Taiwan is located at the plate boundary between the Philippine Sea Plate and the Eurasian Plate, where the Taiwan mountain belt was formed as a result of arc–continent collision (Li, 1976; Teng, 1990). The Luzon Arc moves northeast toward the Asian continent at an average rate of 7 cm per year (Yu et al., 1997). Across the Longitudinal Valley Fault (LVF), the fast speed of the movement decreases dramatically from the arc (Coastal Range) of 7 cm per year to the continent (Central Range) of 2 cm per year (Ching et al., 2011). The highly deformed Lichi Mélange formation, which was the forearc basin and composed of chaotic mudstone intermixed with exotic blocks of various sizes and lithology including fragments of oceanic crust (Chang et al., 2000), may be the substance that absorbs the movement of 5 cm per year (Ching et al., 2011) and provides the suitable environment for the formation of MVs.

Two MVs are located on the hanging wall of the Lichi Fault in the Lichi Mélange and the foothill of the Coastal Range (Figures 1A, B). In the south, MV Lei-Gong-Huo (LGH) appears as an elongated mud shield with an area about 150 m long and 50 m wide (Figures 1C–E). On the top of the mud shield, dozens of mud pools from about 10 cm to 2 m in diameter are distributed along the direction of approximately N15°W (NNW–SSE). The mud pools in the south are distributed linearly, but the northern ones are sparsely distributed (Figure 1C). A total of 46 mud pools in this area were sampled from October 2015 to July 2016 monthly and more than half of them are short-lived ones. Not all mud pools were sampled except the November 2015, May 2016, and July 2016 visiting. Seven groups in the MV LGH are classified according to their locality (Figure 1C). Group A to E are five clusters of mud pools distributed from north to south along the axis of the mud shield, and group F is the mud pools on the western flank. Mud pools in the northwestern corner belong to group G. In the north, MV Luo-Shan (LS; previously also named Yencheng by Shih, 1967) contains dozens of linearly distributed swamp-like mud pools with larger coverage, approximately 1 km in length and 200 m in width, and have a direction of N20°E (NNE-SSW), parallel to the strike of the Lichi Mélange formation and major faults. Due to the single fluid source of MV LS (Chao et al., 2013), only the most active one, LS-A1, has been sampled as the reference site.

MVs LGH and LS emit predominantly methane (>90%), similar to typical sedimentary MVs. However, low CO₂ (<0.2%) in both MVs with higher N₂ concentration (>5%) and mantle helium signals of MV LS (Yang et al., 2003; Yang et al., 2004; Chao et al., 2010) show different gas composition compared with sedimentary MVs. MV LS emits thermogenic methane, but MV LGH emits methane of a mix of thermogenic and microbial sources (Chao et al., 2010; Sun et al., 2010). The Sr isotopes of the fluids in LS show the typical and consistent igneous characteristics for all MVs collected in this area (You et al., 2004; Chao et al., 2013) but LGH has stronger sedimentary contributions with larger ⁸⁷Sr/⁸⁶Sr distribution, although there are only two samples presented (Chao et al., 2013).

MV fluid samples were collected approximately 20 cm right below the gas bubbling area using four 50-cm³ polypropylene (PP) centrifuge tubes. The temperature, pH, and oxidation-reduction potential (ORP) values were obtained on site with a WalkLAB^® TI9000 temperature compensation pH meter. After centrifuge, samples were filtered with 0.45-μm nylon membrane filters. The percentage of mud weight was determined by the weight of the mud after 50°C drying in the oven over the weight before centrifuge. For each sample, approximately 60 ml of the filtered solution was acidified with purified concentrated nitric acid to pH <2 for determination of major, trace elements, and Sr isotopes. Unacidified samples were preserved for the measurement of anion concentrations, total alkalinity, and water isotopes. All samples were kept at 4°C in the refrigerator for later analysis.

Dissolved anions (Cl⁻) were determined using ion chromatography (Dionex^® ICS-3000) with a precision of 5%, and quality assurance was obtained by the diluted international seawater standard IAPSO. Chloride concentration is calculated based on the salinity of IAPSO and the equation established by Millero et al. (2008). Major and trace elements (B, Ba, Ca, Fe, K, Li, Mg, Mn, Na, S, Si, and Sr) were measured using Agilent 5100 inductively coupled plasma optical emission spectrometry (ICP-OES) with a precision of better than 3%. Total alkalinity (TA) was measured with acid titration (Metrohm^® 905 Titrando) with the precision better than 1%, estimated by repeating the analyses of the sample (n = 3) and in-house-prepared bicarbonate standard.

H and O isotopes were obtained by the cavity ring-down spectroscopy (CRDS, H₂O isotope analyzer, Picarro L2140-i) installed at the Exploration and Development Research Institute, CPC Corporation, Taiwan (EDRI). Instrumental fractionation was corrected by measuring two international standards (V-SMOW2 and SLAP2). This correction was certified by a third international standard (GISP). For drift correction and quality control, four USGS standards of known δD and δ¹⁸O values (USGS-45, USGS-47, USGS-48, and USGS-50) were analyzed after six consecutive samples were run. The reproducibility of δD and δ¹⁸O is better than 0.1‰ and 0.03‰ (2σ, n = 6), respectively, by repeating the analyses of some selected samples. All δD and δ¹⁸O isotopic results reported in this study were normalized to V-SMOW2.

The separation and purification of Sr were performed using an extraction chromatography technique, Sr Spec^® resin (Eichrom Technologies, USA). The procedure was modified from the published protocol by Ohno et al. (2008). In brief, 360 ng of Sr, which is enough for triple measurements, from the samples was evaporated to dryness and redissolved in 1 ml 2 M of HNO₃. Acid-cleaned PP columns with an inner diameter of 5 mm were loaded with 1 ml of Eichrom Sr Spec^® resins (100–150 mesh). After 2 ml 2 M of HNO₃ precondition, most of the potential matrix and isobaric interference elements such as Ca, Na, Rb, and Zr were eliminated by the first 6 ml 2 M of HNO₃, and Ba was removed by the following 6 ml 7 M of HNO₃. The presence of Ba in the sample solution causes significant isotope fractionation of the stable Sr in the mass spectrometer (de Souza et al., 2010; Chao et al., 2013; Scher et al., 2014). At last, the Sr fraction was collected by 6 ml of deionized water, evaporated to dryness, and redissolved in 2.4 ml 0.3 M of HNO₃ in sequence. The elution curve shows the recovery of Sr is higher than 98%.

The Sr isotopic composition of fluid samples was determined using a multicollector inductively coupled plasma mass spectrometer (MC-ICP-MS, Neptune, Thermo Fisher Scientific) installed at the Earth Dynamic System Research Center, National Cheng-Kung University. A modified empirical external normalization (EEN, Liu et al., 2012) with the traditional standard-sample-bracketing (SSB) technique was used to correct instrumental mass bias. After column purification, sample solutions were adjusted to a final concentration of 150 ppb Sr with spiked 300 ppb Zr (High-Purity Standards^®) in 0.3 M HNO₃. ⁸⁶Sr, ⁸⁷Sr, ⁸⁸Sr, ⁹⁰Zr. and ⁹²Zr ion-beams were collected simultaneously in a static mode. ⁸⁵Rb and ⁸³Kr were monitored for the correction of ⁸⁷Rb contribution on ⁸⁷Sr and ⁸⁶Kr on ⁸⁶Sr. The mass bias correction for traditional ⁸⁷Sr/⁸⁶Sr ratio was based on the assumption of constant ⁸⁸Sr/⁸⁶Sr ratio (8.375, 209). The monitored ⁹²Zr/⁹⁰Zr ratio was applied for mass bias correction on ⁸⁸Sr/⁸⁶Sr determinations.

The reproducibility of ⁸⁷Sr/⁸⁶Sr and δ⁸⁸Sr in the international seawater standard IAPSO and the inhouse seawater LMN_4-3 is better than 0.000014 and 0.026‰ (2σ, n = 5) with the values of 0.709, 178, 0.378‰, and 0.709, 173, 0.380‰ respectively. The δ⁸⁸Sr results are expressed using conventional notation modified after Krabbenhöft et al. (2009): δ 88 S r = [ S 88 r / S 86 r s a m p l e S 88 r / S 86 r N B S   S R M   987 − 1 ] × 10 3   ( ‰ ) (1) where ⁸⁸Sr/⁸⁶Sr_sample and ⁸⁸Sr/⁸⁶Sr _standard denote the ⁸⁸Sr/⁸⁶Sr ratios in the sample and standard respectively.

Radiogenic Sr isotope, ⁸⁷Sr/⁸⁶Sr, is a robust tool to decipher the source of the fluids. The fluids involved with sedimentary and igneous rocks are the best objects because of the distinct end members. By comparing the correlation between the chemical composition and ⁸⁷Sr/⁸⁶Sr ratio of the fluids in LGH, two major end members are revealed. One is the fluid interacted with sediments, which are possibly inside the Lichi Mélange. The other is the fluid interacted with igneous rocks, which are the andesitic volcanic arc in the east also the basement of the Coastal Range (Figure 1). Sedimentary fluid carrying ⁸⁷Sr/⁸⁶Sr ratio higher than seawater, high Na, Mg, Sr, and Ba concentrations, and low Ca, S, and B concentrations with low δ⁸⁸Sr is distributed in the northwestern area while andesitic fluid carrying low ⁸⁷Sr/⁸⁶Sr ratio, low Na, Mg, Sr, and Ba concentrations, and high Ca, S, and B concentrations with higher δ⁸⁸Sr is located at the southeastern side (Figures 3 and 7; Supplementary Table S1). LGH-D4 and group E are the representatives of andesitic fluids, while groups F and G can be the representatives of sedimentary fluids (Figures 3 and 7). The other groups located on the axis of the mud shield emit fluids mixing with two end members (Figures 3 and 7). However, some sedimentary-fluid-dominated mud pools (i.e., LGH-F1, LGH-G5, LGH-G6, and LGH-G8) do not fall on the mixing line between the two major end members and a trend toward ancient seawater can be extrapolated (Figure 4), indicating contribution of residual seawater to a sedimentary end member.

Chao et al. (2013) reported large ⁸⁷Sr/⁸⁶Sr variations by two MV fluids collected in LGH in the year 2008. This chemical characteristic of two-end-member-mixing has lasted at least since the year 2008. The mud pools distributed on the axis of the mud shield (e.g., LGH-A1, LGH-B1, LGH-C1, and LGH-D1) have relatively large temporal variation (Figures 7 and 8). The representative of two end members (i.e., LGH-D4 and LGH-F1) shows minor to no variations. There was a shift toward the mixing area before November 2015 for mud pools on the axis and lasted until July 2016, the time of the last visit (Figure 8). The northern mud pools LGH-G1 and LGH-G3 emitted more andesitic fluids during April 2016 and May 2016, and went back to its previous condition in June 2016 although the western one, LGH-G3, did not fully recover. The time variation of the dominated source of mud pools on the axis possibly results from response of earthquake activities as indicated in the literature (Bonini et al., 2016; Bonini, 2020; Bonini, 2021). In summary, andesitic fluid dominates the southeastern region of MV LGH, while sedimentary fluid dominates the northwest. The spatial distribution matches the geological background of the volcanic arc in the east and the forearc basin right beneath and in the west.

Generally speaking, sedimentary MVs have simultaneous low Cl concentration with higher O isotope and lower H isotope relative to seawater, indicating dilution of fresh high δ¹⁸O and low δD water by clay dehydration in the source region (Dählmann and de Lange, 2003; Hensen et al., 2004; You et al., 2004; Chen et al., 2020). MVs in western Taiwan emit fluids of classical geochemical characteristics (You et al., 2004; Yeh et al., 2005; Chao et al., 2011; Chao et al., 2013). However, eastern Taiwan MVs contain fluid of both chloride concentration and water isotopes lower than seawater, indicating different mechanisms or fluid source compared with those of the western ones. Considering the hypothesis that MV fluid originates from altered ancient marine pore water, several possible mechanisms, such as meteoric water input, clay dehydration, or water–rock interaction may be involved in altering the pore water.

Marine pore water originally recorded chemical and isotopic composition of the seawater at the time sediment deposited. As the depth, pressure, and temperature increased, the chemical and isotopic composition of the pore water was altered by early diagenesis, the addition of lateral transport fluids or diffusive/advective fluid migrated from deep underneath (e.g., Kopf, 2002; Aloisi et al., 2004; Hensen et al., 2004; Mazzini and Etiope, 2017; Chen et al., 2020). Lichi Mélange, which is the most likely formation where the fluid source is located, is estimated to be deposited from 5 to 6 Ma (Lin et al., 2019). The δ¹⁸O value of the ancient seawater at that time estimated by benthic foraminifer is about 0.2‰–1.0‰ lower than the present value (Zachos et al., 2008; Lear et al., 2015). This result seems to be a reasonable explanation for water isotope compositions of eastern Taiwan MV fluids since the average δ¹⁸O values are −0.55 and −1.27‰ for LS-A1 and LGH, respectively. However, the preservation of original deposited seawater cannot explain the lowering of the chloride concentration and the enrichment of Ca. Chloride has a very long residence time in seawater (87 million years; Berner and Berner, 1987), indicating limited chloride variation in this time scale. The temporal variation of chloride between 5 and 6 Ma, and the present time is too small to be detected by our quantification precision.

Marine pore waters which show similar chemical and isotopic pattern to eastern Taiwan MV fluids are reported in the literature (Perry et al., 1976; Lawrence et al., 1979; Lawrence and Gieskes, 1981; Gieskes et al., 1990; Gieskes et al., 1993; Kastner et al., 1993). They contain lower Cl, Mg, K, δD, and δ¹⁸O but higher Ca relative to seawater. This pattern is interpreted as low-temperature alternation of volcanic ashes and underlying basaltic ocean crust coupled with diffusive exchange with overlying ocean column (Garlick and Dymond, 1970; Lawrence and Gieskes, 1981). Although the major pattern matches, the slope of oxygen isotope decreasing versus chloride decreasing and calcium increasing are different. Adopting the regression trend established by DSDP cores (Lawrence and Gieskes, 1981) to MV LGH, δ¹⁸O value will have an end member of −7‰ or −12‰ based on the Ca concentration or Ca/Cl ratio respectively. The end-member calculation based on Ca and Ca/Cl ratio of LS-A1 is −3‰ and −9‰, respectively. These end members show lower δ¹⁸O than MV fluids and indicate that δ¹⁸O of eastern Taiwan MV fluid is not simply controlled by water−rock interaction of ancient seawater and volcanic materials. A third mechanism or an additional source of the fluid has to be involved. Considering the Cl concentration of the MV fluids is relatively lower than that of marine pore water showing similar characteristics, this mechanism or additional fluid has to dilute the MV fluids.

Meteoric water is one of the possibilities, and it was previously interpreted as a possible source of eastern Taiwan MV fluid (Yeh et al., 2005). The water isotope values of LGH fall right on the local meteoric water line, and the values of LS only deviated a little from the line (Figure 6), making the interpretation look reasonable. However, neither the direct dilution with seawater nor ancient marine pore water by the weighted mean value of meteoric water (Peng et al., 2010) result in the value of the MV fluids. MV LS has a lower chloride concentration but a higher δ¹⁸O relative to LGH, showing conflicting results to meteoric water input. Moreover, water isotopes of meteoric water have strong seasonal variation (Peng et al., 2010), but most fluids from individual mud pools reveal minor temporal variation. Indeed, a diluted sample, LGH-G6_2016–3, is affected by meteoric water input but it is not the major source of MV fluid.

Clay dehydration is one of the common fluid sources for both marine and terrestrial MVs (e.g., Dählmann and de Lange, 2003; Hensen et al., 2004; Chao et al., 2013; Chen et al., 2020). The formation temperature of the eastern Taiwan MV fluids is estimated by major elements (Chao et al., 2011; this study), the existence of the thermogenic methane and equilibrium temperature estimated by methane clumped isotopes of MV LS-A1 (76°C –101°C; Lin et al., 2019) indicate that the source region has reached the temperature of smectite to illite transformation (40°C –150°C; Freed and Peacor, 1989; Ijiri et al., 2018). The addition of clay dehydration fluid leads to the dilution of chloride but increases δ¹⁸O of the MV fluids. This back and forth scenario neither exceeds present seawater value nor MVs in western Taiwan possibly due to a lower degree of dehydration induced by a lower formation temperature (Chao et al., 2011). The large quantify of smectite distributed in present forearc basin (Nayak et al., 2021) indicates sufficient source of fresh water, even the degree of dehydration is low.

In summary, oxygen isotope in eastern Taiwan MV fluids originates from the ancient seawater approximately 5–6 Ma carrying δ¹⁸O values lower than the present value (−0.2‰–1.0‰), decreased by water–rock interaction with volcanic ashes or andesitic basement, and increased by clay dehydration fluid coupled with further chloride dilution.

Water isotopes of LGH fluid show an evaporation trend with the slope close to 2. Normally, the slope of local evaporation line of surface water is around 4 to 5 for normal temperature and humidity conditions. The slope increases with decreasing temperature but no clear relationship with humidity (e.g., Gibson et al., 2008). The surface temperature of eastern Taiwan MV fluids is similar to or only a little higher than the air temperature, from 18°C to 36°C (Supplementary Table S1). The reasonable slope of this temperature is approximately from 3.0 to 4.5 for the surface water (Gibson et al., 2008), still higher than the observed slope. However, a similar slope has been documented for soil moisture (Dincer et al., 1974). The evaporation fractionation of MV fluids may have similar behavior to soil moisture.

Due to strong negative correlation between Na and Ca concentration of the LGH fluid, the end-member composition of sedimentary and igneous fluids can be estimated by the relationship between elements and Na or Ca. Groups F and G are the candidates of sedimentary fluid, while group E and mud pool D4 are the candidates of the igneous fluid. The maximum values of elements which have high positive correlation (r > 0.65; Supplementary Table S2) with Na or Ca are chosen among candidates as the end-member composition, and the minimum values of elements with high negative correlation (r < −0.65) are chosen among candidates. The average values are used for those which have low correlation such as Cl⁻, K, Ba, and so on. Samples fractionated by evaporation are excluded for the average value of δ¹⁸O and δD. The estimated end-member composition of the two fluids is listed in Table 1.

Besides the high Na concentration and ⁸⁷Sr/⁸⁶Sr ratio, sedimentary fluid shows significantly high total alkalinity, Sr, Mg, Ba, and Li with low S, Ca, B, and δ⁸⁸Sr relative to igneous fluid. B concentration in MV fluids is majorly controlled by readsorption by clay minerals (Chao et al., 2011). The sedimentary fluid has end member concentration of B almost an order lower than igneous fluids, possibly resulting from intense adsorption of B and indicating more sorbent presented in the sedimentary fluid-dominated reservoir. Total alkalinity in the MV fluids generally correlated strongly with CO₂ content of emitted gas (Chao et al., 2010; Lavrushin et al., 2015). Higher alkalinity in the sedimentary end member probably indicates it emits gas with higher CO₂ concentration compared with igneous mud pools. Unfortunately, the gas samples were not collected in this study, and the gas composition data in the literatures were collected from the mud pools on the axis, which are mostly dominated by igneous end member. MV LS-A1 shows total alkalinity concentration lower than LGH and emits lesser CO₂ gas as well (Yang et al., 2004; Chao et al., 2010; Sano et al., 2017), supporting the hypothesis of sedimentary-dominated mud pools in LGH emitting more CO₂.

Calcium and sulfur have interestingly strong positive correlation (r > 0.99; Supplementary Table S2; Figure 3), but they have three orders of magnitude differences in concentration. Anaerobic oxidation of methane (AOM) is the major sink of the sulfur in MV fluids (e.g., Chang et al., 2012) and significantly decreased below pH 7 at salinities below 20‰ (Nauhaus et al., 2005). Indeed, igneous end member has lower pH value than sedimentary end-member (Table 1), and sulfur has a moderate negative correlation with pH, indicating sedimentary end member may have stronger AMO and result in lower dissolved S content. However, dissolved sulfate and sulfide were not differentiated in this study because the sulfate concentration is too low to be detected by ion chromatography after dilution. The fundamental mechanism requires further studies.

Stable Sr isotope reveals negative correlation with radiogenic Sr isotope (Supplementary Table S2; Figure 5). Previous studies indicate that stable Sr isotope in the MVs and other fluids in the hydrosphere may be controlled by the source rock, the degree of water–rock interaction, recrystallization of the carbonates, and coprecipitation with the carbonates (Krabbenhöft et al., 2010; Chao et al., 2013; Voigt et al., 2015; Fruchter et al., 2017; Liu et al., 2017; Voigt et al., 2018; Shao et al., 2021). The coprecipitation is the major factor to fractionate stable Sr isotope in the hydrosphere including seawater, river water, and MV fluids, and raise the δ⁸⁸Sr value of the fluid as Sr is removed from the fluid (Krabbenhöft et al., 2010; Chao et al., 2013; Fruchter et al., 2017; Shao et al., 2021). Sedimentary fluid contains higher total alkalinity and lower Ca, resembling carbonate precipitation occurring. However, δ⁸⁸Sr and Sr concentration of sedimentary fluid denote a reverse relationship with carbonate precipitation, higher Sr and lower δ⁸⁸Sr, indicating coprecipitation is not the major factor. The occurrence of clay adsorption increases δ⁸⁸Sr of the fluid as well (Liu and You, 2020). Sedimentary fluid has stronger adsorption as indicated by B concentration and δ¹¹B (Chao et al., 2011), but lower δ⁸⁸Sr in the sedimentary fluid expresses a conflicting result. The adsorption is not the major factor, either. The recrystallization of the carbonate releases high δ⁸⁸Sr into the fluid (Voigt et al., 2015) but high Sr sedimentary fluid contains low δ⁸⁸Sr, illustrating it is not the major one. Radiogenic Sr isotope indicates the source of Sr. The stable Sr isotope of the arc basement is not documented and assumed to be similar to andesite standard AGV-1, 0.32‰ (Charlier et al., 2012) or arc/back-arc hydrothermal fluids, 0.29‰–0.30‰ (Yoshimura et al., 2020). Sediments generally have δ⁸⁸Sr lower than igneous rock (Halicz et al., 2008; Chao et al., 2015). Both ⁸⁷Sr/⁸⁶Sr and δ⁸⁸Sr of the igneous fluid is higher than that of the andesite basement (Table 1; Figure 4), indicating incomplete water–rock interaction of the ancient seawater (⁸⁷Sr/⁸⁶Sr = 0.708, 925 to 070, 894; δ⁸⁸Sr = 0.340 to 0.375; 5 to 6 Ma; DePaolo, 1986; Hodell et al., 1990; Hodell et al., 1991; Paytan et al., 2021) and the andesite basement. With the evidence of negative correlation between radiogenic and stable Sr isotopes of the fluid as well as the source rock, the major fractionation mechanism of δ⁸⁸Sr of the LGH fluid is the source.

Chloride, δ¹⁸O, and δD show interestingly no discrepancy between two end members. They are relatively conservative parameters and generally are good tracers for water masses. Two end members with distinct Sr isotopes but nearly the same chloride, δ¹⁸O and δD, probably indicate a late stage water–rock interaction before migrating upward. As discussed in The source of low Cl, δD, and δ ¹⁸ O fluid section, the chemical characteristic of chloride and δ¹⁸O in MV LGH was originally ancient seawater. Slightly diluted and lowered δ¹⁸O through interaction with volcanic ashes and/or andesitic basement, and further diluted and raised its δ¹⁸O by clay dehydration. If the clay dehydration occurs after the modification of Sr isotopes, the degree of dehydration or the amount of released fresh water will be different between the two end members because of different lithology, and different lithology implies different quantities of dehydrated minerals. If the degree of clay dehydration or the amount of released fresh water is different between two end members, they should have different chloride concentration and δ¹⁸O. On the other hand, ⁸⁷Sr/⁸⁶Sr is sensitive to water–rock interaction as Sr is easy to be leached out even under low temperature (e.g., Menzies and Seyfried, 1979; James et al., 2003; Ma et al., 2010) but insensitive to smectite to illite transformation because of limited Sr release from the interlayer (e.g., Williams and Hervig, 2005). Fractionation of δ¹⁸O is harder than ⁸⁷Sr/⁸⁶Sr by water–rock interaction because O has a much larger pool size than Sr (54 M versus 0.2 mM). Thus, the alternation of the Sr isotopes has to occur after clay dehydration to keep H and O isotopes but vary Sr isotopes. The similar Cl, δ¹⁸O, and δD indicate two end members suffering similar degree of dehydration in the environment having similar quantity of dehydrated materials. This may imply that the two end members originate from the same source regionally. There are two possibilities to have Sr isotope variation after clay dehydration. One is during fluid migration, the two end members stay in different reservoirs with different lithology and the Sr isotopes are fractionated by the rock they are hosted. The other is after the igneous basement collides into the forearc basin and the fluid interacted with the arc basement has altered its Sr isotopes. The later possibility explains the signal of residual seawater in the sedimentary end member.

In summary, the two end members are likely to be originated at similar depth and formation, undergo similar mechanisms to have their chloride diluted and their water isotopes fractionated. At the final stage before migrating to the surface, their major constitutes and Sr isotopes are further altered by the rock they are hosted. This may occur at the source region or in the reservoir with long residence time.

The ratio of dissolved elements is used to estimate the formation temperature of the fluid such as Na/K, K/Na, K/Mg, Mg/Li, and Na/Li ratios (Giggenbach, 1988; Kharaka and Mariner, 1989; Verma and Santoyo, 1997; Can, 2002). Giggenbach (1988) proposes the “Giggenbach ternary diagram” to estimate the equilibrium condition. If the temperature estimated by the K/Na ratio shows similar result to the K/Mg ratio, the fluid is in equilibrium condition. If two temperature estimations have different results, the fluid is in partial equilibrium or immature condition. However, these geothermometer calculations cannot be applied to the fluids that result from the mixing of multiple sources (e.g., Nicholson, 1993). Since at least two fluid sources can be defined in MV LGH, as discussed above, the temperature calculation is only applied to the end-member composition (Table 1). Equations adopted from Giggenbach (1988) and Can (2002) were used: T   ( ° C )     =   1390 1.75   −   log   K N a   −   273.15 (2) T   ( ° C )     =   4410 14.0   −   log K M g   −   273.15 (3) T   ( ° C )     =   1052 1   +   e ( 1.714 × log ( N a K ) + 0.252 ) + 76 (4) where all element concentrations presented in the above equations are in ppm.

Both end members in LGH show lower K/Mg temperature than K/Na, indicating partial equilibrium of the fluid (Table 1; Figure 9). Igneous end member has higher formation temperature than sedimentary end member by all three equations. This may imply that the source or the long residence time fluid reservoir of sedimentary end member has shallower depth or lower geothermal gradient than that of igneous end member. The later inference is the likely scenario because of large heat flow variation across the Longitude Valley (Chi and Reed, 2008; Wu et al., 2013). Sedimentary end member in LGH show similar Na–K-based temperature with MV LS-A1. LS-A1 has a reverse relationship between K/Na and K/Mg temperature. Due to shorter equilibrium time of Mg, K/Mg temperature is generally lower than that of K/Na. Therefore, Na–K-based geothermometers are often used to estimate the source temperature, and K/Mg geothermometer is used to estimate the temperature of the shallower reservoir (Giggenbach, 1988; Verma et al., 2008). However, solute geothermometers have an error range between 20°C and 40°C; thus, the 11°C difference still falls within the error range and can be seen as in an equilibrium condition (Verma et al., 2008). These solute geothermometers of MV LS-A1 show similar results to the temperature estimated by methane clump isotopes (Lin et al., 2019), indicating similar source depth of gas and saline water.

Generally speaking, if a fluid contributed by two reservoirs with distinct chemical compositions lies at depth and its chemical composition changes with time; that means fluids from two reservoirs mix well at certain depth, migrate to surface afterward and the contribution from each reservoir changes with time. Even with some split fluid channels near the surface, the chemical composition of the main crater and the satellites will vary together. However, chemical and isotopic compositions of MV LGH show spatial distribution of two-end-member mixing and the mud pools on the axis denote time variation, indicating the fluids from two sources mix close to the surface. Terrestrial MVs are mostly distributed on the geological fracture structure such as faults and the axis of anticlines (Bonini, 2012; Mazzini and Etiope, 2017). Anticline is a reasonable possibility since two sources may migrate from two flanks of the anticline and converge close to the surface. But the flank of the anticline is actually the compression region and only the axis provide the extension condition, the fluid channels probably will not exist in the flanks of the anticline. Furthermore, whether it is a normal anticline or a fault-related anticline, the axis of the fold should be parallel or subparallel to the strike of the strata nearby. Assuming the distribution trend of the mud pools are able to indicate the direction of the fold axis or fault as indicated in the literature (Mazzini and Etiope, 2017), the trend of mud pool distribution in LGH is not parallel to regional strike of strata (Jiang et al., 2011). This anticline hypothesis is not supported by the field work investigation, either. The result of field work investigation indicates the strike–slip fault is the major geological structure in the study area (Lin et al., 2008) and the folds are in a small scale, not likely to extend to the depth. The scattered distribution of mud pools at the northwestern part of MV LGH may indicate that the fracture structure for the underground fluid pathway may result from the fault. The shape of fault rupture on the surface is not a straight line at the outcrop scale; it ruptures irregularly. Large spatial–temporal variation of chemical composition at the northwestern part of LGH (Figure 3 and Figure 7) indicates complex conduits underground. Geodesic evidence reveals that MV LGH is located on the local extension zone (Ching et al., 2007; Lin et al., 2010) and the fluid channels are under unclamping situation coseismically (Bonini et al., 2016; Bonini, 2021). With the spatial distribution of the mud pools and the extension fractures after 2003 Chengkung earthquake (M_w = 6.8) recorded by Jiang et al. (2011), each group of mud pools seems to result from underground extensional cracks and provides pathways for the fluids. The fluid channels are probably induced by a local normal fault and the orientation of the mud pools may denote the distribution of the fault zone.

MVs in eastern Taiwan are not the typical ones like their counterparts in western Taiwan. They emit gases with low CO₂ (<0.2%), higher N₂ (up to 10%) and He isotopes (up to 1.9 Ra; Yang et al., 2003; Yang et al., 2004; Chao et al., 2010; Sun et al., 2010; Sano et al., 2017), and waters which contain igneous signals illustrated by high Ca/Cl and low ⁸⁷Sr/⁸⁶Sr ratios (You et al., 2004; Chao et al., 2011; Chao et al., 2013; this study). They also lack the geophysical evidence of mud diapir underneath. The volcanic and sedimentary hybrid system named the sediment-hosted geothermal system has its definition on several gas and isotopic compositions (Procesi et al., 2019). Although MVs in eastern Taiwan do not match all the chemical definitions, we suggest they are still able to be classified into this hybrid system on a position very close to the hydrocarbon sedimentary system (typical sedimentary MVs).