Elsevier

Applied Geochemistry

Volume 20, Issue 6, June 2005, Pages 1106-1119
Applied Geochemistry

Radon in soil gas: How to identify anomalies caused by earthquakes

https://doi.org/10.1016/j.apgeochem.2005.01.014Get rights and content

Abstract

Anomalies have been observed in Rn content in soil gas from 3 boreholes at the Orlica fault in the Krško basin, Slovenia. To distinguish the anomalies caused by environmental parameters (air and soil temperature, barometric pressure, rainfall) from those resulting solely from seismic activity, the following approaches have been used: (i) deviation of Rn concentration from the seasonal average, (ii) correlation between time gradients of Rn concentration and barometric pressure, and (iii) regression trees within a machine learning program. Approach (i) is much less successful in predicting anomalies caused by seismic events than approaches (ii) and (iii) if ±2σ criterion is used and is equally successful if ±1σ is used. Approaches (ii) and (iii) did not fail to observe an anomaly preceding an earthquake, but show false seismic anomalies, the number of which is much lower with (iii) than with (ii). Model trees are shown to outperform other approaches. A model has been built which, in the seismically non-active periods when Rn is presumably influenced only by environmental parameters, predicts the concentration with a correlation of 0.8. This correlation is reduced significantly in the seismically active periods.

Introduction

Since the advent of the nuclear era in Slovenia in the sixties, 222Rn and other radionuclides have been systematically monitored in ground and surface waters (Kobal et al., 1978, Kobal et al., 1990, Kobal, 1979, Kobal and Renier, 1987, Kobal and Fedina, 1987, Vaupotič and Kobal, 2001, Vaupotič, 2002, Popit et al., 2002, Popit et al., 2004). The first Rn analyses with the objective of forecasting earthquakes (Ulomov and Mavashev, 1971, Scholz et al., 1973, Mjachkin et al., 1975, King, 1978, King, 1986, Ui et al., 1988, Ohno and Wakita, 1996, Pulinets et al., 1997, Toutain and Baubron, 1999, Planinić et al., 2000, Planinić et al., 2001, Belyaev, 2001, Virk et al., 2001) were carried out in Slovenia in 1982 (Zmazek et al., 2000a). Radon concentrations were determined weekly in 4 thermal water springs, while Cl, SO42-, hardness and pH, were determined monthly. In 1998, this study was extended to other thermal water springs (Zmazek et al., 2000b, Zmazek et al., 2002a, Zmazek et al., 2002b) and also to soil gas (Zmazek et al., 2000c, Zmazek et al., 2002c) at selected, seismically relevant sites, and sampling frequency was increased from once a week to once an hour. In this paper the authors will focus on the Rn concentration in soil gas.

Since April 1999, in 60–90 cm deep boreholes at 6 locations in the Krško basin, Rn concentration in soil gas, barometric pressure and soil temperature have been measured and recorded once an hour, using barasol probes (MC-450, ALGADE, France). Other meteorological data, such as air temperature and rainfall, have been provided by the Office of Meteorology of the Environmental Agency of the Republic of Slovenia, and seismic data by the Office of Seismology of the same agency. Boreholes 1 and 4 are located in the Orlica fault zone, at a distance about 4000 m from each other, while the other boreholes are at distances from 150 to 2500 m on either side of the fault zone (Fig. 1). Air temperature and rainfall were measured at the meteorological station Bizeljsko, approximately 14 km from the boreholes. This paper interprets data only from stations 1 (Krško-1), 5 (Kremen) and 6 (Grmada), because measurements were often interrupted and disturbed at the others. The experimental procedure is reported elsewhere (Zmazek et al., 2002c). As often utilised, for earthquakes Dobrovolsky’s equation (Dobrovolsky et al., 1979) was used to calculate RD i.e., RD = 100.43M, where M is the earthquake magnitude and RD the radius of the zone within which precursory phenomena may be manifested (so-called Dobrovolsky’s radius in km). Earthquakes for which the distance RE between the epicentre and the measuring site was equal or less than 2RD have been used in the interpretation.

Following general practice in this field (Yasuoka and Shinogi, 1997, Singh et al., 1999, Virk et al., 2001), an anomaly in Rn concentration is defined as a significant deviation from the mean value and is then related to seismic activity (Zmazek et al., 2002c). It is often impossible to distinguish an anomaly caused solely by a seismic event, from one resulting from meteorological or hydrological parameters. For this reason, the implementation of more advanced statistical methods in data evaluation (Di Bello et al., 1998, Cuomo et al., 2000, Biagi et al., 2001, Belyaev, 2001, Negarestani et al., 2001, Planinić et al., 2003, Steinitz et al., 2003) is important. In this paper, in addition to Rn anomalies expressed as deviations of Rn concentration from the average seasonal value by more than a multiple of standard deviation, and those based on a positive correlation between the time gradients of barometric pressure and of Rn concentration, the authors have applied regression trees for the first time to forecast earthquakes. Data mining and machine learning methods used for that purpose have been already successfully applied to many environmental problems, as reviewed by Džeroski (2002). Here, Rn concentrations are predicted on the basis of environmental data (air and soil temperature, barometric pressure, rainfall) during seismically non-active periods, and then is tested the hypothesis that the prediction is significantly worsened during seismically active periods is tested.

Section snippets

Methodology of data analysis

Experimental data have been analysed by searching for Rn anomalies (i) defined as deviations in Rn concentration of more than ±×σ (multiple × of standard deviation σ), from the average seasonal value (ii) expressed as those time intervals when time gradients of barometric pressure and Rn concentration have the same sign, and (iii) by using regression trees, which have been confirmed as outperforming other regression methods in this area (Zmazek et al., 2003).

Results and discussion

Experimental results are shown in Fig. 2. In the following sections, these raw data will be analysed by applying the approaches mentioned above. It will be seen that some earthquakes are preceded and accompanied by Rn anomalies (denoted as CA case: correct anomaly related to seismic events), some are not (denoted as NA case: no anomaly observed for an earthquake), and, also, that there are anomalies during seismically non-active periods (denoted as FA case: false anomaly appearing without a

Conclusions

The analysis has shown that Rn anomalies (i) based on deviations of Rn concentration from the seasonal average concentration have been less successful in predicting earthquakes as those (ii) based on the time gradient of Rn concentration and barometric pressure having simultaneously the same sign, and (iii) obtained with model trees. Approaches (i) and (ii) strongly depend on the values of ±×σ and ΔPt thresholds, respectively, while the dependence of approach (iii) on the threshold of (CRn)m/(

Acknowledgements

The study was funded by the Slovenian Ministry of Education, Science and Sport. The authors thank Prof. H. Ui from the Toyama University, Japan, for fruitful discussions and his constructive suggestions.

References (52)

  • J.P. Toutain et al.

    Gas geochemistry and seismotectonics: a review

    Tectonophysics

    (1999)
  • H. Ui et al.

    Anomalously high radon discharge from the Atotsugawa fault prior to the western Nagano Prefecture earthquake (M 6.8) of September 14, 1984

    Tectonophysics

    (1988)
  • B. Zmazek et al.

    Geochemical monitoring of thermal waters in Slovenia: relationships to seismic activity

    Appl. Radiat. Isot.

    (2002)
  • B. Zmazek et al.

    Soil radon monitoring in the Krško basin, Slovenia

    Appl. Radiat. Isot.

    (2002)
  • B. Zmazek et al.

    Application of decision trees to the analysis of soil radon data for earthquake prediction

    Appl. Radiat. Isot.

    (2003)
  • A.A. Belyaev

    Specific features of radon earthquake precursors

    Geochem. Int.

    (2001)
  • P.F. Biagi et al.

    Difficulties with interpreting changes in groundwater gas content as earthquake precursors in Kamchatka, Russia

    J. Seismol.

    (2001)
  • L. Breiman et al.

    Classification and Regression Trees

    (1984)
  • V. Cuomo et al.

    Robust statistical methods to discriminate extreme events in geoelectrical precursory signals: implications with earthquake prediction

    Nat. Hazard

    (2000)
  • G. Di Bello et al.

    Time dynamics of background noise in geoelectrical and geochemical signals: an application in a seismic area of Southern Italy

    Il Nuovo Cimento

    (1998)
  • I.P. Dobrovolsky et al.

    Estimation of the size of earthquake preparation zones

    Pure Appl. Geophys.

    (1979)
  • S. Džeroski

    Applications of KDD methods in environmental sciences

  • L.M. Hubbard et al.

    Time-variation of the soil gas radon concentration under and near Swedish house

    Environ. Int.

    (1996)
  • K.G. Ioannides et al.

    Measurements of 222Rn migration in soil

    J. Radioanal. Nucl. Chem.

    (1996)
  • C.Y. King

    Radon emanation on San Andreas fault

    Nature

    (1978)
  • C.Y. King

    Gas geochemistry applied to earthquake prediction: an overview

    J. Geophys. Res.

    (1986)
  • Cited by (61)

    • Identification of radon anomalies induced by earthquake activity using intelligent systems

      2021, Journal of Geochemical Exploration
      Citation Excerpt :

      However, it is often difficult to distinguish an anomaly solely caused by an earthquake activity from one influenced by meteorological parameters using simple statistical techniques. In this regards, there have been significant attempts made to develop a reliable criteria for identification of radon anomaly using different statistical and computational techniques (Negarestani et al., 2002; Zmazek et al., 2005; Biagi et al., 2001; Torkar et al., 2010; Gregorič et al., 2012; Piersanti et al., 2016; Singh et al., 2017; Tareen et al., 2019; Soldati et al., 2020). For example, Negarestani et al. (2002) utilized layer neural network technique with a dataset of one observation/week, Zmazek et al. (2005) utilized regression-based technique with a dataset of 24 observations/day, Torkar et al. (2010) utilized ANN technique with a dataset of 24 observations/day and Singh et al. (2017) utilized ANN/regression techniques with a dataset of 1 observation/week.

    • Testing of <sup>222</sup>Rn application for recognizing tectonic events observed on water-tube tiltmeters in underground Geodynamic Laboratory of Space Research Centre at Książ (the Sudetes, SW Poland)

      2020, Applied Radiation and Isotopes
      Citation Excerpt :

      Another advantage is its radioactivity, which can be easily detected by physico-chemical methods, such as active methods, e.g. using semiconductor detectors, Lucas chamber and AlphaGUARD (Kozłowska and Kozak, 2006; Przylibski et al., 2010) despite its very small concentration (of the order of 10−16 g/kg of rock) in the upper part of the earth's crust (Conen and Robertson, 2002; Oliveira et al., 2003; Hamada and Miyazaki, 2004; Sakashita et al., 2004; Wu et al., 2004; Brunke et al., 2004; Gupta et al., 2004; Zahorowski et al., 2004; Sesana et al., 2006; Podstawczyńska et al., 2009; Richon et al., 2011; Zafrir et al., 2013). There have been attempts to apply variations in 222Rn activity concentration in soil air, groundwaters or other geofluids (e.g. crude oil, natural gas or CO2 stream) to characterize and predict phenomena such as earthquakes (Steinitz et al., 2003; Fleischer, 1981; Hammond et al., 1981; Fong-Liang and Gui-Ru, 1981; Fong-Liang et al., 1981; Ta-Liang et al., 1981; King, 1984/1985, 1986; Igarashi et al., 1993, 1995; King et al., 1993, 1994, 1996; Virk and Singh, 1993, 1994; Heinicke et al., 1995; Ohno and Wakita, 1996; Al-Hilal et al., 1998; Finkelstein et al., 1998; Fytikas et al., 1999; Castro Morales and LaBrecque, 1999; Heinicke and Koch, 2000; Zmazek et al., 2000, 2005; LaBrecque et al., 2001; Biagi et al., 2001; Planinić et al., 2001, 2004; Virk and Walia, 2001; Tsvetkova et al., 2001; Richon et al., 2003; Huang et al., 2004; Hartmann and Levy, 2005; Choubey et al., 2007, 2009; Crockett et al., 2006; Ghosh et al., 2009; Kuo et al., 2006, 2011; İnan and Seyis, 2010; Crockett and Gillmore, 2010; Yalım et al., 2012; Shi et al., 2013; Negarestani et al., 2014; Petraki et al., 2015), volcanic eruptions (Barillon et al., 1993; Martinelli, 1998; Martinelli et al., 1995; Segovia, 1991; Segovia et al., 1995, 1997; 2001; Connor et al., 1996; Segovia and Mena, 1999; Garcìa et al., 2000; Berlo et al., 2004; Immè et al., 2006), landslide formation (Choubey et al., 2005), mining subsidence (Kies et al., 2004) as well as rock bursts, and gas or rock outbursts in mines (Lebecka et al., 1990; Mnich and Lebecka, 1992; Wysocka, 2010) or to identify active fault zones (Steele, 1981; Toutain and Baubron, 1999; Baubron et al., 2002; Ioannides et al., 2003; Al-Tai et al., 2004; Pizzino et al., 2004; Ramola, 2009; Utkin and Yurkov, 2010) and earth tides (Walia et al., 2003, 2013; Yang et al., 2005; Hartmann, 2005). A wide range of possible applications of radon as a tracer in a variety of processes occurring in the natural environment inspired us to launch joint research.

    View all citing articles on Scopus
    View full text