the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Jul 2025
How to handle uncertainties in modelling due to human reliability issues for nuclear disposals
Abstract. Modelling plays a crucial role in assessing the design for future technical or geological development of a repository for radioactive waste. Models and the application of these models to scenarios are used to weigh different safety-related designs or to assess the suitability of sites. Even if the models simulate and evaluate a system in a geological context that is designed as a passively safe system, the human factor plays a significant role in the overall assessment process and thus in finding a site with the best possible safety in accordance with the Site Selection Act. This influence is not seen at the level of repository operation, as is traditionally viewed with regard to human factors, but rather in the design of the repository – particularly in the decision-making process and the definition of the system's fundamental design parameters. Thus, considerations of human reliability are also of utmost importance for the passively safe system of a repository, especially in the current phase of the search and evaluation process. Given that severe accidents in man-made technological systems depend heavily on the reliability of human behaviour, not only in operation but also in design and conceptualization, considering human reliability aspects is essential for a successful site selection (Straeter 2019).
This article first provides an overview of the technical, organizational, cross-organizational, and individual aspects of human reliability that are crucial in the modelling phase of radioactive waste management. Human aspects include variations in the selection of models, the definition of input parameters, and the interpretation of results as individual or group efforts. Based on a review of relevant guidelines on the topic (VDI 4006), suggestions are presented for dealing with these human factors at different levels. The results of a study on the importance of these factors are presented, which was carried out in the context of the TRANSENS project in cooperation between the University of Kassel and the TU Clausthal in order to demonstrate the importance of the human factor in modelling. Overall, based on these considerations, the AHRIC (Assessment of Human Reliability in Concept phases) method is proposed to assess the negative effects of trust issues in the site selection work processes and to derive mitigating measures (Fritsch, 2025). The method applies to all work processes of the key actors.
- Preprint
(733 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on sand-2025-3', Anonymous Referee #1, 14 Aug 2025
- AC1: 'Reply on RC1', Oliver Straeter, 24 Oct 2025
-
RC2: 'Comment on sand-2025-3', Anonymous Referee #2, 07 Oct 2025
The manuscript addresses the important subject of human reliability aspects in the initial planning and decision stages of selection for nuclear disposal waste. It adequately outlines the individual, collective and organisational as well as cross-organisational factors which influence human reliability in planning and operational processes. In addition, it highlights the aspect of cognitive bias and trust in the respective socio-cognitive processes in the context of site selection. The authors provide recommendations for the assessment and management of human reliability aspects in alignment with the VDI 4006 guideline, highlight the importance and tight coupling of human reliability with system safety and propose a method (AHRIC: Assessment of Human Reliability in Concept phases) for the pre-emptive management of human reliability in critical modelling processes of site selection. The conceptual-methodological part of the manuscript is complemented by the presentation of a case study derived from the TRANSENS project that should provide insights of the suitability and usefulness of the approach for practice.
The manuscript is overall well structured and for the most comprehensible and insightful. However, there are two critical shortcomings that need to be considered and revised by the authors.
First, the manuscript claims to present the results of the deployment of AHRIC in the TRANSENS project. The manuscript provides rather conceptual information as well general procedural (i.e. deployment principles) and intended outcomes of the method than presents specific results. I missed information on the procedure of assessment, on the demographics of participants and the context of measurement. In addition, the manuscript only partly provides an overview of the results on the three postulated dimensions of the method and how these can be associated with the quality of decisions in site selection as a tangible outcome. It is further more not clear, whether the example provided in Section 3.2 and Fig. 6 is a fictional one (for the sake of demonstration) or results that correspond to the suggested empirical evaluation of the method.
Second, there is sparse information on the potential vulnerability of the method against cognitive biases of the participants themselves. As a self-reporting, self-assessment method AHRIC may be susceptible itself on the perpetuation of the same cognitive biases which it claims to assess and counteract. It is sufficient that the authors provide more elaborate information on the method administration and the usage process in order to provide adequate information of the respective available safeguarding aspects and the validity of the method. Further on, the postulated verification modality of the effectiveness of the method (lines 228-229) reads more like “wishful thinking” and not as a valid outcome attained through observation and measured with solid indicators. No information whatsoever has been provided on such effectiveness measures deployed in the TRANSENS project.
In addition, there are some minor aspects that need to be amended:
- Some critical terms are not clearly described and/or rely highly on the respective sources and thus, may be inadequately understood or even misunderstood by the readers (goal-oriented vs task-oriented behaviour, 38-40)
- The method need to be presented in more details. For instance: how many items for each bias/categories in the AHRIC-M/AHRIC-S; availability of statistical quality criteria (e.g. reliability, validity)
- Some abbreviations appear without previous complete description (line 159; line 169)
- Occasionally the terms used appear to be vague/not clear indicating spelling mistakes (line 133)
- Occasionally language use is more colloquial than appropriate in a scientific paper (line 54: “These people”)
- Some rather strong statements would need to be supported by respective citation (if applicable) (line 55-56)
- The term “cognitive biases” is a rather well documented and established term. Seminal researchers on the subject e.g. Tverksy and Kahneman should be cited in the first generic use of the term (line 67)
- In general, an additional proofreading is highly recommended as some sentences are slightly incomprehensible something that is not beneficial for the manuscript.
- AC2: 'Reply on RC2', Oliver Straeter, 24 Oct 2025
Status: closed
-
RC1: 'Comment on sand-2025-3', Anonymous Referee #1, 14 Aug 2025
Title: How to handle uncertainties in modelling due to human reliability issues for nuclear disposals
Authors: O. Straeter, F. Fritsch
Review
General comments:
The paper is addressing the various types of human bias encountered in the safety case for nuclear waste repositories, it shows ways to identify them and avoid them.
In table 1, a few more modelling steps could be added that are regularly required: completeness of data survey (when is enough enough); details and type of documentation (to support the traceability of decisions); implicit assumptions (often hidden in larger models / input files / databases / code packages). It is also advised to put “result evaluation” above “result interpretation”.
Concerning the parameterization of models, it is dangerous (but often observed) that modellers just use the databases coupled to code packages that they have paid for without checking the origin and quality of these parameter sets. Another point with parameterisation is the often individually biased selection of process and their uncertainty to be discussed when it comes to the categories “unknown knowns” and “unknown unkowns”, i.e. how to deal with missing understanding and parameters (ignorance vs. uncertainty).
Specific comments:
In figure 2 (although taken from another activity) ”competence” should also be fed from “evaluation”, in many areas benchmarking (between codes, and also with respect to real field data or experimental results) are well-respected factors to generate trust. Examples are the huge international consortia behind DECOVALEX or JOSA.
The work explains in great detail the biases linked to group structures and behaviour. However, it should also be mentioned that in many circumstance a “four-eye-principle” could on the contrary be beneficial to steps in modelling issues. This is closely connected to the role of “external peers” and review processes.
Line 159: URS should be spelled out and a link to the project given.
Lines 195ff: An example would be very beneficial for the reader to understand what the entries in Figure 5 (which, by the way, could easily be turned into a table) are really meaning; currently this is very generic, data (response scale) interpretation without associated statements is not clear. Figure 6 does obviously focus already on the next step; it is not explained where the p (Success) values are coming form – and why there are only four distinct numbers in addition to 1 and 0. In addition, the computation of the numbers in the right-most column is unclear.
Line 250: What is the meaning of “heurism” in that context?
Technical comments:
Lines 223-226 are strongly redundant to lines 210+. Should be merged.
Line 244: “and” instead of “ans”
An acknowledgement of TRANSENS is missing.
Citation: https://doi.org/10.5194/sand-2025-3-RC1 - AC1: 'Reply on RC1', Oliver Straeter, 24 Oct 2025
-
RC2: 'Comment on sand-2025-3', Anonymous Referee #2, 07 Oct 2025
The manuscript addresses the important subject of human reliability aspects in the initial planning and decision stages of selection for nuclear disposal waste. It adequately outlines the individual, collective and organisational as well as cross-organisational factors which influence human reliability in planning and operational processes. In addition, it highlights the aspect of cognitive bias and trust in the respective socio-cognitive processes in the context of site selection. The authors provide recommendations for the assessment and management of human reliability aspects in alignment with the VDI 4006 guideline, highlight the importance and tight coupling of human reliability with system safety and propose a method (AHRIC: Assessment of Human Reliability in Concept phases) for the pre-emptive management of human reliability in critical modelling processes of site selection. The conceptual-methodological part of the manuscript is complemented by the presentation of a case study derived from the TRANSENS project that should provide insights of the suitability and usefulness of the approach for practice.
The manuscript is overall well structured and for the most comprehensible and insightful. However, there are two critical shortcomings that need to be considered and revised by the authors.
First, the manuscript claims to present the results of the deployment of AHRIC in the TRANSENS project. The manuscript provides rather conceptual information as well general procedural (i.e. deployment principles) and intended outcomes of the method than presents specific results. I missed information on the procedure of assessment, on the demographics of participants and the context of measurement. In addition, the manuscript only partly provides an overview of the results on the three postulated dimensions of the method and how these can be associated with the quality of decisions in site selection as a tangible outcome. It is further more not clear, whether the example provided in Section 3.2 and Fig. 6 is a fictional one (for the sake of demonstration) or results that correspond to the suggested empirical evaluation of the method.
Second, there is sparse information on the potential vulnerability of the method against cognitive biases of the participants themselves. As a self-reporting, self-assessment method AHRIC may be susceptible itself on the perpetuation of the same cognitive biases which it claims to assess and counteract. It is sufficient that the authors provide more elaborate information on the method administration and the usage process in order to provide adequate information of the respective available safeguarding aspects and the validity of the method. Further on, the postulated verification modality of the effectiveness of the method (lines 228-229) reads more like “wishful thinking” and not as a valid outcome attained through observation and measured with solid indicators. No information whatsoever has been provided on such effectiveness measures deployed in the TRANSENS project.
In addition, there are some minor aspects that need to be amended:
- Some critical terms are not clearly described and/or rely highly on the respective sources and thus, may be inadequately understood or even misunderstood by the readers (goal-oriented vs task-oriented behaviour, 38-40)
- The method need to be presented in more details. For instance: how many items for each bias/categories in the AHRIC-M/AHRIC-S; availability of statistical quality criteria (e.g. reliability, validity)
- Some abbreviations appear without previous complete description (line 159; line 169)
- Occasionally the terms used appear to be vague/not clear indicating spelling mistakes (line 133)
- Occasionally language use is more colloquial than appropriate in a scientific paper (line 54: “These people”)
- Some rather strong statements would need to be supported by respective citation (if applicable) (line 55-56)
- The term “cognitive biases” is a rather well documented and established term. Seminal researchers on the subject e.g. Tverksy and Kahneman should be cited in the first generic use of the term (line 67)
- In general, an additional proofreading is highly recommended as some sentences are slightly incomprehensible something that is not beneficial for the manuscript.
- AC2: 'Reply on RC2', Oliver Straeter, 24 Oct 2025
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 297 | 46 | 23 | 366 | 33 | 71 |
- HTML: 297
- PDF: 46
- XML: 23
- Total: 366
- BibTeX: 33
- EndNote: 71
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Title: How to handle uncertainties in modelling due to human reliability issues for nuclear disposals
Authors: O. Straeter, F. Fritsch
Review
General comments:
The paper is addressing the various types of human bias encountered in the safety case for nuclear waste repositories, it shows ways to identify them and avoid them.
In table 1, a few more modelling steps could be added that are regularly required: completeness of data survey (when is enough enough); details and type of documentation (to support the traceability of decisions); implicit assumptions (often hidden in larger models / input files / databases / code packages). It is also advised to put “result evaluation” above “result interpretation”.
Concerning the parameterization of models, it is dangerous (but often observed) that modellers just use the databases coupled to code packages that they have paid for without checking the origin and quality of these parameter sets. Another point with parameterisation is the often individually biased selection of process and their uncertainty to be discussed when it comes to the categories “unknown knowns” and “unknown unkowns”, i.e. how to deal with missing understanding and parameters (ignorance vs. uncertainty).
Specific comments:
In figure 2 (although taken from another activity) ”competence” should also be fed from “evaluation”, in many areas benchmarking (between codes, and also with respect to real field data or experimental results) are well-respected factors to generate trust. Examples are the huge international consortia behind DECOVALEX or JOSA.
The work explains in great detail the biases linked to group structures and behaviour. However, it should also be mentioned that in many circumstance a “four-eye-principle” could on the contrary be beneficial to steps in modelling issues. This is closely connected to the role of “external peers” and review processes.
Line 159: URS should be spelled out and a link to the project given.
Lines 195ff: An example would be very beneficial for the reader to understand what the entries in Figure 5 (which, by the way, could easily be turned into a table) are really meaning; currently this is very generic, data (response scale) interpretation without associated statements is not clear. Figure 6 does obviously focus already on the next step; it is not explained where the p (Success) values are coming form – and why there are only four distinct numbers in addition to 1 and 0. In addition, the computation of the numbers in the right-most column is unclear.
Line 250: What is the meaning of “heurism” in that context?
Technical comments:
Lines 223-226 are strongly redundant to lines 210+. Should be merged.
Line 244: “and” instead of “ans”
An acknowledgement of TRANSENS is missing.