the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 23 Sep 2026
Nonlinear Event-Space Analysis of Radon and Meteorological Parameters for Seismic Anomaly Identification
Abstract. Radon (²²²Rn) dissolved in soil gas serves as a well-established geochemical tracer for crustal deformation and seismic precursory research. However, radon concentration time series recorded at near-surface monitoring stations are systematically contaminated by meteorological parameters, principally atmospheric temperature, humidity, barometric pressure, and precipitation and rigorous characterization of these influences is a prerequisite for reliable earthquake precursor signals. Conventional linear correlation analyses are fundamentally limited in their capacity to capture transient, event-driven, and nonstationary radon responses that are most likely to encode seismically significant precursory signals. This study presents a comprehensive nonlinear event-space analytical framework applied to 11,572 hourly soil-gas radon data recorded at the Jiaosi monitoring station in northeastern Taiwan spanning January 2021 to April 2022. The methodology integrates time-series first-order differencing to isolate event-based change signatures, classification of hike, drop, and normal meteorological change states, and Locally Weighted Scatterplot Smoothing (LOWESS) for nonparametric characterization of radon response curves in change space. In the linear amplitude domain, temperature (r = 0.500), atmospheric pressure (r = −0.384), and humidity (r = 0.158) emerge as the dominant meteorological correlates of radon variability. Critically, these correlations weaken markedly when evaluated in the nonlinear event domain, revealing that temperature, pressure, and humidity primarily modulate quasi-stationary and periodic background radon signals rather than transient excursions. In striking contrast, rainfall which exhibits negligible linear correlation with radon reveals a pronounced U-shaped nonlinear response in event space. The nonlinear rainfall response remained stable across all threshold configurations and accumulation windows, confirming the robustness of the event-space framework. in the event domain that is consistent across all precipitation accumulation windows (24-hour, 72-hour, and 240-hour), and is mechanistically attributable to the rainfall capping effect, piston effect, and soil drainage dynamics. This pattern constitutes the primary source of false positives in radon-based precursor identification. The event-space LOWESS framework substantially enhances the discriminability of meteorologically driven noise from potentially geophysically significant anomalies, providing a computationally accessible complement to Empirical Mode Decomposition–Hilbert-Huang Transform (EMD-HHT) approaches. The methodology has direct implications for operational earthquake precursor detection frameworks and multi-station radon monitoring networks.
- Preprint
(2316 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 04 Nov 2026)
- RC1: 'Comment on egusphere-2026-5420', Anonymous Referee #1, 04 Oct 2026 reply
-
RC2: 'Comment on egusphere-2026-5420', Anonymous Referee #2, 05 Oct 2026
reply
The article presents the results of an analysis of time series representing radon concentrations and meteorological parameters (pressure, temperature, humidity, and precipitation) measured in Taiwan from January 2021 to April 2022. The data consist of synchronous time series with a one-hour sampling interval and a total length of 11,572 data points. The aim of the analysis is to identify precursor effects in radon behavior prior to earthquakes and to distinguish these effects from the influence of meteorological factors that could be misinterpreted as precursors.
From a methodological standpoint, this article proposes an alternative approach to analyzing the same data previously examined using Empirical Mode Decomposition via the Hilbert-Huang transform (https://doi.org/10.1007/s12665-025-12100-y). That earlier study demonstrated that the primary challenge in accounting for the impact of meteorological parameters on the radon time series arises from heavy precipitation; this influence generates radon amplitude anomalies that resemble those preceding earthquakes. It was shown that the effects of pressure and temperature are concentrated primarily within two narrow frequency bands corresponding to 12-hour and 24-hour variation periods. These frequency bands correspond to two specific IMF levels in the Huang decomposition, which were previously recommended for exclusion from the analysis. Consequently, the focus shifted to examining radon variations within IMF levels 5 through 8, excluding both the highest-frequency levels (which carry noise) and the lowest-frequency levels (where variation periods fall outside the frequency range associated with radon precursors). Ultimately, the previously proposed approach could not automate the task of distinguishing between radon anomalies caused by precipitation and those potentially interpretable as earthquake precursors, thereby necessitating expert analysis.
This article proposes a modification to a previously introduced method based on empirical mode decomposition. This modification relies on the use of the LOWESS local polynomial regression method. As a preliminary step, the Pearson correlation matrix was calculated for the increments of radon time series and meteorological parameters (excluding precipitation). Overall, the presentation of the material is overly concise, and many details remain unclear. Further clarification is required regarding the following points:
- The Hilbert-Huang decomposition is not employed in the article; therefore, the method is independent and bears no relation to EMD-HHT, contrary to the statement in the abstract. The conclusion reiterates the claim that the proposed method complements EMD-HHT. This assertion requires further elaboration, specifically regarding how EMD-HHT can be used in conjunction with the LOWESS method.
- Regarding the plots in Fig. 3: what do the variables PC1, PC2, and PC3 in row 7 represent? The label indicates "reduced space," yet the text provides no definition for this term.
- The definition of "Event Classification" includes a constant of 2, measured in "units/hr"—what does this represent? It might refer to a standard deviation unit of the time series after conversion to increments. Regardless of whether this interpretation is correct, these units must be explicitly defined.
- The article fails to specify how the effects of pressure, temperature, and humidity on radon levels are compensated for. The Pearson correlation matrix in Fig. 4 offers no information regarding the compensation method for these meteorological factors. It remains unclear whether compensation occurs before or after curve smoothing. It should be noted that compensation via regression across the entire dataset is a standard linear operation. Since these meteorological parameters are interrelated, the compensation procedure should be multivariate, accounting for their mutual correlations. Furthermore, their influence is frequency-dependent—extending beyond the 12- and 24-hour periods to a broader frequency range—and is non-stationary in nature. Therefore, proper compensation should be performed using a multidimensional frequency-domain filter that accounts for mutual cross-coherences within a sliding time window.
- It is necessary to explain in greater detail exactly what is plotted on the axes of the graphs in Fig. 5; the text must explicitly define what "DeltaMeteorology(Standardized)" represents—currently, one can only guess.
Citation: https://doi.org/10.5194/egusphere-2026-5420-RC2 -
RC3: 'Comment on egusphere-2026-5420', Anonymous Referee #3, 15 Oct 2026
reply
1 General comments
The title “Nonlinear Event-Space Analysis of Radon and Meteorological Parameters for Seismic Anomaly Identification” is misleading as it suggests the application of a method to identify seismically induced anomalies. Alas, no such anomalies are presented in your present work. The reader expects that you follow the workflow, which is nicely outlined in Figure 3, step-by-step. At the very end one would expect a synoptic summary plot which includes the original radon timeseries, the first difference, LOWESS smoothed signal, the meteorological component, “radon variations with meteorological influences minimized”, identified anomalies, and finally a panel showing rain and seismic activity. For the latter I suggest to plot “seismic energy density” (see Wang, 2007; Manga & Wang 2010) instead of magnitudes. You might also want to add a panel with the “instanteneous energy” from Chowdhury et al. (2025) to discuss pros and cons of different non-linear analysis approaches. In conclusion I suggest major revisions of your manuscript.
2 Specific comments
2.1 AI
Please indicate which figures, tables or text have been generated with support of AI and which tools have been used.
2.2 Event classification thresholds
I have problems to understand the logic behind the uniform event classification thresholds of +/- 2 units/hr. Table 3 indicates thresholds as 2°C/hr for temperature, 2 mbar/hr for pressure, 2 %/hr for humidity, and 2 mm/hr for rainfall. Suppose you measured the pressure in the unit Pa/hr, would you then set the threshold to 2 Pa/hr? I could understand a selection based on the standard deviation (or a similar measure) of the meteorological parameters, but this is obviously not the case. So, what is the justification for your threshold selection and what are the physical implications.
2.3 Selected Literature
In the literature review important works in USA and Island are missing. Give reference for the Tashkent case. Give reference for tidal effects on radon (line 163).
In the discussion of your results there is a strong focus on Arora et al. (2017) and Chowdhury et al. (2025), which you cite 28 and 23 times, respectively. It would be interesting to compare your findings with other studies, e.g., Tommasone Pascale et al. (2015) to name just one.
3 Technical corrections
- Avoid repetitions, e.g., line 92/93 and 100/101 and line 426-428 and 475-477 “This implies…”
- Explain abbreviations at their first appearance in the text, i.e., “IMF” in line 63, “IE analysis” in line 173
- Line 310 “metrology” is the science of measurements, here you mean “meteorological” I suppose
- Line 416 “radon” instead of “radar”
- Line 419/420 “The high rainfall can occur with either increased or decreased rainfall, …” the wording might be misleading
- Line 465 “radon” instead of “radar”
- Figures: the font size of Figs. 2, 4, 6, 7, 11, 12 is too small
- 1b add a km-scale
- 4 right: add units to X- and Y-axis
- 8, 9 add units of X- and Y-axis
- 11 add unit of Y-axis
- Figure captions: Figure 3 shows the whole workflow, not only the pre-processing; Figure 6 explain RA
- The following references are listed, but not referenced in the text body: Dobrovolsky, Kulali, Papastefanou, Schumann, Ulomov, Walia 2003.
References
- Chowdhury, S., Walia, V., Lin, S.J., Fu, C.C., Lee, H.F., 2025. Non-linear analysis of soil 222Rn time series recorded at Jiaosi in north-east Taiwan for possible earthquake precursor study. Environmental Earth Sciences 84(3). doi.org/10.1007/s12665-025-12100-y. 77.
- Tommasone Pascale, F., Carbone, P., De Francesco, S., Cuoco, E., Tedesco, D., 2015. Rainstorm-induced soil 222Rn concentration spikes observed in Southern Italy. Environmental Earth Sciences 73(12), 8177–8187. doi.org/10.1007/s12665-014-3976-0.
- Wang, C.Y., 2007. Liquefaction beyond the Near Field. Res. Lett. 78(5), 512–517.
- Wang, C.Y. & Manga, M., 2010. Earthquakes and water. Springer, Berlin-Heidelberg.
Citation: https://doi.org/10.5194/egusphere-2026-5420-RC3 -
CC1: 'Comment on egusphere-2026-5420', Antonino D'Alessandro, 01 Nov 2026
reply
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-5420/egusphere-2026-5420-CC1-supplement.pdf
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 365 | 99 | 75 | 539 | 75 | 59 |
- HTML: 365
- PDF: 99
- XML: 75
- Total: 539
- BibTeX: 75
- EndNote: 59
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript egusphere-2026-5420 regards a comprehensive, nonlinear event-space analytical framework to evaluate hourly soil-gas radon time series recorded in Northeast Taiwan from January 2021 to April 2022. To isolate event-based change signatures, the applied methodology combines first-order differencing of the time series with the classification of meteorological states into hikes, drops, and normal conditions. Nonparametric characterization of radon response curves within the change space is then achieved using Locally Weighted Scatterplot Smoothing (LOWESS). The key findings demonstrate that: (i) linear domain correlations provide a robust baseline; (ii) radon variability is primarily correlated with temperature (r = +0.500), atmospheric pressure (r = −0.384), and soil humidity (r = +0.158); (iii) meteorological parameters drive the quasi-periodic background variability; and (iv) rainfall acts as the primary nonlinear driver of transient radon fluctuations.
I think the results reported in the manuscript could be noteworthy and very interesting for the scientific community studying soil-gas radon, as potential seismic precursor and, more precisely, the influence of meteorological parameters on soil-gas radon measurements. After having carefully read it, I suggest accepting with “minor revisions”.
Main comment
Line 605. The conclusions paragraph I think it is quite too long and maybe redundant, therefore, for improving its readability, I suggest rewriting it to make shorter.
Minor comments
I suggest enlarging the font of text to improve the readability of figures 1, 2, 3, 4, 5, 6, 7, 11, 12.
Figure 1. Improve the captions of map and b and scheme c. For map a add in the caption the legend of acronyms HC SJ, JS and HH; for map b add the graphic scale, locate the radon monitoring station, and enlarge the font of text; for scheme c enlarge the font of text and add in the caption the definition of the radon network units.
Figure 2. Add the date (months) in the X-axis.
Figure 4. Add the unit of measure for radon.
Figure 8. Typo “Temprature”, add the unit of measure for radon change.
Figure 10. Define in the caption text the terms normal, hike, drop.
Figure 11. Add the unit of measure for radon.