EPA-600/4-81-001
February 1981
REMOTE MONITORING OF ORGANIC CARBON IN SURFACE WATERS
by
Michael Bristow and David Nielsen
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Las Vegas, Nevada
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S.'ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NEVADA 89114
-------
EPA-600/4-81-001
February 1981
REMOTE MONITORING OF ORGANIC CARBON IN SURFACE WATERS
by
Michael Bristow and David Nielsen
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Las Vegas, Nevada
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
LAS VEGAS, NEVADA 89114
-------
DISCLAIMEH
Thi s report has been revi ewed by the Envi ronmenta 1 i4onitori ng Systems
Laboratory, U.S. Environmental Protection Agency, and approved for
publication. Mention of trade names or commercial products does not
constitute endorsement or recommendation for use.
i;
-------
CONTENTS
F; 9 u res. . . . . . . . . . . . . . . . . . . . . . .
Tab 1 es . . . . . . . . . . . . . . . . . . . . . . .
Abbreviations and Symbols. . . . . . . . . . . . . .
Page
.888."'" V
. . . . . . . . .. vi;
. . . . . . . . . . viii
L
2.
Introduction
. . . . . . . .
. . . . .
. . . . . .
. . . .
. . . .
Conclusions. . . .
. . . .
. . . . .
. . . . . . . . . .
. . . . .
3.
4.
Recommendations.
. . . . . . .
. . . . . . .
. . . .
.......
Review. . . . . .
. . . . . . . .
'88"""
. . . . . . .
Existing methodology for determining total organics
in surface waters. . . . . . . . . . . . . . . . . . . . . .
Possible methods for remote monitoring total organics
in surface waters. . . . . . . . . . . . . . . . . . . . . .
5. Previous Laboratory Studies on the Fluorescence
of Natural Waters. . . . . . . . . . . . . . .
8""""
6.
Effects of Optical Attenuation on Water
Fluorescence Measurements. . . . . . .
. . . . . . . . . . . . .
Raman correction of airborne fluorescence measurements. . . . .
Raman correction of laboratory fluorescence measurements. . . .
7.
Sample Collection, Analysis and Preservation.
Sample sources. . . . . . . . .
Organic carbon analyses. . . .
Fluorescence analyses. . . . .
Preservation and preparation of
. . . . .
. . . . .
. . . . . .
.....
. . . . .
. . . . . .
. . . . . . . . . .
. . . . .
......
. . . . .
flubrescence samples. .
. . . . .
8.
Data Analysis. . . . . . . . . . . .
. . . . . . . .
. . . . . . .
Linear correlation results. . . . .
Multiple correlation effects. . . .
. . . . .
. . . .
. . . . .
. . . .
. . . . . . . . . .
9.
Differences between Laser Fluorosensor and Spectrofluorometer
Measurements of FR/R . . . . . . . . . . . . . . . . . . . . . .
i i i
1
3
4
5
5
7
9
13
17
20
21
21
22
23
28
36
36
59
60
-------
CONTENTS (Continued)
10. Discussion. . . . . . . . . . . . .
. . . . . . . . . . .
. . . .
Inherent limitations in the FA/R-
DOC correlation method. . . . .
. . . . . .
. . . . . . . . . 67
Calibration of airborne FR/R data in terms of
DOC concentration. . . . . . . . . . . . . . . . .
. . . . .
Improvements in experimental methodology.
Detailed presentation of conclusions.
. . . .
. . . . . . .
. . . . . .8 . .
. . . . .
References. . . . .". . . . .
. . . . . . . .
.........
. . . .
iv
Page
67
70
73
74
77
-------
FIGURES
Page
1
Principle of operation of airborne laser fluorosensor. . .
. . . . 14
2
Corrected fluorescence emission spectra for lake and ultra
p~re water samples. . . . . . . . . . . . . . . . . . . . . . . 15
3
Optical diagram of corrected-spectra spectrofluorometer
(Perkin-Elmerf'WF4).................... .
. . 25
4
5
Tra~smission curve for Cation-X liquid filter. . . . . .
. . . . . 26
Schematic showing water fluorescence and Raman emission
parameters obtained from spectra produced using a
laboratory spectrofluorometer. . . . . . . . . . . . .
. . . . . 27
6
Variation of Fmax/R with time for different preservation
and preparatlon procedures as applied to identical subsamples.
CV is the coefficient of variation (S/i) . . . . . . . . . . . . 30
7
8
Variation of DOC (mg/l) with Fmax/R for 158 samples. . . . . . . . 37
Variation of DOC (mg/l) with FR/R for 158 samples. . . . . . . . . 38
Variation of TOC (mg/l) with Fmax/R for 158 samples. . . . . . . . 39
9
10
Variation of TOC (mg/l) with FR/R for 158 samples. . . . .
. . . . 40
11
Variation of linear correlation coefficient, r, with
wavelength for F1/R versus DOC, F1/R Versus TOC, .
Fl/R versus POC and Fl/R versus turbidity. . . . .
. . . .
. . . 46
12
Variation of turb.idity (NTU) with bandwidth of fluorescence
emission spectrum, BW (nm), for 110 samples. . . . . . . .
. . . 49
13
14
Variation of DOC (mgl1) with Fmax/FR for 158 samples. . . . . . . 54
Variation of DOC (mg/l) with central wavelength of fluorescence
emission band, lcen (nm), for 155 samples. . . . . . . . . . . . 55
v
-------
~e
15
Variation of central wavelength of fluorescence emission band,
Acen (nm), wi th Fmax/FR. . . . . . . . . . . . . . . . . . . . . .
Changes in fluorescence spectrum parameters produced by
progressively diluting an Atchafalaya River sample (#E-1)
with high-purity water. . . . . . . . . . . . . . . . . .
57
56
16
. . . .
17
Optical layout for laboratory simulation of airborne laser
fluorosensor for monitoring surface water organics. . . . . . . .
Emission spectrum showing [OH]-stretch water Raman band
from high-purity water sample at 381 nm, excited by
N2 laser at 337 nm . . . . . . . . . . . . . . . . . .
62 .
61
18
. . . . . .
19
Variation of FR/R from simulated laser fluorosensor
measurements with comparable values measured using a
laboratory spectrofluorometer for a variety of field samples. . . . 63
vi
-------
"
TABLES
1
Summary of Published Fluorescence/Organic Carbon
Correlation Data. . . . . . . . . . . . . . . .
. . . .
. . . . . .
2
Correlation Coefficients for Fluorescence Emission Parameters
versus Water Quality Parameters. . . . . . . . . . . . . .
. . . .
3 Correlation Coefficients for Relationships Between Water
Quality Parameters. . . . . . . . . . . . . . . . . . . . . . . . .
4 Maximum, Mean and Minimum Values for Fluorescence and
Water Quality Parameters. . . . . . . . . . . . . .
. . . . . . . .
vi i
Page
12
42
43
44
-------
Abbreviations
LIST OF ABBREVIATIONS AND SYMBOLS
BOD
BW
CCE
CHLA
COC
COD
COND
CV
DO
DOC
FWHM
GAC
NTU
[OH]
PH
POC
RFM
S
TOC
TRANSP
TURB
UV
x.
Subscripts
C
. CEN
L
LFS
M
MAX
R
SPF
T
f
W
A
o
-
Biochemical oxygen demand
Bandwidth of fluorescence spectrum
Carbon chloroform extract
Chlorophyll .!
Colloidal organic carbon
Chemical oxygen demand
Conductivity
Coefficient of variation
. Dissolved oxygen
Dissolved organic carbon
Full width at half maximum
Granulated activated carbon
Nephelometric turbidity unit
Notation for hydroxyl radical
Sampl e pH
Particulate organic carbon
Rapid fluorometric method
Sample standard deviation
Total organic carbon
Secchi disc transparency
Turbidity
Ultraviolet
Sample mean
.-
Central wavelength, per computer
Central wavelength
Laser (excitation) wavelength
Laser fluorosensor value
Maximum wavelength, per computer
Maximum wavelength
Raman wavelength
Spectrofluorometer value
Value at Toe
Fluorescent substance
Water
Fluorescence wavelength
Value at ooe
printout.
printout.
viii
Same as CEN.
Same as MAX.
-------
Superscripts
D
F
NF
P
Symbols
A
C
DOC
F
H
K
L
N
P
R
T
b
d
h
i
k
m
n
q
r
s
t
y
a
°1
°2
°3
A
a
T
LIST OF ABBREVIATIONS AND SYMBOLS (Continued)
Dissolved substance
Fluorescent substance
Nonfluorescent substance
Particulate substance
-
Proportionality constant in Equation 9.
Proportionality constant in Equation 5
Concentration of dissolved organic carbon
Fluorescence power or intensity
Elevation of sensor above water surface
Diffuse attenuation coefficient
Fluorescence wavelength, per computer printout. Same as A.
Number of samples
Laser output power
Raman power or intensity
Temperature
Carbon conversion factor, as defined in Equation 9
Constant accounting for system and environmental factors
Depth of sample volume below water surface
Dimensionless number
Effective optical attenuation coefficient of water sample
Temperature coefficient
Concentration
Dimensionless number
Pearson product-moment linear correlation coefficient
Dimensionless number
. Dimensionless number
Dimensionless number
Beam attenuation coefficient
Constant, defined in Equation 4
Constant, defined in Equation 6
Constant, defined in Equation 7
Fluorescence wavelength
Excitation cross section
Pulse width
ix
-------
SECTION 1
I NTROOUCTI ON
Waterborne organic substances are of concern because of their recognized
impact on public health, and because of the effects of manmade organics on the
aquatic ecosystem. Alarm has been voiced over the presence of toxic and
potentially carcinogenic organic materials in drinking water. High concen-
trations of organics in natural waters can lead to malodorous and toxic
conditions which reduce the utility of these waters as a drinking water
resource ~nd for recreational purposes. However, a certain concentration of
organic substances of natural origin is essential for maintaining a balance in
the energy and carbon cycles of the ecosystem, particularly with regard to
bacterial growth. Bacteria play an essential role in the biodegradation .of
organic substances, both natural and manmade. However, the large amounts of
oxygen required by this process to degrade high concentrations of organics can
lead to severe oxygen depletion with predictqble effects on higher aquatic
life forms. Alternatively, high concentrations of toxic organics may lead to
total elimination of the local biological community.
Organic carbon determinations are routinely accomplished by making
laboratory analyses on grab samples. This approach is both time-consuming and
costly in terms of manpower and facilities. In addition, because of the
relatively long time required to take grab samples from launches or
helicopters, it is rarely possible to obtain a synoptic record of organic
carbon distribution for a given water surface due to water movement and
diurnal effects.
In contrast, both airborne and satellite remote monitori'ng techniques are
capable of rapidly and cost effectively providing data for certain water
quality parameters from large areas of water surface without influencing the
nature of the sample. The principal limitation of remote sensing in its
present form is the inability to provide information from below the surface
layer. For active as well as passive sensors operating in the optical region,
the depth of this layer is generally on the order of 0.3 to 30 meters, but is
more correctly characterized by the attenuation length, the reciprocal of the
optical attenuation coefficient measured at a specific wavelength. The
concept of a single characteristic attenuation length makes the assumption.
that the substances responsible for this attenuation are homogeneously
distributed.
The subject of this laboratory study is the feasibility of measuring the
concentration of organic carbon in the surface layers of natural waters from
an airborne platform. At the present time, no proven remote sensing technique
exists that is capable of monitoring, directly or indirectly, the organic
carbon content of surface waters. It is therefore intended that the results
1
-------
of this study will be used in the design of a compact integrated airborne
system capable of mapping trends, gradients and anomalies in the distribution
of organic carbon in surface waters.
2
-------
SECTION 2
CONCLUSIONS
The results of this laboratory feasibility study demonstrate that the
Raman-normalized fluorescence emission induced in surface waters by
ultraviolet radiation can be used to provide a unique airborne remote sensing
capability for monitoring the concentration of DOC.
Airborne laser fluorosensors that utilize this principle will provide a
synoptic survey capability for rapidly and cost effectively producing isopleth
maps that show concentrations of surface water DOC. These maps can be used
for delineating gradients, temporal changes and anomalies in the distribution
of total dissolved organics in the surface 1ay~rs of rivers, lakes and coasta"
waters. Anomalous features in the airborne data that cannot be readily
explained on the basis of existing information can then be investigated in
more detail either by means of in situ monitoring or by laboratory analyses of
grab samples. Specific applications include baseline monitoring of pristine
lakes, verification of lake cleanup and restoration, sampling network design,
ecosystem modeling, and defi.ning the location and extent of point and nonpoint
sources of organic pollution of unknown origin. Sources of organic carbon
include, but are not restricted, to harbors, marinas, septic tank leachates,
oil refineries and industrial sites adjacent to waterways, pulp and paper mill
effluents, feed lot runoff, municipal sewage effluents, agricultural and.
silviculture activities, and surface runoff containing organic materials from
both living and decayed natural vegetation. (See Section 10 for a detailed
presentation of conc1usipns)..
3
-------
SECTION 3
RECOMMENDATIONS
Factors to be considered when designing an airborne laser fluorosensor
for monitoring dissolved organic carbon (DOC) are:
(i) The Raman normalized fluorescence signal should be used in
preference to the uncorrected fluorescence signal as an indicator of DOC.
(ii) The Raman normalized fluorescence emission measured at the water
Raman wavelength should be used in preference to the maximum value of the
Raman norm~lized fluorescence emission as the parameter characterizing surface
water DOC.
(iii) Careful consideration should be given to conflicting requirements
when choosing a (laser) wavelength for exciting fluorescence in waterborne
organi cs.
With regard to the operation of an airborne laser fluorosensor it is
suggested that:
(i) The airborne fluorescence data should be regarded as a more reliable
indicator of surface water DOC than of total organic carbon (TOC), because of
the unpredictable and relatively nonfluorescent nature of the particulate
organic fraction (paC), where TOC = DOC + pac.
(ii) The airborne measurements of the Raman non-nalized fluorescence
- emission should be calibrated directly in terms of DOC by making a small
. number of selected ground truth measurements of DOC on samples collected
the sensor flight path concurrent with the airborne survey.
(ii;) The ground truth samples for DOC analysis should be prepared with
precleaned silver membrane filters, treated with high purity acid and sealed
into glass ampoules in those situations where samples cannot be analyzed
directly at the field collection site.
under
4
-------
'SECTION 4
REV I EW
EXISTING METHODOLOGY FOR DETERMINING TOTAL ORGANICS IN SURFACE WATERS
Dissolved, colloidal and suspended particulate organic matter in lake and
river water, which is predominantly natural in origin, can generally be
classified under the headings of humic, fulvic and hymatomelanic acids of
uncertain structure, with the addition of smaller percentages of polysac-
charides, peptides and tannins (Rook, 1977). Although not toxic in low
concentrations, there is evidence that these substances can act as carriers
for more toxic manmade pollutants, both inorganic (Schnitzer, 1971) and
organic (Ogner and Schnitzer, 1970). Evidence now exists that humic and
fulvic. acids in surface waters act as carriers of trace metals (Nriagu and
Coker, 1980). Waters which are highly enriched in organics contain, in
addition to the above mentioned substances, significant concentrations of
small biodegradable molecules, sulphonated 1ignins, sulphonated aromatics and
chlorinated hydrocarbons (Rook, 1977).
Recently, much concern has been expressed over the presence of trace
amounts of halogenated hydrocarbons detected in drinking water (Trussell and
Umphres, 1978). A prime example is chloroform, a trihalomethane, which is
frequently present in parts per billion concentrations. It is now generally
accepted that this substance is a carcinogen in animals (Tardiff, 1977). It
is a1so'recognized that chlorinated hydrocarbons are produced during routine
chlorination from natural dissolved organics such as humic and fulvic acids,
free fatty aci ds, alcohol s, etc., that are present in the source waters (Rook,
1977; Youssefi et a1., 1978).
Another group of organics known as polynuclear aromatic hydrocarbons has
been detected in the water environment. These substances, usually formed in
combustion or other high temperature processes, are also considered to be
highly carcinogenic , e.g. 3,4-benzpyrene (Ande1man and Suess, 1970). Although
their solubility in water is extremely low, these substances are suspected of
being transported by adsorption onto colloidal aggregates of natural organics
or of becomi ng sol ubil i zed by detergents ors i mil ar substances. '
Waterborne organics, whether dissolved or particulate, can be categorized
as originating from either point or nonpoint sources. Typical point sources
are municipal sewage effluents, industrial discharges and waste waters, and
mineral oil pollution from harbors, marinas, oil tankers, shipping and oil
refineries. Nonpoint sources generally can be broken down into three
categories. The first category encompasses runoff waters from highways,
construction sites, natural vegetation, lands under silviculture and
5
-------
agriculture, and landfills such as those associated with mining activities.
Second, groundwater leachates from sources such as septic tank facilities
introduce organics into surface waters. Third, the atmosphere introduces
organics into surface waters, through the medium of rainwater and fallout of
atmospheric particulates (Wakeham, 1977), and also in the form of airborne
litterfall (Gasith and Hasler, 1976).
Total organic carbon (TOC) and its dissolved component (DOC) are now
generally accepted as the primary measurement parameters for gauging the
organic carbon content of a water sample. They are nonspecific water quality
indicators which provide an accurate indication of the concentration of total.
organics present in a water sample (Kehoe, 1977).
As indicators of waterborne organic carbon, TOC and DOC have distinct
advantages over other parameters such as biochemical oxygen demand (BOO) and
chemical oxygen demand (COO) that have been widely used for this purpose in
the past (Kehoe, 1977; Dishman, 1979). The BOD and COO water quality tests
involve measuring the oxygen consumed during a controlled biochemical or
chemical oxidation process and then calculating the corresponding concen-
tration of organic carbon. The TOC and DOC measurements can be readily made
within a few minutes by automated or semi automated instrumentation (Goulden,
1976; Michalek et al., 1977; Dishman, 1979), whereas the BOD and COO
measurements take respectively five days and approximately two to three hours
to accomplish by means of wet chemistry. The BOD test fails to account for
that fraction.of the organic carbon that is not biodegradeable; in many
situations this can be a significant fraction of the total. In addition, the
BOD test is susceptible to a number of problems including the effects of
biological seed acclimatization, pH and toxic substances, and fails to oxidize
certain organic materials. Similarly, the COO test also fails to oxidize
certain refractory organics, and is sensitive to the presence of .inorganic
compounds that are capable of consuming available oxygen.
On the other hand, the TOC and DOC methods are relatively free of
interferences and are capable of nearly 100 percent organic carbon recovery.
Possibly the principal advantage of the TOC and DOC methods over the BOD and
COD methods are their ability to encompass the complete range of
concentrations from grossly polluted waters and discharge effluents through to
high quality waters «1 mg/l); at an organic carbon concentration of 1 mg/l,
accuracy of the TOC method should be better than zS percent. By contrast, BOO
and COD values are not reliable below 10 mg/l. It is therefore not surprising
that universal relationships between TOC and BOO .or TOC and COD do not exist,
although high correlations between TOC and either BOD or COO for specific
waste or fresh waters have been obtained (Stevens and Symons, 1974; Chandler
et al., 1976; Dishman, 1979; Constable and McBean, 1979). It should therefore
be emphasized that, in general, the BOD and COO measurements are unreliable as
indicators of organic carbon unless a correlation can be established for a
specific water type. Similarly, the TOC and DOC measurements should not be
utilized to indicate oxygen demand unless specific relationships have been
established with either BOD or COD.
In a study of the variation of TOC for Lake Superior, Maier and Swain
(1978) have indicated that the use of BOO as a measure of waterborne organics
6
-------
is so insensitive that early signs of irreversible damage to water quality
might go undetected. Similarly, in a 5-year study of the Rhine in West
Germany, DOC was found to be a significantly 10re sensitive indicator of
clean-up progress than was COD, particularly when variations in river flow
rate were taken into account (Sontheimer, 1977).
fhe abil ity to remotely monitor TOC over extensive areas woul d therefore
be advantageous not only because of its cost effectiveness in relation to
direct sampling methods, but also because it is possibly the only way in which
a synoptic record can be obtained at a single time. On a single airborne
mission over a given water body, a remote sensing device for monitoring
organic carbon could make the equivalent of several million TOC analyses on
surface water samples within a period of one hour. By use of relatively cheap
microprocessor technology these airborne data could then be converted in near
real time into a map of the water surface showing isopleths of surface water
organic carbon concentration. Such a map would be used as a general purpose
screening or survey tool for locating and monitoring anomalies, gradients,
point and nonpoint sources, and, in particular, sources of unknown origin.
TOC surveys such as that conducted by Maier and Swain (1978) for Lake
Superior, would be an ideal application for a device of this kind. In
addition, such maps might be used for designing sampling networks or for
indicating how best to further investigate, by direct sampling methods,
sources of waterborne organics of known or unknown origin.
POSSIBLE METHODS FOR REMOTE MONITORING TOTAL ORGANICS IN SURFACE WATERS
Currently, passive remote sensing in the form of multispectral scanner
imagery has achieved only limited success in predicting the concentration of
water quality parameters. Much effort has been expended on formulating
interpretation schemes for-extracting information on concentration of
phytoplanktonic chlorophyll ~ (Miller et al., 1977; Morel and Prieur, 1977;
Wilson et al., 1978; Johnson, 1978) and suspended sediment (Johnson et al.,
1977; Bowker and Witte, 1977; Sydor et al., 1978) in surface waters.
Empirical regression models based on extensive concurrent ground truth data
must be devised for each application in order to reliably interpret the
satellite or airborne data. This is necessary because of the overlap in the
spectral reflectance characteristics of suspended sediment and phytoplankton
in the visible and near infrared spectral regions. Provided that the
concentration of one substance remains dominant, then the model is likely to
hold, but this can be quickly undermined by significant fluctuations in algal
production (especially at the surface) and suspended sediment runoff that
occur on a seasonal basis. Extrapolation of a given interpretation model to
data from a different geographical location is generally not feasible because
each waterbody has its own peculiar distribution of phytoplankton and
suspended sediment types with their. own specific optical absorption and
scattering properties.
An attempt to remotely monitor organics in surface waters has been made
by Davis and Fosbury (1973). They were able to correlate the optical density
of black and white multispectral photography with a number of water quality
parameters. A correlation coefficient of -0.60 was obtained between TOC
7
-------
ground truth data and film optical density from flights over the Houston Ship
Channe 1. No attempts were made to cali brate film densi ty wi th respect to TOC
concentration. Clearly, if such an approach is to be developed, the effects
of atmospheric backscatter, atmospheric transmission and sky reflection would
need to be eliminated and corrections made for the effects of chlorophyll and
suspended sediment. It is possible that this correlation is more circum-
stantial than causal due to the fact that for this one measurement, the solar
backscatter was predominantly from material that was organic in origin.
However, situations frequently exist where this is not true.
More recently airborne lidar devices employing a laser induced
fluorescence approach have shown promise in the area of water quality
monitoring. These devices, known collectively as laser fluorosensors, have
been used to detect surface water oil spills (Kim and Hickman, 1975; O'Neil et
al., 1975; Bristow, 1978), and to generate fluorescence emission spectra that
are characteristic of the oil slick (Fantasia and Ingrao, 1974; O'Neil et al.,
1980). In addition, they have been successfully used to remotely profile
:surface waters for chlorophyll a present in phytoplankton (Bristow et al.,
1979; Farmer et al., 1979), and-for the lignin effluents from pulp and paper
mills (O'Neil et al., 1975; Bristow, 1978). This latter application is of
particular relevance to the present study as it represents a special category
of a highly fluorescent dissolved organic material that can be readily
detected using an airborne laser fluorosensor.
Although no attempt has yet been made to use the laser fluorosensor
method for remotely sensing total organics in surface waters, a number of
studies in the published literature indicate the existence of a relationship
between the concentration of total organics in a given sample and the
intensity of the UV excited fluorescence emission. The following chapter
reviews the res~lts of these studies with the specific purpose of gauging the
potential of any such relationship as a means for remotely monitoring
waterborne organic carbon.
8
-------
SECT! ON 5
PREVIOUS LABORATORY STUDIES ON THE FLUORESCENCE OF NATURAL WATERS
The existence of the blue, UV-induced fluorescence emission from natural,
drinking and polluted waters has been known for many years (Dienert, 1910;
Radley and Grant, 1935; DeMent and Dake, 1942). McLean and Speas (1946)
proposed fluorometric screening of well drinking water for the presence of
bacterial contamination. Work on the fluorescence of sea water has been
reviewed by Kalle (1966), Duurmsa (1974) and Jerlov (1976). This fluorescence
is shown to be restricted principally to coastal waters; deep ocean water is
known to have a very low organic content and to exhibit relatively weak
fluorescence emission. The exact chemical formulations of the substances
responsible for this sea water fluorescence are not well understood although
they are known to be relatively stable. Consequently, this fluorescence
emission can be used to monitor the mixing of river and coastal water with
deep ocean water. In addition, these substances are kn"own to be part of a
much larger group of sea water organics known collectively as gelbstoffe
(yellow stuff). The main sources for these organics are runoff from land
introduced principally as river water, rain water, and from in situ formation
through the activity and decay of biological matter.
Earlier work on the fluorescence of fresh waters has been reviewed by
Smart et al. (1976). Organics of natural origin enter lake and river water in
the same manner as for sea water and can generally be classified as humic,
fulvic and hymatomelanic acids. Black and Christman (1963), Seal et al.
(1964), Levesque (1972) and Hall and Lee (1974) have investigated the
fluorescence spectra of humic compounds present in soil and colored waters and
also the influence that pH has on the wavelength and intensity of this
emission.
A quantitative fluorescence technique for monitoring dissolved waterborne
organics has been developed and is being used for gauging the adsorption
efficiency and lifetime of activated carbon columns used for removing
dissolved organics in drinking water treatment plants (Sylvia, 1973; Sylvia et
al., 1974; 1977; Montalvo and Lee,"1976). This technique, call~d the rapid
fluorometric method (RFM), is based on the existence of a strong correlation
between the fluorescence emission intensity excited in drinking water samples
from a single source and the weight of dissolved organics adsorbed from the
same sample water by a granulated activated carbon (GAC) sample; these trapped
organics are dissolved from the activated carbon using chloroform and the
resultant carbon chloroform extract (CCE) determined gravimetrically.
Alth,ough about six man hours of time are required for each GAC and CCE
determination, the whole process can take at least eight days to perfonm
(Buelowet al. 1973), whereas the RFM measurement can be made in a few
9
-------
I-
I
minutes." It should be noted that, because of the lengthy sample preparation
time, and because the adsorption and extraction processes are not 100 percent
efficient (Stevens and Symons, 1974), the GAC/CCE measurement has not found
favor as an indicator of waterborne dissolved organics. "
A number of studies have .been published on methods for determining the
concentration of pulp and paper mill effluents by performing fluorescence
analyses on either grab or in situ water samples. These effluents contain
high concentrations of dissolved organic substances, many of which are
considered hazardous to the aquatic environment due to their high toxicity and
" tendency to produce conditions of high oxygen demand (Walden, 1976). .
Christman and Minear (1967), Thruston (1970), Wilander, Kvarnas and Lindell
(1974) and Almgren et ale (1975) have established the fluorescence properties
for lignosulphonate effluents from sulfite process mills. Likewise,
Baumgartner etal. (1971) have established the fluorescence properties for
kraft mill effluents. For specific locations it has been shown that the
intensity of the UV-induced fluorescence emission at a specific wavelength is
hi gh ly correlated wi th the effl uent concent rat ion, thereby. provi di ng the basi s
for a rapid in situ tracer method for determining the concentration of these
effluents. Fluorescence emission from humic substances present in these same
waters was found to occur at longer wavelengths, and as such, did not
constitute a significant source of interference (Almgren et al., 1975).
Airborne laser fluorosensors have been successfully used to monitor the
presence and extent of lignosulphonate effluents in surf~ce waters (OINeil et
al., 1975; Bristow, 1978). In addition, it was shown that the intensity of
the fluorescence emission from the organics in the relatively unpolluted
waters upstream from these effluent outfalls could be detected with good
sensitivity using existing technology (Bristow, 1978).
Fluorescence techniques have also been applied to in situ monitoring "of
the dispersal of septic tank leachates in near-shore surface waters. Kerfoot
and Brainard (1978) measured the concentration of leachates using the ratio of
fluorescence to ionic conductivity as the pollutant indicator. System
calibration was achieved by measuring this ratio for a leachate sample of
known concentration~ The fluorescence signal was considered to be due to the
presence of fluorescent detergent whiteners, surfactants and natural
degradation products that are persistent under conditions of darkness and low
dissolved oxygen.
In all of the above water fluorescence studies, no attempts were made to
relate this UV-induced fluorescence emission to established water quality
parameters indicative of organic carbon, such as BOD, COD, TOC and DOC.
However, more recently the results of a number of studies have become
available in which data relating the UV-induced fluorescence emission of water
samples to either TOC or DOC are presented.
Zimmerman and Bandy (1975) and Huro et al. (1977) examined the
relationships between fluorescence and DOC and between fluorescence and TOC,
respectively, for marine waters, whereas Lakshman (1975), Measures et al.
(1975) and Smart et al. (1976) compared sample fluorescence to TOC for a
variety of different fresh water types.
10
-------
The measurements of Symons et ale (1975) were somewhat different and, as
such, nonrepresentative in that they examined drinking water samples from 80
cities in the United States rather than natural waters. In addition, they
were careful to note that their measurements encompassed only the nonvolatile
fraction of TOC, as the generally small volatile fraction was driven qff with
the inorganic C02 when purging the acidified samples. As no mention is made
in the other studies cited above of preparations to monitor the organic carbon
level of the headspace above the sealed samples, it can be assumed that the
reported TOC and DOC measurements were also made on the nonvolatile fraction
of the samples.
The correlation data for these studies are summarized in Table 1 together
with information on the emission and excitation wavelengths employed in the
fluorescence analyses. Also shown are the data of Sylvia et al. (1974, 1977)
in which sample fluorescence is compared to the CCE accum~lated in a GAC
column in a drinking water treatment plant. Reasons why the correlation
coefficients obtained in these studies vary over a range from 0.46 to 0.99 for
the comparisons between the fluorescence and organic carbon data are not
immediately apparent. The most likely explanation is that the ratio of
fluorescent to nonfluorescent organics is not a constant, but varies due to
changes in the concentration and nature of the fluorescent organics that are
unrelated to the changes in the TOC and DOC concentrations.
It is notable that in the studies by Zimmerman and Bandy (1975) and by
Measures et al. (1975), which are aimed at an airborne laser fluorosensor
application for monitoring waterborne organics, the fluorescence emission data
were normalized by means of the concurrent water Raman emission signal. This
normalization procedure corrects for the attenuation of the fluorescence
signal caused by the presence of both dissolved and particulate matter present
in the sample. .
The present laboratory study was therefore initiated with the purpose of
conducting an investigation of the relationships between the fluorescence
characteristics of surface waters and the widely accepted organic carbon water
quality parameters TOC and DOC, and of establishing the possibility of using
these characteristics as a means for remotely monitoring total organics from;
an airborne platform. In particular, attention will be paid to finding the'
best water fluorescence parameter for this purpose and to investigating the
merits of the Raman normalization procedure as a means of correcting both the
laboratory and airborne measured fluorescence data for the effects of optical
attenuation. -
This study is presented in four parts. The first part (Chapter 6) deals
with the effects of optical attenuation on the fluorescence data, and with the
Raman normalization procedure. The second part (Chapter 7) describes the
experimental methodology with some emphasis being placed on methods for
storing and preparing water samples for fluorescence analysis. The third-part
(Chapter 8) examines the correlations obtained between water quality
parameters that are a measure of organic carbon and various fluorescence
emission parameters. The fourth part (Chapter 9) is an investigation of
problems that are encountered when attempting to apply fluorescence-DOC
regression data established in this or similar laboratory studies to the
interpretation of fluorescence data obtained using an airborne laser
fluorosensor in terms of surface water DOC.
11
-------
TABLE 1. SUMfrtARY OF PUBLISHED FLUORESCENCE/ORGANIC CARBON CORRELATION DATA
-------- ----
---------
Fluorescence Parameters
Organic Excl taUon Eml ss I on Number II nca I'
Carbon Wavelength . Wavelength Raman Sample of Correlation
Reference Parameter (nlO) (1110) Corrected Source Sanip I es Coefficient
---
Sylv1~ et al., 1914 Carbon 340 390 "0 From granulated activated carbon 52 0.96
Chlorofonn column In drinking water treatment plant
htract
Sylvia et al.. 1911 Carbon 340 390 "0 Fr~n granulated activated carbon 80 0.93
Ch 1 orofonn column in drinking water treatment plant
Extract
Zimmennan and Dandy, DOC 331 Max. Yes Fresh and marine estuarine waters 39 0.11*
1915
lakstJnan, 1915 TOC 365 460 "0 Agr1cultura' runoff and water extracts 9 0.98
from manure
to-"
N Symons et a1., 1915 Non 340 390 No Drinking water fr~n 80 cities In USA 13 0.69*
Volatile
TOC
Symons et a'., 1975 Non 310 Tota' "0 Drinking water fr~n 80 cities In USA 80 0.66*
Volatile Emission
TOC
Measures et a'., JOC 331 Max. Yes Creek and river water 1 0.13
1915
Smart et al., 1916 TOC 365 510 No Miscellaneous 24 0.98
~nart et al., 1916 TOC 365 510 No F reewayrunof f 11 0.46
Smart et aI., 1916 TOC ]65 510 No Creck water 34 0.87
lIuro et a'., 1911 TOC 340 450 "0 Coas ta 1 martne water 15 0.99
-------------
* Estimated by present authors fran graphical data
-------
SECTION 6
EFFECTS OF OP nCAL ATTENUATION ON WATER FLUORESCENCE MEASUREMENTS
Airborne laser fluorosensors use high power, pulsed, blue or ultravfolet
(UV) lasers to excite fluorescence emission in a sample volume in the water
surface. A fraction of this multidirectional emission is collected by a large
aperture telescope and converted into an electrical signal by an optical
detector. This fluorescence signal, in conjunction with aircraft navigation
data, can be used to prepare contour maps showing the variation in
concentration of the specific water quality parameter under investigation.
The principle of operation for the laser fluorosensor is illustrated in Figure
1 in which the airborne platform is usually flown at a height of several
hundred meters above the water surface. The volume of water interrogated by
the system is approximately defined by the diameter of the laser excitation
spot on the water surface and the penetration depth of the laser beam. For a
homogeneous vertical distribution of dissolved and particulate matter in the
sqmple column, this depth can be represented by the characteristic optical
attenuation length. Assuming that the attenuation lengths at the laser and
fluorescence wavelengths are approximately equal, then it can be shown that
(1-1/e2), i.e. 86.51" of the fluorescence radiation received by an airborne
laser fluorosensor is emitted from the first attenuation length (e.g. see
Friedman and Hickman, 1972).
The spectral nature of this laser-induced fluorescence emission from
typical surface waters can be readily demonstrated using a proprietary.
laboratory spectrofluorometer. The two emission spectra shown in ,Figure 2
were obtained on high purity and lake water samples using a laboratory
spectrofluorometer (Perkin Elmer MPF-4), which has a corrected spectra
attachment that eliminates spectral artifacts that would otherwise be
introduced by the emission and excitation monochromators, the xenon excitation
lamp and the photomultiplier detector. Excitation at 337 nm was employed as
it lies close to the wavelength producing the maximum fluorescence emission in
fresh water samples, and also because it corresponds to that of the widely
available pulsed nitrogen laser. Tnis type of laser has already been used in
an airborne laser fluorosensor system capable of monitoring the fluorescence
of natural waters with good sensitivity (Bristow, 1978), albeit in the absence
of a solar background signal.
These ~'ater fl uorescence spectra exhi bit two features of interest.
First, the lake water sample exhibits a broad blue fluorescence band peaked in
the region of 430 nm whereas the spectrum for the high purity sample is
essentially fluorescence free. Treatment of water samples by passage through
an activated charcoal trap to remove organics also eliminates essentially all
of this fluorescence signal (Sylvia, 1973; Montalvo and Lee, 1976), suggesting
13
-------
Telescope
Field of View
Fluorescence and Raman
Emission from Surface
Water Volume
Fi gure 1.
Principle
of operation
of ai rborne
laser fluorosensor.
14
-------
~
-
U)
z
"'"
~
z
-
"'"
>
j:
c
....
"'"
=
350
Figure 2.
OH STRETCH RAMAN EMISSION BAND
/ OF WATER AT 381 nm
EXCITATION- AT 337 nm WITH 3 nm
BANDWII)TH
EMISSION- SCANNED FROM 350 nm
TO 500nm WITH 3 nm
BANDWIDTH
SURFACE WATER SAMPLE FROM
LAKE MEAD, NEVADA
/
ULTRA PURE WATER SAMPLE
./
380
410
440 470 500
EMISSION WAVELENGTH (nm)
Corrected fluorescence emission spectra for lake
and ultra-pure water samples.
15
-------
that fluorescent inorganics make a negligible contribution to the fluorescence
of natural waters. The sample obtained from Lake Mead in Nevada, is known to
be low in organics having a TOe level on the order of 1 mg/l. This suggests
that a fluorescence signal that can be measured with high sensitivity from
samples low in TOe might be usable as an indicator of surface water.organics
provided that the fluorescence and organic carbon parameters can be related in
a meaningful way.
The other spectral feature, an intense, relatively narrow, constant
amplitude band located at 381 nm and superimposed on the fluorescence
spectrum,. is the Raman emission for the OH vibrational stretching mode of
water. Unlike fluorescence emission, (Stokes) Raman emission is always
displaced by a constant frequency shift to the longwave side of the excitation
wavelength. For water in the liquid phase, this frequency shift is 3,418
cm-l, which, for excitation at 337 nm, corresponds to Raman emission at 381
nm. Although the intensity and spectral shape of the water Raman band are
known to be weakly dependent on water temperature and salinity (Walrafen,
1967; Chang and Young, 1974), changes that can be anticipated in these
parameters for fresh water environments are not expected to affect the
integrated emission inte~sity for the whole Raman band. .
For high purity samples or those relatively low in dissolved and
particulate matter, e.g., most drinking waters, and for a constant excitation
intensity, the amplitude of the Raman signal emitted from a short (1 cm)
pathlength sample remains essentially constant. Stated another way,
self-absorption of the fluorescence and Raman radiation by the sample ;s
negligible over a l-cm pathlength. However, in the presence of more
significant concentrations of dissolved and particulate matter, the intensity
of this Raman signal will be attenuated by absorption and scattering losses.
In the case of the fluorescenc~ signal, changes in intensity may be due either
to this optical attenuation or to changes in concentration of the fluorescent
substances under investigation thereby introducing ambiguities into the
interpretation of the fluorescence data. This attenuation can be significant
in:thecase of samples high in dissolved organics with the result that the
attenuated fluorescence signal will indicate a lower value of TOe than is
actually present.
As will be shown, significant attenuation of the fluorescence emission
due to self-absorption can occur in 1-cm thick laboratory samples that contain
high levels of organics. In an airborne laser fluorosensor application, the
effective volume of sample being interrogated will depend critically. on the
penetration depth of the laser beam through the surface water. As this depth
or attenuation length may vary over a range from, say, 0.1 to 1Q meters or
more for a given water body, very large errors can be incurred in the measured
fluorescence signal. Consequently, as the Raman signal is a property of water
alone, and as the concentration of water is constant for all but grossly
polluted waters, observed variations in its intensity will be due to
variations in the optical attenuation coefficients at the laser and Raman
wavelengths. The Raman signal can therefore be used as an internal reference
standard with which to monitor the effect of changes in these attenuation
coefficients on the concurrent fluorescence signal.
16
-------
Although self evident, it should be emphasized that the water Raman
signal used to monitor changes in sample optical attenuation is emitted from
essentially the same sample volume and at the same time as th~ fluorescence
signal. This is an important consideration in a remote sensing application
when it is realized that marked changes in the attenuation properties of
surface waters can occur over distances of a few meters.
RAMAN CORRECTION OF AIRBORNE FLUORESCENCE MEASUREMENTS
By taking the ratio of the remotely sensed fluorescence emission
intensity or power level to the corresponding Raman value, a parameter is
obtajned that, to a first approximation, varies only as the concentration of
the homogeneously distributed fluorophor (fluorescent substance) and is
independent of changes in the optical attenuation coefficents at the laser
(excitation), fluorescence and Raman wavelengths. Use of this Raman
correction technique has been proposed and successfully demonstrated by .
Bristowet al. (1979) in a similar airborne remote sensing application, which
involved monitoring the concentration of surface water chlorophyll a. It was
shown .that the remotely sensed fluorescence power at any wave1ength-A can be
given by a general expression of the form: .
L p ) nf a f d A
FA = ~ . (kA + kL)
(1)
where P is the power of the laser excitation source at wavelength AL, H is
the fluorescent target to collector lens distance, nf is the concentration
of the fluorophor under investigation, af is the fluorescence conversion
efficiency (excitation cross section) of the f1uorophor, dA isa constant
accounting for known or measurable environmental and system factors, and k;\
and kL are the effective optical attenuation coefficients at the
fluorescence and excitation wavelengths respectively. A number of assumptions
are made in the determination of this expression and these are discussed by
Browell (1977) and Bristow et al. (1979). In the present context, three of
these assumptions are noteworthy. First, FA is the emitted fluorescence
. power integrated over a sample of infinite depth or thickness. In reality,
this means that the total water depth must be significantly larger than the
optical attenuation lengths given by (k4)-1 and (kL)-l. Second, ata .
given sampling location, the concentratlons of all dissolved and particulate
waterborne substances whether fluorescent or nonfluorescent that influence the
values of kA and kL are assumed to remain constant in the vertical
dimension such that kA and kL also remain constant. Third, the
integration of the fluorescence power over an infinitely thick sample requires
that of remain constant over this sample volume. However, for Equation 1 to
be valid in a given remote sensing application, of must also remain constant.
for all sample locations. In the present application, the validity of this
assumption will depend critically on the constancy of the percentage
contribution of each fluorophor to the overall mixture of f1uorophors over the
extent of a given water body. This subject will be examined further in the
di scussion.-
17
-------
A similar expression can be written for the remotely sensed water Raman
power, R, measured at wavelength, AR, such that:
R - C: ) ~:R a: ::J
(2)
where nw is the concentration of water, which for all natural waters is a
constant, Ow is the Raman excitation cross-section for the [OH]-vibrationa1
stretching mode of water, dR is a constant similar to d)., and kR is the
effective optical attenuation coefficient at the Raman emission wavelength.
Using Equations 1 and 2, an expression for the power ratio FAIR is obtained
where:
FA
R
=
nf af dA
OW Ow dR
(kL. + kR)
kL + k).
(3 )
As of, d)., nw, Ow and dR are all considered to be constant, this
expression can be simplified to read:
FA
R
=
(kL + kR)
nf 01
kL + kA
(4)
where 01 is a new constant encompassing all of the above mentioned factors,
and (kL + kR)/(kL + kA) is a factor representing residual attenuation
effects due to differential spectral absorption and scattering. The
expression 'differential spectral absorption and .scattering' refers to the
changes in the magnitude of these optical loss mechanisms that exist between
any two (or more) wavelengths, and by implication, indicates that they are
wavelength dependent. Clearly, for this expression to be useful in the
interpretation of remote sensing data, (kL + kR)/(kL + kA) must remain
constant or, at the most, change only slowly when significant changes occur in
the individual k values. Preliminary airborne laser fluorosensor measurements
(Bristowet al., 1979; Bristow, 1980*) have shown that FAIR provides a
better indication of changes in chlorophyll a concentration than does FA,
suggesting that, to a first approximation, (fL + kR)/(kL + kA) can be
treated as a constant. Equation (4) can then be simplified to read: .
FA
R
=
nf C
(5 )
* Bristow, 1980, U.S. Environmental Protection Agency, unpublished data.
18
-------
where C is an unknown constant that can best be determined from Equation 5
using a number of airborne measurements of FAIR together with concurrent
ground truth measurements of nf obtained on samples from a series of
preselected reference sites under the sensor flightpath. In this situation,
nf will correspond to the concentration of organic fluorophors, which is
c osely related to total organics as represented by TOe (or DOC).
With the proposed UV laser excitation at 337 nm, Raman emission at 381 nm
and fluorescence emission at (say) 430 nm, significant differences in the
magnitude and behavior of the corresponding k values for similar water types
can be expected (Jerlov, 1976). However, all concern over the validity of the
assumption that (kL + kR)/(kL + kA) remains constant can be eliminated
by making both the Raman and fluorescence measurements at the same wavelength;
in this case, (kL + kR)/(kL + kA) becomes equal to unity.
In applications to remotely monitor the concentration of total organics,
three significant advantages are to be gained by normalizing the fluorescence
signal with the concurrent water Raman signal:
(i 1)
(1)
Raman normalization corrects for the attenuation of the fluorescence
signal due to the presence of either inorganic or organic
particulate and dissolved matter. For airborne applications over
deep waters, the effective sample length is determined by the
optical attenuation length, which may vary over a range of 100 to 1
or more. Correspondingly large corrections to ~he fluorosensor data
are then required. In addition, as the factor (kL + kR)/(kL +
kA) is assumed to be a constant, no information is required
concerning the analytical model needed to describe the effective
optical attenuation coefficients, kA. Some controversey exists
over whether the beam attenuation coefficient, the diffuse
attenuation coefficient, or some intermediate form is most
applicable to the laser fluorosensor situation (Browell, 1977).
Raman normalization of the fluorescence data eliminates problems due
to both short-term (within mission) and long-term changes in system
sensitivity that have an equal influence on both the fluorescence
and Raman signals. Such changes include but are not restricted to
pulse-to-pulse fluctuations and long term drift in laser output
power, line voltage changes and amplifier gain drift. .
(iii) As both the fluorescence and Raman signals exhibit a 1/H2
dependence on changes in aircraft altitude, H, above the water
surface target, the ratio FAIR is independent of these changes.
It should be mentioned that (i) and (ii) apply equally to remote sensing,
in situ or laboratory applications for measuring water fluorescence, although
for (i), attenuation effects in 1-cmpathlength samples will be relatively
small requiring corrections to the fluorescence signal no greater than 50
percent.
Although not related to the present airborne laser fluorosensor
application it is suggested that the measured values of R can also be used to
19
-------
measure kA which in this case is the mean of kR and kL appearing in
Equation 2. Both? and H are measurable, and nw, wand dR are constant
so that kA varies as R-~, where kA = (kL + kR)/2 is equivalent to .
the effective attenuation coefficient for some wavelength intermediate between
AL and AR. The proportionality constant relating k>.. to R-l can be
determined by making a number of in situ measurements of kR and kL
together with corresponding airborne measurements of R at the same reference
locations as proposed for establishing constant C in Equation 5.
Implementation of this proposal raises the question as to whether kR and
kL are better described by the beam attenuation coefficient or by the.
diffuse attenuation coefficient. By arranging that an airborne mission
encounter a sufficiently wide range of kA values, it may be possible to
determine which of these two coefficients is a more suitable model of optical
attenuation in a laser fluorosensor application.
RAMAN CORRECTION OF LABORATORY FLUORESCENCE MEASUREMENTS
The laboratory measured values of FAIR are subject to the same
secondary effects of differential spectral absorption and scattering as are
the airborne measured values, but, because of the 90° scattering angle
employed in laboratory spectrofluorometers of the type used in this study,
Equations 1 through 5 are not applicable. In such an instrument, the
fluorescence emission, excited in a small focal volume of sample, is viewed at
90° to the excitation beam in contrast to the 0° scattering angle employed in
a remote sensing configuration. Consequently, the emission emanates from a
single point at a fixed depth below the surface of an optica11y thin sample
(typically in a I-em square sample cell) rather than being integrated over an
infinitely deep sample volume. For emission from a point source at a fixed
depth, it is easily shown .(e.g. see Friedman and Hickman, 1972) that FAIR
can be given by an expression of the form:
FAIR =
nf 02 exp {h (kR - kA)}
(6)
where 62 is a constant analogous to 61 in Equation 5, and h i~ the sample
depth, which, in this case, is the distance from the inner surface of the
sample cell wall to the focal volume located at the center of the cell. In
the same manner as for the remote sensing application, differential spectral
attenuation effects, represented by the exponential term in Equation 6, can be
eliminated by making the fluorescence measurements at the Raman band
wavelength. In this case, the exponential term becomes equal to unity, and
FR/R = nf 62, which is analogous to Equation 5.
On the basis of the foregoing discussion, the fluorescence to Raman
ratio, FAIR, would appear to be a~ ideal candidate parameter for remotely
characterizing the fluorescence emission of surface waters due to its relative
independence from environmental and system factors. The merits of this and
other dimensionless fluorescence parameters for use as remote sensing
indicators of TOC (or DOC) will therefore be examined in some detail in
Chapter 8.
20
-------
SEcn ON 7
SAMPLE COLLECTION, ANALYSIS AND PRESERVATION
SAMPLE SOURCES
A total of 161 water samples was collected for analysis. In order to
avoid surface debris and oil films but, at the same time, obtain samples
representative 'of the water surface, I-liter samples were obtained at a depth
of 1 m. This master sample was then stirred, and split into three nominally
identical subsamp1es for the TOC, DOC and fluorescence analyses, which were
stored in 100-m1 polyethylene bottles.
Fifty samples, relatively low in organics, were obtained from Lake Mead
and Lake Mohave, both manmade reservoirs on the Colorado River bordering
Arizona and Nevada that can 'vary between oligotrophic and mesotrophic
conditions on a seasonal basis. Although the bulk of the water is derived
from the Colorado River, Lake Mead is also supplied by the Muddy River, the
Virgin River and Las Vegas Wash. The latter source, a nutrient-rich stream
contaminated with industrial and municipal sewage wastewatar from the Las
Vegas region, enters Lake Mead through Las Vegas Bay, the receiving end of
which exhibits eutrophic conditions for a large part of the year. Even though
the flow rate for Las Vegas Wash is only 0.5 percent of that for the Colorado
River, it is the principal source of the pollutants in Lake i4ead, and is .
especially rich in nitrogen and phosphorus. Lake Mohave, further down the
Colorado River, receives all of its water from Lake Mead.
The largest set of samples, totaling 107 and ranging widely in organic
content, was obtained from the Atchafalaya River Basin which is a large
shallow depression located within the deltaic plain of the Mississippi River
in southern Louisiana. Approximately 30 percent of the total flow of the
Mississippi River enters the Basin through the Old River Control Structure
where it joins with the flow from the Red River to form the Atchafa1aya River.
A major feature of the region is the Atchafalaya Basin Floodway, which is
defined by a system of levees used to contain periodic inundation by the
Mississippi River. The majority of samples were obtained from sites within
the levees and were often highy colored, containing relatively high levels of
particulate and dissolved organic matter. It is .these conditions that are
responsible for the detritus-based food chain of the waters in the floodway.
In addition the limited light penetration into these turbid waters has a
restraining influence on phytoplankton growth with chlorophyll a levels
generally less than 10 ~g/l. In contrast, 20 samples collected-from
swamplands outside the levees were generally characterized by lower levels of
particulate and dissolved organic matter but by higher levels of chlorophyll
~. The Mississippi River also introduces significant levels of manmade
21
-------
organics into the floodwaters, some of which are not readily biodegradable.
Pesticides and runoff from agricultural activities in and adjacent to the
Basin have been observed to have detrimental effects on fish populations, and
as such, also make contributions to the overall level of organics. Although
there are over 100 oil- and gas-producing wells within this basin, there is
little evidence to suggest that they constitute a significant pollution threat
at the present time.
In addition, four drinking and high purity water samples were included in
the survey with a view to extending the range of measured values.
ORGANIC CARBON ANALYSES
The samples for TOC and DOC analysis were acidified to pH i2 by the
addition of HCl in order both to drive off inorganic carbon in the form of
C02 and to attenuate bacterial activity, and then, prior to analysis, were
stored in the dark at 4°C. All TOC and DOC analyses were made with the TOC
analyzer (Envirotech Dohrmann DC~52) operating in the low level, high
sensitivity sparge mode using 10 ~l-sample injections. Reproducibility of the
measurements was established by making 30 replicate measurements on a Lake
Mead sample relatively low in organic carbon. The mean values were 2.86 I
0.10 mg/l for TOC and 0.91 I 0.05 mg/l for DOC, where the uncertainty
represents the 95 percent confidence limits for the means. The DOC values
were obtained in the same manner as for the TOC values except that the
particulate matter was removed prior to analysis by passing the samples
through Whatman GF/F glass fiber filters. For particles 0.7 ~m or larger in
size, it is claimed that these filters have a 98 percent retention efficency
(Aversa, 1976). Glass fiber filters were selected because they are known to
contain very low concentrations of organic carbon (Parker, 1967; Sharp, 1974),
and because cellulosic membrane filters, such as the Millipore HA, 0.45-~m
pore size filter that is frequently used for separating particulate and
soluble materials in water quality work, have been shown to introduce
significant amounts of organic carbon into the filtrate (Guillard and
Wangersky, 1958; Cahn, 1967; 'Parker, 1967; Hwang et ale 1979). Comparable
silver membrane filters (Selas, FlotronicsDivision, Huntingdon Valley, Pa.)
were rejected in relation to glass fiber filters because of their higher cost,
higher carbon content (Gordon and Sutcliffe, 1974; Sharp, 1974), low
filtration speed (Parker, 1967) and unpredictable load capacity (Sheldon,
1972). In addition, Salonen (1979) has recently shown that silver membrane
filters have a loading capacity that is several times lower than that for
glass fiber filters with comparable retention efficiency. The POC data were
obtained by subtracting the measured DOC values from the corresponding TOC
values.
Sparging or purging the samples prior to analysis with an inert gas
drives off any residual C02 produced during acidification, but, in the
process, also drives off volatile organic substances. The TOC and DOC values
measured in this work should therefore be understood to represent the
less-volatile fraction of their respective total values. Generally, for
surface waters, volatile organics constitute only a small percentage of the
total organics present (MacKinnon, 1979). In addition, the loss of volatiles
22
-------
during sparging should not significantly effect the correlation with the
fluorescence data» as these volatile substances are generally non-aromatic
having low molecular weight» and, as such, are unlikely to exhibit significant
fl uoresc ence.
DOC and TOC blanks subsequently run on high purity water samples stored
in glass containers as against polyethylene containers» with and without
acidification, indicated a carbon background for the acid of about 0.4 mg/l.
This acid background value was therefore subtracted from all measured TOC and
DOC values. Similar tests to determine the background contribution from the
polyethylene sample bottles indicated an approximate increase in carbon
background on the order of 0.05 mg per day but with a large bottle-to-bottle
variability. However» as the samples were usually analyzed within a five-day
period, corrections for the polyethylene background were not made» partly
because of its generally small value and partly because of the uncertainty in
establishing a precise value. As such, this sample-bottle background level
constitutes a potential source of er~or in the TOC and DOC data.
For the 158 samples analyzed,TOC values ranged from 0.2 to 44 mg/l with
a mean value of 7.21 mg/l, DOC values ranged from 0.2 to 19 mg/l with a mean
value of 4.02 mg/l and POC values ranged from 0 to 27 mg/l with a mean value
of 3.47 mg/l. Although the mean value for DOC is marginally greater than that
for p~C, many samples had POC values considerably larger than the
corresponding DOC values, and for the 50 samples from Lakes Mead and Mohave»
the mean value for POC (1.48 mg/l) was 40 percent high~r than the mean value
for DOC (1.05 mg/l). .
FLUORESCENCE ANALYSES
As the ultimate purpose of this feasibility study is to establish a
method for remotely detecting the concentration of total organics for in situ
surface water samples using fluorometric techniques, every effort was made to
ensure that the grab sampl es for fl uorescence anal ysi s remai ned as .
representative of the true field conditions as was practicably possible.
Clearly» remote sensing operations do not allow for any form of sample
preparation or conditioning. A number of measures were therefore taken in
order to ensure that the integrity of the fluorescence properties of the grab
samples was maintained during the period between collection and analysis.
Procedures that minimize the changes in sample fluorescence properties with
time are discussed in the following section. In at least" one case» viz., that
of temperature, replication of field conditions was not possible. This was of
concern because it has been shown that the fluorescence emission is
t~nperature dependent (Smart et al., 1976). The effects of temperature on the
remote sensing of fluorescence emission is also discussed in the following
section. For the purposes of the laboratory correlation study» the
fluorescence measurements were made at a temperature of 21° tloC.
A 11 fl uorometric measurements were made usi ng a propri etary
spectrofluorometer (Perkin Elmer MPF4) capable of producing fluorescence
spectra on a relative quanta scale that have been corrected for the spectral
artifacts introduced by the xenon excitation lamp, the excitation and emission
23
-------
grating'monochromators and the photomultiplier detector (Porro et al., 1973).
The optical layout for this instrument is shown in Figure 3. Corrected
spectra can be made over the wavelength range from 290 nm to 704 nm.
Additional features are the high sensitivity cell holder that has back
reflectors providing a four-fold signal increase, high-gain red-sensitive
emission and reference photomultipliers (Hamamatsu R928), and a Cation-X
short-wavelength cutoff filter. This filter is used to prevent the 337-nm
excitation radiation that is scattered by sample particulates from reaching
the emission detector, and' interfering with the measurements in the 350-nm to
380-nm region. The Cation-X liquid filter consists of a 2-gm/1 aqueous
solution of 2,7-dimethyl-3,6-diazacyclohepta-l,6-diene iodide (Eastern
Chemical Company, Hauppauge, N.V.). The transmission curve for the 2-1/2-mm
thick liquid filter, held between fluorescence-free, fused quartz plates, is
shown in Figure 4 and is essentiall,)l flat for the working region above 360
nm. In addition, this liquid filter appears to be essentially non-fluorescent
in contrast to many proprietary colored glass short wavelength cutoff filters
(Bristow, 1979).
Water fluorescence emission spectra, as typified by those shown in Figure
2, were obtained by exciting the sample at 337 nm and scanning the emission
spectrum from 350 nm to 500 nm. A fixed excitation wavelength of 337 nm was
employed throughout this study partly because this wavelength generally lies
close to that producing the maximum fluorescence emission from fresh water
samples, and partly because of the wide availability of the reliable pulsed
N2 laser, which operates at this wavelength and which is considered suitable
for remote sensing laser fluorosensor applications. Use of an alternate
wavelength such as that corresponding to the pulsed XeF excimer laser at 351
nm should not make any significant changes to the conclusions obtained in this
study. Both excitation and emission monochromators were used with spectral
slit widths adjusted to an equivalent bandwidth of 3 nm in order to optimize
the resolution of the OH-stretch water Raman band at 381 nm, also shown in
Figure 2. A fluorescence emission spectrum was produced for each water
sample, and. the six fluorescence characteristics, indicated in Figure 5 were
obtained in each case. Fmax is the fluorescence emission intensity at the
peak of the spectrum and generally lies in the 41S-nm to 440-nm region. R is
the peak intensity of the Raman band at 381 nm and FR is the corresponding
fluorescence emission intensity at this wavelength. The Rand FR components
of the total emission intensity at 381 nm are separated by interpolation,
using a third-order polynomial curve-fitting procedure, which utilizes the
fluorescence intensity values at 362.5 nm, 370 nm, 397.5 nm and 407.5 nm. The
. other fluorescence spectrum characteristics are aw, the full bandwidth at half
height (FWHM), Amax, the wavelength at the peak intensity, and ~cen' the
wavelength at the center of the fluorescence band as determined from the
midpoint of BW. Due to the broad and relatively flat nature of these spectra,
the exact value of Amax was often difficult to establish, and as a result,
~cen was preferred as the characteristic wavelength. In addition, FA, the
fluorescence intensity at wavelength ~, was measured every 10 nm from 350 nm
to 500 nm. From these latter data, it was possible to establish Fav, the
mean fluorescence intensity for the spectrum between 350 nm and 500 nm, which
is a measure of the total or integrated emission between these wavelength
limits.
24
-------
. EMISSION
MONOCHROMATOR
.N
U'I
EXCITATION
MONOCHROMATOR
RHODAMINE B
QUANTUM
COUNTER
MONITOR
PHOTO
MULTI
PLiER
(R 928)
GRATING
--
--
SAMPLE
PHOTOMULTIPLIER
. (R 928)
CATION X
FILTER
Figure 3.
GRATING
--
BEAM
SPLITTER
--
HIGH SENSITIVTV
SAMPLE COMPARTMENT
Optical diagram of corrected-spectra spectrofluorometer (Perkin-Elmer MPF4).
-------
100
90
80
10
60
50
40
30
20
10
o
340
TRANSMISSION
. (%)
.
/
---- .
/0
.
.
.
ABSORBER.
CATION X
IN WATER
CONCENTRATION -1 gm/l
FILTER PATH LENGTH -- 2 1/2 mm
345
Figure 4.
350
355 360
WAVELENGTH (nm)
365
370
Transmission curve for Cation-X liquid filter.
26
375
-------
N
.....
~
"en
c:
!
c:
-
Q)
>
"~
co
Q)
CC
Nitrogen laser
Excitation line Water Raman Emission
(not to scal'e) Band (OH-Stretch)
/
f
R
AL =337
FR
!
,\ R =381
Fluorescence Emission
from Dissolved Organics
FMAX
I
BW
(FWHM)
I
~
'\MAX '\CEN =435
Wavelength (nm)
Figure 5.
Schematic showing water fluorescence and Raman emission parameters obtained
from spectra produced using a laboratory spectrofluorometer.
-------
Reproducibility of the fluorescence data was established by making 10
replicate measurements of Fmax/R and FR/R on the same unfiltered Lake Mead
sample as was used for evaluating the reproducibility of the TOe and DOC
measurements. The means were 0.68 t 0.01 for Fmax/R and 0.48 t 0.01 for
FR/R, where the uncertainty represents the 95 percent confidence limits for
the means.
PRESERVATION AND PREPARATION OF FLUORESCENCE SAMPLES
The time taken for each fluorometric measurement was on the order of 20
minutes, so that a complete fluorescence analysis of the 161 samples takes at
least 7 days. In addition, a 1-to-2 day delay was generally incurred before
the fluorescence analysis could be started due to the time taken to collect
. and ship the ice-chilled samples from the field site in Louisiana to the
laboratory in Las Vegas, Nevada. Questions were therefore raised concerning
sample contamination and integrity during this shipping and storage period,
and the effects that any such changes might have on the fluorescence data.
Sample contamination by the polyethylene bottles and the possible effects of
continued bacterial activity, oxidation, photodecomposition and colloidal
aggregate formation were of principal concern. Tests were therefore conducted
with the purpose of establishing a sample preservation and storage procedure
. that minimizes changes to the sample fluorescence properties.
Contamination of the samples by hydrocarbons leached from the
polyethylene storage bottles is known to influence the UV absorption spectra
of pure water (Delhez, 1960), and, in the present study, has been shown to
contribute a carbon background on the order of 0.05 mg/day. The effect of the
bottles on the fluorescence measurements was investigated by examining the
fluorescence of an ultrapure water sample stored in an unused 100-ml
polyethylene bottle each day over a 4-day period. Froax/R remained close to
zero, suggesting that the organics that are leached from the polyethylene
bottles over this period are essentially nonfluorescent.
. .
In order to )nvestigate the effects of storage and preparation on sample
integrity, tests. were performed on identical subsamples of a single master
sample from the Atchafalaya River Basin that were .stored, preserved and
prepared accordfng to seven different procedures over a 14-day period. In all
of these procedures, the fraction of subsample that was drawn off each day was
brought up to the laboratory ambient temperature (21 tlOC) just prior to the
fluorescence analysis. This was done because Smart et al. (1976) have shown
that the fl u-orescence of natural waters. is moderately sensiti ve to water
temperature. The seven sample preservation and storage procedures are as
follows:
(i)
Subsample stored in the dark at 4°C and vigorously shaken by
hand each day just prior to drawing off fraction for
fluorescence analysis.
(i i)
Subsample kept at laboratory temperature and under fluorescent
lighting during normal working hours, and vigorously shaken by
hand each day just prior to drawing off fraction for
fluorescence analysis.
28
-------
(iii)
Same as (i), but subsample not shaken.
(vi) Same as (i), but subsample agitated with an ultrasonic (20-
kHz) probe just prior to drawing off fraction for
fluorescence analysis rather than hand shaken. .
(v) Same as (i) but, in addition, subsample initially acidified
with HCl to pH~2.
(vi) Same as (i) but, in addition, subsample heated to 130°F for 1
hour just prior to drawing off fraction for fluorescence
analysis.
(vii) Same as (i) but, in addition, the fractl0n drawn off for
fluorescence analysis was passed through a 0.3 ~m pore-size
cellulosic membrane filter '(Millipore MF-PH).
The results of these tests are summarized graphically in Figure 6 as
plots of Fmax/R against time. Procedure (iv), for which the samples were
stored in the dark at 4°C and agitated with an ultrasonic probe prior to
analysis, exhibited the least variability and drift over the 4-day measurement
period. Ultrasonic agitation was limited in both duration and power level to
ensure that the sample temperature never rose above ambient by more than a few
degrees. This precaution was taken to avoid the possibility of locally
overheating the sample thereby causing thermal damage to the fluorescent
organics. .
The coefficient of variation (S/i) for Fmax/R for procedure (iv) over
the 14-day period was 2.1 percent, suggesting that no significant changes have
occurred wi th the fl uorescent organi cs for sampl es preserved and treated in
this manner. All 161 samples for fluorescence analysis that were examined in
this study were therefore stored and prepared according to procedure (iv)..
Although the data for samples treated according to (i) are not significantly
di fferent from those treated accordi ng to (i v) at the 5 percent 1 evel, (i) was
not employed because of the inherently haphazard nature of hand-shaking as a
method for sample homogenization.
It is noteworthy that samples treated according to procedure (iii) in
which the samples were stored in the dark at 4°C, but neither shaken nor
agitated during storage or preparation, exhibited a lower mean value for
. Fmax/R together with a higher coefficient of variation than did (i) or (iv).
This reduction in Fmax/R is most likely du~ to either settling-out of .
fluorescent particulates or to the formation of aggregations from colloidal
material. The exact manner by which Fmax/R is reduced is not clear as at
least three different optical mechanisms are available that will, in
principle, either reduce Fmax or increase R. First, if either the settled
particulates or the aggregated colloidal material is fluorescent, then a
reduction in Fmax and hence in Fmax/R will result. Formation of colloidal
aggregations effectively reduces the physical and fluorescence excitation
cross section of fluorescent colloidal material exposed to the excitation beam
thereby reducing the measured values of Fmax and Fmax/R below their true
or in situ values. Second, a real or apparent reduction in suspended material
29
-------
8
Fmax
R
6
w
a
5
o
/\
ff ....
.... ,
.... ///
...J.. ..,......,."..~~
.. .... -.. :.. ."
.... .... i'....... -.. : .... ..... .
- ..., 8.:. .. .
;c. .. ... . '(":::7"" ~....
.. ... ~ - , . .
~ ..'
.... ...--' , ,. .
..-... /\t\ V'\: ....""
, ji ~ ~\ ., ,.--
::.../ /1"",,----( \\ ~~ "
'" ,., \ ""'/ \ '\ .'I', f' l
'-I I \ ".../"\," ~ >l'k#'{. ........\ ,
""",, '" \ ,~. " ,---~ ,
/ \ P j' ~.}/--.
, \ . \ \,
\. \ I
\ t 'v'
\.t
,
r'/~
....
....
.....
.....
,
(it Refrig..Dark. Shaken.
CV = 2.6%. i = 6.78
----- (ii) Room Temp. and
Light. Shaken
CV = 6.2%. i = 6.15
--- (iii) Refrig.. Dark. not
Shaken.
CV = 3.8%. i = 6.57
......... (iv) Refrig.. Dark.
Ultrasonic Probe.
CV = 2.1%. x = 6.78
---- (v) Preacidified (pHS2).
Refrig.. Dark,
Shaken.
CV = 6.1%. i = 6.03
P'////- (vi) Refrig.. Dark.
130°f For 1 Hour.
Shaken.
CV = 3.1%. i = 7.01
--- (vii) Refrig.. Dark. Shaken.
0.3 lAm Filtered.
CV = 6.8%. x = 6.03
100
400
200 300
Time After Sampling (Hr.)
Figure 6. Variation of Fmax/R with time for different preservation and
preparation procedures as applied to identical subsamples.
CV is the coefficient of variation (SIx).
-------
in the excitation beam-will reduce the absorption and scattering losses that
are incurred by both the Raman and fluorescence signals. As thes~ losses are
stronger in the UV than in the visible, the reduction in the attenuation of R
will be greater than that of Fmax. This tends to increase R in relation to
Fmax, again resulting in a reduction in the measured value of Fmax/R, but
in this case towards, rather than away from, the true value. Third, an
undi spersed broadband background si gnal that is known to 1 eak from the
excitation monochromator is scattered by both the particulate and colloidal
materials into the emission monochromator. This background signal enhances
the fluorescence emission without influencing the narrow band Raman emission
and therefore acts to increase Fmax/R above its true value. In this case, a
reduction in the scattering losses caused by either particulate settling or by
aggregate formation will reduce the measured value of Fmax. Again, this
results in- a reduction in Fmax/R, and also towards its true value.
The relative contributions that each of these optical effects makes
towards reducing Fmax/R when compounded with the physical effects of
particulate settling or aggregate formation are not apparent from the data for
procedure (iii), shown in Figure 6. However, detailed examination of the
fluorescence versus DOC correlation data to be presented for the 161 samples
of this survey in Chapter 8 will provide some insight into the relative
influence that the latter two (side) effects have on the fluorescence
measurements. -
A plausible explanation for the higher coefficient of variation for the
data of procedure (iii) in relation to that for (i) and (iv) is that random
amounts of particulate and colloidal materials were drawn off each day from
the unstirred subsample thereby inducing fluctuations in the measured values
of Fmax/R by means of the optical mechanisms described above.
A number of other interesting trends and effects can be seen in the
fluorescence plots shown in Figure 6. The sample preserved according to (ii),
i.e., under conditions of ambient temperature and lighting and shaken prior to
anaylsis, exhibited aslow but steady decline in Fmax/R with time away from
an initial value that is close to the initial values for (i) and (iv). A
likely explanation for this trend is that the fluorescent substances are
undergoing degradation due possibly to bacterial activity, photodecomposition
or oxidation.
The additional features contained within procedures (v) and (vi) were
evaluated with the specific purpose of terminating rather than inhibiting
bacterial activity. Procedure (vi), similar to (i) except that the sample was
heated to 130°F for 1 hour each day, gave values for the mean of Fmax/R and
the coefficient of variation that were slightly higher than that for
procedures (i) and (iv), and that exhibited a slight upward trend with time.
This suggests that heating the sample to terminate bacterial activity provides
no additional advantage over inhibiting bacterial activity by storage in the
dark at 4°C and, in addition, may induce a small increase in the fluorescence
properties of the sample. -
Procedure (v) was found to be unacceptable as a method for termi nat i n9
bacterial activity both because of the relatively large coefficient of
31
-------
variation (Six = 6.1 percent), and because the mean value for Fmax/R was
significantly lower than that for procedures (i) and (ii). This latter
phenomenon has been observed by Smart et al. (1976), who showed that changing
sample pH to below 4 produced a marked reduction in the fluorescence of a
natural water sample, in relation to unmodified samples having pH values in
the range between 6.5 and 8.5.
The fluorescence spectra used to determine the optimum sample
preservation procedure and those subsequently obtained for the 161 samples of
the survey handled according to procedure (iv), were all made on unfiltered
samples. This approach was adopted for the obvious reason that, when
performing fluorescence measurements from an airborne platform, one does not
have the option to filter the in situ samples. One might then expect Fmax/R
for unfiltered samples to show a better correlation with TOC than with OOC or
P~C on the assumption that both the dissolved and particulate organic
materials make significant contributions to the measured value of Fmax/R.
With a view to establishing the fluorescence -properties of the dissolved
fraction, the fluorescence spectra were obtained on a day-to-day basis for a
subsample prepared according to procedure (vii), which is identical to -
procedure (i) except that, in addition, the fraction drawn off each day for
fluorescence analysis was passed through a O.3-~m pore-size cellulosic
membrane filter (Millipore MF-PH). As indicated in Figure 6, the mean value
of Fmax/R for this data set is about 11 percent below that obtained using
procedures (i) and (iv), suggesting that, for this specific sample, the
contribution to Fmax/R from the dissolved fraction is on the order of 89
percent. However on a day-to-day basis, the percentage for the fluorescence
of the dissolved fraction varied significantly from a high of 96 percent to a
low of 77 percent, indicating that other factors are influencing the amount of
fluorescent particulate material entering the filtrate in addition to the
filter pore size. The coefficient of variation for this data set was 6.8
percent, which is over three times larger than that for procedure (iv).
Concern that the-variability in Fmax/R for the filtered samples was due
to contamination by e1utable organics known to be present in membrane filters
(Gui11ard and Wangersky, 1958; Cahn, 1967; Parker, 1967) was unfounded.
Tests performed on high purity water samples indicated that they remained
essentially fluorescence free after passage through the filters referred to
above suggesting that any soluble organics contributed by the filters are
essentially non-fluorescent. A rational explanation for the higher
variab1ity in Fmax/R for the filtered samples in relation to the unfiltered
samples might be the presence of relatively high concentrations of colloidal
material in relation to the truly dissolved and particul~te fractions.- Much
evidence is now available indicating that a significant fraction of TOC
present in both marine and fresh waters exists in colloidal form (Reiswig,
1972; Sharp, 1973; Lock et a1., 1977), and that the colloidal organic carbon
(COC) fraction is generally much larger than the P~C fraction (Reiswig, 1972;
Wangersky, 1972; Sharp, 1973). The equivalent sph~rical size range for
particles in this intermediate (colloidal) size category can vary continuously
from those 10-3 ~m in diameter, only just larger than truly dissolved and
discrete molecules, through to those 1 ~ in diameter, ~ore typical of
particles that readily settle in water when subjected to a 1~g gravitational
field.
32
-------
Problems arise because of the tendency of eoe to adsorb onto the filter
mc.erials (Lammers, 1971; Reiswig, 1972; Gordon and Sutcliffe, 1974), even
though the size of the colloidal particle may be smaller than the equi~alent
filter-pore size (Wangersky, 1972). The filtration of samples containing cae
may be further complicated by the tendency of colloidal particles to aggregate
into larger units (Riley, 1970; Reiswig, 1972; Sharp, 1974) either when the
sampl~s are allowed to stand without agitation as occurs during transportation
and storage, or directly at the filter surface. As these particles grow in
size, the probability that they will penetrate a given glass fiber filter will
fall and ultimately approach zero as the particle diameter exceeds the filter
characteristic pore size. In this respect, it is significant that the
fluorescence data for the subsamples, prepared according to procedure (iv) in
which the samples were agitated with an ultrasonic probe just b~fore the
fluorescence analysis, exhibited the lowest day-to-day variability. Clearly,
if the processes governing the adsorption and aggregation of organic colloidal
materials are not fully understood or cannot be easily controlled, then it is
conceivable that random amounts of cae may enter the filtrate for a given.
samp 1 e and filter type after a gi ven storage peri od. Although ultrasoni c
agitation prior to filtration may help to break up colloidal aggregations, the
problem of adsorption onto the filter surface still remains. Partly because of
the relatively poor reproducibility of the Fmax/R data obtained from
filtered samples, but principally because sample filtration is not relevant to
remote sensing of in situ samples, no further attempts were made to measure
the fluorescence properties of the dissolved fraction.
The relatively poor reproducibility of the fluorescence data for the
filtered samples raised doubts concerning the reliability of the DOC data.
Evidence exists that glass fiber filters are able to adsorb significant
amounts of eoe that would otherwise be expected to pass through a filter of a
given pore size (Riley, 1970; Wangersky,1972; Gordon and Sutcliffe, 1974).
This evidence conflicts with claims by the filter manufacturer, who suggest
that their glass fiber filters have negligible adsorptive properties (Meakin
and Pratt, 1972). .
Although it is possible that the glass fiber filters are able to trap a
higher percentage of cae material than are other types of filters, resulting
in lower DOC values, the reproducibility of the DOC data obtained using these
filters appears to be acceptable for this type of measurement. For example,
the 30 replicate TOC and DOC analyses . performed on a Lake Mead sample gave
means and coefficients of variation for TOe of 2.86 mg/l and 9.1 percent,
respectively and, for DOC, of 0.91 mg/l and 14.3 percent, respectively. As
the spread of the DOC data is not markedly different from that of the TOC data
particularly as the DOC value of 0.9 mg/l is close to the usable sensitivity
limit of the TOC analyzer, we conclude that the reproducibility of the DOC
data has not been significantly impaired by use of the glass fiber filters at
least for the samples from Lakes Mead and Mohave. However, as comparable TOe
and DOC reproducibility tests were not performed on samples from the
Atchafalaya Basin, it cannot be concluded that filter-sample interactions were
absent from the DOC measurements made on this group of samples.
Even though the reproducibil ity of the DOC data obtained using glass
fiber filters appears to be within acceptable limits,. it cannot be ruled out
33
-------
that use of non adsorbing filters might result in.a higher correlation between
FmaxlR measured on unfiltered samples and DOC than is reported in this
stu~y. Silver membrane filters would appear to be an obvious alternative as
these filters do not appear to adsorb COC (Wangersky, 1972; Gordon and'
Sutcliffe, 1974). However, care should be taken with the choice of pore size
when using these filters. The O.45-~m filters are known to contain relatively
high and random amounts of carbon that must be removed by precombustion, a
process that increases their effective pore size to about 0.8 ~m (Riley, 1970;
Wangersky, 1975). In contrast, the 0.8-~m and 1.2-~m filters were seen to
exhibit lower carbon blanks and maintain their pore size after combustion.
It is therefore recommended that, when preparing samples for DOC or .
fluorescence analysis using filtration techniques, particularly for waters
high in colloidal matter, the samples be filtered using the aforementioned
0~8-~m pore-size silver membrane filter in spite of the drawbacks indicated
earlier for this filter type. Use of this specific filter type will therefore
combine the DOC and COC fractions into a single overall DOC category that will
also include bacteria smaller than 0.8-~m in size, while simultaneously
avoiding adsorption effects peculiar to glass fiber filters. In addition, use
of this type of filter eliminates problems with the fluorescence measurements
caused by optical scattering from large (>O.8-~m) particles. However, it is
recommended that colloidal aggregations should be broken up prior to
filtration by vigorous agitation, preferably by means of an ultrasonic probe
as described earlier.
An alternative approach to this problem is to simply avoid it by not
performing analyses on filtered samples. Because the separation of sample
organic carbon into the dissolved and particulate fractions by filtration is
both arbitrary in terms of the selected particle cutoff size and the filter
characteristics, and is possibly unreliable due to the formation-of
aggregates, Gordon and Sutcliffe (1973) and Sharp (1973) have suggested that
filtration not be performed and that TOC be the sole parameter characterizing
waterborne organi c carbon. In studi es of the type reported here ~ thi s .
approach wouJd be undesirable for the simple reason that, as will be shown,
the fluorescence parameters exhibit significantly higher correlations with DOC
than with roc. This approach would also appear to be unwise because DOC is a
better overall indicator of the concentration of natural (dissolved) organics
such as humic and fulvic acids than is TOC. The presence of these substances
in source waters is currently of great concern because they are known to be
susceptible to conversion into potentially carcinogenic trihalomethanes during
routine chlorination i~ drinking water treatment plants.
Although the adopted sample preservation and storage procedure maintains
the laboratory measured values of FmaxlR at an acceptably constant level, it
should not be concluded that these values are necessarily the same as those
that would be obtained either by using a fresh grab sample or by making an in
situ measurement, whether made using either a flow-through fluorometer or an
airborne laser fluorosensor.
A number of environmental factors that are known to influence FAIR,
either directly or indirectly, will change as the sample is collected,
transported and stored in the laboratory. Some of the factors that are
-34
-------
subject to change are sample temperature, pH, colloidal state, dissolved
oxygen level and the viability of algae and bacteria. For example, Smart et
ale (1976) have shown that the fluorescence of dissolved organics varies with
temperature according to the relationship FT = Foexp(mTr, where FT is
the fluorescence intensity at temperature T, Fo is the corresponding
emission at ooe and m is a temperature coefficient that is a constant for a
given water type. With m = -O.Ol.Boe-l, the largest value observed, a 15°C
change in sample temperature produces a 25 percent change in the fluorescence
emission intensity.
However all questions concerning sample integrlty and changes in FAIR
that occur between field and laboratory can be avoided if the airborne laser
fluorosensor is calibrated to measure TOe or DOC directly. This can be
achieved by making an airborne measurement of FAIR on. an in situ sample
together with concurrent ground truth determinations of TOe and DOC for a
preselected sampling station site on the water body being surveyed. A
conversion factor relating FAIR to either TOe or DOC is then obtained
directly from Equation 5.
35
-------
SECTION 8
DATA ANALYSIS
Due to the absence of any simple physicochemical relationships between
the organic carbon characteristics of a water sample and its corresponding
fluorescence emission properties, the approach adopted in analyzing the
experimental measurements is strictly statistical with emphasis placed on
establishing the existence of a linear relationship between these two groups
of data through use of the linear correlation coefficient.
LINEAR CORRELATION RESULTS
The emission intensity values Fmax, FR' and R, obtained on the
aforementioned samples using the laboratory spectrofluorometer, were used to
calculate the nondimensional ratios Fmax/R, FR/R, Fmax/FR' Fmax/(R +
FR) and (Fmax -FR)/R for correlation with independent TOC, DOC and POC
measurements obtained on corresponding subsamples. O~her fluorescence
parameters employed in the correlation analyses were the characteristic
. wavelengths Acen and Amax, the bandwidth BW, and (Acen -Amax). In
addition, the fluorescence to Raman ratios FAIR and Fa~/R were calculated.
FA is the fluorescence emission intensity at any specltied wavelength A and
was measured at intervals of 10 nm from 350 nm to 500 nm. Fav represents
the mean fluorescence emission intensity between 350 nm and ~uO nm and is
obtained from the individual FA values by means of the relationship
Fav
N
= ~ i ~1 FA'
where N is the number of spectral bands used in the summation. For the present
measurements, N = 15.
Data used to correlate DOC and TOC with Fmax/R and FR/R are presented
graphically in Figures 7, 8, 9 and 10, which show Fmax/R versus DOC, FR/R
versus DOC, Froax/R versus TOC and FR/R versus TOC respectively.
Examination Of the data in these figures indicates the presence of a
relatively large and discrete group of data points at the low end of the
fluorescence scale. This group represents the 54 samples from Lake Mead and
Lake Mohave together with the four drinking and ultra-pure water samples.
The Pearson product-moment linear correlation coefficient, r, was
calculated for all possible pairings between the fluorescence and organic
36
-------
'"
o
a
""
u
o
go
~
CORRELATION
COEFFICIENT = 0.83
N = [58
.
'"
a
\Q
-
.
.
'"
o
.
-
..
.
'"
o
~
. .
. . .
.
.
.
.. . .
+
0 .
~ + . ..
CD . .. ..
.
. ..
0 . . ..
~ + .+
CD + .
.. .. '.
i . .
.
. .
0 . . .
~ .., .. .
. . ..... ..
+ i'I'..+ ~.. ..
+ . . .
+ . ". .
.... ..:.. ...
0 ~
~ ....
"" .~ ... .
.
... .
£. ..
0 FM/R
~
91.00 1.'0 3.00 4.'0 8'.00 7.'0 9.00 10.'0 1%.00 15.,0 1'.00
'"
a
a
-
Figure 7.
Variation of DOC (mg/l) with Fmax/R for 158 samples.
37
-------
<:>
~
N
U
a CORRELATION
CI a 4
~ COEfF[CIENT = 0.79
III
- N = 158
CI +
~
\Q
-
+ I
Cli
0,
.;j
-, ..
CI
~
~
..
.
. +
CI
~
~
.
4 . 4 ..4 J
o .
~ .. ..
II) .. .. . . I
.
.. ..
0 .
~ .. . ..
. .
\Q + .. .
. . ..
.. .
. . .
. .
0 .. . .
~ .-+t.. .. ..
. +.. . ..
.. .+. ..
. .. .... .
+ . +
+ + .: .
.. + ..
0 ~+.. ... .
~ .. .
..
N .. .t. .. . . .
+
..... .
L. .
0 FR/R
~
9J.oo O.SO 1.20 1.1SO z.~o 3.00 3.80 ~.20 ~"BO 5.~0 6.00
Figure 8.
Variation of DOC (mg/l) with FR/R for 158 samples.
38
-------
<:)
o
-
1D
u
a
g~
1D
..
.
CORRELAT I ON
COEH[CIENT =0.73
N = 1 58
<:)
o
~
<:)
o
In
I"
<:)
o
o
I"
<:)
o
tD
N
CI
o
o
N
CI
o
III
-
<:)
o
~
+
. .
o J
o
II' i .
~ f~ ~
a
'?
C'Q . 00 1..5C 3.00
. ~
.
.
.
..
.
.
.
.
.
.
..
.
+ .
.
.
..
..
..
.
. ..
.
.
+- ~
~
..
.
..'11> . +
+ '\* t
l.~1f. ..+
~t... 4'++
... .
+
++
.
.. ...
+ .
.. ...
.
.
..
FM/R
4.sa
5.00
7.'0
S.OO
10.50
1%.00
13.'0
15.00
Fi gure 9.
Variation of TOC (mg/l) with Fmax/R for 158 samples.
39
-------
co
~
IQ
U
Q CORRELATION
co I-
~ C OEF FIe 1 EN T = 0 . "7 4
on N = 158
..
co
~
0
..
co
~
on
~
co
~
0
'"
co
~
II)
N
CI
~
0
N
CI
C!
on
-
I
I
CI
C!
0
-
..
..
0
C!
1ft 4 .
t-. ..
~ ..
0
C! ~
9>.00 0.60 1.20 I.ea
..
..
.+.
.
.
.
..
..
..
..
..
..
..
+
..
..
.. .. .
~
..
..
.
.
..
..
....
...
.. .
. ..
.
.
...
..+ ~
..
~++.. :.:\+ .~.. +
r... .. .. 4J.:.
.. ..
. ..
2.40
~
..
FR/R
++
.
).00
3.50
..20
..80
:I .40
6.00
Figure 10.
Variation of TOC (mg/l) with FR/R for 158 samples.
40
-------
carbon parameters. These coefficients are shown in Table 2 together with
values denoting the upper and lower limits of the 9S percent confidence
interval for r calculated by converting the distribution of r to normal
statistics using Fisher's z transformation, together with N, the number of
data pairs used in the calculation of r. All values of r.? to.27 with N.? 95
and r > to.22 with N > 148 are significant at the 1 percent level. Similarly,
all values of r > to.21 with N > 95 and r > to.17 with N > 148 are significant
at the 5 percent-level. - - -
Also compared with these fluorescence parameters are chlorophyll ~
(CHLA),pH (PH), dissolved oxygen (DO), conductivity (COND), Secchi disc
transparency (TRANSP) and turbidity (TURB). Correlation coefficients relating
Amax and Acen to the other f1 uorescence i ntensi ty rati as are al so
presented. In addition, an array of coefficients for all possible pairings
between the established water quality parameters for which measurements were
made, is presented in Table 3. Maximum, mean and minimum value for all the
fluorescence and water quality parameters are presented in Table 4 together
with the number of measurements made for each parameter.
Discussion of the correlation data is presented in two parts; the first
part deals with the relative tendencies of the three organic carbon parameters
to correlate with the fluorescence emission data whereas the second part
examines the relative merits of the various fluorescence parameters for use as
remote sensing indicato~ of waterborne organic carbon.
Table 2 indicates that DOC is the water quality parameter most highly
correlated with the various fluorescence parameters, in particular with
Fmax/R, FR/R, (Fmax - FR)/R, Fmax/FR and Fmax/(R + FR). A
similar conclusion can be drawn from Figure 11, where the coefficients for
correlations between FAIR and DOC, FAIR and TOC, FAIR and pac, and
FAIR and turbidity are plotted as a function of wavelength. At wavelengths
of 381 nm and approximately 435 nm, FAIR becomes equivalent to FR/R and
Fmax/R'respectively. The correlation coefficients for DOC versus the
various ,fluorescence parameters lie in the range from 0.78 to 0.84 whereas the
correspondi ng values for TOC versus the same fl uorescence parameters 1 i e in
the range from 0.62 to 0.74. By converting the distribution for these r
values to normal statistics uSing Fisher1s z transformation, a test can be
performed to determine whether the correlation coefficient for DOC versus a
specific fluorescence parameter belongs to the same population as does the
coefficient for TOC versus this same fluorescence parameter. This procedure
shows that correlations involving DOC are significantly different from similar
correlations involving TOC at. the 5 percent level for a sample size of 158.
As DOC is the organic carbon pararpeter exhibiting the highest correlation
with FAIR (r:0.83) together with a number of other fluorescence parameters,
and, as p~C shows only a weak correlation with DOC for which r:0.44 (Table 3),
one can then predict a poor correlation between p~C and FAIR; this is
. confirmed by the curve for p~C versus FAIR in Figure 7, for which r:O.4S.
. This prediction can be made because the fluorescence intensity measurements,
FA, were obtained on unfiltered samples containing both DOC and POC, so that
the relationships between DOC and FAIR and between P~C and FAIR are
mutually dependent. A consequence of these observations is that TOC (=DOC +
41
-------
TABLE 2. CORRELATION COEFFICIENTS FOR FLUORESCENCE EMISSION PARAMETERS VERSUS
WATER QUALITY PARAMETERS
nt/P ff~/~ IFH-FRI/R FM/FR FM/CR.F~I LCEN LHU LC-LH Ihl TURII
III) .11146 .11440 .87114 .8361 .831)6 .!l201 .1914 -. 1442 .044] -.0154 OOC
R .1I]~h . .1929 .11361 .183<; .111'1] .1614 .12411 -.2931 -.1126 -.2582
I I< .11dl .1214 .111]4 .1155 .11'>3 .6810 ."411 -.4]01 -.2641 -.4241
N 1151>1 liSt\) lI<;ill 11581 115'11 11'351 11551 11551 11551 11081
III) .1~bO .1I00l .1131 .1049 .15.15 .1066 .1191 -.2282 .2181 .7.2]6 Toe
R .1:'11" .13"~ .10]2 .6116 .6718 .6181 .6349 -.]111 .12111 .0319
I R .6SQl ....511 .615<; .51111 .S"43 .~121 .5]1" -...98.. -.02116 -.1 SO..
" 115bl 115111 11<;111 11')81 11581 11551 11551 11551 11551 11011
III< .<;)96 .5821 .4QJ3 ."0!!4 ."9h5 ...lIn ."151 -.. 159 .4124 ."124 PUr.
R .41/1 .4611 .:lh 34 .261<; .3" 12 .211>9 .1"20 -.]265 .2105 .2415
I.~ .2112 . )]42 .21111 .114? .2?20 .1226 .19211 -.4621 .11511 .0')41
N 11511 11511 11<;11 11511 11<;11 114HI 11481 11481 11481 11041
III< -.00~9 .0284 -.018<; .0)0<; ..O?.!4 .0412 -.0442 .2144 .1524 .0261 CHL j\
Q -.19~b -.11>21 -.7.nll -.160.3 -.lh'l2 -.1499 -.?321 .0892 -.0386 -.1645
II' -.31lh -.1"11 -.31425 -. ]]911 -.3""9 -.130] -."0"1 -.102" -.2268 -.]4)1
~, 110..1 110101 110..1 110..1 110..1 II 0"1 110101 11041 110101 110..1
"P -.04di' .0951 -.1222 -.2911 -.1115 -.11)) -.1061 .2..21 .4512 .4994 PH
I< -.~)d" -.0991 -.1"1" -.4580 -.352" -.15"1 -.4111 .0526 .2901 .U94
1.1) -.4Ia -.21164 -.4127 -.591Q -.51"" -.'>118 -.(081) -.1414 .10]] .1'>11
+:a ~. 11011 11011 11011 11011 II 011 11011 II 0 II 11011 e 1011 ClOII
N
"I) -.1114 .0060 -.111.. -.381ft -.21124 -..1211 -.350] .0039 .6126 .61S& 00
P -..lnut- -.1'140 -.3<;51 -.5)80 -.4""3 -...b12 -.c;ObS - . 1801 .418) .5512
IP -.4"40 -.1" J] -.5101 -.6604 -.5A'>1 -.1>198 . -."153 - . )6)2 .3113 .410A
N 110..1 11041 11041 11041 11041 11041 11041 110..1 110"1 110..1
I'" .01)b1 .1336 .0131 .011h . OS"',, .0551 -.1045 ...01.. .015" .0465 CO"lO
" -.131" -."""6 -.1,,"4 -.119<; -.1..16 -.1422 -.?9..tI .229] -.1228 -.1<;13
IP -.31"" -.~,)11 -.349'> -.3631 -.12..9 -.\2",5 -...h..4 .0]..5 -.3111 -.3311
., I 911 I '111 I 'HI I 911 1 911 I 911 I 911 I 911 I 911 e 911
1.1) .")J" .<.'9..5 ...99<, .I,&..A . .51.1" .5106 .4951 .5255 -.4055 -.5614 TRj\NSP
... .lbOI .IO~I .1:\&Q .539<; .420;& ...2<'2 .132" . 36 19 -.5568 -.6820
10 .Of>d2 -.1\97~ . Ie; 14 .38"" .1'<;11 .2..1;> .146'> .1858 -.618'+ -.1142
M c 9dl I 9111 I Qal I 9!!1 I 9't1 I 'Iii I e 981 I 981 1 981 I 9/11
.11> -.n()~l .211..11 -.1?51 -.)"11> -.1"" -.~~)) .0099 -.5155 .9051 1.0000 lURII
p -.181:\ .11209 -.10.!"I -.5112 -.3116 -.411\5 -.1151 -.6394 .8649 1.0000
II> -.35.,,6 -.1"..5 -."f>I~ -.""01 -.41..5 -."5~9 -.148& -.1310 .11096 1.0000
. II I UI 11101 11101 11101 III fli 11101 11101 11101 11101 11101
II" .~..o" .'<195 .9..38 .9139 .9103 1.00110 .Q602 -.23"6 .Ob..9 -.2533 LCEN
" .9?69 .a911 .9;>..0 .96..5 .9<;'11 1.0000 .9460 -.]151 -.0901 -.'+185
1.1> .qOIS .>15..9 . AcH.. .951Q .~4~4 1.0000 .9210 -.5012 -.2"20 -.5<;99
" C 1C;,U II C;A I 11<;1\1 11"81 IIS~I el581 11581 11581 11581 11101
II" .934" .94"') .9025 .921? .99111 .9b02 1.0000 -.5511 .2913 .0099 LMj\.
I> .911'> .<"217 .A1>91 .'}OIQ .'1152 ..)..60 1.0000 -.&5511 .1495 - .115 1
I P .A811 .902'> .1\;>5.. .8684 .9M3 .~210 1.0000 -.1359 -.005] -.3486
M 115t11 11')81 C 1,>81 11581 115/11 115111 11581 el581 elSll1 e 1101
ur ani lr are the uwer ani lower limits of the 95\ confidence intCival for the correlation
coefficient r. N is the I\U1tJer of ~les used to fird the conelaUon coefficient r.
-------
TABLE 3.
CORRELATION COEFFICIENTS FOR RELATIONSHIPS BETWEEN WATER QUALITV PARAMETERS
Doe toe poe Cl'iL A. PH 00 CI)t.lO TAA.NSP 111....
Ilk 1.0000 flor
R 1.0000
I A 1.0000
h (1581
IIIl ."448 1.0000 HIe
.. .7937 1.0"00
LU .12!!? 1.0000
~ ( I C) 11 11 '>B 1
1/1< .S6/Ja .8911.1 1.0000 I'o~
I< .4419 .Abcl 1.0000
I." .30S0 .8)59 1.0000
~, 11511 (1511 1 Hill
III< -.0<;!!7 . 1327 .31M 1.01100 r.Ht A
R -.7471> -.0615 .12A5 1.00,,0
I H -.4194 - . 251 I -.06R6 1.00110
N 11 021 11 0 11 1 981 (11)41
"14 .0152 .Ibcn .:\314 .S3Al I.OOO~ PH
~ R -.lld9 -.0210 . \It)4 .3A4b 1.0000
W L" -.1600 -.?~04 -.0')<;6 .?O#':' 1.0000
,.. II 001 1 9';'1 1 9111 110/1 1 110 II
Uf< -.03bO .l'U, .)5');> . <,b<;o .18)) 1.11000 1)(,
.. -.22t.>1 -.oon .1116 .07'1b .b9'>2 1.0/100
II' -.4005 -. 19 3A -.0?41> -. 11"\6 .5191 1.01100
N 11 021 ,11011 ( 'Jill (1011 (100 I 11 041
III' .))98 . 1'142 .10?0 .~M'O' .5505 .:n30 1.0000 r.lI/Ij P
... -.OtlOl. -.0054 .105/1 .41 n I . }'34 .1"'51 1.0000
U' -.:>146 -.(!04' -.09k.. .?11" .2094 -.O'\"\':; 1.0000
N ( 951 ( 9.1 '1 911 ( '11.1 ( ~JI ( Qnl I 91)
Ilk .4215 .2553 .0108 .1'IQl -.2004 -.~I\">l .?4/1b 1.0000 TIJANS"
.. .2445 .0599 -.1'21 ".00:>& -.JA45 -.lAf,O .04~5 1.0000
LH .1\494 -.1'.(11 -.32~~ -.19"\11 -.5424 -.5/00" -.1!>3b 1.0000
N 1 9bl 1 '151 1 921 ( '171 1 941 ( 'AI 1 9;>1 ( 9>11
liP -.0154 .2?3b .4124 .O?"I .4'194 .bl<;f, .114"5 -.Std4 1.0000 'II~;'
" -.2562 .0319 .2415 -.11'>45 .)394 .5<;7? -.1:'13 -.blt?1I 1.0000
LU -.42"1 -. ) SOlo .0541 -.34,\1 01511 .410" -.3311 -.7142 1.0000
N (101!! 11 011 11041 II 041 (1011 11041 ( 'HI ( 9AI III I) 1
ur am lr aloe the uppel: am lONer Hrni ts of the 95% confidence interval for the correlation
coefficient r. N is the l1lunber of samples used to fioo the correlation coefficient r.
-------
TABLE 4. MAXIMUM, MEAN AND MINIMUM VALUES FOR FLUORESCENCE AND WATER
QUALITY PARAMETERS
r~ax. Mean Min. Samp 1 e
Parameter Units Value Value Value Size
Fmax/R 14.25 4.43 0.02 158.
FR/R 5.95 2.36 0.02 158
Fav/R 10.71 2.95 0.01 158
(Fmax-FR)/R 8.30 2.06 0 158
Fmax/FR 2.40 1.69 0.89 158
Fmax/(R+FR) 2.05 1.10 0.02 158
Acen nm 447.74 435.82 423.55 155
Am ax nm 439.69 427.68 415.63 155
Acen-Amax nm 13.60 8.15 . -2.70 155
BW nm 153.90 122.13 113.30 155
DOC mg/l 18.91 4.02 0.18 158
TOC mg/l 43.69 7.21 0.20 158
POC mg/l 26.95 3.47 0 158
CHLA . J1 g/ 1 31.30 7.73 0.20 104
pH pH 8.10 7.34 6.70 101
DO mg/l 9~60 5.01 0.01- 104
CO NO J1rnhos/cm 973.00 270.71 87.00 97
TRANSP m 0.99 0.30 0.10 98
TURB FTU 220.00 50.07 0.61 110
44
-------
POC) will show a weaker correlation with FAIR than will DOC. This is
apparent in Figure 11 where the correlation coefficient curves for TOC
versus FAIR and for DOC versus Ft/R have values of about 0.72 and 0.83
respectively in the blue spectra region.
It should be noted that neither the fluorescence data, the organic carbon
data, nor the various correlations provide any direct evidence as to whether
the fluorescence emission originates from all or part of either the
particulate or dissolved substances or from some mixture of both. This
ambiguity, which exists in spite of the strong correlation between Fmax/R
and DOC (r = 0.83) and the weak correlation between Fmax/R and POC (r =
0.42), arises because the fluorescence measurements were made exclusively on
unfiltered samples. As indicated earlier, filtration of the fluorescence
samples was not performed, mainly because airborne monitoring does not allow
for any such sample preparation, but also because of the aforementioned
difficulties encountered in measuring the fluorescence of filtered samples.
Also, sample' filtration may produce differences between the laboratory
measured values of Fmax/R for the filtered and unfiltered samples due to the
elimination of particulates that can produce either scattering losses in R or
an apparent enhancement in Fmax due to scattering of .a leaked background
signal. These differences are in addition to that caused by the elimination
of the fluorescence contribution from particulate organic substances. Further
discussion of the influence of these scattering side effects on the
fluorescence data are presented below. .
As an example of this ambiguity, a scenario can be envisaged in which
sample fluorescence orginates entirely from a small but significant fraction
of the particulate organics that correlates, not with POC, but with DOC. This
latter condition is necessary because poe, and DOC have been shown to be
poorly correlated (r = 0.44). A.possible candidate for this material might be
eoe which is more likely to vary in unison with DOC than with POC, even though
eoc is included in the POC category by the ability of glass fiber filters to
efficiently trap colloidal material by adsorption. This situation is ..
suggested mainly to caution the reader to avoid assuming that the correlations
obtained in this work necessarily established causal relationships. Clearly,
the relative contributions to Fmax/R from both particulate and dissolved
organics can be determined by making fluorescence measurements on filtered
samples as well as on .unfiltered samples in which the same filter type is used
for both the fluorescence and the DOC analyses. Based on the earlier
discussion, a non-adsorbing silver membrane filter with a pore size of 0.8 ~m
would appear to be ideal for this application, so that all COC below this
cut-off size will enter the filtrate and be counted as DOC. Alternatively,
glass fiber filters might be used throughout, in which case a significant part
of the COC fraction will be trapped on the glass filter and hence be counted
. as POCo
Discussion of the relative merits of the various fluorescence parameters
as airborne predictors of waterborne organic carbon are dealt with on a
case-by-case basis as follows:
45
-------
1.0
r
0.9
0.8
0.7
0.6
0.5
0.2
0.1
o
-0.1
-0.2
340
e
e
I F~.lR vs. DOC. n = 158
e_e-e-.-e-e-e-e-e-e
--e-
~/.
/ F)JR VS. TOC~ n = 158
.~:-1-._.-.-.-.-.-._.~.~_._.
I I .
I
e """J.......
I e-"e-.. F)JR vs. P~C. n =151
e-e
. -e-e-e-e-e-e-e-e
I
I
I
I
I
Water Raman Emission
V Wavelength at 381 nm
-~ - - . - - --
I e,
I e'e, FAIR vs. Turbidity. n = 110
e-"e-e-e-e-e-e-e-e
360
380
400
460
420
440
Wavelength. A (nm)
Figure 11. Variation of linear correlation coefficient, r, with
wavelength for FAIR versus DOC, FAIR versus TOC, FAIR
versus p~C and FAIR versus turbidity.
46
-------
( i i )
(i)
Fmax/R and FAIR above 400 nm
The correlation coefficient curve for FAIR versus DOC shown in
Figure 11 indicates that the optimum value for correlation with DOC
occurs in the region of 430 nm; this wavelength corresponds closely
to the mean values for the peak (Amax) and centre (Acen)
wavelengths of the emission spectra respectively as indicated in
Table 4. As Fmax is defined to be the fluorescence intensity at
Amax' then Fmax/R would appear to be 'the optimum value of FAIR
for use as a DOC indicator.
FAIR below 400 nm
According to Figure 11, the correlations Qetween FAIR and
either DOC or TOC exhibit a marked decline for wavelengths shorter
than 400 nm. Two reasons for thi s dec 1 i ne are suggested.
It is apparent from the fluorescence emission spectrum for the
sample of Lake Mead water shown in Figure 2 that the fluorescence
intensity FA falls rapidly for wavelengths shorter than 380 nm.
This will produce an increase in the percentage measurement error for
FA, thereby tending to reduce the correlation between FAIR and
either DOC or TOC.
However, a more plausible explanation for this decline involves
an experimental artifact introduced by the spectrofluorometer in
conjunction with the effects of sample suspended particulates. In
addition to the desired spectral band at 337 nm, the excitation
monochromator is known to emit a somewhat broader but considerably
. weaker band of scattered light. This leaked or background light
falls off only gradually over a region several hundred nanometers.
wide to the long wavelength side of the desired excitation band. As
the sample fluorescence is monitored at 90° to the excitation beam,
this background does not constitute a probl em in the absence of a .
scattering medium such as a suspension of particulates. In reality
however, samples were frequently turbid such that a significant
fraction of FA measured in the 350-nm to 380-nm region for these
samples was due to this Mie-scattered background light in addition
to the desired contribution from the DOC-related fluorescence. The
influence of this background signal on FA is exacerbated by the
fact that, in the 350-nm to 360-nm region where this background is
the strongest, the sample fluorescence signal is at its weakest as .
indicated by the spectrum for the Lake Mead sample shown in Figure 2.
This background signal therefore increases the rneasured value of FA
particularly in the short wavelength region while leaving the
measurement of R unchanged; as the Raman intensity R is superimposed
on the fluorescence spectra, its measurement is not affected by the
presence of this background signal. The influence of this scattered
background was particularly noticeable in the spectra for samples
high in suspended particulates as a flattening out of the
fluorescence spectrum due to enhancement of FA in the short
wavelength region below 380 nm.
47
. ~
-------
This explanation is confirmed by uSing sample turbidity as an
indicator of suspended particulate concentration and correlating
turbidity with FAIR. As shown in Figure 11, this correlation
becomes strongly positive in the short wavelength region with
r = + 0.92 at 350 nm. If the suspended particulates were simply
attenuating FA rather than acting as a source of scattered
background light, then the correlation between turbidity and Ft/R
wo~ld be strongly negative in contrast to the high positive va ues
obtained for the spectral region below 381 nm.
Another manifestation of the combines effect of the leaked
background signal and the suspended particulates on FA is provided
by the strong correlation (r = + 0.86) between turbidity and BW, the
FWHM bandwidth of the fluorescence spectrum. The distribution of
these data points are shown in Figure 12. The bandwidth of 115 nm at
zero turbidity probably represents the true limiting value in the
absence of the background interference, although the influence of
differential absorption effects, to be discussed later, may also be
present. As the values of FA at shorter wavelengths are enhanced
without .any si gni fi cant change occuri ng at the longer wavelengths,
the net effect is such as to increase BW in direct response to
increases in the concentration of the suspended particulates as
represented by the turbidity measurement.
This background signal, which contributes to the FA
measurement through the presence of suspended particulates,
is therefore considered to be the principal cause for the marked
decline in the correlations between FAIR and either DOC or TOe
at wavelengths below 380 nm. As DOC and turbidity are only weakly
correlated (r = -0.26)~ it is concluded that the correlation between
FAIR and turbidity has been enhanced at short wavelengths at the
expense of the correlation between FAIR and DOC.
It was possible to eliminate this background from the emission
spectra by inserting a 337-nm narrow band interference filter into
the excitation beam before it enters the sample cell. However, due
to the poor transmission «20~) of this and other filters available
for this spectral region, this expedient produced a significant and
unacceptable reduction in FA and the corresponding signa1-to-noise
ratio. The fluorescence analyses of the 161 samples were therefore
performed without the use of thts filter on the assumption that this
background constituted only a small percentage of FA in the region
of Amax, particularly as the turbid samples were generally also
highlY fluorescent. However, it is. possible that the correlation
between FR/R (at 381 nm) and DOC, as indicated in Figure 11, has
been degraded by this instrumental artifact. Due to the presence of
this background, the F~/R data in the 350 nm to 370 nm spectral.
region cannot be considered meaningful in relation to the TOe and DOC
data.
In contrast to the effects of the background scattered light,.
the influence of differential absorption and scattering on FAIR, as
48
-------
o
~
0
C\I
(I)
o ~
~ =>
~ t-
at
o
~
ID
r-
-
0
~
..
In
-
0
~
N
I't
0
~
~
CO
0
G
~
CO
~
ID
ID
CO
~
..
..
CO
~
C\I
C\I
0
I:! ...
~10.0a
.
.
CORRELATION
COEfFIClt:Nr = 0.86
N = llO
...
.
... ...
...
...
. -.
. *' ...~
...
.
... "'
.
...
.
...
. .
...
...
. 1'.
...
.
+
1'
...
.
..
.
~....
...
... ... ~
.. ... +
... 1'...1'+...
+ ~...
. .+"'1+
... #f'
... ;
...
... ...
...
...
...
~1...
..
lZO.oa
lSO.OO
.
ll~.OO
lZS.OO
l30.00
lJS.OO
l40.00
l4'.iJO
BW
lS' .00
lI50.00
Figure 12.
Variation of turbidity (NTU) with bandwidth of fluorescence
emission spectrum, BW (nm), for 110 samples.
49
-------
predicted by the exponential term in Equation 6, and on the
correlation between FAIR and either DOC or TOCis not apparent in
the curves shown in Figure 11.
Optical absorption by natural waters increases rapidly at
shorter wavelengths, particularly below 380 nm, and is closely
related to the DOC level (Dobbs et al., 1972; Mattson et al., 1974;
Briggs et al., 1976; Smart et al., 1976). In this situation, the
attenuation coefficients in Equation 6, kR and kA, will become
strongly dependent on nf and, by implication, on DOC.
Consequently, an increase in DOC, in addition to increasing FA,
results in attenuation by absorption of both FA and the incident
excitation beam. Even though differential absorption of FA with
respect to R is a nonlinear process governed by the exponential term
in Equation 6, its effect, for all but very highly colored waters, is
to introduce a second order correction term into the otherwise linear
relationship between FAIR and DOC. Clearly a trend of this
magnitude that is also systematic with DOC will not have a
significant influence on the correlation between FAIR and DOC,
particularly in view of the relatively wide dispersion in the
existing data as typified by the "d1stributions shown in Figures 7 and
8. It is therefore likely that the effect of differential absorption
is being masked by the scattered background signal that acts to
enhance rath~r than attenuate FA (and FAIR) in the spectral
region below 380 nm.
Attenuation by Mie scattering from sample suspended sediment is
also stronger at shorter wavelengths and is governed by a wavelength
dependence of the form 1/AY where, for waterborne particulates, y
generally has a value in the range between 1 and 2 (Austin, 1974).
With Y = 1.5, this dependence produces a 22% reduction in R at 381
nm in relation to FA at 433 nm, thereby suggesting the existence of
a positive correlation between FAIR and turbidity for wavelengths
above 381 nm. However, the data for this correlation, as shown in
Figure 11, indicate a negative correlation on the order of -0.18 for
the spectral region above 400 nm. Because this value is not
significant at the 5 percent level for a sample size of 110, specific
conclusions should not be drawn from these correlation data for this
spectral region. For wavelengths shorter than 381 nm, Mie
scattering losses in FA in relation to R, should reduce FAIR and
produc~ anegattve correlation with turbidity in contrast to the
positive values shown in Figure 11. Clearly, any tendency in this
direction has been masked by the influence of the scattered
background signal in this spectral region.
A further trend apparent in Figure 11 is the tendency for the
correlation between FAIR and POC to increase at shorter wave-
lengths, reaching a peak at about 360 nm. Two possible reasons for
this trend are suggested. First, POC, which is only poorly corre-
lated with DOC (r = 0.44) and as such relatively independent of DOC,
may make an increasing fluorescence contribution to FA at short
wavelengths thereby raising the correlation between FAIR and POCo
50
-------
(iii)
(iv)
Alternatively, the particulate organic material may be scattering the
leaked background light into the emission monochromator in the same
manner as the total suspended particulate concentration. With the
data presently available, it is not possible to separate out the
contributions that these two side effects make to the correlation
between FAIR and pac.. The fall-off in the correlation between
FAIR and pac below 360 nm is most likely due to the dominance of
the correlation between FAIR and turbidity in this spectral region
caused by scattering of the background signal from suspended
particulates.
FR/R
Although FR/R exhibits a weaker correlation with DOC
(r = 0.79) than does Fmas/R (r = 0.83), there are several reasons
why FR/R is preferable to Fmax/R for use as an airborne indicator
of DOC.
First, the difference between these two coefficients, which is
not significant at the 5 percent level for a sample size of 158, is
apparently due to an experimental artifact introduced by a scattered
background signal that leaks from the excitation monochromator. In .
the absence of this interference, it can be expected that the
correlation between FR/R and DOC will approach that obtained
between Fmax/R and DOC.
The second reason concerns the influence of differential
spectral absorption and scattering on the measurement of Fmax with
respect to those of FR and R. As already discussed, these
secondary effects cause the fluorescence (and Raman) intensity
measurements made at the shorter wavelengths to be attenuated more
than those made at the longer wavelengths. Measured values of
Fmax/R will therefore exhibit the influences of these side effects
in addition to the desired information un DOC concentration.
In contrast, both the ai rborne and labo.ratory measurements of FR/R
are free of these differential spectral effects as predicted by
Equations 4 and 6 respectively.
A final advantage to using FR/R concerns the practicabilities
of making these fluorescence measurements with an airborne laser
fluorosensor. FR/Rrequires only three intensity measurements, one
at 381 nm and one each on either side of the Raman wavelengths in
order that FR and R can be separated by the process of linear
interpolation. Using the ratio Fmax/R as an indicator for DOC
requires the additional measurement of Fmax. This demands that an
additional system channel for optical detection, electronic
monitoring and data recording be included in the airborne system.
(Fmax - FR)/R
This parameter was calculated with the purpose of subtracting
out any background andi nterference effects common to both Fmax/R'
51
-------
and FR/R. The correlation of (Fmax - FR)/R versus DOC gave a
coefficient of +0.84, which is not significantly different fran the
value of +0.83 for the correlation between Fma~/R and DOC. This
correlation is not surprising as Fmax/R and FR/R are themselves
highly correlated such that their difference will also be highly
correlated with DOC provided that (Fmax- FR)/R is not small in
relation to Fmax/R and FR/R. This parameter would therefore
appear to have no particular advantage over Fmax/R or FR/R as an
indicator of DOC.
(v). Fmax/(R + FR)
This ratio has the advantage that its calculation requires only
two spectral measurements; (R + FR) is the total measured intensity
at the Raman wavelength so.that in this case Rand FR do not have
to be separated. A requirement for only two discrete spectral
measurements would reduce to two the number of detection, monitoring
and recording channels required in an airborne laser fluorosensor
in relation to the four needed to calculate Fmax/R.
At low DOC levels, fluorescence emission is generally weak with
FR«R, so that the ratio Fmax/(R + FR) becomes equivalent to
Fmax/R. At high DOC levels where FR»R, the ratio becomes
equivalent to Fmax/FR. The merits of this latter ratio~ which is
also highly correlated with DOC, are discussed in the following
section.
Although there is no conceptual basis for using this parameter
as a DOC indicator, it appears to correlate with DOC as well as any
of the other fluorescence parameters at the S percent significance
level. In addition, it has the advantage of requiring only two.
spectral measurements. .
(vi)
Fav/R
As the curve of r for DOC versus FAIR is essentially flat in
the 400-nm to SOO-nm wavelength region and as the bulk of the.
fluorescence emission occurs in this spectral region, the correlation
of Fav/R versus DOC has a value of 0.81, which, not surprisingly~
is essentially the same as the coefficients for DOC versus Fmax/R
(0.83) and for DOC versus Fx/R (0.83) in this spectral region.
As the calculation of Fav/R requires considerably more spectral
data than does Fmax/R or FR/R, it offers no particular advantages
as an indicator of DOC.. .
(vii) Fmax/FR and Acen
If the sale effect of changing the concentration of DOC (or TOC)
were to simply change the amplitude of the fluorescence emission
spectrum, one could anticipate a zero correlation between DOC and
either Fmax/FR or Acen. In reality, however, the coefficients
for DOC versus Fmax/FR and DOC versus Acen were 0.78 and 0.76,
52
-------
respectively. These correlations exist because the fluorescence
spectrum exhibits changes in location and shape in addition to
amplitude with change in the DOC level.
This relationship is another manifestation of the influence of
differential spectral absorption on the fluorescence measurements.
fhe dependence of Fmax/FR on the concentration of organic
fluorophorst nft measured using a laboratory spectrofluorometer can
be predicted by a relationship similar to Equation 6 such that:
Fmax
~
=
03 exp {h (kR - kmax)}
(7)
where 03 is a constant similar to 01 and 02t and kmax is the
effective optical attenuation coefficient at Amax. As both Fmax
and FR are linearly dependent on nft Fmax/FR does not exhibit
an explicit dependence on nf but rather varies by virtue of the
fact that the attenuation coefficient kR and kmax are both
strongly dependent on nf. Implicit in this discussion, is the
assumption that nf is closely related to DOC.
fhe experimental data relating DOC to either Fmax/FB or
Acen are presented in Figures 13 and 14 respectively. Tne trend in
both cases. is consistent with increasing sample self-absorption of
the short (UV) wavelength component, FR, in relation to the longer
(blue) wavelength component, Fmax' with increasing sample DOC.
Increasing DOC causes Fmax/FR to increase and Acen to shift to
longer wavelengths. Predictably, FmQx/FR and Acen are highly
correlated, with r = + 0.96; the var1ation of Acenwith
Fmax/FR is shown graphically in Fi gure 15.
The limiting value for Fmax/FR of approximately 1.35 for
vanishingly small DOC levels 1S presumably characteristic of spectra
unbiased by differential self-absorption effects. Similarly, the
limiting value of Acen for small DOC levels is approximately
427 nm.
Further evidence confirming the action of differential self
absorption on FA was obtained by progressively diluting a single
sample high in organics and low in suspended sediment with ultra-pure
water. The variations in FmaxlR, FR/R, Fmax/FR and Acen
with percentage dilution are presented in Figure 16. Fmax/R
and FR/R predictably fall to zero for a 100 percent pure-water
sample, whereas Acen and Fmox/FR fall to their limiting
values typical of a sample low in DOC and therefore negligible
self-absorption over a l-cm pathlength.
These same values of Fmax/FR and Acen for the diluted
sample are also presented in Figure 15 together with the individual
53
-------
CJ
~
(It . , .
u
0 CORRELAT I ON - .
CJ 0 COEFFICIENT -0.78
CI
0 N = l 58
-
CJ ~
CI
~
..
Ct
~
~ .
-
..
..
Ct
~
(It
-
. ..
.. ... ..
Ct . ..
CI
~
.
.. + .. .
CI ..
CI \.+
0 _ .+.. ..
.
. ..
CI . .. +.. ..
CI +
0 .. +
.. .. ..
.. .
+ .. .
. .
CI .. ...
~ . . .. -: i
.. .. .+. . ..
.. .. . .t t
. ~ .
+ + + .. ..
. ~ .
CI +. .." ....
~ ... .
(It .~. .. ..
.
.
..; ..
... ~I~ ..
CI FN/FR
~ ..
93." 0.35 1'.13 1.'3 t.~ 1." 1.~ 2.13 2." 2.:5:5 2."
Figure 13.
Variation of DOC (mg/l) with Fmax/FR for 158 samples.
54.
-------
o
o
(\I
u
a
go
GI)
-
o
o
~1
;1
~
o
-
o
o
GI)
o
o
GI)
o
o
..,
o
o
(\I
o
o
°4Z0 .00
CORRELFlT10N
COEFF[CIENf =0.76
N = L 55
+
..
..
.
.
.
+
+
..
+
. .
.
.
.
.
.
..
..
.
..
+
..
+ ..
..
.
++
. .
.
.. ..-.
++.* ..
.
.. ..
..... +~ ..
.,...
.
..
..
.. . .
.. ...
.
. ..
.. ....
.. .
..
.. " .
. .. ..
.
..
...4-
..
.. .
.. .
.
.
..
.
..
..
.
.
... ..
..
.
...t
.+
... ... .
. ...
..
.. ..
.. ..
..
LCEN
441 .00
4Z' .00
432.00
4ZS .00
4Z~.00
435.00
438.00
444.00
450.00
4"'.00
Figure 14. Variation of DOC (mg/l) with central wavelength .of
fluorescence emission band, ~cen (nm), for 155 samples.
55
-------
450.00
447.00
444.00
441.00
438.00
435.00
432.00
429.00
=
=
~
~
CORRELATION COEFFICIENT = 0.96
Y=18.99 X + 403.4
N=158
.
.
.
.
.
UNDILUTED SAMPLE
.
.~ ~HANGE PRODUCED BY
DILUTING SAMPLE E.1
WITH ULTRAPURE WATER
426.00 .
.
423.00 Fmax
-
FR
420.00
1.00 1.15 1.30 1.45 1.60 1.75 1.90 2.05 2.20 2.35 2.50
Figure 15. Variation of central wavelength of fluorescence emission band,
Acen (nm), with Fmax/FRo Also shown are data obtained by
progressively diluting a single Atchafalaya River sample
(#£-1) with high-purity water.
56
-------
9 2.0 Leen
Fmax Fmax (nm)
-
R . FR
8 439
FR
-
R
7 438
6 .. Fmax 437
~F;"
5 . 436
A ..~
0
435
.
3 1.70 434
2
433
1.60
o 10
UNDILUTED
SAMPLE
20
30 40 50 60 70
DILUTION PERCENTAGE (%)
80
90
100
100%
WATER
Figure 16. Changes in fluorescence spectrum parameters produced by
progressively diluting an Atchafalaya River sample ( E-l)
with high-purity water.
57
-------
values for the 158 undiluted samples. The slope of the linear
regression line of DOC on Fmax/FR for the 158 undiluted samples
was 19.0 and that for the dilution data is essentially identical at
19.1. From this observation, it is concluded that the change in
Fmax/FR and the shift in ~cen with change in DOC concentration
is due to optical self-absorption of the fluorescence emission by the
sample rather than to any differences in the chemical or physical
properties of the dissolved organics between samples. In addition
Fmax/FR' like F~/R for ~ <381 nm, is influenced by the leaked
background light which is scattered by sample suspended particulates.
Table 2 indicates a correlation of -0.52 for Fmax/FR versus
turbidity suggesting that FR is being enhanced in relation to
Fmax by the scattered background signal. .
Fmax/FR also exhibited significant correlations with pH and
00 (dissolved oxygen) giving coefficients of -0.46 and -0.54
respectively. Although these correlations do not establish cause and
effect relationships, it is possible that changes in these parameters
might influence the shape and location of the fluorescence spectra.
In.particu1ar, Smart et a1. (1976) have demonstrated that changing
sample pH to values greater than 9 or to less than 4 can have a .
marked effect on the overall emission intensity whereas changes of pH
within this band had only a small effect on sample fluorescence. As
the maximum and minimum values of pH encountered in this study were
8.10 and 6.70 respectively as shown in Table 4, one can expect the
effect of pH on Fmax/FR to be minimal. The negative correlation
between pH and Fmax/FR would seem to imply that increasing pH
either increases FR in relation to Fmax or shifts the whole
fluorescence spectrum to shorter wavelengths. No further
investigations of the influence of pH on F.~ were conducted partly
because water pH cannot be measured from an ai rborne pl atform and
partly because its effect on the correlations between DOC and the
more important ratios, Fmax/R and FR/R, was, at the most, only
marginally significant (see Table 2).
Unlike fluorescence parameters such as FmaxlR that are subject
to interference from differential absorption, Fmax/FRis uniquely.
dependent on this phenomena for its existence as a DOC indicator.
However, like FmaxlR, it is susceptible to scattering interference
from suspended sediment and, similarly, requires four spectral
measurements for its determination.
. Although ~cen appears to be equally good as an indicator of
DOC as is Fmax/FR' its measurement would probably require at
least six spectral measurements in the 400-nm to 450-nm region for
its accurate determination. Fmax/FR and ~cen would therefore
appear to offer no special advantages over FR/R as indicators of
DOC while incurring several disadvantages.
58
-------
MUL TIPLE CORRELATION EFFECTS
Multiple linear correlation effects between specific fluorescence
properties and an array of water quality parameters were also investigated
with a view to determining the cause of the residual variation, (1-r2), not
explained by a simple two-parameter linear relationship between the
fluorescence and organic carbon parameters. Table 3, an array of linear
correlation coefficients for relationships between water quality parameters,
indicates the presence of a number of correlations significant at the 5-
percent level. In particular, correlations between DOC and either chlorophyll
a, dissolved oxygen or turbidity gave coefficients of -0.25, -0.23, and -0.26
respectively. Correlations between the fluorescence and water quality
parameters are given in Table 2. The most significant correlations were seen
to be those between Fmax/FR and either dissolved oxygen, pH or turbidity,
for which the coefficients were -0.54, -0.46 and -0.52 respectively.
Multiple linear regression analysis was applied to all data sets complete
in fluorescence and organic carbon parameters together with chlorophyll ~, pH,
dissol~ed oxygen, conductivity and turbidity. A total of 87 samples from the
Atchafalaya River Basin for which complete data sets existed was used in the
analysis. The multiple correlation analysis employed a stepwise regression
procedure for maximizing the multiple correlation coefficient (Dixon, 1975).
Multiple correlation coefficients for Fmax/R or FR/R versus the array of
water quality parameters, including TOC, DOC and POC, not surprisingly
indicated the principal water quality parameter to be DOC with very small
increments from' dissolved oxygen, chlorophyll a, turbidity and pH. For
example the coefficient for Fmax/R versus DOC is +0.773, increasing to
+0.787 by adding in. dissolved oxygen and to +0.791 by adding in chlorophyll a.
Only multiple correlations with Fmax/FR appeared to show any real
improvement over the two-parameter linear correlation model. The multiple
correlation coefficient for Fmax/FR versus DOC is +0.69, increasing to
+0.82 by adding in turbidity, to +0.85 by adding in dissolved oxygen, and to
+0.86 by adding in conductivity. However, the influence of turbidity can be
discounted as it has been shown to be caused by an instrumental artifact
whereby a leaked background light signal is scattered into the detector by
suspended sediment.
Clearly, the influence of these additional water quality parameters other
than turbidity on the relationships between DOC and the various fluorescence
parameters are not significant, suggesting that other reasons must be found to
explain why these correlations are less than ideal. This subject will be
examined in more detail in the discussion (Chapter 10).
59
-------
SECTION 9
DIFFERENCES BEfWEEN LASER FLUOROSENSOR AND SPECTROFLUOROMETER
MEASUREMENfS OF FR/R
Utilization of the fluorescence-organic carbon regression data to predict
TOC or DOC values from airborne fluorometric measurements assumes that the
fluorescence ratios measured using the laboratory spectrofluorometer, in
particular Fmax/R and FR/R, are. directly equivalent to those values
measured using an airborne laser fluorosensor. With a view to testing this
assumption, an experiment was conducted in which FR/R was measured for a .
representative group of 10 samples selected from the 161 samples of the survey
using both the spectrofluorometer and a laboratory laser fluoroserisor that
simulates the optical characteristics of an airborne system. FR/R was
chosen for the comparison rather than Fmax/R in order to avoid possible
interferences from differential absorption and scattering effects.
The simulated laser fluoro'sensor, shown schematically in Figure 17, used
a pulsed N2 laser excitation source, and was configured so as to monitor the
fluorescence and Raman emission in the 180° scattering mode; the 180°
scattering angle used in this laboratory simulation and the 0° scattering
angle used in airborne measurements for viewing the water Raman emission are
exactly equivalent (Gilson and Hendra, 1970). The effects of laser beam
polarization and emission monochromator polarization artifacts on the laser
fluorosensor measurements of FR/R were eliminated by depolarizing the laser
excitation beam and the fluorescence emission using compensated quartz-wedge
polarization scramblers, as shown in Figure 17. The pulsed emisSion from the
samples was dispersed through a scanning monochromator with 0.8 nm spectral
resolution and detected with a fast response photomultiplier (Hamamatsu R928).
Using a sampling oscilloscope in the non-scanning mode, the peak of the pulse
was converted into a steady signal level and used to produce stripchart
recordings of the emiss;-on spectra by scanning the monochromator from 350 nm
to 500 nm. By careful choice of electronic low-pass filter cutoff frequency
and monochromator scanning rate, a considerable enhancement in the
signal-to-noise ratio was achieved for spectra recorded in this manner. The
[OH]-stretch water Raman band for a high purity water sample recorded with
this simulated laser fluorosensor is shown in Figure 18. The width of this
band (FWHM) at 20°C was 6.4 nm, which is close to the value of 6.3 nm
estimated from the measurements of Walrafen (1967) made at 20°C.
The values of FR/R for the 10 samples obtained using both
spectrofluorometer (SPF) and laser fluorosensor (LFS) systems are plotted
against one another in Figure 19. The two sets of data are related by the
approximate relationship:
60
-------
1/6.8 SCANNING
MONOCHROMATOR
IGCA McPHERSON EU.7001
"
QUARTZ WEDGE
DEPOLARIZER
QUARTZ
CONDENSING
LENS
NZ LASER
20 KW 500 ppa 10naec.
I AVCO C-4001
2cmSAMPLE CElL
en
~
QUARTZ COLLECTOR
LENS
CA nON X
LIQUID LONG
WAVE PASS
FILTER
RED SENSITIVE
PMT IR.9281
LASER BEAM STOP
N2 LASERLINE FILTER
QUARTZ DIFFUSER
SILICON PIN DIODE
rn
2 GhZ SAMPLING
OSCILLOSCOPE
TUNABLE
LOW PASS
FIL TER
"CHART ~ECORDER
Figure 17. Optical layout for laboratory simulation of airborne laser fluorosensor
for monitoring surface water organics.
-------
RELATIVE INTENSITY
70
60
50
40
30
20
10
o
370
375
380
WAVELENGTH (om)
385
390
395
. .
Figure 18. Emission spectrum showing [OH]-stretch water Raman band
from high-purity water sample at 381 nm, excited
by N2 laser at 337 nm.
62
-------
2.8
2.4
(~lFS
LFS--LASER FLUOROSENSOR
SPF--SPECTROFLUOROMETER
2.0
1.6
1.2
0.8
0.4
o
1
3
4
2
.
.
.
(~)lFS
(~)SPF
5
6
Figure 19 ~ Vari ation of FR/R from simul ated 1 aser fl uorosensor
measurements with comparable values measured using a laboratory
spectrofluorometer for a variety of field samples.
63
-------
(FR/R)SPF . =
2.71 (FR/R)LFS
(8 )
The deviation of the individual data points from the'regression line, given by
Equation 8, was found to be reproducible for a given sample and is considered
to be due principally to scattering artifacts introduced into the measurement
of (FR/R)SPF by samp1e-to-samp1e variability in the concentration of
suspended particulates. As already discussed, suspended particulates scatter
an undispersed background signal, which is known to leak from the excitation
monochromator into the emission monochromator. This background signal
increases both the measured values of (FR)SPF and the calculated values of
(FR/R)SPF above their true values. In addition, sample particulates also
produce partial depolarization of the Raman emission, which, in conjunction
with the polarization sensitivity of the emission monochromator, can produce
changes in the measured values of (R)SPF and hence in the calculated values
of (FR/R)SPF.
A number of factors are known to be responsible for the large value of
the ratio of FR/R measured with the spectrofluorometer in relation to that
measured with the laser fluorosensor. These factors are discussed below and
wherever possible, correction factors are calculated with a view to reducing
this ratio toward unity: .
(1)
Spectrofluorometer Spectral Resolution
Although the spectrofluorometer has a lilniting spectral
resolution of 0.2 nm, signal-to-noise requirements dictate the use
of excitation and emission slit widths equivalent to 3 nm when
monitoring the relatively weak water fluorescence and Raman signals.
The true bandwidth at FWHM for the [OH]-stretch Raman band at 381 nm
when excited at 337 nm is 6.3 nm (Wa1rafen, 1967), whereas the
bandwidth measured using 3 nm-wide excitation and emission slits is
on the order of 7.2 nm (Figure 2). By progressively reducing the
slit width towards zero, the true Raman bandwidth of 6.2 nm was
obtained, togethe~ with a 30 percent increase in the Raman band
amplitude, but with a considerable reduction in the signa1-to-noise
ratio. Correction factors of 1.3 and 1.3-1 should therefore be
applied to the measured values of RSPF and (FR/R)SPF respectively.
(ii). Spectrofluorometer Risetime
Due to the combined effects ofspectrofluorometer wavelength
scanning rate, finite chart recorder slewing rate and electronic
Tow-pass filter bandwidth, a full scale (10 inches) chart deflection
signal for the Raman band is attenuated by about 5 percent. The
corresponding multiplicative correction factor is 1.05 falling to 1
as the signal amplitude is reduced to zero. Correction factors of up
to 1.05 and 1.05-1 should therefore be applied to the measured
values of (R)SPF and (FR/R)SPF respectively depending on the
relative amplitude of (R)SPF.
64
-------
(iii) Laser Fluorosensor Induced Sample Photodecomposition
Photodecompos iti on of the di sso 1 ved organics by the focused UV
laser beam was generally found to reduce (FR)lFS by about 5
percent duri ng the time taken to perform a spectral scan of a fresh
sample. A correction factor of 1.05 should therefore be applied to
the measured values of (R)LFS and (FR/R)lFS. Clearly, sample
photodecomposition effects will not occur during the airborne
operation of a laser fluorosensor as the water surface sample volume
would be exposed to only a single laser pulse of considerably lower
energy and power densities than exist for the laboratory laser
fluorosensormeasurements.
( i v)
laser Fluorosensor Fluorescence Decay Time
In contrast to the spectrofluorometer, which uses a constant
output excitation source, the laser is pulsed with pulsewidth on the
same order as the fluorescence lifetime of the dissolved o~ganics.
Consequently the fluorescence pulse observed from the water samples
will be stretched-out such that the peak amplitude is attenuated in
relation to that for the laser excitation pulse. As the Raman
emission process is essentially instantaneous, the Raman pulse
follows the profile of the laser excitation pulse so that its peak
amplitude is unattenuated. . To a first approximation, the effects of
fluorescence pulse-stretching on the amplitude of.(FR)lF5 can be
estimated by multiplying (FR)LFS and (FR/R)LFS by (lf/lL),
the ratio of the fluorescence to laser pulsewidths at FWHM. 1f .
was measured to be about 13 nsec for the majority of samples examined
and "l was about 10 nsec, so that (Tf/Tl) = 1.3. This
correction procedure is based on the assumption that both the.
fluoresence and Raman pulses are Gaussian in shape and concurrent in
time during the oscilloscope pulse height measurement.
(v)
PolariZation Artifacts
liquid water consists of two spectroscopically distinct forms in.
thermal equilibrium. These polymeric and monomeric types influence
the [OH] molecular bond strength and ultimately the nature of the
[OH]-stretch Raman emission (Chang and Young, 1974). In particular,
the spectral characteristics of the Raman emission are highly
dependent on. both the. polarization state of the excitation radiation
and on the scattering configuration used to view the emission (Gilson
and Hendra, 1970). The partial polarization effect introduced by the
diffraction grating monochromators and the 90°-scattering
configuration of the spectrofluorometer are therefore not compatible
with the format envisioned for an air~orne laser fluorosensor. An
airborne laser fluorosensor of necessity operates in a downlooking
mode, which consitutes a 0°-scattering angle, and will likely employ
an essentially unpolarized N2 laser as the excitation source.
Consequently, the combined effects of Raman band depolarization, the
polarization sensitivity of the excitation and emission
65
-------
monochromators and the gOo-scatteri ng confi gurati on employed by the
spectrofluorometer make it difficult to estimate an eQuivalent laser
fluorosensor val ue for FR/R from the val ues measured usi ng a
spectrofluorometer.
By combining all of the above correction factors into a single factor,
viz., (1.3 x 1.05 x 1.30 x 1.05)-1, the constant in Equation (3) can be
reduced from 2.71 to 1.45. It is suggested that the deviation of this new
ratio or conversion factor from unity is due principally to the polarization
artifacts known to influence the spectrofluorometer determination of R.
On the basis of the foregoing discussion, it is clear that the laboratory
spectrofluorometer and airborne laserfluorosensor determinations of FH/R are
not equivalent and that different laboratory and airborne systems will produce
different values for the conversion constant in Equation 8. In addition,
measurements of FR/R made on identical samples but using different
spectrofluorometers will also be different due to the different spectral,
temporal and polarization artifacts introduced by each instrument into the
measurement of R.
66
-------
SECTION 10
DISCUSSION
INHERENT LIMITA nONS IN THE FAIR-DOC CORRELA nON METHOD
Although a highly significant correlation has been found to exist between
FAIR and DOC, these regression data cannot be used to provide a precise
prediction of DOC concentration for other water bodies using airborne
measurements of FAIR because of the relatively wide scatter in the
distribution of DOC with FAIR, as illustrated by the data in Figures 7 and
8.
A number of these outlying data points can be explained on the basis of
poor experimental methodology. Six samples were found to have DOC values
greater than their corresponding TOC values and, in the three of these cases
where the differences were large, the DOC data points were among the 10 most
extreme outliers, whereas the corresponding data points for TOC versus
FmaxlR lay significantly closer to the regression line. This suggests the
existence of a problem with the DOC analyses, probably involving contamination
during filtration of the samples.
Problems with adsorption of COC (colloidal organic carbon) onto the glass
fiber filters during preparation of the DOC samples do not appear to have made
a significant contribution to this scatter. This can be concluded because of
the good DOC reproducibility obtained on 30 replicates of a single Lake Mead
sample, and also because the scatter present in the plot of DOC versus
FmaxlR is also present in the plot of TOC versus FmaxlR, although on a
somewhat smaller scale. In particular, six of the ten most extreme outliers
in Figure 7 with high DOC values correspond with six of the ten most extreme
outliers in Figure 8 with high TOC values.
The principal reasons for the less than ideal correlation between DOC and
a g; ven fl uorescence parameter, say FR/R, can be understood by represent i ng
each parameter as a summation of its constituent parts where these parts are
the product of the concentration of each individual substance and an
appropri ate conversi o'n factor. A general express i on for the concentrat i on of
DOC can be written in the form:
DOC =
q
i~l
(D,F )
ni bi
+
t
. ~ 1
1=q+
(O,NF )
ni bi
(9)
67
-------
where ni is the concentration of the ith organic chemical substance, bi
is a factor that converts ni to an equivalent carbon value, superscripts F
and NF denote whether the substance is respectively either fluorescent or
nonfluorescent, and D is a superscript denoting a dissolved substance. The
number of terms in the fluorescent DOC and nonf1uorescent DOC series,
respectively q and t-q, will depend critically on the type and pore size of
the filter used to separate the DOC fraction from the POC fraction. It is
likely that both q and (t-q) are significantly larger when using filters known
to be nonadsorbing rather than adsorbing for the purpose of separating the POC
and DOC fractions. Similarly, FR/R can be written as:
FR
R
= A
q
l:
i=1
(D,F )
ni a i +
s
l:
i=q+1
(P,F )
ni a i
(10 )
where'A is a proportional ity constant that accounts for system and envi ron-
mental factors such as attenuation coefficients, fields of view, system'
efficiencies, etc., ai is the excitation cross section for the ith
substance for fluorescence emission at Raman wavelength AR when excited at
wavelength AL' and P is a superscript denoting a particulate substance. As
the fluorescence measurements, whether measured from an airborne platform or
in the laboratory, are obtained on unfiltered samples, the division of this
expression into two parts representing the dissolved and particulate fractions
is arbitrary, and is made solely to aid in the comparison with the similar
expressi on for DOC. Consequently the di ssol ved fract i on is composed of the
contributions from q substances whereas the particulate fraction is composed
of the contributions from (s-q) substances where s is the total number of
fluorescent substances present in the sample, both particulate and dissolved.
It is assumed in the formulation of Equation 10, that the fluorescence from
inorganic f1uorophors is negligible. As noted earlier, the fluorescence of
water samples undergoes a large reduction after passage of the samples through
an activated carbon filter, indicating that the fluorescence orginates
predominantly from dissolved organics rather than from inorganic substances.
From an initial examination of these two equations, it is surprising that
a correlation of any significance exists between FR/R and DOC, particularly
as the conversion factors bi and 0i are considered to be unrelated. The
most 1 i ke1 y set of ci rcumstances that supports the observed corre1 ati on
between FR/R and DOC (r = 0.78) are that: ."
(a) the second (and presumable- larger) "tenT! in Equation 9, representing
the nonfluorescent DOC, varies in unison with the first term representing the
fluorescent DOC,
(b) the second term in Equation 10 representing the particulate
f1uorophors is significantly smaller than the first term representing the.
dissolved fluorophors,
and
68
-------
(c) the first term in Equation 10 varies as the first term in
Equation 9.
If all of these conditions are met, then Equations 9 and 10 can be combined,
and reduced to a form similar to Equation 5 where nf is equivalent to DOC.
It should be noted that for condition (c) to be valid, the conversion
factors biand . i must either remain essentially constant or vary in
unison, or alternatively the averaging effect of large values for ni in
combination with small values for bi and 0i for a large number of
substances must be such that the first term in Equation 9 varies closely as
the first tenn in Equation 10. These conditions are most likely to be met for
water within a given drainage basin or lake system where the percentage
contribution from each organic substance or group of substances remains nearly
constant. It is therefore significant that the samples corresponding to 6 of
the 10 extreme outliers in Figures 7 and 8 with high DOC and TOC values
respectively, were collected at sites in swampland regions of the Atchafa1aya
Basin outside the levees that define the f1oodway. Samples from these
regions, unlike those obtained from within the levees, were generally
characterized by relatively high levels of chlorophyll a and relatively low
turbidity. -
The data corresponding to the samples from Lakes Mead and Mohave,
apparent in Figures 7 and 8 as the discrete group of points at the low end of
the fluorescence scale, are somewhat more anomalous in that these samples
exhibit considerable variability in TOC and particularly DOC for essentially
constant fluorescence. In addition, a~l the outliers with high values of DOC
or TOC correspond to samples collected from Lake Mead, although the highest
DOC value in this group can probably be discounted as it exceeds the
corresponding TOC value. .
A 1 though there are no obvi ousexp1 anati ons for the vari abil ity in the.
Lake Mead data, particularly in view of the essentially constant fluorescence
data, several possibilities are discussed. Pleasure craft introduce oil and
gasoline residues onto the lake surface, but these substances can be
discounted as a reason for this variability as it i$ well established that
petroleum based hydrocarbons are highly fluorescent.
Lake Mead, being surrounded by desert low in vegetation and having low
annual rainfall, receives very little organic carbon through runoff from
natural vegetation... In addition, examination of the TOC and DOC data in
relation to the sampling sites suggests that Las Vegas Wash does not introduce
significant 1~ve1s of organic carbon into Lake Mead.
Another possible source fo~ this variability, particularly in the DOC
data, is the phytoplankton population. It is known that both marine and.
freshwater algae exude large amounts of DOC during photosynthesis, principally
in the fonn of carbohydrates, and release similar amounts on their death and
decomposition, and that this algae-related DOC is often a large fraction of
the total DOC level (Wangersky, 1972; 1978). Further amounts of DOC are also
likely to be produced by zooplankton grazing.
69
-------
Although planktonic chlorophyll ~ measurements were not made on the
samples from Lakes Mead and Mohave, it was noticed that the samples
corresponding to the six principal outliers in this group were collected at
sites that are possibly associated with anomalously high levels of chlorophyll
a in relation to nearby deep water sampling sites. Three sites, located close
to a cove, a recreational beach and an island, support significant populations
of benthic algae, whereas the other three sites were located close to anchored
navigation buoys that are known to support colonies of periphytonic algae.
If this explanation for the DOC variability for the Lake Mead samples is
to be accepted, then it is necessary to assume that this algae-related DOC is
essentially nonfluorescent, such that it makes a significant contribution to
the second term in Equation 9 without making a corresponding contribution to
Equation 10. Samples taken from other water bodies that are characterized b~
significant spatial variability in the phytoplankton population might then be
expected to exhibit similar variability in the distribution of DOC while
maintaining a nearly constant fluorescence emission signal.
This suggestion is highly tentative and is based on the unsubstantiated
assumption that algae-related DOC is nonfluorescent. Clearly an investigation
should be made to establish the fluorescence properties of the DOC fraction
associated with the phytoplankton community before any firm conclusions can be
reached.
In relation to this anomaly, it is worth noting that th'e chlorophyll .!
and DOC data for the Atchafalaya Basin samples did not exhibit a direct
relationship. The most likely explanation for the negative correlation.
(r = -0.25) is probably the presence of high levels of detrital organic matter
in many of the Mississippi Floodway samples that is closely related to the DOC
levels. This same detrital matter, manifest as high sample turbidity, .
severely limits light penetration, which in turn inhibits photosynthesis
resulting in depressed chlorophyll ~ levels. Under these circumstances it is
likely that tendencies towards a direct relationship between algae-related DOC
and chlorop.hyll a are being masked.
Finally, the possibility exists that the relatively high variability in
the DOC data (in relation to the TOC data) was caused by some shortcoming in
the DOC analysis, particularly, as has already been noted, the extreme outlier
for the Lake Mead data had a DOC value greater than the corresponding TOC
value. In spite of careful preparations, sample contamination can always
occur during either the handling, acidification, filtration or anaJysis of the
samples, or alternatively by sample leakage past the filter.
CALIBRATION OF AIRBORNE FR/R DATA IN TERMS OF DOC CONCENTRATION
The obvious inconsistencies between Equations 9 and 10 constitute an
inherent limitation to the determination of a universal relationship between
FR/R and DOC. This conclusion is not surprising in view of the wide
diversity of organic substances, both natural and manmade, that are present in
the aquatic environment. .
70
-------
It is suggested that the impact of these differences on the potential for
remote sensing organic carbon be minimized by treating each water body as a
separate entity with its own unique conversion factor relating FR/R to the
DOC concentration. This approach is justified on the assumption that, within
a given water body. the fluorescent organics change io concentration but not
in character. However, this can lead to ambiguities in the interpretation of
the airborne data whereby changes in FR/R may result from changes in DOC or.
alternatively. be caused by the presence of organics with anomalous
fluorescent properties.
In any event, such a situation, involving large or sudden changes in
FR/R over a short distance, would demand that the anomaly be subjected to
more intensive investigation by conventional sampling and analytical
techniques. Clearly the determination of a conversion factor for each water
body by means of an extensive ground truth survey is. not only time-consuming
and costly but negates any advantages to be gained from a subsequent remote
sensing survey. Rather, the airborne measurements for FR/R must be
calibrated directly in terms of DOC by collecting samples for DOC analysis at
a number of key reference sites under the sensor flight path concurrent with
the airborne mission. On receipt of this ground truth DOC data, it will then
be possible to calibrate the airborne survey data for use in preparing maps of
the water surface indicating lines of constant DOC concentration. Because of
the possibility of ambiguities in the interpretation of the data as mentioned
earlier. it is suggested that these maps be employed in a general purpose
screening role to indicate trends and anomalies in the distribution of surface
water organics that can then be used to investigate these anomalies in more
detail by direct sampling methods.
This direct calibration approach has its limitations; first, it requires
that a limited ground truth capability be made available for every airborne
mission, and second, this in si-tu calibration is still subject to the
influence of anomalous samples that can produce excessively weak or strong
fluorescence emission for a given DOC level.
However, these limitations are more than offset by three significant
advantages to be gained by direct calibration of the airborne data at the time
and place of a particular survey. First, as the calibration is obtained at
the same place and time as the survey, it corresponds exactly to the
particular distribution of organics present and to the environmental
conditions to which these organics are being subjected. Second, direct
calibration of the airborne fluorescence data in terms of DOC avoids the
requirement for performing laboratory fluorescence analyses on the reference
samples. This is advantageous because FAIR ;s known to be sensitive to
variations in sample temperature and pH, which will change between field and
1 aboratory, and because of the problems encountered in converting the F ~JR
data measured on a specific laboratory spectrofluorometer to values equlvalent
to those measured using a specific airborne laser fluorosensor. Finally,
direct calibration of the airborne data at the time of a specific survey
eliminates any limitations that might be incurred by using laboratory
calibration data. In particular, it now becomes possible to use different
excitation wavelengths~ subject of course to the availability of a laser with
the desired wavelength.
71
-------
A number of conflicting factors must be considered when selecting the
laser (excitation) wavelength in addition to laser performance parameters such
as the pulse energy, width, and repetition rate, and the beam divergence.
Reasons for choosing near UV excitation (in the region of 337 nm) are:
(a) 337 nm lies close to the optimum wavelength for exciting
fluorescence in the dissolved organics present in most fresh waters. The
maxima for the fluorescence emission spectra obtained on the 161 fresh water
samples of this survey with excitation at ~37 nm were located in a relatively
narrow (24-nm wide) band centered at 428 nm. With excitation shifted to the
blue region, the fluoresce~ce emission spectra will become severely attenuated
and truncated in relation to those produced by near UV excitation such as that
for the Lake Mead sample shown in Figure 2. This 'produces a marked shift in
Amax to longer wavelengths together with a marked reduction in Fmax.
(b) The intensity of the Raman emission exhibits a AL-4 dependence
on the excitation wavelength.
(c) Photomultipliers, which are used to detect the fluorescence and
Raman signals, are generally more sensitive in the near UV and blue spectral
regions.
On the other hand, several reasons exist for employing lasers that
operate in the blue spectral region. These are:
(a) Laser excitation in the blue region results in attenuation
coefficients for the laser, Raman and fluorescence wavelengths that have
values closer to the minimum for attenuation in natural waters than do the
coefficients for near UV excitation. Although these reductions in kL, kR
and kF are not expected to effect a significant change in FAIR, they will
increase FA andR, as predicted by, respectively, Equations 1 and 2. This,
in turn, results in an increase in the sensitivity with which FAIR can be
measured.
(b) By choice of a suitable blue wavelength, the Raman band can be
shifted to the peak of the fluorescence band. The separation of FR and R by
the process of linear interpolation using three spectral measurements is more
reliably performed when the narrow Raman band lies at the peak of the
relatively broad fluorescence band than if it were located on the side of the
fluoresce~ce band as occurs in Figure 2. In addition, the preferred
parameter, FRIR, becomes equivalent to FmaxlR under these circumstances.
(c) Visible rather than UV excitation results in reduced atmospheric
scattering losses to the laser; Raman and fluorescence signals.
(d) Visible rather than UV excitation generally tends to reduce the
background fluorescence induced in the optics and other materials of
construction used in the laser fluorosensor.
(e) By employing excitation in the blue spectral region, it is possible
to use the laser fluorosensorto monitor phytoplan~tonic chlorophyll ~ as well
72
-------
as DOC. rhe optimum wavelength for exciting chlorophyll ~ lies in the region
of 436 nm with emission of 685 nm. . Both fluorescence signals can then be
corrected for variations in optical attenuation using the same Raman signal.
IPROVEMENfS IN EXPERIMENTAL METHODOLOGY
In addition to the inherent limitations to the fluoresence organic carbon
Irrelation technique described above, it is likely that a number of
lortcomings in experimental methodology have tended to reduce the degree of
Irrelation between FR/R and DOC, particularly with regard to the DOC
lalyses. These factors together with possible corrective action are
scussed as follows:
( i i )
. (iii)
(i v)
(i)
. Use of (Whatman GF/F) glass fiber filters in the preparation of
the DOC samples may .lead to variability in the DOC data due to
adsorption of COCo It is suggested that this problem can be avoided
by using nonadsorbing O.S-um pore size silver membrane filters.
This expedient also has the advantage of including the potentially
fluorescent COC material in the DOC fraction. .
HCl added to the DOC and TOC samples for driving off inorganic
carbon (i.n form of C02) and arresting bacterial activity was seen
to introduce a carbon background on the order of 0.4 mg/l. It is
suggested that high purity phosphoric acid be used instead of HCl
(Gordon and Sutcliffe, 1973).
rhe polyethylene bottles used for transporting and storing the
fOC and DOC samples introduced ~ carbon background on the order of
0.05 mg/l per day. Delhez (1960) has noted that the UV.absorption .
of distilled water stored in polyethylene bottles increases \~ith
time, suggesting that soluble organics are being leached from the
plastic. Symons et al. (1975) have suggested that this problem be
avoided by using glass sample vials that are sealed with PTFE-lined
caps, and muffled prior to use at 400°C for at l~ast one hour to
drive off any organic matter that might interfere with the TOC or DOC
analyses. Alternatively, sample contamination can be avoided by
immediately sealing the samples in clean glass ampoules at the
collection site, particularly when the samples must be transported
over large distance~ or stored for extended periods of time prior to
analysis. The latter expedient will eliminate all possibility of
contamination prior to analysis and will also allow the volatile
organic fraction to be retained for those situations that allow
for their inclusion in the analysis.
A nonunifonn distribution of POC in the master sample prior to
partitioning into.subsamples for the TOC, DOC and fluorescence
analyses can produce errors in the TOC measurements. . Care must be
taken to ensure that the master sample is thoroughly stirred or
homogenized prior to subsampling.
73
-------
(vi)
(vii)
L
(v)
The reproducibility of the TOe and DOC determinations) presently
on the order of to.1 mg/l at the 1 mg/l level) can be improved to
to.OI mg/l by using TOC analyzers that employ I-ml sample injections
instead of the 10-~1 sample injections used by the TOe analyzer in
the present study. However) higher precision also demands that
extreme care be taken to avoid contamination of the samples or the
TOC analyzer.
Problems associated with contamination of the filters (glass
fiber or silver membrane) which can occur during manufacture or
during preparation of the DOC samples. should be avoided by
precombustion and careful handling.
Leakage of undispersed stray light from the excitation
monochromator is not a problem when examining the fluorescence of
clear liquid samples. However) when turbid samples are being
examined) a significant background signal is produced by scattering
from the suspended particulates. As the concentration of these
particulates is generally unrelated to the intensity of the sample
fluorescence) this background signal can make a marked and random
contribution to FA) particularly in the region close to the
excitation wavelength.. This is seen to produce a reduction in the
degree of correlation between FA/R and DOC) particularly at shorter
wavelengths; as this background is relatively broad, it does not
influence the measurement of R.
fhis background signal can be eliminated by inserting a narrow
passband filter into the excitation beam centered at the excitation
wavelength. Alternatively the intensity of this stray light can
be reduced by improved baffling within the excitation monochromator
or by using a double monochromator.
Elimination of this background signal by removal of the sample
particulates is, unfortunately, not practicable as the samples would
no longer be representative of those seen by an' airborne 1a~er
fluorosensor. First, filtration will remove fluorescent particulate
material from the s~mple and, setond) even if non-adsorbing silver
. membrane filters were used to avoid 10ss of potentially fluorescent
COC, the suspended colloidal material passing the filter would still
produce a scattered background signal.
OEIAILED PRESENTATION OF CONCLUSIONS
A number of important findings that have implications
operation of an airborne laser fluorosensor for monitoring
organics are presented together with a brief discussion of
involved.
in the design and
surface water
the reasoni ng
(i)
It has been shown that the fluorescence data must be normalized
using the concurrent water Raman intensity data in order to correct
74
-------
( i i )
(iii)
( i v)
for sample-to-sample variations in optical attenuation and to make
the fluorescence data independent of variations in spectrofluorometer
sensitivity. In an airborne laser fluorosensor application, this
data correction procedure becomes an absolute necessity because of
large variations in optical attenuation that occur from place to
place and also because of significant variations that can occur with
system sensitivity, particularly those involving laser output power.
Use of the concurrent water Raman return as a measure of the
variation in laser beam penetration is ideal because it requires no
information concerning the theoretical form of .the attenuation
coefficient, whether it be the beam attenuation coefficient, the
diffuse attenuation coefficient or some intermediate model. .Also,
the Raman emission emanates from essentially the same sample volume
as the fluorescence emission and therefore makes an accurate
correction for changes in the attenuation properties, particularly
when using FR/R as an indicator of DOC. Also, FAIR, unlike F~~
is independent of I/H2 variations in received signal intensity
produced by changes in H, the elevation of the sensor above the water
surface.
As already suggested, the airborne measurement of R can be used
to provide a continuous measurement of kA provided that the
airborne data are calibrated by making a number of concurrent ground
truth measurements of kA at reference sites under the aircraft
flight path. In addition, if a sufficiently wide range of k~
values is encountered, it might also be possible to determine whether
k~, the attenuation coefficient used in the laser fluorosensor
equations, is more appropriately described by the diffuse attenuation
coefficient or by the beam attenuation coefficient.
On the basis of the reported correlation data, the fluorescence
signal to Raman signal ratio, FAIR, appears to be the fluorescence
parameter offering the most promise as a remote sensing indicator of
waterborne organic carbon, specifically DOC. fhe significantly lower
correlations between F~/R and TOC appear to be due to the
variability in p~C concentration in relation to that for DOC in
combination with the fact that the p~C fraction appears to be less
fluorescent than the accompanying DOC. The ability to remotely
monitor DOC rather than TOC may be an advantage as the presence of
DOC in untreated sources of drinking water is currently generating
much concern. In the absence of specific procedures for their
removal, natural dissolved organics such as the humic and fulvic
acids are able to penetrate through a drinking water treatment plant
.and undergo conversion into potentially carcinogenic trihalomethanes
during routine chlorination (Youssefi et al., 1978).
Although Fmax/R exhibits the highest correlation with DOC
(r = 0.83), there are compelling reasons for preferring FR/R as the.
chosen remote sensing parameter for characterizing surface water
organic carbon. The correlation between FR/R and DOC (r = 0.78),
which is not significantly different from that involving Fmax/R at
75
-------
the 5 percent level, was shown to be lower due to the interference
from an instrumental artifact in which undispersed stray light that
leaks from the excitation monochromator is scattered by sample
particulates and becomes irretrievably mixed with the fluorescence
emission. The influence of this relatively broadband background was
most marked in the fluorescence data obtained in the near UV region,
which includes the "Raman band wavelength at 381 nm, but was
considered to be insignificant at the wavelength of maximum emission
close to 430 nm; this broadband background does not, however, affect
the measurement of R, the intensity for the relatively narrow Raman
emission band. It is likely that, in the absence of this near-UV
background, FR/R will exhibit the same degree of correlation with
DOC as does Fmax/R.
There are two significant advantages to using FR/R over
Fmax/R as a measure of surface water DOC in remote sensing
applications. First, because both FR and R are measured at the
same wavelength, interference by differential spectral absorption
and scattering does not occur. The same is not true for Fmax/R as
differential spectral effects between the measurements made at
AR and Amax are known to be present by virtue of the strong
correlatlon between Fmax/FR and DOC (r = 0.78). The second
advantage concerns the fact that, in an airborne laser fluorosensor
system employing discrete detector and recording channels for each
spectral measurement, the determination of FR/R requires only three
spectral detector chann~ls whereas that for Fmax/R requires four
such channels.
76
-------
. REFERENCES
Almgren, T., B. Joseffson and G. Nyquist. 1975. A Fluorescence Method for
Studies of Spent Sulfite Liquor and Humic Substances in Sea Water. Anal.
Chim. Acta, 78:411-422.
Andelman, J. B. and M. J. Suess. 1970. Polynuclear Aromatic Hydrocarbons in
the Water Environment. Bull. Wld. Hlth. Org. 43:479-508.
Austin, R. W. 1974. Problems in Measuring Turbidity as a Water Quality
Parameter. Proceedings of Seminar on Methodology for Monitoring the
Marine Environment, Seattle, Washington. October 16-18, 1973. Report
No. EPA-600/4-74-004, October 1974, Office of Monitoring Systems, U.S.
Environmental Protection Agency, Washington, D.C.
Averso, S. J. 1976.
Glass-Fiber Filter Paper. American Laboratory 8:97-104.
Baumgartner, D. J., M. Ii. Fel dman and C. L. Gi bbons. 1971 . A Procedure for
Tracing Kraft Mill Effluent fram an Ocean Outfall by Constituent
Fluorescence. Water Research, 5:533-544.
Black, A. P. and R. F. Christman. 1963. Characteristics of Colored Surface
Waters. Jour. A.W.W.A. 55:753-770.
Bowker, D. E. and W. G. Witte.
Parameters in the Coastal
on Satellite Applications
November 15-18, 1977.
Briggs, R., J. W. Schofield and P. A. Gorton. 1976. Instrumental Methods of
Monitoring Organic Pollution. Wat. Pollute Control, 75:47-57.
1977. The Use of LANDSAT for Monitoring Water
Zone. Proc. of AIAA 1977 Joint Conference
of Marine Oper~tions, New Orleans, Louisiana,
8ristow, M. P. F. 1979. Fluorescence of Short Wavelength Cutoff Filters.
Appl. Optics 18:952-955.
Bristow, M. P. F. 1978. Airborne Monitoring of Surface Water Pollutants by
Fluorescence Spectroscopy. Remote Sensing of Environment'7:105-127.
Bristow, M., D. Nielsen and R. Furtek. 1979. A Laser-Fluorosensor Technique
for Water Quality Assessment. Proceedings of the 13th International
Symposium on Remote Sensing of the Environment. Environmental Research
Institute of Michigan, Ann Arbor, Michigan. April 23-27, 1979, pp.
397-417. . .
77
-------
Browell, E. V. 1971. Analysis of Laser Fluorosensor Systems for Remote Algae
Detection and Quantification. NASA Technical Note TN 0-8447, NASA,
Washington, D.C., June 1971.
Buelow, R. W., J. K. Carswell and J. M. Symons. 1973.' An Improved Method for
Determining Organics by Activated Carbon Adsorption and Solvent
Extraction-Part I. Jour. A.W.W.A. 65:51-72.
Cahn, R. D.
1967.
Detergents in Membrane Filters.
Science, 155:195-196
Chandler, R. L., J. C. O'Shaughnessy and F. C. Blanc. 1916. Pollution
Monitoring with Total Organic Carbon Analysis. Jour. Water Poll. Control
Fed. 48:2803-2191.
Chang, C. H. and L. 'A. Young. 1914. Seawater Temperature Measurement from
Raman Spectra. AVCO Research Note 960, January, 1974. AVCO Everett
Research Laboratory, Inc., Everett, Massachusetts.
Christman, R. F. and R. A. Minear. 1967.
Sulfonates. Trend in Eng. 19:3-7.
Constable, T. W. and E. R. McBean. 1979. BOD/TOC Correlations and Their
Application to Water Quality Evaluation~ Water, Air and Soil Pollution.
11 :363-375.
Fluorometric Detection of Lignin
Davis, E. M. and W. J. Fosbury. 1973. Application of Selected Methods of
Remote Sensing for Detecting Carbonaceous Water Pollution. Canada
Centre for Inland Waters: Hamilton, Ontario. June 1973. Proceedings
#11: Remote Sensi'ng and Water Resources Management (Edited by K. P. B.
Thompson~ R. K. Lane and S. C. Csallany); American Water Resources
Association, Urbana, Illinois, 1973. '
Del hez, R~ 1960. Ultraviolet Absorption Spectrum of Water Stored in
Polyethylene Bottles. Chemist-Analyst 49:20-21.
DeMent, J. and H. C. Dake. 1942. Fluorescent Chemicals and their
Applications. Chemical Publishing Co., Inc. New York.
Dienert, F. 1910. De la recherche des substances fluorescent dans le
controle de la sterilisation des eaux. Compt. Rend. Acad. Sci. (Paris)
150 : 487 -488. .
Dishman, R. A. ,1979. Automation of Laboratory Organic Carbon Analysis. Am.
Lab. 11:97-111. .
Dixon, W. J. 1975. Biomedical Computer Programs (BMDP).
California Press, Berkeley.
Dobbs, R. A., R. H~ Wise and R.B. Dean. 1972. The Use of Ultra-Violet
Absorbance for Monitoring the Total Organic Carbon Content of Water and
Wastewater. Water Research 6:1173-1180.
Un; versity of
78
-------
Duurmsa, E. K. 1974. The Fluorescence of Dissolved Organic ~atter in the
Sea. Chapter 11 in Optical Aspects of Oceanography. Edited by N. G.
Jer10v and E. Steeman Nielsen. Acadei"ic Press, New York. .
Fantasia, J. F. and H. C. Ingrao. 1974. Development of an Experimental
Airborne laser Remote Sensing System for the Detection and Classification
of Oil Spills. Proceedings of the 9th International Symposium on Remote
Sensing of the Environment, Environmental Research Institute of Michigan,
Ann Arbor, Michigan, April 15-19, 1974, pp. 1711-1745.
Farmer, F. H., C. A. Brown, O. Jarrett, J. W. Campbell and W. Staton. 1979.
Remote Sensing of Phytoplankton D~nsity and Diversity in Narragansett Bay
using an Airborne F1uorosensor. Proceedings of the 13th International
Symposium on Remote Sensing of the Environment. Environmental Research
Institute of Michigan, Ann Arbor, Michigan, April 23~27, 1979. pp.
1793-1805. .
Friedman, E. J. and Hickman, G. D. 1972. laser Induced Fluorescence in
Algae: A New Technique for Remote Detection. Final Report, October
1972. Sponsored by NASA Wallops Island under Contract No. NAS 6-2081.
Sparcom, Inc., Alexandria, Virginia.
Gasith, A. and A. D. Hasler. 1976. Airborne Litterfall as a Source of
Organic Matter in Lakes. Limno1. Oceanogr. 21:25~-258.
Gilson, T. R. and P. J. Hendra. 1970.
Wi1ey-Interscience, New York.
Laser Raman Spectrometry.
Gordon, D. C. and W. H. Sutcliffe. 1973. A New Dry Combustion Method for the
Simultaneous Determination of Total Organic Carbon and Nitrogen in
Seawater. Marine Chemistry, 1:231-244.
Gordon, D. D. and W. H. Sutcliffe. 1974. Filtration of Seawater using Silver
Filters for Particulate Nitrogen and Carbon Analysis. Limnol.Oceanogr.
19:989-993.
Goulden, P. D. 1976. Automated Determination of Carbon in Natural Waters.
Water Research. 10:487-490.
Gui11ard, R. R. L. and P. J. Wangersky. 1958. The Production of
Extracellular Carbohydrates by Some Marine flagellates. Limnol.
Oceanogr. 3:449-454.
Hall, K. J. and G. F. lee. .1974. Molecular Size and Spectral. .
Characterization of Organic Matter in a Meromictic lake. Water Research,
8:239-251.
Huro, K., T. Tanaka and A. Kawahara. 1977. Measurement of Organic Pollutants
in Coastal Sea Water by Fluorescence Method. Bunseki Kagaku 26:574-575.
In Japanese. (Translation obtained from F. E. Hoge, NASA Wallops,
Wallops Island, Virginia)
79
-------
Hwang, C. P., T. H. Lockie and R. R. Munch. 1979. Correction for Total
Organic Carbon, Nitrate, and Chemical Oxygen Demand when Using the
MF-Millipore Filter. Environ. Sci. Te-chnol. 13:871-872.
Jerlov, N. G. 1976. Marine Optics, Elsevier/North Holland Inc., New York.
Johnson, R. W., I. W. Duedall~ R. M., Glasgow, J. R. Proni and T. A. Nelsen.
1977. Quantitative Mapping of Suspended Solids in Wastewater Sludge
Plumes in the New York Bi~ht Apex. Jour. Water Poll. Control Fed.
49:2063-2073.
Johnson, R. W. 1978. Mapping of Chlorophyll .! Distributions in Coastal
Zones. Photogrammetric Engineering and Remote Sensing. 44:617-624.
Kalle, K. 1966. The Problem of Gelbstoff in the Sea. Oceanogr. Mar. Bio1.
Ann. Rev. 4:91-104. -
Kehoe, T. J. 1977.- Detennining toc in Waters. Environ. Sci. Technology
11: 137-139.
Kerfoot, W. B. and E. C. Brainard. 1978. Septic Tank Leachate Detection. A
Technological Breakthrough for Shoreline On-lot System Perfonmance
Evaluation. In: State of Knowledge in Land Treatment of Wastewater.
Edited by H. L. McKim. International Symposium at the Cold Regions
Research and Engineering Laboratory, Hanover, New Hampshire, pp. 325-334.
Kim, H. H. and G. D. Hickman. 1975. An Airborne Laser Fluorosensor for the
Detection of Oil on Water. - In: The Use of Lasers for Hydrographic
Studies, NASA Wallops Island, Virginia, September 12, i973. NASA Special
Publication SP 375, Washington, D.C. pp. 197-202. - -
Lakshman, G. 1975. Monitoring Agricultural Pollution Using Natural
Fl uorescence. Water Resources Research 11: 705-708. -
Lammers, W. T. 1971. Insoluble Material in Natural Waters. In: - Water and
Water Pollution Handbook, Vol. -2. Edited by L. ~. Ciaccio. r~arce1 Dekker
Inc., New York. pp. 593-635. - -
Levesque, M. 1972. Fluorescenc~ and Gel Filtration of Humic Compounds.
Science 113:346-353. -
Soil
Lock, M. A., P. M. Wallis and H. B. N. Hynes.
in Running Waters. Oikos, 29~1~4.
MacKinnon, M. D. 1979. The Measurement of the Volatile Organic Fraction of
TOC in Seawater. Marine Chemistry 8:143-162. -
1977. Colloidal Organic Carbon
Maier, W. J. and W. R. Swain. 1978. Organic Carbon - A Non Specific Water
Quality Indicator for Lake Superior. Water Research 12:523-529.
Mattson, J. 5., C. A. Smith, T. T. Jones and S. M. Gerchakov. 1974.
Contfnuous Monitorfngof Dissolved Organic Matter by UV-Visible
Photometry. Limnol. Oceanogr. 19:530-535.
80
-------
McLean, R. A. and M. L. Speas. 1946. Fluorescence of Drinking Waters as
Evidence of Sewage Pollution. Jour. A.W.W.A., 38:355-360.
Meakin, J. C. and M. C. Pratt. 1972. Whatman Manual of Laboratory
Filtration. W. and R. Ba1ston Ltd., Maidstone, England.
Measures, R. M., J. Garlick, W. R. Houston and D. G. Stephenson. 1975.
Induced Spectral Signatures of Relevance to Environmental Sensing.
Remote Sensing Journal, 1:95-102.
Michalek, S. A., R. A. Dishman and R. B. Thayer. 1977. Capabilities of
On-Line TOC a"nd TOO Analyzers. Progr. Wat. Tech. 9:87-92. Pergamon
Press, Oxford.
Laser
Can.
Miller, J. R., S. C. Jain, N. T. OINeill, W. R. McNeil and K. P. B. Thompson.
1977. Interpretation of Airborne Spectral Reflectance Measurenents over
Georgian Bay. Remote Sensing of Environment 5:183-200.
Montalvo, J. G. and C. G. Lee. 1976. Analytical Notes - Removal of Organics
from Water: Evaluating Activated Carbon. Jour. A.W.W.A. 68:211-215.
Morel, A. and L. Prieur. 1977. Analysis of Variations in Ocean Color.
Limno1. Oceanogr. 22:709-722.
Nriagu, J. O. and R. D. Coker. 1980. Trace Metals in Humic and Fu1vic Acids
from Lake Ontario Sediments. Env. Sc;. Technol. 4:443-446.
Ogner, G. and M. Schnitzer. 1970. Humic Substances:
Phthalate Complexes and their Role in Pollution.
Fu1vic Acid-Dia1ky1
Science, 170:317-318.
OINeil, R., L. Buja-Bijunas, and D. M. Rayner, 1980. Field performance of a
1aserfluoresensor for the detection of oil spills. App1 Opt. ~:863-870.
01Nei1, R. A., A. R. Davis, H. G. Gross and J. Kruus. 1975. A Remote Sensing
Laser Fluorometer. Water Quality Parameters. ASTM STP 573, American
Society for Testing Materials. pp. 424-436.
Parker, B. C. 1967. Influence of Method for Removal of Seston on the
Dissolved Organic Matter. J. Phyco1., 3:166-173.
Porro, T. J., R. E. Anacreon, P. S. Flandreau and 1. S. Fagerson. 1973.
Corrected Fluorescence Spec~ra of Polynuclear Materials and their
Principal Applications. Journal of the A.O.A.C. 56:607-620. .
Radley, J. A. and J. Grant. 1935. Fluorescence Analysis in Ultra-Violet
Light. Chapman Hall Ltd., London.
Reiswig, H. M. 1972. The Spectrum of Particulate Organic Matter of
Shallow-Bottom Boundary Waters of Jamaica. Limnol.Oceanogr.,
17:341-348.
R i 1 ey, G . A .
8:1-118.
1970.
Particulate Organic Matter in Sea Water. Adv. Mar. Bio1.
81
-------
Rook, J. J. 1977. Chlorination Reactions of Fulvic ~cids in Natural Watars.
Env. Sci. Technol. 11:478-482.
Salonen, K. 1979. Comparison of Oifferent Glass Fibre and Silver Metal
Filters for the Determination of Particulate Organic Carbon.
Hydrobiologia 67:29-32.
Schnitzer, M. 1971. Metal-Organic Matter Interactions in Soils and ~aters.
Chapter 13. In: Organic Compounds in Aquatic Environments (Edited by
S. J. Faust and J. V. Hunter). pp. 297-313. Marcel Dekker, New York.
. .
Seal, B. K., K. B. Roy and S. K. Mukherjee. 1964. Fluorescence Emission
Spectra and Structure of Humic and Fulvic Acids. Jour. Indian Chem. Soc.
41:212-214.
Sharp, J. H. 1973. Size Classes of Organic Carbon in Seawater.
Oceanogr., 18:441-447.
limno 1 .
Sharp, J. H. 1974. Improved Analysis for "Particulate" Organic Carbon and
Nitrogen from Seawater. Limnol. Oceanogr. 19:984-989.
Sheldon, R. W. 1972. Size Separation of Marine Seston by Membrane and
Glass-Fiber Filters. Limnol. Oceanogr. 17:494-498.
Smart, P. L., B. L. Finlayson, W. D. Rylands and C. M. Ball. 1976. The
Relation of Fluorescence to Dissolved Organic Carbon in Surface Waters.
Water Research, 10:805-811.
Sontheimer, H. 1977. Samples of Rhine Water Reveal Organic Carbon Content.
Water and Wastes Engineering 14:24-27.
Stevens, A. A. and J. M. Symons. 1974. Measurement of Organics in Drinking
. Water. Proceedings A.W.W.A. Water Quality Technology Conference. pp.
xxiii-1-xxiii-2S, December 2-3, 1973, Cincinnati, Ohio. A.W.W.A.,
Denver,Colorado.
Sydor, M., K. R. Stortz and \L R. Swain. 1978. Identification of
Contaminants in Lake Superior through Landsat I Data. J. Great Lakes
Res. 4:142-148.
Sylvia, A. E. 1973. Detection and Measurement of Microorganics in Drinking
Water. Jour. New England Water Works Assoc. 87:183-199.
Sylvia, A. £0, D. A. Bancroft and J. D. Miller. 1974. Detection and
Measurement of f4ic roorganics in Dri nki ng \4ater by Fl uorescence. . In Proc.
A.W.W.A. Water Quality Technology Conference, Dallas, Texas, December
2-3, 1974.
Sylvia, A. E., D. A. Bancroft and J. D. Miller. 1977. Analytical Note-A
Method for Evaluating Granular Activated Carbon Adsorption Efficiency.
Jour. A.W.W.A., 69:99-102.
82
-------
Symons, J. M., T. A. Bellar, J. K. Carswell, J. DeMarco, K. L. Kropp, G. G.
Robeck, D. R. Seeger, C. J. Slocum, B. L. Smith and A. A. Stevens. 1975.
. National Organics Reconnaissance Survey for Halogenated Organics. Jour.
A.W.W.A~, 67:634-647. .
Tardiff, R. G. 1977. Health Effects of Organics: Risk and Hazard Assessment
of Ingested Chlorofonm. Jour. A.W.W.A. 69:658-661.
Thruston, A. D. 1970. A Fluorometric Method' for the Detenmination of Lignin
Sulfonates in Natural Waters. J. Wat. Pollute Control Fed. 42:1551-1555.
Trussell, R. R. and M. D. Umphres.
Jour. A.W.W.A. 70:604-612.
1978.
The Formation of Trihalomethanes.
Wakeham, S. G. 1977. Synchronous Fluorescence Spectroscopy and its
Application to Indigenous and Petroleum-Derived Hydrocarbons in
Lacustrine Sediments. Environ. Sci. Technology, 11:272-276.
Walden, C. C. 1976. The Toxicity of Pulp and Paper Mill Effluents and
Corresponding Measurement Procedures. Water Research 10:639-664.
Walrafen, G. E. 1967. Raman Spectral Studies of the Effects of Temperature
on Water Structure. J. Chern. Phys. 47:114-126.
Wangersky, P. J. 1972. The Cycle of Organic Carbon in Sea Water. Chimia,
26:559-564.
Wangersky, P. J. 1975. Measurement of Organic Carbon in Seawater. In:
Analytical Methods in Oceanography. Edited by T.R.P. Gibb, Jr. Advances
in Chemistry Series, 147:148-162. American Chemical Society, Washington,
D.C.
Wangersky, P. J. 1978. Production of Dissolved Organic Matter. In: Marine
Ecology, Volume IV, Dynamics, Edited by Otto Kinne. Wiley Intersc;ence,
New York. .
Wilander, A., H. Kvarnas and T. Lindell. 1974. A Modified Fluorometric
Method for Measurement of Lignin Sulfonates and its .ill situ application
in natural waters. Water Research, 8:1037-1045.
. .
Wilson, W. H., R. W. Austin and R. C. Smith. 1978. Optical Remote Sensing of
Chlorophyll in Ocean Waters. Proceedings of the 12th International
Symposium on Remote Sensing of the Environment, April 20-25, 1978,
Manila, Philippines. Environmental Research Institute of Michigan,
Ann Arbor, Michigan. . .
Youssefi, M., S. T. Zenchelsky and S. D. Faust. 1978. Chlorination of
Naturally Occuring Organic Compounds in Water. J. Environ. Sci. Health
AI3:629-637.
Zimmenman, A. and A. R. Bandy. 1975. The Application of Laser Raman
Scattering to Remote Sensing of Salinity and Turbidity. Final Technical
Report under NASA Contract NASI-11D7, August 1975.
83
-------
TECHNICAL REPORT DATA
(Please read !nsl7Uctions on the reverse belore com;iaing)
1. REPORT NO. 12. !3. RECIP!ENT'S '>'CC:::5SION NO.
EPA-600/4-8l-00l
4. ilTLE AND SUBTITLE 5. REPORT DATE
REMO"IE MONITORING OF ORGANIC CARBON IN r- 1 QAl
SURFACE WAIERS 6. PERFORMING ORGANIZATION CODE
7. AUTHORIS) 8. PERFORMING ORGANIZATION REPORT NO.
t4i chae 1 Sri stow and David Nielsen
9. PERFORMING ORGANIZATION NAME AND ADDRESS 10. PROGRAM ELEMENT NO.
Envi ronmenta 1 Monitoring Systems Laboratory
Office of Research and Development 11. CONTRACT/GRANT NO.
U.S. Environmental Protection Agency
Las Vegas, Nevada 89114
12. SPONSORING AGENCY NAME AND ADDRESS 13. TYPE OF REPORT AND PERIOD COVERED
U.S. Envi ronmenta 1 Protection Agency-Las Vegas, NV
Office of Research and Development 14. SPONSORING AGENCY CODE
Environmental Monitoring Systems Laboratory
Las Vegas, NV 89114 EPA/600/07
15. SUPPLEMENTARY NOTES
16. ABSTRACT
rhi s study shows that the intensity of the Raman normalized fluorescence emission
induced in surface waters by ultraviolet radiation can be used to provide a unique
remote sensing capability for airborne monitoring the concentration of dissolved
organic carbon (DOC). Trace concentrations of hydrocarbons, both manmade and natural
in orgin, are the predominant source for this fluorescence. Water, on the other hand,
is nonfluorescent under UV irradiation, but emits an intense Raman band of constant
amplitude relative to the incident 1 i ght. This Raman emission can be used as an
internal reference or normalizing standard with which to correct the fluorescence
emission for the effects of attenuation, for variations in system sensitivity, and for
'changes in sensor elevation. It is suggested that a direct calibration of the airborne
fluorescence data in terms of equivalent DOC concentration be accomplished by making
DOC measurements on samples obtained at a small number of reference sites under the
aircraft flight path at the time of the airborne survey.
Airborne laser fluorosensors that utilize this principle will provide a synoptic
survey capability for rapidly and cost-effectively producing isopleth maps that' show
concentrations of surface water DOC. These isopleths can be used for delineating
gradients, temporal changes and anomalies in the distribution of total dissolved
organics in the surface layers of rivers, lakes and costal waters.
-
17,., KEY WORDS AND DOCUMENT ANAL YSIS
a. DESCRIPTORS b.IDENTIFIERS/OPEN ENDED TERMS 1:. CDSA TI Field/Group
hydrocarbons laser fluorosensing 48G
surface water airborne platforms 63C
ultraviolet irradiation. 680
water po 11 ut ion
18. DISTRIBUTION STATEMENT' 119. SECURITY CLASS (This Reporr) 21. !\IO. OF PAGES
! UNCLASSIFIED i 93
I
RELEASE ro PUBL IC !
I i 20. SECURITY CLASS iT/lis pl1gej 122. PRICE
I
UNCLASSIFIED
EP,t, i=orm 2220-1 (Rey, 4-i7)
PREVIOUS EOITION IS OBSOLETE
------- |