© J.J. Ingremeau et al., Published by EDP Sciences, 2025

1. Introduction

Traces of artificial radionuclides are detected almost every year by European monitoring networks, particularly in northern Europe without these releases being officially declared by any authority. So, their origin remains today unknown. The measured concentrations are very low, ranging from 0.1 to 10 μBq m⁻³, which are near the detection limits of the instruments, and pose no health or environmental risks. 15 events have been reported between 2014 and 2024, occurring mainly during spring or summer [2].

An in-depth investigation of the June 2020 detection event has been carried out by IRSN in a previous publication [1]. The assessed area of the potential source location extends from eastern Estonia to western Russian Federation. This location was corroborated by Health Canada [3]. The IRSN analysis [1] concludes that the source of the release is likely a primary resin from a Pressurized Water Reactor (PWR), with fuel cladding failure and dispersion of fissile material in the primary circuit. The level and the nature of the fission product releases would imply a few months of decay. The study [2] focused on two additional minor releases occurring in 2019 and 2022, which are characterised in particular by the presence of ⁴⁶Sc.

The geographical origin of these two events are entirely consistent with that estimated for the 2020 event. The source location would be in the western part of the Russian Federation, in the Leningrad area.

More recently, in February 2023, another release event was detected. The Norwegian Meteorological Institute assessed the possible source location for this event [4]. Due to the limited number of detections (only 3 stations) and a week-long sampling period, combined with a complex meteorological situation, the estimated area is quite large; however, it includes the Leningrad region and eastern Estonia. Moreover, this zone is the closest possible area to the detection zone, and therefore the most likely one. According to the fact that the greater the distance to the release, the higher the source term must be to remain detectable. So, without increasing dramatically the source term magnitude, the most likely source location is again in the same region.

From a review of the 15 comparable detections between 2012 and 2024 (Tab. 1), it can be established that:

• they do not contain any short-lived isotopes such as iodine, but only intermediate-lived isotopes (with half-lives of a few months) and long-lived isotopes (with half-lives of a few years or more).

• They all contain some Activated Corrosion Products (ACP). In particular, ⁶⁰Co is always detected. For example, ⁵⁹Fe is also frequently detected.

• Just under than half of them contain Fission Products (FP). Before May 2023 event, three of them contained some low volatile fission products such as ¹⁰³Ru.

The simultaneous detection of ACP and low volatile fission products provides strong evidence for the hypothesis of a pressurized water reactor primary resin, involving fuel cladding failure and dispersion of fissile material in the primary circuit. All the other detections, without low volatile FP, could also be compatible with a primary resin origin with different cladding sealing states, but could also come from various operational solid wastes from a primary loop, such as filters, gloves, clothing, tools, etc.

However, for these detection events, excluding those in May 2023 and June 2024, four of them contained traces of ⁴⁶Sc. This was not expected because ⁴⁶Sc is neither a fission product nor a ACP of western PWR. In fact, ⁴⁶Sc is mainly used and produced for medical purposes, or a tracer in the oil industry. It can be also produced in a nuclear reactor but only if the internal structures contain significant amounts of natural Scandium (⁴⁵Sc) or Titanium (⁴⁶Ti), which is not usually the case in power reactors, at least in the western ones. Three hypotheses were explored to explain the origin of ⁴⁶Sc in [2]:

• mixing of various solid wastes, some of which containing ⁴⁶Sc.

• PIK reactor hypothesis (a research reactor that could possibly contain natural scandium in its internal structures). However, this hypothesis was practically excluded by the analysis.

• Incineration of some RBMK graphite.

None of these hypotheses were entirely satisfactory, but they could all theoretically explain the detection.

The May 2023 detection event differs from previous ones by its comprehensive nature. It contains not only ACP and low volatile fission products but also actinides, and even ⁴⁶Sc. The possible release location of this event is presented in Section 2. Section 3 focuses on the source term analysis of the May 2023 event. Following the additional investigations prompted by the May 2023 event, a much more satisfactory explanation for the ⁴⁶Sc origin is proposed in Section 4. Finally, another plausible scenario that could explain the release is discussed in Section 5.

2. Release location and magnitude of the May 2023 event

The May 2023 event represents the most comprehensive detection in Fennoscandia over the last decade. 16 artificial isotopes were detected in 10 different locations in Finland, Sweden and Estonia. This corresponds to 79 “isotopic detections”¹. Detections were reported between May 14^th to June 12^nd, the majority of which occurring between May 22^nd and 29^th.

Most of the detections were associated with extremely low airborne concentrations, below 3 μBq m⁻³, close to detection limits of the instruments.

Only 3 isotopes were measured above this concentration level: ¹⁴⁰Ba² and ²³⁴Th at respectively 9 and 3.4 μBq m⁻³, and ¹³⁷Cs with 4 measurements between 3.0 and 4.5 μBq m⁻³. These extremely low levels of airborne concentrations are detectable thanks to the high-volume sampling stations of the Finnish, Sweden and Estonian monitoring networks. The Finnish and Swedish authorities (STUK and FOI) publish their results on their website. The Swedish detections associated to this event are available in the FOI annual report [5].

The most reliable source location was determined using an inverse modelling approach, which combines results of LdX atmospheric transport model [2] and observed airborne concentrations. The measurements integrated in the inverse modelling process include only data from Sweden and Finland between May 14^th to 29^th. The observations under detection limits are also considered. Only the ¹³⁴Cs data were used to estimate the source term, as it was the most abundant of the radionuclides, except for ¹³⁷Cs. However, using ¹³⁷Cs is challenging because the concentration includes residual background radiation resulting from the resuspension of radioactive particles deposited following the Chernobyl accident in 1986. Therefore, it is difficult to estimate the proportion of the concentration actually linked to a recent release.

Due to the lack of information regarding the exact location of the release, a grid encompassing potential release points was created. For computation time reasons, the domain is divided into a set of 420 grid points with 1 ° ×1° spatial resolution. Each grid point is assumed to be a potential source location.

A minimisation procedure was then applied to estimate the magnitude of the source term associated to each potential grid point. It consists of the minimisation of a log-normal cost function that measures the discrepancy between the simulated airborne concentrations using LdX and the observed data.

To mitigate the influence concentrations that are very small, a threshold θ of 1 μBq/m³ is added to the log-normal cost function. Due to the ill-posed nature of the inverse problem, the cost function also includes a regularisation parameter, λ, to ensure the stability of the numerical solution. This parameter must be determined rigorously since its influence may be significant, especially when the number of measurements are small compared to the number components of the source term to be estimated. The L-curve technique [6] is therefore used to determine the optimal value of λ, obtained at λ = 10⁻⁶.

Fig. 1. Percent of the simulated air concentrations that is within a factor of 5 of the observed values. Blue triangles are nuclear plants located in the area of interest: Leningrad NPP, Petersburg Nuclear physics Institute (Gatchina), Olkiluoto NPP, Loviisa NPP and Smolensk NPP.

The potential release zone was then identified by assessing the agreement at each grid point, between the simulated airborne concentrations derived using LdX and the estimated source term, and the observed data. Factor5 (FAC5) indicator which represents the proportion of the simulated airborne concentrations that are within a factor of 5 of the observed data, was used. The approach is described in details in [2, 7].

Figure 1 shows a map of the FAC5 indicator values interpolated at a 1 ° ×1° spatial resolution, representing a possible area of source location. The purple area indicates the most probable location from which the release originated. Atmospheric dispersion simulations assuming a release from this area are roughly consistent with over 80% of the measurements. A release in the red or orange areas could explain between 50% and 80% of the measurements. This region extends from Estonia to the western Russian Federation and broadly aligns with the source locations identified after the detection events in July 2019, June 2020 and May 2022. The Swedish FOI conducted an analysis and determined that the source location was to “the east or northeast of the Gulf of Finland”, which agrees with our own analysis [5].

Assuming a release at the location (x, y = (29.2; 59.2) – marked by the red triangle in Fig. 2) and corresponding to the most likely source location, the dispersion of ¹³⁴Cs plume is plotted in Figure 2, alongside the monitoring network stations. Monte-Carlo sensitivity analysis [7] was performed to reconstruct the source term at this location. The estimated ¹³⁴Cs source term would be between 2 and 4 GBq, primarily released between 18 and 23 May.

Fig. 2. 134Cs plume dispersion assuming a release at the location (29.2; 59.2). Up – May 22nd – Down May 23rd. Blank dots are air concentration measurements below the detection limit of the instruments.

The plume mainly affected northern Europe including Finland, Sweden, Estonia and Latvia, and did not spread to western or southern Europe. Simulations can be used to predict the arrival time of the plume and the order or magnitude of the peak concentration at each station (see figures in the supplementary material).

3. Source term analysis: May 2023 detection event

Table 1 summarizes the artificial radionuclides reported in the Northern Europe between 2012 and May 2023, with measurement data sourced from the STUK website and the informal radioactivity monitoring network (Ro5).

Due to the significant deposits of ¹³⁷Cs in Northern European countries following the Chernobyl accident in 1986, traces of ¹³⁷Cs are consistently measured in the air because of the resuspension of particles from the ground. As a consequence, there is a persistent ¹³⁷Cs background level in the air across most of the European continent, which slightly fluctuates depending on meteorological conditions but remains very low in all cases. In Table 1, detections that correspond to this usual ¹³⁷Cs background are marked with an “X”, while larger detections, clearly indicating a new release of fission products, are marked with an “X”.

Moreover, no ⁹⁵Zr was detected in May 2017 and March 2020, despite the measurement of its decay product, ⁹⁵Nb. This suggests that ⁹⁵Zr was indeed present, but in quantities too low to be detected by the instruments. These instances are marked with a “?” in Table 1.

Returning to the May 2023 event, as mentioned before, ACP have been detected (⁶⁰Co, ⁵⁹Fe, ⁵⁴Mn), together with volatile FP (caesium) and low volatile FP (¹⁰³Ru, ¹⁰⁶Ru, ¹⁴¹Ce, ¹⁴⁴Ce, ¹⁴⁰Ba). As detailed in [1], the release of such significant quantities of low volatile FP is not possible without a fuel cladding failure.

In addition, this event cannot be attributed to an unlikely fuel melt accident, since such an event would result primarily in the release of fission products, not activated corrosion products (ACP). In addition, more short-lived isotopes, such as iodine, would also be expected in such hypothesis but was not detected at any station.

Therefore, the most plausible explanation for the significant presence of low volatile FP, and ACP is a fuel failure with dispersion of fuel in the primary loop during reactor production. In this case, fissions occur directly in the primary circuit, and the resulting low volatile fission products migrate and are trapped in the primary cleaning primary system, mainly in the ion exchange resin, together with the ACP generated in the primary circuit. The presence of actinides (²³⁴Th and ²³⁵U) provides further evidence of fuel dispersion into the primary loop during reactor operation. In such case, a significant proportion of fuel particles outside the cladding ultimately ends up in the cleaning system, either in the filters or in the ion exchange resin. This is the first time that actinides have been detected in such an event, suggesting that the fuel failure may have been quite significant.

²³⁴Th is the decay product of the most abundant isotope in a nuclear fuel, ²³⁸U. While ²³⁸U decays by emitting an α particle with a half-life of 4 billion years, ²³⁴Th is much more radioactive with a much shorter half-life of 24 days. This suggests that ²³⁸U was present in the release but not detected, firstly because the station only detects gamma emissions, and secondly because of its extremely low activity. Assuming that ²³⁴Th and ²³⁸U are in equilibrium, the ²³⁸U activity can be estimated and compared to the ²³⁵U activity. The activity ratio ²³⁵U/²³⁸U allows a simple calculation to evaluate the enrichment of the fuel. Taking into account the uncertainties, the enrichment is estimated to be between 1.2% and 3.9%. This result excludes reactors using natural uranium. It remains not precise enough to distinguish between RBMK and WWER reactor type.

Additional isotopic ratios are available for this event. The ¹³⁴Cs/¹³⁷Cs ratio is between 0.7 and 1.2 (after applying correction for the ¹³⁷Cs background from Chernobyl accident). These values are typical for spent fuel from a pressurised water reactor (PWR) after a few months of decay, depending on the level of enrichment and burn-up. However, such levels are very unlikely to be observed in a RMBK reactor, where typical values at end of cycle are closer to 0.5 and continue to decrease after shutdown [8].

The ¹⁰³Ru/¹⁰⁶Ru ratio, which is between 0.9 and 1.75, provides an estimate of the decay time. Assuming that ruthenium originates from the fission of dispersed fuel in the primary water, the decay time can be estimated to be less than six months³. However, if there has been a prolonged accumulation of ¹⁰⁶Ru in the resin, or if there has been a slow additional ¹⁰⁶Ru release from the fuel, the amount of ¹⁰⁶Ru in the resin would be higher. In this case a shorter decay time would be required to achieve the observed ratio. Therefore, the real decay time is likely to be significantly less than 6 months.

The same analysis can be carried out for the ¹⁴¹Ce/¹⁴⁴Ce ratio, which is between 0,65 and 1. It suggests that the decay time is probably significantly less than 5 months.

The absence of short-lived isotopes, especially iodine, is also an indication that the decay time is at least of 2 months.

In conclusion, consistently with the June 2020 release, the most likely origin of the May 2023 detection event is a spent primary ion exchange resin from PWR reactor, contaminated after a fuel cladding failure and significant dispersion of fuel in the primary loop. The estimated decay time is between 2 and 5 months.

4. The origin of ⁴⁶Sc

In the four detection events that reported the presence of ⁴⁶Sc, as analysed in [2], ⁴⁶Sc was almost always associated with ACP only.

The presence of ⁴⁶Sc with ACP was observed in February 2023, again in parallel with the absence of FP releases.

Finally, the May 2023 event evidence an opposite trend: a significant release containing several low volatile fission products, together with ⁴⁶Sc. ⁴⁶Sc was detected at two locations (Kotka and Narva-Jõesuu), with the low volatile fission products, demonstrating it is associated to the same release. The most recent detection event occurring in May/June 2024, also contains ⁴⁶Sc associated with ACP and low volatile fission products. In fact, ¹⁴¹Ce had already been detected together with ⁴⁶Sc long before at the August 2016 event, but since it was the only FP present, it was not fully considered in the analysis.

This evidence prompts further investigation into the possible ⁴⁶Sc sources and the identification of an alternative scenario.

According to the IAEA TECDOC “Water Chemistry of WWER Nuclear Power Plants” [9], the primary circuit and internal components of WWER reactors are made of the 08X18H10T steel alloy. This alloy is also used for spacer grids and end caps of fuel pins.

The steel alloy 08X18H10T contains titanium, as indicated by the “T” in its name. When exposed to a fast neutron flux, the ⁴⁶Ti(n, p)⁴⁶Sc reaction occurs, resulting in the production of ⁴⁶Sc. In another IAEA TECDOC [10], a Russian publication specifically identifies the production of ⁴⁶Sc, as an isotope inducing a radiological risk, by two different pathways: by the presence of ⁴⁵Sc impurity in the steel (scandium capture); and a second, by the titanium content in the steel.

Therefore, contrary to what was stated in [2], ⁴⁶Sc is also an activated corrosion product of WWER reactors.

Moreover, its systematic association with other ACP, in addition to the three detection events for which its association with fission products was reported, became fully coherent. The other hypotheses (RMBK graphite or the mixing of different waste) are then much less convincing and can probably be discarded.

In addition, a VVER operator confirms that the ⁴⁶Sc activity in the primary coolant is in the order of magnitude of 0.1−1 kBq/kg [11]. This range of activity results in a cumulated activity in the primary resin in the range of tens to hundreds of GBq after one cycle. After two to five months of decay, the remaining activity in the primary resin is still in the order of 10 GBq since the half-life of ⁴⁶Sc is 84 days. Thus, a fraction of the primary resin inventory of a WWER could contain about 1 GBq of ⁴⁶Sc which, according to the results described in Section 2, is sufficient to explain the detections.

This is another clue that reinforces the hypothesis of a WWER reactor as the origin of the release.

5. What happened? A possible scenario for the release

The analysis so far has focused on determining from which type and part of a nuclear installation the releases could have come from, even if no certainty can be achieved. However, one main question remains unanswered: how could the inventory of a primary resin have been released into the atmosphere?

In fact, after a few years of operation, the spent primary resins are usually transferred to the waste management facility during a reactor shutdown, for dry storage for a few years of decay, and then conditioned for long-term disposal. No atmospheric release of their inventory is expected during any of these operations.

To the author’s knowledge, no incidents related to these releases have been reported by any operator. The only communication of note related the June 2020 event, for which the Russian operator claimed: “No incidents were recorded at the Leningrad nuclear power plant and the Kola nuclear power plant, both stations operate normally, there have been no complaints about the equipment’s functionality” [12].

The Russian authorities monitor the radioactivity in the environment in the vicinity of the nuclear power plants (NPP). The results of the measurements are published annually on their website [13]. Moreover, the declared radioactive releases and the related authorizations for the NPP are presented. In the 2020 report [13], the declared releases from the Leningrad NPP are below the estimated source term of the June 2020 event in [1] (2 orders of magnitude for the ⁹⁵Zr, ¹³⁴Cs and ¹³⁷Cs), and below the authorized limits. In addition, it should be noticed that the estimated source term of June 2020 event is in the order of magnitude of the authorized limits⁴.

Without additional information from the operator concerned, or measurement data, it is almost impossible to understand what happened during these release events. A wide range of hypotheses were investigated in [1] for the June 2020 detection event, but “None of the hypotheses considered fit with all the characteristics of the release”.

Finally, from our analysis, a rather plausible scenario is proposed in this paper. The aim is not to describe what exactly occurred during the event, but to present a plausible scenario that could have led to the atmospheric release of a primary resin, consistent with the measurement data. However, this scenario is based on at least one main assumption and should be considered as speculative and one of several possible explanations.

The starting point of this scenario is to assume an incident during the transfer of a spent primary ion exchange resin from the reactor building to the waste management one. In fact, this transfer is usually performed by hydraulic transfer (use of pressurized water to pull the resin) through some pipes outside the plant buildings during the shutdown of the NPP. During this transfer, a resin blockage may occur, for example in a pipe bend. That type of event is considered to be potentially quite frequent, in the order of one per year for a NPP with several units. Such an incident implies a cutting operation on that pipe to remove the blocked part of the resin and to restore the system. This operation itself, if it is carried out correctly and it is assumed that it is, does not release any significant activity into the atmosphere. Moreover, the decay time at that moment is a few weeks and is not consistent with the detections.

However, such an operation also involves the contamination of some tools and the production of some solid waste, such as plastic vinyl sheeting. This solid waste may contain a small fraction of the original resin inventory. In this scenario, it is assumed that this fraction is of the order of a few percent. If the inventory in a resin is in the order of hundreds to thousands of GBq for ¹³⁷Cs or ⁶⁰Co for example, such a fraction is still compatible with the estimated source term of the various releases.

Then, after a few months of decay, these solid wastes are burned in an incinerator, and the associated source term is released to the atmosphere.

This scenario is fully consistent with all the measurements (isotope list, isotope ratios, source term magnitude, release duration) reported during the June 2020 and May 2023 events (and globally all the 14 mentioned detections). Moreover, it is also fully consistent with the Russian communication indicating that no incident occurred in June 2020 in the Leningrad or the Kola Nuclear Power Plant: the incinerator not being a part of the reactor. It also must be noticed that in this scenario, the resin may come from another reactor. For example, the resin could be from the Kalinin reactor, and the solid waste being burned in Leningrad (the Kalinin nuclear power plant did not have an incinerator [14]).

However, this scenario suffers from a significant limitation: in an incinerator, emissions are expected to be filtered, which would greatly reduce the activity of any released material. Indeed, the Nukem Technologies company claims that the decontamination factor of the incinerator of Leningrad is about 30 000 [15]. With such filtration, the release magnitude would be so low that the incineration would be undetectable. Thus, to explain the detection, this scenario requires a leak in, or a major failure, of the filtration system leading to a decontamination factor close to one. This is a very unlikely hypothesis, that the filtration system could be so inefficient, during so many years.

However, if all systems had been functioning correctly, there would have been no detection. The observed release must therefore have resulted from an abnormal and unlikely situation.

6. Conclusion

Among the numerous detection events reported in Northern Europe in recent years, the May 2023 one stands out as one of the most intriguing and informative for advancing the understanding of the origin of these releases.

Analysis of the event indicates that a release occurred between May 18 and 23, 2023. The levels measured were close to detection limits and did not pose a health risk. Using inverse modelling techniques, the most likely release location was identified within an area extending from eastern Estonia to western Russian Federation, consistent with previous analysis conducted concerning other releases.

An estimate of the release magnitude at the most likely location, suggests an activity at a few GBq of ¹³⁴Cs.

Analysis of the isotopic composition of the release indicates that the most likely source is a primary ion exchange resin of a WWER reactor with fuel cladding failure and dispersion of fissile material in the main coolant, with a decay time between 2 and 5 months after reactor shutdown. However, this conclusion remains uncertain due to the limited data available.

⁴⁶Sc is an activated corrosion product specific to WWER reactors due to the presence of Ti in the steel alloy use in the primary circuit and internal components. The presence of this specific isotope in the May 2023 release provides additional evidence supporting a WWER origin for the release.

In fact, such events are relatively frequent. At least five similar ones have been reported between 2012 and 2024, and possibly as many as to fifteen with different fuel cladding failure states.

A hypothetical release scenario has been proposed that matches all the characteristics of the June 2020 and May 2023 events, although it relies on some unlikely assumptions. However, without further information, the exact origin of these releases remains unknown.

Acknowledgments

The authors express their gratitude to all members of the ro5 network and especially the STUK (Radiation and Nuclear Safety Authority’s of Finland), the FOI (Swedish Defence Research Agency) and the Keskkonnaamet (Estonian Environmental Board) for providing measurement data. They also acknowledge the contribution of IRSN colleagues involved in the analysis of the event.

Funding

This analysis has been self-funded by IRSN and did not receive any other specific funding.

Conflicts of interest

The authors declare that they have no competing interests to report.

Data availability statement

The main data used in this article are airborne concentration detections provided by members of the ro5 network, in particular STUK, FOI and the Keskkonnaamet. These are publicly available on their website and are reported in the supplementary material. The results of the atmospheric dispersion calculations in Section 2 cannot be published for legal reasons, as the LdX code is owned by the IRSN and its results cannot be shared. Furthermore, the meteorological data used for the simulations are the property of Météo-France and cannot be shared. For the other sections of the paper, no associated data have been generated.

Author contribution statement

All the authors have read and approved the final manuscript. JJ. Ingremeau wrote the paper, and conducted the analysis of the Sections 1, 3–6. O. Saunier conducted the analysis of the Section 2.

Supplementary material

Supplementary file supplied by the authors. Access here

References

All Tables

All Figures

	Fig. 1. Percent of the simulated air concentrations that is within a factor of 5 of the observed values. Blue triangles are nuclear plants located in the area of interest: Leningrad NPP, Petersburg Nuclear physics Institute (Gatchina), Olkiluoto NPP, Loviisa NPP and Smolensk NPP.
In the text

	Fig. 2. 134Cs plume dispersion assuming a release at the location (29.2; 59.2). Up – May 22nd – Down May 23rd. Blank dots are air concentration measurements below the detection limit of the instruments.
In the text