Sequestration and Transformation in Chemically Enhanced Treatment Wetlands: DOC, DBPPs, and Nutrients

Bachand, Philip A. M.; Bachand, Sandra M.; Kraus, Tamara E. C.; Stern, Dylan; Liang, Yan Ling; Horwath, William R.

doi:10.1061/(ASCE)EE.1943-7870.0001536

Open access

Technical Papers

Jun 6, 2019

Sequestration and Transformation in Chemically Enhanced Treatment Wetlands: DOC, DBPPs, and Nutrients

Authors: Philip A. M. Bachand, Ph.D. [email protected], Sandra M. Bachand [email protected], Tamara E. C. Kraus, Ph.D. [email protected], Dylan Stern [email protected], Yan Ling Liang, Ph.D. [email protected], and William R. Horwath [email protected]Author Affiliations

Publication: Journal of Environmental Engineering

Volume 145, Issue 8

https://doi.org/10.1061/(ASCE)EE.1943-7870.0001536

PDF

Abstract

This paper examines the effectiveness of chemically enhanced treatment wetlands (CETWs), wetlands that received water treated with coagulants, to remove dissolved organic carbon (DOC), disinfection byproduct precursors (DBPPs), nutrients, and metals from agricultural drain water. Wetlands consisted of controls with no coagulant addition, ferric sulfate dosed, and polyaluminum chloride dosed treatments. CETWs were more effective in removing DOC, DBPPs, phosphate, dissolved organic nitrogen, and metals than control wetlands. Coagulation-treated wetlands removed

245 - 349 g / m^{2} / year

DOC, whereas control wetlands produced

51 g / m^{2} / year

. Wetland passage released DOC in the controls and dosed treatments; this release was directly correlated to temperature, suggesting thermally dependent mechanism(s) were partly responsible. A first-order plug flow reactor kinetic model that considered hydraulic retention time (HRT), temperature, and DOC concentration was tested for wetland DOC processing. Models indicate that operating CETWs at high coagulant dosing and low temperature can result in the lowest DOC release. Operating at the lowest HRT to meet discharge targets helps overcome wetland processes that increase DOC release and provide the smallest footprint needed for treatment.

Introduction

Fertile soils along with a secure source of irrigation water make delta regions an attractive hub for agriculture; however, these areas are often confronted with land subsidence and water quality issues, which are exacerbated by extensive farming (Törnqvist et al. 2008; Verhoeven and Setter 2010; Brown and Nicholls 2015; Minderhoud et al. 2018). In the United States, agriculture is a major contributor to degraded water quality in wetlands (Locke 2004; USEPA 2016), and while subsidence can be simply a nuisance, in the Sacramento–San Joaquin Delta (Delta) of California (United States) it poses a great threat for water security and the region’s economy (Deverel and Leighton 2010; Hoffman 2010; California Department of Water Resources 2012). The extensive levee systems surrounding islands in the Delta were originally built to allow production of terrestrial crops. Over the span of

150 +

years of these farming practices, subsidence has occurred from soil oxidation, wind erosion, and physical compaction, and on some islands subsidence is severe with land surface levels

> 6 m

below sea level (Deverel et al. 2016; Minderhoud et al. 2018).

In current-day Delta, in addition to maintaining and protecting valuable agricultural land (Delta Stewardship Council 2010), these levees guard against saltwater intrusion that would otherwise render this water source temporarily or permanently unusable and protect against water quality degradation due to island flooding events (Ingebritsen et al. 2000; Mount and Twiss 2005). Subsidence causes a larger hydraulic gradient that must be withstood by these levees and increase the probability of levee failure (Mount and Twiss 2005). The Delta is a crucial freshwater resource for California (Moazezi et al. 2017); diversion points from this resource include outtakes for Northern California water agencies, such as Contra Costa Water District, federal and state water project pumps that provide irrigation water for 3 million acres of California’s agriculture and drinking water for 25 million Californians (Lund 2008, 2016; PPIC 2016). Due to stricter water quality regulations on both a national and state level (State Water Resources Control Board 2006; USEPA 2018), the ability to meet potential discharge regulations while maintaining agrarian productivity is a concern for Delta farmers and land managers.

To address subsidence and water quality challenges, chemically enhanced treatment wetlands (CETWs) were implemented and studied on a subsided island in the Delta to offer a potential solution (USGS 2016). The CETWs consisted of in situ coagulation treatments combined with wetlands serving as settling basins for water treatment residuals (flocs); finished water is conceptualized to be exported into the Delta water channels to improve and/or negate water quality degradation from farming practices. Coagulation is anticipated to work synergistically with wetland processes that promote settling of particles through quiescent conditions and vegetation density (Darke and Megonigal 2003). In CETWs, water quality improvement is addressed primarily by coagulation and transfer of material from the water column to the sediment, and subsidence mitigation is addressed by keeping soils saturated, which minimizes oxidation of existing organic material in addition to accumulation of new material in the wetlands in the form of plant biomass and floc (Miller et al. 2008; Deverel and Leighton 2010). The subsidence component of this project is discussed in a related publication (Stumpner et al. 2018); here the authors focus on water quality improvement from coagulation and wetland interactions.

CETW systems offer several advantages over full-scale water treatment plants: CETW are cost and energy saving because they require less infrastructure and do not require floc removal and off-site disposal but instead use in situ floc accumulation to further aid subsidence mitigation; CETW are more easily operable than full-scale plants due to the replacement of sedimentation, clarification, and filtration basins with low maintenance wetlands. Compared to untreated wetlands, CETWs require less land area (that would otherwise be dedicated to crop production) to meet water quality goals and provide a faster time response to varying source water quality than untreated wetlands. Employment of coagulation in series with wetlands is uncommon and requires further examination (Malecki-Brown and White 2009; Malecki-Brown et al. 2009; Lindstrom and White 2011; McCord and Heim 2015). To date, CETWs have been tested at small scales in field mesocosm systems upstream of the Florida Everglades (Bachand et al. 2000) and laboratory mesocosm studies of stormwater treatment in wetlands upstream of Lake Tahoe (Bachand and Associates et al. 2006; Trejo-Gaytan et al. 2006). These studies have included water quality constituents of concern removal efficiency assessments (Bachand et al. 2000; Bachand and Associates et al. 2006; Trejo-Gaytan et al. 2006), design of dosing strategies to mitigate toxicity issues (Lopus et al. 2009; Bachand et al. 2010a), and feasibility testing and design recommendations (Bachand et al. 2010b). The current data on coagulation in semi-natural systems suggest opportunities exist for using CETWs to treat agricultural drainage water; however, studies are still needed to quantify their effectiveness, and identify any difficulties and safety concerns since implementation of these systems are only cost-effective at larger scales and for longer time periods than have currently been studied (P. A. M. Bachand, T. E. C. Kraus, W. R. Horwath, N. Hatch, and S. M. Bachand, “Chemically enhanced treatment wetland design and cost for water quality improvement and land subsidence mitigation in the Sacramento-San Joaquin Delta,” submitted).

In this study, several constituents were monitored to evaluate the effectiveness of the CETWs and to understand the interactions of these systems—dissolved organic carbon (DOC), metals, and nutrients—all of which are considered common problematic pollutants in surface water systems by the USEPA (2018). Additionally, DOC has been identified by the State Water Resources Control Board as a water quality constituent of concern because of its potential to form disinfection byproducts (DBPs) upon chlorination treatment (Xie 2004; Richardson et al. 2007; Kraus et al. 2008). These DBPs have been shown to be carcinogenic, mutagenic, and often correlated to DOC concentration in solution (Barrett et al. 2000; Richardson et al. 2007; Díaz et al. 2008; Hrudey 2009; Krasner et al. 2009). Previous studies have estimated 13%–49% of DOC exported by Delta waters originated within the Delta itself, including substantial contributions from agricultural drainage waters from subsided, organic-rich lands located in the central Delta (Deverel et al. 2007; Kraus et al. 2008). Laboratory studies employing aluminum and iron-based coagulants for treating similar island drain waters have shown DOC removals between 44% and 80% (Mourad 2008; Henneberry et al. 2011). In addition, nutrients and heavy metals can also be removed during coagulation (Aguilar et al. 2002; Lee and Westerhoff 2006; El Samrani et al. 2008; Akbal and Camcı 2010), leading to their use for eutrophication reduction in lakes and reservoirs (Harper et al. 1998; Rydin and Welch 1998; Welch and Cooke 1999; Sherwood and Qualls 2001; Welch and Schrieve 2013). However, removal of nutrients may have important implications on plant nutrient availability, bacterial growth, and wetland processes (Thingstad and Lignell 1997; Poorter and Nagel 2000; Delgado-Baquerizo et al. 2017; Weisman et al. 2018).

This project represents the first full-scale implementation of this hybrid coagulation/wetland technology lasting longer than 1 year. To study the effectiveness of on-site coagulation for constituent removal under field conditions and the interaction of flocs with wetlands subjected to seasonal variations, the authors constructed CETWs under two coagulation treatments—ferric sulfate coagulation and polyaluminum chloride coagulation—and operated these alongside untreated control wetlands. Wetlands were operated for 16 continuous months. During this period, water flow, DOC, DBP formation potentials (DBPFPs), total dissolved nitrogen (TDN), dissolved organic nitrogen (DON), dissolved inorganic nitrogen (DIN), phosphate, dissolved aluminum (DAl), and dissolved iron (DFe) were monitored to assess efficacy of their removal, predict performance, and understand their behavior under varying field conditions. Analyses were made by examining changes in concentration, mass load budgets, and employing plug flow reactor (PFR) kinetic modeling.

Methods

Wetland Design and Treatments

Nine wetland cells were constructed on Twitchell Island in the Sacramento–San Joaquin Delta of California, United States (Fig. 1). Each wetland cell was approximately 36.6 m long (from inflow to outflow) and 12.2 m wide at the bottom with 2:1 side slopes and approximately 0.4 m deep. Wetland cells were separated by berms that were approximately 4 m wide. Where leakage across cells was determined to be significant based on water budget calculations, slurry walls were built between the cells within the berm walls. Drainage ditches along the site perimeter were used to manage groundwater and maintain a slight downward gradient in the cells. Two coagulation treatments were included in this study, an Fe treatment dosed with ferric sulfate coagulant (60% Soln., Kemira Water Solutions, Mojave, California) and an Al treatment dosed with polyaluminum chloride coagulant (PAC, PAX-XL19, Kemira Water Solutions, Finland), and a non-coagulation control (Co). Treatments and controls were placed in a complete randomized block design, with three replicates per treatment and control. Each of the three blocks contained one of each treatment and a control and were spatially clustered north to south to account for any spatial trends in soil biogeochemistry or hydrology [Fig. 1(a)].

Fig. 1. (a) Twitchell Island (38°05′55.19″N 121°39′19.08″W, elevation -21 ft; image Landsat/Copernicus © Google 2018) and schematic of replicated field study design; and (b) diagram of a chemically enhanced treatment wetland. White rectangular box in (a) indicates study field site; designated treatment of each wetland cell is indicated by Fe and Al for the ferric sulfate and polyaluminum chloride treatments, respectively, and Co for the control.

Hydrologic and chemical control and monitoring systems were designed, built, and tested during 2011 and 2012. Coagulant dosing occurred from July 2012 to November 2013 (16 months), except for a 3-week period in October 2012 when the coagulation system was offline due to equipment failures. Untreated source water, consisting of a combination of drainage water from surrounding fields and irrigation water from the San Joaquin River, was pumped into the cells at velocities that averaged

0.61 - 0.76 m / s

flowing through 5-cm (2-in.) pipes. Pumps were operated for a set period each day to meet a HRT of 3–5 days. Coagulant was injected into source water pipes and immediately flash mixed using static mixers. Coagulant dosing was controlled using peristaltic pumps to achieve 60%–80% of DOC removal. Dosing rate is defined in this study by the percentage removal of DOC. Detailed methods for calculating and controlling coagulant dosing can be found in Bachand et al. (2019). Treated water containing flocs flowed through the pipes for about 2 min, uniform for all cells, and then entered wetlands via inflow pipes [Fig. 1(b)]. After entering cells, flocs settled out of the water column in the wetlands as water flowed the length of the cells and exited by skimming action of weirs that flowed into outflow pipes that exported water into drainage ditches [Fig. 1(b)]. Water from the drainage ditches connected to a main drain where water was pumped off the island into the San Joaquin River. Water depth in the cells were maintained by adjusting weirs to target 30–50 cm. Flow rates entering and exiting each cell were measured in 15-min intervals using magnetic flow meters. Wetlands were predominantly vegetated by cattails (Typha spp.). Further details regarding the control system, wetland design, and operation can be found in related design papers (Bachand et al. 2019).

Water Monitoring, Sampling, and Analyses

Weekly water sampling occurred between July 2012 and November 2013. Water was collected from the inflows and outflows of all nine wetland cells, with water entering the untreated controls representing nondosed inflow (In), water collected from the inflow to the Fe-treated and Al-treated wetlands representing postdosed inflow (PD), and water exiting the cells representing outflow (Out) water samples. Samples were collected into 250-mL amber glass bottles [Fig. 1(b)]. Samples were analyzed weekly for DOC and monthly for dissolved aluminum (DAl), dissolved iron (DFe), phosphate (

{PO}_{4}^{3 -}

), total dissolved nitrogen (TDN), and dissolved inorganic nitrogen (DIN, sum of nitrate and ammonium) concentrations. Dissolved organic nitrogen (DON) was calculated by calculating the differences between TDN and DIN. DOC was measured by UV-persulfate digestion (Teledyne-Tekmar Phoenix 8000, standard method 5310-C). TDN was measured by persulfate oxidation and colorimetry (D’Elia and Steudler 1977; Koroleff 1983). Concentrations of ammonium, nitrate, phosphate, and DFe were determined by colorimetry (Murphy and Riley 1958; Pritchard 1967; Verdouw et al. 1978; Viollier et al. 2000; Doane and Horwath 2003). All constituent measurements were performed on samples centrifuged in 2-mL micro-centrifuge tubes at 15,000×g for 20 min. All colorimetric measurements were made using a Shimadzu UV mini 1,240 spectrophotometer. Dissolved Al was measured using inductively couple plasma-atomic emission spectrometry (EPA 1994). Samples for DBPFP were collected in late February, May, and July 2013 to represent winter, spring, and summer, respectively. Samples were collected in 2-L PETG bottles and analyzed for trihalomethane (THM) and haloacetic acid (HAA) formation potentials used as an indicator for DBPPs according to methods described in Hansen et al. (2018).

Ancillary water parameter temperature, dissolved oxygen (DO), and pH were measured in water exiting and entering the wetland cells using a multiparameter water-quality Sonde (YSI 6920) during the weekly sampling events. YSI calibration, maintenance, and data collection proceeded as per manufacturer’s instructions (YSI Incorporated 2010, 2011). Drift check and calibration of YSI Sonde was performed the night before operation.

Water and Mass Budgets, Calculations and Statistical Analyses

Water budgets for each cell were calculated daily using average daily inflow, outflow, evapotranspiration (ET), and precipitation. Seepage losses from infiltration were calculated by difference. Hydraulic retention time for each cell was calculated using cell volumes divided by flow rates at inflows. For control wetlands, PD samples are identical to In samples because water was not dosed. To get the most accurate estimate of constituent removal resulting from coagulant dosing, each PD sample result was paired with an In sample taken closest in time. To evaluate changes in the entire system (coagulation + wetland passage = total system), each In sample result was paired with the Out sample result taken on the same day from the same cell. Constituent annual loads were calculated from measured weekly flow rates and concentration data. The authors compared concentration data, load data, and concentration ratios (concentration normalized to concentration from an earlier step in treatment process) to evaluate coagulation (In – PD), wetland (PD – Out), and total system performance (In – Out). All calculations and data quality assurance and quality control were made in an ACCESS database (Microsoft). Statistical analysis was performed using ANOVA in STATISTICA version 8.0 with a p level

< 0.05

to identify significant differences.

Kinetic Model

Kinetic models, built from analytical mass balance equations, assess changes in constituent concentrations due to microbial activity or chemical reactions, relating changes in concentrations with the concentration of the constituent itself (Metcalf and Eddy 1979; Snoeyink and Jenkins 1980; Kadlec and Knight 1996). These models are used to develop equations for calculating relative changes in constituent concentrations across a system, and to assess the importance of controlling factors such as HRT, dosing rate, and temperature. An analytical model was developed for constituent behavior during wetland passage based on the plug flow reactor (PFR, Fig. S1) model and first-order kinetic wetland constituent transformation:

C_{o u t} = C_{p d} e^{- K (H R T)}

(1)

where

C_{o u t}

is the outflow concentration,

C_{p d}

is concentration at inflow post-dosing, HRT is the hydraulic retention time, and

K

is the rate constant. Eq. (1) was modified to include of temperature effects on microbial reactions based upon a modified Arrhenius relationship (Metcalf and Eddy 1979; EPA 1993):

C_{o u t} = C_{p d} e^{- K_{20} θ^{T - 20} (H R T)}

(2)

where

T

is temperature in °C,

K_{20}

is the rate constant at 20°C, and

θ

is the temperature dependency of the rate. Additionally, the reversibility of coagulation was considered due to constituent desorption after the flocs settled in the wetland cells (Chen et al. 2014). The concentration change due to desorption (

C_{r}

) is assumed to be proportional to the concentration change due to coagulation (

Δ C_{M e}

) and to the percentage of removal that is reversible standardized to 20°C (

%_{M e}

, called here the floc release coefficient). The floc release coefficient is coagulant-dependent and indicated by subscript Me for Al or Fe treatment. After including the temperature dependency of release at 20°C (

θ_{M e}^{T - 20}

), the process can be modeled as

C_{r} = %_{M e} Δ C_{M e} \times θ_{M e}^{T - 20}

(3)

Combining Eqs. (2) and (3) yields a model of wetland outflow constituent concentrations dependent on PD concentration, concentration change due to coagulation, temperature, HRT, and coagulant type

C_{o u t} = C_{p d} e^{- K_{20} θ^{T - 20} (H R T)} + %_{M e} Δ C_{M e} \times θ_{M e}^{T - 20}

(4)

Details on the derivation of the preceding equations can be found in the Supplemental Data file.

A stepwise approach was used to develop the final kinetic model constants. First,

K_{20}

and

θ

were developed for the control using Eq. (2). The rate constant was then used in Eq. (4) to develop the values for the remaining coagulant-dependent constants. This approach enabled us to overlay coagulant-specific effects onto the wetland model, under a simplified assumption of superposition. In Step 1, the K-value in Eq. (1) was developed to model DOC concentration changes across the untreated Co wetlands. In Step 2, temperature effects were incorporated to improve the model for the Co wetlands and calculate rate constants (

K_{20}

) and its temperature dependency [

θ

, Eq. (2)]. Steps 3 and 4 applied the same processes as Steps 1 and 2 but for all wetlands simultaneously. Step 5 overlaid coagulation reversibility [Eq. (3)] onto the Co wetland analytical model [Eq. (2)] and using Eq. (4) calculated

K_{20}

,

θ

,

%_{M e}

and

θ_{M e}^{T - 20}

for all wetlands. Constants were calculated using nonlinear multivariate regression and modeling was performed using STATISTICA version 8.0.

Results

DOC and DBPP Concentrations, Loads and Concentration Ratios

Average constituent concentrations, annual loads and concentration ratios for DOC and DBPPs are shown in Tables 1 and 2. Annually, Al and Fe coagulation removed 71%–72% of DOC loads entering the wetlands, 66%–69% of THM precursors, and 79%–81% of HAA precursors (Table 1). Concentration ratios help to evaluate the degree of removal or release by normalizing concentration at one step to that of a previous step in the treatment process (Table 2). This is relevant to wetland processes because the relationship describing wetland outflow concentrations is proportional to inflow PD concentrations [Eq. (2)]. The ratio

C_{p d} / C_{i n}

is a measure solely of coagulation performance with lower ratios signifying greater removal. These ratios were equivalent for both Al and Fe treatments at 0.27, demonstrating equal removal of DOC by both coagulants (Table 2). The

C_{o u t} / C_{p d}

DOC ratio showed significant differences in wetland performance by treatment (

p < 0.05

), with ratios equal to 1.19, 1.78, and 3.03 for Co, Al, and Fe treatments, respectively, indicating the largest DOC release was from Fe wetlands. Total system performance ratios (

C_{o u t} / C_{i n}

), which include coagulation and wetland processes combined, was 1.19 for the Co, 0.47 for Al and 0.76 for Fe treatments. These system ratios were significantly different by treatment (Table 2) and corresponded to total removal of

- 8 %

(Co), 58% (Al), and 41% (Fe) of inflow DOC loads by the system (Table 1). Overall, annual achievable load removal of DOC by the total system was

245 g / m^{2} / year

under Fe treatment,

349 g / m^{2} / year

under Al treatment and –

51 g / m^{2} / year

in controls (Table 1).

Table 1. Annual loads, average constituent concentrations, loads, and removal for DOC, DBPs, and nutrients for the control (Co), and polyaluminum chloride (Al) and ferric sulfate (Fe) treatments

Analyte	Trt	N	Concentrations and loads							Load removal
			Inflow (In)		Postdosing (PD)		Seepage	Outflow (Out)		Coagulation		Wetland		Total system
			Concentration $Mean \pm STD$	Load ( $g / m^{2} / year$ )	Concentration $Mean \pm STD$	Load ( $g / m^{2} / year$ )	Load ( $g / m^{2} / year$ )	Concentration $Mean \pm STD$	Load ( $g / m^{2} / year$ )	Mean ( $g / m^{2} / year$ )	% inflow	Mean ( $g / m^{2} / year$ )	% inflow	Mean ( $g / m^{2} / year$ )	% inflow
DOC ( $mg / L$ )	Co	161	$15 \pm 9$	643^a	$15 \pm 9$	643^b	89^b	$17 \pm 7$	605^c	0^a	0	$- 51$ ^b	$- 8$	$- 51$ ^a	$- 8$
	Al	165	$15 \pm 9$	601^a	$4 \pm 3$	176^a	37^a	$7 \pm 4$	215^a	424^b	71	$- 76$ ^b	$- 13$	349^c	58
	Fe	162	$15 \pm 9$	600^a	$4 \pm 3$	167^a	39^a	$10 \pm 4$	316^b	433^b	72	$- 188$ ^a	$- 31$	245^b	41
HAA ( $μg / L$ )	Co	8	$1,043 \pm 445$	49^a	$1,043 \pm 445$	49^b	7^a	$1,220 \pm 352$	48^c	0^a	0	$- 6$ ^a	$- 13$	$- 6$ ^a	$- 13$
	Al	9	$998 \pm 437$	46^a	$211 \pm 95$	10^a	2^a	$278 \pm 132$	10^a	36^b	79	$- 3$ ^a	$- 6$	33^b	73
	Fe	8	$1,043 \pm 445$	49^a	$193 \pm 142$	9^a	2^a	$542 \pm 122$	21^b	39^b	81	$- 14$ ^a	$- 29$	25^b	52
THM ( $μg / L$ )	Co	8	$1,433 \pm 527$	67^a	$1,433 \pm 527$	67^b	9^a	$1,582 \pm 418$	62^b	0^a	0	$- 5$ ^b	$- 7$	$- 5$ ^a	$- 7$
	Al	9	$1,389 \pm 510$	63^a	$474 \pm 122$	21^a	4^a	$556 \pm 191$	20^a	42^b	66	$- 3$ ^b	$- 5$	39^b	61
	Fe	8	$1,433 \pm 527$	66^a	$448 \pm 198$	21^a	4^a	$900 \pm 166$	35^ab	46^b	69	$- 18$ ^a	$- 27$	27^b	41
DAl ( $μg / L$ )	Co	20	$90 \pm 99$	4.5^a	$90 \pm 99$	4.5^b	0.6^ab	$68 \pm 57$	2.9^b	0^a	0	1.0^ab	23	1.0^a	23
	Al	21	$87 \pm 98$	3.9^a	$68 \pm 26$	2.9^b	0.3^b	$12 \pm 12$	0.4^a	1.1^b	27	2.1^b	54	3.2^a	81
	Fe	20	$90 \pm 99$	4.1^a	$26 \pm 28$	1.1^a	0.1^a	$25 \pm 24$	1.0^ab	2.9^b	72	0.0^a	0	2.9^a	72
DFe ( $mg / L$ )	Co	41	$2.3 \pm 0.7$	89.0^a	$2.3 \pm 0.7$	89.0^c	16.5^c	$2.3 \pm 1.0$	69.7^c	0^a	0	2.9^ab	3	2.9^a	3
	Al	45	$2.3 \pm 0.7$	84.5^a	$0.2 \pm 0.2$	6.8^a	1.5^a	$0.3 \pm 0.3$	4.5^a	77.8^b	92	0.8^b	1	78.5^b	93
	Fe	43	$2.3 \pm 0.7$	84.2^a	$0.5 \pm 0.5$	17.2^b	5.7^b	$1.1 \pm 1.1$	27.6^b	67.0^b	80	$- 16.1$ ^a	$- 19$	50.9^b	60
DIN ( $mg / L$ )	Co	20	$0.83 \pm 0.57$	37.0^a	$0.83 \pm 0.57$	37.0^a	3.7^a	$0.27 \pm 0.28$	10.7^a	0.0^a	0	22.6^a	61	22.6^a	61
	Al	21	$0.81 \pm 0.56$	34.8^a	$0.83 \pm 0.56$	35.7^a	4.1^a	$0.41 \pm 0.44$	15.5^a	$- 1.0$ ^b	$- 3$	16.1^a	46	15.1^a	43
	Fe	20	$0.83 \pm 0.57$	35.9^a	$0.85 \pm 0.57$	37.0^a	3.2^a	$0.45 \pm 0.53$	17.0^a	$- 1.1$ ^b	$- 3$	16.9^a	47	15.8^a	44
DON ( $mg / L$ )	Co	20	$1.05 \pm 0.53$	46.8^a	$1.05 \pm 0.53$	46.8^b	7.1^a	$0.96 \pm 0.25$	34.0^b	0.0^a	0	5.7^a	12	5.7^a	12
	Al	21	$1.03 \pm 0.52$	44.1^a	$0.47 \pm 0.23$	20.6^a	3.6^a	$0.42 \pm 0.17$	14.5^a	23.5^b	53	2.5^a	6	26^b	59
	Fe	20	$1.05 \pm 0.53$	45.5^a	$0.47 \pm 0.15$	20.5^a	3.9^a	$0.62 \pm 0.21$	22.7^ab	25.0^b	55	$- 6.1$ ^a	$- 13$	18.9^ab	42
TDN ( $mg / L$ )	Co	20	$1.88 \pm 1.09$	83.9^a	$1.88 \pm 1.09$	83.9^a	10.8^a	$1.23 \pm 0.45$	44.7^a	0.0^a	0	28.3^a	34	28.3^a	34
	Al	21	$1.84 \pm 1.08$	78.9^a	$1.30 \pm 0.65$	56.3^a	7.7^a	$0.83 \pm 0.55$	30.1^a	22.5^b	29	18.6^a	24	41.1^a	52
	Fe	20	$1.88 \pm 1.09$	81.4^a	$1.32 \pm 0.53$	57.5^a	7.1^a	$1.06 \pm 0.55$	39.7^a	23.9^b	29	10.7^a	13	34.7^a	43
${PO}_{4}$ ( $mg / L$ )	Co	41	$0.13 \pm 0.03$	4.9^a	$0.13 \pm 0.03$	4.9^b	0.9^b	$0.12 \pm 0.04$	3.8^b	0.0^a	0	0.2^a	4	0.2^a	4
	Al	45	$0.13 \pm 0.03$	4.6^a	$0.03 \pm 0.02$	1.1^a	0.2^a	$0.04 \pm 0.02$	0.9^a	3.5^b	75	0.0^a	$- 1$	3.4^b	75
	Fe	43	$0.13 \pm 0.03$	4.6^a	$0.03 \pm 0.02$	1.0^a	0.2^a	$0.04 \pm 0.02$	1.0^a	3.6^b	78	$- 0.2$ ^a	$- 4$	3.4^b	74
DO ( $mg / L$ )	Co	155	$4.17 \pm 1.70$	172.3^a	$4.17 \pm 1.70$	172.3^a	23.3^a	$3.88 \pm 1.03$	135.1^a	0.0^a	0	13.9^b	8	13.9^b	8
	Al	159	$4.13 \pm 1.73$	165.3^a	$4.00 \pm 1.71$	160.4^a	32.5^b	$4.75 \pm 1.16$	148.6^a	4.9^b	3	$- 20.7$ ^a	$- 13$	−15.8^a	$- 10$
	Fe	157	$4.17 \pm 1.71$	165.7^a	$4.15 \pm 1.62$	167.6^a	26.9^ab	$4.93 \pm 1.49$	159.6^a	$- 1.9$ ^b	$- 1$	$- 18.9$ ^a	$- 11$	−20.8^a	$- 13$
pH	Co	158	$6.47 \pm 0.22$	—	$6.47 \pm 0.22$	—	—	$6.41 \pm 0.23$	—	—	—	—	—	—	—
	Al	162	$6.47 \pm 0.22$	—	$6.42 \pm 0.22$	—	—	$6.47 \pm 0.25$	—	—	—	—	—	—	—
	Fe	159	$6.48 \pm 0.23$	—	$5.78 \pm 0.43$	—	—	$6.15 \pm 0.38$	—	—	—	—	—	—	—

Note: DOC = dissolved organic carbon; HAA = haloacetic acid; THM = trihalomethane; DAl = dissolved Al; DFe = dissolved Fe; DIN = dissolved inorganic nitrogen; DON = dissolved organic nitrogen; TDN = total dissolved nitrogen; ${PO}_{4}$ = phosphate; DO = dissolved oxygen. Trt = treatments: Al = polyaluminum chloride, Fe = ferric sulfate; Co = control; N = sample size; and STD = standard deviation. Seepage loads at inflow, postdosing, and outflow locations take into account evapotranspiration and precipitation. Post-hoc letters indicate significant differences between treatments and control for each constituent and process type.

Table 2. Constituent concentration ratios representing coagulation, wetland, and total system performance for the control (Co), and polyaluminum chloride (Al) and ferric sulfate (Fe) treatments

Analyte	Treatment	Coagulation $C_{p d} / C_{i n}$	Wetland $C_{o u t} / C_{p d}$	Total system $C_{o u t} / C_{i n}$
DOC	Co	N/A	1.19^a	1.19^c
	Al	0.27^a	1.78^b	0.47^a
	Fe	0.27^a	3.03^c	0.76^b
HAA	Co	N/A	1.24^a	1.24^c
	Al	0.21^a	1.34^a	0.28^a
	Fe	0.16^a	4.56^b	0.61^b
THM	Co	N/A	1.17^a	1.17^c
	Al	0.31^a	1.21^a	0.37^a
	Fe	0.26^a	2.83^b	0.68^b

Note: DOC = dissolved organic carbon; HAA = haloacetic acid (HAA); and THM = trihalomethane. Constituent concentrations ( $C_{i n}$ , $C_{p d}$ , $C_{o u t}$ ) at inflow (In), postdosing (PD), and outflow (Out) locations were used to calculate ratios. Ratios $< 1$ indicate net removal of constituent; ratios $> 1$ indicate release of constituent. Post-hoc letters indicate significant differences between treatments and control for each constituent and process type.

Both THM and HAA precursors exhibited similar behaviors as DOC: significant removal from the water column due to coagulation alone, followed by increased concentrations and loads after wetland passage (Table 1), resulting in significant differences across the entire system by treatment (Table 2). Haloacetic acid and trihalomethane precursor concentrations were strongly and linearly correlated to DOC concentrations [

R^{2} > 0.96

,

p ≪ 0.05

, Fig. 2(a)], as were system performance ratios [

R^{2} \geq 0.80

,

p = 0.00

, Fig. 2(b)]. Similar to DOC, coagulation performance for both HAA and THM removal were not significantly different for Al and Fe treatments (Tables 1 and 2). With respect to wetland performance, under the Al treatment

C_{o u t} / C_{p d}

ratios for DOC averaged 1.78 compared to 1.34 and 1.21 for HAA and THM precursors, respectively (Table 2). Under the Fe treatment, wetland performance ratios for DOC averaged 3.03 compared to 4.56 and 2.83 for HAA and THM precursors, respectively. Overall, annual HAA load removal averaged 33 and

25 g / m^{2} / year

and annual THM load removal were 39 and

27 g / m^{2} / year

for Al and Fe systems, respectively (Table 1). Control wetlands typically increased DBPP loads by

5 - 6 g / m^{2} / year

after wetland passage.

Fig. 2. (a) Correlation of haloacetic acid formation potential (HAAFP) and trihalomethane formation potential (THMFP) with dissolved organic carbon (DOC) in concentration; and (b) correlation of HAAFP and THMFP with DOC in outflow/inflow concentration ratio ( $C_{o u t} / C_{i n}$ ). Coefficients of determination ( $R^{2}$ ) and probability values (p) are shown.

Ancillary Parameters, Nutrients and Metals Concentrations and Loads

Concentrations and annual loads for nutrients, metals, and ancillary data are reported in Table 1. Inflow DAl loads from untreated source water ranged between

3.9 a n d 4.5 g / m^{2} / year

, while inflow DFe loads were an order of magnitude higher at

84 - 89 g / m^{2} / year

. Coagulation alone removed 27% and 72% of DAl inflow loads in Al and Fe treatments, respectively. Coagulation removal of DFe was higher at 92% and 80% by Al and Fe treatments, respectively. Control and Al wetlands removed additional DAl and DFe during wetland passage, but Fe wetlands did not.

On average, coagulation removed 75%–78% of phosphate and 29% of TDN loads (Table 1). TDN removal during the coagulation step occurred mainly from DON removal (54%–55%), since coagulation resulted in a slight (

\sim 3 %

) increase of DIN. TDN decreased significantly in all wetlands, resulting in concentration decreases of 0.65, 0.48, and

0.26 mg / L

in the Control, Al, and Fe wetlands, respectively. Reductions of TDN in the wetlands were mainly due to decreasing DIN concentrations; DON concentrations were not greatly affected by wetland passage in the Al and Co wetlands but increased slightly in Fe wetlands. The total system reduced phosphate loads by 4%, 75%, and 74% and TDN loads by 34%, 52%, and 43% for Co, Al, and Fe systems, respectively. Average loads of phosphate greatly decreased in the coagulation systems with removals of

3.4 g / m^{2} / year

, whereas Co systems only removed

0.2 g / m^{2} / year

.

Dissolved oxygen in the non dosed inflow water averaged

4.2 mg / L

, decreased about

0.3 mg / L

within the control wetlands (to

3.9 mg / L

) and increased about 11%–13% in the coagulation wetlands (to

4.7 - 4.9 mg / L

). Passage through the wetlands decreased pH slightly from 6.47 to 6.41 in the Co wetlands (Table 1). Coagulation with Al decreased pH slightly from 6.47 to 6.42, which was reversed during wetland passage resulting in outflow pH of 6.47. A significant drop in pH was observed due to Fe coagulation, decreasing pH from 6.48 to 5.78; this was partially reversed in the wetlands with outflow pH of 6.15.

Kinetic Models

Results for the stepwise development of model constants are shown in Table 3. Eq. (1) predicted DOC changes across the Co wetlands (Step 1,

R^{2} = 0.78

) more effectively than when applied to all the wetlands together (Step 3,

R^{2} = 0.57

). The calculated K value for Step 1 was a smaller negative number than for Step 3, indicating less production of DOC in the Co wetlands than for all wetlands combined. This is consistent with data showing greater DOC production in the Al and Fe wetlands than in the Co wetlands (Tables 1 and 2). The addition of temperature effects [Eq. (2)] slightly increased performance of the model for the Co wetlands (

R^{2} = 0.82

, Step 2) and for all wetlands (

R^{2} = 0.58

, Step 4). The authors expected the model would improve with temperature because data showed concentration ratios correlated to temperature (Fig. 3). Temperature explained 56% of the variance for DOC ratios across Co wetlands and 29% and 53% for Fe and Al wetlands, respectively (

p < 0.05

).

Table 3. Kinetic model constants for dissolved organic carbon processing across the wetlands for the control (Co), and polyaluminum chloride (Al) and ferric sulfate (Fe) treatments

Step	Equation	Wetland(s)	Constant	Mean	SE	$R^{2}$
1	Eq. (1)	Co	$K$	$- 0.0132$	0.0036	0.78
2	Eq. (2)	Co	$K_{20}$	$- 0.0219$	0.0038	0.82
		Co	$Θ$	1.2336	0.1155	—
3	Eq. (1)	Co, Al, Fe	$K$	$- 0.0319$	0.0036	0.57
4	Eq. (2)	Co, Al, Fe	$K_{20}$	$- 0.0366$	0.0035	0.58
		Co, Al, Fe	$Θ$	1.0845	0.0365	—
5	Eq. (4)	Co, Al, Fe	$K_{20}$	$- 0.0219$	0.0038	0.83^a
		Co, Al, Fe	$Θ$	1.2336	0.1155	—
		Fe	$%_{Fe, 20}$	0.7578	0.0331	—
		Fe	$Θ_{Fe}$	1.1507	0.0122	—
		Al	$%_{Al, 20}$	0.2447	0.0332	—
		Al	$Θ_{Al}$	1.0533	0.0235	—

Note: $K$ = rate constants; $K_{20}$ = rate constants at 20°C; $Θ$ = temperature dependency of the rate; $Θ_{Al}$ = temperature dependency of release in Al treatments; $Θ_{Fe}$ = temperature dependency of release in Fe treatments; $%_{Al, 20}$ = coagulant-dependent release coefficients at 20°C in Al treatments; and $%_{Fe, 20}$ = coagulant-dependent release coefficients at 20°C in Fe treatments. Calculated constants are presented in means with standard errors (SE), along with coefficients of determination ( $R^{2}$ ) for model fit and probability value (p-value). Negative $K$ values represent net ecosystem production; positive $K$ values represent net ecosystem uptake. Larger $%_{M e, 20}$ (where Me is Fe or Al) represents larger constituent release from wetlands. Larger $Θ$ indicates larger temperature dependency of constituent processes.

a

R^{2}

for all wetlands combined was 0.83. Within the model,

R^{2} = 0.82

for the Co.

Fig. 3. Correlation of outflow/postdosing DOC concentration ratios ( ${DOC}_{out} / {DOC}_{pd}$ ) with temperature for the polyaluminum chloride (Al) and ferric sulfate (Fe) treatments, and controls (Co). Coefficients of determination ( $R^{2}$ ) and probability values (p) are shown.

Step 5 [Eq. (4)] incorporated DOC released across the wetland as a function of DOC removal by dosing. Since a time lag between the binding of DOC through floc formation and DOC release in the wetlands due to aging and biotic or abiotic processes would be expected, this expression is a simplification. However, DOC removal by coagulation showed a relationship with DOC released from the Al and Fe treatments in excess of what was released concurrently from the Co (Fig. S2). Incorporating this relationship into the model improved the overall fit of this model to

R^{2} = 0.83

(Step 5, Table 3) for all wetlands as compared to

R^{2} = 0.58

for the simpler model used in Step 4. Fig. 4 presents modeled outflow DOC concentrations (

{DOC}_{out}

) compared to measured

{DOC}_{out}

data; modeled

{DOC}_{out}

were predicted using Eq. (4) and measured inputs (DOC PD concentrations, HRT, temperature). The model was able to fit the Co and Al treatment data well (

R^{2} = 0.74 - 0.82

) but represented poorly the Fe treatment data (

R^{2} = 0.33

).

Fig. 4. Measured and modeled outflow dissolved organic carbon (DOC in $mg / L$ ) for the polyaluminum chloride (Al) and ferric sulfate (Fe) treatments, and controls (Co) plotted temporally for study period. Coefficients of determination ( $R^{2}$ ) between measured and modeled concentrations are shown.

The effects of HRT on wetland performance ratios and

{DOC}_{out}

based on model predictions are presented in Fig. 5. The model predicts that wetland performance ratios increased with HRT and the rate of increase (slope) is independent of treatment type [Fig. 5(a)]. Effects of HRT on

{DOC}_{out}

, unlike the ratio, is dependent on treatment because of the dependency of

{DOC}_{out}

on post-dosed DOC concentrations [

{DOC}_{pd}

, Eq. (1)]. With lower

{DOC}_{pd}

concentrations entering the Al and Fe wetlands than the Co wetlands, Al and Fe

{DOC}_{out}

was less sensitive to HRT and had a smaller positive slope than Co [Fig. 5(b)]. This study used Al wetlands to illustrate how temperature, dosing, and HRT affect DOC release during wetland passage [Figs. 5(c–f)]. Concentration ratios increased by nearly 50% due to temperature increase from 15 to 23°C under a 10-d HRT and corresponded to an increase in concentration [Figs. 5(c and d)]. Dosing rates, % removal by coagulation, significantly affected DOC release, with higher dosing rate resulting in increasing

{DOC}_{out} / {DOC}_{pd}

ratios [Fig. 5(e)] but decreasing DOC outflow concentrations [Fig. 5(f)].

Fig. 5. Modeled hydrologic retention time (HRT) plotted against dissolved organic carbon (DOC) in absolute concentrations ( $mg / L$ ) and outflow:postdosing concentration ratios ( ${DOC}_{out} / {DOC}_{pd}$ ) with (a and b) treatment effects; (c and d) temperature (°C) effects; and (e and f) coagulant dosing rates effects. Treatments are indicated by Al for polyaluminum chloride, Fe for ferric sulfate and Co for the control. Linear slopes are shown along corresponding trendline.

Discussion

Coagulation, Wetland, and Total System Performance

Coagulation

Removal of DOC from inflow water followed a seasonal trend, with the greatest removal on a mass basis during winter to spring months (October to May) when inflow DOC concentrations were higher than the other months (Fig. S3). This seasonality was expected since dosing rates were maintained to achieve 60%–80% DOC removal, driving removal trends to follow seasonal source water concentration trends. Also, resulting from the maintained dosing rate, both Fe and Al coagulants performed similarly prior to wetland passage with respect to removal of DOC and nearly all other constituents. Although similar removal of DBPPs was achieved by the two different coagulants, HAA precursors were more effectively removed than THM precursors (Tables 1 and 2). Removal preference for HAA has been demonstrated by other authors as well (Mourad 2008; Tubić et al. 2013) and believed to result because the fraction of DOC responsible for their formation have higher aromaticity (Liang and Singer 2003; Hansen et al. 2018). Coagulation is known to be effective in removing the fraction of the DOC pool that forms DBPs, particularly those associated with high molecular weight, aromatic compounds (Uyak et al. 2008; Matilainen et al. 2010; Roberts and Inniss 2014).

In addition to removing DOC and DBPPs, coagulation was responsible for significant amounts of phosphate and DON removal but did not effectively remove DIN. Ancillary data indicate pH was significantly lowered by Fe coagulation; pH drops are often observed with hydrolyzing metal-salts such as ferric sulfate. The effect arises from hydrolysis product formation shifting the solution equilibrium to

H^{+}

production (Chipperfield 2003; Duan and Gregory 2003). DAl and DFe loads decreased immediately after coagulation in both the Al and Fe treatments, indicating the ability of coagulants to incorporate dissolved metals into their floc structures. These results show the effectiveness of removing most constituents investigated; additionally, it demonstrates successful control and operation of CETW coagulation systems for extended periods using this relatively simplistic and easier to operate design. However, in a CETW system, constituent removal due to coagulation alone does not determine full removal potential as flocs retained within the wetlands will be subjected to natural conditions that may affect their structure and stability.

Control Wetlands

In free surface flow wetlands, constituent removal occurs through three main mechanisms: physical, chemical, and biological (Kent 2000). Wetlands promote physical removal of constituents through cultivation of laminar flow environments, which allows more efficient particle settling and prevention of particle resuspension; biological removal is effected through plant uptake; and chemical removal occurs though sorption by wetland components (soil, biological surfaces, etc.). While natural wetlands often improve water quality, constructed wetlands commonly produce and release DOC from physical disturbances during construction (Barber et al. 2001; Fleck et al. 2004; Díaz et al. 2012; Scholz et al. 2016). Rice fields, similar to seasonally flooded wetlands, have also been shown to produce DOC in both mineral soil (Bachand et al. 2014b) and in organic peat soil systems (Bachand and Associates et al. 2006). Though ineffective at DOC removal, constructed wetlands have been found to reliably remove phosphorous, nitrogen, and metals and to have the ability to decrease toxicity through water polishing (Bachand 1996; Bachand and Horne 1999; Richardson and Qian 1999; Lin et al. 2002; Fisher and Acreman 2004; Vymazal 2014; Mendes et al. 2018). However, the timing of these processes can be affected seasonally due to changes in hydrologic pathways caused by evaporation and transpiration effects (Bachand et al. 2014a, b). By considering these hydrologic processes (e.g., evapotranspiration, precipitation and seepage, Fig. S4), the authors were better able to investigate the effects wetlands and coagulation had on constituent processes.

Similar to what other authors have observed, this study found untreated Co wetlands contributed DOC and DBPPs and removed phosphate and DIN (Table 1). Contributions of DOC was dependent on season, with the majority of DOC release occurring during the spring to early autumn months corresponding to greater biological activity (Qiu et al. 2005; Miller and Fujii 2010; Hansen et al. 2018). Control wetlands were effective at removing TDN in the form of DIN with only minor DON removal. These results indicate that nitrogen cycling within the wetlands was leading to a net loss of N from the water column, likely through a combination of biotic N uptake and denitrification that exceeded N release via mineralization (Van Cleemput et al. 2007). Decreasing pH and DO levels suggested that nitrification, the rate-limiting step for denitrification, was also occurring. Stoichiometry for nitrification predicts about a 4:1 relationship between mass of oxygen consumed for mass of ammonia-nitrified (Metcalf and Eddy 2009; Liu and Wang 2012). Additionally, for each mole of ammonium converted to nitrate, the process creates about 2 moles of

H^{+}

atoms, causing a pH drop. Control wetlands also aided in the removal of DAl and DFe loads, but their removal was modest compared to the coagulation systems (Table 1).

Coagulation Wetlands

Load data indicate that the DOC and DBPP removal efficiency of the hybrid coagulation/wetland system decreased for both coagulation treatments after subjecting flocs to flooded wetland conditions. Release of DOC during wetland passage was greater in the Fe wetlands than Al wetlands (Table 1). Both coagulation wetlands showed year-round DOC release, unlike the seasonal release observed in the Co wetlands that occurred during spring to fall but not winter (Fig. S3). This suggests that other mechanism(s) of DOC release were occurring in the coagulation wetlands. Lowered removal efficiency can be due to a combination of natural (constructed) wetland processes that contribute DOC, as in the case of the Co, and constituent release from flocs. Fe treatment wetlands lowered annual DOC removal from 72% PD to 41% after wetland passage. This result, in addition to significantly higher wetland ratios (Table 2) and wetland loads (Table 1) compared to the control, suggests that Fe-flocs can be a significant DOC source under flooded wetland conditions. Disaggregation of Fe-flocs has been observed by others (Mikutta et al. 2014; Fritzsche et al. 2015). An abiotic laboratory study of Fe-flocs formed using the same Fe coagulant and source water as this study found Fe-flocs withstood crystallization and did not release DOC with changes in pH or redox (Henneberry et al. 2012). DOC release observed in this field study suggests other factors, such as structural changes of the flocs, pin-floc resuspension, or microbial electron transfer reduction, could be responsible for DOC release from these flocs under natural conditions.

To understand better the nature of this floc release, dissolved metal concentrations and their ratio to DOC at outflows were examined and compared with bulk floc measurements and literature findings. DFe concentrations in Fe wetlands increased

0.6 mg / L

(

0.011 mmol / L

, Table 1) from PD to outflow, an order of magnitude lower than the DOC increase (

6 mg / L

or

0.5 mmol / L

). Accordingly, the molar ratio of DOC to iron was 46.5; this high molar ratio suggests floc resuspension was not responsible for carbon release because data from floc measurements indicate bulk Fe-flocs contained molar ratios of 2.5 DOC:Fe. Chen et al. (2014) found that Fe was not released for floc C:Fe molar ratios lower than 4, suggesting neither dissolution nor dispersion of the floc. However, infrared data measured from flocs produced from this study indicate DOC released from within the floc mass was complexed with iron cations (Liang 2016). Considering both macroscale (DOC:Fe molar ratio) and microscale (infrared) data, higher DOC release from Fe wetlands was likely a combination of floc resuspension in addition to contributions of uncomplexed DOC released from the floc itself. Given the low amount of DFe released during wetland passage and high observed DOC:Fe molar ratios (Table 1), release appears in large part due to desorption of DOC from the floc rather than from floc resuspension. However, only a direct measure of the outflow DOC properties can confirm the actual nature of the released DOC and provide a more definitive explanation.

The Al wetlands released less DOC than the Fe wetlands; thus, the efficiency of DOC removal only decreased from 71% PD to 58% after Al wetland passage (Table 1). Even though Bachand et al. (2019) reported similar molar ratios of C:Al under Al dosing as C:Fe under Fe dosing, DAl loads after passage through the Al wetlands did not increase unlike DFe loads in the Fe wetlands (Table 1). This suggests DOC release in Al wetlands occurred through desorption only and/or from wetland contributions; this is supported by infrared data conducted on flocs from this study indicating no Al-DOC interaction (Liang 2016).

Along with DOC, THM and HAA precursors were concomitantly released into the water column during wetland passage in treatments and controls, and like DOC they were more readily released from the Fe wetlands than Al wetlands (Tables 1 and 2). This is not surprising as DBPP concentration is often strongly correlated with DOC concentration as demonstrated here (Fig. 2) and by others (Fuji et al. 1998; Díaz et al. 2008; Mourad 2008; Hansen et al. 2018).

Removal of DON during the coagulation step appeared to have impacted N cycling processes in the coagulation-treated wetlands. Removal of DON by coagulation seemed to have hindered N cycling and uptake in the wetlands as indicated by the lower DON consumption in coagulation treatments (Table 1). However, these mean values were not significantly different from the Co likely due to the small data set in this study (

N \approx 20

, Table 1). However, evidence from concurrent increased pH and DO in the coagulation wetlands would also suggest that nitrification may have been hindered in the coagulation wetlands (Metcalf and Eddy 2009). If nitrification was in fact disrupted, it remains unclear if this resulted from the lower DON concentrations in the water column of coagulation wetlands (less bioavailability) or a direct disruption to the microbial mass was being caused by the flocs. Since nitrification is the limiting step for denitrification, the authors expected to also see decreased opportunities for DIN removal in the coagulation wetlands. While this is also observed, the mean values were again not significantly different between treatments and control. Other factors, in addition to the nitrification limiting step, can also be responsible for lowered denitrification in coagulation wetlands, including high oxygen content (McKenney et al. 2001; Burgin and Groffman 2012) and less available labile organic carbon and nitrogen (Hume et al. 2002a, b; Dodla et al. 2008; Kadlec and Wallace 2008; Metcalf and Eddy 2009), all of which existed in the coagulation wetlands.

Since nitrogen (N) concentrations found in this study were generally low (EPA 2017b), coagulation effects on the N cycle are thought to be minor. Direct and indirect effects on N cycling and removal could have important implications if treating more N-rich source waters. Due to the complex nature of wetlands, the small nitrogen data set in this study and the low N levels contributing uncertainty, the effect of coagulation on N cycling in wetlands is speculative. Nevertheless, these preliminary results suggest that the direct and indirect effects of coagulation on wetland N cycling and its relationship to DIN, DON, DOC, and DO warrants further investigation.

Total System

Even though natural wetlands are known to provide an ecosystem that promotes particle settling and water purification (Kent 2000; Darke and Megonigal 2003), constructed wetlands can have the opposite effect and contribute constituents to the water column. This is often the case in the early stages immediately following wetland construction and have been demonstrated here and by others (Barber et al. 2001; Fleck et al. 2004; Díaz et al. 2012; Scholz et al. 2016). Coagulation can be used to offset these initial constituent contributions, but it is beneficial to understand how flocs formed through coagulation respond to wetland conditions and whether the two systems (coagulation and wetland) complement each other.

Dissolved Al loads decreased in coagulation and wetland steps for both coagulation treatments, showing great potential for removal of this constituent with CETWs. Removal of DFe was also synergistic between coagulation and wetland steps for Al treatments. Although Fe treatments showed opposing effects between coagulation and wetland steps for DFe, there was still net removal of this constituent compared to the Co. Both coagulation treatments showed opposing effects of coagulation removal and wetland release for DOC, DBPPs, and phosphates. Despite greater constituent released from the Fe wetlands compared to Al wetlands, both CETW systems (coagulation and wetland passage) resulted in net lowering of all constituent loads compared to the untreated control (with the exception of DIN, Table 1). Net removal of constituents has also been observed in similar coagulation-wetland systems by others (Malecki-Brown and White 2009; Malecki-Brown et al. 2009; Lindstrom and White 2011). In this study, the coagulation step was the main driver for load removal of all constituents, except for DIN (both coagulation treatments) and DAl (Al treatment only) where the wetland represented a significant removal step. The interaction between coagulation and wetland steps were synergistic for certain constituents and antagonistic for others, demonstrating the complexities found in CETW systems and show environmental fate determination of flocs are an important consideration under these hybrid systems.

Kinetic Models

Water temperature measurements in the wetlands can be used to indicate PFR conditions (Metcalf and Eddy 2009) and deviations from a linear temperature decrease through the wetlands suggest non-ideal PFR conditions (Figs. S5 and S6). Temperature changes were not entirely uniform, indicating preferential flow paths (Martinez and Wise 2003; Min and Wise 2009) and heterogeneous biogeochemistry (Miller and Fujii 2011). Kinetic models can be used to better understand these complex wetland processes. This study used a simple kinetic model to assess differences between treatments and controls through developing rate constants (Table 3). This approach is foundational throughout the literature for assessments of wastewater and water treatments (Metcalf and Eddy 2009) and evaluation of water quality changes across wetlands (Kadlec and Wallace 2008; Bachand et al. 2014a, b).

Wetland kinetic models using a modified Hoff-Arrhenius equation, despite its simplicity, provided important insights into wetland processes. The model assumes the same release rate due to normal wetland processes related to PD DOC concentration, HRT, and temperature for all wetlands. The authors investigated the sensitivity of the model to HRT, treatment wetland type (Co, Fe, Al), temperature, and dosing rate (% removal by coagulation). For the coagulation wetlands, the model also included a DOC release term to account for release from resident flocs.

Model Predictions and Insights

With the model, one gains additional insight into the system useful in planning wetland design, managing operations, and predicting performance by demonstrating outcomes. For instance, minimizing operational HRT can improve system performance (Fig. 5). In Al treatments, outflow total suspended solids (TSS) ranged from about 0.5%–18% of PD TSS regardless of HRT between 2 and 10 days (Fig. S7), but for Fe treatments, HRT more than 5 day was needed to consistently reduce outflow TSS to 15% or less of PD levels. Thus, greater settling times are required for the Fe flocs, resulting in higher HRT. If Al wetlands can be operated at a 2-day HRT, achievable mean

{DOC}_{out}

would be about

7 mg / L

with an average

0.2 mg / L

increase per additional day [Fig. 5(b)]. If Fe-flocs needs 5 days to settle, DOC levels would be over

17 mg / L

at the outflow [Fig. 5(b)]. This demonstrates that if high settling rates can be achieved, then lower DOC and DBPP levels will be discharged because of lower required HRT.

The model also showed that operation of Al wetlands was most beneficial at higher dosing rates (75% removal) because not only did it remove more DOC through coagulation, it also suppressed the effects of DOC release from longer HRT indicated by the smaller positive slope [Fig. 5(f)]. Suppression of DOC release due to increasing HRT was similarly observed for lower temperatures [Figs. 5(c and d)]; this resulted from the temperature dependency of DOC release from the wetlands (Fig. 3). This temperature dependence suggests that the mechanism(s) of release is partly due to thermally dependent processes. Mechanisms such as plant and microbial activity are known to have thermal dependencies and play a part in DOC release (Freeman et al. 2001; Fenner et al. 2005; Jerman et al. 2009), though it is not clear here if these are the responsible release mechanisms. Similar results were seen for Fe dosing (data not shown), although there was greater DOC release from Fe wetlands. Many of these outcomes modeled for DOC would be expected to hold true for DBPPs also since THM and HAA are strongly correlated with DOC (Fig. 2,

R^{2} > 0.80

). Overall, the wetland kinetic model captured 83% of the variance in outflow concentrations for treatments and controls combined (Step 5, Table 3). The model was a good fit for the Co and Al wetlands, but poorly represented the Fe wetlands. The good fit for the Co showed this modeling approach captured wetland processes well, even given the heterogeneity and preferential pathway complexity. Weaker model fits observed for Fe treatments were partially due to the greater variance of DOC concentrations between cells in this treatment (Fig. S8), whereas measured

{DOC}_{out}

from both the Co and Al treatment was replicated during the study and had very similar trends throughout (Fig. S8). This higher variability indicates greater biogeochemical interactions and complexity between the Fe-flocs and wetland processes, demonstrating the need for consideration of additional parameters not measured in this study (e.g., microbial activity, oxidation states).

Practical Implications and Further Considerations

Monitoring discharge concentrations for water quality constituents of concern is the primary method for managing pollutant loading to water bodies for point discharges (EPA 2017a). When those limits cannot be met, or deemed inappropriate, the Clean Water Action Section 303(d) (Impaired Waters and Total Maximum Daily Loads – TMDLs) assists states, territories, and authorized tribes to develop TMDLs to help restore water quality in impaired waterbodies (McCord and Heim 2015). This study shows relatively simplistic hybrid coagulation wetlands can successfully treat agricultural drain water and represents a potential solution for water quality improvement. Field results show that polyaluminum chloride CETWs would be a more effective treatment if DOC and DBPP removal from agricultural drainage water are of primary concern; however, this treatment was also better suited to treat many of the other constituents considered in this study.

For coagulation selection and optimization, jar test are often performed and in laboratory DOC removal studies, ferric sulfate was shown to outperform PAC (Mourad 2008; Henneberry 2012). Laboratory studies subjecting flocs to varying pH and redox conditions to simulate field conditions indicated both types of flocs to be equally resilient and no DOC was observed to be released (Henneberry 2012). Moreover, both coagulants were equally effective in constituent removal during the postdose stage in this field study. The vastly different behavior between the two coagulation treatments under field conditions was unanticipated but reveals the importance of field studies for understanding the environmental fate of flocs. Recent estimates show that approximately 31% of drinking water treatment residuals produced in the United States are disposed of by discharge into surface waters (USEPA 2011), the fate of which are largely unknown after disposal. In addition to discharge to surface water, landfill burial and land application are common management options for treatment residuals. This study showed that coagulant choice can have large implications for the fate of these treatment residuals.

Kinetic models, such as the ones applied here, are useful for informing management practices and predicting outcomes for wetland implementation. These models indicate operation of CETW systems when temperatures are low will limit DOC release from temperature-dependent wetland processes and that higher dosing rates are desirable in keeping outflow DOC concentrations low. To minimalize DOC production from constructed wetlands, HRTs should be kept no higher than required to allow settling of flocs. Furthermore, lower HRTs require smaller operational footprints for the same volume of water treated. Although both coagulation treatments reduced outflow loads of DOC and DBPPs from drain water sources and performed better than control wetlands, the Al CETWs provided a more robust and predictable performance compared to Fe CETWs by the production of faster settling flocs that remained settled under field conditions.

Changes in water quality entering wetlands and floc retention can potentially have important impacts on wetland processes and species. Data here suggested a potential disruption of the mineralization/nitrification/denitrification cycle in wetlands, but this data set was limited and more studies are certainly needed. Effects on other wetland processes not measured here could have also resulted from coagulation treatment. For example, pH is a common determinant of speciation, and ultimately the fate, of many wetland constituents; these changes could have compounding effects and result in toxicity within the wetlands and downstream (Kadlec and Wallace 2008; Wang et al. 2016). Related studies showed that water quality and floc retention in the wetlands did not have any negative impacts on Typha or mosquitofish viability and growth (Ackerman et al. 2015; Liang et al. 2018). While there was a lack of negative impact on studied biological species in this project, other studies have found that exposure to PAC flocs can produce higher aluminum accumulation in submerged aquatic vegetation and cause decreased microbial activity in the surface of soils (Malecki-Brown et al. 2007, 2010; Malecki-Brown and White 2009). In determining the applicability of CETWs for any scenario, all impact factors (in addition to others not mentioned here) should be considered under cost-benefit analyses.

Conclusions

CETW systems have the potential to remove DOC, DBPPs, DAl, DFe, dissolved organic nitrogen, and phosphate in excess of what can be removed by untreated control wetlands. Coagulation, with Al-based and Fe-based coagulants, was primarily responsible for constituent removal from the agricultural drain waters. In contrast, untreated control wetlands were found to produce DOC and DBPPs while the remaining constituents were removed at modest or low levels. Wetland effects were shown to be synergistic with coagulation for DAl and DFe removal in coagulated treatments, except for DFe in the Fe treatment, which showed release during wetland passage. For DOC, DBPPs and phosphate, wetland passage reversed some of the benefits achieved by coagulation. There is evidence to suggest that a portion of the DOC, and associated DBPPs, released from the Fe treatment occurred from the Fe-flocs destabilizing in the wetlands. Release of DOC was found to be temperature-dependent, suggesting thermally dependent mechanisms were partially responsible for the release. Kinetic plug flow reactor models were used to evaluate how relevant parameters affect constituent behavior and determine operational settings to optimize removal and process efficiency for these systems. Based on insights from the model, DOC release from wetlands can be minimized by decreasing HRT, increasing dosing rate, and operating at low temperatures. Low temperatures and high dosing rates had an additional suppression effect on DOC release due to higher HRTs; this was predicted to result in the lowest DOC concentrations leaving the wetlands. Overall, coagulation-treated wetlands improved water quality more than untreated wetlands and Al CETWs provided a more effective and robust treatment for constituent removal than Fe CETWs and control wetlands.

Supplemental Data

Figs. S1–S8 and additional supporting information are available online in the ASCE Library (www.ascelibrary.org).

Notation

The following symbols are used in this paper:

$C_{i n}$: inflow concentration;
$C_{o u t}$: outflow concentration;
$C_{p d}$: concentration at inflow postdosing;
$C_{r}$: concentration change due to desorption;
$K$: rate constant;
$K_{20}$: rate constant at 20°C;
$Δ C_{M e}$: concentration change due to coagulation;
$θ$: temperature dependency of the rate;
$θ_{M e}^{T - 20}$: temperature dependency of release at 20°C, where Me indicates treatment (Fe or Al); and
$%_{M e}$: floc release coefficient, the percentage of removal that is reversible, where Me indicates treatment (Fe or Al).

Supplemental Materials

File (supplemental_data_ ee.1943-7870.0001536_bachand.pdf)

Download
962.88 KB

Acknowledgments

We are grateful for funds provided by the California Department of Water Resources (Agreement No. 46000003886/7W4CA4000003886) and matching funds provided by the USGS Cooperative Water Program. We would like to acknowledge Timothy A. Doane for his lab support during this project; Fred Simms for providing expertise on coagulants and their characteristics and difference; Brett Offerman for helping us work through the logistics of coagulant delivery and use; DWR for providing the field site; Elizabeth Stumpner and Angela Hansen for their field and lab support; Jim Casey and Bruce Gornto for field assistance, field set up recommendations and expertise on field management; Marc Estrade, Jeff Schuyler and Eyasco for setting up the extensive electronic system in the field and helping with troubleshooting; Robert Pedlar and Genevieve Schrader for their support and expertise throughout; Reilly Hossner for her support with this manuscript. Finally, we appreciate the hard work of Dr. Roger Fujii of the USGS who supported Dr. Bachand through co-writing the proposals for this project, worked with DWR to implement this project, and supported the project team efforts throughout. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the US Government.

References

Ackerman, J. T., T. E. Kraus, J. A. Fleck, D. P. Krabbenhoft, W. R. Horwath, S. M. Bachand, M. P. Herzog, C. A. Hartman, and P. A. Bachand. 2015. “Experimental dosing of wetlands with coagulants removes mercury from surface water and decreases mercury bioaccumulation in fish.” Environ. Sci. Technol. 49 (10): 6304–6311. https://doi.org/10.1021/acs.est.5b00655.

Abstract

Introduction

Methods

Wetland Design and Treatments

Water Monitoring, Sampling, and Analyses

Water and Mass Budgets, Calculations and Statistical Analyses

Kinetic Model

Results

DOC and DBPP Concentrations, Loads and Concentration Ratios

Ancillary Parameters, Nutrients and Metals Concentrations and Loads

Kinetic Models

Discussion

Coagulation, Wetland, and Total System Performance

Coagulation

Control Wetlands

Coagulation Wetlands

Total System

Kinetic Models

Model Predictions and Insights

Practical Implications and Further Considerations

Conclusions

Supplemental Data

Notation

Supplemental Materials

Acknowledgments

References

Information

Published In

Copyright

History

Authors

Affiliations

Metrics

Citations

Download citation

Cited by

Figures

Other

Share

Copy the content Link

Share with email

Share

Request Username

Create a new account

Change Password

Password Changed Successfully

Verify Phone

Congrats!