**2. Sample collection radon determination**

Accurate sampling for radon measurements depends on appropriate monitoring wells. Because radon concentration in groundwater relates to emanation rates of geological layers, representative sampling must be from properly constructed wells. A submersible pump is commonly used in monitoring wells for groundwater sampling except artesian wells. Every sampling starts with flushing the stagnant water in the well and especially in the screen zone. Inadequate purging can be a major source of error, because the water sample is a mixture of stagnant water from the well bore, pore water from the filter gravel and groundwater influenced by the natural emanation rate of the aquifer. Fig. 1 shows the radon concentration in the well discharge during continuous sampling in a monitoring well. During the first period of flushing, the radon concentration of the water samples is practically zero and then increases rapidly to 529 pCi/L. The mean radon concentration measured for this monitoring well was 529 ± 19 pCi/L (eleven samples). A minimum of 3 well-bore volumes was purged before taking samples for radon measurements.

Application of Recurrent Radon Precursors

software

Background(cpm), or, Conversion Factor (cpm/pCi)

0

2

4

6

8

10

for Forecasting Local Large and Moderate Earthquakes 163

Fig. 2. Alpha spectrum of radon-222 and its daughter nuclides represented by TRI-CARB

A calibration factor for the LSC measurements of 7.1 ± 0.1 cpm/pCi (Fig. 3) was calculated using an aqueous Ra-226 calibration solution, which is in secular equilibrium with Rn-222 progeny. For a count time of 50 min and background less than 6 cpm, a detection limit below 18 pCi/L was achieved using the sample volume of 15-ml (Prichard et al. 1992).

> Background Conversion Factor

Date of analysis (2003~2011) 2003/3/1 2004/5/1 2005/7/1 2006/9/1 2007/11/1 2009/1/1 2010/3/1 2011/5/1

Fig. 3. Calibration factor and background for LSC measurements

Fig. 1. Radon concentration and electrical conductivity in the well discharge during continuous sampling in a deep observation well

A 40-ml glass vial with a TEFLON lined cap was used for sample collection. After collecting a sample, the sample vial was inverted to check for air bubbles. If any bubbles were present, the sample was discarded and the sampling procedure repeated. The date and time of sampling was recorded and the sample stored in a cooler. The maximum holding time before analysis was 4 days.

For the determination of the activity concentration of radon-222 in groundwater, a modified method described by Prichard and Gesell (1977) was adopted. Radon was partitioned selectively into a mineral-oil scintillation cocktail immiscible with the water sample (Noguchi 1964). The sample was dark-adapted and equilibrated, and then counted in a liquid scintillation counter (LSC) using a region or window of the energy spectrum optimal for radon alpha particles (Lowry 1991).

Radon concentrations were determined by drawing a 15-ml sample directly from a field sample into a clean syringe. Care was taken to prevent aeration of the samples in the process. The samples were then injected beneath a 5-ml layer of mineral-oil-based scintillation solution in 24-ml vials. The vials were vigorously shaken to promote phase contact, dark-adapted and held for at least three hours to ensure equilibrium between radon-222 and its daughters, and then assayed with a liquid scintillation counter. The results were corrected for the amount of radon decay between sampling and assay.

The results of the measurements were determined in units of counts per minute (cpm). It was essential to ensure that only the activity of radon-222 was measured. Using the TRI-CARB software of Packard 1600TR, it was possible to view the alpha spectrum (Fig. 2). The peaks of radon-222 (5.49 MeV), polonium-218 (6.00 MeV) and polonium-214 (7.69 MeV) can be distinguished.

Radon-222

Specific conductance

Specific conductance,

150

160

170

180

190

200

S/cm

Sampling time, min 0 10 20 30 40 50 60 70

A 40-ml glass vial with a TEFLON lined cap was used for sample collection. After collecting a sample, the sample vial was inverted to check for air bubbles. If any bubbles were present, the sample was discarded and the sampling procedure repeated. The date and time of sampling was recorded and the sample stored in a cooler. The maximum holding time

For the determination of the activity concentration of radon-222 in groundwater, a modified method described by Prichard and Gesell (1977) was adopted. Radon was partitioned selectively into a mineral-oil scintillation cocktail immiscible with the water sample (Noguchi 1964). The sample was dark-adapted and equilibrated, and then counted in a liquid scintillation counter (LSC) using a region or window of the energy spectrum optimal

Radon concentrations were determined by drawing a 15-ml sample directly from a field sample into a clean syringe. Care was taken to prevent aeration of the samples in the process. The samples were then injected beneath a 5-ml layer of mineral-oil-based scintillation solution in 24-ml vials. The vials were vigorously shaken to promote phase contact, dark-adapted and held for at least three hours to ensure equilibrium between radon-222 and its daughters, and then assayed with a liquid scintillation counter. The results

The results of the measurements were determined in units of counts per minute (cpm). It was essential to ensure that only the activity of radon-222 was measured. Using the TRI-CARB software of Packard 1600TR, it was possible to view the alpha spectrum (Fig. 2). The peaks of radon-222 (5.49 MeV), polonium-218 (6.00 MeV) and polonium-214 (7.69 MeV) can

were corrected for the amount of radon decay between sampling and assay.

Fig. 1. Radon concentration and electrical conductivity in the well discharge during

Number of wellbore volumes purged 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0

before analysis was 4 days.

be distinguished.

for radon alpha particles (Lowry 1991).

continuous sampling in a deep observation well

100

0

200

100

300

200

400

300

Radon-222 concentration, pCi/L

500

400

600

500

600

Fig. 2. Alpha spectrum of radon-222 and its daughter nuclides represented by TRI-CARB software

A calibration factor for the LSC measurements of 7.1 ± 0.1 cpm/pCi (Fig. 3) was calculated using an aqueous Ra-226 calibration solution, which is in secular equilibrium with Rn-222 progeny. For a count time of 50 min and background less than 6 cpm, a detection limit below 18 pCi/L was achieved using the sample volume of 15-ml (Prichard et al. 1992).

Fig. 3. Calibration factor and background for LSC measurements

Application of Recurrent Radon Precursors

aftershocks).

for Forecasting Local Large and Moderate Earthquakes 165

water started to decrease and continued to decrease for 45 days. Twenty days prior to the earthquake, the radon concentration reached a minimum value of 330 pCi/L and before starting to increase. Just before the earthquake, the radon concentration recovered to the previous background level of 780 pCi/L. The main shock also produced a sharp anomalous coseismic decrease (~300 pCi/L). After the earthquake, some irregular variations were observed, which we interpret as an indication that the strain release by the main shock was not complete and that some accumulation and release of strain continued in the region.

Fig. 5. Map of the epicentral and hypocentral distributions of the mainshock and aftershocks of the 2003 Chengkung earthquake and 1951 mainshocks (star: mainshock, open circles:

The Antung hot spring (Fig. 7) is in a unique tectonic setting located at the boundary between the Eurasian and Philippine Sea plates near the Coastal Range. Four stratigraphic units are present. The Tuluanshan Formation consists of volcanic units such as lava and volcanic breccia as well as tuffaceous sandstone. The Fanshuliao and Paliwan Formations

Verification of radon-222 as the radioisotope responsible for activity in the well water tested was obtained by the repeated counting of three samples from two wells. The half-life of 3.841 days experimentally determined for samples from Well Liu-Ying (I) located in Tainan Plain, Taiwan compares favorably with the accepted value of 3.825 days as shown in Fig. 4. When the counting vials are lack of tightness, radon will escape from counting vials and the half-life times experimentally determined for samples will be apparently shorter. Fig. 4 also shows an example of such a case from Well Wen-Tsu (II) located in Choshui River Alluvial Fan, Taiwan.

Fig. 4. Measurement of half life from semi-logarithmic decay curve

#### **3. Suitable geological conditions to catch recurrent radon precursors**

The 2003 Chengkung earthquake of magnitude (M) 6.8 on December 10, 2003 was the strongest earthquake near the Chengkung area in eastern Taiwan since 1951. The Antung radon-monitoring well (D1, Fig. 5) was located 20 km from the epicenter. Approximately 65 days prior to the 2003 Chengkung earthquake, precursory changes in radon concentration in ground water were observed. Specifically, radon decreased from a background level of 780 pCi/L to a minimum of 330 pCi/L (Fig. 6). Both geological conditions near the Antung hot spring and the vapor-liquid phase behavior of radon were investigated to explain the anomalous decrease of radon precursory to the 2003 Chengkung earthquake.

The production interval of the well ranges from 167 m to 187 m below ground surface and is pumped more or less continuously for water supply purposes. Discrete samples of geothermal water were collected for analysis of radon (Rn-222) twice per week. Liquid scintillation method was used to determine the activity concentration of radon-222 in ground water (Noguchi 1964; Prichard et al. 1992). The radon concentration was fairly stable (780 pCi/L in average) from July 2003 to September 2003 (Fig. 6). Sixty-five days before the magnitude (M) 6.8 earthquake (December 10, 2003), the radon concentration of ground

Verification of radon-222 as the radioisotope responsible for activity in the well water tested was obtained by the repeated counting of three samples from two wells. The half-life of 3.841 days experimentally determined for samples from Well Liu-Ying (I) located in Tainan Plain, Taiwan compares favorably with the accepted value of 3.825 days as shown in Fig. 4. When the counting vials are lack of tightness, radon will escape from counting vials and the half-life times experimentally determined for samples will be apparently shorter. Fig. 4 also shows an example of such a case from Well Wen-Tsu (II) located in Choshui River Alluvial Fan, Taiwan.

> Liu Ying (I) Wen - Tsu (II)

012345678

Time after sampling, t, days

The 2003 Chengkung earthquake of magnitude (M) 6.8 on December 10, 2003 was the strongest earthquake near the Chengkung area in eastern Taiwan since 1951. The Antung radon-monitoring well (D1, Fig. 5) was located 20 km from the epicenter. Approximately 65 days prior to the 2003 Chengkung earthquake, precursory changes in radon concentration in ground water were observed. Specifically, radon decreased from a background level of 780 pCi/L to a minimum of 330 pCi/L (Fig. 6). Both geological conditions near the Antung hot spring and the vapor-liquid phase behavior of radon were investigated to explain the

The production interval of the well ranges from 167 m to 187 m below ground surface and is pumped more or less continuously for water supply purposes. Discrete samples of geothermal water were collected for analysis of radon (Rn-222) twice per week. Liquid scintillation method was used to determine the activity concentration of radon-222 in ground water (Noguchi 1964; Prichard et al. 1992). The radon concentration was fairly stable (780 pCi/L in average) from July 2003 to September 2003 (Fig. 6). Sixty-five days before the magnitude (M) 6.8 earthquake (December 10, 2003), the radon concentration of ground

**3. Suitable geological conditions to catch recurrent radon precursors** 

anomalous decrease of radon precursory to the 2003 Chengkung earthquake.

Fig. 4. Measurement of half life from semi-logarithmic decay curve

lnC (C = radon-222 activity, pCi/L)

3.5

4.0

4.5

5.0

5.5

6.0

6.5

water started to decrease and continued to decrease for 45 days. Twenty days prior to the earthquake, the radon concentration reached a minimum value of 330 pCi/L and before starting to increase. Just before the earthquake, the radon concentration recovered to the previous background level of 780 pCi/L. The main shock also produced a sharp anomalous coseismic decrease (~300 pCi/L). After the earthquake, some irregular variations were observed, which we interpret as an indication that the strain release by the main shock was not complete and that some accumulation and release of strain continued in the region.

Fig. 5. Map of the epicentral and hypocentral distributions of the mainshock and aftershocks of the 2003 Chengkung earthquake and 1951 mainshocks (star: mainshock, open circles: aftershocks).

The Antung hot spring (Fig. 7) is in a unique tectonic setting located at the boundary between the Eurasian and Philippine Sea plates near the Coastal Range. Four stratigraphic units are present. The Tuluanshan Formation consists of volcanic units such as lava and volcanic breccia as well as tuffaceous sandstone. The Fanshuliao and Paliwan Formations

Application of Recurrent Radon Precursors

for Forecasting Local Large and Moderate Earthquakes 167

Fig. 7. Geological map and cross section near the radon-monitoring well in the area of Antung hot spring (Q: Holocene deposits, Lc: Lichi mélange, Plw: Paliwan Formation, Fsl: Fanshuliao Formation, Tls: Tuluanshan Formation, Bl: tuffaceous fault block, D1: radonmonitoring well, : Chihshang, or, Longitudinal Valley Fault, : Yongfeng Fault). See Fig. 6

minimum value of 330 pCi/L twenty days before the earthquake. During this 45-day period (Stage 2), dilation of the rock mass occurred and gas saturation developed in cracks in the

for map location.

consist of rhythmic sandstone and mudstone turbidites. The Lichi mélange occurs as a highly deformed mudstone that is characterized by penetrative foliation visible in outcrop. The Antung hot spring is situated in a brittle tuffaceous-sandstone block surrounded by a ductile mudstone of the Paliwan Formation (Chen & Wang 1996). Well-developed minor faults and joints are common in the tuffaceous-sandstone block displaying intensively brittle deformation. It is possible that these fractures reflect deformation and disruption by the nearby faults. Hence, geological evidence suggests the tuffaceous-sandstone block displays intensively brittle deformation and develops in a ductile-deformed mudstone strata. Ground water flows through the fault zone and is then diffused into the block along the minor fractures.

Fig. 6. Radon concentration data at the monitoring well (D1) in the Antung hot spring. Stage 1 is buildup of elastic strain. Stage 2 is dilatancy and development of cracks and gas saturation. Stage 3 is influx of ground water and diminishment of gas saturation.

Under geological conditions such as those of the Antung hot spring, we hypothesized that when regional stress increases, dilation of the rock mass occurs at a rate faster than the rate at which pore water can flow into the newly created pore volume (Brace et al. 1966; Scholz et al. 1973). During this stage (Stage 2 in Fig. 6), gas saturation and two phases (vapor and liquid) develop in the rock cracks. Meanwhile, the radon in ground water volatilizes and partitions into the gas phase and the concentration of radon in ground water decreases. Thus, the sequence of events for radon data prior to the 2003 Chengkung earthquake (Fig. 6) can be interpreted in three stages. From July 2003 to September 2003 (Stage 1), radon was fairly stable (around 780 pCi/L). During this time, there was an accumulation of tectonic strain, which produced a slow, steady increase of effective stress. Sixty-five days before the magnitude (M) 6.8 earthquake, the concentration of radon started to decrease and reached a

consist of rhythmic sandstone and mudstone turbidites. The Lichi mélange occurs as a highly deformed mudstone that is characterized by penetrative foliation visible in outcrop. The Antung hot spring is situated in a brittle tuffaceous-sandstone block surrounded by a ductile mudstone of the Paliwan Formation (Chen & Wang 1996). Well-developed minor faults and joints are common in the tuffaceous-sandstone block displaying intensively brittle deformation. It is possible that these fractures reflect deformation and disruption by the nearby faults. Hence, geological evidence suggests the tuffaceous-sandstone block displays intensively brittle deformation and develops in a ductile-deformed mudstone strata. Ground water flows through the fault zone and is then diffused into the block along the

Fig. 6. Radon concentration data at the monitoring well (D1) in the Antung hot spring. Stage

Under geological conditions such as those of the Antung hot spring, we hypothesized that when regional stress increases, dilation of the rock mass occurs at a rate faster than the rate at which pore water can flow into the newly created pore volume (Brace et al. 1966; Scholz et al. 1973). During this stage (Stage 2 in Fig. 6), gas saturation and two phases (vapor and liquid) develop in the rock cracks. Meanwhile, the radon in ground water volatilizes and partitions into the gas phase and the concentration of radon in ground water decreases. Thus, the sequence of events for radon data prior to the 2003 Chengkung earthquake (Fig. 6) can be interpreted in three stages. From July 2003 to September 2003 (Stage 1), radon was fairly stable (around 780 pCi/L). During this time, there was an accumulation of tectonic strain, which produced a slow, steady increase of effective stress. Sixty-five days before the magnitude (M) 6.8 earthquake, the concentration of radon started to decrease and reached a

1 is buildup of elastic strain. Stage 2 is dilatancy and development of cracks and gas saturation. Stage 3 is influx of ground water and diminishment of gas saturation.

minor fractures.

Fig. 7. Geological map and cross section near the radon-monitoring well in the area of Antung hot spring (Q: Holocene deposits, Lc: Lichi mélange, Plw: Paliwan Formation, Fsl: Fanshuliao Formation, Tls: Tuluanshan Formation, Bl: tuffaceous fault block, D1: radonmonitoring well, : Chihshang, or, Longitudinal Valley Fault, : Yongfeng Fault). See Fig. 6 for map location.

minimum value of 330 pCi/L twenty days before the earthquake. During this 45-day period (Stage 2), dilation of the rock mass occurred and gas saturation developed in cracks in the

Application of Recurrent Radon Precursors

using formation brine from the Antung hot spring.

2000; Yu & Kuo, 2001; Lee et al., 2003).

**5. Case study** 

for Forecasting Local Large and Moderate Earthquakes 169

Fig. 8. Variation of radon concentration remaining in ground water with gas saturation at 60 ºC

We have monitored groundwater radon since July 2003 at well D1 at the Antung hot spring that is located about 3 km southeast of the Chihshang fault (Fig. 9). The Chihshang fault is part of the eastern boundary of the present-day plate suture between the Eurasia and the Philippine Sea plates. The Chihshang fault ruptured (Hsu, 1962) during two 1951 earthquakes of magnitudes M = 6.2 and M = 7.0. The annual survey of geodetic and GPS measurements has consistently revealed the active creep of the Chihshang fault that is moving at a rapid steady rate of about 2-3 cm/yr during the past 20 years (Angelier et al.,

Fig. 10 shows the radon concentration data since July 2003 at the monitoring well (D1) in the Antung hot spring. Radon-concentration errors are ±1 standard deviation after simple averaging of triplicates. Recurrent groundwater radon anomalies were observed to precede the earthquakes of magnitude *Mw* = 6.8, *Mw* = 6.1, *Mw* = 5.9, and *Mw* = 5.4 that occurred on December 10, 2003, April 1, 2006, April 15, 2006, and February 17, 2008 at the Antung D1 monitoring well. We consider the *Mw* 5.9 earthquake that occurred on April 15, 2006 triggered by stress transfer in response to the 2006 *Mw* 6.1 Taitung earthquake. All the three recurrent anomalous decreases observed at Antung follow the same v-shaped progression and are marked with green inverted triangles in Fig. 10. Environmental records such as atmospheric temperature, barometric pressure, and rainfall were examined to check whether the radon anomaly could be attributed to these environmental factors. The

rock and radon volatilized into the gas phase. Stage 3 started at the point of minimum radon concentration when water saturation in cracks and pores began to increase and radon increased and recovered to the background level. The main shock produced a sharp coseismic anomalous decrease (~300 pCi/L). After the earthquake, some irregular variations were observed, which we attribute to strain release as some accumulation and release of strain continued in the region.

#### **4. In-situ radon volatilization mechanism**

Radon partitioning into the gas phase can explain the anomalous decreases of radon precursory to the earthquakes (Kuo et al., 2006). To support the hypothesis of radon volatilization from ground water into the gas phase, radon-partitioning experiments were conducted to determine the variation of the radon concentration remaining in ground water with the gas saturation at formation temperature (60 ℃) using formation brine from the Antung hot spring. Five levels of gas saturation were investigated, specifically *Sg* = 5 %, 10 %, 15 %, 20 %, and 25 % where *Sg* is gas saturation. Triplicate experiments were conducted for each level of gas saturation. Every experiment started with 40-ml of formation brine. Five levels of headspace volume at 2 ml, 4 ml, 6 ml, 8 ml, and 10 ml were then created above the liquid phase for five levels of gas saturation at 5 %, 10 %, 15 %, 20 %, and 25 %, respectively. Two-phase equilibrium was achieved for each experiment in 30 minutes at the formation temperature (60 ℃).

A kinetic study of radon volatilization from ground water into the gas phase was conducted to determine the time required to reach equilibrium. In the kinetic experiment, formation brine from the Antung hot spring with an initial radon concentration of 479 ± 35 pCi/L was used. Every sample started with 40-ml formation brine and a headspace volume at 6 ml was then created above the liquid phase. A total of five samples were prepared. The radon concentration remaining in ground water was determined at various volatilization times (i.e., 2 min, 5 min, 15 min, 30 min, and 60 min). The time required to reach equilibrium for radon volatilization was only about 5 minutes.

Data from the vapor-liquid two-phase equilibrium radon-partitioning experiments (Fig. 8) were regressed with the two-phase partitioning model to determine Henry's coefficient as follows.

$$\mathcal{C}\_{\mathfrak{o}} = \mathcal{C}\_{\mathfrak{w}} \left( H \times S\_{\mathfrak{g}} + 1 \right) \tag{1}$$

where 0 *C* is initial radon concentration in groundwater precursory to each radon anomaly, pCi/L; *Cw* is the radon minimum in groundwater observed in well D1 during an anomalous decline, pCi/L; *Sg* is gas saturation, fraction; *H* is Henry's coefficient for radon at formation temperature (60 ℃), dimensionless. Fig. 8 shows the regressed line with *H* = 12.8 and R2 = 0.9919 (regression coefficient). Henry's coefficient for radon at 60 ℃ determined for the Antung formation brine (12.8) is higher than the value (7.91) for water at 60 ℃ (Clever, 1979). Fig. 8 can be used to estimate the amount of gas saturation required for various decreases in concentration of radon. For example, the anomalous decrease of radon concentration from 780 pCi/L to 330 pCi/L required a gas saturation of 10 % in cracks in the rock.

Fig. 8. Variation of radon concentration remaining in ground water with gas saturation at 60 ºC using formation brine from the Antung hot spring.
