Science of the Total Environment 607–608 (2017) 1320–1338

Contents lists available at ScienceDirect

Science of the Total Environment

journal homepage: www.elsevier.com/locate/scitotenv

Ambient sediment quality conditions in Minnesota lakes, USA: Effects of

watershed parameters and aquatic health implications
Judy L. Crane
Minnesota Pollution Control Agency, 520 Lafayette Road North, St. Paul, MN 55155-4194, USA

H I G H L I G H T S G R A P H I C A L A B S T R A C T

• Information on ambient sediment

quality has been lacking in the state of
Minnesota.
• Surficial sediments were collected
from 54 lakes for sediment chemistry
parameters.
• Overall sediment quality ranged from
good (43%) to moderate (57%).
• Some contaminants varied significantly
in lakes from different watershed land
uses.
• Ecological health impacts were assessed
for benthic invertebrates and fish.

a r t i c l e i n f o a b s t r a c t

Article history: Surficial sediments were collected from 50 randomly selected Minnesota lakes, plus four a priori reference lakes,
Received 2 April 2017 in 2007. The lakes encompassed broad geographic coverage of the state and included a variety of major land uses
Received in revised form 25 May 2017 in the surrounding watersheds. Sediment samples were analyzed for a suite of metals, metalloids, persistent or-
Accepted 26 May 2017
ganic pollutants, total organic carbon, and particle size fractions. In addition, a small fish survey was conducted to
Available online xxxx
assess PBDEs in both whole fish and fish tissues. Sediment quality in this set of lakes ranged from good (43%) to
Editor: Kevin V. Thomas moderate (57%) based on an integrative measure of multiple contaminants. On an individual basis, some contam-
inants (e.g., arsenic, lead, DDD, and DDE) exceeded benchmark values in a small number of lakes that would be
Keywords: detrimental to benthic invertebrates. The sediments in two developed lakes tended to be more contaminated
Sediment chemistry than sediments in lakes from other major watershed land uses. These differences were often statistically signif-
Metals icant (p b 0.05), particularly for lakes with developed versus cultivated land uses for arsenic, lead, zinc, and nu-
Polycyclic aromatic hydrocarbons merous PAH compounds. Multivariate statistical approaches were used on a subgroup of contaminants to show
Persistent organic pollutants the two urban lakes, as well as a few northeastern Minnesota lakes, differed from the rest of the data set. Back-
Land use
ground threshold values were calculated for data with b 80% nondetects. Source apportionment modeling of
Fish tissue
PAHs revealed that vehicle emissions and coal-related combustion were the most common sources. A general en-
vironmental forensic analysis of the PCDD/F data showed that ubiquitous combustion sources appeared to be im-
portant. BDE-209, a decaBDE, was detected in 84% of lake sediment samples, whereas fish at the top of the food
chain (i.e., predator trophic group) had significantly higher (p b 0.05) mean lipid-normalized concentrations of
BDEs-47, 100, and 153 than lower trophic fish. These results will be used for future status and trends work.
Published by Elsevier B.V.

E-mail address: judy.crane@state.mn.us.

http://dx.doi.org/10.1016/j.scitotenv.2017.05.241
0048-9697/Published by Elsevier B.V.
 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338
1. Introduction
Sediment quality is important for healthy aquatic ecosystems. The
bed sediments provide vital habitat for benthic invertebrates and fish,
as well as nearshore vegetation. Depositional sediments are also a
repository for particle-associated contaminants that enter waterways
through surface water runoff, leaking septic systems, atmospheric
deposition, and historical dumping of waste material and effluent
(Taylor and Owens, 2009). The accumulation of contaminants over
time can contribute to beneficial use impairments (BUIs) in waterways,
including: 1) tumors and other deformities in bottom-dwelling fish,
2) degraded benthic communities, which results in less food for fish
and some wildlife, 3) degraded fish and wildlife habitats, 4) bioaccumu-
lation of contaminants up the food chain, which may result in fish
and wildlife consumption advisories for humans, 5) potential human
health risks from exposure to sediment-derived contaminants, 6) aes-
thetic impairments, and 7) restrictions on navigational dredging and
beneficial use of dredged material (Zarull et al., 2001; Crane and
Hennes, 2016a). Contaminated sediments were a leading contributor
of BUIs in the initial designation of 43 Great Lakes Areas of Concern
(AOCs) by the International Joint Commission (IJC, 1989). Worldwide,
contaminated sediments are of concern in coastal ports and other
urban areas, particularly where past or current commercial and indus-
trial operations have contaminated receiving waters (Taylor and
Owens, 2009).
Although developed countries have implemented point source con-
trols on contaminant sources, reductions in nonpoint sources are a more
challenging problem for environmental agencies to address. The diffu-
sive nature of nonpoint sources can have widespread impacts on
water and sediment quality, as well as human and ecological health.
Local, regional, or national product bans are one way to reduce these
sources. Examples of bans in the U.S. include: organochlorine pesticides
like DDT (USEPA, 2000a), polychlorinated biphenyls (PCBs; Johnson
et al., 2006), and coal tar-based sealants in local and state jurisdictions
(including Minnesota) to reduce environmental contamination by poly-
cyclic aromatic hydrocarbons (PAHs; Crane, 2014). Voluntary or negoti-
ated restrictions on producing certain chemicals have also been utilized,
such as when the Great Lakes Chemical Corporation, the sole U.S. man-
ufacturer of certain flame retardants [i.e., penta-brominated diphenyl
ether (BDE) and octaBDE], voluntarily ceased production in 2004
(USEPA, 2009). Implementation of stormwater best management prac-
tices (Minnesota Stormwater Steering Committee, 2017), improved
farming techniques (Centner et al., 1999; Wortmann et al., 2011), use
of mass transit (IAGLR, 2002), and pollution prevention efforts
(Daughton, 2004) can be effective ways to reduce nonpoint sources of
contamination to waterways. However, the physical/chemical proper-
ties of persistent organic pollutants (POPs) can result in continued cy-
cling in the environment (Jones and de Voogt, 1999; Lohmann et al.,
2007; Wöhrnschimmel et al., 2013), even after source control measures
have been implemented. Thus, these contaminants may continue to
persist in surficial sediments for some time.
Ambient sediment quality pertains to chemical concentrations de-
rived from natural and/or widespread diffuse anthropogenic sources
(e.g., atmospheric deposition). Surficial sediments provide the most re-
cent record by which to assess ambient sediment quality conditions,
such as from legacy organic contaminants. The U.S. Geological Survey
(USGS) identified 686 chemicals of highest priority for ambient moni-
toring of sediment quality for national- and regional-scale studies
(Olsen et al., 2013). Determining ambient sediment quality is important
for assessing current conditions, as well as for future status and trends
work. These types of investigations should be conducted using a ran-
domized statistical design for the selection of sample sites.
The purpose of this study was to collect and evaluate ambient sedi-
ment chemistry and particle size data from a subset of Minnesota
lakes located in the upper Midwest, U.S. Surficial sediments were col-
lected from 50 randomly selected lakes, plus four a priori reference
1321
lakes, during the summer of 2007. The sediments were analyzed for a
large suite of metals, metalloids, persistent organic pollutants, total or-
ganic carbon (TOC), and particle size. These data were used to provide
an indication of the statistical range of more recently deposited analytes
and to determine if these concentrations were influenced by major
watershed land uses. The sediment quality data were evaluated using
chemical indices, making comparisons to benchmark ecological health
values, conducting statistical analyses (including multivariate
methods), and incorporating environmental forensic techniques to
identify sources for certain POPs. To further assess biological impacts
from one class of contaminants, a fish survey of five study lakes was
conducted to evaluate BDE congeners in up to three trophic classes of
fish.
2. Methods
2.1. Sampling design
The selection of study lakes for sediment samples was the same as
those selected for the Minnesota component of the 2007 National
Lake Assessment (NLA) study (Monson and Heiskary, 2008; Crane and
Hennes, 2016a, 2016b). Fifty randomly selected lakes, and four a priori
reference lakes, were selected within five size classes of lake surface
areas (Table A-1). The reference lakes were selected by the U.S. Environ-
mental Protection Agency (EPA) in consultation with state agencies, as
well as by evaluation of aerial photographs (Herlihy et al., 2013). The
lake watersheds encompassed a variety of land uses (Table A-2) and
ecoregions within Minnesota (Fig. 1). Due to either drought conditions
during the summer of 2007 or the inaccessibility of some lakes that did
not have public access, some of the original draw lakes were replaced by
the oversample list of randomly selected lakes.
2.2. Sample collection
2.2.1. Sediment samples
The NLA field crews collected sediment samples during June 26,
2007 through August 22, 2007. The sediment sampling and processing
procedures are described elsewhere (Crane and Hennes, 2016a). Briefly,
a modified K-B corer with Lexan core tubes was used to collect surficial
sediment samples from the deepest location of each lake. The upper
15 cm of the sediment profile was collected and extruded into a large,
pre-cleaned 1 L wide-mouthed glass jar with a Teflon-lined lid. For
most lakes, two cores were collected in close proximity to each other,
and each core was extruded into a separate labelled jar. Four lakes in
the Boundary Waters Canoe Area Wilderness (BWCAW) [i.e., Alruss
(#2), Becoosin (#6), Lamb (#17), and Vesper (#47)] were sampled by
canoe since motorized boats are generally not allowed there. Only one
sediment core sample was collected from each BWCAW lake due to
more difficult field conditions. Field replicates, consisting of two cores,
were collected in close proximity to the original sample to assess field
precision for five lakes [i.e., Fairy (#12), Flat (#15), Long (#19),
Nokomis (#27), and Pine Mountain (#35)]. The samples were stored
in ice-filled coolers and transported to St. Paul, MN. The sediment sam-
ples were composited and processed further at the Minnesota Pollution
Control Agency's (MPCA's) St. Paul, MN office before subsamples were
stored and transported to analytical laboratories. The samples were
processed within one to 23 days of sample collection.
2.2.2. Fish tissue samples
Five NLA lakes [i.e., August (#5), Cass (#7), Jennie (#16), Mayo
(#24), and Nokomis (#27)] were sampled for fish by Minnesota Depart-
ment of Natural Resources (MDNR) staff during June to September
2007. The field sampling methods are described elsewhere (Crane and
Hennes, 2016a). Briefly, fish were collected with gill nets or trap nets
from each of the following trophic guilds: predator (i.e., northern pike
or walleye), omnivore (pan fish; i.e., bluegill or yellow perch), and
1322 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338

Fig. 1. Map of random and reference lake sediment samples in Minnesota.

benthic fish (i.e., brown bullhead, yellow bullhead, or white sucker).
However, benthic fish from Mayo (#24) and Nokomis (#27) were inad-
vertently not collected. Five additional predator fish (i.e., northern pike)
from August (#5) were collected for the field replicate. Weight and
length measurements were taken in the field, the fish were wrapped
in aluminum foil and stored on ice in coolers during transport to St.
Paul, MN, and the samples were stored in a walk-in freezer at the
MPCA. Five comparable fish for a particular species from each guild
were selected and shipped overnight on dry ice to AXYS Analytical Ser-
vices Ltd., located in Sidney, BC, Canada, for processing and analysis. The
samples were transferred to secure storage and kept at −20 °C prior to
extraction. The length, sex, and age of the individual fish species were
determined.
2.3. Analytical methods
2.3.1. Sediment
A large suite of sediment chemistry and particle size parameters
were analyzed on the sediment samples. In order to maximize available
resources, three groups of parameters [i.e., PCB congeners, organochlo-
rine pesticides/Total PCBs, and polychlorinated dibenzo-p-dioxin/furan
(PCDD/F) congeners] were limited to a subgroup of 23–24 lakes
(Crane and Hennes, 2016a). All other parameters were analyzed on
the full group of 54 lakes.
A detailed description of the analytical methods is provided else-
where (Crane and Hennes, 2016a). Briefly, the Minnesota Department
of Health (MDH) analyzed 18 metals and metalloids using inductively
 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338
coupled plasma – mass spectrometry (ICP-MS). MDH also analyzed
mercury by cold-vapor atomic absorption (CVAA). All other chemical
and physical analyses of sediment samples were performed by TDI-
Brooks International (College Station, TX); their subcontract laboratory
(Columbia Analytical Services in Houston, TX) analyzed sediment sam-
ples for PCDD/F congeners. The analysis of 43 parent and alkylated
PAHs, as well as biphenyl, was performed using gas chromatography/
mass spectrometry (GC/MS) in selected ion monitoring (SIM) mode.
The analysis of 28 organochlorine pesticides and metabolites, as well
as Total PCBs and 114 individual and coeluting groups of PCB congeners,
were performed using a high resolution GC with electron capture detec-
tion (ECD). Total PCBs was an analytical parameter included with the
organochlorine pesticide analyses, and it is hereafter referred to as
Total PCBspesticide scan. The analysis of 17 PCDD/F congeners that had
chlorine atoms located in the 2,3,7,8 positions, as well as several homo-
log groups, was conducted using high resolution GC (HRGC) – high res-
olution MS (HRMS). Table A-3 provides the abbreviations for individual
PCDD/F congeners and homolog groups. Fifty-five individual and
coeluting homolog groups of polybrominated diphenyl ether (PBDE)
compounds were quantified, including BDE-209, by GC/MS-Negative
Chemical Ionization (GC/MS-NCI) and GC/MS-SIM. The analysis of TOC
was conducted by the Lloyd Kahn method (Kahn, 1988). Particle size
was determined using a combination of sieves and settling tubes on ho-
mogenized subsamples mixed with a deflocculant solution to disaggre-
gate the samples.
Organic contaminants were quantified different ways. Quantitation
of PAHs, PCB congeners, Total PCBspesticide scan, organochlorine pesti-
cides, and PBDE concentrations were determined using the internal
standard method, and analyte concentrations were corrected for surro-
gate recovery. Quantitation of individual PCDD/F congeners, total
PCDDs, and total PCDFs was achieved in conjunction with the establish-
ment of a multipoint (five-point) calibration curve for each homolog,
during which each calibration solution was analyzed once. Percent
moisture was determined so that all results were reported as dry weight
measurements.
A variety of quality control (QC) samples and quality assurance (QA)
procedures were used by the analytical laboratories (TDI-Brooks Inter-
national, Inc., 2009; Crane and Hennes, 2016a). For metals and metal-
loids, the QC samples included: mixed calibration standard solutions,
analytical blanks, post-digestion spike additions, and one analytical du-
plicate per batch of 20 samples. For organic parameters, the QC samples
for each analytical batch included: a method blank, analytical duplicate,
matrix spike, and matrix spike duplicate. For PAHs, PCBs, and organo-
chlorine pesticides, calibration check standards were interspersed
throughout an analytical batch in order to insure the instrument's integ-
rity. A diluted oil standard was used as a retention index solution for
PAH compounds not found in the calibration solution (e.g., alkylated
PAH homolog groups). An Aroclor mixture was used as a retention
index solution for individual PCBs not found in the calibration solution,
and individual PCB congener retention times were based on pattern rec-
ognition. Identification of PCDD/F congeners and PBDEs is described in
detail elsewhere (Crane and Hennes, 2016a). The QC samples for TOC
and particle size included analytical duplicates.
The analytical QA/QC results were determined to be acceptable for
most parameters, with a few exceptions (Crane and Hennes, 2016a).
Analytical duplicate anomalies were noted for C3-chrysene (lake #19),
C3-phenanthrenes/anthracenes (lake #19R field replicate), and C4-
phenanthrenes/anthracenes (lake #31). In each case, the analytical du-
plicates were below the reporting limits. The analytical laboratory used
one-half the reporting limit in their calculation of the relative percent
differences (RPD) between each sample and analytical duplicate, and
the RPDs were artificially elevated. The spike recoveries of BDE-85 and
BDE-154 in two samples (#5 and 28, respectively) were high, and the
results may be biased high. The spike recoveries for beta-endosulfan
were low for two samples (#7 and 28), and the results may be biased
low. For these four spike anomalies, each of these analytes was below
1323
the reporting limit for the corresponding sediment samples. Additional
QA/QC results are provided elsewhere (TDI-Brooks International, Inc.,
2009; Crane and Hennes, 2016a).
2.3.2. Fish tissue
The fish samples were processed different ways, depending on tro-
phic guild and species type (Crane and Hennes, 2016a). Predator fish
were processed as skin-on fillets (removed scales). Omnivore fish
were processed as skin-on whole fish (removed scales). Benthic fish
were processed as skin-on whole fish for bullheads and skin-on whole
fish (removed scales) for white suckers. For each lake, five fish for
each trophic guild collected were homogenized and composited togeth-
er. The percentage of lipids in each composite sample was determined.
The extraction and analysis of BDEs in fish tissue followed EPA Meth-
od 1614 (USEPA, 2007), with some modifications developed by AXYS
(Crane and Hennes, 2016a). Analysis of the extracts was performed on
a HRMS coupled to a HRGC equipped with a DB-5HT chromatography
column (30 m × 0.25 mm ID and 0.10 μm film thickness). The isotope
dilution/internal standard quantification procedure was used, and the
data were recovery corrected for possible losses during extraction and
cleanup.
The analytical QC samples for fish included three procedural blanks,
a lab-generated reference sample known as the ongoing precision and
recovery sample, a matrix spike, and matrix spike duplicate. The fish tis-
sue samples were re-run after an internal source of contamination was
found in some of the blank and fish tissue samples. The final analyses
passed the laboratories QA/QC parameters. However, the MPCA consid-
ered six of the composite sample results for BDE-99 and four of the com-
posite sample results for BDE-153 as estimated values due to the
detection and quantification of these congeners in the procedural
blanks (Crane and Hennes, 2016a, 2016b).
2.4. Watershed information
Lake, ecoregion, and watershed land use information was obtained
for the study lakes. Five size classes of lakes, as used by the NLA program
(Crane and Hennes, 2016a, 2016b), were represented in the study lakes,
including: A) 4–10 ha: four lakes; B) 10–20 ha: three lakes; C) 20–50 ha:
11 lakes (including two reference lakes); D) 50–100 ha: 13 lakes; and
E) N100 ha: 23 lakes (including two reference lakes; Table A-1). The
September 2008 ecoregion classifications used by the MDNR were uti-
lized (Crane and Hennes, 2016a). The study lakes encompassed five
ecoregions, most of which were located in the North Central Hardwood
Forest and the Northern Lakes and Forest ecoregions. Land use type in-
formation for each lake watershed was obtained from other MPCA staff
involved in the NLA project (Crane and Hennes, 2016a, 2016b). Land use
categories included developed, cultivated (i.e., agriculture), pasture and
open land, forest, and lakes and wetlands (which included the surface
area of the lake). The number of feedlots in each watershed was
noted, too. The major land use categories for the study lakes included:
developed (two lakes); cultivated (20 lakes); forested (24 lakes, includ-
ing the four reference lakes); and lakes and wetlands (eight lakes;
Table A-2).
2.5. Calculated benchmarks
Ecological benchmark values were calculated for groups of chemical
classes, as well as for a compilation of different chemicals. The calcula-
tion methods, including how censored (i.e., nondetect) data were treat-
ed, is provided elsewhere (Crane and Hennes, 2016a). Briefly, the U.S.
EPA's PAH Equilibrium Partitioning Sediment Benchmark (ESB) Toxic
Units (TU) model was used for evaluating ecological risk to aquatic in-
vertebrates. It was calculated for ∑PAH34-TU (Table A-4) using critical
concentrations of PAHs in sediment that are related to final chronic
values (USEPA, 2003; Burgess, 2009; Table A-5) and TOC data. For
PCDD/Fs, aquatic life toxic equivalents (TEQs) were calculated by
1324 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338
multiplying the 17 PCDD/F congener concentrations by their respective
World Health Organization (WHO) toxic equivalency factors (TEFs;
Table A-3) and summing the results. Because TEF values for benthic in-
vertebrates have not been developed, TEF values based on fish were
used (Van den Berg et al., 1998).
Mean probable effect concentration quotients (PEC-Qs) were calcu-
lated on a dry weight basis to distill data from a mixture of contami-
nants into one unitless index (Crane and Hennes, 2007). Briefly,
individual PEC-Qs [i.e., chemical concentration divided by the corre-
sponding PEC value (USEPA, 2000b)] were calculated for chemicals
with reliable PECs [i.e., seven metals, ∑PAH13 (Table A-4), and Total
PCBspesticide scan; MacDonald et al., 2000]. The PAH and PCB concentra-
tions were not normalized to TOC (Crane and Hennes, 2007). Next,
the mean PEC-Q for metals with reliable PECs (i.e., arsenic, cadmium,
chromium, copper, lead, nickel, and zinc) were calculated. Finally, the
mean PEC-Q values for the three main classes of chemicals with reliable
PECs were calculated. In some cases, the mean PEC-Qs were only calcu-
lated using the mean PEC-Qmetals and PEC-Q∑PAH13 data for samples
lacking Total PCBspesticide scan data.
2.6. Comparisons to sediment quality targets (SQTs)
Sediment chemistry data and benchmark values, on a dry weight
basis, were compared to analogous Level I and Level II sediment quality
target (SQT) values for benthic invertebrates (Crane et al., 2000, 2002;
Crane and Hennes, 2007). The SQTs include a suite of eight metals and
metalloids, 13 low molecular weight and high molecular weight PAHs,
∑PAH13, ∑PCBs, 10 legacy organochlorine pesticides, mean PEC-Qs
(Table A-6), and PCDD/F TEQs (Table A-7). Most of the SQTs
were adopted from consensus-based sediment quality guidelines
(MacDonald et al., 2000), for which most of the Level I SQTs correspond
to threshold effect concentrations (TECs) and most of the Level II SQTs
correspond to PEC values. The Level I SQTs are intended to identify con-
taminant concentrations below which harmful effects on benthic inver-
tebrates are unlikely to be observed, whereas the Level II SQTs are
intended to identify contaminant concentrations above which harmful
effects on benthic invertebrates are likely to be observed (Crane et al.,
2000, 2002; Crane and Hennes, 2007). The SQTs adopted from other
sources are identified in Tables A-6 and A-7.
2.7. Statistical analyses
An in-depth description of the statistical methods used to evaluate
the analytical data and calculated benchmarks is presented in Crane
and Hennes (2016a). Summary statistics were run in either Microsoft
Excel 2010 (Microsoft Corporation, Redmond, WA), SigmaPlot 13.0
(Systat Software, Inc., San Jose, CA), or the U.S. EPA's ProUCL 5.0 soft-
ware. Shapiro-Wilk normality tests, Pearson product moment correla-
tions, Spearman rank order correlations, equal variance tests, pairwise
Student's t-tests, Mann-Whitney rank-sum tests, one-way Analysis of
Variance (ANOVA), Kruskal-Wallis ANOVA on ranks, Holm-Sidak and
Dunn's method multiple pairwise comparisons, and principal compo-
nents analysis (PCA) were run in SigmaPlot 13.0. Hierarchical cluster
analysis (HCA) was conducted using PAleontological STatistics (PAST)
version 3.0 software. Box plots, quantile-quantile (Q-Q) plots, outlier
tests at the 5% significance level (i.e., Dixon's outlier test for b25 detect-
ed samples and Rosner's outlier test for ≥25 detected samples), good-
ness of fit tests, and background threshold values were determined
using ProUCL 5.0 software for both detected parameters and those
with b 80% nondetects. The Kaplan-Meier method (Helsel, 2012) was
used in ProUCL 5.0 to calculate total values of PAH compounds and
PCB congeners containing censored data with b80% nondetects
(i.e., the Kaplan-Meier mean was multiplied by the number of PAH com-
pounds or PCB congeners, respectively, to determine total values). The
Kaplan-Meier method (Helsel, 2012) was also used in ProUCL 5.0 to cal-
culate summary statistics and background threshold values for
censored data sets with b80% nondetects. For some other statistical
methods (e.g., correlation analyses), substitution at the full reporting
limit was used when there were b5% nondetects. This low percentage
was selected to reduce bias from substitution methods. The data were
not weighted to make general classifications for statewide lakes,
although the study design could allow this to be done.
PCA is a multivariate statistical technique used to reduce the com-
plexity of data sets to its most important components. PCA takes a
large number of interrelated (i.e., correlated) variables and transforms
the data to a new set of uncorrelated reference variables known as prin-
cipal components. The first principal component accounts for as much
of the variability in the data as possible, and each succeeding compo-
nent accounts for as much of the remaining variability as possible. If
most of the variability between samples can be accounted for by a
small number of principal components (e.g., two or three), then rela-
tionships between multivariate data can be assessed by a two- or
three-dimensional plot of the principal component scores. Since PCA is
sensitive to variance in data sets, all of the detected and calculated
data selected for PCA analyses underwent a range transformation
(also known as the maximum-minimum transformation; Johnson
et al., 2004). This step resulted in unitless data so that a covariance ma-
trix could be run for each PCA. The significance level for hypothesis test-
ing was set at 0.050 with a confidence level of 95%. The selection
method for components was either selected for the average eigenvalue,
minimum eigenvalue (e.g., 1.0), or number of components (e.g., three).
The Henze-Zinker normality test was used with a p value to reject of
0.050. Scree plots, component loading plots, and component score
plots were produced for each PCA analysis. In addition, separate 3D
plots of the component scores were made using SigmaPlot 13.0. Similar
groups of samples were noted on the 3D plots based on the HCA results
of the PCA component scores. The best HCA option for identifying dis-
crete groups was the Paired Group (UPGMA) algorithm with the
Gower similarity index (Crane and Hennes, 2016a).
ProUCL 5.0 was used to estimate several types of background thresh-
old values of the ambient data. In order to perform these analyses, out-
liers and/or separate populations of data from the dominant population
of raw data were first removed. While the outlier tests provided a statis-
tical basis for selecting outliers, visual observation of the Q-Q plots
(Crane and Hennes, 2016a) was also used to select breaks in the data
that might be indicative of additional outliers or a separate population
of data. The 95% one-sided upper tolerance limit (UTL) with 95% cover-
age (hereafter referred to as the UTL or UTL95-95) was the background
threshold value of greatest interest. The UTL95-95 represents the value
below which 95% of the population values are expected to fall with 95%
confidence. This approach resulted in no more than a 5% false positive
error rate (i.e., the chance of falsely classifying a true background con-
centration as a contaminated concentration). When the data were not
normally distributed and the gamma distribution was used, the
gamma UTL upper limits were estimated using the Wilson Hilferty
(WH) and Hawkins Wixley (HW) methods. For censored data with
b80% nondetects, the Kaplan-Meier versions of the UTL95-95 calcula-
tions were performed.
2.8. Environmental forensics
2.8.1. PAHs
Environmental forensic approaches for determining sources of PAHs
in the study lakes are described in detail in Crane and Hennes (2016a).
Briefly, qualitative (i.e., PAH histogram plots), semi-quantitative
(i.e., PAH source ratio plots), and quantitative (i.e., chemical mass bal-
ance model) approaches were used. Histogram plots of parent and
alkylated PAH concentrations (generated in SigmaPlot 13.0) were
used to assess patterns associated with petrogenic (i.e., oil-based)
and/or pyrogenic (i.e., combustion-based) sources of PAHs. In addition,
the following diagnostic source ratios were used to identify petrogenic
and pyrogenic sources of PAHs: fluoranthene:pyrene (F/P) and
 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338
phenanthrene:anthracene (P/A; Budzinski et al., 1997). The following
source ratios were used for determining a range of major emission
sources from automobiles: benzo[b]fluoranthene:benzo[k]fluoranthene
(BbF/BkF) vs. benzo[a]pyrene:benzo[e]pyrene (BaP/BeP; Dickhut et al.,
2000). Double-ratio plots were generated in SigmaPlot 13.0 for these
combinations of PAH source ratios.
The U.S. EPA's Chemical Mass Balance (CMB) 8.2 receptor model
(Coulter, 2004) was used to provide a quantitative assessment of impor-
tant sources of PAHs to the study lakes. This air model has been adapted
for use in the source apportionment of PAHs in sediment (Li et al., 2003;
Van Metre and Mahler, 2010; Crane, 2014; Baldwin et al., 2016). The
model uses an effective variance weighted least square solution to the
CMB equations, provided several model assumptions are met (Li et al.,
2003; Coulter, 2004). Similar methods to Crane (2014) were followed
for this data set. Briefly, the model included ≤ 12 parent PAHs
(∑PAHCMB) identified in Table A-4. A group of published PAH source
profiles listed in Crane (2014) were considered for inclusion in the
model, including the general categories of coal tar-based sealant (CT-
sealant) dust and particulate runoff, asphalt sealant pavement dust,
vehicle-related sources, coal combustion, fuel oil combustion, fireplace
combustion of wood products, and used motor oil.
The model included several source inputs. Each run of the CMB8.2
model attempted to fit the source data to the ambient sediment data
in ≤20 iterations using source elimination. The source profile uncertain-
ty was set at 40% (Li et al., 2003; Van Metre and Mahler, 2010; Crane,
2014; Baldwin et al., 2016), and the minimum source projection was
set at the default value of 0.95. The uncertainty values of the ambient
PAH data from the lake sediment samples were set at 20%. Output
files were generated as comma-separated value files that were
imported into Microsoft Excel 2010 for further analysis.
Several statistical procedures were used to prepare the measured
∑PAHCMB data and for evaluating source profile data for inclusion in
the CMB8.2 model. PAH proportional values (i.e., individual PAH con-
centrations divided by ∑PAHCMB) were calculated and assessed for
normality using the Shapiro-Wilk test. The linear dependence between
PAH proportional values of normally distributed sources and sediment
data were assessed using Pearson product moment correlations
(r) and the statistical significance (p) of the correlations. A Spearman
rank order correlation was run when the data were not normally dis-
tributed. The sediment data were evaluated several different ways to se-
lect the best sources to include in the model. Due to the widespread
geographic coverage of the samples in Minnesota, the best approach
was to run each sample independently through the model with an as-
sortment of source options that were significantly correlated to the sed-
iment data. Graphs comparing the PAH proportional values for the
sources and groups of lakes using those sources were prepared in
SigmaPlot 13.0.
Several statistical measures were used to assess the performance
of each model run. These measures included Chi square, coefficient
of determination (R 2 ) values, the percent mass estimated by the
model, and the T-statistic (Coulter, 2004). Chi square is the weighted
sum of squares of the differences between the calculated and mea-
sured fitting species concentrations; values less than one were
preferred (indicating a very good fit to the data), whereas values be-
tween one and two were also acceptable. R2 values near 1.0 indicated
that the source contribution estimates explained the measured con-
centrations very well, but other R2 values N0.8 were acceptable, too.
Percent mass represented the percent ratio of the sum of the model-
calculated source contribution estimates to the measured mass con-
centration, and ratios approaching 100% were ideal; ratios between
80 and 120% were also acceptable. The T-statistic that appeared
with the source contribution estimates provided additional informa-
tion to assess model performance because it could be indicative of
collinearity among the source profiles when it was b 2.0 (Coulter,
2004). Finally, mean RPD values between measured and modeled
∑PAHCMB concentrations were calculated.
1325
2.8.2. PCDD/Fs
The weight percentage pattern of PCDD/F congeners and its homo-
logs creates a unique “fingerprint” that can be used to identify sources
of these compounds. The weight percentages of 17 PCDD/F congeners
were calculated, and the mean and standard deviation (SD) values
were plotted in a histogram plot. The results were then compared to
published literature values for likely sources. The percent homolog pat-
terns of PCDD/Fs were calculated, and the mean (SD) values were plot-
ted in a histogram plot. The data were further broken out into different
lake groups and assessed for differences due to source type. Another
common forensic technique is to compare patterns in the ratios of
2,3,7,8-TCDD (i.e., PCD2378) concentrations with the homolog of total
TCDDs (i.e., PCD_T4; Quadrini et al., 2015). A plot of this ratio for detect-
ed compounds in individual lakes were compared to published litera-
ture values for sources. All plots were prepared in SigmaPlot 13.0.
3. Results and discussion
3.1. Lake sediment quality
3.1.1. Summary statistics
Summary statistics for SQT-compatible parameters with b80%
nondetects are provided in Table 1. For the other chemical and particle
size parameters, summary statistics are provided in Table A-8 along
with additional data interpretation in the Supplemental Information.
Data from the four a priori reference lakes were pooled with the random
lakes, because a preliminary evaluation of the PAH data showed that the
contaminant concentrations were not statistically significantly different
(p N 0.05) between the reference lakes and random lakes in the same
major land use class (i.e., forested; Crane and Hennes, 2016a). Based
on the small sample size of the a priori reference lakes (n = 4) and
random lakes (n = 50), it was also advantageous to pool these data to
increase the power of the statistical tests.
Most of the SQT-compatible metals, the metalloid arsenic, and PAHs
were detected in all sediment samples (Table 1). However, mercury was
not detected in 59% of samples due to higher reporting limits than ex-
pected, and cadmium was not detected in 11% of lake sediments
(Table 1). The highest mercury concentrations were found in lakes lo-
cated in northeast Minnesota, which was consistent with other research
(Engstrom et al., 2007). Sediment accumulation rates are lowest in
northeast Minnesota lakes where the local geology is characterized by
mostly thin, noncarbonated glacial drift and crystalline bedrock
(Engstrom et al., 2007). Thus, these lakes have lower watershed sources
of sediment to dilute mercury resulting from atmospheric inputs. Most
of the mercury found in Minnesota's lakes and rivers is due to atmo-
spheric deposition, of which 90% comes from other states and countries
(MPCA, 2013). Although the mean mercury concentrations were low
compared to other metals and metalloids (Tables 1 and A-8), this ele-
ment is of high concern because of its potential to transform to methyl-
mercury. Certain bacteria can convert mercury to methylmercury in
aquatic sediments, where it can build-up in the food chain and accumu-
late in fish tissue (Hsu-Kim et al., 2013). Twenty-four of the 54 study
lakes have fish advisories for mercury (Crane and Hennes, 2016a).
SQT-compatible PAHs were ubiquitous in all 54 lakes. The two urban
lakes from the Minneapolis-St. Paul, MN metropolitan area [i.e., Nokomis
(#27) and Snail (#38)] had the highest ∑PAH13, ∑PAH17, and
∑PAH34 concentrations of the study lakes (Fig. A-1), indicating urban
watershed sources of PAHs were important. Mean concentrations of
SQT-compatible PAHs in the 54 lakes varied by two orders of magnitude,
ranging from 4.5 μg/kg dry wt. for dibenzo[a,h]anthracene to 109.7 μg/kg
dry wt. for fluoranthene (Table 1). All PAH ESB Toxic Unit values were
below 1.0 (Table 1), indicating that benthic organisms in these lakes
are not likely to encounter adverse effects through narcosis
(i.e., resulting in the alteration of cell membrane function; Burgess,
2009).
1326 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338

Table 1 the calculation of mean PEC-Qs reduced the contribution of the metal
Summary statistics for SQT-compatible parameters with b80% nondetects in sediment. PEC-Q values.
Parameter N % Nondetects Meana SDa The detection frequency of organochlorine pesticides and metabo-
Metal or metalloid (mg/kg dry wt.)
lites varied widely in 24 study lake sediments (Tables 1 and A-8). Diel-
Arsenic 54 0 10.5 12.6 drin, which is also a common degradate of aldrin, was detected in nine
Cadmium 54 11.1 0.52 0.33 lakes (Table 1). The half-life of dieldrin in temperate soils is about five
Chromium 54 0 35 19.8 years (ATSDR, 2002). Dieldrin had the lowest mean pesticide concentra-
Copper 54 0 25.4 35.6
tion (0.1 μg/kg dry wt.) in Table 1. Total DDT (which included all DDT
Lead 54 0 34.1 30.9
Mercury 54 59.3 0.1 0.1 metabolites) and isomers of three DDT metabolites (i.e., p,p′-DDE,
Nickel 54 0 16 10.9 p,p′-DDD, and o,p′-DDD) were detected the most frequently, ranging
Zinc 54 0 61.4 29.2 from 70.8 to 95.8% detects (Table 1). These compounds also had the
PAHs (μg/kg dry wt.) highest mean pesticide concentrations (Table 1), despite the ban on
2-Methylnaphthalene 54 0 9.8 6.5 DDT in 1972 (USEPA, 1975). These persistent DDT metabolites, along
Acenaphthene 54 0 16.3 10.6 with o,p′-DDE, were also frequently detected in historical sediment
Acenaphthylene 54 0 8.7 28 cores from the Duluth-Superior Harbor, MN (Schubauer-Berigan and
Anthracene 54 0 11.7 30.5
Crane, 1997). DDMU, a breakdown product of DDE (Wetterauer et al.,
Fluorene 54 0 41.1 30.4
Naphthalene 54 0 16.4 11.3 2012), was detected in 66.7% of samples (Table A-8). The DDT isomers
Phenanthrene 54 0 70.4 103.1 (i.e., p,p′-DDT and o,p′-DDT) were detected less frequently (Table 1).
Benzo[a]anthracene 54 0 31.1 77.8 Sediment quality data on ambient DDT and DDT metabolite concentra-
Benzo[a]pyrene 54 0 34.2 96.4
tions in freshwater, surficial sediments are lacking for much of the
Chrysene 54 0 57.1 128.4
Dibenzo[a,h]anthracene 54 5.6 4.5 12.6 world (Ricking and Schwarzbauer, 2012). A recent study of DDTs in Pol-
Fluoranthene 54 0 109.7 282 ish lake sediments also found p,p′-DDE in nearly all surficial samples (n
Pyrene 54 0 78.4 193.2 = 522 samples) followed by p,p′-DDD (n = 502 samples); o,p′-DDT
∑PAH13 54 0 489.3 980.5 and o,p′-metabolites were not evaluated (Bojakowska et al., 2014).
PAH ESB Toxic Units 54 0 0.017 0.019 DDT and its metabolite isomer ratios can be used to assess the envi-
ronmental fate of these compounds (Ricking and Schwarzbauer, 2012).
Total PCBs (μg/kg dry wt.)
Total PCBspesticide scan 24 8.3 16.7 19.6
Five different ratios were calculated, as described further in the Supple-
∑PCBcongeners 24 50 18.8 19.7 mental Information and Table A-9. Interestingly, the presence or ab-
sence of feedlots in the lake watershed was associated with a
PEC-Qs
Metal PEC-Q 54 0 0.23 0.13
significant difference (p b 0.05) in one particular ratio. The median
Mean PEC-Q 54 0 0.11 0.062 value of o,p′-DDX (i.e., o,p′-DDT + o,p′-DDE + o,p′-DDD) as a percent-
age of Total DDT was over three times higher in lakes with feedlots than
Pesticide or metabolite (μg/kg dry wt.)
alpha-Chlordane 24 70.8 0.36 0.63
those lacking feedlots. Additional work would be needed to determine if
gamma-Chlordane 24 62.5 0.38 0.91 historical or current feedlot operations are a potential source of residual
Dieldrin 24 62.5 0.1 0.083 DDT metabolites (particularly o,p′-DDD) to nearby lakes.
o,p′-DDD 24 29.2 2.2 5.8
p,p′-DDD 24 16.7 8 20.6
3.1.2. Comparisons to guidelines
o,p′-DDE 24 54.2 0.36 0.84
p,p′-DDE 24 4.2 12.5 24.8 The sediment chemistry data were compared to ranges of Level I and
p,p′-DDT 24 62.5 0.38 0.47 Level II SQT values (Crane et al., 2000, 2002; Crane and Hennes, 2007).
Total DDT 24 4.2 25 53.7 The percentage of most parameters (e.g., cadmium, mercury, and
cis-Nonachlor 24 29.2 0.24 0.3
benzo[a]pyrene) were below the analogous Level I SQT values
trans-Nonachlor 24 45.8 0.31 0.42
(Table 2). Based on the mean PEC-Q values, nearly 43% of lakes were
PCDD/F TEQs (ng TEQ/kg dry wt.) 23 0 3.8 4.1 below the Level I SQT value of 0.1. This meant that conditions were pro-
SQT = sediment quality target; N = number of samples; SD = standard deviation; PAHs tective of sediment-dwelling organisms for these chemicals, and sedi-
= polycyclic aromatic hydrocarbons; PCBs = polychlorinated biphenyls; PEC-Q = prob- ment quality was good. A short list of chemicals were more highly
able effect concentration quotient; DDD = dichlorodiphenyldichloroethane; DDE = elevated between the Level I and Level II SQT values than other SQT
dichlorodiphenyldichloroethylene; DDT = dichlorodiphenyltrichloroethane; TEQ =
ranges, including: acenaphthene, mean PEC-Qs, Total DDT, and PCDD/
toxic equivalent.
a
The Kaplan-Meier method was used to calculate mean and SD values for censored F TEQs (Table 2). In particular, approximately 57% of lakes had mean
data sets that had b80% nondetects. PEC-Qs within the range of N0.1 to ≤0.6, which indicated moderate sed-
iment quality. Previous work in the St. Louis River AOC, MN showed that
sediment toxicity increased as the mean PEC-Q ranges increased from
Total PCBs, determined two different ways, had similar mean (SD) ≤0.1, N0.1 to ≤0.5, and three higher ranges (Crane et al., 2002). Thus,
concentrations (Table 1). Total PCBspesticide scan were detected in 22 of lake sediment samples between the Level I and II SQT values for mean
24 lakes, and ∑PCBcongeners were estimated in 12 of 24 lakes from PEC-Qs (Table 2) could show a higher incidence of sediment toxicity
PCB congener data (Table 1). The highest concentrations were observed compared to those below the Level I SQT value. Other physical-
in Nokomis (#27) and Snail (#39). The estimated ∑PCBcongener values chemical factors in sediments, such as particle size, TOC, acid volatile
were highly correlated (r = 0.998) to Total PCBspesticide scan values sulfide, and ammonia, can also influence the health of benthic inverte-
(Fig. A-2). Further work is needed to determine if the pesticide scan brates (Crane et al., 2000).
Total PCBs could provide a cost-effective screen for congener-specific The Level II SQT values were exceeded in a small number of sedi-
∑PCBs. The coplanar PCBs were also included in a different version of ment samples for four metals, arsenic, acenaphthylene, sum DDD, and
the PCDD/F TEQs, but they contributed little to this calculation (Crane sum DDE (Table 2). Harmful effects on sediment-dwelling organisms
and Hennes, 2016a). were likely in these samples. The two developed lakes (i.e., lakes #27
As an integrative indicator, the mean PEC-Qs demonstrated that and 38) exceeded the Level II SQT value for lead, which was likely due
metals were the most important component. When calculated with to urban contamination from stormwater runoff and other sources.
metals alone, the metal PEC-Q was twice as high as the mean PEC-Q The Level II SQT exceedances for chromium (lake #3), copper (lake
(Table 1). The inclusion of ∑PAH13 and Total PCBspesticide scan data in #6), and nickel (lakes #5 and 6) occurred in lakes from northeast
 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338 1327

Table 2 organisms showed that sodium arsenite increased in use in Minnesota

Percentage of parameter results within designated SQT ranges. in 1954, and it was used to eradicate aquatic vegetation through 1969
Parameter Percent of Samples (Crane and Hennes, 2016a). Arsenic is likely to remain high in Snail
≤Level I NLevel I SQT NLevel II
Lake for some time. For a lake in Michigan that was treated with sodium
SQT to ≤ Level SQT arsenite in 1957, Siami et al. (1987) developed a model that predicted it
II SQT would take N 100 years for arsenic levels to reach pretreatment concen-
Metal or metalloid (n = 54) trations in the surficial sediments. The source of the high arsenic con-
Arsenic 75.9 20.4 3.7 centration of 67.5 mg/kg dry wt. in Elk Lake (#51) is less certain. Elk
Cadmium 88.9 11.1 0 Lake is a “super-sentinel” deep-water lake located in Itasca State Park
Chromium 70.4 27.8 1.8 (O'Hara et al., 2011), and it is included in a group of 24 lakes being stud-
Copper 87.0 11.1 1.8
Lead 64.8 31.5 3.7
ied by the MDNR to understand the effects of environmental stressors
Mercury 81.5 18.5 0 such as climate change (http://www.dnr.state.mn.us/fisheries/slice/
Nickel 88.9 7.4 3.7 sentinel.html; accessed 3/30/2017). Further sampling is warranted to
Zinc 94.4 5.6 0 evaluate if other areas of Elk Lake have elevated arsenic concentrations
PAHs (n = 54) in the sediment.
2-Methylnaphthalene 96.3 3.7 0
Acenaphthene 16.7 83.3 0 3.1.3. Multivariate statistical analyses
Acenaphthylene 74.1 24.1 1.8
PCA analysis of the 54 lakes was performed to reduce a set of chem-
Anthracene 98.1 1.9 0
Fluorene 90.7 9.3 0 ical variables to the ones that accounted for most of the variance. In turn,
Naphthalene 100 0 0 similiarites and differences among lakes were evaluated using HCA
Phenanthrene 96.3 3.7 0 analysis. The PCA analysis was done for a subset of chemicals that also
Benzo[a]anthracene 94.4 5.6 0 have SQT values, including: five metals (i.e., chromium, copper, lead,
Benzo[a]pyrene 94.4 5.6 0
Chrysene 94.4 5.6 0
nickel, and zinc), the metalloid arsenic (hereafter lumped with the
Dibenzo[a,h]anthracene 98.1 1.9 0 other metals), ∑PAH13, and mean PEC-Qs (Fig. 2). The scree plot, com-
Fluoranthene 96.3 3.7 0 ponent loading plots, and component score plots are provided in the
Pyrene 96.3 3.7 0 Supplemental Information (Figs. A-3 and A-4). Three principal compo-
∑PAH13 96.3 3.7 0
nents accounted for 80.1% of the variance in the data. Each principal
Total PCBs (n = 24) component (PC) is a linear combination of the original variables, after
Total PCBspesticide scan 91.7 8.3 0 each original variable has been centered about its mean. The Eigenvec-
∑PCBcongeners 91.7 8.3 0 tors in PC1 were highest for zinc, mean PEC-Qs, and lead, while nickel
Mean PEC-Q (n = 54) 42.6 57.4 0 and arsenic contributed to PC2. Arsenic, copper, and zinc contributed
the greatest Eigenvectors to PC3. Chromium and ∑PAH13 were reduced
Pesticide or metabolite (n = 24)
from the first three PCs. The unexplained variance ranged from 5.3% for
Chlordanea 91.7 8.3 0
Dieldrin 100 0 0 mean PEC-Qs to 48.1% for chromium. Four significant outliers (p b 0.05)
Sum DDD 75 16.7 8.3 were noted from the component scores, including: lakes #6 (Becoosin),
Sum DDE 50 37.5 12.5
Sum DDT 100 0 0
Total DDT (measured)b 41.7 58.3 0
Endrin 100 0 0
Heptachlor Epoxide 100 0 0
gamma-Hexachloro-cyclohexane (Lindane) 100 0 0
Toxaphenec 100 0

PCDD/F TEQs (n = 23) 13 87 0

SQT = sediment quality target; PAHs = polycyclic aromatic hydrocarbons; PCBs =

polychlorinated biphenyls; PEC-Q = probable effect concentration quotient; DDD =
dichlorodiphenyldichloroethane; DDE = dichlorodiphenyldichloroethylene; DDT = di-
chlorodiphenyltrichloroethane; PCDD/Fs = polychlorinated dibenzo-p-dioxins/dibenzo-
furans; TEQs = toxic equivalents.
a
Chlordane = cis-Nonachlor + trans-Nonachlor + alpha-Chlordane + gamma-
Chlordane.
b
When Total DDT was calculated as Sum DDD + Sum DDE + Sum DDT, it yielded 45.8%
of samples ≤ Level I SQT and 54.2% of samples in the category of NLevel I to ≤ Level II SQT.
c
The reporting limits for toxaphene exceeded the Level I SQT value of 0.1 μg/kg dry wt.,
but were less than the Level II SQT value of 32 μg/kg dry wt. Values can only be expressed
as bLevel II SQT.

Minnesota where these metals are naturally high in geological minerals.

Lake #6 (Becoosin) is located in the protected BWCAW, and would only
receive air-derived anthropogenic inputs of contaminants besides natu-
rally derived runoff from the local watershed. For acenaphthylene, the
Level II SQT value was exceeded in lake #27. The developed lakes
(#27 and 38) exceeded the Level II SQT values for Sum DDD and Sum
DDE, as did lake #32 (Pebble) for Sum DDE.
Arsenic exceeded the Level II SQT value in lakes #38 (Snail) and 51
(Elk). The high arsenic concentration of 73.2 mg/kg dry wt. in Snail
Lake (#38) was due to past treatments of arsenic compounds to control Fig. 2. PCA plot of three principal components for six metals, mean PEC-Q values, and
aquatic macrophytes and algae (Crane and Hennes, 2016a). A review of ∑PAH13 that have corresponding SQT values. The significant outliers (p b 0.05)
historical lake permits to destroy or control aquatic vegetation and identified in the PCA analysis included lakes #6, 27, 38, and 51.
1328 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338

Table 3
Multiple pairwise comparisons by land use categories for selected parameters. Pairs noted with gray shading are significantly different (p b 0.05) from each other.

p < 0.05 (gray shading)

Parameter Dev vs. cul Dev vs. wet Dev vs. for For vs. cul For vs. wet Wet vs. cul
Metal or metalloid
Arsenica,b
Bariuma,b
Chromiuma,b
Coppera,b
Leadb,c
Zincb
PAHs
Acenaphtheneb,c
Acenaphthylened DNT DNT DNT
Benzo[a]anthracened DNT DNT DNT
Dibenzo[a,h]anthracenea,b
Benzo[a]pyrened DNT DNT DNT
Benzo[b]fluoranthenea,b
Benzo[g,h,i]perylenea,b
Benzo[k]fluoranthened DNT DNT DNT
C1-chrysenesa,b
C1-dibenzothiophenesb,c
C1-fluorenesb,c
C1-fluoranthenes/pyrenesd DNT DNT DNT
C1-naphthalenesb
C1-phenanthrenes/anthracenesd DNT DNT DNT
C2-fluorenesb,c
C2-phenanthrenes/anthracenesd DNT DNT
C3-naphthalenesb
Chrysened DNT DNT DNT
Dibenzothiophened DNT DNT DNT
Fluoranthened DNT DNT DNT
Indeno[1,2,3-cd]pyrenea,b
1-Methylnaphthaleneb
2-Methylnaphthaleneb
1-Methylphenanthrenea,b
1,6,7-Trimethylnaphthalenea,b
Naphthaleneb,c
Phenanthreneb,c
Pyrened DNT DNT DNT
∑PAH13d DNT DNT DNT
∑PAH17d DNT DNT DNT
∑PAH34d DNT DNT DNT
Biphenylb,c
PEC-Qs
Metal PEC-Qa,b
Mean PEC-Qa,b
TOCb
Dev = developed; cul = cultivated; for = forested; wet = lakes and wetlands; PAHs =polycyclic aromatic hydrocarbons; DNT = do not test; PEC-Qs = probable effect concentration
quotients; TOC = total organic carbon.
a
Based on natural log transformation of data.
b
The Holm-Sidak method was used for these pairwise multiple comparisons.
c
Based on square root transformation of data.
d
Dunn's method was used for these pairwise multiple comparisons.

27 (Nokomis), 38 (Snail), and 51 (Elk; Fig. 2). This result also matched
the dissimilarity of these lakes from each other and the other lakes as
determined by HCA analysis (Figs. 2 and A-5). All four of these outlier
lakes exceeded the Level II SQT values for metals that were important
in one or more PCs. In addition, the northeast Minnesota lakes of #3 (Ar-
thur), 4 (Aspen), and 5 (August) clustered together (Fig. 2) These lakes
tended to have higher nickel and elevated chromium concentrations
compared to the other lakes. All other lakes, encompassing cultivated,
forested, and lakes and wetlands land uses, were clustered together in
one group (Fig. 2).
3.1.4. Major land use comparisons
Statistical comparisons were run on highly detected sediment data
(i.e., ≤5% nondetects) collected from all the study lakes to test the null
 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338
hypothesis of no significant differences between major watershed land
use groups. Either one-way ANOVAs or Kruskal-Wallis one-way ANOVAs
on ranks were run on either raw or transformed data (Table A-10).
Fifteen parameters (e.g., aluminum, manganese, and vanadium) were
similar (p N 0.05) across land use groups (Table A-10).
Parameters with statistically significant results (p b 0.05) underwent
multiple pairwise comparisons by major land use categories (Table 3).
The mean or median values of thirty-four of 41 parameters (Table 3)
showed statistically significant differences (p b 0.05) between lakes
with developed watersheds (n = 2) versus lakes with cultivated water-
sheds (n = 20 lakes). In all cases, higher concentrations were observed
in the developed lakes for arsenic, lead, zinc, 25 PAHs, ∑PAH13,
∑PAH17, ∑PAH34, biphenyl, metal PEC-Qs, and mean PEC-Qs. Lead,
zinc, and PAHs are common urban contaminants (Callender and Rice,
2000; Van Metre and Mahler, 2005; Jartun et al., 2008). The significant
result for arsenic was due to past treatment of an urban lake
(i.e., Snail—#38) with arsenic compounds to reduce aquatic macro-
phytes (Crane and Hennes, 2016a). Biphenyl occurs naturally in coal
tar, crude oil, and natural gas, and it is released to the environment by
the incomplete combustion of mineral oil and coal in addition to vehicle
emissions (World Health Organization, 1999). Thus, biphenyl has sever-
al common urban sources. The elevated urban PEC-Q values are reflec-
tive of the metals and/or ∑PAH13 values used to calculate them.
For lakes in developed watersheds versus those in watersheds dom-
inated by lakes and wetlands, as well as forests, there were 19 and 18
parameters, respectively, with statistically significant (p b 0.05) results
(Table 3). In all cases, mean or median concentrations were highest in
lakes surrounded by developed land uses. Some other PAH compounds
(e.g., acenaphthylene) had “Do Not Test” results for these comparisons
when the nonparametric Dunn's method was used for the pairwise
comparisons (Table 3). This result occurs for a comparison when no sig-
nificant difference (p N 0.05) is found between the two rank sums that
enclose that comparison. A result of “Do Not Test” should be treated as
if there is no significant difference between the rank sums, even though
one may appear to exist.
The other multiple pairwise comparisons had a lower number of sta-
tistically significant (p b 0.05) results. For lakes in forested versus culti-
vated watersheds, 13 parameters had statistically significant (p b 0.05)
differences (Table 3). These parameters included barium, lead, zinc,
seven PAH compounds, metal PEC-Q, mean PEC-Q, and TOC (Table 3).
In all cases, except barium, the mean or median concentrations were
Table 4
Multiple pairwise comparisons of selected parameters by lake surface area categories. Pairs shaded gray are significantly different (p b 0.05) from each other.
Parameter A vs. E
Metal
Manganesea
PAHs
C1-naphthalenesb,c
C2-naphthalenesc,d
1-Methylnaphthaleneb,c
2-Methylnaphthaleneb,c
1,6,7-Trimethylnaphthalenec,d
Naphthaleneb,c
Biphenylc,d
TOCc
DNT = do not test; PAHs = polycyclic aromatic hydrocarbons; TOC = total organic carbon.
Surface Area Categories: A = 4-10 ha; C = 20-50 ha; D = 50-100 ha; E = >100 ha.
a
Dunn's method was used for this pairwise multiple comparison.
b
Based on square root transformation of data.
c
The Holm-Sidak method was used for these pairwise multiple comparisons.
d
Based on natural log transformation of data.
1329
higher in lakes with forested watersheds. Only a few statistically signif-
icant (p b 0.05) results were noted for lakes in cultivated versus lakes
and wetlands watersheds (Table 3).
For other parameters measured in a subset of lakes (n ≤ 24), statisti-
cal comparisons were run between the developed lakes (#27 and 38)
and the remaining samples if there were ≤ 5% nondetects. The results
are provided in the Supplemental Information. Briefly, statistically sig-
nificant (p b 0.05) differences were observed for Total DDT, p,p′-DDE,
∑PCBcongeners, Total PCBspesticide scan, ∑PCDD/Fs, and most PCDD/F
congeners and homologs (except PCD_T4; Table A-11). For these pa-
rameters, higher mean or median values were observed in the devel-
oped lakes. These results, in addition to the majority of the results for
parameters measured in all of the study lakes, demonstrate how
urban environments have impacted sediment quality in developed
lakes compared to lakes surrounded by other major watershed land
uses.
3.1.5. Lake surface area comparisons
For sediment quality parameters measured in each study lake, statis-
tical comparisons were run between different surface area categories.
The analyses were limited to parameters with ≤ 5% nondetects. The
null hypothesis was no significant differences between the five surface
area categories. Either one-way ANOVAs or Kruskal-Wallis one-way
ANOVAs on ranks were run on either raw or transformed data
(Table A-12). The only parameters with statistically significant (p b
0.05) differences included: TOC, manganese, six low molecular weight
PAHs, and biphenyl (Table A-12). Multiple pairwise comparisons were
then run.
A limited number of statistical differences were observed between
surface area groups. For manganese, surface area categories E and A
were not statistically different (p N 0.05) from each other, while the
other nine pairwise comparisons were given a “Do Not Test” classifica-
tion (Table 4). The surface area categories D versus E were significantly
different (p b 0.05) for five of the six naphthalene and alkylated naph-
thalene compounds, as well as biphenyl (Table 4). In all cases, mean
values were higher in lakes included in surface area category D
(i.e., 50–100 ha). This category included the two developed lakes (#27
and 38), for which one or both of them were statistical outliers at the
5% significance level for these six compounds. For TOC, two pairwise
comparisons were statistically significant (p b 0.05; Table 4). Mean
TOC values were significantly greater (p b 0.05) in class A versus class
p < 0.05 (gray shading)
Seven other pairwise
C vs. E D vs. E comparisons
DNT DNT DNT
1330 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338

E lakes, as well as in class C versus class E lakes. In general, assessing the 12 PAH compounds included in ∑PAHCMB (Crane and Hennes,
lakes by surface area class was less useful than by major land use 2016a). For the final model runs, no samples were excluded from the
categories in the lake watersheds. CMB8.2 model, residual uncertainty was minimized, the T-statistic
that appeared with the source contribution estimates was N 2.0 for the
3.1.6. Environmental forensics major source(s), and the model performance [mean (SD)] was accept-
able for R2 [0.957 (0.027)], Chi Square [0.46 (0.22)], and percent mass
3.1.6.1. PAHs. The use of several types of environmental forensic proce- [96.8% (5.5%); Fig. A-7]. The mean RPD between measured and calculat-
dures demonstrated the importance of pyrogenic sources of PAHs to ed ∑PAHCMB concentrations (Fig. A-7) was 4.7% with a SD of 4.8%; the
the study lakes. Visual observations of PAH histogram plots indicated median RPD was 2.7%.
pyrogenic sources of PAHs were dominant due to higher concentrations The source apportionment of PAHs in each study lake was effectively
of most parent PAHs than their alkylated homologs (Crane and Hennes, accomplished with the CMB8.2 model. Twelve source profiles (Table A-
2016a, 2016b). In addition, high concentrations of perylene in some 13) were considered for the entire lake data set, for which each lake typ-
lakes comprised a major proportion of ∑PAH34 (Fig. A-6), which is in- ically used two or three sources (Crane and Hennes, 2016a, 2016b). Two
dicative of a diagenic source for this PAH compound. In many sediment lakes used one source, four lakes used four sources, and one lake used
studies, perylene does not associate well with pyrogenic PAHs resulting five sources. Lakes that had a common source were grouped together
from atmospheric deposition (Slater et al., 2013). Perylene is produced so that the mean and SD of PAH proportional values for the 12 PAHs
naturally in anoxic sediments, but its precursor has not been identified. used in the calculation of ∑PAHCMB could be compared to the analo-
Some researchers think that perylene originates mainly from in situ gous PAH proportional values of the common source profile. Those
biogenic diagenesis under anoxic conditions and that this process plots with the strongest correlations are provided in Fig. 4, while the re-
is microbially mediated (Silliman et al., 2000; Han et al., 2015). maining plots are given in Fig. A-8. In particular, the mean PAH propor-
Christensen et al. (2007) observed that watersheds dominated by peat tional values for lakes were highly correlated to both the CT-sealant
and bogs can lead to high levels of perylene in lake sediments. North dust: Austin, TX #2 source (Mahler et al., 2010) and the coal emissions
central and northeast Minnesota have the greatest concentration of average source (Li et al., 2003; Fig. 4). For vehicle emission sources, the
peatland in the U.S. (N6 million acres) after Alaska (Wright et al., PAH proportional values for diesel vehicle particulate emissions (Li
1992). Perylene was high in other parts of Minnesota, so there are likely et al., 2003) were similar to the higher molecular weight PAH propor-
other organic material sources important for its formation in sediments. tional values from lakes using this source, while the traffic tunnel air
PAH source ratios were used for semi-quantitative analysis of the source (Li et al., 2003) more closely aligned with lower molecular
data. A double-ratio plot of P/A versus F/P (Fig. 3a) indicated that the weight PAHs from lakes using this source (Fig. 4). For the pine wood
samples were dominated by pyrogenic sources since the F/P ratios soot particles #1 source (Schauer et al., 2001), it was similar to the
were N 1 and the P/A ratios were b10 for most samples (Budzinski PAH proportional values of lakes using this source, except for fluoran-
et al., 1997). For P/A ratios between 10 and 15 in Fig. 3a, these values thene and pyrene. The CMB8.2 model considered the best mix of source
are considered indeterminate relative to source (Gao et al., 2007). A profiles to approximate the ambient sediment data so it was not neces-
double-ratio plot was also generated for BbF/BkF versus BaP/BeP sary for all source PAH compounds to be in sync with the ambient data.
(Fig. 3b), and it was compared to a range of major emission sources The source apportionment results were further generalized into
from automobiles (Dickhut et al., 2000). The two urban lakes (#27 overarching source categories of CT-sealant pavement dust, vehicle
and 38) were not within the range of vehicle emissions being a major emissions, coal-related combustion, and wood combustion. Most
source of PAHs. Four lakes completely intersected the major vehicle study lakes had two to three results in these broad source categories.
emission ranges of both source ratios: #1 (Allen), #37 (Richey), #39 Vehicle emissions and coal-related combustion were tied as the most
(South), and #40 (South Drywood). These lakes represented a wide prevalent sources of PAHs to sediments in 41 lakes. CT-sealant pave-
geographic range of Minnesota, which would indicate atmospheric ment dust was a source of PAHs to sediment in 32 lakes. Van Metre
transport of vehicle emissions provide a widespread source of PAHs to et al. (2012) estimated that PAH emissions from new CT-sealant appli-
Minnesota lake sediments. Since double-ratio plots do not provide a cations each year (~1000 Mg) are larger than annual vehicle emissions
complete understanding of all sources of PAHs, quantitative modeling of PAHs for the U.S. Thus, there could be long-range transport of CT-
was used to provide a rigorous assessment of PAH sources. sealant dust particles that could be deposited via wet and dry deposition
Two-hundred and seventy (270) runs of the CMB8.2 model were to Minnesota lakes outside of urban areas. Wood combustion was a
made using different combinations of source profiles (Table A-13) for source of PAHs in 12 lakes. Most of the study lakes, outside of the

Range of
major emission
sources from
automobiles
16 4.0
Benzo[b]fluoranthene:Benzo[k]fluoranthene

Petrogenic
Phenanthrene:Anthracene (P/A)

14 F/P <1 3.5

P/A >10
12 3.0

10 2.5

8 2.0

6 1.5 Range of major

emission sources
4 Pyrogenic 1.0 from automobiles
F/P >1
P/A <10
2 0.5
0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6

a) Fluoranthene:Pyrene (F/P) b) Benzo[a]pyrene:Benzo[e]pyrene

Fig. 3. a) Double-ratio plot of phenanthrene:anthracene (P/A) vs. fluoranthene:pyrene (F/P) to distinguish between petrogenic and pyrogenic sources, and b) double-ratio plot for PAH
compounds indicative of major emission sources from vehicles.
 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338 1331

Fig. 4. Comparison of PAH proportional values (i.e., individual PAH concentration normalized to ∑PAHCMB concentration) between sources used in the EPA's CMB8.2 model (open circles)
and the mean profile for NLA lakes (closed circles; uncertainty bars indicate one SD) for which that source was used in the model results. PAHs range from low molecular weight to high
molecular weight compounds along the x-axis. Either Pearson's r or Spearman's rho values, associated p values, and number of lakes in each group are given in parentheses; p values b0.05
are significant.

urban core of the Minneapolis-St. Paul, MN metropolitan area, had low
concentrations of ∑PAH13 (i.e., below the Level I SQT value). While
PAH concentrations in most of these lake sediments were of low con-
cern, it was helpful to go through this modeling exercise to create a
baseline source apportionment that could be used for evaluating the re-
sults of federal and state actions in the future (e.g., clean air regulations,
improvements in vehicle performance). Sediments provide a historical
record of contaminant fluxes and radioisotope dating of cores, coupled
with source apportionment modeling, could provide stronger evidence
of changes because of regulations or management actions.
The source apportionment of PAHs, by general source categories,
was also displayed visually for each lake. CT-sealant dust and coal-
related combustion were the dominant sources of PAHs in surficial sed-
iments of the three lakes with the highest ∑PAHCMB concentrations
(i.e., #27, 38, and 32; Fig. 5). In comparison, CT-sealant dust and
particulate wash off were dominant sources of PAHs to stormwater
pond sediments in the Minneapolis-St. Paul, MN metropolitan area, ac-
counting for 67.1% of modeled ∑PAHCMB concentrations (Crane, 2014).
Atmospheric deposition was probably the main transport pathway of
PAHs to the four BWCAW lakes (i.e., #2, 6, 17, 47) where motorized ve-
hicles and boats are prohibited from most of the park. Vehicle emissions
comprised the major source to each BWCAW lake (Figs. 5 and A-9),
demonstrating long-range transport of emission source particles. In ad-
dition, coal-related combustion also contributed small amounts of PAHs
to Alruss (#2) and Becoosin (#6) lakes (Figs. 5 and A-9).
Although the CMB8.2 model provided a rigorous source apportion-
ment of ∑PAHCMB in lake sediments throughout Minnesota, this
model has several important limitations. These limitations include:
1) the model results are sensitive to uncertainty in the input data,
which is often not well defined; 2) the source profiles are limited by a
1332 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338
Fig. 5. General source apportionment of PAHs in each lake.
common number of PAHs (i.e., 12 parent PAHs), and a broader suite of
PAHs would allow more detailed and unique source profiles to be devel-
oped; 3) source profiles were obtained from several literature sources,
which may have varying data quality objectives and adherence to QA/
QC protocols; 4) no source profiles were available for ethanol plants
(21 in Minnesota; http://www.eia.gov/state/analysis.cfm?sid=MN,
accessed 3/31/2017) or for the combustion of biofuels, which may be
more important in the future as these fuels increase in use in Minnesota;
and 5) source results can be misinterpreted if important sources are ex-
cluded from the model (Li et al., 2003; Van Metre and Mahler, 2010;
Crane, 2014). Another limitation with interpreting the results is that
the surficial sediment core samples were not dated, and the sample
lakes are likely to have different sedimentation rates. Thus, the historical
record of PAH accumulation in the top 15 cm of each lake is likely to
vary. A shorter depth interval (e.g., 0–2 cm or 0–5 cm) would provide
a better assessment of more recent sources of PAHs.
3.1.6.2. PCDD/Fs. Multiple, semi-quantitative environmental forensic
methods were used to determine broad source categories of PCDD/Fs.
On a weight percentage basis, OCDD comprised 72.3 ± 3.5% of
∑PCDD/F congeners followed by PCD1234678 at 12.9 ± 1.5%
(Fig. 6a). These two highly chlorinated congeners are indicative of ubiq-
uitous combustion sources, such as wood fires and vehicle exhaust
(Cleverly et al., 1997; Barabas et al., 2004; Shields et al., 2006; USEPA,
2006; Towey et al., 2012). These two congeners also dominated regional
background sources in soils collected near the Tittabawassee River, MI,
for which sections of the river downstream of Midland, MI were con-
taminated with PCDD/Fs from the Dow Chemical Company (Towey
et al., 2010). OCDD comprised 60–80% of ∑PCDD/F concentrations in
flood-plain soils and sediment of the Tittabawassee River upstream of
Midland, MI, and 30–60% in downstream samples (Hilscherova et al.,
2003). Thus, the upstream percentages of OCDD in the Tittabawassee
River flood-plain soils and sediments were consistent with the percent-
ages of OCDD observed in the study lake sediments. OCDF comprised
the third highest weight percentage of PCDD/Fs in this subset of study
lakes at 5.5 ± 2.5% (Fig. 6a). OCDF is a common nonpoint source conge-
ner (Shields et al., 2006; USEPA, 2006; Towey et al., 2012). Because of
these highly chlorinated ubiquitous congeners, the fractions of less
chlorinated PCDD/F congeners are often more useful for distinguishing
unique sources in contaminant site investigations (Towey et al., 2010;
Quadrini et al., 2015). The other PCDD/F congeners each comprised
b3% of the weight percentage of ∑PCDD/F congeners. Since only one
sediment sample was collected from each lake over a wide geographic
area in Minnesota, forensic use of less chlorinated PCDD/F congeners
is more limited.
Another common forensic technique is to compare patterns in ratios
of PCD2378 concentrations to PCD_T4. The ratio of detected PCD2378 to
PCD_T4 for individual lakes is provided in Fig. 6b. The ratios ranged from
0.010 in lake #11 [Eagle (North)] to 0.21 in lake #37 (Richey). Ratios on
the order of 0.06 or less may be indicative of wastewater and
atmospheric sources (Chaky, 2003). Quadrini et al. (2015) noted high
ratios between 0.6 and 1.0 in lower Passaic River, NJ sediments
where 2,3,7,8-TCDD was a by-product of the 2,4,5-trichlorophenol
manufacturing process. The ratios in the study lakes were far below
those observed in the lower Passaic River. PCDD/Fs emitted from resi-
dential burn barrels and open burning of yard waste is the number
one source of these compounds in the U.S. EPA's national emissions in-
ventory (USEPA, 2006). As such, the PCDD/F “fingerprint” can vary de-
pending on the type of waste being burned (e.g., plastics and paper
treated with chemicals, coatings, and inks). Although most burning of
household garbage is illegal in Minnesota, 45% of respondents in a Min-
nesota survey burned household wastes (https://www.pca.state.mn.us/
living-green/dont-burn-your-garbage, accessed 3/31/2017).
The percent homolog patterns of PCDD/Fs provided another forensic
technique for three groups of lakes: lake #36, the two developed lakes,
and the remaining 20 lakes. Compared to the other study lakes, lake #36
(Red Rock) had very low PCDD/F homolog concentrations and a unique
particle size distribution (i.e., very sandy). This lake had a higher
percentage of lower and medium chlorinated PCDD homologs
(i.e., PCD_T4, PCD_T5, and PCD_T6), and a lower percentage of highly
chlorinated PCDD homologs (i.e., PCD_T7 and OCDD) than the other
lakes (Fig. 6c). PCF_T6 was the only PCDF homolog detected in Red
Rock, for which the homolog weight percentage was lower than in the
other lakes (Fig. 6c). A consistent source profile was observed for the
 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338 1333

80 developed lakes, while the mean weight percentage of PCF_T7 was sig-
PCDD/F Congener Weight Percentage
nificantly higher (p b 0.05) in the developed lakes than the other 20
70 lakes (Fig. 6c). These significant differences may be attributed to differ-
15
ent sources. For example, residential trash burning (which could con-
tribute PCD_T6) should be less in the Minneapolis-St. Paul, MN
metropolitan area where residents have access to commercial trash
10
haulers compared to outstate Minnesota where these services may be
lacking. The Minneapolis-St. Paul, MN metropolitan area also has
more vehicle and commercial/industrial emissions than most outstate
5
areas, which could contribute to higher weight percentages of PCF_T7.

3.1.7. Background threshold values

0
The UTL95-95 values were determined for SQT-compatible
OCDD
PCD1234678
OCDF
PCF1234678
PCD123789
PCD123678
PCF2378
PCF123478
PCF234678
PCF123678
PCF23478
PCF12378
PCD123478
PCD12378
PCF1234789
PCD2378
PCF123789
chemicals (Tables 5 and 6), as well as other parameters (Tables A-14
and A-15), to represent ambient sediment quality conditions in this
a) group of Minnesota lakes. For organic parameters, one or both devel-
oped lakes (i.e., #27 and 38) were the most common outliers removed
0.25
from the UTL95-95 calculations. Potential outliers, or separate distribu-
Ratio of PCD2378 to PCD_T4

tions of data, were more variable for the metals and metalloids. Environ-
0.20 mental lake sediment data above the UTL95-95 values would be more
unusual due to natural geologic and/or anthropogenic (e.g., point
0.15 source) influences on the surrounding watershed or in-lake processes.
Since certain metals (e.g., copper, nickel) are naturally elevated in por-
tions of northeast Minnesota, it would be useful to collect more ran-
0.10
domly determined surficial sediment data from this area in order to
calculate regional UTL95-95 values. In addition, the UTL95-95 values
0.05 can be expanded as more randomly collected surficial sediment data
are collected throughout the state.
0.00 The SQT-compatible UTL95-95 values were compared to their corre-
1 7 11 14 15 20 27 32 37 38 39 50 52 53 sponding Level I and Level II SQT values. Several UTL95-95 values
b) Lake # exceeded the corresponding Level I SQT values, but were less than the
corresponding Level II SQT values. This list included: arsenic, cadmium,
chromium, lead, mercury, nickel, zinc, acenaphthene, acenaphthylene,
60
mean PEC-Qs, sum chlordane isomers, Total DDT, sum DDD isomers,
Homolog Weight Percentage

PCDD Homologs PCDF Homologs

50
sum DDE isomers, and PCDD/F TEQs. None of the UTL95-95 values
exceeded the corresponding Level II SQT values. The MPCA is lacking
40 SQT values for several pesticides (e.g., heptachlor, mirex) so the UTL
#36-Red Rock
Developed Mean values are a useful benchmark to assess exceedances of ambient back-
20 Lake Mean
30 ground concentrations. With time, the UTL95–95 values will become
more robust as additional, randomly collected surficial sediment data
20 are used to calculate them. In addition, the SQT and UTL95-95 values
have different purposes, and it is not necessary for the UTL95–95 values
10 to be less than the Level I SQT values. However, this finding may point
out the widespread anthropogenic contamination or natural enrich-
0 ment of certain parameters. This issue should be taken into consider-
PCD_T4

PCD_T5

PCD_T6

PCD_T7

PCF_T4

PCF_T5

PCF_T6

PCF_T7
OCDD

OCDF

ation for remediation decisions at small, uncomplicated sediment sites

c) where individual SQTs or mean PEC-Qs may be used to develop remedi-
ation targets (Crane and MacDonald, 2003).

Fig. 6. a) Mean (SD) weight percentages of PCDD/F congeners, b) the ratio of detected
PCD2378 concentrations to PCD_T4 plotted for each lake, and c) mean (SD) weight
3.2. Fish trophic class comparisons of BDEs
percentages of PCDD/F homologs in lake #36 (Red Rock), the developed lakes, and the
other 20 lakes. For c), the ^ symbol represents four nondetect PCDF homologs in lake Two of the more commonly detected BDE congeners in sediment
#36, and statistically significant (p b 0.05) differences between the developed lakes and samples (i.e., BDEs-47 and 99; Table A-8) were also the dominant con-
20 lake group are noted with an asterisk.
geners in most composite fish tissue samples from lakes #5, 7, 16, 24,
and 27. In addition, BDEs-49, 100, and 153 were commonly detected
in these fish. For each of these five BDE congeners, there were no signif-
mean weight percentages of PCDD/F homologs in 20 study lakes, where icant differences (p N 0.05) in the mean wet weight concentrations be-
OCDD N PCD_T7 N PCD_T6 N PCF_T4 N PCF_T5 (Fig. 6c). OCDD and tween the trophic guilds of fish (i.e., predator, omnivore, benthic). The
PCD_T7 also dominated the homolog weight percentages in the devel- mean concentrations (pg/g wet wt.) were highest for BDE-47, followed
oped lakes (i.e., #27, 38). Interestingly, OCDD also dominated the by BDEs-99, 49, 100, and 153 (Table A-16). BDE-47 is usually found
PCDD/F homolog profile from residential and industrial areas in cen- more frequently than other BDE congeners in humans, fish, and other
tral South Africa (Nieuwoudt et al., 2009).The results of either t-tests biota, followed by BDEs-99, 100, 153, and 154 (USEPA, 2009).
or Mann-Whitney rank sum tests showed that the weight percentages The most common BDEs in the fish tissue samples were also promi-
of two PCDD/F homolog groups had statistically significant differences nent in other environmental studies. BDEs-47, 49, 99, and 100 were also
(p b 0.05). The mean homolog weight percentage of PCD_T6 was sig- prevalent in the atmosphere over Lake Superior and showed a net flux
nificantly higher (p b 0.05) in the other 20 lakes compared to the into the lake near Duluth, MN during the summer of 2011 (Ruge et al.,
1334 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338

Table 5
UTLs of detected SQT-compatible parameters and index values with potential outliers removed at the 5% significance level. UTL units are provided with each chemical group.

Parameter N Potential outliers removed at 5% Normal distributiona Gamma distributiona

significance level (lake ID numbers)
95% Approximately gamma UTL with
95% coverage

95% UTL with 95% WH HW

Coverage

Metal or metalloid (mg/kg dry wt.)

Arsenic 52 38, 51 – 18.2 18.6
Chromium 53 3 – 73.1 75.2
Copper 47 4, 6, 9, 25, 37, 38, 47 27.5
Lead 52 27, 38 67.2
Nickel 51 4, 5, 6 25.1
Zinc 54 121

PAHs (μg/kg dry wt.)

2-Methylnaphthalene 52 27, 38 17.2
Acenaphthene 53 27 – 42.3 44.7
Acenaphthylene 51 24, 27, 38 – 10.9 11.4
Anthracene 52 27, 38 14.9
Benzo[a]anthracene 49 24, 27, 32, 38, 47 – 41.0 43.8
Benzo[a]pyrene 51 27, 32, 38 – 43.5 46.0
Chrysene 49 24, 27, 32, 38, 47 64.3
Fluoranthene 51 27, 32, 38 – 162.3 172.1
Fluorene 50 24, 27, 34, 49 69.9
Naphthalene 53 27 32.8
Phenanthrene 52 27, 38 – 147.7 156.3
Pyrene 49 24, 27, 32, 38, 47 – 106.7 113.7
∑PAH13 51 27, 32, 38 636

PAH ESB Toxic Units 49 27, 31, 32, 38, 46 0.023

Metal PEC-Q 53 38 – 0.48 0.48

Mean PEC-Q 51 6, 27, 38 – 0.21 0.22

PCDD/F TEQs (ng TEQ/kg dry wt.) 21 27b, 38 6.2

UTL = Upper Tolerance Limit; N = number of samples after removing outliers; ID = identification; WH = Wilson Hilferty; HW = Hawkins Wixley; PAHs = polycyclic aromatic hydro-
carbons; ESB = equilibrium partitioning sediment benchmark; PEC-Q = probable effect concentration quotient; PCDD/F = polychlorinated dibenzo-p-dioxins/dibenzofurans; TEQ =
toxic equivalent.
a
UTLs for normally distributed data were preferred over those derived from gamma distributed data. When the data were not normally distributed, the gamma distribution was used to
estimate UTL values. Gamma UTLs were calculated two different ways, for which professional judgment can be used to select the value of interest.
b
Potential outlier was based on professional judgment; the other outlier was based on Dixon's outlier test at the 5% significance level. This test can only select one outlier when there are
b25 samples.

2015). Similarly, BDEs-47 and 99 (along with BDE-209) were the most
abundant congeners in water column samples from all five Great
Lakes (Venier et al., 2014). The prevalence of BDEs-47, 99, and 100 in
fish tissues from the study lakes was consistent with other studies of
fish tissues in western Lake Superior (MPCA, 2006), Lake Michigan
(Kuo et al., 2010), all five Great Lakes for lake trout (Crimmins et al.,
2012), in fillets of 18 species of Great Lakes fish in Canadian waters
(Gandhi et al., 2017), and in European high mountain lakes and Green-
land for liver and muscle tissues of trout (Vives et al., 2004). Concentra-
tions of BDE congeners in lake trout throughout the Great Lakes, in
general, are declining as a result of the elimination of the commercial
production of pentaBDE in the U.S. during 2004 (Crimmins et al.,
2012). Gandhi et al. (2017) found that major lower brominated BDEs
declined by 46–74% between 2006/07 and 2012 in Great Lakes fish fil-
lets from Canadian waters.
BDE-209 was detected in only one composite fish tissue sample, for
which a concentration of 575 pg/g wet wt. was observed in bluegills (an
omnivore species) from Lake Nokomis (#27). Stapleton et al. (2006)
showed that juvenile rainbow trout and common carp have a metabolic
pathway for the debromination of BDE-209, particularly in fish tissue as
opposed to the liver where it accumulates. Gandhi et al. (2017) also ob-
served that deca-BDEs in panfish, like bluegill and pumpkinseed, com-
prised a major contribution to ∑PBDE. The bluegill composite sample
from Lake Nokomis (#27) was based on skin-on whole fish with the
scales removed so BDE-209 likely accumulated in the liver of these
fish. Some of the other fish species examined in this study were based
on skin-on fillets or whole fish, and the fillets would be expected to
have lower concentrations of BDE-209 based on the work of Stapleton
et al. (2006). Kuo et al. (2010) also observed the concentration of
BDE-209 decreased at higher trophic levels, whereas BDEs-47 and 100
biomagnified in a Lake Michigan food web of plankton, Diporeia, lake
whitefish, lake trout, and Chinook salmon. Kuo et al. (2010) suggested
that partial uptake or biotransformation of BDE-209 could explain
their observations.
The fish tissue data were normalized for percent lipids to enable all
species to be treated similarly (Randall et al., 1998). For three trophic
classes of fish, summary statistics were generated for the lipid-
normalized concentrations of five BDE compounds (Table A-17). One-
way ANOVAs were run of the trophic class data. The statistical results
were significant (p b 0.05) for all five BDE congeners (Table A-18), indi-
cating the differences in the lipid-normalized mean values for the tro-
phic groups were greater than would be expected by chance. Next, the
Holm-Sidak method was used for all pairwise multiple comparisons.
For BDEs-47, 100, and 153, there were significant differences (p b
0.05) between the predator and omnivore trophic groups, as well as be-
tween the predator and benthic trophic groups (Table 7). In all cases,
the corresponding lipid-normalized mean BDE concentrations were
higher in the predator fish (Table A-17). There was also a significant dif-
ference (p b 0.05) between the predator and benthic trophic groups for
BDE-49 (Table 7), where the mean value was higher in the predator fish
(Table A-17). These differences were apparent despite various sample
processing procedures. The predator fish analyses were based on skin-
on fillets (removed scales), which excluded the other tissues and organs
included in the skin-on whole fish of the omnivore and benthic fish
 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338 1335

Table 6
UTLs of detected and censored SQT-compatible parameters (b80% nondetects) with outliers removed at the 5% significance level. UTL units are provided with each chemical group.

Parameter N (# Nondetects) % Nondetects Potential outliers removed Normal distribution and Gamma distribution
at 5% significance levels KM estimatesa and KM estimatesa
(lake ID numbers)
95% Approximately gamma
UTL with 95% coverage

95% UTL with 95% coverage WH HW

Metal (mg/kg dry wt.)

Cadmium 54 (6) 11.1 – 1.5 1.6
Mercury 53 (32) 60.4 47 0.26

PAH (μg/kg dry wt.)

Dibenzo[a,h]anthracene 51 (3) 5.9 27, 32, 38 5.2

Total PCBs (μg/kg dry wt.)

Total PCBspesticide scan 22 (2) 9.1 27c, 38 33.7
∑PCBcongenersb 22 (12) 54.6 27c, 38 36.6

Pesticide or metabolite (μg/kg dry wt.)

alpha-Chlordane 24 (17) 70.8 1.8
gamma-Chlordane 23 (15) 65.2 38 – 0.89 0.90
o,p′-DDD 22 (7) 31.8 27c, 38 2.1
o,p′-DDE 22 (13) 59.1 27, 38c – 0.36 0.36
p,p′-DDD 22 (4) 18.2 27c, 38 – 13.9 15.9
p,p′-DDE 21 (1) 4.8 27, 32c, 38c 10.7
p,p′-DDT 24 (15) 62.5 1.5
Total DDT 21 (1) 4.8 27c, 32c, 38 – 30.8 35.0
Dieldrin 24 (15) 62.5 0.29
trans-Nonachlor 23 (11) 47.8 38 0.70
cis-Nonachlor 22 (7) 31.8 27c, 38 0.46

UTL = Upper Tolerance Limit; N = number of samples after removing outliers; ID = identification; KM = Kaplan-Meier; WH = Wilson Hilferty; HW = Hawkins Wixley; PAH = polycyclic
aromatic hydrocarbon; PCBs = polychlorinated biphenyls; DDD = dichlorodiphenyldichloroethane; DDE = dichlorodiphenyldichloroethylene; DDT = dichlorodiphenyltrichloroethane.
a
UTLs for normally distributed data were preferred over those derived from gamma distributed data. When the data were not normally distributed, the gamma distribution was used to
estimate UTL values. Gamma UTLs were calculated two different ways, for which professional judgment can be used to select the value of interest.
b
The Kaplan-Meier method was used to determine means for 12 lake sediment samples with b80% nondetects, which were then multiplied by the number of congeners to yield es-
timated ∑PCBcongeners. For samples with N80% nondetects, the detected congener values and reporting limits for other congeners were summed and expressed as less than ∑PCBcongener
values used in the UTL calculation.
c
Potential outlier was based on professional judgment; all other outliers were based on statistical tests at the 5% significance level.

species. All other pairwise comparisons were not significant (p N 0.05)
(Table 7).
4. Implications
The results of this study will be used to help prioritize government
agency activities related to ambient sediment quality in Minnesota,
and for future status and trends work. For the first time, sediment chem-
istry (i.e., metals and metalloids, PAHs, PCB congeners, organochlorine
pesticides, TOC, and particle size) will be added to the 2017 NLA suite
of lakes nationwide. The Minnesota component of the 2017 NLA sedi-
ment data will provide an opportunity to reassess some of the UTL95-
95 values with additional data collected from a shorter depth interval
(i.e., 0–5 cm). The 2017 data set will include 12 of the lakes sampled
in 2007. If continued long-term, this national status and trends work
can help track the effectiveness of government programs and regula-
tions pertaining to a suite of contaminants. In addition, the nationwide
sediment contaminant data set to be collected from the 2017 NLA
suite of lakes will be freely available for others to use (e.g., calculation
of chemical specific UTL95-95 values by state or region).
As urban development expands in Minnesota, degradation of sedi-
ment quality in those areas is likely to occur. Incorporation of best man-
agement stormwater practices can alleviate some of these issues. In
addition, pollution prevention efforts can reduce the release of contam-
inants into the environment. For example, expansion of Project Green
Fleet, from the Minneapolis-based Environmental Initiative, is
retrofitting vehicles and equipment with diesel engines (e.g., school
buses) to reduce emissions (Crane and Hennes, 2016a). In addition,
statewide clean energy initiatives can result in lower emissions of
PAHs in both urban and rural areas of Minnesota (Crane and Hennes,

Table 7
Multiple pairwise comparisons by fish trophic classes for lipid-normalized BDEs with significant one-way ANOVA results (Table A-18). Pairs shaded gray are significantly different (p b
0.05) from each other as determined by the Holm-Sidak method.

p < 0.05 (gray shading)

BDE congener Predator vs. omnivore Predator vs. benthic Omnivore vs. benthic
BDE-47a
BDE-49a
BDE-99
BDE-100a
BDE-153
BDE = brominated diphenyl ether; ANOVA = analysis of variance.
a
Based on natural log transformation of the data.
1336 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338
2016a). In turn, these efforts can reduce local sources of PAHs to lakes,
although emission sources from outside Minnesota will not be affected.
The BDE congener results in sediment and fish tissue provided addi-
tional relevance for Minnesota's legislative actions to prohibit products
with pentaBDE and octaBDE in 2008, as well as an upcoming ban in
2018 on decaBDE. With recent declines in BDE congeners in Great
Lakes fish, Gandhi et al. (2017) recommend switching regular fish mon-
itoring in the Great Lakes to other in-use flame retardants and targeted
surveillance of BDE congeners. A similar approach may be helpful in
Minnesota lakes, too.
Bans on persistent organic chemicals may take decades to result in
contaminant reductions in surficial sediments. Several banned organo-
chlorine pesticides and their metabolites were still detected in a sub-
group of study lake sediments, even after being banned over 30–
40 years ago. Although PCBs were banned in 1979 (Johnson et al.,
2006), these legacy contaminants are still an important issue in urban
sediments. These recalcitrant compounds continue to cycle through
the environment, and nonlegacy PCBs (such as PCB 11 used in commer-
cial paint pigments; Hu and Hornbuckle, 2010) are an emerging con-
cern. Although Minnesota's 2014 ban on coal tar-based sealants
reduced a source of PAHs to urban waterbodies from new sealant appli-
cations (Crane, 2014), old sealant will continue to abrade and be
transported from parking lots and driveways to waterbodies. A ban on
coal tar-based sealants in Austin, TX in 2006 resulted in decreases of
∑PAH16 sediment concentrations by nearly one-half in 6–8 years in
Lady Bird Lake (Van Metre and Mahler, 2014). Pavlowsky (2013) esti-
mated that it could take ≥20 years to result in an 80–90% decrease in
PAH concentrations in stream and pond sediments in Springfield, MO
if a local ban on coal tar-based sealants was enacted.
Incorporation of buffer strips along public agricultural waterbodies
may help reduce sediment and contaminant loads to agricultural water-
ways. Minnesota's 2015 landmark Buffer Law, which was amended by
the Legislature and signed into law by Governor Mark Dayton on April
25, 2016, designates an estimated 110,000 acres of land for water qual-
ity buffer strips statewide (Crane and Hennes, 2016a). This new law is
very timely, because row crop farming has expanded substantially in
central Minnesota during the past decade due to higher crop prices
(Lark et al., 2015).
Other jurisdictions may compare their ambient sediment quality
data to this study. The most relevant comparisons may be for neighbor-
ing states and provinces. The addition of other ecosystem health indica-
tors, such as sediment toxicity tests and benthic invertebrate
community structure, would provide a well-rounded weight-of-evi-
dence approach for fully assessing ambient sediment quality.
Acknowledgements
The author appreciates the assistance of many individuals responsi-
ble for collecting, processing, and analyzing sediment and fish tissue
samples, in addition to internal staff who provided database, GIS, and
management support. Preliminary discussions about the sediment
quality results with Steve Hennes (MPCA—retired) were helpful. This
project was funded by internal state funds from the MPCA. In addition,
a joint powers agreement with the U.S. Fish and Wildlife Service
allowed most of the sediment samples to be analyzed by their contract
laboratory at a significant cost savings to the MPCA.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.scitotenv.2017.05.241.
References
ATSDR, 2002. Toxicological profile for aldrin/dieldrin. Agency for Toxic Substances and
Disease Registry. Atlanta, GA http://www.atsdr.cdc.gov/toxprofiles/tp1.pdf.
Baldwin, A.K., Corsi, S.R., Lutz, M.A., Ingersoll, C.G., Dorman, R., Magruder, C., Magruder,
M., 2016. Primary sources and toxicity of PAHs in Milwaukee-area streambed sedi-
ment. Environ. Toxicol. Chem. http://dx.doi.org/10.1002/etc.3694.
Barabas, N., Goovaerts, P., Adriaens, P., 2004. Modified polytopic vector analysis to identify
and quantify a dioxin dechlorination signature in sediments. 2. Application to the
Passaic River. Environ. Sci. Technol. 38, 1821–1827.
Bojakowska, I., Stasiuk, M., Gasior, J., Wolkowicz, W., 2014. Assessment of DDT, HCH and
PCB pollution of lake sediments in Poland. Limnol. Rev. 14, 63–74.
Budzinski, H., Jones, I., Bellocq, J., Piérard, C., Garrigues, P., 1997. Evaluation of sediment
contamination by polycyclic aromatic hydrocarbons in the Gironde estuary. Mar.
Chem. 58, 85–97.
Burgess, R.M., 2009. Evaluating ecological risk to invertebrate receptors from PAHs in sed-
iments at hazardous waste sites. U.S. Environmental Protection Agency, Ecological
Risk Assessment Support Center, Cincinnati, OH. EPA/600/R-06/162F https://cfpub.
epa.gov/ncea/risk/recordisplay.cfm?deid=214715.
Callender, E., Rice, K.C., 2000. The urban environmental gradient: anthropogenic influ-
ences on the spatial and temporal distributions of lead and zinc in sediments. Envi-
ron. Sci. Technol. 34, 232–237.
Centner, T.J., Houston, J.E., Keeler, A.G., Fuchs, C., 1999. The adoption of best management
practices to reduce agricultural water contamination. Limnologica 29, 366–373.
Chaky, D.A., 2003. Polychlorinated Biphenyls, Polychlorinated Dibenzo-p-dioxins and Furans
in the New York Metropolitan Area: Interpreting Atmospheric Deposition and Sediment
Chronologies. (Ph.D. Dissertation). Rensselaer Polytechnic Institute, Troy, NY (431 pp.).
http://digitool.rpi.edu:8881/R/TL4XXQ1X6A1ISXT1MFG9TNRJY7UF5135P49DYPVB5G
H1JEGS8A-00444?func=dbin-jump-full&object_id=15897&local_base=GEN01&pds_
handle=GUEST.
Christensen, G.N., Savinov, V., Savinova, T., Alexeeva, L., Kochetkov, A., Konoplev, A.,
Samsonov, D., Dauvalter, V.A., Kashulin, N.A., Sandimirov, S.S., 2007. Screening Stud-
ies of POP Levels in Bottom Sediments From Selected Lakes in the Paz Watercourse.
Akvaplan-niva, Tromsø, Norway (Report number 3665.01). https://helda.helsinki.fi/
handle/10138/38236.
Cleverly, D., Schaum, J., Schweer, G., Becker, J., Winters, D., 1997. The congener profiles of
anthropogenic sources of chlorinated dibenzo-p-dioxins and chlorinated dibenzofu-
rans in the United States. Organohalogen Compd. 32, 430–435.
Coulter, C.T., 2004. EPA-CMB8.2 users manual. United States Environmental Protection
Agency, Research Triangle Park, NC. (EPA-452/R-04-011) http://www.epa.gov/
scram001/models/receptor/EPA-CMB82Manual.pdf.
Crane, J.L., 2014. Source apportionment and distribution of polycyclic aromatic hydrocar-
bons, risk considerations, and management implications for urban stormwater pond
sediments in Minnesota, USA. Arch. Environ. Contam. Toxicol. 66, 176–200.
Crane, J.L., Hennes, S., 2007. Guidance for the use and application of sediment quality tar-
gets for the protection of sediment-dwelling organisms in Minnesota. Minnesota Pol-
lution Control Agency, St. Paul, MN. tdr-gl-04 https://www.pca.state.mn.us/sites/
default/files/tdr-gl-04.pdf.
Crane, J.L., Hennes, S., 2016a. Ambient sediment quality conditions in Minnesota. Minne-
sota Pollution Control Agency, St. Paul, MN. MPCA Document Number tdr-g1-19
https://www.pca.state.mn.us/sites/default/files/tdr-g1-19.pdf.
Crane, J.L., Hennes, S., 2016b. Ambient sediment quality conditions in Minnesota: Appen-
dices. Minnesota Pollution Control Agency, St. Paul, MN. MPCA Document Number
tdr-g1-19a https://www.pca.state.mn.us/sites/default/files/tdr-g1-19a.pdf.
Crane, J.L., MacDonald, D.D., 2003. Applications of numerical sediment quality targets for
assessing sediment quality conditions in a US Great Lakes Area of Concern. Environ.
Manag. 32, 128–140.
Crane, J.L., MacDonald, D.D., Ingersoll, C.G., Smorong, D.E., Lindskoog, R.A., Severn, C.G.,
Berger, T.A., Field, L.J., 2000. Development of a framework for evaluating numerical
sediment quality targets and sediment contamination in the St. Louis River Area of
Concern. Great Lakes National Program Office, United States Environmental Protec-
tion Agency, Chicago, IL. EPA 905-R-00-008 https://www.pca.state.mn.us/sites/
default/files/sqt-slraoc.pdf.
Crane, J.L., MacDonald, D.D., Ingersoll, C.G., Smorong, D.E., Lindskoog, R.A., Severn, C.G.,
Berger, T.A., Field, L.J., 2002. Evaluation of numerical sediment quality targets for
the St. Louis River Area of Concern. Arch. Environ. Contam. Toxicol. 43, 1–10.
Crimmins, B.S., Pagano, J.J., Xia, X., Hopke, P.K., Milligan, M.S., Holsen, T.M., 2012.
Polybrominated diphenyl ethers (PBDEs): turning the corner in Great Lakes trout
1980-2009. Environ. Sci. Technol. 46, 9890–9897.
Daughton, C.G., 2004. Non-regulated water contaminants: emerging research. Environ.
Impact Assess. Rev. 24, 711–732.
Dickhut, R.M., Canuel, E.A., Gustafson, K.E., Liu, K., Arzayus, K.M., Walker, S.E., Edgecombe,
G., Gaylor, M.O., MacDonald, E.H., 2000. Automotive sources of carcinogenic polycy-
clic aromatic hydrocarbons associated with particulate matter in the Chesapeake
Bay region. Environ. Sci. Technol. 34, 4635–4640.
Engstrom, D.R., Balogh, S.J., Swain, E.B., 2007. History of mercury inputs to Minnesota
lakes: influences of watershed disturbance and localized atmospheric deposition.
Limnol. Oceanogr. 52, 2467–2483.
Gandhi, N., Gewurtz, S.B., Drouillard, K.G., Kolic, T., MacPherson, K., Reiner, E.J., Bhavsar,
S.P., 2017. Polybrominated diphenyl ethers (PBDEs) in Great Lakes fish: levels, pat-
terns, trends and implications for human exposure. Sci. Total Environ. 576, 907–916.
Gao, S., Sun, C., Zhang, A., 2007. Pollution of polycyclic aromatic hydrocarbons in China. In:
Li, A., Tanabe, S., Jiang, G., Giesy, J.P., Lam, P.S.K. (Eds.), Persistent Organic Pollutants in
Asia: Sources, Distributions, Transport and Fate. Vol. 7. Elsevier, Amsterdam,
pp. 237–287.
Han, Y.M., Wei, C., Bandowe, B.A.M., Wilcke, W., Cao, J.J., Xu, B.Q., Gao, S.P., Tie, X.X., Li,
G.H., Jin, Z.D., An, Z.S., 2015. Elemental carbon and polycyclic aromatic compounds
in a 150-year sediment core from Lake Qinghai, Tibetan Plateau, China: influence of
regional and local sources and transport pathways. Environ. Sci. Technol. 49,
4176–4183.
 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338
Helsel, D.R., 2012. Statistics for Censored Environmental Data Using Minitab® and R.
second ed. John Wiley & Sons, Inc., Hoboken, NJ.
Herlihy, A.T., Sobota, J.B., McDonnell, T.C., Sullivan, T.J., Lehmann, S., Tarquinio, E., 2013. An
a priori process for selecting candidate reference lakes for a national survey.
Freshwat. Sci. 32, 385–396.
Hilscherova, K., Kannan, K., Nakata, H., Hanari, N., Yamashita, N., Bradley, P.W., McCabe,
J.M., Taylor, A.B., Giesy, J.P., 2003. Polychlorinated dibenzo-p-dioxin and dibenzofuran
concentration profiles in sediment and flood-plain soils of the Tittabawassee River,
Michigan. Environ. Sci. Technol. 37, 468–474.
Hsu-Kim, H., Kucharzyk, K.H., Zhang, T., Deshusses, M.A., 2013. Mechanisms regulating
mercury bioavailability for methylating microorganisms in the aquatic environment:
a critical review. Environ. Sci. Technol. 47, 2441–2456.
Hu, D., Hornbuckle, K.C., 2010. Inadvertent polychlorinated biphenyls in commercial paint
pigments. Environ. Sci. Technol. 44, 2822–2827.
IAGLR, 2002. Linking science and policy for urban nonpoint source pollution in the Great
Lakes region. International Association for Great Lakes Research, Ann Arbor, MI (30
pp). http://www.iaglr.org/scipolicy/nps/nps_iaglr02.pdf.
IJC, 1989. Great Lakes Water Quality Agreement of 1978 (as Amended by Protocol Signed
November 18, 1987). International Joint Commission, Windsor, ON.
Jartun, M., Ottesen, R.T., Steinnes, E., Volden, T., 2008. Runoff of particle bound pollutants
from urban impervious surfaces studied by analysis of sediments from stormwater
traps. Sci. Total Environ. 396, 147–163.
Johnson, G.W., Ehrlich, R., Full, W., 2004. Principal components analysis and receptor
models in environmental forensics. In: Murphy, B.L., Morrison, R.D. (Eds.), Introduc-
tion to Environmental Forensics. Elsevier Academic Press, Burlington, pp. 461–515.
Johnson, G.W., Quensen III, J.F., Chiarenzelli, J.R., Hamilton, M.C., 2006. Polychlorinated bi-
phenyls. In: Morrison, R.D., Murphy, B.L. (Eds.), Environmental Forensics: Contami-
nant Specific Guide. Elsevier, New York, pp. 187–225.
Jones, K.C., de Voogt, P., 1999. Persistent organic pollutants (POPs): state of the science.
Environ. Pollut. 100, 209–221.
Kahn, L., 1988. Determination of total organic carbon in sediment (Lloyd Kahn method).
U.S. Environmental Protection Agency, Region II Environmental Services Division, Ed-
ison, NJ http://www.nj.gov/dep/srp/guidance/rs/lloydkahn.pdf.
Kuo, Y.-M., Sepúlveda, M.S., Hua, I., Ochoa-Acuña, H.G., Sutton, T.M., 2010. Bioaccumula-
tion and biomagnification of polybrominated diphenyl ethers in a food web of Lake
Michigan. Ecotoxicology 19, 623–634.
Lark, T.J., Salmon, J.M., Gibbs, H.K., 2015. Cropland expansion outpaces agricultural and
biofuel policies in the United States. Environ. Res. Lett. 10. http://dx.doi.org/10.
1088/1748-9326/10/4/044003.
Li, A., Jang, J.-K., Scheff, P.A., 2003. Application of EPA CMB8.2 model for source apportion-
ment of sediment PAHs in Lake Calumet, Chicago. Environ. Sci. Technol. 37, 2958–2965.
Lohmann, R., Breivik, K., Dachs, J., Muir, D., 2007. Global fate of POPs: current and future
research directions. Environ. Pollut. 150, 150–165.
MacDonald, D.D., Ingersoll, C.G., Berger, T.A., 2000. Development and evaluation of
consensus-based sediment quality guidelines for freshwater ecosystems. Arch. Envi-
ron. Contam. Toxicol. 39, 20–31.
Mahler, B.J., Van Metre, P.C., Wilson, J.T., Musgrove, M., Burbank, T.L., Ennis, T.E., Bashara,
T.J., 2010. Coal-tar-based parking lot sealcoat: an unrecognized source of PAH to set-
tled house dust. Environ. Sci. Technol. 44, 894–900.
Minnesota Stormwater Steering Committee, 2017. Minnesota stormwater manual. Wiki
Version. Minnesota Pollution Control Agency, St. Paul, MN https://stormwater.pca.
state.mn.us/index.php?title=Main_Page.
Monson, B., Heiskary, S., 2008. Minnesota National Lakes Assessment Project: water mer-
cury concentrations in Minnesota lakes. Minnesota Pollution Control Agency, St. Paul,
MN. wq-nlap-02 https://www.pca.state.mn.us/sites/default/files/wq-nlap1-02.pdf.
MPCA, 2006. Polybrominated diphenyl ethers: the emerging contaminants in the Lake
Superior Watershed (GL2002-184). Environmental Outcomes Division, Minnesota
Pollution Control Agency, St. Paul, MN. MPCA Document Number tdr-g1-14.
MPCA, 2013. Sources of mercury pollution and methylmercury contamination of fish in
Minnesota. Minnesota Pollution Control Agency, St. Paul, MN. MPCA Document Num-
ber p-p2s4-06 https://www.pca.state.mn.us/sites/default/files/p-p2s4-06.pdf.
Nieuwoudt, C., Quinn, L.P., Pieters, R., Jordaan, I., Visser, M., Kylin, H., Borgen, A.R., Giesy,
J.P., Bouwman, H., 2009. Dioxin-like chemicals in soil and sediment from residential
and industrial areas in central South Africa. Chemosphere 76, 774–783.
O'Hara, K., Heiskary, S., Valley, R., Habrat, M., 2011. Sentinel lake assessment report. Elk
Lake (15-0010). Clearwater County, Minnesota. Water Monitoring Section, Minneso-
ta Pollution Control Agency, St. Paul, MN and Section of Fisheries, Minnesota Depart-
ment of Natural Resources, St. Paul, MN. MPCA Document Number wq-2slice15-0010
https://www.pca.state.mn.us/sites/default/files/wq-2slice15-0010.pdf.
Olsen, L.D., Valder, J.F., Carter, J.M., Zogorski, J.S., 2013. Prioritization of constituents for
national- and regional-scale ambient monitoring of water and sediment in the
United States. U.S. Geological Survey, Reston, VA. Scientific Investigations Report
2012–5218. 203 p. Plus Supplemental Tables https://pubs.usgs.gov/sir/2012/5218/.
Pavlowsky, R.T., 2013. Coal-tar pavement sealant use and polycyclic aromatic hydrocar-
bon contamination in urban stream sediments. Phys. Geogr. 34, 392–415.
Quadrini, J.D., Ku, W., Connolly, J.P., Chiavelli, D.A., Israelsson, P.H., 2015. Fingerprinting
2,3,7,8-tetrachlorodibenzodioxin contamination within the lower Passaic River. Envi-
ron. Toxicol. Chem. 34, 1485–1498.
Randall, R.C., Young, D.R., Lee II, H., Echols, S.F., 1998. Lipid methodology and pollutant
normalization relationships for neutral nonpolar organic pollutants. Environ. Toxicol.
Chem. 17, 788–791.
Ricking, M., Schwarzbauer, J., 2012. DDT isomers and metabolites in the environment: an
overview. Environ. Chem. Lett. 10, 317–323.
Ruge, Z., Muir, D., Helm, P., Lohmann, R., 2015. Concentrations, trends, and air-water ex-
change of PAHs and PBDEs derived from passive samplers in Lake Superior in 2011.
Environ. Sci. Technol. 49, 13777–13786.
1337
Schauer, J.J., Kleeman, M.J., Cass, G.R., Simoneit, B.R.T., 2001. Measurement of emissions
from air pollution sources. 3. C1–C29 organic compounds from fireplace combustion
of wood. Environ. Sci. Technol. 35, 1716–1728.
Schubauer-Berigan, M., Crane, J.L., 1997. Survey of sediment quality in the Duluth/Superi-
or Harbor: 1993 sampling results. Minnesota Pollution Control Agency, St. Paul, MN.
EPA 905-R97-005 https://www.pca.state.mn.us/sites/default/files/93mud.pdf.
Shields, W.J., Tondeur, Y., Benton, L., Edwards, M.E., 2006. Dioxins and furans. In:
Morrison, R.D., Murphy, B.L. (Eds.), Environmental Forensics: Contaminant Specific
Guide. Elsevier, New York, pp. 293–312.
Siami, M., McNabb, C.D., Batterson, T.R., Glandon, R.P., 1987. Arsenic sedimentation along
the slope of a lake basin. Environ. Toxicol. Chem. 6, 595–605.
Silliman, J.E., Meyers, P.A., Ostrom, P.H., Ostrom, N.E., Eadie, B.J., 2000. Insights into the or-
igin of perylene from isotopic analyses of sediments from Saanich Inlet, British Co-
lumbia. Org. Geochem. 31, 1133–1142.
Slater, G.F., Benson, A.A., Marvin, C., Muir, D., 2013. PAH fluxes to Siskiwit revisited: trends
in fluxes and sources of pyrogenic PAH and perylene constrained via radiocarbon
analysis. Environ. Sci. Technol. 47, 5066–5073.
Stapleton, H.M., Brazil, B., Holbrook, R.D., Mitchelmore, C.L., Benedict, R., Konstantinov, A.,
Potter, D., 2006. In vivo and in vitro debromination of decabromodiphenyl ether (BDE
209) by juvenile rainbow trout and common carp. Environ. Sci. Technol. 40,
4653–4658.
Taylor, K.G., Owens, P.N., 2009. Sediments in urban river basins: a review of sediment-
contaminant dynamics in an environmental system conditioned by human activities.
J. Soils Sediments 9, 281–303.
TDI – Brooks International, Inc, 2009. Analytical results report. TDI – Brooks International,
Inc., College Station, TX. Catalog Number 3080058. MPCA Document Number tdr-g1-
19b. https://www.pca.state.mn.us/sites/default/files/tdr-gl-19b.pdf.
Towey, T.P., Chang, S.-C., Demond, A., Wright, D., Barabás, N., Franzblau, A., Garabrant,
D.H., Gillespie, B.W., Lepkowski, J., Luksemburg, W., Adriaens, P., 2010. Hierarchical
cluster analysis of polychlorinated dioxins and furans in Michigan, USA soils: evalua-
tion of industrial and background congener profiles. Environ. Toxicol. Chem. 29,
64–72.
Towey, T.P., Barabas, N., Demond, A., Franzblau, A., Garabrant, D.H., Gillespie, B.W.,
Lepkowski, J., Adriaens, P., 2012. Polytopic vector analysis of soil, dust, and serum
samples to evaluate exposure sources of PCDD/Fs. Environ. Toxicol. Chem. 31,
2191–2200.
USEPA, 1975. DDT. A review of scientific and economic aspects of the decision to ban its
use as a pesticide. U.S. Environmental Protection Agency, Washington, DC. EPA-540/
1-75-022 https://nepis.epa.gov/Exe/ZyPDF.cgi/910128JF.PDF?Dockey=910128JF.
PDF.
USEPA, 2000a. Guidance for assessing chemical contaminant data for use in fish advi-
sories. Volume 1. Fish sampling and analysis. third ed. U.S. Environmental Protection
Agency, Office of Water, Washington, DC EPA 823-B-00-007. https://www.epa.gov/
sites/production/files/2015-06/documents/volume1.pdf.
USEPA, 2000b. Prediction of sediment toxicity using consensus-based freshwater sedi-
ment quality guidelines. U.S. Environmental Protection Agency, Great Lakes National
Program Office, Chicago, IL. EPA 905/R-00/007 http://www.cerc.usgs.gov/pubs/
center/pdfdocs/91126.pdf.
USEPA, 2003. Procedures for the derivation of equilibrium partitioning sediment bench-
marks (ESBs) for the protection of benthic organisms: PAH mixtures. U.S. Environ-
mental Protection Agency, Office of Research and Development, Washington, DC.
EPA/600/R-02/013 https://clu-in.org/conf/tio/porewater1/resources/EPA-ESB-
Procedures-PAH-mixtures.pdf.
USEPA, 2006. An inventory of sources and environmental releases of dioxin-like com-
pounds in the United States for the years 1987, 1995, and 2000. U.S. Environmental
Protection Agency, Washington, DC. EPA 600/P-03/002 https://cfpub.epa.gov/ncea/
risk/recordisplay.cfm?deid=159286&CFID=57378536&CFTOKEN=69449246.
USEPA, 2007. Method 1614. Brominated diphenyl ethers in water, soil, sediment, and tis-
sue by HRGC/HRMS. U.S. Environmental Protection Agency, Office of Water and Of-
fice of Science and Technology, Washington, DC. EPA-821-R-07-005 https://www.
nemi.gov/methods/method_summary/10575/.
USEPA, 2009. Polybrominated diphenyl ethers (PBDEs) action plan. U.S. Environmental
Protection Agency https://www.epa.gov/sites/production/files/2015-09/documents/
pbdes_ap_2009_1230_final.pdf.
Van den Berg, M., Birnbaum, L., Bosveld, A.T.C., Brunström, B., Cook, P., Feeley, M., Giesy,
J.P., Hanberg, A., Hasegawa, R., Kennedy, S.W., Kubiak, T., Larsen, J.C., Rolaf van
Leeuwen, F.X., Liem, A.K.D., Nolt, C., Peterson, R.E., Poellinger, L., Safe, S., Schrenk, D.,
Tillitt, D., Tysklind, M., Younes, M., Waern, F., Zacharewski, T., 1998. Toxic equivalency
factors (TEFs) for PCBs, PCDDs, PCDFs for humans and wildlife. Environ. Health
Perspect. 106, 775–792.
Van Metre, P.C., Mahler, B.J., 2005. Trends in hydrophobic organic contaminants in urban
and reference lake sediments across the United States, 1970–2001. Environ. Sci.
Technol. 39, 5567–5574.
Van Metre, P.C., Mahler, B.J., 2010. Contribution of PAHs from coal-tar pavement sealcoat
and other sources to 40 U.S. lakes. Sci. Total Environ. 409, 334–344.
Van Metre, P.C., Mahler, B.J., 2014. PAH concentrations in lake sediment decline following
ban on coal-tar-based pavement sealants in Austin, Texas. Environ. Sci. Technol. 48,
7222–7228.
Van Metre, P.C., Majewski, M.S., Mahler, B.J., Foreman, W.T., Braun, C.L., Wilson, J.T.,
Burbank, T.L., 2012. PAH volatilization following application of coal-tar-based pave-
ment sealant. Atmos. Environ. 51, 108–115.
Venier, M., Dove, A., Romanak, K., Backus, S., Hites, R., 2014. Flame retardants and legacy
chemicals in Great Lakes' water. Environ. Sci. Technol. 48, 9563–9572.
Vives, I., Grimalt, J.O., Lacorte, S., Guillamón, M., Barceló, D., 2004. Polybrominated ether
flame retardants in fish from lakes in European high mountains and Greenland. Envi-
ron. Sci. Technol. 38, 2338–2344.
1338 J.L. Crane / Science of the Total Environment 607–608 (2017) 1320–1338
Wetterauer, B., Ricking, M., Otte, J.C., Hallare, A.V., Rastall, A., Erdinger, L., Schwarzbauer, J.,
Braunbeck, T., Hollert, H., 2012. Toxicity, dioxin-like activities, and endocrine effects
of DDT metabolites—DDA, DDMU, DDMS, and DDCN. Environ. Sci. Pollut. Res. 19,
403–415.
Wöhrnschimmel, H., MacLeod, M., Hungerbuhler, K., 2013. Emissions, fate and transport
of persistent organic pollutants to the Arctic in a changing global climate. Environ.
Sci. Technol. 47, 2323–2330.
World Health Organization, 1999. Concise international chemical assessment document
6. biphenyl. United Nations Environment Programme, the International Labour Orga-
nisation, and the World Health Organization, Geneva, Switzerland (31 pp). http://
www.who.int/ipcs/publications/cicad/en/cicad06.pdf.
Wortmann, C., Morton, L.W., Helmers, M., Ingels, C., Devlin, D., Roe, J., McCann, L., Van
Liew, M., 2011. Cost-effective water quality protection in the Midwest. Heartland Re-
gional Water Coordination Initiative http://extensionpublications.unl.edu/assets/
pdf/rp197.pdf.
Wright Jr., H.E., Coffin, B., Aaseng, N.E. (Eds.), 1992. The Patterned Peatlands of Minnesota.
University of Minnesota Press, Minneapolis, MN (376 pp).
Zarull, M.A., Hartig, J.H., Krantzberg, G., 2001. Contaminated sediment remediation in the
Laurentian Great Lakes: an overview. Water Qual. Res. J. Can. 36, 351–365.