Bomb ¹³⁷Cs in modern honey reveals a regional soil control on pollutant cycling by plants

Mid-20th century atmospheric nuclear weapons tests injected long-lived fission products, notably 137Cs (half-life 30.2 years), into the stratosphere leading to global fallout. Although much attention has focused on immediate and local impacts, the long-term biogeochemical fate of 137Cs in ecosystems far from test sites remains uncertain. Because Cs mimics K in plants, it can be taken up via K transporters and enter food webs. The authors note a lack of recent data on 137Cs in eastern U.S. plants and foods since the late 1980s and pose the question of whether legacy weapons-derived 137Cs continues to be cycled by vegetation and concentrated in honey, and whether regional soil potassium levels control this uptake and transfer.

Prior work documented global fallout and its use as chronological markers in soils and sediments, with assumptions of strong fixation of 137Cs to minerals. Post-Chernobyl studies in Europe detected 137Cs in honey and pollen, generally concluding levels were safe for human consumption, and highlighting honey as an indicator of atmospheric contaminants. However, these studies typically covered small regions with limited geospatial context and mixed sources (Chernobyl and weapons fallout). In the U.S., widespread monitoring focused on milk via the Pasteurized Milk Network (PMN, 1957–2014), providing extensive 137Cs time series but little to no recent published data on 137Cs in eastern U.S. plants or food (including honey) after 1988.

• Samples: 122 discrete raw, unfiltered honey samples, largely from small-scale producers in the eastern U.S.; hive locations identified at U.S. county scale for 110 samples. Additional samples included central U.S. agricultural honeys and one from Cuba.
• Gamma spectrometry: Direct photon counting of 50–200 g per sample using shielded high-resolution intrinsic germanium detectors (Canberra 5030 Broad Energy, 150 cm3; FWHM <1.5 keV at 662 keV). Typical count time ≥2 days (often longer). Spectra collected with Genie 2000 v3.2. Detector efficiency calibrated using certified liquid 137Cs and high-purity KCl standards for each geometry. Typical detection limit ~0.03 Bq kg−1 for a 150 g sample counted 200,000 s.
• Radionuclides measured: 137Cs (662 keV) and 40K; spectra also scanned for 134Cs (605, 796 keV) to assess reactor-derived inputs; no significant 134Cs peaks detected, indicating legacy weapons-source 137Cs.
• Soil data: County-scale 20th-century 137Cs deposition (1953–1972) from models constrained by precipitation and 90Sr data; county-averaged soil K from high-resolution airborne radiometric surveys (U.S. Geological Survey). A separate USDA National Soil Laboratory dataset of NH4Ac-extractable (plant-available) K provided point-based coverage for about half the relevant counties.
• Statistics: For plotting and tests, undetected 137Cs values were assigned 0.015 Bq kg−1 (half the detection limit). Regression analyses used log-normalized transfer factors (honey/soil) tested against soil K; Excel used for regressions; PAST 4.0 for Mann–Whitney tests. Fig. 2 analysis excluded four Florida agricultural honeys.
• Vegetation archives: Native foliage from the Hubbard Brook Experimental Forest archives (1960s onward) and recent collections were analyzed to reconstruct temporal trends of 137Cs in plants.

• Detection prevalence and magnitudes: Detectable 137Cs (≥0.03 Bq kg−1) in 68 of 122 honeys. In the eastern U.S., 36 of 39 honeys from Florida, Georgia, and South Carolina were detectable, versus 12 of 40 from states north of Virginia. Central U.S. agricultural honeys (n=5) were very low (4 undetectable, 1 at the detection limit). One Cuban sample was undetectable.
• Geographic patterns not explained by fallout alone: Despite slightly higher historic fallout in the northeastern U.S., honey 137Cs did not correlate with deposition gradients (r2 < 0.01; p > 0.5). Concentrations spanned nearly three orders of magnitude, with far higher values in the southeastern U.S.; maximum reported up to 19.1 Bq kg−1.
• Soil potassium as primary control: The transfer of 137Cs from soil to honey scaled inversely with county mean soil K (n=106; p = 10−10, power-law fit; uncertainty dominated by within-county K variability). Using NH4Ac-extractable K (n=55), the inverse relationship persisted (p = 0.027). All 26 honeys from counties with soil K <0.17% had detectable 137Cs. Of counties with NH4Ac-extractable K <0.05 cmol(+) kg−1, 13 of 14 honeys were detectable.
• Agricultural effects: Florida honeys from managed orange/pepper groves (common K and N fertilization) averaged 0.2 Bq kg−1 versus 3.4 Bq kg−1 for 18 Florida wildflower honeys; central U.S. agricultural honeys were similarly low, implicating K/N fertilization and NH4+ competition in reducing Cs uptake.
• Comparison with historical milk data: Nationwide PMN milk 137Cs peaked at ~6 Bq kg−1 in late 1963 and fell to <0.6 Bq kg−1 by 1970. Of 122 modern honeys, three exceeded 6 Bq kg−1 and 30 exceeded 0.5 Bq kg−1. The highest Florida honey (2018) exceeded any monthly milk value reported from 1958–2014. Florida vs NYC-area honey differed significantly (Mann–Whitney z=4.70, p=10−5.6); 20/22 Florida honeys detectable vs 7/22 NYC-area.
• Temporal trends in vegetation: Native foliage 137Cs declined from median ~390 Bq kg−1 in the late 1960s to ~4 Bq kg−1 by 2019 (2 orders of magnitude), consistent with radioactive decay and migration below rooting depth. Historical honey in the Southeast likely exceeded 100 Bq kg−1 in the 1960s–1970s.
• Radiation dose context for bees: Estimated total dose rate to bees from honey <0.1 μGy h−1, but 137Cs substantially increased dose relative to background 40K: in one case tripled, in 7 cases more than doubled, and in 41 cases (33%) 137Cs contributed >10% of the honey-derived dose. The 25 highest 137Cs/40K dose ratios were from the southeastern U.S., where soil K is low.

The study demonstrates that legacy weapons-derived 137Cs continues to cycle through vegetation and is biomagnified into honey, with regional variability controlled primarily by soil potassium availability. The lack of correlation with fallout deposition, juxtaposed with a strong inverse relationship with soil K, indicates geologic substrate and climate-driven soil nutrient status govern bioavailability and transfer to plants and pollinators. Agricultural fertilization (K and NH4+) suppresses Cs uptake, explaining lower values in managed cropland honeys. Comparisons with historical milk and vegetation archives reveal large temporal declines in biotic 137Cs, yet modern honey in parts of the Southeast still carries levels that materially increase ionizing dose to bees relative to 40K. Given emerging evidence that insects can be sensitive to low-dose radiocesium ingestion, historical and even some modern levels observed in the Southeast could have imposed ecological effects on pollinators and ecosystem services. The findings underscore that even decades after deposition, soil chemistry modulates radionuclide cycling and exposure in food webs.

Honey serves as a sensitive integrator of regional biogeochemical cycling of legacy 137Cs and reveals a strong first-order control by soil potassium on radionuclide transfer from soils to plants and pollinators in the eastern U.S. Despite nearly two half-lives since peak weapons testing fallout, measurable and sometimes substantial 137Cs persists in honey, particularly in low-K southeastern soils, occasionally doubling or more the ionizing dose relative to natural 40K. The work provides a framework to predict regions with enhanced and long-lasting radionuclide uptake based on soil K and supports using honey as a bioindicator. Future research should refine plant-available K mapping, resolve species-specific uptake differences, assess sublethal effects on pollinators across exposure gradients, and extend approaches to other fission products (e.g., 90Sr, 131I).

• Soil potassium proxies: County-scale airborne radiometric K represents total K, not plant-available K; NH4Ac-extractable K data had limited spatial coverage and point-based sampling, often biased toward agricultural lands.
• Spatial resolution: Hive locations resolved to county scale; within-county soil K variability introduces uncertainty in transfer estimates.
• Sample representativeness: Uneven geographic sampling density with relatively few central U.S. and non-U.S. samples; agricultural vs wildflower sources not uniformly documented.
• Analytical and statistical assumptions: Non-detects assigned half the detection limit may affect transfer factor distributions; regression models assume log-normal behavior; exclusion of certain agricultural samples in specific analyses.
• Biological variability: Plant species differences and varying nectar sources can affect Cs uptake independent of soil K; seasonal and management factors not fully constrained.