Application of artificial neural networks in simulating radon levels in soil gas D. Torkar ⁎, B. Zmazek 1 , J. Vaupotič 1 , I. Kobal 1 Jožef Stefan Institute, Jamova cesta 39, SI-1000 Ljubljana, Slovenia abstract article info Article history: Received 18 June 2009 Received in revised form 9 September 2009 Accepted 28 September 2009 Available online xxxx Editor: D.B. Dingwell Keywords: Radon in soil gas Environmental parameters Earthquakes Correlation Neural networks Simulation Anomalies have been observed in radon content in soil gas from three boreholes at the Orlica fault in the Krško basin, Slovenia. To distinguish the anomalies caused by environmental parameters (air and soil temperature, barometric and soil air pressure, rainfall) from those resulting solely from seismic activity, the following approaches have been used. First, the seismic activity data were eliminated from the dataset and then an artificial neural network (ANN) with 5 inputs for environmental parameters and a single output (radon concentration) was trained with the standard backpropagation learning rule. Then the predictions of Rn concentrations (C p ) generated with this ANN for the whole dataset were compared to measurements (C m ) and three types of anomalies (CA — correct anomaly, FA — false anomaly and NA — no anomaly) have been detected in the signal |C m /C p -1| by varying five parameters describing an anomaly within predefined intervals. An exhaustive search among results was made to find the best ones and thus identifying the best set of parameters. Finally, an attempt was made to shorten the search procedure by training another ANN with numbers of anomalies of each type in the input and five anomaly detection parameters in the output. With these procedures we were able to correctly predict 10 seismic events out of 13 within the 2-year period. © 2009 Elsevier B.V. All rights reserved. Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 2. Experimental basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 3. Methodology of data analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 3.1. Artificial neural networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 3.2. Identification anomalies of radon concentration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 4. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 4.1. Radon concentration prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 4.2. Anomaly detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 5. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 0 1. Introduction Radon is a radioactive noble gas appearing by radioactive decay of radium in the natural radioactive decay chains in the Earth's crust. Only a small fraction of radon (described as the emanation coefficient) enters the space between mineral grains and, thus, has the possibil- ity to travel away from the source, carried either by carrier gases (methane, carbon dioxide) or by water, and eventually reaches the atmosphere (Etiope and Martinelli, 2002). This travel is, in addition to radioactive decay, also subjected to the influence of geochemical and geophysical parameters. Of the three radon isotopes, mostly 222 Rn is usually of our interest, as it appears at a measurable level in our environment because of its relatively long half-life compared with half-lives of the other two isotopes. It has been thus for decades known as one of the potential earthquake precursors (Ulomov and Mavashev, 1971; King, 1978; King, 1986; Ui et al., 1988; Ohno and Wakita, 1996; Planinić et al., 2001; Virk et al., 2001; Yang et al., 2005). Chemical Geology xxx (2009) xxx–xxx ⁎ Corresponding author. Tel.: +386 1 477 37 64; fax: +386 1 477 38 82. E-mail addresses: drago.torkar@ijs.si (D. Torkar), boris.zmazek@ijs.si (B. Zmazek), janja.vaupotic@ijs.si (J. Vaupotič), ivan.kobal@ijs.si (I. Kobal). 1 Tel.: +386 1 477 3900; fax: +386 1 477 3882. CHEMGE-15832; No of Pages 8 0009-2541/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.chemgeo.2009.09.017 Contents lists available at ScienceDirect Chemical Geology journal homepage: www.elsevier.com/locate/chemgeo ARTICLE IN PRESS Please cite this article as: Torkar, D., et al., Application of artificial neural networks in simulating radon levels in soil gas, Chem. Geol. (2009), doi:10.1016/j.chemgeo.2009.09.017