Pollutant Load Reductions for Total Maximum Daily Loads for Highways (2013)

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.

7 chapter three RESULTS AND SYNTHESIS The literature review provided key information on state TMDL programs including the types of BMPs being imple- mented, their effectiveness, design standards, and associated costs. The literature review findings supplemented the infor- mation obtained during the interviews and provided a greater level of detail to present a more comprehensive picture of TMDL and BMP implementation strategies. However, a clear difficulty with the literature is that it describes a vari- ety of different study approaches, BMP terminologies, and reporting methods and it is difficult to compare data directly from separate studies. In some cases, results are conflict- ing and/or may only apply within certain geographic areas or climates (e.g., BMPs in arid climates may not work in temperate climates). An attempt is made to synthesize the findings to the extent possible, although there is a need for further research in standardizing BMP naming conventions (see chapter five for more details). The interviews with the 12 state DOTs were informa- tive and helped establish the current state of knowledge as it relates to TMDL programs around the country. In many cases, several DOT representatives were available from a variety of divisions (e.g., environmental, design, mainte- nance, and engineering) to provide information on different aspects of their stormwater programs. Following each inter- view, a brief summary was prepared to capture the primary discussion points. Only ten DOTs were required to be inter- viewed (in accordance with EA 2012) and the top ten most relevant DOT interviews were summarized and reported. These summaries were sent back to the DOTs for review and all ten responded with comments that were then incor- porated into the text. The interview summaries are provided in Appendix B. Because the findings of the literature review and inter- views complement each other (and in some cases overlap), the results from the two approaches are combined. These combined results are presented throughout this chapter and are ultimately synthesized into common patterns and themes, a BMP matrix/toolbox and idealized timeline for the overall TMDL and NPDES permit development process, and a series of conclusions and areas for further research. In cases where the information provided in the interviews was insufficient or not detailed enough, supplemental results are presented from the literature review. However, in cases where the interviews were more informative than the literature review (e.g., on the topics of DOT institutional structures or DOT participation in TMDL development), the results of the interviews are the primary source. TOTAL MAXIMUM DAILY LOADS BY STATE The number of TMDLs in which the interviewed DOT was a named stakeholder varied widely depending on the state. There is some conflicting information in the literature on the exact numbers of TMDLs in which the DOTs are a named stakeholder versus the numbers provided in the DOT inter- views. For simplicity, the information from the interviews was assumed to be the most up to date. Table 1 summarizes the number of TMDLs in which the DOT is a named stake- holder, the associated pollutants of concern, some basic infor- mation on how TMDLs are implemented in the state, and other related issues (e.g., construction permits, 401 certifica- tions). As shown, some states are not named in any TMDLs (e.g., Kansas) and some have a large number of TMDLs and a vast array of TMDL pollutants of concern (e.g., California). This range of involvement in TMDLs reflects the intentional selection of DOTs with differing maturity levels in their TMDL programs. Where the DOT is a named stakeholder, TMDLs are implemented in various ways such as through NPDES MS4 permits [or variants such as State Pollution Discharge Eliminations System (SPDES) permits in New York or Transportation Separate Storm Sewer System (TS4) permits in North Carolina] and/or through Construction Gen- eral Permits. DOTs may be assigned a specific WLA in cases where they are identified as a significant contributor to the water body impairment and/or the WLA may be shared with other stakeholders. More details are provided in Table 1. It is outside the scope of this synthesis to address all the TMDLs listed in the table. However, we highlight certain TMDLs and present detailed implementation plans for select states (see Total Maximum Daily Load Implementation Plans). INSTITUTIONAL PRACTICES FOR TOTAL MAXIMUM DAILY LOAD IMPLEMENTATION In this section, some DOT institutional practices for TMDL implementation are described. Institutional practices are defined here as any nonstructural DOT practices, policies, and organizational strategies that are used to help the DOT manage its stormwater program to meet TMDL target reduc- tion goals.

8 State DOT No. of TMDLs Where the DOT Is Named as a Stakeholder TMDL Pollutants of Concern Comments Kansas 0 N/A KDOT does not foresee TMDLs being implemented in the near future; it focuses on temporary construction stormwater BMPs only (no retrofits). Ohio 0, but approximately 40 new ones are expected N/A ODOT is indirectly affected by TMDLs through watershed-specific Construction General Permits (CGP) that apply to projects disturbing more than 1 acre of land. Georgia 0, but some are expected soon N/A GDOT must monitor outfalls to 303(d) streams in its MS4 permit areas to determine if the roadway is a significant contributor to the impairment. GDOT expects the monitoring will identify it as a significant contributor, potentially naming it as a stakeholder in new TMDLs. Colorado 2, but many more TMDLs are expected sediment TMDL WLAs are divided into different sources of sediment, but all are assigned to CDOT since their assets are typically the only man-made structures in remote mountainous TMDL watersheds. TMDL goals are driven by comparison with reference watersheds in pristine wilderness areas. New Hampshire 4 chloride/salt TMDL WLAs are allocated among NHDOT (10%), municipalities (35%), and private property owners (55%) based on salt application data provided to the state by each discharger. The initial chloride TMDL study (funded by NHDOT) was the result of a 401 Certification of a Wetland Permit for the I-93 expansion project. New York 5 nitrogen, phosphorus, pathogens TMDLs are implemented through the State Pollutant Discharge Elimination System (SPDES) General Permit for MS4s, which requires a percent load reduction (unique for each TMDL). Virginia 8 in current MS4 permit, with more than 20 expected in the next permit cycle (2013–2018), including the Chesapeake Bay TMDL sediment, nitrogen, phosphorus, PCBs, bacteria, flow TMDLs are implemented through the MS4 permit and VDOT is typically listed under an aggregated WLA with other stakeholders. VDOT only addresses TMDLs in watersheds where they have been designated as having responsibility for the discharge from an MS4 or a construction site. Delaware 21 nutrients, bacteria, PCBs, TSS TMDLs are implemented through the state regulations for sediment and stormwater control, through MS4 permits, and in the case of the Chesapeake Bay TMDL, as a blanket requirement across the state. DelDOT is assumed to meet all requirements if they follow the state regulations. DelDOT also develops pollution control strategies to reduce nutrients in certain watersheds to the level required by TMDLs. Washington State 26, with possibly 17 more in the next MS4 permit fecal coliform, temperature, TSS/turbidity, dissolved oxygen, pH, mercury, arsenic, pesticides, PCBs TMDLs are listed in WSDOT’s MS4 permit; the permit may be modified every 18 months to add any new TMDLs that name WSDOT as a stakeholder. WSDOT’s TMDL requirements and specific action items to address the TMDLs are spelled out in the permit. Minnesota approximately 40–50 approved and pending dissolved oxygen, biota, chloride, fecal coliform, nutrients, turbidity, etc. TMDLs usually have a WLA for construction stormwater (implemented through a CGP) and a WLA for MS4s (implemented through an MS4 permit). Construction stormwater BMPs are required by the CGP, TMDLs, and by requirements of watershed districts and management organizations. If MnDOT complies with the CGP, it is assumed to have met its construction stormwater WLA. TABLE 1 NUMBER OF TMDLs WHERE THE STATE DOT IS NAMED AS A STAKEHOLDER, THE ASSOCIATED POLLUTANTS OF CONCERN, INFORMATION ON HOW TMDLs ARE IMPLEMENTED, AND OTHER RELATED ISSUES (continued on next page)

9 One example of an institutional practice is a collabora- tive approach with the state regulatory agency to TMDL development. As recommended in McGowen et al. (2009), active participation is key because many source and treat- ment reduction approaches will benefit from economies of scale. They also recommend that DOTs participate early in the process to help ensure that WLAs are valid and equi- table. The DOT interview questions probed this issue to evaluate if DOTs are engaging in this process and what, if any, benefits are being realized. The interview results indicated that the level of collabo- ration with state regulatory agencies on TMDL develop- ment varied widely. Some DOT participation was limited to simply commenting on the draft TMDL during the public comment period; others took a more proactive approach by providing data and expertise to the regulatory agency and in some cases working with them to help define the DOT’s contribution to the WLA or to develop specific DOT action items for TMDL implementation. Overall, the message from the DOTs was clear: four of the DOTs that participated in the development of the TMDLs indicated that the most significant success in their TMDL program was the relationship they developed with the state regulatory agency through early participation and collaboration. The benefits were measurable in terms of the greater level of understanding on the part of the regulatory agency translating into more appropriate and achievable compliance goals. It also helped to educate the other stake- holders on the actual impact of the DOT ROW in order to establish more realistic expectations for the role of the DOT in the compliance strategy. Brief state-by-state summaries of DOT participation in the TMDL development process are provided here (further details are available in the interview summaries in Appendix B): • Delaware does not participate in the development of TMDLs owing to a lack of scientific expertise. How- ever, it does comment on draft TMDLs during the pub- lic comment period and participates in the development of pollution control strategies and watershed imple- mentation plans. • New York does not participate in TMDL development except during the public comment period. • Colorado does participate in TMDL development by providing data on sand application rates to address a sediment TMDL. • California does not participate in TMDL development; however, it does provide a wealth of data to the regional water quality control boards (the regulatory agencies) that are responsible for developing the WLA and enforcing the TMDL. The California DOT (Caltrans) also comments on draft TMDLs during the public comment period. • Washington State does participate in TMDL develop- ment to the best of its ability, but is limited by a lack of manpower (currently only one full-time employee is ded- icated to the development process). The Washington State DOT (WSDOT) reviews and provides comments during the public comment stage. It does not currently provide data to the state agency, but may in the future as monitoring efforts are increasing as part of the its new MS4 permit. WSDOT works with the state regulatory agency to ensure that realistic action items for TMDL requirements are written into its MS4 permit. • Minnesota participates in TMDL development by pro- viding data and commenting during the public comment State DOT No. of TMDLs Where the DOT Is Named as a Stakeholder TMDL Pollutants of Concern Comments California 62, with approximately 200 additional in the near future trash, bacteria, metals, selenium, nutrients, dissolved oxygen, chlorophyll A, toxics, pesticides, PCBs, sediment toxicity, PAHs, etc. TMDLs are developed and enforced by numerous regional water quality control boards (regulatory agencies). Caltrans is actively implementing about 40 high-priority TMDLs in cases where they are discharging a pollutant of concern from the roadway with the potential to impact water quality. North Carolina estimated more than 100, although DOT not named in all fecal coliform, nutrients, turbidity, biological integrity, etc. TMDLs are implemented through a TS4 permit that reflects the unique nature of the transportation corridor. The permit describes a step-by- step process for meeting the TMDL requirements in cases where NCDOT is assigned a specific WLA. Additional “TMDL-like” requirements occur in select coastal estuaries and drinking water supply reservoirs that require nutrient management strategies, so there are BMPs being implemented in both TMDL watersheds and (similarly regulated) non-TMDL watersheds. N/A = not available; PCBs = polychlorinated biphenol; PAH = polycyclic aromatic hydrocarbons. TABLE 1 (continued)

10 period. The Minnesota DOT (MnDOT) requests an indi- vidual WLA for smaller TMDLs as governed (in part) by a Memorandum of Understanding with the state regula- tory agency, which outlines the procedure for providing the agency with data (acres in ROW) and reviewing/ commenting on TMDLs. • Virginia participates in TMDL development through its environmental division and MS4 consultant; how- ever, the bulk of its analysis comes after the TMDL is approved through characterization studies. These stud- ies verify the ROW acreage and estimate baseline pol- lutant loads and potential load-reduction strategies using a modeling approach. • New Hampshire participates in TMDL development by providing data on salt application rates to the state reg- ulatory agency (for chloride TMDLs). The DOT also provided funding for a TMDL study along I-93 result- ing from a wetland mitigation permit. • North Carolina is probably the most proactive interviewed DOT in terms of participation in the TMDL development process. The North Carolina DOT (NCDOT) typically provides data, expertise, and sometimes funding to the state regulatory agency to ensure that the process is based on accurate and scientific assessments. However, the state uses its own modeling tools to develop the WLA. NCDOT’s proactive collaborative approach has provided many benefits including the ability to help define the DOT’s contribution to the WLA and form a reason- able TMDL strategy. Further, this approach has helped inform the development of a TS4 permit, the first of its kind in the country. The TS4 permit reflects the unique characteristics of the linear highway (as opposed to a municipal network) and an understanding of the trans- portation corridor within the urban setting. Another institutional TMDL implementation strategy is the development of watershed partnerships outside of the DOT’s ROW. The watershed partnerships are based on the concept that there are multiple land owners and sources of pollutants within a watershed. TMDL implementation strate- gies, therefore, must address pollutant loads from all sources to improve the chances of successfully meeting the TMDL. One major impetus for developing watershed partnerships is that pollutant loadings from highway surfaces may be rel- atively insignificant compared with those from the broader TMDL watershed, especially in cases where the DOT’s assets comprise only a small fraction of the watershed. In these cases, even if the impervious highway surface converts a greater fraction of the rainfall to runoff (e.g., 75%–80%) and often carries higher concentrations of specific pollutants compared with other land uses, the pollutant loads gener- ated by the highway are still minor compared with loads from other land uses within the watershed (Driscoll et al. 1990). These authors even suggest that where typical high- way pollutant concentrations are significantly lower, as with nutrients, highway runoff could essentially be excluded from the analyses in cases where water quality issues are related to nutrient loads. In support of this, a study by Hon et al. (2003) concluded that highways are an insignificant source of mass loadings of fecal coliform, total phosphorus, total nitrogen, and pathogens because the runoff volume from the highways is so small compared with the runoff from other watershed land uses. Similarly, a report by the East–West Gateway Coordinating Council (2000) suggests that land uses surrounding the highway facility have a far greater impact on the characteristics of the stormwater runoff from highway surfaces. Based on this evidence, developing watershed partner- ships would appear to be a more cost-effective TMDL strat- egy, particularly if the ROW is too constrained for BMP implementation (e.g., in ultra-urban corridors) or in cases where BMP implementation within the ROW is insufficient to achieve the required pollutant load reductions. A more distributed watershed approach to stormwater management would allow managers to acquire a more complete under- standing of the overall water quality conditions in a receiv- ing stream and the stressors that affect those conditions; in addition, a watershed approach can save managers time and money (Oregon State University et al. 2006), especially in cases where design, construction, and maintenance costs can be shared among partners. One example of an off-site partnership is stream restora- tion. This practice may be an important opportunity for DOTs to achieve sediment load-reduction goals. In Maryland, the Maryland State Highway Administration has worked with the Maryland Department of the Environment (the state regu- latory agency) to increase the efficiency credits of such activi- ties to at least 310 lb of sediment per 100 linear feet of stream restored. Furthermore, this number does not consider flood- plain connectivity as part of the stream restoration efforts, which may provide additional benefits in terms of nutrient and TSS reductions. Based on the interviews, it was found that a majority of the DOTs believed that their current efforts do not repre- sent the most cost-effective strategies, and most indicated that they believed watershed partnerships with other stakeholders are important for achieving watershed scale requirements through the development and implementation of cost-effective TMDL strategies. However, only 6 of the 12 DOTs had a standard policy for initiating collaboration with other parties (e.g., local governments, adjacent property owners, and water- shed groups) to address stormwater requirements. In some cases, these partnerships already exist as standard operational agreements between the state and local DOTs for the opera- tion and maintenance of the joined highway systems, such as the entrances to limited access Interstates and primary col- lectors, tunnel facilities, and bridges. For example, an Inter- Jurisdictional Agreement (IA) between New Castle County, Delaware, and the Delaware DOT (DelDOT) outlines the roles

11 and responsibilities of each party in inspecting, repairing, and maintaining MS4s, bridges, stormwater management basins and ponds, etc., within their respective jurisdictions. In particular, the IA states that the county and DelDOT shall consider TMDLs and various pollution control strategies as provided in a consent decree. For more details on the IA, see DelDOT (2001). Not all DOTs, however, are interested in entering into watershed partnerships. For example, NCDOT interview responses indicated caution when considering a partnership that may potentially leave them beholden to another entity for achieving on-going compliance with a WLA assigned to the DOT system. However, they also indicated that the cost- effectiveness of practices available to other watershed part- ners such as exclusion fencing of streams to keep cattle from eroding stream banks and introducing pathogens is signifi- cantly greater than traditional DOT practices. On the whole, though, there were more examples of successful partner- ships; for example, the New Hampshire DOT’s (NHDOT’s) technology transfer with private and local salt applicators; and MnDOT’s, Colorado DOT’s (CDOT’s), and DelDOT’s Cooperative Agreements with other stakeholders. LITERATURE REVIEW ON HIGHWAY BEST MANAGEMENT PRACTICES PERFORMANCE STUDIES This section provides key information from the literature review on highway BMP performance studies. The literature on BMP performance studies is considerable, but here the focus is only on those studies that met the criteria matrix. In general, BMP performance is difficult to measure. There are several potential metrics to use, some more scientifically valid than others, but there is no standardized way of reporting BMP performance in the literature. Further, the selection of a performance metric may have a pronounced bearing on how a BMP’s performance is perceived (Lenhart and Hunt 2011). Percent removal has been a common metric for BMP per- formance. Although there are drawbacks to using this metric (Wright Water Engineers and Geosyntec Consultants 2007), it is in part relied on because it is ubiquitous. However, we also present results from the literature that utilize other metrics, such as effluent quality (a more scientifically valid approach according to Wright Water Engineers and Geosyntec Consul- tants 2007) and load reduction to present a more comprehen- sive picture of BMP effectiveness. Texas DOT Study (Hon et al. 2003)—Process Framework for Identifying and Prioritizing Water Quality Improvement for Meeting TMDLs in Texas A Texas DOT study by Hon et al. (2003) documents a pro- cess framework for identifying and prioritizing water quality improvement for meeting TMDLs in Texas. As part of this study, existing BMPs were assessed for their effectiveness in treating highway runoff. A summary of their results is pre- sented in Table 2. It shows the relative ranking of BMP effec- tiveness (low, medium, or high) based on pollutant removal efficiencies. Hon et al. (2003) concluded that most of the BMPs ana- lyzed are effective at removing TSS and certain metals from TSS TKN Nitrate TP Total Zinc FecalColiform Sand Filters: Austin Sand Filter Delaware Sand Filter High High Medium Medium Low Low Medium Medium Medium High Medium High Extended Detention Basin Medium Low Low Medium High Low Wet Basin High Low Medium Low High High Infiltration Basin High High High High High High Infiltration Trenches High High High High High High Vegetated Swales Low Medium Low Low High Low Vegetated Buffer Strips Medium Low Low Low High Varied From Hon et al. (2003). Rankings are defined as follows. For TSS, low = 0%–50%, medium = 51%–75%, and high = >75%. For all other constituents, low = 0%–30%, medium = 31%–65%, and high = >65%. TKN = total Kjeldahl nitrogen; total phosphorus. TABLE 2 RELATIVE RANKING OF BMP EFFECTIVENESS FOR VARIOUS CONSTITUENTS BASED ON POLLUTANT REMOVAL EFFICIENCIES

12 highway runoff but are less effective at removing nutrients and bacteria. In some cases, the BMPs most likely exported nutrients as a result of fertilizer applications in vegetated BMPs and nitrification in filtration BMPs. A second analysis of the data from the same report examined vegetated filter strips at the edge of the pavement in California and Texas even when those strips were not designed as BMPs. The study concluded that the vegetated filter strips achieve a substantial reduction in pollutant concentrations (similar to engineered systems) in highway runoff. However, the reductions in bac- teria and nutrient concentrations in vegetated BMPs were not sufficient; that is, the concentrations did not meet water qual- ity standards for the main causes of water quality impairments in Texas, which are bacteria and low dissolved oxygen. Other conventional stormwater controls were equally ineffective at treating bacteria and nutrients. Table 3 shows the vegetated filter strip data for a variety of constituents. Caltrans Studies: BMP Retrofit Pilot Program Final Report, Phase I–IV Gross Solids Removal Devices Pilot Studies A BMP retrofit pilot program study by Caltrans (2004) eval- uated the performance of various structural BMPs for treat- ing stormwater runoff from existing Caltrans facilities. The study is part of a pilot study that installs and implements a range of BMPs along freeways. Although several difficulties were encountered along the way, the program was largely a success, and several successful BMPs are now operating throughout many portions of urban southern California. All of the tested devices (except for inlet inserts) were success- fully sited without compromising the safety of the traveling public or Caltrans personnel. In terms of BMP performance, the results are summarized as the expected effluent quality for TSS, total phosphorus, and total zinc that would be achieved if each of the BMPs were subject to runoff with influent concentrations equal to that observed during the study (see Table 4). Expected efflu- ent concentrations of 0 are assumed for infiltration devices that have no discharge to surface waters. As indicated in Table 4, total phosphorus concentrations in the effluent were greater than influent concentrations for the wet basin, biofiltration swale, and biofiltration strips. For the wet basin, the Caltrans (2004) report does not offer an expla- nation for the lack of nutrient removal. However, in the case of the biofiltration swale and strips, the higher effluent concentra- tions were attributed to natural leaching of phosphorus during the dormant season from the salt grass vegetation used in these BMP types. The report suggests that a mixture of drought- tolerant native grasses is preferable to a salt grass monoculture. The report also presents pollutant removal efficiencies for the pilot study BMPs based on a comparison of the expected effluent concentrations with the Water Quality Design Storm Concentrations [event mean concentration (EMC) for pilot study]. These results are summarized in Table 5. Finally, the Caltrans (2004) report presents a series of graphs that compare the predicted effluent concentration and load reductions (including the effects of infiltration) for vari- ous BMP technologies for treatment up to the design capac- ity (i.e., design storm). The BMPs are presented across a range of relative life-cycle costs. The life-cycle costs were developed by adding the present value (assuming a 20-year life cycle and a 4% discount rate) of normalized expected operation and maintenance costs to the normalized adjusted construction costs. The term “normalized” in this instance means that the costs were normalized by the water qual- ity volume. The construction costs were adjusted to allow for additional site-specific and ancillary costs that may be encountered during future BMP retrofits. Note that infiltra- tion trenches and infiltration basins are assumed to have an effluent concentration of zero and a 100% load reduction Constituent Highway Median EMC (mg/L) Vegetated Area EMC (mg/L) Difference (%) TSS 100 20 80 NO3 1.2 0.8 33 TKN 2.1 1.3 38 Total N 3.3 2.1 36 Total P 0.27 0.18 33 Zinc 120 30 75 Fecal Coliform (cfu/100 mL) 2,500 800 68 Hon et al. (2003). NO3 = nitrate; TKN = total Kjeldahl nitrogen, N = nitrogen; P = phosphorus. TABLE 3 COMPARISON OF HIGHWAY MEDIAN EVENT MEAN CONCENTRATIONS (EMCs) AND VEGETATED AREA EMCs FOR VEGETATED FILTER STRIPS IN CALIFORNIA AND TEXAS AT THE EDGE OF THE PAVEMENT

13 TABLE 4 EXPECTED EFFLUENT CONCENTRATIONS FOR BMP TYPES From Caltrans (2004). EDB = extended detention basin; MCTT = multi-chambered treatment train; CDS = continuous deflection separator. Device TSS (Influent 114 mg/L) Total Phosphorus (Influent 0.38 mg/L) Total Zn (Influent 355 ug/L) From Caltrans (2004). TN = total nitrogen; TP = total phosphorus; TZn = total zinc; TCu = total copper; TPb = total lead; N/A = not available. BMP Type TSS TN TP TZn TCu TPb TABLE 5 REPRESENTATIVE POLLUTANT REMOVAL EFFICIENCIES (Percent) FOR PILOT STUDY BMPs

14 for the water quality design storm because they have no effluent discharge. An example graph for TSS is shown in Figure 2. Similar plots for nitrate, total Kjeldahl nitrogen, dissolved P (phosphorus), particulate P, particulate Zn (zinc), dissolved Zn, particulate Cu (copper), dissolved Cu, particu- late Pb (lead), and dissolved Pb are provided in the report (Caltrans 2004). Another study by Caltrans examined Gross Solids Removal Devices (GSRDs), a nonproprietary device to capture trash, vegetation, and debris of relatively large size (collectively “gross solids”) (Caltrans 2005a). GSRDs can be incorporated into existing or future highway drainage systems. Along with other strategies, they have been used successfully in Califor- nia to address trash TMDLs. (a) (b) FIGURE 2 Predicted TSS effluent concentration (a) and TSS load reduction (b) for various BMP technologies across a range of relative 20-year life-cycle costs. MCTT = multi- chambered treatment train; OWS = oil-water separator; EDB = extended detention basin; CDS = continuous deflection separator. Error bars indicate the reliability of the estimated effluent concentrations and load reductions (from Caltrans 2004). Infiltration trenches and infiltration basins have TSS concentrations of 0 and load reductions of 100% because they have no discharge to surface waters.

15 Constituent Influent EMC (mg/L) Influent Load (kg) Effluent EMC (mg/L) Effluent Load (kg) Removal (%) TSS 204 5705 3.50 94.0 98 Turbidity 53.0 750 4.60 62.6 92 COD 90.6 2474 11.0 286 88 TOC 32.0 692 12.6 261 62 Nitrate 1.24 20.6 0.474 7.40 64 TKN 1.59 33.8 0.591 11.9 65 Phosphorus 0.356 7.96 0.126 2.72 66 Zinc 0.143 3.85 0.008 0.214 94 Iron 3.25 70.2 0.175 3.71 95 From Keblin et al. (1998). EMC = event mean concentration; COD = chemical oxygen demand; TOC = total organic carbon; TKN = total Kjeldahl nitrogen. This study (Caltrans 2005a) is one of several pilot studies (Phases I–IV) that test the effectiveness of GSRDs of varying configurations and slopes. In this case, the Phase IV GSRD had a 35 degree slope and captured approximately 44% of the gross solids by volume during the first storm season. The low capture efficiency was primarily the result of the large momentum of the stormwater runoff, which forced the gross solids out of the GSRD. As a result, the Phase IV GSRD did not meet the TMDL criteria or Caltrans goals. Other results from Phases I–III of the pilot study, however, with different configurations of the GSRDs, showed much better capture efficiencies of generally >85% based on either wet weight or wet volume of gross solids. The interested reader is referred to Caltrans 2003a, b, and 2005b for more details. Keblin et al. (1998)—The Effectiveness of Permanent Highway Runoff Controls: Sedimentation and Filtration Systems A study by Keblin et al. (1998) focused on the performance of a sedimentation and filtration treatment system in the Edwards Aquifer Recharge zone in Texas. The study includes (1) mon- itoring and evaluating a single sedimentation/filtration facil- ity in Austin, Texas; and (2) evaluating the factors that affect sedimentation in a prototype detention basin. The results are summarized in terms of influent EMC and load, effluent EMC and load, and the percent removal, and pertain only to the runoff that entered the system (i.e., bypass was excluded, see Table 6). They also calculated the percent removal based on the total runoff that drained from the watershed (Table 7). TABLE 6 PERCENT REMOVAL FOR A SEDIMENTATION/FILTRATION SYSTEM FOR ONLY THE RUNOFF ENTERING THE SYSTEM (i.e., excluding bypass) Constituent Watershed Load (kg) Bypass + Final Effluent Load (kg) Removal (%) TSS 7132 1520 79 Turbidity 937 250 73 COD 3092 905 71 TOC 865 434 50 Nitrate 25.7 12.5 51 TKN 42.2 20.3 52 Phosphorus 9.95 4.71 53 Zinc 4.81 1.18 76 Iron 87.8 21.3 76 The total includes the runoff entering the system plus the bypass water. From Keblin et al. (1998). COD = chemical oxygen demand; TOC = total organic carbon; TKN = total Kjeldahl nitrogen. TABLE 7 PERCENT REMOVAL FOR A SEDIMENTATION/FILTRATION SYSTEM BASED ON THE TOTAL RUNOFF FROM THE WATERSHED

16 The total included the runoff entering the system plus the bypass water in order to identify the total load to the receiv- ing waters. As shown, accounting for the bypass reduces the performance (i.e., percent removal) of the sedimentation and filtration system owing to the poorer hydraulic performance (compare Tables 6 and 7). The results from the sedimentation and filtration facility suggest the following: 1. Sedimentation and filtration is an excellent form of treatment for runoff captured in the system; however, the poor hydraulic performance of this particular (sand) filter reduces the facility’s capture capacity and increases the quantity of untreated runoff that bypasses the system. 2. Results from the prototype experiments show that deten- tion time is more important than outlet design for achiev- ing satisfactory removal of constituents in runoff. 3. Treatment by sedimentation alone is comparable to sedimentation and filtration when adequate and con- sistent detention times are achieved. In addition, the Keblin et al. (1998) report summarizes removal efficiencies for seven dry detention ponds from a range of geographic locations across the United States (Table 8). Although direct comparison of these data is not pos- sible because of different methodologies and pond designs, the efficiencies provide an idea of the range of performance for dry detention ponds. In particular, the detention ponds are most effective at removing particulate constituents (e.g., TSS) in runoff and less effective at removing soluble con- stituents (e.g., NO3-N) (Keblin et al. 1998). This is supported by the data in Table 8, which show that removal efficiencies on average are among the highest for TSS and among the lowest for NO3-N. International Stormwater BMP Database The International Stormwater BMP Database, first developed in 1996 by ASCE and available at www.bmpdatabase.org, is an important source of scientifically based BMP performance monitoring and reporting protocols. Currently, the database features more than 400 studies from approximately 150 data providers from around the country including universities, municipalities, state agencies, private entities, and others. The database is continually being updated. Not all of these studies are related to highway applications. However, through the website the user may retrieve studies by individual data providers, which includes about ten transportation-related agencies. Much of the highway-specific data in the database are from Caltrans, although other state DOTs (e.g., Florida, Delaware, North Carolina, Texas, Minnesota, and Washington State) have also provided data. For this synthesis, we provide an overview of perfor- mance results for various BMP categories from the most recent analysis of the database by Geosyntec Consultants and Wright Water Engineers (2012). As noted in the analysis report, a variety of screening criteria were applied to make sure that the data sets and BMP designs are reasonably rep- resentative (see full report for details). The report presents a large amount of data. In the interest of space and applicabil- ity to highways, the data tables from the report were tailored and condensed to make the information more user-friendly for highway managers. For example, we focused only on the most prevalent TMDL pollutants of concern for state DOTs [TSS, total nitrogen (TN), total phosphorus (TP), fecal coli- form, and metals] (see Table 1) and excluded BMPs that are not considered relevant for highways (green roofs) or lack specificity (composite and treatment train BMPs). We also present only the median influent and effluent concentra- tions, the 95% confidence interval, and the statistical signifi- cance (increase or decrease) between the inflow and outflow median concentrations. The interested reader is referred to the report for additional statistics and information. The results are presented in Table 9. Here we assume that “effective” BMP performance is synonymous with sta- tistically significant decreases between median influent and effluent concentrations. Significant differences are in dark shaded bold (decrease) and light shaded italic (increase) to help the reader quickly assess the relative effectiveness of Detention Pond TSS TOC TN NO3-N TP Pb Zn Lakeridge, VA 14 — 10 9 20 — 10 London, VA 29 — 25 — 40 39 24 Stedwick, MD 70 — 24 — 13 62 57 Maple Run, Austin, TX 30 30 35 52 18 29 38 Oakhampton, Baltimore, MD 87 — — 10 26 — — Lawrence, KS 3 3 — 20 19 66 65 Greenville, NC 71 10 26 2 14 55 26 From Keblin et al. (1998); after Stanley (1996). TOC = total organic carbon; TN = total nitrogen; TP = total phosphorus; Pb = lead; Zn = zinc. TABLE 8 REMOVAL EFFICIENCIES FOR SEVEN DRY DETENTION PONDS (%)

17 BMP Type Median (95% Conf. Interval)* Median (95% Conf. Interval)* Median (95% Conf. Interval)* Median (95% Conf. Interval)* Median (95% Conf. Interval)* Median (95% Conf. Interval)* Median (95% Conf. Interval)* TSS In mg/L TSS Out mg/L TN In mg/L TN Out mg/L TP In mg/L TP Out mg/L FC In # / 100 mL FC Out # / 100 mL TZn In ug/L TZn Out ug/L TCu In ug/L TCu Out ug/L TPb In ug/L TPb Out ug/L Grass Strip 43.1(36.0, 45.0) 19.1 (16.0, 21.5) 1.34 (1.06, 1.50) 1.13 (1.00, 1.23) 0.14 (0.11, 0.15) 0.18 (0.15, 0.20) NA NA 103.3 (86.0, 120.0) 24.3 (16.0, 26.0) 24.52 (19, 26) 7.30 (6.4, 7.9) 8.83 (6.6, 11.5) 1.96 (1.30, 2.20) Bioretention 37.5 (29.2, 45.0) 8.3 (5.0, 9.0) 1.25 (1.06, 1.35) 0.90 (0.74, 0.99) 0.11 (0.08, 0.12) 0.09 (0.07, 0.10) NA NA 73.8 (62.0, 83.5) 18.3 (7.7, 25.0) 17.0 (11.0, 23.0) 7.67 (4.60, 9.85) 3.76 (2.49, 5.5) 2.53 (2.50, 2.50) Bioswale 21.7 (16.2, 26.0) 13.6 (11.8, 15.3) 0.75 (0.60, 0.92) 0.71 (0.63, 0.82) 0.11 (0.09, 0.12) 0.19 (0.17, 0.20) 4720 (2120, 5500) 5000 (2600, 6200) 36.2 (30.0, 40.0) 22.9 (20.0, 26.6) 10.86 (8.70, 13.20) 6.54 (5.7, 7.7) 3.93 (2.80, 5.00) 2.02 (1.80, 2.29) Detention Basin 66.8 (52.3, 76.1) 24.2 (19.0, 26.0) 1.40 (1.03, 1.57) 2.37 (1.75, 2.69) 0.28 (0.25, 0.30) 0.22 (0.19, 0.24) 1480 (789, 1900) 1030 (500, 1900) 70.0 (40.0, 95.0) 29.7 (17.1, 38.2) 10.62 (7.78, 14.00) 5.67 (4.0, 6.8) 6.08 (3.86, 8.0) 3.10 (2.15, 4.30) Manufactured Device 34.5 (30.0, 36.8) 18.4 (15.0, 19.9) 2.27 (1.98, 2.65) 2.22 (1.90, 2.41) 0.19 (0.16, 0.22) 0.12 (0.10, 0.13) NA NA 87.7 (79.0, 95.0) 58.5 (52.8, 63.5) 13.42 (11.90, 14.70) 10.16 (7.94, 11.0) 8.24 (6.77, 9.56) 4.63 (3.80, 5.16) Manufactured Device-F** NA NA NA NA NA NA 478 (200, 1300) 1890 (200, 3000) NA NA NA NA NA NA Manufactured Device-P** NA NA NA NA NA NA 2210(900, 3000) 2750 (1400, 5000) NA NA NA NA NA NA Media Filter 52.7 (45.9, 58.2) 8.7 (7.4, 10.0) 1.06 (0.85, 1.25) 0.82 (0.68, 0.99) 0.18 (0.16, 0.19) 0.09 (0.08, 0.10) 1350 (725, 2300) 542 (200, 625) 77.3 (68.2, 86.0) 17.9 (15.0, 20.0) 11.28 (10.0, 12.68) 6.01 (5.1, 6.6) 10.5 (8.02, 11.79) 1.69 (1.30, 2.00) Porous Pavement 65.3 (45.0, 80.3) 13.2 (11.0, 14.4) NA NA 0.15 (0.12, 0.16) 0.09 (0.08, 0.09) NA NA 57.6 (49.6, 66.0) 15.0 (12.5, 16.8) 13.07 (11.45, 15.3) 7.83 (6.80, 8.10) 4.30 (3.28, 5.47) 1.86 (1.38, 2.21) Retention Pond 70.7 (59.0, 79.0) 13.5 (12.0, 15.0) 1.83 (1.60, 1.98) 1.28 (1.19, 1.36) 0.30 (0.27, 0.31) 0.13 (0.12, 0.14) 1920 (970, 2650) 707 (200, 1160) 53.6 (49.0, 59.0) 21.2 (20.0, 23.0) 9.57 (8.0, 10.0) 4.99 (4.06, 5.0) 8.48 (6.80, 9.41) 2.76 (2.00, 3.00) Wetland Basin 20.4 (16.6, 24.4) 9.06 (7.0, 10.9) 1.14 (1.04, 1.28) 1.19 (1.04, 1.21) 0.13 (0.11, 0.14) 0.08 (0.07, 0.09) 13000 (5080, 21000) 6140 (230, 11800) 48.0 (40.6, 53.2) 22.0 (16.7, 24.3) 5.61 (4.36, 6.34) 3.57 (3.00, 4.00) 2.03 (1.57, 2.24) 1.21 (1.00, 1.55) Wetland Channel 20.0 (17.0, 22.0) 14.3 (10.0, 16.0) 1.59 (1.38, 1.78) 1.33 (1.05, 1.56) 0.15 (0.13, 0.17) 0.14 (0.13, 0.17) NA NA 23.0 (16.0, 30.0) 15.6 (11.0, 20.0) 4.52 (3.80, 5.10) 4.81 (3.61, 5.2) 2.94 (1.90, 4.20) 2.49 (1.40, 3.11) *Computed using the BCa bootstrap method described by Efron and Tibishirani (1993). **For bacteria, manufactured devices are broken down into inlet insert/filtration (Manufactured Device − F) and physical settling/straining devices (Manufactured Device − P). NA = Limited or no data available. <values> Hypothesis testing shows statistically significant decreases for this BMP category. <values> Hypothesis testing shows statistically significant increases for this BMP category. TABLE 9 INFLUENT/EFFLUENT SUMMARY STATISTICS FOR TOTAL SUSPENDED SOLIDS (TSS), TOTAL NITROGEN (TN), TOTAL PHOSPHORUS (TP), FECAL COLIFORM (FC), TOTAL ZINC (TZN), TOTAL COPPER (TCU), AND TOTAL LEAD (TPB) AS ADAPTED FROM GEOSYNTEC CONSULTANTS AND WRIGHT WATER ENGINEERS (2012). DATA ARE FROM THE INTERNATIONAL STORMWATER BMP DATABASE (www.bmpdatabase.org).

18 one BMP versus another. However, these results should be used cautiously. As stated previously, the results are not nec- essarily all derived from highway studies. Furthermore, the statistics may not be meaningful in cases where the influent concentrations are already relatively clean to begin with or in cases where the results include many nondetects (although this is more of an issue for dissolved metals, which are not included here) (Geosyntec Consultants and Wright Water Engineers 2012). Nonetheless, the results present a reason- able basis for appraising the general performance of different BMP categories for TMDL planning purposes. As shown in Table 9, the data support the view that TSS and total metals are relatively easy to treat with almost any type of BMP. This was also generally the case in the other literature sources reviewed (e.g., see Table 20 in chapter four). Note, however, that this may not necessarily be the case with dis- solved metals, which are generally more difficult to remove. Similarly, nutrients are also relatively difficult to treat. In some cases the BMP may be a net contributor of nutrients (e.g., grass strips and bioswales show a statistically signifi- cant increase in total phosphorus) (Table 9). Media filters and retention ponds were the only BMPs effective at treating both nitrogen and phosphorus. Fecal coliform concentrations were highly variable, which is typical of bacteria in surface waters during storm flows. The database provides two additional manufactured device options for fecal coliforms: inlet insert/ filtration devices (“F”) and physical settling/straining devices (“P”). Neither is effective at treating fecal coliform; however, media filters and retention ponds showed a statistically sig- nificant reduction in fecal coliform median concentrations. Overall, media filters and retention ponds appear to be effec- tive for all of the TMDL pollutants assessed, including TSS, nutrients, fecal coliform, and total metals. Other Studies Several other studies were identified with highway BMP per- formance data, mostly related to specific innovative practices being developed and utilized in certain areas of the country. For example, two reports, one each by Barrett (n.d.) and Eck et al. (2012), show significant pollutant reductions from permeable friction courses, a roadway material 25 to 50 mm thick that is applied on top of regular impermeable pavement to enhance driving safety during wet weather conditions and improve water quality. The Barrett (n.d.) study found statistically sig- nificant concentration reductions of 92% for suspended solids, 90% for total lead, 51% for total copper, and 74% for total zinc relative to a conventional asphalt surface in Texas. The Eck et al. (2012) study found similar reductions in suspended solids (90%) and lower effluent concentrations of phosphorus, copper, lead, and zinc for permeable friction courses in Texas and North Carolina. The water quality benefits were similar in both states and lasted through the design life of the pavement. A study by WSDOT (2006) examined an ecology embank- ment (sometimes referred to as a media filter drain) test system located along a highway shoulder in western Wash- ington State. Ecology embankments are linear flow-through treatment devices designed for highway side-slopes, medi- ans, borrow ditches, or other linear depressions, particularly in areas with a limited ROW. The monitoring results showed that treatment goals for total suspended solids (80% removal) and total phosphorus (50% removal) were generally met or exceeded for the sampled storm events. In addition, the ecology embankment was found to provide “enhanced treat- ment” of dissolved zinc (median removal of 80% to 90%) and dissolved copper (median removal of about 40%), mean- ing that it performed better than basic treatment facilities. The median removal for total zinc was 85% to 90% and the median removal for total copper was 86%. Finally, a comprehensive study by MnDOT (2005) pro- vides performance data for total suspended solids and total phosphorus for a variety of BMPs. Although they present removal efficiencies similar to the previous studies, they also provide estimates of TSS and phosphorus removed over 20 years as a function of water quality volume for dry deten- tion basins, wet basins, constructed wetlands, infiltration trenches, bioinfiltration filters, and sand filters. An example is shown in Figure 3. Further details are available in the report (MnDOT 2005). NONSTRUCTURAL BEST MANAGEMENT PRACTICE PERFORMANCE Nonstructural BMPs for decades have been qualitatively recognized as having the potential to contribute to water quality improvements, but quantification of those contribu- tions remains elusive. Nonstructural BMPs may be consid- ered as source controls rather than as treatment facilities because their application is generally distributed through- out a watershed and their intent is to prevent pollutants from entering the watershed drainage system rather than removing them once they are moving with stormwater or snowmelt runoff. Nonstructural BMPs are programmatic in the sense that their successful implementation requires ongoing efforts to identify implementation opportunities, select appropriate tech- nologies, develop implementation plans, and budget the nec- essary resources to sustain the plans. Nonstructural BMPs that may be applicable to transportation environments include the following: • Street sweeping • Catch basin cleaning • Naturescaping and tree planting • Stream restoration and reclamation • Vegetation management • Anti-icing management • Downspout disconnection (rest areas, maintenance facilities, etc.) • Erosion control on construction sites

19 • Spill prevention and response plans • Education/awareness for the public and employees. Effectiveness of any of these BMPs depends on a num- ber of factors that have been recognized for over a decade (NVPDC 1996). These include how widely applicable the BMP can be applied, how widely the BMP is applied, how long the BMP is sustained, the effectiveness of the technol- ogy used, watershed characteristics, and site-specific hydrol- ogy. Because each of these can vary widely by transportation agency, watershed, and state, quantification of performance for TMDLs is challenging. Table 10 summarizes the quantitative performance met- rics identified in the literature search. None of these metrics appear to be broadly transferable nationwide largely because of the variations in conditions mentioned previously. How- ever, they do provide reference points on how quantification has been addressed to date and the challenges that remain. Street sweeping effectiveness is highly variable and dependent on sweeping technology and sweeping frequency (NVPDC 1996; HECI 2006). Generally, sweeping technolo- gies have improved and older studies tend to report less effec- tive performance. Climate, including frequency and magnitude of rain storms and the extent of a winter season affected by snowfall, will also affect performance. One study in Table 10 showed a range of annual load reduction for TSS, phospho- rus, and metals. The differences in reductions reflect, in part, the effectiveness in picking up different size fractions to which the other pollutants may be attached. A spreadsheet model estimated that load reductions though the source of these estimates was not apparent (HECI 2006). One limi- tation of metrics that simply measure the mass of material removed is that trash is generally included in that mea- surement, but may not contribute to TSS or other pollutant TMDLs (assuming there is not a separate TMDL for trash). The cost-effectiveness of street sweeping in a transportation context, with frequency a major driving variable, was not addressed. The effectiveness of catch basin cleaning has been both measured and modeled as shown in Table 10. In one model- ing effort, reductions in annual loads were tied to cleaning frequency with virtually no reduction with infrequent clean- ing (every 5 years) and up to 45% reduction in loading when cleaning every three months. Removal effectiveness is tied to the size of the solids storage and the rate at which the storage fills. As the storage approaches its capacity, previously cap- tured material is more likely to be resuspended with larger storms. Other assessments shown in the table measure per- formance in terms of mass removed. As with street sweeping, trash is generally included in these measurements. The state of Maryland has developed performance metrics in terms of an equivalency to structural BMPs. It is beyond the scope of this synthesis to evaluate this methodology, but it does represent an agreement between the Maryland State High- way Administration (MDSHA) and the regulatory authorities regarding the effectiveness of this nonstructural BMP. Documentation of the effectiveness of tree planting, or more generically “naturescaping,” is limited. An older report provided examples of the reduction in runoff con- centrations for nutrients with a program of reforestation (NVPDC 1996). The specific numerical values are intended to be illustrative of what may be achieved by converting an FIGURE 3 Estimated TSS removed in 20 years for dry detention basins with the 67% confidence interval (from MnDOT 2005).

20 urbanized land use back to a more natural landscape. Tree planting is also included in the Maryland Accounting Proto- col (MDSHA 2012). Stream restoration is considered a nonstructural BMP because it can reduce the loading of excess sediment and, potentially, other pollutants by reducing erosion and by pro- viding the means to capture sediment in riparian vegetation and the floodplain. One site-specific study in the Denver, Colorado, area estimated that 90 to 220 pounds of phos- phorus is immobilized per mile of stream reclamation every year (CCBWQA 2011). That is, by reducing the erosion and unraveling of the stream the sediment and corresponding phosphorus remained in the stream banks and bed, while a reconnection to the floodplain allows capture of phosphorus from upstream sources. Stream restoration is also included in the Maryland Accounting Protocol (MDSHA 2012). Quantitative measures of other nonstructural BMPs were limited or not found in the literature review. Vegetation man- agement efforts are targeted at not over-fertilizing or irrigating vegetated areas and reducing pesticide use within the transpor- tation corridor, thereby reducing avoidable discharges of nutri- ents and pesticides. Similarly, anti-icing management efforts are targeted to reducing over-application of sand and chemi- cals while protecting the public safety. One excellent example BMP/Study TSS Nutrients (phosphorus, nitrogen) Metals (copper, lead, zinc) Street Sweeping Performance modeled with sweeping frequency from twice weekly to biweekly (HECI 2006) 45 to 70% reduction in annual loads 35 to 60% reduction in phosphorus annual loads 25 to 60% reduction in annual loads Spreadsheet model assuming sweeping frequency of 6 to 12 times per year (HECI 2006) 25 to 35 lb removed per lane mile per year <0.1 lb phosphorus removed per lane mile per year <0.1 lb removed per lane mile per year Catch Basin Cleaning Spreadsheet model with cleaning frequency from every 3 months to once in 5 years (HECI 2006) 0 to 45% reduction in annual loads Alameda County study with cleaning frequencies from monthly to annual (HECI 2006) 8 to 70 lb removed per cleanout; generally a decline in removal per cleanout with higher frequency Spreadsheet model with frequency ranging from every 4 days to every 40 days (HECI 2006) 35 lb removed per cleanout; constant regardless of frequency 0.00003 lb phosphorus removal per catch basin per cleaning 0.0000012 to 0.0000064 lb removal per catch basin per cleaning MD NPDES Accounting Protocol (specifics on pollutants not given) (MDSHA 2012) 2000 lb removed by cleaning is equivalent to 0.4 impervious acres of structural BMP treatment Tree Planting Reforestation (NVPDC 1996) Nonpoint source runoff concentrations for phosphorus reduced from 0.205 to 0.15 mg/L and for nitrogen reduced from 0.139 to 0.078 mg/L as examples MD NPDES Accounting Protocol (specifics on pollutants not given) (MDSHA 2012) 1 acre of planting is equivalent to 0.38 impervious acres of structural BMP treatment Stream Restoration Site-specific evaluation (CCBWQA 2011) 90–220 lb phosphorus immobilized per mile per year MD NPDES Accounting Protocol (specifics on pollutants not given) (MDSHA 2012) 100 linear feet of restoration is equivalent to 1 impervious acre of structural BMP treatment TABLE 10 NONSTRUCTURAL BMP PERFORMANCE MEASURES

21 of anti-icing management is a pilot project implemented by NHDOT. The pilot project tests advanced technology to measure the temperature of the road surface and automati- cally manage the rate of salt application. Early results were described during the interview as very promising in terms of reducing the amount of salt applied (20% reduction), thereby reducing the chloride loads to receiving streams and water bodies. The next phase of this effort is intended to apply this technology to private snow removal operations, which are significantly greater in terms of total impervious cover (local roads and parking lots) and total material applied to the pave- ment surface. Downspout disconnection effectiveness is lim- ited in a transportation context because the area covered by roofs is limited. Erosion control at construction sites and spill prevention and response plans are generally already covered under MS4 requirements and other permitting programs. Further cost- effective practices in these areas beyond existing require- ments may be identified for TMDL compliance, but may be limited. Education and awareness campaigns for the general public and state transportation employees may have some benefits, but these have not been quantified. For example, campaigns encouraging driving habits that result in the reduced deposi- tion of pollutants related to cars on the roadway may improve water quality. However, response rates to public education campaigns may be as low as 1.5% to 8% (HECI 2006). Effective strategies for TMDL compliance may include application of multiple structural and nonstructural BMPs. Among the greatest needs for TMDL managers is not only how to measure effectiveness of individual measures, but how to account for measure interactions. BEST MANAGEMENT PRACTICE DESIGN STANDARDS BMP design standards have generally evolved slowly for DOTs. The original stormwater quality BMPs developed in the mid-1980s (Schueler 1987) are similar to those being implemented today. Some new practices have been intro- duced (e.g., bioretention) and some of the older practices are now being recognized as providing a water quality benefit (e.g., street sweeping) (EPA 2011; MDE 2011a, b). How- ever, while several of the DOTs indicated some use of new or innovative structural or nonstructural BMPs on DOT ROWs such as biochar, iron filings, polyacrylamide, etc., the major- ity of BMPs being implemented are detention and extended detention basins, grass swales, and infiltration. Nonetheless, within the past 5 years the pace of change in BMP design standards as implemented by most state stormwater programs has begun to accelerate. This coincides with the EPA’s accel- erated implementation of the NPDES permit program and local TMDL development. As stated previously, DOTs have traditionally used practices designed to reduce the peak runoff rate (Oregon State University et al. 2006). These practices were the first to evolve and included BMPs such as an extended deten- tion to allow for settling of solids. All the DOTs interviewed included detention and extended detention basins on the list of the most common BMPs. However, research is gener- ally revealing these practices to be limited in their ability to address many of the TMDL pollutants of concern; for exam- ple, nutrients and metals (Hon et al. 2003). Nine of 12 DOTs (New Hampshire, New York, Colorado, Delaware, Washington State, California, Georgia, North Carolina, and Virginia) indicated that they have a drainage and/or BMP design manual, either developed specifically for the DOT applications, modified from the state storm water agency manual to fit DOT ROW applications, or adopted directly from the state stormwater agency. In some cases, the DOTs indicated that the state is in the process of adopt- ing or has recently adopted new BMP design standards as a result of TMDLs and NPDES permits; for example, Virginia, Maryland, Delaware, and New York. The changes in BMP standards are incorporating runoff volume reduction as a new BMP performance metric consistent with the EPA’s adoption of the recommendations of the National Research Council report Urban Stormwater Management in the United States (NRC 2008). Achieving volume-reduction goals is intended to replicate the pre-development (i.e., before new urban structures, not before human disturbance) runoff character- istics in terms of volume and duration of flows. The greatest challenge noted by many DOTs during the interviews is to establish which of these new BMPs are appropriate for linear ROW application. BEST MANAGEMENT PRACTICE MAINTENANCE Maintenance of the stormwater infrastructure and storm- water BMPs is a fundamental element of long-term compliance with TMDLs and local stormwater management programs. Ensuring on-going maintenance of the growing inventory of structural BMPs is both a financial and an administrative challenge for state stormwater programs. No DOTs indicated that they were experiencing maintenance audits or enforce- ment pressure from their respective state programs or the EPA. However, all the states are required to ensure that the MS4 permittees, including the DOTs, are meeting the substantial requirements for maintenance (EPA 2007). Another consideration is the ability to manage the rou- tine maintenance of BMPs as part of the regular maintenance practices already programmed into the DOT budgets. Reg- ular ROW maintenance including grass mowing and litter pick-up has served to meet the routine maintenance needs of the traditional detention and retention basins. The new gen- eration of water quality BMPs, such as bioretention and other vegetated practices with subsurface components, require a

22 different though generally similar level of effort for routine maintenance. Based on the DOT interviews, it is the uncer- tain financial and manpower estimates associated with the infrequent maintenance and the potential for major overhaul of the BMPs that in part fuels the resistance of implementing the new practices. Routine maintenance is generally described as procedures expected to be performed on a regular basis to maintain the proper working order of a BMP, such as vegetation manage- ment, trash and debris removal, and minimal grading and repairs. Infrequent maintenance activities are described as those tasks anticipated to be performed periodically but less frequently than routine maintenance. Examples of infrequent maintenance include accumulated sediment removal and dis- posal; soil media, mulch, and riprap replacement; and larger scale grading and repairs. Only four DOTs (Delaware, California, Georgia, and Min- nesota) indicated that they conducted any systematic obser- vations of BMPs located within the ROW to help predict the projected service life or maintenance interval. This would be considered a very inexpensive form of monitoring that can support the adaptive management format of evaluating BMPs to help inform the location and design of the practices. This effort would also serve to identify the potential need for the infrequent maintenance before it reaches the more expen- sive overhaul stage. Four of the 12 DOTs (Georgia, Washington State, North Carolina, and California) indicated having a maintenance pro- cedure manual that addresses stormwater BMPs. NCDOT, for example, has developed a Stormwater Control Inspection and Maintenance Manual that identifies routine and emer- gency maintenance procedures, reporting and record keep- ing protocols, and specific activities and checklists for each BMP approved for use on the DOT ROW. One DOT indicated that a maintenance procedure manual is under development with a specific section dedicated to stormwater BMPs, and two DOTs indicated that they are adapting a standard DOT drainage system maintenance guide to address routine BMP maintenance. Aside from BMP maintenance, the primary drivers of DOT maintenance expenditures are related to pavement fail- ure (Arika et al. 2006). The two leading causes of pavement failure are (1) traffic volume (the pavement surface is no lon- ger able to absorb and transmit the wheel loading to the sub- grade because of increases in traffic and/or traffic loads), and (2) poor drainage (the surface and subsurface drainage sys- tem must keep the surface and subsurface water sufficiently below the pavement and subgrade). Although DOTs are not able to reduce traffic volume, they can control the manage- ment of runoff and groundwater. This was described by sev- eral DOTs during the interviews as a significant concern with using stormwater retention BMPs, which by definition retain water within or adjacent to the ROW despite design standards that promote the opposite, namely directing the runoff away from the road surface and ROW as quickly as possible. HIGHWAY BEST MANAGEMENT PRACTICES COSTS Based on the DOT interviews, BMP implementation costs (capital, O&M, land acquisition, life cycle) are generally not closely tracked. In some instances, they are not tracked at all, or are difficult to track, owing to the lack of an adequate tracking system. Some DOTs were clearly better at tracking than others. They typically tracked capital costs and/or O&M costs only (only two of the 12 DOTs interviewed tracked land acquisition costs and only one tracked life-cycle costs). However, none of the costs were assessed at the individual BMP level. Capital and/or O&M costs were often rolled into project budgets or larger programmatic budgets (e.g., NPDES or TMDL implementation). North Carolina and Washington State were probably the two most comprehensive cost trackers. The NCDOT tracks design, capital, and O&M costs for retrofit projects because of an aggressive retrofit requirement in their TS4 permit (70 retrofits required over the 5-year permit cycle). The WSDOT completes an Environmental Mitigation Cost Study every 3 years that tracks capital, design, and land acquisition costs for BMP implementation. A copy of the latest study is available in WSDOT (2009). WSDOT is also developing a Highway Activity Tracking System that will include some tracking of O&M costs. NPDES compliance budgets, if known, were typically in the $2 to 4 million range for the smaller states and about $90 million for Caltrans. Actual cost data on highway BMPs are difficult to obtain as they are not reported systematically and may have lim- ited application in other areas owing to differences in BMP design standards, materials costs, and other project-specific elements. Nonetheless, through the literature review, some cost data were located. One goal is to ultimately provide life-cycle cost informa- tion that refers to the total project cost across the life span of a BMP, including design, construction, and O&M costs (Arika et al. 2006). In this context, the term “life-cycle” refers to 10-year and 20-year cost estimates over the service life of the BMP. In the following sections we present design, con- struction, and O&M costs individually and then summarize the information at the end with life-cycle cost estimates that include all of the components mentioned previously. There is some overlap, as life-cycle costs include design, construc- tion, and O&M costs. However, the intent is to build up from individual cost elements to a more comprehensive picture of total life-cycle costs for the benefit of DOT managers who may wish to focus either on a particular cost element or on the broader life-cycle costs. Several reports are highlighted

23 here; however, the reader is encouraged to refer to the full reports for additional details. Design Costs There was generally little information available in the litera- ture on design costs. Here we present highlights from URS Corporation (2010), which is a study of stormwater control measures (SCMs) (i.e., BMPs) related to bridge infrastruc- ture for the NCDOT. The report determines the cost of imple- menting effective treatments for existing and new bridges over waterways in North Carolina. SCMs are grouped into four main categories: Level I treatment, Level II treatment, design-related, and maintenance. Level I treatment SCMs remove pollutants that enter storm water runoff as it travels over the bridge deck (i.e., pavement abrasion, atmospheric deposition, leaching of metals from vehicles), and Level II treatment SCMs reduce erosion by stabilizing soil, dissipat- ing energy, and promoting diffuse flow of bridge runoff. Here we focus on Level I and II treatment SCMs as they are the most relevant to DOTs (URS Corporation 2010). Although the study relates specifically to bridges, it is also reason- able to consider these costs as generally reflective of the financial resources needed to retrofit the nonbridge highway environment, as well (Andrew McDaniel, NCDOT, personal communication, July 19, 2012). In the report, SCM design costs are expressed as a per- centage of construction costs and are based on the cost of retrofits. The retrofits are expected to incur higher SCM design and construction costs than larger newer construc- tion projects because of the need for additional site visits, surveying, and utilities investigations. The report concludes that design costs for SCMs associated with new construc- tion projects were approximately 40% of the design costs for SCM retrofit projects. Based on this assumption, the design budgets were adjusted to yield new construction SCM design costs, which were compared with preliminary construction cost estimates to assess the relationship between design and construction costs. This relationship, which is based on a variety of Level I SCMs, is presented in Figure 4. Construction Costs Data on construction costs are available in three reports by URS Corporation (2010), Arika et al. (2006), and Caltrans (2004). In the URS Corporation (2010) report, construction costs for Level I SCMs are based on actual costs from previ- ously constructed NCDOT SCM projects or preliminary con- struction estimates for planned and ongoing projects. These costs are presented in Table 11. The relationship between the construction costs and the cor- responding water quality volume (represented as impervious acres) was used to develop regression equations for predict- ing construction costs based on impervious drainage area. An example of a regression equation is presented in Figure 5 for a bioretention basin. Additional plots for dry detention basins, filtration basins, stormwater wetlands, level spread- ers, and environmental site design–dry detention basins can be found in the report (URS Corporation 2010) and the equations are summarized in Table 11 (additional equations available in the report for level spreader and dry detention environmental site design). The second report (Arika et al. 2006) presents construc- tion costs for common BMPs as a series of equations based on data from MnDOT (2005): • Dry Pond CC = 97.338  WQV-0.3843 • Wet Pond CC = 230.16  WQV-0.4282 • Constructed Wetland CC = 53.211  WQV-0.3576 FIGURE 4 Comparison of design cost percentage to construction cost for a variety of stormwater control measures (from URS Corporation 2010).

24 • Infiltration Trench CC = 44.108  WQV-0.1991 • Sand Filter CC = 389.00  WQV-0.3951 • Bioretention CC = 0.0001  WQV + 9.00022 • Grass Swales CC = 21.779  ln(A) - 42.543 Where CC is the construction cost expressed in dollars per unit of water-quality volume (WQV) or BMP area A (ac). The WQV is defined as the volume of water the facility will treat and can be calculated as follows: WQV P Rv A=  4356012    Where P is the design precipitation depth (in.), Rv is the ratio of runoff to rainfall in the watershed, and A is the watershed area (acres). Arika et al. (2006) also present a graph showing the construction costs as a function of WQV for various BMPs (Figure 6). The report by Caltrans (2004) presents site-specific BMP costs that were reviewed by a technical work group to develop “generic” retrofit costs that could reasonably be applied to other BMP retrofit projects. The costs were developed by reviewing the specific construction items for each site, elimi- nating those that were atypical, and reducing the costs that were considered to be in excess of what would “routinely” be encountered in a retrofit situation. Construction costs were adjusted to account for ancillary site-specific costs based on the discussions of the technical work group. Construction costs were also normalized to the water quality volume (i.e., cost per WQV or $/m3). The results are presented in Table 12. For more details, see Caltrans (2004). The Caltrans (2004) report also presents mean unit con- struction costs ($/m3 of WQV) calculated by a third party cost work group from data collected in a nationwide sur- vey. One set of columns lists the statistics from the Caltrans pilot study (Caltrans 2004), a second set lists statistics of all nationwide data (excluding Caltrans), and a third set gives statistics only from BMP construction by the MDSHA (see Table 13). The MDSHA BMP projects were selected for comparison because the costs are combined with broader highway reconstruction costs and therefore are thought to represent greater cost savings than retrofit programs in other states. However, the Caltrans (2004) authors noted a number of limitations to the MDSHA dataset such as (1) the lack of separate line-item costs for these BMPs, which implies the BMP Number of Systems Analyzed Average Cost Equation (y = total construction cost, x = impervious drainage area in acres) Bioretention Basin n = 9 $76,748 y = 27903x + 48785 Dry Detention Basin n = 5 $41,541 y = 1064.5x + 39592 Filtration Basin n = 7 $107,650 y = 30959x + 46156 Stormwater Wetlands n = 3 $75,407 y = 25157x 7191 Swales n = 2 $12,483 N/A From URS Corporation (2010). N/A = not available. TABLE 11 SCM CONSTRUCTION COST DATA FIGURE 5 Comparison of bioretention basin construction cost to impervious drainage area (from URS Corporation 2010).

25 data could not be independently verified; and (2) the cost database is small and only contains between one and five examples of each BMP type. As shown in the table, MDSHA costs are much lower than the Caltrans pilot study costs and are generally simi- lar to nationwide costs. Despite the limitations of the Mary- land dataset, the data support the view that integrating BMP implementation costs into larger highway projects does result in significant cost savings. In addition, the water quality vol- umes for the MDSHA sites are substantially larger than the Caltrans pilot study sites, suggesting that the drainage area of the BMP can be a significant source of cost savings owing to economies of scale (Caltrans 2004). Operation and Maintenance Costs O&M costs are those post-construction costs necessary to ensure or verify the continued effectiveness of a BMP (Arika et al. 2006). These costs are summarized in several reports highlight here. In the URS Corporation report (2010), O&M costs are estimated to include both routine maintenance and infrequent maintenance. Routine maintenance is based on procedures expected to be performed on a regular basis to maintain the proper working order of an SCM, such as vege- tation management and trash and debris removal. Infrequent maintenance includes, for example, sediment removal; soil media, mulch, and riprap replacement; and larger scale grad- ing and repairs. Typical routine and infrequent maintenance costs are provided in Table 14 for various Level I and II treat- ment SCMs. In Arika et al. (2006), maintenance costs are expressed as a fraction of construction cost, which varies depending on the BMP type. EPA (2004) presents maintenance costs as an annual percentage of construction costs for common BMPs as follows: • Dry Pond <1% • Wet Pond 3% to 6% • Constructed Wetland 3% to 6% • Infiltration Trench 5% to 20% • Infiltration Basin 1% to 3% • Sand Filter 11% to 13% • Bioretention 5% Figure 7 shows the maintenance costs for various BMPs as related to the WQV treated using a period of analysis of 20 years and a discount rate of 7% from EPA (2004). Caltrans (2004) summarizes annual maintenance in terms of equipment and materials cost and average labor hours performed for each of the tested devices during a retrofit pilot study. Labor hours were generally high for all devices; however, they do not correspond to the effort that would rou- tinely be required to operate the piloted BMPs or reflect the design lessons learned during the course of the study. The largest maintenance item for each of the BMPs was gener- ally vegetation management. Maintenance costs are summa- rized in Table 14. Note that sand filters were relatively cheap to maintain because (unlike the other BMPs) they were constructed of concrete. This conflicts with the relatively expensive maintenance costs of sand filters from Arika et al. (2006) (see Figure 6), which may reflect geographic differ- ences between the Arika study (from Minnesota) and the Caltrans study, and/or differences in maintenance proce- dures in these two states. FIGURE 6 Construction cost for selected stormwater BMPs (from Arika et al. 2006).

26 Life-Cycle Costs As stated earlier, life-cycle costs represent the sum total of design, construction, and O&M. Life-cycle costs are generally not well documented. However, four reports by Arika et al. (2006), URS Corporation (2010), King and Hagan (2011), and Caltrans (2004) were identified with some comprehensive life-cycle costs and methodologies for calculating them. It can be noted that these reports use different approaches for life-cycle cost estimating and gen- erally assess different types of BMPs. Therefore, a direct comparison of the data from all four reports is not possible. However, in a few cases, the reports examine the same BMP type and the cost estimates can be variable. As an example, construction costs for swales range from $12,000 to $57,818 (compare Tables 14, 15, and 16). This variability is likely the result of the different calculation methods used in the reports and perhaps real differences in costs by geo- graphic region (these studies include four different states: California, Maryland, Minnesota, and North Carolina). Highlights from each report, including summary tables and cost equations, are presented here. As discussed previously, there is some overlap with the previous sections as life- cycle calculation includes design, construction, and O&M cost elements. In Arika et al. (2006), life-cycle costs are provided for dry ponds, wet ponds, constructed wetlands, infiltration trenches, From Caltrans (2004). The number of BMPs are shown in parentheses; e.g., EDB (4). EDB = extended detention basin; IB = infiltration basin; WB = wet basin; MFSTF = media filter, storm filter; MFSA = media filter sand Austin type; MCTT = multi-chambered treatment train; BSW = biofiltration swale; BSTRP = biofiltration strip; IT/STRP = infiltration strip. TABLE 12 ADJUSTED CONSTRUCTION COSTS BY BMP TYPE (1999 Dollars)

27 From Caltrans (2004). TABLE 13 COMPARISON OF MEAN UNIT CONSTRUCTION COSTS AND WQVs FROM NATIONWIDE SURVEY TO ADJUSTED MEAN UNIT CONSTRUCTION COSTS AND WQVs IN CALTRANS RETROFIT PILOT PROGRAM (1999 dollars) From Caltrans (2004). Maintenance is mainly for vegetation management except for sand filters, which are concrete. TABLE 14 BMP ACTUAL ANNUAL MAINTENANCE EFFORT FOR CALTRANS BMP RETROFIT PILOT PROGRAM

28 infiltration basins, sand filters, bioretention areas, and veg- etated swales. The life-cycle costs are presented as a series of equations where they are the sum of construction costs and design costs, which include design, permitting, erosion control, and contingency costs. Additional equations are pro- vided for computing (1) the construction cost as a function of the Water Quality Volume (or Qv in this case), (2) the main- tenance cost as a function of the construction cost and the multi-year discount factor, and (3) the multi-year discount factor over a specific period of analysis. In addition, indi- vidual cost data (i.e., construction costs, design costs, main- tenance costs, etc.) are presented for a variety of drainage area sizes. A sample of the life-cycle cost data is presented in Figure 8. For life-cycle cost estimates for other BMP types see the full report (Arika et al. 2006). A report by the URS Corporation (2010) also presents life-cycle cost estimates based on a 10-year BMP service FIGURE 7 Present worth annual maintenance costs as a function of water quality volume for various BMPs. References can be found in Arika et al. (2006). FIGURE 8 Life-cycle cost estimates for a dry pond. References can be found in Arika et al. (2006).

29 life. Typical 10-year life-cycle cost estimates include capital costs (construction plus design) and operating costs (which includes maintenance). A summary of the 10-year cost data is provided in Table 15 for Level I and II treatment SCMs (as defined earlier). Notes and references associated with the table can be found in URS Corporation (2010). Comprehensive life-cycle cost data are provided in a report by King and Hagan (2011). They present planning level unit cost estimates per acre of impervious area treated. The data are designed for use in Maryland to assist counties in com- paring the effectiveness of BMP implementation strate- gies based on costs as well as their potential contributions to meeting county TMDL targets. Therefore, the data may be of limited use outside of Maryland, but the report provides a basis for extrapolating to other areas and presents a method for devising similar life-cycle cost estimates in other locali- ties [for full details, see King and Hagan (2011)]. A summary from the report is presented in Table 16, which presents both the total costs over a 20-year BMP service life and the aver- age annualized cost over 20 years. Costs are further broken down into pre-construction costs, construction costs, land costs (i.e., land acquisition), and post-construction costs. Table references and assumptions used are available (see King and Hagan 2011). Additional life-cycle cost estimates are presented in Cal- trans (2004). In this report, life-cycle costs include construc- tion costs and O&M costs assuming a 20-year BMP service life and a 4% discount factor. As discussed previously, con- struction costs (and in this case O&M costs) were adjusted to account for ancillary site-specific costs, and life-cycle costs are normalized by water quality volume (i.e., cost/WQV or $/m3). The cost data are presented in Table 17. Additional details are available in the Caltrans report (2004). TOTAL MAXIMUM DAILY LOAD IMPLEMENTATION PLANS A number of state DOTs, although not all, have developed TMDL implementation plans to address their WLA require- ments. This section provides some detailed examples of implementation plans from California, Washington State, Colorado, and New York to address a variety of TMDL pol- lutants of concern (metals, toxics, fecal coliform, dissolved oxygen, temperature, PCBs, sediment, nutrients, pathogens, etc.). It is outside the scope of this synthesis to address the TMDL implementation plans in all 12 of the interviewed states; however, by presenting selected highlights, we hope to identify some key patterns and themes common to all implementation plans. California—Los Angeles River Metals TMDL Caltrans is a named stakeholder in a Los Angeles River metals TMDL issued in 2005. Its implementation plan is described in Caltrans (2010a). The TMDL assigns a wet weather WLA specifically to Caltrans, while the dry weather WLA is aggregated among all the permittees under a joint See text for definition of Level I and Level II (from URS Corporation 2010). TABLE 15 TYPICAL COSTS FOR LEVELS I AND II TREATMENT STORMWATER CONTROL MEASURES (SCMs) FOR APPLICATION TO BRIDGES

30 From King and Hagan (2011). TABLE 16 SUMMARY UNIT PLANNING LEVEL STORMWATER COST ESTIMATES PER IMPERVIOUS ACRE TREATED TABLE 17 LIFE-CYCLE COST OF BMP TECHNOLOGIES (1999 dollars)

31 NPDES MS4 permit. Within the Los Angeles River water- shed, Caltrans operates and maintains 275.5 road miles and a number of park-and-ride facilities and maintenance stations. Together, these assets comprise only 1.3% of the watershed area. The general TMDL approach includes implementation of (1) structural BMPs, (2) maintenance activities, (3) source control, and (4) special studies and monitoring plans. These four approaches are described individually here. Structural BMPs are widely implemented throughout the Los Angeles River watershed. As of October 2010, 115 struc- tural BMPs had been completed, including 101 GSRDs, three biofiltration swales, an EDB, a sand filter, and an infiltration device. An additional 72 structural BMPs were in progress, including 65 GSRDs, five Austin sand filters, and an infil- tration basin. Another 481 BMPs are in the planning phase, including 241 GSRDs, 98 media (sand) filters, 64 biofiltration swales, 41 biofiltration strips, 24 infiltration basins, 11 deten- tion basins, and one infiltration trench. Pollutant load reduc- tions are calculated based on the following assumptions: (1) the average drainage area for each BMP is 1 acre; (2) the annual rainfall is 14.78 inches, which is used to determine the WQV to be treated annually; and (3) BMP removal effi- ciencies are the same as those identified in the DOT’s BMP Retrofit Study (Caltrans 2004). Structural BMPs are selected based on a list of approved treatment BMPs identified in the Statewide Stormwater Management Plan (Caltrans 2003b). The list is continually expanded through Treatment BMP Technology Reports (e.g., Caltrans 2010b) that identify and evaluate new BMP technologies for potential use in the high- way environment. An additional 287 treatment BMPs are planned for implementation in the Los Angeles River Basin based on corridor studies that identify potential BMP oppor- tunity sites along the roadway corridor. Caltrans also implements maintenance activities to com- ply with the Los Angeles River metals TMDL. These include stenciling of storm drain inlets to help educate the public about stormwater runoff pollution. For the dry weather WLA, the DOT conducted a study to identify dry weather runoff from its facilities. In total, 61 instances of dry weather flows were observed in the Los Angeles River watershed, mostly in cases where commercial and residential properties outside of the DOT ROW were contributing dry weather run-on. Cal- trans is coordinating with municipalities to eliminate these sources. Most dry weather flows attributed to the DOT were the result of broken irrigation lines. Overall, Caltrans pro- duces little to no dry weather flow and is currently meeting the dry weather WLA for metals. Caltrans also participates in source control to address the TMDL. These include brake pad partnerships, roadside land- scape measures, annual element (a soil stabilization protocol), and enhanced street sweeping. The brake pad partnership is funded by Caltrans and others including municipal stormwater permittees, brake pad manufacturers, environmental groups, government agencies, and community members and designed to reduce copper in vehicle brake pads. Current brake pads con- tain up to 20% copper, but following state legislation signed in 2010 (SB 346), copper is being phased out in brake pads to no more than 5% by 2021, and no more than 0.5% by 2025. Effec- tive in 2014, the law also restricts cadmium, hexavalent chro- mium, lead, mercury, and asbestos, which will substantially reduce other metals entering the Los Angeles River. Roadside landscape measures include the use of vegetation for erosion control, which helps trap sediments and their attached metals at the source. Annual element, a soil stabilization protocol, is another practice to prevent erosion, typically using rock rip- rap, willow planting, ice plant planting, pavement, and mulch- ing. Enhanced street sweeping is currently being considered to reduce the amount of metals. However, the amount of metals removed during street sweeping is not known, but is likely to be highly variable based on site conditions. Finally, Caltrans has partnered with other watershed stake- holders to develop special studies and monitoring plans. The purpose of the special studies is to evaluate water quality objectives and ensure that the WLAs are appropriate for site- specific conditions. Special studies cover such topics as recal- culating metals criteria using new data, analyzing metal loads deposited through atmospheric deposition, and developing a water effects ratio [a ratio of the toxicity of metals in labora- tory dilution water to the toxicity of metals in site water (EPA 1994)] to revise the WLA targets. Results from these studies are submitted to the regulatory agency during the re-opening of the TMDL. For an example of a special study, see Larry Walker Associates (2008). A Coordinated Monitoring Plan was also developed by Caltrans and other stakeholders to col- lect data to evaluate the uncertainties and assumptions made during development of the TMDL, assess compliance with the WLAs, and evaluate potential management scenarios. The monitoring program has a three-tiered TMDL implementa- tion approach: long-term monitoring (Tier I), targeted moni- toring of tributaries with repeated exceedances (Tier II), and monitoring of source control efforts (Tier III). Data collected at ambient monitoring stations are used to gauge the effec- tiveness of TMDL implementation efforts. For more details of the Coordinated Monitoring Plan, see Los Angeles River Metals TMDL Technical Committee (2008). In summary, Caltrans implements four basic strategies to comply with the Los Angeles River metals TMDL: (1) struc- tural BMPs, (2) maintenance activities, (3) source control, and (4) special studies and monitoring plans. Using all of these approaches, Caltrans expects to meet the wet weather WLA by 2028. California—Marina Del Rey Harbor Toxics TMDL Caltrans is also a named stakeholder in a toxics TMDL for the Marina del Rey Harbor. Its implementation plan is docu- mented in city of Los Angeles et al. (2011). The plan was jointly developed with two other municipalities. The DOT’s

32 assets comprise only 1% of the watershed. The plan proposes a two-pronged approach including institutional BMPs (vehicle brake pad product replacement, enhancing street sweeping, education and outreach, catch basin cleaning, and downspout retrofits) and structural BMPs (mainly infiltration practices for housing developments). The watershed also has a bacteria TMDL, for which the county of Los Angeles installed three low flow diversions and five tree wells; and a trash TMDL, for which approximately 100 catch basin opening screen covers are to be installed by 2011 (it is not clear if they have been actually installed). Many of these BMPs also serve the dual purpose of reducing toxics. Although Caltrans does not play a major role in the TMDL implementation plan, the plan is a good example of a collaborative watershed-based approach in which BMPs have multi-pollutant benefits to address multiple TMDLs. Caltrans has many additional TMDL implementation strat- egies for a wide variety of pollutants as they are currently named in more than 60 TMDLs. It is outside the scope of this synthesis to address them all. However, the interested reader is referred to CDM (2007), an implementation plan for the Malibu Creek bacteria TMDL; a report by Sobelman et al. (n.d.) that documents trash management strategies for compliance with trash TMDLs in Ballona Creek and the Los Angeles River; and finally to the Caltrans interview sum- mary in Appendix B, which presents additional information on TMDL implementation strategies. Washington State For WSDOT, the implementation plan is written directly into its NPDES municipal stormwater permit, which lists all of the TMDLs and the associated “action items” for each along with implementation timelines. The action items represent require- ments above and beyond the permit obligations. WSDOT actively works with the state regulatory agency [State of Washington Department of Ecology (SWDE)] to ensure that the action items and timelines are appropriate and effective for meeting TMDL goals. The current NPDES permit, last modi- fied in March 2012, lists 26 TMDLs and numerous associ- ated action items (see Appendix 3 in SWDE 2012). Most of the TMDLs are for fecal coliform; however, other pollutants addressed include temperature, TSS, dissolved oxygen, pH, mercury, arsenic, polychlorinated biphenyls (PCBs), and chlorinated pesticides. TMDLs are grouped into two basic categories: (1) those TMDLs for which compliance with the action items constitutes compliance with the assigned WLA, and (2) those TMDLs for which compliance with the permit obligations that address TMDL listed pollutants constitutes compliance with the WLA. Some examples of action items from the current NPDES permit as described in SWDE (2012) are provided here: • For the Nisqually River tributaries fecal coliform and dissolved oxygen TMDL, WSDOT is required to install a pet waste station, maintain WSDOT controlled tide gates every other year, and participate in annual adap- tive management meetings. • For the Teanaway River temperature TMDL, WSDOT is required to maintain roads and roadside stormwater conveyance ditches to prevent entry of sediment into area waterways. • For the Walla Walla River watershed fecal coliform, PCBs, chlorinated pesticide, temperature, pH, and dis- solved oxygen TMDL, WSDOT is required to (1) re- route 97% of the traffic volume on US-12 to a plateau located well above the Walla Walla River, (2) imple- ment infiltration and/or dispersion BMPs to address the pollutants covered under this TMDL, and (3) follow the current Integrated Roadside Vegetation Management Plan (IRVM) (South Central Region, Area 4) (WSDOT 2012) within the Walla Walla TMDL boundary. The IRVM is designed to enhance roadside vegetation by providing stable, sustainable plant communities; reduce maintenance costs; and improve weed control. For more details on the IRVM, see WSDOT (2012). • For the Spokane River watershed dissolved oxygen TMDL, WSDOT is required to conduct a stormwater dis- charge inventory within its ROW and inside the NPDES coverage area, as well as identify phosphorus and ammo- nia sources. If phosphorus and ammonia sources are found, WSDOT will apply BMPs from their Stormwater Management Program Plan or SWMPP (see Appendix 7 of SWDE 2012) or perform remediation to eliminate the sources. A common theme running through all of the WSDOT action items for fecal coliform TMDLs is the implementation of a programmatic approach. This approach was developed by WSDOT to systematically address fecal coliform TMDLs; it includes activities such as fecal coliform source identification, inventory of highway discharge locations, illicit discharge detection and elimination, and identification of fecal coli- form maintenance issues within the TMDL boundary. A flow chart of the programmatic approach is provided on page 53 of the current NPDES permit (SWDE 2012). If the program- matic approach finds bacteria discharges within WSDOT’s ROW that are above natural background levels, WSDOT will implement BMPs from its SWMPP or perform remediation to remove the bacteria source. For run-on sources of bacte- ria from outside the ROW, WSDOT will notify SWDE and work with them, the local jurisdiction, and any other parties involved to resolve the issue. It can be noted, however, that WSDOT does not generally consider itself a source of bacte- ria (other than minor contributions from pet walking areas at rest stops and bird roosting areas under bridges), but rather a conveyance of bacteria (WSDOT 2011a). For a more detailed listing of WSDOT’s TMDL action items, see Appendix 3 of the current NPDES permit (SWDE 2012). WSDOT is also unique among state DOTs in that its High- way Runoff Manual (WSDOT 2011b) has specific TMDL

33 considerations for 21 of the most commonly used runoff treatment BMPs available for highway applications. These are broadly grouped into infiltration, dispersion, biofiltra- tion, wet pool, oil control, and phosphorus control BMPs. For example, one type of wet pool BMP, a constructed storm- water treatment wetland, is listed as the “preferred” BMP for reducing dissolved metals and TSS/turbidity, but is to be avoided for reducing fecal coliform. In general, infiltration BMPs (e.g., infiltration pond, infiltration vault, drywells, and permeable pavement) and dispersion BMPs (natural dis- persion and engineered dispersion) are considered the most desirable for TMDL situations because they are effective across a broad range of pollutants, including fecal coliform, nutrients, oil and grease, dissolved metals, TSS/turbidity, polycyclic aromatic hydrocarbons (PAHs), and several others. For more details, see WSDOT’s Highway Runoff Manual (WSDOT 2011b). Colorado—Straight Creek Sediment TMDL The Colorado DOT (CDOT) is a named stakeholder in two sediment TMDLs. One of the TMDLs is for Straight Creek in a pristine mountainous area west of Denver near I-70, where it passes through the Eisenhower Memorial Tunnel. The TMDL was developed in 2000 by the state regulatory agency [the Colorado Department of Public Health and the Environment (CDPH&E)]. It identifies two main sources of sediment: wash-off of applied traction sand on I-70 and ero- sion of the cut and fill slopes of the I-70 approach to the Tunnel (CDPH&E 2000). The load allocation compliance strategy for CDOT is written directly into the TMDL and consists of (1) revegetating at least 70% of the cut and fill slopes to 70% potential cover; (2) cleaning and maintaining of 12 existing sedimentation basins, a holding pond, and sedi- ment control structures on the I-70 roadway; and (3) remov- ing at least 25% of the traction sand applied yearly to the I-70 roadway within the TMDL boundaries (CDPH&E 2000). In addition, the TMDL study specifies several water qual- ity targets designed to protect aquatic life. These include (1) increasing the median particle substrate size, (2) decreas- ing the in-stream pool volume filled with fine sediment, (3) protecting the morphology of the stream, and (4) show- ing an improvement in the Brook Trout population. Sediment control activities have been ongoing in the Straight Creek watershed for decades and pre-date the devel- opment of the TMDL in 2000 (CDOT 2002). Therefore, it is difficult to distinguish between specific TMDL strategies and continuation of previous sediment clean-up efforts. Accord- ing to CDOT’s Sediment Control Action Plan developed in 2002, CDOT implemented a Straight Creek Erosion Control Project in 2000–01 (one of several erosion control projects dating back to 1979), which included seeding and mulch- ing of 107 acres on cut and fill slopes, planting 1,000 tree tublings, cleaning 7,466 tons of sand and sediment off the highway, and paving approximately 2,000 linear feet of ditches to reduce erosion and provide a surface for sweep- ing. The Sediment Control Action Plan also proposes several sediment mitigation strategies, all of which would fulfill the Straight Creek TMDL requirements. The most rigorous sce- nario includes implementation of an enhanced maintenance program as well as nonstructural controls (e.g., removal of sand deposits, semi-annual sweeping and ditch cleaning, cleanup of sediment basins and traps, and revegetation) and structural controls (e.g., basins and traps to capture sediment, paving of shoulder areas, and valley pan drains to control and route highway runoff). In addition, a Straight Creek Clean Up Committee that includes representatives from CDOT met eight times between 1998 and 2000 to satisfy issues on the TMDL goals, targets, and implementation, with meet- ings scheduled at least once annually beyond that (CDPH&E 2000). There has also been a significant change in CDOT’s winter road maintenance materials, with a trend away from the use of salt and sand in the 1990s toward sand/slicer mix- tures and liquid deicer salts in the 2000s (CDOT 2011). However, despite these efforts, a 2008 compliance evalu- ation of the Straight Creek Sediment TMDL by the U.S. For- est Service and the EPA concluded that Straight Creek was not meeting the water quality goals of the TMDL although many of the required sediment control practices had been completed (CDOT 2011). Specifically, as outlined in the report, the TMDL water quality targets related to median particle size and fisheries populations had either not been attained or evaluations of the data have not been conclusive. Further, the shift to sand/slicer mixture and liquid deicer salts has resulted in increased chloride concentrations and loads in Straight Creek. The current TMDL data collection and evalu- ation is underfunded and inconsistent, making it difficult to quantify improvements and assess if water quality goals are being met (CDOT 2011). The report makes several recom- mendations for achieving TMDL objectives: (1) a coherent monitoring and evaluation plan, (2) consistent annual fund- ing, and (3) nonparametric analysis to address high annual variability when seeking trends. CDOT’s own nonparamet- ric analysis of data from 1992–1998 versus 1999–2006 has shown some improvement in aquatic life support categories over time; however, this analysis was not accepted by the U.S. Forest Service and state regulators (Holly Huyck, personal communication, May 9, 2012). New York NYSDOT is a stakeholder in several TMDLs for nitrogen, phosphorus, and pathogens. The TMDLs were developed by the New York State Department of Environmental Con- servation (NYSDEC) and are referenced in the current SPDES General Permit for MS4s (available in Appendix C of NYSDOT 2012). Stormwater retrofit programs (an implemen- tation strategy for all TMDLs) must demonstrate pollutant reductions in accordance with the SPDES Permit or, in one case (the New York City East of Hudson River TMDL for

34 phosphorus, see Table 18), in accordance with the Croton Watershed Phase II Phosphorus TMDL Implementation Plan. Stormwater retrofit program plans are required to be submitted to NYSDEC to meet permit requirements. A brief summary of general TMDL implementation strategies for each TMDL is presented in Table 18. In addition, NYSDOT has developed two main policies for phosphorus TMDLs: (1) all new development projects must have erosion and sediment control plans in accordance with NYSDOT specifications, and (2) post-construction stormwater management practices must be consistent with the New York State Stormwater Management Design Man- ual, Chapter 10—Enhanced Phosphorus Removal Standards (CWP 2010). [For more details on NYSDOT’s stormwater management program to address TMDL requirements, see NYSDOT (2012), pp. 50–57]. Maryland—Chesapeake Bay TMDL MDSHA is working to meet the requirements of the Chesapeake Bay TMDL, issued December 29, 2010. The requirements are established in Maryland’s Watershed Implementation Plan for the Chesapeake Bay Total Maximum Daily Load (WIP I), issued December 3, 2010. SHA has land coverage in three sectors: (1) minor processed wastewater, (2) septic, and (3) regulated urban stormwater. MDSHA has coordinated with two state agencies respon- sible for establishing the TMDL [Maryland Department of the Environment (MDE) and the Department of Natural Resources] as well as the U.S. Fish and Wildlife Service, the U.S. Army Corps of Engineers, and the EPA. Because the vast majority of the state of Maryland is in the Chesa- peake Bay watershed, a statewide perspective is essential. As resources have permitted, MDSHA has also participated in workshops, webinars, and meetings with county and other local officials, as well as local watershed groups to identify partnering activities to achieve the TMDL No specific requirements have been imposed on MDSHA for non-MS4 areas. However, specific WLAs (see Table 19) have been issued by the MDE for MDSHA compliance with the Chesapeake Bay TMDL and are the MDSHA components of the overall limits of pollutants in regulated urban stormwater to meet water quality standards by 2025. The 2017 SHA target load is 60% of the reduction requirement based on the MDE 2009 baseline progress scenario. The impervious surfaces goal in the table specifies the proportion of impervious surface area that must be treated by BMPs in the Phase I and Phase II coun- ties by 2017. The WLAs are expressed as “delivered” (DEL) or “edge-of-stream” (EOS). The EOS load is the amount of pol- lutant that enters a stream near a pollutant source; DEL loads are the proportion of the EOS load that ultimately discharges to the Chesapeake Bay. The DEL load is generally lower than the EOS load because of losses of the pollutant (e.g., the result of deposition, removal by algae and plants, and denitrification) during transport by means of streams and rivers. Requirements for regulated urban stormwater represent the largest TMDL challenges for MDSHA. The department TMDL Watershed TMDL Pollutant TMDL Implementation Strategies New York City East of Hudson River Watershed Phosphorus Map the drainage system to help track suspected illicit discharges, develop a stormwater retrofit program that demonstrates phosphorus reductions, inspect and maintain the MS4 drainage features, remediate degradation sites where the MS4 creates a potential for adverse impacts to NYC’s drinking water supply1; implement a standardized turf management and procedures policy Greenwood Lake Watershed Phosphorus Develop a stormwater retrofit program that demonstrates phosphorus reductions; implement a standardized turf management practices and procedures policy Onondaga Lake Watershed Phosphorus Create a poster to help educate the public on ways to reduce phosphorus in the watershed2, develop a stormwater retrofit program that demonstrates phosphorus reductions; implement a standardized turf management practices and procedures policy Oyster Bay Watershed Pathogens Develop a stormwater retrofit program that demonstrates pathogen reductions Peconic Bay Watershed Pathogens Develop a stormwater retrofit program that demonstrates pathogen reductions 27 Long Island Shellfishing Impaired Embayments Pathogens Develop a stormwater retrofit program that demonstrates pathogen reductions Peconic Bay Watershed Nitrogen Develop a stormwater retrofit program that demonstrates nitrogen reductions; implement a standardized turf management practices and procedures policy 1NYSDOT worked with NYSDEC and the New York City Department of Environmental Protection (NYCDEP) to remediate these sites, most ly through stabilization efforts or redirection of drainage to reduce or eliminate direct stormwater discharges into water bodies. Examples of BMPs utilized included swales, check dams, re-grading, sediment basins, slope stabilization (include soil bioengineering), outlet protection, and the removal of paved or concrete swales. 2Poster is available here: https://www.dot.ny.gov/divisions/engineering/environmental-analysis/repository/PHOSPHORUS_POSTER.pdf. TABLE 18 TMDL IMPLEMENTATION STRATEGIES BY NEW YORK STATE DOT TO ADDRESS PHOSPHORUS, PATHOGEN, AND NITROGEN TMDLs

35 maintains MS4 permit coverage for the SHA roadway storm drain systems in nine Maryland MS4 Phase I counties and in two MS4 Phase II counties. The WIP only applies to these counties. The Maryland Assessment Scenario Tool, a Geo- graphic Information System (GIS)-based watershed plan- ning software, is used to forecast and evaluate MDSHA’s progress in achieving the overall limits of pollutants (TSS, phosphorus, and nitrogen) that can be discharged to the bay and still meet water quality standards. SHA has phased implementation of TMDL responsibili- ties over several milestones with the ultimate goal of achiev- ing the requirements by 2025. Achievement will be evaluated, in part, using MDE guidance: Accounting for Stormwater Waste Load Allocations and Impervious Acres Treated (draft, June 2011), otherwise referred to as the NPDES Account- ing Protocol. The milestones include implementation of a combination of structural and nonstructural BMPS by 2013, 2017, and 2025 for 10%, 60%, and 100% implementation, respectively. BMPs include bioretention/rain gardens, dry detention, extended detention, stormwater retrofits, catch basin clean- ing, urban filtering, urban infiltration, stream restoration, tree planting, vegetated open channels, wet ponds, wetlands, bio- swales, forest conservation, and outfall stabilization. Quan- titative goals for each measure for each milestone have been established. For structural BMPs, the measure is expressed in terms of the drainage area restored. For nonstructural BMPs, the measure is uniquely defined for each BMP. For exam- ple, goals for catch basin cleaning are expressed as pounds of material collected and goals for stream restoration are expressed as linear feet of stream restored. For the nonstructural BMPs, Maryland Assessment Sce- nario Tool incorporates equivalencies from the NPDES Accounting Protocol to convert BMP implementation quan- tities to impervious areas treated. For example, the account- ing protocol defines 1 ton of material collected from catch basin cleaning as equivalent to 0.4 impervious acre treated by structural BMPs. Similarly, it defines 100 linear feet of stream restoration as equivalent to 1 impervious acre treated by structural BMPs. MDSHA is also developing a custom application in a GIS environment that will track and generate reports for vari- ous parameters. Some potential reports may include TMDL 2-year milestone progress, MS4 database annual delivery, the MDSHA business plan data, bay expenditures data, and implementation status. The application will be housed in the SHA Enterprise GIS environment. Part of the rationale for developing a custom applica- tion is that MDSHA would like to address several technical discrepancies it believes exist in currently available track- ing and assessment tools. One of the areas being addressed is the method of accounting for nutrient credits for various nonstructural BMPs such as street sweeping and catch basin cleaning. The department is collaborating with MDE on these issues as it moves forward with its TMDL responsibilities. Internally, SHA has convened a workgroup/oversight committee to bring all design, construction and operations functions within the SHA together to discuss the require- ments, develop strategies, and address programmatic and funding gaps. Training was developed and given to all seven SHA district offices including design, construction, and maintenance managers, and TMDL liaisons have been des- ignated for each district to address local implementation and coordination. Budget needs for this TMDL are estimated to be $508.2 million for the fiscal years 2012 through 2016. However, prior to the start of fiscal 2012, only $78.2 million had already been planned for this period leaving a large gap in funding. MDSHA represents one of the transportation modes within the Maryland DOT, which estimates that the cost for all modes will be $1.5 billion. Maryland DOT recognizes TMDL implementation as a top priority, but also balances the safety of the traveling public in funding the implementation. TN (lb/yr) TP (lb/yr) TSS (lb/yr) Impervious Surfaces (I/II) SHA Phase I/II MS4 WLA WLA (DEL) 433,358 25,336 — — SHA Phase I/II MS4 WLA (EOS) 764,772 43,574 27,270,536 — 2017 SHA Target Load (EOS) (60% WLA Reduction) 825,095 50,611 30,782,560 30%/20% DEL = delivered; EOS = edge of stream. TABLE 19 SHA WLA AND IMPERVIOUS TREATMENT REQUIREMENTS FOR REGULATED URBAN STORMWATER SECTOR