare unit area loading rates for different land uses (Granato 2013). The procedure for completing the SELDM analysis, the reference input statistics, and the example results are provided for state DOT practitioners to replicate this analysis for a specific watershed. The SELDM model runs a Monte Carlo analysis to determine the contribution from a defined site of loads on downstream water quality using a mass simulation approach (Granato and Jones 2015). SELDM was designed to model highway runoff, and—through user-defined inputs— additional land uses can be easily analyzed. For this analysis, only land use loading rates were analyzed; however, SELDM also has the functionality to assess the effects of implementing BMPs and the impacts on receiving water bodies (Granato and Jones 2015, Granato and Jones 2017). Users should review the online SELDM reference directory (https://doi.org/10.5066/f7bg2m33) for additional documentation on SELDM and its applicability to highways and TMDL analyses. The following procedures outline the steps to complete the SELDM analysis and to compare land use unit area loading rates (i.e., annual yields). As an example, unit area loading rates were compared for TSS, TP, and total zinc in the Pacific Northwest (EPA Ecoregion 15). For each step, the input values evaluated in this example are defined. • POC: Based on TMDL development or water body impairment status (see the first section of Chapter 2), determine the target pollutant(s) for the analysis. – Pacific Northwest Example: TSS, TP, and total zinc were selected as representative pollutants. • Land Use Characteristics: Determine the land use types and percentage of impervious areas for each land use type within the watershed. This can be determined for local condi- tions by overlaying local land use geographic information systems mapping data with imper- vious areas. Alternatively, representative impervious areas percentages for the considered land uses have been compiled in the Land Use Imperviousness section that follows. These percentages can be used as an initial estimate. The range of impervious cover by land use can be large (Table 13), and it should be quantified at the watershed scale whenever possible. – Pacific Northwest Example: The estimated median impervious cover by land use pre- sented in Table 13 was used for the following land use types: Commercial (86 percent), Industrial (73 percent), Residential Low Density (15 percent), Residential Medium Density (32 percent), Residential High Density (49 percent), Roadways (60 percent), and Roadways (100 percent). The additional Roadway 100 percent impervious land use

Significance of Stormwater Pollutants 35 was added to account for differences in roadway definition. The 100 percent impervious category is representative of roads with curbs, gutters, and a piping system, whereas the 60 percent impervious category is representative of an entire highway corridor (includ- ing medians, shoulders, and drainage swales). • Water Quality Statistics: For each land use type, define the runoff concentration central tendency and distribution statistics for the POCs. SELDM requires mean, standard deviation, and skew coefficient inputs. These values should be entered as log10 or natural log values. For the pollutants quantified in the previous Land Use section, natural log input statistics were calculated and organized by land use type. They were then used as SELDM inputs as discussed in the Land Use Runoff Concentration Statistics section that follows. Log-transformed statis- tics account for the general positive skew of water quality data. Moreover, local monitoring data that quantifies land use runoff concentrations may be used as an alternative to these values. The data sets presented in the Land Use Runoff Concentration Statistics section are occasionally limited for certain land use types and constituents. Users should consult the sample number, site count, and non-detect statistics in gauging the robustness of the data set for specific applications. – Pacific Northwest Example: TSS, TP, and zinc land use runoff statistics were defined using the values presented in Tables 14 to 20. • SELDM Site Definition: For each of the identified land use types, a different SELDM model run is completed. Within the Highway Site: Identify Site Characteristics tab, the location and hydraulic inputs are defined. The latitude and longitude input parameters are used by the model to associate nearby rain gauges and stream statistics. To calculate the unit area loading rates, the drainage area is specified as 1 acre and the impervious fraction is defined based on the results of Step 2. The drainage area length, slope, and basin development factor inputs can be given representative inputs and do not affect volumetric-based calculations. – Pacific Northwest Example: The coordinates central to the Pacific Northwest (45.7°, –120.7°) were defined at a representative location for the entire region. • Precipitation Statistics: Based on the specified project location, SELDM can be used to identify associated rain gauges. The precipitation statistics can be aggregated by EPA rain zone, ecoregion, selected nearby rain gauges, or user-defined values. Local statistics can be used for site scale investigations. On the Synoptic Storm-Event Precipitation Statistics tab, the user selects the desired precipitation statistics data set. – Pacific Northwest Example: A rain zone average was selected for EPA Rain Zone 15, averaging all included rain gauges to develop a representative data set for the entire Pacific Northwest. • Analyzed Constituents: Within the Water Quality Menu tab, user-defined constituents are entered. Under the Highway-Random heading, new constituents can be defined with the statistics developed in Step 3. – Pacific Northwest Example: For each land use type and pollutant pair (e.g., TSS, TP, and zinc), a constituent was defined in SELDM. The highway land use type used the all AADT statistics defined in Table 14. • Run SELDM Model: The SELDM model is run after all inputs are defined. After the model has run, the annual loading rates for each year are provided in the output files. This informa- tion can be aggregated to determine the annual yield (lb/acre/year) exceedance probabilities for each land use type. The model is replicated for each land use type and updated to redefine the corresponding land use concentrations. When replicating the model, the Copy Current option can be used on the Highway Site: Identify Site Characteristics tab to redefine site characteristics and only adjust needed parameters. – Pacific Northwest Example: For each land use type and constituent, the exceedance probability plots were created from the SELDM output as shown in Figure 19.

36 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff The results of the example SELDM analysis that quantified the unit area pollutant loads for the Pacific Northwest are shown in Figure 19. The exceedance probability represents the likeli- hood in any given year that the load from that land use will be greater than the specified value. By comparing the magnitude of annual loading rates for land uses present in the drainage area, one can assess the anticipated contribution to receiving water load. The higher impervious frac- tion land use types generally had higher unit area loading rates overall with the 100 percent impervious highway exhibiting the highest zinc and TSS loads. However, the effects of runoff concentration on TP concentrations were evident with industrial and commercial land use load- ing rates having similar values to the 100 percent impervious roadway, even with lower calcu- lated runoff volumes. In addition, the accurate characterization of land use imperviousness and runoff concentrations at the watershed scale can improve the estimates of land use loading rates. Nevertheless, any replicated analyses performed (Figure 19) and then scaling up to the land use areas within the subject watershed can assist state DOTs with preparing for TMDL development and advocating for equitable WLAs. Land Use Imperviousness The representative impervious areas that were based on land use type are presented in Table 13 per a review of 30 literature studies conducted by Granato (2010B). A wide range of impervi- ous values are evident within land use types. Granato (2010B) noted that discrepancies may be due to how land use types are defined based on including or excluding open space as part of other land use types. Also, the values presented in Table 13 can be used to develop preliminary estimates for land use loading rates but should be verified with watershed-specific calculations when feasible. The estimated medians reported in Table 14 for commercial, residential, and industrial areas are similar to the assumed impervious fractions that were used in the develop- ment of runoff curve numbers (Cronshey et al. 1986). The residential land use density classifi- cations correspond approximately with the following lot sizes: • Low Density: Greater than 1 acre per lot • Medium Density: 0.25–1 acre per lot • High Density: Less than 0.25 acre per lot Commercial Industrial Res. Low Density Res. Med. Density Res. High Density Open Space Highway 100% Imp. Highway 60% Imp. 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 0255075100 A nn ua l L oa d (lb /a c/ yr ) Exceedance Probability (%) TP 0 0.5 1 1.5 2 2.5 3 0255075100 Zinc 0 200 400 600 800 1,000 1,200 1,400 1,600 0255075100 TSS Figure 19. Annual unit area loading rate exceedance plots by land use for the Pacific Northwest (EPA Rain Zone 15) (Res. = residential; Imp. = impervious; Med. = medium) (Source: Geosyntec Consultants).

Significance of Stormwater Pollutants 37 Land Use Runoff Concentration Statistics The land use runoff concentration statistics presented in Tables 14 to 20 are organized by land use type and provide the necessary input statistics for defining a water quality constituent in SELDM. SELDM can be used to generate unit loading rates for the POCs identified in Table 2 for various land uses. SELDM allows users to determine the loading rates based on several key user inputs, such as location, runoff concentration, and imperviousness. Highway runoff concentration sta- tistics were characterized for each pollutant in Table 2 from the HRDB, NSQD, and BMPDB. Land Use Percent Imperviousness Min Estimated Mediana Max Highways/Transportation 32 60 91 Commercial 35 86 100 Industrial 30 73 96 Institutional 30 65 91 Residential, All 6 32 90 Residential, Low Density 6 15 50 Residential, Medium Density 23 32 40 Residential, High Density 30 49 80 Open Space, All 0 5 21 Open Space, Forest 0 5 15 Open Space, Parks, and Grassland 1 9 21 Agriculture, All 1 5 62 Agriculture, Crops, and Pasture 1 5 14 Agriculture, Livestock, and Poultry 42 52 62 a50th percentile values were assumed for studies identified within a reported range. Table 13. Impervious area by land use based on literature values reported in Granato (2010B). Constituent Unit Sites Samples Non-Detect (%) Ln Mean Ln SD Ln Skew Geomean TSS mg/L 193 4,160 <1 4.117 1.288 -0.384 61.4 TN mg/L 12 389 0 0.347 0.742 0.372 1.41 TKN mg/L 164 3,052 3 0.270 0.920 -0.293 1.31 NO2,3 as N mg/L 123 2,265 3 -0.633 1.017 -0.389 0.531 TP mg/L 184 3,325 6 -1.833 1.095 -0.069 0.160 DP mg/L 71 1,361 21 -3.057 1.376 -0.138 0.047 Aluminum µg/L 1 16 0 8.482 0.684 -0.368 4825 Arsenic µg/L 100 1,764 28 0.162 1.198 0.329 1.18 Cadmium µg/L 162 2,613 23 -0.711 1.577 0.509 0.491 Copper µg/L 196 3,692 5 3.035 1.164 -0.145 20.8 Iron µg/L 61 1,028 2 7.453 1.627 -0.338 1724 Lead µg/L 187 3,411 9 2.934 2.082 0.125 18.8 Manganese µg/L 20 185 0 4.035 1.197 0.022 56.5 Mercury µg/L 7 39 100 -2.508 0.475 0.122 0.081 Zinc µg/L 200 3,722 1 4.774 1.184 -0.280 118 E. coli MPN/100 mL 1 22 0 6.989 1.741 -0.480 1085 Fecal Coliform MPN/100 mL 50 404 5 7.517 2.352 -0.215 1838 DO mg/L 4 18 0 2.069 0.257 -0.469 7.91 BOD mg/L 50 602 4 2.057 0.983 0.599 7.82 Cl- mg/L 51 2,249 1 3.787 2.221 0.588 44.1 TDS mg/L 47 918 4 3.965 0.955 -0.526 52.7 Note: Ln = natural log; SD = standard deviation; Geomean = geometric mean = exp(Ln Mean). Table 14. Highway land use (all AADT) runoff concentration data set parameters.

38 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff Constituent Unit Sites Samples Non-Detect (%) Ln Mean Ln SD Ln Skew Geomean AADT 0–25K TSS mg/L 32 665 1 3.802 1.530 0.016 44.8 TKN mg/L 28 492 16 -0.262 1.177 -0.117 0.770 NO2,3 as N mg/L 13 207 2 -0.816 0.828 0.397 0.442 TP mg/L 28 433 16 -2.251 1.451 0.226 0.105 DP mg/L 4 46 7 -2.552 0.532 0.034 0.078 Arsenic µg/L 19 339 56 -0.386 1.525 0.048 0.680 Cadmium µg/L 27 446 51 -2.048 2.088 0.235 0.129 Copper µg/L 35 651 10 2.239 1.156 0.051 9.38 Iron µg/L 13 241 3 6.931 1.994 -0.009 1024 Lead µg/L 33 621 21 2.088 2.208 -0.032 8.07 Manganese µg/L 3 32 0 3.423 1.334 -0.199 30.7 Zinc µg/L 35 639 2 4.104 1.104 0.029 60.6 Fecal Coliform MPN/100 mL 2 4 0 10.384 2.478 0.365 32323 BOD mg/L 6 66 8 1.649 0.612 0.502 5.20 Cl- mg/L 12 423 4 3.115 2.256 0.249 22.5 TDS mg/L 1 4 0 3.761 0.253 -1.021 43.0 AADT 25–50K TSS mg/L 15 266 <1 4.069 1.343 0.208 58.5 TKN mg/L 13 164 1 0.631 0.804 -0.022 1.88 NO2,3 as N mg/L 12 123 <1 -0.250 0.793 -0.457 0.779 TP mg/L 18 266 7 -1.935 1.226 0.543 0.144 DP mg/L 4 41 22 -2.437 0.987 -0.245 0.087 Arsenic µg/L 10 151 17 0.240 1.043 0.387 1.27 Cadmium µg/L 17 226 13 -1.026 1.208 0.188 0.358 Copper µg/L 21 301 10 2.931 0.922 -0.148 18.74 Iron µg/L 12 193 0 7.491 1.236 0.289 1792 Lead µg/L 21 301 4 2.647 1.609 0.166 14.12 Manganese µg/L 6 70 0 3.676 0.918 0.358 39.5 Zinc µg/L 21 313 <1 4.569 1.290 0.208 96.4 Fecal Coliform MPN/100 mL 1 9 0 6.526 1.026 -1.316 683 BOD mg/L 5 39 3 1.413 0.785 -0.275 4.11 Cl- mg/L 10 605 0 3.978 2.181 0.895 53.4 AADT 50–100K TSS mg/L 27 598 <1 4.449 1.226 -0.600 85.5 TKN mg/L 21 335 <1 0.506 0.738 -0.116 1.66 NO2,3 as N mg/L 11 120 0 -0.195 0.986 -0.288 0.822 TP mg/L 20 329 6 -1.768 0.960 -0.440 0.171 DP mg/L 3 28 32 -2.618 1.042 -0.136 0.073 Arsenic µg/L 9 152 32 0.074 1.022 0.985 1.08 Cadmium µg/L 23 334 13 0.224 2.260 0.308 1.25 Copper µg/L 27 561 2 3.453 1.024 -0.065 31.6 Iron µg/L 9 203 0 8.412 1.010 -0.334 4502 Lead µg/L 27 550 3 4.079 2.397 -0.129 59.1 Manganese µg/L 2 28 0 4.063 1.041 0.462 58.1 Zinc µg/L 27 501 <1 5.104 0.922 -0.197 165 Fecal Coliform MPN/100 mL 3 15 0 7.710 1.541 0.070 2230 BOD mg/L 8 112 <1 2.580 0.839 -0.010 13.2 Cl- mg/L 10 568 <1 3.613 2.104 0.728 37.1 Table 15. Highway land use by AADT runoff concentration data set parameters.

Significance of Stormwater Pollutants 39 Constituent Unit Sites Samples Non-Detect (%) Ln Mean Ln SD Ln Skew Geomean AADT 100k+ TSS mg/L 45 873 <1 4.416 1.285 -1.379 82.8 TKN mg/L 38 567 2 0.544 0.873 -0.128 1.72 NO2,3 as N mg/L 22 290 <1 -0.173 0.789 0.101 0.841 TP mg/L 43 559 2 -1.510 0.940 -0.136 0.221 DP mg/L 8 54 2 -1.923 1.333 0.461 0.146 Arsenic µg/L 23 325 23 0.542 0.961 0.404 1.72 Cadmium µg/L 46 586 6 0.000 1.261 0.740 1.00 Copper µg/L 48 779 6 3.620 1.100 -0.252 37.3 Iron µg/L 16 150 0 8.292 1.304 -0.405 3993 Lead µg/L 49 717 1 3.870 1.855 0.140 47.9 Manganese µg/L 9 55 0 4.834 1.068 -0.211 125.7 Mercury µg/L 4 15 100 -2.190 0.486 -1.055 0.112 Zinc µg/L 49 796 0 5.277 1.179 -0.959 196 Fecal Coliform MPN/100 mL 6 56 0 7.961 2.191 0.423 2868 BOD mg/L 12 118 2 1.866 1.025 1.368 6.46 Cl- mg/L 13 583 <1 4.381 2.188 0.635 79.9 TDS mg/L 1 4 0 4.931 1.312 -0.170 139 Ln = natural log; SD = standard deviation; Geomean = geometric mean = exp(Ln Mean). Table 15. (Continued). Constituent Unit Sites Samples Non-Detect (%) Ln Mean Ln SD Ln Skew Geomean TSS mg/L 124 2,155 <1 3.984 1.263 -0.374 53.7 TN mg/L 43 769 0 0.241 0.791 -0.606 1.27 TKN mg/L 104 1,608 3 0.117 0.810 0.087 1.12 NO2,3 as N mg/L 110 1,787 2 -1.027 1.095 -0.937 0.358 TP mg/L 140 2,200 3 -1.586 1.048 0.247 0.205 DP mg/L 33 704 17 -3.272 1.676 0.057 0.038 Arsenic µg/L 25 262 57 0.280 1.058 0.459 1.32 Cadmium µg/L 61 851 57 -1.119 1.633 0.297 0.327 Copper µg/L 98 1,396 8 2.406 1.061 0.086 11.1 Iron µg/L 4 109 0 5.891 0.931 -0.169 362 Lead µg/L 86 1,101 18 2.080 1.402 -0.182 8.00 Manganese µg/L 1 20 10 3.815 0.589 0.024 45.4 Mercury µg/L 20 155 95 -1.991 0.972 1.517 0.137 Zinc µg/L 110 1,643 2 4.207 1.264 0.182 67.2 E. Coli MPN/100 mL 10 166 4 7.383 3.591 0.336 1608 Fecal Coliform MPN/100 mL 39 526 4 8.378 2.292 -0.363 4348 DO mg/L 3 43 0 2.049 0.423 -2.974 7.76 BOD mg/L 65 1,099 7 2.096 0.862 0.177 8.13 Cl- mg/L 17 271 11 2.075 1.512 -0.382 7.97 TDS mg/L 55 887 <1 4.388 0.937 0.831 80.5 Ln = natural log; SD = standard deviation; Geomean = geometric mean = exp(Ln Mean). Table 16. Commercial land use runoff concentration data set parameters.

40 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff Constituent Unit Sites Samples Non-Detect (%) Ln Mean Ln SD Ln Skew Geomean TSS mg/L 97 1,225 <1 3.987 1.481 -0.373 53.9 TN mg/L 13 144 3 0.550 0.883 -0.409 1.73 TKN mg/L 88 973 6 0.343 0.881 0.258 1.41 NO2,3 as N mg/L 76 856 5 -0.643 1.022 -0.420 0.526 TP mg/L 104 1,203 4 -1.490 1.067 0.048 0.225 DP mg/L 16 301 24 -2.418 1.553 -0.021 0.089 Aluminum µg/L 3 24 4 7.275 1.782 -0.352 1444 Arsenic µg/L 27 179 73 -1.017 1.833 0.060 0.362 Cadmium µg/L 58 692 49 -1.006 1.952 0.199 0.366 Copper µg/L 91 1,109 10 2.377 1.399 -0.365 10.8 Iron µg/L 2 18 0 8.318 1.303 0.093 4097 Lead µg/L 80 875 21 2.170 1.799 -0.088 8.8 Mercury µg/L 23 186 95 -2.109 1.175 1.449 0.121 Zinc µg/L 97 1,166 2 4.219 1.460 -0.693 68.0 E. Coli MPN/100 mL 7 61 5 7.043 1.811 -0.581 1145 Fecal Coliform MPN/100 mL 42 344 8 7.669 2.431 -0.127 2141 DO mg/L 9 51 0 2.006 0.327 -1.493 7.44 BOD mg/L 67 742 6 2.145 0.930 -0.124 8.55 Cl- mg/L 11 165 12 2.543 2.392 0.767 12.7 TDS mg/L 49 619 7 4.826 1.539 1.238 125 Ln = natural log; SD = standard deviation; Geomean = geometric mean = exp(Ln Mean). Table 17. Industrial land use runoff concentration data set parameters. Constituent Unit Sites Samples Non-Detect (%) Ln Mean Ln SD Ln Skew Geomean TSS mg/L 21 287 0 3.511 0.910 -0.153 33.5 TN mg/L 1 3 0 0.498 0.277 -0.475 1.65 TKN mg/L 6 73 0 0.215 0.512 0.317 1.24 NO2,3 as N mg/L 21 275 <1 -1.052 1.124 -0.821 0.349 TP mg/L 21 191 <1 -2.185 0.759 -0.352 0.112 DP mg/L 2 7 57 -2.968 0.295 0.054 0.051 Arsenic µg/L 3 8 13 2.023 1.266 1.315 7.56 Cadmium µg/L 4 24 17 -0.943 1.017 0.888 0.390 Copper µg/L 6 39 0 2.271 0.980 0.561 9.69 Lead µg/L 4 26 4 1.920 0.927 1.048 6.82 Zinc µg/L 21 248 0 3.878 0.705 -0.661 48.3 Fecal Coliform MPN/100 mL 2 34 0 7.834 2.163 -0.618 2525 BOD mg/L 4 66 0 2.009 0.726 0.238 7.45 Cl- mg/L 17 165 4 3.492 1.653 0.019 32.8 TDS mg/L 3 49 0 4.422 0.606 -0.295 83.3 Ln = natural log; SD = standard deviation; Geomean = geometric mean = exp(Ln Mean). Table 18. Institutional land use runoff concentration data set parameters.

Significance of Stormwater Pollutants 41 Constituent Unit Sites Samples Non-Detect (%) Ln Mean Ln SD Ln Skew Geomean TSS mg/L 250 4,011 <1 4.018 1.440 -0.289 55.6 TN mg/L 44 766 3 0.461 0.810 -0.325 1.59 TKN mg/L 224 3,084 2 0.302 0.878 -0.240 1.35 NO2,3 as N mg/L 213 3,112 2 -0.670 1.050 -0.731 0.512 TP mg/L 271 4,146 2 -1.466 1.044 -0.234 0.231 DP mg/L 39 1,192 5 -2.631 1.308 -0.440 0.072 Arsenic µg/L 64 673 56 0.034 1.210 0.090 1.03 Cadmium µg/L 125 1,658 55 -1.356 1.801 0.066 0.258 Copper µg/L 216 2,792 10 2.448 1.349 -0.190 11.6 Iron µg/L 14 347 3 6.918 1.583 -0.596 1010 Lead µg/L 184 2,373 19 2.128 1.768 -0.128 8.40 Manganese µg/L 5 225 7 4.103 0.944 -0.063 60.5 Mercury µg/L 44 360 80 -5.844 5.334 0.441 0.003 Zinc µg/L 230 3,213 3 4.240 1.388 -0.457 69.4 BOD mg/L 161 1,887 5 2.121 0.928 0.055 8.34 E. Coli MPN/100 mL 15 150 4 7.346 2.110 -0.454 1550 Fecal Coliform MPN/100 mL 82 851 4 7.951 2.934 -0.299 2839 DO mg/L 16 104 0 2.028 0.245 -0.428 7.60 Cl- mg/L 28 464 15 2.138 1.893 0.203 8.48 TDS mg/L 115 1,566 <1 4.406 0.910 0.249 81.9 Ln = natural log; SD = standard deviation; Geomean = geometric mean = exp(Ln Mean). Table 19. Residential land use runoff concentration data set parameters. Constituent Unit Sites Samples Non- Detect (%) Ln Mean Ln SD Ln Skew Geomean TSS mg/L 42 589 <1 4.258 1.508 -0.597 70.7 TN mg/L 10 156 19 0.577 0.742 -0.227 1.78 TKN mg/L 36 428 1 0.367 0.850 -0.001 1.44 NO2,3 as N mg/L 36 469 4 -0.565 1.150 -0.269 0.568 TP mg/L 9 228 11 -2.405 1.485 -0.303 0.090 DP mg/L 44 607 2 -1.164 1.056 -0.445 0.312 Arsenic µg/L 15 174 56 0.447 1.224 0.104 1.56 Cadmium µg/L 22 256 53 -1.001 1.297 0.176 0.367 Copper µg/L 30 388 10 2.585 1.123 -0.120 13.3 Iron µg/L 5 27 0 5.843 1.556 -0.704 345 Lead µg/L 30 365 15 2.591 1.481 -0.278 13.34 Mercury µg/L 9 97 91 -2.016 0.908 1.792 0.133 Zinc µg/L 35 431 0 4.496 1.108 -0.478 89.7 E. Coli MPN/100 mL 2 14 0 6.942 2.117 -0.578 1035 Fecal Coliform MPN/100 mL 14 181 3 8.424 2.663 -0.677 4556 DO mg/L 3 12 0 2.116 0.161 -0.631 8 BOD mg/L 24 244 7 2.349 0.871 0.141 10.5 Cl- mg/L 3 108 7 1.593 1.969 1.017 4.92 TDS mg/L 24 336 <1 4.371 1.179 -0.386 79.1 Ln = natural log; SD = standard deviation; Geomean = geometric mean = exp(LnMean). Table 20. Open space land use runoff concentration data set parameters.

42 Approaches for Determining and Complying with TMDL Requirements Related to Roadway Stormwater Runoff These data were further analyzed to estimate land use summary statistics for each constituent for user input into the model (Tables 14–20). These inputs, along with location to determine the associated ecoregion and rain zone, can then be used to estimate the loads for the selected high- way runoff constituents. As an example, unit area loads were estimated using SELDM for the states of California, Washington, Florida, and Massachusetts. Table 52 in Appendix B identifies the typical average annual rainfall determined by the model based on the location of each unit area. Table 53 in Appendix B presents the SELDM modeling results summarized as average annual unit area loads for selected water bodies, including the Los Angeles River in California, Lake Campbell in Washington, the Lower St. Johns River in Florida, and the Charles River in Massachusetts. These types of summary results can be used by state DOTs with highways within the tributary watersheds to roughly estimate preliminary baseline loads. Various compliance strategies for state DOTs to reduce loads from highways are presented in Chapter 4. Additional guidance on implementing stormwater BMPs with space-constrained rights-of-way can be found in NCHRP Report 728: Guidelines for Evaluating and Selecting Modifications to Existing Roadway Drainage Infrastructure to Improve Water Quality in Ultra- Urban Areas (Geosyntec Consultants et al. 2012).