**4. Origin and mechanisms of radon anomalies**

Most of the researchers define radon anomaly as the positive deviation that exceeds the mean radon level by more than twice the standard deviation.

The origin and the mechanisms of the radon anomalies and their relationship to earthquakes are yet poorly understood, although several in-situ and laboratory experiments have been performed and mathematical modelings have been proposed. The radon observed in case of anomalies correlated with geophysical events may be considered as having two possible

Radon as Earthquake Precursor 151

monitoring station while the underground is depleted in radon. Since the half-life of radon is of 3.82 d, it would take a time larger than what is observed actually for those lower parts to supply by radioactive decay only. A much longer radon gas column than the few meters involved by mere diffusion should be involved. However, the most of the radon would decay away before reaching the detection system. The motion of pore fluids would first of all increase the radon concentration temporarily and then exhaust the available resources for

When radon concentrations are measured in continuous mode for a long time and with a time resolution of at least one hour, it is possible to classify the observed radon anomalies according to different trends. Typically two shapes can be considered that *Friedman, 1991*  classified group A and group B anomalies. Probably the physical process is different for the

The group A shows a very slow increase (or decrease) of the radon concentration with a rate less than 0.1% per hour. This kind of anomaly can be linked to a continuous increase in

The second group B is characterized by a fast increase (or decrease) in the radon concentration with a rate of about 1% per hour. Often a fast increase is followed by a rather constant radon concentration. Sometimes anomaly spikes with fast radon change immediately followed by a fast change in the opposite direction. These two kind of B anomalies could be linked to different physical processes or simply to different time scales

The B-type anomalies can be a local effect, which depends on certain local parameters, or they can be an epicentral effect. In this case the epicentral area must be supposed as origin of

> max 1 1 *Rn dC <sup>V</sup>*

From data of V and the epicentral distance *d* a very rough correlation can be found for

Thus by considering *V* from an observed anomaly a rough estimation of the distance from

Earthquake prediction means to forecast place, time and magnitude of an earthquake. From the analysis of a wide variety of radon data available from different countries and

*log(V) = -2log(d) + 4 [V] = hours-1, [d] = km* (15)

*C dt* (14)

For spike anomalies the maximum velocity of the radon concentration change is:

*CRn* the difference between radon concentration before and after the fast change

further increase.

**4.1 Radon anomaly shapes** 

two groups of anomalies.

(*Friedmann, 1991*).

the fast change in stress.

where *CRn* is radon concentration

the epicentre can be made.

**5. Forecasting relations** 

the time of the fast change (in hours)

*d*>70km, according to a relation of the form:

stress, until the rock fracture occurency.

origins. Either it is produced in depth origin or, once produced locally, it is displaced by other interstitial fluids whose motion is triggered by geodynamical events. Both possibilities have been discussed so far but the local origin hypothesis seems to be the most reliable, sustained by experiments too. In order to relate radon anomalies to earthquake occurrence, several scenarios have been proposed. Accordingly to the dilatancy-diffusion model (*Scholz, 1973; Planinic et al, 2001*) the radon anomalies could be related to mechanical crack growth in the volume of dilatancy or to changes in groundwater flow. Consequently either opening of new cracks, widening or closing of old cracks or redistribution of open and closed cracks can happen. In dry rocks opening or closing of cracks will lead to significant changes of the diffusion coefficient of radon. Volumetric changes in the rock will also lead to a subsurface gas flow and therefore to an additional radon transport. If the new open cracks are filled with water the increased water-rock interface leads to an increase in the transfer of radon from the rock matrix to the water. If water filled cracks close, the water will be compressed to another subsurface volume where the emanation from the rock to the water may be different. All these effects result in pressure and water level variations of the relevant aquifer. This also can lead to changes in the mixing ratios for the water which can be observed at the earth's surface. Finally gas flows can also move some groundwater and again all previously discussed mechanisms which are consequences of the redistribution of water in the earth's crust can take effects. This scenario has the drawback that an unreasonably large change in stress or strain is required far away from the epicentre.

An alternative mechanism is the stress corrosion theory, first proposed by *Anderson and Grew (1977)*. It attributes the radon anomalies to slow crack growth controlled by stress corrosion which should precede any mechanical cracking in wet environment. According to the mechanism of stress corrosion radon anomalies may depend on strain rate and local conditions such as rock type, porosity, elasticity, pattern of micro-cracks, degree of saturation, temperature, stress intensity factor and hygroscopic properties.

If the local parameters of the rocks are assumed to be responsible for the radon anomalies an explanation is needed to justify how very small changes in the stress field can produce such effects. From the theory of the earthquake preparation process it could be derived that in a region where the stress reaches a level which is not very far away from rock failure, very small changes in the stress fields result in considerable changes in certain rock parameters. If this theory holds, earthquake sensitivity could be expected only in areas which are highly pressed, for example near fault zone systems, not necessarily seismic active.

According to another kind of mechanism, the compression mechanism proposed by *King (1978)*, the anomalous radon concentration may be due to an increase in crustal compression, impending an earthquake, that squeezes out the soil-gas into the atmosphere at an increased rate. Radon anomalies have been observed at large distance from the earthquake epicentre, resulting from changes in the immediate vicinity of recording station, rather than at the distant focal region. This is accomplished if it is assumed that changes in stress or strain are propagated from the rupture zone to the radon station, leading to variations also in porosity, emanating power or flow rate of the local groundwater, near the radon monitoring station.

When the diffusion constant of radon in a soil of average porosity and moisture content is considered, the calculation shows that radon cannot be detected at a distance larger than a few meters. When radon is pumped by an upward moving carrier, whose motion is of the order of a few microns per second, it of course increases its concentration near the

origins. Either it is produced in depth origin or, once produced locally, it is displaced by other interstitial fluids whose motion is triggered by geodynamical events. Both possibilities have been discussed so far but the local origin hypothesis seems to be the most reliable, sustained by experiments too. In order to relate radon anomalies to earthquake occurrence, several scenarios have been proposed. Accordingly to the dilatancy-diffusion model (*Scholz, 1973; Planinic et al, 2001*) the radon anomalies could be related to mechanical crack growth in the volume of dilatancy or to changes in groundwater flow. Consequently either opening of new cracks, widening or closing of old cracks or redistribution of open and closed cracks can happen. In dry rocks opening or closing of cracks will lead to significant changes of the diffusion coefficient of radon. Volumetric changes in the rock will also lead to a subsurface gas flow and therefore to an additional radon transport. If the new open cracks are filled with water the increased water-rock interface leads to an increase in the transfer of radon from the rock matrix to the water. If water filled cracks close, the water will be compressed to another subsurface volume where the emanation from the rock to the water may be different. All these effects result in pressure and water level variations of the relevant aquifer. This also can lead to changes in the mixing ratios for the water which can be observed at the earth's surface. Finally gas flows can also move some groundwater and again all previously discussed mechanisms which are consequences of the redistribution of water in the earth's crust can take effects. This scenario has the drawback that an

unreasonably large change in stress or strain is required far away from the epicentre.

saturation, temperature, stress intensity factor and hygroscopic properties.

pressed, for example near fault zone systems, not necessarily seismic active.

radon monitoring station.

An alternative mechanism is the stress corrosion theory, first proposed by *Anderson and Grew (1977)*. It attributes the radon anomalies to slow crack growth controlled by stress corrosion which should precede any mechanical cracking in wet environment. According to the mechanism of stress corrosion radon anomalies may depend on strain rate and local conditions such as rock type, porosity, elasticity, pattern of micro-cracks, degree of

If the local parameters of the rocks are assumed to be responsible for the radon anomalies an explanation is needed to justify how very small changes in the stress field can produce such effects. From the theory of the earthquake preparation process it could be derived that in a region where the stress reaches a level which is not very far away from rock failure, very small changes in the stress fields result in considerable changes in certain rock parameters. If this theory holds, earthquake sensitivity could be expected only in areas which are highly

According to another kind of mechanism, the compression mechanism proposed by *King (1978)*, the anomalous radon concentration may be due to an increase in crustal compression, impending an earthquake, that squeezes out the soil-gas into the atmosphere at an increased rate. Radon anomalies have been observed at large distance from the earthquake epicentre, resulting from changes in the immediate vicinity of recording station, rather than at the distant focal region. This is accomplished if it is assumed that changes in stress or strain are propagated from the rupture zone to the radon station, leading to variations also in porosity, emanating power or flow rate of the local groundwater, near the

When the diffusion constant of radon in a soil of average porosity and moisture content is considered, the calculation shows that radon cannot be detected at a distance larger than a few meters. When radon is pumped by an upward moving carrier, whose motion is of the order of a few microns per second, it of course increases its concentration near the monitoring station while the underground is depleted in radon. Since the half-life of radon is of 3.82 d, it would take a time larger than what is observed actually for those lower parts to supply by radioactive decay only. A much longer radon gas column than the few meters involved by mere diffusion should be involved. However, the most of the radon would decay away before reaching the detection system. The motion of pore fluids would first of all increase the radon concentration temporarily and then exhaust the available resources for further increase.

#### **4.1 Radon anomaly shapes**

When radon concentrations are measured in continuous mode for a long time and with a time resolution of at least one hour, it is possible to classify the observed radon anomalies according to different trends. Typically two shapes can be considered that *Friedman, 1991*  classified group A and group B anomalies. Probably the physical process is different for the two groups of anomalies.

The group A shows a very slow increase (or decrease) of the radon concentration with a rate less than 0.1% per hour. This kind of anomaly can be linked to a continuous increase in stress, until the rock fracture occurency.

The second group B is characterized by a fast increase (or decrease) in the radon concentration with a rate of about 1% per hour. Often a fast increase is followed by a rather constant radon concentration. Sometimes anomaly spikes with fast radon change immediately followed by a fast change in the opposite direction. These two kind of B anomalies could be linked to different physical processes or simply to different time scales (*Friedmann, 1991*).

The B-type anomalies can be a local effect, which depends on certain local parameters, or they can be an epicentral effect. In this case the epicentral area must be supposed as origin of the fast change in stress.

For spike anomalies the maximum velocity of the radon concentration change is:

$$V = \frac{1}{\delta \mathcal{C}} \left( \frac{d\mathcal{C}\_{Rn}}{dt} \right)\_{\text{max}} \ge \frac{1}{\tau} \tag{14}$$

where *CRn* is radon concentration

*CRn* the difference between radon concentration before and after the fast change

the time of the fast change (in hours)

From data of V and the epicentral distance *d* a very rough correlation can be found for *d*>70km, according to a relation of the form:

$$\log(V) = -2\log(\text{d}) + 4 \qquad \text{[V]} = \text{hours} \cdot \text{s}^{-1}, \text{ [d]} = km \tag{15}$$

Thus by considering *V* from an observed anomaly a rough estimation of the distance from the epicentre can be made.

#### **5. Forecasting relations**

Earthquake prediction means to forecast place, time and magnitude of an earthquake. From the analysis of a wide variety of radon data available from different countries and earthquakes with *M*<3, *Rikitake* proposed an empirical relation between the time interval *t*  between radon anomaly and earthquake occurrence and magnitude of an earthquake (*Rikitake, 1976*):

$$\text{Log t} = 0.76M - 1.83 \tag{16}$$

Radon as Earthquake Precursor 153

where *d* is in km and *M* is the magnitude of the earthquake. It means that a magnitude *5* earthquake will be detected by means of precursory phenomena at a distance not greater

By collecting and analyzing radon anomaly data *Hauksson et and Goddard (1981)* found a similar relation. It is important that all these relations do not differ by more than *30%* in *d*

The constants in (23) may differ for different areas, however it is a good over all approximation. Of course we can expect that certain directions from the future epicentre are favored compared to others. The limit (23) must be seen as the limit for the favored

While precursor time *t* (in days) related to the magnitude *M* and the epicentral distance *d* (in

Long term series analyses have revealed a relation between the amplitude and duration of the gaseous anomaly and the magnitude *M* of the expected earthquake (*Barsukov et al.,* 

where *K* is a correction factor and *S* is the area of the peak anomaly, thus the shape of the

Several radon investigations have been carried out all over the world. Measurements of this gas both in soil and in groundwater have shown that spatial and temporal variations can

In the following we report some examples of studies among the numerous ones performed around the world with the purpose to relate abnormal radon emission to seismic events. The pioneering work on radon investigation in groundsoil was performed at an active fault zone for two years (*Hatuda, 1953*). Radon concentration in soil gas was measured and anomalous radon concentrations were reported before the strong earthquake (M=8) of

Some years later Tanner (*Tanner, 1959*) evidenced the importance of the influence of the meteorological parameters on radon measurements and in 1964 he suggested that radon could be used as tracer to discover uranium deposits or to predict earthquakes (*Tanner,* 

possibility of detecting a precursor anomaly at a distance *d* (in Km) to be

*Mmin=(2.3*

*M*

Another relation was proposed by *Martinelli, 1992* for which:

peak is a diagnostic parameter for the forthcoming seismic event.

**6. Radon anomalies and earthquakes: Some cases** 

provide information about geodynamical events.

Tonankai (December 1944, Japan).

*4*. But the most interesting result is that all observed precursors are limited by a straight line which coincides practically with a computed deformation of *10-8*. Summarizing the results in only one formula it is possible to estimate the magnitude-limit Mmin for the

*0.2)log(d)-(0.4*

*M=2.4logd – 0.43 –0.4* (24)

*logdt = 0.63M + (-) 0.15* (25)

*M K S* (26)

*0.3)* (23)

than *142* km.

for *M*

directions.

*1984*):

*1964*).

km) can be estimated as follows:

The relation was modified by *Fleischer* depending on the time interval: (*Rikitake, 1976*):

$$\text{Log } t \doteq M - 2.16 \quad \text{for } 0, 1 \le t \le 7 \text{ days} \tag{17}$$

$$\text{Log t} \equiv 0.62M - 1.0 \qquad \text{for t} \gg 7 \text{ days} \tag{18}$$

Starting from the radon diffusion equation and analyzing radon data from many countries , *Ramola et al (1988)* proposed an empirical relation to predict the magnitude of strong earthquakes (*M*>5) :

$$M \equiv 2\log\left(\lambda\_{Ru}\Lambda\mathcal{C}\_{R\eta}\mathcal{K}\Gamma\right) - 15.26\tag{19}$$

where *CRn* is the anomalous variation of radon concentration, *T* rise time for radon anomaly and *K* is a constant (*3.96 x 10-17*).

Several models were suggested in the past to evaluate the size of the area subject to changes in the tensional state. The models are based on assumption of homogeneity and isotropy of the ground or little heterogeneity around the focal zone.

In particular, *Dobrovolsky et al.(1979)* proposed some relations, taking into account an ellipsoidal inclusion with a 30% of heterogeneity with respect to the surrounding ground. He obtained the following relations that connect the magnitude *M* and the maximum distance *R* that the deformation can reach with the amplitude of the deformation *E*:

$$\begin{aligned} E &= \frac{10^{1.5M - 9.18}}{R^3} & M &< 5.0\\ E &= \frac{10^{1.3M - 8.19}}{R^3} & M &\ge 5.0 \end{aligned} \tag{20}$$

On the basis of these relationships deformations that can generate an anomaly were evaluated to be of the order of 10-8 (*Hauksson, 1981*).

Since radon anomalies seem to have a local origin, it is important to consider a relationship between the magnitude and the distance to the epicentre.

If the maximum possible distance *d* between the epicentre of a forthcoming earthquake and the spring which can be influenced by this earthquake is proportional to the volume of the pre-stressed lithosphere or to the energy of the earthquake respectively, a relation holds of the form (*Friedmann, 1991*):

$$M \equiv a \bullet \log(d) + b \qquad a, b \equiv \text{const} \tag{21}$$

From known relations between magnitude *M* and the volume of the focal zone the *a*-value can be determined to be about *2*. *Dobrovolsky et al.(1979)* observed that precursory phenomena are not observed beyond the distance *d*, thus to estimate roughly the radius of the effective precursory manifestation zone, they proposed the formula:

$$d \equiv 10 \exp\left[0.43\,\text{M}\right] \tag{22}$$

earthquakes with *M*<3, *Rikitake* proposed an empirical relation between the time interval *t*  between radon anomaly and earthquake occurrence and magnitude of an earthquake

Starting from the radon diffusion equation and analyzing radon data from many countries , *Ramola et al (1988)* proposed an empirical relation to predict the magnitude of strong

where *CRn* is the anomalous variation of radon concentration, *T* rise time for radon

Several models were suggested in the past to evaluate the size of the area subject to changes in the tensional state. The models are based on assumption of homogeneity and isotropy of

In particular, *Dobrovolsky et al.(1979)* proposed some relations, taking into account an ellipsoidal inclusion with a 30% of heterogeneity with respect to the surrounding ground. He obtained the following relations that connect the magnitude *M* and the maximum

*E M*

*E M*

<sup>10</sup> 5.0

<sup>10</sup> 5.0

On the basis of these relationships deformations that can generate an anomaly were

Since radon anomalies seem to have a local origin, it is important to consider a relationship

If the maximum possible distance *d* between the epicentre of a forthcoming earthquake and the spring which can be influenced by this earthquake is proportional to the volume of the pre-stressed lithosphere or to the energy of the earthquake respectively, a relation holds of

From known relations between magnitude *M* and the volume of the focal zone the *a*-value can be determined to be about *2*. *Dobrovolsky et al.(1979)* observed that precursory phenomena are not observed beyond the distance *d*, thus to estimate roughly the radius of

*M = a • log(d) + b a,b = const* (21)

*d= 10 exp 0.43 M* (22)

The relation was modified by *Fleischer* depending on the time interval: (*Rikitake, 1976*):

*M = 2log(*

*Rn*

distance *R* that the deformation can reach with the amplitude of the deformation *E*:

1.5 9.18 3 1.3 8.19 3

*M*

*M*

*R*

*R*

the effective precursory manifestation zone, they proposed the formula:

*Log t = 0.76M – 1.83* (16)

*CRn/KT) –15.26* (19)

(20)

*Log t = M –2.16* for 0,1<t<7 days (17)

*Log t = 0.62M –1.0* for t> 7 days (18)

(*Rikitake, 1976*):

earthquakes (*M*>5) :

anomaly and *K* is a constant (*3.96 x 10-17*).

the ground or little heterogeneity around the focal zone.

evaluated to be of the order of 10-8 (*Hauksson, 1981*).

the form (*Friedmann, 1991*):

between the magnitude and the distance to the epicentre.

where *d* is in km and *M* is the magnitude of the earthquake. It means that a magnitude *5* earthquake will be detected by means of precursory phenomena at a distance not greater than *142* km.

By collecting and analyzing radon anomaly data *Hauksson et and Goddard (1981)* found a similar relation. It is important that all these relations do not differ by more than *30%* in *d* for *M4*. But the most interesting result is that all observed precursors are limited by a straight line which coincides practically with a computed deformation of *10-8*. Summarizing the results in only one formula it is possible to estimate the magnitude-limit Mmin for the possibility of detecting a precursor anomaly at a distance *d* (in Km) to be

$$M \pm \text{Mmin} \equiv (2.3 \pm 0.2) \log(d) \text{-} (0.4 \pm 0.3) \tag{23}$$

The constants in (23) may differ for different areas, however it is a good over all approximation. Of course we can expect that certain directions from the future epicentre are favored compared to others. The limit (23) must be seen as the limit for the favored directions.

Another relation was proposed by *Martinelli, 1992* for which:

$$M \equiv 2.4 \log d \quad -0.43 \quad \text{-0.4} \tag{24}$$

While precursor time *t* (in days) related to the magnitude *M* and the epicentral distance *d* (in km) can be estimated as follows:

$$
\log dt = 0.63M + \text{(-)}\ 0.15\tag{25}
$$

Long term series analyses have revealed a relation between the amplitude and duration of the gaseous anomaly and the magnitude *M* of the expected earthquake (*Barsukov et al., 1984*):

$$M = K\sqrt{S} \tag{26}$$

where *K* is a correction factor and *S* is the area of the peak anomaly, thus the shape of the peak is a diagnostic parameter for the forthcoming seismic event.

#### **6. Radon anomalies and earthquakes: Some cases**

Several radon investigations have been carried out all over the world. Measurements of this gas both in soil and in groundwater have shown that spatial and temporal variations can provide information about geodynamical events.

In the following we report some examples of studies among the numerous ones performed around the world with the purpose to relate abnormal radon emission to seismic events.

The pioneering work on radon investigation in groundsoil was performed at an active fault zone for two years (*Hatuda, 1953*). Radon concentration in soil gas was measured and anomalous radon concentrations were reported before the strong earthquake (M=8) of Tonankai (December 1944, Japan).

Some years later Tanner (*Tanner, 1959*) evidenced the importance of the influence of the meteorological parameters on radon measurements and in 1964 he suggested that radon could be used as tracer to discover uranium deposits or to predict earthquakes (*Tanner, 1964*).

Radon as Earthquake Precursor 155

In Bhatsadam, Maharashtra, India, major earthquakes occurred during August 1983 - July 1984. In that region radon concentration was measured by *Rastogi et al.(1986).* They found an increase in radon concentration during March–April 1984 when seismicity was high enough. Precursory phenomena of radon in earthquake sequence were observed by *Rastogi et al. (1987)* and by other groups at the Osmansagar reservoir, Hederabad, India during January– February, 1982 (*Rastogi et al.,1987*). An earthquake with a magnitude of 3.5 occurred on January 14, 1982 with subsequent seismic events. There was an increase of radon

*Singh et al. (1991)* performed a daily radon monitoring in soil-gas in Amritsar from 1984 to 1987. They recorded radon anomalies before different earthquakes: June 1988 (M=6.8); April, 26, 1986 (M=5.7); July 1986 (M=3.8); Kangra earthquake March 1987 (M=7) and May

*Virk and Singh (1994)* carried out daily measurements of radon in soil-gas and groundwater at Palampur since 1989 and radon anomaly was recorded simultaneously in both soil- gas and groundwater. Weekly integrated data also showed abnormal radon behaviour during first week of October, 1991 at different recording stations. These recorded anomalies were correlated with an earthquake of magnitude 6.5 occurred in Uttarakashi area in October 1991.

*Al-Hilal et al. (1998)* recorded groundwater radon data for two years, during 1993 and 1994 at monthly intervals, from two selected monitoring sites of the northern extension of the Dead Sea Fault System. The results showed that measured radon concentrations fluctuate around the mean value, showing some variations with peak values, about two or three times the mean value, preceding some seismic events. It is possible to consider those anomalies related to changes in crustal strain and thereby to indicate a probable relation with the local seismicity. Nevertheless, the authors conclude that this does not necessarily means that it is possible to relate univocally these radon peaks to seismic event occurrence, but rather, it

may indicate the possibility of using groundwater radon variations as a useful tool.

In soil radon gas was monitored by *Friedmann et al. (1988)* in a network of ve monitoring sites along 200 km at the North Anatolian Fault Zone, Bolu. They observed an increase in radon concentration during the strong earthquake (M=5.7) on July 5, 1983. In order to search some relation between earthquakes and radon concentration variations, more recently *Inceoz et al (2006)* performed a radon investigation at the North and East Anatolian fault system. They found that radon anomaly was quite signicant in particular over the fault line but not

Also the Aksehir fault zone was investigated, by *Baykara and Dogru (2006)* and *Yalim et al. (2007),* trough radon measurements in well water. Although the observed radon levels could be related to several seismic activity that at the fault region occurred with high magnitude, the authors did not infer correlation between seismic activity and radon concentration. Radon concentration in thermal water was investigated by *Erees et al. (2006,2007)* at two thermal springs at the Denizli basin site and signicant radon anomalies were observed

before earthquakes with magnitude between 3.8 and 4.8.

concentration in soil gas during February due to those high seismic activities.

**6.2 India** 

1987 (M= 5).

**6.3 Syria** 

**6.4 Turkey** 

away from this line.

The first evidence of radon in groundwater as precursor of earthquakes was observed in Tashkent (*Ulomov, 1967*). The author observed that the radon concentration in a spring near Tashkent increased constantly before the M=5.2 earthquake on April 15, 1966.

Afterward many studies have been performed about radon anomalies and earthquakes.

In the following some examples are reported on ground radon monitoring in the most seismic regions in the world.

#### **6.1 Japan**

As already cited, studies performed by *Hatuda (Hatuda 1953),* at an active fault zone evidenced anomalous radon concentration before the strong earthquake (M=8) of Tonankai. Radon anomalies were recorded before the Nagano Prefecture earthquake (M= 6.8) on September 14, 1984 (*Hirotaka et al., 1988*). The authors observed a gradual increase in radon counts three months before the quake and a remarkable increase two weeks before the shock.

For about twenty years an extensive network of groundwater radon monitoring has been operated mainly by the University of Tokyo and the Geological Survey of Japan for the purpose of earthquake prediction in eastern Japan. In figure 1. a significant example of radon anomaly is reported (*Igarashi et al., 1995)*. The authors performed radon concentration analysis in a well 17 m deep from November 1993 to March 1995 and observed stable radon concentration of 20 Bq/l at the end of 1993. The radon concentration started to increase gradually from October 1994 reaching 60 Bq/l on November 1994, three times that in the same period one year before. Furthermore, a sudden increase of radon concentration, recorded on 7 January was followed by a sudden decrease on 10 January, 7 days before an earthquake of magnitude 7.2. After the earthquake, the radon concentration returned to the pre-October 1994 levels. The main result of this example is that it is possible to observe strange behavior before an anomaly. This, for instance, as in this case, must be preceded by a continuous increasing in the background level till its manifestation. Naturally it depends on the geodynamical evolution of the area

Fig. 1. Radon concentration data at the well in the southern part of Nishinomiya city, Japan [From . *Igarashi et al., 1995*]
