Radon-222 (222Rn) is a naturally occurring radioactive noble gas, formed in an intermediate step in the radioactive uranium decay chain. 222Rn is highly soluble in water, which allows it to move freely with groundwater and fluids within the Earth's crust. The release rate of 222Rn is primarily contingent on radium-226 concentration in rocks and the surface area of rocks/minerals1. Variations in surface area, influenced by factors such as porosity, cracks, and fractures, impact the contact area between pore fluid (water or air) and rock, thereby influencing the release of 222Rn. Consequently, the 222Rn concentration can serve as a tracer for determining subsurface rock types and providing insights into groundwater aquifers.

222Rn has been recognized as a precursor of earthquakes and an indicator of crustal deformation2,3,4,5,6,7. Recently, monitoring networks have been established for soil 222Rn in Italy and China. Machine learning techniques have been applied to analyze the collected data, marking a recent advancement8,9. Active faults, often maturing into well-defined structures with gouge in the fault core, cataclasite, and fracture zones, exhibit an exceptionally high specific surface area in the damaged zone, e.g., 3.3 × 106 m2 per unit area of the fault surface10. Water in the crust, such as groundwater and hot springs selectively traverse permeable zones, turning fault zones into conduits for groundwater. The concentration of 222Rn in the crust, especially in groundwater, is a powerful indicator of active faults, allowing for the estimation of fault activity. By mapping 222Rn in groundwater, Tsunomori11 identified large active faults and concluded that the faults may be larger than previously estimated. Furthermore, a correlation between 222Rn concentration and stress evolution was discerned through spatiotemporal variations before and after a large earthquake12. These investigations mapped groundwater 222Rn concentration on land, revealing active fault locations and estimating their activity. However, the variations in 222Rn concentration around submarine faults have never been reported or discussed. In this study, we attempted to generate a 222Rn concentration map for the Yatsushiro submarine faults, believed to have been activated by the 2016 Kumamoto earthquake. This report presents the 222Rn concentration findings and discusses the relationship between 222Rn concentration and seismicity in the vicinity of the submarine faults.

Yatsushiro submarine faults and sampling

The Yatsushiro submarine faults constitute an active fault group situated in the Yatsushiro Sea, southwest of Kyushu, Japan. This fault network represents the southernmost segment of the Futagawa-Hinagu fault zone, the epicenter of the 2016 Kumamoto Earthquake13,14 (Fig. 1). Earthquakes larger than M6.0 have repeatedly occurred in this Yatsushiro Sea segment, raising concerns regarding the potential hazard of another large seismic event in the near future15.

Figure 1
figure 1

Map of active faults around the Yatsushiro Sea. Red solid and dashed lines show active faults on land and submarine active faults, respectively. Black circles with focal mechanism solutions represent the epicenters of the 2016 Kumamoto earthquakes. The map was generated using Generic Mapping Tools (GMT), ver.6.528.

Historical seismic activities in the Yatsushiro Sea segment date back to 744 and 1619 AD16,17, documented through seafloor acoustic surveys and sedimentary rock dating14,18. The Yatsushiro submarine faults exhibit a steep to nearly vertical orientation, with discernible folding structures in the jog sections connecting the Yatsushiro Sea segment to the adjacent Hinagu segment18. A seismicity pattern analysis post the 2016 Kumamoto earthquake indicates that the jog section is under high-stress status19, suggesting an elevated potential for a large earthquake in the area, given the high seismicity rate and active crustal deformation.

The Hakuho Maru Expedition KH18-3 research cruise, conducted from July 27th to 30th, 2018, aimed to elucidate seafloor environmental changes resulting from active faults and submarine landslides triggered by the 2016 Kumamoto Earthquake in the Yatsushiro Sea. Additionally, the expedition sought to create a 222Rn concentration map for the Yatsushiro submarine faults. During the expedition, we collected two types of cores using a multiple corer and a piston corer, strategically positioned directly above or near active submarine faults. The multiple corer retrieved sediments below the seafloor in one-meter tubes, accompanied by seawater just above the seafloor, referred to as bottom water. Upon retrieval of the multiple corer onto the deck, the seawater was sampled to measure the 222Rn concentration. For deeper sediment sampling, a long piston corer was employed. In the piston cores, we collected upwelling pore water, which was obtained by self-loading and accumulated at the core's top after standing for a day. Subsequently, the 222Rn concentrations in the piston core water were measured.

Seismicity around Yatsushiro Sea

To analyze the seismicity in the Yatsushiro Sea, earthquake parameters (event location and magnitude) were collected from the earthquake catalog of the Japan Meteorological Agency spanning from January 1996 to February 2020. Seven circular observation regions (regions 1–7) were delineated along the central axis of the Yatsushiro Sea, each possessing a 5 km radius (Fig. 2a). These circular regions exhibited an approximate 5-km overlap in radius with their nearest counterparts, collectively covering the majority of the Yatsushiro Sea. A total of 1853, 1719, 1731, 1338, 1687, 1740, and 1027 earthquakes were incorporated into our calculations for regions 1–7, respectively.

Figure 2
figure 2

(a) Regions where the seismicity study was conducted. The red circles indicate the six regions for the seismicity investigation along the Yatsushiro Sea. Analysis results showcasing (b) b-value, (c) moment release, and (d) event count for each specified region. The map was prepared using GMT, ver.6.528.

Utilizing this dataset, the b-value and moment release for the specified regions were determined, as outlined in the Method section. Figures 2b, c and d, along with S1, summarize the b-value, moment release, and event count over the entire period. Regions 2 and 3 exhibited lower b-values and higher moment release. Given the substantial seismic event count in each zone, the associated uncertainty in b-value calculation is comparatively low. Specifically, when considering both the b-value and event count, the standard deviation of the b-value ranges from 0.018 to 0.024, a range significantly smaller than the temporal variations observed in b-values. Notably, the central area in the Yatsushiro Sea, encompassing regions 2 and 3, displayed heightened seismic activity. These results suggest that regions 2 and 3 may possess heightened potential for stress accumulation.

Further analysis was conducted on seismic data pre and post the 2016 Kumamoto earthquake (Figs. 3, S2, S3). Following the earthquake, there were noticeable changes in the spatial distribution of b-value and moment release were observed. Specifically, b-values in regions 2 and 3 decreased post-earthquake. A higher moment release was observed in the southern area (regions 6 and 7) before the earthquake, whereas post-earthquake, the central area (regions 2 and 3) exhibited heightened moment release. These findings indicate an intensified crustal deformation process, including increased seismic activity in the central area of the Yatsushiro Sea, subsequent to the 2016 Kumamoto earthquake.

Figure 3
figure 3

Analysis results showcasing b-value, moment release, and event count for each specified area (a) before and (b) after the 2016 Kumamoto earthquake.

Results for core analysis and 222Rn concentrations

Results obtained from the examination of piston cores are depicted in Fig. 4 and Supplementary Fig S4. The sediment composition in the Yatsushiro Sea primarily includes mud and sand, along with some volcaniclastics. In the southwestern region of the Yatsushiro Sea, conglomerate beds were observed at sites PC7 and PC10 (Fig. S4 a–e). Cores, especially those with conglomerate layers, exhibit a higher presence of upwelling water. Within the conglomerate layer of PC 10, minor granitoid gravels were detected, predominantly consisting of sandy gravels. These granules likely originate from Goshoura Island, located to the northwest of PC10, where Granodiorite is partially exposed in sedimentary rocks. Sediments obtained by multiple cores consist entirely of soft mud (Fig. S4 f.). The mud collected from all sampling locations is similar.

Figure 4
figure 4

Geological columnar sections of piston core (PC) and corresponding 222Rn concentrations (PC samples) and bottom water samples from nearly the same location (multiple core (MC) samples). “n.m.” indicates that the measurement was not conducted due to the inability to obtain the seawater samples. Conglomerate layers are observed in cores from sites PC7 and PC10.

Figure 5
figure 5

Results of 222Rn concentration measurements around Yatsushiro submarine faults14 in (a) upwelling water from piston core (PC) and (b) bottom water from multiple core (MC). The map was generated using GMT, ver.6.528.

The concentration of 222Rn in seawater samples obtained using multiple cores (MC) and piston cores (PC) was measured. 222Rn concentrations were determined using a radon and thron monitor (RTM 1688, SARAD GmbH, Germany) attached to a gas–water separation bottle (Fig. S5). The detailed measurement technique is discussed in the Methods section.

Results of 222Rn concentrations for the MC and PC samples are illustrated in Fig. 5a and b and Table S1. The 222Rn concentrations in the PC samples surpassed those in the MC samples, with notably higher concentrations in sediments containing conglomerate layers.

Typically, 222Rn is rarely detected in seawater. However, we observed 222Rn concentrations in bottom water (MC samples), particularly at sites MC7, 8, 9, 11, and 12, where relatively high 222Rn concentrations were noted. MC12 and PC10 represent the same sites, both containing a thick conglomerate layer. The central area in the Yatsushiro Sea around sites MC7, 8, 9, and 11 is a region exhibiting high 222Rn concentrations despite the absence of conglomerate beds. The high 222Rn concentration regions around MC7, 8, 9, and 11 align with the area seismically most active (2 and 3). The area is situated close to the segment boundary between the Yatsushiro Sea segment and the Hinagu segment (Fig. 1), and the region is considered to be under high stress18,19. The high-stress condition can also be inferred from geological structures such as pressure ridge formations and their displacement by active faults in the area18.


High concentrations of 222Rn in bottom water were observed at sites above conglomerate beds and in high-stress condition areas near the boundary between the Yatsushiro Sea and the Hinagu segments. Various factors may influence the 222Rn concentration in bottom water, including (1) different radium concentrations in rock types, (2) water flow from terrestrial sources, (3) water flow rates from sediments, and (4) crustal deformation involving changes in porosity and/or fracture formation1.

The amount of radium-226, which decays into 222Rn, in host sediments varies based on mineral composition. However, the sediments beneath the Yatsushiro Sea are quasi-homogeneous20, with a minor presence of granitic rock only in the gravel layer of PC10, which could contain a relatively high concentration of uranium-238. Therefore, the observation results suggest that the higher 222Rn concentration is not solely attributed to differences in rock types. The sample containing a conglomerate layer, PC10, may slightly elevate 222Rn concentration.

The primary source of terrestrial water in the region is the Kuma River, the only major river flowing into the Yatsushiro Sea (Fig. 1). Nikpeyman et al.21 noted that the 222Rn concentration in water discharged from the Kuma River is relatively high along the coastal region near the estuary of the Kuma River in the Yatsushiro Sea. Higher 222Rn concentrations were observed in Izumi Bay on the southern side of the Yatsushiro Sea, attributed to submarine groundwater discharge (SGD). However, the offshore regions of the Yatsushiro Sea show minimal impact from 222Rn originating from both the Kuma River and SGD in Izumi Bay. In addition to measuring 222Rn concentrations, they conducted simulations to assess the seasonal influence of 222Rn, including the spread of river water discharging from the sea surface to the seabed. Simulation results confirm that 222Rn from the river predominantly affects the northern area of the Yatsushiro Sea. Our sampling locations, situated in the central and southern regions of the Yatsushiro Sea at a depth of around 50 m, receive negligible 222Rn contribution from terrestrial water.

Crustal deformation plays a pivotal role in the release of 222Rn through alterations in permeability and specific surface area induced by crack formation and fracturing. The fracture zone, encompassing cracks and fractured grains, exhibits a higher water–rock contact area compared to the host rock22. Fault zones with flowing water display elevated 222Rn concentrations due to increased release from rocks into the surrounding water11. Permeability changes resulting from stress-induced crack formation can also influence the flow rate of upwelling water, leading to the discharge of highly concentrated 222Rn water onto the seafloor.

Both MC and PC samples indicate that conglomerate beds tend to induce higher 222Rn concentrations. These conglomerate layers, characterized by high porosity and increased water content in the pore space, undergo compaction, facilitating the easy upwelling of pore water with elevated concentrations of 222Rn. The observed 222Rn concentration and related results align with water flowing through porous conglomerate sediments. The flow rate in conglomerates, with higher permeability and porosity compared to muddy and sandy sediments, would be faster. The steady state concentration of 222Rn in bottom water at the seafloor is maintained by inflow from sediments, outflow due to migration and diffusion in the bottom water, and its decay. The faster flow of a larger volume of pore water through conglomerates results in a higher supply of 222Rn to bottom water than in sediments without conglomerates. The 222Rn concentration at site MC 12, including conglomerates, would be higher due to faster upwelling through porous sediments. Minor granitoid gravels around PC10 and MC12 may also contribute to increased 222Rn levels.

In the study area, elevated 222Rn concentrations were observed in the vicinity of sites MC 7, 8, 9, and 11. Sediments around these sites primarily consist of sand and mud, with no conglomerates. 222Rn from river water is negligible around the area. However, the region with higher 222Rn concentration aligns with a large number of seismic events, high moment releases and low b-values (Fig. 2), indicative of seismically active and higher stress concentration. Continuous release of 222Rn by seismic events, passing through a highly permeable fault zone with water to the seafloor, could contribute to the observed 222Rn levels. Following the 2016 Kumamoto earthquake, a notable increase in both the number of seismic events and moment release occurred in the vicinity (Fig. 3). Considering the half-life of 222Rn (3.8 days), our measurements reflect the crustal condition after the 2016 Kumamoto earthquake. The heightened seismic activity likely played a role in the increased in-water 222Rn levels observed in the central area of the Yatsushiro Sea, specifically around sites MC 7 through 11, suggesting a potential post-earthquake influence on 222Rn concentration in the region.

This study establishes the relationship between crustal activity and 222Rn concentration. To further validate the correlation between seismic activity and 222Rn levels, regular measurements of radon concentration are imperative. Ongoing monitoring will contribute to a comprehensive understanding of the dynamic interplay between crustal processes and 222Rn release in marine environments.


The investigation of 222Rn concentrations in the bottom water of the Yatsushiro submarine faults has unveiled elevated levels in highly permeable sediments and regions characterized by intense tectonic and seismic activity. An examination of minor seismic events in the Yatsushiro Sea post the 2016 Kumamoto earthquake indicates a heightened seismic activity. It is evident that seismic events led to the release of 222Rn through crack formations, with subsequent movement and upwelling occurring through permeable active faults. Furthermore, the active fault zone itself plays a role in the presence of 222Rn in the passing seawater. These discoveries posit that 222Rn concentrations can serve as an effective indicator of crustal activity not solely on land but also beneath the sea.


Measurement of in-water 222Rn concentration

The methodology employed for determining the 222Rn concentration in water aligns with that detailed by Kawabata et al.12 Initially, a 1.0-L volume of seawater was transferred into a 2.2-L bottle, which was subsequently sealed with a screwcap housing two tube connectors (see Fig. S1). To expedite the establishment of gas–liquid equilibrium, the bottled seawater underwent vigorous hand agitation for a duration of three minutes. Following this, the air in the gas phase within the bottle was dehumidified utilizing a DRIERITE® drying column and was then directed into a 222Rn monitor via an internal pump. The 222Rn monitor continuously acquired the energy spectrum of alpha particles emitted by daughter elements of radon-222 over a 15-min period for each measurement. Three consecutive sets of measurements were conducted for each sample, with the maximum values from the second or third measurements deemed as the 222Rn concentration data in the gas phase. This approach accounted for incomplete equilibrium conditions during the initial measurement stage.

We estimated the 222Rn concentration in the aqueous phase by considering the time elapsed from the date of sampling. The equation for the conservation of 222Rn in gas and water in the measurement system is:


where Ca and Cw denote the 222Rn concentrations in the gas phase and water phase, respectively, before agitation, and Ca(eq) and Cw(eq) are the 222Rn concentrations in the gas and water phases at equilibrium, respectively; Va and Vw represent the volumes of the gas and water phase, respectively.

The 222Rn equilibrium constant23, Kd is expressed as:

$$Kd=\frac{{C}_{w}\left(eq\right)}{{C}_{a}\left(eq\right)}=0.15+0.403{\text{exp}}(-0.0502 T)$$

where, by assuming Ca to be 0, we estimated the 222Rn concentration in water prior to measurement (Cw) using Eq. (1):


We calculated 222Rn concentration in water at sampling time, Cw0, using the 3.8 d half-life of 222Rn, when measurement start time deviated from the sampling time.

Analysis of seismicity (b-value and moment release)

In our investigation, the determination of the b-value in the study regions followed the Gutenberg–Richter relation24. The frequency-magnitude distribution can be depicted in a seismogenic domain as:


Here, M represents the magnitude, N indicates the number of earthquakes with a magnitude within M ± 0.1, and both a and b are constants determined through regression analysis of seismic activity. The 90% confidence interval for the slope coefficient was interpreted as the uncertainty of the b-value. This parameter signifies the proportion of small to large seismic events, with its variation within a region often reflecting the specific characteristics of its seismic activity. A low b-value in a region suggests significant differential stress, indicating proximity to the end of its seismic cycle25.

To accurately determine the standard deviation of the b-value, we employed a methodology26 that entails the following steps:

$$\sigma = bN^{{ - {1}/{2}}} ,$$

where N represents the total number of seismic events utilized for estimating the b-value. Upon examining the b-values depicted in Fig. 2b and the event counts shown in Fig. 2d across regions 1–7, the respective standard deviations (σ) are calculated to be 0.020, 0.019, 0.018, 0.022, 0.022, 0.020, and 0.024, respectively.

Furthermore, we calculated seismic moment of all events in each region and added them up to obtain the moment release. The magnitude-moment relation is


where Mw is moment magnitude and M0 is seismic moment in dyne-cm27.