Size and resin fractionations of dissolved organic matter and characteristics of disinfection by-product precursors in a pilot-scale constructed wetland

Controlling the formation of disinfection by-products (DBPs) is a major issue in the drinking water industry, and understanding the characteristics of DBP precursors in treatment processes for micro-polluted raw water is key to improving water quality. In this study, a sampling program was undertaken to investigate the fate of dissolved organic matter (DOM) and the characteristics of DBP precursors in a pilot constructed wetland imitating the Yanlong Lake ecological project. Using XAD resin adsorption and ultrafiltration techniques, the dissolved organic carbon, UV254, and DBP formation potential (DBPFP) were measured in different DOM fractions in raw water and wetland effluents. After the constructed wetland treatment, the low molecular weight fraction (<3 kDa) of DOM and DBPFP generally showed a decreasing trend along the water path, while the high molecular weight fraction (>3 kDa) of DOM increased. The specific DBPFP (SDBPFP) was much higher in the <1 kDa fraction than in the other fractions. Although the hydrophobic fraction of DOM was the most abundant in all stages of the wetland treatment, the SDBPFP of the hydrophilic fraction was higher than that of the hydrophobic fraction. Furthermore, compared with raw water, the DOC, UV254 and DBPFP in the treated wetland effluents increased; however, all of the chemical DOM fractions exhibited decreased SDBPFP in accordance with a decrease in the specific ultraviolet absorbance during wetland treatment. These conclusions indicate that the DOM produced by thewetland system may generate DBPs less readily compared with the DOM of raw water.


INTRODUCTION
In China, due to rapid development of industry and agriculture, approximately 60% of urban drinking water sources are polluted to varying degrees (Wang & Wang ). To ensure the safety and quality of drinking water and respond to sudden pollution incidents, construction of storage reservoirs incorporating artificial wetlands and other ecological mitigation measures has become an economically favorable alternative to energy-intensive engineered treatment plant approaches to improving the quality of raw water (Haynes   () also demonstrated that aromatic compounds react easily with chlorine to form DBPs such as THMs and HAAs. Yang et al. () used a vertical subsurface flow constructed wetland and surface wetland tandem system to treat micro-polluted raw water in the Yangtze River and found that the THM formation potential (THMFP) of the system effluent increased by 20.52% compared with the influent water. DOM in the roots and leaves of plants also has an effect on DBP production, and soluble microbial products and aromatic proteinaceous substances (polyphenols) can enhance the THMFP (Wei et al. ). In addition, aquatic animals often breed prolifically in ecological engineering sites, and the amino acids, proteins, and fats contained in their metabolites can also act as precursors for DBPs (Sun et al. ). However, although previous studies have indicated that the DOM produced by plants and animals in wetlands may act as precursors for DBPs, field investigations and characterization of DOM in constructed wetlands have received only limited research attention.
DOM from wetlands is a heterogeneous mixture of complex organic materials including humic substances, hydrophilic acids, proteins, lipids, carboxylic acids, polysaccharides, amino acids, and hydrocarbons (Leenheer & Croué ; Cheng et al. ). It is impossible to identify and investigate these compounds individually; therefore, the preferred method for evaluating DPB precursors is classification of DOM in water bodies based on a given characteristic and measurement of the reaction behavior of the DOM with this characteristic. Resin adsorption (RA) and ultrafiltration have been widely and successfully applied in characterizing the chemical and physical properties of DOM from natural waters (Wei et al. ). In particular, the XAD-8 and XAD-4 resins have been widely used to separate DOM into hydrophobic (HPO) and hydrophilic (HPI) components (Leenheer ).
The purpose of this work is to evaluate the behaviors and characteristics of DBP precursors in three types of wetlands that form an ecological engineering system for a water source with micro-polluted raw water, by analyzing the molecular weight distribution and chemical fractionation of DOM. Dissolved organic carbon (DOC) and UV 254 values were measured to investigate the variation patterns of different molecular weight and chemical fractions. Additional experiments were conducted to evaluate the DBP formation potential (DBPFP) of each weight and chemical fraction in raw water and wetland effluent.

Constructed wetlands
Experiments were carried out in three different types of constructed wetlands: a surface wetland, submerged plant pond, and ecological pond, hereafter named wetland A, B, and C. The constructed wetlands ( Figure 1) were located within a pilot plant on one of the inflow streams of a drinking water reservoir in Yancheng City, Jiangsu Province, China, designed to imitate the Yanlong Lake ecological project (a replicated field-scale study). In this pilot plant, raw water from the Viper River is driven into a high water tank by a lifting pump and then passed through wetland A, B, and C successively under gravity. A schematic diagram of the composite constructed wetland is presented in Figure 1.  Table 2 were transplanted from a local natural field to the wetlands. In addition, a non-classical biological manipulation technique was adopted in wetland C: silver carp and bighead carp were added to control the algae density in the raw water and prevent eruption of cyanobacteria, which are detrimental to water quality. The total density of carp was 30 g m À3 , and the quantity ratio of silver carp to bighead carp was 2:1. Some of the wetland parameters and installation compositions are listed in Table 2. After planting and enriching with fish, the influent flow was gradually increased from 0.3 to 0.6 m 3 days À1 over three months, and the plants were allowed to grow freely. When the plants were well established and the constructed wetlands stabilized, the investigation commenced. The influent flow to the wetlands system was set to 0.6 m 3 days À1 , corresponding to a theoretical hydraulic retention time of 21 days, based on the actual hydraulic retention time of Yanlong Lake of approximately 21 days and 10 hours (Xu et al. ). There were four sampling points at the entrance and exit of each wetland, as shown in Figure 1.
All data in this study were collected in April 2018.  fraction, the effluent from the XAD-8 resin containing HPI acids, bases, and neutral compounds was then passed through the XAD-4 resin. The TPI fraction was obtained by eluting the XAD-4 resin with the same eluent used to wash the XAD-8 resin. The soluble organics passing through the XAD-4 resin were the HPI fraction, containing HPI bases and neutral compounds (not retained on either the XAD-8 or XAD-4 resin). All separated water samples were adjusted to pH 7.0 (±0.2) using HCl or NaOH.

Procedure of DOM fractionation
Another set of water samples was fractionated to different molecular weight classes using a stirred ultrafiltration cell (Millipore, 8400) with YM disc membranes (Amicon, nominal molecular weight cut-offs are 0.5, 1, 3 and 10 kDa). All membranes were rinsed with ultrapure water to ensure a residual DOC concentration of <0.2 mg·L À1 .
The DOC mass balance of our size fractionation had been controlled in (100 ± 5)% following the cleaning procedure used by Guo & Santschi () and Wei et al. (). A schematic of the specific separation steps is presented in Figure 2(b). Then, DOC, UV 254 , and DBPFP were measured for all fractionated samples.

DOM and DBPFP measurements
A total of five water samples were collected in precleaned 250 mL glass bottles at each sampling location on the sampling date in April 2018. The samples were then mixed in a larger bottle and cooled immediately in an ice cooler.
After filtration through a 0.45 μm cellulose membrane filter, the raw water and wetland effluent samples were stored in a dark room at 4 C to minimize changes in the constituents. When ready for analysis, the samples were   Table 3.

RESULTS AND DISCUSSION
Fraction distributions of the DOM The molecular weight distributions of DOM in raw water and wetland effluents are presented in Figure 3(a). The low molecular weight fraction (<3 kDa) of DOM generally   components, and that the aromatic structure present in HPO substances is the main precursor of THMs (Rook ).
Therefore, the SUVA results showed that the wetland treatments increased the quality of aromatic materials but decreased the aromaticity, which may reduce the DBP level produced per unit mass.

Size and chemical fractions of DBPFP
The THMFP and HAA formation potential (     This indicated that the DBPFP of DOM generated in the wetlands is smaller than that of the raw water. In other words, our findings implied that the DOM produced by the constructed wetland system may not readily generate DBPs during chlorination.