WAER QUALITY CAPTURE VOLUME FOR STORMWATER BMP and LID DESIGNS

Transcription

1 Guo, James C. Y. Urbonas, B. and MaKenzie K. (2012) Water Quality Capture Volume for LID and BMP Designs, HEENG- 1539, aepted for publiation on ASCE J of Hydrologi Engineering, Ot, 2012 WAER QUALITY CAPTURE VOLUME FOR STORMWATER BMP and LID DESIGNS James C.Y. Guo, Ph.D., P.E, Ben Urbonas, P.E., D.WRE, Ken MaKenzie, P.E. ABSTRACT This paper summarizes the methodology and proedure developed for determining the Water Quality Capture Volume (WQCV) for stormwater BMP and LID faility designs. WQCV is diretly related to the loal rainfall pattern, watershed imperviousness, and drain time applied to the BMP/LID faility. The performane of a BMP faility is evaluated using the rainfall-runoff ontinuous simulation that omputes the long term runoff volume-based and event-based apture ratios using the priniple of water volume balane among rainfall amount, hydrologi losses, runoff volume aptured in the BMP faility, and bypass flow. For a regional study, the proedure assoiated with the ontinuous simulation an produe optimized design values for WQCV. Typially, but not always, the optimal runoff volume and event apture ratios vary between 80 and 90%. This WQCV design and analysis proedure is more robust than the one used to estimate the WQCV s using the regression equations outlined in ASCE Manual of Pratie No. 87 and WEF Manual of Pratie No. 23. The omputer model, WQ-COSM, was also developed as a freeware for evaluating the performane of a BMP faility or produing regional design harts. The model aepts the standard hourly and 15- minute rainfall data format provided by the National Climati Data Center. Hourly data are typially available for major metro areas in the United States for a period of 20 to 60 years. Key Words: Stormwater BMP, LID, water quality apture volume, drain time, sedimentation, detention pond, retention pond. INTRODUCTION Sine 1977, the Clean Water At (CWA 1977) is the federal law the United States governing surfae water pollution ontrol. The at sets the riteria to mitigate releases of any form of pollutant (i.e., dissolved and suspended) into waters of the USA. The impat of 1977 CWA on urban planning is revealed in the development of new approahes that inlude storm water best management praties (BMP) and low impat development (LID) designs (EPA Reports 1983, 1986). Correspondingly, urban drainage systems are expanded into a 3-layer asading flow system that inlude a miro system designed for water quality enhanement, a minor system designed for dealing with 2 to 5-yr flooding water through street gutters, and a major system sized to mitigate the 50- to 100-yr extreme events through natural waterways, man-made hannels, streets, detention systems, et. Figure 1 presents an example illustrating how this 3-layer asading flow system is blended into a street layout. A miro system is sized to apture and to treat the runoff volumes generated from the diretly onneted impervious areas under small, frequently ourring rainfall events while the minor and major systems follow the onventional approah to ope with the 2- to 100-yr and larger extreme events. During a signifiant storm event, the miro-system basin interepts the initial storm water runoff volumes from impervious surfaes immediately upstream. What is not interepted overtops the basin and flows to the street gutter(s), downstream storm sewers, hannels and reeiving waters. Many reative onepts for stormwater BMP and LID have been onverted into miro system designs using sub-base filtering and infiltrating proesses, inluding sand filters, infiltration basins, rain gardens, bio-swales, retention ponds, wetlands, extended detention basin et. (EPA National Menu for Stormwater BMP 2011). 1

2 Figure 1 Casading Flows in Street Drainage System A ommon pratie in storm water management is to inorporate stormwater quality BMPs and their water uality ontrol basins (WQCB) into flood mitigation systems. In doing so, the operation of suh a urban drainage system provides a full-spetrum ontrol that range from small to extreme rainfall events (Guo 2009, Guo and Cheng 2008). The hallenge in the design of stormwater full-spetrum ontrol system lies in how to selet the design events for the miro, minor, and major levels of protetion. The rudimentary question is how big a BMP faility is big enough to treat storm water while at the same time how small to make it to be ost effetive. This tradeoff presents a new hallenge to stormwater professionals. Reommendations on how to size stormwater BMP failities vary greatly, inluding the 1-year 24-hour storm, the 80 to 95 th perentile storms, the runoff depth of 25.4 mm (one-inh) from the new development areas et. (International BMP Database, 2010). Similarly, there are several empirial and regulatory methods in use for the seletion of WQCB storage volume, varying from a high standard to apture the runoff depth of 76 mm (3 inhes) for LID designs in the Washington D.C. areas (UFC-LID 2004) to the ost effetive approah using water quality apture volume (WQCV) for BMP designs in the Denver region (UDFCD Vol ). Although these empirial methods appear inonsistent, they all agree that the WQCV shall be in the same magnitude as the street flush volume or the 3 to 6 month event. As a result, there is an urgent need to have proedures that an systematially relate the field experiene of urban runoff volume to loal rainfall patterns. For design events with a return period of less than one year, approahes using annual data series are not appliable. As a result, it is neessary that the rainfall analysis for BMP/LID designs shall swith from extreme value methods to the analyses of omplete data series. In 1996, a long-term event-based simulation sheme was derived to selet the WQCV based on the optimized runoff apture ratio (Guo and Urbonas 1996). Using 30 to 40 years of ontinuous rainfall reorded in the Cities of Seattle WA, Saramento CA, Phoenix, AZ, Denver, CO, Cininnat OH, Tampa, Fl, Boston MA, and Little Rok, AK, a set of regression equations were derived and then reommended for sizing stormwater BMP failities (ASCE Manual, 1998). Although these empirial formulas were further examined by the statistial model using the exponential distribution, it is lear that the event-based simulation approah does not best represent a ontinuous runoff time series flowing through the proposed BMP/LID faility (Guo and Urbonas 2002a and 2002b). In this study, a rainfall-runoff ontinuous simulation tehnique was developed to first onvert a long-term ontinuous rainfall series into a ontinuous runoff series, and then to route the inremental runoff depths as a time series through the BMP/LID faility to alulate the long term runoff apture ratios. As expeted, the more the apture volume suh a faility has, the higher the runoff apture ratio will be. The optimal design magnitude for BMP failities an then be derived using the priniple of diminishing returns. This proedure has been oded into a omputer model: Water Quality Capture Optimization and Statistis Model (WQ-COSM) whih is a freeware available at and 2

3 CONVERSION OF RAINFALL INTO RUNOFF VOLUME Rainfall data are often reorded at a rain gage as a point value. When the tributary area is small (<25 square km), the point rainfall depth an represent the soure of runoff without a depth-area adjustment (NOAA Hydro-40, 1984). A ontinuous rainfall reord is omposed of individual events separated by a dry period. A rainfall event separation time of six hours is reommended to identify individual rainfall events (Drisoll et al. 1989). With a pre-seleted rainfall event separation time, a 30 to 40-year rainfall reord an be divided into thousands of individual events. For eah single event, its rainfall duration is divided into several time steps aording to the pre-seleted time inrement used in the numerial simulations as: T ( N ( (1) t Where T(= duration for i-th rainfall event, N(= number of time steps, and Δt= time inrement suh as one hour or 15 minutes for eah time step used in omputations. The hydrologi abstration for eah event inludes soil depression and infiltration losses. Soil infiltration rates an be quantified using the rational method or Horton s soil infiltration formula, depending on the modeling onveniene (WQ-COSM 2011). If the lumped model is preferred, the runoff oeffiient an be hosen based on the watershed imperviousness perent. The inremental soil infiltration loss is estimated as: F ( (1 C)[ p( pi ] where 1 j N ( (2) Where ΔF(=j-th inremental soil infiltration loss during i-th rainfall event in depth per watershed area, C= runoff oeffiient seleted based on watershed imperviousness perentage p(= inremental rainfall depth at j-th time step, and p i = inipient runoff depth suh as 2.0 to 2.5 mm as reommended (EPA report 1986). If the detailed modeling tehnique is preferred, the storm runoff from a watershed an be divided into two independent flows from the pervious and impervious sub-areas respetively. Eah sub-area has its own depression and infiltration losses alulated at eah time step as: F( [ f ( f o f ) e ( jt ) ] t where 1 j N( (3) Where f = final infiltration rate, f o =initial infiltration rate and α= soil infiltration deay fator. During a ontinuous simulation proess, the reovery of these parameters used in the loss funtions is linearly modeled as a yle over a user defined period suh as 1 to 14 days, depending on the loal limate ondition (Rossman 2005). For instane, in arid limates a drying time of 1 to 3 days would be appropriate, while 7 to 14 days would be more appropriate in a limate with ool temperatures and muh rainfall. Eq 3 desribes how the soil olumn beomes saturated over time. Integration of Eq (3) represents the soil anteedent moisture defiit as: F( f ( f jt f o ) ( jt ) [1 e ] where 1 j N( (4) Where F(=infiltration amount at j-th time step during i-th event in depth per watershed. The numerial proedure for modeling hydrologi losses is to firstly fill up the anteedent soil moisture defiit (EPA SWMM 2011). As soon as the aumulated rainfall amount exeeds Eq (4), the depression loss begins to be filled up. Numerially, the array of inremental depression loss is determined step by step through the early inremental rainfall depths as: d( 1) max{0, min[( p( 1) F( 1)), D]} (5) 3

4 d( max{0, min[ k j k1 p( k) F( k j1 k1 d(, D k j1 k1 d( k)]} for 2 j N( (6) Where d(= j-th inremental depression loss during i-th event in unit depth per watershed area, d(1)= initial depression loss, p(1)= first inremental rainfall depth in i-th event, F(1)= initial infiltration loss, k= k-th time step to fill up depression loss, and D= watershed depression loss suh as 10 mm (0.4 inh) for pervious area and 2.5 mm (0.1 inh) for impervious area (UDFCD 2001). Eq (5) defines the initial inremental depression loss. Under a mild rainfall event, p(1)<f(1), Eq (5) leads to d(1)=0 while an intense event, Eq (5) may result in d(1)=d. Likely, d(1)lies between these two extreme onditions. Eq 6 generates the array of inremental depression losses, d(, after the anteedent soil moisture defiit has been satisfied. After having the depression storage areas been ompletely filled up, the inremental depression loss for eah time step is redued to zero; and the inremental runoff depth will be produed as: k j k j r( max[0, p( k) F( d( k)] k1 k1 for 1 j N( (7) in whih r(=j-th inremental runoff depth for i-th rainfall event. Eq 7 implies that the watershed surfae remains dry until the aumulative rainfall depth beomes greater than the sum of depression and infiltration losses. RUNOFF ROUTING THROUGH A WQCB The key fator for flow routing through a WQCB is the drain time. For detention and retention BMPs, the seletion of drain time is losely related to the sedimentation proess in the WQCB. If sediment harateristis are not known, a 12- hour settling time is reommended for a wet WQCB and a 24-hr settling time is reommended by Waugh, et al (2002) for a dry WQCBs. At the same time UDFCD (2011) reommends a 40-hr drain time for dry extended detention basins. As a result, the drain time for an extended detention basin an typially range from 24- to 48-hr, depending on the loal design riterion. Drain times for a surharge WQCB nested in a retention system typially range from 12 to 24 hr (UDFCD 2011). Drain times of 12 hr are reommended for LID designs suh as rain gardens and infiltration swales, while drain times of 24 hr are reommended for sand filters (Urbonas, 1999). With the pre-seleted storage volume and drain time, the average release rate for the proposed WQCB is alulated as: P p q (8) Td Where q= average release rate in depth per time, P p = WQCB s brimful storage volume in depth per watershed area, and T d = drain time. Throughout the i-th event, the basin is loaded with the inremental runoff depth for eah time step as: P ( 1) min[0, P ( i 1) r( 1) qt] v (9) o P ( min{0, min[ P, P ( j 1) r( qt]} where 2 j N( v p v (10) Where P v (= aumulated storage volume in WQCB, and P o (i-1)= residual volume from the previous event. Eq (9) sets the i-th initial ondition for flow routing through the WQCB. Eq (10) warrants that the maximum aumulated storage volume does not exeed the basin size and the minimum volume is not less than the dry ondition. After the rainfall event ends, the basin ontinues being drained as: 4

5 k k 1 k1 P0 ( max[0, Pv ( N( ) kqt ] (11) Numerially, Eq (11) will be operated till the WQCB beomes emptied or the next event omes, whihever omes first. Between two events, all hydrologi loss parameters are refreshed through the user-defined reovery yle. During the i-th flow routing proess, the WQCB either had a omplete intereption if the event produes a runoff volume less than the WQCB s maximum apaity or the WQCB is overtopped with an untreated or partially treated bypass flow. The maximum runoff treatment apaity in a WQCB is no more than its brim full volume plus the water volume flowing through the WQCB during the storm event as: P ( P q T( m (12) p Where P m (= maximum runoff treatment apaity for i-th event in depth per watershed area. R ( min{ P m (, j N ( j1 [ r( q t ]} 1 j N( (13) Where R (= treated runoff volume in depth per watershed and p o (= initial water volume at the beginning of i-th event. Eq 13 means that the runoff apture volume for i-th rainfall event is either the runoff volume flowing through the WQCB or the maximum WQCB s treatment apaity, whihever is smaller. Applying Eq (13) to the entire rainfall reord, the overall runoff volume apture ratio (RVCR) is alulated as: R V i1 i M jn ( i1 im R ( j1 r( (14) Where R v = runoff volume apture ratio (RVCR), and M= number of individual rainfall events in the ontinuous rainfall reord. Eq (14) gives a fair assessment on the WQCB s performane if the rainfall reord is not skewed with extreme events. In fat, it is likely that several large events in the data base an numerially dominate the outome from Eq (14). To avoid a bias analysis, the runoff event apture ratio (RECR) is also developed to ount the number of events that were ompletely aptured with no bypass flow. For the entire rainfall reord, the overall RECR is defined as: m( 1 if j N ( j1 [ r( q t] P m ( or m( 0 (15) R V i M m( i 1 M (16) Where m( = ounter for i-th event, M= total number of events in reord, and R v = runoff event apture ratio (RECR). 5

6 A WQCB is designed to apture and to treat runoff from the small urban runoff events and relatively small athments. During Runoff volume will be aptured and treated up to WQCB s apaity and then overtopping will our. Usually, the overtopping flows arry diluted or partially treated sediment loads. In pratie, the RECR is more meaningful than the RVCR. This is beause the bulk of urban pollutants, inluding sediment are transported by frequent, small events, into natural water bodies. The design of a miro system is best aimed at the targeted water quality enhanement rather than the flood mitigation provided in the minor and major systems to deal with extreme storm events that reate signifiant drainage and flooding problems, but it would be prohibitively expensive to apture and treat. DERTERMINATION OF WQCV FOR BMP DESIGNS To onstrut a design hart for WQCV at a site, repeat the above-desribed proedure for a range of WQCB storage volumes. Both RVCR and RECR an be alulated as a rising urve as the storage volume inreases. Figure 2 is an example derived for the Denver region, Colorado. The watershed imperviousness for Figure 2 is set to be 50%. The drain time of 12 hours is used to generate 1869 individual events from the 40-year ontinuous hourly rainfall data reorded at the Stapleton Airport rain gage. The soil losses are desribed with f 0 = 76 mm/hr (3 inhes/hr), f =12.5 mm/hr (0.5 inh/hr) and α= 3 per hour, and D=10 mm for pervious area, and 2.5 mm for impervious area. The distribution of RECR is alulated and plotted in Figure 2. The x-axis represents various basin storage volumes normalized by the index rainfall depth. The index rainfall depth is hosen as a ut-off value to exlude very large events in the rainfall data base (EPA SWMM5). In general, the 99.5th perentile rainfall depth an serve this purpose. Both RECR and RVCR represent the non-exeedane probability urve that inreases from zero to unity. Before the break-even point as shown in Figure 2, the urve begins with sharply inreasing returns and then ends with flat diminishing returns. As a result, the breakeven point is reommended as the water quality apture volume (WQCV) for the design of stormwater BMP and LID basins. Figure 2 WQCV Developed for Denver Area with 50% Imperviousness and 12-hr Drain Time 6

7 WQCV depends on watershed s imperviousness and WQCB s drain time. For a seleted drain time, the same proedure used in Figure 2 an be repeated for a range of imperviousness perentages. As shown in Figure 3, the RECR-based design hart is developed for Denver s WQCV using 12-, and 24-hr drain times. The average RECR is 0.87 for the 12-hr urve and 0.85 for the 24-hr urve. Of ourse, similar WQCV design urves an be produed for any metropolitan areas using the omputer model: WQ-COSM. Basin Size in m per watershed RECR based WQCV for Denver Area, Colorado % 20% 40% 60% 80% 100% Watershed Imperviousness Perent 12 hr 24 hr Figure 3 Denver s Water Quality Capture Volumes Developed for 12- and 24-hr Drain Times CONCLUSION Like many BMP projets, this researh study has been a ontinuous effort, starting in Extensive rainfall and runoff field data were analyzed over the last twenty five years under supports of Urban Watersheds Researh Institute and Urban Drainage and Flood Control Distrit, Denver, Colorado. This paper presents a proedure to derive the loalized Water Quality Capture Volume (WQCV) for stormwater BMP and LID designs. The latest development in the ontinuous rainfall-runoff modeling tehniques modifies the regression formulas and design harts outlined in ASCE Manual of Pratie No. 87 and WEF Manual of Pratie No. 23. Although the rainfall event separation time is reommended to be 6 hours (Drisoll et al, 1989), the loal design requirement may have a different riterion. From the operational point of view, set the rainfall event separation time equal to WQCB s drain time to warrant that flow routing begins with an empty basin. When sedimentation is the primary pollutant removal mehanisim, WQCB s drain time is often seleted to be not shorter than the settling time to allow the targeted solids to be effetively trapped. Derivation of WQCV is a proedure to relate the urban runoff volume to the loal rainfall pattern and then routing this runoff though a WQCB. WQCV depends on watershed s imperviousness and the BMP/LID basin s brim-full drain time. The longer the drain time is, the higher the WQCV will be. The apture ratio at the point of diminishing returns typially, but not always, varies between 80 and 90% for both the event-based or 7

8 volume-based approahes. At the same time, the RVCR-based WQCV an demand signifiantly more storage volume than the RECR-based WQCV while the added environmental benefits of the inreased size of the BMP/LID faility have not been yet determined and appear to be of marginal value. This is beause urbanization impats on geomorphi and eologial response in reeiving waters are losely related to the inreased frequeny of small runoff events, rather than the large events that have produed flooding flows before urbanization. The lengthy omputational proedure has been oded into a window-based omputer model: Water Quality Capture Optimization and Statistis Model (WQ-COSM). WQ-COSM is a freeware available at and at The model aepts the standard hourly and 15-minute rainfall data format provided by the National Climati Data Center ( Hourly data are typially available for most muniipal areas in the United States for a period of 20 to 60 years and 15-minute data for most major metro areas for periods of 15 years. The omputer model, WQ-COSM, is an effetive tool that provides a onsistent basis to develop site or region-speifi WQCV analyses and design harts using the loal rainfall reords. REFERENCES ASCE Manual of Pratie No. 87 and WEF Manual of Pratie No. 23, (1998). Urban Runoff Quality Management, Amerian Soiety of Civil Engineers, New York. CWA (1977). Clean Water At, Drisoll, E.D., Palhegy G.E., Streker, E.W. and Shelley, P.E. (1989). Analysis of Storm Events Charateristis for Seleted Rainfall Gauges Throughout the United States. U.S. Environmental Protetion Ageny, Washington, D.C. EPA Report (1983). Results of the Nationwide Urban Runoff Program, Final Report, U.S. Environmental Protetion Ageny, NTIS No. PB , Washington, DC, EPA Report (1986). Methodology for Analysis of Detention Basins for Control of Urban Runoff Quality, U.S Environmental Protetion Ageny, EPA440/ , September. EPA National Menu of Stormwater Best Management Praties (2011), EPA SWMM (2011), Stormwater Management Model supported by EPA. Guo, J.C.Y.(2009) Retrofitting Detention Basin for LID Design with a Water Quality Control Pool, ASCE J. of Irrigation and Drainage Engineering, Vol 135, No 6, Otober. Guo, J.C.Y. and Cheng, J. Y.C.(2008) Retrofit Stormwater Retention Volume for Low Impat Development (LID), ASCE J. of Irrigation and Drainage Engineering, Vol 134, No 6, Deember. Guo, J.C.Y. (2002a). Overflow Risk of Storm Water BMP Basin Design, ASCE J. of Hydrologi Engineering, Vol 7, No. 6, Nov. Guo, J.C.Y. and Urbonas, B. (2002b). Runoff Capture and Delivery Curves for Storm Water Quality Control Designs, ASCE J. of Water Resoures Planning and Management, Vol 128, Vo. 3, May/June. Guo, J.C.Y. and Urbonas, B. (1996). "Maximized Detention Volume Determined by Runoff Capture Ratio, ASCE J. of Water Resoures Planning and Management, Vol 122, No 1, Jan. 8

9 Roesner, L., Urbonas, B, and Guo, J.C.Y (1996). "Hydrology for Optimal -Sizing of Urban Runoff Treatment Control System", J. of Water Quality International, London, SW1H9BT, UK, February. International BMP Database (2010). User s Guide: Stormwater Best Management Praties (BMP) Database supported by Water Environment Researh Foundation, Federal Highway Administration, Environment and Water Resoures Institute of the Amerian Soiety of Civil Engineers, U.S. Environmental Protetion Ageny, Amerian Publi Works Assoiation, Release Version 3.2, August. NOAA Hydro-40 (1984). Depth-Area Ratios in the Semi Arid Southwest United Statee, NWS NOAA TM Hydro- 40, US Department of Commere, Silver Spring, Maryland. Rossman, L. A. (2005) Storm Water Management Model User s Manual. Version 5, Water Supply and Water Resoures Division, National Risk Management Researh Laboratory, Cininnat OH. UDFCD (2001). Urban Stormwater Design Drainage Manuals, Vol 1 and 2, published by Urban Dainage and Flood Control Distrit, Denver, Colorado. UDFCD (2011). Urban Stormwater Design Drainage Manuals, Vol 3, published by Urban Drainage and Flood Control Distrit, Denver, Colorado. UFC-LID (2004). Unified Faility Criteria for Low Impat Development, Report: UFC , published by Department of Defense, Washington D.C., USA, Otober. Urbonas, Ben R. (1999). Design of a Sand Filter for Stormwater Quality Enhanement, Water Environment Researh, Volume 71, Number1, Jan-Feb, Alexandria, VA. Waugh, P.D., Jones, J.E., Urbonas, B.R., MaKenzie, K.A., and Guo, J. C.Y. (2002) Denver Urban Storm Drainage Criteria Manual., ASCE J. of Urban Drainage, Vol 112, No 56. WQ-COSM (2001), User s manual for Water Quality Capture Optimization and Statistis Model (WQ-COSM), APPENDIX II C= runoff oeffiient D=watershed depression loss d(= j-th inremental depression loss in unit depth per watershed area, d(1)= initial depression loss, f = final infiltration rate, f o =initial infiltration rate F(=infiltration amount at j-th time step during i-th event in depth per watershed F(1)= initial infiltration loss, k= k-th time step M= number of individual rainfall events in the reord m( = ounter for i-th event, N(= number of time steps fro i-th rainfall event, P p = WQCB s brim full storage volume in depth per watershed area, P m (= maximum runoff treatment apaity for i-th event p i = inipient runoff depth 9

10 p(= inremental rainfall depth at j-th time step, p(1)= first inremental rainfall depth in i-th event, P v (= aumulated storage volume in WQCB, P v (= aumulated storage volume in WQCB, P o (i-1)= residual volume from the previous event q= average release rate in depth per time, R (= treated runoff volume in depth per watershed R v = runoff volume apture ratio (RVCR) R v = runoff event apture ratio (RECR). r(=j-th inremental runoff depth for i-th rainfall event T d = drain time. T(= duration for i-th rainfall event, α= soil infiltration deay fator. Δt= time inrement used in omputations. ΔF(=j-th inremental soil infiltration loss during i-th rainfall event in depth per watershed area 10