1651 Water Quality Design Storms for Stormwater Hydrodynamic Separators Victoria J. Fernandez-Martinez 1 and Qizhong Guo 2 1 Rutgers University, Department of Civil and Environmental Engineering, 623 Bowser Road, NJ 08854 2 Corresponding Author, Rutgers University, Department of Civil and Environmental Engineering, 623 Bowser Road, NJ 08854; Phone: 732-445-4444; Fax: 732-445-0577; E-mail: Qguo@rci.rutgers.edu Abstract When measure of the device performance is based on laboratory data, NJCAT calculates the average annual removal efficiency using the NJDEP-specified weighting factors. In this study, ten years of precipitation records were used to quantify frequency distributions of runoff volumes/depths as well as runoff peak flow rates. The SWMM model was used to simulate the runoff events produced by the given precipitation events. The model results indicate that the weighting factors, based on frequency distribution of the peak runoff rates normalized by the peak runoff rate generated by the uniform-intensity water quality design storm, are close to the NJDEP-specified weighting factors. The impact of using two different water quality design storms, uniform vs. non-uniform intensity distribution, to size an interim-certified hydrodynamic separator was also evaluated. The SWMM model was used to continuously simulate solids loading to the treatment device. The lab-measured relationship between flow rate and removal efficiency was used to specify the removal rate of solids within the storm event. The model results also indicate that sizing with the uniform-intensity design storm would achieve a removal efficiency close to the one verified. Sizing with the non-uniform intensity design storm, although could be taken as a conservative approach, would achieve the removal efficiency considerably higher than the one verified. 1. Introduction In the State of New Jersey the Stormwater Management Rules (NJDEP 2004a) require Stormwater Best Management Practices (BMPs), for major new developments, to achieve an annual average of Total Suspended Solids (TSS) removal efficiency equal or greater than 80%, and for major re-developments, 50% TSS removal. Regulations require this efficiency as an annual average. Therefore, the treatment devices do not have to be designed to achieve 80 or 50 percent of TSS reduction for every individual storm event that occurs in a year. The rules (NJAC 7:8-5.5) also specify a water quality design storm which consists of 1.25 inches of rainfall falling non-uniformly over two hours. This design storm substituted the one specified in NJAC 5:21-7.6 and 7:13-2.8(b), defined as 1.25 inches of rainfall falling in two hours with a uniform intensity of 0.625 in/hour. According to the proponents of the non-uniform quality storm, although the previous one provided sound basis for the design of stormwater
1652 quality treatment facilities that require an estimate of the total runoff volume that will flow to the facility it did not provide information for the design of those facilities that required an accurate estimate of the runoff peak flow rate that would flow through them. The non-uniform water quality storm provides estimates of total runoff volume as well as peak runoff rate for the same storm event. Based on the described regulations the New Jersey Best Management Practices Manual (NJDEP 2004b) indicates that the design of flow-through devices, such as the hydrodynamic separators, must be based on the runoff peak flow rate generated by the current non-uniform water quality design storm. Therefore under such specification, project engineers will size the devices to achieve the required solid removal efficiency when the treatment flow rate, in the drainage area, equals the runoff peak flow rate generated by the water quality design storm, with non-uniform intensity distribution. This study intends to compare the hydrodynamic separator sizing methodology based on the uniformly distributed quality design storm with the sizing methodology based on the nonuniformly distributed design storm. Two approaches were taken to determine the removal efficiency that a hydrodynamic separator could achieve when sized with different design storms. The first approach consisted of calculating the removal efficiency of the device using weight factors which are based on the frequency distribution of the runoff depths and peak runoff rates generated by precipitation over 10 years. The second approach consisted of continuous simulation of the solids removal performance of a hydrodynamic separator during 10 years of precipitation events. 2. Methodology 2.1. Curve Fitting of Hydrodynamic Separator Performance Many hydrodynamic separator manufacturers have tested the solid removal performance of their devices at full scale and under laboratory controlled conditions. The results from many of those tests have been submitted to the New Jersey Corporation of Advanced Technology (NJCAT) for verification and performance of many devices has been certified by the New Jersey Department of Environmental Protection (NJDEP). When this study was started, the NJCAT verification report most recently available corresponded to the FloGard Dual-Vortex Hydrodynamic Separator (model DVS-48), manufactured by Kristar Enterprises Inc. Therefore, the FloGard was the device selected for evaluating the impacts of using different water quality design storms to size the hydrodynamic separators. The removal efficiency of the device was tested following the Total Suspended Solids Laboratory Test Procedure developed by NJDEP (2003). Tests were run at flow rates of 25, 50, 75, 100 and 125 % of the FloGard DVS-48 treatment flow rate (280 gpm) (NJCAT 2007). Figure 1 shows the TSS removal efficiencies measured at all the five tested flow rates and at three TSS influent concentrations (100 mg/l, 200 mg/l, and 300 mg/l). TSS was measured as Suspended Sediment Concentration (SSC). It can be observed that the solids removal efficiency
1653 of the device decreases as the operating flow rate increases and the removal efficiency does not vary significantly with the influent concentrations. An exponential regression of the measured removal efficiencies vs. the operating flow rates yields the following mathematical expression for the model DVS-48: where RE = removal efficiency (%) Q = operating flow rate (gpm) Figure 1. Measured FloGard SSC Removal Efficiencies vs. Operating Flow Rates 2.2. Collection of Precipitation Data Hydrodynamic separators on site receive stormwater runoff under non-uniform flow rates that change in small time steps. Such variations affect the removal efficiency of the device since the removal performance of the device depends on its influent flow rate, as discussed above. For that reason precipitation data with small-time-interval readings was needed for the intended simulation. The National Climatic Data Center (NCDC) reports precipitation readings every 15 minutes, for 19 stations in New Jersey. At the time of this study, reliable readings with even smaller time steps were not available from national or regional climatologic services.
1654 Trenton, New Jersey was the meteorological station selected to collect 10 years of precipitation data from the NCDC website since New Jersey Rainfall Intensity-Duration-Frequency Curves were developed from Trenton rainfall data between 1913 and 1975 (NJDEP 2004b). The available data at the mentioned website covered the period between 1977 and 2001. The daily data of each year was summed in order to calculate the annual precipitations to select 10 years in a row with an annual precipitation close to the long term annual mean precipitation in New Jersey, which is 40.24 inches, according to the Office of the New Jersey State Climatologist. Years 1981 to 1990 were selected, with annual rainfalls of 45.0, 37.1, 46.2, 46.1, 27.4, 37.0, 50.9, 33.5, 49.4, and 42.2 inches, respectively. Average annual precipitation during the studied period was 41.48 inches, close to the long term average. 2.3. Determination of Weighting Factors Stormwater Runoff Continuous Simulation EPA Storm Water Management Model (EPASWMM), version 5.0, was used to simulate the peak flow rates generated by the collected precipitation data over a defined subcatchment. The properties of the subcatchment introduced in SWMM intended to represent highly impervious drainage areas, such as parking lots, where hydrodynamic separators are usually installed. The subcatchment was defined as having 80% of impervious area, 20% of pervious area, evaporation of 0.1 in/day, impervious depression storage of 0.1 inches, pervious depression storage of 0.2 inches and slope of 0.1%. The infiltration loss was calculated in SWMM using the Horton Method. The drainage area of the subcatchment, which would be an input to SWMM, was calculated assuming that the runoff peak flow rate, generated by the former New Jersey water quality design storm (uniform intensity of 0.625 in/hour), was equal to the treatment operating rate of the FloGard DVS-48, i.e., 280 gpm (0.63 cfs). Applying the Rational Method and assuming a runoff coefficient of 0.8, the subcatchment area obtained was 1.26 acres. From the SWMM modeling results, the uniform design storm generated a peak flow of 0.638 cfs in a drainage area of 1.26 acre. Such peak flow was slightly higher than the FloGard treatment operating rate. For that reason a smaller drainage area of 1.25 acre was used. This drainage area generated a peak flow rate of 0.63 cfs. Frequency Distribution of Runoff Volumes/Depths After providing SWMM with the collected precipitation data and defining the subcatchment area to study, the following steps were performed: 1. The model was run using the collected 10-year precipitation data. 2. A rank ordered report was obtained from SWMM. The report contained the magnitude of the total volume of each runoff event, the starting date and time, the duration, its exceedance frequency and the return period in years. 3. The runoff volume was divided by the area of the subcatchment in order to obtain the stormwater runoff depth.
1655 4. The runoff volumes generated by the uniform and non-uniform design storms were determined by SWMM and the correspondent depths were calculated. Since the obtained depths had almost the same values, an average design storm runoff depth was calculated. 5. The runoff depths generated by the collected precipitation data were expressed as a percentage of the average depth generated by the design storms. 6. The obtained values were ranked from the lowest to the highest in order to obtain the probability (P) of values being equal to or less than the ranked one. The probability was calculated using the Weibull plotting position formula. 7. The runoff depths, expressed as a percentage of the average runoff depth generated by the design storms, were plotted versus their probability of no-exceedance. Frequency Distribution of Runoff Peak Rates In order to obtain frequency distribution of the peak runoff flow rates the following steps were performed: 1. The runoff peak flow rate that the water quality design storm of uniform intensity would generate over the defined area was obtained using SWMM. 2. The runoff peak flow rate that the water quality design storm of non-uniform intensity would generate over the defined area was obtained with SWMM. 3. The peak runoff flow rates generated by the collected precipitation data over 10 years were simulated by SWMM. 8. A rank ordered report was obtained. The report contained the magnitude of peak flow rates of each storm event, the starting date and time, the duration, the exceedance frequency and the return period in years. 4. The runoff peak flow rates were expressed as percentages of the peak flow rates obtained from the simulated design storm events. Two sets were obtained. 5. The obtained values were ranked from the lowest to the highest in order to obtain the probability of values being equal to or less than the ranked one. The probability was calculated using the Weibull plotting position formula. 6. The runoff peak flow rates, expressed as percentages of the peak runoff rate generated by the uniformly distributed design storm, were plotted versus their probability of noexceedance. 7. The runoff peak flow rates, expressed as percentages of the peak runoff rate generated by the non-uniformly distributed design storm, were plotted versus their probability of noexceedance. Weighting Factors Figure 2 shows the frequency distribution of the runoff depth and the peak flow rates, generated by the collected 10-year precipitation data, expressed as percentages of the average runoff depth as well as the peak flow rates generated by the uniform and non-uniform design storms, respectively.
1656 Figure 2. Frequency Distribution of Runoff Depths and Peak Flows Table 1 shows the cumulative frequency distribution of the runoff depths produced by the collected precipitation data, expressed as percentages of the average runoff depth generated by the water quality design storms (0.930 inches). The same table also shows the frequency distribution of peak runoff flow rates expressed as percentages of the peak flow rates generated by the uniformly and the non-uniformly distributed water quality design storms (0.63 cfs and 2.85 cfs, respectively). The difference between two adjacent cumulative frequencies is the probability of occurrence for range between two adjacent design values, and it is defined as the weighting factor. Five weighting factors for the ranges of 0-25, 25-50, 50-75, 75-100 and 100-125 % of the design value were calculated. These ranges of flows were evaluated by the FloGard manufacturer, as required by the NJDEP lab testing procedure. Table 2 shows the obtained weighting factors. The probability (thus the weighting factor) for the range larger than 125% of the design value was added to the probability (thus the weighting factor) for the range between 100-125%.
1657 Table 1. Cumulative Frequency Distribution of Runoff Depth and Peak Flow Rates Generated by Precipitation over Ten Years (expressed as percentages of the design values) % of Design Value Runoff Depth Cumulative Frequency (%) Peak Flow Rate (uniform design storm) Peak flow rate (non-uniform design storm) 25 48.00 13.00 84.05 50 66.81 60.00 94.60 75 78.00 78.00 96.87 100 85.75 81.50 98.00 125 90.31 83.50 98.60 Table 2. Weighting Factors of Runoff Depth and Peak Flow Rates Generated by Precipitation over Ten Years (expressed as percentages of the design values) % of Design Value Runoff Depth Weighting Factors Peak Flow Rate (uniform design storm) Peak flow rate (non-uniform design storm) 25 0.48 0.13 0.84 50 0.19 0.47 0.10 75 0.11 0.18 0.03 100 0.08 0.04 0.01 125 0.14 0.18 0.02 2.4. Calculation of Weighted Removal Efficiency from Weighting Factors The FloGard weighted removal efficiency was calculated using each set of the weighting factors. Each weighting factor was multiplied by the removal efficiency measured from the laboratory tests at the corresponding flow rate (25, 50, 75, 100, and 125% of the device treatment rate, respectively). The total weighted removal efficiency is the sum of the five products.
1658 2.5. Continuous Simulation of Removal Efficiency Stormwater Runoff Simulation The stormwater runoff from two subcatchments of different sizes was simulated using SWMM. The area of each subcatchment was specified according to two sizing criteria. The first criterion considered the treatment flow rate of the FloGard DVS48 as equal to the peak flow rate generated by the water quality design storm with a uniform intensity of 0.625 in/hr. The second criterion considered the treatment rate of the FloGard DVS-48 as equal to the peak flow rate generated by the water quality design storm with a non-uniform intensity distribution. The areas were calculated using the Rational Method and assuming a runoff coefficient of 0.8. The first sizing criterion yielded a subcatchment area of 1.25 acres and the second sizing criterion yielded a subcatchment area of 0.25 acre. The subcatchments were meant to represent a typical watershed were the hydrodynamic separators are usually installed. The subcatchments were defined as having 80% of impervious area, 20% of pervious area, 0.1 in/day of evaporation, 0.1 inches of depression storage in the impervious zone, 0.2 inches in the pervious zone, and a 0.1 % slope. The infiltration loss was calculated in SWMM using the Horton Method. Water Quality Simulation The pollutant buildup was simulated in SWMM assuming a linear accumulation of the dust on the subcatchment surfaces (Huber and Dickinson 1988). According to New Jersey BMPs Manual (NJDEP 2004b), an annual accumulation of 200 pounds per acre can occur in a commercial area. Dividing this value by 365 days a daily accumulation rate of 0.55 lbs/day was obtained. This value was used in the linear dust accumulation equation. SWMM simulates the washoff process of the accumulated pollutants through the following expression: where C t = pollutant washed off at time t (pounds) P = quantity of constituent still available on the surface at time t (pounds) C w = washoff coefficient r = runoff rate (cfs) w = modeling exponent The parameters C w and w are inputs to SWMM by the user. The remaining material on the surface at the end of a time step is calculated by SWMM during the simulation.
1659 In this study, three different cases were simulated by varying the values of the parameters C w and w in order to generate concentration distributions that reflected the first flush effect (FF) during the storm event. However there is no standard quantitative definition of the first flush effect. Different authors proposed different mass-based indicators. For this reason, three pairs of coefficients were selected to generate pollutant washoff distributions that would reflect the most common definitions of the mass-based first flush effect. Case 1 simulates a FF of 80% total pollutant load transported by the first 20% of the total runoff volume (Sansalone and Chad 2004). Case 2 simulates a first flush effect of 80% of the total pollutant load transported by the first 30% of the total runoff volume (Bertrand-Krajewski, et al. 1998) and case 3 simulates a first flush effect of 80% of the total pollutant load transported by the first 40% of the total runoff volume. Among all the continuously simulated storm events, three were selected to evaluate distribution of the solids mass (solids concentration multiplied by runoff volume). These events were selected because their peak flow frequency exceedance was close to 50%, according to SWMM statistic report for subcatchment 1 (1.25 acres). That is, they are average storm events. Different exponent-coefficient combinations were tried until the desired distributions were achieved. Table 3 shows the selected coefficients for the simulation and the resulted distribution of the solids mass. Table 3. Study Cases of Pollutants Washoff Case W C w % Volume at 80% mass 1 1.6 50 20 2 1.7 50 30 3 1.9 50 40 A simulation of the removal efficiency of the device was additionally performed for the case when 80% of the mass is transported by 80% of the volume, i.e., the runoff mass is uniformly distributed through the event. The washoff process was modeled using an exponent (w) equal to 2.5 and a washoff coefficient (C w ) of 2.0. Removal Efficiency Continuous Simulation As described above, the laboratory test of the FloGard Dual Vortex showed that its removal efficiency depends on the flow rate. If the flow rate is less than the hydraulic capacity (560 gpm for the tested model DVS-48), the fitted relationship between the removal efficiency and the flow rate is assumed to be applicable. After the hydraulic capacity is exceeded, the removal efficiency is assumed to be zero since the removal efficiency of the treated part of the flow is small and the bypassed part will not receive any treatment. However, the use of positive and
1660 zero removal efficiency may not be appropriate if a significant bottom sediment resuspension occurs under the flow beyond the tested maximum flow rate. The removal efficiency could actually be negative if a severe sediment resuspension occurs during high flows that would pass through an online system. At the end of the event the total mass of pollutant that entered the device will equal the sum of the calculated influent mass during each time step of the entire storm (= the simulated flow rate multiplied by the simulated concentration and the time interval). The total mass removed from the storm water runoff will equal the sum of the mass removed during each time step of the entire storm (= the influent mass multiplied by the removal efficiency at the specific flow rate at that time step). The removal efficiency of the device for the entire event is obtained by dividing the total mass removed by the total mass that entered the device. See Fernandez (2008) for additional modeling details. 3. Results and Discussion 3.1. Calculated Weighted Removal Efficiency The weighted removal efficiencies calculated using three different sets of weighting factors are listed in Table 4. Table 4. FloGard DVS-48 Weighted Removal Efficiencies based on Frequency Distribution of Runoff Depths and Peak Flow Rates, Expressed as Percentages of Design Values. % Operating Rate Av Removal Efficiency Runoff Depth Weighting Factors Removal Efficiency Runoff Peak Rate, Uniform Design Storm Weighting Factors Removal Efficiency Runoff Peak Rate, Nonuniform Design Storm Weighting Factors Removal Efficiency 25 74.6 0.48 35.81 0.13 9.70 0.84 62.66 50 67.6 0.19 12.84 0.47 31.77 0.10 6.76 75 55.6 0.11 6.12 0.18 10.00 0.03 1.67 100 43.8 0.08 3.50 0.04 1.75 0.01 0.44 125 38.7 0.14 5.42 0.18 6.97 0.02 0.77 Total weighted removal efficiency 63.69% 60.19% 72.30%
1661 The results shown in Table 4 suggest that the weighted removal efficiency (63.69%) calculated from the weighting factors based on frequency distribution of runoff depths is close to that (60.19%) calculated from the weighting factors based on frequency distribution of peak runoff flow rates expressed as percentages of the peak runoff rate generated by the uniform intensity design storm. However, the weighted removal efficiency (72.30%) calculated from the weighting factors based on frequency distribution of peak runoff flow rates expressed as percentages of the peak runoff rate generated by the non-uniform intensity design storm is considerably larger than that normalized by the uniform intensity design storm (60.19%). It is approximately 12% larger in terms of absolute difference. The weighted removal efficiency (60.19%) calculated from the weighting factors based on frequency distribution of peak runoff flow rates expressed as percentages of the peak runoff rate generated by the uniform intensity design storm is strikingly similar to that (60%) calculated using the NJDEP-specified weighting factors (NJCAT 2007). The performance claim verified by NJCAT reads: The FloGard Dual-Vortex Hydrodynamic Separator, Model DVS-48, at a flow rate of 280 gpm (0.63 ft3/s), has been shown to have a 60% total suspended solids (TSS) removal efficiency, measured as suspended solids concentration (SSC) (as per the NJDEP methodology for calculation of treatment efficiency) using NJDEP specified material with an average d50 particle size of 70 microns, an average influent concentration of 202 mg/l and 100% initial sediment loading in laboratory studies using simulated stormwater. The weighted removal efficiency of 60% was calculated using the NJDEP-specified weighting factors. 3.2. Continuously Simulated Removal Efficiency Tables 5 and 6 summarize the results of simulated total influent and effluent mass and the correspondent removal efficiency for two different drainage areas (1.25 acres and 0.25 acres, respectively). See Fernandez (2008) for detailed model results. The SWMM continuous simulation indicates the FloGard model DVS-48 can achieve an average annual removal efficiency of approximately 70%, on a drainage area of 1.25 acres sized from the peak runoff flow rate generated by the design storm with uniform intensity. The simulation shows that the same device can achieve a removal efficiency of approximately 90% over a drainage area of 0.25 acres sized from the peak runoff flow rate generated by the design storm with non-uniform intensity. The use of uniform rainfall distribution is already a conservative criterion with the simulated approximately 70% removal efficiency that is larger than the verified 60%. The use of non-uniform rainfall distribution is even more conservative with the simulated approximately 90% removal efficiency that is much larger than the verified 60%.
1662 Table 5. Simulated Influent Mass, Effluent Mass and Removal Efficiency for the FloGard DVS-48 in Subcatchment 1 (1.25 acres) Subcatchment 1 Case Influent Mass (lbs) Effluent Mass (lbs) Removal Efficiency (%) 1 2352.78 692.09 70.58 2 2352.45 713.70 69.66 3 2352.45 765.94 67.44 Runoff mass uniformly distributed 2345.91 1042.16 55.58 Table 6. Simulated Influent Mass, Effluent Mass and Removal Efficiency for the FloGard DVS-48 in Subcatchment 2 (0.25 acres) Subcatchment 2 Case Influent Mass (lbs) Effluent Mass (lbs) Removal Efficiency (%) 1 477.61 47.45 90.06 2 477.61 48.56 89.83 3 477.60 50.73 89.38 Runoff mass uniformly distributed 474.76 121.69 74.36 For the drainage area of 1.25 acres sized with the uniform-intensity design storm and in the case where the runoff mass was assumed to be uniformly distributed within the storm event, the continuous simulation yielded a removal efficiency of 55.58%, which is reasonably close to the removal efficiency (60%) calculated using the NJDEP-specified weighting factors and verified by NJCAT. This result is consistent with the assumption behind application of the weighting factors that the runoff mass is uniformly distributed through the event. It should be cautioned that the above-described SWMM modeling results may not fully represent the field performance of the device. Size and density of the particles used in the lab testing and the lab-measured flow rate-removal efficiency relationship that was an input to the model may not accurately represent the field conditions.
1663 4. Conclusions The current NJ stormwater rules define the water quality design storm as 1.25 inches of rainfall non-uniformly distributed over two hours. This is a significant change from the previous definition of uniform rainfall distribution. The current definition leads to a higher peak runoff rate in small watersheds in comparison to the previous definition although the runoff volume remains essentially the same. The runoff frequency analysis suggests that setting design treatment flow rate equal to the peak flow rate generated by the uniform-intensity water quality design storm is more consistent with the use of existing weight factors in calculating the average annual solids removal efficiency. Use of the non-uniform intensity water quality design storm to size the treatment device is probably too conservative. The SWMM modeling of the solids removal performance also suggests that sizing the treatment device with the uniform intensity water quality design storm is more appropriate than sizing with the non-uniform intensity design storm. 5. Acknowledgements The first author was sponsored by the Fulbright Program while the research was conducted at Rutgers University. The support is gratefully acknowledged. 6. References Bertrand-Krajewski, J. L., Ghassam, C. and Agnes, S. (1998). Distribution of pollutant mass vs. volume in stormwater discharges and the first flush phenomenon. Water Research, 32(8), 2341-2356 Fernandez-Martinez, V. J. (2008). Water Quality Design Storms for Stormwater Hydrodynamic Separators, MS Thesis, Rutgers University, New Brunswick, NJ, May. Huber, W and Dickinson, R (1988). Storm Water Management Model Version 4, User s Manual. EPA/600/3-88/001a (NTIS PB88-236641/AS), US Environmental Protection Agency, Athens, GA. New Jersey Corporation of Advanced Technology (NJCAT) (2007). FloGard Dual-Vortex Hydrodynamic Separator, Verification Report, August. New Jersey Department of Environmental Protection (2003). Total Suspended Solids Laboratory Testing Procedure, December. New Jersey Department of Environmental Protection (2004a). N.J.A.C. 7:8 Stormwater Management Rules, February. New Jersey Department of Environmental Protection (2004b), New Jersey Stormwater BMP Manual, April. Sansalone, J. and Chan, M (2004) First flush concepts for suspended and dissolved solid in small impervious watersheds. Journal of Environmental Engineering, 130 (11), 1301-1314.