- Email: [email protected]
Assessing removal efficiency of dissolved organic matter in wastewater treatment using fluorescence excitation emission matrices with parallel factor analysis and second derivative synchronous fluorescence
Bioresource Technology 144 (2013) 595–601
Contents lists available at SciVerse ScienceDirect
Bioresource Technology journal homepage: www.elsevier.com/locate/biortech
Assessing removal efficiency of dissolved organic matter in wastewater treatment using fluorescence excitation emission matrices with parallel factor analysis and second derivative synchronous fluorescence Huibin Yu a, Yonghui Song a,⇑, Xiang Tu a, Erdeng Du b, Ruixia Liu a,⇑, Jianfeng Peng a a b
State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Sciences, Beijing 100012, China State Key Laboratory of Pollution Control and Resource Reuse, Tongji University, Shanghai 200092, China
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
Assess removal efficiency of DOM in
WWTP by EEM with PARAFAC. Develop SDSF approach compared
with EEM–PARAFAC to analyze DOM fractions. Verify availability of SDSF for DOM removal and optimize of sampling sites.
a r t i c l e
i n f o
Article history: Received 20 May 2013 Received in revised form 2 July 2013 Accepted 5 July 2013 Available online 12 July 2013 Keywords: Dissolved organic matter (DOM) Parallel factor analysis (PARAFAC) Second derivative synchronous fluorescence (SDSF) Multivariate analysis Wastewater
a b s t r a c t To assess removal efficiency of dissolved organic matter (DOM) in the wastewater treatment plant (WWTP), fluorescence excitation emission matrices (EEM) with parallel factor analysis (PARAFAC) and second derivative synchronous fluorescence (SDSF) were used for the characterization of DOM in wastewater. In A2/O process, tryptophan-like and tyrosine-like materials were removed to the much greater extent than that of fulvic-like. The protein-like material might be mostly decomposed by anaerobic and aerobic bacteria in anaerobic/anoxic and oxic zones. C1, C2, I276, I286, AP1 and AP2 were much better in tracing variations of tryptophan-like and tyrosine-like materials than C3, I329 and AF1 in tracing fulviclike. The number of sampling sites should be reduced, as concentration variations of DOM components were subtle among sampling sites in the oxic zone and secondary sedimentation tank. SDSF may be a useful tool as PARAFAC to monitor removal efficiency of DOM fractions from wastewater in the WWTP. Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction Reuse and recycling of treated wastewater effluent have played an increasingly prominent role in ensuring water quality, quantity, ecology and security during the last decades (Henderson et al., 2009; Ishii and Boyer, 2012). In wastewater treatment schemes, ⇑ Corresponding authors. Tel./fax: +86 10 84928380. E-mail addresses: [email protected] (Y. Song), [email protected] (R. Liu). 0960-8524/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biortech.2013.07.025
ongoing successive managements, such as online monitoring, data processing, warning and emergency are essential to ensure efficiency and reliability of wastewater treatment performance. Although an online monitoring system has been applied to monitor water quality, the monitoring parameters are usually limited to total organic carbon, pH, conductivity, turbidity, color, dissolved oxygen, ammonia, nitrate and so on (Henderson et al., 2009). Dissolved organic matter (DOM) is a heterogeneous mixture of aromatic, amino and aliphatic organic compound containing oxy-
596
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
gen, nitrogen and sulfur functional groups (Chen et al., 2003). It is ubiquitous in natural and engineered aquatic systems, and has a multifunctional role in the aquatic environment (Murphy et al., 2011). DOM presents upcoming challenges in engineered systems due to its impact on all water treatment processes. Concentration and composition of DOM can dominate coagulation efficiency, membrane fouling, disinfection byproduct formation, oxidant demand, and microbial growth (Ishii and Boyer, 2012). Moreover the analysis and characterization of DOM in wastewater or treated wastewater have become important, for DOM can influence on the further treatment of wastewater and also on the fate and toxicity of organic or/and inorganic pollutants in receiving water. Fluorescence spectroscopy has commonly been used to discriminate between DOM fractions (Hudson et al., 2008; Osburn et al., 2012). It has widely been applied in the water industry as a potential monitoring technique. Fluorescence monitoring is attractive as it is a rapid, inexpensive and reagentless technique that requires no sample preparation prior to analysis (Henderson et al., 2009). It can be utilized as a online monitoring tool for water quality assessment in natural water bodies and process control in water treatment, e.g., specific pollutants in industrial wastewater, oil in water and disinfection byproduct formation potentials in drinking water treatment (Liu et al., 2007; Borisover et al., 2009; Seredyn´skaSobecka et al., 2011). The most significant advance for fluorescence technique in recent years has been the ability to simultaneously scan excitation and emission wavelengths through a set light path length of aqueous sample to generate a fluorescence excitationemission-matrix (EEM) (Baker et al., 2003). Chemometric techniques, including parallel factor analysis (PARAFAC), selforganizing map, principal filter analysis, partial least squares regression, etc., have been applied to interpret EEM. PARAFAC can mathematically discriminate chemical independent but spectral overlapping fluorescence components and the maximum intensities (Fmax) of these components are used to trace variability in DOM fluorescence in wastewater treatment plant (WWTP) (Henderson et al., 2009). Thus the online monitoring instruments may test the ability of the PARAFAC for monitoring WWTP performance through predictive algorithms in real-time. A new algorithm called EEMizer has been developed, which aims to automate the use of PARAFAC (Bro and Vidal, 2011). Synchronous fluorescence spectroscopy can provide a narrow, sharp and simple spectrum, compared with broad fluorescence emission/excitation spectra obtained in conventional fluorescence measurement
(Abdelal et al., 2009; Yu et al., 2012). However synchronous fluorescence spectroscopy has an extremely restricted scope of application in the analysis of DOM, because its selectivity is reduced by extensive spectral overlap (Mozo-Villaías, 2002). Derivatives can amplify narrow band and avoid broad band, for the amplitude of the derivative signal is inversely proportional to the band width of the original spectrum (Mozo-Villaías, 2002; Abdelal et al., 2009). The combination of synchronous fluorescence spectroscopy and derivatives is of great value to increasing sensitivity, which presents a further strategy for reduction of extensive spectroscopic overlap and rejection of matrix interference (Abdelal et al., 2009). Derivative synchronous fluorometric spectroscopy technique has been most widely utilized in the analysis of mixture of polycyclic aromatic hydrocarbons, rodenticides, inorganic ions and humic substances (Mozo-Villaías, 2002; Kumar et al., 2005; Abdelal et al., 2009; Yu et al., 2012). However, few studies have been carried out to assess removal efficiency of DOM fractions from wastewater in WWTP using this technique. In this study, second derivative synchronous fluorescence (SDSF) approach compared with mature EEM-PARAFAC, is firstly developed to characterize DOM fluorescence properties of wastewater in WWTP. Multivariate statistics was used to verify the availability of SDSF for the assessment of DOM removal in wastewater treatment process and optimization of the sampling sites. 2. Methods 2.1. Sample collection Gaobeidian WWTP is located in the southeastern Beijing City of China. It treats 1,000,000 m3 d1 wastewater from both domestic and industrial sources. A2/O, a conventional active sludge process, is used in the WWTP for integrated removal of carbon, nitrogen and phosphorus, which incorporates anaerobic/anoxic and oxic zones in sequence (Wang et al., 2011, 2012). There are four rectangular aerated regions with six aeration tanks in each rectangular aerated region and three corridors in a given aeration tank. Each corridor has a length of 96 m, a width of 9.3 m and a depth of 6 m, and fifty percents of the first corridor is used as the anaerobic/anoxic zone (Fig. 1). Water samples were collected at 22 sampling sites in Gaobeidian WWPT in 31 December 2011. Sites 1–3 are located in the primary sedimentation tank, sites 4–8 in the anaerobic/anoxic zone,
Fig. 1. Simplified wastewater treatment process diagram of an aerated region in Gaobeidian WWTP. Sites 1–3 in the primary sedimentation tank, sites 4–8 in the anaerobic/ anoxic zone, sites 9–19 in the oxic zone and sites 20–22 in the secondary sedimentation tank.
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
sites 9–19 in the oxic zone and sites 20–22 in the secondary sedimentation tank (Fig. 1). In each sampling site, duplicate samples were collected with a Wildco Kemmerer 1.2 L sampler and transferred into EE BOD bottles. Samples were filtered through glass fiber filters (Whatman GF/F, 0.45 lm, pre-combusted at 450 °C for 4 h). The filtrate was collected into precombusted glass amber bottles and stored in the dark at 4 °C until analyzed. All samples were analyzed within 2 days of collection. 2.2. Fluorescence spectroscopy Fluorescence spectra were acquired on a Hitachi Fluorescence Spectrophotometer (F-7000) equipped with the fluorescence solutions 1.00.000 for data processing. Excitation and emission slit widths were set to 5 nm bandpass and PMT Voltage to 700 V. Synchronous spectra were measured by scanning simultaneously both the excitation (ex), varied from 260 to 550 nm, and emission (em) wavelengths, while maintaining a constant and optimized wavelength difference 4k = kem kex = 55 nm (Yu et al., 2012). To obtain fluorescence EEMs, excitation wavelengths were determined every 5 nm from 200 to 450 nm; emission wavelengths were determined from 280 to 550 nm at intervals of 5 nm (Osburn et al., 2012). Appropriate instrument-specific corrections of the excitation and emission were performed and all fluorescence spectra were corrected for inner-filtering effects before Quinine-sulfate calibration against scatter (Stedmon and Bro, 2008; Murphy et al., 2011). All spectra were subtracted from their respective procedural blanks.
597
2.5. Statistical analyses Hierarchical cluster analysis was applied on spectroscopic parameters deduced from PARAFAC and SDSF using Furthest Neighbor with Euclidean Distances. The variations of DOM components in wastewater were determined from the hierarchical cluster analysis using linkage distance. Discriminant analysis was performed on spectroscopic parameters based on the standard mode to construct the best discriminant functions to confirm the clusters determined by the hierarchical cluster analysis, to verify the parameter accuracy of indicating DOM removal and to optimize the sampling sites. All statistical analyses were performed using SSPS 16.0 and OriginPro 8.0 for windows.
3. Results and discussion 3.1. EEM spectroscopic properties
SDSF was achieved numerically in a desk-top computer, using FL Solution Software (Hitachi High-Technologies Corporation 1998, 2008) program, which allowed fast, easy handling of data, such as derivative spectra, smoothing, algebraic operations and spectrum correction. The interval in the process was 1 nm. Noise reduction was mainly obtained using accumulation of spectra. The SDSF was smoothed by weight 75 points in order to remove excess noise and improve resolution (Kumar et al., 2005).
There were five distinct fluorescence peaks observed in the EEM spectrum of the water sample in site 3 (Influent sample) (Fig. 2a). These have been defined according to Coble (1996) as follows: Peak A – fulvic-like (kex/em = 237–260/400–500 nm), Peak C – fulvic-like (kex/em = 300–370/400–500 nm), Peak B – tyrosine-like (kex/em = 216–237/300–325 nm), Peak T1 – tryptophan-like (kex/em = 275– 296/340–380 nm) and Peak T2 – tryptophan-like (kex/em = 216– 237/340–380 nm) (Chen et al., 2003; Hudson et al., 2008). Wastewaters, mainly consisting of sewage effluents (Hudson et al., 2008), poultry and livestock wastes and landfill leachates (Baker and Spencer, 2004) have been detected to be rich in peaks T and B derived from soluble microbial products and other aromatic proteins. In the wastewater treatment process the intensities of peaks T and B reduced to a much greater extent than those of the fulvic-like A and C peaks (Fig. 2b). This indicates that peaks B and T in untreated wastewater, derived from anthropogenic activity, are associated with fresher and less degraded material with a high potential for oxidation, which may be used as indicators to monitor DOM removal in wastewater treatment performance (Hudson et al., 2008; Henderson et al., 2009).
2.4. PARAFAC model
3.2. Parallel factor analysis
Fluorescence EEM was modeled using PARAFAC which uses an alternating least squares algorithm to minimize the sum of squared residuals across the dataset and estimate the underlying structure of the EEM (Harshman and Lundy, 1994; Bro, 1997). The data signal is decomposed into a set of trilinear terms and a residual array:
PARAFAC modeling of the water samples revealed that the EEM of DOM could be characterized by three different fractions (Fig. 2c– e). Component 1 (C1) exhibited primary and secondary peaks with the kex/em = 230 and 275/345 nm, component 2 (C2) also showed primary and secondary peaks with the kex/em = 220/305 and 345 nm and component (C3) included two peaks with the kex/ em = 245 and 323/425 nm. Following the literature compilation made by Chen et al. (2003) and Stedmon and Bro (2008), C1 was associated with the tryptophan-containing matter that might be related to a biological production and activity of microorganisms. C2 was likely related to tyrosine-like material (simple aromatic proteins), and C3 was termed as fulvic-like material. The fates of fluorescence components (1–3) extracted by the FARAFAC technique were tracked using their maximum fluorescence intensities (Fmax) during the wastewater treatment process. Fmax is considered to be proportional to the relative concentrations of different components (Henderson et al., 2009; Goldman et al., 2012). In the primary sedimentation tank the mean Fmax values of tryptophan-like component (Fmax of C1: 4212.83 ± 771.51) and tyrosine-like (Fmax of C2: 3392.01 ± 671.31) were much higher than that of fulvic-like (Fmax of C3: 861.89 ± 57.626) (Fig. 3a), which expatiated that tryptophan-like and tyrosine-like materials were considered as represen-
2.3. Derivative method
xijk ¼
F X aif bjf ckf þ eijk
ð1Þ
f 1
where xijk is the intensity of the ith sample at the jth variable (emission mode) and at the kth variable (excitation mode) in the F-component model; aif is directly proportional to the concentration of the fth analyte of the ith sample, bjf and ckf are scaled estimates of the emission and excitation spectra at wavelengths j and k, respectively, for the fth analyte, and eijk is the residual noise, representing the variability not accounted for by the model; i = 1, 2, . . ., I, j = 1, 2, . . ., J, k = 1, 2, . . ., K. The calibrated and corrected fluorescence data was processed in Matlab software. The Domfluor toolbox 1.7 for Matlab was used for the implementation and validation of the PARAFAC model (Stedmon and Bro, 2008; toolbox is freely available). The number of fluorescence components was found by a validation method including split-half and residual analysis.
598
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
Fig. 2. EEM spectroscopy (a and b) and EEM of PARAFAC components (c–e) of DOM of wastewater in Gaobeidian WWTP.
tative components of DOM in the wastewater. In the wastewater treatment process, the variations of tryptophan-like and tyrosinelike materials were much larger than that of fulvic-like (Fig. 3a), indicating that the protein-like material was removed to a much higher extent than that of the fulvic-like material. The mean removal efficiencies of tryptophan-like reached to 58.78 ± 2.41%, 37.49 ± 10.22% and 4.68 ± 1.58% in the anaerobic/
anoxic, the oxic and secondary sedimentation zones, respectively. The average removal efficiencies of tyrosine-like attained to 54.98 ± 3.79%, 98.31 ± 3.85% and nearly zero in the three successive zones. The mean removal efficiencies of fulvic-like were nearly less than 10% in the three successive zones (Fig. 3b). These revealed that protein-like material was mostly degraded by anaerobic and aerobic bacteria in the anaerobic/anoxic and oxic zones.
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
Fig. 3. Column and whiskers plots represent Fmax variations (a) and mean removal efficiency (b) of Fmax of PARAFAC components across the wastewater treatment process. Column plots show the Fmax values and whiskers delimit standard deviation values. A2, anaerobic/anoxic zone; O, oxic zone; SS, secondary sedimentation tank.
Moreover A2/O is inefficient to remove fulvic-like material, which is consistent with the previous studies (Goldman et al., 2012).
3.3. Derivative fluorescence spectroscopy There were a strong peak at around 275 nm (protein-like and/or amino-like) and two weak shoulders at around 329 and 367 nm (fulvic-like) in the synchronous fluorescence spectroscopy of DOM from the wastewater sample (Fig. 4a), which was attributed to strong overlaps between the three bands. After the second derivative transformation of the synchronous fluorescence spectroscopy, the protein peak was well separated into two peaks which appeared at 276 and 286 nm (Fig. 4b and c). The former was defined as the tryptophan-like material, while the latter as tyro-
599
sine-like material (Gräzel and Thomas, 1976; Reynolds, 2003). The fulvic-like weak shoulders became relatively distinct (Fig. 4b and c). All normalized SDSF spectra of DOM from the wastewater samples were presented in Fig. 4d. The fluorescent intensities of tryptophan-like peaks (I276) in the SDSF spectra showed a very significant positive linear correlation with C1 (Figure not shown), suggesting that I276 could be a substitute for C1 as an indicator of the tryptophan-like content in the wastewater treatment process. The intensities of tyrosine-like peak (I286) exhibited a very close positive linear correlation with C2, testifying that I286 might be alternative to the tyrosine material content. C3 had a good positive correlation with the intensities of the fulvic-like peaks at the wavelength of 329 nm (I329), while a poor correlation with the fulvic-like peak at the wavelength of 367 nm (I367). This revealed that I329 was the better substitute for C3 to characterize the fulvic-like material than I367. While the fluorescent intensity relies on a fixed wavelength only, the absolute area, a measurement of integrate where intensity changes with wavelengths, can reflect expanded spectroscopic information over a broad wavelength range (Hur et al., 2009; Yu et al., 2012). Based on the two peaks and two shoulders, four relative fluorescence regions were assigned to tryptophan-like (P1), tyrosine-like (P2), fulvic-like I (F1) and fulvic-like II (F2), each of which corresponded to the wavelength of 265–283, 283–291, 324–345 and 375–450 nm, respectively. There had a better positive relationship between C1 and the absolute area in the P1 region (AP1) (Figure not shown), demonstrating that AP1 could be indicative of tryptophan-like component in the wastewater treatment process. A significant positive relationship existed between C2 and the absolute area in the P2 region (AP2), certifying that AP2 could reflect the concentration of tyrosine material. There appeared to be a better relationship between C3 and the absolute area in the F1 (AF1), while a poor relationship between C3 and the absolute area in the F2 region (AF2). This illustrated that AF1 was a better indicator for characterizing the fulvic-like material than the AF2.
Fig. 4. Synchronous fluorescence spectroscopy (a), second Synchronous fluorescence spectroscopy (b), normalized derivative spectroscopy (c) of DOM in site 3. Normalized derivative spectra for all 22 sampling sites (d).
600
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
3.4. Multivariate statistics 3.4.1. Parameters distribution and descriptive statistics Prior to hierarchical cluster analysis and discriminant analysis, spectroscopic parameters (C1, C2, C3, I276, I286, I329, AP1, AP2 and AF1) were needed to conform to a normal distribution; hence the normality of the distribution of each parameter was checked by analyzing kurtosis and skewness (Zhou et al., 2007). For the original parameters, AF1, C3 and I329 showed normal distributions, whereas the other parameters were positively skew with kurtosis coefficients observably greater than zero (Figure not shown). After log transformation of these parameters, all skewness and kurtosis values evidently decreased with ranges from 1.29 to 1.22 and 0.87 to 0.70, respectively, which were less than the critical values. 3.4.2. Hierarchical cluster analysis An initial exploratory approach involved the application of hierarchical cluster analysis on the spectroscopic parameters. The cluster analysis produced a dendrogram grouping the 22 wastewater samples into five clusters at the rescaled distance (RD) < 1.0 (Fig. 5) and the difference among the clusters was significant (Zhou et al., 2007). Cluster i included sites 1–3 located in the primary sedimentation tank, cluster ii comprised sites 4–8 in the anaerobic/anoxic zone, cluster iii involved sites 9–10 in the oxic zone, cluster iv covered sites 11–18 in the oxic zone, and cluster v had sites 19–22 in the secondary sedimentation tank (except for site 19). These indirectly indicated that the concentrations of the tryptophan-like material decreased in the order of primary sedimentation tank > anaerobic/anoxic zone > oxic zone > secondary sedimentation tank, so were the tyrosine-like and fulvic-like materials (Fig. 3a). In the cluster i, the two sub-clusters (2, 1 and 3) indirectly illustrated an obvious alteration of the content of DOM fractions in the wastewater (Fig. 3a). In the cluster ii, a sub-cluster containing site 8 indirectly suggested that the concentration of DOM in site 8 were lower than those of the other sites (sites 4–7), which was worth noting that site 8 was close to the oxic zone (Fig. 1). In the cluster
Table 1 Discriminant function coefficients for the canonical discriminant analysis of the alterations of DOM fractions. Spectroscopic parameters
C1 C2 C3 I276 I286 I329 AP1 AP2 AF1
Function 1
2
3
4
2.189 2.352 0.077 10.888 8.239 0.529 6.242 5.401 0.794
0.540 0.847 0.470 4.565 12.153 0.326 8.179 9.395 0.902
0.129 0.218 1.147 5.482 16.349 0.695 0.892 12.292 0.514
0.075 0.144 0.249 5.305 1.124 0.734 3.805 2.059 0.375
iii, sites 9 and 10 were close to the anaerobic/anoxic zone (Fig. 1), where the transitional zone of anaerobic and aerobic processes occurred, i.e., facultative aerobic zone. 3.4.3. Discriminant analysis The discriminant analysis of the alterations of DOM fractions was performed using the 9 spectroscopic parameters after classification into the five clusters obtained from the hierarchical analysis. Sites were the dependent variables and the spectroscopic parameters were the independent variables. Wilks’ lambda and Chi-square for each discriminant function varied from 0.00 to 0.81 and from 3.02 to 154.00, respectively, at p < 0.00 to 0.81, which illustrated that the discriminant analysis was very credible and effective. Four discriminant functions with the spectroscopic parameters were obtained from the standard modes of the discriminant analysis (Table 1), which produced a classification matrix (Table not shown) with close to 90.9% correct assignments and casewise statistics (Table not shown) using the nine discriminant parameters. The absolute values of variable coefficients in the discriminant functions are higher, and the variables are more accurate (Zhou et al., 2007). The absolute values of C3, I329 and AF1 coefficients close to zero were distinctly much less than those of I276, I286, AP1, AP2, C1 and C2, indicating that the latter variables were more
Fig. 5. Dendrogram showing the clustering of the wastewater samples.
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
accurate parameters for discriminating tryptophan-like and tyrosine-like alterations than the former variables for discriminant fulvic-like alterations. This suggested that SDSF might be very appropriate for a chemometric technique as PARAFAC to indicate alterations of DOM fractions in wastewater. Based on above results, hierarchical cluster analysis provided a classification of the variations of DOM fractions, which might aid in designing monitoring plan of DOM with an adjustment of the sampling sites. For instance, the number of sampling sites might be decreased in the clusters iv and v, i.e., in the oxic zone and secondary sedimentation tank. Furthermore, the discriminant analysis proved to be a useful tool to recognize the discriminant parameters in the alterations of DOM fractions. As PARAFAC, SDSF was very rapid, inexpensive and accessible and might be used to monitor DOM fractions from wastewater in the WWPT. 4. Conclusion SDSF is an available tool as EEM-PARAFAC for characterizing organic matter in the WWTP performance. Tryptophan-like and tyrosine-like materials are dominant components of DOM in wastewater, which are removed to the much larger extent than that of fulvic-like material. Tryptophan-like and tyrosine-like materials may be mostly degraded by anaerobic and aerobic bacteria in the aeration tank. I276, I286, AP1, AP2, C1 and C2 are more accurate parameters for monitoring alterations of protein-like material than C3, I329 and AF1 for fulvic-like. The number of sampling sites should be decreased in the oxic zone and secondary sedimentation tank. Acknowledgements This work was financially supported by National Major Program of Science and Technology for Water Pollution Control and Governance (Fund number, 2012ZX07202-005, PR China), National Natural Science Foundation of China (21107103), China Postdoctoral Science Foundation (Fund number, 2012M510515). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.biortech.2013. 07.025. References Abdelal, A., El-Enany, N., Belal, F., 2009. Simultaneous determination of sulpiride and its alkaline degradation product by second derivative synchronous fluorescence spectroscopy. Talanta 80 (1), 880–888. Baker, A., Spencer, R.G.M., 2004. Characterization of dissolved organic matter from source to sea using fluorescence and absorbance spectroscopy. Environ. Pollut. 124 (3), 57–70. Baker, A., Inverarity, R., Charlton, M., Richmond, S., 2003. Detecting river pollution using fluorescence spectrophotometry: case studies from the Ouseburn, NE England. Environ. Pollut. 124 (3), 57–70.
601
Borisover, M., Laor, Y., Parparov, A., Bukhanovsky, N., Lado, M., 2009. Spatial and seasonal patterns of fluorescent organic matter in Lake Kinneret (Sea of Galilee) and its catchment basin. Water Res. 43 (6), 3104–3116. Bro, R., 1997. PARAFAC: tutorial and applications. Chemometr. Intell. Lab. 38 (2), 149–171. Bro, R., Vidal, M., 2011. EEMizer: automated modeling of fluorescence EEM data. Chemometr. Intell. Lab. 106 (1), 86–92. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitationemission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. Technol. 37 (24), 5701–5710. Coble, P.G., 1996. Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy. Mar. Chem. 51 (4), 325–346. Goldman, J.H., Rounds, S.A., Needoba, J.A., 2012. Applications of fluorescence spectroscopy for predicting percent wastewater in an urban stream. Environ. Sci. Technol. 46 (8), 4374–4381. Gräzel, M., Thomas, J.K., 1976. In: Wehry, E.L. (Ed.), Modern Fluorescence Spectroscopy 2. Plenum Press, New York. Harshman, R.A., Lundy, M.E., 1994. Parafac: parallel factor analysis. Comput. Stat. Data Anal. 18 (1), 39–72. Henderson, R.K., Baker, A., Murphy, K.R., Hambly, A., Stuetz, R.M., Khan, S.J., 2009. Fluorescence as a potential monitoring tool for recycled water systems: a review. Water Res. 43 (4), 863–881. Hudson, N., Baker, A., Ward, D., Reynolds, D.M., Brunsdon, C., Carliell-Marquet, C., Browning, S., 2008. Can fluorescence spectrometry be used as a surrogate for the biochemical oxygen demand (BOD) test in water quality assessment? An example from South West England. Sci. Total Environ. 391 (1), 149–158. Hur, J., Lee, D.H., Shin, H.S., 2009. Comparison of the structural, spectroscopic and phenanthrene binding characteristics of humic acids from soils and lake sediments. Org. Geochem. 40 (10), 1091–1099. Ishii, S.K.L., Boyer, T.H., 2012. Behavior of reoccurring PARAFAC components in fluorescent dissolved organic matter in natural and engineered systems: a critical review. Environ. Sci. Technol. 46 (1), 2006–2017. Kumar, V., Sharma, V.K., Kalonia, D.S., 2005. Second derivative tryptophan fluorescence spectroscopy as a tool to characterize partially unfolded intermediates of proteins. Int. J. Pharm. 294 (1–2), 193–199. Liu, R.X., Lead, J.R., Baker, A., 2007. Fluorescence characterization of cross flow ultrafiltration derived freshwater colloidal and dissolved organic matter. Chemosphere 68 (7), 1304–1311. Mozo-Villaías, A., 2002. Second derivative fluorescence spectroscopy of tryptophan in proteins. J. Biochem. Bioph. Methods 50 (2–3), 163–178. Murphy, K.R., Hambly, A., Singh, S., Henderson, R.K., Baker, A., Stuetz, R., 2011. Organic matter fluorescence in municipal water recycling schemes: toward a unified PARAFAC model. Environ. Sci. Technol. 45 (7), 2909–2916. Osburn, C.L., Handsel, L.T., Mikan, M.P., Paerl, H.P., 2012. Fluorescence tracking of dissolved and particulate organic matter quality in a river-dominated estuary. Environ. Sci. Technol. 46 (16), 8628–8636. Reynolds, D.M., 2003. Rapid and direct determination of tryptophan in water using synchronous fluorescence spectroscopy. Water Res. 37 (13), 3055–3060. Seredyn´ska-Sobecka, B., Stedmon, C.A., Boe-Hansen, R., Waul, C.K., Arvin, E., 2011. Monitoring organic loading to swimming pools by fluorescence excitation emission matrix with parallel factor analysis (PARAFAC). Water Res. 43 (6), 3104–3116. Stedmon, C.A., Bro, R., 2008. Characterizing dissolved organic matter fluorescence with parallel factor analysis: a tutorial. Limnol. Oceanogr. 6, 572–579. Wang, J., Zhang, J., Xie, H., Qi, P., Ren, Y., Hu, Z., 2011. Methane emissions from a fullscale A/A/O wastewater treatment plant. Bioresour. Technol. 102 (9), 5479– 5485. Wang, Y., Geng, J., Peng, Y., Wang, C., Guo, G., Liu, S., 2012. A comparison of endogenous processes during anaerobic starvation in anaerobic end sludge and aerobic end sludge from an anaerobic/anoxic/oxic sequencing batch reactor performing denitrifying phosphorus removal. Bioresour. Technol. 104 (1), 19– 27. Yu, H., Song, Y., Xi, B., Zhang, M., He, X., 2012. Application of derivative synchronous fluorescence spectroscopy (DSFS) to indicate salinisation processes of saline soils in semi-arid region. Ecol. Indic. 18 (7), 532–539. Zhou, F., Guo, H., Liu, Y., Jiang, Y., 2007. Chemometrics data analysis of marine water quality and source identification in Southern Hong Kong. Mar. Pollut. Bull. 54 (6), 745–756.
Contents lists available at SciVerse ScienceDirect
Bioresource Technology journal homepage: www.elsevier.com/locate/biortech
Assessing removal efficiency of dissolved organic matter in wastewater treatment using fluorescence excitation emission matrices with parallel factor analysis and second derivative synchronous fluorescence Huibin Yu a, Yonghui Song a,⇑, Xiang Tu a, Erdeng Du b, Ruixia Liu a,⇑, Jianfeng Peng a a b
State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Sciences, Beijing 100012, China State Key Laboratory of Pollution Control and Resource Reuse, Tongji University, Shanghai 200092, China
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
Assess removal efficiency of DOM in
WWTP by EEM with PARAFAC. Develop SDSF approach compared
with EEM–PARAFAC to analyze DOM fractions. Verify availability of SDSF for DOM removal and optimize of sampling sites.
a r t i c l e
i n f o
Article history: Received 20 May 2013 Received in revised form 2 July 2013 Accepted 5 July 2013 Available online 12 July 2013 Keywords: Dissolved organic matter (DOM) Parallel factor analysis (PARAFAC) Second derivative synchronous fluorescence (SDSF) Multivariate analysis Wastewater
a b s t r a c t To assess removal efficiency of dissolved organic matter (DOM) in the wastewater treatment plant (WWTP), fluorescence excitation emission matrices (EEM) with parallel factor analysis (PARAFAC) and second derivative synchronous fluorescence (SDSF) were used for the characterization of DOM in wastewater. In A2/O process, tryptophan-like and tyrosine-like materials were removed to the much greater extent than that of fulvic-like. The protein-like material might be mostly decomposed by anaerobic and aerobic bacteria in anaerobic/anoxic and oxic zones. C1, C2, I276, I286, AP1 and AP2 were much better in tracing variations of tryptophan-like and tyrosine-like materials than C3, I329 and AF1 in tracing fulviclike. The number of sampling sites should be reduced, as concentration variations of DOM components were subtle among sampling sites in the oxic zone and secondary sedimentation tank. SDSF may be a useful tool as PARAFAC to monitor removal efficiency of DOM fractions from wastewater in the WWTP. Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction Reuse and recycling of treated wastewater effluent have played an increasingly prominent role in ensuring water quality, quantity, ecology and security during the last decades (Henderson et al., 2009; Ishii and Boyer, 2012). In wastewater treatment schemes, ⇑ Corresponding authors. Tel./fax: +86 10 84928380. E-mail addresses: [email protected] (Y. Song), [email protected] (R. Liu). 0960-8524/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biortech.2013.07.025
ongoing successive managements, such as online monitoring, data processing, warning and emergency are essential to ensure efficiency and reliability of wastewater treatment performance. Although an online monitoring system has been applied to monitor water quality, the monitoring parameters are usually limited to total organic carbon, pH, conductivity, turbidity, color, dissolved oxygen, ammonia, nitrate and so on (Henderson et al., 2009). Dissolved organic matter (DOM) is a heterogeneous mixture of aromatic, amino and aliphatic organic compound containing oxy-
596
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
gen, nitrogen and sulfur functional groups (Chen et al., 2003). It is ubiquitous in natural and engineered aquatic systems, and has a multifunctional role in the aquatic environment (Murphy et al., 2011). DOM presents upcoming challenges in engineered systems due to its impact on all water treatment processes. Concentration and composition of DOM can dominate coagulation efficiency, membrane fouling, disinfection byproduct formation, oxidant demand, and microbial growth (Ishii and Boyer, 2012). Moreover the analysis and characterization of DOM in wastewater or treated wastewater have become important, for DOM can influence on the further treatment of wastewater and also on the fate and toxicity of organic or/and inorganic pollutants in receiving water. Fluorescence spectroscopy has commonly been used to discriminate between DOM fractions (Hudson et al., 2008; Osburn et al., 2012). It has widely been applied in the water industry as a potential monitoring technique. Fluorescence monitoring is attractive as it is a rapid, inexpensive and reagentless technique that requires no sample preparation prior to analysis (Henderson et al., 2009). It can be utilized as a online monitoring tool for water quality assessment in natural water bodies and process control in water treatment, e.g., specific pollutants in industrial wastewater, oil in water and disinfection byproduct formation potentials in drinking water treatment (Liu et al., 2007; Borisover et al., 2009; Seredyn´skaSobecka et al., 2011). The most significant advance for fluorescence technique in recent years has been the ability to simultaneously scan excitation and emission wavelengths through a set light path length of aqueous sample to generate a fluorescence excitationemission-matrix (EEM) (Baker et al., 2003). Chemometric techniques, including parallel factor analysis (PARAFAC), selforganizing map, principal filter analysis, partial least squares regression, etc., have been applied to interpret EEM. PARAFAC can mathematically discriminate chemical independent but spectral overlapping fluorescence components and the maximum intensities (Fmax) of these components are used to trace variability in DOM fluorescence in wastewater treatment plant (WWTP) (Henderson et al., 2009). Thus the online monitoring instruments may test the ability of the PARAFAC for monitoring WWTP performance through predictive algorithms in real-time. A new algorithm called EEMizer has been developed, which aims to automate the use of PARAFAC (Bro and Vidal, 2011). Synchronous fluorescence spectroscopy can provide a narrow, sharp and simple spectrum, compared with broad fluorescence emission/excitation spectra obtained in conventional fluorescence measurement
(Abdelal et al., 2009; Yu et al., 2012). However synchronous fluorescence spectroscopy has an extremely restricted scope of application in the analysis of DOM, because its selectivity is reduced by extensive spectral overlap (Mozo-Villaías, 2002). Derivatives can amplify narrow band and avoid broad band, for the amplitude of the derivative signal is inversely proportional to the band width of the original spectrum (Mozo-Villaías, 2002; Abdelal et al., 2009). The combination of synchronous fluorescence spectroscopy and derivatives is of great value to increasing sensitivity, which presents a further strategy for reduction of extensive spectroscopic overlap and rejection of matrix interference (Abdelal et al., 2009). Derivative synchronous fluorometric spectroscopy technique has been most widely utilized in the analysis of mixture of polycyclic aromatic hydrocarbons, rodenticides, inorganic ions and humic substances (Mozo-Villaías, 2002; Kumar et al., 2005; Abdelal et al., 2009; Yu et al., 2012). However, few studies have been carried out to assess removal efficiency of DOM fractions from wastewater in WWTP using this technique. In this study, second derivative synchronous fluorescence (SDSF) approach compared with mature EEM-PARAFAC, is firstly developed to characterize DOM fluorescence properties of wastewater in WWTP. Multivariate statistics was used to verify the availability of SDSF for the assessment of DOM removal in wastewater treatment process and optimization of the sampling sites. 2. Methods 2.1. Sample collection Gaobeidian WWTP is located in the southeastern Beijing City of China. It treats 1,000,000 m3 d1 wastewater from both domestic and industrial sources. A2/O, a conventional active sludge process, is used in the WWTP for integrated removal of carbon, nitrogen and phosphorus, which incorporates anaerobic/anoxic and oxic zones in sequence (Wang et al., 2011, 2012). There are four rectangular aerated regions with six aeration tanks in each rectangular aerated region and three corridors in a given aeration tank. Each corridor has a length of 96 m, a width of 9.3 m and a depth of 6 m, and fifty percents of the first corridor is used as the anaerobic/anoxic zone (Fig. 1). Water samples were collected at 22 sampling sites in Gaobeidian WWPT in 31 December 2011. Sites 1–3 are located in the primary sedimentation tank, sites 4–8 in the anaerobic/anoxic zone,
Fig. 1. Simplified wastewater treatment process diagram of an aerated region in Gaobeidian WWTP. Sites 1–3 in the primary sedimentation tank, sites 4–8 in the anaerobic/ anoxic zone, sites 9–19 in the oxic zone and sites 20–22 in the secondary sedimentation tank.
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
sites 9–19 in the oxic zone and sites 20–22 in the secondary sedimentation tank (Fig. 1). In each sampling site, duplicate samples were collected with a Wildco Kemmerer 1.2 L sampler and transferred into EE BOD bottles. Samples were filtered through glass fiber filters (Whatman GF/F, 0.45 lm, pre-combusted at 450 °C for 4 h). The filtrate was collected into precombusted glass amber bottles and stored in the dark at 4 °C until analyzed. All samples were analyzed within 2 days of collection. 2.2. Fluorescence spectroscopy Fluorescence spectra were acquired on a Hitachi Fluorescence Spectrophotometer (F-7000) equipped with the fluorescence solutions 1.00.000 for data processing. Excitation and emission slit widths were set to 5 nm bandpass and PMT Voltage to 700 V. Synchronous spectra were measured by scanning simultaneously both the excitation (ex), varied from 260 to 550 nm, and emission (em) wavelengths, while maintaining a constant and optimized wavelength difference 4k = kem kex = 55 nm (Yu et al., 2012). To obtain fluorescence EEMs, excitation wavelengths were determined every 5 nm from 200 to 450 nm; emission wavelengths were determined from 280 to 550 nm at intervals of 5 nm (Osburn et al., 2012). Appropriate instrument-specific corrections of the excitation and emission were performed and all fluorescence spectra were corrected for inner-filtering effects before Quinine-sulfate calibration against scatter (Stedmon and Bro, 2008; Murphy et al., 2011). All spectra were subtracted from their respective procedural blanks.
597
2.5. Statistical analyses Hierarchical cluster analysis was applied on spectroscopic parameters deduced from PARAFAC and SDSF using Furthest Neighbor with Euclidean Distances. The variations of DOM components in wastewater were determined from the hierarchical cluster analysis using linkage distance. Discriminant analysis was performed on spectroscopic parameters based on the standard mode to construct the best discriminant functions to confirm the clusters determined by the hierarchical cluster analysis, to verify the parameter accuracy of indicating DOM removal and to optimize the sampling sites. All statistical analyses were performed using SSPS 16.0 and OriginPro 8.0 for windows.
3. Results and discussion 3.1. EEM spectroscopic properties
SDSF was achieved numerically in a desk-top computer, using FL Solution Software (Hitachi High-Technologies Corporation 1998, 2008) program, which allowed fast, easy handling of data, such as derivative spectra, smoothing, algebraic operations and spectrum correction. The interval in the process was 1 nm. Noise reduction was mainly obtained using accumulation of spectra. The SDSF was smoothed by weight 75 points in order to remove excess noise and improve resolution (Kumar et al., 2005).
There were five distinct fluorescence peaks observed in the EEM spectrum of the water sample in site 3 (Influent sample) (Fig. 2a). These have been defined according to Coble (1996) as follows: Peak A – fulvic-like (kex/em = 237–260/400–500 nm), Peak C – fulvic-like (kex/em = 300–370/400–500 nm), Peak B – tyrosine-like (kex/em = 216–237/300–325 nm), Peak T1 – tryptophan-like (kex/em = 275– 296/340–380 nm) and Peak T2 – tryptophan-like (kex/em = 216– 237/340–380 nm) (Chen et al., 2003; Hudson et al., 2008). Wastewaters, mainly consisting of sewage effluents (Hudson et al., 2008), poultry and livestock wastes and landfill leachates (Baker and Spencer, 2004) have been detected to be rich in peaks T and B derived from soluble microbial products and other aromatic proteins. In the wastewater treatment process the intensities of peaks T and B reduced to a much greater extent than those of the fulvic-like A and C peaks (Fig. 2b). This indicates that peaks B and T in untreated wastewater, derived from anthropogenic activity, are associated with fresher and less degraded material with a high potential for oxidation, which may be used as indicators to monitor DOM removal in wastewater treatment performance (Hudson et al., 2008; Henderson et al., 2009).
2.4. PARAFAC model
3.2. Parallel factor analysis
Fluorescence EEM was modeled using PARAFAC which uses an alternating least squares algorithm to minimize the sum of squared residuals across the dataset and estimate the underlying structure of the EEM (Harshman and Lundy, 1994; Bro, 1997). The data signal is decomposed into a set of trilinear terms and a residual array:
PARAFAC modeling of the water samples revealed that the EEM of DOM could be characterized by three different fractions (Fig. 2c– e). Component 1 (C1) exhibited primary and secondary peaks with the kex/em = 230 and 275/345 nm, component 2 (C2) also showed primary and secondary peaks with the kex/em = 220/305 and 345 nm and component (C3) included two peaks with the kex/ em = 245 and 323/425 nm. Following the literature compilation made by Chen et al. (2003) and Stedmon and Bro (2008), C1 was associated with the tryptophan-containing matter that might be related to a biological production and activity of microorganisms. C2 was likely related to tyrosine-like material (simple aromatic proteins), and C3 was termed as fulvic-like material. The fates of fluorescence components (1–3) extracted by the FARAFAC technique were tracked using their maximum fluorescence intensities (Fmax) during the wastewater treatment process. Fmax is considered to be proportional to the relative concentrations of different components (Henderson et al., 2009; Goldman et al., 2012). In the primary sedimentation tank the mean Fmax values of tryptophan-like component (Fmax of C1: 4212.83 ± 771.51) and tyrosine-like (Fmax of C2: 3392.01 ± 671.31) were much higher than that of fulvic-like (Fmax of C3: 861.89 ± 57.626) (Fig. 3a), which expatiated that tryptophan-like and tyrosine-like materials were considered as represen-
2.3. Derivative method
xijk ¼
F X aif bjf ckf þ eijk
ð1Þ
f 1
where xijk is the intensity of the ith sample at the jth variable (emission mode) and at the kth variable (excitation mode) in the F-component model; aif is directly proportional to the concentration of the fth analyte of the ith sample, bjf and ckf are scaled estimates of the emission and excitation spectra at wavelengths j and k, respectively, for the fth analyte, and eijk is the residual noise, representing the variability not accounted for by the model; i = 1, 2, . . ., I, j = 1, 2, . . ., J, k = 1, 2, . . ., K. The calibrated and corrected fluorescence data was processed in Matlab software. The Domfluor toolbox 1.7 for Matlab was used for the implementation and validation of the PARAFAC model (Stedmon and Bro, 2008; toolbox is freely available). The number of fluorescence components was found by a validation method including split-half and residual analysis.
598
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
Fig. 2. EEM spectroscopy (a and b) and EEM of PARAFAC components (c–e) of DOM of wastewater in Gaobeidian WWTP.
tative components of DOM in the wastewater. In the wastewater treatment process, the variations of tryptophan-like and tyrosinelike materials were much larger than that of fulvic-like (Fig. 3a), indicating that the protein-like material was removed to a much higher extent than that of the fulvic-like material. The mean removal efficiencies of tryptophan-like reached to 58.78 ± 2.41%, 37.49 ± 10.22% and 4.68 ± 1.58% in the anaerobic/
anoxic, the oxic and secondary sedimentation zones, respectively. The average removal efficiencies of tyrosine-like attained to 54.98 ± 3.79%, 98.31 ± 3.85% and nearly zero in the three successive zones. The mean removal efficiencies of fulvic-like were nearly less than 10% in the three successive zones (Fig. 3b). These revealed that protein-like material was mostly degraded by anaerobic and aerobic bacteria in the anaerobic/anoxic and oxic zones.
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
Fig. 3. Column and whiskers plots represent Fmax variations (a) and mean removal efficiency (b) of Fmax of PARAFAC components across the wastewater treatment process. Column plots show the Fmax values and whiskers delimit standard deviation values. A2, anaerobic/anoxic zone; O, oxic zone; SS, secondary sedimentation tank.
Moreover A2/O is inefficient to remove fulvic-like material, which is consistent with the previous studies (Goldman et al., 2012).
3.3. Derivative fluorescence spectroscopy There were a strong peak at around 275 nm (protein-like and/or amino-like) and two weak shoulders at around 329 and 367 nm (fulvic-like) in the synchronous fluorescence spectroscopy of DOM from the wastewater sample (Fig. 4a), which was attributed to strong overlaps between the three bands. After the second derivative transformation of the synchronous fluorescence spectroscopy, the protein peak was well separated into two peaks which appeared at 276 and 286 nm (Fig. 4b and c). The former was defined as the tryptophan-like material, while the latter as tyro-
599
sine-like material (Gräzel and Thomas, 1976; Reynolds, 2003). The fulvic-like weak shoulders became relatively distinct (Fig. 4b and c). All normalized SDSF spectra of DOM from the wastewater samples were presented in Fig. 4d. The fluorescent intensities of tryptophan-like peaks (I276) in the SDSF spectra showed a very significant positive linear correlation with C1 (Figure not shown), suggesting that I276 could be a substitute for C1 as an indicator of the tryptophan-like content in the wastewater treatment process. The intensities of tyrosine-like peak (I286) exhibited a very close positive linear correlation with C2, testifying that I286 might be alternative to the tyrosine material content. C3 had a good positive correlation with the intensities of the fulvic-like peaks at the wavelength of 329 nm (I329), while a poor correlation with the fulvic-like peak at the wavelength of 367 nm (I367). This revealed that I329 was the better substitute for C3 to characterize the fulvic-like material than I367. While the fluorescent intensity relies on a fixed wavelength only, the absolute area, a measurement of integrate where intensity changes with wavelengths, can reflect expanded spectroscopic information over a broad wavelength range (Hur et al., 2009; Yu et al., 2012). Based on the two peaks and two shoulders, four relative fluorescence regions were assigned to tryptophan-like (P1), tyrosine-like (P2), fulvic-like I (F1) and fulvic-like II (F2), each of which corresponded to the wavelength of 265–283, 283–291, 324–345 and 375–450 nm, respectively. There had a better positive relationship between C1 and the absolute area in the P1 region (AP1) (Figure not shown), demonstrating that AP1 could be indicative of tryptophan-like component in the wastewater treatment process. A significant positive relationship existed between C2 and the absolute area in the P2 region (AP2), certifying that AP2 could reflect the concentration of tyrosine material. There appeared to be a better relationship between C3 and the absolute area in the F1 (AF1), while a poor relationship between C3 and the absolute area in the F2 region (AF2). This illustrated that AF1 was a better indicator for characterizing the fulvic-like material than the AF2.
Fig. 4. Synchronous fluorescence spectroscopy (a), second Synchronous fluorescence spectroscopy (b), normalized derivative spectroscopy (c) of DOM in site 3. Normalized derivative spectra for all 22 sampling sites (d).
600
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
3.4. Multivariate statistics 3.4.1. Parameters distribution and descriptive statistics Prior to hierarchical cluster analysis and discriminant analysis, spectroscopic parameters (C1, C2, C3, I276, I286, I329, AP1, AP2 and AF1) were needed to conform to a normal distribution; hence the normality of the distribution of each parameter was checked by analyzing kurtosis and skewness (Zhou et al., 2007). For the original parameters, AF1, C3 and I329 showed normal distributions, whereas the other parameters were positively skew with kurtosis coefficients observably greater than zero (Figure not shown). After log transformation of these parameters, all skewness and kurtosis values evidently decreased with ranges from 1.29 to 1.22 and 0.87 to 0.70, respectively, which were less than the critical values. 3.4.2. Hierarchical cluster analysis An initial exploratory approach involved the application of hierarchical cluster analysis on the spectroscopic parameters. The cluster analysis produced a dendrogram grouping the 22 wastewater samples into five clusters at the rescaled distance (RD) < 1.0 (Fig. 5) and the difference among the clusters was significant (Zhou et al., 2007). Cluster i included sites 1–3 located in the primary sedimentation tank, cluster ii comprised sites 4–8 in the anaerobic/anoxic zone, cluster iii involved sites 9–10 in the oxic zone, cluster iv covered sites 11–18 in the oxic zone, and cluster v had sites 19–22 in the secondary sedimentation tank (except for site 19). These indirectly indicated that the concentrations of the tryptophan-like material decreased in the order of primary sedimentation tank > anaerobic/anoxic zone > oxic zone > secondary sedimentation tank, so were the tyrosine-like and fulvic-like materials (Fig. 3a). In the cluster i, the two sub-clusters (2, 1 and 3) indirectly illustrated an obvious alteration of the content of DOM fractions in the wastewater (Fig. 3a). In the cluster ii, a sub-cluster containing site 8 indirectly suggested that the concentration of DOM in site 8 were lower than those of the other sites (sites 4–7), which was worth noting that site 8 was close to the oxic zone (Fig. 1). In the cluster
Table 1 Discriminant function coefficients for the canonical discriminant analysis of the alterations of DOM fractions. Spectroscopic parameters
C1 C2 C3 I276 I286 I329 AP1 AP2 AF1
Function 1
2
3
4
2.189 2.352 0.077 10.888 8.239 0.529 6.242 5.401 0.794
0.540 0.847 0.470 4.565 12.153 0.326 8.179 9.395 0.902
0.129 0.218 1.147 5.482 16.349 0.695 0.892 12.292 0.514
0.075 0.144 0.249 5.305 1.124 0.734 3.805 2.059 0.375
iii, sites 9 and 10 were close to the anaerobic/anoxic zone (Fig. 1), where the transitional zone of anaerobic and aerobic processes occurred, i.e., facultative aerobic zone. 3.4.3. Discriminant analysis The discriminant analysis of the alterations of DOM fractions was performed using the 9 spectroscopic parameters after classification into the five clusters obtained from the hierarchical analysis. Sites were the dependent variables and the spectroscopic parameters were the independent variables. Wilks’ lambda and Chi-square for each discriminant function varied from 0.00 to 0.81 and from 3.02 to 154.00, respectively, at p < 0.00 to 0.81, which illustrated that the discriminant analysis was very credible and effective. Four discriminant functions with the spectroscopic parameters were obtained from the standard modes of the discriminant analysis (Table 1), which produced a classification matrix (Table not shown) with close to 90.9% correct assignments and casewise statistics (Table not shown) using the nine discriminant parameters. The absolute values of variable coefficients in the discriminant functions are higher, and the variables are more accurate (Zhou et al., 2007). The absolute values of C3, I329 and AF1 coefficients close to zero were distinctly much less than those of I276, I286, AP1, AP2, C1 and C2, indicating that the latter variables were more
Fig. 5. Dendrogram showing the clustering of the wastewater samples.
H. Yu et al. / Bioresource Technology 144 (2013) 595–601
accurate parameters for discriminating tryptophan-like and tyrosine-like alterations than the former variables for discriminant fulvic-like alterations. This suggested that SDSF might be very appropriate for a chemometric technique as PARAFAC to indicate alterations of DOM fractions in wastewater. Based on above results, hierarchical cluster analysis provided a classification of the variations of DOM fractions, which might aid in designing monitoring plan of DOM with an adjustment of the sampling sites. For instance, the number of sampling sites might be decreased in the clusters iv and v, i.e., in the oxic zone and secondary sedimentation tank. Furthermore, the discriminant analysis proved to be a useful tool to recognize the discriminant parameters in the alterations of DOM fractions. As PARAFAC, SDSF was very rapid, inexpensive and accessible and might be used to monitor DOM fractions from wastewater in the WWPT. 4. Conclusion SDSF is an available tool as EEM-PARAFAC for characterizing organic matter in the WWTP performance. Tryptophan-like and tyrosine-like materials are dominant components of DOM in wastewater, which are removed to the much larger extent than that of fulvic-like material. Tryptophan-like and tyrosine-like materials may be mostly degraded by anaerobic and aerobic bacteria in the aeration tank. I276, I286, AP1, AP2, C1 and C2 are more accurate parameters for monitoring alterations of protein-like material than C3, I329 and AF1 for fulvic-like. The number of sampling sites should be decreased in the oxic zone and secondary sedimentation tank. Acknowledgements This work was financially supported by National Major Program of Science and Technology for Water Pollution Control and Governance (Fund number, 2012ZX07202-005, PR China), National Natural Science Foundation of China (21107103), China Postdoctoral Science Foundation (Fund number, 2012M510515). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.biortech.2013. 07.025. References Abdelal, A., El-Enany, N., Belal, F., 2009. Simultaneous determination of sulpiride and its alkaline degradation product by second derivative synchronous fluorescence spectroscopy. Talanta 80 (1), 880–888. Baker, A., Spencer, R.G.M., 2004. Characterization of dissolved organic matter from source to sea using fluorescence and absorbance spectroscopy. Environ. Pollut. 124 (3), 57–70. Baker, A., Inverarity, R., Charlton, M., Richmond, S., 2003. Detecting river pollution using fluorescence spectrophotometry: case studies from the Ouseburn, NE England. Environ. Pollut. 124 (3), 57–70.
601
Borisover, M., Laor, Y., Parparov, A., Bukhanovsky, N., Lado, M., 2009. Spatial and seasonal patterns of fluorescent organic matter in Lake Kinneret (Sea of Galilee) and its catchment basin. Water Res. 43 (6), 3104–3116. Bro, R., 1997. PARAFAC: tutorial and applications. Chemometr. Intell. Lab. 38 (2), 149–171. Bro, R., Vidal, M., 2011. EEMizer: automated modeling of fluorescence EEM data. Chemometr. Intell. Lab. 106 (1), 86–92. Chen, W., Westerhoff, P., Leenheer, J.A., Booksh, K., 2003. Fluorescence excitationemission matrix regional integration to quantify spectra for dissolved organic matter. Environ. Sci. Technol. 37 (24), 5701–5710. Coble, P.G., 1996. Characterization of marine and terrestrial DOM in seawater using excitation-emission matrix spectroscopy. Mar. Chem. 51 (4), 325–346. Goldman, J.H., Rounds, S.A., Needoba, J.A., 2012. Applications of fluorescence spectroscopy for predicting percent wastewater in an urban stream. Environ. Sci. Technol. 46 (8), 4374–4381. Gräzel, M., Thomas, J.K., 1976. In: Wehry, E.L. (Ed.), Modern Fluorescence Spectroscopy 2. Plenum Press, New York. Harshman, R.A., Lundy, M.E., 1994. Parafac: parallel factor analysis. Comput. Stat. Data Anal. 18 (1), 39–72. Henderson, R.K., Baker, A., Murphy, K.R., Hambly, A., Stuetz, R.M., Khan, S.J., 2009. Fluorescence as a potential monitoring tool for recycled water systems: a review. Water Res. 43 (4), 863–881. Hudson, N., Baker, A., Ward, D., Reynolds, D.M., Brunsdon, C., Carliell-Marquet, C., Browning, S., 2008. Can fluorescence spectrometry be used as a surrogate for the biochemical oxygen demand (BOD) test in water quality assessment? An example from South West England. Sci. Total Environ. 391 (1), 149–158. Hur, J., Lee, D.H., Shin, H.S., 2009. Comparison of the structural, spectroscopic and phenanthrene binding characteristics of humic acids from soils and lake sediments. Org. Geochem. 40 (10), 1091–1099. Ishii, S.K.L., Boyer, T.H., 2012. Behavior of reoccurring PARAFAC components in fluorescent dissolved organic matter in natural and engineered systems: a critical review. Environ. Sci. Technol. 46 (1), 2006–2017. Kumar, V., Sharma, V.K., Kalonia, D.S., 2005. Second derivative tryptophan fluorescence spectroscopy as a tool to characterize partially unfolded intermediates of proteins. Int. J. Pharm. 294 (1–2), 193–199. Liu, R.X., Lead, J.R., Baker, A., 2007. Fluorescence characterization of cross flow ultrafiltration derived freshwater colloidal and dissolved organic matter. Chemosphere 68 (7), 1304–1311. Mozo-Villaías, A., 2002. Second derivative fluorescence spectroscopy of tryptophan in proteins. J. Biochem. Bioph. Methods 50 (2–3), 163–178. Murphy, K.R., Hambly, A., Singh, S., Henderson, R.K., Baker, A., Stuetz, R., 2011. Organic matter fluorescence in municipal water recycling schemes: toward a unified PARAFAC model. Environ. Sci. Technol. 45 (7), 2909–2916. Osburn, C.L., Handsel, L.T., Mikan, M.P., Paerl, H.P., 2012. Fluorescence tracking of dissolved and particulate organic matter quality in a river-dominated estuary. Environ. Sci. Technol. 46 (16), 8628–8636. Reynolds, D.M., 2003. Rapid and direct determination of tryptophan in water using synchronous fluorescence spectroscopy. Water Res. 37 (13), 3055–3060. Seredyn´ska-Sobecka, B., Stedmon, C.A., Boe-Hansen, R., Waul, C.K., Arvin, E., 2011. Monitoring organic loading to swimming pools by fluorescence excitation emission matrix with parallel factor analysis (PARAFAC). Water Res. 43 (6), 3104–3116. Stedmon, C.A., Bro, R., 2008. Characterizing dissolved organic matter fluorescence with parallel factor analysis: a tutorial. Limnol. Oceanogr. 6, 572–579. Wang, J., Zhang, J., Xie, H., Qi, P., Ren, Y., Hu, Z., 2011. Methane emissions from a fullscale A/A/O wastewater treatment plant. Bioresour. Technol. 102 (9), 5479– 5485. Wang, Y., Geng, J., Peng, Y., Wang, C., Guo, G., Liu, S., 2012. A comparison of endogenous processes during anaerobic starvation in anaerobic end sludge and aerobic end sludge from an anaerobic/anoxic/oxic sequencing batch reactor performing denitrifying phosphorus removal. Bioresour. Technol. 104 (1), 19– 27. Yu, H., Song, Y., Xi, B., Zhang, M., He, X., 2012. Application of derivative synchronous fluorescence spectroscopy (DSFS) to indicate salinisation processes of saline soils in semi-arid region. Ecol. Indic. 18 (7), 532–539. Zhou, F., Guo, H., Liu, Y., Jiang, Y., 2007. Chemometrics data analysis of marine water quality and source identification in Southern Hong Kong. Mar. Pollut. Bull. 54 (6), 745–756.