The resulting methane totals were generally slightly lower in countries well covered by the monitoring network, while uncertainty reduction was modestly greater. The method does not directly repair the transport model; instead, it uses radon to diagnose how transport errors affect methane observations and prevents those errors from being mistaken entirely for emission biases. Atmospheric inversions infer surface emissions by comparing measured concentrations with concentrations simulated by a transport model.

The differences, known as model-data mismatches, are conventionally attributed through Bayesian optimization to errors in the assumed fluxes, observational uncertainty and transport-model uncertainty. This can become problematic when the transport model systematically misrepresents atmospheric mixing. If a simulation disperses a tracer too rapidly through the planetary boundary layer, it may predict concentrations that differ from observations for reasons unrelated to the underlying emissions.

An inversion that does not recognize such errors can compensate by adjusting fluxes in the wrong direction. The new analysis targets that problem with radon-222, a radioactive gas produced in soils and released to the atmosphere. Radon has an e-folding lifetime of about 5.5 days, similar to the timescale for ventilating the planetary boundary layer, and it is chemically inert apart from radioactive decay.

Those properties allow radon concentrations to accumulate when the lower atmosphere is stably stratified but prevent long-lived accumulation aloft. The investigators used hourly methane and radon observations from continental towers, coastal stations, mountain sites and one urban site in the British Isles, northern France, the Netherlands, Germany and Switzerland. Afternoon observations were selected for most sites to favor well-mixed conditions; nighttime data were used at two mountain sites to avoid thermally generated slope winds.

Simulations were generated with the Stochastic Time-Inverted Lagrangian Transport model, or STILT, driven by meteorological fields from the European Centre for Medium-Range Weather Forecasts. For each site and hour, 100 numerical particles were released and traced backward for up to 10 days to map the surface areas influencing the observations. Methane priors combined anthropogenic emissions from agriculture and waste management, fossil-fuel use, natural terrestrial sources and ocean sources.

Daily process-based radon fluxes were calculated using high-resolution German soil-moisture data and coarser data elsewhere. The two gases were inverted jointly with a non-diagonal covariance matrix that allowed their model-data mismatches to vary together. The standard analysis used a correlation coefficient of 0.6, the median value measured across the 17 sites.

Transport-model performance varied sharply with atmospheric stability. At Karlsruhe, for example, the standard deviation of the methane mismatch when the modeled boundary-layer height was below 500 metres was 134 percent larger than under higher mixing heights; for radon, it was 144 percent larger. Across all sites, one-third of afternoon observations occurred under stable conditions, with mismatch variability 105 percent greater for methane and 128 percent greater for radon than in well-mixed conditions.

The two tracers tended to diverge from their simulations in the same direction, consistent with shared errors in vertical mixing. At Karlsruhe, the correlation between their mismatches was 0.82 for afternoon data and 0.75 for other hours. The median across all sites was about 0.6, although the weakest association occurred at a site strongly affected by local methane emissions that the coarse transport model could not resolve.

Several tests supported the interpretation that common transport errors, rather than correlated flux errors, drove much of the relationship. Anthropogenic sources accounted for roughly 80 percent of methane emissions across the relevant central European domain, whereas radon comes mainly from soil, so their source errors were expected to be largely independent. Moreover, replacing the STILT radon simulations with simulations from FLEXPART or NAME weakened the methane-radon association: the afternoon coefficient of determination fell from 0.67 with STILT to 0.44 and 0.34, respectively.

The investigators also considered peatlands, where wet soil can suppress radon release while promoting methane production. Such opposing effects would tend to anti-correlate the two flux errors, and peatlands contributed less than 2 percent of methane emissions in the study domain. With realistic uncertainty assumptions, the dual-tracer analysis changed country-total methane estimates by only a few percent.

Germany’s annual mean was nearly identical to the methane-only result because lower estimates in southern Germany offset higher values in northern Germany. In the Netherlands, Belgium, the United Kingdom and Ireland, annual mean methane totals were 1 to 3 percent lower than in the methane-only inversion, while corresponding radon totals were 0 to 8 percent lower. The inversion also reduced model-data mismatch bias by a median of 69 percent for methane and 58 percent for radon.

For Germany, the methane-only result increased the annual prior total by 75 percent and reduced its statistical uncertainty by 85 percent; the radon-only result increased the prior radon total by 45 percent and reduced uncertainty by 88 percent. The dual-tracer method achieved slightly greater methane uncertainty reduction, indicating that radon supplied information not available from methane alone. The importance of radon’s prior flux uncertainty became clearer in sensitivity experiments.

The standard inversion assigned 30 percent uncertainty to the annual total radon flux over the European model domain, corresponding to roughly 60 percent uncertainty for Germany. In an intentionally unrealistic sensitivity scenario, the uncertainty for Germany’s annual radon total was set to 5 percent. Under that condition, the dual- and methane-only methane estimates for Germany diverged by 8 percent.

Tighter constraints made the system less willing to alter radon fluxes to erase its mismatch, effectively turning radon into a stricter diagnostic of transport performance. That diagnostic reduced methane adjustments during periods when the model performed poorly and concentrated them around conditions the model represented more accurately. Differences reached roughly 10 percent only under the unrealistically tight 5 percent scenario.

The method also has boundaries that the results make explicit. It requires colocated methane and radon measurements, a monitoring network far denser than currently exists in many regions, and reliable maps of soil radon emissions. It is expected to work best where transport errors mainly reflect deficient vertical mixing rather than strong local sources that a regional model cannot resolve.

In peatland-dominated areas, methane and radon prior-flux errors may need to be represented explicitly rather than assumed to be independent. An evaluation at two German sites excluded from the inversion showed the trade-off clearly: the methane-only analysis reduced their mean mismatch by 67 percent, whereas the dual-tracer inversion accepted a 3 to 7 percent higher mismatch because it interpreted part of that discrepancy as a shared transport bias. Radon therefore did not make every observation fit better; it changed the attribution of the mismatch.

The study suggests that better radon flux maps and expanded monitoring networks could make methane inversions more reliable, while broader tests with other trace gases would determine how widely the approach can be applied.

Subject of Research: Use of radon observations to represent transport-model error in European methane emission inversions

Article Title: Can radon help to improve methane emission estimates? Results from a dual-tracer inversion

Article References: Maier, F., Rödenbeck, C., Karstens, U., Koch, F.-T., Gachkivskyi, M., Smerald, A., & Gerbig, C. (2026). Can radon help to improve methane emission estimates? Results from a dual-tracer inversion. Atmospheric Chemistry and Physics, 26(20), 14699-14723. https://doi.org/10.5194/acp-26-14699-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-14699-2026

Keywords: methane emissions, radon-222, atmospheric inversion, vertical mixing, transport models, central Europe, model-data mismatch, Bayesian analysis, greenhouse gases, radon monitoring, radon, help

The research team, led by Aloyce Isaya Amasi of the Nelson Mandela African Institution of Science and Technology together with George Herbert Zinga and Shovi Furaeli Sawe of the Tanzania Atomic Energy Commission, sampled five widely used hot springs distributed across five administrative regions: Chemka/Kikuletwa in Kilimanjaro, Lake Manyara in Manyara, Misughaa in Singida, Songwe in Songwe, and Kilambo in Mbeya. These sites span a remarkable geological range, from the volcanically active Eastern Rift to the central cratonic regions of the country, and each spring serves a different mix of bathers, recreational swimmers, and tourists. Water samples were collected directly at the emergence points and analyzed using an AlphaGUARD professional radon monitor equipped with an AquaKIT arrangement, a reference-grade instrument that measures radon activity concentrations in water through a closed-loop degassing and detection cycle. The choice of instrumentation matters: AlphaGUARD systems have been validated against international reference standards, giving the Tanzanian measurements a level of technical credibility that allows direct comparison with datasets from Europe, Asia, and elsewhere in Africa.

The results revealed striking spatial variability that surprised even the researchers. Radon activity concentrations ranged from a low of 0.56 plus or minus 0.39 becquerels per liter at Songwe to a high of 103.34 plus or minus 44.91 becquerels per liter at Misughaa, a difference of nearly two orders of magnitude across the country. Intermediate values were recorded at Chemka, which averaged 4.96 plus or minus 0.76 becquerels per liter, Kilambo at 9.82 plus or minus 7.58 becquerels per liter, and Lake Manyara at 13.84 plus or minus 2.34 becquerels per liter. Statistical analysis confirmed that the differences among sites were highly significant, with a probability value below 0.001, meaning the variation is almost certainly geological in origin rather than a product of measurement noise or sampling chance. Within individual spring systems, particularly Misughaa and Kilambo, concentrations also fluctuated considerably from one sub-location to another, hinting at complex subsurface plumbing beneath each emergence point.

That internal variability is where the study becomes scientifically rich. The authors attribute the heterogeneity to localized hydrogeological controls, including fracture-mediated groundwater flow, groundwater-rock interaction, and radon degassing that occurs before the water reaches the surface. Radon-222 has a half-life of only 3.8 days, so its concentration in any given water sample is a snapshot of recent contact with radium-bearing rock. Water that travels quickly through fractured crystalline basement picks up radon efficiently, because fractures expose fresh rock surfaces and shorten the transit time during which the gas can decay or escape. Conversely, water that lingers in open pools or shallow reservoirs loses radon to the atmosphere through partitioning at the air-water interface, a process whose efficiency depends strongly on temperature and salinity. The exceptionally high values at Misughaa, where individual samples reached up to 179 becquerels per liter, suggest that this spring taps a radium-rich source zone with rapid fracture-dominated ascent, while the low Songwe values indicate either depleted source rock or extensive pre-emergence degassing.

Against international safety benchmarks, the picture is largely reassuring but not uniformly so. The parametric value of 100 becquerels per liter established by Council Directive 2013/51/Euratom applies to water intended for human consumption, and every site mean fell below that threshold except Misughaa, whose site mean marginally exceeded the guideline and whose individual samples climbed well past it. It is important to note that thermal spring water is not typically consumed as drinking water in the way municipal supplies are; the relevant exposure scenario involves incidental ingestion during bathing and, more significantly, inhalation of radon released from warm water into the often-enclosed spaces of bathhouses and spa facilities. Nevertheless, the Misughaa exceedance matters because it identifies a specific location where sustained human contact with elevated radon levels is a realistic and recurring condition, and where targeted monitoring rather than blanket national regulation would be the proportionate response.

To translate concentrations into health consequences, the team calculated annual effective doses using accepted radiological protection models grounded in the recommendations of the International Commission on Radiological Protection and the exposure factor frameworks of the United States Environmental Protection Agency and the United Nations Scientific Committee on the Effects of Atomic Radiation. Effective dose combines the amount of radiation absorbed with sensitivity weightings for different organs and tissues, yielding a single figure in millisieverts that can be compared against international public exposure limits. The calculations showed that estimated annual effective doses from incidental ingestion remained below internationally recommended public exposure limits at all sites, although the higher end of the dose distribution occurred at Misughaa under frequent exposure scenarios, meaning that people who bathe regularly at that spring accumulate doses meaningfully larger than occasional visitors. The dose estimates also depend on assumptions about ingestion volume, exposure frequency, and the fraction of radon that transfers from water to air, and the authors were careful to frame their results as scenario-based estimates rather than individualized dosimetry.

The Tanzanian findings slot into a growing global literature on radon in thermal waters, and the comparisons are instructive. Studies in western Turkey, northern Pakistan, southern Poland, Slovenia, Hungary, South Africa, and India have documented thermal spring radon concentrations spanning similar ranges, with fracture-controlled geology repeatedly emerging as the dominant predictor of elevated values. The South African study of the Montagu thermal spa resort, for instance, measured radon levels and associated effective doses at a tourist destination structurally analogous to Tanzania’s most visited springs, while work in Slovenia and Hungary emphasized that inhalation in enclosed bathing halls can rival or exceed ingestion as a dose pathway. What distinguishes the new study is not the magnitude of its measurements but the fact that, until now, Tanzania’s substantial geothermal resource base had generated no published radon data whatsoever, leaving regulators at the Tanzania Atomic Energy Commission without an empirical foundation for environmental radiation protection decisions in this sector.

The practical implications extend beyond radiation safety into the economics of geothermal tourism. Tanzania has been actively developing its geothermal resources for both energy production and wellness tourism, and hot springs such as Chemka/Kikuletwa have become destinations for international travelers seeking natural bathing experiences. The new baseline dataset allows the country to integrate radiological monitoring into its emerging geothermal regulatory framework from the outset, rather than retrofitting safeguards after problems arise. The authors argue that their findings support both environmental radiation protection and sustainable geothermal resource management, and the data availability statement indicates that the essential measurements are published within the article while raw datasets can be requested from the corresponding author for academic and non-commercial validation purposes. For a country whose radiation protection infrastructure is still maturing, having defensible, instrument-verified numbers is the prerequisite for every subsequent policy step.

What comes next is likely to be a broader national survey. The five springs sampled here represent a first tranche, and Tanzania hosts dozens of additional thermal manifestations along the Western and Eastern Rift branches whose radon signatures remain unmeasured. The strong link between fracture density and radon concentration observed at Misughaa suggests that future work could combine radon mapping with structural geology and hydrogeological modeling to predict which springs warrant priority monitoring. Repeated seasonal sampling would also clarify whether radon levels fluctuate with recharge patterns, water table position, or seismic activity, a relationship documented in earlier studies of thermal springs in seismically active regions. For now, the message for bathers is measured rather than alarming: the overwhelming majority of Tanzanian thermal waters deliver radon doses well within internationally accepted limits, but the Misughaa spring demonstrates that local geology can push specific sites past guideline values, and that the only way to know which waters are safe for frequent use is to keep measuring them. In the quiet chemistry of hot spring water, Tanzania has found both a public health signal and a new window into the fractured plumbing of the East African Rift.

Subject of Research: Radon-222 concentrations in Tanzanian thermal spring waters and the associated radiological doses to users

Article Title: Assessment of radon concentrations in selected Tanzanian thermal waters and associated radiological risks to users: implications for environmental radiation protection

Article References: Amasi, A. I., Zinga, G. H., & Sawe, S. F. (2026). Assessment of radon concentrations in selected Tanzanian thermal waters and associated radiological risks to users: implications for environmental radiation protection. Environmental Geochemistry and Health, 48(14), Article 584. https://doi.org/10.1007/s10653-026-03468-2

Image Credits: AI Generated

DOI: 10.1007/s10653-026-03468-2

Keywords: radon-222, thermal springs, Tanzania, geothermal water, natural radioactivity, annual effective dose, radiation protection, hydrogeology, groundwater, hot springs, radon monitoring, East African Rift