**5. Conclusion**

190 Earthquake Research and Analysis – Statistical Studies, Observations and Planning

Fig. 7. A comparison of different approaches for the identification of radon anomalies: a) standard deviation (I); b) the relationship between radon exhalation and barometric

pressure (II); c) artificial neural networks (III); and d) model trees (IV).

Since the appropriate interpretation of field measurements plays an important role in any research, the purpose of this work was to combine and evaluate the different approaches applied by our research group for differentiating the radon anomalies caused by increased seismic activity from those caused solely by environmental parameters. The application of four different approaches – standard deviation from the related mean value (I), the correlation between time gradients of barometric pressure and radon concentration (II), artificial neural networks (III) and decision trees (IV) – was presented. Radon anomalies based on approach (I) have been less successful in predicting earthquakes than those based on the other three approaches. Secondly, approaches (I) and (II) greatly depend upon the values of the ±xσ and Δ*P*/Δ*t* thresholds, respectively, while the dependence of approaches (III) and (IV) on the threshold of (m-*C*Rn/p-*C*Rn)1 is very weak. The number of false anomalies for approach (II) points to the disturbance of radon exhalation by other environmental parameters and not just by barometric pressure. The assumption that radon exhalation is only directly influenced by barometric pressure is further suggested by different forms of radon transport at compression and dilatation zones (Ghosh et al., 2009). Promising results are achieved by applying approaches (III) and (IV), which make it possible to simultaneously incorporate all of the available environmental parameters. Furthermore, in using these techniques, the relation between radon concentration and environmental parameters does not necessarily have to be presumed linear. And finally, in taking into account the scale of the earthquake magnitudes observed during the time of radon measurements, one may speculate that the performance of the applied approaches would be better in the case of stronger earthquakes.
