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.
57 Chapter 5 Ferric Oxide Filter Sizing Methodology Sizing a ferric oxide-sand filter to meet load reduction goals is different than sizing for ponds or other filtration devices. Although future study is warranted to confirm the outcome of this study, it appears that ferric oxide-sand filters will likely be able to achieve a consistent treated concentration for a range influent pollutant concentrations and infiltration rates (which are a function of hydrology and the treatment system design) that are similar to the concentrations and rates measured in this study. Erickson et al., 2012, demonstrated that iron can react quickly and that iron-sand mixtures with contact times of approximately 5 to 7 minutes could remove phosphate. This is confirmed for metals with this current study as treatment performance did not change notably at Highway 36/61 for a range of events with a range of infiltration rates. It can be inferred that the rate of ferric oxide-metal binding is much faster than the infiltration rate of sand. Performance also did not change measurably over the course of one event monitored at Highway 36/61. Sizing will largely be based upon the volume of water desired treated with the primary limitation being the duration of ponding that can be provided before oxygen is reduced to zero and anaerobic conditions are established. However, the use of another media with a hydraulic conductivity higher than sand may require the inclusion of performance targets for sizing if the contact time is short enough to impact removal. Although not observed in this current study, it is possible that sites with high metals concentrations may reduce performance measured as percent removal. Sites with high hydraulic conductivity media and elevated metals concentrations were not considered in the sizing methodology below. Several methods are available for sand filter sizing. Urbonas, 1999, described in detail how to size sand filters for treatment volume and solids removal. Taylor et al., 2014 provide performance and life cycle costs for sand filters based upon sand filter design (e.g., size and detention storage) and watershed characteristics. These and other sizing and design methodologies may be useful and should be reviewed prior to construction of a ferric oxide-sand filter. However, a ferric oxide-sand filter is different than a sand filter in that performance appears to be unrelated to filtration rates and influent concentrations and the prospect of ferric oxide fouling requires additional attention to dissolved oxygen. Also, given the limited space typically available in the highway right of way, ferric oxide sand filters that are sized to filter at the same rate as the largest storm event (e.g., no storage) peak discharge rate will need to be larger than systems with some storage. Hence, a deterministic approach is recommended for ferric oxide-sand filters largely because of the added complexity of dissolved oxygen and the potential for fouling. This approach uses a continuous flow record rather than a stochastic approach that provides the probability of occurrence with respect to storm event volume and event duration. With a deterministic approach, the volume of water that is treated and the volume of water that is bypassed is determined for each stage (e.g., hour) of an event for a given ferric oxide-sand filter configuration. Both approaches are valid. A deterministic approach requires more data and a larger evaluation effort but will likely be more accurate with respect to expected bypass volumes. 5.1 Sizing Methodology A sizing methodology is provided below that is based upon several factors with a load reduction goal and water volume treatment target being the primary determinants of two sizing parameters for ferric oxide
58 sand filter system: (1) bed area, and (2) above-bed detention storage. The determinants of filter bed area and detention storage are filter bed material (ferric oxide and sand) hydraulic conductivity (see values in Table 3-11) and the rate of oxygen consumption in stormwater when ponded above the filter. The methodology for calculating oxygen consumption was provided in Figure 3-5 Dissolved oxygen in ponded water at Woodlynn Avenue during two storm events in 2018. Table 3-5 and site-specific data will need to be collected to calculate expected oxygen consumption rates and the time it takes for oxygen to decline to zero. At near zero oxygen, iron has the potential to be reduced resulting in a loss of iron in the filter bed and a reduction in performance (Stumm and Morgan, 1996.). To accurately size the filter bed area and storage volume, a continuous runoff record is needed to calculate the fraction of water treated by the bed and the fraction that will bypass. A continuous record is necessary as the infiltration rate will change with height above the bed and the volume of water that bypasses the filter will be dependent upon the level of water in the bed. The ferric oxide sand filter sizing steps are provided below. Step 1. Identify the Load Reduction or Treatment Volume Goal 1.1. A continuous runoff record (modeled or measured) will be needed for the tributary watershed 1.2. If a load reduction goal is required as part of a regulatory program (e.g., TMDL), the volume of water that needs to be treated can be determined by knowing the stormwater concentration and the estimated reduction with treatment through the ferric oxide filter (see Table 3-11). This treatment volume may be all or some fraction of the runoff from the tributary watershed. Note that a certain fraction of the storm events that deliver runoff to the treatment cell will not generate outflows (see Figure 3-8). For these events, treatment efficiency is 100 percent. 1.3. If a load reduction goal has not been identified and other factors such as space availability or cost determine the storm water volume that can be treated, then the volume of water treated will be determined by these limiting factors. Step 2. Identify Maximum Allowable Ponding Time When storm water is ponded above the filter bed, dissolved oxygen is consumed (see Chapter 3), and the extent of dissolved oxygen loss will depend upon the length of time that the water is ponded above the filter bed (e.g., the residence time), the organic matter decay rate, the dissolved and particulate organic matter concentration, and temperature. Using the continuous runoff record (hourly time step is recommended) and a proposed treatment system configuration (filter bed area, filter bed depth, and overflow elevation), dissolved oxygen can be modeled in the ponded water assuming that the ponded water is completely mixed. Calculation steps are provided below. 2.1. Data needs
59 An estimate of the ultimate biological oxygen demand (UBOD as mg O2 L-1) and the storm water oxygen consumption (mg O2 L-1 d-1) rate will be needed. Ultimate biological oxygen demand and the oxygen consumption rate can be determined by collecting stormwater at the project site and submitting the water for laboratory analysis or estimates can be calculated from stormwater chemistry (ultimate biological oxygen demand and decay were estimated for this current study, see Table 3-5). Collection of several storm events is recommended. Also, it may be valuable to install a dissolved oxygen probe at the proposed construction site to develop a record of dissolved oxygen concentration in stormwater prior to treatment and ponding. 2.2. Equations to calculate dissolved oxygen in the ponded water Dissolved oxygen in the ponded water can be calculated as follows: ð¶ ð¶ â ð ð¶ â ð ð¶ â ð ð¶ â ðð Eq. 4 Where: C = dissolved oxygen (mg L-1) in the ponded water Co = initial dissolved oxygen (mg L-1) in the ponded water V = ponded water volume Cd = oxygen consumption rate (mg L-1 t-1) Cin = dissolved oxygen concentration in stormwater Vin = volume of water flowing into the treatment cell per time step Vout = volume of water flowing out of the treatment cell per time step Dissolved oxygen consumption in the treatment cell per time step can be calculated as follows: Cd = UBOD*(1-exp(-K*t))*1.04(T-25) Eq. 5 Where: Cd = oxygen consumption rate (mg L-1 t-1) UBOD = ultimate biological oxygen demand (mg O2 L-1) K = oxygen consumption rate (t-1) or ultimate biological oxygen demand decay rate (t-1) t = time (hour time step is recommended). T = temperature (oC). Assume that the water is completely mixed above the treatment cell bed.
60 2.3. Filtration rate The filter bed infiltration rate is calculated using Darcyâs Law and the depth of water ponded above the filter bed. This assumes that there are no tail water effects. The filtration rate is calculated as: Q= KA*ïH/L Eq. 6 Where: K= hydraulic conductivity (ft t-1) A = filter bed surface area (ft2) ïH = fractional height of water level above sand filter. ïH/L = 1 for water levels less than or equal to the filter bed depth, and for water above the bed ïH is the depth above the bed surface. L = filter bed depth (ft) Q/A = filtration rate (ft t-1) A timestep of one quarter to one hour is recommended. 2.4. Ponding depth Ponding depth is calculated from a water balance of flow into and out of the treatment cell. Storage in the filter is provided by the filter bed as well as the volume above the bed which is a function of the above-bed geometry and the elevation of the overflow outlet. Calculation of filter bed storage and ponding depth is calculated in two steps, the first being the filter bed storage, and when the filter bed pore spaces are full of water the water storage above the filter can be calculated. The depth in the filter bed at each time step is calculated. The first step is to calculate the total volume in the system at each time step: ð ðð â ð¼ â ð´ ð¼ â ð´ ð Eq. 7 Where: A = Bed area (ft2) V0 = Initial volume (ft3)âin the filter bed and above the bed Ib = Bed filtration rate (in t-1) IS = Infiltration rate to soils (in t-1) VIN = Storm water volume into cell (ft3) during time step (t)
61 If VT is less than the pore space volume of the filter bed which is given by: ð ð â ð· â ð´ Eq. 8 Then the depth in the filter bed is ð· â Eq. 9 Where: P= porosity (unitless) Vo = Initial volume (ft3) Ib = Bed filtration rate (in t-1) Is = Infiltration rate to soils (in t-1) Vin = Storm water volume into cell (ft3) during time step (t) Db = Filter bed material depth (ft) When Df => Db, ponding occurs and Db = (VT â VB)/A Eq. 10 The depth of water above the filter bed may be defined by a volume to depth relationship if the side slopes of the filter are not vertical. Step 3. Optimization Optimization is conducted by modifying two parameters: (1) the overflow elevation, and (2) filter bed area. The goal is to minimize the filter bed area and total footprint while also minimizing the frequency at which dissolved oxygen drops to zero in the filter bed. Sensitivity analysis may also be conducted for ultimate biological oxygen demand, hydraulic conductivity, and infiltration. An example of this process is provided in the following section. 5.2 Sizing Example Using a hydrologic model built for the Highway 36/61 watershed, one year of runoff was simulated using SWMM (Huber and Dickinson, 1988) and precipitation data collected at the ferric oxide-sand treatment system in 2018 as well as the Minneapolis-St. Paul International Airport rain gauge. This continuous record was used to calculate a hydrologic balance for a ferric oxide-sand treatment system using the equations in Chapter 5.1. Six different ferric oxide-sand filter conditions were evaluated: (1) ferric oxide-sand filter bed size of 0.125, 0.25, and 0.5 acres, and (2) overflow outlet depth set at 1 and 2 feet above the sand filter bed. Dissolved oxygen in water ponded above the sand filter was calculated using the equations in Chapter 5.1. Coefficients and assumptions included a range of ultimate biological oxygen demand, a hydraulic conductivity of 1.94 inches per hour, bed depth of 1.2 feet, porosity of 0.45, ultimate biological oxygen demand decay coefficient of 0.05 (hr-1), water temperature of 25 oC, and dissolved oxygen water saturation of 8 mg L-1.
62 A compilation of the different simulations are shown in Table 5-1 and an example of the outcome of one of the simulation iterations (ferric oxide-sand filter bed area of 0.25 acres, overflow outlet depth of 1 foot, and ultimate biological oxygen demand of 25 mg L-1) is shown in Figure 5-1. Using the continuous flow record enables the simulation of water level and hence the volume of water that bypasses the filter bed (Figure 5-1a and Figure 5-1b) throughout each storm event. The duration that water ponds above the filter bed determines how much oxygen is lost as a result of ultimate biological oxygen demand (Figure 5-1c). In this example, dissolved oxygen did not drop below 0 mg L 1, however, if ponding time is long and the ultimate biological oxygen demand high, oxygen drops below 0 mg L 1 for this same treatment cell configuration. Clearly, oxygen cannot drop below zero, but this is an indicator of reduced conditions and the potential for iron reduction, solubilization, and release. When oxygen approaches or reaches zero other stormwater or filter bed compounds (e.g., sulfate, nitrate, ferric oxide and oxides of manganese) will be reduced. However, modeling the redox condition of each of these compounds can be complex, and for the practitioner, modeling oxygen is straightforward and will provide an understandable indication of potentially reducing conditions that could foul the ferric-oxide sand filter bed. Table 5-1 shows the outcome of this sizing exercise (note that this is different from the NCHRP 792 tool). Cell size, overflow outlet placement, and ultimate biological oxygen demand affect how much water is treated and how much is bypassed, as well as how long low dissolved oxygen conditions (e.g., dissolved oxygen below 0 mg L 1) persist. The table includes aerial water load which is the depth of water above the filter that is treated annually. A larger aerial water load signifies more work being done by the filter per unit area. This statistic does not include water that is bypassed. Finally, a dissolved metal load reduction estimate is provided that is based upon the monitoring results at Highway 36/61. It can be seen that smaller filters with an overflow outlet that is set higher above the filter bed can treat more water per unit filter bed surface area. Overall, this is a more economical option as less area is used for the filter. However, sites with higher ultimate biological oxygen demand may have persistently low dissolved oxygen. This may affect performance, release metals, and bleed iron out of the filter bed leading to a loss of ferric oxide media over time. The practitioner will need to identify the frequency of low dissolved oxygen conditions that can be tolerated. A smaller filter bed area with the same watershed area will treat less water and the dissolved metal load reduction will be less. Hence, the sizing will depend upon the available space and the desired load reduction.
63 Figure 5-1 Ferric oxide-sand filter sizing example showing (a) water level within and above the ferric oxide sand filter bed, (b) flows through the overflow outlet, and (c) dissolved oxygen in water ponded above the filter bed.
64 Table 5-1 Outcome of the ferric oxide-sand treatment cell sizing exercise. CellÂ andÂ WatershedÂ ParametersÂ HydrologyÂ PotentialÂ DissolvedÂ LoadÂ ReductionÂ (gÂ yearâ1)Â OverflowÂ DepthÂ (ft)Â UBODÂ (mgÂ Lâ1)Â CellÂ SizeÂ (ac)Â VolumeÂ TreatedÂ (ft3)Â VolumeÂ BypassedÂ (ft3)Â PondingÂ TimeÂ (hrs)Â TimeÂ (hr)Â DO<0Â mgÂ Lâ1Â MaximumÂ WaterÂ LevelÂ (ft)Â AerialÂ WaterÂ LoadÂ (ft)Â ZnÂ CuÂ PbÂ 1Â 10Â 0.125Â 297183Â 252909Â 186Â 0Â 2.2Â 55Â 141Â 13Â 0.7Â 0.250Â 488828Â 61265Â 106Â 0Â 2.2Â 45Â 232Â 22Â 1.1Â 0.500Â 547653Â 2439Â 25Â 0Â 2.2Â 25Â 260Â 24Â 1.2Â 25Â 0.125Â 297183Â 252909Â 186Â 0Â 2.2Â 55Â 141Â 13Â 0.7Â 0.250Â 488828Â 61265Â 106Â 0Â 2.2Â 45Â 232Â 22Â 1.1Â 0.500Â 547653Â 2439Â 25Â 0Â 2.2Â 25Â 260Â 24Â 1.2Â 50Â 0.125Â 341099Â 208993Â 162Â 32Â 2.2Â 63Â 162Â 15Â 0.8Â 0.250Â 488828Â 61265Â 106Â 18Â 2.2Â 45Â 232Â 22Â 1.1Â 0.500Â 547653Â 2439Â 25Â 3Â 2.2Â 25Â 260Â 24Â 1.2Â 2Â 10Â 0.125Â 386237Â 163860Â 222Â 0Â 3.2Â 71Â 183Â 17Â 0.9Â 0.250Â 529407Â 20685Â 113Â 0Â 3.2Â 49Â 251Â 24Â 1.2Â 0.500Â 550093Â 0Â 25Â 0Â 2.3Â 25Â 261Â 25Â 1.2Â 25Â 0.125Â 434466Â 115627Â 193Â 18Â 3.2Â 80Â 206Â 19Â 1.0Â 0.250Â 529407Â 20685Â 113Â 4Â 3.2Â 49Â 251Â 24Â 1.2Â 0.500Â 550093Â 0Â 25Â 0Â 2.3Â 25Â 261Â 25Â 1.2Â 50Â 0.125Â 434466Â 115627Â 193Â 72Â 3.2Â 80Â 206Â 19Â 1.0Â 0.250Â 529407Â 20685Â 113Â 28Â 3.2Â 49Â 251Â 24Â 1.2Â 0.500Â 550093Â 0Â 25Â 3Â 2.3Â 25Â 261Â 25Â 1.2Â Notes:Â AverageÂ dissolvedÂ concentrationÂ reductionsÂ areÂ asÂ follows:Â ZnÂ (16.77Â ÂµgÂ Lâ1),Â CuÂ (1.58Â Â ÂµgÂ Lâ1),Â PbÂ (0.078Â ÂµgÂ Lâ1).Â AerialÂ waterÂ loadÂ =Â volumeÂ treatedÂ dividedÂ byÂ totalÂ cellÂ area.Â