Abstract
Disinfection by-products (DBPs), generated from the reaction of disinfectants with DBP precursors, have been found to pose unintentional risks to human health. Considering that the concentration and speciation of DBPs formed during disinfection will be affected by the content and composition of dissolved organic matter (DOM), widespread concern about the characteristics of DBP precursors in water sources have been prompted. Three-dimensional excitation–emission matrix (EEM) can quickly and efficiently determine the properties and composition of DOM in water, and thus is generally used to investigate the origin of DBP precursors in water sources. This study overviews the fluorescent properties of different DBP precursors, summarizes the application of different EEM interpretation methods in DBP precursors and analyses the key factors affecting the correlation between the fluorescent components and DBP precursors (e.g., natural organic matter, algal organic matter, effluent organic matter and organic matter derived from other sources). A series of factors, including composition of fluorophores, bromide concentration, spatio-temporal characteristics and disinfectant types, could impact the correlation between DBP formation potential and fluorescent components. As for future research needs, it is of significance to select suitable fluorescence analysis methods and investigate the combination of EEM with other characterization technologies based on different situations.
HIGHLIGHTS
The main substances and sources of DBP precursors in water are introduced.
Fluorescence analysis methods for DBP precursor are summarized.
The fluorescence properties of DBP precursors from different sources are analyzed.
The correlation between fluorescent substances and DBP precursors are discussed.
INTRODUCTION
Disinfection of drinking water could effectively control pathogenic microorganisms and prevent the outbreaks of waterborne diseases (Calderon 2000; Chen et al. 2021). However, the chemical risks arising from the interaction of disinfectants with organic substances in water have been widely concerned since the first detection of hazardous disinfection by-products (DBPs) in chlorinated drinking water (Bellar et al. 1974). Epidemiological studies have shown that DBPs could cause health risks and the related issues include bladder cancer, rectal cancer, colon cancer and unfavorable pregnancy defects (Richardson et al. 2007; Bond et al. 2012; Postigo et al. 2018). So far, more than 700 DBPs have been discovered in drinking water (Richardson et al. 2007; Yang & Zhang 2016), and the dominant carbonaceous DBPs (C-DBPs) such as trihalomethanes (THMs) and haloacetic acids (HAAs) have been widely regulated (Richardson 2011). However, relevant research indicated that unregulated nitrogenous DBPs (N-DBPs), including haloacetonitriles (HANs) and haloacetamides (HAMs), exhibited higher toxicity than the regulated C-DBPs (Muellner et al. 2007; Wagner & Plewa 2017). In addition, some unregulated phenolic and aromatic DBPs have also been reported to be more toxic (Liu & Zhang 2014; Han et al. 2021).
The concentration and speciation of DBPs during disinfection would be influenced by the nature and composition of dissolved organic matter (DOM) in water. Natural organic matter (NOM) is the primary DBP precursors in water sources, leading mainly to the generation of C-DBPs. However, water sources impacted by algal blooms or municipal wastewater discharges have also been utilized in some cases to meet the increasing demand for water resource due to the growth of population. Unlike terrestrial-derived NOM with low organic-nitrogen content (mostly less than 5% of dissolved organic carbon [DOC] by weight), effluent organic matter (EfOM) derived from wastewater treatment plants (WWTPs) and algal organic matter (AOM) produced by cell exudation or lysis are composed of macromolecules and cellular debris with a nitrogen-enriched protein fraction, and could serve as significant N-DBP precursors (Leenheer et al. 2007; Shah & Mitch 2012; Liu et al. 2022; Wang et al. 2023). In addition, some anthropogenic micro-pollutants including pesticides, industrial chemicals, pharmaceuticals and personal care products (PPCPs) could also form DBPs during the disinfection process (Chu et al. 2016; Tang et al. 2020; Dong et al. 2021).
The determination and characterization of DBP precursors could assist in optimizing treatment processes to improve the removal efficiency of DBP precursors (Krasner et al. 2007; Chu et al. 2013; Wang et al. 2021b; Luo et al. 2022). Methods used for DOM characterization mainly include spectroscopic methods, chromatographic methods, mass spectrometric methods and bulk parameter determination (Matilainen et al. 2011). General parameters such as DOC, ultraviolet (UV) absorbance, and specific UV absorbance (SUVA) could be applied to indicate the overall levels of organic DBP precursors (Chowdhury et al. 2009). However, the composition and characteristics of DOM, which is a complex mixture of humic substances, carbohydrates, amino acids, as well as other organic fractions, cannot be fully determined by these bulk parameters (Her et al. 2003; Peleato & Andrews 2015). Besides, chromatographic methods need select appropriate separation methods and mass spectrometric methods hardly interpret the data due to complexity of NOM (Matilainen et al. 2011), whereas fluorescence methods have faster online detection capability and could be more easily explained (Peleato et al. 2018).
EEM can show the composition and relative content of organic substances with different properties in water and is very sensitive to the change of chemical properties of organics in water during disinfectant oxidation (Bieroza et al. 2012; Pifer & Fairey 2012; Yang et al. 2015a, 2015b; Nguyen et al. 2023). The information obtained by EEM spectrum is generally interpreted through the following methods including peak picking (Coble 1996), fluorescence indices (McKnight et al. 2001), fluorescence regional integration (FRI) (Chen et al. 2003), principal component analysis (PCA) (Peiris et al. 2010), and parallel factor analysis (PARAFAC) (Stedmon et al. 2003; Yang et al. 2015a, 2015b). Peak picking and fluorescence indices use the fluorescence intensity and intensity ratio of the representative positions to establish a relationship with DBP precursor (Massalha et al. 2018; Roccaro et al. 2020), while FRI using large masses of fluorescence data as well as PCA and PARAFAC providing significant dimensionality reduction based on chemometric methods have also been used in different water bodies (Peleato & Andrews 2015; Jutaporn et al. 2021). The neural network (NN) approach allowing for non-linear dimensionality reduction of fluorescence spectra without explicit constraints has less use in the DBP field (Bieroza et al. 2011; Peleato et al. 2018). In general, the development of modern data analysis methods also provides more diversified choices for fluorescence data analysis and is widely applied to characterize DBP precursors in water sources (Fernández-Pascual et al. 2023).
A large amount of studies have used EEM to characterize DBP precursors in different water sources and analyze the change of DBP precursors arising from different treatment processes. However, systematical review is still lacking. This paper aims to (1) discuss the principle of EEM application in characterizing DBP precursors, (2) analyze the correlation between DBP precursors and fluorescent components, (3) overview different data analysis methods used for DBP precursor analysis. Besides, future research needs concerning improvement of fluorescence methods for DBP precursor analysis are proposed.
SOURCE OF DBP PRECURSORS
Humic acid and fulvic acid, accounting for 50–90% of the total DOC, are hydrophobic substances and serve as the main precursors of THMs (Thurman et al. 1989). Compared to fulvic acid, humic acids have more active sites that can react with chlorine and thus yield more THMs (Peters et al. 1980), and the active sites including reactive phenolic structures and high aromatic carbon content contribute to the formation of organohalides during chlorine disinfection (Singer 1999). Amino acids, an important hydrophilic fraction of NOM, have also been proven to serve as precursors of THMs and HAAs, and their concentrations in treated water varied from a few to several hundred nmol/L (Hureiki et al. 1994; Bond et al. 2009). The amino acids with side groups, like activated aromatic rings, amino or sulfur, react more easily with chlorine (Hureiki et al. 1994). In addition, these nitrogen enriched compounds such as amino acids and amines may also contribute to the formation of highly toxic N-DBPs (Bond et al. 2012). Thus, the concentration and speciation of DBPs formed during disinfection will be affected by the composition of DOM, which is related to the source of organic matter. In general, humic acid and fulvic acid accounted for most of NOM and were more likely to generate C-DBPs during disinfection (Shah & Mitch 2012), while soluble microbial products (SMPs) produced by bacteria during wastewater treatment as well as AOM liberated from algae are comprised of nitrogen-enriched compounds and serve as significant N-DBP precursors (Westerhoff & Mash 2002).
A summary of traditional DOM peaks of three different organic sources: (a) NOM; (b) EfOM; and (c) AOM. Adapted from Chen et al. (2017a, 2017b); Guo et al. (2014); Wang et al. (2015).
A summary of traditional DOM peaks of three different organic sources: (a) NOM; (b) EfOM; and (c) AOM. Adapted from Chen et al. (2017a, 2017b); Guo et al. (2014); Wang et al. (2015).
DATA ANALYSIS METHOD FOR FLUORESCENCE CHARACTERIZATION OF DBP PRECURSORS
Fluorescence data of DOM in different water sources are inconsistent and complex, consisting of hundreds of fluorescence intensities under the different excitation/emission (Ex/Em) pairs. Specific data analysis methods can provide more key information better concerning DBP precursors in water.
Analysis methods based on peak intensity and intensity ratio
Peak picking
Peak picking is a quantitative method based on selecting the fluorescence intensity at a certain excitation/emission wavelength of the regions of interest, and the fluorophore types of the regions in the original data are identified through pre-defined region classification (Coble 1996). The corresponding chemical compositions of different regions are shown in Supplementary material, Figure S1. In addition, the ratios between fluorescence intensities are often used to express changes in the composition of organic matter in water. The information of typical fluorescence intensity ratios is listed in Supplementary material, Table S1.
The maximum fluorescence intensity within the peak region was usually applied to connect with DBP formation potential (DBPFP) even though the location of the maximum may vary due to different water quality (Hao et al. 2012b; Park et al. 2019). Thus, the existence and intensity of some peaks identified by this method can be used as the characteristic factor of DBP precursors in water (Hua et al. 2007; Hao et al. 2012b). In addition, the differences in peak positions can provide information on the changes in the properties of DBP precursors during the biochemical changes in samples (Peiris et al. 2011; Luo et al. 2022). Compared to the individual peak intensity, the peak intensity ratio can reflect the relative content of different components in organic matter and reveal changes in the composition of organic matter after different biochemical processes, which can provide an understanding of DBP generation (Ritson et al. 2014; Hohner et al. 2016; Cai et al. 2020).
Fluorescence regional integration
Fluorescence spectra consist of thousands of data points of fluorescence intensity depending on the wavelength, but quantitative technique based on peak picking and other applications use only one to three data points ignoring other available information (Chen et al. 2003; Peleato & Andrews 2015). FRI is a quantitative technique by integrating all fluorescence intensity of divided region of an EEM. An EEM map is divided into five main regions as shown in Supplementary material, Figure S2. Normalized excitation–emission area volumes beneath each region can express the contents of each region representing the type of substance and percent fluorescence response can show the relative content of each sort of substance in the sample.
The change of integrated intensity in different regions in FRI can provide insight into the location of DBP precursors in EEM (Johnstone & Miller 2009). Some studies tend to establish regression analysis between the different DBP produced after disinfection and the fluorescence intensity integral of different regions to find the source of DBP precursors and predict the formation of DBPs (Roccaro et al. 2020). In addition, FRI can reflect the changes of components in different regions during water treatment which can be used to analyze the changes of DBP precursor sources after treatment (Meng et al. 2016; Fan et al. 2020).
Analysis methods based on data dimensionality reduction
Principal component analysis
PCA is a multivariate method that decomposes a data matrix into a series of linear terms and a residual matrix (Wold et al. 1987). PCA can transform lots of interrelated variables into independent new principal components (PCs) with the directions of maximum variance of the combined columns of the data matrix to provide significant dimensionality reduction (Persson & Wedborg 2001). The PCs are calculated in order of decreasing variance to contribute to using the first few variables to realize the variance interpretation of a data matrix consisting of a large number of variables. According to the different input variables, the EEM data analysis methods based on PCA are mainly divided into two types, including the fluorescence intensity corresponding to each excitation/emission wavelength and the fluorescence variables obtained by other data analysis methods. The loadings of PCs are directly linked to the fluorophores present and thus provide a basis for interpreting structural differences.
The scores of PCs based on PCA can be applied to predict the formation of DBPs in different source water. Each PC resulting from dimensionality reduction of EEM is composed of different fluorescent substances in water rather than individual component in the fluorescence spectra. However, the high loading values of spectral regions of PCs can be used to identify representative substances of each PC, which can provide insight into the types of DBP precursors (Peleato & Andrews 2015; Peleato et al. 2018). In addition, PCA is also applied to track the correlation between some fluorescence parameters like peak intensity and intensity ratio gained by other analysis methods and DBPFP (Nguyen et al. 2021).
Parallel factor analysis
PARAFAC is also a multivariate analysis technique that decomposes the data matrix into a set of trilinear terms and a residual array (Stedmon et al. 2003). Compared with two-way PAC, three-way PARAFAC analysis can obtain more adequate, robust and interpretable models to account for the three-dimensional nature of EEMs by using constraints of unimodality and non-negativity. In PARAFAC models, the overlapping fluorescence signals formed by different organics in water are decomposed into several interpretable components representing a group of fluorophores of similar, specific fluorescence properties (Stedmon & Markager 2005). The tutorials based on different toolboxes for MATLAB have been provided to help practical application of PARAFAC to fluorescence datasets (Stedmon & Bro 2008; Murphy et al. 2013). A diagram summarizing the different necessary steps in the fluorescence-PARAFAC analysis was shown in Supplementary material, Figure S3. These analysis methods help provide a reasonable model reflecting the composition of organics in water, but it is worth noting that the fluorescence intensity of the components obtained from the analysis only reflects the relative concentration of organics in the water.
The loadings at different excitation and emission wavelengths displayed by components separated by PARAFAC can be used to indicate the type and source of DBP precursors and the maximum fluorescence intensity of components can be used to indicate the relative content of DBP precursors. The separated fluorophore signal representing a single substance in water had higher correlation with DBPFPs than the conventional indicators like DOC and SUVA (Jutaporn et al. 2021). In addition, combined parameters from different fluorescence components including addition and ratio between components were more flexible and accurate in predicting the formation of DBP (Pifer & Fairey 2014; Xia et al. 2018). Compared to other methods, PARAFAC tend to require more data samples to ensure the credibility of the model. The collection of sample data over a long period of time made the application of PARAFAC more accurate and sensitive in the long-term monitoring process (Jutaporn et al. 2021; Li et al. 2021a, 2021b).
Artificial neural networks
An artificial neural network (ANN) is a powerful mathematical model composed of a series of the single processing elements (nodes, neurons) connected with each other. In an active neuron, each input vector is multiplied by its weight and the product is summed with the bias, and the output is generated through the transfer function which is generally a non-linear function (Bieroza et al. 2011). Environmental conditions such as pH or temperature, inner filter effects (IFEs) and the Rayleigh and Raman scatter can all affect the analysis results of PCA and PARAFAC based on linear dimensionality reduction, while some constraints eliminating these effects may limit the overall precision of reconstruction based on the condensed representation (Peleato 2022). Therefore, an NN approach based on non-linear regression may reduce the dimensionality of fluorescence spectra without explicit constraints, which gradually is used for accounting for fluorescence high-dimensionality and superposition of EEM. In addition, the fluorescence parameters provided by other analytical methods such as PCA and PARAFAC can be used as the input of the ANN model to provide the prediction of DBP precursors. The application of these methods is summarized in Table 1.
The principle and drawbacks of different data analysis methods for fluorescence characterization of DBP precursors
Method . | Principle . | Drawbacks . | Critical references/tutorial . |
---|---|---|---|
Peak picking |
| Coble (1996), Hao et al. (2012a) | |
FRI | Mathematical integration |
| Chen et al. (2003), Li et al. (2020) |
PCA | Alteration of the linear space |
| Bridgeman et al. (2011), Stedmon et al. (2003) |
PARAFAC | Least squares method |
| Murphy et al. (2013), Peleato & Andrews (2015), Stedmon & Bro (2008) |
SOMs | Artificial neural network |
| Bieroza et al. (2009), Li et al. (2020) |
Autoencoder neural network | Artificial neural network |
| Peleato et al. (2018) |
CNNs | Artificial neural network |
| Peleato (2022) |
Method . | Principle . | Drawbacks . | Critical references/tutorial . |
---|---|---|---|
Peak picking |
| Coble (1996), Hao et al. (2012a) | |
FRI | Mathematical integration |
| Chen et al. (2003), Li et al. (2020) |
PCA | Alteration of the linear space |
| Bridgeman et al. (2011), Stedmon et al. (2003) |
PARAFAC | Least squares method |
| Murphy et al. (2013), Peleato & Andrews (2015), Stedmon & Bro (2008) |
SOMs | Artificial neural network |
| Bieroza et al. (2009), Li et al. (2020) |
Autoencoder neural network | Artificial neural network |
| Peleato et al. (2018) |
CNNs | Artificial neural network |
| Peleato (2022) |
The current ANN methods used for DBP analysis include self-organizing maps (SOMs), autoencoder NN and convolutional neural networks (CNNs). SOM is an unsupervised classification algorithm and two-layered ANN, which could provide dominant fluorescence features and the relationship between the sample distribution and the specific excitation–emission wavelengths. The main fluorophores and comparison of relative concentrations of DOM gained by SOMs could provide a reference for the source of DBP precursors (Bieroza et al. 2009). Autoencoder NN could obtain the lower dimensional variables representing the characteristics of fluorescence data and use the obtained variables to realize the prediction of the formation of DBP (Peleato et al. 2018). Peleato (2022) found that the use of CNNs can significantly improve the accuracy of prediction for all DBP species and the use of heat maps can identify spectral regions associated with DBP precursors without relying on complex dimensionality reduction.
CHARACTERIZATION OF DBP PRECURSORS FROM DIFFERENT SOURCES
The analysis of DBP precursors in NOM
The most common components in NOM like humic and fulvic acids generally originate from the prolonged degradation of structural polymers lacking nitrogen, like lignin and cellulose, which serve as C-DBP precursors, particularly THMs and HAAs (Shah & Mitch 2012). Therefore, most research has correlated the fluorescence parameters representing humic and fulvic acids with C-DBP formation potential (C-DBPFP). Previous research found that C-DBPFP was more closely associated with the maximum fluorescence intensity of fulvic-like and humic-like components compared to tyrosine-like or tryptophan-like components (Watson et al. 2018; Jutaporn et al. 2021; Wang et al. 2021b; Xu et al. 2021). In addition, stronger correlations were also observed between the reduction of C-DBPFP and the removal of humic-like component rather than the protein-like components (Jutaporn et al. 2020; Wang et al. 2021b). However, both humic-like and protein-like components were used as precursors of THM, thus the correlation between the sum of fluorescence intensity of the two components and THM formation potential (THMFP) was stronger than fluorescence intensity of the single component (Wang et al. 2021a). In addition, microbial humic-like substances were also the important N-DBP precursors shown by the higher correlation between dichloroacetonitrile formation potential (DCAN-FP) and microbial humic-like components (Watson et al. 2018; Li et al. 2021a, 2021b).
Amino acids and protein-like components, accounting for 20–40% in NOM, are also related to N-DBP formation potential (N-DBPFP). Some research found that maximum fluorescence intensity from amino acid-like and protein-like components showed a significant and high correlation with HAN formation potential (HANFP) and NDMA formation potential (NDMAFP) (Yang et al. 2008, 2015a, 2015b; Jutaporn et al. 2020). However, it is worth noting that humic-like components sometimes can show a higher correlation with HANFP than amino acid-like or protein-like components, indicating the universality of the sources of HAN precursors (Watson et al. 2018; Li et al. 2021a, 2021b). In contrast, amino acid-like and protein-like components are the more effective indicators for detecting NDMA precursors than HAN precursors in some polluted surface water (Yang et al. 2015a, 2015b). In addition to the correlation with a class of total DBP, the correlation between different halogen substituted DBPs and fluorescent components was related to the amount of substituted bromine. The bromide ion can be oxidized to hypobromous acid (HOBr) competing with free chlorine to react with DOM, therefore, some research found that chlorine substituted DBP had a positive correlation with the fluorescence parameters, but gradually became a negative correlation with the increase of the amount of substituted bromine (Watson et al. 2018; Li et al. 2021a, 2021b). Jutaporn et al. (2021) also found that the presence of bromide ions in water will weaken the predictive power of organic surrogate parameters, indicating that the correlations from fluorescent DOM fractions should be used with caution. In addition, the NOM in surface water is very vulnerable to seasonal climate changes, and the formation of DBP also fluctuates greatly under its influence. Li et al. (2021a, 2021b) found that the enhanced microbial activity in autumn resulted in the higher concentrations of C-DBP precursors and the increased discharges of organic matter from industrial and municipal wastewater in autumn brought the higher concentrations of N-DBP precursors, which resulted in non-significant correlation between fluorescence components and THMFP for the overall data for one year. Yang et al. (2015a, 2015b) also found that the relationship between THMFP and humic-like components was much strengthened by examining separately for each month compared with the overall data for one year because of the exclusion of the temporal variations in the chemical composition of DOM. In addition, compared with the untreated surface water, the reduced fluorescence intensity after primary and secondary water treatment is more conducive to express the high correlation with the DBPFPs (Watson et al. 2018; Li et al. 2021a, 2021b). Thus, the correlations between the same type of fluorescent substances and DBPs are sometimes inconsistent in different studies. The low concentration of a certain type of DBP or the low concentration of related fluorescent substances and the concentration of bromine ions in water will affect the correlation between fluorescence parameters and DBPFP (Pifer & Fairey 2012; Watson et al. 2018; Jutaporn et al. 2021; Li et al. 2021a, 2021b; Wang et al. 2021b; Xu et al. 2021). Several representative studies are summarized in Table 2 and Supplementary material, Table S2.
Applications of EEM for analyzing the NOM from different source waters and various DBPs
Source water of NOM . | Fluorescence measurement method . | Main fluorescent substances . | Correlation analysis and key conclusion . | Reference . |
---|---|---|---|---|
Huangpu River, China | PARAFAC | Tyrosine-like Tryptophan-like Terrestrial humic-like Microbial humic-like |
| Li et al. (2021a, 2021b) |
Dongjiang River, China Roosevelt lake, USA Suwannee River, USA | FRI | Tryptophan-like SMP-like |
| Yang et al. (2008) |
Missouri Lake, USA | PARAFAC | Terrestrial humic-like |
| Hua et al. (2010) |
Suwannee River, USA | Picking peek | Humic-like Tyrosine-like Tryptophan-like |
| Saipetch & Yoshimura (2019) |
Iowa River, USA | FRI | Five regions difference |
| Johnstone & Miller (2009) |
Manas River, China | PARAFAC | Terrestrial humic-like Tryptophan-like |
| Wang et al. (2021a) |
Northern China reservoir | PARAFAC | Fulvic-like and humic-like UV/visible humic-like Ubiquitous soil fulvic-like |
| Xu et al. (2021) |
DWTPs from South Carolina, USA | PARAFAC | Fulvic acids-like Humic-like Tryptophan-like |
| Yang et al. (2015a, 2015b) |
DWTPs from Shanghai (China) | PARAFAC | Tryptophan-like Protein-bound-like Tyrosine-like Humic-like |
| Wang et al. (2021b) |
Lu Jhu Reservoir, Southern Taiwan | PARAFAC | Fulvic acid-like Humic-like Humic-like |
| Hidayah et al. (2017) |
Chapel Hill DWTP, USA | PARAFAC | Anthropogenic humic-like Terrestrial humic-like Protein-like |
| Jutaporn et al. (2020) |
Pong River, reservoir rainwater, Thailand | PARAFAC | Microbial humic-like |
| Jutaporn et al. (2021) |
Suwannee River, USA | PARAFAC | Terrestrial humic-like Anthropogenic humic-like Protein-like |
| Watson et al. (2018) |
Otonabee River, Ontario, Canada | PARAFAC | Terrestrial humic-like |
| Peleato et al. (2016) |
Source water of NOM . | Fluorescence measurement method . | Main fluorescent substances . | Correlation analysis and key conclusion . | Reference . |
---|---|---|---|---|
Huangpu River, China | PARAFAC | Tyrosine-like Tryptophan-like Terrestrial humic-like Microbial humic-like |
| Li et al. (2021a, 2021b) |
Dongjiang River, China Roosevelt lake, USA Suwannee River, USA | FRI | Tryptophan-like SMP-like |
| Yang et al. (2008) |
Missouri Lake, USA | PARAFAC | Terrestrial humic-like |
| Hua et al. (2010) |
Suwannee River, USA | Picking peek | Humic-like Tyrosine-like Tryptophan-like |
| Saipetch & Yoshimura (2019) |
Iowa River, USA | FRI | Five regions difference |
| Johnstone & Miller (2009) |
Manas River, China | PARAFAC | Terrestrial humic-like Tryptophan-like |
| Wang et al. (2021a) |
Northern China reservoir | PARAFAC | Fulvic-like and humic-like UV/visible humic-like Ubiquitous soil fulvic-like |
| Xu et al. (2021) |
DWTPs from South Carolina, USA | PARAFAC | Fulvic acids-like Humic-like Tryptophan-like |
| Yang et al. (2015a, 2015b) |
DWTPs from Shanghai (China) | PARAFAC | Tryptophan-like Protein-bound-like Tyrosine-like Humic-like |
| Wang et al. (2021b) |
Lu Jhu Reservoir, Southern Taiwan | PARAFAC | Fulvic acid-like Humic-like Humic-like |
| Hidayah et al. (2017) |
Chapel Hill DWTP, USA | PARAFAC | Anthropogenic humic-like Terrestrial humic-like Protein-like |
| Jutaporn et al. (2020) |
Pong River, reservoir rainwater, Thailand | PARAFAC | Microbial humic-like |
| Jutaporn et al. (2021) |
Suwannee River, USA | PARAFAC | Terrestrial humic-like Anthropogenic humic-like Protein-like |
| Watson et al. (2018) |
Otonabee River, Ontario, Canada | PARAFAC | Terrestrial humic-like |
| Peleato et al. (2016) |
DWTP, drinking water treatment plants; DCAN, dichloroacetonitrile; DCAA, dichloroacetic acid; TOX, total organic halogen; TCM, trichloromethane; TCAA, trichloroacetic acid; CH, chloral hydrate; NDMA, N-nitrosodimethylamine; HK, haloketones; 1,1-DCP, 1,1-dichloropropanone; 1,1,1-TCP, 1,1,1-trichloropropanone; BDCM, bromodichloromethane; DBCM, dibromochloromethane; MX, halofuranones.
The analysis of DBP precursors in AOM
Different from terrestrial NOM precursors that are highly aromatic and more hydrophobic, AOM produced by algal cells mainly consists of low aromatic and high hydrophilic components with high nitrogenous content, which are the important sources of C-DBP and N-DBP precursors (Hua & Reckhow 2007; Huang et al. 2009). The molecular weight (MW) distribution of AOM is mainly distributed in a range from <1 to >100 kDa and the hydrophilic part with less aromatic content usually has a smaller MW, which serves as the precursor having relatively high formation potential for both C-DBPs and N-DBPs (Fang et al. 2010b; Hua et al. 2019). Unlike the UV absorbance responding poorly to AOM samples, EEM has been gradually used to characterize DBP precursors in AOM to obtain more comprehensive information (Hua et al. 2018).
Different from NOM that mainly contain fulvic acid-like fluorescence and humic acid-like fluorescence, protein-like and amino acid-like fluorophores accounted for a larger proportion of AOM, which had stronger correlation with N-DBPFP and C-DBPFP (Yang et al. 2011; Chen et al. 2017a, 2017b; Ma et al. 2018). Previous research mainly focused on two types of samples including organic matter extracted from cultured algal cells and algal-rich natural water. EEM can distinguish the source of fluorophores from algal-rich natural water and analyze their impact on the formation of DBP. Visentin et al. (2020) found that SMP-like and aromatic protein-like components produced by algae growth had stronger correlation with THM and HAA compared with humic acid-like fluorophores from nature water. The DBP precursors of cultured algae derived organics are mainly determined by analyzing the types of main fluorescent substances and DBPFP at different algae growth stages. AOM-containing organic nitrogen-like SMP-like and aromatic protein-like components will react with chlorine to produce organic chloramine, a crucial middle intermediate that promotes the formation of N-DBPs like NDMA and HAN (Zhang et al. 2016; Liu et al. 2022). It is worth noting that the release of extracellular protein in exponential phase and stationary phase may result in the increase of HANs formation (Yang et al. 2011; Zhang et al. 2016). In addition, the formation of DBPs during the chlorination process was studied in some research to investigate the effects of Br− concentrations, higher concentrations and bromine incorporation factor (BIF) of total DBPs including THMs, HANs, and HNMs were observed in the presence of Br− because the substitution reaction efficiency of HOBr/OBr− with the fluorescent precursors was higher than that of HOCl/OCl− (Yang et al. 2011; Zhang et al. 2016; Chen et al. 2017a, 2017b).
The differences of DOM composition in extracellular organic matter (EOM) and intracellular organic matter (IOM) will also result in the differences in the change of fluorescence intensity and DBPFPs under different chlorine/chloramine doses. Previous studies have reported that the fluorescence intensity of EOM and IOM was mainly attributed to the SMP-like component, which could be a more significant source of NDMA than humic acid-like component, thus, the higher proportion of SMP-like region of EOM led to more NDMA formation and sustained response with increasing chlorine dose (Liu et al. 2022). However, another study indicated that the higher organic nitrogen concentrations in IOM were more conducive to the formation of haloacetaldehydes (HALs) and HANs since chlorination of organic nitrogen including free amino acids and aliphatic amines was known to produce HANs and HALs (Fang et al. 2010b; Yang et al. 2010). The selection of different disinfectant types also has a great impact on the formation of DBP, which may become an important variable in the correlation analysis between fluorescence parameters and DBPFP. Firstly, the concentration of THMs formed during chloramination decreased significantly due to the lower oxidation capacity, whereas the formation potentials of CH, 1,1-DCP, and 1,1,1-TCP increased significantly without the subsequent reactions (Fang et al. 2010a; Zhang et al. 2016). Secondly, previous studies indicated that both monochloramine and organic nitrogenous compounds contributed to the nitrogen in N-DBPs such as HAN and NDMA (Fang et al. 2010b; Yang et al. 2011; Liu et al. 2022). In addition, the facilitation of bromide on the formation of DBPs is more significant during the chloramination processing (Zhang et al. 2016). A few of the studies are summarized in Table 3 and Supplementary material, Table S3.
Applications of EEM for analyzing the AOM from different algal sources and various DBPs
Algal sources . | Fluorescence measurement method . | Main fluorescent substances . | Correlation analysis and key conclusion . | Reference . |
---|---|---|---|---|
Microcystis aeruginosa | FRI | SMP-like |
| Chen et al. (2017a, 2017b) |
Microcystis aeruginosa | PARAFAC | Tryptophan-like Amino acid-like Humic-like Fulvic-like |
| Ma et al. (2018) |
Cyanobacteria-impacted lakes | PARAFAC | SMP-like Aromatic protein -like |
| Visentin et al. (2020) |
Microcystis aeruginosa, Chlorella vulgaris | FRI | SMP-like Aromatic protein -like |
| Yang et al. (2011) |
Microcystis aeruginosa | FRI | SMP-like Humic-like |
| Liu et al. (2022) |
Cyanobacteria-impacted river | Peak picking | Tryptophan-like Low MW amino acids-like |
| Park et al. (2019) |
Chlorella sp. HQ | FRI | Tyrosine-like Tryptophan-like SMP-like |
| Liu et al. (2017) |
Microcystis aeruginosa | FRI Peak picking | Humic-like Protein-like SMP-like |
| Chen et al. (2018) |
Microcystis aeruginosa | Peak picking | Organic nitrogen-like Organic carbon-like |
| Fang et al. (2010b) |
Algal sources . | Fluorescence measurement method . | Main fluorescent substances . | Correlation analysis and key conclusion . | Reference . |
---|---|---|---|---|
Microcystis aeruginosa | FRI | SMP-like |
| Chen et al. (2017a, 2017b) |
Microcystis aeruginosa | PARAFAC | Tryptophan-like Amino acid-like Humic-like Fulvic-like |
| Ma et al. (2018) |
Cyanobacteria-impacted lakes | PARAFAC | SMP-like Aromatic protein -like |
| Visentin et al. (2020) |
Microcystis aeruginosa, Chlorella vulgaris | FRI | SMP-like Aromatic protein -like |
| Yang et al. (2011) |
Microcystis aeruginosa | FRI | SMP-like Humic-like |
| Liu et al. (2022) |
Cyanobacteria-impacted river | Peak picking | Tryptophan-like Low MW amino acids-like |
| Park et al. (2019) |
Chlorella sp. HQ | FRI | Tyrosine-like Tryptophan-like SMP-like |
| Liu et al. (2017) |
Microcystis aeruginosa | FRI Peak picking | Humic-like Protein-like SMP-like |
| Chen et al. (2018) |
Microcystis aeruginosa | Peak picking | Organic nitrogen-like Organic carbon-like |
| Fang et al. (2010b) |
TCNM, trichloronitromethane; HNMs, halonitromethanes; CNCl, cyanogen chloride; BIX, biological factor; DCA, dichloroacetaldehyde; org-N, organic nitrogen.
The analysis of DBP precursors in EfOM
Domestic wastewater from human excreta and other wastes is the source of high dissolved organic nitrogen (DON) and DOC content including proteins and carbohydrates, which usually account for more than 50% of the organics in the raw sewage. Although the biological treatment in WWTPs appeared to be capable of removing some of the DON that could potentially form a nitrogenous class of DBPs, EfOM still had higher DON and DOC contents than NOM in natural water due to some degradation by-products (Krasner et al. 2009; Huang et al. 2012). In addition, SMPs from bacterial exudates are composed of macromolecules and cellular debris that contain protein (nitrogen-enriched) and polysaccharides, which also serve as the precursors of some N-DBPs like NDMA and HANs (Krasner et al. 2009; Shah & Mitch 2012).
Different from protein that can easily be biologically treated, humic acid-like and fulvic acid-like substances were easy to pass through the biological treatment process and were dominant on the spectrogram of the effluent sample due to their non-biodegradable property (Wu et al. 2013; Liu et al. 2016). However, compared with NOM mainly consisting of humic-like substances in surface water, the proportion of protein-like materials and anthropogenic humic-like DOM in EfOM were higher than NOM, which may become DBP precursors (Cohen et al. 2014; Jutaporn et al. 2020). During the process of chlorine and chloramine disinfection, humic-like compounds in EfOM usually showed a higher correlation with THM because of the presence of a large number of electron-rich reactive sites that can be easily oxidized (Tungsudjawong et al. 2018; Ike et al. 2019; Kiattisaksiri et al. 2020), while the SMP-like fluorophores in EfOM had the lower reactivity with chlorine than humic-like fluorophores, which resulted in lower DBP yields despite the higher organic content in the effluent (Jutaporn et al. 2020; Kiattisaksiri et al. 2020), However, the formation of some N-DBPs, including HAN and NDMA, also shows a stronger correlation with the fluorescence parameter of microbial derived proteins or amino acids. Roccaro et al. (2020) found that protein-like materials were correlated with the NDMAFP and specific NDMA precursors. Moreover, the reduction of selected EEM peaks was also correlated with the removal of NDMA precursors. Kiattisaksiri et al. (2020) also found that SMP-like peaks showed relatively strong correlation with trichloronitromethane (TCNM) yield compared with other components.
Numerous efforts have been made to understand the main sources of various DBP precursors in EfOM, and some works have tried to link the fractions of DOM in municipal WWTPs gained by resin isolation techniques with the characterization of organics by EEM. Han & Ma (2012) found that the more fulvic acid-like and humic acid-like components in the fluorescence spectrogram of hydrophilic acids (HIA) and hydrophobic acids (HOA) fractions were considered to be the main reason for the high generation of THM, while aromatic amino acid-like and SMP-like materials were also important DBP precursors in HOA fraction. In addition, some research has analyzed the main sources of DBP precursors by comparing the DBP formation and main fluorescent components of different fractionation fractions. For instance, Zhang et al. (2009) found more aromatic proteins and higher THMFP and HAAFP from HOA fractions, which indicated that aromatic proteins were more responsible for the formation of DBPs than fulvic acids. Wu et al. (2013) found that the HOA seemed to have higher THMFP than the hydrophilic neutrals (HON) due to the additional protein-like material. In addition, some research also found that the presence of free or combined aromatic amino acids in the HOA fraction may result in the production of highly genotoxic DBPs during the chlorination of wastewater with high levels of ammonia nitrogen (Wang et al. 2007; Wu et al. 2013). Ozone is one of the commonly used disinfectants in the treatment of EfOM (Du et al. 2020), but DBPs formed from ozonation disinfection have gradually been concerned in wastewater treatment (Wert et al. 2008; Liu et al. 2015). When using ozone disinfection, the changes of integrated fluorescence were applied to establish predictive relationships with different DBPs including bromate and aldehydes and carboxylic acids (ACAs). In addition, Liu et al. (2015) achieved a good prediction of the formation of ACA by using relative changes of integrated fluorescence in different peek regions. The removal of total fluorescence was also found to be correlated with the formation of NDMA during ozonation, with a higher regression slope observed in wastewater with higher concentrations of NDMA precursors and TN (Sgroi et al. 2016). A few of the studies are summarized in Table 4 and Supplementary material, Table S4.
Applications of EEM for analyzing the EfOM from different biological treatment and various DBPs
Biological treatment process . | Fluorescence measurement method . | Main Fluorescent substances . | Correlation analysis and key conclusion . | Reference . |
---|---|---|---|---|
Membrane bioreactor | FRI Peak picking | Protein-like Fulvic-like Humic-like Aromatic protein -like peek |
| Ma et al. (2016) |
Membrane bioreactor | FRI Peak picking | Aromatic protein –like SMP-like |
| Roccaro et al. (2020) |
Aeration tank | FRI | Tryptophan-like Amino acid-like Humic-like Fulvic-like |
| Liu et al. (2015) |
Sequencing batch reactor | Peak picking | Humic-like SMP-like |
| Wang et al. (2018) |
Four WWTPs | FRI | Fulvic acid-like Humic acid-like SMP-like |
| Han & Ma (2012) |
Activated sludge treatment | Peak picking | Fulvic-like Humic-like Aromatic protein -like |
| Zhang et al. (2009) |
Four WWTPs | FRI | Total fluorescence |
| Sgroi et al. (2016) |
Four WWTPs | Peak picking | Tyrosine-like Tryptophan-like SMP-like |
| Needham et al. (2017) |
Membrane bioreactor | FRI Peak picking | Humic-like SMP-like Aromatic protein -like |
| Ma et al. (2014) |
Four WWTPs | PARAFAC | Tyrosine-like Tryptophan-like Fulvic acid-like |
| Xu et al. (2020) |
Oxidation ditch | Peak picking | Tyrosine-like Tryptophan-like Microbial humic-like |
| Wu et al. (2013) |
Activated sludge process | FRI | Total fluorescence |
| Ruffino et al. (2020) |
Biological treatment process . | Fluorescence measurement method . | Main Fluorescent substances . | Correlation analysis and key conclusion . | Reference . |
---|---|---|---|---|
Membrane bioreactor | FRI Peak picking | Protein-like Fulvic-like Humic-like Aromatic protein -like peek |
| Ma et al. (2016) |
Membrane bioreactor | FRI Peak picking | Aromatic protein –like SMP-like |
| Roccaro et al. (2020) |
Aeration tank | FRI | Tryptophan-like Amino acid-like Humic-like Fulvic-like |
| Liu et al. (2015) |
Sequencing batch reactor | Peak picking | Humic-like SMP-like |
| Wang et al. (2018) |
Four WWTPs | FRI | Fulvic acid-like Humic acid-like SMP-like |
| Han & Ma (2012) |
Activated sludge treatment | Peak picking | Fulvic-like Humic-like Aromatic protein -like |
| Zhang et al. (2009) |
Four WWTPs | FRI | Total fluorescence |
| Sgroi et al. (2016) |
Four WWTPs | Peak picking | Tyrosine-like Tryptophan-like SMP-like |
| Needham et al. (2017) |
Membrane bioreactor | FRI Peak picking | Humic-like SMP-like Aromatic protein -like |
| Ma et al. (2014) |
Four WWTPs | PARAFAC | Tyrosine-like Tryptophan-like Fulvic acid-like |
| Xu et al. (2020) |
Oxidation ditch | Peak picking | Tyrosine-like Tryptophan-like Microbial humic-like |
| Wu et al. (2013) |
Activated sludge process | FRI | Total fluorescence |
| Ruffino et al. (2020) |
Abbreviations: TONO, total N-nitrosamine; AEA, anti-estrogenic activity.
The analysis of DBP precursors in other organic sources
In addition to the direct study on DOM in surface water, the original DOM formed by the decay of plants is increasingly recognized as an important source of DBP precursors in watersheds (Mikkelson et al. 2013). During the rainy period, the decaying litter and the upper soil horizons in the surface of the watersheds will leach organic matter and inflow into surface water (Palmer et al. 2001; Ritson et al. 2014). The quantity and quality of DOM in leachates from decaying litter composed of different types of plants may affect the formation potential and toxicity potency of DBPs of various species (Reckhow et al. 2004; Jian et al. 2016). In addition, the chemical composition of DOM in leachates can also be affected by environmental conditions including sunlight, microbial activity, wildfire combustion and anthropogenic activity, resulting in the changes of DBP formation (Majidzadeh et al. 2015; Lee et al. 2019). Plant litter is composed of complex mixtures of organic components, including lignin, aliphatic biopolymers and amino sugars having a higher nitrogen content, which are considered the major precursors of C-DBPs and N-DBPs in drinking water disinfection (Kalbitz et al. 2006; Pellerin et al. 2010; Chow et al. 2011; Jian et al. 2016). Many studies have applied EEM to characterize the organic matter released from plant litter to explore the types of DBP precursors.
The organic matter leached from plant litter (LLOM) is mainly composed of humic-like and protein-like fluorescent components. Consequently, this index of sole humic or protein fluorescence may not give an accurate picture of DOC characteristic. When both humic and protein-like fluorescence were considered, Ritson et al. (2014) found that the ratio of peak C to peak T was correlated with DOC removal efficiency via coagulation, which can provide a reference for the formation of DBP after coagulation. Consistent with surface water, since the formation of C-DBPs including THM and HAA dominated, the humification index (HIX) and humic-like fluorescence parameters could well reflect different characteristics of the total DBP precursors from soil and litter leaf sources (Lee et al. 2019). However, different from surface water, protein-like fluorescence from LLOM had a higher proportion than NOM from surface water, resulting in lower C-DBP like THM and higher N-DBP like DCAN and TCNM (Jian et al. 2016). In addition to the vegetation type, litter decomposition stage and leaf age also affect the composition of LLOM because of the role of natural processes such as photolysis and microbial biodegradation. The amino acids and proteins as labile materials were preferentially degraded during the initial stages of exposure in the field, while the remaining humic-like materials generated at first and then degraded under solar exposure and biodegradation, which resulted in the reduction of total concentrations of THM, CH, and TCNM from aged leaves (Chow et al. 2011; Cuss & Gueguen 2015; Jian et al. 2016).
Due to the occurrence of wildfire events, the changes of organic components of different vegetation types after wildfire combustion were also noticed. Previous research has reported that thermal effects of wildfire on DOM in watersheds may reduce the coagulation efficiency of DOM and increase the DOM reactivity to yield toxic N-DBPs (Cawley et al. 2017; Cai et al. 2020). The relationship between fluorescence properties of extractable organic matter affected by wildfire and DBPFP was inconsistent in different literatures because of the influence of temperature conditions and vegetation types. Generally, the proportion of tyrosine-like and humic acid-like FRI showed a decreasing trend after wildfires, which indicated the general reduction of the reactivity to form THM (Majidzadeh et al. 2015; Wang et al. 2015; Uzun et al. 2020a, 2020b). In addition, many studies showed that the reactivity of LLOM to form HAN had been enhanced in most wildfire tests, and some research indicated that the polymerization of N into a condensed aromatic structure, resulting in the formation of dissolved black nitrogen compounds, is likely a major factor in the increased reactivity of HAN formation (Knicker 2007; Wang et al. 2015; Hohner et al. 2016; Cawley et al. 2017). According to the change of fluorescence spectrum, the increase of aromatic protein-like fluorescence and SMP-like fluorescence in some literatures can also explain the increase of HAN formation (Wang et al. 2015; Uzun et al. 2020a). Wang et al. (2015) also found the higher reactive NDMA precursors at moderate burning temperatures (200 − 500 °C) and the subsequent degradation of NDMA precursors in higher levels burning (>510 °C). At the same time, it is worth noting that under the influence of different regions, the increase of bromide ions after combustion in some coastal areas will lead to the increase of the proportion of brominated DBP (Wang et al. 2015; Uzun et al. 2020a). Different from wildfires, Uzun et al. (2020b) conducted experiments on specified fires and found that low-intensity and low severity specified fires can consume humic-like substances and SMP-like fluorescent substances, thus reducing the concentration of DBP precursors and not leading to the formation of additional N-DBPs and brominated THMs and HAAs with higher toxicity.
Applications of EEM for analyzing the DOM from other sources and various DBPs
Water sample source . | Fluorescence measurement method . | main related fluorescent substances . | Correlation analysis and key conclusion . | Reference . |
---|---|---|---|---|
Postfire soil samples, Tuolumne River Watershed | FRI Peak picking | Tryptophan-like |
| Wang et al. (2015) |
Unburned and burned litter, Congaree National Park | FRI | Tyrosine-like Fulvic acid-like Humic acid-like SMP-like |
| Majidzadeh et al. (2015) |
Surface peat samples, Exmoor National Park, UK | Peak picking | Tyrosine-like Humic-like |
| Ritson et al. (2014) |
Soil and leaf litters, Han River basin | PARAFAC | Terrestrial humic-like Tyrosine-like Microbial humic-like |
| Lee et al. (2019) |
Fresh litter and associated decomposed litter, oak woodland | Intensity ratio | Fluorescence index |
| Chow et al. (2011) |
Rain from Postfire area, Northern California Coastal | FRI | Tyrosine-like Tyrosine-like Fulvic-like Humic-like SMP-like |
| Uzun et al. (2020a) |
River samples from Postfire area Cache la Poudre River | Intensity ratio | Fluorescence index |
| Hohner et al. (2016) |
Foliage samples | PARAFAC | Tyrosine-like Tryptophan-like Terrestrial humic-like Microbial humic-like |
| Jian et al. (2016) |
Heated soil samples, Cache la Poudre Watershed | PARAFAC | Terrestrial humic-like Humic-like and Black nitrogen-type moieties |
| Cawley et al. (2017) |
Forest materials, Georgetown,South Carolina | FRI | Humic-like SMP-like |
| Uzun et al. (2020b) |
Biochar, Chuncheon, South Korea | PARAFAC | Terrestrial humic-like Microbial humic-like Tryptophan-like |
| Lee et al. (2018) |
Soil samples from biochar addition | Intensity ratio | HIX FI β/α |
| Cai et al. (2020) |
Surface water and six different PM samples | PARAFAC | Two terrestrial humic-like Autochthonous Tryptophan-like |
| Lee et al. (2021) |
PM2.5, PM10, total PM samples and Rainwater samples | FRI | Tyrosine-like Tryptophan-like SMP-like |
| Hou et al. (2018) |
Water sample source . | Fluorescence measurement method . | main related fluorescent substances . | Correlation analysis and key conclusion . | Reference . |
---|---|---|---|---|
Postfire soil samples, Tuolumne River Watershed | FRI Peak picking | Tryptophan-like |
| Wang et al. (2015) |
Unburned and burned litter, Congaree National Park | FRI | Tyrosine-like Fulvic acid-like Humic acid-like SMP-like |
| Majidzadeh et al. (2015) |
Surface peat samples, Exmoor National Park, UK | Peak picking | Tyrosine-like Humic-like |
| Ritson et al. (2014) |
Soil and leaf litters, Han River basin | PARAFAC | Terrestrial humic-like Tyrosine-like Microbial humic-like |
| Lee et al. (2019) |
Fresh litter and associated decomposed litter, oak woodland | Intensity ratio | Fluorescence index |
| Chow et al. (2011) |
Rain from Postfire area, Northern California Coastal | FRI | Tyrosine-like Tyrosine-like Fulvic-like Humic-like SMP-like |
| Uzun et al. (2020a) |
River samples from Postfire area Cache la Poudre River | Intensity ratio | Fluorescence index |
| Hohner et al. (2016) |
Foliage samples | PARAFAC | Tyrosine-like Tryptophan-like Terrestrial humic-like Microbial humic-like |
| Jian et al. (2016) |
Heated soil samples, Cache la Poudre Watershed | PARAFAC | Terrestrial humic-like Humic-like and Black nitrogen-type moieties |
| Cawley et al. (2017) |
Forest materials, Georgetown,South Carolina | FRI | Humic-like SMP-like |
| Uzun et al. (2020b) |
Biochar, Chuncheon, South Korea | PARAFAC | Terrestrial humic-like Microbial humic-like Tryptophan-like |
| Lee et al. (2018) |
Soil samples from biochar addition | Intensity ratio | HIX FI β/α |
| Cai et al. (2020) |
Surface water and six different PM samples | PARAFAC | Two terrestrial humic-like Autochthonous Tryptophan-like |
| Lee et al. (2021) |
PM2.5, PM10, total PM samples and Rainwater samples | FRI | Tyrosine-like Tryptophan-like SMP-like |
| Hou et al. (2018) |
Summary of the application of EEMs in correlation between DBPFP and DOM in different water sources.
Summary of the application of EEMs in correlation between DBPFP and DOM in different water sources.
FUTURE RESEARCH NEEDS OF EEM FOR DBP STUDIES
Further expansion of fluorescence analysis methods
The application of EEM in DBP precursors is mostly limited to establish the correlation between fluorescence intensity and DBPFP. Some unexploited fluorescence parameters like the apparent fluorescence quantum yield, stokes shift and excited energy state could characterize the MW, hydrophilicity and chemical structure of DOM (Xiao et al. 2016, 2018b). Therefore, these parameters could be used to explain the distinction of DBPFP from different water sources due to tendency characteristics of different DBP precursors (Mao et al. 2016). In addition, common fluorescence analysis methods tend to use a single or all data points of fluorescence intensity to analyze the composition of DBP precursors. Due to the presence of data points unrelated to DBP precursors, selecting more representative DBP precursor related regions contributes to precise analysis of the precursors of different DBPs (Trueman et al. 2016). Marhaba et al. (2009) also expected to screen out the fluorescence intensity variables that best represent the variation characteristics of DBP precursors by removing the scatter and focusing on peak areas. The development of machine learning may provide a feasible approach for screening out intended data points. Some new analytical methods, such as fluorescence quotient, have been used to divide regions related to the properties of organic matter like hydrophobicity and functional behavior of DOM (Xiao et al. 2018a, 2018b). Thus, selecting more representative DBP precursor related regions on the EEM map based on some new methods like fluorescence quotient also requires further research.
Further expansion of identifying the precursors of emerging DBPs by EEM
Some emerging DBPs, such as iodinated THMs, HAMs, haloanisoles, chlorophenols, and halofuranones, have been found to exhibit high toxicity at extremely low concentrations (Richardson & Postigo 2016; Xie et al. 2016; Chen et al. 2017a, 2017b; Gilca et al. 2020). Previous studies mainly focused on the formation mechanisms and detection methods of emerging DBPs, whereas the use of three-dimensional fluorescence in analyzing potential organic precursors has rarely been considered. Potential fluorescent precursor for emerging DBPs in different water sources also requires further research for comparison with fluorescent precursors of conventional DBPs. In addition, some emerging organic micro-pollutants have been proven to serve as the precursors for many DBPs during chlorine-based disinfection processes (Shao et al. 2023), and some water quality parameters like NOM, temperature, pH value, and the type of ions could have an essential impact on DBPs formation during the chlorination of these micro-pollutants (Kali et al. 2021). Some organic micro-pollutants also exhibit fluorescence properties (Li et al. 2021a, 2021b), and the concentration in water was strongly correlated with specific fluorescence parameters. Thus, the fluorescence properties of micro-pollutants can be linked to the formation of DBP to expand subsequent research.
Further expansion of identifying by-products of other disinfectants by EEM
Disinfectants that produce organic DBPs generally include chlorine, chloramine, and ozone. Similar to ozone, the research on by-products of other oxidants mainly focused on the production of by-products caused by the oxidation of bromine ions or iodine ions to produce hypobromic acid or hypoiodic acid (Jiang et al. 2016; Yang et al. 2019; Wang et al. 2020). Currently, the formation of ozone oxidation by-products like bromate and ACA can be predicted by changes in fluorescence intensity (Liu et al. 2015), but the research on by-products of other oxidants mainly focused on the reaction mechanism with halogen ions, without involving the analysis of EEM fluorescence spectrum (Huang et al. 2016; Wang et al. 2020). However, other oxidants like ferrate, persulfate and permanganate have been used to reduce the production of DBPs produced by chlorine, and the influence on the fluorescence properties of DOM has also been revealed in some research. Permanganate had a strong removal effect on humic-like fluorescence to achieve the removal of formation of different DBP including trichloromethane (TCM), DCAN, dichloroacetamide (DCAcAm) and TCNM (Chu et al. 2011). UV/H2O2 could break down the fluorescent and aromatic components of DOM like fulvic-like and humic-like fluorophores and transform DOM with high MW into low MW species, which resulted in the increasing THMFP following chlorination or chloramination but the reduction of I-THMs following chlorination (Zhang et al. 2018). Ferrate showed the ability to remove humic-like components and thus achieve a reduction in THM formation (Li et al. 2021a, 2021b). Overall, the effect of oxidants on fluorescent DOM tends to be positive, which is beneficial for controlling the subsequent production of DBP. The effect of oxidants on fluorescence properties can be combined with other possible by-products to expand subsequent research.
CONCLUSIONS
The application of EEM for characterizing DBP precursors has developed for a long time and still has more research potential for some information like the apparent quantum yield and fluorescence lifetime with less use. In this review, analytical methods of EEM for characterizing DBP precursors, differences of main fluorescent components and DBP precursors in different water sources have been discussed to help researchers better use EEM for the study of DBP precursors in water. It is necessary to select appropriate fluorescence analysis methods according to different water quality conditions and analysis purposes. The application of ANNs based on non-linear regression is rarely used but has prospects for wider application in the future.
Some discussions and analyses in this paper are expected to further provide insights into the application of EEM in DBP precursors. The content and properties of fluorescent components in different source waters are different, leading to the inconsistency of the main related DBP precursors. Therefore, it is important to use different methods to more comprehensively evaluate the correlation between DBP precursors and fluorescent components, which contribute to screening out the key fluorescent parameters related to DBP precursors.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the National Natural Science Foundation of China (No. 52325001; 51578389; 51778445), the National Major Science and Technology Project of China (No. 2015ZX07406-004; 2017ZX07201-005), the Shanghai City Youth Science and Technology Star Project (No. 17QA1404400), State Key Laboratory of Pollution Control and Resource Reuse Foundation (NO. PCRRE16009), and Fundamental Research Funds for the Central Universities (1380219169).
AUTHOR CONTRIBUTIONS
J.H. conceptualized and wrote the original draft. R.X., R.Z., Z.W., F.J., C.Y., and R.Q. wrote, reviewed, and edited the article. W.C. supervised, wrote, reviewed, and edited the article..
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.
CONFLICT OF INTEREST
The authors declare there is no conflict.