Field Test of BMPs Using Granulated Ferric Oxide Media to Remove Dissolved Metals in Roadway Stormwater Runoff (2019)

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

15 Chapter 3 Field Testing of Ferric Oxide Media Filters: Monitoring Results 3.1 Woodlynn Avenue Ferric Oxide Filter 3.1.1 Monitoring Events A total of 14 storms were sampled from July, 2017 through September, 2018. The Woodlynn Avenue system is designed to capture runoff in a storage cell with a maximum capacity of 6,826 ft3. Water is stored in the cell and then slowly treated. Sampling was designed such that water was collected in individual one liter bottles at the inlet and outlet throughout the entire hydrograph of the storm event as well during the entire period of storage, treatment, and discharge. In this manner, the average water chemistry (event mean concentration) of water entering the cell could be matched with the average chemistry of the water leaving the cell. If the inlet or outlet samples only collected part of the event hydrograph, samples collected for that event were discarded. The hydrology and general chemistry data are provided in the following sections. The total and dissolved metals data for the first event (July 2017) were not used and are not shown as a result of metals contamination in the equipment blank for a majority of the metals parameters monitored. 3.1.2 Hydrology The ferric oxide treatment cell at Woodlynn Avenue is designed to capture and store runoff and then slowly treat the captured water. The small and completely impervious watershed typically fills the cell rapidly during a storm event while storage and treatment may occur for upwards of 20 hours. This approach minimizes the treatment cell footprint. At maximum storage, the surface area of the cell is 3,903 ft2 while the ferric oxide sand filter bed area is 528 ft2, and with a maximum head of 3 feet outlet flow control is necessary and is provided by a cap and orifice. It can be seen in Figure 3-1 that the orifice is moderating flow out of the cell as the flow rate does not change notably with water level in the cell.

16 Figure 3-1 Water level above the ferric oxide filter bed, storage, and cumulative outflow for an individual storm event on August 15, 2017 at the Woodlynn Avenue ferric oxide filter. A hydrologic summary is provided for each storm event with water sample collection (Table 3-1). The inflow pipes to the cell are short and have a steep slope and as a consequence the combined effect of rapidly changing velocity, tailwater, and other factors the total inflow was deemed unreliable and was calculated from watershed area and a local precipitation gauge maintained by the Ramsey Washington Metro Watershed District. Several checks were conducted to confirm the accuracy of this estimate including the relationship between event total rainfall and total outflow (r2 = 0.97) as well as water level, and calculated storage. Table 3-1 also provides calculated estimates of filtration through the ferric oxide filter bed as well as infiltration into the soils and surrounding subsoils. The treatment cell was designed to promote soil infiltration by extended detection of stormwater. With the exception of one outlier, the average filtration rate ranged from 3.99 to 11.62 inches per hour. The elevated filtration rate for the July 25, 2018 storm was likely due to the low precipitation volume and the likelihood that flow was unsaturated leading to high hydraulic conductivity. Infiltration rates to subsoils was similar to but lower than filtration rates. 0.0 0.5 1.0 1.5 2.0 2.5 0 1000 2000 3000 4000 5000 6000 8/13/2017 19:12 8/14/2017 0:00 8/14/2017 4:48 8/14/2017 9:36 8/14/2017 14:24 8/14/2017 19:12 Le ve l i n  Ce ll  (ft ) Vo lu m e  (ft 3 ) Cumulative Flow Out (ft3) Cell Storage Volume (ft3) Average Level Above Sand Filter (ft)

17 Table 3-1 Summary of hydrologic measurements for the Woodlynn Avenue ferric oxide treatment cell. Event Start  Date  Precipitation  (in)1  Total  Inflow (ft3)  Total  Outflow  (ft3)  Maximum  Water Depth  in Cell (ft)6  Average Water  Depth in Cell (ft)  Inundation  Duration (hr)2  Average  Filtration  Rate (in hr ‐1)3  Soil Infiltration  Rate (in hr‐1)4  7/17/17  1.26  8279  4301  2.28  1.02  14.0  6.98  3.90  8/3/17  0.35  2300  1053  1.48  0.62  6.0  3.99  2.46  8/13/17  0.95  6242  3793  2.39  1.20  13.5  6.39  2.67  8/16/17  0.49  3219  1813  0.94  0.43  7.5  5.49  2.07  9/25/17  0.61  4008  2237  1.46  0.49  8.5  5.98  2.34  10/2/17  1.33  8738  6302  2.65  1.30  19.5  7.35  1.91  5/29/18  1.38  9067  5021  1.80  1.08  13.5  8.45  4.21  6/16/18  3.68  24157  15465  (5)  (5)  (5)  (5)  (5)  7/1/18  0.79  5191  3710  1.78  0.93  9.5  8.88  2.07  7/12/18  2.16  14192  7793  2.92  1.60  16.8  10.57  6.51  7/25/18  0.22  1445  1146  0.51  0.12  0.5  52.11  5.87  8/3/18  1.08  7101  3452  1.59  0.91  6.8  11.62  7.12  8/24/18  1.11  7293  2953  2.08  1.32  11.0  6.10  6.06  9/20/18  2.60  (5)  (5)  3.01  2.35  (5)  (5)  (5)  1. Source is a rain gauge at the Ramsey Washington Metro Watershed District.  2. Inundation duration defined as the time that the water level above the sand filter was greater than zero. Depth of sand filter bed is  approximately 1.1 ft.  3. Filtration rate is calculated using a sand filter area of 528 ft2.   4. Soil infiltration rate = [(total inflow volume ‐ total outflow volume)/cell area inundated]/inundation duration.  5. Equipment error.  6. Water depth measured from the surface of the filter bed.   3.1.3 Monitoring Results and Treatment Performance: Metals Quantification of the dissolved metals removal capacity of the treatment system at Woodlynn Avenue (e.g., a swale-type best management practice with ferric oxide media) was the primary monitoring objective. A second object was the identification of factors (chemical and physical) that influence treatment performance if there was a clear difference in performance among the monitoring storm events. Total metals were also measured in this study. The dissolved metals fraction is operationally defined by filtration through a 0.45 µm polyethersulfone filter, and particulates removed by the filter may also include colloidal particles or other small particles with adsorbed metals. Since ferric oxide, which is often a hydroxide in the aqueous environment, has the capacity to bind a large range of particles and not just metals, it was hypothesized that the total metals removed performance of ferric oxide-sand filters in this study would exceed the performance of sand filters that did not incorporate media such as ferric oxide (Barrett, 2010). Plots of inflow and outflow concentrations for total and dissolved metals are provided in Figure 3-2 and the event mean concentration data are summarized in Table 3-2 and Table 3-3. For the plots, values reported as below detection were set equal to the detection limit. The plots do not include data that were identified as part of the data quality review as qualified and flagged as unusable. These data were most often removed as a result of a detected concentration in the equipment blank that was within five times

18 the sample reported value (i.e., the blank value multiplied by five). In some cases, data were removed if the dissolved value was greater than the total value. Dissolved metals removal was metal specific (Figure 3-2). Dissolved arsenic and zinc were consistently removed while chromium, lead and copper varied from removal, to no change between the inlet and the outlet, and for some storm events a net increase over the influent concentrations (i.e., release). For every storm event, dissolved nickel was released from the treatment cell. Dissolved iron was released for a majority of the events. In contrast, there was consistent total metals removal (Figure 3-3) for every storm event for arsenic, chromium, lead, iron and zinc. For these metals, it is notable that the treated discharge concentration was largely the same for each storm event regardless of the inflow concentration. For copper there was removal, no removal, and for two events it appeared that there was some release. Nickel was the one metal consistently higher in the outlet compared to the inlet. The lack of nickel removal is somewhat surprising as another laboratory column study (Genc- Fuhrman et al., 2007) using granular ferric hydroxide and iron coated sand demonstrated nickel as well as zinc, copper, arsenic, and chromium removal. Ionic strength and pH can have an effect on the surface charge of ferric oxide as well as nickel speciation (Ponthieu et al., 2006), however, the range of pH and ionic strength for runoff at Woodlynn should not change ferric oxide surface charge notably. The pH of the laboratory column study (Genc-Fuhrman et al., 2007) was 7.8 and nickel is primarily Ni+2 at pH 9 and below, hence, dissolved nickel speciation at Woodlynn should be similar to the study by Ponthieu. The release of dissolved iron from the treatment cell indicates that reducing conditions were established to some degree in the ferric oxide-sand filter bed and this may have liberated nickel. Carbon dioxide has been shown to bind to ferric oxide species called ferrihydrite resulting in a carbonate like species binding with iron (Hausner et al., 2009). This reaction may have the effect of displacing metals bound to ferric oxide. According to this study, when carbon dioxide pressure is reduced, carbon dioxide can desorb, and it may be hypothesized that binding capacity would be restored. Ferric oxide adsorbs metals selectively with greater affinity for lead, zinc, chromium, and zinc compared to nickel (Schultz et al., 1987). Overall, the dissolved metals removal performance is likely being affected by the design of this treatment system that includes extended detention and ponding above the ferric oxide-sand filter bed and also possibly the inclusion of organic planting soils above the filter bed. This is discussed in Chapter 3.1.5.

19 0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) b) Dissolved Cr removal release 0 1 2 3 4 0 1 2 3 4 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) f) Dissolved Cu removal release Figure 3-2 Event mean concentrations of dissolved metals measured at the inlet and outlet of the Woodlynn Avenue ferric oxide-sand filter. 0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) a) Dissolved As removal release 0.0 1.1 2.2 3.4 4.5 0.0 1.1 2.2 3.4 4.5 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) d) Dissolved Ni removal release 0 6 13 19 25 0 6 13 19 25 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) e) Dissolved Zn removal release 0.00 0.05 0.10 0.15 0.20 0.00 0.05 0.10 0.15 0.20 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) c) Dissolved Pb removal release 0 15 30 45 60 0 15 30 45 60 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) g) Dissolved Fe removal release

20 Figure 3-3 Event mean concentrations of total metals measured at the inlet and outlet of the Woodlynn Avenue ferric oxide-sand filter. 0.0 0.2 0.4 0.6 0.8 0.0 0.2 0.4 0.6 0.8 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) a) Total As removal release 0.0 1.5 3.0 4.5 6.0 0.0 1.5 3.0 4.5 6.0 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) b) Total Cr removal release 0 1 2 3 4 0 1 2 3 4 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) c) Total Pb removal release 0.0 1.5 3.0 4.5 6.0 0.0 1.5 3.0 4.5 6.0 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) d) Total Ni removal release 0.0 2.5 5.0 7.5 10.0 0.0 2.5 5.0 7.5 10.0 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) f) Total Cu removal releas e 0 15 30 45 60 0 15 30 45 60 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) e) Total Zn removal release 0 750 1500 2250 3000 0 750 1500 2250 3000 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) g) Total Fe removal release

21 Table 3-2 Dissolved metals event mean concentrations at the inlet and outlet of the Woodlynn Avenue ferric oxide-sand filter. Event  Start Date  As  Cr  Cu  Pb  Ni  Zn  Fe  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  8/3/17  0.081  <0.081  0.310  0.214  1.853  2.710  0.033  0.046  0.635  2.38  13.2  4.652  14.6  55.4  8/13/17  0.099  <0.081  0.152  0.148  1.168  1.490  0.019  0.028  0.336  1.33  6.95  <4.19  11.5  29.3  8/16/17  0.080  0.053  0.161  0.208  1.390  1.640  0.021  0.032  0.347  1.47  12.6  3.99  10.4  36.5  9/25/17  0.139  0.096  0.105  0.220  1.911  2.530  0.0202  0.0302  0.318  2.15  6.452  2.542  30.1  30.1  10/2/17  0.106  0.084  0.356  0.228  0.4822  1.682  0.028  0.047  <0.285  1.39  3.952  5.122  5.31  49.2  5/29/18  0.177  0.110  0.268  0.360  1.496  3.250  <0.073  <0.073  0.549  3.00  10.63  4.43  17.6  33.0  6/16/18  0.100  <0.081  0.283  0.628  0.939  2.885  <0.073  <0.073  0.338  3.90  4.81  6.684  7.77  17.9  7/1/18  0.119  <0.081  0.375  0.174  1.320  3.540  <0.073  <0.073  0.334  3.88  4.52  <1.61  9.69  15.3  7/12/18  0.142  <0.081  <0.628  <0.628  3.750  2.885  <0.073  <0.073  0.599  3.90  8.20  1.607  16.5  17.9  7/25/18  0.188  <0.081  0.534  0.240  3.960  6.340  <0.073  <0.073  1.140  4.23  10.2  3.27  11.0  9.6  8/3/18  0.153  <0.081  0.37  <0.202  3.006  2.740  0.1593  <0.073  0.570  2.91  15.6  <3.84  51.7  27.0  8/24/18  <0.081  <0.081  0.494  0.120  1.008  3.160  0.073  <0.073  0.241  2.16  4.80  1.00  8.56  25.4  9/20/18  0.074  <0.081  0.542  0.671  0.444  1.690  0.073  0.0537  0.160  1.76  5.14  2.21  7.39  42.6  Average5  0.118  0.062  0.328  0.271  1.85  2.91  0.049  0.041  0.439  2.65  9.03  2.67  15.5  29.9  Average6  0.118  0.062  0.328  0.271  1.75  2.81  0.043  0.039  0.439  2.65  8.52  3.16  15.5  29.9  Note: all values are ug L‐1  1. Average of the South Inlet and North Inlet when both samples are collected during a storm event.  2. Data not useable. Reported value is within 5 times the equipment blank.  3. South Inlet sample dissolved concentration greater than total. Sample not used to calculate average concentration in stormwater inflows.  4. Data not useable. Dissolved concentration greater than total concentration.  5. Average excluding qualified data. Average value calculated using regression on order in ProUCL5.1 when the parameter included values  below the reporting limit.  6. Average including qualified data. Average value calculated using regression on order in ProUCL5.1 when the parameter included values  below the reporting limit.  7. Data not useable. Reported value is within 5 times the equipment blank. Sample value was less than the equipment blank indicating the  sample may not have been contaminated. 

22 Table 3-3 Total metals event mean concentrations at the inlet and outlet of the Woodlynn Avenue ferric oxide-sand filter. Event  Start Date  As  Cr  Cu  Pb  Ni  Zn  Fe  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  8/3/17  0.169  <0.081  0.822  0.305  3.45  3.04  0.527  0.105  1.23  2.44  27.5  5.03  378  146  8/13/17  0.131  0.094  0.582  0.333  1.89  1.80  0.292  0.092  0.61  1.60  14.2  4.19  232  136  8/16/17  0.118  0.058  0.769  0.322  2.34  1.86  0.288  0.082  0.80  1.66  23.4  4.37  200  120  9/25/17  0.303  0.096  1.87  0.333  5.06  2.68  1.21  0.092  1.93  2.58  36.2  23.4  1140  119  10/2/17  0.353  0.098  2.49  0.369  4.202  2.022  1.42  0.130  2.26  1.73  42.1  5.422  1400  162  5/29/18  0.730  0.151  4.81  0.633  8.66  3.71  3.77  0.204  3.74  3.33  58.3  4.14  2680  331  6/16/18  0.450  0.081  4.54  0.628  6.38  3.16  2.68  0.098  2.95  4.35  52.1  3.14  1960  139  7/1/18  0.152  <0.081  0.735  0.231  2.13  3.93  0.440  0.157  0.64  3.86  13.0  2.17  236  116  7/12/18  0.229  <0.081  0.913  <0.628  5.17  3.16  0.640  0.098  1.00  4.35  17.5  3.063  413  139  7/25/18  0.246  <0.081  1.18  0.314  5.01  4.41  0.736  <0.073  1.64  4.25  27.8  1.90  374  49  8/3/18  0.247  <0.081  1.68  0.521  4.58  3.07  0.923  0.127  1.24  3.17  31.0  <3.84  569  106  8/24/18  <0.081  <0.081  0.551  0.196  1.49  2.91  0.272  <0.073  0.37  2.98  8.22  1.70  146  81  9/20/18  0.133  <0.081  1.18  0.237  1.84  1.88  0.602  0.090  0.68  2.19  14.3  2.19  339  145  Average4  0.252  0.071  1.70  0.365  4.00  2.97  1.06  0.105  1.47  2.96  28.1  4.85  774  138  Average5  0.252  0.071  1.70  0.365  4.02  2.89  1.06  0.105  1.47  2.96  28.1  4.77  774  138  1. Average of the South Inlet and North Inlet when both samples collected during a storm event.  2. Data not useable. Reported value is within 5 times the equipment blank.  3. Data not useable. Reported value is within 5 times the equipment blank. Sample value was less than the equipment blank indicating the  sample may not have been contaminated.  4. Average excluding qualified data. Average value calculated using regression on order in ProUCL5.1 for parameters with below reporting  limit data.  5. Average including qualified data. Average value calculated using regression on order in ProUCL5.1 for parameters with below reporting  limit data.   3.1.4 Monitoring Results: General Chemistry In addition to metals, several parameters were monitored that had the potential to affect performance or may provide indications of function (Table 3-4). The intent was to monitor pH continuously in the treatment bed, however, the in-situ pH meter fouled and measurements were conducted in the laboratory after August, 16, 2017. Dissolved oxygen (light dissolved oxygen probe) was measured in-situ in the ferric oxide-sand filter bed during the entire monitoring period.

23 Table 3-4 General parameter event mean concentrations at the inlet and outlet of the Woodlynn Avenue ferric oxide-sand filter. Event  Start  Date  Total Alkalinity  (mg L‐1 as  CaCO3)  DOC           (mg L‐1)  Cl              (mg L‐1)  Hardness  (mg  L‐1 as CaCO3)  TSS            (mg L‐1)  VSS            (mg L‐1)  Specific  Conductance  (µs cm‐1)  Average  pH (s.u.)  Average  DO In‐ Situ (mg  L‐1) Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  7/17/17  <10.0  33.5  5.2  7.0  <1.0  2.0  25  35  40.6  2.0  12  1.5  (3)  (3)  (2)  5.6  8/3/17  10.8  31.0  7.2  6.0  1.3  <1.0  17  26  16.4  1.8  6.4  1.4  30.2  57.0  (2)  7.6  8/13/17  <10.0  33.5  3.5  2.0  1.2  <1.0  17  21  23.1  3.1  4.3  0.6  26.5  48.0  (2)  7.5  8/16/17  11.0  24.0  3.7  3.3  1.2  <1.0  12  24  13.0  1.9  6.2  1.8  25.5  50.0  (2)  7.6  9/25/17  11.5  41.0  3.6  4.2  <1.0  <1.0  25  30  48.8  1.4  10  1.2  24.0  61.0  7.6  7.0  10/2/17  <10.0  28.0  1.2  2.1  1.0  2.3  35  60  317  <0.6  23  <0.6  20.0  67.0  7.1  6.9  5/29/18  17.8  47.0  5.3  5.6  2.5  2.3  48  30  70.4  2.6  16  1.2  70.5  115  7.5  3.3  6/16/18  14.0  31.0  4.8  3.9  1.2  2.0  64  30  86.0  1.7  21  <0.6  35.2  68.0  6.8  (3)  7/1/18  13.0  62.0  4.1  6.2  <1.0  1.3  17  30  6.8  2.0  3.8  2.0  29.2  76.5  6.1  6.9  7/12/18  10.0  36.0  7.3  5.8  1.6  1.3  25  32  14.0  3.3  5.8  2.0  31.5  76.7  6.8  5.3  7/25/18  17.0  44.0  14  11  3.1  4.2  24  32  25.0  1.4  10  1.2  56.0  133  7.2  7.1  8/3/18  21.4  36.0  5.2  4.8  <1.0  1.1  19  30  33.0  1.5  12  <0.6  35.9  80.1  7.5  6.3  8/24/18  <10.0  23.0  3.8  6.3  <1.0  <1.0  5  30  5.9  7.3  3.2  6.5  33.0  81.0  6.9  7.1  9/20/18  8.68  17.5  1.6  2.3  <1.0  <1.0  34  30  14.0  0.8  7.7  1.6  19.6  40  6.4  8.6  Average4  11.5  34.8  5.0  5.0  0.99  1.2  26.3  31.4  51.0  2.2  10  1.1  33.6  73.4  7.0  6.7  1. Flow weighted average of the south and north inlets when both samples collected during a storm event.  2. Data not recorded due to pH probe failure.  3. Data not recorded.  4. Average value calculated using regression on order in ProUCL5.1 for parameters with below reporting limit data.  Alkalinity was greater at the outlet compared to the inlet, indicating that alkalinity is generated in the ferric oxide-sand bed. The parent material of the ferric oxide material is elemental (zero valent) iron which rusts and transforms to ferric oxide. This rusting process creates alkalinity. Alkalinity generation also appears to buffer the pH as the average pH across all of the storm events was 7.0. The increase in alkalinity at the outlet also explains the increase in specific conductance, however, specific conductance was overall lower than may be expected for surface water runoff. Chloride and hardness were largely unchanged. The ferric oxide-sand filter was very effective at removing total suspended solids (96 percent removal) and volatile suspended solids (89 percent removal). It appears that total suspended solids removal may be enhanced for sand filters with ferric oxide compared to standard sand filters. Barrett, 2010, reported on six Austin sand filters with total suspended solids removal averaging 86 percent with a range of 74 to 95 percent. Similarly, total suspended solids removal by Delaware sand filters was 79 and 83 percent (Watershed Protection Techniques, 1994). It is also notable that dissolved organic carbon was not removed by the Woodlynn Avenue treatment cell. This is in contrast to the Highway 36/61 site (see Chapter 3.2.4). Iron has an affinity for dissolved organic carbon and lack of removal may be indicator of the condition of ferric oxide in the filter bed. Also, it seems possible that dissolved organic carbon is assisting with the export of dissolved metals (particularly nickel) by acting as a carrier.

24 3.1.5 Monitoring Results: Chemistry Measurements for a Storm Hydrograph 3.1.6 Dissolved Oxygen The Achilles heel of iron-based media is the elevated solubility of iron in the reduced state (Fe (II)) and the limited affinity for metals and other constituents such as phosphate once iron is reduced (Stumm and Lee, 1961, Fuchs et al., 2018). A significant fraction of oxidized iron is solid, immobile, and binds a range of stormwater constituents. Stormwater that has dissolved and particular organics exert biological oxygen demand which can be significant. Figure 3-4 and Figure 3-5 show oxygen fluctuation in the Woodlynn Avenue treatment cell ferric oxide-sand bed during and between storm events. Once the cell fills with water, the loss of oxygen is rapid and at a constant rate. This indicates that the rate of oxygen consumption is a function of ponded stormwater constituent decay rather than the loss of oxygen due to oxygen consumption in the filter bed. Since outflow is constant, if the filter bed were responsible for oxygen loss, there would be consistent but lower oxygen at the outlet rather than a continuous oxygen decline. Hence, it can be seen that the longer the ponding time, the lower the oxygen and the greater the potential for iron reduction. Given that organic carbon decay (e.g., BOD) and oxygen consumption occur largely linearly during individual storm events, oxygen loss during ponding can be quantified and hence predicted from stormwater chemistry, hydrology, and water temperature. These parameters can be collected at proposed ferric oxide filters. Ultimate biological oxygen demand and decay rates were estimated for selected storms at the Woodlynn Avenue ferric oxide filter after the end of the storm event, when the detention storage was at maximum, and oxygen decay was linear. Using stormwater chemistry (estimated ultimate biological oxygen demand), water temperature, and ponding time, decay rates were calculated (Table 3-5). A similar calculation could be conducted by collecting storm event samples that are then sent to the laboratory and analyzed for ultimate biological oxygen demand. Ultimate biological oxygen demand was not available at Woodlynn Avenue and was estimated from dissolved organic carbon and volatile suspended solids. The decay rate was then calculated for each storm event as a function of oxygen consumed during a period of record, time, and water temperature. First order decay rates ranged from 0.02 to 0.10 hr-1 and averaged 0.05 hr-1. Figure 3-4 Change in dissolved oxygen in the Woodlynn Avenue ferric oxide-sand filter bed during the (1) 2017 and (b) 2018 monitoring periods. 0 2 4 6 8 10 12 14 05/10/18 06/09/18 07/09/18 08/08/18 09/07/18 10/07/18 DO  (m g  L‐1 ) b) 0 2 4 6 8 10 12 14 06/24/17 07/24/17 08/23/17 09/22/17 10/22/17 11/21/17 DO  (m g  L‐1 ) a)

25 Figure 3-5 Dissolved oxygen in ponded water at Woodlynn Avenue during two storm events in 2018. Table 3-5 Inputs used to estimate UBOD decay rates at the Woodlynn Avenue ferric oxide- sand filter. Event  Start  Date  DOC  (mg L‐1)  VSS    (mg L‐1)  TOC     (mg L‐1)  Ultimate  BOD       (mg O2 L‐1)  Average  Temperature  (oC)  Oxygen  Consumed  (mg O2 L‐1)  Average  Oxygen  Consumption  Rate            (mg O2 L‐1 hr‐1)  Time  (hr)  K (hr‐1)  7/17/17  5.2  12.0  8.8  23.7  24.0  6.29  0.54  11.7  0.03  8/3/17  7.2  6.4  9.1  24.6  19.1  1.75  0.41  4.3  0.02  8/13/17  3.5  4.3  4.8  12.9  17.7  5.05  0.41  12.2  0.06  8/16/17  3.7  6.2  5.6  15.0  21.4  0.92  0.33  2.8  0.03  9/25/17  1.2  22.7  8.0  21.7  18.1  4.80  0.74  6.5  0.05  10/2/17  4.8  12.6  8.5  23.0  23.3  3.38  0.54  6.2  0.03  5/30/18  4.8  12.6  8.5  23.0  23.3  3.38  0.54  6.2  0.03  7/2/18  4.1  3.8  5.2  14.1  22.7  6.08  0.90  6.8  0.09  7/13/18  7.3  5.8  9.0  24.4  25.0  3.30  0.66  5.0  0.03  7/26/18  14.0  9.9  17.0  45.8  28.7  2.30  4.60  0.5  0.09  8/6/18  5.1  10.0  8.1  21.9  22.1  6.04  0.97  6.3  0.06  8/27/18  3.8  3.2  4.8  12.9  19.1  4.99  0.71  7.0  0.10  Notes:                    TOC = POC + DOC  Particulate Organic Carbon (POC) = 0.3 * VSS (Koch et al., 2000)  Ultimate BOD (mg/L) estimated as TOC (mg L‐1)*2.7 (see Thomann and Muller, 1987)  Standard Temperature (oC) = 25  K (UBOD decay coefficient) normalized to 25o C using 1.04^(T‐25)  Dissolved Oxygen Consumption (mg L‐1 hr‐1) = UBOD*(1‐exp(‐K*t))*1.04^(T‐25)  The 8/24/18 and 9/20/18 storm event not used in this analysis because of variable oxygen concentrations.  0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 0 1 2 3 4 5 6 7 8 9 21:36 0:00 2:24 4:48 7:12 9:36 12:00 W at er  D ep th  (f t) DO  (m g  L‐1 ) Hour  a) May 30, 2018 dissolved oxygen 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0 1 2 3 4 5 6 7 8 9 10 2:24 4:48 7:12 9:36 12:00 14:24 16:48 W at er  D ep th  (f t) Do  (m g  L‐1 ) Hour b) June 26, 2018 dissolved oxygen

26 3.1.7 Multi-Sample Event For the September 24, 2018 storm event, the autosampler flow pacing was set to collect individual one liter samples throughout the entire storm event. Samples were composited into 7 or 8 groups and submitted for chemical analysis. Hence, concentrations of metals and general parameters would be traceable throughout the event. It was hypothesized that treatment efficiency may change throughout the course of the storm event with first flush effects, residence time, and other factors such as variable pH and dissolved oxygen. A first flush effect is clearly visible in Figure 3-6 with total copper, nickel, zinc, and iron at higher concentrations at the start of the storm event. Total and volatile suspended solids were also higher at the start of the event. While the first flush effect was evident for particulate metals, dissolved metals were not higher in the inlet at the start of the storm event. Dissolved copper, nickel, and iron were higher at the outlet at the start of the storm event despite the lack of first flush effects for the dissolved form of these metals. There was a net release of dissolved copper, iron, and nickel (e.g., concentrations were higher at the outlet) during this event. Release was greatest at the start of the event and this suggests that metals release was not necessarily a function of the chemistry in the ferric oxide-sand filter bed during the storm event. Total volatile suspended solids removal declined as the event progressed, however, this may have been a reflection of inlet concentration. Dissolved organic carbon was higher at the outlet throughout the storm event. It is possible that dissolved natural organic matter acts as a metals carrier and is partly responsible for export of metals at the Woodlynn Avenue ferric oxide-sand filter.

27 Figure 3-6 Change in metals and general parameter concentrations during a September 24, 2018 storm event at the Woodlynn Avenue ferric oxide-sand filter. 0.0 0.8 1.6 2.4 3.2 4.0 9/24 9:07 9/24 9:36 9/24 10:04 9/24 10:33 9/24 11:02 Co pp er  (u g  L‐1 ) Outlet: Dissolved Inlet: Dissolved Outlet: Total Inlet: Total (a) 0 200 400 600 800 1,000 9/24 9:07 9/24 9:36 9/24 10:04 9/24 10:33 9/24 11:02 Iro n  (u g  L‐1 ) Outlet: Dissolved Inlet: Dissolved Outlet: Total Inlet: Total (b) 0.0 0.6 1.2 1.8 2.4 3.0 9/24 9:07 9/24 9:36 9/24 10:04 9/24 10:33 9/24 11:02 Ni ck el  (u g  L‐1 ) Outlet: Dissolved Inlet: Dissolved Outlet: Total Inlet: Total (c) 0 8 16 24 32 40 9/24 9:07 9/24 9:36 9/24 10:04 9/24 10:33 9/24 11:02 Zi nc  (u g  L‐1 ) Outlet: Dissolved Inlet: Dissolved Outlet: Total Inlet: Total (d) 0 7 14 21 28 35 9/24 9:07 9/24 9:36 9/24 10:04 9/24 10:33 9/24 11:02 Co nc en tr at io n  (m g  L‐1 ) Inlet: Volatile Suspended Solids Inlet: Total Alkalinity Outlet: Volatile Suspended Solids Outlet: Total Alkalinity (e) 0 1 2 3 4 5 9/24 9:07 9/24 9:36 9/24 10:04 9/24 10:33 9/24 11:02 Di ss ol ve d  Or ga ni c C ar bo n  (m g  L‐1 ) Inlet Outlet (f)

28 3.2 Highway 36/61 Ferric Oxide Filter 3.2.1 Monitoring Events A total of 13 storms were sampled from August, 2017 through October, 2018. The Highway 36/61 system differs from the Woodlynn Avenue system as the sand bed is notably larger with a surface area of 22,800 square feet and dead storage of 0.6 feet. If water ponds more than 0.6 feet additional inflows are bypassed. Rarely was water ponded above the ferric oxide-sand filter bed. The sampling design included water collection in individual one liter bottles at two inlets (see Figure 3-7) and one outlet throughout the entire hydrograph of the storm event. In this manner, the average water chemistry (event mean concentration) of water entering the cell could be matched with the average chemistry of the water leaving the cell. If the inlet or outlet samples were only collected for part of the event hydrograph, samples collected for that event were discarded. Hydrology and general chemistry data are provided in the following sections. 3.2.2 Hydrology Water flow was measured at three locations: (1) HWY Inlet East, (2) HWY Inlet West, and (3) HWY Outlet and water levels were recorded simultaneously using pressure transducers placed a five locations across the cell (see Figure 2-4 for monitoring locations). For most storm events, stormwater inflow rates were not high enough to result in ponding above the filter bed and for those events with ponding above the bed duration was short (Figure 3-7). Despite the large surface area of the filter bed, stormwater flows appeared to spread over the entire surface with no evidence of short circuiting (Figure 3-7b). The low hydraulic conductivity of the ferric oxide-sand filter bed promoted even distribution and treatment across most of the treatment cell filter bed. Components of the water balance as well as calculated filtration rates (average and peak) and losses to pore spaces or infiltration to subsurface soils are provided in Table 3-6. Losses to pore spaces and infiltration to subsoils (this may also be evaporative losses) were lumped together. A plot of inflows to outflows (Figure 3-8) shows good agreement of these two measurements but also suggests that about 1,792 ft3 of inflows were lost for each storm event and if inflows are less than 1,792 ft3 there may be little to no outflow. Assuming a porosity of 0.45 for the ferric oxide-filter bed, this volume is approximately 14 percent of the bed pore space. Examination of the water balance demonstrates that consistent and accurate flow measurements were taken throughout the course of this study, most importantly however, these measurements have value in that in-situ hydraulic conductivity could be calculated.

29 Figure 3-7 Water level (a) within and above the ferric oxide sand filter bed and (b) for the individual pressure transducers for a single storm event on May 29 and 30. Table 3-6 Summary of hydrologic measurements for the Highway 36/61 ferric oxide treatment cell. Event  Start Date  Event  Rain  Total (in)  Flows (ft3)  In‐Situ Ferric Oxide  Filter Measurements  Filtration Though Cell  Losses to  Pore Space  or Infiltration  (ft3)  Direct  Rainfall   East  Inflow   West  Inflow   Total  Inflow   Outlet   Average  Level  (ft)4  Inundation  Duration  (hr)2  Average  Rate (ft3  hr‐1 )  Average  Rate (in  hr‐1)3  Peak  Rate (in  hr‐1)  8/16/17  0.55  1045  15645  338  17028  15925  0.32  15.50  1027  0.53  3.75  1103  9/18/17  0.36  684  6566  160  7410  3542  0.10  13.00  272  0.14  0.68  3868  9/24/17  0.79  1501  13838  317  15656  14023  0.27  14.50  967  0.50  2.60  1633  10/6/17  0.45  855  9273  295  10423  8845  0.29  9.50  931  0.48  1.73  1578  10/7/17  0.22  418  5128  57  5603  4956  0.16  10.75  461  0.24  0.61  647  10/14/17  0.46  874  7203  182  8259  6590  0.17  18.75  351  0.18  1.31  1669  10/21/17  0.37  703  4768  119  5590  2784  0.10  10.75  259  0.13  0.71  2806  5/29/18  1.32  2508  12607  369  15484  15396  0.25  16.00  962  0.51  2.18  88  6/16/18  1.21  2299  12647  401  15347  13118  (1)  (1)  (1)  (1)  (1)  2230  7/1/18  0.92  1748  10006  298  12052  10829  0.19  13.50  802  0.42  2.08  1223  9/20/18  0.38  722  2163  185  3070  2622  0.05  5.00  524  0.28  0.15  449  10/8/18  1.46  2774  15107  515  18396  16388  0.22  32.25  508  0.3  1.9  2008  10/26/18  0.35  665  1897  84  2646  900  0.03  29.25  31  0.02  0.2  1747  1. Data not collected.  2. Inundation duration is defined as the time in which water level is greater than zero in the cell during a storm event.  3. Filtration rate based upon a cell surface area of 22,800 ft2.  4. Average level with zero being the bottom of the ferric oxide filter.  0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 5/10/18 6/9/18 7/9/18 8/8/18 9/7/18 10/7/18 11/6/18 Av er ag e  W at er  Le ve l ( ft) Filter Bed  Surface 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 16:48 19:12 21:36 0:00 2:24 4:48 7:12 9:36 W at er  Le ve l ( ft) L1 L2 L3 L5 L6 b)a)

30 Figure 3-8 Comparison of inflow and outflows from the Highway 36/61 ferric oxide treatment cell. 3.2.2.1 Hydraulic Conductivity The consistent and accurate hydrology data, including outflow and water level in the filter bed, enabled the estimation of in-situ hydraulic conductivity from storm event peak water level (average of the five sensors) and corresponding peak outflow rate, both of which were measured every 15 minutes and corresponded without significant lag (e.g., peak flow corresponded to peak level). From these measurements, hydraulic conductivity was calculated for 12 of the 13 storm evens measured. For each event the peak water level and peak outflow rate were matched and used to calculate the hydraulic conductivity for that event (Table 3-7). Darcy’s law was used to estimate hydraulic conductivity: 𝑄 𝐾 ∗ 𝐴 ∗ 𝐻𝐿 Eq. 1 𝐹𝑖𝑙𝑡𝑟𝑎𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒 𝑄𝐴 Eq. 2 𝐾 𝐹𝑖𝑙𝑡𝑟𝑎𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒 𝐻 Eq. 3 Where: Q = peak flow (ft3 s-1) K= hydraulic conductivity (ft s-1) A = filter bed surface area (ft2) H = fractional height of peak water level above sand filter L = filter bed depth (ft) 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 Ou tfl ow  (f t3 ) Inflow (ft3) 1,792

31 The fractional height of the peak water level above the sand filter is one if the water level does not rise above the filter bed, and if water is above the bed the fractional height is total water level measured from the filter bed bottom divided by the filter bed depth. It should be noted that there was a clear relationship between peak water level and peak flow (Figure 3-9) up to a water depth of approximately 1.2 feet, and above that level there was a jump in flow relative to water level. However, the hydraulic conductivity and hence the flow rate should have been constant for water levels below the filter bed. This suggests that the underdrain system, which consisted of slits cut into drainpipe, may have been restricting flows more than the sand for most of these storm events. The calculated hydraulic conductivity ranged from 5.22 x 10-5 in s-1 to 8.79 x 10-4 in s-1 with an average of 5.41 x 10-4 in s-1 or 1.95 in hr-1. The average hydraulic conductivity is similar to sandy loam hydraulic conductivity of 3.35 x 10-4 in s-1 (Wanielista et al., 1997). The average hydraulic conductivity may provide a starting point for sizing a ferric oxide-sand filter however laboratory testing of sand-ferric oxide mixes expected to be used for a particular site is recommended. The sand gradation used for the Highway 36/61 filter was, as particle size and percent finer: 9.5 mm, 100 percent; 4.75 mm, 90 to 100 percent; 2.0 mm, 45 to 90 percent; 0.425 mm, 5 to 35 percent; and 75 µm, 0 to 3 percent. The zero valent iron gradation used for this filter was, as sieve size and percent finer: #4, 100 percent; #8, 95 to 100 percent; #16, 75 to 90 percent; #30, 25 to 45 percent; and #50, 0 to 10 percent. Table 3-7 Calculation of hydraulic conductivity for the Highway 36/61 ferric oxide-sand filter. Event Start Date1  Peak Water  Level (ft)  Peak Flow  (ft3 s‐1)  Peak Filtration  Rate (in hr‐1)  H L‐1   K (in s‐1)  8/16/17  1.49  1.98  3.75  1.24  8.40E‐04  9/18/17  0.36  0.36  0.7  1.00  1.88E‐04  9/24/17  1.23  1.37  2.6  1.03  7.03E‐04  10/6/17  0.94  0.91  1.7  1.00  4.80E‐04  10/14/17  0.67  0.69  1.3  1.00  3.63E‐04  10/21/17  0.38  0.38  0.71  1.00  8.27E‐04  5/29/18  1.11  1.15  2.2  1.00  2.39E‐04  7/1/18  1.05  1.10  2.1  1.00  5.78E‐04  9/20/18  1.55  2.03  3.85  1.29  8.27E‐04  9/20/20182  1.84  2.56  4.85  1.53  8.79E‐04  10/8/18  1.02  0.98  1.86  1.00  5.16E‐04  10/26/18  0.09  0.10  0.19  1.00  5.22E‐05  Notes:  Total Sand Filter Area (ft2)  22800    Sand Filter Depth (ft)  1.2   

32 Figure 3-9 Peak water level and peak flow measured at Highway 36/61. 3.2.3 Monitoring Results and Treatment Performance: Metals The primary objective of monitoring at Highway 36/61 was to identify the capacity of this type of treatment system (e.g., a vault-type ferric oxide-sand filter) to remove dissolved metals from highway runoff and identify chemical and physical factors that may affect treatment performance. Quantification of metals as “dissolved” is operationally defined by filtration through a 0.45 µm filter. Particulates removed by the filter may also include colloidal particles or other small particles with adsorbed metals. Since ferric oxide, which is often a hydroxide in the aqueous environment, has the capacity to bind a large range of particles and not just metals, it was hypothesized that the total metals removed performance of ferric oxide-sand filters in this study would exceed the performance of sand filters that did not incorporate media such as ferric oxide (Barrett, 2010). It should be noted, however, that there is a pre-treatment pond directly upstream of the primary inflow to the ferric oxide-sand filter and this reduced loading by solids and metals attached to particulates. Plots of inflow and outflow concentrations for dissolved and total metals are provided in Figure 3-10 and Figure 3-11 and the event mean concentration data are summarized in Table 3-8 and Table 3-9. For the plots, results reported as below detection were set equal to the detection limit. The plots do not include data that were identified as qualified and flagged as unusable as part of the data quality review. These data were most often removed as a result of a detected concentration in the equipment blank that was within five times the sample reported value (i.e., the blank value multiplied by five). In some cases, data were removed if the dissolved value was greater than the total value. Dissolved metals removal was metal specific (Figure 3-10). Dissolved arsenic, chromium, lead, iron, and zinc were consistently removed while copper was removed less consistently and for a majority of the storm events dissolved copper at the outflow was similar to the concentration of dissolved copper at the 0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.5 1.0 1.5 2.0 Pe ak  F lo w  (f t3 ) Peak Water Level (ft) Filter Bed Surface

33 Table 3-9. For the plots, results reported as below detection were set equal to the detection limit. The plots do not include data that were identified as qualified and flagged as unusable as part of the data quality review. These data were most often removed as a result of a detected concentration in the equipment blank that was within five times the sample reported value (i.e., the blank value multiplied by five). In some cases, data were removed if the dissolved value was greater than the total value. Dissolved metals removal was metal specific (Figure 3-10). Dissolved arsenic, chromium, lead, iron, and zinc were consistently removed while copper was removed less consistently and for a majority of the storm events dissolved copper at the outflow was similar to the concentration of dissolved copper at the inlet. Dissolved nickel appeared to pass through the ferric oxide-sand filter untreated as the inflow and outflow concentrations were largely equal. For arsenic, chromium, zinc, lead, and iron, outlet concentrations were not affected by changes in inlet concentrations. Unlike the Woodlynn Avenue ferric oxide-sand filter, dissolved metals release did not occur except for a few events and nickel. Total metals removal occurred for all metals (Figure 3-11), however, total arsenic, chromium, lead, iron, and zinc were consistently removed while copper was removed less consistently and for a few storm events total copper at the outflow was similar to the concentration of total copper at the inlet. Since dissolved nickel was not removed by the treatment system, there was limited total nickel removal. For arsenic, chromium, zinc, lead, iron, and to a lesser extent copper, outlet concentrations were not affected by changes in inlet concentrations.

34 Figure 3-10 Event mean concentrations of dissolved metals measured at the inlet and outlet of the Highway 36/61 ferric oxide filter. 0.0 0.3 0.6 0.9 1.2 0 0.3 0.6 0.9 1.2 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) a) Dissolved As removal release 0.0 1.5 3.0 4.5 6.0 0.0 1.5 3.0 4.5 6.0 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) b) Dissolved Cr removal release 0.0 2.5 5.0 7.5 10.0 0.0 2.5 5.0 7.5 10.0 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) c) Dissolved Cu removal release 0.0 0.6 1.3 1.9 2.5 0.0 0.6 1.3 1.9 2.5 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) d) Dissolved Ni removal release 0 20 40 60 80 0 20 40 60 80 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) e) Dissolved Zn removal release 0.00 0.06 0.13 0.19 0.25 0.00 0.06 0.13 0.19 0.25 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) f) Dissolved Pb removal release 0 200 400 600 800 0 200 400 600 800 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) g) Dissolved Fe removal release

35 Figure 3-11 Event mean concentrations of total metals measured at the inlet and outlet of the Highway 36/61 ferric oxide filter. 0.0 0.4 0.8 1.2 1.6 0.0 0.4 0.8 1.2 1.6 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) a) Total As removal release 0 2 4 6 8 0 2 4 6 8 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) b) Total Cr removal release 0 5 10 15 20 0 5 10 15 20 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) c) Total Cu removal release 0 1 2 3 4 0 1 2 3 4 In le t ( ug  L‐ 1 ) Outlet (µg L‐1) d) Total Pb removal release 0.0 1.5 3.0 4.5 6.0 0.0 1.5 3.0 4.5 6.0 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) e) Total Ni removal release 0 30 60 90 120 0 30 60 90 120 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) f) Total Zn removal release 0 500 1000 1500 2000 0 500 1000 1500 2000 In le t ( µg  L‐ 1 ) Outlet (µg L‐1) g) Total Fe removal release

36 Table 3-8 Dissolved metals event mean concentrations at the inlet and outlet of the Highway 36/61 ferric oxide-sand filter. Event  Start  Date  As  Cr  Cu  Pb  Ni  Zn  Fe  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  8/16/17  0.414  0.256  2.15  0.401  7.63  1.72  0.090  0.041  0.69  0.63  12.1  <4.19  56.2  32.9  9/18/17  0.763  0.393  2.68  0.164  8.47  3.16  0.195  0.035  2.32  2.04  61.1  3.79  186  69.5  9/24/17  0.622  0.372  5.10  0.672  3.60  3.45  0.106  0.035  1.35  1.50  22.7  3.87  178  52.4  10/6/17  0.449  0.200  2.21  0.903  2.30  1.48  0.171  0.052  0.62  0.62  20.8  2.81  69.1  23.5  10/7/17  0.465  0.213  3.77  0.386  4.15  1.54  0.131  <0.036  0.90  0.88  16.6  4.412  200  36.1  10/14/17  0.483  0.187  3.31  0.272  4.36  2.11  0.178  0.042  1.13  1.45  22.6  3.20  151  33.6  10/21/17  0.499  0.167  1.84  0.198  5.44  1.99  0.214  0.032  1.40  1.40  34.6  <4.04  168  41.9  5/30/18  0.827  0.399  1.03  0.335  3.03  3.67  0.131  <0.073  1.09  2.19  11.3  4.00  125  13.2  6/18/18  0.965  0.392  0.75  0.246  2.14  2.68  0.089  <0.073  1.13  1.59  10.1  1.21  456  9.55  7/2/18  0.964  0.407  0.86  0.314  1.97  1.91  0.090  <0.073  1.14  1.24  9.14  2.38  457  45.2  9/20/18  0.695  0.580  1.33  0.334  2.91  2.80  0.219  <0.073  1.26  1.54  14.3  2.39  536  69.5  10/8/18  0.515  0.244  1.01  0.616  1.89  1.45  0.200  0.037  0.84  0.84  8.92  0.99  141  24.0  10/26/18  0.653  0.398  0.39  0.086  2.67  2.10  0.092  <0.073  1.92  2.95  11.8  5.29  101  8.35  Average  0.640  0.324  2.03  0.379  3.89  2.31  0.147  0.0693  1.21  1.45  19.7  2.93  217  35.4  Note: All concentration are as µg L‐1.  1. Inlet concentration is a flow weighted average of the HWY Inlet West and HWY Inlet Pond concentrations.  2. Not included in average calculation. Equipment blank detected value within 5 times the reporting limit. Dissolved zinc greater than total zinc.  3. Average value calculated using regression on order in ProUCL5.1 for parameters with below reporting limit data.    

37 Table 3-9 Total metals event mean concentrations at the inlet and outlet of the Highway 36/61 ferric oxide-sand filter. Event  Start  Date  As  Cr  Cu  Pb  Ni  Zn  Fe  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  Inlet1  Outlet  8/16/17  0.68  0.28  3.46  0.57  13.5  2.36  1.69  0.231  1.61  0.82  38.7  <4.19  836  210  9/18/17  1.16  0.46  5.29  0.32  18.9  3.66  2.75  0.155  4.24  2.41  101  5.06  1964  440  9/24/17  0.86  0.49  7.05  0.81  9.55  3.93  2.07  0.212  2.47  1.66  52.4  5.25  1475  347  10/6/17  0.76  0.25  5.11  1.08  8.20  1.90  3.55  0.371  2.07  0.86  52.6  3.07  16742  279  10/7/17  0.73  0.25  5.86  0.65  11.2  1.82  1.97  0.078  2.29  1.08  56.2  2.213  1491  180  10/14/17  0.76  0.24  5.93  0.45  15.2  2.40  2.82  0.106  2.74  1.45  73.2  3.80  1774  238  10/21/17  0.69  0.19  3.66  0.28  13.2  2.20  2.56  0.070  2.41  1.48  77.0  <4.04  1490  176  5/30/18  1.44  0.52  2.55  1.24  6.34  5.07  1.83  0.283  2.61  2.56  26.9  4.89  1086  641  6/18/18  0.90  0.49  1.24  0.77  3.15  3.32  0.64  0.113  1.47  2.07  17.7  1.54  1399  363  7/2/18  1.09  0.48  1.25  0.57  3.10  2.69  0.60  0.118  1.47  1.31  16.5  1.95  1396  351  9/20/18  0.97  0.80  2.73  0.73  6.10  3.04  1.55  0.102  2.13  1.64  31.3  3.35  1844  531  10/8/18  0.64  0.29  1.46  0.94  2.81  1.62  0.93  0.122  1.23  0.90  14.7  1.60  626  184  10/26/18  0.85  0.57  1.23  0.48  5.28  2.24  1.06  <0.073  2.71  3.05  27.5  <5.29  101  669  Average  0.89  0.41  3.60  0.68  8.95  2.79  1.85  0.1534  2.26  1.64  45.1  3.224  1290  355  Note: All concentration are as µg/L.  1. Inlet concentration is a weighted average of the HWY Inlet‐West and HWY Inlet‐Pond concentrations.  2. Not included in average calculation, total iron duplicate relative percent difference 87 percent.  3. Not included in average calculation. Equipment blank detected value within 5 times the reporting limit. Dissolved zinc greater than total  zinc.  4. Average value calculated using regression on order in ProUCL5.1 for parameters with below reporting limit data.     3.2.4 Monitoring Results: General Chemistry The general chemistry data collected in conjunction with the metals analysis provide a further understanding of the function of the Highway 36/61 ferric oxide-sand filter and how it contrasts with the Woodlynn Avenue filter (Table 3-10). Alkalinity increases as water passes through the treatment cell, and this is likely a function of the iron rusting process that generates alkalinity. It is likely that alkalinity is generated between storm events (see Chapter 3.2.6) and as a consequence the pH of stormwater entering the treatment bed is buffered. Hardness and chloride and specific conductance also increase with treatment. It is possible that other anions and cations (e.g., sulfate, potassium) not measured are removed as the chloride and hardness increase appears larger than would be expected by the overall release of dissolved solids from the filter bed (specific conductance can be used to estimate dissolved solids). It also appears possible that the ferric oxide-sand filter is acting like an anion exchange column by removing and storing chloride during the winter months and during snowmelt and releasing chloride during the non- winter season. The Highway 36/61 ferric oxide sand filter is also removing dissolved organic carbon which may be beneficial to downstream waters by reducing the biological oxygen demand of the treated stormwater. This is also in contrast to Woodlynn Avenue, perhaps suggesting that there is more remaining ferric oxide binding capacity at the Highway 36/61 site. Total suspended solids removal was 86 percent and volatile

38 suspended solids removal was 82 percent. Although the treatment performance of the Highway 36/61 site was worse than Woodlynn Avenue on a percentage basis, the treated concentration of total suspended solids and volatile suspended solids were similar for both sites. Total suspended solids removal at Highway 36/61 was likely less than at Woodlynn Avenue as the pre-treatment pond upstream of the Highway filter removed solids. Table 3-10 General parameter event mean concentrations at the inlet and outlet of the Highway 36/61 ferric oxide-sand filter treatment system. Event  Start Date  Total  Alkalinity  (mg L‐1 as  CaCO3)  DOC         (mg L‐1)  Cl (mg L‐1)  Hardness  (mg L‐1 as  CaCO3)  TSS           (mg L‐1)  VSS            (mg L‐1)  Specific  Conductance  (µs cm‐1)  Average  pH (s.u.)  Average DO  (mg L‐1 )  In1  Out  In1  Out  In1  Out  In1  Out  In1  Out  In1  Out  In1  Out  East Inlet  or In‐Situ  In‐Situ  8/16/17  26  60  4.2  2.6  7.2  15  37  45  17  2.6  5.0  0.9  75  160  (2)  4.4  9/18/17  46  150  15  7.5  26  310  65  310  34  2.8  16  2.8  196  1389  7.4  5.9  9/24/17  34  110  10  5.1  16  89  59  120  39  6.2  14  2.9  134  523  6.8  3.6  10/6/17  22  58  22  2.7  4.6  12  33  30  40  2.6  9.7  0.8  76  142  6.7  4.7  10/7/17  41  90  6.1  2.8  14  63  41  90  27  1.5  10  1.0  131  389  6.8  3.8  10/14/17  51  120  8.2  3.7  22  180  88  220  42  3.0  18  2.2  179  715  7.0  4.8  10/21/17  47  120  7.9  3.4  20  160  130  200  44  1.1  18  1.1  161  782  7.1  4.2  5/30/18  50  150  8.3  5.1  380  520  58  160  15  4.5  7.0  3.0  1494  2160  6.7  4.5  6/18/18  49  150  8.3  4.1  360  490  79  110  18  3.2  5.3  <0.65  1487  1890  6.9  (2)  7/2/18  61  130  8.3  4.1  280  260  98  90  11  3.0  5.6  2.0  1112  1102  6.8  5.2  9/20/18  29  140  9.6  6.4  42  170  130  170  26  6.0  16  8.0  254  890  6.2  1.0  10/8/18  52  100  5.2  2.8  33  110  57  97  9  8.5  3.3  0.8  231  622  6.3  5.7  10/26/18  120  250  11  4.7  170  990  140  710  20  2.6  10  <1.4  714  3480  6.4  8.6  Average  48  130  9.6  4.2  110  260  78  180  26  3.7  11  2.03  480  590  6.8  4.7  1. Inlet is a weighted average concentration of the Pond and the West inlet.  2. Data not collected.  3. Average value calculated using regression on order in ProUCL5.1 for parameters with below reporting limit data.    3.2.5 Dissolved Oxygen and pH Dissolved oxygen (Figure 3-12 and Figure 3-13) and pH (Figure 3-14) were measured continuously in the ferric oxide-sand filter bed. Similar to the Woodlynn Avenue site, dissolved oxygen declined from the start to the end of the storm event. However, dissolved oxygen never reached zero during 2017 and 2018.

39 Figure 3-12 In-situ dissolved oxygen measurement at the Highway 36/61 ferric oxide-sand filter bed in 2017 and 2018. Because ponding occurred rarely at the Highway 36/61 treatment cell, there was less time for organic matter decay and oxygen consumption. After examining several individual storm events, it was observed that dissolved oxygen was low coming into the treatment cell rather than dissolved oxygen being high at the start of the storm event and then declining linearly as was observed at Woodlynn Avenue. The storm event on May 29 to May 30, 2018 demonstrates how dissolved oxygen was often at the lowest level at the start of the storm event and prior to ponding in the filter bed (Figure 3-13). As the storm event progressed and with water levels increasing, dissolved oxygen rose. This does not indicate that dissolved oxygen consumption did not occur, it demonstrates however that reduced ponding time can prevent low oxygen and ferric oxide fowling. At peak water level (Figure 3-13), it would take approximately 6 hours for the majority of water to drain out of the cell. It can be seen that dissolved oxygen fluctuated within a given range during this period rather than declined. It appears that dissolved oxygen dynamics at Highway 36/61 is largely determined by upstream activity. Although pH varied between storm events within a range circa 6 to 7, pH did not appear to fluctuate notably during a single storm event, nor did the length of the event affect pH. The maintenance of pH within 6 to 7 in the ferric oxide filter bed is signficant in that metals speciation in this range is amenable to binding by ferric oxides (Smith, 1999). 0 2 4 6 8 10 12 14 6/24/17 8/3/17 9/12/17 10/22/17 12/1/17 Di ss ol ve d  Ox yg en  (m g  L‐ 1 ) 0 2 4 6 8 10 12 14 04/20/18 06/09/18 07/29/18 09/17/18 11/06/18 Di ss ol ve d  Ox yg en  (m g  L‐1 )

40 Figure 3-13 Dissolved oxygen dynamics within the Highway 36/61 ferric oxide-sand bed during a single storm event in 2018. Figure 3-14 Example of pH dynamics within the Highway 36/61 ferric oxide-sand bed during a select number of storm events in 2018. 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 0 1 2 3 4 5 6 7 8 9 5/29/18 10:48 5/29/18 20:24 5/30/18 6:00 W at er  Le ve l ( ft)  o r F lo w  (f t3 s‐1 ) DO  (m g  L‐1 ) Dissolved Oxygen Average Water Level Flow‐HWY East Inlet top of filter bed 5.5 5.7 5.9 6.1 6.3 6.5 6.7 6.9 7.1 7.3 7.5 0 5 10 15 20 pH  (s u)  H ig hw ay  3 6/ 61  Fi lte r B ed Hour During Storm 5/28/18 6/16/18 9/20/18 9/20/18 10/8/18 10/26/18

41 3.2.6 Multi-Sample Event For a storm event spanning October 9 to October 10, 2018, the autosampler flow pacing was set to collect individual one liter samples throughout the entire storm event. Samples were placed into 7 or 8 groups enabling metals and general parameters to be traceable throughout the event. It was hypothesized that treatment efficiency may change throughout the course of the storm event with first flush effects, residence time, and other factors such as change in pH and dissolved oxygen. However, the change in dissolved and total metals at the outlet from the start to the end of the event was minimal. This confirms what was also shown in Figure 3-14, that the concentration of dissolved and total metals post-treatment is largely independent of the inlet concentration for the range of metals concentrations experienced in this study. This suggests that treatment performance may be best evaluated with the use of the outlet concentration reported as part of this study rather than applying a percent removal as the percent removal was a function of inlet concentration rather than performance. Several general parameters are plotted in Figure 3-16 for the October 9 through 10 storm event. There was a first flush effect for these parameters and in general a steady decrease in concentration was observed as the storm event progressed. It should be noted that the inlet concentration in this figure was a flow weighted average of the HWY Pond and HWY West inlets. The concentration of chloride, alkalinity, and to a lesser degree hardness at the outlet decreased over the course of the storm event and this also corresponded to a concentration decline at the inlet. The outlet concentrations were higher than the inlet, and it is interpreted that chloride and alkalinity were washed out as the storm progressed. In contrast, dissolved organic carbon and volatile suspended solids were largely constant at the outlet as ferric oxide readily binds these weakly negative organic compounds. It appears that there is a concentration based limit to dissolved organic carbon and volatile suspended solids removal.

42 Figure 3-15 Change in total and dissolved metals at the two inlets and one outlet of the Highway 36/61 ferric oxide-sand filter during the course of one storm event. 0.0 0.5 1.0 1.5 2.0 2.5 10/9 5:01 10/9 18:51 10/9 20:19 10/9 21:27 10/9 22:29 10/9 23:11 10/9 23:51 10/10 0:31 Di ss ol ve d  M et al s O ut le t ( µg  L‐ 1 ) 0.0 0.5 1.0 1.5 2.0 2.5 10/9 5:01 10/9 18:51 10/9 20:19 10/9 21:27 10/9 22:29 10/9 23:11 10/9 23:51 10/10 0:31 To ta l M et al s O ut le t ( µg  L‐ 1 ) Nickel Lead Copper Chromium Arsenic Zinc 0 2 4 6 8 10 12 0.0 0.5 1.0 1.5 2.0 2.5 10/9 0:41 10/9 17:32 10/9 20:31 10/9 22:16 10/9 22:44 10/9 23:32 10/10 0:03 10/10 1:19 Zi nc  (µ g  L‐1 ) Di ss ol ve d  M et al s H W Y  Po nd        (µ g  L‐1 ) 0 5 10 15 20 0.0 0.7 1.4 2.1 2.8 3.5 10/9 0:41 10/9 17:32 10/9 20:31 10/9 22:16 10/9 22:44 10/9 23:32 10/10 0:03 10/10 1:19 Zi nc  (µ g  L‐1 ) To ta l M et al s H W Y  Po nd  (µ g  L‐1 ) 0 5 10 15 20 25 30 35 40 0 6 12 18 24 30 10/9 0:15 10/9 16:01 10/9 18:53 10/9 21:50 10/9 22:31 10/9 23:36 10/10 5:00 Zi nc  (µ g  L‐1 ) Di ss ol ve d  M et al s H W Y  W es t  (µ g  L‐1 ) 0 10 20 30 40 50 60 70 80 90 0 7 14 21 28 35 10/9 0:15 10/9 16:01 10/9 18:53 10/9 21:50 10/9 22:31 10/9 23:36 10/10 5:00 Zi nc  (µ g  L‐1 ) To ta l M et al s H W Y  W es t ( µg  L‐ 1 )

43 Figure 3-16 Change in general chemical parameters in the inlet and the outlet of the Highway 36/61 ferric oxide-sand filter during the course of one storm event from October 9 to October 10, 2018. 3.3 Comparison of Woodlynn and Highway 36/61 The dissolved and total metals removal performance at Highway 36/61 was notably better than the performance at Woodlynn Avenue. For dissolved metals, there was a significant reduction in arsenic, chromium, copper, lead, zinc, and iron at Highway 36/61, and for Woodlynn Avenue only arsenic and zinc were significantly reduced (Table 3-11). For Highway 36/61, significant total metals removal was had for all metals measured. For Woodlynn Avenue, significant metals removal was had for all total metals measured except for copper and nickel. It should be noted that inlet concentrations were higher for Highway 36/61 for both total and dissolved metals for all of the metals monitored. Hence, performance may have been hindered by the lower influent concentrations at Woodlynn Avenue. Treated dissolved metals concentrations were similar for Woodlynn Avenue and Highway 36/61. Treated total metals concentrations were similar for Woodlynn Avenue and Highway 36/61 for copper, lead, and zinc. It is possible that treatment efficiency appeared worse at Woodlynn Avenue compared to Highway 36/61 because of the lower influent concentrations, but there may also be other factors. The filtration rate at Woodlynn was approximately 4 to 12 inches per hour (after removing one outlier) while the average storm event filtration rate at Highway 36/61 was measurably less and ranged from 0.02 to 0.53 inches per hour. At Woodlynn Avenue, dissolved oxygen also had the potential to decline to lower minima during extended detention and this may have affected performance. The export of dissolved iron at Woodlynn Avenue may be an indication of the different redox conditions of the two treatment cells. Dissolved organic carbon was not removed at Woodlynn Avenue whereas dissolved organic carbon was removed at Highway 36/61. Dissolved organic carbon may be acting as a carrier for dissolved metals at Woodlynn Avenue. The pass-through of dissolved organic carbon may be an indication of limited remaining binding capacity at Woodlynn. 0 2 4 6 8 10 10/9 00:41 10/9 17:32 10/9 20:31 10/9 22:16 10/9 22:44 10/9 23:32 10/10 00:03 10/10 01:19 Co nc en tr at io n  (m g  L‐1 ) Highway Inlet Pond: VSS HWY Outlet: VSS HWY Inlet Pond: DOC HWY Outlet: DOC 0 50 100 150 200 250 10/9 00:41 10/9 17:32 10/9 20:31 10/9 22:16 10/9 22:44 10/9 23:32 10/10 00:03 10/10 01:19 Co nc en tr at io n  (m g  L‐1 ) HWY Inlet Pond: Chloride HWY Outlet: Chloride HWY Inlet Pond: Hardness HWY Outlet: Hardness HWY Inlet Pond: Total Alkalinity HWY Outlet: Total Alkalinity

44 Table 3-11 Comparison of average total and dissolved metals removal at Woodlynn Avenue and Highway 36/61. Metal  Fraction  Woodlynn Avenue  Highway 36/61  Inlet      (ug L‐1)  Outlet  (ug L‐1)  Percent  Removal  Hypothesis Test     (p value)  Inlet  Outlet  Percent  Removal  Hypothesis  Test             (p value)  Arsenic  Dissolved  0.118  0.062  47%  <0.000  0.640  0.324  49%  <0.000  Chromium  Dissolved  0.328  0.271  17%  0.146  2.03  0.379  81%  <0.001  Copper  Dissolved  1.85  2.91  ‐57%  0.978  3.89  2.31  41%  0.009  Lead  Dissolved  0.049  0.041  16%  0.073  0.147  0.069  53%  <0.001  Nickel  Dissolved  0.439  2.65  ‐504%  1  1.21  1.45  ‐19%  0.0847  Zinc  Dissolved  9.03  2.67  70%  <0.000  19.7  2.93  85%  <0.000  Iron  Dissolved  15.5  29.9  ‐93%  0.820  217  35.4  84%  0.001  Arsenic  Total  0.252  0.071  72%  <0.000  0.89  0.41  54%  <0.000  Chromium  Total  1.70  0.365  79%  <0.000  3.60  0.68  81%  <0.000  Copper  Total  4.00  2.97  26%  0.075  8.95  2.79  69%  <0.000  Lead  Total  1.06  0.105  90%  <0.000  1.85  0.153  92%  <0.000  Nickel  Total  1.47  2.96  ‐102%  0.999  2.26  1.64  28%  0.021  Zinc  Total  28.1  4.85  83%  <0.000  45.1  3.22  93%  <0.000  Iron  Total  774  138  82%  0  1290  355  73%  <0.000  Notes:  Null hypothesis: mean inlet concentration < mean outlet concentration, p <0.05 indicates a rejection of the null hypothesis.  Alternative hypothesis: mean inlet concertation > mean outlet concentration.  Two sample t‐test with alpha = 0.05 used for data without values below the reporting limit.  Either Wilcoxon‐Man‐Whitney or Gehan with alpha = 0.05 used for data with values below the reporting limit.  Statistical tests conducted using two sample hypothesis testing in ProUCL 5.1.