the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Mar 2025
Research on the Extraction of Pre-Seismic Anomalies in Borehole Strain Data of the Mado Earthquake Based on the SVMD-Informer Model
Abstract. Earthquake is a major natural disaster triggered by the accumulation and release of crustal stress, and the accurate extraction of pre-seismic anomaly signals is crucial to improve the earthquake prediction capability. In this study, an anomaly detection method for borehole strain data based on the combination of Segmented Variational Modal Decomposition (SVMD) and Informer network is proposed, and a pre-seismic anomaly extraction study is carried out for the 2021 Mado Ms7.4 earthquake in Qinghai. The SVMD method effectively solves the memory limitation problem of traditional Variational Modal Decomposition (VMD) when dealing with large-scale data through the sliding-window mechanism, and at the same time maintains the correlation between the data. The Informer network significantly reduces the computational complexity of the long-series prediction and realizes the high-precision one-time long-time series prediction by utilizing its ProbSparse self-attention mechanism and self-attention distillation. By analyzing the borehole strain data from the Mengyuan station, this study identifies the accelerated anomaly accumulation phenomenon in the two stages before the Mado earthquake: in the first stage, the number of anomalous days shows an accelerated growth starting from about three months before the earthquake (February 13, 2021); in the second stage, the anomalous accumulation tendency is further intensified since the second month before the earthquake (the end of March, 2021), and the accumulation curve shows a typical S-shape growth characteristic. The results are highly consistent with the time windows of the index of microwave radiation anomaly (IMRA), outward long-wave radiation (OLR) and geoelectric field anomalies, and with the subsurface-to-atmosphere multilayer anomalies (e.g., Benionff strain, CO concentration, electron concentration anomalies, etc.), which indicate that the borehole strain anomalies are closely related to the gestation process of the Mado earthquake. This study provides a new method for the extraction of pre-seismic anomalies based on machine learning, and provides an important basis for understanding earthquake precursors.
- Preprint
(1644 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-1130', Anonymous Referee #1, 05 Apr 2025
Since the M6.9 Menyuan earthquake occurred close to the Menyuan station, I would like to ask: why do the authors interpret the signals recorded at this borehole strain gauge station as precursors to the Maduo (Mado) earthquake, which occurred more than 400 km away? Unless the authors can clearly explain how this signal can be attributed to the Maduo earthquake rather than to unrelated phenomena or local seismic events, I find the current conclusions unconvincing.
Specific comments:
- Introduction: The section cites too many references, each describing different precursor phenomena from strain gauges, making the narrative overly verbose and lacking synthesis. Please consider summarizing the key findings of these previous studies to provide a more coherent and concise background.
- Line 64: Could you provide more details about the borehole strain gauges used to record earthquake precursor signals, such as installation depth, sampling frequency, and observation duration?
- Line 85: Please elaborate on the cited literature. For example, what findings did Liu et al. (2014) present? What specific anomalies were observed before and after the Lushan earthquake? Did the PBO program detect reliable precursory signals for particular earthquakes?
- Line 79: You mention that “Zhu et al. (2020) observed the Wenchuan earthquake precursors by analyzing the eigenvalues and eigenvectors of the borehole strain data.” Could you clarify how these mathematical features are physically related to pre-seismic processes?
- Line 121: What does SVMD stand for? Also, the manuscript lacks a clear explanation of how SVMD improves upon conventional VMD. Please provide comparative analysis in terms of computational efficiency, memory usage, or signal decomposition performance.
- Line 127: Please provide the full forms of GRU and LSTM when first mentioned.
- Lines 134 & 180: Please correct “Mado earthquake” to “Maduo earthquake”.
- Line 135: Why was the Menyuan station selected for detecting precursors of the Maduo earthquake? A significant M6.9 earthquake occurred near Menyuan on January 1, 2022. It is more plausible that the signals from the Menyuan station correspond to the local mainshock rather than to the Maduo earthquake over 400 km away.
- Line 140: When decomposing the borehole strain data, is there a risk that potential precursor signals might be lost or distorted in the process?
- Lines 190–195: As noted by the authors, I believe the signals described may be related to the Menyuan earthquake in 2022 rather than the Maduo earthquake in 2021.
- Line 230: You use a 7-day sliding window with a 1-day step for anomaly detection. Please justify this choice and provide statistical or empirical evidence supporting the selected window size and step.
- Line 334. The study uses strain data from January 2020 to May 2021. However, this relatively short time span may not sufficiently capture long-term pre-seismic cycles. Please discuss the temporal adequacy of the dataset.
- Line 340: Key model parameters (e.g., SVMD bandwidth = 2000, convergence threshold = 1e-7, Informer layers = 128) are presented without any sensitivity analysis. The authors should evaluate how parameter changes affect signal decomposition quality (e.g., signal-to-noise ratio). Additionally, please include uncertainty bounds in prediction intervals (e.g., in Figure 7) to better reflect model confidence and variability.
- Line 340: The patterns shown appear more indicative of the coseismic effects of the Maduo earthquake, rather than its precursors. Please consider adding information in Figure 6 to clarify and support the presence of pre-seismic signals.
- Line 380: The two-stage S-shaped anomaly accumulation (Figure 8) is interesting, but no underlying mechanism is provided. Moreover, the upward trend of pre-seismic acceleration could plausibly be associated with the Menyuan earthquake in January 2022 instead of the Maduo event.
- Line 415: The study references various anomalies (CO, TBB, electron density), but does not sufficiently address the possibility of false positives caused by anthropogenic or environmental factors. Please discuss the limitations of borehole strain data, including potential sensitivity to non-tectonic influences such as groundwater fluctuations or temperature changes.
Citation: https://doi.org/10.5194/egusphere-2025-1130-RC1 -
RC2: 'Comment on egusphere-2025-1130', Anonymous Referee #2, 07 Apr 2025
The methodology of this work is based on the article published in Scientific Reports | (2023) 13:20095. In general terms, I think the authors should clarify some aspects of their manuscript. The objective of this study is to identify any precursor signals to the Maduo earthquake. Some suggestions and clarifications are listed below:
In the introduction:
1) In the first part, the authors could be more specific about the different works they cite.
2) The authors should clarify the use, operation, and scope of borehole strain gauges in terms of the signals they measure.
3) The paragraph "By mounting strain gauges deeper in the bedrock, borehole strain gauges are able to continuously record stress and strain data, making them a key tool for monitoring crustal deformation" is unclear since it does not specify what type of borehole strain gauge they use and how they ensure the results mentioned. Furthermore, it does not specify how they physically justify the scaling suggested in the paragraph: "The high-resolution recordings provided by borehole strain gauges allow us to capture small changes in strain, thus providing accurate data to support a deeper understanding of crustal deformation processes."
4) On line 80, they should specify the methodology that led to the discovery of precursors by analyzing eigenvectors and eigenvalues. Where did they get it?
5) Correct Maduo instead of Mado on lines 134, 140, 180, 335, and elsewhere.
In the Results and Discussion section,
6) The authors used the SVMD-Informer network to extract preseismic anomaly signals from the Mado earthquake from well deformation data at Mengyuan Station. The authors state that anomalies associated with the earthquake were recognized when the raw data exceeded the corresponding upper or lower limits. The question is: What is the physical basis for determining the criteria described in paragraphs 366 to 368?
7) If an earthquake is considered to correspond to a phase transition, are the results shown in this analysis associated with the critical state or preparation mechanism of the seismic process?
The conclusion section must be improved
Citation: https://doi.org/10.5194/egusphere-2025-1130-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 116 | 27 | 8 | 151 | 8 | 7 |
- HTML: 116
- PDF: 27
- XML: 8
- Total: 151
- BibTeX: 8
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1