Descriptive Statistics
Soil gas radon concentrations were determined between 2020 and 2021 at the Institute of Nuclear Sciences' backyard. The soil gas radon concentrations varied from 2178.6 to 22706.3 with a mean value of 10655 Bq m-3. As reported by many authors (Ciotoli et al. 2014; Zafrir et al. 2016), global soil gas radon concentrations are in the range of kBq m-3, and our results are consistent with the literature. In a previous study (İçhedef et al. 2013), soil gas radon concentrations were determined between 0.1 and 261.1 Bq m-3 around the Tuzla fault, which passing through the west of Izmir city center. The radon levels of this study are relatively low and vary within a narrow range. In another work, soil gas radon concentrations measured around the Manisa Fault varied between 0.2 and 35.2 kBq m-3, and the average radon concentration was 4.84 kBq m-3 (Taşköprü et al. 2023). This fault passes northeast of the İzmir city center and is approximately 30 km from our measurement location. Descriptive statistics for the radon data are given in Table 2. The mean and median values are very close, while the mod has three values: 676.2, 7888.0, and 12620.9. Kurtosis is calculated as 0.5159 in the range between − 3 and + 3, and skewness is 0.4465, which should be close to zero, indicating that the radon data fit a normal (Gaussian) distribution.
Additionally, the Kolmogorov-Smirnov test was applied to check normality, and there was no significant deviation from the normal distribution (p = 0.75). As a result, the radon time series has some outliers, but these extreme values do not destroy normality. The descriptive statistical analysis was conducted in R and Rstudio (RStudio Team 2019; Team R C 2020), and figures were prepared with the package ggplot2 (Wickham et al. 2019).
Table 2
Descriptive statistics for radon
| Radon (Bq m− 3) |
Min | 2178.6 |
Max | 22706.3 |
Median | 10442.3 |
Mean | 10655.4 |
Mod | 7888.0 |
12620.9 |
6761.2 |
Standard Dev. | 3585.5 |
Skewness | 0.52 |
Kurtosis | 0.45 |
Test of Normality | 0.75 |
Boxplots, histograms, Q-Q plots, and ECDF plots of 222Rn activity in Fig. 2 show the detailed analysis of the radon data distribution.
Radon Time Series
A time series was obtained from two-year soil gas radon data, as shown in Fig. 3. It seems clear that the raw radon data (black line in Fig. 3) varied over a more comprehensive range during the study period, and large variations were observed, caused mainly by sharp and marked decreases in the measured radon activities. It is not easy to evaluate raw radon data regarding seismicity without a connected scatterplot containing many peaks. Therefore, it is difficult to determine the dominant trend. To obtain the best-fit trend, we added a smooth curve and its uncertainty in the form of point-wise confidence intervals, shown in grey. Locally estimated scatterplot smoothing, or LOESS, a nonparametric method for smoothing a series of data in which no assumptions are made about the underlying structure of the data, was applied to the raw data. LOESS uses local regression to fit a smooth curve through a connected scatterplot, and this approach is effective when there are outliers in the dataset. The LOESS methodology includes techniques for constructing confidence intervals around the curve.
The trend line (Blue line) and confidence intervals (shown in yellow) in Fig. 3 give a clearer idea of the trend of radon. In a further analysis, we graphed a chart containing the radon trend and earthquakes and tried to interpret this graph (Fig. 4).
At the beginning of the study, three earthquakes, ranging in magnitude from 4.8 to 5.2, occurred near Manisa. The epicenters of these earthquakes were approximately 50 km northeast of our measurement location. The radon concentration shows an increasing trend with a minimal slope in this period (during the first 30 measurements) and then a slight decrease followed by a sharper increase between the 30th and 50th measurements. It is noteworthy that no earthquakes occurred for approximately 4-month period between the 10th and 50th measurements. The increased trend was over after three earthquakes occurred in 25 days (50th to 60th measurements). The radon behavior changed after these earthquakes, and a decreasing trend was observed for approximately 3 months from the beginning of July to the end of October. At the end of this period, the catastrophic Samos earthquake (ML=6.6) and its aftershocks occurred. Nearly 3000 aftershocks were recorded in the first 11 days following the earthquake. Among these, only five earthquakes with magnitudes ranging from 4.7 to 6.6, capable of inducing radon anomalies, were selected. After the Samos earthquake, the radon trend line exhibited a relatively sharp increase. Compared to other earthquakes, it is observed that radon exhibited a marked decreasing trend following the Samos earthquakes, followed by a subsequent increase. A decrease was subsequently observed before the first earthquakes occurred in the Aegean Sea near Seferihisar in early February 2021.
A decrease was subsequently observed before the first earthquakes occurred in the Aegean Sea near Seferihisar in early February 2021. Then, the radon trend line continued to increase with a decreasing slope. This trend continued similarly until the two earthquakes in Menderes (Izmir) occurred on 19.05.2021. These earthquakes are the last earthquakes that occurred during the study period. The radon trend line gradually decreased during the period following these two earthquakes and followed a horizontal trend during the last 2 months of the study.
During the two-year period of measurements, two separate groups with radon concentration anomalies and anomaly mechanisms were identified. The first group is related to earthquakes numbered 4, 5, 6, 14, and 15, where radon concentration increased before these earthquakes and decreased afterward. The second group is associated with earthquakes numbered 7, 8, 9, 10, 11, 12, and 13, wherein radon concentration exhibited the opposite trend, increasing before these earthquakes and decreasing afterward. A comparable observation applies to earthquakes 1, 2, and 3; nonetheless, as these seismic events occurred at the outset of the investigation, no remarks can be made about radon alterations preceding them. Tarakçi et al. (2014) worked on stress-related pre-seismic soil gas radon variations in Tuzla fault another active fault near Izmir city. They reported that radon concentrations show significant differences in compression and expansion regions. Soil gas radon concentration levels increase before earthquakes and decrease towards the time of earthquake occurrence in a compression seismic area. Conversely, radon levels do not show any changes before earthquakes and increase during earthquake occurrences in a dilation area. Their study area is a strike-slip fault with different characteristics compared to the İzmir Fault Zone (IF), which is a normal fault. In another study, results of the 26-year continuous observation data show different groups marked by significant large amplitude changes related to earthquakes. The first group data indicates that radon concentrations increased before the Xiuyan Mb 5.0 and MW 5.1 earthquakes and decreased afterward. The second group has the opposite behavior as the radon concentration increased sharply in the month following the earthquake. The last group shows a V-shaped progression before the earthquakes and initially low values after the earthquakes in 2013 (Zhou et al. 2020). Trend changes in radon may be related to fault characteristics and earthquake features such as depth, magnitude, distance, etc.
Radon is an inert gas that is not a chemical compound in nature. This feature makes it unique since its concentrations in soil gas are controlled by physical factors such as earthquakes, volcanic eruptions, and meteorological factors. Therefore, it is essential to consider meteorological factors such as atmospheric pressure, soil air temperature, relative humidity, rainfall, etc. These data were provided by the Turkish State Meteorological Service, which has a meteorological station approximately 1 km from the measurement station. The radon measurement dates were used to transform the meteorological data. Additionally, comparisons were made between the radon concentrations, average temperature, average pressure, and total precipitation data for consecutive years. This study, initiated in February 2020 and completed in February 2022, included a comprehensive review of these factors. Therefore, the first 12 months were compared with the results of the next 12 months. Remarkably, the soil gas radon concentrations exhibited comparable ranges in both years; the median values were close to each other. However, visualizing the data via a violin chart (Fig. 5) revealed distinct distribution differences. In the second year of the study, the radon concentrations ranged widely and had many more outliers than the first year. On the other hand, there were no significant changes in soil temperature or atmospheric pressure from year to year; significant differences were detected only in total precipitation. The total rainfall, which was 480 mm in the first year (2020), reached 824 mm in the second year (2021). This may also explain the differences in the distributions of radon concentrations among consecutive years.
The Pearson correlation coefficients between the soil gas radon concentration and meteorological parameters are shown in Fig. 6. Although the correlation coefficients are relatively low, the obtained results are consistent with the literature. First, there is a positive correlation between precipitation and relative humidity (r=-0.65). Similarly, there was an inverse correlation between soil temperature and relative humidity (r = 0.65). The soil temperature was positively correlated (r = 0.31) with the radon concentration. An inverse correlation (r=-0.11) was observed between radon and rainfall, similar to the findings of Ramola et al., 2008, where the correlation was r=-0,16. Briefly, a negative correlation was found between radon concentration and other parameters except temperature.