Aquatic Organic Matter Fluorescence
111
of disinfectant by-product (DBP) formation. Korshin et al. (1999) demonstrated that differential fluorescence–based indexes (namely the ratio of normalized fluorescence intensities
at 500 and 450 nm using an excitation of 320 nm) used in conjunction with known treatment parameters (chlorine dose, chlorine reaction time, temperature, and NOM properties)
can be used to determine DBP formation and speciation. Korshin et al. (1999) noted that
the observed changes in fluorescence properties of the NOM after chlorination are consistent with the breakdown of NOM caused by both changes in the aromatic fluorophores and
the conformations and molecular weights of NOM molecules. In conclusion, Korshin et al.
(1999) hypothesized that the degradation of the aromatic chlorine attack sites, breakdown
of the NOM, and release of DBPs occur simultaneously. Świetlik and Silorska (2004) used
total luminescence studies, synchronous fluorescence techniques, and EEMs, to monitor
the effects of chlorine dioxide and ozone on NOM. Their research showed that oxidation
of NOM with chlorination dioxide resulted in a decrease in aromaticity and fragmentation
of NOM fractions. Ozonation of NOM resulted in the formation of a significant amount of
ozonation by-products. In both cases NOM fractions were shown to exhibit high reactivity.
Yang et al. (2008) used EEMs to obtain fluorescence intensity data from 16 organic matter fractions isolated from a variety of sources including river water, wastewater effluents,
water treatment works, lake water, and groundwater. As part of their work the researchers employed the fluorescence regional integration approach first developed by Chen
et al. (2003) to analyze the fluorescence intensity data generated by the EEMs. Using this
approach they were able to show relationships among fluorescence intensity data, organic
matter properties, and DBP formation during chloramination. Specific UV absorbance
(SUVA) values at 254 nm also correlated with DBPs formed during chloramination and
these correlations were found to be significantly higher than those derived from the EEM
data using fluorescence regional integration.
Research carried out in this field over the past 2 years has been concerned with data
mining, data analysis, and use of multivariate analysis methods of the fluorescence spectra.
Recently, Peiris et al. (2010) reported the use of principal component analysis (PCA) of
EEMs for the performance monitoring of pretreatment stages (e.g., biological filtration),
and to identify fouling events in membrane-based drinking water treatment processes. This
work demonstrated that principal component score plots could be related to high fouling
events resulting from elevated levels of particulates and/or colloidal material. The impact
of this work could result in identification of key “foulants” and the provision of an early
warning system that allows the implementation of suitable countermeasures.
Bieroza et al. (2009b) investigated the use of different multivariate analysis methods and
artificial neural networks (ANNs) for the decomposition and calibration of EEMs obtained
from DOM present in drinking waters. This research is the first to evaluate and compare
the application of different data mining methods, including multiway analysis and artificial
neural networks, for the analysis of EEMs. This research characterized the organic matter fluorescence properties and its removal in drinking water treatment. PARAFAC methodology and self-organizing maps were employed to analyze the EEM data, in order to
obtain information about the organic matter present. From this it was possible to reduce the
111
of disinfectant by-product (DBP) formation. Korshin et al. (1999) demonstrated that differential fluorescence–based indexes (namely the ratio of normalized fluorescence intensities
at 500 and 450 nm using an excitation of 320 nm) used in conjunction with known treatment parameters (chlorine dose, chlorine reaction time, temperature, and NOM properties)
can be used to determine DBP formation and speciation. Korshin et al. (1999) noted that
the observed changes in fluorescence properties of the NOM after chlorination are consistent with the breakdown of NOM caused by both changes in the aromatic fluorophores and
the conformations and molecular weights of NOM molecules. In conclusion, Korshin et al.
(1999) hypothesized that the degradation of the aromatic chlorine attack sites, breakdown
of the NOM, and release of DBPs occur simultaneously. Świetlik and Silorska (2004) used
total luminescence studies, synchronous fluorescence techniques, and EEMs, to monitor
the effects of chlorine dioxide and ozone on NOM. Their research showed that oxidation
of NOM with chlorination dioxide resulted in a decrease in aromaticity and fragmentation
of NOM fractions. Ozonation of NOM resulted in the formation of a significant amount of
ozonation by-products. In both cases NOM fractions were shown to exhibit high reactivity.
Yang et al. (2008) used EEMs to obtain fluorescence intensity data from 16 organic matter fractions isolated from a variety of sources including river water, wastewater effluents,
water treatment works, lake water, and groundwater. As part of their work the researchers employed the fluorescence regional integration approach first developed by Chen
et al. (2003) to analyze the fluorescence intensity data generated by the EEMs. Using this
approach they were able to show relationships among fluorescence intensity data, organic
matter properties, and DBP formation during chloramination. Specific UV absorbance
(SUVA) values at 254 nm also correlated with DBPs formed during chloramination and
these correlations were found to be significantly higher than those derived from the EEM
data using fluorescence regional integration.
Research carried out in this field over the past 2 years has been concerned with data
mining, data analysis, and use of multivariate analysis methods of the fluorescence spectra.
Recently, Peiris et al. (2010) reported the use of principal component analysis (PCA) of
EEMs for the performance monitoring of pretreatment stages (e.g., biological filtration),
and to identify fouling events in membrane-based drinking water treatment processes. This
work demonstrated that principal component score plots could be related to high fouling
events resulting from elevated levels of particulates and/or colloidal material. The impact
of this work could result in identification of key “foulants” and the provision of an early
warning system that allows the implementation of suitable countermeasures.
Bieroza et al. (2009b) investigated the use of different multivariate analysis methods and
artificial neural networks (ANNs) for the decomposition and calibration of EEMs obtained
from DOM present in drinking waters. This research is the first to evaluate and compare
the application of different data mining methods, including multiway analysis and artificial
neural networks, for the analysis of EEMs. This research characterized the organic matter fluorescence properties and its removal in drinking water treatment. PARAFAC methodology and self-organizing maps were employed to analyze the EEM data, in order to
obtain information about the organic matter present. From this it was possible to reduce the