Quantifying the Fist-Flush Phenomenon: Eects of Fist-Flush on Wate Yield and Quality D. B. Matinson* and T.H. Thomas** *Depatment of Civil Engineeing, Univesity of Potsmouth, Potland Building, Potland Steet, Potsmouth, PO1 3AH, UK, (E-mail: bett.matinson@pot.ac.uk) **School of Engineeing, Univesity of Wawick, Gibbet Hill Rd, Coventy CV4 1AL, UK, (E-mail: t.h.thomas@wawick.ac.uk) ABSTRACT Fist-flush divesion is inceasingly ecognised as a useful intevention to educe both suspended and dissolved contaminate loads in ainwate systems. Such fist flush systems ely on the ealy ain to wash the oof befoe wate is allowed in the stoe. While thee is almost univesal acceptance that this is beneficial, thee is no ageement on just how much wate is to be diveted and the eset of the device aely consideed. In a pape deliveed at the 12 th IRCSA confeence the authos pesented a numbe of field measuements and deived an exponential decay constant fo the fist-flush phenomenon based on ainfall depth. This pape builds on these esults by applying this decay constant, and a time constant fo debis accumulation deived fom the same data, to a watebalance model. The esults show that most cuent fist-flush devices used in the field have a poo pefomance; howeve it is possible to emove up to 85% of incoming mateial while etaining 85% of the wate if the device is designed caefully. Bette mateial emoval pefomance is possible but only at the expense of lowe wate yield; similaly wate yield can be impoved by educing oveall mateial emoval. The key to good pefomance is found to be to use a slow device eset combined with a lage wate divesion, though not as lage as had been initially feaed. A design pocedue is discussed along with pactical technical constaints, possibilities and cuently available techniques. KEYWORDS Accumulation; mass balance; fist-flush; wash-o; wate quality; wate yield. INTRODUCTION Fist-flush divesion is inceasingly ecognised as a useful intevention to educe contamination in ainwate systems. Such fist-flush systems ely on the initial ainfall in a stom to wash the oof befoe wate is allowed into the main stoe. Fist flush systems have a numbe of advantages ove filtation: They ae not sensitive to paticle size, which is paticulaly impotant when the small size of oof dust is consideed They will emove dissolved contaminants as well as suspended ones, which is impotant if tace mineals such as lead and zinc ae poblematic While thee is almost univesal acceptance that this is beneficial and impessive esults have been shown fo the eectiveness of fist flush devices on wate quality in ainwate tanks (Abbott et al., 2007; Ntale and Moses, 2003), thee is no ageement on just how much wate should be diveted, o whethe such divesion should be based on volume, ainfall depth, ainfall duation o ainfall intensity. In a pape pesented at the 12 th confeence in New Delhi (Matinson and Thomas, 2005), the authos pesented a elation fo wash-o based on the exponential decay function deived by Sato
and Boyd (1972) and detemined the appopiate constants fo oof uno based on a seies of measuements of oof uno. The measuements pesented showed a wide vaiation but allowed the geneation of a simple ule-of-thumb fo fist-flush behaviou: Fo each mm of fist flush the contaminate load will halve This ule emains a useful simplification, but the inteactions between the undelying physical pocesses and equipment pefomance ae complex, and a moe detailed appoach is equied to popely design fist-flush devices. This pape descibes the esults of a seies of wate balance models used to simulate the eect of fist-flush devices on the wate quality and wate yield of a oofwate havesting system and pesents an empiical fomula and pocedue to calculate the necessay design paametes fo fistflush divetes. As space is limited the specific deivation of the equation is not detailed in this pape, howeve the behaviou discoveed is descibed and the ationale behind the deivation is pesented. The undelying detail is the subject of a jounal pape cuently in pepaation and is also descibed in Matinson (2008). The natue of the wok is necessaily mathematical, howeve it is hoped that pactitiones inteested mainly in sizing systems will find the esults useful. METHODOLOGY The flow of contaminants and wate though the system The pefomance of a fist-flush system depends on the physical pocesses involved in the accumulation of mateial on the oof and on the flow of wate and contaminating mateial o the oof and though the system. These flows inteact with fist-flush device and the stoage tank and ae summaised in Figue 1 and descibed below. Contamination emoval (L l ) Wate loss (V l ) Oveflow (V ov ) Rainfall () (V Roof FF device, ) Tank Aea (A ) Runo (V ) Design divesion ( d ) Reset time (t ) Modified uno Volume (V S ) Withdawals (V w ) Mateial Accumulation accumulation time (t a,99 ) Maximum accumulation (L max ) Runo Contamination load (L ) Figue 1: Contaminant and wate flow though a RWH system with fist-flush 1. Mateial accumulates on the oof ove time. Tank inflow Contamination load (L, ) Withdawn Contamination (L w ) 2. Duing a ainfall event, ain falls on the oof, collects some of the accumulated mateial, which mixes with the wate. Both the uno wate (V ) and uno contaminant (L ) flow into the fistflush device, the contaminant concentation educing with ainfall. 3. The fist-flush device divets a cetain amount of the ainfall depending on its volume and allows the emainde to ente the tank. The design divesion ( d ) of the fist-flush device epesents the maximum that can be flushed, howeve once the ain has stopped, a well designed divete will slowly eset ove time and so when it next ains the device may not be completely
eset. As a esult, the actual fist-flush divesion () will incease with antecedent dy peiod until the full design divesion is eached afte the complete eset time (t ). 4. Once the fist-flush device has diveted the appopiate amount, the uno wate is then allowed to ente the tank. Thus thee will be educed wate flow (V, ) and a educed contaminant flow (L, ) deliveed to the tank. Steps 3 and 4 fom the fist inteaction between physical pocesses and equipment pefomance. The most obvious is the fist-flush divesion inteacts with the change in contaminant level to poduce the eduction in contaminants enteing the tank ove the couse of a ain event. A lesse undestood inteaction is the accumulation of contaminants on the oof and the esetting of the divete which esults in changes in the level of contaminants enteing the tank fom one ainfall event to the next. Geneally, it is assumed that the eset is fast enough that the divete will completely eset befoe the next ainfall event. This is not necessaily the case, no is it desiable. 5. Finally, Wate is and contaminant is withdawn fom the tank o allowed to oveflow. This second inteaction is impotant, pimaily as the educed wate flow (V, ) inteacts with the tank volume, use demand behaviou, oveflow, and volumetic diseconomies of scale to substantially change the volumetic eiciency of the total system geneally fo the bette. Physical pocesses Mateial wash-o. The wash-o of contaminants is well descibed by the exponential decay function deived by Sato and Boyd (1972). The function is based on the assumption that the ate of emoval of mateial washed o a suface is popotional to the amount of mateial pesent on the suface and the ainfall intensity. As discussed in Matinson and Thomas (2005), the Sato-Boyd function can be simplified and stated in tems of accumulated ainfall: k w L L0 e Equation 1 Whee; L is the contaminant load emaining; L 0 is the initial contaminant load; k w is the wash-o constant (mm -1 ); is the accumulated ainfall (mm). Mateial accumulation. The accumulation of mateial on a oof between ain events has two components: Deposition of mateial Removal of deposited mateial by wind etc. An assotment of accumulation functions ae used and the most commonly applied was developed by Shaheen (1975). It consides mateial deposition to be linea and emoval to follow the same ules as fist flush: k a t L Lmax 1 e Equation 2 Whee L is the contaminate load; L max is the maximum contaminate load that can be sustained by the suface, o moe specifically the equilibium load whee the deposition and emoval pocesses balance; and k a is the accumulation constant (h -1 ). Fo simplicity, this can be expessed as an accumulation time (t a ) which is the time equied to achieve a cetain faction of L max e.g. t a,90 is the time needed to achieve a 9 of L max. Divete paametes Fist-flush design divesion. The design divesion ( d ) is the maximum ainfall a fist-flush divete is capable of emoving. In most cases, this will be when the divete has fully emptied.
Fist-flush device eset time. The device eset time (t ) is the time it takes fo the fist-flush divete to eset itself usually by emptying. The device eset may be linea, e.g using a slow elease valve o by use behaviou such as egulaly emoving a set volume of wate; based on tubulent emptying, e.g fom a weep hole in the bottom of the divete; o by lamina emptying, e.g by seepage though a poous substance. In the case of lamina emptying, the device will neve completely eset so the eset time must be consideed in the same way as accumulation time. i.e t,90 is the time needed to achieve a 9 of complete eset. Fist-flush divesion. The divesion () is the actual ainfall that a divete emoves fo a paticula ainfall event. If the antecedent dy peiod is longe than the eset time this will be the entie design divesion, if it is shote, the divesion will be less than the design divesion. Pefomance measues Fist-flush divetes change the inlet steam; educing the contaminant load, but usually also educing the wate deliveed to the tank. The moe wate that is diveted by the fist fist-flush device, the cleane the wate deliveed to the tank will be, howeve geate divesion will also mean less wate will be deliveed to the tank. Balancing these factos is key to ational fist-flush device design. Removal eiciency. The emoval eiciency ( ) of a fist-flush system is a measue of how well it emoves contaminants fom the incoming wate steam. It can simply be defined as the atio of contaminant emoved by the fist-flush system (L l ) to the total contaminant load washed o the oof (L ): Ll Equation 3 L The measue can eithe be applied ove an individual stom o ove a numbe of stoms. In this pape, the emoval eiciency is applied to the entie time seies to give the oveall pefomance of a paticula system. Volumetic eiciency. The volumetic eiciency ( v ) is a measue of how little wate is wasted by the fist-flush system. It can be measued in two places; the tank inlet ( v,i ) and the tank outlet ( v,o ). The most intuitive loss to conside is that at the inlet; howeve in eality it is the loss at the tank outlet eflecting the eduction in available withdawals that is the eal loss to the use. The eiciency when measued at the tank outlet dies significantly fom the inlet and is usually highe. Volumetic eiciency at the tank inlet ( v,i ) can be calculated by simply dividing the sum of uno afte fist-flush divesion (V, ) by the sum of the uno without divesion (V ), as the oof aea is the same fo both, the v,i can simply be calculated using the ainfall () and fist-flush divesion (): V, Equation 4 V Volumetic eiciency at the tank outlet is calculated by using V, in place of V in a mass balance and dividing the total withdawals fom the system with the fist-flush divete (V w, ) by the total withdawals fom a sepaate mass balance without fist-flush divesion (V w ) Vw, vw, Equation 5 V w
The mass balance model The pocesses descibed above wee used in a mass balance that modelled the mateial accumulation and washo, oof uno, fist-flush divesion, tank stoage and use behaviou. Moe specific detail egading the technicalities of the model and equations used can be found in Matinson (2008) The model used fifteen minutely data which was obtained fom the US National Climatic Data Cente (NCDC poduct DS3260) epesenting a numbe of climate types and ainfall pattens as shown below in Table 1. The data was chosen to eflect single wet season and bimodal ainfall distibutions in both high and low ainfall aeas. A typical tempeate climate with medium ainfall without maked seasonality was also included fo compaison. Table 1: Data souces State Town Köttek climate type Mean annual ainfall (mm) Pueto Rico Coozal Am 1 900 Texas Big Lake BSh 480 Califonia Blue Canyon Dsb 1 700 Hawaii Kekaha As 550 Rhode Island Newpot Cfa 1 200 Rainfall Patten The esults of each simulation wee a emoval eiciency and volumetic eiciency fo the design divesion and eset time selected. A seies of simulations wee caied out fo each location vaying divete paametes and othe system paametes such as use demand, demand patten and stoage volume. All volumes wee non-dimensionalised by dividing by the aveage daily uno (ADR) fom the oof and so the esults ae scalable. Each paamete was vaied sepaately fom a standad system whee the tank volume was 10 x ADR and nominal demand was 0.8 ADR. Based on the sampling epoted in New Delhi and some futhe analysis of this data, accumulation time was taken as 99% of maximum in 25 days and wash-o as halving fo each millimete. RESULTS AND DISCUSSION Eiciency tade-os and the eect of system paametes To gauge the tade-os between emoval eiciency and volumetic eiciency, the eset was set to match the accumulation and the design divesion was vaied. The emoval eiciency and the volumetic eiciency at the tank inlet and outlet wee noted and plotted. Typical esults ae shown in Figue 2. The figues show the esults fom only one location (Coozal), howeve vey simila pattens of esults wee obtained fom all locations. The esults consistently show that fist-flush divesion is moe volumetically eicient with lage tanks and with smalle demand while demand patten was found to have a negligible eect. The volumetic eiciency is also consistently geate at the tank outlet than fo the incoming steam. A paticulaly inteesting esult is shown in Figue 2c which shows that matching eset to accumulation is not the optimal solution and that a faste eset time can yield bette pefomance.
Volumetic eiciency Reset time (days) v v v 10 10 10 5 5 5 5 10 0.5 ADR demand at inlet 0.8 ADR demand at inlet 1.5 ADR demand at inlet 0.5 ADR demand with tank 0.8 ADR demand with tank 1.5 ADR demand with tank 2mm flush 7mm flush 5 10 2 ADR tank at inlet 10 ADR tank at inlet 50 ADR tank at inlet 2 ADR tank with tank 10 ADR tank with tank 50 ADR tank with tank 2mm flush 7mm flush 5 10 1 hou eset at inlet 1 day eset at inlet 25 day eset at inlet 1 hou eset with tank 1 day eset with tank 25 day eset with tank 2mm flush 7mm flush a. Vaying demand b. Vaying tank size c. Vaying eset time Figue 2: Removal vs volumetic eiciency gaphs Iteative optimisation and empiical ules Figue 3 plots the eect on volumetic eiciency esulting fom a numbe of mass balance simulations whee eset time is vaied until a taget emoval eiciency is achieved fo a cetain FF divesion. 10 25 8 20 6 15 4 10 2 5 0 10 20 design divesion (mm) 0 0 5 10 design divesion (mm) 9 emoval; inlet steam 6 9 emoval 6 emoval 9 emoval; outlet of 10 day tank 6 9 emoval; outlet of 100 day tank 6 6 emoval; inlet steam day tank 6 emoval; outlet of 10 day tank 0 day tank 6 emoval; outlet of 100 day tank Figue 3: Optimal design divesions and eset times
v,o v,o The divesion gaphs quickly ise to a peak and slowly fall with the design divesion coesponding to the peak value changing slightly with the pesence o absence of a tank and only vey slightly with tank size. Theefoe, fo a given location, optimising fo any paticula emoval eiciency esults in a vey simila design divesion egadless of othe system paametes. Thee is only a small penalty fo going above the optimum divesion poviding an appopiate eset time is used but making the device too small has a lage penalty. Vey poo eiciencies have been epoted fo small devices with quick esets fo example Gadne et.al (2004) epoted volumetic eiciencies as low as 62% with a 0.5mm divete and a eset of about 20 minutes. The eset times show a emakably linea elationship with design divesion. The similaities between optimal design divesion and the linea elation between eset time and design divesion make it possible to geneate an empiical fomula that appoximates optimal conditions. A seies of cuve fitting execises was pefomed on the optimal design divesion (in mm) ove the 5 data sets simulated and the esulting fomula is found to be: d, opt 1 1.2ln 1 Equation 6 And the eset time (in days) can be found by: 19000 1 2.4 t d 0.24e a Equation 7 Unfotunately, this optimum still equies eset times in the ode of seveal days. Devices ae available that use slow elease valves (Rain Havesting, n.d.) and seepage though poous media is also a possibility (Knight, 2005), howeve faste eset ae desiable as they ae fa less technically challenging. Futhe iteative simulation limiting the eset time to paticula values poduces the esults shown in Figue 4 which shows that lowe eset times may be used, though with some penalty in volumetic eiciency. 10 10 84.08% 91.02% 86.43% 75.22% 5 5 2 4 6 8 10 0.1 1 10 100 t 2 day 3/4 h t (hous) 8 9 Figue 4: Dieence in pefomance between a slow and fast eseting optimised fist-flush device
The penalty can be easonably significant; howeve the simplicity aoded by the faste eset may make the tade-o in pefomance desiable. This suggests a moe pactical method of optimising a fist-flush divete wheeby the longest eset technically pactical is used and the design divesion calculated fom Equation 7 which can be ewitten in tems of design divesion and with the eset time, moe conveniently expessed in hous: d a 450 000 t 1 0.24e 2.4 Equation 8 a t 2.4 Applying d 0.24e Equation 8 to small eset times of an hou does show 450 000 1 that typical design divesions of 0.5mm ae too small to be eective which is confimed by simulations of such small divesions which show they have a emoval eiciency of less than 35%. A moe sensible divesion is ove 2mm which both calculation and simulation yield a emoval eiciency of about 9, howeve the quick eset means that volumetic eiciency is only about 75% which in some situations may be poblematic. In these cases, a smalle emoval eiciency will need to be accepted o a slowe eset specified. CONCLUSION The mass balance models used in this study have shown that diveting the fist pat of the ainstom can educe the incoming contamination by 9 while deliveing 85% of the wate as measued afte the stoage tank. Geate mateial emoval is possible but with significant loss of wate. The ainfall that needs to be diveted is geneally lage than the capacity of most devices cuently being used, howeve it is not as lage as had been initially feaed in Matinson and Thomas (2005). Ideal eset times ae significantly lowe than the technically unachievable matching of eset and accumulation and futhemoe, eset times that ae elatively staightfowad can be used with a elatively small pefomance penalty. Design pocedue 1. Establish the desied emoval eiciency 2. Establish the maximum eset time technically possible 3. Use figue 4 to confim the emoval eiciency has a volumetic eiciency in an acceptable ange. Note: Figue 4a is based on the assumption of a tank that povides about 8 of the building s wate and total wate demand that is about 8 of available uno. Smalle tanks and geate demand will make the divete less eicient in wate delivey, lage tanks and smalle demand will impove the divete s wate delivey. 19000 1 2.4 4. Use t d 0.24e Equation 7 to detemine the optimised fistflush design divesion fo building. a 5. Multiply this divesion by the oof aea to obtain the volume of wate that needs to be diveted. REFERENCES Abbott, S., Caughley, B., Wad, A., Gowan, G. & Ashwoth, J. (2007) An Evaluation of Measues fo Impoving the Quality of Roof-Collected Rainwate. In Poceedings of the 13th Intenational Rainwate Catchment Systems Confeence. Sydney: Intenational Rainwate Catchment Systems Confeences. Gadne, T., Baisden, j. & Milla, G. (2004) Rainwate Fist Flush Devices Ae They Eective? In Poceedings of the Sustainable Wate in the Uban Envionment. Bisbane.
Knight, M. (2005) Design of a Fist Flush Divete fo Domestic Rainwate Havesting, Unpublished undegaduate poject epot Univesity of Wawick, Matinson, D. B. (2008) Impoving the Viability of Roofwate Havesting in Low-Income Counties, Unpublished doctoal thesis, Univesity of Wawick, Coventy. Matinson, D. B. & Thomas, T. H. (2005) Quantifying the Fist Flush Phenomenon. In Poceedings of the 12th Intenational Confeence on Rain Wate Catchment Systems. New Delhi: Intenational Rainwate Catchment Systems Association. Ntale, H. K. & Moses, N. (2003) Impoving Quality of Havested Rainwate by Using Fist Flush Inteceptos/Retaines. In Poceedings of the 11th Intenational Confeence on Rainwate Catchment Systems. Texcoco. Rain Havesting (n.d.) Fist Flush Wate Divetes (Bochue), Rocklea: Rain Havesting Pty Ltd. Sato, J. & Boyd, G. (1972) Wate Pollution Aspects of Steet Suface Contaminants, (EPA-R2-72- 081). U.S. Envionmental Potection Agency. Shaheen, D. G. (1975) Contibutions of Uban Roadway Usage to Wate Pollution, (EPA-600/2-75-004). U.S. Envionmental Potection Agency.