Fluorescence and Dissolved Organic Matter
41
is the colloidal fraction, which consists of suspended solids operationally considered as
solutes (Morel and Gschwend, 1987). Colloidal organic matter in natural waters is composed of living and senescent organisms, cellular exudates, and partially to extensively
degraded detrital material, all of which may be associated with mineral phases (Lead and
Wilkinson, 2007). Distinctions between dissolved (filtered), colloidal, and particulate
organic matters are important when measuring the optical properties of a sample. In some
instances, samples can be analyzed without filtration. For example, the determination of
chlorophyll associated with algae by fluorescence has long been used to study algal distributions and ecosystem dynamics in lakes and marine waters (Berman, 1972). Other
examples include the application of fluorescence to the study of colloids and nanoparticles
(Fatin-Rouge and Buffle, 2007). In general, however, spectroscopic methods, such as fluorescence and UV-visible absorption, are sensitive to the presence of particulate material
in a sample, and, to obtain usable spectroscopic data, samples need to be appropriately
filtered or optical data need to be corrected for the presence of particles (Karanfil et al.,
2005; Saraceno et al., 2009).
DOM fluorescence data can be collected and presented in a variety of ways. The simplest
approach is to measure fluorescence intensity of a single excitation–emission wavelength
pair. This approach is the one used by modern in situ instruments designed to measure signals associated with DOM, chlorophyll, or specific fluorescent compounds, such as rhodamine. The commonly reported parameter fluorescence index (FI; Cory and McKnight,
2005) is simply the ratio of the emissions intensities determined at wavelengths of 470 and
520 with an excitation wavelength of 370 nm. Data can also be collected and presented as
emission spectra measured at a single-excitation wavelength, or as the absorption or excitation spectrum at a single-emission wavelength. Another useful approach is the presentation
of synchronous spectra, wherein emission data are offset by a constant amount from excitation wavelength (i.e., λ + Δλ). The offset in wavelength is determined by the fluorescence
properties of the fluorophores of interest (Cabaniss and Shuman, 1987; Liu et al., 2006;
Ziegmann et al., 2010). In theory, this approach allows measurement of fluorophores of
interest and may provide less ambiguous data when properly constrained.
With the advent of modern spectrofluorometers, data are commonly collected and presented as excitation–emission matrices (EEMs). EEMs spectra contain a large amount of
information and are usually displayed graphically (Figure 2.2). The large amount of data
associated with EEMs spectra are often presented as the intensities of individual excitation–
emission pairs or peaks associated with different compound classes, an approach commonly
referred to as “peak picking.” The most common peak assignments are given in Table 2.1
and graphically illustrated in Figure 2.2c. Statistically based approaches, such as hierarchal
clustering (Jiang et al., 2008), partial least squares regression (Persson and Wedborg, 2001;
Hall et al., 2005), principal component analysis (Persson and Wedborg, 2001; Hall and
Kenny, 2007), and parallel factor analysis (PARAFAC; Stedmon et al., 2003), use all of the
data contained in an EEM spectrum to identify spectral features and determine contributions to the spectrum of different areas or ex/em regions of fluorescence. These approaches
employ curve fitting techniques and assume linear behavior of the different components of
41
is the colloidal fraction, which consists of suspended solids operationally considered as
solutes (Morel and Gschwend, 1987). Colloidal organic matter in natural waters is composed of living and senescent organisms, cellular exudates, and partially to extensively
degraded detrital material, all of which may be associated with mineral phases (Lead and
Wilkinson, 2007). Distinctions between dissolved (filtered), colloidal, and particulate
organic matters are important when measuring the optical properties of a sample. In some
instances, samples can be analyzed without filtration. For example, the determination of
chlorophyll associated with algae by fluorescence has long been used to study algal distributions and ecosystem dynamics in lakes and marine waters (Berman, 1972). Other
examples include the application of fluorescence to the study of colloids and nanoparticles
(Fatin-Rouge and Buffle, 2007). In general, however, spectroscopic methods, such as fluorescence and UV-visible absorption, are sensitive to the presence of particulate material
in a sample, and, to obtain usable spectroscopic data, samples need to be appropriately
filtered or optical data need to be corrected for the presence of particles (Karanfil et al.,
2005; Saraceno et al., 2009).
DOM fluorescence data can be collected and presented in a variety of ways. The simplest
approach is to measure fluorescence intensity of a single excitation–emission wavelength
pair. This approach is the one used by modern in situ instruments designed to measure signals associated with DOM, chlorophyll, or specific fluorescent compounds, such as rhodamine. The commonly reported parameter fluorescence index (FI; Cory and McKnight,
2005) is simply the ratio of the emissions intensities determined at wavelengths of 470 and
520 with an excitation wavelength of 370 nm. Data can also be collected and presented as
emission spectra measured at a single-excitation wavelength, or as the absorption or excitation spectrum at a single-emission wavelength. Another useful approach is the presentation
of synchronous spectra, wherein emission data are offset by a constant amount from excitation wavelength (i.e., λ + Δλ). The offset in wavelength is determined by the fluorescence
properties of the fluorophores of interest (Cabaniss and Shuman, 1987; Liu et al., 2006;
Ziegmann et al., 2010). In theory, this approach allows measurement of fluorophores of
interest and may provide less ambiguous data when properly constrained.
With the advent of modern spectrofluorometers, data are commonly collected and presented as excitation–emission matrices (EEMs). EEMs spectra contain a large amount of
information and are usually displayed graphically (Figure 2.2). The large amount of data
associated with EEMs spectra are often presented as the intensities of individual excitation–
emission pairs or peaks associated with different compound classes, an approach commonly
referred to as “peak picking.” The most common peak assignments are given in Table 2.1
and graphically illustrated in Figure 2.2c. Statistically based approaches, such as hierarchal
clustering (Jiang et al., 2008), partial least squares regression (Persson and Wedborg, 2001;
Hall et al., 2005), principal component analysis (Persson and Wedborg, 2001; Hall and
Kenny, 2007), and parallel factor analysis (PARAFAC; Stedmon et al., 2003), use all of the
data contained in an EEM spectrum to identify spectral features and determine contributions to the spectrum of different areas or ex/em regions of fluorescence. These approaches
employ curve fitting techniques and assume linear behavior of the different components of