As part of the earthquake precursor detection program, changes in groundwater levels, temperatures, and chemical composition related to earthquakes have been documented for decades (Toutain et al., 1997; Roeloffs, 1998; Montgomery and Manga, 2003; Wang and Manga, 2010; Skelton et al., 2014; Manga and Wang, 2015; De Luca et al., 2018; Martinelli, 2020). Because of the feasibility of using automated monitoring systems, earthquake-related changes in water levels, temperatures, and radon concentrations are the most commonly reported responses (King, 1981; Kitagawa et al., 1996; Manga and Wang, 2015; Martinelli, 2020). Hence, these monitoring methods of groundwater also play key roles in precursor detection in China, and a nationwide network of monitoring wells located along strain-sensitive locations has been constructed for this purpose.

In contrast, groundwater geochemical changes induced by earthquakes have been documented relatively rarely (Martinelli, 2020; Shi et al., 2020) because manual sampling and expensive, time-consuming laboratory analyses are required (Wang and Manga, 2010). However, such studies are highly important because earthquake-related chemical changes are considered to be sensitive to crustal stress and to be beneficial for earthquake prediction (Thomas, 1988; Wakita, 1996; Poitrasson et al., 1999; Manga and Wang, 2015; Martinelli and Dadomo, 2017; Shi et al., 2020; Kopylova et al., 2022). For instance, based in part on hundreds of hydrological anomalies including geochemical changes, an imminent prediction was made before the 1975 M 7.3 Haicheng earthquake in China (Wang et al., 2006). Following the 1995 M7.2 Kobe earthquake, several papers reported precursory changes in the concentrations of radon, chlorine, and sulfate ions in groundwater (Igarashi et al., 1995; Igarashi and Wakita, 1995; Tsunogai and Wakita, 1995). Claesson et al. (2004) observed simultaneous changes in trace elements, major elements, and isotopes before and after major earthquakes. Barberio et al. (2017) and Rosen et al. (2018) found that some trace elements exhibited responses to the seismic sequence in central Italy. Skelton et al. (2019) found that changes in major ions and hydrogen and oxygen isotopes were associated with three different earthquakes in Iceland. Shi et al. (2020) found that trace element concentrations showed coseismic decreases following the Tonghai earthquake in the Yunnan Province of China, but no significant changes occurred in major ions and hydrogen and oxygen isotope concentrations.

Overall, according to previous geochemical research, most documented studies investigated the major elements (Tsunogai and Wakita, 1995; Toutain et al., 1997; Kingsley et al., 2001; Woith et al., 2013) and hydrogen and oxygen isotopes related to destructive great earthquakes (Taran et al., 2005; Skelton et al., 2019), and only a few studies focused on trace elements’ responses to the earthquake (Rosen et al., 2018; Shi et al., 2020) despite changes in trace elements being more sensitive to small earthquakes than the commonly used chemical constituents. In this paper, we continuously collected 36 water samples of the hot spring from August 2021 to August 2022, and each water sample contained 41 trace elements which contributed to the comparison of the different responses to earthquakes and the discussion on the possible mechanisms and their implications for future seismic monitoring.

North China has experienced a period of strong earthquakes frequently over the last few decades. From 1966 to 1998, a series of strong earthquakes occurred, such as the Xingtai M7.2 earthquake in 1966, the Bohai M7.4 earthquake in 1969, the Haicheng M7.3 earthquake in 1975, and the Tangshan M7.8 earthquake in 1976. Since the Zhangbei M6.2 earthquake in 1998, there has been no earthquake above M6 for 20 years. The seismic activity in North China has been relatively quiet, and there would be a potential danger of damaging earthquakes in the next few years.

Zhangjiakou-Bohai Fault Zone is a group of NW-W orderly active fault zones with high seismic activity in North China; the fault zone starting from the northern margin of Taihang Mountain in the west is distributed along the junction area of Yanshan Mountain and the North China Basin and enters the Bohai Sea in the east. In this zone, many violent earthquakes above M6 have occurred (Yang et al., 2022), so it is an important fault zone for earthquake monitoring and research.

Beijing is the capital of China with a large population, and it is located at the intersection of the North China Plain and Zhangjiakou-Bohai seismic belt, where active faults have developed. These faults mainly included the NW orderly Nankou-Sunhe Fault Zone and two groups of NE orderly active fault zones named Huangzhuang-Gaoliying Fault Zone and Shunyi-Liangxiang Fault Zone. All earthquakes occurred along these fault zones or where they intersect (Figure 1), and all these earthquakes were less than M6. Therefore, it is of great scientific and practical significance to carry out small earthquake monitoring and forecasting in the Beijing area.

To better monitor the seismic activity in the capital area, the Chinese Earthquake Administration (CEA) started to deploy the Beijing Metropolitan Digital Seismic Network (BSN) in the late 90s. Currently, the network consists of broadband, borehole, and surface short-period stations that cover Beijing and the surrounding areas densely and uniformly (Jiang et al., 2008; Li and Huang, 2014). Since 2001, 107 stations have been installed in the Chinese capital region, which is the most advanced and densest digital seismic network in Mainland China. A large number of high-quality waveforms have been recorded, which provides an opportunity to improve the location accuracy of small earthquakes in the region (Jiang et al., 2008).

Beijing hosted the 2022 Winter Olympic Games from February 4th to 20th. During this time, continuous monitoring of possible earthquake areas was strengthened. On 3 February 2022, an earthquake with a magnitude of ML3.3 occurred in the Chaoyang district of Beijing (Figure 1). In this study, we conducted the geochemical monitoring study of the Songshan hot spring located in the Yanqing district of Beijing; the spring had an epicentral distance of ∼75 km from the Chaoyang earthquake (Figure 1). Geographically, this spring is exposed to the valley of Dahaituo Mountain which is composed of granite with fissure development (Yang et al., 2001), and fault activities were good channels for deep groundwater migration. The groundwater of Songshan spring comes from the granite of Dahaituo Mountain; there is no lake, river, or surface reservoir in the area.

Songshan Spring is an artesian type of rising spring, and the water is from the granite rock body. The spring has 41°C–42°C water temperature, high fluoride content (15.93 mg/L), and high silica content (54 mg/L); the water type is Na-SO4 (Yang et al., 2001). The main observation items of Songshan Spring were water mercury and there were high correspondences between mercury anomalies and earthquakes (Yang et al., 2001).

In addition, a seismic groundwater monitoring well named Wuliying is ∼10 km near the Songshan spring. The Wuliying well is located on the subsidence transition zone at the northern edge of the Yanqing Basin, with compressive faults developing around the well area (Gao and Xing, 2011). The groundwater level observation is one of the items of the Wuliying well and is usually recorded at 1-min intervals by a SWY-2 digital instrument developed and produced by the Institute of Crustal Dynamics, China Earthquake Administration. The water-level sensor has a range of 0–50 m, resolution of 1 mm, sampling interval of 1 s, precision of 0.02 percent, and accuracy of 0.05 percent.

Songshan spring has no water level monitoring; we checked water level changes in the Wuliying well near the spring for hydrological responses to earthquakes. The groundwater level in the Wuliying well continuously increased from 1.3 to 1.8 m between July 2021 and August 2021 (Figure 2). No abnormal changes in groundwater level were found related to any earthquakes (Figure 2). The water level change also has no relationship to precipitation as there is only large precipitation during summer (from June to August every year).

The results of analyses of the 36 water samples are listed in the Supplementary Material, (Concentrations of 41 trace elements of each water sample in the Songshan spring). The PAAS-normalized 15 rare earth elements (REE) display remarkable positive Eu anomalies and flat distribution patterns (Figure 3A). Eu concentrations showed a significant increase from 0.002 to 0.007 mg/L from February 2022 to August 2022 (the end of sampling in August) (Figure 3B). These changes were more evident when data smoothing using the mathematic method (rLOWESS, robust Locally Weighted Scatterplot Smoothing) (Figure 3C). Eu had a marked increase after early February.

Apart from REE, the other 26 trace elements show different changes over the monitored period. Based on element correlative clustering, these elements’ constitutes can be broadly classified into four groups (Figure 4). The first group consists of Li, Sc, Ti, and Pb elements which cluster each other in the correlation network (pink circles in Figure 4). The second group consists of Cu, Nb, Th, Zn, Tl, and U elements clustering in grey circles in Figure 4. The third group consists of Ga, Mo, Ba, Rb, Cs, and Sr elements clustering in green circles in Figure 4. The fourth group consists of Be, Mn, Cr, Co, Y, Bi, Cd, Sb, Ni, and V elements dispersing in the correlation network (white circles in Figure 4).

The elements of the first group had upward changes (Figure 5): Li, Sc, Ti, and Pb concentrations continued to rise from December 2021 to February 2022 and peak, then remained stable at high values from March to May 2022, and decreased from June to August 2022 (several months after the earthquake). These changes were quite marked when using the smoothing methods (LOWESS, rLOWESS, LOESS, and rLOESS) (Figure 5).

The elements of the second group had significant downward changes (Figure 6): Cu, Nb, Th, Zn, and Tl contents continued to decline from August 2021 to January 2022, then remained stable at minimum values from February to August 2022 (Figure 6). U contents declined from August 2021 to October 2021, then maintained stable values from November 2021 to February 2022, and the last decreased from March to August 2022 (Figure 6).

The elements of the third group had upward and then downward changes (up-convex shape) (Figure 7): elemental concentrations continued to rise from August 2021 to January 2022 and peaked, then remained stable at high values from February to April 2022, and the last decreased from May to August 2022 (Figure 7).

The elements of the fourth group had no regular changes, and the relationship with earthquakes was not clear.

Chaoyang ML3.3 earthquake was an epicentral distance of ∼75 km from Songshan hot spring (Figure 1). The seismic energy density produced by this earthquake at the Songshan hot spring was ∼0.0019 J/m3 (calculation according to the method by Wang, 2007), which is the same order of magnitude with the minimum energy density (10–4 J/m3) that was required to trigger hydrological responses based on global observations (Wang, 2007). However, there were no changes in the groundwater level of the Wuliying well near the Songshan spring groundwater level to the Chaoyang earthquake, which means that the Chaoyang ML3.3 earthquake might have been too small to cause the hydrological responses of Songshan spring, and no additional water into the wellbore could be expected.

According to the above results, the changes in trace elements have a strong temporal correlation with earthquakes, which means that trace elements were probably sensitive to the small earthquake, but different kinds of trace elements have particular responses to the Chaoyang earthquake, for instance, the Eu element of rare earth elements had obvious upward changes after the Chaoyang earthquake. The first group elements’ (Li, Sc, Ti, and Pb) concentrations had upward changes 2 months before the Chaoyang earthquake, then returned to their initial concentration a few months after the earthquake. The second group elements’ (Cu, Nb, Th, Zn, Tl, and U) concentrations had downward changes a few months before the Chaoyang earthquake, then maintained stable minimum values until the end of sampling.

This study has the following three conclusions.

1. Although the Chaoyang ML3.3 earthquake had a small magnitude and longer epicenter distance, some trace elements in Songshan spring are likely to be sensitive to crustal minor strain by the earthquake. The Eu element of rare earth elements had obvious upward changes after the earthquake. The first group consisting of Li, Sc, Ti, and Pb elements had significantly upward changes, and the second group consisting of Cu, Nb, Th, Zn, Tl, and U elements had downward changes before the earthquake.