Hydrogeological Earthquake Precursors: A Case Study From the Kamchatka Peninsula

The relevance of hydrogeological precursors (HGPs) study is justified by the need to obtain reliable information about the spatio-temporal manifestations and the relationships of HGPs with the parameters of subsequent earthquakes for seismic forecasting. In the review the data on repeated manifestations of HGPs before strong earthquakes obtained from long-term observations in five deep wells on the Kamchatka Peninsula (Far East of Russia) are presented. The analysis of the correlation of HGPs occurring in several wells is carried out in comparison with earthquake parameters characterizing both earthquake sources (magnitude, linear size of the source) and the impact of earthquakes in the area of wells (specific energy density in wave, intensity of shaking). It is shown that the manifestation of HGPs in several wells is observed before earthquakes with M w = 6.6–7.8 at epicentral distances up to the first hundreds of km to observation wells in the near and intermediate zones of the sources with the ratio of the epicentral distances and the source sizes no more than 1–5. A feature of our study was the use of certain types of HGPs in water-level changes for predictive assessments of the strong earthquakes in the Kamchatka Peninsula. The review presents precursors in water-level changes detected in real time and the corresponding earthquake forecasts, which were recognized as successful according to the conclusions of the expert council on earthquake prediction.


INTRODUCTION
Studies of hydrogeological precursors of earthquakes, manifested in changes in physical and chemical parameters of groundwater HGPs have been carried for decades (Roeloffs, 1988;Thomas, 1988;Wang and Manga, 2010;King, 2018;King and Manga, 2018;Skelton et al., 2019). The relevance of studying such phenomena is justified by the need to obtain reliable information about the spatio-temporal manifestations and the relationships of HGPs with the parameters of subsequent earthquakes for seismic forecasting. The basis for studying the HGPs for predicting earthquakes is the systematization of their recurring manifestations according to the data of long-term observations in individual wells. Unfortunately, there are few such reliable data on HGPs. In the above-mentioned works, it is reported mainly about "supposed" precursors in the parameters of groundwater. The lack of a sufficient number of such "supposed" precursors does not allow them to be used for seismic forecasting and obtaining of such new data is the primary task of further research.
The review presents data on HGPs obtained from long-term hydrogeochemical and level observations in five wells on the Kamchatka Peninsula. These data are poorly known, because most of the previous publications about HGPs in wells of Kamchatka were in Russian. Brief information on hydrogeochemical observations in the Kamchatka Peninsula is given in (Wang and Manga, 2010). In Biagi et al. (2000a) and Biagi et al. (2000b) some anomalies in the chemical composition of groundwater and gas before the Kamchatka earthquakes of the 1990s were described.
Our review is aimed to a description at both of data on HGPs in the Kamchatka wells, and an experiment for predicting earthquakes in real time using some types of HGPs. The data on HGPs in previously works of authors and other researchers along with the graphical presentation of the HGPs (see Supplementary Materials S1) provide detailed information on the observation wells, methods of observations and the features of HGPs manifestations in the individual wells. These materials form the basis of generalization about the relationship of the considered HGPs with parameters of the subsequent earthquakes.
An important feature of our long-term study of HGPs in water-level changes is the combination of observations in wells with an experiment on the use of certain types of HGPs for predictive assessments of the strong earthquakes in the Kamchatka Peninsula. This review presents examples of the precursors in water-level changes detected in real time, and forecasts of strong earthquakes recognized as successful according to conclusions of the expert council on earthquake prediction working in the Kamchatka Krai.

OBSERVATIONAL DATA, PRECURSORS, COOPERATION WITH THE EXPERT COUNCIL
The Kamchatka Peninsula ( Figure 1A) is part of the Kamchatka Krai in the Russian Far East. It is located at the junction of the Pacific oceanic plate with the Eurasian and North American continental plates. This region is prone to frequent strong earthquakes which cause damaging ground motion and devastating tsunami. The strategies for reducing negative impacts from the earthquakes include, inter alia, the monitoring studies aimed at establishing the precursors and forecasting the time of the strong earthquakes.
Kamchatka Branch of the Geophysical Survey of the Russian Academy of Sciences (KB GS RAS) conducts long-term groundwater-level and chemical composition monitoring at a network of wells in the Petropavlovsk-Kamchatsky testing site ( Figure 1B) in order to search for hydrogeological precursors of earthquakes and other effects of seismicity in groundwater parameters . The data on the wells with repeated HGPs occurrences are presented in Supplementary Materials Table S1 and Figure S1. In all considered wells, water is not pumped out, as well as other man-made activities do not affect their natural state.
During the visits to the flowing wells GK-1, M-1, and G-1, the employees of the Laboratory of hydroseismology carried out flow rate measurements and water sampling every three days in 1986-1999. Free gas sampling was conducted from the GK-1 well. Water samples were tested for pH as well as for the anion and cation concentrations. The gas content was analyzed chromatographically. Water and gas parameters were determined with an accuracy of 2-10% (Khatkevich and Ryabinin, 2006). After 1999, the system of hydrogeochemical observations had changed due to an increase in the observation interval, a decrease in the set of determined parameters of the groundwater composition and experiments with the installation of equipment into wellbores that violate the natural hydrodynamic and hydrogeochemical regime of wells. Therefore, we consider the hydrogeochemical data only for the period 1986-1998, for which uniform time series were obtained under the conditions of the natural regime of observation wells.
In piezometric wells E-1 and YuZ-5 the instruments installed in 1996 provide highly sensitive water-level and atmospheric pressure recording up to ±0.1 cm and ±0.1 hPa, respectively, with a frequency of 5-10 min . In 1987-1994, water-level measurements in the E-1 well were carried out daily with an accuracy of ±0.5 cm, variations in atmospheric pressure were recorded by a barograph with an accuracy of 1 hPa (Kopylova, 2001). The seismic effects in water-level changes are identified with due regard of the influence of atmospheric pressure, precipitation, Earth's tides, seasonal trend, and other factors .
The primary outcome of long-term observations is that it was revealed hydrogeochemical and hydrogeodynamic phenomena preceding strong local earthquakes ( Figure 1A). The data on earthquakes preceded by HGPs manifested in two to four observation wells are presented in Supplementary Materials Table S2.
In this paper, we consider the following effects as HGPs originally detected retrospectively: i) visually distinguishable and statistically confirmed anomalies in the ionic and dissolved gas composition of the groundwater in the flowing wells GK-1, M-1 and G-1 (Kopylova et al., 1994;Biagi et al., 2000a;Biagi et al., 2000b;Khatkevich and Ryabinin, 2006;Kopylova and Taranova, 2013); ii) higher-rate water-level decreases repeatedly manifested before earthquakes with M ≥ 5.0 in the E-1 well which were identified as the precursors PS1 and PS2 (Kopylova, 2001;Kopylova and Boldina, 2012;; iii) substantial deviation in water-level behavior from seasonal average pattern in the YuZ-5 well ; Supplementary Materials Figure S4).
Hydrogeological precursors in the groundwater ion and gas composition in three wells are presented in Supplementary Materials Figure S3.
In the GK-1 well, a decline in chloride ion concentration was observed within one to nine months before six earthquakes (Supplementary Materials Table S2 and Figure S3A, left diagram). In the cases of the earthquakes of January 1, 1996 and December 5, 1997, the decreases in chloride ion concentration were changed into the sharp increases lasting 4-5 months. The increase in the variance and the change in the average concentrations of free gases were observed during 2 months before the earthquake of March 2, 1992 (Supplementary Materials Figure S3A, right diagram).
In the M-1 well, the concentration of bicarbonate ion decreased before five earthquakes (Supplementary Materials  Table S2 and Figure S3B). In four cases, simultaneous increases were detected in the concentrations of sulfate ion, calcium, and sodium. Within one month before the earthquake of March 2, 1992, water mineralization increased by 25%, and the hydrogeochemical type of water has changed due to the increase in the concentration of sulfate ion and the decrease in the concentration of bicarbonate ion (Kopylova et al., 1994).
The changes in the concentrations of chloride ion, sulfate ion, bicarbonate ion as well as sodium and calcium were observed in the G-1 well before the earthquakes of January 1, 1996 and December 5, 1997 (Supplementary Materials Figure S3S) (Khatkevich and Ryabinin, 2006).
The maximum durations of hydrogeochemical anomalies (T 1 ) and lead times before earthquakes (T 2 ) in individual wells are presented in Supplementary Materials Table S2. The T 1 and T 2 values coincide for wells GK-1 and G-1. At the same time, the M-1 well is characterized by the appearance of relatively short-term hydrogeochemical anomalies (T 1 4 weeks) during 4-21.5 weeks before earthquakes. This feature of the well is associated with a high rate of water exchange in the wellbore due to the high water discharge (q 1.5 dm 3 /s) compared to wells GK-1 and G-1 (q 0.1 and <0.001 dm 3 /s).
According to digital observations 1996-2016 at well E-1, two types of the pre-seismic signals are identified in daily rate of water-level decreases: PS1 developing weeks prior to the earthquakes (Supplementary Materials Figures S4,S6) and PS2 with a manifestation time of 5-6 years (Supplementary Materials Figure S7). The PS1 signal manifests itself in a more rapid water-level decline with increased daily rate before the earthquakes with M ≥ 5.0 at epicentral distances d e ≤ 350 km. It was found that an expected earthquake can occur within a time of about one month after the end of the PS1 (in 90% of the cases) or during the PS1 development (in 10% of the cases).
According to the manual level measurements in 1987-1994, seven cases of a water-level decreases with amplitudes 4-47.4 cm were revealed during 3-36 weeks before earthquakes with M w 5.6-7.5 considered as hydrogeodynamic precursors (Kopylova, 2001).
The PS2 signal appears as a more rapid long-term water-level decline with increased daily rate preceding and accompanying the groups of the Kamchatka strong earthquakes . This type of HGP was observed in 1991-1997 (six 1992-1997 earthquakes with M w 6.9-7.8 at epicentral distances up to 300 km) (Kopylova, 2001) Figure S7).
In the YuZ-5 well, we have revealed retrospectively two cases of the anomalous changes in the water-level before the earthquakes of December 5, 1997 with M w 7.8 and January 30, 2016 with M w 7.2 (Kopylova and Boldina, 2019; Supplementary Materials Figure S5).
The durations of hydrogeological precursors in different wells prior to individual earthquakes (T 1 ) ranged from 4 to 39 weeks, i.e., from about 1 to 9 months, and do not show any relationship with earthquake magnitudes (Supplementary Materials Figure  S2A). For the M-1 well, there is an increase in the lead time of the hydrogeochemical anomaly (T 2 ) in the range 1-5 months with an increase in the magnitude of subsequent earthquake (Supplementary Materials Figure S2B).
The hypothetical mechanisms of the hydrogeochemical and hydrogeodynamic precursors in observation wells of the Kamchatka Peninsula were considered in previous publications and briefly presented in Supplementary Materials S2.
The expert councils for earthquake prediction have been working in Kamchatka Krai for many years providing official seismic forecasts based on seismological, geophysical, hydrogeological, and other types of observations (Chebrov et al., 2011;Chebrov et al., 2013). Observation data at the wells of the Petropavlovsk-Kamchatsky test site are also used in the practice of such expert councils. Our collaboration with the Kamchatka branch of the Russian Expert Council for Earthquake Forecasting and Seismic Hazard and Risk Assessment (KB REC) which determines the correspondence between the released forecasts and the occurrence of the actual earthquakes provides the unique possibility to demonstrate practical significance of the some types of HGPs for real-time forecasting of the Kamchatka earthquakes.
Long-term data were obtained on the occurrences of the precursory signal PS1 in water-level changes in the E-1 well. The retrospective forecasting 1996-2012 earthquakes on the base of PS1 according the approach (Gusev, 1974;Kopylova, 2001;Chebrov et al., 2011;Chebrov et al., 2013) is presented in Supplementary Materials Table S3.
When PS1 is detected in real time, the forecast is made for 1-2 months (Supplementary Materials Figure S8). The observational data and the timing of the forecast are shown in Supplementary Materials Figure S6.
In accordance with the KB REC assessments our forecasts made on base of PS1 in 2004-2016 were true in terms of the time, location, and magnitude of the six events with M w 5.3-7.2 ( Figure 1A).

RELATIONSHIP BETWEEN HYDROGEOLOGICAL PRECURSORS AND EARTHQUAKE PARAMETERS
When using HGP in individual well to predict earthquakes, it is necessary to know about the relationship between the HGP in that well and the parameters of subsequent earthquakes (magnitude, epicentral distance), as well as to estimate the expected seismic impact in the observation area.
We have analyzed the relationship between considered HGPs in individual wells and the parameters of the subsequent earthquakes ( Figure 2). As the parameters characterizing the earthquakes, we considered the ratio between magnitude M w and epicentral distance to the well (d e ). The value of the specific density of seismic energy in the wave (e) was used as a parameter of the earthquake impact in the area of the observation well. The e value is proportional to the square of the seismic wave velocity and can be applied as a metric for some co-and post-seismic processes such as soil liquefaction and undrained consolidation of sedimentary deposits (Wang and Chia, 2008;Wang and Manga, 2010). The e values were estimated by the formula logd e 0.48M w −0.33loge−1.4 (Wang, 2007;Wang and Chia, 2008) and used to evaluate the range of e variations for the earthquakes preceded by the HGPs in the wells of the Kamchatka Peninsula. Previously in (Wang, 2007;Wang and Chia, 2008;Wang and Manga, 2010;Kopylova and Boldina, 2020), the e values were applied to assess co-and postseismic phenomena in ground-and surface waters. In this work, the e values are used to assess the possible seismic impact of earthquake in the observation area with manifestations of hydrogeological precursors.
The earthquakes with M w 6.5-7.8 before which there have been precursory anomalies in the groundwater chemical composition in two to three wells occurred at the epicentral distances d e 95-308 km, or 2.1-3.7 maximum linear sizes of earthquake sources according to (Riznichenko, 1976) (Figure 2A). During these earthquakes which were accompanied by the ground shaking with intensity I 4.5-5.5 on MSK-64 scale (Supplementary Materials Table  S2) the e values were 0.1-0.3 J/m 3 . Close values of e were obtained for the two earthquakes, before which a precursor appeared in water-level changes in the YuZ-5 well ( Figure 2B).
As mentioned above, long-term data were obtained on the occurrences of the precursory signal PS1 in the E-1 well before earthquakes with M ≥ 5.0, d e ≤ 350 km highlighted retrospectively ( Figure 2B, Supplementary Materials Table S3). Figure 2B also shows 1996-2012 earthquakes with M ≥ 5.0, d e ≤ 350 km before which PS1 did not appear for the period of retrospective analysis (red crosses). In the range of magnitudes M 5.0-6.5, the PS1 signal appears in 44% of cases (blue crosses), whereas before earthquakes with M w 6.6-7.8, PS1 manifested itself almost always.
Frontiers in Earth Science | www.frontiersin.org November 2020 | Volume 8 | Article 576017 Before the six 2004-2016 earthquakes, PS1 was identified in real time and advanced forecasts were issued ( Figures 1A, 2B). We believe that with pronounced manifestations of PS2 and PS1 in water-level changes in the E-1 well, as well as other types of considered HGPs, a seismic forecast is possible for the strongest earthquakes in the adjacent segment of the Kamchatka seismic focal zone. Such earthquakes can be characterized by the magnitudes M w ≥6.6, the d e /L ratio below 5, and the e value above 0.1 J/m 3 (Figure 2).

CONCLUSION
(1) Hydrogeological precursors manifesting themselves in the changes of chemical composition and pressure of groundwater at depths from hundreds of meters to a few first km are demonstrated on the case study from the Kamchatka Peninsula with a low population density and lack of industrial enterprises. The wells where hydrogeological precursors were detected are characterized by the absence of the influence of groundwater development and other anthropogenic factors which can disrupt the natural state of groundwater during the preparation of earthquakes. Unfortunately, in many cases when studying hydrogeological effects before earthquakes, an important aspect of the technogenic impact on the regime of observation wells is not sufficiently taken into account (King, 2018;King and Manga, 2018;Wang et al., 2018). When discussing the relatively long-term manifestations of hydrogeological precursors from urbanized and densely populated countries, it is necessary to consider technogenic factors in more detail, as well as climatic factors and their influence on the processes of groundwater filtration. Otherwise, ideas about the forms of hydrogeological precursors and their connections with the parameters of subsequent earthquakes may be distorted. (2) A correlation was detected between the manifestations of hydrogeological precursors in considered wells and the parameters of the subsequent earthquakes ( Figure 2). In the Kamchatka Peninsula, the hydrogeological precursors were mainly observed before the earthquakes with M w 6.6-7.8 at the epicentral distances of 80-300 km from the wells. These earthquakes caused ground shaking with intensity four to six on MSK-64 scale; seismic energy density during these events in the regions of the wells ranged from 0.1 to 4.5 J/m 3 . The HGPs were mainly observed in the near and intermediate field zones of the earthquake sources (d e /L 0.9-3.7). The data obtained on the HGPs manifestations can be useful in the study of the phenomena in the groundwater preceding strong earthquakes of other seismically active regions and other geophysical fields associated with changes in water pressure and physical properties of water-saturated rocks.
(3) In the case of the E-1 well, we revealed the increased sensitivity of fluid pressure during the preparation of the strong earthquakes which manifested itself in the more rapid water-level falls at an increased rate both before separate earthquakes (PS1 precursory signal) and before groups of the  strong earthquakes (PS2 precursory signal). This indicates that during the preparation of strong (M w ≥ 6.6) earthquakes, a decrease of water pressure in gas-saturated groundwater occurs over a period of time from weeks to several years. Such a process can occur with an increase in the capacity of waterbearing rocks, as well as due to phase transitions of gas and an increase in the density of groundwater in water-bearing rocks and in the wellbore. (4) The retrospective analysis of the PS1 manifestations has shown that with the increase in the magnitudes of the predicted earthquakes from M ≥ 5.0 to M w ≥ 5.8, the efficiency of PS1 for the seismic forecast increases from J 1.4 to J 2.4 (Supplementary Materials Table S3). This indicates that PS1 is a useful precursor of the strong earthquakes which may improve the forecasts of such earthquakes by a factor of 1.4-2.4 compared to random guessing. At the same time, the relatively low statistical estimates of the correlation between the PS1 and the subsequent earthquakes make the PS1 applicable for seismic forecasting only if combined with the other observation data and other precursors. The correlations between the other precursors and the subsequent earthquakes in Kamchatka are also low and not exceeding the values obtained for the PS1 (Serafimova and Kopylova, 2010;Chebrov et al., 2011;Chebrov et al., 2013). This highlights the need for developing new methods for analyzing of the prognostic data for increase the accuracy and reliability of earthquake forecasting. The data on the joint occurrence of different precursors before earthquakes from the archives of the expert councils may help much in providing more objective estimates of the efficiency of the precursors for seismic forecasting.
(5) Since 2001, an experiment has been conducted on the use of the PS1 precursor, and since 2012, together with the PS2 precursor, to predict earthquakes in real time by submitting forecasts to the KB REC. According to the KB REC conclusions, successful predictions of the location, time, and magnitude were made for six 2004-2016 earthquakes with M w 5.3-7.2 ( Figures 1A, 2B).
We believe that progress in the study of hydrogeological and other types of earthquake precursors for seismic forecasting can be achieved with closer collaboration of specialists observing precursors with expert councils for earthquake prediction.

AUTHOR CONTRIBUTIONS
All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.

FUNDING
The study was funded by the Kamchatka branch of the Geophysical Survey of the Russian Academy of Sciences (KB GS RAS) with the partial participation of the Russian Foundation for Basic Research, grant 18-05-00337.