the influence of land-based activities on …the influence of land-based activities on contaminants...
TRANSCRIPT
THE INFLUENCE OF LAND-BASED ACTIVITIES ON CONTAMINANTS IN RIVERINE
SYSTEMS
by
Rebecca J. Fauver
(Under the direction of James T. Peterson)
ABSTRACT
River water throughout Georgia was analyzed for organic contaminants including organochlorine
(OC) chemicals (OC pesticides and polychlorinated biphenyls) and perfluorinated compounds
(PFCs) over two years during low and high stream-flow periods. Concentrations for OC
chemicals and PFCs ranged from 0 – 114 ng/L and 0 – 47 ng/L, respectively. Chemical
concentrations were related to land-use and environmental variables with hierarchical linear
models. Urban land cover influenced PFC concentrations in the river while agricultural land-use
influenced OC chemical concentrations in the river water. Samples near an urban area or with a
high percentage of urban in the riparian area of the stream had higher concentrations of PFCs.
Run-off from row crop agricultural fields was a strong source for legacy-use OC chemicals as
OC concentrations increased in the water column with an increase in turbidity and river
discharge in areas with a high percentage of row crop agriculture within riparian area of the
stream.
INDEX WORDS: organic contaminants, rivers, urban, agriculture, land-use, modeling,
PFCs, PFOA, PFOS, PCBs, OC pesticides
THE INFLUENCE OF LAND-BASED ACTIVITIES ON CONTAMINANTS IN RIVERINE
SYSTEMS
by
Rebecca Fauver
B.S., The Ohio State University, 2006
A Thesis Submitted to the Graduate Faculty of The University of Georgia in Partial
Fulfillment of the Requirements for the Degree
MASTER OF SCIENCE
ATHENS, GEORGIA
2008
© 2008
Rebecca J. Fauver
All Rights Reserved
THE INFLUENCE OF LAND-BASED ACTIVITIES ON CONTAMINANTS IN RIVERINE
SYSTEMS
by
REBECCA JANE FAUVER
Major Professor: James Peterson Committee: Nathan Nibbelink
Aaron Fisk Electronic Version Approved: Maureen Grasso Dean of the Graduate School The University of Georgia December 2008
iv
ACKNOWLEDGEMENTS
I would like to acknowledge my major professor, Jim Peterson, for his guidance throughout my
career at the University of Georgia. I would like to thank my committee members, Aaron Fisk,
for motivating me, getting me started with field and laboratory work, and keeping me going, and
Nate Nibbelink for imparting his spatial wisdom to me. Thanks to Gregg Tomy for all his help,
knowledge, and wisdom in and outside of academia. I would like to thank my family for
providing me with the love and support that has enabled me to pursue my education. Thanks to
everyone who helped me in the field: Jason Meador, Scott Craven, Colin Shea, Alison Price,
Matt Mundy, and Will Bickerstaff, and to those who helped me in the lab: Stephanie Verkoeyen,
Jaclyn Brush, Brett Ziter, Sandra Ellis, Kerri Pleskach and Bruno Rosenberg. Thanks to the
Great Lakes Institute for Environmental Research at the University of Windsor and the
Department of Fisheries and Oceans, Freshwater Institute at the University of Manitoba for
providing me with field supplies and equipment as well as laboratory space, supplies, and
guidance every step of the way. Thanks to Julie Wilson and Shannon Albeke who helped me
problem solve and learn tools and techniques in GIS. Thanks also to various students, staff, and
faculty who have helped me along the way.
v
TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS...............................................................................................................iv
CHAPTER
1 INTRODUCTION AND LITERATURE REVIEW .................................................1
REFERENCES ..........................................................................................................4
2 THE INFLUENCE OF LAND-USE ON RIVERINE DELIVERY OF
PERFLUORINATED COMPOUNDS TO ESTUARINE SYSTEMS .....................6
INTRODUCTION .....................................................................................................7
METHODS ................................................................................................................8
RESULTS ..................................................................................................................16
DISCUSSION............................................................................................................19
LITERATURE CITED ..............................................................................................39
3 THE INFLUENCE OF LAND-USE ON ORGANOCHLORINE CHEMICAL
DELIVERY TO ESTUARINE SYSTEMS...............................................................43
INTRODUCTION .....................................................................................................44
METHODS ................................................................................................................45
RESULTS ..................................................................................................................53
DISCUSSION............................................................................................................57
LITERATURE CITED ..............................................................................................77
4 CONCLUSIONS.........................................................................................................80
vi
LITERATURE CITED ..............................................................................................83
APPENDIX A
SUPPLEMENTAL TABLES FOR CHAPTER 3 ...................................................85
LITERATURE CITED ............................................................................................89
1
CHAPTER 1
INTRODUCTION AND LITERATURE REVIEW
The world, US and Georgia’s human populations are increasing rapidly [1]. From 1990-
2000 the population of Georgia increased by 26.4%, twice the national increase of 13.2% [1].
An increase in population results in an increase in the need for water resources in agriculture,
industry and municipalities, and can cause many unintentional, negative impacts to aquatic
systems. Thus, it is necessary to monitor human-inhabited watersheds for anthropogenic
contaminants to understand the impact on aquatic ecosystems.
Human-populated freshwater and estuarine watersheds often contain a wide range of
human-generated contaminants ranging from pesticides and pharmaceuticals to plasticizers and
flame retardants [2, 3]. These contaminants can persist in the environment and some may have
toxicological effects on aquatic life [4]. Many of these contaminants are not currently measured
in standard contaminant monitoring programs, but they have been found in a number of streams
throughout the U.S. [2, 4-6]. Additionally, the majority of these chemicals currently do not have
established maximum allowable levels in natural waters [2].
Contaminants found in aquatic environments can often be linked to the land-use in that
watershed [4, 5]. A wide range of contaminants enter rivers through surface run-off or industrial
and municipal wastewater treatment plant discharges, and if persistent, can be transported
significant distances, eventually reaching estuaries at the mouth of the rivers. Upstream
pollution, including point and non-point sources, tends to be the primary contributor to estuary
contamination [4, 7]. Therefore human activity within an estuarine watershed likely influences
2
the types and concentrations of anthropogenic contaminants found in the estuary [8]. Highly
urbanized watersheds are likely to deliver greater amounts of industrial contaminants such as
perfluorinated compounds while predominantly agricultural watersheds are likely to deliver both
legacy and current-use pesticides. However, organic contaminants are not only derived from
within the watershed. Long and short range delivery via atmospheric transport accounts for a
portion of contaminant delivery as well [9-12]. The types and concentrations of contaminants
found in riverine systems are likely to be influenced by changes in stream discharge due to water
use and development as well as seasonal discharge patterns. Knowledge of the fate and transport
of anthropogenic contaminants is necessary to determine the potential sources of contamination
to estuarine ecosystems and to develop best management strategies to reduce the impact of
anthropogenic contaminants to the aquatic environment.
A need exists for contaminant monitoring in the Southeastern United States due the
historical high use of legacy organochlorine chemicals and increased use of present-day
industrial, municipal, and agricultural chemicals in this region. In addition, there is a lack of
historical monitoring for anthropogenic contaminants in the environment in the Southeast. This
study will increase our understanding of contaminant fate in the environment by modeling
contaminant delivery through the Altamaha and Ogeechee Rivers in southeast Georgia, USA.
The Altamaha River basin has been affected by industrial and urban growth, while the Ogeechee
River basin is dominated by agriculture. We determined concentrations for legacy-use chemicals
such as polychlorinated biphenyls (PCBs) and organochlorine (OC) pesticides such as DDT as
well as current municipal and industrial chemicals, perfluorinated compounds (PFCs) at various
locations within each watershed. Models were developed relating contaminant concentrations to
landscape and environmental variables, such as water temperature, river discharge and land-use.
3
The objectives of this study were to: 1) identify the types and concentrations of organic
contaminants present in the Altamaha and Ogeechee River watersheds and where they enter
estuaries; 2) identify the likely sources of these contaminants as related to land-uses; and 3)
determine the influence of seasonal stream-flow patterns on the types and concentrations of
contaminants delivered to the estuarine systems.
Perfluorinated compounds differ from OC chemicals in that PFCs are currently used
today, thus PFCs have direct sources while legacy OC chemicals in the environment are likely
because of historical use and therefore indirect sources. In addition, their properties differ in that
PFCs are not lipophillic, unlike OC chemicals, and therefore their behavior in the environment
and impact on wildlife is different. PFCs are a group of fully fluorinated organic chemicals that
have unique properties which allow them to repel both oil and water [13]. The carbon – fluorine
bond is very strong and makes these chemicals resistant to degradation in the environment [14].
PFCs are currently used in commercial and industrial applications and likely enter the
environment through production and use in urban environments. Their chemical stability may
allow long-range transport in aquatic ecosystems. Chapter 2 will focus on PFCs, including their
presence and concentrations in Georgia rivers and their relationship with landscape and
environmental variables. OC chemicals are legacy-use compounds including OC pesticides and
PCBs bind to soils and sediments, are stable and persistent in the environment, and are known to
bioaccumulate in organisms and biomagnify through food webs. Little is known about the
presence and concentrations of OC chemicals in Georgia despite their historical use in this area
[15]. Chapter 3 will focus on OC chemicals, including their spatial distribution and
concentrations and their relationship with environmental and landscape variables coupled with
their physical-chemical properties.
4
LITERATURE CITED
1. Bureau, U. C., Global Population Profile: 2002. In International Population Reports: 2002.
2. Kolpin, D. W.; Furlong, E. T.; Meyer, M. T.; Thurman, E. M.; Zaugg, S. D.; Barber, L. B.; Buxton, H. T., Pharmaceuticals, hormones, and other organic wastewater contaminants in US streams, 1999-2000: A national reconnaissance. Environmental Science & Technology 2002, 36, (6), 1202-1211.
3. De Wit, C. A., An overview of brominated flame retardants in the environment. Chemosphere 2002, 46, 583-624.
4. Iannuzzi, T. J.; Armstrong, T. N.; Thelen, J. B.; Ludwig, D. F.; Firstenberg, C. E., Characterization of chemical contamination in shallow-water estuarine habitats of an industrialized river. Part 1: Organic compounds. Soil & Sediment Contamination 2005, 14, (1), 13-33.
5. Oros, D. R.; Jarman, W. M.; Lowe, T.; David, N.; Lowe, S.; Davis, J. A., Surveillance for previously unmonitored organic contaminants in the San Francisco Estuary. Marine Pollution Bulletin 2003, 46, (9), 1102-1110.
6. Ternes, T. A.; Joss, A.; Siegrist, H., Scrutinizing pharmaceuticals and personal care products in wastewater treatment. Environmental Science & Technology 2004, 38, (20), 392a-399a.
7. Armstrong, T. N.; Iannuzzi, T. J.; Thelen, J. B.; Ludwig, D. F.; Firstenberg, C. E., Characterization of chemical contamination in shallow-water estuarine habitats of an industrialized river. Part II. Metals. Soil & Sediment Contamination 2005, 14, (1), 35-52.
8. King, R. S.; Beaman, J. R.; Whigham, D. F.; Hines, A. H.; Baker, M. E.; Weller, D. E., Watershed land use is strongly linked to PCBs in white perch in Chesapeake Bay subestuaries. Environmental Science & Technology 2004, 38, (24), 6546-6552.
9. Bidleman, T. F.; Leone, A. D., Soil-air exchange of organochlorine pesticides in the Southern United States. Environmental Pollution 2004, 128, (1-2), 49-57.
10. Harner, T.; Wideman, J. L.; Jantunen, L. M. M.; Bidleman, T. F.; Parkhurst, M. J., Residues of organochlorine pesticides in Alabama soils. Environmental Pollution 1999, 106, (3), 323-332.
11. Hung, H.; Blanchard, P.; Halsall, C. J.; Bidleman, T. F.; Stern, G. A.; Fellin, P.; Muir, D. C. G.; Barrie, L. A.; Jantunen, L. M.; Helm, P. A.; Ma, J.; Konoplev, A., Temporal and spatial variabilities of atmospheric polychlorinated biphenyls (PCBs), organochlorine (OC) pesticides and polycyclic aromatic hydrocarbons (PAHs) in the Canadian Arctic: Results from a decade of monitoring. Science of the Total Environment 2005, 342, (1-3), 119-144.
5
12. Shen, L.; Wania, F.; Lei, Y. D.; Teixeira, C.; Muir, D. C. G.; Bidleman, T. F., Atmospheric distribution and long-range transport behavior of organochlorine pesticides in north America. Environmental Science & Technology 2005, 39, (2), 409-420.
13. Kissa, E., Fluorinated Surfactants and Repellents. 2nd ed.; Marcel Dekker: New York, 2001.
14. Banks, R. E., B.E. Snartm J.C. Tatlow, Organofluorine Chemistry Principles and Commercial Applications. Plenum Press: New York, 1994.
15. Moore, J. W., S. Ramamoorthy, Organic Chemicals in Natural Waters - Applied Monitoring and Impact Assessment. Springer-Verlag: New York, 1984.
6
CHAPTER 2
THE INFLUENCE OF LAND-USE ON RIVERINE DELIVERY OF PERFLUORINATED
COMPOUNDS TO ESTUARINE SYSTEMS1
1 The influence of land-use on riverine delivery of perfluorinated compounds to estuarine
systems. Fauver, R.J., A.T. Fisk, N.P. Nibbelink, G. Tomy, K. Pleskach, J.T. Peterson. To be
submitted to Environmental Science and Technology
7
INTRODUCTION
Perfluorinated compounds (PFCs) are a group of fully fluorinated organic molecules with
a unique set of properties that allow them to repel both oil and water which make them desirable
for commercial and industrial purposes [1]. PFCs are commonly found in surface protection for
carpets, upholstery, fabrics, and fast food wrappers due to their repulsion properties, and they are
also used as industrial surfactants and firefighting foams [1]. The strength of the fluorine–carbon
bonds causes this group of chemicals to be very stable and persistent in the environment [2].
Their widespread use and persistence in the environment has lead to their detection in abiotic and
biotic samples from throughout the world, including remote areas such as the Arctic [3-5]. PFCs
are known to bioaccumulate and biomagnify through food webs [5-7] and protective aqueous
concentration values for perfluorooctane sulfonate (PFOS) have been established for aquatic
species and resident piscivorous birds [8]. Determined toxic concentrations of PFOS from
laboratory toxicity tests for aquatic species are typically higher than observed PFC
concentrations in the environment [9], but their persistence and bioaccumulation potential make
them a continuing concern.
Perfluorinated compounds can enter aquatic systems directly during production and
industrial and municipal use or via atmospheric deposition. As well, PFCs entering wastewater
treatment plants from industry or municipalities are not removed and thus are released into the
environment with waste water discharge [10]. PFCs have been detected in aqueous samples
around manufacturing plants [11], high use areas [12], and non-spill surface waters in the US
[13-15]. However, monitoring efforts are typically conducted at one point in time, revealing a
snapshot of the PFC contamination at a given site. PFCs are persistent in the environment, and
are able to dissolve in water, therefore they are likely to be present in rivers and be delivered
8
downstream with little degradation [2]. PFC contamination is likely to vary at the same site
under seasonally varying environmental conditions, such as water temperature or river discharge.
In addition, concentrations of PFCs are likely to vary among sites as use and disposal of these
compounds are concentrated in urbanized areas which are localized points within the watershed.
Anthropogenic contaminants in aquatic systems are associated with land-based activities
[16-18]. Despite the varying land-use patterns in Georgia, there have been no comprehensive
studies on the types and concentrations of contaminants in its surface waters or an assessment of
the influence of stream flow or other environmental variables. Additionally, there have been no
studies of the delivery of riverine contaminants to coastal estuaries in Georgia. Therefore, the
objectives of our research were threefold: 1) to quantify seasonal patterns and concentrations of
PFCs in riverine systems; 2) to determine the impact of land-based activities and seasonal
variation on PFC concentrations in riverine systems; and 3) to develop predictive models using
the land-use and environmental variables collected.
METHODS
Chemicals and standards
The suite of native and mass-labeled PFCs and their nomenclatures used in this study
(Table 2.1) were obtained from Wellington Laboratories (Guelph, ON, Canada) with the
exception of 13C2-perfluoronanoic acid (PFNA) and 18O2-perfluorooctane sulfonate (PFOS),
which were a gift from Sheryl Tittlermier (Health Canada, Ottawa, ON). Optima-grade methanol
and water used in this research were purchased from Fisher Scientific.
9
Study sites and sample collection
Twelve sampling locations were chosen along two Atlantic Slope river systems in
Georgia, USA; the Altamaha (and its major tributaries) and the Ogeechee Rivers (Figure 2.1).
Sampling locations were chosen by stratified random sampling. Strata were determined by
dominant land-use and distance to the estuary. Points along the river were ranked by percent
row crop agriculture and percent urban area and were put in strata of high and low agriculture
and urban land-use, mixed land-use, and near and far distance from the estuary. Ten points were
randomly chosen as sampling locations from these strata and two additional sites upstream of the
salt-wedge on each river were systematically chosen to determine freshwater contaminant
contributions to estuaries and to avoid any potential interactions with physical chemical
properties of the contaminants to salinity (Figure 2.1). Stream flows in both river systems vary
substantially by season, with flows generally greatest during winter and spring months due to the
high average precipitation in Georgia from January through March, whereas they are generally
lowest during late summer and fall due to the low average precipitation from September through
November [19]. Water samples were collected during low and high flow seasons over two years
to account for variation due to time and river discharge.
Water samples were collected at each of twelve sites in the Altamaha and Ogeechee
watersheds within a four week period, hereafter known as sample period, during fall 2006, spring
2007, fall 2007, and spring 2008 (n=3 samples/site/season). Two liters of water were collected
for each of 3 replicate samples by dipping clean polypropylene sampling bottles into the river
approximately 20 cm below the surface. Field blanks consisted of two liters of Optima-grade
water and were prepared by pouring water into the sample bottles in the field. Water samples
(including blanks) were spiked in the field with 10 µL of a recovery internal standard (RIS,
10
Table 1), packed on ice and transported to the Fisheries Laboratory at the University of Georgia
where they were stored at 4°C until extraction (within two weeks of collection). Water
temperature was measured in the field at the time of sampling and river discharge was calculated
from the nearest USGS gauging station for the day on which sampling took place.
Contaminant extraction and instrument analysis
The targeted PFCs were extracted from water through Oasis hydrophilic-lipophilic-
balance (HLB; 20 ml, 1 g, 60 um) solid-phase extraction cartridges (Waters Corporation,
Milford, MA, USA) [20, 21]. The HLB cartridges were pre-treated with 5 mL of Optima grade
methanol, and then sample water was simultaneously filtered (1.0 µm glass fiber, Pall
Corporation, East Hills, NY) and pumped through the cartridge by a peristaltic pump (flow rate,
25 mL/min). Cartridges were processed by eluting first with 5 mL of Optima grade water which
was discarded, followed by 15 mL of Optima grade methanol which was collected and reduced
in volume (500 µL) by a gentle stream of nitrogen and centrifuged (2,000 rpm for 10 minutes)
[12]. A 200 µL sample was removed and fortified with instrument performance internal standard
(IPIS, 2 µL of 1 ng/µL solution, Table 1) and analyzed by liquid chromatography tandem mass
spectrometry (LC/MS/MS).
An Agilent 1100 Series high-performance liquid chromatography system (Agilent
Technologies, Palo Alto, CA, USA) equipped with a vacuum degasser, binary pump,
autosampler, and a Discovery C18 analytical column (length, 5.0 cm; inner diameter, 2.1 mm;
particle size, 5µm; Supelco, Oakville, ON, Canada) were used for all separations and analysis.
Water and methanol were used for the mobile-phase at a flow rate of 300 µL/min, and a sample
injection volume of 3 µL. A gradient began with 20% methanol, increased to 95% in 9.5 min,
11
and held for 2 min [12]. The mobile-phase composition returned to starting conditions in 5 min.
The column was allowed to equilibrate for 5 min between runs. The detection of the
perfluorinated compounds was performed with a Sciex API 2000 triple-quadrupole mass
spectrometer (MDS Sciex, Concord, ON Canada) in the negative-ion electrospray mode using
multiple-reaction monitoring. The optimized parameters were as follows: Ionspray voltage, -
1,200 V; curtain gas flow, 15.00 arbitrary units (a.u.); sheath gas flow, 30.00 a.u.; turbo gas flow,
35.00 a.u.; temperature 525°C; focusing potential, -360 V; and collision-assisted dissociation gas
flow, 8 a.u. The reactions monitored are provided in Table 1.
Quality Assurance/Quality Control
Inherent problems exist in quantifying PFCs by LC/MS/MS in environmental samples.
High background signals of perfluorinated compounds from solvents used, carryover between
injections, and lack of appropriate isotopically labeled standards have been well documented in
the literature [4, 22]. To account for these difficulties, two types of blanks were used in this
study, an instrument blank and a method blank. Instrument blanks were injections of methanol
run after every five samples and were used to determine contamination from the LC/MS/MS
system and any possible carryover contamination [12]. Method blanks consisted of Optima-
grade water and were extracted along with the field samples and were used to monitor
contamination occurring during the extraction procedure [12].
Ion signals of PFOA were detected consistently in all of our blanks and the intensity of
the signal was similar between instrument and method blanks, suggesting contamination during
extraction was less important than contamination from the instrument itself. The background
signal of PFOA could be reduced 10-fold by reducing the column equilibration time between
12
sample injections. For all other PFCs, method blanks always had higher signals than instrument
blanks, indicating that contamination during extraction and work-up was more substantial.
The recoveries were (mean ± standard error, n = 6) of 13C2-perfluodecanoic acid (PFDA),
13C4-PFOA, 13C5-PFNA, and 13C4-PFOS in the samples were 54.2% ± 15.0%, 40.1% ± 7.5%,
79.2% ± 11.9%, 78.1% ± 14.1%, respectively. PFCs were blank corrected by subtracting the
signal from extraction blanks from the sample signals. Native PFCs in the samples were recovery
corrected based on the recovery of the labeled surrogate with the nearest retention time (Table 1).
Method detection limits were determined by known amounts of PFOS and PFOA spiked into the
method blanks that were analyzed previously and found to have non-detectable concentrations of
PFCs (i.e. the response of PFCs was not greater than the instrument blanks). Separate injections
of the spiked extracts then were made, and the ion signals obtained for each PFC were adjusted
to estimate concentrations that would give a signal to noise ration of 5:1. In this manner, method
detection limits of PFOA (0.6 ng/L), PFNA (0.7 ng/L), PFDA (0.2 ng/L), perfluorododecanoic
acid (PFDoDA, 0.08 ng/L), and PFOS (0.6 ng/L) were estimated based on a 2-L sample.
Spatial data
The watersheds above each sampling location were delineated with ArcHydro in ArcGIS
9.1 using a 30-m digital elevation model (DEM) raster obtained from the National Elevation
Dataset [23] and the Altamaha and Ogeechee river system hydrography from the USGS National
Hydrography Dataset [24].
PFCs are associated with industrial and commercial uses [1] and therefore, various
metrics of urban land-use and industrial and municipal point source discharges were determined
for each sampling location. Urban and wetland land-uses were the landscape variables of
13
primary interest for this analysis. The percentage of urban land-use within each watershed was
calculated by combining low and high urban land cover using Landsat-derived land cover data
[25]. Percent of urban and wetland land cover types also were determined for a 1 km buffer on
each side of the stream to obtain riparian land-use, defined as riparian urban or riparian wetland.
Municipal and industrial discharge locations in each watershed were totaled to determine the
number of point source discharges upstream of each sampling site [26]. Additionally, stream
distance was measured from each sampling location to the nearest high density urban area (city
population > 10,000) and the nearest industrial or municipal discharge in order to assess the
influence of proximity to point source on PFC concentrations.
Statistical Analysis
Many watershed level and site specific variables were considered for use in explaining
patterns in perfluorinated compounds (PFCs). To avoid biased parameter estimates and standard
errors caused by multicolinearity, only uncorrelated predictor variables (r2 < 0.30) were included
in the candidate models.
The relationship between the concentration of PFCs in the water column and
environmental variables were initially evaluated using linear regression. However, multiple
samples and chemicals collected at a site during different sample periods were likely to be
autocorrelated as were multiple samples collected during a single sample period. Similarly,
concentrations are likely to vary among the types of PFCs. Therefore, a global linear regression
model containing all predictor variables was initially fit (Table 2) and an analysis of variance
(ANOVA) on the residuals was conducted. The ANOVA indicated substantial autocorrelation
among chemicals (F = 246.38; df = 9, 22; P <0.0001), among sample periods (F = 3.34; df = 3,
14
22; P = 0.02), and among sites (F=0.1.49; df = 11, 22; P = 0.13). To account for the
autocorrelation, hierarchical models were used to examine the relationship between
environmental variables and the concentration of PFCs in the water column. Hierarchical
models differ from more familiar regression techniques in that autocorrelation is incorporated by
including random effects [27]. For this study, the random effects associated with site, sample
period, and chemical represent the differences in concentration between sites, between sample
periods, or between types of chemicals, respectively, unaccounted for by covariates.
An information-theoretic approach [28] was used to evaluate the relative plausibility of
the candidate relating the concentration of PFCs in the water column to landscape and
environmental variables. Each candidate model represented hypotheses regarding the
concentrations of PFCs found in the water column deriving from either near or diffuse sources.
A global model with all predictors was initially constructed and 10 candidate models were
constructed from the global model (Table 2.2). To assess relative plausibility of each candidate
model, Akaike’s Information Criterion (AIC) [29] with the small-sample bias adjustment (AICc)
[30] was calculated. AICc is an entropy-based measure used to compare candidate models for the
same data [28], with the best fitting model having the lowest AICc. AICc penalizes a model for
complexity due to the number of parameters included. The number of parameters for each model
was calculated by summing the number of fixed and random effects, plus an error term. The
relative plausibility of each candidate model was assessed by calculating Akaike weights as
described in Burnham and Anderson [28]. These weights range from 0 to 1 with the most
plausible model having the highest weight. A confidence set of models, analogous to a
confidence interval for a parameter estimate, was reported instead of basing all inferences on the
single best approximating model. The ratio of Akaike weights for two candidate models can be
15
used to assess the degree of evidence for one model over another [31]. The confidence set of
models included only those candidate models with Akaike weights that were within 10% of the
highest weight, similar to the 1/8 rule of thumb suggested by Royall [32]. Only those parameter
estimates in the confidence set of models were interpreted. The precision of estimates was
determined by calculating 95% confidence intervals, an estimate with a confidence interval
overlapping zero was considered imprecise. Prior to model selection, the variance was
partitioned by fitting a random effects ANOVA with the site, chemical and sample period
random effects to determine where the variation was originating. Goodness-of-fit of each
candidate model was determined by examining a normal probability plot of the residuals. All
hierarchical and random effects models were fit in SAS PROC MIXED with the maximum
likelihood (ML) estimation specified (SAS Institute 2004).
Prior to evaluating the fit of the candidate models, the relative fit of several variance
structures were evaluated for the hierarchical model random effects using the global (all
predictors) model. The first set of variance structures modeled each random effect as varying
among sampling periods, the second among sites, the third among sites and sample periods. Each
model also included an additional random effect associated with each PFC type to account for
the overall differences in concentrations among chemicals. AICc was used to assess the relative
fit of each error structure. The best approximating error structure then was used during the
evaluation of the relative plausibility of the candidate models. An ANOVA of the residuals for
the best fitting error structure also was conducted to determine if autocorrelation had been
accounted for by the error structure.
16
RESULTS
Environmental data
River discharges during the study were below long term averages for each season due to
a long-term drought, but the pattern mirrored the normal pattern of low discharge in the summer
and fall and high discharge in the winter and spring (Figure 2.2). The observed discharges in the
Altamaha watershed ranged from 2 – 59 m3/s in the fall with the lowest discharge in the
Ohoopee River and greatest discharge in the Oconee River and 39 – 403 m3/s in the spring with
the lowest discharge in the Ohoopee River and greatest discharge in the Altamaha River (Table
2.3). The average fall discharge in the Ogeechee River was 4.0 m3/s and average spring
discharge was 91.7 m3/s (Table 2.3). The water temperature was typically greater in the fall and
lower in the spring for all rivers except the Ohoopee River where the average water temperature
was 18.8 °C in the fall and 19.9 °C in the spring (Table 2.3). The calculated landscape metrics
varied by river. The Ocmulgee River had the highest percent urban land-use in the whole
watershed (14%) as well as the one km riparian area (3%; Table 2.4). The Ocmulgee River site
had the shortest stream distance to the nearest high density urban area (8 km), followed by the
Oconee River (21 km; Table 2.4). The Altamaha and Ogeechee Rivers each had an average of
over 200 km to the nearest urban area while the Ohoopee River did not have any high density
urban areas in the watershed (Table 2.4).
PFC Concentrations
Nine perfluorinated compounds were analyzed for in water ranging from below detection
limits to 47.22 ng/L (Table 2.5); with PFOS and PFOA detected at each sample location and
having the highest concentrations among all PFCs. The average concentrations for PFOS and
17
PFOA were greater in the fall (0.98 – 15.35 ng/L for PFOS and 0 – 15.99 ng/L for PFOA) than
in the spring (0.18 – 6.28 ng/L for PFOS and 0.70 – 11.96 ng/L for PFOA; Table 2.5). PFNA,
PFDA, PFUA, PFDoDA, and PFOSA were found in the Altamaha River watershed during both
fall and spring samples. PFNA and PFDA were detected in low concentrations in fall samples in
the Ogeechee River but were not detected in the spring samples. Downstream sites, closer to the
estuary, had higher concentrations than upstream sites in both the Altamaha and Ogeechee River
watersheds. The greatest PFC concentrations were detected in the sites that were closest to high
density urban areas: the Ocmulgee River site and the downstream Ogeechee River sites.
Statistical Analysis
Plots of residuals from the global model relating PFC concentrations to environmental
variables indicated that the residuals were non-normal and heteroscedastic. To remedy the
problem, PFC concentrations were log10 transformed and the models were refit using the
transformed data. An examination of the residuals from the global model with the transformed
data indicated no departures from normality therefore all models were fit with the transformed
data.
The best approximating variance structure for the global model relating PFC
concentration to watershed level and site specific environmental predictor variables included an
intercept that randomly varied by chemical, and additional random effects corresponding to site
and sample period additively. The ANOVA of the residuals from this model indicated no
detectable dependence between chemicals (F = 0.01; df = 9, 22; P = 1.0), sample periods (F =
0.05; df = 3, 22; P = 0.98), and sites (F = 0.62; df = 11, 22; P = 0.81). However, heterogeneity in
error variance among chemical types was detected. The problem was unable to be remedied by
18
fitting a heterogeneous variance model [33]. Heterogeneity in error variance can bias
comparisons of variance components, but has little effect on fixed effect parameter estimates
[34]. Therefore, the best approximating error structure described above was used with all
candidate models in model selection and restricted our inferences to the fixed effects parameter
estimates.
The most plausible model of PFC concentrations contained distance to nearest urban
area, distance to point sources, riparian urban, and riparian wetland (Table 2.6). This model was
two times more likely than the next best model, which contained all variables in the best model
plus river discharge and water temperature. The confidence set of models included these two, a
model containing distance to urban, river discharge, water temperature, and a distance to urban
by river discharge interaction, and a model containing distance to urban, distance to nearest point
source, riparian urban, river discharge, and water temperature (Table 2.6).
Distance to the nearest high density urban area was negatively related to PFC
concentrations across all models in the confidence set (Table 2.7; Figure 2.3). Estimates from
the best model indicate that concentrations for the various PFCs vary widely and were related to
distance to nearest urban area (Figure 2.4). PFOA had the greatest concentration at near high
density urban sites, 2.6 ng/L, and steadily decreased to 0.28 ng/L as the distance to a high density
urban area increased (Figure 2.4). PFOS concentrations were the next greatest at 0.70 ng/L and
exhibited a similar pattern of decreasing to 0.07 ng/L as distance to high density urban increased
(Figure 2.4). All other PFCs were near zero at all distances to high density urban areas (Figure
2.4). Each model including riparian urban land-use indicated a positive relationship with water
column PFC concentrations (Table 2.7). As the percent of urban land-use in the riparian area
increased, PFOAconcentrations increased steadily from 0.58 to 12.6 ng/L and were higher than
19
PFOS concentrationswhich ranged from 0.16 to 3.14 ng/L (Figure 2.5). All other PFC
concentrations were near zero at all riparian urban percentages (Figure 2.5). The Akaike
importance weights for distance to high density urban, distance to point source discharge,
riparian urban, and riparian wetland were almost twice as great as the next most important
parameters (river discharge and water temperature) indicating strong support for the former
(Table 2.8).
DISCUSSION
Perfluorinated compounds were detected in surface waters at all sites sampled in
southeast Georgia; PFOA had the greatest overall prevalence followed by PFOS while all other
compounds were lower, if detected. PFOA and PFOS are the terminal degradation products for
higher carbon chain length PFCs but also have specific uses, which likely explain their greater
overall prevalence [35]. Average PFOS and PFOA concentrations that we measured were
comparable or lower than average concentrations reported for surface waters at non-spill
locations in Michigan and New York (range 2-5 ng/L; maximum 29 ng/L), the Great Lakes (15-
70 ng/L), and North Carolina (1.14-200 ng/L) [13-15]. We observed substantial seasonal
variation of PFC concentrations with PFOA concentrations in the fall (October) averaging 6.7
ng/L compared those in the spring (March), 2.7 ng/L. The distinct variation in concentration
between seasons suggest that properly characterizing PFC concentrations at a single location
would require multiple samples collected at different times of the year. We also observed large
differences in PFC concentrations among locations with site-specific average values ranging
from 0.7 ng/L to 16.0 ng/L. Greater PFC concentrations occurred in sample locations that were
near an urban area or that contained a relatively high concentration of urban land-use within the
20
watershed. The complex spatial and temporal variation in PFC concentrations suggests that
delivery of PFCs to streams is related to many factors. In order to more thoroughly understand
the whole picture and potential ecological effect of PFCs, we need to know the concentrations of
PFCs and how these concentrations change through time.
Urban land cover within the riparian zone and distance from a high density urban area
were the variables with the greatest support in the contaminant models, suggesting that urban
centers were the most important land-based source for the concentration of PFCs in the river
system. Urban land-use is a probable source for PFCs because these compounds have industrial
and municipal uses, such as surfactants and surface repellents, and thus, they are likely to enter
the stream via urban centers [1]. In particular, the amount of urban land cover near streams
within the watershed was more important for PFC concentrations than diffuse or whole
watershed urban sources of PFCs. There was little support for the number of discharge pipes in
the watershed, suggesting that overall industrial and municipal discharges were not the strongest
sources of PFC concentrations in riverine systems. We hypothesized that industrial and
municipal waste water treatment plants (WWTPs) were a source for PFCs to aquatic systems.
PFCs are likely to enter WWTPs because they are commonly used in household products (e.g.,
stain guards for carpets and upholsteries) and are not degraded in WWTPs [10, 36, 37]. The lack
of a strong relationship between industrial and municipal discharges and PFC concentrations
may be because our indirect measurements of discharge (total number and distance to nearest
industrial and municipal discharge location) were not sufficient to measure potential for
contaminant input. A better measure of inputs from WWTPs might be the actual volume of
waste water discharged, or the proportion of the WWTP discharge to that of the river. Municipal
and industrial WWTPs are not required to report their actual daily discharge to regulatory
21
agencies, therefore actual WWTP discharge data does not exist for this region. While our study
found urban centers to be a strong predictor of PFC concentration in the rivers, measuring both
constituents and volume of discharge from WWTPs will be essential for determining their role as
a PFC source to surface waters.
The downstream delivery of PFCs depends both on proximity to the source and on how
the chemicals partition in the environment. PFCs are persistent in aquatic environments but their
concentrations decreased with the distance to an urban area. The persistence of perfluorinated
chemicals may allow them to be transported downstream, but we hypothesize that concentrations
decrease downstream due to binding to sediment [38, 39], uptake by biota [5, 7, 40-42], or
dilution, which may occur as tributaries from non-source areas enter the river. In the current
state of manufacturing, use, and disposal of PFCs, streams within 100 river km of an urban
center are of the most concern for these chemicals. Ecologically sensitive areas that are near a
high density urban area should receive the most attention. For instance, the Ogeechee River
estuary in our study is within 100 river km of an urban center and concentrations of the
carboxylic acids PFOA, PFNA and PFDA were measured in greater concentrations in the salt-
wedge of this river (16.0 ng/L, 3.9 ng/L, and 5.5 ng/L, respectively) than in the Altamaha River’s
salt-wedge where there are no high density urban areas within 300 river km (concentrations,
PFOA = 7.4 ng/L, PFNA = 3.2 ng/L, and PFDA was not detected). Furthermore, we observed a
longitudinal gradient of PFOA concentrations increasing downstream, as concentrations in the
Ogeechee River increased from 0.7 ng/L upstream to 16.0 ng/L downstream near the estuary.
Conversely, we observed a pattern of decreasing concentrations downstream for PFOS. The
pattern of increasing PFOA and PFOS concentrations were observed downstream of suspected
PFC sources in previous studies where PFOA increased steadily but PFOS increased at a lesser
22
rate [11] or eventually decreased [12] further downstream. PFOS in the water column may
decrease downstream as it binds to sediment with a high organic carbon content [39], which may
explain the decrease in concentration downstream; however the steady downstream increase in
PFOA suggests sorption to sediments for this chemical is unlikely. Although PFOS
concentrations tend to decrease longitudinally and, in general, PFC concentrations decrease
downstream from an urban center, the potential for riverine delivery of sediment-bound PFCs
exists during a flood event where sediment is transported downstream.
Implications for estuaries
Estuaries near a high density urban area, such as a port city, should be considered a
higher priority for monitoring for PFCs since these chemicals are derived primarily from urban
sources. We found higher concentrations of the fluorinated carboxylic acids in the downstream
Ogeechee River site than the less urbanized downstream Altamaha River site. However, it
appears that PFOA may gradually increase in concentration downstream, so even distant
upstream sources such as the urbanization around Macon, GA may have an impact on the PFOA
concentrations found in the estuary. Upstream sources may influence the concentration of PFCs
in the estuary during high discharge events, delivering PFCs that are in the water column or
bound to soils and sediment. Overall, it appears that distant upstream sources do not currently
influence PFC concentrations in the estuaries. However if PFCs are continually manufactured
and used, but not degraded in the wastewater treatment process, the impact of distant sources on
PFCs found in the estuaries may be greater in the future. To control PFC contamination to
surface waters, including estuaries, PFCs must be broken down before they enter the stream.
Presently, PFCs are not degraded in wastewater treatment plants, therefore including an
23
additional phase designed specifically to degrade these (and other persistent compounds) may
reduce PFC contaminants from entering the stream and transporting downriver to the estuaries.
We were unable to assess the ecological significance of PFC concentrations to wildlife
because adverse affects to organisms exposed to low concentrations of PFCs for long durations,
such as those observed in nature, have not been assessed. Perfluorinated compounds may cause
adverse ecological affects through long-term, ambient exposure in the water or other exposure
pathways. PFOS and PFOA have been determined to bind to blood plasma proteins [43], and
laboratory toxicity tests found that fish exposed to high concentrations of PFOS (exposure, 21 d;
concentrations, 0.03 – 1 mg/L) and PFOA (exposure, 39 d; concentrations, 0.3 – 100 mg/L)
exhibited a decrease in reproduction [9, 44]. More work is needed to determine adverse affects
on aquatic biota from long-term exposure to low concentrations of PFOA and PFOS, or exposure
to a mixture of PFCs during sensitive life-stages.
This study showed significant seasonal variation in PFC concentrations indicating that
synchronization of sampling for PFCs with potentially affected wildlife, especially during life-
stages, is critical for a comprehensive understanding of the ecological impacts of PFC
contamination. In addition, it is important to understand the pathway of PFC exposure to biota to
understand how to monitor a system for potential adverse affects to wildlife. In our study, we
determined the concentrations of PFCs in the water column, however PFCs bound to suspended
sediment or bed sediment may also be a likely pathway for exposure. By understanding where
chemicals are located in the system, we may gain a better understanding of how these chemicals
will behave in the ecosystem, enter the food web and what types of organisms they are likely to
affect. Sediment may also be a likely pathway for exposure, and should be investigated. Prior to
this work, little was known about the presence of these chemicals in Georgia’s coastal rivers.
24
We have shown how an understanding of the spatial distribution of PFCs and their relationship
to land use characteristics has helped us to focus on likely sources and begin to characterize
relevant spatial and temporal scales at which to improve our future understanding of ecosystem
level effects of PFC contamination. Future work on where these chemicals occur in the system
will further aid our understanding of their behavior in the ecosystem, of where they enter the
food web, and the types of organisms they are likely to affect.
25
Table 2.1. List of native and labeled perfluorinated compounds used as recovery standards in this study. Reactions monitored are given in parenthesis.
Native PFC analyzed Recovery Internal Standard (RIS)
Labeled Instrument Performance Internal Standard
(IPIS) Perfluorooctanoic acid (PFOA) (413/ 369), (413/ 169)
13C4 – PFOA (417/ 372), (417/ 169)
13C2 – PFOA (415/ 370), (415/ 169)
Perfluorooctanesulfonate (PFOS) (499/ 99), (499/ 80)
13C4 – PFOS (503/ 99), (503/ 80), (503/ 131)
18O2 – PFOS (503/ 103), (503/ 84)
Perfluorononoic acid (PFNA) (463/ 419), (463/ 169)
13C5 – PFNA (468/ 423), (468/ 169)
13C2 - PFNA (465/ 420), (465/ 169)
Perfluorodecanoic acid (PFDA) (513/ 269), (513/ 469)
13C2 – PFDA (515/ 269), (515/ 470)
13C2 - PFNA (465/ 420), (465/ 169)
Perfluoroundecanoic acid (PFUA) (563/ 519), (563/ 169)
13C2 – PFDA (515/ 269), (515/ 470)
13C2 - PFDoA (615/ 570), (615/ 169)
Perfluorododecanoic acid (PFDoA) (613/ 569), (613/ 169)
13C2 – PFDA (515/ 269), (515/ 470)
13C2 - PFDoA (615/ 570), (615/ 169)
26
Table 2.2. Summary of perfluorinated compound concentrations (ng/L; mean ± 1 standard error) by site and season. N=6 for each sample unless denoted by an asterisk (*) where these samples, n = 5. PFOA = perfluorooctanoic acid; PFNA = perfluorononanoic acid; PFDA = perfluorodecanoic acid; PFUA = perfluoroundecanoic acid; PFDoDA = perfluorododecanoic acid; PFOS = perfluorooctane sulfonate; PFOSA = perfluorooctane sulfonamide; NEt-PFOSA = N-ethylperfluorooctane sulfonamide; NMe-PFOSA = N-methylperfluorooctane sulfonamide ; BDL = below detection limit.
Site Season PFOA PFNA PFDA PFUA PFDoDA PFOS PFOSA NEt-PFOSA NMe-
PFOSA
Altamaha Fall 7.36 ± 2.15 3.22 ± 1.53 BDL BDL 2.9 ± 2.61 5.15 ± 2.60 0.07 ± .06 0.12 ± 0.12 BDL
1 Spring 2.27 ±0.04 0.73 ± 0.73 BDL BDL BDL 1.43 ± 0.70 0.03 ± 0.01 BDL BDL
Altamaha Fall 3.47 ± 0.59 1.63 ± 0.66 0.18 ± 0.11 0.50 ± 0.37 0.56 ± 0.56 6.46 ± 0.70 0.49 ± 0.10 BDL BDL
2 Spring 2.71 ± 1.04 1.42 ± 0.40 0.05 ± 0.03 0.15 ± 0.09 1.10 ± 0.53 2.84 ± 0.99 0.13 ± 0.07 0.04 ± 0.03 BDL
Altamaha Fall* 3.65 ± 0.70 3.47 ± 1.59 1.11 ± 0.75 BDL BDL 8.23 ± 1.73 0.51 ± 0.26 0.15 ± 0.15 BDL
3 Spring 3.36 ± 0.33 0.36 ± 0.36 0.06 ± 0.06 0.44 ± 0.44 BDL 1.97 ± 0.79 BDL BDL BDL
Altamaha Fall 6.7 ± 1.03 1.29 ± 0.93 1.04 ± 0.47 0.29 ± 0.29 0.38 ± 0.38 4.4 ± 0.66 1.01 ± 0.9 BDL BDL
4 Spring 2.05 ± 0.70 0.38 ± 0.38 0.11 ± 0.11 BDL BDL 3.35 ± 0.93 0.1 ± 0.08 BDL BDL
Ocmulgee Fall 11.96 ± 3.51 6.08 ± 3.07 1.11 ± 0.52 BDL BDL 15.35 ± 3.42 0.06 ± 0.04 BDL BDL
1 Spring 5.38 ± 0.98 1.9 ± 1.19 0.1 ± 0.1 BDL 0.87 ± 0.68 6.28 ± 1.27 0.08 ± 0.07 BDL 0.28 ± 0.28
Oconee Fall 5.8 ± 0.96 BDL BDL BDL BDL 1.11 ± 0.62 0.04 ± 0.03 BDL BDL
1 Spring 3.46 ± 0.75 BDL BDL BDL 0.12 ± 0.12 0.76 ± 0.49 0.05 ± 0.05 0.6 ± 0.59 BDL
Oconee Fall 1.35 ± 0.06 1.62 ± 0.83 0.23 ± 0.23 BDL BDL 3.46 ± 0.85 0.22 ± 0.18 BDL 0.01 ± 0.01
2 Spring 1.29 ± 0.30 0 0.22 ± 0.22 BDL BDL 0.82 ± 0.35 0.26 ± 0.16 BDL BDL
Ogeechee Fall 15.99 ± 6.47 3.9 ± 1.37 5.48 ± 2.91 BDL BDL 5.34 ± 0.88 BDL BDL 0.10 ± 0.10
1 Spring 4.09 ± 1.07 BDL BDL BDL BDL 1.31 ± 0.64 BDL BDL BDL
Ogeechee Fall 11.65 ± 0.63 7.26 ± 1.39 13.67 ±5.63 BDL 0.88 ± 0.88 3.45 ± 0.63 0.24 ± 0.11 BDL BDL
2 Spring 3.58 ± 0.66 BDL BDL 0.51 ± 0.51 0.26 ± 0.26 3.65 ± 1.30 BDL BDL BDL
Ogeechee Fall 9.2 ± 1.59 6.34 ± 2.28 10.48 ± 4.35 BDL BDL 5.29 ± 1.39 0.14 ± 0.06 BDL BDL
3 Spring 3.27 ± 0.40 BDL BDL BDL BDL 5.86 ± 1.72 BDL BDL BDL
Ogeechee Fall* 0.69 ± 0.32 BDL BDL BDL BDL 1.25 ± 0.53 BDL BDL BDL
4 Spring 0.7 ± 0.3 BDL BDL BDL 0.46 ± 0.46 0.70 ± 0.30 BDL BDL BDL
Ohoopee Fall* BDL 1.2 ± 1.2 BDL BDL 0.43 ± 0.43 0.98 ± 0.37 BDL 0.31 ± 0.31 BDL
1 Spring 1.49 ± 1.00 BDL BDL BDL 0.89 ± 0.89 0.18 ± 0.18 BDL BDL BDL
27
Table 2.3. Environmental interpretation of predictor variables used in candidate models relating the local and watershed-level factors to perfluorinated compound concentrations in the water column.
Predictor Variable Interpretation/Hypothesis Percent urban in watershed Contaminants were derived from cumulative
effect of urban land use. Percent urban in riparian area Contaminants were derived principally from
nearby urban sources. Distance to nearest high urban Contaminants were derived from urban point
source via stream path. Number of discharge pipes in watershed
Contaminants were derived from cumulative impact of point source discharges on contaminants found in the river.
Distance to nearest discharge pipe Contaminants were derived from industrial or municipal point source via stream path.
River discharge High discharge may dilute the concentration of contaminants; it may also increase the transport of these chemicals due to the high flow of water. A lower discharge may then concentrate chemicals.
Water temperature Highly associated with season, high temperatures during fall sampling and low temperatures during spring sampling. Possible interactions with chemical properties such as vapor pressure or water solubility which may affect contaminant concentration in the water column.
River discharge × Number discharge pipes
The concentration of chemicals from discharge point sources will be diluted with high river discharge.
River discharge × urban riparian The concentration of chemicals from near urban sources will be diluted with high river discharge.
28
Table 2.4. Mean (standard deviation) and range of water temperature and river discharge summarized by water body in each season sampled. Water temperature C River Discharge m3/s
River Spring Fall Spring Fall Mean (SD) Min-Max Mean (SD) Min-Max Mean (SD) Min-Max Mean (SD) Min-Max Altamaha 17.8 (0.9) 16.1-21.9 24.0 (1.7) 21.3-27.2 403.4 (122.5) 148.4-574.8 51.1 (6.5) 43.3-61.4 Ocmulgee 13.5 (0.9) 12.8-14.1 22.5 (2.5) 20.7-24.3 69.7 (44.5) 38.2-101.1 12.4 (3.2) 10.2-14.7 Oconee 13.9 (2.2) 12.1-16.6 21.4 (1.7) 19.8-23.4 229.6 (305.6) 9.6-679.6 59.9 (98.3) 7.4-207.3 Ohoopee 19.9 (4.4) 16.8-23 18.8 (3.3) 16.5-21.1 39.8 (38.2) 12.7-66.8 2.0 (1.3) 1.0-2.9 Ogeechee 13.9 (3.6) 8.6-18.1 19.8 (3.3) 15.0-23.9 91.7 (42.5) 41.9-171.9 4.0 (1.6) 2.7-7.7
29
Table 2.5. Mean and range (in parentheses) of landscape characteristics for sample locations, summarized by major water body.
River Sampled
No. point source discharges
Watershed Urban (percent)
Riparian Row Crop (percent)
Watershed Agriculture (percent)
Riparian Urban (percent)
Riparian Wetland (percent)
Distance to urban area (km)
Distance to point source (km)
Altamaha Watershed Altamaha River 350
(334-357) 4.1
(1.7-10.1) 6.7
(6.5-6.9) 10.9
(10.5-11.2) 1.2
(1.1-1.3) 16.7
(15.8-17.5) 223
(143-301) 36
(5-72) Ocmulgee River 176 14 2.1 3.4 3.3 10.1 8 7 Oconee River 116
(106-125) 2.4
(1.6-3.2) 3.3
(2.0-4.6) 5
(3.1-7.0) 0.7
(0.6-0.7) 12.4
(11.0-13.7) 21
(15-27) 22
(21-24) Ohoopee River 16 1.6 9.3 15 0.2 21.4 602 32 Ogeechee Watershed Ogeechee River 23
(7-29) 3.7
(0.2-8.0) 9.3 (
2.4-11.8) 15.1
(3.5-19.0) 0.1
(0.0-0.2) 21.1
(11.76-24.9) 226
(85-602) 129
(81-228)
30
Table 2.6. Predictor variables, number of parameters (K), Akaike’s Information Criterion with the small-sample bias adjustment (AICc), ∆AICc, and Akaike weights (w) for the set of candidate models (i) for predicting concentration of perfluorinated compounds. Candidate Model K AICc ∆AICc wi Distance to urban, distance to point source, riparian urban, riparian wetland 9 4471.6 0.00 0.448
Distance to urban, distance to point source, riparian urban, riparian wetland, river discharge, water temperature 11 4473.0 1.36 0.227Distance to urban, river discharge, distance to urban x river discharge, water temperature 9 4474.5 2.84 0.108Distance to urban, distance to point source, riparian urban, river discharge, water temperature 10 4476.0 4.36 0.051Watershed urban, river discharge, watershed urban x river discharge 8 4476.3 4.65 0.044Distance to urban, distance to nearest point source, watershed urban, riparian urban, riparian wetland, number of point source discharges in watershed, river discharge, water temperature, distance to nearest effluent × discharge, distance to urban x discharge, watershed urban × river discharge 16 4476.7 5.07 0.036Watershed urban, river discharge, water temperature watershed urban × river discharge 9 4477.0 5.35 0.031Distance to urban, distance to point source, river discharge, water temperature, distance to nearest effluent × discharge, distance to urban × discharge 11 4477.9 6.24 0.020Watershed urban, number of point source discharges in watershed 7 4478.4 6.79 0.015
Watershed urban, number of point source discharges in watershed, river discharge, water temperature 9 4479.0 7.41 0.011watershed urban, river discharge, water temperature 8 4479.2 7.63 0.010
31
Table 2.7. Estimates, standard errors (in parentheses), and upper and lower 95% confidence intervals (CI) of fixed and random effects for the candidate set of hierarchical linear models relating environmental variables to perfluorinated compound concentrations using the natural log transformed concentration data.
Effect Estimate Lower CI Upper CI Distance to urban, distance to point source, riparian urban, riparian wetland Fixed Effects Intercept -3.54560 (0.63000) -5.55060 -1.54060 Distance to urban -0.00119 (0.00036) -0.00190 -0.00049 Distance to point source 0.00165 (0.00129) -0.00087 0.00418 Riparian urban 34.23390 (10.22650) 14.17090 54.29690 Riparian wetland 4.26520 (1.49460) 1.33300 7.19740 Random effects Intercept 2.01640 (0.95840) 0.94920 6.80810 Site 0.02564 (0.01836) 0.00912 0.22030 Sample period 0.16460 (0.13290) 0.05330 2.18490 Distance to urban, distance to point source, riparian urban, riparian wetland, river discharge, water temperature Fixed Effects Intercept -3.80710 (0.66320) -5.91780 -1.69640 Distance to urban -0.00127 (0.00036) -0.00198 -0.00055 Distance to point source 0.00189 (0.00134) -0.00073 0.00452 Riparian urban 32.24290 (10.28810) 12.05900 52.42670 Riparian wetland 3.82030 (1.51390) 0.85020 6.79040 River discharge -0.00029 (0.00031) -0.00090 0.00031 Water temperature 0.02093 (0.01485) -0.00819 0.05006 Random effects Intercept 2.00280 (0.95100) 0.94330 6.75190 Site 0.02557 (0.01849) 0.00902 0.22700 Sample period 0.09248 (0.08710) 0.02638 2.56010 Distance to urban, river discharge, distance to urban x river discharge, water temperature Fixed Effects Intercept -2.69070 (0.58420) -4.55000 -0.83140 Distance to urban -0.00166 (0.00042) -0.00248 -0.00085 River discharge -0.00084 (0.00046) -0.00174 0.00007 Water temperature 0.02159 (0.01481) -0.00747 0.05065 Distance to urban × discharge 0.00001 (0.00001) -0.00001 0.00001 Random effects Intercept 2.00490 (0.95210) 0.94430 6.76020 Site 0.05564 (0.03212) 0.02311 0.26970 Sample period 0.09319 (0.08788) 0.02656 2.59710
32
Table 2.7 continued. Estimates, standard errors (in parentheses), and upper and lower 95% confidence intervals (CI) of fixed and random effects for the candidate set of hierarchical linear models relating environmental variables to perfluorinated compound concentrations using the natural log transformed concentration data.
Effect Estimate Lower CI Upper CI Distance to urban, distance to nearest point source, riparian urban, river discharge, water temperature Fixed Effects Intercept -3.01690 (0.60770) -4.95090 -1.08290 Distance to urban -0.00142 (0.00045) -0.00229 -0.00054 Distance to point source 0.001485(0.00164) -0.00174 0.00471 Riparian urban 16.95890 (10.41780) -3.47950 37.39720 River discharge -0.00037 (0.00032) -0.00100 0.00027 Water temperature 0.024340 (0.01514) -0.00537 0.05405 Random effects Intercept 2.00210 (0.95070) 0.94310 6.74910 Site 0.05029 (0.02920) 0.02080 0.24690 Sample period 0.08101 (0.07849) 0.02256 2.63900
33
Table 2.8. Akaike importance weights for parameters from all candidate models relating PFC concentrations to environmental variables. Importance weights were estimated as the sum of Akaike weights from individual candidate models containing the parameter.
Model Parameter No. of candidate
models Importance
weights Distance to urban 6 0.8892 Distance to nearest point source 5 0.7808 Riparian urban 4 0.7610 Riparian wetland 3 0.7103 River discharge 9 0.5368 Water temperature 8 0.4620 Distance to urban × river discharge 3 0.1638 Watershed urban 6 0.1463 Watershed urban × river discharge 3 0.1104 Number of point sources in watershed 3 0.0605 Distance to nearest point source × river discharge
2 0.0553
34
Figure 2.1. Sampling locations (triangle) in the Altamaha and Ogeechee River watersheds in Georgia, USA.
35
Figure 2.2. Discharge for the USGS gauge station, 02226000, Altamaha River at Doctortown, Georgia observed during the sampling periods, bold line, and the long-term monthly averages, dotted line.
36
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
Distance tourban
Distance topoint source
Riparian urban
Riparianwetland
Stan
dard
ized
Par
amet
er E
stim
ate
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
Distance tourban
Distance topoint source
Riparian urban
Riparianwetland
Stan
dard
ized
Par
amet
er E
stim
ate
Figure 2.3. Parameter estimates from the best model relating perfluorinated chemical concentrations to environmental and landscape variables were standardized to the mean to show the relative effect each parameter has on the concentrations of PFCs in the water column.
37
Figure 2.4. Back transformed estimates of the relationship between the distance to the nearest urban area and the concentration of PFCs for each of nine chemicals as well as the sum of the chemicals based on the best fitting hierarchical linear model. PFOA = perfluorooctanoic acid; PFNA = perfluorononanoic acid; PFDA = perfluorodecanoic acid; PFUA = perfluoroundecanoic acid; PFDoDA = perfluorododecanoic acid; PFOS = perfluorooctane sulfonate; PFOSA = perfluorooctane sulfonamide; NEt-PFOSA = N-ethylperfluorooctane sulfonamide; NMe-PFOSA = N-methylperfluorooctane sulfonamide.
38
Figure 2.5. Predicted PFC concentrations from the back-transformed estimates of the relationship of PFC concentrations to the percent of urban land-use in the riparian area. PFOA = perfluorooctanoic acid; PFNA = perfluorononanoic acid; PFDA = perfluorodecanoic acid; PFUA = perfluoroundecanoic acid; PFDoDA = perfluorododecanoic acid; PFOS = perfluorooctane sulfonate; PFOSA = perfluorooctane sulfonamide; NEt-PFOSA = N-ethylperfluorooctane sulfonamide; NMe-PFOSA = N-methylperfluorooctane sulfonamide.
39
LITERATURE CITED
1. Kissa, E., Fluorinated Surfactants and Repellents. 2nd ed.; Marcel Dekker: New York, 2001.
2. Banks, R. E., B.E. Snartm J.C. Tatlow, Organofluorine Chemistry Principles and Commercial Applications. Plenum Press: New York, 1994.
3. Giesy, J. P.; Kannan, K., Perfluorochemical surfactants in the environment. Environmental Science & Technology 2002, 36, (7), 146a-152a.
4. Martin, J. W.; Smithwick, M. M.; Braune, B. M.; Hoekstra, P. F.; Muir, D. C. G.; Mabury, S. A., Identification of long-chain perfluorinated acids in biota from the Canadian Arctic. Environmental Science & Technology 2004, 38, (2), 373-380.
5. Tomy, G. T.; Budakowski, W.; Halldorson, T.; Helm, P. A.; Stern, G. A.; Friesen, K.; Pepper, K.; Tittlemier, S. A.; Fisk, A. T., Fluorinated organic compounds in an eastern Arctic marine food web. Environmental Science & Technology 2004, 38, (24), 6475-6481.
6. Houde, M.; Bujas, T. A. D.; Small, J.; Wells, R. S.; Fair, P. A.; Bossart, G. D.; Solomon, K. R.; Muir, D. C. G., Biomagnification of perfluoroalkyl compounds in the bottlenose dolphin (Tursiops truncatus) food web. Environmental Science & Technology 2006, 40, (13), 4138-4144.
7. Kannan, K.; Tao, L.; Sinclair, E.; Pastva, S. D.; Jude, D. J.; Giesy, J. P., Perfluorinated compounds in aquatic organisms at various trophic levels in a Great Lakes food chain. Archives of Environmental Contamination and Toxicology 2005, 48, (4), 559-566.
8. Newsted, J. L.; Jones, P. D.; Coady, K.; Giesy, J. P., Avian toxicity reference values for perfluorooctane sulfonate. Environmental Science & Technology 2005, 39, (23), 9357-9362.
9. Ankley, G. T.; Kuehl, D. W.; Kahl, M. D.; Jensen, K. M.; Linnum, A.; Leino, R. L.; Villeneuve, D. A., Reproductive and developmental toxicity and bioconcentration of perfluorooctanesulfonate in a partial life-cycle test with the fathead minnow (Pimephales promelas). Environmental Toxicology and Chemistry 2005, 24, (9), 2316-2324.
10. Boulanger, B.; Vargo, J. D.; Schnoor, J. L.; Hornbuckle, K. C., Evaluation of perfluorooctane surfactants in a wastewater treatment system and in a commercial surface protection product. Environmental Science & Technology 2005, 39, (15), 5524-5530.
11. Hansen, K. J.; Johnson, H. O.; Eldridge, J. S.; Butenhoff, J. L.; Dick, L. A., Quantitative characterization of trace levels of PFOS and PFOA in the Tennessee River. Environmental Science & Technology 2002, 36, (8), 1681-1685.
12. Konwick, B. J.; Tomy, G. T.; Ismail, N.; Peterson, J. T.; Fauver, R. J.; Higginbotham, D.; Fisk, A. T., Concentrations and patterns of perfluoroalkyl acids in Georgia, USA surface waters near and distant to a major use source. Environmental Toxicology and Chemistry 2008, 27, (10), 2011-2018.
40
13. Sinclair, E., S. Taniyasu, N Yamashita, K Kannan, Perfluorooctanoic acide and perfluorooctane sulfonate in Michigan and New York waters. Organohalogen Compounds 2004, 66, 4069-4073.
14. Nakayama, S.; Strynar, M. J.; Helfant, L.; Egeghy, P.; Ye, X. B.; Lindstrom, A. B., Perfluorinated compounds in the Cape Fear Drainage Basin in North Carolina. Environmental Science & Technology 2007, 41, (15), 5271-5276.
15. Boulanger, B.; Vargo, J.; Schnoor, J. L.; Hornbuckle, K. C., Detection of perfluorooctane surfactants in Great Lakes water. Environmental Science & Technology 2004, 38, (15), 4064-4070.
16. Berka, C.; Schreier, H.; Hall, K., Linking water quality with agricultural intensification in a rural watershed. Water Air and Soil Pollution 2001, 127, (1-4), 389-401.
17. Herlihy, A. T.; Stoddard, J. L.; Johnson, C. B., The relationship between stream chemistry and watershed land cover data in the mid-Atlantic region, US. Water Air and Soil Pollution 1998, 105, (1-2), 377-386.
18. Cuffney, T. F.; Meador, M. R.; Porter, S. D.; Gurtz, M. E., Responses of physical, chemical, and biological indicators of water quality to a gradient of agricultural land use in the Yakima River Basin, Washington. Environmental Monitoring and Assessment 2000, 64, (1), 259-270.
19. National Weather Service, N. O. a. A. A., In 2008.
20. So, M. K.; Taniyasu, S.; Yamashita, N.; Giesy, J. P.; Zheng, J.; Fang, Z.; Im, S. H.; Lam, P. K. S., Perfluorinated compounds in coastal waters of Hong Kong, South China, and Korea. Environmental Science & Technology 2004, 38, (15), 4056-4063.
21. Taniyasu, S.; Kannan, K.; So, M. K.; Gulkowska, A.; Sinclair, E.; Okazawa, T.; Yamashita, N., Analysis of fluorotelomer alcohols, fluorotelorner acids, and short- and long-chain perfluorinated acids in water and biota. Journal of Chromatography A 2005, 1093, (1-2), 89-97.
22. van Leeuwen SPJ, K. A., van Bavel B, de Boer J, Lindstrom G., Struggle for quality in determination of perfluorinated contaminants in environmental and human samples. . Environ Sci Technol 2006, 40, 7854-7860.
23. USGS, National Elevation Dataset. In USGS, Ed. 1999.
24. USGS, National Hydrography Dataset. In USGS, Ed. 2001.
25. Natural Resources Spatial Analysis Laboratory. 1998 vegetation/land cover map of Georgia. In Institute of Ecology, University of Georgia, Athens, GA, 2003.
26. USEPA, EPA/OW Industrial Facilities Discharge Database for CONUS. In Office of Water, U. S. E., Ed. Washington, D.C., 1998.
41
27. Snijders, T. a. B. R., Multilevel analysis: an introduction to basic and advanced multilevel modeling. Sage: Thousand Oaks, California, 1999.
28. Burnham, K. P. and Anderson, D. R. , Model selection and inference: a practical information-theoretic approach. 2 ed.; Springer-Verlag: New York, 2002.
29. Akaike, H., Information theory and an extension of the maximum likelihood principle. In Second International Symposium on Information Theory, Csaki, B. N. P. a. F., Ed. Akademiai Kiado: Budapest, Hungary, 1973; pp 267-281.
30. Hurvich, C. M.; Tsai, C. L., Regression and Time-Series Model Selection in Small Samples. Biometrika 1989, 76, (2), 297-307.
31. Anderson, D. R.; Burnham, K. P.; Thompson, W. L., Null hypothesis testing: Problems, prevalence, and an alternative. Journal of Wildlife Management 2000, 64, (4), 912-923.
32. Royall, R. M., Statistical evidence: a likelihood paradigm. Chapman and Hall: New York, 1997.
33. Wolfinger, R. D., Heterogeneous Variance - Covariance Structures for Repeated Measures. Journal of Agricultural, Biological, and Environmental Statistics 1996, (1), 205-230.
34. Sokal, R. R. and F. J. Rohlf, Biometry: the principles and practice of statistics in biological research. Freeman: New York, 1995.
35. Ellis, D. A.; Martin, J. W.; De Silva, A. O.; Mabury, S. A.; Hurley, M. D.; Andersen, M. P. S.; Wallington, T. J., Degradation of fluorotelomer alcohols: A likely atmospheric source of perfluorinated carboxylic acids. Environmental Science & Technology 2004, 38, (12), 3316-3321.
36. Loganathan, B. G.; Sajwan, K. S.; Sinclair, E.; Kumar, K. S.; Kannan, K., Perfluoroalkyl sulfonates and perfluorocarboxylates in two wastewater treatment facilities in Kentucky and Georgia. Water Research 2007, 41, (20), 4611-4620.
37. Sinclair, E.; Kannan, K., Mass loading and fate of perfluoroalkyl surfactants in wastewater treatment plants. Environmental Science & Technology 2006, 40, (5), 1408-1414.
38. 3M Company, Soil adsorption/desorption study of potassium perfluorooctanesulfonate (PFOS). In Laboratory Project E00-13311, AR226-1030a030, E. D., Ed. 3M Environmental Laboratory: St. Paul, MN, USA, 2001.
39. Higgins, C. P.; Luthy, R. G., Sorption of perfluorinated surfactants on sediments. Environmental Science & Technology 2006, 40, (23), 7251-7256.
40. Gulkowska, A.; Jiang, Q. T.; So, M. K.; Taniyasu, S.; Lam, P. K. S.; Yamashita, N., Persistent perfluorinated acids in seafood collected from two cities of China. Environmental Science & Technology 2006, 40, (12), 3736-3741.
42
41. Holmstrom, K. E.; Berger, U., Tissue distribution of perfluorinated surfactants in common guillemot (Uria aalge) from the Baltic Sea. Environmental Science & Technology 2008, 42, (16), 5879-5884.
42. Moody, C. A.; Martin, J. W.; Kwan, W. C.; Muir, D. C. G.; Mabury, S. C., Monitoring perfluorinated surfactants in biota and surface water samples following an accidental release of fire-fighting foam into Etohicoke Creek. Environmental Science & Technology 2002, 36, (4), 545-551.
43. Jones, P. D.; Hu, W. Y.; De Coen, W.; Newsted, J. L.; Giesy, J. P., Binding of perfluorinated fatty acids to serum proteins. Environmental Toxicology and Chemistry 2003, 22, (11), 2639-2649.
44. Oakes, K. D.; Sibley, P. K.; Martin, J. W.; MacLean, D. D.; Solomon, K. R.; Mabury, S. A.; Van Der Kraak, G. J., Short-term exposures of fish to perfluorooctane sulfonate: Acute effects on fatty acyl-CoA oxidase activity, oxidative stress, and circulating sex steroids. Environmental Toxicology and Chemistry 2005, 24, (5), 1172-1181.
43
CHAPTER 3
THE INFLUENCE OF LAND-USE ON ORGANOCHLORINE CHEMICAL DELIVERY TO
ESTUARINE SYSTEMS1
1 The influence of land-use on organochlorine chemical delivery to estuarine systems. Fauver,
R.J., A.T. Fisk, N.P. Nibbelink, G. Tomy, B. Rosenburg, J.T. Peterson. To be submitted to
Environmental Science and Technology
44
INTRODUCTION
Organochlorine (OC) chemicals including pesticides like DDT, heptachlor, chlordane,
and dieldrin and the industrial chemical group, polychlorinated biphenyls (PCBs) have a long
history of use in the Southeastern United States [1]. OC pesticides were widely produced and
used from the 1940s through the 1970s when they were banned from use due to the growing
evidence of toxic effects to wildlife. At the same time, PCBs were manufactured and used
extensively for industrial use [2]. Over 600 million kilograms of PCBs were used in North
America between 1930 and 1979, of which about 15% entered the environment through use,
disposal, and accidental spills [3, 4]. Although production and use in North America has ceased,
OC chemicals are routinely found in environmental samples today. OC chemicals are lipophillic
in nature which allows them to bioaccumulate in biota and transfer to higher trophic level
organisms through food webs [5-9]. There is concern over their presence in the environment as
some of these compounds are hormone disrupting chemicals, affecting endocrine and
reproductive systems in humans and wildlife [10-12].
Organochlorine chemicals are hydrophobic therefore readily binding to sediment and
soils in aquatic environments [1, 13]. OC contaminants are found in higher concentrations in
suspended sediments than in the water column or bed sediment, but are much higher in biota
than suspended sediment [14]. OC chemicals were used heavily until the 1970s when use and
production ceased in North America, but are still found in surface waters, sediments, and biota
worldwide. OC pesticides that were sprayed on agricultural fields may still reside in the soils,
and pesticides and PCBs that have been atmospherically transported may remain in soils for
years to come due to their persistence and inability to be naturally degraded [3]. Therefore, OC
contaminants may enter the stream via surface run-off from soils with OC pesticide and PCB
45
residues. In addition, atmospheric transport may account for some of the OC contaminant
concentrations in aquatic systems. Physical-chemical properties of OC contaminants may
influence their presence, concentration, and transport in aquatic systems as well. Contaminants
with low volatility, low hydrophobicity, and high water solubility are more likely to be found in
the water column.
Unfortunately, there is a lack of information on the current status of environmental OC
chemicals and their likely sources and transport in water bodies of the southeastern US despite
the historical high use of OC pesticides and PCBs in this region [1]. Therefore, we studied the
OC contaminants in surface waters in Georgia with the following objective: to determine the
types and concentrations of OC contaminants in rivers and estuaries; relate concentrations to
land-use and environmental factors; and to build a predictive model of OC contaminant presence
and concentrations in the surface waters draining into coastal Georgia.
METHODS
Chemicals and standards
Optima grade acetone, hexane, dichloromethane, and water were purchased from Fisher
Scientific. Tribrominated benzene and PCB 85 spiking standards were purchased from
Chromatographic Specialties (Brockville, Ontario, Canada) and diluted to 125 µg/L and 200
µg/L, respectively, to use as a known addition recovery spike for organochlorine contaminant
analysis. A reference standard of Aroclor PCB mixture was purchased from Chromatographic
Specialties (Brockville, Ontario, Canada) and diluted to 525 µg/L.
46
Study sites and sample collection
Twelve sampling locations were chosen along two river systems in Georgia, USA, the
Altamaha (and its major tributaries) and the Ogeechee Rivers (Figure 3.1). Sampling locations
were chosen by stratified random sampling where strata were determined by dominant land-use
and distance to the estuary. Points along the river were ranked by percent row crop agriculture
and percent urban area and were put in strata of high and low agriculture and urban land-use,
mixed land-use, and near and far distance from the estuary. Ten points were randomly chosen
from these strata and were used for the field sampling locations. Two additional sites were
added, one at the salt-wedge of each river. Field samples were taken in October 2006 and 2007
during the low flow season and in March 2007 and 2008 during the high flow season to assess
variation in contaminant delivery due to stream flow.
Field sampling methods
Fifty liters of water were extracted through XAD (Amberlite XAD 1180 ion-exchange
resin) via the Infiltrex 300 inline filter and extractor (Axys Technologies, Sidney, BC, Canada).
Water was pumped at approximately 1 L/minute through a wound glass wool filter to remove
particulate matter and then passed through a teflon column packed with 40 g of XAD. The glass
wool filters were pre-cleaned by soxhlet extracting with ultra pure dichloromethane for 16 hours
to avoid OC chemical contamination. Columns were packed at the Whitehall Fisheries
Laboratory at the University of Georgia, Athens, Georgia, USA. Teflon columns were washed
and rinsed with acetone and hexane before they were packed with XAD. First, XAD was treated
with Optima grade methanol by soaking for 15 minutes to wet the resin. The resin was
transferred to Optima grade water and packed in the columns, topped with water and sealed to
47
retain wetness. Columns were spiked in the field with 100 µL of 125 µg/L tribrominated
benzene recovery standard before water was extracted through the column. Four liters of Optima
water were pumped through the XAD column for each field blank, three or four field blanks per
sample period.
Contaminant extraction and analysis
Samples were extracted at the Organic Analysis Laboratory at the Great Lakes Institute
for Environmental Research, University of Windsor, Windsor, ON, Canada.
Extraction of organochlorine compounds from the XAD resin was done by eluting the
resin with 200 ml Optima grade acetone (fraction 1). A second elution with 200 ml of 50:50
hexane/dichloromethane solution (fraction 2) was collected separately, reduced in volume and
then added to fraction 1. The combined fractions were reduced to approximately 25 ml and
combined with 25 ml of saturated sodium chloride (NaCl) solution to polarize the water phase
and drive organochlorine compounds into the solvent phase. A liquid-liquid extraction was
performed with the NaCl solution by adding 100 ml Optima grade hexane, vigorously shaking,
allowing to settle and removing the hexane fraction. This extraction was repeated three times to
extract most of the contaminants from the water/acetone phase have been transferred to the
hexane phase. The combined hexane solution was reduced via roto-evaporation to
approximately 2 ml and treated with reduced copper to remove sulfur compounds which hinder
the chromatography. The copper treated solution underwent a florisil clean-up procedure which
divided the sample into three fractions by eluting with hexane (fraction 1), followed by 85:15
hexane/dichloromethane (fraction 2), and followed by 60:40 hexane/dichloromethane (fraction
3). The three fractions were kept separated, evaporated to 200 µL and transferred to 2 mL gas
48
chromatograph auto sample vials with 200 µL inserts for gas chromatography electron capture
detection (GC/ECD) analysis.
A method blank and a reference sample were extracted with each set of samples to
determine recovery. All samples were spiked with 50 µL of PCB 85 standard before extractions,
the method blank was spiked with 10 µL of tribrominated benzene and the reference was spiked
with 100 µL of Aroclor PCB mixture and extracted with the field samples.
The samples were analyzed with Varian 3800 GC/ECD with an 8400 autosampler
(Missisauga, Ontario, Canada) at the Freshwater Institute, Department of Fisheries and Oceans,
Winnipeg, Manitoba, Canada. Contaminants were separated on a 60 m DB-5 with a 60 m XLB
confirmatory column (characteristics are 0.25 µm phase thickness and 0.25 mm ID; J&W
Scientific, Chromatographic Specialties, Brockville, Ontario, Canada).
Quality Assurance/Quality Control
The extraction for OC chemicals was conducted at the Organic Analysis Laboratory at
the Great Lakes Institute for Environmental Research which is an accredited lab with the
Canadian Association for Environmental Analytical Laboratories. Organochlorine contaminant
analytical analysis was conducted at the Freshwater Institute, Department of Fisheries and
Oceans at the University of Manitoba. This lab is currently participating in the Northern
Contaminants Program check sample program, the International Atomic Energy Agency's
Marine Environmental Quality Laboratory check sample program, and Quality Assurance for the
Intercontinental Atmospheric Transport of Anthropogenic Pollutants to the Arctic project by the
Process Research Branch, Science and Technology Section, Environment Canada.
49
In this study, six field samples were extracted alongside a method blank, the method
blank extracted with the set of samples was used to blank correct that set of samples. Recoveries
of the PCB 85 standard were used to assess the recovery of OC contaminants from the field
samples. Recoveries for PCB 85 averaged 80% (standard error, 13.2%) and therefore all OC
concentrations were corrected to 80% recovery.
Spatial data
The watersheds above each sampling location were delineated with ArcHydro in ArcGIS
9.1 using a 30-m digital elevation model (DEM) raster obtained from the National Elevation
Dataset [15] and the Altamaha and Ogeechee river system hydrography from the USGS National
Hydrography Dataset [16]. PCBs and OC pesticides are associated with past industrial and
agricultural uses [2], respectively, and therefore various metrics of urban and row crop land-uses
as well as industrial and municipal point source discharges were determined for each sampling
location. Urban and row crop agricultural land-uses were the landscape variables of primary
interest for this analysis, and therefore various metrics were calculated using these data. The
percentage of row crop agricultural and urban (low and high urban land cover) land cover was
calculated for each sample location watershed using Landsat-derived land cover data [17].
Percentages of urban and row crop land cover types also were determined for a 1 km buffer on
each side of the stream to obtain riparian land-use, defined as riparian urban or riparian row crop.
Additionally, stream distance was measured from each sampling location to the nearest high
density urban area (city population >10,000) in order to assess the influence of proximity to an
urban source on contaminant concentrations.
50
Physical chemical properties
The physical chemical properties of a contaminant may influence the presence,
concentration, and transport of that contaminant in waterbodies. These characteristics can be
used to predict concentrations of OC contaminants in riverine systems. A literature search for
physical chemical properties of OCs that would influence either source or delivery of the
chemicals through riverine systems was conducted for all chemicals analyzed in this study.
Chemical databases available to the public were also searched [18, 19]. Any physical chemical
property that was not described from these sources was estimated using similarly structured
chemicals. Physical chemical properties investigated were water solubility, log Kow, and vapor
pressure due to their influence on chemical presence in the water column (Appendix A).
Statistical Analysis
Many landscape, site specific, and physical chemical property variables were considered
for use in explaining patterns in OC concentrations in the water column. To avoid biased
parameter setimates and standard errors caused by multicolinearity. To avoid multicolinearity,
Pearson correlations were run on all pairs of predictor variables prior to modeling and only
uncorrelated predictor variables (r2 < 0.30) were included in the candidate models.
The relationship between the concentration of organochlorine chemicals (OCs) in the
water column and environmental variables were initially evaluated using linear regression.
However, multiple samples and chemicals collected at a site during different sample periods
were likely to be autocorrelated as were multiple samples collected during a single sample
period. Similarly, the relations between environmental variables and concentrations are likely to
vary among the types of OCs. Therefore, I initially fit a global candidate model containing all
51
predictor variables (Table 3.1) and conducted an analysis of variance (ANOVA) on the residuals.
The ANOVA indicated substantial autocorrelation among sample periods (F = 523.94; df = 3,
132; P < 0.0001), among sites (F = 18.05; df = 11, 132; P < 0.0001), and among chemicals (F =
56.41; df = 118, 132; P < 0.0001). To account for the autocorrelation, I used hierarchical models
to examine the relationship between environmental variables and the concentration of OC
chemicals in the water column. Hierarchical models differ from more familiar regression
techniques in that autocorrelation is incorporated by including random effects [20]. For this
study, the random effects associated with sample periods, sites, sample years, or chemicals
represent the differences in concentration between sample periods, between sites, among sample
years, or between types of chemicals, respectively, unaccounted for by covariates.
An information-theoretic approach [21] was used to evaluate the relative plausibility of
the candidate models relating the concentration of OC chemicals in the water column to
environmental variables. Each candidate model represented hypotheses regarding the
concentrations of OC chemicals related to either near or diffuse sources or urban or agricultural
land-uses. A global model with all predictors was initially constructed and 13 candidate models
were constructed from the global model (Table 3.1). To assess relative plausibility of each
candidate model, Akaike’s Information Criterion (AIC) [22] with the small-sample bias
adjustment were calculated (AICc) [23]. AIC is an entropy-based measure used to compare
candidate models for the same data which penalizes a model for complexity due to the number of
parameters used [21], with the best fitting model having the lowest AICc. The number of
parameters used for calculated AICc included both fixed and random effects, plus an error term.
The relative plausibility of each candidate model was assessed by calculating Akaike weights as
described in Burnham and Anderson [21]. These weights range from 0 to 1 with the most
52
plausible model having the highest weight. A confidence set of models, analogous to a
confidence interval for a parameter estimate, was reported instead of basing all inferences on one
single best model. The ratio of Akaike weights for two candidate models can be used to assess
the degree of evidence for one model over another [24]. A confidence set of models included
only those candidate models with Akaike weights that were within 10% of the highest weight,
similar to the 1/8 rule of thumb suggested by Royall [25]. The precision of parameter estimates
was determined by evaluating their 95% confidence intervals, an estimate with a confidence
interval overlapping zero was considered imprecise. All inferences were made using parameter
estimates included in the confidence set of models. Akaike importance weights were calculated
for each parameter used in contrasting models to determine the relative importance of that
parameter in the models. Prior to model selection, a random effects ANOVA with the site,
chemical, and sample period was fit to partition variation in chemical concentrations. Goodness-
of-fit of each candidate model was determined by examining a normal probability plot of the
residuals. All hierarchical and random effects models were fit in SAS PROC MIXED with the
maximum likelihood (ML) estimation specified (SAS Institute 2004).
Prior to evaluating the fit of the candidate models, the relative fit of several different
variance structures was evaluated for the hierarchical model random effects using the global (all
predictors) model. The first variance structure included a random effect for sampling period, the
second for sample sites, the third included a random effect for both site and sample period
additively, the fourth for sample year, and the fifth included a random effect for both site and
sample year. Each hierarchical model also included a randomly varying intercept among
chemicals. To assess the relative fit of each variance structure, AICc were calculated. The best
approximating variance structure was then used during the evaluation of the relative plausibility
53
of the candidate models. An ANOVA of the residuals for the best approximating variance
structure was conducted to determine if autocorrelation had been accounted for by the error
structure.
RESULTS
Spatial and Physical/chemical properties
River discharges during the study were below long term averages for each season due to
a long-term drought, but the pattern mirrored the normal pattern of low discharge in the summer
and fall and high discharge in the winter and spring (Figure 3.2). The observed discharges in the
Altamaha watershed ranged from 2 – 59 m3/s in the fall with the lowest discharge in the
Ohoopee River and greatest discharge in the Oconee River and 39 – 403 m3/s in the spring with
the lowest discharge in the Ohoopee River and greatest discharge in the Altamaha River (Table
3.2). The average fall discharge in the Ogeechee River was 4.0 m3/s and average spring
discharge was 91.7 m3/s (Table 3.2). The water temperature was typically greater in the fall and
lower in the spring for all rivers except the Ohoopee River where the average water temperature
was 18.8 °C in the fall and 19.9 °C in the spring (Table 2). Turbidity measurements varied
among sites and seasons, with no distinct seasonal pattern. The turbidity measurements ranged
from 4-60 NTU and were generally greatest in the Altamaha and Oconee Rivers, least in the
Ohoopee and Ogeechee Rivers (Table 3.2).
The landscape metrics varied by river. The Ocmulgee River had the greatest percent
urban land-use in the whole watershed (14%) as well as the one km riparian area (3%; Table
3.3). The Ocmulgee River site had the shortest stream distance to the nearest high density urban
area (8 km), followed by the Oconee River (21 km; Table 3.3). The Altamaha and Ogeechee
54
Rivers each had an average of over 200 km to the nearest urban area while the Ohoopee River
did not have any high density urban areas in the watershed (Table 3.3). Conversely, the
Ocmulgee and Oconee Rivers had the lowest percent row crop agriculture in the whole
watershed (3-5%) and in the riparian area (2-3%) while the Altamaha, Ohoopee, and Ogeechee
Rivers contained 10-15% agriculture in the whole watershed and 6-9% in the riparian area (Table
3.3).
Water samples were collected at 12 sites during the fall and spring over two years in
replicates of three resulting in 144 samples which were analyzed for OC contaminants including
OC pesticides and PCBs. Twenty-seven OC pesticides and 91 PCBs were detected in at least
one sample period. The concentrations of OC pesticides were generally lower than PCBs; the
average OC pesticide concentration found in the water ranged from 0.02 – 0.10 ng/L (Figure 3.3)
while the average PCB concentration ranged from 0.2 – 2.0 ng/L (Figure 3.4). Seasonal
variation was observed in both OC pesticides and PCBs. The concentrations of OC pesticides
were higher in spring samples ranged from 0.0 – 11.9 ng/L while concentrations in fall samples
were lower, ranging from 0.0 – 1.4 ng/L. The Ocmulgee River had the highest average OC
pesticide concentrations in the fall (0.077 ng/L), which was five times greater than the next
highest average OC pesticide concentration. The highest observed concentration of OC
pesticides was on the Altamaha River during spring sampling with a p,p’-DDT concentration of
11.9 ng/L (Figure 3.3). The Altamaha and Ogeechee Rivers had similar OC pesticide
concentrations and no longitudinal pattern of concentrations increasing downstream was
apparent.
PCB concentrations in spring samples were generally higher than fall samples for PCBs
as well; spring PCB samples ranged from 0.0 – 114.1 ng/L while fall PCB samples ranged from
55
0.0 – 41.5 ng/L (Figure 3.4). The highest observed PCB concentration was 114 ng/L for PCB 70
on the Altamaha River during spring sampling. The Oconee River sites in the Altamaha
watershed and the Ogeechee River had the lowest average PCB concentrations, averaging less
than 0.4 ng/L, while the Ocmulgee and Altamaha River samples contained the highest PCB
concentrations (Figure 3.4). The Altamaha River samples had greater PCB concentrations than
the Ogeechee River samples, site specific averages ranged from 0.30 – 2.85 ng/L and 0.19 – 0.78
ng/L, respectively (Figure 3.4). Both the Altamaha and Ogeechee Rivers appeared to exhibit a
longitudinal pattern where PCB concentration generally increased downstream.
Statistical Analysis
Plots of residuals from the global model relating OC chemical concentrations to
environmental variables indicated that the residuals were non-normal and heteroscedastic. The
data was normalized by a log10 transformation and the models were refit. An examination of the
residuals from the global model with the transformed data indicated normality therefore all
models were fit with the transformed data.
The best approximating variance structure for the global model relating OC chemical
concentrations to watershed level and site specific environmental predictor variables included an
intercept that randomly varied by chemical, and two additional random effects corresponding to
site and sample period additively. The analysis of variance of the residuals from this model
indicated no detectable dependence among sample periods (F = 0; df = 3, 132; P = 0.99), sites (F
= 0.07; df = 11, 132; P = 1.0), and chemicals (F = 0.02; df = 118, 132; P = 1.0). However,
heterogeneity in error (among replicate samples) variance among chemical types was detected.
The problem was unable to be remedied by fitting a heterogeneous variance model [26].
56
Heterogeneity in error variance can bias comparisons of variance components, but has little
effect on fixed effect parameter estimates [27]. Therefore, the best approximating error structure
described above was used with all candidate models in model selection and restricted our
inferences to the fixed effects parameter estimates.
The most plausible model of OC chemical concentrations contained riparian row crop
agriculture, log Kow, turbidity, log Kow and turbidity interaction, water solubility, water
temperature, water solubility and water temperature interaction, vapor pressure, vapor pressure
and water temperature interaction, river discharge, and a riparian row crop agriculture and river
discharge interaction (Table 3.5). This model was 51 times more likely than the next best
approximating model, which was the global model containing all predictor variables. There was
very little support for any other candidate model, therefore the best approximating model was the
only model contained in the confidence set (Table 3.6).
The most important environmental variables in the model were discharge, which was
negatively related to OC concentrations, and water temperature and turbidity, which were both
positively related to OC concentrations (Table 3.5; Figure 3.5). Water solubility and log Kow
were the most important physical chemical parameters in the model and were both negatively
related to PFC concentrations in the watershed (Table 3.5; Figure 3.5). The riparian row crop
and river discharge interaction was positively related to OC contaminant concentrations in the
water column (Table 3.5; Figure 3.5). The concentration of contaminants decreased from
0.0006ng/L under low discharges (less than 15 m3/s) to less than 0.0005 ng/L under high
discharges (greater than 600 m3/s) for low and average percent row crop agriculture and all other
variables held at average observed values, but the OC concentration increased to 0.002 ng/L
under high discharges when there was a high percent row crop agriculture present (Figure 3.6).
57
The water temperature and water solubility interaction was positively related to OC contaminant
concentration (Table 3.7). The concentration of OC contaminants decreased to near zero with an
increase in water solubility in low, average, and high water temperatures, with all other variables
held at average observed values (Figure 3.7). The log Kow and turbidity interaction parameter
was negatively related to OC concentrations (Table 3.6). As turbidity increased, low log Kow
contaminant concentrations increased from 0.0005 to 0.008 ng/L, but under average log Kow
values, contaminant concentrations only increased from 0.0003 to 0.002 ng/L, and under high
log Kow values, contaminant concentrations only increased from 0.0001 to 0.0007 ng/L (Figure
3.8).
DISCUSSION
Legacy OC pesticides and PCBs were detected in measurable quantities in the Altamaha
and Ogeechee Rivers under a wide range of discharges, water temperatures, and land-uses.
Concentrations of OC pesticides and PCBs were detected in similar concentrations to surface
waters around the world, typically less than 100 ng/L in lakes and rivers [28-30]. OC
contaminant concentrations were generally greater in the spring (March) in contrast to the fall
(October) and greater at sites with high amounts of agricultural land-use near the stream. Greater
concentrations may be due to the increased precipitation this region receives during the spring
which would cause surface run-off from row crop agriculture near the stream.
There was much greater support for models containing agriculture land use compared to
urban related sources (e.g., urban land use, number of municipal discharges), suggesting that
agricultural land-uses were more important sources of OC contaminants in surface waters in
Southeast Georgia than urban sources. PCB contamination in surface water has historically been
58
linked to urban land-use [31] while OC pesticides have been linked to row crop agriculture.
However, due to the persistence of OC contaminants in the environment coupled with
atmospheric delivery and the lack of current use, these chemicals have become widespread and
weakly linked to sources. OC pesticides and PCBs bind strongly to soils [13] and resist
degradation in the environment [3]. We hypothesize that trace amounts of OC chemicals are
bound to soils in agricultural fields where they were once sprayed or deposited atmospherically.
Thus, agriculture fields with OC pesticide residues and deposited PCB pollutants bound to their
soil now appear to serve as a source of OC contaminants to the stream.
OC concentrations in the water column were most strongly related to amount of row crop
agriculture that was immediately adjacent to the stream. The nature of the relation, however, was
influenced by discharge at the time of sampling. When row crop agriculture fields in the areas
adjacent to the streams were low to average, the concentrations of OCs were negatively related to
discharge at the time of sampling. In contrast, the relation between OC concentrations and
discharge was strongly positive when row crop land use was relatively high. We hypothesize
that increases in discharge are due to precipitation and subsequent surface run-off from the land
adjacent the stream. Run-off from agricultural fields with legacy OC pesticide residue or
atmospherically delivered PCBs bound to the soil may deliver OC contaminants with the soil and
serve as a source for OC contamination to the river. Without a strong source like high
percentage of riparian agriculture, transport of OC contaminants would have to take place over a
greater distance, which is unlikely due to the nature of these chemicals to bind strongly to soils.
Therefore, in areas without riparian row crop fields, an increase in river discharge dilutes any
contaminants present in the water column. The positive relation between OC contaminant
concentrations and turbidity is consistent with our hypothesis. Surface run-off from non-forested
59
areas like urban or agricultural fields transports suspended soil to the stream and causes
increased turbidity in the water column [32]. OC concentrations also increased at a higher rate
with turbidity for contaminants with a low log Kow. All OC chemicals in this study are
considered to be hydrophobic due to their low partitioning into water as seen by their high log
Kow values, but the less hydrophobic chemicals (i.e. low logKow chemicals) may dissociate from
suspended particles in highly turbid waters and enter into the water phase. Consequently, surface
run-off from riparian agricultural fields is a major contributor to contaminant delivery to the
stream.
Implications as source to estuaries
Rivers appear to serve as conduits for persistent OC contaminant delivery from terrestrial
sources to estuaries. The concentrations of OC chemicals in estuaries in the US and Asia are
typically less than 50 ng/L [14, 28, 33-35], which are higher than concentrations found in this
study that were 1.03 ng/L, on average. PCB concentrations generally increased downstream in
the Ogeechee and Altamaha Rivers, indicating that these chemicals are persistent in aquatic
systems and are building up downstream as they enter the estuaries. The highest PCB
concentrations in the Ogeechee River were found downstream near the estuary at an average of
0.47 ng/L. However the highest concentrations of PCBs in the Altamaha River were detected
near the confluence of the Ocmulgee and Oconee Rivers at an average of 2.9 ng/L but the
Altamaha River site near the estuary had an average concentration of PCBs of 1.3 ng/L which
was greater than the remaining sites within the watershed. OC contaminants are environmentally
stable due to the inability of natural soil and aquatic biota to degrade the compounds at a
significant rate [3, 36], therefore OC contaminants that are bound to soils are likely to remain for
60
years to come. Riparian agricultural fields are a chronic source for OC chemicals to the stream,
and for that reason, management efforts should be concentrated on increasing the forested
riparian zones near row crop agriculture.
The PCB concentrations detected near estuaries in this study are above the Ambient
Water Quality guidelines for PCBs as determined by the US EPA, which was established to
protect freshwater and marine aquatic life. The water quality criteria for the sum of PCBs found
in the water is 0.1 ng/L and the protective guideline for PCB 105 is 0.09 ng/L [37]. We routinely
detected higher concentrations for the sum of PCBs and the PCB 105 congener than the
protective water quality guidelines. Low protective water quality guidelines are established for
these chemicals because many are considered endocrine disruptors and affect reproduction in
wildlife as well as bioaccumulate in organisms and transfer through food webs [37]. However, it
should be noted that the protective standards are very conservative measures to protect sensitive
species and were established one to two orders of magnitude lower than protective water quality
guidelines for Canada [37]. Adverse affects to OC chemicals typically occur at concentrations
three to six orders of magnitude greater than detected in the water, but concentrations in the
suspended sediments and benthic sediments are two to three orders of magnitude greater than
that in the water and concentrations in benthic biota are four orders of magnitude higher than
water [14]. Concentrations of OC chemicals in the freshwater entering the estuary are higher
than the protective guidelines, indicating that potential adverse affects to aquatic organisms may
be occurring.
Prior to this study, little was known about the presence and distribution of OC chemicals
in Georgia waters. This work has begun to show spatial distributions and temporal trends of
legacy OC contaminants in Georgia’s coastal rivers. The relationship of OC contaminant
61
concentrations to land-based activities was quantified through predictive modeling aids the
understanding of how these chemicals behave in the environment. This work has determined
that run-off from agricultural fields is the most likely source for legacy OC contaminants;
therefore future work should focus on how to prevent these chemicals from entering aquatic
systems from land-based sources.
62
Table 3.1. Environmental interpretation of predictor variables used in candidate models relating the local and watershed-level factors to organochlorine chemical concentrations in the water column.
Predictor Variable Interpretation/Hypothesis Percent urban in watershed Contaminants were derived from cumulative effect of urban land use. Percent urban in riparian area Contaminants were derived principally from nearby urban sources. Distance to nearest high urban Contaminants were derived from urban point source via stream path. Percent agriculture in watershed Contaminants were derived from cumulative effect of agricultural land use. Percent agriculture in riparian area Contaminants were derived from nearby agricultural land use. Number of discharge pipes in watershed
Contaminants were derived from cumulative impact of point source discharges on contaminants found in the river.
River discharge High discharge may dilute the concentration of contaminants; it may also increase the transport of these chemicals due to the high flow of water. A lower discharge may then concentrate chemicals.
Water temperature Highly associated with season, high temperatures during fall sampling and low temps during spring sampling. Possible interactions with physical chemical properties that affect chemical presence in water.
Turbidity Contaminants may bind to suspended particles in water and increase contaminant concentrations found in the water column.
Log Kow Low Log Kow (<1) indicates the contaminant is hydrophilic, high Log Kow (>4) indicates the contaminant is hydrophobic. The more hydrophobic, the more likely to bind to organic sediment. Long range transport may be limited because contaminant is likely to bind to sediment. Less likely to enter river via run off due to binding to soil.
63
Table 3.1 continued. Environmental interpretation of predictor variables used in candidate models relating the local and watershed-level factors to organochlorine chemical concentrations in the water column.
Predictor Variable Interpretation/Hypothesis Water solubility Higher water solubility will allow more of a contaminant to be found in the river water, but
solubility depends on pressure and temperature and each compound reacts differently to changes in temperature. Long range transport possible for water soluble chemicals. Water soluble chemicals are more likely to enter the river system in run off or leaching.
Vapor pressure A contaminant with a low vapor pressure will likely be found in its liquid or solid forms. Contaminants with a high vapor pressure are often referred to as volatile and will not likely remain in their liquid or solid forms. Vapor pressure increases with an increase in temperature. Long range transport less likely with high vapor pressure chemicals because they will not remain in the water phase.
Water solubility × water temperature
As water temperature increases, water solubility increases, increasing the concentration of the chemical.
Vapor pressure × water temperature
As water temperature increases, vapor pressure increases, which may decrease the concentration of the chemical in the water.
Log Kow × turbidity Contaminants with a high log Kow will be more likely to bind to sediment and suspended particles in the water, increasing the concentration in the water column.
River discharge × Watershed Urban
The concentration of chemicals from the urban land-use in the watershed will be diluted with high river discharge
River discharge × urban riparian
The concentration of chemicals from near urban sources will be diluted with high river discharge
64
Table 3.2. Mean, standard deviation (SD), and range of water temperature, turbidity, and river discharge at each river sampled. Water temperature C
River Spring Fall Mean (SD) Min-Max Mean (SD) Min-Max Altamaha 17.8 (0.9) 16.1-21.9 24.0 (1.7) 21.3-27.2 Ocmulgee 13.5 (0.9) 12.8-14.1 22.5 (2.5) 20.7-24.3 Oconee 13.9 (2.2) 12.1-16.6 21.4 (1.7) 19.8-23.4 Ohoopee 19.9 (4.4) 16.8-23 18.8 (3.3) 16.5-21.1 Ogeechee 13.9 (3.6) 8.6-18.1 19.8 (3.3) 15.0-23.9
Turbidity NTU Spring Fall Mean (SD) Min-Max Mean (SD) Min-Max Altamaha 18.9 (4.7) 12.9-28.6 22.9 (12.5) 10.0-37.6 Ocmulgee 42.9 (22.8) 26.7-59.0 18.5 (20.2) 4.2-32.7 Oconee 38.2 (26.4) 16.2-70.0 13.0 (1.7) 3.1-22.1 Ohoopee 8.7 (6.2) 4.4-13.1 4.4 (2.5) 2.7-6.2 Ogeechee 13.5 (14.6) 1.8-47.8 12.8 (10.2) 3.6-22.4
River Discharge m3/s Spring Fall Mean (SD) Min-Max Mean (SD) Min-Max Altamaha 403.4 (122.5) 148.4-574.8 51.1 (6.5) 43.3-61.4 Ocmulgee 69.7 (44.5) 38.2-101.1 12.4 (3.2) 10.2-14.7 Oconee 229.6 (305.6) 9.6-679.6 59.9 (98.3) 7.4-207.3 Ohoopee 39.8 (38.2) 12.7-66.8 2.0 (1.3) 1.0-2.9 Ogeechee 91.7 (42.5) 41.9-171.9 4.0 (1.6) 2.7-7.7
65
Table 3.3. Mean and range (in parentheses) of landscape characteristics for each sample location, summarized by major water body.
River Sampled
No. point source discharges
Watershed Urban (percent)
Riparian Row Crop (percent)
Watershed Agriculture (percent)
Riparian Urban (percent)
Riparian Wetland (percent)
Distance to urban area (km)
Distance to point source (km)
Altamaha Watershed Altamaha River 350
(334-357) 4.1
(1.7-10.1) 6.7
(6.5-6.9) 10.9
(10.5-11.2) 1.2
(1.1-1.3) 16.7
(15.8-17.5) 223
(143-301) 36
(5-72) Ocmulgee River 176 14 2.1 3.4 3.3 10.1 8 7 Oconee River 116
(106-125) 2.4
(1.6-3.2) 3.3
(2.0-4.6) 5
(3.1-7.0) 0.7
(0.6-0.7) 12.4
(11.0-13.7) 21
(15-27) 22
(21-24) Ohoopee River 16 1.6 9.3 15 0.2 21.4 602 32 Ogeechee Watershed Ogeechee River 23
(7-29) 3.7
(0.2-8.0) 9.3 (
2.4-11.8) 15.1
(3.5-19.0) 0.1
(0.0-0.2) 21.1
(11.76-24.9) 226
(85-602) 129
(81-228)
66
Table 3.4. Candidate models relating environmental variables to organochlorine chemical concentrations, number of parameters (K), log likelihood (LogL), AICc, ∆AICc, and weights of evidence, wi for the relationship between organochlorine contaminants and environmental variables.
Candidate Model K AICc ∆AICc wi Percent agriculture in the riparian area, log Kow, turbidity, log Kow × turbidity, water solubility, water temperature, water solubility × water temperature, vapor pressure, vapor pressure × water temperature, river discharge, percent agriculture in the riparian area × river discharge
16 52026.1 0.00 0.965
Percent agriculture in the riparian area, percent agriculture in the watershed, percent urban in the riparian area, percent urban in the watershed, distance to nearest high urban, log Kow, turbidity, log Kow × turbidity, vapor pressure, water temperature, vapor pressure × water temperature, water solubility, water solubility × water temperature, river discharge, distance to nearest high urban × river discharge, riparian urban x river discharge, watershed urban x river discharge, riparian row crop × river discharge
23 52032.6 6.60 0.036
Distance to nearest high urban, percent urban in the riparian area, log Kow, tubidity, log Kow × turbidity, vapor pressure, water temperature, vapor pressure × water temperature, river discharge, distance to nearest high urban × river discharge, percent urban in the riparian area × river discharge, water solubility, water temperature × water solubility
18 52067.6 41.56 0.000
Percent agriculture in the watershed, log Kow, turbidity, log Kow × turbidity, water solubility, water temperature, water solubility × water temperature, vapor pressure, vapor pressure × water temperature, river discharge
15 52078.6 52.51 0.000
Percent urban in the watershed, number of point sources in watershed, log Kow, turbidity, log Kow × turbidity, vapor pressure, water temperature, vapor pressure × water temperature, water solubility, water solubility × water temperature, river discharge, percent urban in the watershed × river discharge
17 52079.5 53.50 0.000
Distance to nearest high urban, percent urban in the riparian area, log Kow, turbidity, log Kow × turbidity, vapor pressure, water temperature, vapor pressure × water temperature, river discharge, distance to nearest high urban × river discharge, percent urban in the riparian area × river discharge
16 52093.9 67.87 0.000
Percent urban in the watershed, number of point sources in watershed, log Kow, turbidity, log Kow × turbidity, vapor pressure, water temperature, vapor pressure × water temperature, river discharge
14 52106.2 80.14 0.000
67
Table 3.4 continued. Candidate models relating environmental variables to organochlorine chemical concentrations, number of parameters (K), log likelihood (LogL), AICc, ∆AICc, and weights of evidence, wi for the relationship between organochlorine contaminants and environmental variables.
Candidate Model K AICc ∆AICc wi Distance to nearest high urban, percent urban in the riparian area, turbidity, vapor pressure, water temperature, vapor pressure × water temperature, river discharge, distance to nearest high urban × river discharge, percent urban in the riparian area × river discharge, water solubility, water solubility × water temperature
17 52117.6 91.52 0.000
Watershed urban, number of point sources in watershed, turbidity, vapor pressure, water temperature, vapor pressure × water temperature, river discharge, watershed urban × river discharge, water solubility, water solubility × water temperature
15 52129.7 103.75 0.000
Distance to nearest high urban, riparian urban, log Kow, turbidity, log Kow × turbidity, river discharge, distance to nearest high urban × river discharge, riparian urban x river discharge
13 52137.6 111.55 0.000
Percent agriculture in watershed, log Kow, turbidity, log Kow × turbidity, water solubility, water solubility × water temperature, vapor pressure, vapor pressure × water temperature
13 52170.0 144.04 0.000
Percent agriculture in riparian area, log Kow, turbidity, log Kow x turbidity, water solubility, water temperature, water solubility × water temperature, vapor pressure, vapor pressure × water temperature
14 52170.1 144.04 0.000
Distance to nearest high urban, percent urban in the riparian area, log Kow, turbidity, log Kow × turbidity, water solubility, water temperature, water solubility × water temperature, vapor pressure, vapor pressure × water temperature
15 52171.8 145.81 0.000
Distance to nearest high urban, percent urban in the riparian area, log Kow, turbidity, log Kow × turbidity, water temperature, vapor pressure, vapor pressure × water temperature
13 52198.0 171.99 0.000
68
Table 3.5. Parameter estimates and upper and lower 95% confidence intervals for the best approximating hierarchical model the relationship between organochlorine contaminants and environmental variables. Parameter Estimate Lower CI Upper CI Fixed effects Intercept -2.1804 (0.4851) -3.7241 -0.6367 Percent agriculture in riparian area -0.8970 (1.1428) -3.1371 1.3431 Log Kow -0.1665 (0.0556) -0.2755 -0.0575 Turbidity 0.0298 (0.0033) 0.0234 0.0362 Log Kow × turbidity -0.0030 (0.0005) -0.0039 -0.0021 Water solubility -0.7666 (0.1469) -1.0546 -0.4787 Water temperature 0.0252 (0.0034) 0.0173 0.0330 Water solubility × water temperature
0.0173 (0.0030) 0.0095 0.0252
Vapor pressure -22.4790 (10.9949) -44.0302 -0.9278 Water temperature × vapor pressure
0.6073 (0.3346) -0.0486 1.2633
River discharge -0.0031 (0.0003) -0.0037 -0.0025 Percent agriculture in riparian area × river discharge
0.0400 (0.0054) 0.0294 0.0507
Random effects Intercept 0.4592 (0.0610) 0.3597 0.6069 Site 0.0173 (0.0078) 0.0084 0.0539 Sample period 0.3202 (0.2270) 0.1147 2.6655 Residual 1.3304 (0.0147) 1.3020 1.3597
69
Figure 3.1. Sampling locations (triangle) in the Altamaha and Ogeechee River watersheds in Georgia, USA.
70
Figure 3.2. Average daily discharge for the USGS gauge station, 02226000, Altamaha River at Doctortown, Georgia observed during the sampling periods, bold line, and the long-term monthly averages, dotted line.
71
Figure 3.3. Average organochlorine pesticide concentrations during the spring (light gray) and fall (dark gray) sampling seasons at each site.
72
Figure 3.4. Average polychlorinated biphenyl concentrations during the spring (light gray) and fall (dark gray) sampling seasons at each site.
73
Riparian row crop x River discharge
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
Riparian row crop
log Kow
Turbidity
log Kow x Turbidity
Water solubility
Water temperature
Water solubility
x
Water temperature
Vapor pressure
Water temperature x
vapor pressu
re
River discharge
Stan
dard
ized
Par
amet
er E
stim
ate
Riparian row crop x River discharge
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
Riparian row crop
log Kow
Turbidity
log Kow x Turbidity
Water solubility
Water temperature
Water solubility
x
Water temperature
Vapor pressure
Water temperature x
vapor pressu
re
River discharge
Stan
dard
ized
Par
amet
er E
stim
ate
Figure 3.5. Parameter estimates from the best approximating model were standardized to demonstrate the relative effect each parameter has on the concentration of organochlorine chemicals in the water column.
74
Figure 3.6. The relationship between discharge and organochlorine chemical concentration predicted from the back transformed estimates from the best fitting hierarchical model. The thin solid line represents a high percentage of riparian row crop agriculture, the dashed line represents a low percentage of riparian row crop agriculture, and the thick black line represents the average observed percentage of row crop agriculture.
75
Figure 3.7. The relationship between water solubility and organochlorine chemical concentration predicted from the back transformed estimates from the best fitting hierarchical model. The thin solid line represents a high water temperature, the dashed line represents a low water temperature, and the thick black line represents the average observed water temperature.
76
Figure 3.8. The relationship between turbidity and organochlorine chemical concentration predicted from the back transformed estimates from the best fitting hierarchical model. The thin solid line represents a high log Kow, the dashed line represents a low log Kow, and the thick black line represents the average observed log Kow.
77
LITERATURE CITED
1. Moore, J. W., S. Ramamoorthy, Organic Chemicals in Natural Waters - Applied Monitoring and Impact Assessment. Springer-Verlag: New York, 1984.
2. NRC, Polychlorinated biphenyls. In National Academy of Sciences: Washington, DC, 1979.
3. Cookson Jr., J. T., Bioremediation engineering: designand application. McGraw Hill: New York, 1995.
4. Erickson, M., Analytical Chemistry of PCBs. 2 ed.; CRC-Lewis Publishers: New York, 1997.
5. Borga, K.; Fisk, A. T.; Hoekstra, P. F.; Muir, D. C. G., Biological and chemical factors of importance in the bioaccumulation and trophic transfer of persistent organochlorine contaminants in arctic marine food webs. Environmental Toxicology and Chemistry 2004, 23, (10), 2367-2385.
6. Braune, B. M.; Outridge, P. M.; Fisk, A. T.; Muir, D. C. G.; Helm, P. A.; Hobbs, K.; Hoekstra, P. F.; Kuzyk, Z. A.; Kwan, M.; Letcher, R. J.; Lockhart, W. L.; Norstrom, R. J.; Stern, G. A.; Stirling, I., Persistent organic pollutants and mercury in marine biota of the Canadian Arctic: An overview of spatial and temporal trends. Science of the Total Environment 2005, 351, 4-56.
7. Campbell, L. M.; Muir, D. C. G.; Whittle, D. M.; Backus, S.; Norstrom, R. J.; Fisk, A. T., Hydroxylated PCBs and other chlorinated phenolic compounds in lake trout (Salvelinus namaycush) blood plasma from the Great Lakes Region. Environmental Science & Technology 2003, 37, (9), 1720-1725.
8. Fisk, A. T.; de Wit, C. A.; Wayland, M.; Kuzyk, Z. Z.; Burgess, N.; Robert, R.; Braune, B.; Norstrom, R.; Blum, S. P.; Sandau, C.; Lie, E.; Larsen, H. J. S.; Skaare, J. U.; Muir, D. C. G., An assessment of the toxicological significance of anthropogenic contaminants in Canadian arctic wildlife. Science of the Total Environment 2005, 351, 57-93.
9. Hoekstra, P. F.; O'Hara, T. M.; Fisk, A. T.; Borga, K.; Solomon, K. R.; Muir, D. C. G., Trophic transfer of persistent orgranochlorine contaminants (OCs) within an Arctic marine food web from the southern Beaufort-Chukchi Seas. Environmental Pollution 2003, 124, (3), 509-522.
10. Patlak, M., A testing deadline for endocrine disrupters. Environmental Science & Technology 1996, 30, (12), A540-A544.
11. Sonne, C.; Leifsson, P. S.; Dietz, R.; Born, E. W.; Letcher, R. J.; Hyldstrup, L.; Riget, F. F.; Kirkegaard, M.; Muir, D. C. G., Xenoendocrine pollutants may reduce size of sexual organs in East Greenland polar bears (Ursus maritimus). Environmental Science & Technology 2006, 40, (18), 5668-5674.
78
12. Soto, A. M.; Chung, K. L.; Sonnenschein, C., The Pesticides Endosulfan, Toxaphene, and Dieldrin Have Estrogenic Effects on Human Estrogen-Sensitive Cells. Environmental Health Perspectives 1994, 102, (4), 380-383.
13. Hutzinger, O., Chemistry of PCBs. Westport Publishing Group: Englewood Cliffs, NJ, 1974.
14. Pereira, W. E.; Domagalski, J. L.; Hostettler, F. D.; Brown, L. R.; Rapp, J. B., Occurrence and accumulation of pesticides and organic contaminants in river sediment, water and clam tissues from the San Joaquin River and tributaries, California. Environmental Toxicology and Chemistry 1996, 15, (2), 172-180.
15. USGS, National Elevation Dataset. In USGS, Ed. 1999.
16. USGS, National Hydrography Dataset. In USGS, Ed. 2001.
17. Natural Resources Spatial Analysis Laboratory. 1998 vegetation/land cover map of Georgia. In Institute of Ecology, University of Georgia, Athens, GA, 2003.
18. Öberg, T., Prediction of physical properties for PCB congeners from molecular descriptors. Internet Journal of Chemistry 2001, 4, (11).
19. University of Akron, D. o. C., The Chemical Database. In 2007.
20. Snijders, T. a. B. R., Multilevel analysis: an introduction to basic and advanced multilevel modeling. Sage: Thousand Oaks, California, 1999.
21. Burnham, K. P. and Anderson, D. R., Model selection and inference: a practical information-theoretic approach. 2 ed.; Springer-Verlag: New York, 2002.
22. Akaike, H., Information theory and an extension of the maximum likelihood principle. In Second International Symposium on Information Theory, Csaki, B. N. P. a. F., Ed. Akademiai Kiado: Budapest, Hungary, 1973; pp 267-281.
23. Hurvich, C. M.; Tsai, C. L., Bias of the Corrected Aic Criterion for Underfitted Regression and Time-Series Models. Biometrika 1991, 78, (3), 499-509.
24. Anderson, D. R.; Burnham, K. P.; Thompson, W. L., Null hypothesis testing: Problems, prevalence, and an alternative. Journal of Wildlife Management 2000, 64, (4), 912-923.
25. Royall, R. M., Statistical evidence: a likelihood paradigm. Chapman and Hall: New York, 1997.
26. Wolfinger, R. D., Heterogeneous Variance - Covariance Structures for Repeated Measures. Journal of Agricultural, Biological, and Environmental Statistics 1996, (1), 205-230.
27. Sokal, R. R. and F. J. Rohlf., Biometry: the principles and practice of statistics in biological research. Freeman: New York, 1995.
79
28. Yu, M.; Luo, X. J.; Chen, S. J.; Mai, B. X.; Zeng, E. Y., Organochlorine pesticides in the surface water and sediments of the Pearl River Estuary, South China. Environmental Toxicology and Chemistry 2008, 27, (1), 10-17.
29. Laabs, V.; Amelung, W.; Pinto, A. A.; Wantzen, M.; da Silva, C. J.; Zech, W., Pesticides in surface water, sediment, and rainfall of the northeastern Pantanal basin, Brazil. Journal of Environmental Quality 2002, 31, (5), 1636-1648.
30. Konstantinou, I. K.; Hela, D. G.; Albanis, T. A., The status of pesticide pollution in surface waters (rivers and lakes) of Greece. Part I. Review on occurrence and levels. Environmental Pollution 2006, 141, (3), 555-570.
31. Borja, J.; Taleon, D. M.; Auresenia, J.; Gallardo, S., Polychlorinated biphenyls and their biodegradation. Process Biochemistry 2005, 40, (6), 1999-2013.
32. Gordon, N. D., T.A. McMahon, B.L. Finlayson, Stream Hydrology: An introduction for ecologists. John Wiley & Sons Ltd.: West Sussex, England, 1992.
33. Iannuzzi, T. J.; Armstrong, T. N.; Thelen, J. B.; Ludwig, D. F.; Firstenberg, C. E., Characterization of chemical contamination in shallow-water estuarine habitats of an industrialized river. Part 1: Organic compounds. Soil & Sediment Contamination 2005, 14, (1), 13-33.
34. So, M. K.; Taniyasu, S.; Yamashita, N.; Giesy, J. P.; Zheng, J.; Fang, Z.; Im, S. H.; Lam, P. K. S., Perfluorinated compounds in coastal waters of Hong Kong, South China, and Korea. Environmental Science & Technology 2004, 38, (15), 4056-4063.
35. Zhou, J. L.; Maskaoui, K.; Qiu, Y. W.; Hong, H. S.; Wang, Z. D., Polychlorinated biphenyl congeners and organochlorine insecticides in the water column and sediments of Daya Bay, China. Environmental Pollution 2001, 113, (3), 373-384.
36. Weigel, J. Q. W., Microbial reductive dehalogenation of polychlorinated biphenyls. FEMS Microbial Ecology 2000, 32, 1-15.
37. USEPA, Ambient water quality criteria for polychlorinated biphenyls (PCBs). In Agency, E. P., Ed. 1992.
80
CHAPTER 4
CONCLUSIONS
Currently used perfluorinated compounds (PFCs) and historically used organochlorine
(OC) pesticides and polychlorinated biphenyls (PCBs) were detected in the surface waters of
southeast Georgia. Organic contaminants enter the environment through manufacturing, use,
disposal, and accidental spills and remain in the environment due to their inability to be degraded
naturally [1, 2]. OC contaminants were more ubiquitous in the surface waters, being present in
greater quantities and at more locations than PFCs. In addition, OC contaminants were
associated with agricultural land-use while PFCs were highly related with urban sources.
Environmental factors such as discharge, water temperature, and turbidity were measured as
surrogates of surface run-off and potentially influential for OC chemical delivery.
Perfluorinated compounds have many industrial and municipal uses such as industrial
surfactants, fire-fighting foam, and surface repellents for fabrics and carpets [3]. PFCs have
been detected in effluent from wastewater treatment plants, indicating they are not broken down
by traditional treatment [4-6]. The PFCs detected most often in this and other studies were
perfluoroctanoic acid (PFOA) and perfluorooctane sulfonate (PFOS) which have direct uses and
are also terminal degradation products of higher carbon chain length PFCs [7]. Near urban
sources were most influential to PFC contaminant presence and concentration in surface waters.
Distance to the nearest urban area was the most and important predictor of downstream
concentration of PFCs in river systems. PFCs are stable in the aquatic environment so they are
81
able to be transported downstream once they enter the river, however, as more tributaries enter
the river, dilution occurs which causes PFC concentrations to decrease further from an urban
point source. To prevent and control PFC contamination in surface waters, we must be able to
break down PFCs before they enter surface waters perhaps by including an additional phase to
industrial or municipal wastewater treatment plants.
OC chemicals were banned from use in the United States and Canada in the 1970s and
1980s but are routinely found in the environment. OC contaminants released into the
environment 30-40 years ago remain in the system because they bind strongly to soils and
sediments where no biodegradation takes place [1, 8]. In addition, OC contaminants are in flux
in the environment, volatilizing into the air where long and short range transport may take place
therefore no distinct land-use sources are currently present [9, 10]. Density of row crop
agricultural fields within one km of the stream was the most influential land-use predicting OC
contaminant concentrations in the river. Specifically, greater OC concentrations were detected
during high discharge events at sites with more row crop agriculture near the streams in the
watershed. In addition, greater OC concentrations were detected in highly turbid waters. Both
high discharge and turbidity are related to surface run-off from precipitation events within the
catchment area [11]. Surface run-off from agricultural fields which may have a reservoir of
legacy use OC chemicals bound to the soils is delivering OC contaminants to the stream.
Increasing forested riparian buffer zones may prevent OC contaminants from running off into the
stream.
PFCs are currently used for a variety of industrial and commercial purposesand there is a
general understanding of how they are entering aquatic systems, therefore we were able to
develop a useful model to predict PFC concentrations among varying riverine sites. On the other
82
hand, OC chemicals are not currently used and thus a direct source is less apparent for this group
of chemicals. However, a greater understanding of OC chemical properties exists and by
including this information in the model, we were able to understand how the chemical properties
may affect the source or delivery for OC contaminants. Physical-chemicals properties are not
known for each of the different types of perfluorinated chemicals so we were not able to include
these parameters in our analysis. The more that we know about a system and the more useful
information we can include in the models, the better the models will be. By including multiple
working hypotheses on how the system works and using an information theoretic approach [12],
the best approximating model given the data can be determined. AIC is an entropy-based
approach which penalizes a model the number of parameters it includes, therefore the best
approximating model is a combination of the best fitting model and the simplest model.
Prior to this research, little was known about anthropogenic contaminant presence,
concentrations, or transport through Georgia waters. Spatial distributions of OC chemicals and
PFCs as well as temporal trends have been determined in coastal Georgia rivers through this
research. Seasonal stream flow patterns influence the concentration of PFCs and OC chemical
concentrations in the water column which demonstrates the importance of multi-season
monitoring of anthropogenic contaminants. The quantified relationships between OC and PFC
contaminants with environmental variables has improved our understanding of how these
chemicals are behaving in the environment by allowing us to focus on likely sources and
delivery mechanisms. Future efforts on PFCs should focus on where these chemicals occur in
the aquatic system and their likely exposure pathways to wildlife while future work should focus
on the mitigation and prevention of OC contaminants from entering the aquatic system.
83
LITERATURE CITED
1. Cookson Jr., J. T., Bioremediation engineering: designand application. McGraw Hill: New York, 1995.
2. Erickson, M., Analytical Chemistry of PCBs. 2 ed.; CRC-Lewis Publishers: New York, 1997.
3. Kissa, E., Fluorinated Surfactants and Repellents. 2nd ed.; Marcel Dekker: New York, 2001.
4. Boulanger, B.; Vargo, J. D.; Schnoor, J. L.; Hornbuckle, K. C., Evaluation of perfluorooctane surfactants in a wastewater treatment system and in a commercial surface protection product. Environmental Science & Technology 2005, 39, (15), 5524-5530.
5. Loganathan, B. G.; Sajwan, K. S.; Sinclair, E.; Kumar, K. S.; Kannan, K., Perfluoroalkyl sulfonates and perfluorocarboxylates in two wastewater treatment facilities in Kentucky and Georgia. Water Research 2007, 41, (20), 4611-4620.
6. Sinclair, E.; Kannan, K., Mass loading and fate of perfluoroalkyl surfactants in wastewater treatment plants. Environmental Science & Technology 2006, 40, (5), 1408-1414.
7. Ellis, D. A.; Martin, J. W.; De Silva, A. O.; Mabury, S. A.; Hurley, M. D.; Andersen, M. P. S.; Wallington, T. J., Degradation of fluorotelomer alcohols: A likely atmospheric source of perfluorinated carboxylic acids. Environmental Science & Technology 2004, 38, (12), 3316-3321.
8. Hutzinger, O., Chemistry of PCBs. Westport Publishing Group: Englewood Cliffs, NJ, 1974.
9. Hung, H.; Blanchard, P.; Halsall, C. J.; Bidleman, T. F.; Stern, G. A.; Fellin, P.; Muir, D. C. G.; Barrie, L. A.; Jantunen, L. M.; Helm, P. A.; Ma, J.; Konoplev, A., Temporal and spatial variabilities of atmospheric polychlorinated biphenyls (PCBs), organochlorine (OC) pesticides and polycyclic aromatic hydrocarbons (PAHs) in the Canadian Arctic: Results from a decade of monitoring. Science of the Total Environment 2005, 342, (1-3), 119-144.
10. Bidleman, T. F.; Leone, A. D., Soil-air exchange of organochlorine pesticides in the Southern United States. Environmental Pollution 2004, 128, (1-2), 49-57.
11. Gordon, N. D., T.A. McMahon, B.L. Finlayson, Stream Hydrology: An introduction for ecologists. John Wiley & Sons Ltd.: West Sussex, England, 1992.
12. Akaike, H., Information theory and an extension of the maximum likelihood principle. In Second International Symposium on Information Theory, Csaki, B. N. P. a. F., Ed. Akademiai Kiado: Budapest, Hungary, 1973; pp 267-281.
84
APPENDIX A
SUPPLEMENTAL TABLES FOR CHAPTER 3
85
Appendix A. Organochlorine contaminant physical chemical properties [65, 66].
OC Chemical
Molecular Weight
(g)
Vapor Pressure (mm Hg)
Water Solubility
(mg/L) log Kow 1,2,3,4 tetrachlorobenzene 215.89 0.0420 0.0000 4.60 1,2,4,5 tetrachlorobenzene 215.89 0.0420 0.0000 4.60 pentachlorobenzene 250.34 0.0080 0.0000 5.18 hexachlorobenzene 284.78 0.0023 0.0018 5.50 alpha-hexachlorcyclohexane 290.83 0.0017 0.0000 3.94 beta-hexachlorocyclohexane 290.83 0.0003 0.0000 3.78 gama-hexachlorocyclohexane 290.83 0.0003 0.0000 4.14 heptachlor 373.32 0.0530 0.0150 5.30 octachlorostyrene 379.71 0.0023 0.0018 5.50 oxychlordane 423.77 0.0014 0.0000 3.45 trans-chlordane 409.78 0.0005 0.0137 6.00 cis-chlordane 409.78 0.0004 0.0137 6.00 trans-nonachlor 444.23 0.0005 0.0000 4.28 cis-nonachlor 444.23 0.0004 0.0000 4.28 heptachlor epoxide 389.32 0.0000 0.0000 5.40 dieldrin 380.91 0.0000 0.0000 5.20 ortho,para-DDE 318.03 0.0009 0.0126 5.80 para,para-DDE 318.03 0.0000 0.0000 6.51 ortho,para-DDD 320.05 0.0000 0.0000 6.02 para,para-DDD 320.05 0.0000 0.0000 6.02 ortho,para-DDT 354.49 0.0000 0.0073 6.20 para,para-DDT 354.49 0.0000 0.0016 6.20 mirex 545.55 0.0001 0.0000 6.90 pentachloroanisole 280.37 0.0006 0.0000 5.45 endosulfan 406.92 0.0000 0.0100 3.83 methoxychlor 345.65 0.0000 0.0000 5.08 endrin 380.91 0.0000 0.0604 5.20 PCB 1 188.66 0.0014 4.8300 4.53 PCB 3 188.66 0.0105 1.3400 4.61 PCB 10/4 223.09 0.0026 2.1300 4.98 PCB 5 223.10 0.0016 0.9970 5.02 PCB 6 223.09 0.0016 0.6100 5.02
86
Appendix A Continuted.
OC Chemical
Molecular Weight
(g)
Vapor Pressure (mm Hg)
Water Solubility
(mg/L) log Kow PCB 7 223.09 0.0014 1.1500 5.16 PCB 8 223.09 0.0021 1.1700 5.09 PCB 16/32 257.54 0.0005 0.1885 5.53 PCB 17 257.54 0.0004 0.0833 5.76 PCB 18 257.54 0.0011 0.4000 5.55 PCB 19 257.54 0.0011 0.3240 5.48 PCB 22 257.54 0.0003 0.0875 5.42 PCB 24 257.54 0.0006 0.0833 5.67 PCB 25 257.54 0.0003 0.0888 5.75 PCB 26 257.54 0.0003 0.2530 5.76 PCB 27 257.54 0.0006 0.2050 5.54 PCB 28 257.54 0.0002 0.2700 5.62 PCB 31 257.54 0.0004 0.1430 5.69 PCB 33 257.54 0.0001 0.1150 5.87 PCB 40 292.01 0.0001 0.0156 6.18 PCB 41/71 292.01 0.0001 0.0356 6.13 PCB 42 292.01 0.0001 0.0298 6.21 PCB 44 292.01 0.0001 0.0382 5.81 PCB 45 292.01 0.0002 0.1460 5.99 PCB 46 292.01 0.0002 0.1460 5.97 PCB 47 292.01 0.0001 0.0226 6.29 PCB 48 292.01 0.0001 0.0164 6.13 PCB 49 292.01 0.0000 0.0322 6.22 PCB 52 292.01 0.0002 0.0153 6.09 PCB 60/56 292.01 0.0001 0.0182 6.07 PCB 64 292.01 0.0001 0.0256 6.21 PCB 66 292.01 0.0000 0.0159 6.31 PCB 70 292.01 0.0000 0.0410 6.23 PCB 74 292.01 0.0000 0.0050 6.67 PCB 82 326.45 0.0000 0.0067 6.82 PCB 83 326.45 0.0000 0.0082 6.78 PCB 84 326.45 0.0000 0.0542 6.04 PCB 85 326.45 0.0000 0.0078 6.61 PCB 87 326.45 0.0000 0.0070 6.85 PCB 91 326.45 0.0000 0.0221 6.70 PCB 95 326.45 0.0000 0.0135 6.55 PCB 97 326.45 0.0000 0.0074 6.67
87
Appendix A Continued.
OC Chemical
Molecular Weight
(g)
Vapor Pressure (mm Hg)
Water Solubility
(mg/L) log Kow PCB 99 326.45 0.0000 0.0037 7.21 PCB 101 326.45 0.0000 0.0154 6.80 PCB 105 326.45 0.0000 0.0034 6.94 PCB 110 326.45 0.0000 0.0073 6.22 PCB 114 326.45 0.0000 0.0037 6.95 PCB 118 326.45 0.0000 0.0039 7.12 PCB 128 360.88 0.0000 0.0004 7.31 PCB 178/129 378.10 0.0000 0.0012 7.56 PCB 130/176 378.10 0.0000 0.0023 7.51 PCB 131 360.88 0.0000 0.0012 7.25 PCB 132 360.88 0.0000 0.0081 7.04 PCB 134 360.88 0.0000 0.0009 7.25 PCB 144/135 360.88 0.0000 0.0045 7.18 PCB 136 360.88 0.0000 0.0045 7.12 PCB 137 360.88 0.0000 0.0013 7.44 PCB 138 360.88 0.0000 0.0015 7.44 PCB 141 360.88 0.0000 0.0021 7.19 PCB 146 360.88 0.0000 0.0009 7.12 PCB 149 360.88 0.0000 0.0042 7.28 PCB 151 360.88 0.0000 0.0035 7.16 PCB 153 360.88 0.0000 0.0010 7.75 PCB 156 360.88 0.0000 0.0004 7.60 PCB 201/157 395.33 0.0000 0.0006 7.90 PCB 158 360.88 0.0000 0.0012 7.37 PCB 170 395.33 0.0000 0.0004 7.94 PCB 171 395.33 0.0000 0.0022 7.83 PCB 172/197 412.55 0.0000 0.0003 8.07 PCB 174 395.33 0.0000 0.0010 7.76 PCB 175 395.33 0.0000 0.0008 7.81 PCB 177 395.33 0.0000 0.0015 7.79 PCB 179 395.33 0.0000 0.0045 7.58 PCB 180 395.33 0.0000 0.0004 7.92 PCB 183 395.33 0.0000 0.0005 7.86 PCB 185 395.33 0.0000 0.0008 7.93 PCB 187 395.33 0.0000 0.0007 7.79 PCB 189 395.33 0.0000 0.0008 8.00
88
Appendix A Continuted.
OC Chemical
Molecular Weight
(g)
Vapor Pressure (mm Hg)
Water Solubility
(mg/L) log Kow PCB 190 395.33 0.0000 0.0004 7.91 PCB 191 395.33 0.0000 0.0003 7.93 PCB 193 395.33 0.0000 0.0003 7.92 PCB 194 429.78 0.0000 0.0003 8.68 PCB 195 429.78 0.0000 0.0002 8.35 PCB 196/203 429.78 0.0000 0.0001 8.33 PCB 198 429.78 0.0000 0.0002 8.37 PCB 199 429.78 0.0000 0.0003 8.29 PCB 200 429.78 0.0000 0.0003 8.20 PCB 205 429.78 0.0000 0.0001 8.47 PCB 206 464.22 0.0000 0.0000 9.14 PCB 207 464.22 0.0000 0.0000 8.91 PCB 208 464.22 0.0000 0.0000 8.16
PCB 209 498.66 0.0000 0.0000 8.27
89
LITERATURE CITED
1. University of Akron, Department of Chemistry, The Chemical Database. In 2007.
2. Öberg, T., Prediction of physical properties for PCB congeners from molecular descriptors. Internet Journal of Chemistry 2001, 4, (11).