Earthquake prediction remains one of the most challenging and important problems faced by the scientific communities. During the last few decades, potential earthquake precursors, including 222Rn, Cl, SO42− and stable isotope ratios (δ2H and δ18O) have been monitored in numerous fault and volcanic zones for earthquake prediction1,2,3,4,5,6. Such precursor signals are anticipated by the formation of micro cracks in fault zones or by groundwater mixing due to crustal dilation1,6. Stresses that develop prior to an earthquake are thought to be responsible for the release and accumulation of certain constituents that may be useful as tracers or precursors of these tectonic forces. However, these potential precursors have not been extensively used to forecast tectonic or volcanic activities because abnormal increases of these components also occur due to other environmental processes, including changing meteorological conditions. Among the potential earthquake precursors, 222Rn in soil and groundwater has shown the highest sensitivity because of its radioactive nature and origin in the subsurface7,8,9,10,11,12. However, 222Rn still suffers from interferences by meteorological phenomena and tidal forces10,13,14.

In order to test the concept of a dual isotopic tracer, we monitored both 222Rn and 220Rn every hour in a limestone cave (Seongryu Cave, Korea) from May 18, 2010 to June 17, 2011 (see Methods). To our knowledge, this represents the first study to evaluate this 222Rn-220Rn isotope pair in an underground environment as a precursor of earthquakes. Seongryu cave, which is ~250 Ma in age, is located in Seonyu Mountain (elevation: 199 m) in the eastern part of Korea (Fig. 1). The cave is at 20 m above the mean sea level and is ~330 m in length, 1–13 m in height and has an entrance of ~1 m2. The main cave contains many branches, including three lakes, of which the two located near the entrance are affected by an outside stream15. More detailed description of this cave is available in Oh and Kim16.

Figure 1
figure 1

Location of the radon isotopes (222Rn and 220Rn) monitoring site in Seongryu Cave.

(a) Location of the radon isotopes monitoring site and the layout of the cave. Grey areas denote lakes areas where the cave bottom is covered by water. (b) Vertical cross-section of the cave. This map was modified from Oh and Kim16.

For the 220Rn measurements, the position of the air-inlet above the cave floor is critical because 220Rn decays almost immediately (~5 minutes for 97% decay) after emanation from the source. Since 220Rn in the soil air should be in equilibrium with its parent 224Ra (even in the skin layer) due to its short half-life, variations in the 220Rn activity of soil air, including that of the porewater in rocks and soils would not be useful. Therefore, the air-inlet must be positioned at a height where the inputs of 220Rn from general environmental processes are minimal, but also where large inputs from earthquakes are detectable. In addition, the monitoring location should be isolated from the outside air, which can also generate 220Rn anomalies due to wind-driven skin flow through rocks and soils. The atmosphere in the first 130 m of Seongryu Cave is influenced by the outside air, while the atmosphere in the inner cave is almost stagnant15. Thus, in this study we chose an air-inlet position 0.2 m above the cave floor and 180 m from the entrance (Fig. 1). At this position, the noise from general meteorological driving forces was minimal and the meteorological conditions were relatively constant for both 220Rn and 222Rn. The activity of 220Rn was not detectable when the air-let was positioned 1.5 m above the cave floor due to its short half-life (see Supplementary Fig. S1). The activities of 220Rn were higher at the entrance site than at the monitoring site due to the influence of wind-driven skin flow (see Supplementary Fig. S2). At the monitoring site, 222Rn was both significantly stable and enriched relative to the entrance site (see Supplementary Fig. S2). Furthermore, since the activity ratios of 222Rn to 220Rn are high in the normal natural environment, there is a possibility of counts from 222Rn spilling over to 220Rn (see Methods). Previous study16 has shown that without a correction for this spillover, one could observe erroneous positive correlations between 222Rn and 220Rn.

Over the monitoring period, 222Rn activity was on average higher during the summer (avg. 645 ± 194 Bq m−3) than during the winter (avg. 140 ± 172 Bq m−3) (Fig. 2a), which is typical of 222Rn observations in caves around the world17,18. This seasonal variation in 222Rn activity is known to be due to the difference in air ventilation intensity. In contrast, 220Rn activity was on average higher in the winter (avg. 9.7 ± 10.1 Bq m−3) than in the summer (avg. 1.2 ± 3.7 Bq m−3) (Fig. 2b). In particular, we noted that both 222Rn and 220Rn showed high peaks in February 2011 that are decoupled from the general seasonal patterns (Fig. 3). Outside weather parameters (temperature, relative humidity and pressure) showed large variations (−13.8–35.2 °C, 7.5–97.9% and 990.9–1033.6 mbar, respectively) relative to the inside air (10.7–15.7 °C, 94–99.9% and 999.0–1034.8 mbar, respectively).

Figure 2
figure 2

Variations in the activities of 222Rn and 220Rn in Seongryu Cave from May 2010 to June 2011.

(a) Hourly variations in 222Rn activity. (b) Variations in 4-hour averaged 220Rn activity. (c) Variations in air temperature during the monitoring period, both inside and outside the cave. (d) Energy (unit: erg = 10–7 J) of earthquakes with magnitudes greater than M6.0 in Japan and Malaysia during the monitoring period. The seasonal (spring–winter) mean and mean ± 2σ (σ: standard deviation) values in (a) and (b) are shown for 222Rn and 220Rn activities. Numbers in parentheses in (d) denote distance (km) from the monitoring site. Due to a lack of a precise calibration for 220Rn, activities are presented in ‘arbitrary units’.

Figure 3
figure 3

Relationships between the daily averages of 222Rn and 220Rn activities and weather parameters outside and inside the cave.

Red circles denote the anomalous data above an upper limit of 99% prediction interval. Values denote the dates of the anomalous data in February 2011. Anomalous data were excluded from the correlation coefficients (r2). (a) Relationship between 222Rn activity and ΔT (temperature difference between the outside and inside of the cave). (b) Relationship between 220Rn activity and ΔT. (c) Relationship between the activities of 222Rn and 220Rn. The solid line denotes a regression line and the dashed lines denote the 99% prediction interval. (d) Relationship between the daily precipitation and activities of 222Rn and 220Rn.

With the exception of the anomalous peaks (red circles in Fig. 3), the daily average of 222Rn activities showed a significant positive correlation (n = 234, r2 = 0.66) with the daily average of temperature (density) difference (the inside temperature is subtracted from the outside temperature) (Fig. 3a), which is typical of cave air behavior17,18. In general, the 222Rn activities in the outside air are approximately two orders of magnitude lower than those in cave airs. Thus, the lower 222Rn activities in winter could be due to greater ventilation of the denser outside air. In contrast, during the summer, 222Rn is trapped inside the cave due to the atmospheric stratification. In addition to the density difference, in some regions, the land surface humidity affects the activity of 222Rn in the underground air because pore space can be affected by water vapor condensation19. As such, we observed a positive correlation (n = 396, r2 = 0.57) between 222Rn activity and the relative humidity of the outside air. The temperature and relative humidity of the inside air remained fairly constant and they correlated weakly with 222Rn activities (r2 = 0.27 and r2 = 0.32, respectively). Precipitation and outside pressure showed no significant correlations with 222Rn activity (r2 < 0.1 and r2 = 0.12, respectively). Thus, the variations in 222Rn activities in the cave seem to be predominantly affected by variations in the outside temperature and humidity.

In contrast to 222Rn, the daily average of 220Rn activities, except for anomalous peaks, showed a negative correlation (n = 234, r2 = 0.51) with temperature difference (Fig. 3b) and relative humidity (r2 = 0.41). However, poor correlations were observed between the activities of 220Rn and outside pressure and precipitation (r2 = 0.15 and r2 < 0.1, respectively). 220Rn activities in the winter were higher than those in the summer, resulting from higher ventilation which results in rapid advection of the pore air (with 220Rn already in equilibrium with 224Ra) in the cave, in the cold season.

The significant 222Rn and 220Rn anomalies observed in February 2011 (Fig. 3a–c) cannot be explained by normal meteorological variations, including episodic precipitation events (Fig. 3d). Thus, we consider that these anomalies may have been precursors of the Tohoku-Oki Earthquake, which occurred approximately one month later (Fig. 2). A recent study20 showed that the Tohoku-Oki Earthquake was preceded by a series of small earthquakes that started on 13 February 2011. Our results showed that the 222Rn alone could not distinguish the February anomalies from the summer peaks (Fig. 3a), however, there are clear anomalous signals based on 220Rn alone or the combined 222Rn vs. 220Rn plots (Fig. 3b,c).

In general, carrier gases (CO2, CH4, Ar and He) play a critical role in controlling the migration and transport of trace gases (e.g., 222Rn) towards the surface21,22. From our results, we assume that the degassing of carrier gases peaked on February 12, 2011, reduced continuously until March 1, 2011 and then almost stopped on the day of the Tohoku-Oki Earthquake. Many studies have reported that carrier gases anomalies occurred days-weeks before earthquakes23.

Based on our results, we present three points of evidence to indicate that the 220Rn-222Rn isotope pair may be an excellent precursor of earthquakes: (1) 220Rn peaks during the anomalous period were much higher than those during normal periods over the year. The observed anomalies cannot be explained by any normal environmental conditions during the monitoring period (Figs 2 and 3); (2) A positive correlation was observed between 220Rn and 222Rn during the anomalous period, perhaps due to the venting of carrier-gases (e.g., CO2) from the sub-surface. On the other hand, negative correlations were observed more generally (Fig. 3c); (3) The peak hours of the 220Rn and 222Rn anomalies were episodic and decoupled from normal diurnal patterns (see Supplementary Fig. S3). In general, 222Rn showed a diurnal fluctuation pattern, particularly in spring and fall when temperature differences between day and night were largest, although this pattern is not seen for the short-lived 220Rn.

Although the monitoring site in this study is ~1200 km distant from the epicenter of the Tohoku-Oki Earthquake, the earthquake impacted the Korean Peninsula in a number of ways. The Korean Peninsula is located on the Eurasian tectonic plate, which extends to Japan. As a result of the Tohoku-Oki Earthquake it was estimated to have moved eastward by 1.2–5.6 cm24. In addition, 46 out of 320 monitoring wells in Korea showed changes in water level, temperature and electrical conductivity as a result of the earthquake25,26. Therefore, it is not surprising that the radon isotope anomalies observed in Seongryu Cave may represent precursors of this extremely large earthquake.

On the basis of our observations of 222Rn and 220Rn, we suggest that a network of 222Rn-220Rn monitoring stations should be constructed to further verify the potential of this method for forecasting the locations and strengths of pending earthquakes. In order to filter out other environmental forcing factors, meteorological parameters and potential carrier gases (CO2, CH4) should also be monitored. Most importantly, we have to carefully select suitable natural or artificial cave systems for these stations. We can easily develop more sensitive 220Rn monitoring systems, data transmission setups to remote laboratories and institute a canary program to automatically detect potential earthquake signals.


A detailed description of the RAD7 radon monitor is available in Burnett et al.27 and Lane-Smith et al.28. Briefly, the RAD7 uses a silicon alpha detector to determine the daughters of 222Rn and 220Rn, 218Po (t1/2 = 3.05 min; 6.00 MeV), 214Po (t1/2 = 164 μs; 7.67 MeV) and 216Po (t1/2 = 0.15 s; 6.78 MeV). The surface of the detector uses electrostatic attraction to capture Po+ ions using an electric potential of 2000 to 2500 volts and the alpha detector counts 218Po, 216Po and 214Po alpha decays. We used both 214Po and 218Po peaks for the 222Rn measurements. An air filter is used at the entrance of the RAD7 to prevent dust particles and charged ions from entering the radon chamber. The internal air pump of the RAD7 (flow rate: 1 L min–1) was activated for 1 minute every 5 minutes to reduce maintenance labor in the humid cave air. In order to maintain relative humidity of <10%, which is necessary for a constant detection efficiency of the RAD7, a desiccant column and a passive moisture exchanger (DRYSTIK, Durridge Co.) were coupled to the air path of the RAD7. During the measurement period, a new desiccant column was replaced every 3 or 4 weeks.

In order to obtain accurate 220Rn activity data in the presence of extremely high 222Rn levels, we corrected for the ‘spillover effect’ of 222Rn to 220Rn by using the method of Chanyotha et al.29. Briefly, we assume that the efficiency of 220Rn detection is a quarter of that for 222Rn to account for thoron decay sample in the intake system (volume of sample tube + drying unit) and internal cell of the RAD7. For the correction of ‘spillover effect’, the spill factor is assumed to be 0.015, which is an average value calibrated and measured by Durridge. The spillover from one of the radon channels (C: 214Po) into the thoron channel (B: 216Po) can be corrected by the analysis software (Capture) provided by Durridge. Due to a lack of a precise calibration for 220Rn, the activity data are presented in ‘arbitrary units’. While there is uncertainty in the calibration of absolute 220Rn activities, it does not affect the interpretation of our results since the same procedures and conditions were held constant during the measurement period. The atmospheric parameters (temperature, relative humidity and pressure) in the cave were measured hourly using external sensors (MSR145, MSR electronics) and stored in a data logger. Outside weather parameters were obtained from the Korea Meteorological Administration (KMA).

Additional Information

How to cite this article: Oh, Y. and Kim, G. A radon-thoron isotope pair as a reliable earthquake precursor. Sci. Rep. 5, 13084; doi: 10.1038/srep13084 (2015).