Meaningful dose in geological disposal safety cases: human intrusion and normal evolution as competing exposure pathways

1 Introduction

Geological disposal is designed to make long-lived radioactive waste unavailable to ordinary human and environmental systems. Its safety case rests on passive isolation, waste-form stability, engineered barriers, geochemical retention, groundwater transport modelling and biosphere dose assessment (IAEA, 2011; 2012). Under normal evolution, the pathway from waste inventory to receptor dose is intentionally long, delayed and attenuated.

Human intrusion is treated differently. Future human actions cannot be predicted reliably over thousands to hundreds of thousands of years, and international guidance therefore discourages speculative forecasting of future societies while allowing stylised cases based on present-day knowledge and technologies (NEA, 1995; IAEA, 2012; ICRP, 2013). This caution is justified. No defensible safety assessment can claim to forecast future motives, institutions, technologies, land use or memory over geological timescales. A recent Swiss assessment likewise treats future human actions as a separate class of stylised safety scenarios and evaluates, among other cases, direct borehole penetration of a spent-fuel disposal canister and waste packages in a low- and intermediate-level waste cavern (Nagra, 2024).

Yet the distinction between normal evolution and human intrusion can hide a second question. Which pathway is more likely to produce a meaningful dose over a specified period? The question is not which pathway is more central to the compliance case. It is which pathway is more capable, probabilistically and physically, of converting isolated inventory into protection-relevant exposure.

We argue that the most modelled pathway is not necessarily the most exposure-efficient pathway. A normal-evolution assessment may demonstrate compliance with a dose criterion over long periods without showing that normal evolution is the pathway most likely to produce meaningful exposure. The common-mode problem is one of safety-case architecture: compliance under normal evolution, cumulative boundary-bypass probability, conditional consequence after physical intersection, and likelihood of meaningful dose are different objects. The safety claim remains incomplete unless these objects are kept separate.

Although developed primarily from spent-fuel geological-disposal examples, the pathway-efficiency framework also applies to intermediate-depth and other underground repositories where isolation depends in part on depth and human intrusion may create an accessible exposure configuration. By assessing both a spent-fuel repository at about 430 m depth and a LILW repository at about 180 m depth within the same report, Posiva’s FHA-OLA study illustrates the transferability of the framework while showing that consequences remain specific to waste type, inventory and repository design (Posiva, 2021).

The present paper complements an earlier analysis of the boundary between repository robustness and longer-term environmental consequence after canister-intersecting drilling (Pescatore, 2026) by addressing a distinct question: the probability architecture through which normal evolution and human intrusion can be compared within safety cases.

2 Meaningful dose and pathway-efficiency framework

This section sets out the framework on which the rest of the argument depends. It first defines the quantity to be compared, namely the probability that a pathway produces a dose at or above a stated meaningful-dose threshold, and it then examines how the choice of that threshold conditions the comparison. The two questions are separable in exposition but not in effect: the definition fixes what is being measured, whereas the threshold determines which pathway the measurement will find controlling.

2.1 Defining pathway exceedance probability

Let D* denote a meaningful-dose threshold. The threshold may be regulatory, diagnostic or protective: for example, a dose constraint, reference level, public-dose benchmark or other value chosen to indicate protection relevance. The comparison proposed here is:

P D ≥ D *

That is, the probability that a pathway produces a dose equal to or greater than the selected meaningful-dose threshold at least once during the assessment period.

For normal evolution, the pathway can be schematically decomposed as:where R denotes release from the waste form, engineered barrier or disposal system, and A denotes arrival in an accessible receptor or biosphere configuration. In a full safety assessment, these terms contain many processes: canister failure or degradation, waste-form dissolution, solubility limitation, sorption, matrix diffusion, groundwater transport, decay, dilution, discharge location, biosphere transfer and receptor behaviour (IAEA, 2012). Meaningful dose requires the pathway to survive this attenuating sequence.

P NE D ≥ D * = P R × P A R ) × P ( D ≥ D * A

For human intrusion, the analogous expression is:where I denotes intrusive activity and X denotes physical intersection of waste or contaminated material. Intrusive activity depends on future land use, technology, memory, institutions and site attractiveness. Physical intersection depends on repository depth, footprint, package geometry, waste distribution and intrusive practice. Conditional dose depends on whether waste-bearing material is brought to the surface, handled, dispersed, abandoned, remediated or incorporated into later land use (SKB, 2010; Posiva, 2021).

P HI D ≥ D * = P I × P X I ) × P ( D ≥ D * X

Figure 1 sets these two decompositions side by side, from a shared inventory to the common meaningful-dose threshold.

The comparison is therefore not between a “normal” scenario and an “abnormal” scenario as labels. It is between two conditional pathways to meaningful dose. A limiting exposure pathway is the pathway that, conditional on initiation, has the highest probability of producing a dose at or above D* . This need not be the pathway with the most probable initiating event; its dominance may instead result from bypassing, shortening or weakening the safety functions that normally separate inventory from exposure.

The two exceedance probabilities need not have identical evidential or epistemic foundations. In many assessments, P_HI(D ≥ D) combines a stochastic representation of whether and when intrusion occurs with epistemic uncertainty in the assumed rate and future conditions, whereas P_NE(D ≥ D) is generated principally from uncertainty in the scenarios, models and parameters governing the calculated dose history. A deterministic reference case does not itself define an exceedance probability, but its selected scenario and parameter values may embody an unstated judgement about where the calculation lies within the relevant uncertainty range; both quantities should therefore be interpreted conditionally on the assumptions and probability representations from which they are derived.

The exceedance probability is not proposed as a replacement for expected-risk or probability-weighted measures where these are used. Such measures integrate probability and consequence and may support compliance or comparative risk assessment, but they can obscure whether a low expected value arises from a low conditional dose or from a severe conditional dose assigned a small initiating probability. Where the governing metric is a dose criterion rather than a risk criterion, P(D ≥ D*) serves a different but complementary purpose: it retains the threshold structure and keeps visible the probability of protection-relevant exposure. In either setting, the exceedance formulation is an additional diagnostic rather than a substitute for the governing metric.

2.2 Choosing the meaningful-dose threshold

The threshold D* is not a single canonical value but a choice, and the choice has consequences for the comparison. In practice D* may be drawn from at least three sources. It may be a regulatory dose constraint, such as the 0.3 mSv/y value used in several national frameworks, which fixes an upper bound on acceptable individual dose. It may be an ICRP reference level appropriate to the exposure situation invoked—planned, existing or emergency—recognising that an intrusion bringing waste to the surface may be judged under a different situation category than a normal-evolution groundwater release (ICRP, 2013). Or it may be a lower diagnostic or optimisation benchmark, chosen not as a limit but as a marker of protection relevance below the constraint.

These sources differ in status, and the difference matters because the two pathways respond to D* with markedly different sensitivities. Human intrusion, once intersection occurs, creates a concentrated and largely unattenuated source; consequently, the conditional exceedance probability P(D ≥ D* |X) may remain high across a broad range of thresholds and be comparatively insensitive to the value of D. Low thresholds may be exceeded in most intersection cases, and a comparatively high conditional exceedance probability may persist as D rises. Normal evolution behaves differently: dose arises only after release, transport and dilution have acted, so the resulting receptor dose is comparatively low; accordingly, the conditional exceedance probability P(D ≥ D* | A) falls steeply as D* rises and becomes zero once D* exceeds the maximum dose attainable through the attenuated pathway.

The consequence is asymmetric. As D* is raised, the exceedance probability for normal evolution reaches zero at a lower threshold than that for human intrusion; beyond that point, only human intrusion remains capable of exceeding D. Conversely, lowering D brings normal evolution back into contention without eliminating human intrusion. For a mobile radionuclide such as I-129, whose normal-evolution release may itself be relatively probable, normal evolution may then rival human intrusion in exceedance probability. For a strongly retained actinide such as Pu-239, normal evolution produces essentially no meaningful dose across the protection-relevant threshold range, and human intrusion dominates the comparison. The choice of D* is therefore not a calibration performed before the comparison; it enters the comparison itself, since a defensible change in D* can change which pathway dominates P(D ≥ D*).

Figure 2 illustrates this asymmetry schematically: the human-intrusion exceedance probability declines comparatively slowly across a broad range of thresholds, whereas the normal-evolution probability falls steeply and approaches zero once D* exceeds the largest dose the attenuated pathway can produce. The crossing marks a region of comparable exceedance probability; beyond it, only human intrusion remains capable of exceeding D* under the stylised assumptions shown (Figure 2).

3 Compliance and pathway dominance

The framework of the preceding section has an immediate consequence for the way compliance results are read. A demonstration that calculated dose remains below a criterion under normal evolution is a statement about one pathway; it is not a statement about which pathway is most likely to deliver meaningful dose. The two claims are separated below, first as they arise in the structure of a conventional safety assessment and then as they appear in a documented regulatory review.

3.1 Compliance under normal evolution

Most geological-disposal safety assessments are structured to demonstrate compliance under normal or expected evolution. A calculated dose below a criterion such as 0.3 mSv/y over long assessment periods is an important result. It shows that the modelled repository system, under the assumptions of normal evolution, satisfies the regulatory performance criterion. However, this result answers a compliance question. It does not answer the pathway-competition question: which pathway is most likely to produce a meaningful dose over the same period?

Over early periods such as 10,000 years, normal evolution is intended to make meaningful dose essentially unavailable. Waste packages should remain intact or highly retentive; engineered barriers should delay release; the geosphere should retard transport; dilution and dispersion should reduce concentrations; and biosphere transfer should further mediate receptor exposure. In qualitative terms, the design objective is:

P NE D ≥ D * ≈ 0

for protection-relevant values of D* over early assessment intervals. The approximate zero is not a claim that every formal calculated dose vanishes. It is a protection statement: the designed system is intended to prevent meaningful dose from becoming available through normal evolution during that period.

3.2 Early-time pathway dominance: the Finnish example

The Finnish TURVA-2012 construction-licence review illustrates the contrast. STUK reported inadvertent human intrusion as the only disturbance scenario producing dose within the 10,000-year dose window (Sagar, 2015). This does not make intrusion predictable. It shows that a pathway treated as stylised and analytically separate may nevertheless be the more likely route by which repository material is reconnected with dose-relevant exposure during the period in which human dose is explicitly assessed.

The same logic extends to longer periods. Normal evolution may remain compliant when judged against a regulatory dose criterion, whereas the cumulative opportunity for human intrusion cannot decrease as the assessment period lengthens. Compliance under normal evolution therefore should not be used as evidence that normal evolution is the dominant meaningful-dose pathway. Posiva’s later FHA-OLA analysis reinforces the same point by treating both immediate driller exposure and a contaminated-landfill construction-worker exposure following canister-intersecting drilling, with very high conditional doses for spent-fuel material brought to the surface (Posiva, 2021).

4 Time-dependent boundary-bypass opportunity

Where the preceding sections treat the comparison at a fixed point, the present section introduces its temporal dimension. Boundary-bypass opportunity does not attenuate with time in the manner of release, transport and decay; it accumulates. The first subsection sets out the cumulative rate representation that expresses this accumulation, and the second states, in four parts, what such a representation can and cannot be asked to support.

4.1 Cumulative rate representation

Boundary-bypass opportunity accumulates with time. For a homogeneous Poisson process with annual hit rate B, the expected number of hits over time T is μ = BT, and:

P N ≥ 1 = 1 − exp − μ

For a non-homogeneous Poisson process with time-dependent annual hit rate B(t), the corresponding expressions are:

μ T = ∫ 0 T B t dt

P N T ≥ 1 = 1 − exp − μ T

These expressions are not predictions of future society. They state the cumulative implication of any specified annual canister- or package-hit rate (Helton, 1993). Their relevance here is limited but important: cumulative boundary-bypass probability is a different object from compliance under normal evolution and from conditional consequence after intersection. Because μ(T) and P[N(T) ≥ 1] are cumulative quantities, neither can decrease as the assessment period lengthens. A long period over which normal evolution remains compliant may therefore be the same period over which cumulative boundary-bypass opportunity becomes increasingly significant. This cumulative framing provides the link between boundary bypass and pathway efficiency: even when the annual hit rate is small, the cumulative opportunity for such intersection is not the same object as the calculated dose from normal evolution.

Earlier safety assessments already contain the ingredients for cumulative reasoning. In SR-97, SKB selected canister penetration by drilling as the human-caused scenario whose consequences warranted further examination because it directly breached a canister while remaining inadvertent, technically possible, practically feasible and plausible. SR-97 then combined a repository-area probability, a conditional probability of canister penetration and an assumed deep-borehole rate, and reported a cumulative probability of 1 for drilling through a canister “at some time” over 100,000 years under an illustrative rate of 200 deep boreholes per year in relevant crystalline basement (SKB, 1999). More generally, Poisson representations of drilling-intrusion probability are well established in the performance-assessment literature (Helton, 1993).

A regulatory precedent for combining intrusion-event representation with cumulative exceedance display is provided by the Waste Isolation Pilot Plant. Its compliance assessment used complementary cumulative distribution functions of normalised releases over 10,000 years and represented inadvertent drilling through a Poisson event process using a drilling rate prescribed by the regulatory framework on the basis of historical drilling activity in the region. The exceedance curves concerned cumulative normalised release rather than dose, and their purpose was regulatory compliance rather than comparison of exposure pathways. Nevertheless, the WIPP precedent shows that stipulated intrusion rates and cumulative exceedance displays can be used transparently without presenting them as forecasts of future society (U.S. EPA, 1996; Helton et al., 2000).

The same reasoning applies at the broader level of intrusive activity, not only to direct canister penetration by drilling. In their WIPP study, Hora and von Winterfeldt (1997) treated inadvertent human intrusion as a significant, possibly the most significant, long-term threat to repository performance and reported expert estimates of approximately 1%–10% cumulative probability of human intrusion over 10,000 years. They also cautioned against reducing intrusion to a single drilling event or aggregating heterogeneous intrusion modes into one probability, because different modes imply different physical source configurations and consequences. This is precisely the distinction required for pathway analysis: the protection-relevant question is not only whether a borehole intersects a package, but whether, over time, intrusive activity creates an accessible exposure configuration capable of producing meaningful dose.

A further reason not to treat the intrusion rate as fixed is that the social conditions controlling intrusion opportunity, especially memory and institutional continuity, are themselves time-dependent. Pettersson (2025) notes the irony that the Nordic KAN 1.3 work on information transfer to future generations, although it anticipated many later Records, Knowledge and Memory themes, has itself largely fallen into oblivion within a few decades. This near-present example reinforces the deep-time point made by Hora and von Winterfeldt (1997): memory loss, reinterpretation and institutional discontinuity are active components of intrusion opportunity, not merely remote abstractions.

4.2 Limitations of the rate representation

The Poisson expressions above are useful precisely because they are modest: they translate a specified annual hit rate into a cumulative probability without pretending to forecast future society. That modesty is also their limit, and four cautions follow.

First, intrusion modes that differ physically should, where practicable, be represented by separate mode-specific rates B_*m*(t). A shallow excavation, a deep borehole and a resource-extraction operation intersect different parts of the repository and create different source configurations; each should therefore be linked to its own P(X_*m* | I_*m*) and P(D ≥ D* | X_*m*). Aggregating these modes before their pathway-specific probabilities and consequences have been evaluated discards exactly the information needed for protection-relevant comparison (Hora and von Winterfeldt, 1997). Aggregation can still occur later, but only after the distinct pathways have been preserved.

Second, B(t) is not stationary. Memory loss, reinterpretation and institutional discontinuity are themselves time-dependent, so the very quantity that would make a homogeneous rate defensible—a stable social configuration—is the quantity least likely to persist over the relevant intervals (Pettersson, 2025).

Third, land-use and technology change act on the rate and on the conditional consequence at once: a practice that raises the probability of intersection may also alter what happens to intersected material.

Fourth, and as a consequence of the first three, a rate calculation can provide a conditional representation of P I under the specified B t, but is silent on P X ∣ I and P D ≥ D * ∣ X. The cumulative framing therefore supports the argument for keeping these terms analytically distinct; it does not license folding them back into a single number.

SR-97 provides an early illustration of this interplay. Its human-actions analysis combined technical screening with societal framework scenarios and treated knowledge of the repository, societal continuity, institutional control and land-use restrictions as determinants of intrusion opportunity. In the quantitative drilling case, it then separated the probability that the repository had been forgotten, the probability of drilling within the repository area, the conditional probability of penetrating a canister, and the resulting dose consequence (SKB, 1999). Although not expressed in the notation used here, this structure anticipates the distinction among P I, P X ∣ I, and P D ≥ D * ∣ X, while also showing why a single stationary aggregate rate cannot represent the full pathway.

5 Radionuclide and source-configuration illustrations

The consequences of the framework differ by radionuclide and, no less importantly, by the physical form in which intrusion delivers material to a receptor. The three illustrations that follow are chosen so as to separate these two influences. The first two contrast retention-limited and mobility-limited radionuclides under the same pathway comparison; the third addresses the source configurations that intrusion itself creates and that normal evolution does not produce.

5.1 Strongly retained actinides

The pathway comparison becomes clearer when applied to contrasting radionuclide classes. For strongly retained actinides such as plutonium or americium isotopes, normal evolution requires canister failure, fuel dissolution, actinide mobilisation, transport through buffer and geosphere, limited sorption or retention, biosphere discharge and uptake. Geological disposal is designed precisely to make this chain ineffective. If a borehole intersects a canister or waste package, however, Pu-bearing fuel or contaminated material can be physically brought to the surface as cuttings, core fragments, drilling slurry, fuel particles or contaminated residues. For Pu-239, intrusion may therefore be the dominant plausible route to biosphere exposure because it bypasses the retention mechanisms on which normal-evolution safety depends.

The asymmetry is one of mechanism rather than of magnitude alone. Normal-evolution performance for these radionuclides rests on retention steps acting in series, so that the weakening of any one is in part compensated by those that remain. Intrusion does not weaken the sequence; it removes it. What is bypassed is therefore not a margin but an architecture, and the conditional consequence given intersection is largely insensitive to the very features on which the normal-evolution case is built.

5.2 Mobile long-lived radionuclides

For mobile long-lived radionuclides such as I-129 or Cl-36, the argument is different. Normal evolution remains a credible release pathway because these radionuclides can be mobile over long periods. But even here, normal evolution is mediated by waste-form release, engineered barriers, groundwater transport, discharge location, dilution and biosphere transfer. Intrusion can instead create a localised surface or near-surface source without the same delay, dilution and dispersion. The compact distinction is therefore: for Pu-239, human intrusion matters because it bypasses retention; for I-129, it matters because it bypasses the delay, dilution and dispersion between failed canister and biosphere.

Two consequences follow. First, the pathway comparison for these radionuclides cannot be settled by mobility arguments alone, since both routes can deliver material to the biosphere and are distinguished by timing and dilution rather than by feasibility. Second, the quantity that matters conditionally is the concentration delivered at the point of exposure rather than the inventory released, because what separates the two routes is precisely the attenuation that normal evolution supplies and intrusion omits.

5.3 Intrusion-created source configurations

The importance of source configuration is further illustrated by Posiva’s FHA-OLA analysis, which considers both immediate driller exposure and a later construction-worker case in which drilling detritus from a canister-intersecting event is disposed of in a shallow uncovered landfill at the drill site (Posiva, 2021). For spent-fuel intrusion, conditional doses remain protection-relevant over very long times and are controlled by different radionuclides at different epochs: actinides such as Am-241, Pu-238 and Pu-239 dominate at early times in both exposure configurations, while longer-lived contributors including Ra-226, Nb-94 and Sn-126 become relevant later in the landfill case. The significance of the Posiva cases is not that such intrusions are predictable, but that once spent-fuel material is assigned to an accessible exposure configuration, the conditional dose structure differs substantially from that of normal-evolution groundwater pathways—both in the radionuclides that matter and in the timescales over which they matter.

By creating drill core, cuttings, slurry, contaminated detritus or landfill-like mixtures, intrusion may make radionuclides dose-relevant that are absent, negligible, or non-controlling in normal-evolution groundwater pathways. Ag-108m provides a useful example: it is not normally a principal radionuclide in groundwater-mediated normal-evolution dose assessments, but it appears among the radionuclides contributing to dose maxima in Posiva’s FHA-OLA drilling and contaminated-landfill cases. This reinforces the central point that human intrusion changes both pathway efficiency and source configuration. Normal evolution asks which radionuclides survive release, transport, dilution and biosphere transfer; human intrusion asks which radionuclides matter once waste-bearing material is physically relocated into an accessible exposure configuration.

6 Implications

Two orders of implication follow from the preceding analysis. The first is conceptual and concerns what a safety assessment is taken to demonstrate; the second is procedural and concerns what can be added to development and review without waiting for that conceptual shift. They are presented in that order because the practical steps are intelligible as consequences of the conceptual point, although they do not depend on its acceptance.

6.1 Conceptual implications

Two implications follow. First, safety assessments should compare pathways to meaningful dose, not only scenario categories. This means distinguishing P(I), P(X |I), P(D ≥ D* |X) and P_HI(D ≥ D) rather than collapsing them into a single plausibility judgement. Similarly, normal evolution should be decomposed into release, arrival and receptor exposure. Only then can P_NE(D ≥ D) and P_HI(D ≥ D*) be compared on a protection-relevant basis over a specified assessment period.

Second, the source configuration created by intrusion should be named explicitly. A breached canister, drill core, contaminated cuttings, drilling slurry, contaminated soil, abandoned borehole, landfill-like mixture and water pathway are different exposure configurations. They may involve different emergency or existing exposure interpretations and different controlling radionuclides. If human intrusion is a potentially dominant meaningful-dose pathway, then depth, repository footprint, package orientation, package density, site attractiveness, resource potential, information preservation and marker strategies are not peripheral variables. They affect P(N ≥ 1), P(X |I) and ultimately P_HI(D ≥ D*). A geothermal or thermal-water exploitation borehole is a particularly important configuration because it may combine boundary bypass, hydraulic contact, repeated use and public exposure. ¹

The framework is not limited to deep geological disposal of spent fuel. With waste- and design-specific parameterisation, it also applies to intermediate-depth repositories and other underground disposal concepts in which human intrusion may bypass isolation. Posiva’s assessment of its SNF and LILW repositories illustrates the common analytical structure: intrusive activity, physical intersection and conditional exposure should remain distinct even when the resulting doses and controlling radionuclides differ markedly (Posiva, 2021).

6.2 Practical implications for development and review

The argument implies four checks that can be applied directly in safety-case development and review. They are stated here as steps rather than principles so that they can be incorporated into practice:

• Bound the cumulative boundary-bypass opportunity. Estimate μ(T) = ∫₀ ^T B(t) dt and P[N(T) ≥ 1] = 1 − exp[−μ(T)] over each assessment period, and report them separately from the normal-evolution dose calculation rather than merging the two.
• Compare conditional exceedance probabilities by radionuclide class. Compute P(D ≥ D* | A) for the normal-evolution pathway and P(D ≥ D* |X) for the intrusion pathway for the controlling radionuclides, since the two pathways are dominated by different nuclides at different epochs.
• Identify the limiting exposure pathway explicitly. Determine which pathway, conditional on initiation, has the highest probability of producing a dose at or above D*, and state it as a finding rather than allowing the most fully modelled pathway to be treated as dominant by default.
• Name the intrusion-created source configuration. Specify whether the exposure arises from a breached canister, drill core, contaminated cuttings, slurry, contaminated soil or a landfill-like mixture, because the configuration governs the controlling radionuclides, the timescale and the applicable exposure-situation interpretation.

None of these steps requires claiming to forecast future society. They can be implemented through stylised scenarios and explicit, protection-oriented assumptions about intrusion opportunity, physical intersection, and conditional exposure. They draw on information already present in many safety cases—or on quantities that may be derived or bounded from that information—and together keep compliance, cumulative bypass opportunity, conditional consequence, and meaningful-dose likelihood visible as distinct analytical objects.

7 Conclusion

Human intrusion and normal evolution are usually separated because they belong to different epistemic and regulatory categories. That separation is justified, but it should not prevent a direct protection-relevant comparison. A normal-evolution safety assessment may demonstrate compliance with a dose criterion over one million years. That is an important safety-case result, but it is not a demonstration that normal evolution is the most likely pathway to meaningful exposure. If the conditional exposure following intersection is severe, human intrusion may become the dominant meaningful-dose pathway even while normal evolution remains compliant.

The safety case may therefore be complete as a compliance demonstration while remaining incomplete as a protection claim. That incompleteness is the common-mode architectural problem identified here. Human-intrusion analysis should therefore be revisited as pathway analysis: normal evolution asks whether release, transport and exposure survive attenuation; human intrusion asks whether intrusive activity, physical intersection and exposure follow boundary bypass. Expressing both pathways in a common probability language, under clearly stated assumptions, makes their respective protection significance visible and defines a new challenge for safety-case development and review: to confront the possibility that compliance, robustness and protection may not always coincide.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. During revision of this manuscript, the author used ChatGPT (OpenAI) for editorial assistance, including language refinement, clarification of wording, structural feedback, and consistency checks for terminology, notation, and references. The tool was not used to generate data or perform scientific analyses. All AI-assisted suggestions were critically reviewed, verified where necessary against the cited sources, and revised by the author, who takes full responsibility for the final content.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Michael Kühn, GFZ German Research Centre for Geosciences, Germany

Reviewed by: Francisco J. Bedoya Rodríguez, University of Santiago de Cali, Colombia

Zico Pratama Putra, National Research and Innovation Agency (BRIN), Indonesia

Footnotes

An ANDRA technical note on drilling plausibility and radiological-consequence assumptions for deep disposal explicitly retained thermal-water exploitation supplying a public spa among representative drilling scenarios, alongside other configurations involving contaminated cuttings, abandoned boreholes and package interception (ANDRA, 2000). Conditional on contact with waste-bearing or contaminated material, such a pathway may be more exposure-efficient than a one-off occupational encounter with contaminated cuttings.