the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 Sep 2025
Statistical Evaluation of Well Water Anomalies as Potential Precursors to Large Earthquakes
Abstract. Macroscopic anomalies such as well water fluctuations and unusual animal behaviour have been recorded in old documents and folklore accounts as possible earthquake precursors. However, no reliable earthquake prediction method based on such anomalies has ever been established. One of the reasons is that the same phenomena often occur independent of earthquakes. This study conducted a re-examination using shallow well water observation data recorded for eleven years by a volunteer observation network in Japan. While some critics have claimed that it was impossible to predict an earthquake using well water anomalies, especially shallow ones, the results suggest that some plausible anomalies may serve as true precursors. On the other hand, many anomalies were observed without associated earthquakes, and earthquakes without associated anomalies were also uncovered. The old documents and folklore may contain elements of truth, while unrecorded failures are hidden behind successful accounts. Credible anomalies were limited, and major earthquakes rarely occurred within the same focal region. Furthermore, maintaining systematic observations has been challenging. Even an eleven-year observation period was too short to empirically verify the feasibility of an earthquake prediction method for disaster mitigation. Therefore, continuous support is needed for empirical research on earthquake precursors.
- Preprint
(4355 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 30 Oct 2025)
-
RC1: 'Comment on egusphere-2025-4444', Anonymous Referee #1, 06 Oct 2025
reply
-
AC1: 'Reply on RC1', Yoshiaki Orihara, 19 Oct 2025
reply
We thank the reviewer for the time and effort spent reviewing our manuscript.We sincerely appreciate the comments, which have helped us significantly improve the manuscript. Our response is as follows.# Regarding the verification of the accuracy of measurement data collected more than 30 years ago.We had briefly mentioned this in Lines 41–44 and 50–54, but we acknowledge that the explanation was insufficient. We would therefore like to add the following details. According to the Catfish Club, groundwater levels were measured using digital water level gauges with an accuracy of ±0.5–1 mm. A float-type digital water level gauge was used for wells with a diameter of 15 cm or more, while a tester-type gauge was used for wells with a smaller diameter. A rain gauge and a barometer were also installed at each observation site.The observation wells were former drinking water wells, 5–20 m deep, and measurements were conducted once or twice per day. As self-recording devices were expensive, measurements were visually read from non-self-recording digital devices. We consider the measurement error to have been small, as standardized instruments were used and — although the values were transcribed by hand — the water level readings themselves were digital.Regarding temporal accuracy, the “Black and White” Score Table was compiled on a daily basis. Therefore, a time discrepancy of even a few minutes would not have affected the results. The recorded data were sent by postcard to the Catfish Club headquarters, where they were plotted as graphs. The president, the secretary, and an experienced observer jointly evaluated whether anomalies were present. Taking into account the characteristics of each well, they judged non-rainfall-induced and non-pressure/pumping-related rises or falls as earthquake-related anomalies.Although the exact criteria of the Catfish Club were not documented, we determined that the process was objective and unbiased because it involved three independent experts rather than a single individual. Therefore, we consider the daily “Black and White” records to be sufficiently objective data for re-examination.# Regarding the analysis parameters, such as the subjective selection of a 7-day thresholdWe agree that the manuscript lacked an explanation for choosing the 7-day threshold. We will therefore add the following clarification: In this study, a threshold of 7 days was adopted from the perspective of practical disaster mitigation. Specifically, in Japan, in preparation for the next Nankai Trough mega-earthquake, the government plans to issue an “Emergency Information for the Nankai Trough Earthquake (Mega-Earthquake Alert)” if any anomalous phenomena are detected, and citizens may be urged to evacuate up to one week (7 days) in advance (1). This 7-day period is based on a Cabinet Office survey of local governments regarding the amount of time required before serious social disruption occurs once an evacuation advisory is issued (2). In other words, 7 days was chosen as the maximum duration for which citizens are expected to be able to sustain a precautionary evacuation.(1) Cabinet Office, Government of Japan (2019). Guidelines for Formulating Disaster Risk Management Measures Based on Various Nankai Trough Earthquake Scenarios (1st Edition).(2) Working Group on Contingency Planning for Various Scenarios of the Nankai Trough Earthquake (2019). Final Report on Examination of Disaster Countermeasures for the Various Scenarios of the Nankai Trough Earthquake. Cabinet Office, Japan.# Regarding the variability of resultsIt is true that the results exhibit variability. Not all wells showed anomalies before earthquakes, and even in wells where anomalies were observed, they did not appear before every target earthquake. However, this study explicitly aimed to reveal such variability by analyzing data from multiple locations over a long observational period. Statistical tests confirmed that, in some wells, pre-earthquake anomalies were significant. Nevertheless, even in those wells, anomalies sometimes occurred without an associated earthquake, and earthquakes also occurred without preceding anomalies. There were wells where no anomalies were ever detected, as well as wells where anomalies were observed but not associated with any earthquake. We believe it is scientifically meaningful that this study was able to quantitatively reveal and characterize such variability.# Regarding the lack of a reasonable hypothesisWe acknowledge that the Catfish Club newsletters, which include the “Black and White” Score Table, do not provide sufficient information to derive a clear physical hypothesis. Rather, the purpose of this paper is to emphasize that continued empirical research is necessary to elucidate the mechanisms—specifically, why some wells exhibit anomalous pre-earthquake water-level changes while others do not.# Regarding the use of old dataIf old data were considered fundamentally unacceptable, historical earthquake research would be impossible. This study demonstrates that even decades-old observational records can yield new scientific insights when subjected to modern statistical analysis.Line 25–28 – Duration prior to earthquakesRegarding Orihara et al. (2014), the observation period before the anomaly appeared was approximately 3 years and 5 months.Line 29 – Clarification on the “Catfish Club”The Catfish Club is not a private volunteer group, but an officially organized volunteer collaboration group.Line 30 – Clarification on the research institute and its relation to the ClubThe Hot Springs Research Institute of Kanagawa Prefecture is a governmental research institution of Kanagawa Prefecture. It has supported and collaborated with the Catfish Club, but the Club itself is independently operated.Line 32 – Definition of “short-term”In earthquake prediction research, “short-term” generally refers to the timescale of a few days to several weeks.Line 35 – How much earlier anomalies were observedFor the 1978 M7.0 Izu Ohshima Kinkai EQ, precursor anomalies were observed, with the duration of the anomalies (days before the event) ranging from 3 to 24 days. Specifically, observations at the TKY sites (e.g., TKY01, TKY03, TKY12, TKY14) showed short-term anomalies ranging from 3 to 9 days, while the longest duration of 24 and 15 days was recorded at SZK03 and SZK05, respectively.For the 1978 M7.4 Miyagi-Oki EQ, anomalies were observed 21 days before the event at station MYG03 and from 17 days to 14 days before and 2 days before at MYG04.Line 39 – “No anomalies and no EQs”This situation was described simply for the purpose of explaining the structure of a contingency table.Line 40 – Further description locationWe will add: “The tables are further described in Section 2.1.”Line 42 and 55 – Removal of effects from rainfall, pressure, pumpingThree experts examined available information on external factors and assessed whether the observed anomalous changes could be attributed to rainfall, atmospheric pressure variations, or pumping activities.Line 59 – Binary nature of the tablesYes, the score tables only include binary records (Black and White).Line 60 – Total number of available stationsThe original database contained 234 stations, of which 108 were selected for this study.Line 61 – Wells with temperature anomaliesThe two stations were hot spring wells owned by local residents. The temperature anomalies were evaluated by the same experts who assessed the groundwater-level changes.Line 61 – Extra “were”Thank you for pointing this out. We have removed the extra “were” and corrected the sentence as follows:“Most of the anomalies presented in this paper indicated unusual groundwater level changes, except for two stations where changes in the groundwater temperature were observed.”Line 64 – “Few days”The exact number of days could not be verified from the available data. We will remove this sentence from the manuscript.Line 64 – Number and distance of anomalous stationsThe anomalies were observed at two stations shown in Fig. 4b. The distance between these two stations was approximately 150 km.Line 69 – Definition of “close together”We agree that this wording was ambiguous. In this study, “close” refers to a distance within approximately 100 m.Line 70 – Independence of nearby stationsWe did not combine such stations. For example, KNG46 and KNG47 were ~40 m apart. Both observed a precursor anomaly before the M7.0 earthquake on 14 January 1978, but only KNG46 observed a false anomaly from 14–17 December 1977. Therefore, they were treated as independent stations.Line 70 – The total number of stationsWe have not yet counted that number. We kindly ask for a little more time to provide the exact value.Line 81 – Missing “to”Thank you for pointing this out. We have corrected the sentence to:“By applying the above criteria, the number of selected EQs was reduced to 18 (see Table 1).”Line 86 – Rationale for 7-day thresholdIt is generally considered that larger earthquakes tend to be associated with longer lead times of macroscopic anomalies. In this study, we focused on earthquakes expected to cause severe damage. As explained above, we adopted the 7-day threshold based on its relevance to practical disaster mitigation policy in Japan.Line 95 – Inclusion/exclusion rule for missing daysAs the referee pointed out, we did not refer to reasonable explanation. Since the lead time was set to 7 days, we revise the rule as follows: If there were lack of measurements for 8 consecutive days, including the day of the earthquake occurrence, the station should be regarded as an undetectable station, and if there was measurement for even 1 day, it should be regarded as a detectable station.Line 95 – Frequency of missing daily dataThe number of missing days per station could be calculated by subtracting the number of observed days (Table 2) from the total observation period (3710 days). If a missing day occurred immediately before a target earthquake, we will apply the revision mentioned above.Line 113 – Extra “like”Thank you for catching this. We have corrected the sentence to:“The anomalies did not exhibit seasonal variations at any of the 46 stations.”Table 1The high hit rate immediately after the Club’s establishment, followed by a marked decline thereafter, is acknowledged. Two earthquakes that showed a 100% hit rate shortly after the Club was founded (Nos. 2 and 3 in Table 1) correspond to stations MYG03 and MYG04, as shown in Figure 4. However, for the 19 January 1981 M7.0 event, no anomalies were observed at MYG04, and MYG03 was not in operation at that time. MYG01, which might have been expected to detect anomalies for the 19 January 1981 M7.0 and the 1 December 1986 M6.0 events, did not detect them. It is possible that MYG01 was intrinsically a well that does not exhibit earthquake precursors or there were other possibilities. For these reasons, we contend that an 11-year observation period remains insufficient to resolve such issues conclusively.That only KNG63 exceeded the 30 cm threshold for the January 1978 event does not necessarily undermine the entire dataset; rather, it suggests that even in the pre-instrument era one well may have recorded detectable precursory changes. Conversely, for the remaining 23 stations the availability of measuring instruments may have been a key factor in detecting anomalies. The principal contribution of this study is to show—using data—that while single success stories may be transmitted and remembered, many failures are not; our analysis makes that bias explicit.Citation: https://doi.org/
10.5194/egusphere-2025-4444-AC1
-
AC1: 'Reply on RC1', Yoshiaki Orihara, 19 Oct 2025
reply
-
RC2: 'Comment on egusphere-2025-4444', Anonymous Referee #2, 16 Oct 2025
reply
General characteristics of the manuscript
The manuscript is devoted to the analysis of materials from daily observations at 108 near-surface wells (depths up to 5-20 m), which were carried out by volunteers over 11 years from February 1, 1977 to March 31, 1987. The source materials are tables containing daily assessments of the presence or absence of anomalies in water level changes in the form of black or white symbols, respectively. This volunteer group was called the "Catfish Club" and was organized by the Kanagawa Prefectural Hot Spring Research Institute, Japan.
The author selected 18 earthquakes (Table 1) based on a specific criterion (p. 4) that takes into account the relationship between earthquake magnitude and its epicentral distance to the stations. It was assumed that these earthquakes could be accompanied by precursor anomalies in water level changes.
The alarm time was determined from the date of the anomaly in water level changes plus 7 days. When anomalies were observed consecutively, all anomaly days and the continuous 7 days were considered alarm days. If another anomaly appeared before the end of the 7-day interval, the alarm days continued. The diagram for dividing time periods into “false alarm days” and “successful alarm days” is shown in Fig. 1.
Further statistical processing of observation materials at individual stations for the purpose of assessing the effectiveness of the manifestation of anomalous changes in water levels (black symbols) is traditional and is presented in the section 3 and in Table 2.
As a result, the author argues that some well water anomalies were statistically significant precursors of earthquakes. However, numerous "missed targets" occurred.
The overall conclusion of the study is quite ambiguous and boils down to the fact that macroscopic well water anomalies (rises and falls in water level with amplitudes of tens of centimeters) contain elements of truth, along with a large number of recorded "false alarms" and "missed earthquakes (targets)."
General impression of this work
I was interested to learn about the history of volunteer work in Japan on monitoring near-surface groundwater to detect anomalies preceding strong earthquakes. I also found the presentation of a methodology for statistically processing formalized data on the presence/absence of "anomalies—black symbols"—in a well network in relation to earthquakes (Table 1) helpful.
At the same time, there are a number of comments regarding the lack of description of the observation wells, methods of processing primary information by experts to decide on the presence of an anomaly (black symbol) and many other details that were later clarified during specialized observations of changes in the level of groundwater.
- It is currently generally accepted that the effects of individual earthquakes do not manifest themselves in changes in the water level of near-surface groundwater. The exception is co- and post-seismic effects resulting from the liquefaction of loose sedimentary deposits during noticeable shaking during strong earthquakes.
However, the manuscript lacks data on the composition of water-bearing rocks in individual wells, as well as on the intensity of shaking during individual earthquakes.
It follows from the text and the bibliography that the author is unfamiliar with the results of precision well water level observations (observation frequency ≥1 Hz), which are widely conducted in Japan, the USA, China, Russia, and other countries since the end of the twentieth century.
These high-frequency observations have shown that the effects of strong earthquakes can include not only precursors but also changes in groundwater level/pressure due to the dynamic action of seismic waves, as well as co-seismic effects due to changes in the static stress state of water-bearing rocks. When a well is located in the near zone of a strong earthquake, co-seismic effects in changes in water levels can be significant and long-lasting. Therefore, the time of observation of the effect of changes in the water level at the well with an accuracy of minutes-seconds is of great importance аs well as indicating the time of earthquakes.
These aspects of water level/pressure fluctuations during strong earthquakes are not reflected in the manuscript under review. However, a critical approach to the source data used for statistical analysis is necessary in this work. The study doesn't pay enough attention to linking the timing of earthquakes and recorded anomalies. Therefore, the nature of the anomalies, even if they occurred, is unknown.
Unfortunately, this article promotes the false and outdated notion that all anomalies in well water level changes are precursors to earthquakes. Therefore, this manuscript in its present form is not recommended for publication.
However, with some revision, this material may be of interest to the scientific community interested in the historical aspect of studying earthquake precursors and the need for statistical processing of data on them.
Recommendations
- The title of the work «Statistical Evaluation of Well Water Anomalies as Potential Precursors to Large Earthquakes» should be changed to remove the mention of precursors.
The term “precursors” can be replaced by “identified anomalies” or another identical one, referring to the black symbols in the Score Table, which is more consistent with reality, taking into account modern ideas about the impact of seismicity on groundwater and significant shortcomings in the observation methodology in 1977-1987.
2. The Introduction should provide an overview of current understanding of seismicity effects in groundwater based on well observations in Japan and around the world. This will allow for a critical assessment of the work of volunteers and experts between 1977 and 1987.
3. It would be desirable to provide an additional table with more detailed characteristics of individual points and observation times at each station. Table 2 provides only their coordinates.
4. It is advisable to compare the earthquake data from the JMA in Table 1 with data from international catalogs. Specifically, the number of earthquakes with Mw = 6.0 or greater for the region under consideration from the NEIS GSUS catalog was 30 events (?).
- It is necessary to supplement the list of references with modern works on the problem under consideration.
After eliminating the aforementioned uncertainties regarding the stations, the method for identifying "anomalies," and clarifying the relationship with earthquake timing and intensity, the results of the statistical analysis could have scientific significance.
-
AC2: 'Reply on RC2', Yoshiaki Orihara, 27 Oct 2025
reply
We thank the reviewer for the time and effort spent reviewing our manuscript. We sincerely appreciate your comments, which will help us significantly improve the manuscript.1. Temporal Resolution of Observation DataThe well water level data, referred to as the "Black and White Table" data, was collected on a daily basis with measurements taken only one to two times per day. Therefore, it lacks the necessary resolution to identify rapid co-seismic changes.2. Revision of Section 3.4: "How did well water anomalies appear?"We will significantly revise and expand Section 3.4 to incorporate a more critical and comprehensive discussion of the well water anomalies. The following key points will be added to the discussion:Initial Candidate Selection vs. Outcome: The well water anomalies identified by the three experts were initially considered candidates for earthquake precursors. However, through our quantitative analysis, we will explicitly show that a large proportion of these signals were not true earthquake precursors.False Alarm Rate Quantification: We will demonstrate the low success rate of the historical alarm system. The ratio of successful alarm days to all alarm days was 0.14 (14%) (as noted on Line 115). This means that 86% of the alarm days constituted false alarms.Statistical Caution: While the statistical test (p-value below 0.05) was met at 23 out of 108 observation stations (approximately 21%), we will emphasize that this statistical significance alone does not completely guarantee that the anomalies observed at these stations prior to earthquakes were "true" earthquake precursors.Hydrological Constraints: A critical point is that many of the wells used by the Catfish Club were unconfined groundwater wells. Detecting volumetric strain changes as water level changes is significantly more challenging in unconfined aquifers (e.g., Bredehoeft, 1967).Confined Well Possibility: Conversely, an old newsletter suggests the MYG03 well, which showed precursory water level anomalies confirmed for both target earthquakes, was correlated with atmospheric pressure changes (The Catfish Club, 1978). This correlation suggests the possibility that MYG03 was a confined groundwater well, where strain-related signals were detected.Conclusion: Despite the limitations in historical data (such information from newsletters was regrettably limited), the possibility remains that the collection of candidates observed by the Catfish Club did include genuine "true" earthquake precursor signals. We will conclude by noting that to clarify those problems the empirical verification of earthquake precursor phenomena requires a continuous, long-term research system and dedicated support.
Bredehoeft, J. D., Response of well-aquifer systems to earthtides, J. Geophys. Res., 72, 3075-3087, (1967).
The Catfish Club, Hot Springs Research Institute of Kanagawa Prefecture, Japan.: The Miyagiken-oki Earthquake of 1978, Bull. Hot Springs Res. Inst., 10(1), Catfish Letters 5, (1978).Citation: https://doi.org/10.5194/egusphere-2025-4444-AC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,006 | 33 | 15 | 1,054 | 13 | 7 |
- HTML: 1,006
- PDF: 33
- XML: 15
- Total: 1,054
- BibTeX: 13
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Orihara has presented an interesting theory of using shallow well water fluctuations as precursors for earthquake events. They have compared data collected by a volunteer group in late 1970s and 80s to the occurrence of earthquakes during that period.
Based on my review of the manuscript, I do not believe that the analysis is comprehensive enough to support the author’s conclusions. There is no review of the accuracy of measurements, which had been collected over 3 decades ago. The analysis parameters such as selecting a threshold of 7 days, are fairly subjective. From the author’s own analysis, the results have high variability. Moreover the author has presented no reasonable hypothesis for the reasons behind fluctuations in shallow well water levels and earthquake occurrences. In this reviewer’s opinion, the data is too old, the collection and analysis too subjective, and the conclusions too inconclusive to warrant publication of this manuscript.
I have listed some additional comments below: