1 Introduction

1.1 Background

The Comprehensive Nuclear-Test-Ban Treaty (CTBT) is verified primarily through a multi-technology network of sensors which detect the key explosion phenomenology associated with a nuclear detonation – seismic (S), hydroacoustic (H) and infrasound (I) waveforms; and atmospherically-transported radioactive material created during the nuclear event, known as radionuclides (RN) (United Nations, 1996). All of these monitoring technologies are subject to a real-world problem of ‘backgrounds’ – above and beyond that associated with the monitoring equipment itself. In the same way that earthquakes, mining blasts, sonic booms and landslides all produce background signals to the waveform monitoring technologies (SHI), the atmospheric radioactive discharges of many civil nuclear facilities create a background signal for RN monitoring technologies.

This work, conducted by a collaboration of scientists from the UK, US and Sweden has explored the radionuclide signatures associated with an operating nuclear power plant, located at Hartlepool in northern England (UK). Through the use of high-resolution gamma (\({{\gamma }}\))-ray spectroscopy, environmental measurements of atmospheric radioxenon, stack measurements of emissions and atmospheric transport modelling (ATM), the Xenon and Environmental Radionuclide Analysis at Hartlepool (XENAH) collaboration has investigated the radionuclide fingerprint of Hartlepool Nuclear Power Station, an advanced gas-cooled reactor (AGR) nuclear power plant (NPP). This paper outlines the measurement equipment deployed, describes the analysis conducted, and provides initial results.

1.2 Radionuclide Monitoring

Anthropogenic radionuclides have been monitored in the atmosphere since the start of the nuclear weapons testing era in the 1940s. Sampling and measurement technology has improved such that a network of around 80 sampling stations covering the globe with 24-hour sampling periods can provide sufficient coverage to be able to detect a 1 kt atmospheric nuclear explosion within 14 days with 90 % probability (Medici, 2001). This network, combined with other monitoring technologies, is known as the International Monitoring System (IMS) (Matthews & Schulze, 2001; Goodwin et al., 2024) (shown in Fig. 1). 40 of the 80 monitoring sites will operate noble gas monitoring capabilities which provide extra sensitivity when particulate radionuclides are unlikely to be released, such as in an underground nuclear test (UGT) (Bowyer et al., 1996). Since the rollout of the IMS the network has detected signals, including radionuclides, likely from DPRK nuclear explosive tests (Ringbom et al., 2014).

Fig. 1
figure 1

The International Monitoring System Radionuclide network showing current and planned particulate monitoring stations (black circles) and particulate/noble gas combined monitoring stations (black/yellow circles). The network is supported by 16 radionuclide laboratories (not shown)

Full size image

Detection of specific radionuclides in the atmosphere is the only of the CTBT verification methods that can determine that an explosion was nuclear rather than, for example, a chemical explosion (Medici, 2001). In the post-Limited Test-Ban Treaty (LTBT) era, where nuclear tests tend to be conducted underground, arguably the most likely detection of radionuclides will be noble gases such as xenon. There are four radioactive isotopes and long-lived isomers of xenon (known as ‘radioxenon’) that are considered relevant to the IMS (\(^{\text {133}}\)Xe, \(^{\text {135}}\)Xe, \(^{\text {131m}}\)Xe and \(^{\text {133m}}\)Xe) with half-lives in the range of hours to weeks that are produced in sufficient quantities to be measurable once the atmosphere has dispersed the material. The IMS routinely monitors the atmosphere for the presence of these four isotopes of radioxenon and the data is analysed by the International Data Centre (IDC) of the CTBT in Vienna and National Data Centres (NDCs) of State Signatories to the Treaty.

Since the early 2000s, isotopes of radioxenon have been detected relatively constantly on the IMS Noble Gas network. One of the sources of radioxenon is the industrial requirement to dissolve and chemically separate an irradiated uranium target in order to produce \(^{\text {99}}\)Mo to generate \(^{\text {99m}}\)Tc for medical imaging in hospitals. These Isotope Production Facilities (IPFs) are responsible for a large amount of the radioxenon in the atmosphere (Doll et al., 2014; Saey et al., 2010; Hoffman & Berg, 2018). There are less than 10 locations worldwide currently conducting such processes however, the quantities of radioxenon released as a result of dissolving irradiated uranium mean the impact on the IMS is significant (Hoffman et al., 2009; Generoso et al., 2022).

Previous work has shown that radioxenon detections on a non-IMS system at AWE (Aldermaston, UK) have been attributable to the Institute for Radioelements (IRE)—a medical isotope production facility in Belgium which has shared stack monitoring data (Goodwin et al., 2021b). Just a few isotope production facilities provide stack emission data to the nuclear explosion monitoring community, meaning a deeper understanding of how facilities contribute to the radioxenon background is required (see Section 3.2 on Stack Monitoring).

In contrast, the magnitude of emissions of nuclear power reactors are usually much lower, but the 320+ facilities worldwide (mostly concentrated in the Northern Hemisphere) mean their cumulative impact is considerable (Kalinowski & Tuma, 2009; Kalinowski, 2023). Recent work using ATM to estimate the \(^{\text {133}}\)Xe background in the East Asian region concluded that more than 50 % of samples were dominated by nuclear power plant contributions during a 6-month periodFootnote 1 (Kusmierczyk-Michulec et al. 2022). Not only do some of these facilities contribute to the global radioxenon background, but some power stations that are close to IMS stations are responsible for frequent radionuclide detections (Miley & Eslinger, 2023). Figure 2 shows the number of detections of CTBT-relevant radionuclides using data reviewed by the IDC for all of calendar year 2021.

Fig. 2
figure 2

Histograms showing the number of detections of each CTBT-relevant particulate radionuclide detected during the year of 2021. Top: activation products, Bottom: fission products. This does not include radioxenon detections

Full size image

There are a variety of systems developed for the IMS to enable radioxenon monitoring, including the SAUNA (Ringbom et al., 2003; Aldener et al., 2023) and SPALAX (Fontaine et al., 2004; Cagniant et al., 2017) and newer systems such as Xenon International (Hayes et al., 2015; Brander et al., 2022). Generally these systems make use of adsorption on materials such as activated charcoal that have a high adsorption coefficient for noble gases, to separate xenon from other gases in the atmosphere, then measure the radioactive isotopes present in the purified xenon sample using beta-gamma (\({{\beta -\gamma }}\)) coincidence spectrometry or high-purity germanium gamma spectrometry. Data from the IMS is processed by the IDC and National Data Centres (NDCs) nominated by each State Signatory to the Treaty. Figure 3 shows frequency distributions of \(^{\text {133}}\)Xe detections at six IMS facilities. The activity concentrations (measured in milli-Becquerels per cubic metre of air) vary considerably depending on the region and proximity to local nuclear facilities. The European sites DEX33 (Freiburg, Germany), SEX63 (Stockholm, Sweden) and NOX49 (Spitsbergen, Norway) show significantly higher activities than those in the southern hemisphere (AUX09, Darwin, Australia), Eastern Asia (JPX38, Takasaki, Japan) and in the Pacific Ocean (USX77, Wake Island, USA).

Fig. 3
figure 3

Frequency histograms (and smoothed frequency polygons) showing the measured activity concentrations above the critical limit (L\(_{\text {c}}\)) of \(^{\text {133}}\)Xe at six IMS noble gas monitoring stations. The data shown corresponds to the year of 2021 and analysis was completed using the UK National Data Centre radionuclide analysis pipeline (Goodwin et al., 2021a). Note the horizontal axes scales are not equal

Full size image

1.3 Atmospheric Transport Modelling (ATM)

Atmospheric Transport (and Dispersion) Modelling techniques are used to determine the origin of material detected at atmospheric radionuclide monitoring sites. To do this meteorological data is provided as an input to ATM models to simulate the transport of the material(s). Typically, meteorological data is obtained as 4-dimensional fields of variables such as wind, temperature and precipitation from numerical weather prediction models which combine observations of the weather with physical models of the atmosphere. Depending on the distance between the detection site and anticipated emission(s), a user can combine meteorological data sources with varying degrees of fidelity with a range of ATM models (with configurable degrees of temporal and spatial resolution) to analyse and interpret the RN data. ATM simulations can be performed in two modes:

  • Forwards (in time): from an emission point, where the simulated ‘contribution’ can be calculated at a sensor point over an average period of time – useful for estimating a ‘background’ signal from known civil radionuclide sources or simulating the dispersion of material from a known ‘source’, such as an announced nuclear test. A series of ‘ATM Challenges’ have examined the ability of different combinations of meteorological data and ATM to model the movement of radioactive material from known ‘background’ sources in the atmosphere (Eslinger et al., 2016; Maurer et al., 2022).

  • Backwards (in time): from a measurement point, where the simulation release period corresponds to the collection period of a sampler station. This is useful to identify an area that may include the source location. Any given area associated with a sample is time-dependent and so backwards simulations are often represented by a series of data points with a time and spatial field (Goodwin et al., 2022).

For the purposes of this work, both forwards and backwards ATM simulations are required. This project involves both calculating the movement of the emitted material from Hartlepool (forward simulations) and identifying the source(s) from a given detection on an array of sensors (backwards simulations).

1.4 Xenon and Environmental Radionuclide Analysis at Hartlepool (XENAH)

A group of Scientists from the UK, US and Sweden are working together as part of the XENAH collaboration. The scientists are predominantly involved in the verification work of the CTBT and this joint research project with EDF Energy UK enables an insight into the workings of an operational nuclear power station. EDF Energy UK have provided stack filter samples for analysis at AWE (Aldermaston, UK) and PNNL (Richland, WA, US), allowed the installation of a stack monitoring system at Hartlepool Power Station and supported the research with the provision of operational status data from the reactor.

2 Hartlepool Nuclear Power Station

Hartlepool power station is situated on the northeast coast of the UK and operates two identical AGRs. The station has been in active operation since 1983, generating around 600 MW(e) per reactor (around 1560 MW(th) each). The AGR design consists of a graphite-moderated reactor core and uses pressurised CO\(_{\text {2}}\) as the primary coolant.

The mean diameter of the active core is approximately 9.3 m and 8.2 m high, with total core diameter being approximately 11.9 m and 12.7 m high. The reactor core consists of columns of circular cross-sectional graphite bricks with interstitial square cross-sectional graphite bricks. There are 324 on-lattice fuel channels formed by bores in the larger circular bricks. Each bore has a diameter of 0.27 m and the distance between each channel is approximately 0.47 m. 81 boronated steel control rod channels are in a one-in-four array in the smaller square brick columns. Within each of the 324 fuel channels are eight stacked fuel elements, each containing 36 clustered fuel pins arranged in concentric rings of 18, 12 and 6 pins within a graphite sleeve. The stainless-steel-cladded fuel pins are approximately 1 m in length with a diameter of 14.48 mm and contain stacked ceramic UO\(_{\text {2}}\) pellets of either 3.2% or 3.78% \(^{\text {235}}\)U. The total core inventory of uranium is approximately 130 tonnes. The primary coolant is driven around the core by eight gas circulators which each have a constant-speed motor running at 3000 rpm, resulting in a total gas mass flow of 3600 kg s\(^{\text {-1}}\). In normal, full-power operation, the primary coolant operates at 39 bar with temperatures at the bottom of the active core around 270 \(^{\text {o}}\)C and 650 \(^{\text {o}}\)C at the top. Heat is deposited into eight steam generators by the pressurised CO\(_{\text {2}}\). Demineralised water acts as the secondary coolant and is fed to each steam generator in a closed-loop system at a rate of 60 kg s\(^{\text {-1}}\). The primary and secondary loops are separated within the steam generator to ensure no contamination is spread to the steam-side equipment from the active primary coolant. The steam generators operate at approximately 139 bar and produce steam at 570 \(^{\text {o}}\)C. Figure 4 shows a schematic of an AGR.

Fig. 4
figure 4

Schematic of an AGR, showing the primary reactor core where coolant (CO\(_{\text {2}}\)) is passed through the reactor and the heat is transferred to water before being used to generate electricity in the turbine hall

Full size image

Fuel typically remains in the core for around eight years, with average discharge irradiation of 32 GWd/MTU.Footnote 2 The on-load average fuel irradiation is around 15.5 GWd/MTU. The reactor typically remains at full power for 22 weeks before undergoing a controlled reactor shut down for off-load, depressurised batch refuelling. During this time the reactor’s primary coolant inventory is exhausted to atmosphere through a filtered route before 20–22 of the highest burn-up fuel assemblies are replaced. Outages across the two reactors are staggered to avoid overlap of shutdown periods. This results in five shutdown periods per year across the two reactors.

Fig. 5
figure 5

Bypass circuit overview showing how the gas is routed during normal operations (yellow) and during a blowdown (blue), with various processing steps along the way

Full size image

The purpose of the bypass circuit is to control the chemical composition of the primary coolant. In normal, at-power operation approximately 22 kg s−1 of reactor gas (around 1% of coolant inventory) is tapped from the inlet plenum at the bottom of the reactor. This gas firstly passes through the bypass/blowdown filter which is a woven stainless steel particulate filter. Gas then passes through the Recombination Unit which converts carbon monoxide back into carbon dioxide. After passing through the Recombination Unit, gas then passes though the drier, which removes excess moisture created from the breakdown of methane. Methane and carbonyl sulphide (COS) are injected into the reactor gas after the drier. These gasses are added in trace amounts to help preserve the graphite core and reduce fuel pin deposition. Once excess moisture has been removed from the primary coolant by the drier, the gas then passes though another particulate filter before re-entering the reactor at the inlet plenum. This continuous, at-power process is shown in yellow in Fig. 5.

Prior to reactor blowdown, the bypass (yellow) circuit is isolated. Reactor gas is no longer tapped from the inlet plenum and conditioned. Instead it remains within the pressure vessel until a blowdown is initiated. The blowdown circuit is shown in blue in Fig. 5. During a reactor blowdown, gas passes from the reactor firstly through the bypass/blowdown filter before going though the Iodine Absorption Plant (IAP), which includes a pre-filter and charcoal bed. The prefilter captures particulate down to 0.1 micron and the charcoal removes iodine before the gas is then vented to atmosphere. A mass flow computer is situated just prior to the IAP which allows measurements of released gas to be taken. During the first part of this process, reactor gas vents to atmosphere via the ‘major blowdown stack’. Installed at this stack is the STAX system (see Section 3.2). Towards the end of the blowdown, the blowdown route is re-configured and reactor gas vents to atmosphere from the ‘minor blowdown stack’. This route is not monitored by the STAX system.

During both at-power and blowdown operations, the primary coolant is monitored by the Gaseous Activity Monitoring (GAM) system. The GAM system taps reactor gas from boiler inlets and measures in-core coolant activity before re-circulating the gas back to the reactor vessel. The system allows for a real-time analysis of the in-core reactor gas using a high-purity germanium (HPGe) detector system. Routine in-circuit monitoring of the primary coolant activity is undertaken to assess the condition of the nuclear fuel. Monitoring is also undertaken during the planned depressurisations through the filtered route prior to refuelling. Typical at-power in-core activity is calculated as 150 kBq kg−1 (of CO\(_{\text {2}}\)) for \(^{\text {133}}\)Xe and 50 kBq kg−1 (of CO\(_{\text {2}}\)) for \(^{\text {135}}\)Xe. Data from this system can be used in conjunction with data from the Pi mass flow computer to estimate total inventory release, irrespective of whether the Major or Minor blowdown routes are in service.

Figure 6 shows a spectrum collected from a routine measurement of the Reactor 1 gas. The measured reactor gas contains various fission and activation product noble gas radionuclides (including radioxenons: \(^{\text {133}}\)Xe (fis), \(^{\text {135}}\)Xe (fis), \(^{\text {135m}}\)Xe (fis), \(^{\text {138}}\)Xe (fis), \(^{\text {125}}\)Xe (act); radiokryptons: \(^{\text {85m}}\)Kr (fis), \(^{\text {87}}\)Xe (fis), \(^{\text {88}}\)Kr (fis); and \(^{\text {19}}\)O (act), \(^{\text {23}}\)Ne (act), \(^{\text {41}}\)Ar (act), \(^{\text {88}}\)Rb (fis)).

Fig. 6
figure 6

In-core gas \({{\gamma }}\)-ray spectrum with an acquisition time of 15 min. The key \({{\gamma }}\) lines are annotated with the radionuclide responsible for the emission

Full size image

During a refuelling or inspection process, the reactor is depressurised. The GAM system can monitor the in-core isotopic activities over time (see Fig. 7). When the reactor is depressurised following power-down, the activity begins to drop. However, because the pressure of the detector system remains constant, the drop in activity is due to radioactive decay of the radionuclides present in the coolant gas — hence why the activity does not drop in the same fashion as the reactor vessel pressure in Fig. 7. This in-core state-of-health monitoring system provides an insight for the nuclear reactor physicists working at the site. It is possible to monitor for fuel failures and the presence of certain activation products can indicate the presence and quantity of other gases in the system. The GAM data can be compared with simulations of the reactor using nuclear reactor modelling software such as the Oak Ridge Isotope Generation and Depletion (ORIGEN) software package. Initial studies suggest the isotopic activities derived from computer simulations (Burnett et al., 2021) agree well with the measurements of the GAM system. As with most nuclear power reactors, the isotopic activity ratios associated with reactor start-up can often look like those of a nuclear explosion, however the activities are usually considerably lower (see Fig. 8).

Fig. 7
figure 7

GAM data (activity concentration for \(^{\text {133}}\)Xe and \(^{\text {135}}\)Xe) from a reactor depressurisation during a blowdown event, prior to refuelling of one of the reactors at Hartlepool. The pressure is the absolute reactor pressure and is different from the GAM system pressure, which is why they follow different trends

Full size image

In Fig. 8, a model of an AGR is used in the ORIGEN software which simulates the reactions taking place in a nuclear reactor (Rearden & Jessee, 2016), shown as the isotopic activity ratios for specific radioxenons that are used to discriminate between nuclear explosions and non-explosive sources (Kalinowski et al., 2010). The radioxenon isotopic activity ratios are extracted and plotted for various stages of the power cycle. A measurement point from the reactor building blowdown route (known as ‘R6’) is included (red diamond) which does not agree with the estimates for the start-up (yellow circles) and operation (purple circles) phase of the power cycle; however it is closer to the simulated ratios from activation of air. The plot shows various simulated and measured data for air activation, including ORIGEN-simulated values (green circles) and measurements performed by Klingberg et al. at the High-Flux Isotope Reactor (HFIR) in Texas, US (Klingberg et al., 2013). The irradiations used various neutron energies, from thermal to 14 MeV, which are shown by the triangular datapoints in Fig. 8.

This indicates that there is some similarity between the isotopic ratios from the Hartlepool AGR and the irradiated natural xenon (in order of decreasing natural abundance: \(^{\text {132}}\)Xe, \(^{\text {129}}\)Xe, \(^{\text {131}}\)Xe, \(^{\text {134}}\)Xe, \(^{\text {136}}\)Xe, \(^{\text {130}}\)Xe, \(^{\text {128}}\)Xe, \(^{\text {124}}\)Xe, \(^{\text {126}}\)Xe). The neutron activation route leads to more favourable production of \(^{\text {131m}}\)Xe over \(^{\text {133m}}\)Xe, when compared to fission, hence shifting the ratio from the right to the left of the plot.

It is feasible that small amounts of air (which include stable isotopes of xenon) are irradiated by the operational reactor, creating radioxenon isotopes through activation. This process can result in isotopic activity ratios which deviate from expected ‘fission-like’ ratios. The in-core and stack-emission datasets are crucial to undertanding this process, providing a ‘live’ picture of what is going on in and around the reactor during operation and refuelling.

Fig. 8
figure 8

Calculation of the radioxenon isotope ratios shows values right of the nuclear explosion discrimination line during AGR start-up and left during routine AGR operation

Full size image

3 XENAH Measurement Campaigns and Initial Results

The XENAH measurement campaigns have been designed to utilise newly available technologies, in order to reveal more information on the impact of an AGR nuclear power station on nuclear explosion monitoring networks. This required input from the main collaborating organisations (AWE (UK), EDF (UK), PNNL (UK) and FOI (Sweden)) as well as support from the Met Office (on atmospheric transport simulations) and UK academic institutions – STFC, The University of Leeds and Durham University.

The collaboration involves real measurements of radionuclides at various stages of the Hartlepool nuclear power station operational process – at source, at release point and in the environment. Figure 9 summarises the three tranches of measurement work completed as part of this project.

Fig. 9
figure 9

Summary of radionuclide measurement tranches for XENAH

Full size image

These measurements are supplemented by important information and further calculations, including:

  • Plant Information – provided by EDF to enable the group to attribute results to real-world processes (see Fig. 10).

  • In-core measurements – Hartlepool Nuclear Power Station operates an in-core gas activity monitoring system which provides useful comparison data for both stack measurement release data and local sample measurement data.

  • Meteorological data and ATM – Since the XENAH measurements will be focused within a fairly small region, high-resolution meteorological data is preferred, which is provided by the Met Office, who also develop and maintain the NAME model for atmospheric transport calculations.

Fig. 10
figure 10

Power profile for reactors 1 and 2 at Hartlepool Power Station for the year of 2020. Refuelling and statutory outages are scheduled on alternating reactors to maintain power output

Full size image

The work of the XENAH collaboration is intended to reveal new information about the radionuclide emissions from an AGR. Like many nuclear power facilities, regulators ensure the operators monitor and declare the quantities of certain radionuclides emitted to the environment, however these are not necessarily the same radionuclides relevant to the nuclear explosion monitoring community, or those that are detected regularly on the IMS. Treaty-verification measurements are also typically performed at much lower activity levels than those required by health and safety regulations. Figure 10 shows the power profile for both reactors at Hartlepool during the year of 2020. There are around 4–5 outages per year, each time resulting in a ‘blowdown’ of the reactor, emitting radionuclides to the atmosphere. The XENAH collaboration will publish data obtained before, during and after the blowdowns in the form of reactor information and in-core data, stack monitoring data, measurement of facility stack filters and environmental monitoring in the region. Sections 3.1-3.3 discuss the three specific tranches of work taking place.

3.1 Filter Measurements

During the blowdown of a reactor at Hartlepool Nuclear Power Station, the coolant gas is extracted from the reactor core and released to the atmosphere. Before it is emitted, the gas is directed through a filtered route, in order to remove or abate as much of the radioactive isotope content as possible. These filters include papers and charcoal maypacks. Charcoal is particularly ideal for trapping iodine and xenon, whereas the paper filters are used to trap other particulate-based isotopes. These samples are used to inform the regulator(s) (ONR/EAFootnote 3) that the facility is operating within regulations. There are multiple filtered routes for gas processing at Hartlepool, including from the fuel pond area and the gas circulation monitoring facility (GCMF); radionuclides produced during reactor operations that are released to the environment are most likely processed by one of these filtered routes. As part of this study, a selection of samples from the different filtration routes on various blowdowns were re-measured using HPGe-based detector systems at AWE and PNNL.

During the initial measurements at AWE and PNNL, the samples were formed into a reproducible geometry (see Fig. 11) and measured on different HPGe detectors. Initial measurements focussed on re-measurements of samples a few months after a blowdown, however recent work has included measuring charcoal samples just a few days after they were removed from the nuclear site, revealing many shorter-lived radionuclides.

Fig. 11
figure 11

Samples received from EDF Hartlepool, packaged into plastic containers ready for gamma spectrometry measurements

Full size image

This part of the project also aims to make use of ultra-sensitive gamma spectroscopy measurements techniques, such as gamma-gamma coincidence measurements, as well as utilising systems in low-background environments, such as the Shallow Underground Laboratory (SUL) at PNNL, or the AWE HPGe system at the Boulby Deep Underground Laboratory in the UK (Scovell et al., 2018). Lower-background measurement systems can provide much lower detection limits, making it possible to detect trace-level radionuclides that may otherwise not be detected. Activation products including \(^{\text {60}}\)Co, \(^{\text {44}}\)Sc, \(^{\text {54}}\)Mn, \(^{\text {108m}}\)Ag and \(^{\text {110m}}\)Ag have been detected in the samples measured so far (Petts, 2024).

3.2 Stack Monitoring

The stack monitoring aspect of the XENAH collaboration involved the installation of a commercial monitoring system, provided through the Source Term Analysis of Xenon (STAX) project run by PNNL (Metz et al., 2022). The modified NGM-2000 system, produced by VF Nuclear, utilises a HPGe detector and can measure the four main radioxenon isotopes. The only modification of the system from the base unit is the continuous airflow through the measurement cell. Gas from the stack is pumped though the cell in front of the detector, and a spectrum is acquired every 15 min. This system is similar to stack monitoring systems deployed at other nuclear facilities worldwide. The stack monitoring system installed at Hartlepool Power station is connected to the R6 major blowdown route at a tapping point within the stack just prior to the exit, after the filters. This system has been in continuous operation since March 2022.

Spectra are automatically analysed by the software running the system (Gamma Vision™). The flow rate though the stack is also monitored, and using the measured activity and flow rate, the overall release rate in Bq h−1 is calculated for every 15-minute interval. An example STAX data set using the February 2023 blowdown of Reactor 2 is shown in Fig. 12. This release event shows the measured \(^{\text {133}}\)Xe release rate as determined by the STAX system and the in-core monitoring system. The in-core GAM system is described in Section 2. Reactor coolant gas from the primary circuit is sampled and passed through a 1 L stainless steel Marinelli beaker. Activity of gaseous fission products are measured using a HPGe detector. These activities are scaled using a measured flow of the blowdown taken just prior to the iodine absorption plant to give a release rate in Bq h−1, calculated in 1-hour intervals. During a blowdown prior to a refuelling activity, toward the end of the operation the exhaust route is changed from the major blowdown stack to the minor blowdown stack. As the STAX system is installed on the Major route, the final portion of the blowdown is not measured by this system. However, the in-core monitoring system can measure activities throughout and calculate a release rate since the measured flow prior to the absorption bed is common to both the major and minor routes. In addition to the automated STAX analysis, each spectrum was reanalysed manually by a gamma spectroscopy expert using the Genie2000\(^{\text {TM}}\) analysis software. The agreement between the automated and expert analysis gives good confidence that the automated spectral analysis is working well and is reliable. This is important since there are many thousands of spectra requiring analysis, which is performed automatically by specially-developed analysis software. The expert analysis was performed using hourly average data, which is why there is some discrepancy when the flow rate drops suddenly, and this is to be expected.

Fig. 12
figure 12

STAX data system results showing the measured \(^{\text {133}}\)Xe activity for the February 2023 blowdown. Graph includes automatic results directly from the system (‘STAX Analysis’, 15-minute resolution, grey squares), results from an expert review of the raw data (‘Expert Analysis’), results calculated from the in-core gas measurements (‘In-core Analysis’) and results from the expert review of hourly spectral data (‘Expert 1 h Analysis’). Note the grey squares are mostly hidden by the red circles, since the values are very similar

Full size image

The STAX system is designed for the analysis of radioxenon in the stack. The automated analysis has a library that is set for only the four main radioxenon isotopes (\(^{\text {131m}}\)Xe, \(^{\text {133m}}\)Xe, \(^{\text {133}}\)Xe and \(^{\text {135}}\)Xe). The manual analysis of the spectra showed more isotopes present than just these four isotopes, and in addition, \(^{\text {41}}\)Ar, \(^{\text {85m}}\)Kr, \(^{\text {125}}\)Xe, \(^{\text {127}}\)Xe, \(^{\text {135m}}\)Xe and \(^{\text {79}}\)Kr were also observed in all or a subset of the spectra. Analysis of the full dataset will be reported in an upcoming publication. The equipment installed and a photograph of it in-situ are shown in Fig. 13.

Fig. 13
figure 13

Stack monitoring device (NGM-2000, VF Nuclear) and (right) the same equipment installed at EDF Hartlepool in the Gas Circulation Monitoring Facility (GCMF)

Full size image

3.3 Radioxenon Monitoring

In order to assess the effect of radioxenon emissions from AGRs on IMS-like monitoring systems, the XENAH collaboration deployed three ‘SAUNA Q\(_{\text {B}}\)’ atmospheric samplers to locations in northern England. As shown in Fig. 14, the SAUNA Q\(_{\text {B}}\) is a smaller, more mobile analogue of the SAUNA II (and more recently, SAUNA III) systems integrated into the IMS. The SAUNA Q\(_{\text {B}}\) samplers use several stages of gas separation, including molecular sieves to remove moisture and silver zeolite (Ag-ETS-10) traps to purify xenon from air samples using thermal desorption, gas chromatography to quantity the volume of xenon in a sample and \({{\beta -\gamma }}\) coincidence spectrometry to quantify the radioactive fraction, using a NaI(Tl) and 18 cm\(^{\text {3}}\) plastic scintillator gas cell (Ringbom et al., 2023). The system operates a 12-hour sampling cycle, collecting around 1\(-\)1.5 cm\(^{\text {3}}\) of xenon gas which corresponds to an air volume of 11.5\(-\)17.2 m\(^{\text {3}}\). The detector system energy calibration is maintained using an automatic check-source adjustment, which is useful when the systems are installed in facilities which may not be temperature-controlled.

Whilst the SAUNA Q\(_{\text {B}}\) is slightly less-sensitive than a typical IMS system, their smaller size, reduced power requirements and ease of installation means they can be easily rolled out and incorporated into ‘arrays’. The benefits of radioxenon sensor arrays for verification activities have been recently reported (Ringbom et al., 2023). Each system can be monitored remotely using state-of-health data gathered by the system’s on-board pressure, temperature and gas flow sensors (amongst others). All data (including sample measurement data) is stored in an on-board database, which can be cloned, archived, queried and disseminated between the organisations running the systems to enable analysis.

Figure 15 shows the locations of each SAUNA Q\(_{\text {B}}\) deployed as part of the XENAH experiment, as well as the locations of operating nuclear reactors in the UK (as of the time of writing). The locations were chosen based on their proximity to Hartlepool Power Station as well as ATM sensitivity predictions for radionuclide emissions from the site (see Fig. 16). A study of the sensitivity of each site was conducted by simulating releases from Hartlepool NPP four times per day, each day for all of 2021 to determine the fraction of releases likely to be detectable; this study was conducted using HYSPLIT (Stein et al., 2015; Draxler & Hess, 1998) with meteorological GDAS data from NOAA (NCEI). This meteorological data has a 0.25 degree spatial resolution, which is fairly coarse for the scale of a local sensor array. As a result, the sensitivity predictions were calculated for indication only. Each system was fielded for approximately one year during which time many samples were collected containing isotopes of radioxenon; some of these could be associated to emissions from Hartlepool Power Station (in conjunction with source measurements using stack monitors), and some were deemed to be from other sources (e.g. IRE in Belgium and other nuclear facilities in the UK, Europe and further afield). The design of the array allowed for enhanced source location estimations based on detection times, magnitudes, and ATM. This capability is greatly reduced when measurements are made using a single, highly sensitive system (Ringbom et al., 2023). Multiple measurements (both positive detections and statistical non-detects) at several sampling points provide a great deal of information and have been shown to aid in source parameter reconstructions using various tools (De Meutter & Hoffman, 2020; Eslinger et al., 2022; Goodwin et al., 2022).

Fig. 14
figure 14

A SAUNA Q\(_{\text {B}}\) system during acceptance testing at AWE. The system is connected via network connection to a laptop which can be used to access the onboard database and view system state-of-health data

Full size image
Fig. 15
figure 15

Map showing the British Isles, with the operating UK nuclear power stations sites (as of 2022) labelled with yellow circles (north to south: Torness, Hartlepool, Heysham, Sizewell), and the XENAH Q\(_{\text {B}}\) array sites labelled with green triangles

Full size image
Fig. 16
figure 16

ATM sensitivity calculations, using simulated emissions (4 per day for a whole year) from EDF Hartlepool and expected detection limit of 0.3 mBq m\(^{\text {-3}}\) for \(^{\text {133}}\)Xe. Simulations performed using the HYSPLIT atmospheric transport and dispersion model with GFS 0.25\(^{\text {o}}\) meteorological data. The sensitivity calculations were used for indication only

Full size image

The analysis results from the SAUNA Q\(_{\text {B}}\) systems provide activity concentrations for four radioxenon fission product isotopes (\(^{\text {133}}\)Xe, \(^{\text {135}}\)Xe, \(^{\text {133m}}\)Xe, \(^{\text {131m}}\)Xe) at a 12-hourly time resolution. Figure 17 shows the \(^{\text {133}}\)Xe activity concentration at all three SAUNA Q\(_{\text {B}}\) sites at the start of the measurement campaign. Numerous detections during this period prompted further analysis, with several plumes of radioxenon identified. Further analysis, using atmospheric transport simulations of known nearby emitters who provide STAX data (such as IRE), as well as simulations of sites that do not share data, has made it possible to determine the likely source of each detection labelled in Fig. 17. Through the period of measurements shown here, excluding the time before Q\(_{\text {B}}\)-2 and Q\(_{\text {B}}\)-3 were operating, there is only one detection that occurs at a single sensor (that can be attributed to Hartlepool) whereas most detections are ‘array detections’ where plumes are passing across the entire array, producing concurrent detections.

The \(^{\text {133}}\)Xe detections across the array beginning 11\(^{\text {th}}\) April 2022 agree well with simulations from IRE, however the plume beginning 22\(^{\text {nd}}\) April 2022 (labelled **) is not expected, based on IRE data and is therefore likely from a different source. Some NDCs use forward simulations from all possible emitters (such as nuclear power stations, nuclear research reactors and medical isotope production facilities) to help identify the most likely source. Figure 18 shows the possible emitters considered in Europe and surrounding areas, and shows which emitters contributed to the samples most, based on forward ATM simulations from each emitter, multiple times per day. Forward ATM simulations suggest that the XENAH Q\(_{\text {B}}\) array was most sensitive to emissions from facilities in Scandinavia, Eastern Europe and Russia, although the lack of ground-truth emissions data makes a more precise determination difficult. There are over 500 detections of \(^{\text {133}}\)Xe on the XENAH Q\(_{\text {B}}\) array during the period of operation, so a more detailed analysis is planned with a dedicated write-up, in order to present these results.

Fig. 17
figure 17

Activity concentration time series for \(^{\text {133}}\)Xe at the beginning of the SAUNA Q\(_{\text {B}}\) measurement campaign. The Q\(_{\text {B}}\)-1 system was activated around mid-March and the other systems came online in late-March. Possible sources have been assigned to the most prominent detections. The dashed lines denote the critical limit of detection (\(L_c\))

Full size image
Fig. 18
figure 18

A locations of emitters used to simulate background contributions to all monitoring network stations [red circles: NPPs, green triangles: MIPFs, blue squares: IMS stations] and B emitters that contribute to the XENAH Q\(_{\text {B}}\)-1 system during the period identified by ** in Fig. 17. The larger the circle, the greater the sensitivity of an emitter to the Q\(_{\text {B}}\)-1 system. N.B. Local emitters were overwhelmingly more sensitive, however ground-truth data meant they could be overlooked in this analysis. The locations used in this analysis are not all operational nuclear reactor sites, and may include nuclear sites closed in the last few years

Full size image
Fig. 19
figure 19

Flow diagram showing the approach to the Q\(_{\text {B}}\) system data

Full size image

Spectra from the SAUNA Q\(_{\text {B}}\) systems were analysed using the beta-gamma matrix (BGM) method (Ringbom & Axelsson, 2020); this method has been shown to decrease the number of false positive detections compared to earlier methods. Screening of this dataset is different to how NDCs might screen IMS data, since the signal to identify is, in this case, a nuclear power reactor rather than a nuclear explosion. As such, the screening of results for this analysis has used a different approach. Backwards ATM simulations provide an opportunity to identify a possible source region for each sample and assess which samples may be associated with one another (Goodwin et al., 2022). Source reconstruction tools can be used to estimate the source parameters (release location, magnitude and release times) based on detection/non-detection data from the SAUNA Q\(_{\text {B}}\) systems and backwards ATM.

With the environmental radioxenon monitoring dataset completed at the time of writing, the effort now moves to data analysis. Recent work has shown how radioxenon sampler arrays are a powerful tool and having more than one sensor can improve the likelihood of identifying a possible source (Ringbom et al., 2023). Currently underway is a systematic analysis of the data, following the approach outlined in Fig. 19.

Fig. 20
figure 20

ATM simulation snapshots showing the transport and dispersion of a plume of material, based on a prolonged release from Hartlepool NPP. The subfigures include NPPs (Hartlepool and Heysham) as red stars and the three SAUNA Q\(_{\text {B}}\) sites as blue circles. Each subfigure corresponds to the labelled time period

Full size image

As part of the analysis of data from the SAUNA Q\(_{\text {B}}\) systems, it is sometimes necessary to run forward simulations from the likely emitter, with an estimated emission profile, in order to compare the simulation output with measurement data. On the scale of this measurement array, using high-resolution meteorological data and a dense model sampling grid are some of the methods for improving the fidelity of such calculations. Figure 20 shows plots of a simulated release of material developing over time (a forward simulation), generated using the Met Office dispersion model, NAME (Jones et al., 2007). Met Office meteorological data was used, which has 1.5 km spatial resolution for mainland UK and surrounding areas – much higher than the global meteorological data used for global network analysis. The input data for this particular simulated emission used STAX data from the emissions following the planned power outage (for reactor refuel) that took place around March 2022. The output of these simulations backs-up the finding that the detection of \(^{\text {133}}\)Xe at Q\(_{\text {B}}\)-1 is a result of a Hartlepool release.

4 Discussion

A multi-approach measurement campaign has been completed involving a group of scientists from the UK, US and Sweden. The measurements demonstrate radionuclide signatures associated with the operation of an AGR nuclear power station. The use of cutting-edge technology in the field of \({{\gamma }}\)-ray spectroscopy, stack monitoring, environmental radioxenon measurements and atmospheric transport modelling provides an insight into the effect of a nuclear power plant on the International Monitoring System of the CTBT.

There are many other types of nuclear reactors in use around the world. According to World Nuclear Association there are around 300 operable Pressurised Water Reactors (PWRs), 60 Boiling Water Reactors (BWRs), 47 Pressurised Heavy Water Reactors (PHWRs), 11 Light-Water Graphite Reactors (LWGRs), 8 Advanced Gas-cooled Reactors (AGRs) and a handful of other types (see Table 1). This means that AGRs make up less than 2 % of global nuclear power reactors. This work demonstrates the impact of an AGR on the IMS, but similar work is required to better understand the impact of water-cooled reactors, which is likely to be much greater based on the number of reactors in use (Kalinowski & Pistner, 2006).

Table 1 Operable nuclear power reactor types, showing the number of each type in operation as well as fuel, coolant and neutron moderator materials (adapted from World Nuclear Association www.world-nuclear.org)
Full size table

The outlook for nuclear reactors as a source of power generation is bright, with multiple government-backed R&D projects into new technology underway in the UK and around the world. Some of the most promising new designs of small modular reactors (SMRs) include High-Temperature Gas Reactors (HGTRs) and Very-High-Temperature Reactors (VHTRs), which use similar principles to the AGR. Understanding the impact of upcoming reactor designs (such as Molten Salt Reactors (MSRs)) as well as current PWR sites will be an important part of future work (Eslinger et al., 2021; Johnson et al., 2021).

One of the most important lessons learned from this work is that an AGR is not a continuous emitter of radionuclides. The primary coolant circuit is a closed loop with little opportunity for the escape of radionuclides during standard reactor operation. Other reactor types may release more regularly than an AGR, which tends to release radionuclides only during depressurisation, required for refuelling or regulatory inspections.

The approach taken in this work to make measurements of radionuclides at the source (using process samples from the site or data shared by the operators), from the stack and remotely in the environment (SAUNA Q\(_{\text {B}}\) systems) has provided insightful data which contributes to a better understanding of the impact of an AGR power station. The provision of samples from the power station provided awareness of radionuclides that are detectable in small quantities and may not be detectable on sampling networks, even on a local scale. Measuring the gaseous effluent using a stack monitor has been extremely useful, since most of the authors of this work have utilised similar data from other facilities. Stack monitoring, particularly through the STAX project, provides a ground-truth measurement which can be used as an input to ATM simulations. Finally the deployment of an array of radioxenon sensors is a relatively new concept and in this case proved very useful. Given the location of the reactor, known meteorological conditions and options for locating sensor equipment, using multiple less-sensitive systems was useful, rather than a single high-fidelity system with a shorter sampling time integral. This was for two reasons, firstly, it provided better spatial coverage of the region, with some detections being attributed to Hartlepool which were only detected on one system. Secondly, the use of an array made it more straightforward to analyse background signals. Whilst there is little directional information from this array (12-hour samples in an array <100 km apart) it is often clear when a plume is from a remote area (see Fig. 17). The array layout can make it easier to investigate possible sources.

To gain a better understanding of the atmospheric radionuclide background, more collaboration is required between the nuclear verification science community and the civil nuclear community. The approach taken by this collaboration could be applied to other nuclear sites in order to determine the effect of different reactor types on the IMS, and the impact on detectability of a nuclear test. The group of collaborators on this project aim to publish multiple papers with results from the measurement data collected to date.