Eagles-Smith et al. 2016 — Fish Hg spatial and temporal patterns across Western US + Canada (USGS broad synthesis)

This USGS-led broad synthesis compiles total mercury (THg) concentrations in 96,310 individual fish from 4,262 unique locations across 15 U.S. states, 3 Canadian provinces, and 2 Canadian territories — covering 206 species sampled 1969-2014. It is the largest freshwater-fish-Hg dataset in the Western North American literature. Across all fish, total THg in muscle ranges from 0.001 to 28.54 µg/g wet weight with a geometric mean of 0.170 ± 0.001 µg/g ww. Crucially:

• 30% of individual fish samples exceed the EPA Fish Tissue Residue Criterion for methylmercury (0.30 µg/g ww), the threshold designed to protect noncommercial-fish-consuming humans
• 17% of site-level mean concentrations exceed the same criterion
• 3.9% of fish samples exceed the FDA action level of 1.0 µg/g ww for commercial seafood
• 34% of whole-body samples exceed 0.20 µg/g ww, an impairment threshold associated with potential effects on fish themselves

The paper develops a relativized THg-estimation framework that accounts for fish-species, fish-size, habitat, and ecoregion variability so that comparisons across watersheds and ecoregions are robust. Riverine habitats are consistently higher than lacustrine; semi-arid and arid regions (Great Basin, Desert Southwest) higher than temperate. Fish THg concentrations are not correlated with sediment THg concentrations at the watershed scale, though they are weakly correlated with sediment MeHg concentrations — indicating that factors influencing MeHg production (microbial methylation activity, sulfate, dissolved organic carbon, redox) are more important than inorganic-Hg loading for predicting fish-MeHg exposure.

Key numbers

Dataset scope:

• 96,310 individual fish records
• 4,262 unique sites
• 206 species (sampled 1969-2014)
• 15 U.S. states: Alaska, Washington, Oregon, Idaho, California, Nevada, Arizona, New Mexico, Colorado, Utah, Wyoming, Montana, North Dakota, South Dakota (the paper summarizes “15 States” while the data-compilation list enumerates 14 by name; the 15th appears in the source databases but is not explicitly named in §2.1)
• 3 Canadian provinces: British Columbia, Alberta, Saskatchewan
• 2 Canadian territories: Yukon, Northwest Territories
• Habitats: lakes, ponds, and reservoirs; rivers and streams; wetlands; canals
• Foraging guilds (5): piscivore, generalist, generalist invertivore, benthivore, planktivore

THg distribution:

• Range: 0.001 to 28.54 µg/g ww (~4.5 orders of magnitude)
• Geometric mean (all fish, all sites): 0.170 ± 0.001 µg/g ww (very narrow SE on n=96K)
• Site-specific geometric mean range: 0.006 to 2.98 µg/g ww (average across sites 0.124 ± 0.002)
• HUC-8 watershed least-squares mean range (relativized for species, length, site, year): 0.011 to 1.45 µg/g ww

Exceedance proportions:

Variability drivers (per the mixed-effects linear models):

• Habitat type (lake vs river vs stream): significant, riverine > lacustrine (lake-to-river difference 21% in piscivores to 61% in benthivores)
• Fish species: highly significant (accounted for 40% of total variance; piscivores have higher THg via biomagnification)
• Fish size: positive correlation within species (THg accumulates with age/size)
• Site within ecoregion: significant (accounted for 32% of total variance — site-level Hg cycling chemistry varies)
• Year: significant temporal trends (variable direction across ecoregions)
• Ecoregion (Level 1, 10 ecoregions): largest single source of regional variation
• Sediment THg concentration: not significantly correlated with fish THg at the watershed scale; sediment MeHg is weakly correlated

Methods

Database compilation: Original THg fish-tissue data assembled from USGS, state, federal, and Canadian monitoring databases. Tissue types standardized to skinless boneless fillet (76.8% of original data) using literature-derived conversion factors. Wet-weight basis (dry→wet conversion using moisture-content data where available, default 76% moisture).

Length standardization: THg concentrations within species standardized to the species’ median length using linear mixed-effects models with fork length as a fixed effect and species × site as random effects (Eagles-Smith and Ackerman 2014 method). For species with sparse data, similar-species groupings used.

Statistical models: Tiered linear mixed-effects models (4 levels). Initial descriptive (site-specific geometric means), then size-standardized comparisons across habitats and foraging guilds, then watershed-scale relativized THg estimates accounting for species/length/habitat. Final tier compared ecoregions. Getis-Ord G* hotspot analysis (ArcGIS v10) identified spatial clusters of high/low Hg.

Detection limits: Analytical detection limits varied considerably across the merged datasets (0.001-0.1 µg/g ww). Only 91/96,310 values (0.09%) were below reported DL; these were included as reported.

Speciation: THg in fish muscle is ≥95% MeHg per established literature (Bloom 1992). The paper treats THg as a proxy for MeHg, consistent with EPA’s Fish Tissue Residue Criterion approach. The paper acknowledges that this proxy is not appropriate when applied to non-muscle tissues (whole body, liver, gut), where the MeHg/THg ratio is lower.

Implications

What the literature documents (this paper specifically):

1. The Western US/Canada freshwater-fish dataset is the largest single Hg-in-fish-tissue evidence base ingested into the wiki to date. The 30% / 17% / 3.9% exceedance fractions against the EPA Fish Tissue Residue Criterion (0.30 µg/g ww MeHg, set by Borum et al. 2001 to protect noncommercial-fish-consuming humans) and the FDA action level (1.0 µg/g ww commercial seafood) are documented in PDF §3.1. The two thresholds serve different purposes per the source: the EPA criterion targets individual-consumption health protection; the FDA action level governs commercial-seafood marketability.

2. MeHg ≈ THg in fish muscle is the paper’s working assumption (PDF “Methods”/Speciation, citing Bloom 1992), where THg in muscle is ≥95% MeHg. The paper explicitly notes the proxy does not hold for whole-body, liver, or visceral tissues where MeHg/THg ratios are lower.

3. Sediment THg is a poor predictor of fish THg at the HUC-8 watershed scale (uncorrelated, p=0.07, N=418 paired HUCs), while sediment MeHg is weakly correlated (p=0.0005, N=216) — per PDF §3.4. The paper interprets this to mean factors governing microbial Hg methylation (sulfate, dissolved organic carbon, redox conditions, microbial community structure) drive fish-Hg more than inorganic-Hg loading does.

4. Ecoregion is the largest single regional source of variation in fish THg. Semi-arid and arid ecoregions (Great Basin, Desert Southwest) show consistently higher fish THg than temperate forests, per PDF §3.4 and the relativized HUC-8 quintile map (Fig. 4).

5. Piscivores carry the highest fish-Hg burdens via biomagnification (e.g., Sauger, Northern Pikeminnow, Walleye, White Bass, Striped Bass, Northern Pike, Lake Trout, Smallmouth Bass, Largemouth Bass, Whiterock Bass — top 10 LS-mean species per PDF §3.2 Fig. 2). Lowest LS-mean concentrations are in Broad Whitefish, Pumpkinseed, Dolly Varden, Mountain Whitefish, Tui Chub, Brook Trout, Coho Salmon, Redside Shiner, Rainbow Trout, and Slimy Sculpin — generally planktivores, invertivores, and generalists.

6. The dataset is explicitly freshwater-only. Per PDF §1 final paragraph: “This assessment did not address marine environments or risk to human and wildlife health” — those assessments are in companion papers in the same Sci Tot Env special issue (Ackerman et al. 2016, Davis et al. 2016, Jackson et al. 2016, Lepak et al. 2016). The Eagles-Smith 2016 findings do not directly support claims about marine fish food chains or commercial baby-food fish, which use predominantly marine species; the synthesis page should treat this as freshwater evidence and route marine-fish-Hg synthesis to the companion-paper sources.

Courses: Excellent case study for spatial epidemiology + mixed-effects modeling. The 96,310-record database + 10-ecoregion Level-1 stratification is a canonical worked example of relativizing biomonitoring data across heterogeneous species/habitat/site/year structure.

App: For the consumer app’s per-meal MeHg risk scoring against freshwater wild-caught fish (sport-caught and subsistence), this dataset establishes that 30% of individual fish sampled across the Western US/Canada exceeded the EPA 0.30 µg/g ww criterion. The app should not extrapolate these proportions to commercial marine-fish products without additional evidence.

Microbiome: Not addressed by the paper. The microbial Hg-methylation drivers cited in PDF §1 (Gilmour et al. 1992, 2013; Hall et al. 2008) are upstream of the dataset’s scope — a microbiome cross-reference belongs on a synthesis page, not here.

Wiki pages updated on ingest

• mercury-total
• mercury-methyl
• fish
• freshwater-fish
• fish-freshwater
• fish-containing-baby-foods
• epa-fish-tissue-residue-criterion-mehg-03 (to be created)
• fda-seafood-mercury-action-level-1ppm (to be created)

Verification notes

2026-05-29 merge-enhance pass (Claude Opus 4.7, manual-fetch ingest cycle), source re-verified against the PDF in raw/Manual Fetch Kimi /. Corrections applied:

• Geometric-mean range corrected from “0.001 to 28.4 µg/g ww” to “0.001 to 28.54 µg/g ww” (PDF §3.1 first paragraph: “individual concentrations ranged from 0.001 to 28.54 μg/g ww”).
• Foraging-guild list corrected from “piscivore, generalist invertivore, omnivore, benthivore, planktivore” to “piscivore, generalist, generalist invertivore, benthivore, planktivore” (PDF §2.2 Statistical Analyses, where “omnivore” appears nowhere and “generalist” is defined as “diet composed of both fish and invertebrate prey”).
• Geography corrected. The prior version mis-attributed Yukon Territories, Northwest Territories, and British Columbia to “15 U.S. states” while listing North Dakota and South Dakota separately under Saskatchewan. PDF §2.1 enumerates: US states (Alaska, Washington, Oregon, Idaho, California, Nevada, Arizona, New Mexico, Colorado, Utah, Wyoming, Montana, North Dakota, South Dakota — 14 named, paper summary cites “15”); Canadian provinces (British Columbia, Alberta, Saskatchewan); Canadian territories (Yukon, Northwest Territories).
• Sediment-correlation nuance added. The prior version reported “fish THg not correlated with sediment THg” but omitted the source’s accompanying finding that fish THg is weakly correlated with sediment MeHg (PDF abstract and §3.4). Both findings are now stated together.
• matrices frontmatter corrected from [whole-food, exposure-modeling] to [freshwater-fish, fish-muscle]. The paper measures THg in fish-muscle tissue of freshwater fish (76.8% skinless boneless fillet, 19.9% whole body converted to fillet equivalents, 3.3% skin-on fillet); it is not an exposure-modeling study.
• products frontmatter expanded from [fish-containing-baby-foods] (broad-context only) to include fish-freshwater (direct evidence — HMTc Cat 6 Row 1) in addition. The Eagles-Smith dataset is the largest direct evidence base for freshwater-fish-Hg in the wiki corpus; routing it solely to fish-containing-baby-foods (where it is broad-format context, since the paper explicitly does not address marine fish or commercial baby-food fish) understated its evidentiary role.
• HUC-8 watershed range (0.011-1.45 µg/g ww least-squares-mean) added to Key numbers; this is the size-and-species-relativized range from the third-tier model (PDF §3.4), a more interpretable cross-watershed comparison than the raw site means.
• Variance attribution added: species accounted for 40% and site for 32% of total variance (PDF §3.4 paragraph beginning “Site effects commonly account for…”).

No content corrections were applied to claims I could not independently verify against the PDF text I read; the values in the exceedance table (30% / 17% / 3.9% / 34% / 20%) reproduce the PDF §3.1 numbers exactly, as do the species-effect descriptions and the habitat-difference percentages.

2026-05-29 fresh-context audit application (Agent subagent verdict REVISE). Findings independently verified and applied where correct:

• Audit subagent flagged Implications §3: the prior version stated “HMTc certification cannot use sediment-Hg measurements as a supplier-screening proxy for fish-product Hg. Supplier audits must measure fish-tissue directly.” Verified against PDF §3.4 — the paper documents the sediment-THg-vs-fish-THg decoupling (p=0.07, N=418) and the weak sediment-MeHg correlation (p=0.0005, N=216) but makes no claim about supplier-audit practice. The “supplier audits must measure fish-tissue directly” claim was a Part 2 firewall slip into HMTc-program design. Corrected: §3 now reports the watershed-scale decoupling as the paper documents it, without prescribing HMTc auditor practice.

• Audit subagent flagged Implications §5: the prior version stated “HMTc Cat 1 fish-containing baby food should preferentially source from lower-trophic-level fish to control MeHg exposure.” Verified against PDF §3.2 — the paper documents species-level biomagnification gradients (top-10 and bottom-10 LS-mean species lists) but makes no sourcing or procurement recommendation. The “should preferentially source from” prescription was a Part 2 firewall slip into HMTc procurement guidance. Corrected: §5 now reports the species LS-mean rankings as the paper documents them, with the actual top-10 and bottom-10 species lists from Fig. 2, without prescribing sourcing strategy.

• Audit subagent flagged Implications §6: the prior version stated “marine fish food chains have higher overall MeHg loads because of methylation in deep-ocean waters and species-specific bioaccumulation.” Verified against PDF §1 — the paper is explicitly freshwater-only and points to four companion papers in the same special issue for the marine-fish, human-health, and wildlife-health assessments. The “methylation in deep-ocean waters” claim was a cross-source synthesis assertion the source does not make. Corrected: §6 now states the paper’s explicit freshwater scope and points to the companion papers as the appropriate evidence base for marine-fish-Hg, rather than asserting marine-fish-MeHg claims this source does not support.

• Audit subagent flagged Implications §1 framing: “FDA 1.0 µg/g action level is for COMMERCIAL seafood and the EPA 0.30 µg/g criterion is the relevant standard for individual-consumption health protection.” Verified against PDF §3.1 — the paper explicitly attributes the EPA criterion’s purpose to protect “the health of humans who eat noncommercial fish” (Borum et al. 2001). The wiki’s framing is faithful to the paper’s own characterization, but the rewrite tightens the language to describe the two thresholds’ distinct purposes per the source, rather than implying one regime is more “relevant” in an absolute sense.

• Audit subagent flagged the “5 orders of magnitude” framing: 28.54/0.001 = 28,540, which is ~4.46 orders of magnitude, not strictly 5. Corrected to “~4.5 orders of magnitude” in the THg-distribution Key numbers entry.

• Audit subagent flagged matrices: [freshwater-fish, fish-muscle] for the controlled-vocabulary check, noting the taxonomy snapshot does not enumerate matrices. Independently checked: fish-muscle is the de facto matrix slug used across 8+ existing fish-tissue source pages in the wiki corpus (grep ^matrices: over wiki/sources/*.md). The snapshot’s omission of a matrices section is a separate gap in the snapshot, not a defect in this page. Recorded as false positive; no change applied. Refresh of docs/gpt-collaboration/taxonomy-snapshot.md to include a Matrices section is a separate task outside the audit scope.

• Audit subagent gave Check 1 (Numerical fidelity) ✅ clean on all 18 transcribed values, Check 3 (Speciation/methods) ✅ clean, Check 4 (Part 12 brand firewall) ✅ clean. The page enters the audited-revised state with all numerical claims verified, all firewall slips corrected, and one documented false positive on the matrices-vocabulary check.

Page history

The five most recent substantive edits to this page. The full version history lives in git; when DOI minting comes online (see schema docs), each entry below will also link to a version-pinned DataCite DOI.