Accepted Manuscript Critical Analysis of Commonly Used Fluorescence Metrics to Characterize Dissolved Organic Matter Julie Ann Korak, Aaron D. Dotson, R. Scott Summers, Fernando L. Rosario-Ortiz PII:
S0043-1354(13)00936-6
DOI:
10.1016/j.watres.2013.11.025
Reference:
WR 10335
To appear in:
Water Research
Received Date: 29 July 2013 Revised Date:
14 October 2013
Accepted Date: 18 November 2013
Please cite this article as: Korak, J.A., Dotson, A.D., Summers, R.S., Rosario-Ortiz, F.L., Critical Analysis of Commonly Used Fluorescence Metrics to Characterize Dissolved Organic Matter, Water Research (2013), doi: 10.1016/j.watres.2013.11.025. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
440
Peak Intensity Specific Peak Intensity
400 380 360 340 320 300 280 260 240
300
350
400
AC C
EP
Fluorescence Index
450
500
550
Emission Wavelength, nm
TE D
Peak Loca,on
Valid Assump,ons?
420
Excitation Wavelength, nm
Which Metrics?
What Methods? Inner Filter Correc,ons
Linear Rela,onship with DOC
Conserva,ve Mixing
ACCEPTED MANUSCRIPT 1
Critical Analysis of Commonly Used Fluorescence Metrics to Characterize
2
Dissolved Organic Matter
3
Julie Ann Korak1, Aaron D. Dotson2, R. Scott Summers1 and Fernando L. Rosario-Ortiz1*
5
Department of Civil, Environmental and Architectural Engineering, 428 UCB,
RI PT
1
4
University of Colorado-Boulder, Boulder, CO 80309, USA 2
6
University of Alaska Anchorage, Anchorage, AK 99515, USA
SC
7
Civil Engineering Department, ENGR 201
8
*Corresponding author.
[email protected]; Phone: 303-492-7607
M AN U
9 10 11
Abstract
The use of fluorescence spectroscopy for the analysis and characterization of dissolved organic
13
matter (DOM) has gained widespread interest over the past decade, in part because of its ease of use
14
and ability to provide insight of bulk DOM chemical characteristics. However, the lack of standard
15
approaches for analysis and data evaluation has complicated its use. This study utilized comparative
16
statistics to systematically evaluate commonly used fluorescence metrics for DOM characterization to
17
provide insight into the implications for data analysis and interpretation such as peak picking methods,
18
carbon-normalized metrics and the fluorescence index (FI). The uncertainty associated with peak
19
picking methods was evaluated, including the reporting of peak intensity and peak position. The linear
20
relationship between fluorescence intensity and dissolved organic carbon (DOC) concentration was
21
found to deviate from linearity at environmentally relevant concentrations and simultaneously across
22
all peak regions. Comparative analysis suggests that the loss of linearity is composition specific and
23
likely due to non-ideal intermolecular interactions of the DOM rather than the inner filter effects. For
AC C
EP
TE D
12
1
ACCEPTED MANUSCRIPT some DOM sources, Peak A deviated from linearity at optical densities a factor of 2 higher than that of
25
Peak C. For carbon-normalized fluorescence intensities, the error associated with DOC measurements
26
significantly decreases the ability to distinguish between compositional differences. An in-depth
27
analysis of FI determined that the metric is mostly driven by peak emission wavelength and less by
28
emission spectra slope. This study also demonstrates that fluorescence intensity follows conservation
29
of mass principles, but the fluorescence index does not.
30
1. Introduction
SC
RI PT
24
Dissolved organic matter (DOM) is comprised of a wide range of compounds that originate
32
from terrestrial or aquatic sources as well as from anthropogenic inputs, such as wastewater effluent
33
(Leenheer, 2009). Composition is spatially and temporally dependent and associated with source
34
material variation and environmental transformation processes (e.g., redox and photochemical
35
processing). DOM is highly heterogeneous and is often characterized by bulk parameters (e.g.,
36
dissolved organic carbon (DOC), light absorption/fluorescence, aromaticity, hydrophobicity and
37
functional groups) (Leenheer, 2009).
TE D
M AN U
31
Fluorescence spectroscopy measures the fraction of DOM that both absorbs and emits light,
39
and the fluorescent DOM fraction is often associated with aromaticity (Senesi et al., 1991). Observed
40
fluorescence can be a function of individual fluorophores contributing superimposed signals or a
41
function of intramolecular charge-transfer mechanisms between interacting fluorophores (Del Vecchio
42
and Blough, 2004).
AC C
43
EP
38
1.1. Application of Fluorescence Spectroscopy for DOM Characterization
44
Fluorescence spectroscopy has been used extensively to characterize differences in DOM
45
signatures in natural systems (Hudson et al., 2007; Fellman et al., 2010). Recently, fluorescence has
46
been applied to engineered treatment systems to characterize DOM changes (Win et al., 2000; Cheng 2
ACCEPTED MANUSCRIPT et al., 2004; Swietlik and Sikorska, 2004; Allpike et al., 2005; Beggs et al., 2009; Dwyer et al., 2009;
48
Kraus et al., 2010; Baghoth et al., 2011; Beggs and Summers, 2011; Bieroza et al., 2011; Beggs et al.,
49
2013). Early applications of fluorescence included two-dimensional emission, excitation and
50
synchronous scans and focused on spectral peak intensity and positions. Analysis of two-dimensional
51
scans also led to the development of indices such as the fluorescence index (FI) and humification index
52
(HIX) (Zsolnay et al., 1999; McKnight et al., 2001; Ohno, 2002). Recent analysis of DOM
53
fluorescence has focused on the use of three dimensional excitation-emission matrices (EEMs), as
54
illustrated in Figure 1. Peak picking methods have driven a significant portion of quantitative
55
fluorescence analysis (Coble, 1996). Furthermore, parallel factor analysis (PARAFAC) has been used
56
to deconvolute EEMs into components, but relies on collecting large datasets of EEMs (Stedmon et al.,
57
2003; Stedmon and Bro, 2008; Murphy et al., 2013).
M AN U
SC
RI PT
47
A wide range of metrics has been used to analyze EEMs. The simplest EEM characterization
59
method is qualitative observation of the presence or absence of regional intensity (Bu et al., 2010). A
60
fluorescence regional integration (FRI) method was developed that sums all intensities within a region
61
(Chen et al., 2003; Zhou et al., 2013). Peak picking is a quantitative method that records the peak
62
intensities within pre-defined regions of interest. Common peak regions relevant to freshwater include
63
two humic-like (A, C) and two heterocyclic nitrogen/polyphenolic/protein-like peaks (B, T) and are
64
outlined in Table S-1 (Coble, 1996; Leenheer, 2009). Peak picking has been executed using two
65
distinct methods. The first method reports the intensity of a stationary point (fixed wavelength) within
66
each region of interest (Leenheer, 2009; Lønborg et al., 2010; Murphy et al., 2010; Romera-Castillo et
67
al., 2011). The second approach is to define a region of interest and systematically extract the
68
maximum intensity within the region even though the location of maximum may vary between samples
69
and replicates (Senesi et al., 1991; Wang et al., 2009; Beggs and Summers, 2011; Bieroza et al., 2011;
70
Esparza-Soto et al., 2011; Guéguen and Cuss, 2011; Yang et al., 2013). This method will be referred to 3
AC C
EP
TE D
58
ACCEPTED MANUSCRIPT as an “algorithm-based” approach herein. Many studies that employ this method also record the
72
excitation and emission wavelength at which the maximum is located and use this information to infer
73
changes in DOM composition (Del Vecchio and Blough, 2004; Wang et al., 2009; Bieroza et al., 2011;
74
Esparza-Soto et al., 2011; Yang et al., 2013). Furthermore, much in the same way as specific UV
75
absorbance (SUVA) has been used to characterize DOM (Weishaar et al., 2003), carbon-normalized
76
peak intensities (intensity divided by DOC), have been used to characterize fluorescence per unit DOC,
77
often termed specific intensities (Alberts et al., 2002; Alberts and Takács, 2004; Cheng et al., 2004;
78
Jaffe et al., 2004; Allpike et al., 2005; Cumberland and Baker, 2007; Hudson et al., 2007; Wu et al.,
79
2007; Dwyer et al., 2009; Fellman et al., 2010; Beggs and Summers, 2011; Fleck et al., 2013).
M AN U
SC
RI PT
71
FI has been correlated to DOM aromaticity and is often used as a surrogate for DOM origin,
81
i.e., allochthonous or autochthonous (Win et al., 2000; McKnight et al., 2001; Cheng et al., 2004;
82
Swietlik and Sikorska, 2004; Allpike et al., 2005; Beggs et al., 2009; Dwyer et al., 2009; Kraus et al.,
83
2010; Baghoth et al., 2011; Beggs and Summers, 2011; Bieroza et al., 2011; Beggs et al., 2013). FI
84
was originally defined for fulvic acids isolated from a range of DOM sources, but has since been
85
applied to whole water analyses (McKnight et al., 2001; Beggs et al., 2009; Kraus et al., 2010; Beggs
86
and Summers, 2011). FI was initially defined as a ratio of emission intensities at 450 nm and 500 nm
87
when excited at a wavelength of 370 nm (Coble, 1996; McKnight et al., 2001). The 450 nm point was
88
chosen because it was between the peak emission for Pony Lake and Suwannee River end members.
89
The 500 nm index represented the point where emission intensity for microbial end members was half
90
of its maximum intensity. Due to instrument specific corrections that shift the emission maxima to
91
longer wavelengths by about 15 nm (red-shifted), the indices for FI were modified to the ratio of 470
92
nm to 520 nm (Cory et al., 2010).
AC C
EP
TE D
80
93
Past work has made great advances in standardizing the procedures by which EEMs are
94
collected, corrected and normalized (Lawaetz and Stedmon, 2009; Murphy et al., 2010). Murphy et. 4
ACCEPTED MANUSCRIPT al., (2010) focused on the use of instrument-specific correction factors, correcting for inner filter
96
effects and a comparison of Raman and quinine sulfate normalization methods. The study also
97
compared the reproducibility of single intensities and intensity ratios through an inter-laboratory
98
comparison. These, however, are only a few of the commonly used metrics that have been employed,
99
and the literature lacks a thorough discussion of metric limitations and a validation of underlying
100
RI PT
95
assumptions.
Many metrics rely on an underlying assumption that fluorescence intensity and DOC
102
concentration are linearly related. Using this assumption, fluorescence can be used as a surrogate for
103
DOC concentration in monitoring applications by developing site-specific correlations (Bergamaschi et
104
al., 2012; Carpenter et al., 2012; Goldman et al., 2012). Online monitoring is one application where
105
there are limitations inhibiting the use of inner filter corrections (IFCs), which in turn affects linearity
106
assumptions. Identifying metrics that do not require IFCs would benefit such applications.
107
illustrate another example, for the ratio of two intensities (i.e., peak C/ peak T, FI) to be an intrinsic,
108
compositional metric over a range of DOC concentrations, the peak intensities must vary
109
proportionally across the range of concentrations. Fundamental theory supports a linear response
110
between fluorescence intensity and DOC concentration given a number of constraints described herein
111
(Parker and Rees, 1962; Kubista et al., 1994). According to the Beer-Lambert law, at low
112
concentrations absorption is directly proportional to concentration when the composition (decadic
113
molar extinction coefficient) and cell path length are constant (Lakowicz, 2006). Non-linear absorption
114
behavior often occurs in the presence of scattered light from turbidity, fluorescing species, and
115
molecular aggregation (Lakowicz, 2006). Once organic molecules are in the excited singlet state, the
116
fluorescence quantum yield dictates the fraction of absorbed energy that is released through a radiative
117
fluorescing mechanism versus non-radiative or phosphorescing energy dissipation mechanisms. The
118
quantum yield is dependent on the fluorophore composition and surrounding environment. The 5
To
AC C
EP
TE D
M AN U
SC
101
ACCEPTED MANUSCRIPT 119
presence of quenchers, enhancers, or the formation of excimers can all impact the quantum yield
120
(Lakowicz, 2006). Past research has demonstrated that DOM behaves as a complex mixture where
121
charge transfer interactions play an important role (Del Vecchio and Blough, 2004). For given application, a variety of metrics have been used to investigate the same process. For
123
example, across a range of DOM coagulation studies, some report changes in peak intensities
124
(Afcharian et al., 1997; Cheng et al., 2004; Swietlik and Sikorska, 2004; Bieroza et al., 2009; Dwyer et
125
al., 2009; Gone et al., 2009; Kraus et al., 2010; Baghoth et al., 2011; Bieroza et al., 2011), some report
126
specific intensities (Cheng et al., 2004; Allpike et al., 2005; Dwyer et al., 2009; Beggs and Summers,
127
2011), and others use indices such as the FI (Beggs et al., 2009; Kraus et al., 2010; Beggs and
128
Summers, 2011). With the increasing use of EEMs to study DOM, the literature lacks a comprehensive
129
study on metric sensitivity and applicability. This study utilized comparative statistics to systematically
130
evaluate the commonly used fluorescence metrics to provide insight into the implications for data
131
analysis and interpretation such as peak picking methods, carbon-normalized metrics and the
132
fluorescence index (FI). The intent is to aid researchers in understanding limitations and pitfalls of
133
different DOM fluorescence data analysis tools. Application of the concepts described here will
134
provide a strong foundation for future experimental design and data analysis to allow for more
135
consistent practices and better comparison across literature. While this study relied on well-
136
characterized DOM standards and some surface waters, the results highlight important considerations
137
for any DOM analysis by evaluating spectral features common to many DOM samples. Metrics were
138
evaluated both with and without IFCs. The article will highlight metrics where inner filter effects can
139
be considered negligible with the remaining analysis documented in the Supplemental Information.
140
2. Materials and Methods
141
AC C
EP
TE D
M AN U
SC
RI PT
122
2.1. DOM sources 6
ACCEPTED MANUSCRIPT Four DOM isolates provided by the International Humic Substance Society and four surface
143
waters were used in this study. The DOM isolates include: Suwannee River NOM (SRNOM,
144
1R101N), Suwannee River Fulvic Acid (SRFA, 1S101F), Suwannee River Humic Acid (SRHA,
145
2S101H), and Pony Lake Fulvic Acid (PLFA, 1R109F). One surface water, Big Elk Meadows Lake
146
(BEM), was used throughout the study and is from a mountain lake in Colorado located at about 7500
147
feet above sea level that naturally has a high DOC concentration, low conductivity and low alkalinity.
148
The other four waters, Boulder Reservoir, Betasso Water Treatment Plant influent, Barr Lake and
149
Boulder Wastewater (WW) effluent, were analyzed to augment the FI discussion. All samples were
150
filtered with GF/F 0.7µm filters (Whatman) that were muffled at 550°C for 4 hours and rinsed with
151
250 mL of 18MΩ-cm lab-grade water (LGW). The DOM isolates and BEM sources were
152
dissolved/diluted with 10 mM, pH 7.5 phosphate buffer using LGW to 8 different concentrations
153
targeting the linear absorbance range according to the Beer-Lambert law (Lakowicz, 2006). The final
154
pH of the dilutions was 7.5±0.15.
M AN U
SC
RI PT
142
After dilution, the solutions had DOC concentrations that ranged from 0.83-40.5 mgC L-1, as
156
measured by high-temperature combustion in the presence of platinum catalyst and non-dispersive
157
infrared detection using a non-purgeable organic carbon method (TOC-VCSH, Shimadzu, MD).
158
Samples were analyzed with repeated injections with the criteria that the peak area coefficient of
159
variation was less than 2% or the standard deviation is less that 0.2, which ever was less. The
160
instrument was calibrated with potassium hydrogen phthalate and the study-wide accuracy and
161
precision of 5 mgC L-1 standards were both within 5%. The minimum reporting limit was 0.16 mgc L-1
162
following EPA method 415.3.
163
AC C
EP
TE D
155
2.2. Fluorescence Spectroscopy Analysis.
164
Fluorescence EEMs were collected using a spectrofluorometer (John Yvon Horiba FluoroMax-
165
4, NJ). All samples were measured in triplicate with a new sample aliquot for each EEM. Fluorescence 7
ACCEPTED MANUSCRIPT intensity was measured during emission scans (300 nm to 550 nm every 2 nm) at set excitation
167
wavelengths in 10 nm increments from 240 nm to 500 nm. A 5 nm bandpass for excitation and
168
emission wavelengths and 0.25 s integration time were used. Fluorescence data were corrected
169
following published methods (Murphy et al., 2010). The EEMs were collected in signal divided by
170
reference mode and incorporated instrument-specific correction factors. Unless specifically stated
171
otherwise for the parallel metrics analysis, all EEMs were corrected for primary and secondary inner
172
filter effects using a UV-Vis absorbance spectrum collected in a quartz cell with a 1 cm path length
173
(Cary-100, Agilent Technologies, CA) (Lakowicz, 2006). EEMs were blank subtracted and Raman
174
normalized based on the Raman peak area for LGW collected at an excitation wavelength of 350 nm
175
(Lawaetz and Stedmon, 2009). Corrected EEMs are presented in Raman Units (RU). Lamp scans to
176
verify monochromator calibration and cuvette contamination checks with LGW were performed daily.
177
EEMs were corrected and analyzed using MATLAB (Mathworks, MA) software. Specific
178
fluorescence intensities were calculated by dividing the fluorescence intensity by the DOC (RU L mgc-
179
1
180
nm (Cory et al., 2010). Local total optical density (ODTotal) was defined as the sum of absorbance
181
measurements at a fluorescence peak excitation and emission wavelengths, because this quantity
182
determines the magnitude of inner filter corrections applied to the measurements.
SC
M AN U
EP
TE D
). FI was calculated as the ratio of emission intensities at 470 nm and 520 nm at an excitation of 370
2.3. Statistical Methods
AC C
183
RI PT
166
184
Coefficient of variation (CV) was used as a comparative metric when triplicate reproducibility
185
was evaluated. For carbon-normalized fluorescence metrics, error propagation using the root mean
186
square approach for the quotient of two measurements was carried out to account for the uncertainty of
187
both measurements (Taylor, 1997). Fluorescence in the Peak B region for the three Suwannee River
188
isolates was not evaluated due to a lack of local intensity; fluorescence intensity was less than 0.02 RU
189
and did not increase with concentration. Regression and ANOVA analyses were performed at a 95% 8
ACCEPTED MANUSCRIPT confidence level. Model adequacy was judged by evaluating the significance of model terms and a
191
residuals analysis. The coefficient of determination, R2, was not used to judge model adequacy because
192
it cannot identify systematic residuals or curvature.
193
3. Results and Discussion
194
3.1. Method Comparison for Peak Intensity Reproducibility
RI PT
190
An intra-laboratory comparison was performed to compare the reproducibility of both peak
196
picking methods to determine if the algorithm-based approach introduced greater uncertainty
197
compared to the fixed wavelength approach. The fixed wavelength method evaluated the CV between
198
triplicates at a fixed point at the center of each peak region (Table S-1, S-2). Intensities at the center of
199
Peak A and C regions (Table 1) were the most reproducible with CVs less than 1% for most samples.
200
As concentration and fluorescence intensity decreased (0.13
Peak C ODTotal 0.09-0.13 0.18-0.26 0.14-0.20 0.20-0.25 >0.07
Peak T ODTotal 0.20-0.28 0.39-0.58 0.31-0.43 >1.90 >0.17
RI PT
DOC mg L-1 4.3-6.5 13.2-20.7 7.1-9.6 4.7-6.0 >8.2
AC C
EP
TE D
M AN U
SC
DOM SRNOM PLFA SRFA SRHA BEM
ACCEPTED MANUSCRIPT
Figure 1. Excitation emission matrix for BEM. Intensity is presented in Raman units (RU). Boxes outline the four peak regions (A, B, C and T) with labels above. Hollow circles indicate the location of the peak intensity in each region. The two points used in the I470/I520 Fluorescence Index (FI) metric are represented by filled circles connected by a line at an excitation of 370 nm. 440
2
400
1.8
380
FI
1.6 1.4
C
340
1.2
SC
360
1
320
0.8
300
B
T
280
A
260 350
400 450 500 Emission Wavelength, nm
AC C
EP
TE D
240 300
M AN U
Excitation Wavelength, nm
RI PT
2.2
420
550
0.6 0.4 0.2
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Figure 2. Peaks C (a) and T (b) intensity as a function of DOC for five DOM sources. Data is IFC corrected and error bars representing the standard deviation are smaller than marker size.
ACCEPTED MANUSCRIPT
Figure 3. Peak A and C intensity as a function of DOC concentration for SRNOM. All data are inner filter corrected and the lines represent a linear model fit to the lowest 5 DOC concentrations. Error bars representing the standard deviation between triplicate measurements may be smaller than the marker.
RI PT
3
2
SC
1.5
1
0.5
M AN U
Peak Intensity (RU)
2.5
Peak A Peak C
0 0
2
4
6 8 DOC (mg /L)
AC C
EP
TE D
c
10
12
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Figure 4. Specific Peak C intensity for 5 DOM sources within linear concentration range with IFCs applied. The inner (shorter) error bar represents the uncertainty associated with only fluorescence. The outer (wider) error bar is the uncertainty associated with both fluorescence and DOC measurements.
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Figure 5. Local curvature as a function of location relative to the peak emission. FI is the ratio of two intensities (I1/I2) spaced 50 nm apart with I1 occurring at the lower wavelength. The x-axis refers to the position of I1 relative to the emission peak. A distance of 0 nm means that the I1 is located at the peak and the second is located 50nm higher. The markers indicate where FI is calculated using the I470/I520 ratio according to Cory et. al. (46). All FIs are measured at an excitation wavelength of 370 nm.
ACCEPTED MANUSCRIPT
Figure 6. Normalized emission spectra at an excitation of 370nm plotted according to a) emission wavelength and b) relative to peak emission wavelength. The lines in plot a) indicate the location of the I470 and I520 and in b) the range of emission wavelengths where the slopes calculated 50nm apart are statistically the same. b)
RI PT
a)
0.8
SC
0.6
0.2
M AN U
0.4
PLFA SRFA 0 450 500 Emission Wavelength (nm)
550
EP
TE D
400
AC C
Normalized Intensity (RU)
1
ACCEPTED MANUSCRIPT
Figure 7. Fluorescence index and Peak C intensity as a function of composite ratio between SRHA and PLFA. Error bars representing the standard deviation may be smaller than the marker.
1.50
0.9
RI PT
FI Peak Intensity
1.32
SC
0.6
1.23
0.5
1.14
0
20
40 60 %PLFA (1− %SRHA)
AC C
EP
TE D
1.05
Peak C Intensity (RU)
0.8
M AN U
Fluorescence Index
1.41
80
0.3
0.2 100
ACCEPTED MANUSCRIPT
Highlights
EP
TE D
M AN U
SC
RI PT
Fluorescence data analysis methods for characterizing DOM are evaluated. Fluorescence-‐DOC linear relationship deviates due to quantum yield change. Fluorescence Index largely driven by emission peak location not curvature. Peak intensity follows conservative mixing but fluorescence index does not.
AC C
• • • •
ACCEPTED MANUSCRIPT
Critical Analysis of Commonly Used Fluorescence Metrics to Characterize Dissolved Organic Matter 1
RI PT
Julie Ann Korak1, Aaron D. Dotson2, R. Scott Summers1 and Fernando L. Rosario-Ortiz1* Department of Civil, Environmental and Architectural Engineering, 428 UCB, University of Colorado-Boulder, Boulder, CO 80309, USA 2
Civil Engineering Department, ENGR 201
SC
University of Alaska Anchorage, Anchorage, AK 99515, USA
M AN U
*Corresponding author.
[email protected]; Phone: 303-492-7607
Table S-1. Defined peak regions based on the values in literature (22). Peak
Range Ex: 240-270 nm Em: 380-470 nm Ex: 260-290 nm Em: 300-320 nm Ex: 300-340 nm Em: 400-480 nm Ex: 260-290 nm Em: 326-350 nm
A B
TE D
C T
Type
Humic-like
Tyrosine-like Humic-like
Tryptophan-like
Table S-2. Excitation and emission wavelength coordinates for the center of each peak region.
AC C
EP
Peak Ex (nm) Em (nm) A 260 426 B 280 310 C 320 440 T 280 338
Sp. A
Sp. B
M AN U
SC
SUVA
Sp. C
Sp. T
TE D
RI PT
DOC
RU L mgc-1 RU L mgc-1 0.12 ± 0.006 0.06 ± 0.003 0.10 ± 0.005 0.02 ± 0.0017 0.14 ± 0.007 0.03 ± 0.002 0.10 ± 0.005 0.02 ± 0.004 0.08 ± 0.004 0.05 ± 0.003 0.11 0.03 0.05 0.04 0.1 0.09 ---
FI 1.46 ± 0.007 1.29 ± 0.007 1.31 ± 0.011 1.08 ± 0.008 1.46 ± 0.007 1.34 1.4 1.83 1.95
EP
UV 254
RU L mgc-1 RU L mgc-1 0.30 ± 0.015 0.04 ± 0.002 0.28 ± 0.014 -0.41 ± 0.02 -0.27 ± 0.013 -0.25 ± 0.012 0.05 ± 0.005 0.31 0.02 0.15 0.04 0.32 0.09 ---
IHSS Peak wavelength Aromatic at Ex=370 Carbon nm % 461 ± 1.4 12 474 ± 1.4 23 474 ± 0.6 24 484 ± 1.9 31 463 ± 2.4 -468 -468 -462 -444 --
Table S- 3. Summary of compositional metrics for IHSS isolates and individual waters used for FI analysis. For the IHSS isolates and BEM, an average value across with linear range is reported with its standard deviation.
DOM Source (cm-1) mgc L-1 L mgc-1 m-1 PLFA --2.56 ± 0.03 SRNOM --3.74 ± 0.16 SRFA --3.97 ± 0.12 SRHA --6.20 ± 0.27 BEM --2.18 ± 0.05 Betasso WTP* 0.3 7.7 4 Boulder Reservoir 0.07 3.5 2 Barr Lake 0.11 6.2 1.8 Boulder WW 0.11 n/a n/a *diluted before analysis Aromatic Carbon values from (Thorn et al., 1989)"
AC C
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Peak Intensity Peak A CV%
Peak B CV%
30 20 10 0
0
5 10 Mean Signal (RU)
40 30 20 10 0
15
0
Peak C CV% 50 Coefficient of Variance (%)
M AN U
Coefficient of Variance (%)
0.5 1 1.5 Mean Signal (RU)
40 30 20 10
5 Mean Signal (RU)
10
TE D
0
2
Peak T CV%
50
0
SC
40
RI PT
50 Coefficient of Variance (%)
Coefficient of Variance (%)
50
40 30 20 10 0
0
1 2 3 Mean Signal (RU)
4
AC C
EP
Figure S-1. Coefficient of variance between triplicate measurements as a function of mean signal intensity for each peak center with inner filter corrections applied.
ACCEPTED MANUSCRIPT
Deviations from the Fluorescence-DOC Linearity Assumption 5 Points
9 Points Peak A % Difference
0.00
Peak T
−0.10 0.01 0.00 −0.01
Peak C
Peak A Peak C
Residual Intensity (RU)
−0.20 0.10
Peak T
SRNOM
0.00
1
3
4 1
4
7 0
15
7 Points
8 Points
0
−10
−4
−10
−30
4
10
10
0
0
−10
−4
−10
−30
10
10
10
0
−5
−5
1
Peak A
% Difference
0.00
Peak T
−0.20 0.04
1
TE D
0.01
8
14 1
11
22 0
21
2
Peak T
3 1
3
4 1
DOC Concentration (mgc/L)
30
5
9
8 Points
0
−5
−10
−4
−20
−30
4
10
20
0
−5
−10
−4
−20
−40
6
5
10
2
−3
−5
1
−10 14 1
8
3 Points
% Difference
EP 0.00
1
15
11
−20 22 0
21
42
DOC Concentration (mgc/L)
Peak A
Peak A Peak C
Residual Intensity (RU)
0.00
−0.00 0.01
Peak T
7 Points
10
5 Points
−0.01 0.00
AC C
BEM
0.00
4 Points
−20 7 0
10
−2
42
Peak C
3 Points 0.01
4
4
DOC Concentration (mgc/L)
−0.01
−20 4 1
3
6 Points
Peak C
Peak A Peak C
Residual Intensity (RU)
−0.20 0.20
Peak T
PLFA
0.20
−0.02
0
M AN U
6 Points
10
DOC Concentration (mgc/L)
DOC Concentration (mgc/L) 0.60
9 Points
10
−10
30
6 Points
4
RI PT
6 Points
SC
5 Points 0.20
4 Points
5 Points
6
4
5
2
0
0
−2
−4
−5
4
2
4
1
−1
1
−2
−4
−2
15
5
15
5
0
5
−5
1
2
3
−5
1
3
4
−5
1
5
9
DOC Concentration (mgc/L)
Figure S-2. Residual analysis for SRNOM, PLFA and BEM showing the upper and lower limits of the linearity threshold and full range of DOC concentrations. The left column presents the residuals in Raman units and the right column presents the residuals as a percentage of the measured intensity.
ACCEPTED MANUSCRIPT
9 Points
6 Points Peak A % Difference
0.00
Peak T
−0.04 0.02 0.00 −0.01
Peak C
Peak A Peak C
Residual Intensity (RU)
−0.10 0.04
Peak T
SRHA
0.00
1
3
5 1
3
6 0
12
DOC Concentration (mgc/L) 6 Points
7 Points
9 Points
−0.01
Peak A
Peak C
Peak T
1
4
7 1
5
−5
−10
−10
−20
−40
5
10
10
0
0
−10
−5
−10
−30
1
40
40
0
10
10
1
−20 5 1
3
6 Points
% Difference
0.00
0
10−1
19
38
TE D
DOC Concentration (mgc/L)
3
−20 6 0
7 Points
12
25
9 Points
5
5
10
−3
−3
−10
M AN U
Peak A Peak C
Residual Intensity (RU)
0.00 −0.10 0.01
Peak T
SRFA
−0.20 0.10
20
DOC Concentration (mgc/L)
0.60 0.20
9 Points
10
−1
25
7 Points
10
RI PT
7 Points
SC
6 Points 0.10
−10
−10
−30
10
10
10
3
0
−10
−5
−10
−30
10
10
10
0
0
−5
−10
1
4
−10 7 1
5
−20 10 −1
19
38
DOC Concentration (mgc/L)
AC C
EP
Figure S-3. Residual analysis for SRHA and SRFA showing the upper and lower limits of the linearity threshold and full range of DOC concentrations. The left column presents the residuals in Raman units and the right column presents the residuals as a percentage of the measured intensity.
ACCEPTED MANUSCRIPT
8
1.6
7
1.4
6
1.2
4 3 2 Peak A Peak C
0.6
10
20 DOC (mgc/L)
30
0.2 0
40
0
2
M AN U
0
SRHA 2.5
4 DOC (mgc/L)
6
8
SRFA
TE D
1
0.5
Peak Intensity (RU)
5
1.5
4
3
2
1
0
2
EP
Peak A Peak C
0
Peak A Peak C
6
2
Peak Intensity (RU)
0.8
0.4
1 0
1
SC
5
RI PT
BEM 1.8
Peak Intensity (RU)
Peak Intensity (RU)
PLFA 9
4
6 8 DOC (mgc/L)
10
Peak A Peak C 0
0
5
10 DOC (mgc/L)
15
AC C
Figure S-4. Peak A and C intensity as a function of DOC concentration for PLFA, BEM, SRHA and SRFA. The lines represent a linear model fit to the lowest 5 DOC concentrations. Error bars representing the standard deviation between triplicate measurements may be smaller than the marker.
ACCEPTED MANUSCRIPT
1.6
0.4
1.4
0.35
1.2
0.3
0.8 0.6 0.4
0.2 0.15 0.1
Peak B Peak T 0
10
20 DOC (mgc/L)
30
0.05
40
0
0
2
M AN U
0.2 0
0.25
SC
1
RI PT
BEM 0.45
Peak Intensity (RU)
Peak Intensity (RU)
PLFA 1.8
4 DOC (mgc/L)
Peak B Peak T 6
8
AC C
EP
TE D
Figure S-5. Peak B and T intensity as a function of DOC concentration for PLFA and BEM. The lines represent a linear model fit to the lowest 5 DOC concentrations. Error bars representing the standard deviation between triplicate measurements may be smaller than the marker.
ACCEPTED MANUSCRIPT
SRNOM
SRHA
0.3
0.14
0.15
0.1
0.05
0
2
4
6 8 DOC (mgc/L)
10
12
SRFA
0.4 0.35
0.04
0
0
2
4
0.3 0.25
6 8 DOC (mgc/L)
10
TE D
Peak T Intensity (RU)
0.06
0.02
0.45
0.2 0.15 0.1
0
EP
0.05 0
0.08
M AN U
0
0.1
SC
Peak T Intensity (RU)
Peak T Intensity (RU)
0.2
RI PT
0.12
0.25
5
10 DOC (mgc/L)
15
AC C
Figure S-6. Peak T intensity as a function of DOC concentration for SRNOM, SRHA and SRFA. The lines represent a linear model fit to the lowest 5 DOC concentrations. Error bars representing the standard deviation between triplicate measurements may be smaller than the marker.
ACCEPTED MANUSCRIPT
BEM SRHA SRFA
DOM Source SRNOM PLFA BEM SRHA
Max. DOC mg L-1 4.3 13.9 9.3 40.4 8.2 3.02 12.0 4.6 18.7
Peak C Slope Regression p y=0.107x-0.021