the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 03 Jun 2024
A Holocene alpine seismic chronicle from Lake Aiguebelette (NW French Alps)
Abstract. Lake sediments are valuable archives and can help construct a chronology of event deposits induced by seismic events. Such a chronology can be used to better understand the recurrence times between seismic events over longer periods than those covered by historical seismicity catalogs. However, only a few studies in lake palaeoseismology have focused on areas with moderate seismicity. This study aims to improve the catalog of paleoseismological archives in the front of the Western Alps. In this part, new multi-proxy data from the sedimentary archives of Lake Aiguebelette (France) allow the identification of 32 homogenites (thickness ≥ 0.5 cm) interpreted as of coseismic origin over the Holocene. An age model based on short-lived radionuclides, paleomagnetic data and radiocarbon ages constrains the chronology of sedimentary deposits in the deep basin of Lake Aiguebelette.
Among these homogenites, several were deposited at time intervals compatible with historical seismic events. To correlate the historical seismic events likely to have generated the event deposits identified in the sedimentary sequences of the deep basin of Lake Aiguebelette, the Earthquake Sensitivity Threshold Index (ESTI) method is used. Historical seismicity catalogs with uncertainties about intensities and epicenter coordinates for earthquakes make correlations to event deposits difficult. To better understand which seismic events may have been archived, a relative comparison was conducted between the pseudospectral acceleration (PSA) values calculated for each event in the FCAT-17 seismic catalog and for two distinct frequencies.
Based on this PSA approach, for higher frequencies (5 Hz), the contribution of nearby and moderate events is significantly stronger than that of strong and distant events in the lake sequence of Aiguebelette. Thus, the chronicle established based on the event deposits archived in Lake Aiguebelette sediment is interpreted as representative of local events (epicentral distance to the lake < 50 km). Recurrence intervals between the deposition of event layers do not follow a specific distribution (log-normal, Weibull, gamma or exponential) but might be a combination of several distributions. This suggests possible coexistence of several processes over the Holocene, impacting the evolution of the seismicity in this area.
- Preprint
(3675 KB) - Metadata XML
-
Supplement
(875 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on nhess-2024-83', Katleen Wils, 28 Jun 2024
The manuscript by Banjan et al. shows an interesting lake paleoseismic study in a very poorly known (from a paleoseismic point of view) region, showing the complexity of such settings. While showing its value, this study also underscores the need for more lacustrine studies in order to definitely validate or reject some of the proposed hypotheses. The paper itself is written in a clear way and is particularly well illustrated. I do have the impression that some of the available literature on Alpine lakes as well as other lakes with similar issues around the world have not been included, as I indicated throughout the annotated manuscript attached. Additionally, while most of the techniques and concepts applied in this paper are established and validated in lake paleoseismic studies around the world, I have some questions with regards to some of the specific approaches and interpretations here. I list my main concerns below, and added some smaller remarks in the attached file. Nevertheless, I do believe all the necessary data is presented in the paper, and that it could still become significant once these concerns are addressed.
- Although log-ratios are often used for interpreting XRF data, it is advisable to use centered log-ratios (as also explain by the reference cited here in the methodology) - especially if you base your core-to-core correlations partly on the absolute abundance of a certain element in the core. It is not clear if you used this approach, as the single element XRF values are not plotted in figure 6 (as was mentioned in the text).
- While the full age model is shown in the manuscript, it is not entirely clear how it was constructed. This should be placed more upfront. Additionally, integrating such a large number of dating techniques for the long core (especially also the varve counts, which is not clear how you incorporated these), using the very simplistic clam model is probably not the ideal way. I strongly advice to use Bacon instead, which also does not require to manually define outliers.
- You seem to add a high confidence level for seismic triggering to any event deposit thicker than 2 cm, even though not all of them actually are, even in historical times. It is not clear how this is incorporated in the discussion, and if you left out these deposits for the section on recurrence statistics. And if not, how does this affect your results, as there could likely be more of those event deposits non-seismically triggered?
- Why do you consider PSA values? It is extremely difficult to determine threshold values based on quantitative ground motion parameters, as so far only intensity data was used and is available. Although I strongly encourage to move away from intensity, I do question why PSA is used instead of for example PGA or PGV. Considering the high uncertainties, why not start with an IPE test and evaluate local intensities at the lake rather than just epicentral intensities? And when you then move to PSA, why specifically 0.5 and 5 Hz? Are there any indications as to why these would be more relevant than others? As far as I understand, low-frequency shaking is amplified in sediments, so the conclusion that is made here on 5 Hz being more relevant seems questionable and not supported by other literature. Additionally, what do you consider as site effects (e.g. Vs30 values)? These should be used as input for the GMPEs I assume, and would also strongly impact the outcomes.
- I agree that historical earthquake catalogs are often incomplete. However, using this as the only argument to discard the potential for the Basel earthquake to have resulted in an event deposit seems questionable. Even more so considering the ESTI approach does not rule out a deposit at all. Are there any other indications as to why you do not consider this earthquake as a likely trigger? If not, I don't see an objective reason to discard it.
- With distance, high-frequency shaking is consistently attenuated faster than low-frequency shaking. So the fact that PSA values are lower for 5 Hz for distant earthquakes compared to closer earthquakes, and not so much for 0.5 Hz, does not really present a solid argument as to why nearby events would more likely trigger the event deposits in Lake Aiguebelette. This entirely depends on which frequency content lake sediments respond to (related to my previous point). Of course, the SA5 would become attenuated faster than the SA0.5, so there will always be a stronger difference with distance for PSA5.
- You do not discuss on why some event deposits consist only of homogenites, and other homogenites plus turbidites. This clearly involves different depositional processes, and might help further distinguish seismic from non-seismic deposits? A seismic seiche without turbidite seems to rather implausible, in my opinion.
- Throughout the paper, you never consider the process of surficial remobilization for turbidite deposition. Although the thicker event deposits might still relate to slope failures, especially the thinner ones might rather be attributed to this surficial process. The fact that event deposit material originates from within the lake only strengthens the possibility for surficial remobilization. It would be beneficial to elaborate on this, as in this particular case, sedimentation rate would not be relevant for turbidity current generation and thus further advocating for a seismically quiet period in stage 2.
- For the recurrence statistics calculations, it would make more sense to focus only on the long core to avoid having a change of recording sensitivity in the recent period that is covered by one of the pilot cores.
- How do your periods of seismic quiescence and activity relate to other Alpine regions? For example, also in Carinthia, some clustering of events and quiet periods have been proposed.
-
AC1: 'Reply on RC1', Mathilde Banjan, 08 Sep 2024
Dear Reviewer,
We thank you for your valuable feedback and constructive comments on our study.
Below, you will find our responses to each of your questions and remarks. Additionally, a PDF containing all our answers to the comments is attached as a supplementary file.
Thank you again for your time and consideration.
Best regards,
Mathilde Banjan and co-authors
The manuscript by Banjan et al. shows an interesting lake paleoseismic study in a very poorly known (from a paleoseismic point of view) region, showing the complexity of such settings. While showing its value, this study also underscores the need for more lacustrine studies in order to definitely validate or reject some of the proposed hypotheses. The paper itself is written in a clear way and is particularly well illustrated. I do have the impression that some of the available literature on Alpine lakes as well as other lakes with similar issues around the world have not been included, as I indicated throughout the annotated manuscript attached. Additionally, while most of the techniques and concepts applied in this paper are established and validated in lake paleoseismic studies around the world, I have some questions with regards to some of the specific approaches and interpretations here. I list my main concerns below, and added some smaller remarks in the attached file. Nevertheless, I do believe all the necessary data is presented in the paper, and that it could still become significant once these concerns are addressed.
1- Although log-ratios are often used for interpreting XRF data, it is advisable to use centered log-ratios (as also explain by the reference cited here in the methodology) - especially if you base your core-to-core correlations partly on the absolute abundance of a certain element in the core. It is not clear if you used this approach, as the single element XRF values are not plotted in figure 6 (as was mentioned in the text).
As expressed in the caption below Fig.6, the core-to-core correlations for the AIG17III and AIG16-05 sediment cores are based on IRM data (magnetic data) and not XRF data.
While we understand the importance of centered log-ratios (CLR) for compositional data analysis to avoid matrix effect, simple log-ratios are effective for general trends and comparisons in geochemical data (Weltje et al., 2015). In Fig.4, the general trend along depth suffices for a visual comparison of geochemical variation across different facies. Therefore, we used a simple ln ratio (cf. lines 235-239).
2- While the full age model is shown in the manuscript, it is not entirely clear how it was constructed. This should be placed more upfront. Additionally, integrating such a large number of dating techniques for the long core (especially also the varve counts, which is not clear how you incorporated these), using the very simplistic clam model is probably not the ideal way. I strongly advice to use Bacon instead, which also does not require to manually define outliers.
The full age-depth model (Fig.8) shown in the manuscript is based on:
- short-lived radioelements (cf. lines 271-279, Fig.7);
- paleomagnetic data (cf. lines 293-302, SM-3);
- radiocarbon dating (data available in Table 2 from Banjan et al. (2023)).
These different time-constraints are highlighted in the section 3.5. of the manuscript (lines 159-166).
The varves counts is visible and available in Fig.7 (page 16), but just for the upper part of the core in comparison to short-lived radionuclides, as varves are not continuous over the whole sediment sequence, thus they are not used for age modeling.Regarding the choice of Clam, different programs exist for age-depth modelling (Bacon, Bchron, Clam and OxCal included). Clam R packages employ classical statistical methods (Blaauw, 2010), whereas the three others use Bayesian statistics (Bacon: Blaauw and Christen, 2011; Bchron: Haslett and Parnell, 2008; OxCal: Bronk Ramsey, 2009).
Wright et al. (2017) compared the relative performance of these different programs for age modelling and found that no single modelling package outperforms all others, but an ensemble approach can exploit different model strengths to produce a ‘consensus’, illustrating that choosing the ‘best model’ is not a simple task (Lacourse and Gajewski, 2020). Uncertainty estimation differs considerably among models and Bayesian age–depth models mainly improve the assessment of uncertainties of age–depth models (Trachsel and Telford 2017). As we have less than two dates per millennium (18 ages for 14kyr), we selected a program which employs classical statistical methods: Clam R package as explained in the manuscript. Moreover this Clam age model is already published for AIG long core (Banjan et al., 2022), thus we do not want to change the chronology.3- You seem to add a high confidence level for seismic triggering to any event deposit thicker than 2 cm, even though not all of them actually are, even in historical times. It is not clear how this is incorporated in the discussion, and if you left out these deposits for the section on recurrence statistics. And if not, how does this affect your results, as there could likely be more of those event deposits non-seismically triggered?
The confidence level for seismic triggering of event deposits thicker than 2 cm is based on a multi-proxy approach, not only on thickness. As detailed in section 5.1. of the paper, the interpretation of these deposits as seismically induced is supported several characteristics:
- Visual observations of homogenite (Hm) and turbidite-homogenite (Tu+Hm) facies;
- Grain-size data trend in a Passega diagram;
- Geochemical signatures (XRF logarithmic ratios) indicating remobilization of sediment from lake slopes;
- High Anisotropy of Magnetic Susceptibility (AMS) foliation values, which have been associated with seiche effects in previous studies;
- Low Isothermal Remanent Magnetization (IRM) amplitudes.The 2 cm threshold was chosen because it allows for reliable AMS measurements, which require a minimum sample volume. However, the paper does acknowledge varying confidence levels for different deposit thicknesses (Fig. 9), with deposits between 0.5 and 2 cm thick still considered of seismic origin, with moderate confidence.
Regarding the recurrence statistics, the analysis in section 5.3. focuses on event layers ≥0.5 cm thick, which includes both the high and moderate confidence deposits. This approach balances the need for a robust dataset with the acknowledgment of uncertainty in interpretation.
The paper does consider the possibility of non-seismic triggers, particularly for thinner deposits. The discussion in section 5.3. explicitly addresses the potential impact of changing sedimentation rates on the frequency of recorded events, acknowledging that this could lead to misinterpretation if not carefully considered.Future work could include sensitivity analyses to assess how the inclusion or exclusion of moderately confident seismic deposits affects the recurrence statistics. Additionally, expanding the multi-proxy analysis to thinner deposits could help refine the criteria for seismic attribution across all deposit thicknesses.
4- Why do you consider PSA values? It is extremely difficult to determine threshold values based on quantitative ground motion parameters, as so far only intensity data was used and is available. Although I strongly encourage to move away from intensity, I do question why PSA is used instead of for example PGA or PGV. Considering the high uncertainties, why not start with an IPE test and evaluate local intensities at the lake rather than just epicentral intensities? And when you then move to PSA, why specifically 0.5 and 5 Hz? Are there any indications as to why these would be more relevant than others? As far as I understand, low-frequency shaking is amplified in sediments, so the conclusion that is made here on 5 Hz being more relevant seems questionable and not supported by other literature. Additionally, what do you consider as site effects (e.g. Vs30 values)? These should be used as input for the GMPEs I assume, and would also strongly impact the outcomes.
Here are several points in order to clarify the use of PSA values and selected methodology:
The use of Pseudo-Spectral Acceleration (PSA) values over Peak Ground Acceleration (PGA) or Peak Ground Velocity (PGV) was driven by the need to discuss a possible frequency-dependent response of lake sediments. PSA provides a measure of ground motion that can be specifically tied to different frequencies, making it interesting for understanding the sediment response at various frequencies. The use of PSA values is compatible with the dynamic behavior of sediments, which is key for identifying the potential for event deposit formation. This approach is consistent with methodologies employed in other studies focused on understanding sediment responses to seismic shaking (Strasser et al., 2013; Avşar et al., 2016; Moernaut, 2020).
The uncertainties associated with the development of Intensity Prediction Equations (IPEs) are of the same order as those for Ground Motion Prediction Equations (GMPEs), as demonstrated by Bakun and Scotti (2006), for France. This consideration supports the approach of using PSA values, which provide a more detailed analysis of the frequency content of seismic waves, over traditional intensity measures.
Regarding site effects, these were not directly considered (cf. lines 440-442). Considering specific site effects for Lake Aiguebelette would require a study, which is beyond the current scope of our work. A detailed analysis similar to that of Shynkarenko et al. (2023) would be necessary to fully integrate these factors. Consequently, our approach focuses on comparing relative PSA values rather than absolute values. It is important to note that site effects could indeed influence the frequency content of seismic waves, potentially affecting the results. This is why our analysis emphasizes relative differences, which helps to mitigate the impact of site-specific factors on our overall conclusions.Site-specific factors such as Vs30 values (shear wave velocity in the top 30 meters of soil) are key for accurate ground motion prediction and were taken into account in our GMPE calculations. The GMPEs used (Akkar et al., 2014; Bindi et al., 2017) incorporate these site effects to provide a more accurate representation of ground shaking at the lake site. In each GMPE, different soil and rock conditions are included and in each case the same trend is observed respectively for low (0.5 Hz) and high (5 Hz) frequencies.
Regarding the choice of frequencies, while the decision to use 0.5 Hz and 5 Hz may seem arbitrary, it was guided by the need to represent both low and high-frequency content. These specific frequencies are commonly used as proxies in the literature to capture the effects of low and high-frequency seismic waves (e.g., Atkinson, 2008). The selection was not meant to be an absolute measure but rather a practical approach to see how historical earthquakes compare across a spectrum of frequency content. Lower frequencies (0.5 Hz) are relevant for assessing long-period waves that travel further and have significant effects on sediment layers. Higher frequencies (5 Hz) are relevant for assessing the impact of near-field, high-energy seismic waves that can trigger immediate and localized sediment disturbances. Lake sediments can respond differently to various frequencies of ground shaking, and modeling PSA at these two frequencies is representative of a spectrum of seismic energy that could influence sediment deposition processes. This approach of using multiple frequencies to assess seismic response is supported by studies such as Kremer et al. (2017), who emphasize the importance of considering both low and high-frequency content when evaluating seismic shaking impacts on lake sediments. Their work demonstrates that different frequency ranges can trigger distinct sedimentary responses, which aligns with the approach of examining both 0.5 Hz and 5 Hz PSA values.
Based on the PSA approach, the results indicate that high-frequency shaking, particularly at 5 Hz, is more relevant for triggering event deposits in Lake Aiguebelette, and thus, it is not always the low-frequency shaking that is most effective (Fig. 11).
5- I agree that historical earthquake catalogs are often incomplete. However, using this as the only argument to discard the potential for the Basel earthquake to have resulted in an event deposit seems questionable. Even more so considering the ESTI approach does not rule out a deposit at all. Are there any other indications as to why you do not consider this earthquake as a likely trigger? If not, I don't see an objective reason to discard it.
Another rationale (lines 460-463) for discarding the potential for the 1356 CE Basel earthquake to have resulted in an event deposit is visible in Fig.11.
For the 1356 CE Basel earthquake, the epicentral distance to the site is 255.3 km with a Mw of 6.5 +/- 0.54. We note that the 1887 CE Ligurian earthquake has a comparable epicentral distance of 258.4 km and a Mw of 6.7 +/- 0.59. If the Basel earthquake had resulted in an event deposit, it is highly probable that the Ligurian earthquake would have caused one as well, yet it did not.
6- With distance, high-frequency shaking is consistently attenuated faster than low-frequency shaking. So the fact that PSA values are lower for 5 Hz for distant earthquakes compared to closer earthquakes, and not so much for 0.5 Hz, does not really present a solid argument as to why nearby events would more likely trigger the event deposits in Lake Aiguebelette. This entirely depends on which frequency content lake sediments respond to (related to my previous point). Of course, the PSA5 would become attenuated faster than the PSA0.5, so there will always be a stronger difference with distance for PSA5.
Here are two arguments (energy distribution and historical evidences) as to why nearby events would more likely trigger the event deposits in Lake Aiguebelette:
(1) The energy distribution of an earthquake typically has more high-frequency content near the source. This results in higher ground acceleration and shaking intensity near the epicenter, which is more likely to disturb and rework sediments to create event deposits, near the source. Near-field ground motion often includes higher peak ground accelerations and velocities that can be more effective in causing sediment disturbance (Boore, 2003).
(2) Historical earthquakes show a higher likelihood of near-field events causing significant geological and sedimentary disturbances compared to far-field events. Studies on lake sediments in seismically active regions often show a higher correlation between event deposits and nearby earthquakes. This observation supports the idea that the intensity and characteristics of near-field shaking are more effective in triggering these deposits (Goldfinger et al., 2008).
Additionally, the response of Lake Aiguebelette’s sediments could be influenced by site-specific factors such as sediment composition, layering, and water depth. These factors can determine the sensitivity of the sediments to different frequencies of seismic shaking. Further site-specific studies (that are not included in this article, but will be the object of future research) are necessary to understand the dynamic response of the lake sediments to seismic activity accurately (Shynkarenko et al., 2023).
7- You do not discuss on why some event deposits consist only of homogenites, and other homogenites plus turbidites. This clearly involves different depositional processes, and might help further distinguish seismic from non-seismic deposits? A seismic seiche without turbidite seems to rather implausible, in my opinion.
When sampling sediment cores, it is possible to capture either the turbidite or the homogenite, depending on the samping location. Due to variations in current strength and topography, sediment gravity flows can transport and deposit sediments unevenly across a basin. As a result, one core might archive a turbidite layer while another core nearby might capture a homogenite layer instead (Stow and Smillie, 2020; Piper, 1978). This variability in sediment deposition is a well-recognized phenomenon in sedimentology (Stow et al., 2001).
The occurrence of homogenites without accompanying turbidites in some records can be attributed to the nature of the depositional environment and the energy of the triggering event. Seismic seiches can produce homogenites without a significant turbidite if the shaking is sufficient to rework the sediment but not strong enough to trigger a full-scale turbidity current (Mulder and Alexander, 2001).The absence of a turbidite does not necessarily rule out a seismic origin for the event deposit. The differential preservation of these deposits in various cores can be related to the spatial variability (of the processes) within the basin.
It is necessary to add that a seismic seiche without a turbidite is not implausible as shown in Lake Bourget (Chapron et al., 1999).8- Throughout the paper, you never consider the process of surficial remobilization for turbidite deposition. Although the thicker event deposits might still relate to slope failures, especially the thinner ones might rather be attributed to this surficial process. The fact that event deposit material originates from within the lake only strengthens the possibility for surficial remobilization. It would be beneficial to elaborate on this, as in this particular case, sedimentation rate would not be relevant for turbidity current generation and thus further advocating for a seismically quiet period in stage 2.
It would be beneficial to include a point about surficial remobilization in the discussion section of this paper for turbidite deposition. During seismically quiet periods, surficial processes could generate turbidity currents, leading to the deposition of thinner turbidites. This could support the observation of sediment layers that are not directly linked to significant seismic events during the historical and recent times. This point will be added to the discussion section.
However, in the Passega diagram (Fig.5) we can see that the data for events thicker than 0.5 cm follow the same trend, suggesting a common depositional process.
9- For the recurrence statistics calculations, it would make more sense to focus only on the long core to avoid having a change of recording sensitivity in the recent period that is covered by one of the pilot cores.
The sensitivity of recording seismic events in lake sediments is influenced more by the variation in sedimentation rates rather than the specific cores used. By including both the long core and the pilot cores, the goal is to provide a more robust understanding of the seismic record. As mentioned in the manuscript, sedimentation rates can vary and significantly impact the sensitivity of the lake sediments to record seismic events (Wilhelm et al., 2016; Rapuc et al., 2018).
To mitigate any potential bias in recording sensitivity, we carefully correlated the pilot cores with the long core using various proxies and age-depth models. This approach ensures that the data from the pilot cores are seamlessly integrated with the long core, providing a continuous and consistent seismic record. The pilot cores, such as AIG20-01, provide valuable high-resolution data for recent periods, complementing the long core data and enhancing the overall accuracy of the seismic chronicle.
Previous studies have demonstrated the importance of using multiple cores to enhance the resolution and accuracy of seismic event chronologies in lacustrine environments. For instance, Strasser et al. (2013) and Moernaut et al. (2014) highlight the benefits of integrating data from various cores to capture a complete and detailed record of seismic events over different time scales.
10- How do your periods of seismic quiescence and activity relate to other Alpine regions? For example, also in Carinthia, some clustering of events and quiet periods have been proposed.
We identified three main stages:
- From -9890 to -2000 yr cal CE: A period of relatively consistent seismic activity;
- From -2000 to 0 yr cal CE: A period of seismic quiescence (with no event layers ≥0.5 cm thick recorded);
- From 0 to 2017 yr cal CE: A period of increased seismic activity.
The study by Daxer et al. (2022) in Carinthia (Austria), also identified periods of clustered seismic activity and quiescence. Phases of enhanced regional seismicity are observed, including two high-frequency periods at ca. 12.8 ka BP and ca. 3.5 ka BP.
While the timing of these periods does not exactly match our results, it supports the observation that seismic activity in the Alpine region could be characterized by alternating periods of increased seismic activity and relative quiescence.Our observation of increased seismic activity from 0 to 2017 yr cal CE is compatible with the trend of increased seismicity in the late Holocene observed in several Alpine lakes. Kremer et al. (2017) highlight an increase in mass movement occurrences in Swiss lakes during the past 4000 years, which they partly attributed to increased seismicity. While we can see some broad similarities with other Alpine records (periods of increased seismicity and periods of relative quiescence), the specific timing and duration of active and quiet periods can vary regionally. This highlights the importance of such studies in the Alpine region to better understand the spatial and temporal variations in seismic activity.
The increase in sedimentation rate could be linked to enhanced erosion driven by human activities, such as agriculture and deforestation, which can lead to more frequent and thicker sediment deposits (Arnaud et al., 2016)
_______
References:
Atkinson, G. M. (2008). Ground-motion prediction equations for eastern North America from
a referenced empirical approach: Implications for epistemic uncertainty. Bulletin of the
Seismological Society of America, 98(3), 1304-1318.
Arnaud, F., Poulenard, J., Giguet-Covex, C., Wilhelm, B., Révillon, S., Jenny, J. P., ... &
Sabatier, P. (2016). Erosion under climate and human pressures: An alpine lake sediment
perspective. Quaternary Science Reviews, 152, 1-18.
Bakun, W. H., & Scotti, O. (2006). Regional intensity attenuation models for France and the
estimation of magnitude and location of historical earthquakes. Geophysical Journal
International, 164(3), 596-610.
Blaauw, M. (2010). Methods and code for ‘classical’ age-modelling of radiocarbon
sequences. Quaternary geochronology, 5(5), 512-518.
Chapron, E., Beck, C., Pourchet, M., & Deconinck, J. F. (1999). 1822 earthquake‐triggered
homogenite in Lake Le Bourget (NW Alps). Terra Nova, 11(2‐3), 86-92.
Goldfinger, C., Grijalva, K., Bürgmann, R., Morey, A. E., Johnson, J. E., Nelson, C. H., ... &
Gràcia, E. (2008). Late Holocene rupture of the northern San Andreas fault and possible
stress linkage to the Cascadia subduction zone. Bulletin of the Seismological Society of
America, 98(2), 861-889.
Kremer, K., Wirth, S. B., Reusch, A., Fäh, D., Bellwald, B., Anselmetti, F. S., ... & Strasser,
M. (2017). Lake-sediment based paleoseismology: Limitations and perspectives from the
Swiss Alps. Quaternary Science Reviews, 168, 1-18.
Lacourse, T., & Gajewski, K. (2020). Current practices in building and reporting age-depth
models. Quaternary Research, 96, 28-38.
Moernaut, J., Daele, M. V., Heirman, K., Fontijn, K., Strasser, M., Pino, M., ... & De Batist, M.
(2014). Lacustrine turbidites as a tool for quantitative earthquake reconstruction: New
evidence for a variable rupture mode in south central Chile. Journal of Geophysical
Research: Solid Earth, 119(3), 1607-1633.
Mulder, T., & Alexander, J. (2001). The physical character of subaqueous sedimentary
density flows and their deposits. Sedimentology, 48(2), 269-299.
Piper, D. W. (1978). Turbidite muds and silts on deepsea fans and abyssal plains.
Sedimentation in submarine canyons, fans, and trenches.
Rapuc, W., Sabatier, P., Andrič, M., Crouzet, C., Arnaud, F., Chapron, E., ... & Von
Grafenstein, U. (2018). 6600 years of earthquake record in the Julian Alps (Lake Bohinj,
Slovenia). Sedimentology, 65(5), 1777-1799.
Shynkarenko, A., Cauzzi, C., Kremer, K., Bergamo, P., Lontsi, A. M., Janusz, P., & Fäh, D.
(2023). On the seismic response and earthquake-triggered failures of subaqueous slopes in
Swiss lakes. Geophysical Journal International, 235(1), 566-588.
Stow, D., & Smillie, Z. (2020). Distinguishing between deep-water sediment facies: turbidites,
contourites and hemipelagites. Geosciences, 10(2), 68.
Strasser, M., Monecke, K., Schnellmann, M., & Anselmetti, F. S. (2013). Lake sediments as
natural seismographs: A compiled record of Late Quaternary earthquakes in Central
Switzerland and its implication for Alpine deformation. Sedimentology, 60(1), 319-341.
Trachsel, M., & Telford, R. J. (2017). All age–depth models are wrong, but are getting better.
The Holocene, 27(6), 860-869.
Wilhelm, B., Nomade, J., Crouzet, C., Litty, C., Sabatier, P., Belle, S., ... & Anselmetti, F. S.
(2016). Quantified sensitivity of small lake sediments to record historic earthquakes:
Implications for paleoseismology. Journal of Geophysical Research: Earth Surface, 121(1),
2-16.-
RC2: 'Reply on AC1', Katleen Wils, 04 Nov 2024
I would like to thank Mathilde Banjan and co-authors for their detailed replies to my earlier comments on their manuscript. I went through them, and in most cases agree that simply some further clarification in the main text of the paper would suffice to address my concerns. However, there are a few points where I would like to come back to which I feel haven't been fully addressed, as indicated below. I’ve used the same numbering as earlier on, so that the discussion can be easily followed.
1. I agree that log-ratios are indeed sufficient to simply check for downcore variability, but as currently presented in the manuscript (L251), you mention ‘single XRF element (Ca and K) variations are used to confirm first correlations’. As counts are always relative to a unit sum, the increase of one element will result in the decrease of another element, and therefore, interpretations between cores cannot be confidently made. So either you will indeed have to use ratios here (and then also show them in the manuscript, as you did for ln(Ca/Ti) in figure 4), or alternatively, a clr transformation is required.
3. Indeed, various confidence intervals for earthquake triggering have been considered in the study using multiple proxies. However, as the authors mention, a summary is given in figure 9 and this shows that the thickness seems to be the main factor to add high confidence (as is also mentioned in its caption). If this is not the case, the figure is probably misleading considering that indeed, not all >2 cm deposits result from earthquakes as shown by the historical record. This should be further clarified in the text.
4. I fully agree that shaking is dependent on frequency, as already hypothesized by many other studies as the authors are also aware of. But, considering this dependency is irrelevant when not considering the proper site effects, which includes vs30 values. I understand that a full site effect study of the lake is beyond the scope of this paper and probably only possible in the future, the authors seem to have tested several possibilities, and this should be mentioned in the manuscript (at least the vs30 values that were used for the plots). To get reasonable outcomes, I would expect values similar to those measured by Shynkarenko et al. to be considered, but at this point, this cannot be evaluated. However, the outcomes here are rather surprising: “high-frequency shaking, particularly at 5 Hz is more relevant for triggering event deposits in Lake Aiguebelette”, and have currently not been addressed by the authors. Soft sediments are known to amplify low-frequency accelerations (see e.g. in Van Daele et al. 2019, or also the Shynkarenko paper the authors already referred to), as their spectral peak is typically located below 5 Hz. So this result definitely requires further discussion and/or revision. This discussion should also be considered for points 5 and 6 in my earlier review.
8. I agree that adding surficial remobilization to the discussion will increase the value of the paper and its outcomes. I do want to emphasize that whether or not a turbidite results from surficial remobilization or a slope failure, the depositional process is still the same (i.e. from a turbidity current), so I would not expect this to affect the Passega diagram.
9. I fully agree that it is important to consider multiple cores for meaningful recurrence statistics. However, it is important that the recording threshold stays the same. For example, in Moernaut et al. (2014) as mentioned by the authors, it is clear that not every site has the same shaking sensitivity. This complicates interpretations when combining them, potentially even making them invalid. As some of the events are not recorded in your long core but are in the pilot core, I suspect a similar thing happening here – even though sedimentation rate seems to have a stronger impact on earthquake sensitivity here (which should not be the case when surficial remobilization is considered, see point 8). I strongly recommend to make separate recurrence analyses for different timeframes, mentioning the recording thresholds of each to allow the readers to evaluate record sensitivity, or use a (thickness?) threshold before applying statistical analysis on the correlated long and pilot cores.
Thank you again for providing me with the opportunity to review this manuscript.
Reference:
Distinguishing intraplate from megathrust earthquakes using lacustrine turbidites
Van Daele, C. Araya-Cornejo, T. Pille, K. Vanneste, J. Moernaut, S. Schmidt, et al.
Geology 2019 Vol. 47 Issue 2 Pages 127-130
DOI: 10.1130/g45662.1Citation: https://doi.org/10.5194/nhess-2024-83-RC2 -
AC2: 'Reply on RC2', Mathilde Banjan, 22 Apr 2025
Dear Reviewer,
We thank you for your additional feedback and constructive comments on our study.
Below, you will find our responses to each of your questions and remarks (they are also available in a pdf attached as supplement).
Thank you again for your time and consideration.Best regards,
Mathilde Banjan and co-authorsI would like to thank Mathilde Banjan and co-authors for their detailed replies to my earlier comments on their manuscript. I went through them, and in most cases agree that simply some further clarification in the main text of the paper would suffice to address my concerns. However, there are a few points where I would like to come back to which I feel haven't been fully addressed, as indicated below. I’ve used the same numbering as earlier on, so that the discussion can be easily followed.
- I agree that log-ratios are indeed sufficient to simply check for downcore variability, but as currently presented in the manuscript (L251), you mention ‘single XRF element (Ca and K) variations are used to confirm first correlations’. As counts are always relative to a unit sum, the increase of one element will result in the decrease of another element, and therefore, interpretations between cores cannot be confidently made. So either you will indeed have to use ratios here (and then also show them in the manuscript, as you did for ln(Ca/Ti) in figure 4), or alternatively, a clr transformation is required.
We agree that XRF data are relative and should be normalized to avoid misinterpretation. While the statement in the manuscript (L251) might imply that raw single-element counts were used for correlation, our actual methodology involved normalized log-ratios, such as ln(Ca/Ti), which are appropriate for assessing downcore variability and inter-core comparisons. To address this, we will:
- Revise the text in Section 3.3 to explicitly state that log-ratios were employed for core-to-core comparisons.
- Add a supplementary figure illustrating the log-ratio variations (e.g., ln(Ca/Ti), ln(K/Al)) for the relevant cores.
- Indeed, various confidence intervals for earthquake triggering have been considered in the study using multiple proxies. However, as the authors mention, a summary is given in figure 9 and this shows that the thickness seems to be the main factor to add high confidence (as is also mentioned in its caption). If this is not the case, the figure is probably misleading considering that indeed, not all >2 cm deposits result from earthquakes as shown by the historical record. This should be further clarified in the text.
While deposit thickness provides a practical threshold for higher-confidence interpretations, it is only one component of our multi-proxy approach. Figure 9 summarizes confidence levels derived from multiple criteria, including AMS foliation, grain-size patterns, geochemical ratios, and IRM amplitudes. We acknowledge that the figure caption and text might emphasize thickness as a contributing factor.
To address this, we will:
- Revise the caption of Figure 9 to highlight that thickness is one of several factors contributing to confidence levels.
- Update Section 5.1 to explicitly state that confidence levels are determined through a combination of proxies, with thickness serving as a practical threshold for reliable AMS measurements and other analyses.
- I fully agree that shaking is dependent on frequency, as already hypothesized by many other studies as the authors are also aware of. But, considering this dependency is irrelevant when not considering the proper site effects, which includes vs30 values. I understand that a full site effect study of the lake is beyond the scope of this paper and probably only possible in the future, the authors seem to have tested several possibilities, and this should be mentioned in the manuscript (at least the vs30 values that were used for the plots).
We agree with the reviewer that the shaking frequency is influenced by site effects, and that Vs30 is an important parameter. While a detailed site effect study is beyond the scope of this paper, it is part of planned future work, as noted in lines 502–505 of the manuscript: “However, it should be noted that the approach developed in this paper is relatively preliminary as it does not, for example, consider local conditions that could lead to changes in seismic motion for certain frequency ranges (Ergin et al., 2004; Maufroy et al., 2015, Courboulex et al., 2020) and thus could alter the proposed interpretations.”
For the GMPE-based PSA analysis presented in this study, we used a Vs30 value of 800 m/s, which corresponds to standard reference rock conditions and is consistent with expected velocities for molasse formations in the region. In the case of Lake Aiguebelette, the Quaternary cover on the slopes consists only of till and is limited to 0-5 m in thickness (according to field observation and geotechnical investigations), which supports the use of this higher Vs30 value (personal communication with Aurore Laurendeau and Ludmila Provost). We also conducted tests with lower Vs30 values (400 m/s) to assess sensitivity, and the results did not show substantial differences in the PSA range relevant to this study. Since the sensitivity to Vs30 within this range (400–800 m/s) remains limited for our conclusions, we chose to keep the representation simple. However, we will add a sentence in the manuscript to specify the Vs30 value used and clearly acknowledge this assumption.
To get reasonable outcomes, I would expect values similar to those measured by Shynkarenko et al. to be considered, but at this point, this cannot be evaluated. However, the outcomes here are rather surprising: “high-frequency shaking, particularly at 5 Hz is more relevant for triggering event deposits in Lake Aiguebelette”, and have currently not been addressed by the authors. Soft sediments are known to amplify low-frequency accelerations (see e.g. in Van Daele et al. 2019, or also the Shynkarenko paper the authors already referred to), as their spectral peak is typically located below 5 Hz. So this result definitely requires further discussion and/or revision. This discussion should also be considered for points 5 and 6 in my earlier review.
This result arises from an empirical approach. At this stage, we do not attempt to quantitatively model site effects or ground motion amplification in the basin, as the necessary site-specific parameters are currently not available.
The aim was to test in a relative way, whether the recent event record appears more consistent with seismic signals rich in high-frequency content (>2–3 Hz) or with lower-frequency energy (<1–2 Hz), which typically characterizes large distant earthquakes. The good match between observed event deposits and ground motion proxies in the 5 Hz range suggests that the lake’s sedimentary slopes are more responsive to higher-frequency shaking.
This may imply that the lacustrine environment of Lake Aiguebelette is not particularly susceptible to low-frequencies, which is consistent with the absence of distal, high-magnitude earthquakes in the record. Had there been strong amplification at lower frequencies due to site effects, we would expect more frequent recording of large distal events. The fact that this is not the case suggests that either site amplification in the low-frequency range is limited, or that slope instability thresholds are reached under high-frequency loading.
We agree that a more detailed site response analysis would be a valuable next step, and we now explicitly state in the revised manuscript that this question remains open. However, as a first-order interpretation, our empirical observation supports the idea that the lake’s subaqueous slopes are more sensitive to near-field events with richer high-frequency content.
Regarding the unexpected result of 5 Hz relevance, with more detail:
- Sediment Response: Lake basin morphology and sediment characteristics may amplify higher-frequency shaking in certain conditions. For instance, thin homogenites may form preferentially under high-frequency shaking that remobilizes finer-grained surficial slope sediments (Sabatier et al, 2022).
- Comparative Data: Previous studies (Kremer et al., 2017) present variability in frequency sensitivity across alpine lake settings and display frequency-selective sensitivities to seismically-induced event layers. We suggest that Lake Aiguebelette is favoring the record of close and moderate earthquakes rather than distant and strong seismic events. We acknowledge that this frequency bias could be further investigated through a quantitative analysis of slope susceptibility to seismic motion, this point will be clearly stated in the revised manuscript.
We will expand Section 5.2 to:
- Include a discussion of frequency-specific sediment responses, referencing relevant literature (e.g., Kremer et al., 2017; Shynkarenko et al., 2023).
- Acknowledge the need for future site-specific studies to refine these interpretations.
- I agree that adding surficial remobilization to the discussion will increase the value of the paper and its outcomes. I do want to emphasize that whether or not a turbidite results from surficial remobilization or a slope failure, the depositional process is still the same (i.e. from a turbidity current), so I would not expect this to affect the Passega diagram.
We agree that surficial remobilization is a key process, particularly during seismically quiet periods. While the depositional mechanism (turbidity currents) remains consistent, triggers may vary, including surficial sediment reworking. In the Passega diagram, the alignment of data for deposits >0.5 cm supports a shared depositional process.
We will:
- Expand the discussion in Section 5.3 to include the role of surficial remobilization in forming thinner turbidites during quiescent periods.
- Make it clear that while the depositional processes are similar, the interpretation of the triggering origin (seismic or not) relies on a combination of multiple proxies.
- I fully agree that it is important to consider multiple cores for meaningful recurrence statistics. However, it is important that the recording threshold stays the same. For example, in Moernaut et al. (2014) as mentioned by the authors, it is clear that not every site has the same shaking sensitivity. This complicates interpretations when combining them, potentially even making them invalid. As some of the events are not recorded in your long core but are in the pilot core, I suspect a similar thing happening here – even though sedimentation rate seems to have a stronger impact on earthquake sensitivity here (which should not be the case when surficial remobilization is considered, see point 8). I strongly recommend to make separate recurrence analyses for different timeframes, mentioning the recording thresholds of each to allow the readers to evaluate record sensitivity, or use a (thickness?) threshold before applying statistical analysis on the correlated long and pilot cores.
We acknowledge the concern regarding potential differences in recording sensitivity between the long core (AIG17III) and the pilot core (AIG16-05). In this study, these two cores were selected for their complementary characteristics: AIG17III covers the full Holocene sedimentation, while AIG16-05 provides coverage of the most recent ~1.2 meters, including the historical period.
Due to coring limitations and sediment disturbance in the upper part of AIG17III, several historical event layers are only recorded in AIG16-05. Importantly, below 1.2 meters all event layers identified in AIG16-05 are also present in AIG17III. This confirms a consistent recording threshold between both cores for the mid-to-late Holocene section.
To ensure consistency in our analysis:
- We apply a threshold of ≥0.5 cm deposit thickness across all cores, which serves as a proxy for minimum sensitivity.
- We will acknowledge in the revised Section 5.3 that this threshold may lead to slightly lower detection sensitivity in the older parts of the record, and potentially higher sensitivity in the last few centuries due to improved preservation and higher sedimentation rates in the upper 1.2 m of AIG16-05.
Combining the two records with the same threshold (≥0.5 cm) allows us to maximize temporal coverage and maintain coherence in the interpretations of the results. Separate analyses would reduce statistical significance, particularly for recent centuries, given the overlap and continuity observed between the two cores. Because of the region’s moderate seismicity and the limited number of events in the recent centuries, performing a statistical analysis separately on the upper 1.2 m of AIG16-05 would not provide sufficient resolution for a representative recurrence assessment. The combination of both cores increases coverage of the event layers, while being cautious in the interpretation (revised section 5.3., in the new version of the manuscript). The integration of these cores will be discussed in the revised manuscript as well as the implication of sensitivity variation on the most recent part of the record.
Thank you again for providing me with the opportunity to review this manuscript.
Reference:
Distinguishing intraplate from megathrust earthquakes using lacustrine turbidites
Van Daele, C. Araya-Cornejo, T. Pille, K. Vanneste, J. Moernaut, S. Schmidt, et al.
Geology 2019 Vol. 47 Issue 2 Pages 127-130
DOI: 10.1130/g45662.1Thank you again for this constructive review. Please find below the references cited in our responses.
References:
Courboulex, F., Mercerat, E. D., Deschamps, A., Migeon, S., Baques, M., Larroque, C., ... & Hello, Y. (2020). Strong site effect revealed by a new broadband seismometer on the continental shelf offshore nice airport (southeastern france). Pure and Applied Geophysics, 177, 3205-3224.
Ergin, M., Aktar, M., & Eyidogan, H. (2004). Present-day seismicity and seismotectonics of the Cilician Basin: Eastern Mediterranean Region of Turkey. Bulletin of the Seismological Society of America, 94(3), 930-939.
Kremer, K., Wirth, S. B., Reusch, A., Fäh, D., Bellwald, B., Anselmetti, F. S., ... & Strasser, M. (2017). Lake-sediment based paleoseismology: Limitations and perspectives from the Swiss Alps. Quaternary Science Reviews, 168, 1-18.Sabatier, P., Moernaut, J., Bertrand, S., Van Daele, M., Kremer, K., Chaumillon, E., & Arnaud, F. (2022). A review of event deposits in lake sediments. Quaternary, 5(3), 34.
Shynkarenko, A., Cauzzi, C., Kremer, K., Bergamo, P., Lontsi, A. M., Janusz, P., & Fäh, D. (2023). On the seismic response and earthquake-triggered failures of subaqueous slopes in Swiss lakes. Geophysical Journal International, 235(1), 566-588.
-
AC2: 'Reply on RC2', Mathilde Banjan, 22 Apr 2025
-
AC1: 'Reply on RC1', Mathilde Banjan, 08 Sep 2024
-
RC3: 'Comment on nhess-2024-83', Anonymous Referee #2, 11 Mar 2025
The paper presents a paleoseismic record for Lake Aiguebelette, France, that extends back to ~10 ka BCE, based on the identification of event layers within sediment cores attributed to earthquakes, in combination with a robust age-depth model. Key to the paleoseismic interpretations is defining and applying a multiproxy sediment model, using XRF logarithmic geochemical ratios, D50 and D90 grain-size data, high AMS foliation values, and low IRM amplitudes, to identify seismically-triggered event layers. Thirty-two out of 55 identified event layers are inferred to be of coseismic origin, all being ≥0.5 cm in thickness. ‘Good’ and ‘moderate’ confidence levels are assigned to the ≥2 cm and between 0.5-2 cm event layers, respectively. The event layer-derived record is compared to the historical seismic record for the Lake Aiguebelette region and used to develop a paleoearthquake record expanded to 10 ka BCE that includes more moderate, local seismic activity. Modelled curves are fitted to histogram distributions of the number of seismic occurrences versus time gap between depositional events for two time periods of the post-10 ka BCE record. The paper potentially is of interest to an international audience and is within the scope of NHESS.
In the Introduction, the authors correctly identify that “to build a seismic chronicle based on sediment archives, a mandatory step is to demonstrate that event layer deposition is related to a seismically-induced process, in general by using comparison with historical seismicity...” (lines 48-50). However, they do not follow this advice by testing their proxy event layer data against probable seismically-related event layers identified from the combination of the presented age-depth model and the SisFrance and FCAT-17 seismic catalogs. There may be sedimentological characteristics common between the various event layers ≥0.5 cm, but the key unanswered question is whether there is a unique seismic triggering mechanism that is initiating the event layer sedimentation process. That the presented example of the multiproxy is based on a 5 cm thick bed does not provide the reader with confidence of a unique seismic origin for the much more numerous, thinner layers, between 0.5-2 cm. Nor is confidence provided by the widely-scattered, grain-size data shown on Fig. 5 that is compared to background sediments and those from, I believe, the thickest event layer in the record (83.6 cm). Importantly, there is no assessment of other plausible mechanisms for the thinner event layers, including severe flood events are the obvious candidate. Instead, all event layers ≥0.5 cm thick are inferred to represent seismic events and, surprisingly, there is no mention of what alternative triggering process(es) explains the deposition of the 13 layers <0.5 cm. Although ‘good’ and ‘moderate’ confidence levels are applied to these inferences, they are based solely on event layer thickness and are literally presented as ‘suggestions’ that are not supported independently by data. There is no mention if the 32 individual event layers inferred to be of coseismic origin are common to every core, several cores, or only one core. Strangely, and without justification, in Section 5.3 (lines 516-517), the full dataset of 32 event layers (>0.5 cm thick), plus an additional 13 that are <0.5 cm, are considered as seismically induced and used on Fig. 12, Table 3 and in this part of the discussion. What may be important evidence supporting the coseimic explanation for the event layers, >0.5 cm thick, does not appear until late in the paper (lines 565-567). Overall, the lack of testing to demonstrate that the event layers are an effective proxy for seismic events in Lake Aiguebelette is a fundamental weakness of the paper that the authors must address.
This paper, therefore, requires major revision to substantiate that the event layers, >0.5 cm thick, in Lake Aiguebelette are exclusively a proxy for seismic events as the authors infer them to be. This requires presenting defendable reasons why extreme flood events, and perhaps other processes, could also not form the event layers or that thicker flood-related event layers can be distinguished objectively from thinner seismically-related ones. This assessment also needs to use core-to-core frequency data for the event layers that can be presented diagrametically (i.e., Fig. 6), and summarized in a new Table. In particular, data should be presented in the body of the paper that makes it straightforward for the reader to determine if each event layer is common to every core, several cores, or only one core. The authors also need to consider if the latter information can be used to support the confidence level of the seismic interpretations.
There are a number of moderate and minor issues that the authors need to address to improve the content and presentation of the paper, as seen from the numerous editorial comments annotated on the returned pdf, and numbered comments keyed to annotations on the manuscript are in a separate pdf.
-
AC3: 'Reply on RC3', Mathilde Banjan, 22 Apr 2025
Dear Reviewer,
We appreciate your detailed comments, suggestions and corrections throughout the manuscript. We will take them into account in the revised version of our manuscript.
Below, you will find our responses to the global questions and remarks grouped by topics. Minor corrections related to spelling, typos, and grammar highlighted in the annotated manuscript will be taken into account, and the final version will be reviewed by a native English speaker.
In addition, our responses to the 34 numbered comments annotated in the manuscript are also provided below.
All responses are also compiled in an attached PDF for easier readability.
Thank you again for your time and consideration.Best regards,
Mathilde Banjan and co-authors- Event layers as seismic proxies
We would like to emphasize that, unlike most lacustrine paleoseismological studies conducted in tectonically active regions, this study is set in a context of low to moderate seismicity. This setting reduces the frequency and intensity of shaking events capable of triggering thick and easily identifiable deposits, making the detection and attribution of event layers more challenging. The main challenge is that comparison between event layers and historical earthquakes is difficult, as very few significant events (Mw > 5) have been recorded within 30 km of the lake.
However, it also makes the resulting chronicle particularly valuable: it provides rare long-term data in a region where instrumental and historical seismic records are sparse. This context justifies the use of a conservative, multiproxy-based approach to distinguish seismic from non-seismic events and highlights the originality and relevance of our contribution.
You highlight that "to build a seismic chronicle based on sediment archives, a mandatory step is to demonstrate that event layer deposition is related to a seismically induced process" and question whether this has been rigorously applied.
This study aims to do this by using a multi-proxy approach that integrates:- Visual observation of homogenite (Hm) and turbidite-homogenite (Tu+Hm) facies, which can be associated with earthquake-triggered mass transport deposits in lacustrine environments (Beck, 2009; Campos et al., 2013; Howarth et al., 2014);
- Grain-size data trends in a Passega diagram (Fig. 5), to distinguish remobilized versus background sedimentation processes (Beck et al., 1996; Moernaut et al., 2014);
- Geochemical signatures (XRF logarithmic ratios) indicating sediment remobilization from subaquatic slopes, as opposed to catchment-derived detrital input, thus excluding flood deposits as the likely origin (Chapron et al., 2004; Wirth et al., 2013) ;
- High AMS foliation values, a recognized indicator of seiche effects (Strasser et al., 2007; Beck et al., 2022);
- Low Isothermal Remanent Magnetization (IRM) amplitudes, linked to seismically induced redeposition (Yakupoğlu et al., 2022; Banjan et al., 2023).
This suite of criteria has been established in previous paleoseismic studies (Moernaut et al., 2007; Wirth et al., 2013; Beck, 2009) and is detailed in Section 4.2 of the manuscript (lines 195-232). Additionally, the highest confidence level is assigned only to deposits ≥2 cm, as they provide reliable AMS measurements, and these are consistently linked to seismically triggered deposits in comparable studies (Campos et al., 2013).
- Consideration of alternative triggering mechanisms
You suggest that extreme flood events could also be responsible for the observed event layers and question why we do not systematically rule out non-seismic origins. In Section 5.1, we address this issue and demonstrate why floods are not a plausible alternative trigger:
- Lake Aiguebelette has no major river inflows, reducing the probability of large flood-induced deposits (line 482). Moreover, no sedimentological evidence or historical record suggests that extreme floods have ever deposited significant layers in this lake.
- Bathymetric control:
- The main tributary (La Leysse) enters the northern basin (line 75, Preprint).
- The southern basin (where AIG17III and seismic core analysis is focused) has independent, small catchments (<1 km²) and is hydrologically and morphologically separate from the north (lines 218–222, 482–484).
- The bathymetry (Fig. 2) shows that the southern basin is relatively isolated, with intermediate depths and no major inflow connection mentioned or mapped to the northern basin.
The northern sub-basin acts as a sediment trap for the main tributary, La Leysse. The cores used for seismic interpretation (AIG17III, AIG16-05, AIG16-06) are all located in the southern basin, isolated from this fluvial input. This is due to the lake’s internal bathymetry being insufficient to transport coarse sediment from the north to the deep southern basin, where the seismic cores were taken. In contrast, the catchments feeding the southern basin are <1 km² and have limited sediment flux. - No historical records of extreme floods correspond to the depositional age of the identified seismically triggered historical event layers, further reinforcing that seismic shaking is the more likely trigger.
- The geochemical signatures of the biggest event layer deposited at the Younger Dryas - Holocene transition (XRF data) suggest sediment remobilization from lake slopes, rather than a riverine source, which is expected for flood deposits. Additionally, the Passega diagram (Fig. 5) provides a comparison of grain-size trends between event layers and background sedimentation. The clustering of seismic event layers along a distinct depositional trajectory suggests a shared seismic trigger.
- Confidence levels and event layer thicknesses
Your concern regarding confidence levels is addressed through several points summarized below.
The event layers ≥2 cm are classified with "high confidence" because:
- They consistently display high AMS foliation values (Fig. 4), indicative of seismically triggered seiche effects.
- They align with historical earthquakes from the SisFrance and FCAT-17 seismic catalogs (Table 3, SM-6).
- They are found in multiple cores at corresponding depths, reinforcing regional correlation (Fig. 6).
For event layers between 0.5-2 cm, we acknowledge a "moderate confidence" level, but we also validate their seismic origin based on:
- Passega diagram grain-size trends, which show alignment with thicker seismically attributed deposits (Banjan et al., 2023).
- XRF signatures, demonstrating sediment remobilization from the lake slopes rather than flood-related inputs. For example, the correlation coefficient between Ca and Ti (XRF data) is negative in the continuous background sediment above and below the thickest event layer, whereas it is positive in the thickest event layer (an additional figure will be added to the revised manuscript). This effect results from the grain size or sediment source being different in each facies. The geochemical profile of the event layers supports an origin by slope remobilization within the lake rather than catchment-derived flooding. Event layers are enriched in Ti and Fe, suggesting their remobilisation from the lacustrine slopes and background laminated sediment, while showing a relative depletion in K and Zr (mostly associated with detrital input during flood events). This observation aligns with previous lacustrine studies (Giguet-Covex et al. (2010); Sabatier et al. (2017, 2022)) where flood layers exhibit high K and Zr content due to mineral-rich erosion from the watershed. In contrast, the event layers (with sufficient thickness to have more than two geochemical measurements) share the geochemical signature of in-lake remobilized sediment. In addition, the morphological isolation of the southern basin from the main tributary excludes a flood origin.
Thus, we do not assign confidence arbitrarily but on a structured, multi-criteria approach.
- Core-to-core frequency comparisons
You request further clarity on the spatial distribution of event layers across cores. This information is already provided in Supplementary Materials (SM-2, SM-5), where correlations between cores (AIG17III and AIG16-05) are mapped. To further improve clarity, we will include a summary table in the revised manuscript (similar to Table 1) highlighting the presence of each event layer across multiple cores. The figure SM-2 will be placed in the main text as explained in the reply to point 11 below.
- Consideration of thinner event layers (<0.5 cm)
We agree that the deposition mechanisms of thinner layers (<0.5 cm) should be explicitly discussed. These deposits are excluded from the recurrence analysis (Section 5.3, line 516), as we acknowledge that their origin is uncertain. However, we emphasize that the 32 identified seismically triggered event layers (≥0.5 cm) remain the focus of our study, as they meet the criteria for seismic attribution.- Additional comments
Numbered comments keyed to the manuscript
- 1. “little or no flood record” This statement is made without elaboration or reference, but has an important bearing later in the paper. It needs to be substantiated somewhere.
We will explain that the geochemical signatures and basin morphology support the absence of flood record. This point will be discussed in detail in Section 5.1, as already indicated in our response above.
- 2. Provide the lengths of the recovered core sediments.
We will include the lengths of each recovered core in the “Methods” section for clarity.
- 3. This statement is incorrect. Supplementary material SM1 shows a summary and the log for only AIG17III. For a paper focused on core sediments as the primary data source, you need to show all of the cores in the body of the paper. Also, see comment 16.
We will make sure the log of the cores are included in the revised manuscript. All these logs are available in Banjan’s PhD.
- 4. Make it clear that these statements are inferred based on studies of other lakes in the region, not this specific paper.
We will state which interpretations are based on previous studies from other lakes and which are specific to our observations at Lake Aiguebelette.
- 5. Lines 181-185 are out of place. This should be presented in section 4.2.
Lines 181–185 will be moved to Section 4.2, as they are more relevant there.
- 6. Lines 205-218 are too detailed and difficult to follow. Simplify and focus on the key points.
We think it is important to describe precisely how event deposits were characterized. Each criterion will be presented individually in a dedicated sub-paragraph for clarity. We will shorten long sentences and rework the structure, as suggested.
- 7. The reviewer is not experienced with magnetic susceptibility foliation data. However, the difference between background and the exemplified event layer seems very slight (between 1.005-1.01 and 1.02-1.025, respectively). Can one range really be considered as ‘high’ relative to the other? Also, showing the numbers as ≤1.01 and ≥1.02 is deceptive because each range is very narrow! You need to say more about the significance of these differences in foliation data to make this more convincing.
Foliation values of 1.01 and 1.02 correspond to 1% and 2% anisotropy, respectively, meaning the difference represents a doubling in anisotropy. The foliation measured in homogenites is typically between 2 and 5 times higher than in the background sedimentation. The sensitivity of the MFK1 device used to measure magnetic susceptibility and its anisotropy is better than 5×10⁻⁸ SI, while bulk susceptibility in our samples ranges from 50 to 100×10⁻⁶ SI. These values allow us to detect meaningful differences in sediment fabric between event layers and background deposits.
- 8. There is no comment 8.
- 9. Ok, there might be a common depositional mechanism, but is it representative of a common triggering mechanism that puts the sediment into the water column? Can this be determined objectively?
We will specify that the observation of similar grain-size trends between thinner and thicker event layers suggests a possible common triggering mechanism. While this cannot be demonstrated with certainty, it reflects our interpretation based on consistent patterns in the data. This reasoning is common in paleoseismological studies, where multiple indicators are used to build confidence in interpretations.
- 10. This caption does not say this explicitly, but I believe that the thickest event layer in the deep basin is 83.6 cm thick (SM1 and SM6). In either case, the layer may have been interpreted by Banjan et al. (2023) as seismically induced, but neither the presence of this thick event layer nor the Bajan et al. interpretation are proof that every event layer between 0.5-2 cm thick was seismically induced.
We will modify the caption as follows: "Blue dots: grain-size data of the homogenite constituting the thickest event layer of the deep basin sequence, presented in Banjan et al. (2023)." We agree that the comparison with this thick deposit is not proof of a seismic origin for thinner deposits. This limitation will be acknowledged in the revised manuscript.
- 11. The core-to-core correlations are fundamental to the paper, but are not presented in any detail. These correlations also need to be presented after the age-depth modelsince the chronology is critical to the correlating process. The new locaton should be after 4.4.3. Fig. 6 only shows correlations based on the IRM data. No chronology data has been presented for AIG17III at this point in the paper, yet it is being drawn upon for these correlations. In other words, the order of the presented materials is incorrect and needs to be modified.
We will use the SM-2 figure and revise it to present a clear core-to-core correlation based on IRM, XRF, and chronological data. This new figure will be included in the main manuscript and placed within a new Section 4.3 titled “Core-to-core correlation,” which will follow Section 4.4 on the age-depth model. This reorganization supports the correlations with chronological context.
- 12. “To improve the count of the event layers”? What do you mean and how does this work?
We will explain that the correlation with a longer core sequence allows us to detect more event layers, particularly in the recent historical period, improving the record of event layers.
- 13. Fig. 6 shows chronology in years CE, but the ages in SM-2 (Correlations between cores from Lake Aiguebelette deep basin) are in yr BP. For your purposes, using yr CE seems to make the most sense to use. Be consistent in the chronology format between the paper and the supplementary materials.
Keeping both is necessary in this study. The use of yr cal BP (before present) is consistent with standard practice in sedimentology and age-depth modeling, particularly with radiocarbon-based chronologies and long Holocene records. We use the yr cal CE format when discussing historical and instrumental seismic events, for comparison with calendar-dated archives such as FCAT-17 or historical records. We will make sure that the use of both formats is clearly explained in the revised manuscript.
- 14. Are these laminae composed of the triplet of layers that represent a varve? If yes, then consider referring to varve ‘layers’ that are composed of triplets of ‘laminae’.
We prefer the term “lamina” rather than “varve”, as the latter implies an annual deposition, which requires evidence to validate. We will consistently use 'lamina' to avoid overstating the chronological resolution before demonstrating that their deposition is annual to seasonal.
- 15. This paragraph is an inadequate summary of the chronology of event layers. The varves were presented in section 4.1 and the radioisotopic data in section 4.4.1. Why are these themes being introduced here as something new? Modify/reorganize as necessary.
We understand the reviewer’s concern. The presence of varves is indeed established earlier in the manuscript (lines 288–291, Section 4.1), based on previous studies and on the observation of seasonal lamination. However, Section 4.4.1 is the appropriate place to present all dating results, including radionuclide profiles, as it provides the complete overview of the age-depth constraints. This section is structured to present the results and interpretations based on all dating methods used, and it cannot be moved or split, as it consolidates all chronological information. The discussion section later recalls these elements briefly to connect the methodological basis with the broader interpretation of the event layer chronology.
- 16. The “SM-2 Correlations between cores from Lake Aiguebelette deep basin”, which is in Supplementary material, absolutely needs to appear in the paper.
We will move the content from SM-2 into the revised version of the manuscript.
- 17. There is no Table 2.
This refers to Table 2 from Banjan et al. (2023). We will add the reference to make this clear in the revised manuscript.
- 18. No information in the body of the manuscript is given on whether the event layers are common to every core, several cores, or only one core. This is key information that the authors are ignoring and which needs to be used in assessing the interpretations of the various event layers.
This information is available in SM2 and SM5. SM2 will be moved to the main manuscript (revised version). SM5 contains a detailed table that indicates whether each event layer is present in one or both cores, along with the thickness and depth of each event layer. These elements provide the necessary information to support our interpretation.
- 19. Cores AIG16-05 and AIG16-06 are about 100 m from AIG17III yet have different stratigraphies from the latter. See Fig. 2. Are you sure each event layer has a seismic origin?
We will explain that although AIG16-05 and AIG16-06 are located only ~100 m from AIG17III, differences in slope, micro-basin morphology, and sediment transport can lead to variations in deposit thickness or preservation. Sediment availability is a limiting factor: some slope failures may remobilize sediment and only deposit material in one part of the basin. Comparing information across cores helps improve the event layers identification. This variability is taken into account when evaluating the seismic origin through a multi-proxy approach of each deposit and will be mentioned in the revised version of the manuscript.
- 20. This statement is based on circumstantial evidence and, as the authors indicate, is a suggestion. The ‘good confidence level’ of a seismic origin is not supported by data.
In the referred sentence, we made it clear that the 'good confidence level' assigned to event layers >2 cm is based on our interpretation of converging proxy evidence. Specifically, the combination of grain-size data, geochemical markers, high AMS foliation, and low IRM amplitudes supports our assessment that these deposits are likely seismically triggered. This multiproxy agreement across cores forms the basis for our classification and will be emphasized more clearly in the revised manuscript with a reference to Figure 9.
- 21. Again, this is another suggestion. The ‘moderate confidence level’ of a seismic origin also is not supported by data.
In the same way, we will refer this label to Figure 9. This figure will subsequently be placed before these mentions.
- 22. This is a reasonable decision.
This decision will therefore remain unchanged.
- 23. The font size on this chart is far too small to read.
We will make the font size on this chart bigger to improve readability in the revised manuscript.
- 24. Does this mean that one of the events is not recorded in the lake, but would be expected? What does this say about the event layer methodology? You need to comment on this.
We agree that the ESTI approach alone does not allow for definitive matching between event layers and specific historical earthquakes, particularly in a region with limited and incomplete historical records. Our goal with the ESTI method is not to provide a one-to-one correlation, but to identify periods where seismic triggering is plausible based on the archive's temporal resolution and the regional seismicity. The absence of a deposit associated with a known earthquake (or vice versa) highlights the archive's sensitivity limits. We will include a sentence to make this methodological limitation clearer in the new version of the manuscript: “The absence of a sedimentary deposit corresponding to a known regional earthquake (or the presence of an event layer, that was not deposited at a time compatible with a known earthquake) highlights both the sensitivity threshold of the sediment archive and the limitations of the ESTI method, which is not intended to present clear one-to-one correlations but to assess the plausibility of seismic triggering across time windows of sufficient resolution.”
- 25. What is the chance that this event layer does not represent a seismic event at all? This also needs to be considered and mentioned. You also need to make the case for this specific event layer being seismically-triggered if the hypothesis that it represents an unrecorded earthquake is being be stated.
We agree that we cannot exclude a non-seismic origin. This interpretation is presented as a plausible working hypothesis, in line with the approach used throughout this study.
We will rephrase as follows:
The event layer dated between 1327 and 1372 yr cal CE may not correspond to the 1356 CE Basel earthquake but could instead reflect a local moderate earthquake that was not archived in historical catalogs. Given the limited reliability and spatial coverage of historical seismic records for the 14th century, and in the absence of a better alternative, a local seismic origin remains a plausible hypothesis, though it cannot be confirmed with certainty.- 26. This only works if the triplet laminations within varves are present within the overlying sediments, but you do not say they are before making this interpretation. If the varves are not present, then I don’t follow how you are able to make this interpretation.
In the case of the 1760-1824 yr cal CE deposit, we observe a succession of laminated structures immediately above and below the event layer in core AIG16-05 (see Figure 3). These laminations show a geochemical triplet consistent with seasonal deposition (Si-rich in spring, Ca-rich in summer, Al-rich in winter), as described by Giguet-Covex et al. (2020). This continuity of laminated sediment at the top and bottom of the event layer allows us to suggest that the deposit was emplaced between a summer and a winter lamina, supporting the interpretation of an autumn event. We already describe in lines 470-476, the presence of Ca-rich and Al-rich laminae below and above the event layer provides geochemical evidence of seasonal deposition (see Figure 3). These features allow us to constrain the timing of the event to between summer and winter, based on varves continuity. This interpretation is specifically written in the discussion section, which seems to be the right place to explore interpretations in more depth.
There are two additional points that support our interpretation, already mentioned in the manuscript:(1) Core AIG17III presents laminated sediment interpreted as varves from the present day back to the Little Ice Age. Core AIG16-05, deposited in the same basin, also contains evident lamination with geochemical characteristics compatible with varves.
(2) Core AIG20-01, retrieved from the same deep basin, shows laminated sediments that are clearly identified as varves, validated through radionuclide dating. As described in the manuscript, the short-lived radionuclide profiles confirm annual deposition and support the varve chronology. These varves present the same geochemical characteristics observed in AIG16-05.- 27. You need to explain this in more detail.
We will explain this in more detail as follows: It also confirms that this lake is more sensitive to high frequencies of the PSA content. This means that seismic waves rich in high-frequency energy, such as those produced by nearby, moderate-magnitude earthquakes, are more likely to induce sediment remobilization in Lake Aiguebelette than low-frequency, long-period ground motions typically associated with larger but more distant events. This point is supported by the calculated PSA values (Fig. 11), where nearby events show relatively higher spectral accelerations at 5 Hz than distant events. The absence of deposits corresponding to strong but remote earthquakes reinforces this interpretation. The sedimentary archive appears more responsive to local ground motions with higher frequency content.
- 28. Not clear. What “situation” are you talking about?
We agree that the word "situation" is vague. For the 14th century, the lack of reliable seismic records makes interpretation difficult. Please see the answer to the following comment 29 (below), with the new rephrasing.
- 29. Make a more definitive interpretation/statement about the 1327-1372 yr cal CE event layer. Reword lines 490-492.
We will reword lines 490-492 as suggested in the annotated manuscript, which improves the clarity of the sentence. The revised version will read: “1327-1372 yr cal CE: The interpretation is difficult for the 14th century, with gaps in the records. As discussed before, even though the major Basel earthquake may correspond in age with the event layer, it is highly probable that a local moderate earthquake would not be archived in the SisFrance database.”
- 30. The seismically-triggered layers have been defined to be 0.5-2cm and >2 cm, as presented on lines 344-351. On what basis can it now be decided to consider that all of the event layers ≤0.5 cm thick were also seismically triggered? It is not defendable to do so late into the paper. Rethink? Reword?
We are not assigning a seismic origin to all event layers ≤0.5 cm thick. As clearly stated in lines 516-518 of the manuscript, we are exploring two hypotheses: (1) considering all identified event layers as seismically induced, and (2) taking into account only the ≥0.5 cm thick event layers, for which greater confidence can be placed in their seismic origin. These scenarios are tested to evaluate the archive’s sensitivity and the impact of different inclusion thresholds. They are presented as working hypotheses and not as definitive conclusions. This discussion appears in the dedicated discussion section, which seems to be the appropriate place to explore such interpretations and push the limits of scientific reasoning based on the available data.
- 31. The text here summarizes three stages that are shown on Fig. 12D, which implies that the stages are defined based on shifts in the sedimentation rate. Stage 1 spans about 8000 yr, which is about twice that of stages 2 and 3. Stage 1 also consists of two obvious sub-stages between -9500 to -6000 and -6000 to -2000 CE. Why are there not four stages? The text needs to clearly indicate how these stages are defined.
We will write a clearer explanation: the three-stage structure was defined based on shifts in both sedimentation rate and event layer frequency. We will clarify that Stage 1 cannot be divided into two sub-stages, as both sedimentation rate and event layer frequency do not allow it.
- 32. Be clear that you are talking about event layers between 0.5-2 cm thick, which I think you are. However, there are four event layers <0.5 cm between -2000-0 CE. Couldn’t these be from floods? I am not sure that the absence of event layers that could be floods really helps support the inference that the layers between 0.5-2 cm were in fact formed by seismic events.
Yes, we are talking about event layers ≥0.5 cm thick. This will be clearly stated in the revised manuscript. Regarding the four event layers <0.5 cm between -2000 and 0 CE, their limited thickness and the lack of proxy resolution prevent us from determining their origin. We do not interpret these as seismically induced. What is clearly stated in the current manuscript (Figure 12, Table 3) is that no event layers ≥0.5 cm are recorded during this -2000 to 0 CE interval, meaning no reliable marker of seismically induced deposition is archived in the sediments. This is explicitly stated in lines 526–528: “In the event layers chronicle, a period of quiescence is visible between -2000 and 0 yr cal CE (Fig. 12; 13), where no event layer ≥ 0.5 cm is archived in the lake.”
- 33. How the mean sedimentation rates were determined needs to be explained briefly somewhere. Also, lacustrine sediments are subject to compaction over time, which is not apparent from an age-depth model. The determination of these rates need tconsider this or at least it needs to be acknowledge that these are apparent rates that do not consider sediment compaction over time.
The sedimentation rates are derived from the constructed age-depth models (using CLAM age-depth modeling). Sedimentation rates are calculated by determining the change in depth over the change in age between successive points on the curve, expressed in centimeters per year (cm/yr). These rates do not account for post-depositional compaction.
The mean sedimentation rates were calculated over specific time intervals based on the outputs of these models. We will clearly explain that these rates are apparent and do not consider sediment compaction over time. This information will be clearly stated in the revised manuscript.- 34. The utility of the modelled curves fitted to histogram distributions of the number of seismic occurrences versus time gap between depositional events (see lines 582-614) is outside of the expertise of the reviewer. Nevertheless, the reviewer is not sure that anything meaningful is represented by these curves. Delete?
We consider the modelled fit curves to be relevant, as the specific distributions observed (time-gap frequencies between events) are characteristic of seismically triggered deposits and recurrence patterns (Goldfinger et al., 2012; Hubert-Ferrari et al., 2020; Strasser et al., 2013). These curves help to evaluate temporal clustering, recurrence variability, as well as potential episodic seismic activity.This is already addressed in the manuscript (lines 584-594), where we present how different statistical laws (Weibull, lognormal, gamma, and exponential) correspond to specific types of seismic behavior. These modelled distributions, applied to time-gap frequencies between events, enable us to assess whether the observed recurrence intervals reflect a known pattern.
Goldfinger, C., Nelson, C. H., Morey, A. E., Johnson, J. E., Patton, J. R., Karabanov, E. B., ... & Vallier, T. (2012). Turbidite event history—Methods and implications for Holocene paleoseismicity of the Cascadia subduction zone (No. 1661-F). US Geological Survey.
Hubert-Ferrari, A., Lamair, L., Hage, S., Schmidt, S., Çağatay, M. N., & Avşar, U. (2020). A 3800 yr paleoseismic record (Lake Hazar sediments, eastern Turkey): Implications for the East Anatolian Fault seismic cycle. Earth and Planetary Science Letters, 538, 116152.
Strasser, M., Monecke, K., Schnellmann, M., & Anselmetti, F. S. (2013). Lake sediments as natural seismographs: A compiled record of Late Quaternary earthquakes in Central Switzerland and its implication for Alpine deformation. Sedimentology, 60(1), 319-341.
Conclusion
Your global comment highlights the importance of rigorously establishing seismic attribution for event layers, and we have shown that our methodology fits this requirement. Our use of multi-proxy sedimentological analysis, historical seismic correlations, and core-to-core comparisons provides a robust basis for identifying seismically triggered event layers. We will make clarifications in the revised manuscript (and include a summary table of event layers) to ensure that the conclusions are additionally substantiated.We appreciate your feedback and hope that this response addresses your concerns.
Best regards,
Mathilde Banjan and co-authors
-
AC3: 'Reply on RC3', Mathilde Banjan, 22 Apr 2025
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 618 | 108 | 168 | 894 | 54 | 43 | 52 |
- HTML: 618
- PDF: 108
- XML: 168
- Total: 894
- Supplement: 54
- BibTeX: 43
- EndNote: 52
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1