Long-term observations of atmospheric radioactivity in the Arctic are a central component of the Arctic Monitoring and Assessment Programme (AMAP) (AMAP, 2002, 2009, 2015). The geographical locations of air filter stations across the AMAP region, including Iceland, the Canadian Arctic, northern Finland, Sweden, and Norway, with the Svalbard archipelago, are mapped in Fig. 1b. To the best of our knowledge, no coordinates are publicly available for nuclear aerosol samplers placed within the Russian Arctic (AMAP, 2009, 2015). Consequently, these stations are not shown on the map.

Air monitoring of radioisotopes is carried out at the Stanisław Siedlecki Polish Polar Station (77°00^′ N, 15°33^′ E, 10 m above sea level), one of the few permanently operating units north of the Arctic Circle (Fig. 1b) (Burakowska et al., 2021). The Polish Polar Station, located 300 m from the shore of Isbjørnhamna Bay in the Hornsund fjord of SW Spitsbergen, Svalbard archipelago (Fig. 1a and c), was established during the International Geophysical Year in 1957. Over decades, the site has evolved into a modern interdisciplinary scientific platform conducting research projects to better understand the functioning of the Arctic environment and the changes it undergoes (Wawrzyniak and Osuch, 2020).

A high-performance air intake system for the AZA-1000 model was launched at Hornsund in 2002 to enable continuous operation at low temperatures (Fig. 1d) (Burakowska et al., 2021). Aerosols are captured on Petrianov FPP-15-1.5 filters – made of chlorinated polyvinyl chloride (CPVC) fibre – that offer a high retention capacity for particles at least 0.3 µm in diameter. The filter efficiency for aerosols with diameters between 0.3 and 1.25 µm at linear air velocities of 0.25–0.4 ms-1 ranges from 96 % to 99 % (Kierepko et al., 2016; Nalichowska et al., 2023). The high airflow (ca. 450 m3h-1) ensures the collection of a representative portion of aerosols in a single air filter sampled over a week from about 50 000–100 000 m3 of pumped air. This research scope extended from 2007–2021, including the 2011–2013 gap in data collection, during which a substantial number of weekly filters were not available.

Since July 1978, systematic, continuous measurements and observations have been conducted at the Hornsund meteorological site (indexed by the international numbering system 01003; https://oscar.wmo.int/surface/, last access: 1 April 2026; in accordance with World Meteorological Organisation (WMO) standards (Wawrzyniak and Osuch, 2020). A detailed description of the measurements and instruments is provided in a collective work edited by Marsz and Styszyńska (2013), while selected information on meteorological variables and sensors is presented in Table S1 in the Supplement.

The Svalbard archipelago experiences the highest air temperatures at northern latitudes, and the observed climate changes are the most pronounced on Earth (IPCC, 2019). An analysis of the climatological dataset spanning 40 years (1978–2018) offers a comprehensive overview of the meteorological conditions in the Hornsund region (Wawrzyniak and Osuch, 2020). The mean annual air temperature reaches -3.7 °C. The coldest month is March, with a mean air temperature of -10.2 °C, and the warmest month is July, with a mean air temperature of 4.6 °C. The West Spitsbergen Current contributes to a relatively moist climate, as evidenced by an annual precipitation amount of 478 mm. However, the interior of Spitsbergen is much drier (at around 200 mm). Winds blowing from the east along the Hornsund fjord are prevailing, and the wind speed regime shows lower average values in summer months (minimum 4.0 ms-1 in June) but higher average values in winter months (maximum 7.1 ms-1 in February). Such variability results from extreme cyclone activity that often occurs during Arctic winters (Rinke et al., 2017). The polar night lasts 104 d (31 October–11 February), while the polar day extends for 117 d (24 April–18 August).

This assessment encompassed daily data on air temperature (°C), sum of precipitation (mm), atmospheric pressure (hPa), relative humidity (%), cloudiness (octa), visibility (marine scale), sunshine duration (h), wind speed (ms-1), and wind direction (°), collected between 2007 and 2021. For most meteorological parameters, daily mean values were calculated from 3 hourly measurements (eight values per day, between 00:00 and 21:00 UTC). Precipitation was measured every 6 h (12:00, 18:00, 00:00, and 06:00 UTC of the next day), whereas the daily sum of total solar radiation from the Campbell–Stokes recorder was obtained at midnight (Wawrzyniak and Osuch, 2020). The wind rose, changes in mean air temperature, and the sum of precipitation, determined daily in Hornsund for 2007–2021, are depicted in Fig. 2a, b, and d.

A series of weekly air filters was supplied from the Polish Polar Station in Hornsund, Svalbard archipelago, to the National Centre for Nuclear Research (NCBJ) in Świerk, Poland. Before measurement, each sample was desiccated using halogen infrared heaters (500 W, 220–230 V) to determine the dry mass of the collected dust. Following this, the suspended particulate matter concentration (SPM, µgm-3) was estimated at weekly intervals (Fig. 2c). Lastly, all filters were compressed individually to a diameter of 4.5 cm and a thickness of ca. 5 mm. Gamma-ray spectra were measured using a Canberra set, comprising HPGe detectors placed inside low-background shielding chambers with 10 cm-thick lead walls. Detector efficiencies range from 35 %–45 %, and energy resolutions are 1.9–2.0 keV for 1.33 MeV photons from 60Co. Spectrometer calibration was performed utilising a certified source with well-known radioisotope activities. Data acquisition was facilitated by the Genie-2000 software (Mirion Technologies). Correction factors were applied to all results to account for decay during sampling, decay from the end of sampling to the start of measurement, and decay during measurement. The self-absorption correction proved negligible due to the samples' low height. The minimum detectable activity concentration for gamma spectrometry (MDCγ, µBqm-3) was calculated according to the Currie law (Currie, 1968). Values of MDCγ reached 0.1–10 µBqm-3, depending on the radionuclide. The present study incorporated weekly results on the volume activities (C, µBqm-3) of 7Be, 137Cs, and 210Pb, arranged into a time series spanning 2007–2021.

The primary phase of the analysis determined the activity concentrations of 238Pu, 239+240Pu, and 241Am by alpha spectrometry. Considering the likely low quantities of artificial alpha emitters in the Hornsund atmosphere over the past 2 decades, it was hypothesised that the weekly aerosol sample would yield 238Pu, 239+240Pu, and 241Am levels below the minimum detectable activity concentrations for alpha spectrometry (MDCα, nBqm-3). A similar conclusion has been drawn by researchers investigating anthropogenic radioisotopes in the ground-level air of European cities during the 21st century (Kierepko et al., 2016; Lujaniene et al., 2012; Nalichowska et al., 2023). Therefore, weekly air filters from Hornsund were combined to represent approximately one-quarter of a year for the 2007–2010 and 2014–2021 intervals. Due to incomplete collection of weekly air filters from 2011–2013, only one biennial sample was prepared for 2011–2012, while two semi-annual samples were obtained in 2013. Typically, the aggregate samples consisted of 9–14 weekly air filters, with corresponding total air volumes ranging from 630 000–996 319 m3.

The radiochemical separation of analytes from matrix components (such as organic matter, aeolian dust, anthropogenic particles, or filter material) followed the general procedure outlined by La Rosa and co-workers at the IAEA Laboratories in Seibersdorf (La Rosa et al., 1992), as well as the method implemented at the Institute of Nuclear Physics, Polish Academy of Sciences (IFJ PAN) in Kraków (Mietelski et al., 2000, 2008). This approach had previously been tested in air filter investigations (Kierepko et al., 2016; Nalichowska et al., 2023). All non-concentrated reagents were prepared by dilution with ultrapure deionised water (Milli-Q^®, 18.2 MΩcm; Merck KGaA). The sample was initially incinerated at 600 °C in a muffle furnace. Next, internal tracers of 242Pu (ca. 0.0041 Bq per sample) and 243Am (ca. 0.0077 Bq per sample) were dosed (SRM 4334j and SRM 4332e, respectively; NIST). Near-complete sample mineralisation was achieved using concentrated acids (HF, HNO3, HCl, and H3BO3), yielding a filtered 1 M HNO3 solution. The plutonium fraction was separated from the 8 M HNO3 feed solution via anion exchange chromatography with AmberChrom™ 1×8 resin (100–200 mesh; Sigma-Aldrich), preceded by adjustment of the Pu(IV) oxidation state. Given thorium's ability to adsorb onto the resin bed, its elution was performed prior to plutonium to prevent interference between plutonium and thorium isotopes during alpha-spectrum acquisition. Pu(IV) was washed out by passing through the column a mixture of 0.1 M HF–0.1 M HCl. The main effluent (8 M HNO3) was expected to contain the americium analyte, significant quantities of lead, possible traces of thorium, rare-earth elements (REE), and other matrix impurities. Experience has shown that purification steps are necessary for effective isolation and measurement of the americium fraction. Removal of lead was achieved by retention on SR resin (100–150 µm particle size; Triskem International) from an 8 M HNO3 solution. Alkali metals, alkaline earth metals, and anionic components were eliminated by subsequent precipitation of calcium oxalate at pH∼2.5 and iron hydroxide(III) at pH∼9. Repurification from any residual traces of thorium was then conducted using TEVA resin (100–150 µm particle size; Triskem International). The Am(III) was separated from the REE by anion-exchange chromatography on AmberChrom™ 1×8 resin (100–200 mesh; Sigma-Aldrich) in a methanol-acid medium. The isolated isotopes of Pu and Am were individually co-precipitated with NdF3 from aqueous solutions and deposited onto Resolve^® filters (0.1 µm pore size; Triskem International).

Separation of the interfering analytes, 237Np and 242Pu, from a few alpha sources proved essential for further assessment. Thus, a simplified version of the procedure by La Rosa et al. (2005) was employed. In the initial stage, the membrane filter was wet-digested with concentrated acids (H3BO3, HCl, HClO4, and HNO3). Next, plutonium and neptunium were adjusted to the required oxidation state (IV) and separated via column chromatography utilising an AmberChrom™ 1×8 resin (100–200 mesh; Sigma-Aldrich). The 8 M HNO3 feed solution was passed through the column, resulting in strong retention of Np(IV) and Pu(IV). Traces of thorium were removed with a 10 M HCl solution, which additionally converted the column from nitrate to chloride form. Pu(III), reduced from Pu(IV), was selectively eluted using 0.1 M NH4I–9 M HCl, whereas Np(IV) was washed out with a mixture of 0.1 M HF–0.1 M HCl. Finally, alpha sources of Pu and Np were individually re-prepared.

For data acquisition, alpha sources were positioned adjacent to passivated implanted planar silicon detectors (PIPS^®, Mirion Technologies) inside the vacuum chambers of the AlphaAnalyst™ 7200 spectrometer (Mirion Technologies). The PIPS^® detector, with a 450 mm2 active area, has an efficiency of approximately 40 % and an energy resolution of 18 keV for 5.486 MeV alpha particles from 241Am. All alpha spectra were processed in OriginPro^® 2024b (OriginLab Corporation) using the Peak Analyser tools for baseline correction, peak detection, integration, or fitting. The latter option was applied to resolve overlapping spectral lines by fitting bigaussian curves. MDCα values – calculated based on the Currie law (Currie, 1968) – typically varied from 0.01–0.1 nBqm-3 for 238,239+240Pu and 241Am. The mean recoveries of Pu and Am fractions after chemical processing equalled 80.7±2.5% and 43.5±2.2%, respectively. The obtained dataset represented the activity concentrations (C, nBqm-3) of 238,239+240Pu and 241Am, largely determined in quarters between 2007 and 2021.

For each radioisotope except 237Np, basic descriptive statistics were computed according to the time intervals at which the radioisotopes were measured. These included the minimum (min), 25th percentile (Q1), median, 75th percentile (Q3), maximum (max), interquartile range (IQR), mean, and standard deviation (SD). To spot outliers in the dataset, Tukey's interquartile range method with a multiplier k=3 was employed (Hoaglin, 2003). Meteorological indicators, the SPM factor, and gamma emitters served as supporting variables to facilitate a comprehensive evaluation of the annual dynamics of artificial actinides in surface air. Consequently, the subsequent stages of statistical analysis involved converting daily and weekly data into quarterly formats. Data aggregation was performed using the following mathematical operations: a.

arithmetic mean for daily air temperature, atmospheric pressure, visibility, relative humidity, and cloudiness,

For tracking aerosol fate dispersion, three-dimensional air-mass trajectories were simulated using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model developed by the National Oceanic and Atmospheric Administration (NOAA, U.S. Department of Commerce). The system has various applications in the domain of propagation and dispersion modelling of forest smoke, hazardous materials, chemicals, and radioactive substances, whether in gaseous or particulate form, including radionuclide decay (Stein et al., 2015). Backward simulations were performed for a final location at 77°00^′ N, 15°33^′ E (Hornsund), whereas a forward trajectory was generated from an initial location at 72°55^′ N, 54°01^′ E (Novaya Zemlya). All trajectories were reconstructed using archived meteorological data from the Climate Data Centre (CDC) and the Global Data Assimilation System (GDAS) repositories, both accessed via the NOAA Air Resources Laboratory (ARL) server.

This study examined radioisotopes divided into two categories, here referred to as “target” and “background”. The first group consisted of subject analytes (238Pu, 239+240Pu, 241Am), whereas the second group, mostly reported by Burakowska et al. (2021), provided supplementary information (7Be, 137Cs, 210Pb). Table 1 summarises the descriptive statistics for both groups. As expected, naturally occurring radionuclides (7Be, 210Pb) in the lower atmosphere of Hornsund reached higher activity concentrations than the artificial ones (137Cs, 238Pu, 239+240Pu, 241Am). The mean and median values exhibited significant dispersion for each radionuclide, reflected by elevated standard deviations and interquartile ranges (Table 1). Estimating upper limits for outliers (>Q3+3×IQR) was a pivotal phase in the assessment. All outlier points were highlighted in the time-series graphs (Figs. 3 and 4) and treated as potential radiological episodes. The outcomes were further displayed as combined violin and scatter plots, demonstrating non-normal distributions for 7Be, 137Cs, 210Pb, 238Pu, 239+240Pu and 241Am (Fig. 5).

The chemical protocol applied in this study was founded on the premise that neptunium occurs at levels below the detection limits of alpha spectrometry for the majority of environmental samples. According to Kelley et al. (1999), deposited activity of 237Np (T1/2=2.144×106 years) is approximately three orders of magnitude lower than that of 239Pu for global fallout. Analyses of three composite samples from Hornsund unexpectedly revealed a distinct 237Np peak that partially overlapped with the 242Pu peak, as shown in Fig. S3. Along with plutonium elution, neptunium is removed from the resin bed; however, no issues are usually experienced during subsequent measurements. This research identified 237Np signals in the 2nd half of 2013, the 1st quarter of 2014, and the 4th quarter of 2018. Overlapping spectral lines were resolved by fitting bigaussian curves, allowing for count integration (Fig. S3). Unfortunately, calculating the activity concentration of 237Np was not feasible because no neptunium tracer was used during the radiochemical procedure. As exemplified in Fig. S4, Pu+Np analytes were successfully separated by an approach that definitively confirmed the presence of 237Np in suspected samples. The 237Np signal in the 4th quarter of 2018 coincided with a 238Pu outlier.

Detection of 237Np was not reported by other researchers in recent aerosol investigations that employed similar radiochemical procedures without separating plutonium and neptunium fractions (Lujaniene et al., 2012; Nalichowska et al., 2023). This prompted us to conduct thorough analyses of the blanks, checking not only the reagents used but also the filter material and equipment in direct contact with the samples examined. The first type of blank comprised only procedural reagents (called “blank”), while the second was a simulated composite sample consisting of 12 cleanly pressed Petrianov filters (called “blank filter”). It was concluded that the presence of radioisotopes in aerosol samples from Hornsund was not due to laboratory contamination or contamination of the spectrometric unit (Table S2). Further quality control was performed on the reference materials IAEA-447 and IAEA-385 to determine the activity concentrations of the target radionuclides 238Pu, 239+240Pu, and 241Am. The calculated values aligned with the certified values (Table S3).

Spearman's correlation coefficient was selected to minimise the influence of outliers and to account for the non-normal distribution of radioactivity data in Hornsund air (Fig. 5) (Mukaka, 2012; Schober and Schwarte, 2018). The correlation outcomes are summarised in Tables 2–4. The input database used to compute the correlation matrix is provided in Tables S4 and S5.

Mutual correlations were at least moderate or strong for 7Be and 210Pb against air temperature, sum of precipitation, relative humidity, cloudiness, E and NE winds; the latter only for 210Pb (Table 2). Negative monotonic relationships appeared for 7Be and 210Pb versus air temperature, sum of precipitation, relative humidity, and cloudiness, whereas positive monotonic relationships occurred for 7Be and 210Pb with E wind as well as between 210Pb and NE wind. Moreover, 7Be and 210Pb were not affected by the SPM factor (Table 3). These findings corroborated the conclusion concerning the atmospheric transportation and behaviour of 7Be and 210Pb, which were carried into Hornsund by prevailing easterly winds and scavenged by precipitation. Interestingly, quarterly data provided sufficient time resolution to identify correlations between climatic factors and natural radioisotopes. The case of 137Cs seems highly questionable at this point, given that the large number of weekly filters have not shown 137Cs levels above the detection limit since 2013 (Table S5). Artificial radioisotopes exhibited no statistically significant response to climatic conditions, except for 239+240Pu versus sunshine duration (Table 2). Further evaluation revealed a positive but weak correlation between 239+240Pu and the SPM parameter (Table 3). Dust concentrations usually peaked in the 4th quarter (Fig. 2c, Table S5), coinciding with enhanced cyclone activity during winter months (Wawrzyniak and Osuch, 2020). Thus, the SPM contribution evidenced a partly local origin for 239+240Pu, likely due to dust redistribution in winter and minimal soil resuspension in summer. This also explained the weak negative correlation of 239+240Pu with sunshine duration (239+240Pu signals were higher during polar night months than during polar day months).

Strong positive correlations between 7Be and 210Pb were observed annually and in specific quarters (1st, 2nd, and 4th) (Table 4). Strong correlations were also noted for 239+240Pu with 7Be, 239+240Pu with 210Pb, and 137Cs with 7Be in the 1st quarter, as well as between 239+240Pu and 241Am in the 3rd quarter (Table 4). All of these relationships were positively monotonic, except for 137Cs versus 7Be. It seems that horizontal air-mass movement significantly contributed to transporting 239+240Pu alongside 7Be and 210Pb during the 1st quarter. The linkage between 241Am and 239+240Pu in the 3rd quarter might reflect a similar supply pattern, although this was difficult to define. Moreover, a weak positive correlation of 239+240Pu against 238Pu was found annually (Table 4), suggesting that a small part of 238Pu was likely governed by the natural mechanisms identified for 239+240Pu. Consistent seasonal fluctuations in the quarterly median values were evident pairwise between 7Be and 210Pb, as well as for 239+240Pu and 238Pu (Fig. 6). No clear seasonality was noticed for 137Cs and 241Am.

The activity ratios (IR) of artificial actinides were investigated to infer their provenance (Table S7 and Fig. 7), taking into account reference signatures for specific nuclear events provided in Table S6. These ratios are known to vary with reactor type, neutron flux and energy, nuclear fuel burn-up duration, or, for fallout from nuclear detonations, bomb type and yield (Oughton et al., 2001). Typically, 238Pu/239+240Pu and 241Am/239+240Pu ratios in nuclear reactors (NPP) and fuel reprocessing plants (NRP) are about an order of magnitude higher than those for global fallout (GF) or isotopic production grade (PPP) (Table S6).

The 238Pu/239+240Pu values ranged from 0.021–8.3, whereas the 241Am/239+240Pu values varied from 1.5–460 in Hornsund air (Table S7). Noteworthy, both maxima were observed in the 1st quarter of 2019. The 238Pu/239+240Pu ratios exhibited considerable variability throughout 2007–2021, mainly fluctuating within the reference limits of GF (+ SNAP 9A) and Chernobyl fallout (Fig. 7). These results indicated a mixed plutonium origin, influenced by both GF (+ SNAP 9A) and releases from nuclear power plants or nuclear reprocessing plants. Moreover, only one point corresponded to the GF (+ SNAP 9A) level, while several points from 2009–2013 and 2017–2020 displayed 238Pu/239+240Pu ratios significantly above the known reference signatures (Fig. 7). The latter issue persisted in the 241Am/239+240Pu data, demonstrating an unprecedented enrichment of 241Am relative to established activity ratios for prior nuclear events (Fig. 7 and Table S7).

Radioactive contamination in the Arctic has occurred at two different scales (AMAP, 2002, 2009): a.

widespread contamination associated with global nuclear weapons testing (88 of which took place at the Arctic test site of Novaya Zemlya), releases from Sellafield and Cap de la Hague, and failures at Chernobyl or Fukushima,

A recent study of air filter samples suggested a mixed origin for plutonium isotopes in Poland and revealed the highest 238Pu/239+240Pu ratios of 33.7±3.6 in the 1st quarter of 2000 for Białystok and 1.29±0.25 in the 1st quarter of 2002 for Kraków (Kierepko et al., 2016; Nalichowska et al., 2023). The 238Pu/239+240Pu values determined in aerosol samples collected in Vilnius over 2005–2006 varied from 0.028–0.042, whereas 241Am/239+240Pu values fell between 0.19 and 0.65, both aligning with the GF (+ SNAP 9A) references. However, a level as high as 1.2±0.1 for 238Pu/239+240Pu was detected in Lithuanian air during March–April 2011 (Lujaniene et al., 2012). Results from surface air in Hornsund between 2007 and 2021 showed deviations similar to those of 238Pu/239+240Pu in Białystok, Kraków, and Vilnius, but not for 241Am/239+240Pu in Vilnius.

The provenance of 238Pu enrichment in the European atmosphere has been widely debated. The following rationale, suggested by researchers, may also offer further insight into Hornsund's situation. Potential sources of plutonium in recently measured aerosol samples are hypothesised to include fuel particles from the Chernobyl or Fukushima accidents and emissions from nuclear facilities (i.e., nuclear power plants, nuclear research centres, or nuclear fuel reprocessing plants) (Wershofen and Arnold, 1998). Hirose and Povinec (2015) proposed that global desert dust events, biomass burning, and sea spray are additional likely delivery patterns. Saharan and Asian dust transport has been classified as the most significant mechanism for redistributing aerosols across mid-latitudes. However, Saharan dust displays isotope ratios typical of global fallout (Chamizo et al., 2010). The Chernobyl ecosystem was affected by several major wildfires, notably in 1992, 1999, 2000, 2002–2004, 2006, 2010, 2015, 2016, 2018, and 2020 (Masson et al., 2021). The dispersion of fly ash particles from biomass combustion has been shown to extend over vast distances, both horizontally (up to several thousand kilometres) and vertically (several kilometres), capable of penetrating the tropopause and reaching the lower stratosphere (Hirose and Povinec, 2015). The contribution of plutonium from sea salt to atmospheric plutonium deposition has been found negligible compared to soil's (less than 0.3 %) (Hirose et al., 2003). The maximum calculated input of sea-spray plutonium to the atmosphere did not exceed 0.006 nBqm-3, which is 2–3 orders of magnitude less than that of other plutonium “feeder” mechanisms.

The maximum activity concentration of 239+240Pu in Hornsund coincided with the highest 239+240Pu level in Krakow, both noted within the 3rd quarter of 2015. This may indicate a common source behind the 239+240Pu peaks; however, the corresponding 238Pu/239+240Pu signatures differed. Plutonium in Krakow could not be attributed to spent nuclear fuel, while the values of 238Pu/239+240Pu and 241Am/239+240Pu obtained in Hornsund (0.426±0.017 and 2.46±0.12, respectively) were consistent with Chernobyl reference ratios (Table S7 and Fig. 7). As reported by Evangeliou et al. (2016), about 10.9 TBq of 137Cs, 1.5 TBq of 90Sr, 7.8 GBq of 238Pu, 6.3 GBq of 239Pu, 9.4 GBq of 240Pu, and 29.7 GBq of 241Am were released from two major fires in the Chernobyl Exclusion Zone (CEZ) in April and August 2015. The labile elements escaped more easily from the CEZ, whereas the larger refractory particles were removed more efficiently from the atmosphere, primarily affecting the CEZ and its surroundings. During the summer fire, about 75 % of the labile and 59 % of the refractory radionuclides were exported from the CEZ, with the majority depositing in Belarus and Russia. Although the evolution and fate of the plutonium and americium plume modelled by Evangeliou et al. (2016) suggest no spread to Spitsbergen, scheduled measurements of 240Pu/239Pu mass ratios should ultimately verify the Chernobyl wildfires' impact on the radioactive signals recorded at Hornsund in 2015. The elevated 239+240Pu activity concentration in Krakow was likely due to intense resuspension associated with a high number of local fires in Poland that could remobilise the global fallout plutonium present in forest litter (Nalichowska et al., 2023).

The above reasoning fails to explain the remarkable enrichment of 241Am between 2007 and 2021, the frequent inputs of 238Pu during 2009–2013 and 2017–2020, or the single injections of 237Np in 2013, 2014, and 2018, all detected at the Hornsund station (Figs. 7 and S3). Correlation assessment indicated that a significant portion of the 238Pu and 241Am must have been introduced into the atmosphere via non-natural processes (Table 2). Identifying the exclusive nuclear emissions of 238Pu, 241Am, and 237Np is challenging due to the low time resolution of quarterly filters and the scarcity of data on artificial actinides from other stations in the years examined.

238Pu is characterised by sufficient decay heat to power a deep-space satellite or a space probe via a radioisotope thermoelectric generator (RTG). The well-documented source of 238Pu alone is the SNAP 9A high-atmospheric burn-up in 1964 (Krey, 1967). It is doubtful, however, that 238Pu from the SNAP 9A accident has been present in the stratosphere over the last few decades (Hirose and Povinec, 2015). Since the early 2010s, the US government has initiated efforts to restart the production of 238Pu heat-source material for NASA's deep-space missions (Urban-Klaehn et al., 2021; Zillmer et al., 2022). The process involves irradiating a 237Np target with neutrons, converting 237Np into 238Pu. New production relies on reactors at both Oak Ridge National Laboratory (ORNL) and Idaho National Laboratory (INL). These recent nuclear operations have not been associated with reported releases of either 238Pu or 237Np. Radioactive 241Am is commonly used in commercial applications, such as precision measuring devices and smoke detectors. The 241Am activity in various smoke detector models ranges from 7.4–740 kBq, distributed under the class exemption (NUREG-1717, 2001; NUREG/CP-0001, 1978; OECD, 1977). The estimated total number of smoke detectors currently in use for residential purposes in the USA is approximately 100 million, with a total 241Am activity of 12 TBq. The high cost of legal storage for spent devices, coupled with homeowners' lack of awareness, has led to disposal with regular waste or common e-waste without controls. The processing of electrical waste, particularly isotopic smoke detectors, has been postulated to account for the 241Am release in 2021 over central Poland (Długosz-Lisiecka and Isajenko, 2024). Recycling of these devices with general waste, or their combustion at incineration plants, is becoming increasingly popular each year. Importantly, 241Am may also be a generator of 237Np, as the latter is a decay product of 241Am.

Recent nuclear episodes raise new challenges for radioecological research. The presence of 106Ru in European air in autumn 2017 could be linked to an accident at the Mayak nuclear fuel reprocessing plant, involving the separation of large quantities of 106Ru from spent fuel rods. However, alternative explanations include the development of a nuclear engine for powering a small missile or drone (Mietelski and Povinec, 2020). Given that the incident followed an unannounced nuclear operation, the exact scenario for widespread air contamination remains unclear. The Nyonoksa accident in August 2019 (which aligns with official statements about the development of new nuclear-powered missiles in the Russian Federation) may have occurred during preparations to test a nuclear engine driven by either a liquid radioisotope generator 42Ar/42K or a small nuclear reactor (Mietelski and Povinec, 2020). To date, no reports have indicated radioactive leakage into the atmosphere from Nyonoksa. The examples above suggest undeclared nuclear activities. The purpose and types of radioisotopes used are unknown.

A series of transport simulations of nuclear aerosols through the troposphere was performed by the HYSPLIT model, focusing solely on the 1st quarter of 2019. During this period, the highest activity concentration among the examined artificial actinides was found for 241Am. The dispersion of 241AmO2 was investigated over a 72 h timeframe, with aerosols predominantly propagating at altitudes up to 1 km. Trajectory modelling was initiated on 1 January and continued until 31 March, as depicted in Figs. 8 and S5. Noteworthy transport pathways from northern Europe and Asia via Novaya Zemlya were observed between January and February (Fig. 8). Nuclear aerosols were carried at low altitudes within the troposphere (often below 100 m), leading to weak dilution and intense deposition. Forward trajectories obtained in early February 2019 indicated substantial material shifts from the Novaya Zemlya region to the aerosol sampling location (Fig. S5). Simulations conducted in early January, late February, and March displayed additional backward trajectories from northern directions (Figs. 8 and S5), which appeared unlikely to transport nuclear aerosols highly enriched in 241Am or other artificial radioisotopes. Reconstructing aerosol propagation during the remaining period – characterised by lower radioisotope activity concentrations than in the 1st quarter of 2019 – was beyond the scope of this paper but is planned to be continued in upcoming scientific exercises.