Water of Eastern Taiwan mud volcanoes: part II—trace element distribution of Lei-Gong-Huo mud volcano

Mud volcano (MV) is one of the most important channels for deep seated volatile materials to migrate to Earth’s surface in the sedimentary basins at subduction zones. It plays a meaningful role of elemental budget in hydrosphere. Waters of MV fluid emitted from 46 mud pools in MV Lei-Gong-Huo (LGH) in eastern Taiwan were sampled from October 2015 to July 2016 monthly. Previous research indicates that the chemical characteristic of fluid expelled by MV LGH is dominated by two-end-member mixing of water–rock interaction with sedimentary and igneous rocks (Chao et al., Front Earth Sci 9:750436, 2022). Trace elements of 165 mud pool samples, including Al, Co, Ni, Cu, Zn, As, Br, Rb, Mo, Sb, I, Cs, Pb and U, were measured. The results show that waters of MV LGH reveal similar chemical characteristics to low-temperature ridge-flank hydrothermal spring and marine pore water in anoxic conditions coupled with the addition of clay dehydration. Principle component analysis (PCA) with previous published variables (Chao et al., Front Earth Sci 9:750436, 2022) indicates the important factors are the source of water–rock interaction, the surface ion exchange, the redox condition of mud pool, and the evaporation. In summary, the concentration of trace elements in the waters expelled by MV LGH is in the range of nM to pM. Their behaviors are dominated by secondary factors during and after migration such as surface ion exchange in response to changes in pressure and temperature, redox state of the mud pool, and evaporation. Trace elements in MV water are not suitable for fluid source or provenance studies.


Introduction
The mud volcano is one of the most effective conduits for the fluids as well as volatile elements to migrate through thick sediments to the surface. It is a diapiric structure resulting from tectonic compaction where soft, thick, and fine-grained mud is rapidly deposited. The generation of hydrocarbon gas weakens the formations and increases the buoyancy of the fluid. Development of the mud diapir or the fault offers the channels for the fluid to migrate upward to the Earth's surface (Milkov 2000;Dimitrov 2002;Kopf 2002;Mazzini and Etiope 2017). The MVs, including terrestrial and marine ones, are distributed on tectonically convergent region such as the Mediterranean Sea, the Alpine-Himalayan orogenic belt, the Caribbean Sea, and the Pacific Ring (Dia et al. 1999;Etiope et al. 2002Etiope et al. , 2011aEtiope et al. , 2011bPlanke et al. 2003;Shakirov et al. 2004;Yang et al. 2004;You et al. 2004;Lavrushin et al. 2005Lavrushin et al. , 2015Mazzini et al. 2007Mazzini et al. , 2009Deville and Guerlais 2009;Farhadian Babadi et al. 2019). There are about 10 times more submarine MVs than terrestrial ones (Milkov 2000;Mazzini and Etiope 2017).
The materials expelled by MVs are composed of gas and mud. The composition of gas is mainly methane (> 90%) Chao Terrestrial, Atmospheric and Oceanic Sciences (2023) 34:1 with nitrogen, argon, carbon dioxide, ethane, and higher hydrocarbon gases (Etiope et al. 2002(Etiope et al. , 2007(Etiope et al. , 2009Shakirov et al. 2004;Yang et al. 2004;Chao et al. 2010;Sun et al. 2010). Globally, more than 76% of MVs emit thermogenic methane and only 4% emit microbial methane. The rest of the MVs emit mixed sourced gases (Etiope et al. 2009). The water separated from the mud originates from marine pore fluids. It may have experienced water-rock interaction and/or have been influenced by clay dehydration. Additionally, it may pass through halide formation and may mix with groundwater, surface runoff, and meteoric water near land surface (Bray and Karig 1985;Dia et al. 1999;Dahlmann and de Lange 2003;You et al. 2004;Mazzini et al. 2009). The chemical and isotopic composition may further vary through degradation, migration, and the conditions of mud pools/ salsa during and after migration (Chao et al. 2011(Chao et al. , 2013 this study). The mud breccias are mainly derived from the ambient sediments surrounding the fluid reservoirs with the addition materials from the fluid channel during migration. They are mostly clay minerals, such as smectite, illite, kaolinite and chlorite as well as other minerals like quartz and calcite (Dia et al. 1999;Kopf and Deyhle 2002;Farhadian Babadi et al. 2019). The three components are not necessarily from the same sources and can be decoupled from each other Mazzini et al. 2018).
The evidence of major elements and 87 Sr/ 86 Sr indicates a two-end-member mixing scenario between igneous and sedimentary rocks dominants the chemical characteristic of waters expelled by MV LGH (Chao et al. 2022). Additionally, the predominantly gas, methane (> 90%), discharged by MV LGH is composed of a mixed thermogenic and microbial sources (Chao et al. 2010;Sun et al. 2010). Both gas and water show multiple origins, probably resulting from complex mélange formation at the source region.
There is limited data on dissolved trace element concentration in MV water (Chao et al. 2011(Chao et al. , 2021Lavrushin et al. 2015). The MV is an important geologic structure for the elements flowing back to the earth's surface (e.g. You et al. 1993You et al. , 2004Dimitrov 2002;Kopf 2002). They are one of the sources for seawater elemental budget. Marine mud volcanoes directly emit water with high concentration of trace elements such as Li, B and I into the ocean (You et al. 1993;Kopf 2002;Scholz et al. 2010). Although it is difficult to separate the trace element contribution in river water from weathering, water discharged by terristrial mud volcanoes does show notable influence on river water chemistry (e.g. Chao et al. 2015).
The two-end-member mixing chemical characteristic of MV LGH provides a good chance to study the distribution of trace elements from two types of waterrock interaction. In this study, water samples from 46 mud pools at MV LGH were collected from October 2015 to July 2016 monthly. Trace elements including Al, Co, Ni, Cu, Zn, As, Br, Rb, Mo, Sb, I, Cs, Pb and U were measured carefully and confidently. The chosen trace elements are well studied in the laboratory and on the field observations (Garlick and Dymond 1970;Seyfried and Bischoff 1979;Lawrence and Gieskes 1981;Wheat et al. 2002;Wheat and Mottl 2000), and thus could be compared to those from this study. The results are going to be combined with previous reported data (Chao et al. 2022), and the systematic measurements will help to understand the behaviors of the trace elements in MV waters from two distinct lithologies (sedimentary and igneous).

Geological background and site description
The Taiwan mountain belt formed as a result of arc-continent collision between the Philippine Sea Plate and the Eurasian Plate and one of the possible suture zones is the Longitudinal Valley, situated between the Central Range and the Coastal Range (Teng 1990). The highly deformed Lichi Mélange formation, which was the forearc basin and composed of chaotic mudstone mixed with fragments of oceanic crust (Chang et al. 2000), may provide the suitable environment and materials for the formation of MVs. All of the eastern Taiwan MVs are distributed along the hanging wall of Lichi Fault and the foothill of the Coastal Range on Lichi Mélange.
Mud Volcano Lei-Gong-Huo is the southmost MV in eastern Taiwan ( Figs. 1 and 2). It has a shield appearance with a mud covered area approximate 150 m long and 50 m wide. Dozens of mud pools from about 10 cm to 2 m in diameter are distributed along the shield axis with the direction of approximately N15ºW (NNW-SSE). Seven groups in MV LGH are classified according to their locality (Fig. 3). Five clusters of mud pools, groups A to E, are distributed from north to south along the axis of the mud shield. Group F is the mud pools on the western flank. Group G is the mud pools in the northwestern corner. Previous study indicates groups D and E, located in the southeast, are dominated by the fluid experienced water-rock interaction with igneous rock (igneous fluid) while groups F and G, situated in the northwest, are dominated by the fluid experienced water-rock interaction with sedimentary rock (sedimentary fluid). The evidence of Sr isotopes also indicates possible residual ancient seawater emitted by the mud pools in group F (Chao et al. 2022).

Trace element composition in the MV water
A total of 46 mud pools in MV LGH were sampled from October 2015 to July 2016 monthly. Detailed sampling was performed on November 2015, May 2016 and July 2016. Only large, active, or long lasting mud pools were sampled on the rest of the visiting dates. Data of on site parameters such as temperature, pH, and oxidation-reduction potential (ORP) as well as major dissolved components, O, H, and triple Sr isotopes have been reported and discussed in the article by Chao et al. (2022).
Trace elements (Al, As, Br, Co, Cs, Cu, I, Mo, Ni, Pb, Rb, Sb, U, and Zn) were measured with Agilent 7500cx quadrupole inductively coupled plasma mass spectrometer (ICP-Q-MS). Due to high Cl, Na, Ca in the samples, matrix-matched calibration standards were prepared, the uncertainty is less than 3% and the quality assurance was obtained through 3 reference materials, SLRS-5, CASS-4, and NASS-5. The results Br, I, and Rb fall in the range of 10% certified values and the results of other elements fall in the range of 20% certified values except Co, Ni, Cu, Zn and As, which are deviated by isobaric interferences. Although Y and La show good results of reference materials, good correlation between Y and Sr as well as La and Ba indicate strong tailing effect on or hydride interference in both elements. The result of 2 elements is discarded. Numerical normalizations are applied to As, Co, Cu, Ni, and Zn due to low signal to noise ratio in collision mode. 75 As has isobaric interference from 35 Cl 40 Ar + . 59 Co has interference from 43 Ca 16 O + . 63 Cu has strong interference from 23 Na 40 Ar + ; therefore, 65 Cu is used in this study.
Cu has interference from 130 Ba 2+ and 25 Mg 40 Ar + . The separated test of matrix matched sample spiking with Ba and Mg indicates 130 Ba 2+ is the major source of interference and the contribution from MgAr is less than 1%. 60 Ni has interference from 44 Ca 16 O + and 66 Zn has interference from 26 Mg 40 Ar + and 132 Ba 2+ . Similar to 65 Cu, major contribution of interference is double charge of Ba, not MgAr. The 43/59, 43/60, 130/65, and 132/66 ratios of samples spiking with concentration matching Cl, Na, Ca and Ba were monitored before and after each 10 samples run. The intensity and ratios were applied to remove the potion from the interferences. The intensity of m/z 75 was deduced by the intensity of m/z 77 and multiplying by the 35 Cl/ 37 Cl ratio for the purpose of correcting the contribution from 35 Cl 40 Ar + . Before the Cl correction, the intensity of m/z 77 contributed by Se was removed by monitoring m/z 82. After numerical normalization, the results of reference materials fall in 20% range of certified values for the 5 elements.

Statistical analysis
Principle component analysis (PCA) is a technique for analyzing large datasets and reducing the dimensionality of a dataset. It is used to examine the variability of the chemical results in waters of MV LGH. A varimax rotation with normalized data is applied to maximize a variance and to obtain a pattern of loading for each factor.

Results and discussion
The ICP-MS measureable elements sorted in order of the average molar concentration are Br, I, Al, Rb, Ni, Zn, Cu, Co, As, Mo, Cs, Sb, Pb and U (Additional file 1: Table S1).
The results of elements Rb, Br, I and U are comparable to previous report which was collected in the year 2008 (Chao et al. 2011). Br, I, and Al are in the range of μM. Rb is in the range of sub-μM. The rest of the elements are in the range of nM to sub-nM. Br shows conservative behavior with limited variation, less than 5% of relative standard deviation. Cu, As, Rb, Mo, I, Cs, and U show variations smaller than 2 orders of magnitude. Al, Co, Ni, Zn, Sb, and Pb have great variations with more than 2 orders of magnitude.

Characteristic of trace elements
Waters expelled by MVs reveal similar chemical characteristics to deep marine pore fluids. They originate from ancient seawater and have been altered by early diagenesis, clay dehydration, and water-rock interaction (e. g. Dia et al. 1999;Dählmann and de Lange 2003;Hensen et al. 2004;You et al. 2004;Mazzini et al. 2007Mazzini et al. , 2009Ray et al. 2013;Farhadian Babadi et al. 2019;Chen et al. 2020). Seawater is an important reference for chemical composition of MV waters. Chloride is one of the most important dissolved ions to decipher the source of the fluid other than seawater and the behavior of other elements because of its conservative characteristic. The average Al, Co, Cu, I and Pb over Cl ratios show at least ten folds higher than those of seawater (Additional file 2: Table S2). The average Ni, Zn, As, Sb and Cs over Cl ratios are slightly higher and the average Br/Cl is only 8% higher than those of seawater. The average Rb/Cl and Mo/Cl ratios are lower than those of seawater, and the average U/Cl ratio has less than 10% of the seawater value. Elements/Cl ratios higher than those of seawater indicate elements are released from the andesitic basement or sediments. Lower ratios indicate mechanisms that remove elements from the seawater such as uptake by alternation of the minerals, co-precipitation with secondary minerals or adsorption because of the variation of temperature, pressure, or redox conditions. Due to high Ca/Cl and low Na/Cl and 87 Sr/ 86 Sr ratios, waters of MV LGH are speculated to derive from waterrock interaction of volcanogenic materials and have similar chemical characteristics to marine pore fluids near oceanic basement or low-temperature ridge-flank hydrothermal springs (Garlick and Dymond 1970;Seyfried and Bischoff 1979;Lawrence and Gieskes 1981;Wheat et al. 2002;Wheat and Mottl 2000). The laboratory simulation experiment and field investigation results show As/ Cl, Rb/Cl, Cs/Cl, and Pb/Cl ratios higher than those of seawater in high temperature hydrothermal experiments (up to 350 ℃; You et al. 1996). Al/Cl ratio is higher than that of seawater in dissolution batch experiments at room temperature (e.g. Dixit et al. 2001). Mn/Cl, Mo/Cl, Co/ Cl, Ni/Cl, and Zn/Cl ratios are higher than those of seawater, but Cu/Cl and U/Cl ratios are lower than those of seawater in low temperature ridge-flank hydrothermal spring waters (up to 64 ℃; Wheat et al. 2002). Br and I are released into pore fluids by the decomposition of organic matter with increasing degrees of diagenesis, raising Br/ Cl and I/Cl ratios with depth (e.g. Gieskes and Mahn 2007).
Most of the element/Cl ratios match the pattern except Cu/Cl, Rb/Cl, and Mo/Cl ratios. Mo and U are redox sensitive elements and can be removed from the water under anoxic conditions easily. The results of on site ORP measurements indicate anoxic conditions in the MV waters (Chao et al. 2022) although Mo concentration shows no correlation with ORP and U has only low positive correlation (Additional file 3: Table S3. Mo removal requires H 2 S (e.g. Brüske et al. 2020), and H 2 S is not documented in the literature (Yang et al. 2004;Chao et al. 2010;Sun et al. 2010), which is a possible explanation of no correlation with the ORP. Low Cu/Cl in the spring water is inferred as being precipitated into sulfide phase (Wheat et al. 2002).

The concentration of S in MV
LGH is low possibly due to anaerobic oxidation of methane (AOM; Chang et al. 2012;Tu et al. 2017). With sufficient supply of methane, AOM is a common mechanism to remove sulfate in the MV water (e.g. Stadnitskaia et al. 2005;Wallmann et al. 2006;Lichtschlag et al. 2010). The absence of S may result in higher Cu content in the MV water. Rb and Cs are mobile elements in the high temperature hydrothermal fluid (Seyfried et al. 1998), and Cs is more mobile than Rb because the release of Cs from sediment is prior to Rb as temperature rises (You et al. 1996). Indeed, the Cs/Cl ratio is higher than Rb/Cl ratio relative to seawater, supporting the results of hydrothermal experiments (You et al. 1996;Seyfried et al. 1998). However, Rb/Cl ratio of MV LGH water is lower than that of seawater, indicating the uptake of Rb from water. Similar results have been reported in MVs from western Taiwan and Azerbaijan (Chao et al. 2011;Lavrushin et al. 2015) and low-temperature submarine springs (Wheat and Mottl 2000;Hulme et al. 2010). This uptake of Rb may couple with K uptake during smectite-illite transformation (Plank and Langmuir 1988;Hüpers and Kopf 2012) or water rock interaction with basalt (70 ℃; Seyfried and Bischoff 1979) at low temperature. Rb and Cs have strong positive correlation (r = 0.85), indicating the Rb uptake is stopped and Rb is released with Cs in the MV water. The evidence of high B and Li concentration as well as elevated δ 18 O in MV LGH waters (Chao et al. 2022) also supports the occurrence of smectite-illite transformation (e.g. Sokol et al. 2019;Baldermann et a., 2020;Chen et al. 2020). In summary, the general pattern of trace element distribution in waters of MV LGH is similar to low temperature submarine springs under anoxic conditions with the occurrence of clay dehydration.

Principle component analysis
Variables, including on site measurements, major elements, and Sr isotopes from Chao et al. (2022), are used with trace elements of this study for principle component analysis. Water isotopes are not used for PCA because not all of the samples have their water isotopes measured, and they are not correlated with any other parameters (Additional file 3: Table S3). There are 6 factors with eigenvalue higher than 1.5 (Table 1), and the accumulated variance of 6 factors is 68%.
The first factor (F1) from the PCA result has the highest variance by 29.71% and high positive loading for Li, Na, Mg, Sr, Ba, and 87 Sr/ 86 Sr and moderate positive loading for TA. B, S, Ca, and δ 88 Sr have high negative loading and Ni has moderate negative loading (Table 1; Fig. 4). 87 Sr/ 86 Sr is a robust tracer in hydrosphere indicating the rock that hosts the fluid reservoir. The high loading of 87 Sr/ 86 Sr in F1 clearly indicates that this factor can be assigned to the source of the fluid. Chao et al. (2022) illustrate that chemical characteristic of two-end-member mixing dominates chemical and Sr isotope distribution in the waters of MV LGH. The two end members are named sedimentary and igneous fluids, which are the fluid experienced water-rock interaction with sedimentary and igneous rocks respectively. The positive loading indicates the variables which are dominated by sedimentary fluid while negative loading indicates the variables which are dominated by igneous fluid. Groups F and G mud pools have high factor scores while D4 and group E mud pools have very negative scores (Additional file 2: Table S2; Figs. 5 and 6). The factor scores imply that sedimentary fluid dominates in mud pools of groups F and G in the northwest while igneous fluid dominates in mud Page 7 of 13 Chao Terrestrial, Atmospheric and Oceanic Sciences (2023) 34:1 pools of D4 and group E in the southeast (Fig. 6). The two-end-member characteristic are under stable condition through the whole study period while the mud pools located on the axis of mud shield (e.g. groups A, B and C) show relatively large temporal variation (Fig. 6). The temporal variation on the axis may imply the variation of end member contribution but the mechanism is not clear since the temporal resolution of the sampling is too low to detect the event.
Factor 2 (F2) has the variance by 11.88% with high positive loading for alkaline metal elements K, Rb and Cs and moderate positive loading for S, Ca, Cu, and Ba. TA and Na show moderate negative loading of F2 (Table 1). TA, B, Na, S, Ca, Cu, Rb, and Ba are both affected by F1 and F2 (Fig. 4). Mud pools with negative factor scores of F2 are distributed off the shield axis (i. e. D4 and F1; Figs. 5 and 7) and have relatively stable scores temporally (Fig. 7). Mud pools on the shield axis Table 1 Factor loadings and eigenvalues after varimax rotated normalization in waters of MV LGH Loadings higher than 0.7 or lower than − 0.7 are in bold Accumulated variance of 6 factors is 68% a Total suspended solids, also called as mud weight in Chao et al. (2022)  Page 8 of 13 Chao Terrestrial, Atmospheric and Oceanic Sciences (2023) 34:1 show positive scores and high score mud pools are distributed in the north with great variation spatially and temporally (Figs. 5 and 7). By comparing mud pools with great score variation (e.g. LGH-A7_2016-6 versus LGH-A7_2016-7, LGH-B2_2015-1 versus LGH-B2_2015-2, LGH-B2_2016-3 versus LGH-B3_2016-7, Page 10 of 13 Chao Terrestrial, Atmospheric and Oceanic Sciences (2023) 34:1 LGH-C1_2015-1 versus LGH-C1_2016-3), the high loading elements K, Rb and Cs show variations up to 26%, 66% and 80% respectively. However, major element Ca and Na, which show medium loadings, do not display co-variation with K because F1 has a greater influence on these elements. Additionally, 87 Sr/ 86 Sr ratios denote no loading on F2; indicating F2 does not represent another source. F2 is probably a secondary mechanism that occurs during fluid migration. Potassium can be affected by ion exchange reaction with the sediments due to the change in pressure and temperature (Wheat and Mottl 2000). This mechanism may occur to Rb and Cs because of similar ion radius. Factor score variation of the mud pools may indicate the variation of the mud flux. Mud pools on the flank show relatively low but stable fluxes and the axis pools display great variations. Factor 3 (F3) has the variance by 10.26% with high positive loading for Si and Fe and moderate positive loading for total suspended solids (TSS) and As. ORP and U have moderate negative loading of F3 (Table 1). The negative correlation between ORP and TSS is interpreted as expelling activity (Chao et al. 2021). If the gas flux is higher than the threshold, the water and mud in the mud pool mix well. This scenario results in higher TSS in the sampling tubes and the ORP of the water will be low due to anaerobic oxidation of methane. The low ORP reduces the solubility of U but raises the solubility of Fe. More suspended particles and intensive blending in the mud pool release Si into aqueous phase. Therefore, F3 explains the activity of gas bubbling in the mud pool.
Factor 4 (F4) shows high positive loading for Mo and moderate positive loading for Sb. Mn has high negative loading of F4 (Table 1). Similar to F3, Mn, Mo and Sb are redox sensitive elements. Mn releases into aqueous phase under reducing condition but Mo is removed from aqueous phase. Although both U and Mo are removed from aqueous phase under anoxic condition, they show distinct differences in their removal pathways (Brüske et al. 2020). The release and uptake of Mn is also different from Fe. Therefore, F4 represents different redox pathways other than F3.
Factor 5 (F5) shows high positive loading for Co and Pb. F5 may represent the mechanism causing Co and Pb to vary. Factor 6 (F6) shows high positive loading for Cl and Br (Table 1). Cl and Br are conservative elements in MV LGH water. F6 implies evaporation and meteoric water dilution. Mud pool G5_16-3 has the lowest Cl and Br concentrations as well as the most negative factor scores (Additional file 2: Table S2).

Comparison with MV LS and MVs in Azerbaijan
Trace element distribution of MV waters obtained by ICP-MS is limited in the literatures. MV LS, which is located in the north, shows lower chloride concentration and slightly higher δ 18 O (Δ 18 O LS-LGH = 1.00 ‰; Chao et al. 2022) than that of MV LGH, possibly resulting from more dehydrated water addition. However, MV LS water shows lower to similar trace element over chloride ratios to MV LGH water (Chao et al. 2021; Additional file 2: Table S2). The proposed differences made by different degree of clay dehydration or sedimentary contribution in LGH water are not clear from the point of view of trace elements. Additionally, possible great variation of trace elements in two-end-member mixing dominated MV LGH waters is not detected since MV LGH does not display significant greater standard deviation than that of MV LS. As discussed in Sect. 3.2, trace element concentrations lower than μM scale are easily altered by secondary mechanisms during fluid migration or the condition of the mud pool. They are not suitable for fluid source tracing or provenance studies.
The average values of trace element over chloride ratios in Azerbaijan MV waters show higher Al/Cl, Zn/Cl/ Br/ Cl, Mo/Cl, Sb/Cl, Pb/Cl, and U/Cl ratios, similar Cu/Cl, and lower Rb/Cl and Cs/Cl ratios relative to LGH waters (Lavrushin et al. 2015;Baldermann et al. 2020; Additional file 2: Table S2). Azerbaijan MVs are distributed on the area similar to Taiwan Island. The major source lithology of MV fluid is sedimentary rock. The addition of brines or dissolution of evaporates occurs in some MVs. It is difficult to compare MV LGH water to Azerbaijan MVs directly due to different spatial scale and lithology. However, the similar pattern of Al, Cu, Zn, As, Sb, Cs, and Pb enrichment and depletion of Rb related to seawater may indicate the general pattern of MV fluid worldwide. The discrepancy between Eastern Taiwan and Azerbaijan MVs may be induced by different lithology, degree of clay dehydration, proportion of water-rock interaction, degradation during migration, and condition of mud pool/ salsa. Global database of the chemical composition of MV waters displays great diversity due to the variation of primary and secondary factors involved in the MVs (Nikitenko and Ershov 2021).

Conclusion
A total of 165 water samples from 46 mud pools of MV LGH in eastern Taiwan were collected from October 2015 to July 2016 monthly. Their dissolved trace elements, including Al, Co, Ni, Cu, Zn, As, Br, Rb, Mo, Sb, I, Cs, Pb and U, were analyzed. The results of Br, Rb, I and U concentrations agree with limited previous findings