tracers in subsurface upflow wetland

8/13/2019 Tracers in subsurface upflow wetland

1/9

ORIGINAL PAPER

A tracer study for assessing the interactions between hydraulicretention time and transport processes in a wetland systemfor nutrient removal

Ni-Bin Chang Zhemin Xuan Martin P. Wanielista

Received: 14 December 2010 / Accepted: 22 July 2011 / Published online: 11 August 2011 Springer-Verlag 2011

Abstract In this study, a new-generation subsurface up-ow wetland (SUW) system packed with the uniquesorption media was introduced for nutrient removal. Toexplore the interface between hydraulic and environmentalperformance, a tracer study was carried out in concert witha transport model to collectively provide hydraulic reten-tion time (7.1 days) and compelling evidence of pollutantfate and transport processes. Research ndings indicate thatour pollution-control media demonstrate smooth nutrientremoval efciencies across different sampling port loca-tions given the appropriate size distribution conversantwith the anticipated hydraulic patterns and layered struc-ture among the sorption media components. The sizablecapacity for nutrient removal in this bioprocess conrmsthat SUW is a promising substitute for an extension of traditional on-site wastewater treatment systems.

Keywords Constructed wetlands Tracer study Transport model Biogeochemical science

Introduction

Constructed wetlands have been widely applied for avariety of wastewater treatments because of their lowconstruction, operation, and maintenance costs, excellentpollutant removal performance, and lower energy con-sumption. As promising sustainable infrastructure alterna-tives for wastewater treatment in the twenty-rst century,

constructed wetlands can remove pollutants with the aid of the complex interactions among water, substrate, plants,and microorganisms through a variety of physical, chemi-cal, and biological mechanisms. Physical functions aremainly ltration by the substrate layer, plant-root zone, andsedimentation. Chemical reactions consist mainly of chemical precipitation, adsorption, ion exchange, and oxi-dationreduction (REDOX) reactions. Biochemical reac-tions mainly refer to microorganisms degrading andremoving pollutants in aerobic, anoxic, and anaerobicconditions. The performance of constructed wetlands relieson many factors. In addition to wetland dimension, wasteloading and other external conditions, hydraulic retentiontime (HRT) and ow patterns have a signicant inuenceon removal efciency. A tracer study is the most directapproach to obtain the information of internal hydraulics inconstructed wetlands.

Tracer studies are used to determine the direction andvelocity of water movement by monitoring the matter orenergy carried by water. They can also indirectly identifyother hydraulic parameters such as dispersivity, porosity,and hydraulic conductivity through further analyses. Anideal tracer should be representative, meaning it followsthe same path as the water and is easily detected, inex-pensive to analyze, and has low toxicity, high solubility,and low background concentration in the natural environ-ment. The three most popular choices for tracers are iso-topes, ions, and dyes. Isotope technology is a common toolused to determine the fate and transport of isotopic nitrogenin nutrient treatment. Kadlec et al. [ 1] used the stableisotope 15 N introduced as ammonium in two subsurface-ow (SSF) wetlands receiving primary meat processingwater and found little of the tracer in gas emissions(* 1%). The majority of the tracer was found in plants(648%) and sediments (2837%). Ronkanen and Klve

N.-B. Chang ( & ) Z. Xuan M. P. WanielistaDepartment of Civil, Environmental, and ConstructionEngineering, University of Central Florida, Orlando,FL 32816, USAe-mail: [email protected]

1 3

Bioprocess Biosyst Eng (2012) 35:399406DOI 10.1007/s00449-011-0578-z
http://-/?-http://-/?-


2/9

[2, 3] used the stable isotope 18 O/ 16 O ratio and tracer teststo help prevent or reduce short-circuiting and dead zones inpeatlands. Although isotope technology has high accuracy,it is expensive. Ionic compounds, especially bromide, havebeen widely used as groundwater tracers [ 4, 5]. Maos-zewski et al. [ 6] used instantaneously injected bromide toevaluate hydraulic characteristics for a duckweed pond inMnio w, Poland. The tracer study result showed that thewastewater owed along three different ow-paths to theexit. But the concentration of bromide in natural watersmay be present well above detection; thus, large quantitiesof ionic tracers would be required to surpass the back-ground levels. Another shortcoming is that, bromide istypically analyzed in a laboratory; portable probes avail-able for direct measurements in the eld are less reliable. Incomparison, dyes have advantages of low detection limits,zero natural background, and relatively low cost. One of the most popular dyes is Rhodamine WT (RWT). Dierbergand DeBusk [ 7] compared RWT and lithium chloride(LiCl) for their suitability as hydraulic tracers in wetlandsand found that both tracers returned more than 95% of theinjected amount in submerged aquatic vegetation-domi-nated mesocosms. They concluded that although RWT wasnot as stable as LiCl at initial concentrations less than60 mg/L, the reduction in the recoveries did not affect theaccuracy of key hydraulic parameters. Lin et al. [ 8] per-formed hydraulic tracer tests in California, USA, to eval-uate the suitability of RWT as a tracer for wetland studiesand to determine HRT distribution of the wetlands. Theyevaluated the performance of RWT by comparing thebreakthrough curve (BTC) of RWT to that of bromide in apilot-scale test and found that the BTCs of RWT andbromide were equal. Giraldi et al. [ 9] used RWT as asquare input signal to develop their numerical wetlandmodel.

In addition to a tracer test that is able to track down theoverall hydraulic pathways, another important consider-ation is the uid dispersion function to help illuminate localtransport behavior in the constructed wetlands. Two theo-retically ideal reactors, the plug ow reactor (PFR) andcontinuously stirred tank reactor (CSTR), represent the twoextreme ow patterns, the no mixing in PFR versusperfect mixing in CSTR. Zero mixing of PFR will resultin an unchanged tracer spike owing out of the system afterone nominal HRT. In a CSTR reactor, however, the traceris assumed to uniformly distribute throughout the entirecell immediately after entering the system. As a result, anexponentially declining tracer output curve will beobserved at the outlet. Obviously, constructed wetlands areneither PFR nor CSTR, but somewhere in between. Aparticular skewed bell-shaped response curve observed inconstructed wetlands can be possibly depicted by a tanks-in-series (TIS) model [ 10]. This type of model provides a

more exible way to simulate the hydraulic performance inconstructed wetlands with varying internal structures. Inaddition to TIS, three other ow models, including plugow with dispersion [ 11 , 12], one-dimensional transportwith inow and storage [ 13, 14], and computational uiddynamics [ 1517] also appear in the wetland literature.With less complexity, the hydraulic parameter number of tanks in series model (NTIS), an extension of the TISmodel, is deemed as the most popular ow model in thiseld.

A thorough review of the HRT determination and TISquantication for free water surface (FWS) wetlands andhorizontal subsurface ow (HSSF) wetlands can be foundin Kadlec and Wallace [ 18]. However, little work has beenpublished to estimate the parameter number of TIS insubsurface upow wetland (SUW), especially from aglobal versus local perspective. The objectives of thisstudy are thus to (a) conduct a tracer study to identify theactual HRT and the ow patterns based on the retentiontime distribution (RTD) function in an SUW for treatingon-site wastewater, and (b) understand the interactiverelationships of ow patterns and transport processesbetween sub-cells due to the varying effect of the mediapermeability. The results will generate advanced insightsinto hydraulic dispersion in relation to biogeochemicalprocesses to bridge the gap between hydraulic design andbiogeochemical science in the knowledge base of con-structed wetlands.

Materials and methods

Site description and experimental design

A performance-based and passive constructed wetlandsystem (i.e., an SUW) following a septic tank was createdat University of Central Florida for nutrient removal in2008. Four parallel 1.52 m wide 9 3.05 m long 9 0.91 mdeep cells were constructed in the test bed, three withdifferent plant species and one with no species to serve asthe control case (Fig. 1) [19]. For the purpose of ground-water protection, each cell was encased from surroundingsoil by a wooden frame containing an impermeable linerat bottom, a gravel substrate (30.48 cm thick), fabric

Outflow

Inflow

G Medium

PC Medium

Gravel

Sand

Fig. 1 Prole view of wetlands [ 19]

400 Bioprocess Biosyst Eng (2012) 35:399406

1 3
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


3/9

interlayer (DuPont TM Weed Control Fabrics), a 15.24 cmsand layer, a 30.48-cm thick pollution control media (50%Citrus grove sand, 15% tire crumbs, 15% sawdust and 20%lime stone; PC media hereafter) layer, growth media (75%expanded clay, 10% vermiculite and 15% peat moss; Gmedia hereafter), and selected plants (Fig. 1). The tracerexperiment was conducted in one of the four wetland cellswith Canna planted for the enhancement of nutrientremoval because Canna outperformed the other two cellsplanted with blue ag and bulrush [ 19]. All three plantedcells exhibited better removal efciencies than the controlcell with no plant species [ 19]. The upstream septic efuentwas the inow for nal treatment at the selected wetlandcell, and the nutrient removal efciencies were measuredbased solely on the inow and outow at the wetland cell.After completion of nutrient removal testing, the same cellwas used to support a tracer test for approximately 2 moreweeks. The hypothesis of wetland hydraulics in this studyis that the gravel layer will be fully saturated rst due to itslarger pore space. The water level will then rise graduallyand ow out of the cell after passing through the sand andPC media layers. Two customized oxygenators (PVC pipewrapped with fabric at bottom) introduced oxygen into thegravel layer to enhance the nitrication at the bottom of thewetland cell [ 19]. The samplers in each wetland cell wereinstalled at 33, 67, and 100% along the longitudinaldirection of the wetland. Vertically, the samplers werelocated at the interface between the different layers(Fig. 1).

Nutrient analysis

Water quality in the wetland system was monitored on aweekly basis from September 2 to 30, 2009. A 24-hcomposite sample (a representative sample obtained bycombining multiple samples at regular intervals) was takenat every sampling port in proportion to the actual ow loadduring the 24-h period [ 19]. Dissolved oxygen (DO), pH,and temperature were measured on site by HACH HQdeld case [ 19]. In addition to parameters requiring a grabsample analysis, ammonia-nitrogen (NH 3 -N), nitrite-nitro-gen (NO 2 -N), nitrate-nitrogen (NO 3 -N), organic nitrogennitrogen (ON-N), total nitrogen (TN), soluble reactivephosphorusphosphorus (SRP-P), and total phosphorus(TP) were measured by an external certied laboratory[19]. The containers used for nutrient sampling wereimmediately sealed, labeled, and measured by the externalcertied laboratory within the same day.

Tracer study

Due to the advantages of low detection limits, zero naturalbackground, low cost, and easy operation, RWT was

selected as the water tracing dye to determine the hydraulicpatterns and HRT of the SUW system. RWT, a syntheticred to pink dye having brilliant uorescent qualities, amolecular formula of C 29 H29 N2 O5 ClNa 2 , and a CASNumber of 37299-86-8, is also known as Acid Red #388. Itis often used as a tracer in water to determine the rate anddirection of ow and transport. In our study, the RWTliquid (20% solution) was purchased from Keystone Ani-line Corporation. An RWT solution with 0.04 g activeingredient was injected into the pipe before the inlet of theSUW. The tracer HRT was then calculated followingHeadley and Kadlecs practical guide [ 23]. Ten sets of water samples collected from nine sampling ports weremeasured in July 2010. Water samples of 50 mL werecollected from each sampling port by using a peristalticpump. The grab samples with the Rhodamine dye weremeasured by Aquauor TM (Turner Designs 998-0851)handheld uorometer. The linear detection range for bothdyes is 0.4300 PPB (active ingredient). Because RWTuorescence is susceptible to photolysis and sensitive totemperature, samples should be collected in glass bottlesand kept in the dark prior to analysis. In addition, thesolution with known concentration was analyzed on sitefor calibration prior to the sample measurement. Theminimal root mean square error divided by the peak height of the distribution was evaluated to determine thegoodness of t for the tracer study. Eventually, the RWTdistribution was demonstrated by the 3-dimensional(3D) data visualization software package, Voxler (Goldensoftware, Inc.).

Dispersion model

For an SSF wetland, the nominal HRT can be obtainedusing Eq. 1:

sn V nQ

/ l w hnQ

; 1where sn = nominal HRT (days), l = wetland length (m),w = wetland width (m), h = wetland depth (m),V n = nominal wetland volume (m 3 ), Q = ow rate (m 3 /d),and / = porosity (unitless).

Danckwerts [ 20] rst extensively analyzed RTD, whichis generally determined by introducing an impulse of tracerand measuring the time that the tracer spends at the wet-land interior. For an impulse tracer addition into a steadilyowing system, the tracer HRT (often referred to as thetracer HRT or mean HRT ( s), can be achieved by theEq. 2. A detention time distribution (DTD) is commonlyused to represent the time that various fractions of waterspend in the wetlands (Eq. 3) [18]. The variance ( r 2 ) is thesquare of the spread of the distribution, expressed in unitsof (time) 2 (Eq. 4):

Bioprocess Biosyst Eng (2012) 35:399406 401

1 3
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


4/9

s R 10 tC t dt R 10 C t dt ; 2

f t C

R 10 C dt ; and 3

r 2

R 10 t s

2C t dt

R 10 C t dt :

4

As presented above, an SUW wetland system is regarded asa process between PFR and CSTR, which are deemed astwo extremes of the hydraulic patterns in unit operation.Therefore, many wetland systems were modeled as anumber, N , of stirred tank reactors in series. In ttingexperimental data, the form for the TIS, shown in Eq. 5[21], becomes the exponential distribution, which can beconsidered as single CSTR when N = 1 and the PFR when N = ? . As a shape factor, the shape of the responsecurve is quite sensitive to change the value of N . The gammadistribution, working as a robust tool, allows modelers tochange the N value from a discrete integer to a continuousvariable [ 18]. Thus, in our curve-tting process, a gammadistribution term took the place of the term ( N - 1)! inEq. 5(see Eq. 6). The GAMMADIST and SOLVER are availablein Microsoft Excel TM to help perform a trade-off in order toobtain the best t by changing the variables N and s:

gt N N

s N N 1!t N 1 exp N

t s : 5

where N = number of tanks (unitless), t = detention time(d), s = HRT (d); and

gt N N

s N C N t N 1 exp N

t s ; 6

where C( N ) = gamma function of N R 10 exp x x N 1d x.Permeability of substrate

Two general types of permeability test methods are com-monly performed in the laboratory: the constant headmethod and the falling head method [ 22]. In this study, theconstant head method [ 22] was used to determine the

permeability of sand and PC media. The permeability wasconverted based on water at a test temperature of 20 C.

Results and discussions

Tracer HRT

Computational procedure for calculating the tracer HRTmay be carried out based on the record in Table 1. The

measured RTD curve (Fig. 2) results from an impulseaddition of 0.04 g RWT relative to the efuent concen-trations of RWT at the wetland cell outlet. The recovery of RWT was calculated as 91.4%, with a main peak on thethird day. The tracer HRT ( s) (Table 1) can be calculatedby taking 4,376.1 and dividing it by 618.6. Hence, thetracer HRT ( s) is equal to 7.1 days, much less than thenominal HRT of 12.5 days.

Curve tting

By using GAMMADIST and SOLVER, the value of N (=1.25) may be obtained (Fig. 3). Because the tracerHRT mentioned previously is much less than the expectednominal HRT, fewer samples were collected in the peak period. The low sampling frequency during the peak periodreduced the corresponding weight of evaluation indexes inthe model tting process, which eventually led to a math-ematical trade-off with a better t in the tail rather than thepeak area.

Distribution of tracer in the wetland

The distribution of tracer in the SUW was plotted byVoxler (Golden software). This robust program candisplay the data in a variety of formats: 3D volrender,isosurfaces, contours, 3D slices, orthographic and obli-que images, scatter plots, stream lines, and vector plots.A prole view of the tracer distribution in wetlandduring an 18-day period (Fig. 4) shows ten small imagesrepresenting a series of detailed ow sequences of waterwith dye illumination. On each small image, the inlet ison the bottom left and the outlet is on the upper right.The second-day small image indicates that the tracerowed with water throughout the bottom layer andmoved upward along two vertical edges of the wetland(i.e., front and back edge) within 2 days. The blue colorin the middle shows that the tracer had not yet becomehomogenous in that region, which means there might besome hydraulic retardation with time in the middle of the wetland. Because most of the tracer at the top layercame from bottom rather than horizontal movement, theoverall observation conrmed our concept of the hy-drolic upow design. In the third day, the tracergradually faded away at the inlet side and continued torise at the outlet side. From day 4 to day 8, there was arising progress of tracer in the middle and then the peak of the tracer moved out of the outlet from day 9 to day11. Finally, the remaining tracer owed gently out of the wetland from day 11 to day 18. Globally, the spa-tiotemporal tracer distribution provides strong supportfor the upow hypothesis in biosystem engineeringdesign.


1 3
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


5/9

Response curve in a local view

The response curves at different sampling ports (Fig. 5)illustrate that permeability decreases from the gravel layerto the sand layer but increases from the sand layer to thePC layer, which coordinates, unies, and harmonizes the

ow patterns and reduces the turbulent ow impact. Thehigher the permeability value, the more stable the responsecurve, implying that biolm can grow better in the PC layerafter receiving nutrients carried over from the sand layer.After passing through the PC media, the disorderly owpattern reverted to normal (same as that of the bottomsamples). Nearly the same parameter number ( N ) and HRTvalues were obtained from these top sampling ports. Theresponse curve in bottom samples presented a similartrend: a peak at the second day with a long tail. The middlethree samples displayed three totally different patterns,however, probably caused by the interaction between thelow permeability of the sand layer and the irregular poresize of the fabric beneath it. Overall, our innovative PCmedia demonstrates not only a powerful media for nutrientremoval, but also a modulator in terms of hydraulic per-formance due to the appropriate size distribution among themedia components [ 19] (US Patents 78745591, 7824551B2, and 7897047).

The original purpose of the fabric liner at the bottom of the sand layer was to separate the gravel layer from thesand layer above. The sand is added beneath the PCmedium to better remove pathogens and total suspendedsolids (TSS). The turbulent ow caused by the fabric layeroccasionally increases the HRT in the middle layer (adelayed peak or multi-peak) and the level of contactbetween water and sand layer, which is the main reasonwhy the E. coli removal efciency reached as high as99.92% in our previous study [ 19]. Further, the addition of a fabric layer also provides a secondary effect of horizontalwater distribution, which works as a semi-permeablemembrane allowing upow paths as uniformly distributedas possible. Although the patterns were totally differentacross those ports, the values of s at the same elevationwere quite close except at the middle port at bottom, whichis partially due to the hydraulic retardation effect.

From a design perspective, the results demonstrate thatthe middle column of the wetland did not effectively pro-vide homogeneous upow. In our case, the height to length(h:l) ratio might be more important than the length to width(l:w) ratio, which should be highlighted as a new consid-erable parameter for future SUW design. A shorter length(i.e., a smaller l:w ratio) can be applied for the futuredesign, which also supports the conclusion in our previousmodeling work [ 24]. Thus, we suggest that a shorter lengthand a larger depth of the SUW system can be the next-generation SUW for both xed and mobilized modules inreal-world applications.

Nutrient removal

In our study, composite samples were collected acrossports M and B (Fig. 1) assuming that water quality would

Table 1 Computational procedure for calculating the tracer HRT

t (d) C (t ) (ppb) C (t ) dt tC (t ) dt

0.0 0.0

2.0 73.1 146.2 292.4

3.0 105.1 105.1 315.3

4.0 39.3 39.3 157.3

6.0 32.4 64.8 388.88.0 30.0 60.1 480.5

9.0 30.0 30.0 270.1

11.0 24.2 48.3 531.5

13.0 18.5 37.1 481.8

16.0 20.1 60.4 966.7

18.0 13.7 27.3 491.8

Summation 618.6 4,376.1

Fig. 2 Measured RTD curve entailing relationship between RWTconcentration and time at the outlet

Fig. 3 Model tting between DTD function and NTIS


1 3
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


6/9

not vary too much in these two layers. Because of the rootzone heterogeneity, sampling across three ports at the toplayer between the growth and PC media was carried outseparately. Water quality analysis spatially across thesethree layers and ports (Table 2) shows a systematic trend of nitrication and denitrication layer by layer from thebottom to the top layers.

The nutrient removal in SUW reveals the afnitybetween transport observations and fate phenomena

(Table 2). In concert with the upow pattern, the con-centration of TN, ON-N, and NH 3 -N decreased layer bylayer. In particular, ON-N and TN were lower in themiddle top ports (i.e., T2) due to the delayed ow, evi-denced by the hydraulic retardation effect in the tracerstudy (Fig. 4). Therefore, a deeper SUW is recommendedin the future to reach a longer HRT and achieve a higherON-N removal efciency. A more sufcient aeration viathe use of more oxygenators could enhance nitricationand improve NH 3 -N removal, however. Such a syner-gistic modication could further promote overallperformance.

Final remarks

The Buckingham Pi theorem in the area of dimensionalanalysis may be further applied to draw on a relationshipbetween the nominal HRT and real HRT that is a functionin terms of several key parameters such as wetlandgeometry, owrate, and gravitational acceleration, theo-retically. By using a dimensional analysis, we may thus

derive an expression based on the following assumedconditions: the variables that affect real HRT (denoted asHRT r hereafter), which may or may not be exactly thesame as tracer HRT, would be the l, h, V n , Q , and g . Notethat sn = V n /Q and V n = l 9 w 9 h. Porosity is not con-sidered in this dimensional analysis due to its heterogenousnature across the multi-layered media in an SUW system.With this setting, we have

f HRT r; l; h; V ; Q; g

0

7

In this case, there are n = 4 variables and m = 2dimensions. We can easily nd four variables that wecannot form into a dimensionless group; therefore,

Wp 1; p 2; p 3 ; p 4 0 8Using V n and Q as the primary variable to retrieve the

information of sn (nominal HRT) in relation to the HRT r,we have

p1 V a 1n Qb1 HRT r 9p2

V a 2n Q

b2 l

10

p3 V a 3n Qb3 h 11p4 V a 4n Qb4 g 12

Working with p1 , p2 , p3 , and p4, we have all thecoefcients of a i and bi to be identied sequentially asbelow:

p 1 : L 0T 0 L 3a 1 L

3

T b1

T a1 1 and b1 1

Fig. 4 Prole view of the tracerdistribution in wetland ( left vesmall images 2, 3, 4, 6, and8 days; right ve 9, 11 13, 16,and 18 days). Vertical scale isconcentration (ppb)


1 3
http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-


7/9

p 2 : L 0T 0 L 3a 2 L 3

T b2

L a2 1=3 and b2 0p

p 3 : L 0T 0 L 3a 3 L

3

T b3

L a3 1=3 and b3 0

L 0T 0 L 3a 4 L

3

T b4 L

T 2 a4 5=3 and b4 2

Hence, p1 HRT r V n =Q ; p2 l ffiffiffiffiV n1=3p ; p3 h ffiffiffiffiV n

1=3p ; and p4 V 5=3n gQ2 ,

Given that V n /Q = sn , we end up having

HRT r sn u p2 ; p3; p4 snu l

ffiffiffiffiffiV n1=3p ;

h

ffiffiffiffiV n1=3p ;

V 5=3n gQ2

sn u l


h


V 5=3n gQ2 C sn ;

in which C is a coefcient that bridges the nominal HRTand real HRT. When the real HRT can be investigated by a

Gravel layer (Permeability: 0.5cm/s)

Sand layer (Permeability: 0.018cm/s)

PC media layer (Permeability: 0.029cm/s)

G media layer

N = 1.25 = 7.71

N = 1.00 = 7.61

N = 1.25 = 7.07

N = 2.87 = 10.61

N = 4.23 = 9.90

N = 1.55 = 9.82

N = 1.39 = 6.56

N = 2.18 = 9.86

N = 1.26 = 5.16

Fig. 5 Comparative analysis of response curves at different sampling ports for reasoning the interfaces between the global and local hydraulicpatterns


1 3


8/9

tracer study, C may be identied numerically in a labora-tory as HRT r is assumed to be measurable by using thetracer. Obviously, p2 and p3 are dimensionless groupsassociated with the geometry of a wetland and p is adimensionless group associated with the Froude number

v=

ffiffiffiffiglp

Q=

l w

=

ffiffiffiffiglp

.

Conclusions

A tracer study was carried out to conrm upow hydraulicpatterns in concert with a transport model to assess the nutrientremoval efciencies across different locations of the SUWsystem. Linking HRT with the N value associated with theTISmodel collectively provides insightful observations on main-taining nitrication and denitrication mechanisms over dif-ferent parts of the SUW system. We concluded that the tracerHRT was 7.1 days of the entire SUW, whereas the average Nvalue is 1.25 days within the SUW system. The distribution of tracer in the SUW system was plotted by Voxler to providetypical visualized evidence of the upow patterns over spaceand time. A series of local tracer response curves help dem-onstrate that permeability of media is a major factor affectinglocal patterns in a layered structure. Althougheach layer hasitsownhydraulic andenvironmental function, thePCmedia layerpacked on the top of the sand layer demonstrates a smoothhydraulic performance due to the appropriate size distributionamong the sorption media components. With regard to thedimensions of theSUW system, in our study the h:l ratiomightbe more important than the l:w ratio, which should be high-lighted as a new considerable parameter for future SUWdesign. Thus, we suggest a shorter length and larger depthSUW system should be the next-generation design.

References

1. Kadlec RH, Tanner CC, Hally VM, Gibbs MM (2005) Nitrogenspiraling in subsurface-ow constructed wetlands: implicationsfor treatment response. Ecol Eng 25:365381

2. Ronkanen AK, Klve B (2007) Use of stabile isotopes and tracersto detect preferential ow patterns in a peatland treating muni-cipal wastewater. J Hydrol 347:418429

3. Ronkanen AK, Klve B (2008) Hydraulics and ow modelling of water treatment wetlands constructed on peatlands in NorthernFinland. Water Res 42:38263836

4. Harman J, Robertson WD, Cherry JA, Zanini L (1996) Impactson a sand aquifer from an old septic system: nitrate and phos-phate. Ground Water 34:11051114

5. Wang B, Jin M, Nimmo JR, Yang L, Wang W (2008) Estimatinggroundwater recharge in Hebei Plain, China under varying landuse practices using tritium and bromide tracers. J Hydrol356:209222

6. Maoszewski P, Wachniew P, Czupryn ski P (2006) Hydrauliccharacteristics of a wastewater treatment pond evaluated throughtracer test and multi-ow mathematical approach. Pol J EnvironStud 15(1):105110

7. Dierberg FE, DeBusk TA (2005) An evaluation of two tracers insurface-ow wetlands: rhodamine-WT and lithium. Wetlands25(1):825

8. Lin AY-C, Debroux J-F, Cunningham JA, Reinhard M (2003)Comparison of rhodamine WT and bromide in the determinationof hydraulic characteristics of constructed wetlands. Ecol Eng20:7588

9. Giraldi D, DeMichieli Vitturi M, Zaramella M, Marion A, Ian-nelli R (2009) Hydrodynamics of vertical subsurface ow con-structed wetlands: tracer tests with rhodamine WT and numericalmodeling. Ecol Eng 35:265273

10. Kadlec RH, Knight RL (1996) Treatment wetlands. CRC Press,Boca Raton

11. Wang HG, Jawitz JW (2006) Hydraulic analysis of cell-network treatment wetlands. J Hydrol 330:721734

12. Fogler HS (1992) Elements of chemical reaction engineering, 2ndedn. Prentice Hall, Englewood Cliffs

13. Valocchi AJ (1985) Validity of the local equilibrium assumptionfor modeling sorbing solute transport through homogeneous soils.Water Resour Res 21:808820

14. Martinez CJ, Wise WR (2003) Analysis of constructed treatmentwetland hydraulics with the transient storage model OTIS. EcolEng 20(3):211222

15. Fan L, Reti H, Wang W (2008) Application of computationaluid dynamic to model the hydraulic performance of subsurfaceow wetlands. J Environ Sci 20(12):14151422

16. Persson J (2000) The hydraulic performance of ponds of variouslayouts. Urban Water 2:243250

17. Worman A, Kronnas V (2005) Effect of pond shape and vege-tation heterogeneity on ow and treatment performance of con-structed wetlands. J Hydrol 301:123138

18. Kadlec RH, Wallace SD (2009) Treatment wetlands, 2nd edn.Taylor and Francis Group, Boca Raton

19. Chang NB, Xuan Z, Daranpob A, Wanielista M (2011) A sub-surface upow wetland system for removal of nutrients andpathogens in on-site sewage treatment and disposal systems.

Environ Eng Sci 28(1):112420. Danckwerts PV (1953) Continuous ow systems: distribution of residence times. Chem Eng Sci 2:113

21. Levenspiel O (1972) Chemical reaction engineering, 1st edn.Wiley, New York

22. ASTM (D2434-68) Standard test method for permeability of granular soils (constant head), ASTM international

23. Headley TR, Kadlec RH (2007) Conducting hydraulic tracerstudies of constructed wetlands: a practical guide. EcohydrolHydrobiol 7:269282

24. Xuan Z, Chang NB, Daranpob A, Wanielista M (2010) Modelingthe subsurface upow wetland (SUW) system for wastewaterefuent treatment. Environ Eng Sci 27(10):879888

Table 2 Spatial distribution of nitrogen concentration, l g/L [ 19]

NH 3 -N NO 2 -N ? NO 3 -N ON-N TN

Septic efuent 56,061.8 7.5 13,670.3 69,739.5

Port B 27,159.0 7.0 5,325.3 32,489.0

Port M 16,653.3 8.5 5,069.8 21,730.0

Port T1 839.8 5.0 3,684.8 4,526.8

Port T2 1,169.8 16.8 1,497.0 2,681.3Port T3 1,185.0 8.8 3,607.0 4,799.3

TN total nitrogen, ON-N organic nitrogen, NH 3 -N ammonia nitrogen, port B the composite of three bottom samples, port M the compositeof three middle samples


1 3
http://-/?-http://-/?-


9/9

Reproduced withpermission of the copyright owner. Further reproductionprohibited without permission.

tracers in subsurface upflow wetland

Documents