Radon and carbon dioxide emissions from the surface rupture zone produced by the 1920 Haiyuan M8.5 earthquake in Northwest China

Abstract

A M8.5 earthquake struck the Haiyuan area in 1920 and resulted in a 240-km long surface rupture zone along the Haiyuan fault (HYF). In this study, to determine the spatial relationship between soil gases and fault activity on a regional scale, radon (Rn) and carbon dioxide (CO₂) concentrations in soil gas were measured in situ with a grid spacing of 5 ∼ 10 km along the Haiyuan surface rupture zone (HYSRZ). Ordinary Kriging (OK) and Inverse Distance Weight (IDW) interpolation were used to investigate spatial variations in Rn and CO₂ concentrations. Synchronous soil gas anomalies were identified in southern Haiyuan County and at the southern end of the HYSRZ, and were correlated with higher slip rate, larger deformation, and larger degree rupture size than other segments. Moreover, soil gas anomalies were spatially highly coupled with seismically active zones and low b value areas, suggesting that the permeability and porosity of gas emissions are enhanced by higher stress. Two seismic velocity profiles across the gas anomaly area confirmed that low-velocity bodies exist at 0 ∼ 10 km below the surface, as revealed by magnetotelluric (MT) sounding results, indicating that gas emissions are associated with the migration of deep fluids. Our results provide new insight into the source of fluids in the HYSRZ and offer support for the design of a continuous geochemical measurement network in this area.

Keywords:

• Radon (Rn)
• carbon dioxide (CO₂)
• spatial interpolation
• b value
• crustal velocity structure
• spatial variation
• fault activity

Introduction

The Earth is an open and dynamic system in which gases continuously escape to the Earth’s surface, especially via active fault zones (King 1986; Ciotoli et al. 1999; Du et al. 2008). This outgassing process is controlled to a large degree by crustal stress fields, deep gas source, and faults (Toutain and Baubron 1999; King et al. 2006; Chen et al. 2022), which drive variations in permeability and porosity (Fu et al. 2005; Yang et al. 2018). Soil gas anomalies provide indicators of fractures and active fault zones that act as channels for the upward migration of deep-crust- and mantle-derived gases (Lombardi and Voltattorni 2010; Zhou et al. 2010; Fu et al. 2017; Iovine et al. 2018). Moreover, coseismic ruptures/fractures enhance near-surface permeability and favour gas emissions, which can persist for decades after seismic events, with higher gas concentrations often correlated with elevated stress and strain in the surface rupture zone (Ciotoli et al. 2007; Chiodini et al. 2011; Zhou et al. 2017; Dong et al. 2023). Therefore, geochemical analysis techniques for soil gas are commonly applied to investigate fault activity. Radon (Rn) and carbon dioxide (CO₂) are the most significant components of endogenic gas, and are sensitive to fault geometry and seismicity activity (Guerra and Lombardi 2001; Yuce et al. 2017; Benà et al. 2022). Concentration variations in CO₂ and Rn in soil gas are spatially correlated with fault activity in seismotectonic environments (Guerra and Lombardi 2001; Ciotoli et al. 2007; Al-Hilal and Abdul-Wahed 2016; Prasetio et al. 2023) and deep-seated faults related to frequent seismic activity (Zhou et al. 2010; 2017; Iovine et al. 2018; Sun et al. 2021). In general, geochemical profiles across faults can effectively reveal soil gas variation associated with fault activity along different fault segments (Han et al. 2014; Ciotoli et al. 2016; Sun et al. 2017; Yang et al. 2021), and survey profiles with a finer spatial resolution of 0.1 ∼ 2 km determined local gas geochemistry (Sun et al. 2016). The relationship between gas behaviour and tectonic activity at a regional scale can be revealed by a regional geochemical survey with a sampling density of 1 ∼ 6 samples per km² (Guerra and Lombardi 2001; Ciotoli et al. 2007; Li et al. 2013; Fu et al. 2017).

A M8.5 earthquake struck the Haiyuan area in 1920 and resulted in a 240-km long surface rupture zone along the Haiyuan fault (HYF). Along the Haiyuan surface rupture zone (HYSRZ), field observations revealed significant variation in degassing associated with fault behaviour and stress state along each segment of the fault (Jiang et al. 2000). The degree of fault healing was revealed by a difference in Rn concentrations on both sides of the HYF, with elevated ³He/⁴He ratios indicating the role of mantle degassing (Sun et al. 2016). Mercury (Hg) and Rn concentrations closely correlate with seismic activity and stress state along the HYF (Zhou et al. 2017). Moreover, various gas components are strongly emitted along the surface rupture zone of the southeastern HYF (Zhou et al. 2011). In these past studies, although local scale investigations of the soil gas–tectonic activity relationship have been performed in different survey profiles, geochemical profile lengths across the HYF ranged from 100 to 300 m; however, the width of the HYSRZ along the southeastern segment exceeds 500 m (Institute of Geology, China Earthquake Administration Earthquake Agency of Ningxia Hui Autonomous Region 1990). A macro-scale soil gas investigation in HYSRZ has not been reported; especially, few studies on soil gas variation associated with seismic activity, stress state and deep structures across the HYF have been discussed. In this study, the relationship between spatial variation of soil gases and fault activity was determined by measuring Rn and CO₂ concentrations at the regional scale. The results of this study provide new insight into the relationship between the crustal velocity structure and soil gas geochemistry along the HYSRZ, and offer support for the design of a continuous geochemical measurement network in this area.

Seismo-geological background

A series of complex arc-shaped structural belts are found in the contact zone between the northeastern margin of the Tibeten Plateau and the western margin of the Ordos Block. The HYF was initiated in the pre-Caledonian and remains active. The Haiyuan fault zone (HYFZ) extends along a NWW–NNW arc. Owing to significant strike-slip movement and very complex geometric structure, the HYFZ is composed of multiple secondary shear faults, pull-apart basins, and compression uplifts (Deng et al. 1986; Zhang et al. 1987; 1988; Burchfiel et al. 1991; Lasserre et al. 1999; Liu-Zeng et al. 2007). It can be divided into four segments—SF1, SF2, SF3, and SF4 (Figure 1)—each of which exhibits unique characteristics and slip rate (Table 1; Institute of Geology, China Earthquake Administration Earthquake Agency of Ningxia Hui Autonomous Region 1990).

Figure 1. Tectonic location and topographic map of the study region. (a) Tectonic map of the Tibeten Plateau, Ordos Block, and vicinity; the square shows the location of the study area. (b) Geography of the study area with the locations of sampling sites (triangle), profiles (black line), main faults (red line), and earthquakes epicenters (red circles).

Table 1. Haiyuan fault (HYF) segment characteristics.

No.	Length (km)	Mean horizontal displacement of river system (m)		Late Holocene slip rate (mm/a)	Fracture zone width (m)
		B grad	C grad
SF1	55	41.5	14.3	5.19–6.92	10–200
SF2	7–8	29.5	–	3.69–4.92	7–8
SF3	5–6	–	–	–	3–5
SF4	73	55	17.7	6.88–9.17	100–500

Table 2. Statistics of soil gas carbon dioxide (CO₂) and radon (Rn).

CO2 (%)					Rn (kBq/m^3)
Max.	Min.	Ave.	Skew	Kurt	Max.	Min.	Ave.	Skew	Kurt
1.94	0.05	0.60	1.44	1.99	84.30	0.72	33.09	0.84	0.24

Global Positioning System (GPS) observations across the HYF indicate that fault displacement differs significantly between the parallel and vertical directions (Li et al. 2017; Song et al. 2019). The rate variation is concentrated within 50 km on both sides of the fault, and there is no significant strain accumulation on either side, indicating that the fault has not yet entirely healed since the 1920 M8.5 Haiyuan earthquake (HYEQ) [Cui et al. 2016]. The inter-seismic creep velocity distribution from Interferometric Synthetic Aperture Radar (InSAR) shows that SF1 and SF4 exhibit obvious afterslip, while the central segments (SF2 and SF3) exhibit relatively little afterslip (Sun et al. 2017).

According to the crustal structure from magnetotelluric (MT) profiling, there are obvious differences in resistivity in the HYEQ focus area (Xi’anzhou), and a low resistance body exists near the hypocentre (Zhan et al. 2004). Moreover, there is a dipping low-resistivity belt within a high-resistivity background beneath the eastern end of the HYF (near Xiaokou). High resistivity layers from 1 to >20 km depth are interlayered with low-resistivity zones (Zhan et al. 2017). The fine velocity structure of the crust reveals a low-velocity layer at 25 ∼ 30 km depth in the HYEQ region, which is attributed to fluids (Xin et al. 2020).

In the HYFZ area, from Yanchi to Xiaokou (Figure 1), at least 5 earthquakes of M8.5 occurred in the Holocene (Liu and Zhou 1985), 8 earthquakes of M > 5 have occurred since 1920, and >30 earthquakes of M > 3.0 have been recorded since 1970.

Materials and methods

Field measurements

Field measurements were performed over 7 days with stable meteorological conditions in July 2019. Concentrations of Rn and CO₂ in soil gas were measured in situ at 69 sampling sites within a 5 ∼ 10 km-spaced grid along the HYSRZ (Figure 1). To avoid the effect of biological action on the concentrations of soil gases, all sites were in locations with similar loess cover thickness, vegetation density, and humidity. Soil gas samples were collected at depths of ∼80 cm using a hollow steel sampler with 12 holes at the end (Figure 2)[Chen et al. 2018] The concentration of Rn was analysed by a RAD7 Radon Detector with the detector limit of 67 Bq/km³ and calibration error of 10%. The concentration of CO₂ was measured with an ATG-C60 CO₂ detector (Adtech, China), which has a detector limit of 0.001% and calibration error of 4%.

Figure 2. The schematic diagram of the gas collection system (modified from Chen et al. 2018).

Spatial interpolation

To distinguish populations of soil gas concentrations related to seismic and tectonic activities, the upper and lower limits of the Rn and CO₂ anomalies were determined using a quantile–quantile (Q–Q) plot (Kafadar and Spiegelman 1986; Cheng et al. 1994).

Rn and CO₂ concentrations were processed by Ordinary Kriging (OK) and Inverse Distance Weight (IDW) interpolation in order to determine the spatial variation of gas geochemistry. The prediction value in the IDW interpolation method is determined by the distance between the measured points and the prediction points, as indicated by Equation 1(1) $$C_{o\left(i\right)}=\sum _{i=1}^n\frac{C_{f\left(i\right)}}{d_i^p}/\sum _{i=1}^n{d}_i^{-p}$$(1) (Daly, 2006). (1) $$C_{o\left(i\right)}=\sum _{i=1}^n\frac{C_{f\left(i\right)}}{d_i^p}/\sum _{i=1}^n{d}_i^{-p}$$(1) where n is the measured or predicted sequence length, and $C_{f(i)}$ and $C_{o(i)}$ are the measured and predicted concentration values of gas component i, respectively; d represents the Euclidean distance between the measured sites and the predicted sites; the power of the Euclidean distance, p, is set to 2 in this paper; the search radius is defined as 12 points.

Kriging interpolation is an unbiased and optimal method that relies on the theory of Semivariance and structural analysis (Kidner 2003). The OK interpolation formula is as follows: (2) $$C_{o\left(i\right)}=\sum _{i=1}^n{\lambda }_i{C}_{f\left(i\right)}$$(2) where λ_i is the weight coefficient that can be determined by semivariance based on unbiased minimum variance, the spherical model was chosen as the semivariance in this paper. According to statistics, the measured value of Rn and CO₂ did not follow a normal distribution (Table 2). Consequently, the original soil gas data underwent a logarithmic transformation and the converted data was subjected to spatial interpolation.

To evaluate the variance difference between the predicted value interpolated by OK and IDW, the F-test was performed on the spatial interpolation results of Rn and CO₂ (Table 3). The mean absolute error (MAE, Equation 3(3) $$\mathit{\text{MAE}}=\frac{1}{n}\sum _{i=1}^n\left|C_{f\left(i\right)}-C_{o\left(i\right)}\right|$$(3) ) and root mean square error (RMSE, Equation 4(4) $$\mathit{\text{RMSE}}=\sqrt{\frac{1}{n}\sum _{i=1}^n{\left(C_{f\left(i\right)}-C_{o\left(i\right)}\right)}^2}$$(4) ) were used to evaluate the accuracy of the spatial interpolations of OK and IDW (Table 4): (3) $$\mathit{\text{MAE}}=\frac{1}{n}\sum _{i=1}^n\left|C_{f\left(i\right)}-C_{o\left(i\right)}\right|$$(3) (4) $$\mathit{\text{RMSE}}=\sqrt{\frac{1}{n}\sum _{i=1}^n{\left(C_{f\left(i\right)}-C_{o\left(i\right)}\right)}^2}$$(4)

Table 3. F Test of the difference between the predicted value and measured value of soil gas carbon dioxide (CO₂) and radon (Rn) interpolated by ordinary kriging (OK) and inverse distance weight (IDW).

	significance level α = 0.05
	CO2 (%)		Rn (kBq/m^3)
	OK	IDW	OK	IDW
Ave.	0.07	−0.0002	0.07	0.14
Var.	0.07	0.00004	1.68	3.31
df	68	68	68	68
F	1761		0.51
P	0.000(<0.05)		0.0003(<0.05)

Table 4. Evaluation of spatial interpolation precision.

Interpolation method	Raster size (km)	Rn (kBq/M^3)		CO2 (%)
		MAE	RMSE	MAE	RMSE
IDW	0.45 × 0.45	3.91	16.02	0.87	2.26
Kriging	0.3 × 0.3	0.70	2.14	10.58	20.5

IDW, Inverse Distance Weight interpolation; MAE, mean absolute error; RMSE, root mean square error; Rn, radon; CO₂, carbon dioxide.

Results

The concentrations of Rn and CO₂ in soil gas along the HYSRZ (Table 5) varied widely. Rn concentrations ranged from ∼0.72 to 84.30 kBq/m³, with a mean value of 33.09 kBq/m³, which is much higher than those in the surface rupture zones of the Madoi M_S7.4 earthquake in 2021 (range from 2.10 to 39.17 kBq/m³; Wang et al. 2023), the M_w 7.0 central Italy earthquake in 1915 (a median value of 14.8 kBq/m³; Ciotoli et al. 2007), and the M_S8.0 Wenchuan earthquake (a mean value of 22.78 kBq/m³; Zhou et al. 2010). The CO₂ concentrations ranged from 0.05% to 1.95%, with a mean value of 0.60%.

Table 5. Measured concentrations of soil gas carbon dioxide (CO₂) and radon (Rn) along the haiyuan fault zone (HYFZ).

Site No.	Lon (°E)	Lat (°N)	Date	Radon (measured, kBq/m^3)	CO2 (measured, %)	Radon (predicted by OK) (kBq/m^3)	CO2 (predicted by IDW, %)
1	106.1811	36.0356	2019.7.1	44.50	0.64	44.47	0.64
2	106.1331	36.0508	2019.7.1	48.40	0.76	48.97	0.76
3	106.0958	36.0286	2019.7.1	84.30	1.43	83.67	1.43
4	106.0906	36.0400	2019.7.1	48.40	0.33	48.97	0.33
5	106.0628	36.0308	2019.7.1	21.50	0.17	21.50	0.17
6	105.9747	35.9839	2019.7.1	42.60	1.03	42.53	1.03
7	105.9058	35.9992	2019.7.1	26.30	0.69	26.30	0.69
8	106.2197	36.0578	2019.7.1	18.40	0.35	18.63	0.35
9	106.2011	36.1764	2019.7.2	13.40	0.87	1.47	0.87
10	106.1616	36.1708	2019.7.2	30.90	0.46	31.14	0.46
11	106.1184	36.1570	2019.7.2	82.80	1.49	82.64	1.49
12	106.0915	36.1530	2019.7.2	36.00	0.97	37.90	0.97
13	106.0650	36.1342	2019.7.2	58.80	0.48	58.98	0.48
14	106.0818	36.1407	2019.7.2	81.40	1.23	79.58	1.22
15	106.1374	36.3258	2019.7.2	64.40	1.86	64.82	1.86
16	106.0818	36.3067	2019.7.2	29.60	0.69	29.70	0.69
17	106.0104	36.2941	2019.7.2	34.10	0.72	34.05	0.71
18	106.0013	36.2826	2019.7.2	27.30	0.55	27.35	0.55
19	105.8266	36.0923	2019.7.3	1.01	0.05	1.01	0.05
20	105.8858	36.1244	2019.7.3	35.10	0.44	35.07	0.44
21	105.9324	36.1234	2019.7.3	0.72	0.06	0.99	0.06
22	105.9857	36.1302	2019.7.3	30.70	1.40	30.70	1.40
23	105.9557	36.2665	2019.7.3	28.10	0.49	28.10	0.49
24	105.8716	36.2537	2019.7.3	21.90	0.50	21.89	0.50
25	105.8210	36.2363	2019.7.3	13.10	0.46	13.12	0.46
26	105.7782	36.1852	2019.7.3	16.60	0.98	16.60	0.98
27	105.8667	36.4614	2019.7.4	18.50	1.09	18.44	1.09
28	105.8194	36.4601	2019.7.4	9.49	0.22	9.52	0.22
29	105.7824	36.4313	2019.7.4	18.00	0.34	17.82	0.34
30	105.7203	36.4136	2019.7.4	15.40	0.30	15.48	0.30
31	105.7478	36.4264	2019.7.4	1.29	0.19	1.79	0.19
32	105.7369	36.4303	2019.7.4	9.34	0.26	9.32	0.26
33	105.7499	36.6303	2019.7.4	8.17	0.20	8.45	0.20
34	105.7405	36.6060	2019.7.4	19.90	0.34	19.87	0.34
35	105.7189	36.5739	2019.7.4	16.60	0.54	16.62	0.54
36	105.6865	36.5496	2019.7.4	30.00	0.25	30	0.26
37	105.6247	36.5543	2019.7.5	69.30	1.00	69.03	1.00
38	105.6068	36.5070	2019.7.5	73.90	0.50	73.49	0.50
39	105.6160	36.4956	2019.7.5	67.20	1.90	66.90	1.90
40	105.6138	36.4664	2019.7.5	30.70	0.72	30.83	0.72
41	105.5688	36.4331	2019.7.5	36.30	0.58	36.31	0.58
42	105.7078	36.4006	2019.7.5	20.80	0.33	20.82	0.33
43	105.7001	36.3933	2019.7.5	26.20	0.45	26.22	0.69
44	105.6744	36.3833	2019.7.5	23.40	0.24	23.38	0.25
45	105.6552	36.3648	2019.7.5	20.90	0.51	20.95	0.51
46	105.5423	36.3573	2019.7.5	26.20	0.69	25.92	0.44
47	105.6125	36.3204	2019.7.5	46.40	0.47	46.31	0.47
48	105.6111	36.7128	2019.7.6	43.60	0.08	43.52	0.08
49	105.5570	36.6676	2019.7.6	14.70	0.23	14.89	0.23
50	105.5215	36.6514	2019.7.6	26.10	0.18	26.17	0.19
51	105.5150	36.6258	2019.7.6	67.60	0.66	57.54	0.66
52	105.4389	36.5318	2019.7.6	17.40	0.33	18.28	0.33
53	105.4269	36.4876	2019.7.6	24.60	0.30	24.74	0.30
54	105.4385	36.5521	2019.7.6	65.60	0.95	61.23	0.94
55	105.4533	36.5752	2019.7.6	35.57	0.55	39.12	0.60
56	105.4528	36.5767	2019.7.6	53.20	0.79	44.90	0.75
57	105.4583	36.5914	2019.7.6	32.20	0.98	32.67	0.98
58	105.4840	36.6075	2019.7.6	12.40	1.95	13.4	1.94
59	105.3423	36.6770	2019.7.7	38.30	0.39	37.84	0.39
60	105.3204	36.6652	2019.7.7	21.60	0.26	22.76	0.26
61	105.3155	36.6537	2019.7.8	42.40	0.60	41.32	0.59
62	105.2555	36.6729	2019.7.7	43.20	0.74	43.11	0.74
63	105.2325	36.6776	2019.7.7	29.00	0.25	29.31	0.25
64	105.2619	36.5719	2019.7.7	47.50	0.37	47.27	0.37
65	105.2630	36.5868	2019.7.7	39.50	0.44	39.84	0.44
66	105.1997	36.6226	2019.7.7	23.20	0.31	23.20	0.31
67	105.2211	36.6375	2019.7.7	25.90	0.31	25.94	0.31
68	105.2357	36.6528	2019.7.7	38.90	0.12	38.74	0.12
69	105.2817	36.6551	2019.7.7	12.70	0.60	12.76	0.60

The upper thresholds of Rn and CO₂ anomalies from the Q–Q plots were 43 kBq/m³ and 0.75%, respectively (Figure 3). The proportions of lower values of Rn (<10 kBq/m³) and CO₂ (<0.2%) were 8.70% and 5.8%, respectively, indicating that most gas samples were not significantly diluted by air.

Figure 3. Anomaly thresholds of soil gas (a) radon (Rn) and (b) carbon dioxide (CO₂) estimated from quantile–quantile (Q–Q) plots.

The spatial distributions of Rn and CO₂ concentrations were similar when using both the OK and IDW interpolation methods. However, according to the F-test (Table 3), the variance of the difference between the predicted value and measured values of Rn and CO₂ obtained by the two methods was significantly different (p < 0.05). Therefore, based on the smallest MAE and RMSE values, the CO₂ interpolation results using IDW interpolation were the most accurate，while the Rn results using OK interpolation(Table 4, Table 5).

Generally, the CO₂ and Rn anomalies were spatially consistent, except for slight differences in size between the distribution areas. Rn anomalies were identified in two areas (Figures 4, 5): (1) southern Haiyuan County, from the surface rupture zone of Caiyuan to Xiaoshan; and (2) the southern end of the HYSRZ, from north of Xiaokou to Heicheng.

Figure 4. Spatial distribution of soil gas radon (Rn) in the Haiyuan surface rupture zone.

Figure 5. Spatial distribution of soil gas carbon dioxide (CO₂) in the Haiyuan surface rupture zone.

Discussion

Correlation of gaseous emissions with fault behaviour and seismic activity

Tensional faults and fractures provide the best channels for deep gas emission. On this basis, anomalies of Rn and CO₂ analyses along the fault segments of SF1 and SF4 can be attributed to higher rupture rates, tensile deformation, and higher fault slip rate. Specifically, the Xiaoshan–Caiyuan segment (SF4) of the HYSRZ is characterised by a series of echelon tensile fractures and normal faults (Institute of Geology, China Earthquake Administration Earthquake Agency of Ningxia Hui Autonomous Region 1990). Several NW trending faults are present near Xiaokou (SF1), and the entire region is severely fragmented. These findings on soil gas in this area support the observation that the gas concentration in the tension zone is often higher compared to other areas (Jiang & Yan 1990). Moreover, the mean displacement of the river system, fault slip rate in the late Holocene, and fracture zones of the SF1 and SF4 are wider than those of other segments (Table 1). The concentration of radon and mercury detected in the geochemical profile in HYF is directly correlated with fault activity. Specifically, the SF4 segment has the highest level of shear fault activity, with Rn concentration being tenfold greater than the background value. Moreover, there is a notable disparity in the levels of Rn in SF1 as opposed to SF4, with substantially lower amounts observed in SF2 and SF3. The cross-fault findings confirm the results of this investigation (Meng et al. 1997; Sun et al. 2016), indicating a significant relationship between variations in gas concentrations and fault activity. In addition, CO₂ anomalies were also detected in the Xi’anzhou area, which may be related to the lush vegetation and relatively dense population. The lowest concentrations of Rn and CO₂ were recorded in the middle segment of the HYF, from Caowa to Caixiang, where faults have not been active since the late Pleistocene confirmed by a field geological investigation profile in Wangjia (near SF3, Institute of Geology, China Earthquake Administration Earthquake Agency of Ningxia Hui Autonomous Region 1990).

The geochemical characteristics of gases are also closely associated with seismic activity (Pizzino et al. 2004; Li et al. 2017). To show a correlation between stress state and soil gas variation, b value were calculated by a coefficient of the Gutenberg–Richter equation that can reflect stress accumulation in active fault zones (Scholz 1968; Yi et al. 2013). In this study, Rn and CO₂ anomalies (Figure 6) were mainly distributed around the western and southeastern segments of the HYF, along which earthquakes more frequently occur compared with other segments, and almost all earthquakes of M > 3 have occurred in the gas anomaly area since 1970. Moreover, the Rn and CO₂ anomalies were spatially concordant with areas of low b values (<0.7 is set as the threshold for the high stress state in study area) in western Haiyuan County and northern Xiaokou (Figure 6). The spatial distribution of low b-values corresponds to regions of high gas concentration, as illustrated in the Generalized HYF (Zhou et al. 2017), the Xiqinling Fault (Su et al. 2013), and the Tangshan area (Lu et al. 2022). The permeability and porosity of the underground media are enhanced by frequent seismic activity and high stress state, which is conducive to the upward migration of deep gas.

Figure 6. Spatial distributions of b value and gas anomalies in the Haiyuan surface rupture zone.

The low concentrations of soil gases near the HYEQ epicentre may be related to the sandy soil overburden layer of the Yanchi Basin, which is not conducive to gas occurrence; moreover, gas emissions may be obstructed by fault healing because the elapsed time of the HYEQ is > 100 years. Finally, relatively weak seismic activity and slip rates in this middle segment of the HYF are smaller than those of other segments (Sun et al. 2017), which is consistent with the low concentrations of soil gases.

Crustal velocity structure and gas migration

Two seismic exploration profiles were established across the HYF and area of gaseous anomalies (Figure 1). The A–B profile of S-wave velocity (Figure 7a) revealed low-velocity bodies at 0 ∼ 10 km depth in the area of the Rn and CO₂ anomalies. A similar phenomenon was observed in the area of the C–D profile (Figure 7b), which coincides with the distribution of low-resistivity in the upper crust (Zhan et al. 2004). Local low-resistance bodies (resistivity from 60 to 90 Ω • m) were also found at 15 to 40 km depth below the epicentre of the HYEQ (near profile C–D). Similarly, the deep electrical structure of the Xiaokou area (near profile A–B) also showed obvious low-resistance layers of ∼10 Ω • m at depths of < 5 km in the upper crust and ∼25 km in the lower crust. (Zhan et al. 2017). The layers of low velocity and high conductivity result from fluid enrichment.

Figure 7. Cross-sectional image of S-wave velocity variation along the a–B (a) and C–D (b) profiles.

Deep fluids may control the distribution of low-velocity and high-conductivity bodies in the lower crust of the Haiyuan area (Xin et al. 2020), and the low-velocity and high conductivity bodies provide a source of soil gases. The HYF extends down more than 30 km into the lower crust (Zhan et al. 2004), providing more channels for deep fluid migration to surface. Moreover, moderate and small earthquakes with focal depths of 0 ∼ 15 km increase permeability, further facilitating deep fluid movement.

The influence of environmental parameters on soil gas

In addition to seismological characteristics, the concentration and migration of soil Rn are affected by many environmental factors, such as thoron concentrations (Jaishi et al. 2014), meteorological parameters, terrain, and vegetation (Sundal et al. 2008). Usually, the presence of abnormally high radon concentration with a low level of thoron activity indicates that the radon originates from a deeper source. Conversely, higher concentrations of thoron with low radon activity suggested that they have been transported to the surface from a shallower source (Giammanco et al. 2007). In long-term continuous soil Rn measurements, it has been confirmed that air temperature, air flows, and air pressure play an important role in soil radon concentrations (Sundal et al. 2008). Moreover, Soil moisture conditions and wind speed were also significant factors in the soil Rn migration (Fujiyoshi et al. 2006). Elevation, soil temperature, soil water content and normalized difference vegetation index (NDVI) were investigated to determine the relationship between Rn concentration and environmental parameters in HYSRZ. The correlation coefficient shown that the spatial distribution of Rn concentration in HYSRZ is not significantly affected by environmental conditions and almost all correlation coefficients are less than 0.1 (Table 6). Therefore, the findings demonstrate that seismogeological parameters were the primary influence on soil gas concentration in HYSRZ. Nevertheless, meteorological factors must be considered if continuous Rn investigations are to be conducted in the HSFZ in the future. Machine learning techniques, such as the layered neural network (LNN) and Random Forest (RF), can be used to estimate Rn concentration affected by environmental parameters and identify the anomalies caused by seismic activity (Negarestani et al. 2002; Naskar et al. 2023).

Table 6. Pearson correlations between Rn concentration and environmental parameters.

	Elevation	Soil temperature	Soil water content	NDVI
correlation coefficient vs. Rn	−0.01	0.09	−0.09	0.02

Conclusions

In this study, the relationship between the spatial variation of soil gases and fault activity along the HYSRZ at a regional scale was determined by measuring Rn and CO₂ concentrations. Combined with geophysical and seismic data, our results support three main conclusions. (1) Rn and CO₂ anomalies have a similar spatial distribution, and are concentrated in southern Haiyuan County and at the southern end of the HYSRZ. This suggests that CO₂ may act as a carrier for Rn and/or they may be derived from the same sources. (2) Geochemical variations in soil gases are closely related to fault behaviour. In particular, tensional faults and fractures, frequent seismic activity, and low b values (i.e. high tress) result in higher concentrations of Rn and CO₂. (3) Low-velocity and high conductivity bodies beneath the HYSRZ facilitate the migration of deep-Earth fluids, including CO₂ and Rn. （4）The spatial distribution of Rn concentration in HYSRZ is not significantly influnced by environmental paramters. These results offer support for the design of a continuous geochemical measurement network in this area.

Owing to topographical limitations and poor transportation systems, the spacing of sampling sites in some areas have the unreasonable problem, which may lead to the soil gas variation may be not very accurate. Moreover, as thoron concentration was not measured during this study, the relationship between Rn concentration and thoron activity cannot be performed in this paper. The S-wave velocity deprived from campaign seismic stations may not reflect the subtle feature of deep crustal structure in study area, a follow-up study on relationship between soil gas variation and crustal structure can be performed by high-density electrical method and the seismic wave velocity structure in a large scale.

Data availability statement

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [grant number 42071230], Ningxia Natural Science Foundation [grant number 2022AAC03698], and the innovative teams of the Earthquake Agency of Ningxia Hui Autonomous Region [grant number CX2023-2].