Full text of DOI:10.1002/etc.5620211019

Environmental Toxicology and Chemistry, Vol. 21, No. 10, pp. 2157–2164, 2002
Printed in the USA
0
730-7268/02 $9.00 ϩ .00
LEVEL AND EXTENT OF MERCURY CONTAMINATION IN OREGON, USA,
LOTIC FISH
S
PENCER A. PETERSON,*† ALAN T. HERLIHY,‡ ROBERT M. HUGHES,§
AMES M. ROBBINS
U.S. Environmental Protection Agency, National Health and Ecological Effects Research Laboratory, Western Ecology Division,
00 Southwest 35th Street, Corvallis, Oregon 97333
Oregon State University, Department of Fisheries and Wildlife, Corvallis, Oregon 97339, USA
Dynamac, 200 Southwest 35th Street, Corvallis, Oregon 97333, USA
U.S. Department of Energy, Albany Research Center, Albany, Oregon 97321-2198, USA
KATHRYN L. MOTTER,§ and
J
࿣
†
2
‡
§
࿣
(
Received 17 September 2001; Accepted 18 March 2002)
Abstract—We conducted a probability survey of 154 Oregon, USA, stream and river sites to assess the spatial extent of mercury
(
Hg) contamination in fish tissue. Samples consisted of whole-fish analyses of both small (
Ͻ
120 mm) and large (
Ͼ
ϭ
120 mm) fish
0.029), large
¯
at each site, when both were present. Overall, Hg levels (
piscivores (
␮
g/g) in small fish (
X
ϭ
0.031; standard deviation [SD]
¯
¯
ϭ
X
ϭ
0.284; SD
ϭ
0.175), and large invertivores (
X
0.055; SD
ϭ 0.047) were found within fairly narrow ranges;
always above detection (0.0025
␮
g/g) and almost always below 0.4 ␮g/g. Given the great ecoregion diversity across Oregon, the
narrow range in fish tissue Hg levels suggests that atmospheric transport is an important vehicle for Hg distribution. In small fish,
Hg levels were almost always low and showed little meaningful difference among fish taxa. In large fish, Hg levels were significantly
related to fish length. Piscivores (pikeminnow and bass) had significantly higher Hg levels, and the slope of their Hg level/length
relationship was much steeper than for invertivores. Salmonids, the most commonly occurring fish taxon in Oregon, exceeded 0.1
␮
g Hg/g (deemed protective for fish-eating mammals) in an inferred 15% of stream lengths where they occurred. Pikeminnows
and bass were found at fewer sites, but they exceeded 0.1
␮
g Hg/g in an inferred 96 and 70%, respectively, of stream lengths
where they occurred.
Keywords—Mercury
Fish tissue
Probability sampling Regional assessment
INTRODUCTION
directly downwind from power plants and heavy industrial
sites. Thus, bioaccumulation of Hg in these areas is not sur-
prising. Porcella [9], while discussing the bioaccumulation of
Hg, called for better temporal data on Hg accumulation in
biological populations. We agree and contend further that a
better understanding of the spatial patterns of elevated Hg
concentrations in biological populations is at least as important
to help understand where individual fish/species and site con-
centrations fit relative to regional population patterns.
Therefore, in 1997 and 1998, we conducted a survey to
determine patterns of the level and extent of Hg concentrations
in lotic fishes across Oregon. We focused our analysis not on
individual fish or fish species but rather on major fish groups,
that is, large versus small fish, piscivore versus invertivore
trophic guild, and large (boatable) rivers versus small (wade-
able) streams. Sampling sites were selected on a probability
basis such that results of the survey could be inferred statewide
to estimate the proportion of stream lengths exceeding various
Hg consumption thresholds. The purpose of this paper is to
describe the whole-fish concentration patterns of Hg in stream
and river fish groups across Oregon to determine what pro-
portion (based on 1:100,000-scale mapped stream length) of
associated lotic systems in the state might pose a risk to pi-
scivorous mustelids and to discuss possible Hg sources.
Methylmercury (a neurotoxin) contamination of the aquatic
environment is of increasing concern to wildlife and human
health professionals [1]. Aquatic organisms, particularly fish,
bioaccumulate and bioconcentrate methylmercury from both
contaminated water and the food chain [2] but predominantly
through food intake [3]. Thus, predatory fishes commonly ac-
cumulate higher concentrations of mercury than do herbivo-
rous ones, and mercury concentration increases as the fish age
and grow [4]. Nearly all Hg accumulated in fish tissue (95–
9
9%) is in the form of methylmercury [4,5]. Therefore, con-
sumption of fish by mammals greatly increases their exposure
to methylmercury and the potential neurological effects of the
substance.
It has been long recognized that Hg contamination results
from certain industrial and mining activities. It was assumed
that elimination of these point-source discharges would solve
most Hg contamination problems. However, that assumption
has been challenged since a number of U.S. states, Canadian
provinces, and the U.S. Environmental Protection Agency
(
U.S. EPA) have begun monitoring Hg in fish tissue. Recent
discoveries of high Hg levels in fish and lake sediment from
remote areas [6,7] implicates a much more widespread source:
atmospheric deposition [8].
Most of the previously described, remote study sites were
APPROACH
Probability sampling design
*
To whom correspondence may be addressed
A combined total of 187 site visits were made during 1997
and 1998 to collect fish and determine their Hg content (Fig.
1). Sites were drawn, using a spatially balanced, randomized
(
peterson.spencer@epa.gov).
Mention of trade names or commercial products does not constitute
endorsement or recommendation for use.
2
157

2
158
Environ. Toxicol. Chem. 21, 2002
S.A. Peterson et al.
Fig. 1. Location of sampling sites and their fish tissue mercury levels (
␮
g Hg/g fish tissue) in the western aggregate and eastern aggregate
ecoregions of Oregon (Cascade Mountain crest divides the two ecoregions). Critical incremental mercury thresholds from Yeardley et al. [12]
are given for piscivorous birds (0.02), mustelids (0.1), and humans (0.2).
sampling, from all lotic courses in Oregon that appeared on
the 1:100,000-scale Digital Line Graph database of the U.S.
Geological Survey [10]. Actual site selection followed a pro-
cedure that recognized the continuous nature of lotic systems
otter). Their [12] numbers were derived from the Great Lakes
Water Quality Initiative (GLWQI) model [13]. More conser-
vative or more liberal thresholds would have resulted in greater
or lesser amounts, respectively, of estimated Oregon stream
length adversely affected by Hg-contaminated fish.
[
11]. The procedure controls for spatial dispersion of the sam-
ples, considers their variable spatial density, and allows for
the occasional need to conduct nested subsampling. This prob-
ability-selected set of lotic sites represents an estimated target
population of 59,655 km (37,069 mi) of streams and rivers
throughout Oregon. The selection process has nonzero, joint
inclusion probability densities that permit rigorous design-
based, expansion factor–derived inference and variance esti-
mates for the population of interest.
A major benefit of the statistical sampling design is its
ability to infer regional conditions, such as the quantity, qual-
ity, and extent of ecological resources, from a limited number
of sampling sites. To estimate the potential risk associated with
Hg-contaminated fish tissue, we focused on the proportion (%)
and length (km) of stream/river courses exceeding critical Hg
threshold values [12].
Sample sites
The Cascade Mountain crest (USA) divides the state into
western and eastern sections. The western section, with the
exception of the Willamette valley, is mostly mountainous and
forested. Most of the state’s human population resides in the
heavily farmed Willamette valley. Western Oregon receives
much more precipitation than eastern Oregon. The increased
precipitation and mountainous terrain contributes to an in-
creased number of streams in this region. By sharp contrast,
eastern Oregon is sparsely populated and semiarid and has few
flowing waters, especially in the southeastern part of the state.
Thus, most of the sample sites in eastern Oregon are located
in the Blue Mountains and eastern slope of the Cascades. Be-
cause fewer rivers than streams exist, rivers were sampled with
a higher probability than streams. Thus, while not evident from
Figure 1, the sample site distribution results from the natural
spatial density of streams and rivers throughout the state and
Critical mercury thresholds
We used 0.1 g Hg/g of fish tissue, calculated by Yeardley
et al. [12], as our critical threshold for mustelids (mink and

Level and extent of mercury in Oregon
Environ. Toxicol. Chem. 21, 2002
2159
Table 1. Sample sizes for fish tissue analyses in 1997–1998 Oregon
Environmental Monitoring and Assessment Program sample; big fish:
from proportionally increased sampling of river sites to com-
pensate for their lower density.
total length
Ͼ120 mm, small fish: total length Ͻ120 mm
Sample collection and analysis
Aggregate ecoregion
Eastern
Details of wadeable stream and nonwadeable electrofishing
sampling procedures are described elsewhere [14]. All caught
fish were identified and enumerated. Subsets of the fish were
retained for mercury analysis. Two types of composite samples
were collected. One composite sample consisted of 50 to 200
g of fish species whose adults are small (
second composite sample was prepared from fish of the same
species whose adults are larger ( 120 mm) (e.g., catostomids,
salmonids, centrarchids, large cyprinids, and ictalurids). Any
composite samples containing fish that varied in size by
were dropped from our analyses.
Fish were measured, double plastic bagged, labeled, and
iced for shipment to the analytical laboratory the next day. On
Western
Total
Number of sample sites
Big fish
Streamsa
Rivers
Total
31
33
64
20
10
30
51
43
94
Ͻ120 mm). The
Ͼ
Small fish
Streams
Rivers
Total
43
28
71
18
10
28
61
38
99
Ͼ20%
Number of fish samplesb
Big fish
Invertivoresc
Piscivores
Total
76
17
93
38
8
46
114
25
139
receipt, samples were frozen at
of preparation for analysis.
Ϫ15ЊC and stored until time
Small fish
Laboratory procedures
Minnows
Sculpins
22
44
5
11
13
4
33
57
9
Sample preparation and digestion. Our primary interest
was in assessing the proportion of stream/rivers that might
pose a risk to wildlife due to the consumption of Hg-contam-
inated fish. Since wildlife consume whole fish, including the
skeleton and viscera, and because the U.S. EPA recommends
that total Hg concentrations be determined in state fish con-
taminant monitoring surveys [15], we measured total Hg con-
centrations in whole-fish samples. In addition, since 95 to 99%
of the Hg in fish is methylmercury [16], our results are reported
as total wet-weight Hg concentrations.
Whole fish were partially thawed and blended. Blending of
the sample with deionized water was at a ratio of approxi-
mately 1:1 until the mixture appeared to be completely ho-
mogeneous. Subsamples were extracted immediately to pre-
vent lipid separation. Dry-weight equivalent subsamples of 0.5
g were digested in 10 ml of high-purity, subboiled distilled
nitric acid (High Purity Standards, Charleston, SC, USA) ac-
cording to the CEM Run Fish Digest program [17] with the
d
Other
Total
71
28
99
a
Streams: wadeable; fish capture used a backpack elecrofisher. Rivers:
boatable; fish capture used a raft mounted electrofisher.
One same species five-fish composite sample was analyzed at 67%
of the sites. At all other sites, composite samples contained two to
four fish of the same species or individual fish when multiple species
were captured.
b
c
Invertivores include bullhead (Ameirus sp.), minnow (Richardsonius
balteatus
;
Rhinichthys osculus), sculpin (Cottus sp.), yellow perch
C. colum-
(
Perca flavescens), sucker (Catostomus macrocheilus
;
bianus), trout (Oncorhynchus clarki; O. mykiss; Salmo trutta), and
whitefish (Prosopium williamsoni). Piscivores include bass (Mi-
cropterus dolomieui; M. salmoides) and northern pikeminnow
(
Ptychocheilus oregonensis).
d
Other includes small trout and mixed-species composite samples.
ment Standards, Ottawa, ON, Canada). Mercury recovery from
the DORM-2 standard reference material (4.62 0.26 ppm)
Ϯ
following modifications: maximum power
centage power 100%, ramp time 25 min, pressure
psi, temperature 210 C, and hold time 15 min. After each
flask cooled, 10 ml of Ultrex 30% hydrogen peroxide (Fisher
Scientific, Houston, TX, USA) were added and the contents
redigested in the microwave using the same modified fish di-
gestion program. After the second digestion and cooling, the
contents were quantitatively transferred to 50-ml volumetric
flasks and diluted to volume with deionized water.
Manual mercury analysis. Samples were manually analyzed
for total Hg by cold vapor atomic absorption (CVAA) using
a Perkin-Elmer 5100 (Perkin-Elmer, Norwalk, CT, USA).
Analysis was according to [18] with minor modifications; a
bubbler vessel with a fritted glass stopper was substituted for
the 300-ml biochemical oxygen demand (BOD) bottle. Quality
ϭ
1,200 W, per-
was 107% with a 4.3% relative standard deviation (indicating
very thorough digestion).
ϭ
ϭ
ϭ
ϭ
200
Њ
ϭ
All digestates were analyzed in replicate. Analyses were
repeated in replicate if differences greater than 5% were en-
countered. Peak height and area data were simultaneously col-
lected over a 110-s read time using the Perkin-Elmer WinLab
software package [19]. Peak-height total mercury results were
recorded in
The instrument detection limit for this manual mercury ap-
paratus was 0.0025 g/g of wet-weight fish tissue.
␩g and converted to ␮g/g wet-weight fish tissue.
␮
Data analysis. Omernik level III ecoregions [20] were com-
bined into two classes (aggregate western and aggregate east-
ern ecoregions) using the crest of the Cascades as the dividing
line (Fig. 1). Fish species were combined into taxa (e.g., min-
now, sculpin) and trophic guilds (piscivores included bass and
northern pikeminnow; invertivores included all other species;
see Table 1 footnotes for further descriptions). Although some
large trout are piscivores, their diet is mostly invertebrates for
the fish sizes analyzed in this study.
We used an analysis of covariance (ANCOVA) to test for
differences in Hg levels for the main effects of ecoregion and
fish class. Determining the ecoregion and fish class main ef-
fects with all eight ecoregions and nine fish species would
produce estimates with poor statistical confidence, so we lim-
assurance was a critical component of our analysis. A 100 ␩g/
ml National Institute of Standards and Technology (Gaithers-
burg, MD, USA) traceable Hg standard preserved in 1% hy-
drochloric acid plus 0.01% potassium dichromate (NBS 136c)
was used for calibration. The National Institute of Standards
and Technology Hg standard reference material 1641c (100:
L) was used as a secondary standard. The average Hg recovery
from fish tissue was estimated from recovery of Hg from
DORM-2–certified fish tissue (Institute for National Measure-

2
160
Environ. Toxicol. Chem. 21, 2002
S.A. Peterson et al.
ited the main effects to two ecoregion classes (aggregate west-
ern vs aggregate eastern) and two fish guilds. Big fish (
39) were tested for the difference between piscivores and
invertivores, while small fish ( 99) were tested for the
n
ϭ
1
n
ϭ
difference among sculpins, minnows, and other species (all
small fish were invertivores) (Table 1). In the ANCOVA mod-
el, watershed size and fish length were used as covariates.
Mercury concentrations were log10 transformed before anal-
ysis. All analyses were run using PROC GLM in SAS௡ [21].
In the big-fish statistical model, we tested for the signifi-
cance of the two main effects: trophic guild, ecoregion, and
their interaction as well as the significance of the covariates.
Since we expected Hg to be related to fish size, we calculated
the least-squares means and standard errors for each effect
because this removes the effects of the covariates and nor-
malizes the means to a common denominator. Most (70%) of
the big fish analyzed were single individuals. Because Hg con-
centration is strongly influenced by fish size, we excluded any
sample if the coefficient of variation of the lengths of the
Fig. 2. Relationship between mercury concentration (
fish and watershed area for piscivore and invertivore trophic guilds.
No significant relationship was observed with area ( 0.1) in the
ANCOVA model or univariately within either trophic guild (
␮g/g) in large
p
Ͼ
2
r
ϭ 0.02
2
for piscivores;
r ϭ 0.02 for invertivores).
individual big fish in the composite sample was
Ͼ20%. This
eliminated 13 of the 152 big-fish Hg samples. We ran the same
ANCOVA on the small-fish data with the exception of fish
length, which was not part of the small-fish data, and fish
families were used instead of trophic guilds. We analyzed 139
big-fish samples collected from 94 different sites representa-
tive of approximately 27,800 km of streams and rivers in
Oregon (Table 1). The samples were about evenly divided
between wadeable streams and boatable rivers, and about twice
as many samples were in the western ecoregion than in the
eastern ecoregion. Invertivore species samples were much
more common than piscivore species samples (114 vs 25). We
also analyzed composite small-fish samples from 99 different
sites representing approximately 28,300 km of streams and
rivers in Oregon. For various reasons, some portions of some
streams could not be assessed (insufficient sample size, sample
loss, or lack of funds for analysis).
we ran separate ANCOVA analyses for piscivores and inver-
tivores. Results from the separate ANCOVA models showed
that big-fish piscivores had least-squares mean Hg concentra-
tions of 0.147 and 0.225
␮g Hg/g in eastern and western
Oregon, respectively, four to five times higher than the mean
eastern and western Oregon invertivore concentrations of
0
.035 and 0.042 ␮g/g (Table 2). These mean trophic guild
concentration differences are well outside their respective 95%
confidence intervals and show that piscivores had significantly
higher fish tissue Hg concentrations than invertivores.
Mercury concentrations in piscivores and invertivores in
the western aggregate ecoregion were slightly higher than in
the eastern aggregate ecoregion. However, the differences from
the two ecoregions were not significant at the 0.05 level for
either piscivores (
F ϭ 2.01, p ϭ 0.17) or invertivores (F ϭ
RESULTS AND DISCUSSION
Big fish
It appeared that a relationship existed between big-fish tis-
sue Hg concentration and watershed area. This appearance
occurred because piscivores showed much higher Hg concen-
trations than invertivores and piscivores were found only in
2
watersheds with areas
Ͼ
100 km (Fig. 2). However, when the
data were analyzed taking trophic guild into account, no mer-
cury–watershed area relationship was observed, either by sim-
ple linear regression or in the ANCOVA model (
F ϭ 0.03, p
ϭ
0.874). Thus, it seems that stream/river size had little effect
on big-fish tissue Hg concentration. Therefore, watershed area
was removed from the ANCOVA model in subsequent anal-
yses.
Fish size, however, had a very strong relationship to big-
fish tissue Hg concentration (Fig. 3). Thus, we used an AN-
COVA model to test for ecoregion and trophic guild effects
taking into account fish size (length). The ANCOVA assumes
a common slope between the dependent variable (Hg) and the
covariate (length) among all the study classes. A visual ex-
amination of Figure 3 and a
t
test of the slopes showed that
Fig. 3. Relationship between large-fish tissue mercury concentration
(␮g/g) and fish length or mean fish length (in the case of composite
the mercury–length relationship (slope) was very different be-
tween piscivores and invertivores but that the relationship was
very similar for eastern versus western aggregate ecoregions.
Therefore, the ecoregion effect could be analyzed by AN-
COVA, but the trophic guild effect could not. Because of this,
samples), shown as piscivore and invertivore trophic guilds. (logHg
ϭ
ϭ
ϭ
Ϫ
Ϫ
Ϫ
ϫ
1.75
1.69
2.32
ϫ
ϫ
0.00142Length for aggregate eastern invertivores, logHg
0.00153Length for aggregate eastern invertivores, logHg
ϫ
0.00582Length for aggregate western piscivores, logHg ϭ Ϫ1.74
0.00431Length for aggregate western piscivores).

Level and extent of mercury in Oregon
Environ. Toxicol. Chem. 21, 2002
2161
Table 2. Least-squares means, 95% confidence intervals, and significance for fish tissue mercury concentrations in Oregon (USA) streams/rivers
Least-squares mean mercury
Effect
concentration (
␮
g/g)
95% confidence interval
ANCOVA model significance
Big fish (total length
Ͼ
120 mm)a
Invertivores
Aggregate ecoregion
Western
Eastern
F
F
ϭ
ϭ
1.72,
2.01,
p
p
ϭ
ϭ
0.19
0.17
0.042
0.035
0.035–0.049
0.027–0.044
Piscivores
Aggregate ecoregion
Western
Eastern
0.225
0.147
0.161–0.315
0.090–0.239
Small fish (total length
Ͻ
120 mm)
Aggregate ecoregion
Western
Eastern
Fish taxa
Minnow
Sculpin
F
F
ϭ
ϭ
2.42,
2.23,
p
p
ϭ
ϭ
0.12
0.11
0.025
0.015
0.017–0.036
0.010–0.020
0.025
0.014
0.020
0.017–0.037
0.010–0.020
0.010–0.042
Other
Ecoregion
ϫ
fish taxa
F
ϭ
0.49,
p
ϭ
0.61
interaction
a
Fish length was used as a covariate in the analysis of covariance (ANCOVA) model for big fish and was highly significant for both the
invertivore ( 10.7, 0.002) and piscivore ( 13.9, 0.001) models.
F
ϭ
p
ϭ
F
ϭ
p ϭ
1
.72,
p
ϭ
0.19; Table 2). The relationship between fish tissue
tissue exceeds the mustelid threshold of 0.1 ␮g Hg/g is due
Hg concentration and fish length was strong and significant
for both the invertivore ( 10.7, 0.002) and the piscivore
13.9, 0.001) models.
mostly to large fish, specifically, pikeminnow, bass, and trout.
Cumulative distribution functions (CDFs) present all the data
for the different trophic guilds and for piscivores in the western
and eastern ecoregions of Oregon (Figs. 5 and 6, respectively).
Figure 4 illustrates the proportion of stream/river length where
the mustelid threshold level is exceeded relative to the oc-
currence of specific taxa. For pikeminnow, bass, and trout, the
proportions amount to 96.3, 70.3, and 15.3%, respectively,
relative to their occurrence. Note, however, that trout were
estimated to occur in up to 34,000 km of Oregon stream/river
courses, while pikeminnow occurred in only about 5,000 km.
Therefore, the potential environmental risk due to trout con-
sumption, while lower based on Hg concentration, is far more
widespread than that for other species. The estimated propor-
tion of stream/river length where minnows and sculpins ex-
ceeded the mustelid Hg threshold was relatively small (15.0
and 2.4%, respectively) compared to the occurrence of these
F
ϭ
p ϭ
(
F
ϭ
p ϭ
Small fish
In general, Hg concentrations in small-fish composite sam-
ples were markedly lower than they were in big-fish samples
(
Table 2). As for big fish, the relationship between small-fish
tissue Hg concentration and watershed area was not significant
0.22, 0.641), so this component was removed from
(
F
ϭ
p ϭ
the ANCOVA model in subsequent analyses. Minnows had a
slightly higher Hg concentration than other taxa, but it was
not significant at the 0.05 level (
). The Hg concentrations were very low in all three taxa
0.014–0.025 g Hg/g; Table 2). As for big fish, higher Hg
F ϭ 2.23, p ϭ 0.11) (Table
2
(
␮
concentration occurred in the western ecoregion, but the dif-
ference between the two ecoregions was not significant at the
0
.05 level(
F
ϭ
2.42,
p
ϭ
0.12) (Table 2). Least-squares mean
taxa. A high percentage (ϳ87%) of assessed stream/river
mercury concentration in the western and eastern ecoregions
was 0.025 and 0.015, respectively. We found no significant
relationship for the interaction between small fish taxa and
ecoregion (
F ϭ 0.49, p ϭ 0.61) (Table 2).
Population estimates
Since fish trophic guild (piscivore/invertivore), fish length
big/small), and fish taxa all influence the amount of Hg in
(
fish tissue, it is most appropriate to infer statewide estimates
of critical thresholds being exceeded from our data on the basis
of fish taxon. Therefore, we report the concentration of Hg in
fish by taxon and estimate the amount of stream/river length
exceeding critical mustelid threshold levels for the various fish
Fig. 4. Estimates of stream/river length in Oregon by fish taxa where
whole-fish concentrations exceeded a threshold value of 0.1 g Hg/
(
Fig. 4). Grouping the various fish taxa into their respective
␮
trophic guilds demonstrates the similarity in Hg concentrations
between small fish and large invertivores as well as the dis-
similarity of these two groups with the large piscivores (Fig.
g wet weight. Extent was estimated using all the Oregon Environ-
mental Monitoring and Assessment Program fish catch data relative
to a state total target stream/river length of 59,566 km. The not-
assessed category indicates sites where fish were caught during sam-
pling but specimens were not analyzed for mercury content because
of insufficient sample size, sample loss, or lack of funds for analysis.
5
).
The amount of stream/river length in Figure 4 where fish

2
162
Environ. Toxicol. Chem. 21, 2002
S.A. Peterson et al.
for the final wildlife value of 1,300
p
g Hg/L of water made
it more predictive for wildlife protection.
Mercury advisory levels for fish have been under discussion
for years with most states issuing advisories at levels lower
than the Food and Drug Administration action level for com-
mercially sold fish (1.0
Health Organization 0.5
For example, Oregon has adopted a 0.35-
␮
␮
g Hg/g of fish tissue) or the World
g Hg/g of fish tissue advisory [23].
g Hg/g level in
␮
edible fish tissue relative to humans (Oregon Human Resources
News. 1997. Elevated levels of mercury in sport-caught bass
and squawfish from the Willamette River. Oregon Health Di-
vision News Letter, February 13, 1997, Oregon Health Divi-
sion, Portland, OR, USA). In 1995, the U.S. EPA recom-
mended lowering the human health safe level for Hg exposure
to 0.1
␮
g Hg/kg of human body weight per day based on the
g-Hg/g rec-
Faroe Islands (Denmark) study [24]. The 0.1-
␮
ommendation from U.S. EPA was disputed by several entities
as overly protective. However, a recent National Academy of
Science report supports the U.S. EPA’s contention that the
human health safe level should be lowered to 0.1
d [25]. The U.S. EPA’s 2001 human health criteria for Hg in
fish tissue was set at 0.30 g/g wet weight [26]. Our objective
was not to defend any of the variously established critical Hg
thresholds. We used the Yeardley et al. [12] mustelid threshold
simply as a marker for our fish tissue data because the number
is of recent derivation [12] and based on a validated model
␮g Hg/kg/
Fig. 5. Cumulative distribution function (CDF) of total fish tissue
mercury ( g/g) for large invertivore, large piscivore, and small fish
relative to their proportion of occurrence throughout stream/river
lengths in Oregon.
␮
␮
length exceeds the Yeardley et al. [12] mustelid Hg threshold
(
0.1
tivores and small fish is both similar and small (
).
␮
g Hg/g) in large piscivores, while that for large inver-
[
22]. In addition, concern is growing [23–25] about the con-
ϳ
2–3%) (Fig.
tinued use of higher safe levels for Hg in fish tissue. If the
reader has interest in some Hg threshold other than the one
we have used as a marker, Figure 5 can be used to estimate
the amount of stream length that might be adversely impacted
by that or any other Hg threshold of interest.
5
The 0.1-␮g-Hg/g mustelid threshold for fish tissue could
be overly protective since apparently unaffected mustelid pop-
ulations commonly are found in proximity to fish populations
whose body burden of Hg exceeds the Yeardley et al. [12]
threshold level that they derived from the GLWQI model.
However, Wolfe and Norman [22] recently validated the model
supporting the GLWQI guidelines in a field study at Clear
Lake (CA, USA). They indicated that the model as originally
applied was too conservative but that modifications adapted
If, for example, the 0.35-␮g-Hg/g level, currently used by
Oregon as the advisory level relative to human consumption
of fish, was selected, approximately 17% of lotic courses in
the state contain large piscivores exceeding that level (Fig. 5).
However, the amount of Oregon stream/river length affected
by the Oregon 0.35
␮g Hg/g relative to human consumption
(
fillets) likely would be greater than our whole-fish data in-
dicates. While it has been reported that fillet and whole-fish
concentrations of Hg were nearly the same in Maine lakes
[
12], that seems unusual and not intuitive. Most sources find
the greatest amount of Hg in fillets, even though higher con-
centrations might exist in liver, spleen, and brain tissue [27,28].
Our own experience, based on a single analysis (unknown
sucker species), indicated that Hg muscle concentration was
4
3% greater than whole-fish concentration. The single analysis
is hardly definitive, but it supports the concept [27,28] that
whole-fish Hg concentrations would be lower than those of
fillets given the dilution effect of the large visceral mass of
most fish. Goldstein et al. [29] found that carp (Cyprinus car-
pio) and channel catfish (Ictalurus punctatus) muscle con-
tained 42 and 39% higher Hg concentrations, respectively, than
whole fish.
Until now, Oregon has had Hg sampling data for only about
2
% of the state’s 5,000 water bodies (Shipley S. 1999. Water
monitoring spotty. Statesman Journal. Salem, OR, February
). This is because the state, like most other states and some
4
federal agencies, focuses its limited resources only on those
areas that it suspects might be contaminated with Hg. Sampling
in this manner cannot produce reliable regional, statewide, or
resource-type estimates of condition [30,31]. However, the sta-
tistically based sampling design of our study permitted us to
Fig. 6. Cumulative distribution function (CDF) of total fish tissue
mercury (␮g/g) for large piscivorous fish in the eastern aggregate and
western aggregate ecoregions of Oregon (USA) relative to their
proportion of occurrence in the total stream/river lengths of each
ecoregion.

Level and extent of mercury in Oregon
Environ. Toxicol. Chem. 21, 2002
2163
sample only 168 sites to represent approximately 60,000 km
of Oregon streams/rivers. Thus, we can infer the sample results
for various fish taxa to all the assessed stream population rep-
resented by the samples (Figs. 4 to 6). With sufficient sample
size, we can further break out our analyses into important
components such as we have done for ecoregion, trophic guild,
and fish taxon (Table 2). Thus, proper sample design can use
limited resources to develop various kinds of environmental
resource condition estimates with known precision and ac-
curacy.
As stated previously, based on least-squares means analysis,
Hg concentrations in big piscivores in the western aggregate
ecoregion were about 35% higher than in the eastern aggregate
ecoregion (Table 2 and Fig. 6). Big invertivores in the western
aggregate ecoregion also contained higher concentrations
communication, U.S Geological Survey, Columbia, MO,
USA). Thus, assessment of environmental risk to ospreys
should be based on Hg concentrations in their preferred prey
(suckers), not on the basis of worst-case scenarios such as
northern pikeminnow.
The mean concentrations of Hg (
(large invertivore 0.055, SD
0.284, SD 0.175; small fish
␮
g/g) in whole-fish tissue
0.047; large piscivore
0.031, SD 0.029) through-
ϭ
ϭ
ϭ
ϭ
ϭ
ϭ
out Oregon are substantially lower than those reported for other
regional surveys [11,12,34–36]. At least two major reasons
for this exist. First, most other regional surveys are not similar
to the one employed in this study, that is, most target sampling
sites to where contamination is known or suspected or to larger
rivers [34–36]. This approach skews results toward high Hg
concentrations in fish tissue and, in general, is not represen-
tative of regional conditions [31,32]. Second, our survey fo-
cused on lotic systems of all sizes, not lakes or rivers. Lakes
tend to act as Hg sinks relative to lotic systems, and larger
lotic systems contain larger piscivores (Fig. 2) that contain
higher concentrations of Hg. Thus, results of our Oregon sur-
vey cannot be related directly to recently reported national
contaminant studies [34–36]. Studies [34–36] of large river
sites are not representative of regional or national lotic sys-
tems, most of which are small streams. In addition, in general,
Hg concentrations in large rivers are skewed toward higher
levels by higher ambient concentrations and larger fish. The
most recent study [34] has changed the reporting focus from
whole-fish contaminant levels to fillets, making direct com-
parison with the results of our whole-fish survey inappropriate.
Probability-based sampling provides the full range of con-
ditions existing across a region. Thus, the mean result will
represent an objective regional baseline condition, not a mean
biased toward contaminated sites. Probability sampling also
permits the sample results to be inferred to the entire popu-
lation of interest, with known confidence (e.g., all streams in
the state depicted on 1:100,000-scale maps). No scientifically
credible means exists of extrapolating handpicked, site-specific
data across a region of apparent inclusion [30,31]. The prob-
ability-based survey approach does not provide all the infor-
mation one might want to know about a site proposed for
remediation. However, CDF curves (Figs. 5 and 6) developed
from such survey data allow one to identify the position and
thus to rank any site in the region within the overall condition
of that region. In addition, CDF curves permit the drawing of
consumption thresholds at any point on the curve, making them
extremely useful for a variety of purposes. Whittier et al. [37]
provides a complete explanation concerning the construction
and use of CDFs in ecological resource surveys. In summary,
probability survey results such as ours provide an assessment
of current status, the ability to identify multiple consumption
thresholds, and a tool for assessing regional change over time.
(
about 17%) of Hg (Fig. 5). As a group, small fish contained
quantitatively much lower Hg concentrations than big ones,
but they also showed the same pattern of higher Hg concen-
trations (about 40%) in the western aggregate ecoregion. A
possible reason for the consistently higher Hg concentrations
in western fish groupings might be related to atmospheric
transport of Asian contaminants and rapid rain-out effect as
the air mass reaches the Oregon coast [32,33].
Relationships between fish tissue Hg and other environ-
mental stressors (e.g., water chemistry, land cover in water-
shed) were tested using Spearman rank correlation coefficients.
In small fish, Hg was not strongly related to any measured
water chemistry variables (acid-neutralizing capacity [ANC],
pH, nitrate, sulfate, chloride, or dissolved organic carbon
[
DOC]) or watershed land cover variable (percentage forest,
2
percentage agriculture, or percentage urban). All the
0.07 ( 119). Relationships between big-fish Hg and
stressors were somewhat higher, with the strongest relationship
r s were
Ͻ
n ϭ
2
being Hg with chloride (
(
cant (
r
ϭ
0.16) and Hg with % urban land
0.10). While these associations are statistically signifi-
p ϭ Ͻ0.001) at a sample size of 137, they explain at
2
r ϭ
most only 16% of the variance in big-fish-tissue Hg concen-
trations. Relationships between Hg and high-DOC/low-pH wa-
ters have been reported in the literature [34]. In our data,
relationships between Hg and DOC, pH, and percentage wet-
2
lands all had an
r
of
Ͻ
0.09, most likely due to the fact that
these variables had a fairly restricted range in our study. Except
for one high-DOC site (25 mg/L), all the DOC values were
Ͻ
4.4 mg/L. Similarly, the lowest pH in any of our study
streams was 6.2, and wetlands ranged only from 0 to 0.6%.
Since our study is a probability study of all Oregon streams
and rivers, low-pH/high-DOC streams/rivers must be very rare
in Oregon, and thus we saw little relationship between pH/
DOC and Hg.
Our results confirm the findings of most previous investi-
gators that piscivorous fish contain considerably higher con-
centrations of Hg than do invertivores and that large pisciv-
orous fish tend to have higher Hg concentrations in their tissue
than do smaller ones [7,12]. Therefore, it appears that unless
a specific interest exists in small fish and small-fish consumers,
surveys of Hg in fish tissue would be best served by focusing
on the worst-case scenario: medium to large piscivorous fish.
However, the significance of small fish and small-fish con-
sumers should not be minimized because of the huge number
of stream kilometers that support them. In addition, it should
be noted that risks to wildlife consumers of Hg-contaminated
fish should not be based on Hg concentrations in all fish. For
example, ospreys eat primarily suckers (C.J. Schmitt, personal
Acknowledgement—The U.S. Environmental Protection Agency fund-
ed this research as part of the Environmental Monitoring and As-
sessment Program supported by Cooperative Agreement CR-824682-
0
1 between the U.S. EPA and Oregon State University and by contract
68-C6-0005 with Dynamac Corporation. This paper has been sub-
jected to the U.S. EPA’s peer and administrative review and approved
for publication. We thank several people who assisted with the prep-
aration of this paper: M. Hails-Avery, K. Baxter, P. Cook, G. Foster,
R. Hjort, J. Lazorchak, A. Nebeker, A. Olsen, C. Schmitt, T. Short,
S. Urquhart, R. Yeardley, and three anonymous reviewers whose com-
ments helped considerably in the final preparation.

2
164
Environ. Toxicol. Chem. 21, 2002
S.A. Peterson et al.
REFERENCES
mercury in water by cold vapor atomic absorption spectrometry.
Method 245.1, Rev 3.0. Technical Report. Cincinnati, OH.
9. Perkin-Elmer. 1998. Winlab Instrument Control Software and
Utilities, Ver 2.61. Norwalk, CT, USA.
0. Omernik JM. 1995. Ecoregions: A framework for managing eco-
systems. The George Wright Forum 12:35–51.
1
2
. Roelke ME, Schultz DP, Facemire CF, Sundlof SF, Royals HE.
991. Mercury contamination in Florida panthers. Florida Panther
Interagency Committee, Gainesville, FL, USA.
. Fujiki M, Hirota R, Yamaguchi S. 1977. The mechanism of meth-
ylmercury accumulation in fish. In Peterson SA, Randolph KK,
eds, Second U.S. Japan Experts’ Meeting on the Management
of Bottom Sediments Containing Toxic Substances. Tokyo, Ja-
pan. EPA-600/3-77-083. U.S. Environmental Protection Agency,
Corvallis, OR, pp 89–95.
1
2
2
2
1
1. SAS Institute. 1989. SAS௡/STAT User’s Guide, Ver 6, 4th ed.
Cary, NC, USA.
2. Wolfe M, Norman D. 1998. Effects of waterborne mercury on
terrestrial wildlife at Clear Lake: Evaluation and testing of a
predictive model. Environ Toxicol Chem 17:214–227.
3. Carpenter H. 1997. Mercury in the midwest: State public health
agency perspective. In Mercury in the Midwest: Current Status
and Future Directions. EPA/905/R-98/003. Research Report.
U.S. Environmental Protection Agency, Chicago, IL.
3
4
. Phillips GR, Buhler DR. 1978. The relative contributions of meth-
2
2
ylmercury from food or water to rainbow trout (Salmo gairdneri
in a controlled laboratory environment. Trans Am Fish Soc 107:
53–861.
. Grieb TM, Driscoll CT, Gloss SP, Schofield CL, Bowie GL, Por-
cella DB. 1990. Factors affecting mercury accumulation in fish
in the Upper Michigan Peninsula. Environ Toxicol Chem 9:919–
)
8
4. Nadis S, Kondro W, Malakoff D, Kaiser J. 1998. Science scope—
Researchers seek consensus at mercury summit. Science 281:
1
779.
5. Kaiser J. 2000. Mercury report backs strict rules. Science 289:
71–372.
9
30.
2
2
5
6
. Bloom NS. 1992. On the chemical form of mercury in edible fish
and marine invertebrate tissue. Can J Fish Aquat Sci 49:1010–
3
6. U.S. Environmental Protection Agency. 2001. Water quality cri-
terion for protection of human health: Methylmercury. EPA/823/
R-01/001. Technical Report. Washington, DC.
7. Foster EP, Drake DL, DiDomenico G. 2000. Seasonal changes
and tissue distribution of mercury in largemouth bass (Microp-
terus salmoides) from Dorena Reservoir, Oregon. Arch Environ
Contam Toxicol 38:78–82.
8. Wiener JG. 1997. Mercury in fish and aquatic food webs. In
Mercury in the Midwest: Current Status and Future Directions.
EPA/905/R-98/003. Research Report. U.S. Environmental Pro-
tection Agency, Chicago, IL.
29. Goldstein RM, Brigham ME, Stauffer JC. 1996. Comparison of
mercury concentrations in liver, muscle, whole bodies, and com-
posites of fish from the Red River of the North. Can J Fish Aquat
Sci 53:244–252.
1
017.
. Haines TM, Komov VT, Jagoe CH. 1994. Mercury concentrations
in perch (Perca fluviatilis) as influenced by lacustrine physical
and chemical factors in two regions of Russia. In Watras CJ,
Huckabee JW, eds, Mercury Pollution: Integration and Synthesis.
Lewis, Boca Raton, FL, USA, pp 397–407.
. Stafford CP, Haines TA. 1997. Mercury concentration in Maine
sport fishes. Trans Am Fish Soc 126:144–152.
. Rada RG, Wiener JG, Winfrey MR, Powell DE. 1989. Recent
increases in atmospheric deposition of mercury to north-central
Wisconsin lakes inferred from sediment analyses. Arch Environ
Contam Toxicol 18:175–181.
. Porcella D. 1997. Historic mercury deposition and biotic accu-
mulation. Mercury in the Midwest: Current Status and Future
Directions. EPA-905-R-98-003. Chicago, IL, USA.
2
2
7
8
9
30. Peterson SA, Larsen DP, Paulsen SG, Urquhart NS. 1998. Re-
gional lake trophic patterns in the northeastern United States:
Three approaches. Environ Manag 22:789–801.
1
1
0. U.S. Geological Survey. 1989. Digital line graphs from 1:
00,000-scale maps; Users Guide 2. Reston, VA.
1
1. Herlihy AT, Larsen DP, Paulsen SG, Urquhart NS, Rosenbaum
BJ. 2000. Designing a spatially balanced, randomized site selec-
tion process for regional stream surveys: The EMAP mid-Atlantic
pilot study. Environ Monit Assess 63:95–113.
2. Yeardley RB Jr, Lazorchak JM, Paulsen SG. 1998. Elemental fish
tissue contamination in northeastern U.S. lakes: Evaluation of an
approach to regional assessment. Environ Toxicol Chem 17:
31. Peterson SA, Urquhart NS, Welch EB. 1999. Sample represen-
tativeness: A must for reliable regional lake condition estimates.
Environ Sci Technol 33:1559–1565.
3
2. Jaffe DA, Anderson T, Covert D, Kotchenruther R, Trost B, Dan-
ielson J, Simpson W, Berntsen T, Karlsdottir S, Blake D, Harris
J, Carmichael G, Itsushi U. 1999. Transport of Asian air pollution
to North America. Geophysical Research Letters 26:711–714.
3. Berntsen TK, Karlsdottir S. 1999. Influence of Asian emissions
on the composition of air reaching the North Western United
States. Geophysical Research Letters 26:2171–2174.
4. Brumbaugh WG, Krabbenhoft DP, Helsel DR, Wiener JG, Echols
KR. 2001. A national pilot study of mercury contamination of
aquatic ecosystems along multiple gradients: Bioaccumulation in
fish. Biological Science Report-2001-0009. U.S. Geological Sur-
vey, Columbia, MO.
1
1
3
3
1
875–1884.
3. U.S. Environmental Protection Agency. 1993. Great Lakes water
quality initiative criteria documents for the protection of wildlife
(
9
proposed)—DDT, mercury, 2,3,7,8-TCDD, PCBs. EPA/822/R-
3/007. Technical Report. Washington, DC.
1
1
1
4. U.S. Environmental Protection Agency. 1998. Field operations
and methods for measuring the ecological condition of wadeable
streams. EPA/620/R-94/004F. Technical Report. Washington, DC.
5. U.S. Environmental Protection Agency. 1999. Mercury update:
Impact on fish advisories. EPA-823-F-99-016. Fact Sheet. Wash-
ington, DC.
6. Wiener JG, Spry DJ. 1996. Toxicological significance of mercury
in freshwater fish. In Beyer WN, Heinz GH, Redmon-Norwood
AW, eds, Environmental Contaminants in Wildlife: Interpreting
Tissue Concentrations. Lewis, Boca Raton, FL, USA, pp 297–
3
3
5. Schmitt CJ, Brumbaugh WG. 1990. National contaminant bio-
monitoring program: Concentrations of arsenic, cadmium, copper,
lead, mercury, selenium, and zinc in U.S. freshwater fish, 1976–
1
984. Arch Environ Contam Toxicol 19:731–147.
6. Schmitt CJ, Zajicek JL, May TW, Cowman DF. 1999. Organo-
chlorine residues and elemental contaminants in U.S. freshwater
fish, 1976–1986: National Contaminant Biomonitoring Program.
Rev Environ Contam Toxicol 16:43–104.
3
39.
7. CEM Corporation. 1998. MARS 5, Operation Manual, Rev 3,
/21/98. Matthews, NC, USA, p 118.
8. U.S. Environmental Protection Agency. 1994. Determination of
37. Whittier TR, Paulsen SG, Larsen DP, Peterson SA, Herlihy AT,
Kaufmann PR. 2002. Indicators of ecological stress and their
extent in the population of northeast lakes: A regional-scale as-
sessment. BioScience 52:235–247.
1
1
1