the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Jun 2025
Seismo-acoustic and GNSS monitoring of a record-breaking storm in the Black Sea: Evidence of climate change and intensifying natural hazards
Abstract. In August 2024, a devastating storm struck Romania’s Black Sea coast, setting new precipitation records and highlighting the increasing frequency of extreme weather events. This study explores the integration of non-conventional sensors (seismic, GNSS, infrasound, and satellite data) with ERA5 meteorological reanalysis to monitor storm dynamics. High-frequency (>30 Hz) seismic signals captured precipitation, while microseismic bands (0.1–1 Hz) reflected wave-induced ground motion. Infrasound data, analyzed using unsupervised learning, revealed distinct storm phases and showed strong spectral correlation with recorded ground motion, pointing to coupled atmosphere-lithosphere processes induced by the storm. The infrasound array also detected over 1,100 signals in the 0.6–7 Hz band, matching lightning discharges observed by geostationary satellites. GNSS-derived estimates of precipitable water vapor tracked atmospheric moisture buildup and showed clear correlation with intense rainfall, including potential precursory signals days before peak precipitation. This study highlights the value of integrating diverse, non-traditional datasets to enhance the resolution and depth of storm analysis. Their combined use offers a more holistic understanding of storm evolution and supports the development of improved early-warning systems in vulnerable coastal regions.
- Preprint
(3338 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (extended)
-
RC1: 'Comment on egusphere-2025-1842', Anonymous Referee #1, 29 Jul 2025
reply
This manuscript gathers together seismic, infrasound, GNSS and meteorological data for a rainfall episode in Romania in late August 2024. The authors claim that this approach can lead to a “holistic understanding of storm evolution”; however, I can not see in the manuscript which is the contribution of putting together these different kind of datasets to such a better understanding. In my opinion, the manuscript only shows that strong rainfall episodes can be identified in seismic, infrasound and GNSS data, a fact that is widely known.
In my opinion, this work do not deserve publication in a high-rated journal as NHESS: my most important concerns have to do with:
1) The real contribution of analyzing different dataset to a better understanding of the storm evolution is not really explained
2) I think that having access to direct measurements of meteorological parameters is an essential point for this kind of studies . In this contribution, the detailed seismic or infrasonic data is compared to 1-hour long estimations of rainfall derived from large-scale models. Why not to compare each seismic/infrasound station with the closer meteorological site??
3) The authors state in the introduction that the precipitation totals reached values of 120-220 mm, corresponding to an exceptional episode. However, the total precipitation graphics in Figures 4 and 5 show maximum hourly values not reaching 3.5 mm, for a total amount of precipitation of about 30-40 mm, values that would correspond to a modest rainfall episode. I guess that this probably arise from the use of a general precipitation model, but I think that it do not makes sense to discuss the correlation of different datasets to non-representative meteorological variables.
4) Meteorological data is reduced to the ERAS global precipitation model. The possible contribution of wind to seismic and infrasound noise or the relationship between humidity and GNSS-derived water content is not commented at all.
5) The correlation between different parameters is only discussed qualitatively all along the manuscript, and in some cases it is unclear. I have the feeling that all along the discussion, only those results leading to “positive” correlations are commented, ignoring those showing contradictions. Some examples include:
- the main peak in precipitation in Fig 4 a matches pretty well the seismic amplitudes, but neither of the other peaks have a good correlation (just overriding panels a and c this becomes evident)
- panels a) and b) in Fig. 8 show a clear similarity, but panel c) has a different peak and panel d) a very different pattern; this is not discussed in the text
- The purple clusters in Fig. 9 b) correspond to very different precipitation values; again, this is not commented/discussed, just stating that “These segments exhibit clear matches’ with storm evolution”
- The area encompassing more lightning strikes in Fig. 10, (located to the SW of AGIR) has a low number of detections and no backazimuthal determination seems to be calculated form inland strikes
- Is is hard to detect any trend in the colored points in Fig 11b
6) The authors claim that using K-means clustering has been a “key aspect of the analysis”. However, it is difficult for me to see which are these aspects. In Fig. 9, I would say that the spectrogram contains much more information that the clustering graph. Besides, no explanation on how and why the parameters of this clustering have been chosen
7) Concerning seismic data, during years a discussion has been open between those relating the microseismic peak amplitudes to air pressure or oceanic waves. However, it is now widely accepted that the amplitude of the microseismic peak is related to oceanic waves. On contrary, it is less clear which is the contribution of open waters and coastal zones to the primary and secondary peaks. If the amplitudes of the microseismic peak is discussed, this is the subject that should considered; Figures 6 and 7 just document that during stormy days, the microseismic noise is higher; this is a well-knonw feature, which could be observed in most of the seismic stations distributed worldwide
8) Concerning infrasound data, no explanation of which is the utility of parameters as spectral centroid, flux etc. is provided. Non-specialist readers need to know if these are parameters routinely calculated to discuss specific characteristics of the signal. The close relationship between infrasound and seismic data, widely documented my many contributions, is not commented at all.
9) The authors do not seem to be aware that the infrasound recordings related to lightning are in fact the acoustic waves generated by the associated thunders. Different works, including some of those included in the manuscript references list, have showed that the acoustic waves generated by thunders are recorded systematically by nearby seismic stations.
10) If seismic and infrasound data is used, the contribution of other sources of vibration and/or sound should be considered; which is the contribution of anthropogenic noise to each of the stations?
In my opinion, if the manuscript goal is to prove that the integration of multiple sensors has a clear utility to study storm evolution, much work is needed, including a better analysis of the existing data and a modelling effort. The work done by the authors can be useful to show that a strong storm can be detected not only by meteorological instruments but also by other sensors. However, this is something that is well-known by researches in each of the different fields and, in my opinion, does not deserve publication in NHESS.
Citation: https://doi.org/10.5194/egusphere-2025-1842-RC1 -
AC1: 'Reply on RC1', Laura Petrescu, 08 Aug 2025
reply
There are two points in the reviewer’s comments that we believe are important to clarify at this stage.
- On the claim that these methods are “widely known” for storm monitoring
This overstates the current state of practice. While the individual physical principles are understood, several of the specific applications in our study are relatively new and far from being routine in meteorological or hazard monitoring:
- GNSS-derived PWV is still transitioning from research to operational use, particularly in regions like Eastern Europe, where radar coverage is sparse.
- High-frequency seismic noise envelopes (>30 Hz) have only recently been proposed for rainfall and storm tracking, with just a handful of case studies in the literature (e.g. Dias et al., 2023; Rindraharisaona et al., 2022; Coviello et al., 2024).
- Infrasound for lightning detection is not yet standardized or widely integrated into weather monitoring systems.
To our knowledge, no previous work has combined these data types in a coordinated analysis of a real extreme storm. The dense and continuous coverage of GNSS and seismic networks, often denser than meteorological and hydrological stations, makes their integration especially promising for hazard applications.
- On the suggestion that a joint quantitative model is necessary
A strict cross-sensor correlation is not scientifically appropriate here because each dataset is sensitive to a different physical mechanism and operates on different spatial and temporal scales:
- GNSS-PWV: moisture buildup hours before the event
- High-frequency seismic noise: local raindrop impact energy
- Microseisms: wave-seafloor coupling from ocean swell
- Infrasound: pressure fronts and lightning activity
Even direct meteorological measurements (ex: radar, barometers, disdrometers) do not always align perfectly, for the same reason: they observe different parts of the system. The aim here is a qualitative, physically informed integration to monitor the event from multiple perspectives, which we believe is the right approach at this stage. More advanced modeling is a logical next step, but requires further methodological development.
We will address the remaining detailed comments in the formal review phase. At this stage, we wanted to emphasize that both the qualitative integration of non-traditional sensors in meteorological and hydrological monitoring, and the analysis of this extreme, recent event, are timely and well aligned with the scope of NHESS.
Citation: https://doi.org/10.5194/egusphere-2025-1842-AC1 - On the claim that these methods are “widely known” for storm monitoring
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 428 | 39 | 10 | 477 | 10 | 26 |
- HTML: 428
- PDF: 39
- XML: 10
- Total: 477
- BibTeX: 10
- EndNote: 26
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1