Short-Term Temporal Variability of Radon in Finnish Dwellings and the Use of Temporal Correction Factors
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThanks for submitting the manuscript “Short-term temporal variability of radon in Finnish dwellings and the use of temporal correction factors”.
The manuscript offers valuable insights into the reliability of short-term measurements in Finnish homes and the application of temporal correction factors.
Please find enclosed the PDF attachment for my queries about novelty, statistical rigor, quantification and validation of key assumptions.
Comments for author File: Comments.pdf
Author Response
Comment 1: Please add a discussion about the novelty of this research and add more information about the hazards of radon för human health and its impacts on the environment.
Reply 1: Our manuscript is intended for publication in the special issue Atmospheric Radon Concentration Monitoring and Measurements (2nd Edition) of the journal Atmosphere, in which each article addresses the topic of radon. Therefore, we consider it unnecessary for every article in the special issue to extensively discuss the health effects of radon and its general occurrence in our living environment.
On the other hand, as suggested by Reviewer #1, it is important to consider in more detail the novelty of our research. We have added the following text to the discussion section:
"The novelty of this study lies in the probabilistic method applied, which allowed for the estimation of the proportion of false negatives in the results, as well as the determination of action levels for different measurement durations using the actual temporal distribution of the results, rather than, for example, an assumed log-normal distribution. In the method presented by Tsapalov and Kovler, the action level is set based on log-normally distributed temporal variation in such a way that the reference level is not exceeded in 95% of the measurement locations. When comparing the action levels estimated in this study—where a 5% false negative rate is accepted (Table 8)—with the results of Tsapalov and Kovler, it is observed that the action levels from this study and those reported by Tsapalov and Kovler for unoccupied dwellings are very similar. For occupied dwellings, Tsapalov and Kovler present lower action levels, suggesting greater temporal variation in their dataset." (Line 559 onwards)
Comment 2: The calibration of RadonEye Plus2 monitors was performed at 300, 700, and 1,500 Bq/m3. Some homes can exhibit low (<50 Bq/m3) or extremely high (>2000 Bq/m3) levels of radon. How would you verify such levels of concentration?
Reply 2: This is a good question and often a topic of discussion in metrology—namely, whether it is justified to extrapolate calibrations. According to standard IEC 61577-4:
“The [System for Test Atmospheres with Radon] STAR should be capable of being used for testing equipment with values of radon volumic activity varying from 1/3 to 2/3 of each range of measurement indicated on the equipment under test. A value situated between 100 Bq/m³ and 1 000 Bq/m³ can be adopted for the activity of ²²²Rn […], enabling one identical operating point to be obtained for all the equipment under test. Other values can be agreed upon between the operator of the STAR and the manufacturer of the instrument.”
It can thus be stated that at least the calibration concentrations of 300 and 700 Bq/m³ fulfilled the standard's requirement for a calibration concentration at the lower end of the radon concentration range. The highest calibration point used in this calibration did not comply with the standard testing conditions (3000–6000 Bq/m3). However, as referenced in the manuscript, such a calibration for the devices had already been performed earlier and reported in the article:
Turtiainen, T.; Kojo, K.; Laine J-P.; Holmgren, O.; Kurttio, P. Improving the assessment of occupational exposure to radon in above-ground workplaces. Radiat. Prot. Dosim. 2021, 196 (1–2): 44–52. https://doi.org/10.1093/rpd/ncab127.
Comment 3: The Spearman test shows a slight negative correlation between AARC and GSD for 2-day measurements, but Pearson 's r suggests no linear association. Such discrepancies should be resolved.
Reply 3: Agreed. I added the missing word ‘linear’ on line 343:
“For the same data set, results of the Pearson correlation indicated that there is no linear association between AARC and GSD (r = 0.011, p = 0.938).”
My assumption is that the slight negative correlation observed in the 2-day measurement results by Spearman’s test is a fluke, as it cannot be seen in the 1- or 3-day measurements. It should be noted that the correlation was observed specifically between the AARC and the variation of the correction factors, not between the annual mean and the value of the correction factors. In the latter case, the value of the temporal correction factor would depend on the annual mean radon concentration, and therefore, a different correction factor value would need to be applied to different measured concentration. The fact that the variation in correction factors may be slightly greater at low concentrations does not prevent combining measurement results from different households for further analysis.
I have added the following explanatory sentences: “Even if the observed correlation is genuine, it pertains specifically to the relationship between the annual mean and the variability of the correction factors, not between the annual mean and the magnitude of the correction factors. A slightly greater variability in correction factors at lower annual mean concentrations does not preclude the aggregation of correction factor values obtained from different households for further analysis” (Line 349).
Comment 4: The 1% false-negative threshold (90-100 Bq/m3) is arbitrary. Please clarify what the acceptable lung cancer thresholds are to better align with radon protection principles.
Reply 4: This is true—the given probabilities for false-negatives (0.5%, 1%, 2%, 3%, 5%) are illustrative examples and are not based on any established standard. The detection limit of a measurement is typically set at the 95% confidence level. In other words, it is estimated that a measurement result below a certain threshold is, with 95% certainty, truly below that level—allowing for a 5% false negative rate.
When setting an action level, however, no such standard exists. The action level is always a matter of policy or agreement, influenced by various factors including scientific evidence, political considerations, and health impacts. I have revised the manuscript beginning at line 560 to state:
“Currently, no established standard defines the acceptable proportion of false negatives when determining an action level. For some individuals, a 5% probability of a false negative may be entirely acceptable, whereas for others—particularly those who work from home or have small children—only the most reliable result is acceptable.”
The cancer risk from radon is generally considered to follow a linear no-threshold (LNT) dose–response relationship. In other words, all radon exposure contributes to an increased relative risk of lung cancer. Because smokers have a higher baseline risk for lung cancer, their absolute risk from the same radon exposure is also greater than that of non-smokers. The EU reference level of 300 Bq/m³ is based on a policy agreement influenced by factors such as the cancer risk associated with radon exposure, typical indoor radon concentrations in the building stock, estimated mitigation costs, and previously used reference levels. It is therefore not an established lung cancer threshold. The probabilities given for false negatives specifically describe the likelihood of exceeding the reference level.
Comment 5: The claim that September is a "poor" month for measurements relies on a single year's data. A multi-year analysis is needed to justify excluding September.
Reply 5: Thank you for this observation. While reflecting on this, I realized that the measurement month was actually September 2023—not 2024—since the measurements were conducted from summer 2023 to summer 2024. I have corrected the text starting from line 501. September 2023 and September 2024 were very similar—both warm and rainy—so this does not affect the text other than the numerical values. I have made the following corrections:
Line 326: In contrast, September 2023 appeared to be a relatively poor month for measurements, as the large variability of the correction factors only decreases at the very end of the month.
Line 602: The study revealed significant radon level fluctuations in September, indicating that this period warrants further investigation in future years and, if necessary, reconsideration for inclusion in the official measurement season.
Comment 6: Related work on air pollution concentrations is recommended to be cited as the recent development, i.e. Spectrochim Acta B. 225, 107124 (2025).
Reply 6: It is true that the concentrations of indoor air pollutants in buildings can vary significantly over time, meaning that this phenomenon is not unique to radon gas. For example, the concentration of mold spores in indoor air can fluctuate by several orders of magnitude. To keep the number of references reasonable, we have limited the citations in this article to those specifically addressing radon.
Comment 7: The study recommends the geometric mean for GM estimation but does not address how its sensitivity to skewness affects high-radon homes. A robustness check is needed to validate this choice.
Reply 7: This likely refers to the statement on line 333: “The probability distributions of kC values evaluated for different measurement durations and starting day or month were strongly right-skewed. Therefore, the GM was chosen to describe the average value, and the GSD, which is definitionally multiplicative and thus relative, was chosen to describe the magnitude of dwelling-specific relative variation.” Furthermore, Line 413 states: “During this measurement season, the distribution of three-day correction factors (N = 12,073) is strongly right-skewed, yielding median, GM, and AM correction factors of 0.901, 0.933, and 1.024, respectively (Figure 6).”
We considered the same issue (usage of log-normal distribution in simulations), as the Kolmogorov–Smirnov test for the logarithmic transformations indicated that the log-transformed values exhibited a statistically significant difference from normal distribution. However, the observed effect size D was small (0.066), suggesting that the deviation of the distribution from normality was minor. Therefore, there was no reason to avoid reporting the GM and GSD in Table 4; rather, this enabled comparison of the current results with those of previous studies.
Since the actual distribution of correction factors deviated from a log-normal distribution—particularly at the extremes, as indicated by the Q–Q plot—we used empirical distributions derived from the data in all simulations, rather than assuming log-normality. This is certainly a very robust method. This approach is explained beginning at line 255: “For this purpose, correction factor values derived from three-day measurements initiated between October 1 and May 12 (N = 12,073) were aggregated, and an empirical probability distribution was constructed in increments of 0.1 per mille.”
Reviewer 2 Report
Comments and Suggestions for Authors1.Consider increasing the number of references appropriately to make the article more convincing.
2.In the “Introduction” section, although the indoor radon problem in Finland and its importance have been summarized, some comparisons on the status of radon contamination on a global scale, as well as the latest research results on the effects of radon on human health, could be added to further enhance the depth and breadth of background information in the paper.
3.In the conclusion section, you can add a summarizing chart, such as a flowchart or a schematic diagram, to visualize the main findings of the study as well as the direction of future research, so that readers can quickly grasp the core content of the paper.
4.The paper mentions methods of calculating correction factors, but does not adequately explain why certain methods (e.g., geometric mean) are superior to others. It is recommended that the theoretical basis for the selection of these methods be detailed in the methods section and their advantages and disadvantages be further explained in the discussion section.
5.Differences in radon concentrations between different building types. It is suggested that the "Materials and Methods" section supplement the discussion of sample selection and explain how data analysis can mitigate potential biases.
Author Response
Comment 1: Consider increasing the number of references appropriately to make the article more convincing.
Reply 1: We have added two key references that demonstrate the significance of radon for health in Europe and worldwide. A more extensive literature review would certainly be a valuable addition to our work, especially in the Discussion section. Therefore, we have added a comparison of the action values (starting at line 565) obtained in this study and those reported earlier (Tsopalov and Kovler). We hope you agree on the usefulness of this comparison and that it contributes to the reporting of this study.
Comment 2: In the “Introduction” section, although the indoor radon problem in Finland and its importance have been summarized, some comparisons on the status of radon contamination on a global scale, as well as the latest research results on the effects of radon on human health, could be added to further enhance the depth and breadth of background information in the paper.
Reply 2: Our manuscript is intended for the special issue "Atmospheric Radon Concentration Monitoring and Measurements (2nd Edition)" of the journal Atmosphere, where every article addresses some aspect of radon. Therefore, we believe it is unnecessary for each contribution to extensively discuss radon's health effects and its general prevalence in our environment. Nonetheless, we have added two key references that highlight the significance of radon for health in Europe and worldwide.
Comment 3: In the conclusion section, you can add a summarizing chart, such as a flowchart or a schematic diagram, to visualize the main findings of the study as well as the direction of future research, so that readers can quickly grasp the core content of the paper.
Reply 3: I completely agree that illustrative images are an excellent addition to the articles. However, I must apologize, as I am not very skilled at creating various graphics, to be honest. Therefore, I have strived to keep the Conclusions section as brief as possible, so that it can be read quickly.
Comment 4: The paper mentions methods of calculating correction factors, but does not adequately explain why certain methods (e.g., geometric mean) are superior to others. It is recommended that the theoretical basis for the selection of these methods be detailed in the methods section and their advantages and disadvantages be further explained in the discussion section.
Reply 4: Section 3.4 compares the reliability of different correction factors. Table 6 shows the proportion of measurement results that align with the true annual average, using accuracy criteria of ±10%, ±15%, ±20%, etc., when applying various calculations for the generic correction factor value. It is clear from the table that the highest percentage of annual average estimates meeting the criteria is achieved using the kC(median) and klnC(regression) generic correction factors. The results in Table 6 were obtained through Monte Carlo simulation and actual empirical distributions of the correction factor values. The method is explained in lines 255–266 in Section 2.4.
Comment 5: Differences in radon concentrations between different building types. It is suggested that the "Materials and Methods" section supplement the discussion of sample selection and explain how data analysis can mitigate potential biases.
Reply 5: That's a good point. We could not recruit any homeowners form apartment buildings. In detached houses, semi-detached houses, and terraced houses, the average radon concentration is clearly higher than in apartment buildings. The comparison was made with similar homes in the national survey, and the dwellings in this study corresponded well to the nationally estimated average concentrations. Consequently, these results pertain only to detached, semi-detached, and terraced houses—not apartment buildings.
This study does not provide new insights into the variations in radon levels in apartment buildings. Since apartment building walls are typically very tall (causing wind-induced pressure gradients) and such buildings can exhibit a strong stack effect, it is entirely possible that the radon concentration variation in ground-facing apartment units is even more pronounced than what has been investigated here. STUK has a case study site where we examined the spatial and temporal variations of an apartment on the ground floor of an apartment building. The instantaneous radon concentration depended strongly on wind direction and varied between 38 and 2800 Bq/m³. However, these results have not been published anywhere, so unfortunately, we cannot reference them.
I reorganized and revised the text in Materials and Methods as follows: …Southern Finland is slightly underrepresented in this study, while Western and Inland Finland are overrepresented. On the other hand, the years of construction, foundation types, ventilation methods, and radon levels represent the national housing stock excellently. No blocks of flats were included in this study; therefore, the results apply only to detached, semi-detached, and terraced houses.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe work is interesting, but it would be more opinion-forming if additional studies could be conducted throughout the year - not just during specific seasons. This would provide better insight into how radon levels change throughout the year. Alternatively, one could refer to source information that would bring more understanding to this area. Do the authors know of any longer-term radon emissions and their correlation with shorter-term studies? In my opinion, these studies cannot be easily extended to other locations due to too many variables influencing this phenomenon.
The study provides a detailed analysis of short-term variability in radon concentrations, which is not extensively covered in existing literature, especially within the context of Finnish homes. This specificity to regional conditions is significant, as radon behavior can vary widely based on geographic and climatic factors. The findings regarding significant radon level fluctuations in September challenge the traditional measurement season, suggesting that guidelines may need to be updated. As climate conditions change, the paper addresses a gap concerning the adequacy of existing measurement protocols and offers a data-driven basis for revising these protocols to include more responsive measures. The study attempts to refine the approach to estimating the annual average radon concentration (AARC) through the use of temporal correction factors. This part addresses the lack of tailored correction methods that accommodate short-term variability and produce more accurate estimates of long-term exposure.
The paper evaluates the reliability of short-term radon measurements for assessing the need for remediation, a topic that has not been extensively explored in Finnish contexts. It provides empirical data showing that current short-term methods may not suffice as replacements for long-term measurement techniques, which is crucial for public health considerations. This research contributes to the development of generic correction factors that can be applied to short-term measurements, allowing for better estimates of annual average radon concentrations. It investigates the effectiveness of these factors and proposes improvements, addressing a prevalent issue in the methodology of radon assessments that previous studies may not have adequately considered.
While the study involves measurements from 55 dwellings, increasing the sample size would improve the robustness and generalizability of the results. A larger and more diverse sample could capture a wider range of building types, geographical locations, and demographic factors influencing radon levels. Developing a standardized protocol for the execution of short-term measurements and the interpretation of their results could minimize variability and improve the reproducibility of findings across different studies and contexts.
The conclusions assert that short-term radon measurements (2–5 days) cannot reliably replace established long-term measurement methods. This is supported by the evidence demonstrating that if a maximum false-negative rate of 1% is accepted, the action level for these short-term measurements should be set between 90–100 Bq/m³, indicating that many short-term tests could produce erroneous assessments of radon risk. The year-long measurement conducted in 55 Finnish dwellings allowed for the analysis of hourly radon concentration data and the calculation of the likelihood of erroneous short-term assessments, directly informing this conclusion.
The references cited pertain directly to radon research, measurement methodologies, health implications, and European regulatory frameworks. For instance, references discussing the health risks associated with radon exposure and estimates on lung cancer prevention provide necessary context for the study's importance.
Author Response
Comment 1: The work is interesting, but it would be more opinion-forming if additional studies could be conducted throughout the year - not just during specific seasons.
Reply 1: This is exactly what this study examined. Instead of conducting brief measurements during different seasons, year-long measurements were carried out, with radon concentrations recorded hourly by the measuring device.
Comment 2: While the study involves measurements from 55 dwellings, increasing the sample size would improve the robustness and generalizability of the results. A larger and more diverse sample could capture a wider range of building types, geographical locations, and demographic factors influencing radon levels.
Reply 2: This is true, and a discussion of it begins on line 579. It should be remembered that the hourly, year-long data in this study is considerably more comprehensive than that of any previous research, so there is no reason to withhold these results from publication due to limited data.