Role of fluid on seismicity of an intra-plate earthquake zone in Western India: an electrical fingerprint from magnetotelluric study

The magnetotelluric (MT) investigation carried out in Koyna Seismogenic Zone (KSZ), an intra-plate earthquake region in Western India, along an E–W profile brings out moderately conductive (~ 700–1000 Ωm) near vertical features within the very high resistive (> 20,000 Ωm) granite/granite-gneiss basement. Occurrences of these anomalous moderate conductors are corroborated with sensitivity analysis. The alignment of earthquake hypocenters along the resistive–conductive boundary signifies the moderate conductor as basement fault. The conversion of resistivity values to the ratio of seismic P- to S-wave velocity (vp/vs) suggests that the moderate conductivity of the fault zone (as compared to the surrounding basement) appears due to the presence of fluid in the fault zone. Geophysical evidences reveal ~ 2.5–3.6 vol% fluid in the fault zone with ~ 1.8–2.6% interconnected porosity, which migrates along the structural boundary and invades the mechanically strong basement to nucleate the brittle failure within it. The present study proposes two mechanisms for the seismicity in the Koyna region. First: the meteoric water circulation due to the loading–unloading of nearby Koyna reservoir acts as potential fluid source for this triggered seismicity, which has also been suggested by previous studies. Second: the fluid circulation due to a deep-seated source. The present MT study brings out a conductive feature below 20 km depth which is thought to be emerged due to the dehydration of amphibole bearing rocks. The fluid generated from dehydration might act as a probable source to the triggered seismicity; since the conductive feature has a linkage to the upper crust.


Introduction
Potential role of fluid, originated from mantle dehydration reaction (Fulton and Saffer 2009) or partial melt associated with aqueous fluid (Schilling and Partzsch 2001), is observed in earthquake processes (Miller 2013) along plate boundaries and volcanic arcs (Irwin and Barnes 1980) at lower crustal to upper mantle depth.However, in the upper crustal level, fluid associated with artificial reservoir impoundment (Gupta 2002(Gupta , 2011;;Gupta et al. 2017), fluid injection (Chen et al. 2017) are primary causes for triggering earthquakes in an intraplate region.Koyna Seismogenic Zone (KSZ) in Western India is such an intra-plate earthquake zone (Fig. 1).Micro-earthquakes were observed in KSZ in conjunction with the dam construction on Shivajinagar Lake (Koyna reservoir in Fig. 1) in the early 1960s (Pandey and Chadha 2003).However, the impoundment of Koyna reservoir (Fig. 1) in 1962 boosted up the micro-seismicities and KSZ started to experience earthquakes of M ≥ 3, after the water column attained the average height in 1963 (Guha et al. 1968;Pandey and Chadha 2003), and leads to the occurrence of a M ~ 6.3 earthquake (Gupta et al. 1969) in Koyna (Fig. 1).These seismic activities have been continuing for the last five decades.
The occurrence of the devastating December 10, 1967 Koyna earthquake (M 6.3) became special interest for the scientific community as it occurred in a nearly non-seismic (Gupta et al. 1969) intra-plate zone.Therefore, a set of possible concepts for the earthquake genesis came forward, e.g.reservoir triggered seismicity (RTS) by Narain and Gupta (1968a, b), Gupta et al. (1969) and Gupta (1983); rift valley structure by Krishna Brahmam and Negi (1973); strain released from a probable fault along west coast of India by Athavale and Mohan (1976); vertical movement of crustal block by Kailasam et al. (1976) etc.However, based on the correlation between earthquake occurrence frequency and RTS characteristics (such as increment rate of reservoir water level, highest point that the reservoir water level achieves, duration of retention of high water level), Gupta et al. (1969Gupta et al. ( , 1972aGupta et al. ( , b, 2002)); Gupta andRostogi (1974, 1976); Gupta (1983) suggested that the continuous seismic activities in Koyna region are related to RTS.Though these studies have suggested the continuous earthquakes as RTS events, the triggering mechanism is still little understood (Gupta et al. 2015).
Electrical resistivity, a physical parameter, has a vital role in studying such fluid-triggered earthquakes.Fluids (aqueous fluids/partial melts) are associated with very low electrical resistivity and a small amount of fluid within a rock matrix considerably reduces its bulk resistivity.In addition, fluids strongly affect the deformation mechanisms and rheology of the rocks (Burgmann and Dresen 2008).Therefore, estimation of subsurface electrical resistivity will help to determine the zone of saturation and its interconnectivity; constrain the subsurface fluid volume and the rheology (Becken et al. 2011).Therefore, the electrical investigation in KSZ is crucial.However, a very few electrical and electromagnetic (EM) investigations (Kailasam et al. 1976;Patro and Sarma 2009;Sarma et al. 2004) are carried out till date to get electrical signature of the subsurface of KSZ.These limited studies were carried out in large scale, which just touched the Koyna region with few stations and therefore, provide very little electrical information of KSZ.As magnetotelluric (MT), a passive EM geophysical technique, can give electrical image with a wide depth range and proves to be effective in basalt covered areas, we, therefore, have chosen MT method to delineate KSZ in detail.A broad band MT survey is carried out along an E-W profile across the KSZ (profile-K, Fig. 1).The acquired data are inverted two dimensionally (2D) to evaluate the detailed sub-basalt upper crustal geoelectrical structures along the E-W profile.The derived electrical signatures are then analysed for the geophysical evidences of fluid role on triggering earthquakes/brittle failure and related seismicity.

Geologic and tectonic setting
The study area, KSZ is situated on an N-S elevated escarpment called Western Ghats that runs parallel to the west coast of India (Fig. 1).Almost all the area is covered by ~ 67.4 Ma old Deccan basalt (Duncan and Pyle 1988), which comprised of sequential lava flows.Each basaltic lava flow comprises brecciated layer followed by a vesicular and/or amygdaloidal layer and then a fine-grained massive basalt layer (Roy et al. 2013).The Deccan basalts are underlain by Archean crystalline granite/granitegneiss basement (Sarma et al. 2004).The Donachiwada fault (Gupta et al. 1999), associated with the December 10, 1967 Koyna earthquake (M 6.3), is striking approximately in an NNE-SSW direction (black dashed lines, Fig. 1).From the fault plane solution study (Langston and Franco-Spera 1985) Donachiwada fault is found as strike slip in nature and surficial evidences indicated towards a left-lateral movement (Talwani 1997).Drilling in this NNE-SSW trending Donachiwada fault suggested a dip amount of 60° WNW (Gupta et al. 1999).The wellbore breakouts and drilling-induced tensile fractures (DITFs) observed in a 3 km deep borehole (KFD1, Goswami et al. 2020) in KSZ (Fig. 1) constrain N9°W ± 17° mean orientation for the maximum horizontal principal stress, S Hmax (Goswami et al. 2020) which is subparallel/oblique to  (Gupta et al. 2015) which approximately covers the region between the Koyna reservoir in the north and the Warna reservoir in the south (red coloured dashed blok).The white circles represent the MT stations (profile-K, near M 6.3 December 10, 1967 earthquake).The black circles represent the positions of exploratory boreholes drilled to instal borehole seismometer (Gupta et al. 2017).The red circle represents the position of the 3-km-deep pilot borehole (Goswami et al. 2020) the strike of Donachiwada fault.The direction of S Hmax is consistent with the NW oriented crustal stress field in the Southern India as determined from rock mechanics studies (Gowd et al. 1992).

Magnetotelluric data and analysis
MT data are acquired in a rolling array pattern with constantly maintained remote site for most of the survey duration.ADU07e instrument is deployed to acquire the orthogonal electric and magnetic field for the period range 0.001-100 s.The MT data at few locations were affected due to electrical noise as this region is traversed by many high-tension power lines and contaminated with other cultural noise.We, therefore, adopt the systematic processing work flow of Borah et al. (2015) to handle the noise and improve the S/N ratio of the datasets (see Fig. 2 for example of data quality).The consistency of the processed resistivity and the phase data is examined with D + approach of Beamish and Travassos (1992) assigning 10% and 5% error to apparent resistivity and phase data, respectively (Fig. 2).

Dimensionality and regional strike analysis
In MT method, the observed data are being checked for dimensionality before adopting a particular modelling approach to convert the observed MT impedances to a true resistivity subsurface section (Simpson and Bahr 2005).The rotationally invariant dimensionality parameter Swift Skew, κ (Swift 1967) is especially sensitive to the static shift effect (occur due to local, near surface inhomogeneities), hence, leads to the incorrect interpretation of one dimensional (1D) or 2D structure as three dimensional (3D) structure (Jones 2012).Therefore, we have tested the dimensionality of our data set with Bahr Phase Sensitive Skew, η (Bahr 1988); a dimensionality measurement based on the phases of the impedance tensor and thus, unaffected by static shift effect.According to Bahr (1988Bahr ( , 1991)), η > 0.3 represents a 3D structure.Figure 3a describes dimensionality of the data in terms of η at each period of a station along the profile.It is observed that, for most of the sites, the values of η are within the threshold limit of 0.3 up to 10 s, suggesting 1D/2D nature of the subsurface beneath the profile.The scattered high (> 0.3) η values observed in the short period (0.001-0.01 s) are the indication of noise contamination in the impedance data.Further, we also estimate the phase tensors (PT) ellipses (Caldwell et al. 2004) filled with β-values (skew angle) for each period of all MT sites along the profile.It is observed that for most of the sites, the PT plots for period < 10 s are circular to elliptical in shape and associated with β-values that are close to zero, which indicates a 2D subsurface up to 10 s.PT ellipses plot at three periods for all the MT stations are presented in Fig. 3b.Both η and PT analysis mostly suggest that the subsurface along the profile is 2D up to 10 s.
Accumulation of electric charge at the interface of near surface inhomogeneities produces secondary electric field and that distorts the measured MT responses; therefore, such MT responses are unable to represent the regional conductivity structure precisely.This problem of distortion can be minimized by applying tensor decomposition technique (Becken and Burkhardt 2004;Groom and Bailey 1989;McNeice and Jones 2001;Smith 1995Smith , 1997) ) which separates local scale galvanic effect from the MT responses and in the 2D situation, gives regional geoelectrical strike direction.Therefore, in our study, we have adopted the single-site multi-frequency approach (McNeice and Jones 2001) which is based on the decomposition method of Groom and Bailey (1989) for regional geoelectrical strike estimation.It may be seen that the difference (RMS error) between the 2D decomposed responses (McNeice and Jones 2001) and the observed Fig. 2 Example of data in terms of apparent resistivity and phase with increased S/N ratio after systematic processing (Borah et al. 2015) show a good consistency through D + approach (solid blue and red lines) of Beamish and Travassos (1992).Remaining data of all the sites are given in Additional file 1: Fig. S1  (Caldwell et al. 2004), along with the inherent 90° ambiguity in strike estimation for the period band 1-10 s.The red and grey wedges of rose diagram represent the two possible mutually perpendicular geoelectrical strike direction data throughout the period ranging from 1 to 10 s for almost all the sites is low (≤ 2), which strongly validates the 2D assumption (Fig. 3c).Although galvanic distortion distorts the observed electric field amplitude, the phase between electric and magnetic (horizontal) field is virtually unaffected (Caldwell et al. 2004).We, therefore, also adopted the PT approach (Caldwell et al. 2004) to estimate the strike direction.The main advantage of PT analysis is that, it does not require the assumption about the dimensionalities of the subsurface conductivity structures like the other techniques (Caldwell et al. 2004).
Dimensionality analysis (both η and PT) of MT data and RMS misfit between decomposed responses and the observed data (Fig. 3) suggests a 2D subsurface up to period 10 s, therefore, to estimate the regional strike we consider the period range 1-10 s (mostly upper crustal depths).For our data set a predominant geoelectrical strike in NE-SW direction (~ N33°E) is obtained from both McNeice and Jones (2001) and PT analysis (Caldwell et al. 2004) for the period range 1-10 s (Fig. 4).The Donachiwada fault (Gupta et al. 1999;Shashidhar et al. 2013) follows the same trend as calculated strike (see Fig. 1) and hence we finalize the strike N33°E for the profile.After getting the proper strike direction, we rotate the measured impedance tensors of all MT sites to align the XY components of impedance tensor (Z XY ) along the strike.This aligned XY-component after rotation is assigned to E polarization (TE mode) and the rotated YX component is assigned to B polarization (TM mode).

Two-dimensional (2D) inversion
In the present study, we use the 2D inversion scheme Non-Linear Conjugate Gradient (NLCG) algorithm of Rodi and Mackie (2001) to bring out the 2D upper-crustal geoelectrical structures along profile-K.A homogeneous half space of 100 Ωm background resistivity having exact topography variation along the profile is taken as the initial model to start the iterative NLCG algorithm.Arabian Sea is situated 25-30 km away from the western most MT sites (Fig. 1).Hence, to tackle the effect of the highly conductive water mass (ocean) on the measured impedance tensor, we fix the ocean with its exact bathymetry and position in the initial 100 Ωm background grid and assigned a resistivity value of 0.33 Ωm to the fixed ocean.
The non-uniqueness of MT inversion (Constable et al. 1987;Jones and Hutton 1979;Oldenburg et al. 1984;Parker and Whaler 1981) can be constrained with regularization, which search for a model that has minimum misfit between the observed and predicted data and also tries to keep the model smooth as possible (Tikhonov and Arsenin 1977).In NLCG algorithm, the regularization parameter (τ) influences both the data fit and the model roughness.If τ is small, data fit will be good but the model will not be a smooth one and vice versa.Therefore, to get a reliable τ value which can take care of both the data fit and model roughness, we applied different τ values and carried out the inversion keeping the other parameters constant.The data fit (RMS error) is then plotted against the model roughness for each τ value (Fig. 5).According to Hansen (1998), this plotting will give an L-shaped trade-off curve and the reliable τ value which has a good agreement with data fit and model roughness will take its position at the knee of the curve.From the L-curves plotted for profile-K (Fig. 5), τ = 7 is considered to be the reliable value.
Both TE and TM mode data for the period range 0.001-10 s are inverted jointly applying NLCG algorithm (Rodi and Mackie 2001), taking regularization parameter,  τ = 7 and assigning 20% error floor to apparent resistivity and 5% error floor to phase and get the smooth model after 100 iterations.The static shift coefficients (the real scaling factors by which the apparent resistivity curves are shifted relative to each other due to near surface inhomogeneities) at each station along the profile are computed during inversion and are corrected from the responses of this model by carrying out the inversion for further 30 more iterations.The final inverted 2D model obtained after 130 iterations yielded an RMS error of 2.53.Before evaluating this derived 2D model, we have computed the depth of investigation (DOI) of each station (Fig. 6) using the average phase of TE and TM mode data associated with 10 s period using the formula derived by Borah and Patro 2019.It is seen that all the stations have more than 40 km of DOI (Fig. 6) and therefore, we consider our final inverted 2D model along profile-K up to 25 km depth (Fig. 7; vertical exaggeration, VE = 2.5) for further analysis and interpretation.
The limit up to which the measured MT data are reconstructed by the model responses of the inversion output is expressed numerically in terms of RMS misfit.The overall RMS misfits at each station along the profile are presented in Fig. 8a (top) along with the static shift coefficients Fig. 8a (bottom).It may be seen that there is good agreement between static shift and RMS misfit; high RMS values coincide with high static shift values (Fig. 8a).To have a vivid view of the data misfit, we have plotted the measured data and computed responses at each station along the profile as pseudo-section (Fig. 8b).The observed and computed responses fit well for both TE and TM data (Fig. 8b).

Results
It is observed that the 2D inversion of MT data along profile-K (Fig. 7) brings out the relatively conductive (≈ 30 Ωm) Deccan basalt cover on the top and very high resistive (≥ 20,000 Ωm) Archean granite/granite-gneiss basement (Sarma et al. 2004) beneath.The thickness of the Deccan basalt delineated from the 2D inversion is consistent with the scientific drilling result (Gupta et al. 2017) (Fig. 7).The very high resistive basement is cut by anomalous near vertical, moderately conductive (≈ 700-1000 Ωm) features (C 1 and C 2 in Fig. 7) which separates the basement into different blocks.
In order to test the reliability of C 1 , and C 2 (Fig. 7), we have carried out sensitivity analysis.First, the resistivity values of C 1 , and C 2 obtained in the static shift corrected final inversion model (Fig. 7) are changed to the resistivity of the adjacent part.For this set up, the inversion is run for further 100 more iterations to examine whether the signatures of C 1 , and C 2 are retrieved or not.After completion of 100 iterations, it is seen that the expressions of C 1 , and C 2 reappear (Fig. 9a).Secondly, forward responses are generated on the final static shift corrected inverted model with the resistivity of C 1 and C 2 changed to the resistivity of the adjacent part.These forward responses are then compared with the model responses of the inversion output along with the data.The comparison study of the responses at those sites (K18, K03) which fall near the moderately conductive zones shows that the data fit degrades if resistivities of C 1 , and C 2 are changed to high resistivity of the adjacent part (Fig. 9b).Therefore, the sensitivity analysis suggests that the near vertical moderately conductive features C 1 , and C 2 appear within the highly resistive basement are not artefacts of 2D inversion; rather required by the data.

Discussion and conclusion
In a subsurface resistivity section, a fault zone, in general, exhibits itself as a conductive feature compared to the surrounding region; and the reason for this conductive signature can be attributed to several aspects.Since, our study area is seismically active and the fault zone has a tremendous role on earthquake genesis, therefore, the moderately conductive near vertical features across the basement (C 1 , C 2 in Fig. 7) might be considered as basement fault zones.For better precision, we have projected the earthquake hypocenters (M > 1), that are falling within a 500 m radius zone centred by the profile, on the 2D resistivity section (Fig. 10a).We have also projected the Donachiwada fault (fault configurations are taken from Gupta et al. 1999), as the profile runs across the fault (Fig. 10a).The correlation of Donachiwada fault with C 1 (Fig. 10a), as well as the alignment of hypocenters along the resistive-conductive boundary (Fig. 10a), clearly indicates that the moderately conductive feature C 1 is Donachiwada fault.Besides, the association of the M 6.3 December 10, 1967 earthquake (Chandra 1976) with the resistive-conductive boundary (Fig. 10a) strongly supports our observation.For the moderately conductive signature C 2 , we have a few earthquake hypocenters aligned along the resistive-conductive boundary below a depth of 10 km (Fig. 10a).This observation signifies that the signature C 2 might represent a less active (as compared to the Donachiwada fault) basement fault in Koyna region.
The most favourable explanation for the conductive features obtained in the inversion output of a seismically active zone is the presence of fluid in the fault zone (Mackie et al. 1997;Ogawa and Honkura 1997;Unsworth et al. 1997Unsworth et al. , 1999)), since a little amount of fluid associated with a rock matrix is capable of reducing its bulk resistivity considerably (Gueguen and Palciauskas 1994).However, the consideration of fluid in the fault zone based on the conductivity contrast is still ambiguous because, similar kind of scenario is observed in case of mineralization by chemical precipitation, i.e. ore coated fracture (Becken et al. 2011) and clayey fault gauge (Palacky 1987;Eberhart-Phillips et al. 1995).Therefore, to elucidate the proper cause for the observed conductivity of C 1 and C 2 we have considered the corresponding ratio of seismic Pto S-wave velocity (v p /v s ratio).We convert the resistivity values of each grid point of the resistivity section to its Several studies have established that there is a reduction in seismic wave velocities (both v p and v s ) within a fault zone because of the reduction in shear modulus of the faulted rocks (Nur and Simmons 1969;Dixit et al. 2014).However, when the faulted rocks are saturated with fluid, v p increases significantly and v s remains the same or slightly decreases because of the presence of fluid (Nur and Simmons 1969;O'Connell and Budiansky 1974;Dixit et al. 2014).Such situation results in increase of v p /v s in the fluid-saturated fault zone.Therefore, for the shearing of dry rock, one would expect the hypocenters to coincide with regions of low-v p and low-v s and the association of hypocenters with high v p /v s (high v p and low v s ) likely signifies the water saturation and high pore pressure within the fault zones (Dixit et al. 2014).
From the v p /v s section (Fig. 10b) it is observed that the fault zones C 1 , and C 2 (or the earthquake hypocenters) are associated with the regions having high v p /v s values (i.e.high v p and low v s ) w.r.t the surroundings.Therefore, this observation directly points that the conductivity contrast observed in the fault zones (C 1 , C 2 ) is because of the presence of the fluid in the fault zone and rules out the effects of other factors.On the other hand, the evidence of hydrous alteration along the fault zone (observed from drilling result, Gupta et al. 2017) signifies the water channelization through the fault (Misra et al. 2017) and therefore, supports the presence of fluid in the fault zone.
The resistivity of aqueous fluids depends on temperature and salinity (Becken et al. 2011).In KSZ, the  (Gupta et al. 1999) with C 1 ; the association of the December 10, 1967 (M 6.3) earthquake which occurred near to profile-K and the alignment of projected earthquake hypocenters (black circles), falling within 500 m radius zone centred by the profile with the resistive-conductive boundaries reveal the presence of basement fault.b Conversion of resistivity to v p /v s ratio using log-linear relationship of Borah and Patro (2022).Association of the earthquake hypocenters with the high v p /v s zone than the surroundings signifies the presence of fluid Fig. 11 Plotting of Hashin-Shtrikman upper bound (Hashin and Shtrikman 1962) for two-phase system composed of rock matrix (≥ 20,000 Ωm) and fluid (~ 18 Ωm) provides ~ 2.5-3.6 vol% fluid for ~ 700-1000 Ωm observed resistivity average estimated temperature for the depth range 6-10 km is 140-195 °C (Gupta et al 2015).The pressure exerted by the granitic basement (of average density ~ 2.65 g cm −3 Gupta et al. 2017) at this depth range is 1558-2597 bars.Therefore, for this temperature and pressure range the estimated average density of pure water (Burnham et al. 1969) at 6-10 km depth range is ~ 0.988 g cm −3 .The hydrochemical study on the well water of Koyna region (Reddy et al. 2017) suggests a very low salt (Cl − ) content of maximum 30 mg L −1 (or 0.003 wt%).Therefore, for the temperature range 140-195 °C, the average resistivity of aqueous fluid ( ρ f ) of density 0.988 g cm −3 containing 0.003 wt% salt (Cl − ) will be ~ 18 Ωm (Sinmyo and Keppler 2017).For this two-phase system of very highly resistive rock matrix (≥ 20,000 Ωm) and a low resistive fluid phase (~ 18 Ωm), the Hashin-Shtrikman upper bound (Hashin and Shtrikman 1962)  The minimum interconnected porosity (Φ) filled with fluid of resistivity ρ f (~ 18 Ωm) that is required to explain the observed bulk resistivity ( ρ b ) of ~ 700-1000 Ωm solely due to the fluid phase (i.e. S w = 1, S w is the pore space filled by water with a resistivity ρ f ) can be estimated using Archie's law, ρ b = ρ f � −m ; where m is the cementa- tion exponent of the rock and usually ranges between 1 to 2. The lower bound of m (m = 1) is more capable of furnishing reasonable and desirable porosity estimations for fault zones (Unsworth et al. 1997).Therefore, considering that only the fluid conductivity (18 Ωm) contributes to the bulk rock conductivity (700-1000 Ωm), we estimate the corresponding interconnected porosity and found that ~ 1.8-2.6%interconnected porosity is responsible for the observed bulk resistivity of ~ 700-1000 Ωm of C 1 and C 2 .Several previous studies have pointed out the correlation of the observed seismicity with the loading and unloading of nearby Koyna reservoir.Therefore, the meteoric water circulation through existing faults in the basement, due to the loading of the reservoir, may consider as the potential fluid source for the fault zone (Fig. 12).However, the present MT investigation has sensed a conductive feature of ~ 40 Ωm below a depth of 20 km (Fig. 12).The existence of this anomalous conductive feature is affirmed by the sensitivity analysis (Fig. 9a).This conductive zone has a linkage to the upper crust through C 2 (Fig. 12).This conductive zone hints towards the probable presence of a fluid source; since aqueous fluids can be found at mid to lower crustal depths due to the dehydration of amphibole bearing rocks at ≥ 600 °C ( Vanyan and Gliko 1999).Amphibole, one of the important hydrous mineral in the continental middle crust (Wang et al. 2012) is abundant (35-40%) in the depth range of 15-30 km (Christensen and Mooney 1995).The temperature-depth profile of KSZ (Gupta et al. 2015) suggests a temperature of ~ 600 °C at 25-30 km depth (considering ~ 50 mWm −2 surface heat flux and ~ 0.5 μWm −3 heat production for the granite-gneiss) which is favourable for the dehydration.Therefore, there is a possibility for dehydrated fluid source below 20 km depth and upward fluid movement from this source due to buoyancy (Becken and Ritter 2012), through the conductive pathway (C 2 ), to provide a linkage between the fluid source and upper crust (Fig. 12).In this perspective, C 2 might be considered as fluid migration path (from lower crustal depth to upper crustal depth) rather than a basement fault.
It is interesting to observe that the hypocenters are aligned along the resistive-conductive block boundaries and associated with the resistive basement blocks (Figs. 10 and 12).It is observed that the fault zone fluid at hydrostatic pressure suppresses seismicity and over-pressurised fluid tends to induce seismicity (Becken and Ritter 2012).This observation signifies that the hydraulically interconnected fault zone fluid at hydrostatic pressure makes the fault zone conductor too weak to accumulate the necessary shear stress for brittle failure (Ritter et al. 2005;Unsworth et al. 2000).Generally, in case of fluidinduced seismicity, the high pressurized fluid (sourced in the conductive zones due to the loading of reservoir and deeper fluid source) invades the critically stressed surrounding resistive formations, lowers the shear stress required for failure, and thereby facilitates the mechanical failure within the rheologically strong regions nearby (Jiracek et al. 2007;Sleep and Blanpied 1992).
The observed alignment of the earthquake hypocenters within the resistive basement block of the resistiveconductive block boundaries (Fig. 12) can therefore be explained by the invasion of fluid (Gurer and Bayrak 2007) from conductive mechanically weak permeable zones (C 1 , C 2 ) to resistive (less permeable) mechanically strong zone along the structural boundary (Fig. 12).This migration induces high fluid pressure that leads to brittle rock failure (Becken et al. 2011) within the high resistivity basement block near the resistive-conductive boundaries (Fig. 12).The entire process depends on the stress concentration and the fluid penetration depth into the resistive rocks (Becken et al. 2011).

Fig. 1
Fig.1Location map showing the geology (input taken from Geological Survey of India) of KSZ.The seismic activity of KSZ is restricted to ~ 30 km × 20 km area(Gupta et al. 2015) which approximately covers the region between the Koyna reservoir in the north and the Warna reservoir in the south (red coloured dashed blok).The white circles represent the MT stations (profile-K, near M 6.3 December 10, 1967 earthquake).The black circles represent the positions of exploratory boreholes drilled to instal borehole seismometer(Gupta et al. 2017).The red circle represents the position of the 3-km-deep pilot borehole(Goswami et al. 2020)

Fig. 3
Fig. 3 Dimensionality analysis for the acquired data in KSZ. a Bahr Phase Sensitive Skew (Bahr 1988) plot at each period of each site along the profile.An overall low skew value (< 0.1) is observed at the shorter periods with few outliers.Bahr Skews have intermediate values (0.1-0.3) up to 10 s represent a 2D earth up to 10 s. b Phase tensor, PT (Caldwell et al. 2004) filled with β-value plots at three periods (0.1 s, 1 s and 10 s) for all the sites along the profile.Up to period 10 s PTs are circular to elliptical in shape with β-values close to zero which represents a 2D subsurface up to 10 s. c RMS misfit plot between 2D decomposed responses (McNeice and Jones 2001) and observed data.An overall low RMS value (≤ 2) is observed for period range 1-10 s of all the sites validating 2D subsurface for this period range

Fig. 5
Fig. 5 L-curve obtained for profile-K with the systematic variation of regularization parameter (τ).Here τ = 7 is fixed as reliable regularization parameter for the profile

Fig. 6
Fig. 6 DOI (Borah and Patro 2019) estimated from the average phase of TE and TM mode data of period 10 s for each station along the profile

Fig. 9 a
Fig. 9 a Comparison of the 2D inversion output along profile-K with the output from sensitivity analysis.Both the outputs represent similar signatures which prove the existence of the near vertical moderately conductive features (C 1 , and C 2 ).b Comparison between model responses and the forward responses (obtained without C 1 , and C 2 , observed in 2D inversion output of profile-K) with data for the sites near to C 1 , and C 2 (K18, K03).Data fit degrades when C 1 , and C 2 are absent which proves the existence of C 1 , and C 2

Fig. 10 a
Fig. 10 a Correlation of the projected Donachiwada fault(Gupta et al. 1999) with C 1 ; the association of the December 10, 1967 (M 6.3) earthquake which occurred near to profile-K and the alignment of projected earthquake hypocenters (black circles), falling within 500 m radius zone centred by the profile with the resistive-conductive boundaries reveal the presence of basement fault.b Conversion of resistivity to v p /v s ratio using log-linear relationship ofBorah and Patro (2022).Association of the earthquake hypocenters with the high v p /v s zone than the surroundings signifies the presence of fluid suggests ~ 2.5-3.6 vol% of interconnected fluid (Fig. 11) to explain the observed bulk resistivity ( ρ b ) of ~ 700-1000 Ωm for C 1 , and C 2 .

Fig. 12
Fig. 12 2D inversion output from joint inversion of TE and TM data along profile-K using NLCG algorithm of Rodi and Mackie (2001) up to 25 km depth brings out an anomalous conductive feature of ~ 40 Ωm below 20 km depth.This conductive feature might signify the presence of dehydrated fluid which has a probable role on the observed seismicity in KSZ along with the meteoric water circulation due to the nearby Koyna reservoir