Opportunistic magnetotelluric transects from CSEM surveys in the Barents Sea

Magnetotelluric (MT) data allow for electrical resistivity probing of the Earth’s subsurface. Integration of resistivity models in passive margin studies could help disambiguate non-unique interpretations of crustal composition derived from seismic and potential ﬁeld data, a recurrent issue in the distal domain. In this contribution, we present the ﬁrst marine MT data in the Barents Sea, derived from industrial controlled-source electromagnetic (CSEM) surveys. We characterize data quality, dimensionality, depth penetration and elaborate an analysis strategy. The extensive MT database consists of 337 receivers located along seven regional transects, emanating from ∼ 70 000 km 2 of 3-D CSEM surveys acquired for hydrocarbon exploration from 2007 to 2019. High-quality MT data are extracted for periods ranging from 0.5 to 5000 s. The data show no apparent contamination by the active source nor effects related to large time-gaps in data collection and variable solar activity. Along receiver proﬁles, abrupt lateral variations of apparent resistivity and phase trends coincide with major structural boundaries and underline the geological information contained in the data. Dimensionality analysis reveals a dichotomy between the western domain of the SW Barents Sea, dominated by a single N–S electromagnetic strike and the eastern domain, with a two-fold, period-dependent strike. 35 receivers show 3-D distortion caused by nearby bathymetric slopes, evidenced by elevated skew values. We delineate geographical areas where the 2-D assumption is tenable and lay the foundation for future MT modelling strategies in the SW Barents Sea. We performed 2-D MT inversion along one of the regional transects, a ∼ 220-km-long, E–W proﬁle encompassing a major structural high and sedimentary basin approaching the continent-ocean transition. The resistivity model reveals low crustal resistivity values (1–10 (cid:2) .m) beneath the deep sedimentary basins, in marked contrast with high resistivity values (1000–5000 (cid:2) .m) of the thick crystalline crust on the structural high. We interpret this abrupt lateral resistivity variation as a rapid transition from a thick, dry continental crust to a hyperextended and hydrated crustal domain. Integration of resistivity with seismic velocity, density and magnetic susceptibility models may further reﬁne these structural models and the underlying tectonic processes in the SW Barents Sea margin. Our methodology is applicable globally where 3-D CSEM surveys are acquired and has a large potential for harvesting new knowledge on the electrical resistivity properties of the lithosphere.


I N T RO D U C T I O N
Over 30 yr of geophysical investigations have deciphered the architecture and composition of the lithosphere in the Barents Sea (Faleide et al. 1991;Breivik et al. 1998;Mjelde et al. 2002;Clark et al. 2013;Gernigon et al. 2014;Klitzke et al. 2015;Indrevaer et al. 2018). Reflection and refraction seismic, potential field and borehole data have harvested considerable knowledge on palaeoprocesses at play in the Wilson Cycle in the North Atlantic and Arctic realm (Wilson 1966;Faleide et al. 2018;Chenin et al. 2019;Gernigon et al. 2019;Lundin & Doré 2019). However, sparse data coverage, imaging challenges and overlapping geophysical properties, often lead to non-unique interpretation of deep lithologies and geometries; a recurrent problem in distal passive margin studies (Peron-Pinvidic et al. 2016;Peron-Pinvidic & Manatschal 2019).
In this context, we introduce for the first time the use of the magnetotelluric (MT) method (Tikhonov 1950;Cagniard 1953) and subsequent probing of the Earth's electrical resistivity in the Barents Sea. The scarcity of MT studies in passive margin settings testifies to the costly acquisition during days, and sometimes weeks of recordings in remote offshore areas (Heinson et al. 2005;Myer et al. 2013;Hoversten et al. 2015;Jegen et al. 2016;Corseri et al. 2017). We overcome this challenge by taking an opportunistic approach, extracting MT data from an existing marine controlledsource electromagnetic (CSEM) database acquired by the industry in the Barents Sea from 2007 to 2019. The intensification of hydrocarbon exploration activities over the past decade led to the acquisition of large CSEM (Eidesmo et al. 2002) surveys by electromagnetic geoservices (EMGS), covering ∼70 000 km 2 of the seafloor of the Barents Sea (Fig. 1). 3-D CSEM surveys (Fig. S1) are primarily designed for hydrocarbon exploration and prospect de-risking (Fanavoll et al. 2014;Baltar & Barker 2015;Berre et al. 2020), contributing to large hydrocarbon discoveries such as Wisting, the northernmost and shallowest oil discovery to-date on the Norwegian continental shelf (Granli et al. 2017). Marine CSEM is the active-source pendant of offshore MT; when the active source is far enough from the receiver or not emitting, MT signal can be extracted from receivers deployed on the seabed (Fig. S1).
The overarching objective of this study is to establish the useability and usefulness of this atypical, 'CSEM-derived' MT data in the context of the SW Barents Sea. The approach should also be applicable worldwide, like in the Gulf of Mexico and offshore Brazil, where extensive CSEM surveying occurred. An efficient integration of the MT method and its derivative products in geological interpretation of the Barents Sea requires a comprehensive assessment of the strengths and weaknesses of this unconventional geophysical data set. In this contribution, we purposely focus on the geophysical results and down-play the geological interpretation. The geological significance of the electrical resistivity models from MT data will naturally follow this work while integrating seismic, potential field and borehole data for more completeness (Corseri et al. 2017). In this paper, we address four specific topics: (1) MT data processing, quality and effective removal of the active source signature, (2) dimensionality analysis and identification of MT responses deviating from the 2-D assumption, (3) depth penetration assessment and (4) elaboration of an effective MT modelling scheme. We illustrate each observation and conclusion with carefully selected data examples, representative of this large regional database of 337 MT receivers. The inversion strategy and interpretative power of a resistivity model derived from 2-D inversion are exemplified by a ∼220-km-long transect (EW3 in Fig. 1).

MT experiment and receiver selection
MT is a passive, remote-sensing geophysical method developed in the 1950s by Cagniard (1953) and Tikhonov (1950). The method utilizes the naturally occurring electromagnetic (EM) field as a passive source to retrieve electrical resistivity properties from shallow depths down to the lower mantle. Electrical conduction in the Earth's lithosphere is mainly driven by the presence of (1) free fluid in the rock matrix through ionic conduction, (2) partial melt by semi-conduction and (3) connected mineral matrix like graphite, massive sulfides and other ore bodies through electronic conduction (Simpson & Bahr 2005;Selway 2012;Karato & Wang 2013).
From 2007 to 2019, EMGS has acquired ∼70 000 km 2 of multiclient CSEM data in the Barents Sea. We have selected 337 receivers of EMGS' database located along seven regional transects (Fig. 1). The selected regional transects fulfill three requirements, as they: (1) cover major geological elements of the SW Barents Sea from deep basins, major structural highs to platforms areas ( Fig. 1), (2) ensure as much as possible the continuity of receiver spacing along each transect, crossing most contiguous 3-D CSEM surveys ( Fig. 1) and (3) are in the vicinity of ocean bottom seismometer (OBS) profiles and deep seismic reflection lines. Once the transect orientation and extent were decided, the closest receivers were chosen. We selected receivers so the spacing is roughly uniform, ranging from 3 to 9 km in most cases. This spatial sampling is adequate to target regional crustal architecture and composition.
Following these criteria, we constituted a database of 337 receivers along four East-West (EW1, EW2, EW3 and EW4) and three north-south (NS1, NS2 and NS3) striking regional transects. This database formed the basis of the investigation presented in this study (Fig. 1).

Manual masking of time-series affected by the active source and other type of noise
In typical 3-D CSEM surveys (Fig. S1), receivers are deployed on the seabed to record the electromagnetic response of an active towed source (MacGregor & Tomlinson 2014). Depending on the spatial extent of the survey, receivers are laid on the seafloor from 5 d up to 3 weeks.
The first step of the MT processing was to discard the time window when the active source was actively emitting in the vicinity of the receiver. The start and end of active source times were defined by visual inspection of the amplitude of the source signal strongest harmonic, reaching values above the ambient electromagnetic field. These active source times were manually marked on spectrograms of each of the 337 receivers (Fig. 2a), effectively masking out hours of recording time. This pre-processing step is critical to ensure that the active source signal frequencies, typically ranging 0.1-10 Hz in the Barents Sea, do not contaminate the naturally occurring EM recordings that we exploit for MTs. In addition, we manually masked other 'bad' times due to external factors like bottom currents, swell and internal noise linked to circuitry and hardware failure. We used EMGS' software SBLwiz to perform the masking step on spectrograms for each receiver.  (Jakobsson et al. 2020) and MT receiver locations. 337 MT receivers (coloured circles) are located along seven regional transects. Four regional transects are oriented in an E-W direction (EW1, EW2, EW3-EW4), three in a N-S direction (NS1, NS2 and NS3). Data from the two receivers highlighted by white circles along EW4 were chosen to illustrate the MT processing workflow shown in Fig. 2  Methodology for extracting MT data and comparison between two neighbouring receivers acquired with a 7-month time gap during two 3-D CSEM surveys. (a) Spectrograms of the recorded magnetic field (channel Hx) of receivers 01Rx054a and 01Rx077a. The colour scale is the same for both spectrograms with high magnetic field amplitude in red and, low amplitude in blue. Receiver locations are indicated by white circles along transect EW4 in Fig. 1. Time ranges highlighted by dark bands are masked-out from further processing as they contain an active source signal signature. The lower panels display apparent resistivity and phase responses for each receiver rotated north, showing high-quality, smoothly varying MT data for periods of 1-2000 s and the striking consistency and similarity throughout acquisition years 2007-2008. (b) Flowchart of the step-by-step processing procedures applied to the MT data sets from active source time masking, receiver grouping for multistation robust processing, data analysis to modelling strategies. Egbert (1997) Robust multistation MT data processing (Egbert 1997) uses a statistical approach to determine coherent signal and noise characteristics of an array of receivers. The processing scheme is well-suited to the large receiver grids with km-spacing such as the 3-D CSEM surveys in the Barents Sea (Eidesmo et al. 2002;MacGregor & Tomlinson 2014). We grouped receivers according to their recording time overlap and respective distance (Fig. 2). In theory, the higher the overlap in recording times and the further the distance between receivers in each group, the better the outcome of the processing. Our receiver database originates from different concatenated 3-D survey vintages ( Fig. 1), which naturally limits the possibility for receiver grouping. In turn, comparing the processed MT data from contiguous receivers acquired in different years (or campaigns) would yield a robust validation of the processing outcome and underline the quality of the MT data set (Fig. 2a). The data processing sequence ( Fig. 2b) was performed with SBLwiz where the receiver grouping routine is automated to efficiently tackle large 3-D receiver grids (Markhus et al. 2015). We intended to extract useable MT signal with period ranging from 0.6 to 5000 s.

Robust multistation MT data processing following
The error bars calculation of the impedance tensor rest on the assumption of Gaussian error distribution (Egbert 1997). The absolute orientation of the receivers on the seabed can be estimated from CSEM data using various optimization scheme as well as magnetic components (Mittet et al. 2007;Key & Lockwood 2010). In this work, we used a representative average of 10-15 orientation estimates per receiver provided by EMGS. The outcome of data processing is the MT impedance tensor Z (Sims et al. 1971) that relates the horizontal components of the electric and magnetic fields (E x , E y , H x and H y ) at the seafloor (Simpson & Bahr 2005). The MT impedance tensor along receiver transects can be graphically represented by its apparent resistivity ρ a and phase curves (Simpson & Bahr 2005). Along receiver transects, these two quantities were plotted as pseudo-sections (Fig. 3). We also used these two graphical representations to quality control the outcome of MT processing using two main criteria: (1) smoothness of the apparent resistivity and phase curves against periods ( Fig. 2a) and (2) their consistency along the receiver lines and across survey vintage (Fig. 3).

MT responses from CSEM surveys in the Barents Sea
The processing of 337 EM time-series from multivintage 3-D CSEM data in the SW Barents Sea yielded high-quality MT data, paving the way for meaningful subsurface resistivity value assessment in the study area. A majority of the receivers gave smooth-varying, apparent resistivity, and phase values with low statistical errors, for periods ranging from 0.8 to 1500 s (Figs 2 and 3). In some instances, such as the two westernmost receivers of EW4 near the Senja Fracture Zone (Fig. 1), useable MT signal was retrieved at 5000 s. We observe that short period data <1s are noisy (Fig. 3), a natural phenomenon in marine environment where the conductive seawater act as low-pass filter on MT signal. The masking of CSEM source times in the pre-processing proved efficient and removed active source contamination from the MT data. If contaminated, spikes would be observed in the MT responses at active source harmonic frequencies, such as 0.2-0.4-0.6 Hz (Fig. 2). About 10 receivers had very noisy MT responses or could not be processed due to noisy or faulty channels; they were subsequently removed from the database.
We found that neighbouring MT receiver responses were remarkably consistent across contiguous 3-D survey. As an illustration, we did not expect any major resistivity variations within the 5 km distance separating receivers 01Rx054a and 01Rx077a of EW4 located in the same basin but acquired with a one-year gap (Figs 1 and 2a). This result and the overall outcome of MT responses emphasize the continuous quality of EM instruments and acquisition campaigns in the SW Barents Sea over the past decade (2007-2019; Fig. 1). This observation can be extended to the entire database and validates the experiment conducted in this study and increases confidence in the geological information contained in the MT data.
The quality of the MT data set can be also inspected in pseudosections along transects NS1 and EW3 (Fig. 3). The apparent resistivity and phase responses show generally smoothly varying curves with periods and along the profiles, from receiver to receiver despite acquisition campaigns spanning 2008-2016. On both Z xy and Z yx components, we note an abrupt apparent resistivity trend change along EW3 at the transition from the Loppa High (increasing apparent resistivity with periods) and Bjørnøyrenna Fault Complex to the Veslemøy High (decreasing apparent resistivity with periods). These trends translate into lower phase values <45 • on the Loppa High and >45 • in the westernmost basins along EW3. Along NS1, we observe a similar behaviour of MT responses at the transition Stappen High-Bjørnøya Basin and Bjørnøya Basin-Bjørnøyrenna Fault Complex (Fig. 1). The inspection of MT data along pseudo-sections confirms that MT is sensitive to major structural boundaries in the SW Barents Sea.
The map in Fig. 4(a) depicts apparent resistivity ρ yx values at T = 523 s for all processed impedance tensors rotated to north and corroborates the latter statement about MT sensitivity to major tectonic boundaries. The transition from the structural highs to the distal basins is a major conductivity boundary, a two-order of magnitude resistivity jump striking roughly N-S. Going eastward towards the Norvarg Dome, another important W-E transition is seen, from a highly resistive N-S oriented 'belt' (>10 2 .m) to more intermediate resistivity values (10 1 -10 2 .m) in the eastern Norwegian Barents shelf.

Dimensionality and strike analysis
We performed the data analysis and its visualization with the MTPy software package, an open-source python library for MT data analysis, developed by Geoscience Australia (Krieger & Peacock 2014;Kirkby et al. 2019).
The electromagnetic, or geoelectric, strike is the preferential direction of the current flow in the subsurface (Simpson & Bahr 2005). As the database covers a wide part the SW Barents Sea encompassing various structural elements, we decisively gathered receivers by relevant geological areas (Fig. 5a) and conducted the strike analysis within them. The 2-D assumption holds when a constant strike angle can be confidently identified for a wide range of frequencies.
There is an inherent 90 • ambiguity in the determination of the strike angle. A common approach for lifting the ambiguity is to liken electromagnetic and geological strike (Heinson et al. 2005;Robertson et al. 2015).
The results of the strike analysis show an emphatic dichotomy between an eastern and western domain of the SW Barents Sea (Fig. 5a). The western domain is characterized by strong tendency toward a single, dominant electromagnetic strike in the range of −20 • N to 20 • N at all periods whereas the eastern domain shows a period-dependent, two-fold, strike angle between the period ranges 10 1 -10 2 s and 10 2 -10 3 s (Figs 5a and S2). The strike angle estimates for each geological area covered by our MT data set in the western domain is summarized in Table 1.
In Table 1, we purposely lift the inherent 90 • ambiguity in the strike angle estimate and use two arguments to justify the choice. First, the crustal architecture follows the N-S orientation of the western Barents Shelf, structurally inherited from the orientation of the Caledonian Orogeny (Ritzmann & Faleide 2007). We consider unlikely that telluric currents would massively flow E-W, across such a long-lived tectonic boundary in the North Atlantic. Secondly, the apparent resistivity map (Fig. 4a) exhibits a clear N-S trend of regional conductivity structures.
We use the MT phase tensor and its graphical representation as an ellipse to carry out the dimensionality analysis of our MT data (Caldwell et al. 2004). In a 1-D earth, the MT phase tensor plots as a symmetric ellipse, hence a circle. In a 2-D earth, the regional strike of the 2-D conductivity structure aligns with the major axis of the ellipse. Ellipticity and skew angle β are coordinate-invariant of the phase tensor and measure of the dimensionality in the data (Caldwell et al. 2004;Bibby et al. 2005). In a 2-D case, the skew angle equals zero. A skew angle β > 3 • or ←3 • indicates 3-D-distortion in the data.
From the analysis of the phase tensors and skew values along each transect, we extract the following patterns: (1) The phase tensors analysis shows high ellipticities and skew values for 30 receivers of NS1, NS2 and EW1 (Fig. 6) located near the seafloor slopes of Bjørnøya (Fig. 1). The three southernmost receivers of NS1 and NS2, located ∼50 km off the coast of mainland Norway, shows a similar distortion. This 3-D distortion is likely an offline effect of the nearby steep bathymetric/topographic highs (Fig. 1). The 2-D assumption is not tenable in these areas and requires a 3-D modelling approach (Fig. 5b). (2) In the eastern domain, the MT signal shows a marked two-fold, period-dependent strike angle and azimuth of the phase tensor ellipse major axis (Figs 5a and S2). Although skew angles and ellipticities mostly fall under the 1-D-2-D criteria, this period-dependency of the strike angle is a violation of the 2-D criteria and therefore should be tackled in 3-D.

Maximum depth penetration
To evaluate the penetration depth at receiver location, we used the Niblett-Bostick transformations (Jones 1983) and computed the maximum signal penetration depth at the longest period per receiver. The 'Niblett-Bostic' depth is equivalent to the skin depth corrected by a factor of 1/ √ 2. The contoured maximal depth estimate map (Fig. 4b) shows variation from >100 km signal penetration for receivers located above the Loppa High to <20 km signal penetration beneath the two westernmost receivers of EW3. The diffusive EM fields attenuate quicker in deep conductive sedimentary basins than in resistive crystalline basement shallowing at structural highs. To compensate for the quick signal dissipation, we processed useable MT signal at longer periods in the distal basins (Fig. 1). As an example, the two westernmost receivers of EW4 located south of the Sørvestsnaget Basin, closest to the continent-ocean transition, have useable MT signal down to 5000 s, allowing for diffusion down to 30-50 km. Most of the receivers within the westernmost basins along EW2, Figure 4. (a) Apparent resistivity maps ρ yx at T = 523 s. All MT responses are rotated to geographic north. Note the large, extrapolated areas between receiver transects. (b) Niblett-Bostick signal depth penetration at longest periods. Contours are hand-drawn based on the data analysis results and known geological trends (dashed lines). MT receivers are depicted by black triangles. Note that some receivers are missing because they are either too noisy or could not be processed due to faulty or missing channels. The dashed lines separate the structural elements of the SW Barents Sea (Fig. 1) (Fig. 4b), which permits crustal and uppermost lithospheric mantle resistivity probing.

MT modelling strategy in the SW Barents Sea
Based on previous observations on dimensionality and strike directions, we summarize the MT modelling strategy to adopt in different geological domains of the SW Barents Sea (Fig. 5b). A 3-D approach is required in the eastern domain and areas of the western domain within a ∼120 km radius of Bjørnøya and ∼80 km from mainland Norway. The rest of MT responses fulfill the 2-D assumption and 2-D modelling is preferred along EW transects (Fig. 5b).

2-D MT inversion along EW3
We inverted the MT data using MARE2DEM, an open-source, freely available, 2-D parallel adaptive finite element code that uses an unstructured triangular mesh (Key 2016). In the Occam inversion scheme, we seek the smoothest model that fits the data (deGroot-Hedlin & Constable 1990). Therefore, MARE2DEM inversion incorporates a two-step process: (1) we minimize the data misfit term and let the inversion converge to an acceptable data fit and (2) we then minimize the roughness term to find the smoothest model that fits the data within an acceptable misfit. The 2-D modelling inversion procedure is applied to EW3, discarding receivers from the eastern domain above Norvarg Dome (Fig. 5) based on the strike analysis (Section 3.2). In the 2-D assumption, the horizontal electric and magnetic fields are orthogonal, and the impedance tensor can be decoupled into two independent modes. The transverse electric   (Simpson & Bahr 2005). The inversion tests were run with starting model consisting of half-spaces with resistivity ranging from 1 to 500 .m. However, the final inverted models were hardly start-model dependent as they converged to similar RMS misfit ranging from 1.5 to 2 with little variability in the main anomalies. A target RMS of 1.8 was chosen to balance model roughness and data misfit. TE and TM modes were jointly inverted. Apparent resistivity data were contaminated with a 7 per cent error floor and 2 • for phase data. This choice stems from MT data collected recently in the region, offshore mid-Norway and Svalbard and assigned, respectively 5 and 10 per cent error floor for MARE2DEM (Myer et al. 2013;Selway et al. 2020). Bathymetry is included in the 2-D inversion model. However, bathymetry profiles are quasi-flat and smoothly varying, typically 250-450 m below sea surface along the selected transects (Fig. 1). All inversions were unconstrained, populating the ∼50 000 triangular mesh cells with isotropic resistivity values from 10 -1 to 10 4 .m. Isotropic resistivity parametrization proved to be a sufficient assumption to reach a good fit to the observed MT data along EW3.
All MT responses were rotated to geoelectric strike (0 • E, Table 1), with TE and TM modes subsequently assigned. The 2-D inversion process converged within 20 iterations (∼10 hr for ∼46 000 triangular cells) to a RMS misfit of 1.89. Higher residuals are observed for the longest periods but overall, the data misfit is randomly distributed across the data space (Figs 7 and S3). Therefore, we expect minimal inversion artefacts in the preferred 2-D resistivity model (Fig. 8) along EW3. We identify three robust resistive anomalies R1 to R3 and two conductors C1 and C2. The resistivity model along EW3 is split into a resistive eastern half, from ∼100 to 220 km distance along the line (the Loppa High) and a conductive western half from 0 to ∼100 km, into the deeper Cretaceous-Palaeocene basins (Fig. 1).
In the conductive half, from 0 to ∼115 km along the line, C1 characterizes a wide conductive zone spanning 1-10 .m from seafloor down to 30 km depth. Within C1, we note a decreasing resistivity trend with depth, in some instances below 1 .m. An isolated intermediate resistive anomaly R1 (∼50 .m) is observed east of the Veslemøy High, spanning depth 10-30 km (Fig. 8).
In the resistive half, from ∼115 to 210 km along the line, R2 is a 40-100 km thick and highly resistive anomaly (>10 3 .m). R2 is as shallow as ∼2 km and reaches 100 km depth, East of the Loppa High. R2 exhibits a decreasing E-W thickness trend, from 100 km East of Loppa High to ∼40 km thick at the transition to the Bjørnøyrenna Fault Complex where the anomaly abruptly ends. R3 is embedded into R2 between 160 and 180 km along the line. R3 is a ∼40 km thick, ∼20 km wide vertical contact with intermediate resistivity values 50-100 .m. C2 is the deepest conductor (1-10 .m), underlying anomaly R2. Although R3 appeared to be a robust feature in the inversion tests, we cannot rule out that it is caused by a singularity in the MT response of one receiver located above. In the receiver gap above the Loppa High, from ∼125 to 150 km, R2 is less well constrained but high resistivities are required by MT data on each side of the gap (Fig. S3). In addition, we can expect continuity of R2 beneath the Loppa High as a homogeneous block of thick continental crust.

Regional MT transects from multivintage CSEM receiver grids
A decade of CSEM data acquisition for hydrocarbon exploration in the Barents Sea led to the acquisition of several years' worth of time recordings of the natural electromagnetic field. We gained access to EMGS' time-series and proved that high-quality MT signal could be carefully extracted from surveys designed for CSEM data acquisition. Despite evolving instrumentations, survey geometries and vessels through the last decade (2007-2019), robust multistation processing (Egbert 1997) of our database produced consistent highquality MT responses regardless of acquisition year (Figs 3 and 4a). This study firmly establishes the scientific value of these MT data from multivintage, 3-D CSEM surveys in the SW Barents Sea.
In our database, the active source signature does not contaminate the MT transfer function after manually masking the 'bad' times ( Fig. 3). Manual masking is a tedious pre-processing step, but we found it necessary to ensure the removal of recordings affected by other type of noise, like resonating effect of the antenna under excitation of bottom currents or internal circuitry and hard disk failures. An automated masking of the active source times could be implemented using the source navigation information by imposing a user-defined time threshold before and after the nearest source passage. However, source navigation information and specifications are often considered confidential information by operators and may be reluctant to share them.
The selected data set consists of seven transects widely distributed across the Norwegian Barents Sea (Fig. 1). The data coverage allows for a quantitative evaluation of the dimensionality and distortion effect in MT data across the region. We found that a large part of the MT data in the SW Barents Sea, from the Sørvestsnaget Basin to the East of Loppa High, fulfill the 2-D assumptions for MT data (Fig. 5). Inversion and robustness model tests can be achieved efficiently in 2-D at a low-computational cost. 3-D distortion effects are identified close to bathymetric highs (Bjørnøya) and coastal areas   (Fig. 4b). C1, C2 are robust conductive anomalies of the inverted model. R1, R2 and R3 are robust resistors required by the observed data. The top crystalline crust, Moho and the lithosphere-asthenosphere boundary (LAB) horizons courtesy of P. Klitzke (Klitzke et al. 2015). SvB, Sørvestsnaget Basin; BFC, Bjørnøyrenna Fault Complex. of mainland Norway (Fig. 6). In the SW Barents Sea, the eastern domain (Fig. 5a) exhibits period-dependent geoelectric strikes, violating the 2-D criterion (Fig. S2 in supplementary information). We speculate that the superposition of different tectonic textures and/or crustal domain can be the cause for it (Shulgin et al. 2020).
As a general result from our methodology applied on MT data in the SW Barents Sea, we recommend for any 2-D lithospheric-scale experiment to (1) use MT responses with periods down to ∼2000 s for deep penetration, (2) avoid the vicinity of bathymetric highs (Bjørnøya and coastal area of northern Norway) due to 3-D effects and (3) prioritize E-W oriented transects in the western domain (Fig. 5) for 2-D inversion and robust resistivity model interpretation (Fig. 8).
However, the MT data quality would benefit from longer recording times in the westernmost, distal margin (Bjørnøya and Sørvestsnaget basins and Veslemøy High in Fig. 1) where the decay of the EM signal is quicker due to a thicker, conductive sedimentary cover (Fig. 4). Therefore, a dedicated MT survey in the SW Barents margin ought to be designed for longer receiver deployments, perhaps over 7 d, in the distal parts of the margin. For an equivalent period range, Myer et al. (2013)

Resolving power of the MT data in the Barents Sea
Compacted Palaeozoic (meta-)sediments are exposed at shallow depth in the platforms and structural highs of the SW Barents Sea (Fig. 1). There, the crystalline crust reaches intracratonic thickness of >40 km (Klitzke et al. 2015). Both compacted (meta-)sediments and dry crystalline crust are highly resistive. In such domains, we showed that the extracted MT signal could penetrate down to 100-200 km (Figs 4b and 8). MT imaging is achieved on a complete lithospheric scale.
At the period of ∼1000 s, the MT signal loses ∼50 km of depth penetration at the transition from structural highs to deep basins (Fig. 4b). Up to 18-km-thick Cenozoic to Palaeozoic sedimentary strata infill the deep Bjørnøya Basin, where potential fields and seismic reflection data reveal a hyper-extended crust (Gernigon et al. 2014). Along NS1 and EW2, the maximal signal penetration varies between 20 and 50 km within the Bjørnøya Basin, enabling full crustal investigation and uppermost mantle resistivity probing.
In the Sørvestsnaget Basin ( Fig. 1), the conductive sedimentary cover thickens to West (Kristensen et al. 2018) and consequently deteriorates the depth penetration of the EM signal (Fig. 4b). There, imaging down to the base of the crystalline crust in the westernmost area requires high-quality MT data for periods >1500 s. This was barely achieved in the two westernmost receivers of EW3 (Figs 1  and 4b). We managed to extract MT signal at 5000 s in the two westernmost receivers of EW4 (Figs 1 and 4b) allowing for full crustal reconstruction in the most distal part of the margin, near the continent-ocean transition (Fig. 1). (Fig. 8) resistivity anomalies along EW3 were tested for robustness by replacing the resistive features by conductive ones (and vice versa) in the final model, locking these anomalous regions and letting the inversion converge again. If the final misfit is negatively affected by the locked region, we then conclude that the anomaly is a requirement of the data (Beka et al. 2017;Wannamaker et al. 2017). After experimenting different start model resistivity and confirming the robustness of prominent anomalies, we consider that the 2-D resistivity model along EW3 can be safely interpreted.

Interpretation of 2-D resistivity model along EW3
To guide our 2-D resistivity model interpretation along EW3, we use the Top Crystalline Crust, Moho and lithosphere-asthenosphere boundary (LAB) Barents-Kara Sea grids (Fig. 8) from Klitzke et al. (2015). The structural grids are based on publicly available geophysical data and regional tomographic models with low spatial resolution (∼5-50 km horizontally) and large interpolation.
From seafloor to Top Crystalline Crust, the sedimentary cover shows a gradual resistivity increase (∼1-10 .m) consistent with depth compaction trends and subsequent removal of pore fluids.
The C1, R1, R2 and R3 anomalies are located within the continental crust as defined by the interval between top crystalline Crust and Moho interfaces. C2 is the only anomaly identified within the upper mantle solely.
R2 and R3 evidently correlate with the thickest part of the crystalline crust at the Loppa High. The loss of resolution with depth may explain why R2 and R3 are leaking deeper into the mantle. An alternative explanation could be that the thermal state of the lithospheric mantle evolves laterally from cold and resistive (R2 to R3 anomalies) in the proximal domain to hot and conductive (C2 anomaly, 1-10 .m) in the distal domain. This hypothesis is corroborated by an increase of the calculated heat flow at top basement level, from ∼35 to 80 mW m -2, at the BFC-Loppa High transition (Klitzke et al. 2016). Indeed, increasing mantle temperature translates into lower viscosity and higher electrical conductivity (Pommier et al. 2013). At this stage, it is important to stress that the Moho is an acoustic velocity discontinuity that does not necessarily translate into a discontinuity of electrical resistivity properties. A 1:1 correlation between seismically defined Moho and electrical resistivity changes should therefore not necessarily be sought.
Intracrustal resistivity variations are expected in the SW Barents Sea as the onshore continuation of the crust exhibits structurally inherited thrusts and suture zones (Ritzmann & Faleide 2007;Gernigon et al. 2014;Shulgin et al. 2020). Given its unusual narrow shape, R3 should be interpreted with caution as it could be caused by a singularity in one MT response but we speculate that R3 originates from a vertical contact inherited from the Laurentia-Baltica plate suture. Integration of resistivity model with other geophysical parameters like density and magnetic susceptibility is needed to support this interpretation.
R1 is a crustal anomaly with intermediate resistivity values (∼50 .m). R1 is seemingly leaking down to the uppermost mantle at 30-40 km depth (Fig. 8) but we attribute this to decreasing sensitivity with depth (Fig. 4b) and smoothing effect. West of the Loppa High, the crust is surprisingly conductive, 0.5-10 .m (anomaly C1) and in sharp contrast with the thick, highly resistive crust at the Loppa High. Resistivity values below 1 .m at crustal depth can be explained by the presence of conductive fluids, partial melts, or interconnected mineralization in metamorphic aureole (Myer et al. 2013;Pommier et al. 2013;Corseri et al. 2017). The two latter hypotheses can be discarded as there is no evidence of anomalously high crustal heat flow (for partial melt) nor intrusive bodies around the Veslemøy High (Gernigon et al. 2014;Klitzke et al. 2016). Our favoured interpretation for C1 is that the resistive crust is either absent or too thin to be detected by MT and is associated with the presence of a conductive fluid phase such as sea water. However, this hypothesis contradicts the interpreted crustal thickness by Klitzke et al. (2015) west of the Loppa High, commonly ∼15-20 km thick with a minimum of ∼10 km east of the Veslemøy High. It is very unlikely that a ∼10 km thick, 'R2-type' resistive crust would be undetected by MT. Our interpretation of C1 requires a much thinner crust and is corroborated by thinning factor ranging 3-5 calculated by Clark et al. (2014), Gac et al. (2018) and Gernigon et al. (2014) in the rifted basins of the SW Barents Sea and that the structural grids of Klitzke et al. (2015) have a low (∼5-50 km) spatial resolution (Fig. 8).
The downward migration of conductive fluids through extensional faults to the upper mantle is a documented phenomenon in rifted margins (Bayrakci et al. 2016) and we note that at least five extension episodes are reported since Carboniferous times in the SW Barents Sea (Tsikalas et al. 2021). Evidence for intense hydration of the extended continental crust, upper mantle and its effect on electrical resistivity should be sought to further develop and test this interpretation.

C O N C L U S I O N S
The present work establishes the usefulness of MT data for crustal and lithospheric studies in the SW Barents Sea. The particularity of the MT database lies in the opportunistic utilization of a regional CSEM database, acquired for hydrocarbon exploration purposes over several campaigns from 2007 to 2019. Even as a by-product of the CSEM surveys, the MT data processed for 337 receivers along seven regional profiles, are high quality for periods ranging 0.5-5000 s. This is the first time that such a methodology is applied on a regional scale. We characterize the dimensionality of the database through EM strike angle and phase tensor analysis to show that the SW Barents Sea can be divided into a western, N-S dominated strike domain and an eastern domain with a two-fold, period dependent strike angles. After a systematic data analysis, we bring forward a 'best-practice' approach to MT data modelling in the SW Barents Sea to extract robust resistivity models of the lithosphere. The interpretative power of the method is illustrated by a resistivity model from 2-D inversion along a regional, 220 km-long transect extending from the Sørvestsnaget Basin to the Loppa High. The final 2-D-inverted model shows an abrupt thinning of highly resistive continental crust combined with the presence of conductive fluids at crustal depth, west of the Loppa High-Bjørnøyrenna Fault Complex transition. The MT database in the SW Barents can be used to constrain plate tectonic (transform margins and Caledonian orogeny) and geodynamic processes (regional uplift, mantle viscosity and glacial isostasy adjustment). The potential for new knowledge is considerable and we believe that this work lays the foundations for an acceptance of MT as a complementary data for crustal and upper mantle investigations in the Barents Sea and Arctic basins. The methodology presented in this contribution can also be applied globally where regional 3-D CSEM surveying exists, like in the Gulf of Mexico and offshore Brazil.

A C K N O W L E D G E M E N T S
This work is funded by the Research Council of Norway through the industry PhD (project 298994) and SkatteFUNN (project 308897) schemes. JIF and SP also acknowledge support from the Research Council of Norway through its Centers of Excellence funding scheme, project number 223272 (CEED) and 228107 (ARCEx). The authors would like to thank EMGS for allowing to use and publish magnetotelluric data extracted from their multiclient database. RC is particularly grateful to F. Roth, M. Hansen, P. Gabrielsen, J.P.
Morten, O.M. Pedersen, E. Bjørdal for enabling this project. L.L. Uri and many EMGS colleagues are acknowledged for their implication in the development of EM technology and data acquisition in the Barents Sea. K. Key kindly provided us with the latest version of MARE2DEM. RC appreciated the technical support from the Geo-IT team at the Department of Geosciences (UiO). P. Klitzke kindly supplied his Barents-Kara regional grids. Finally, RC wishes to thank his VBPR and UiO colleagues for stimulating discussions. The authors acknowledge constructive comments from two anonymous reviewers that helped improving the initial version of the manuscript.

DATA AVA I L A B I L I T Y
The data underlying this paper were provided by EMGS ASA under license/by permission. Data will be shared upon reasonable request to the corresponding author subject to permission of EMGS ASA.

S U P P O RT I N G I N F O R M AT I O N
Supplementary data are available at GJ I online. Figure S1. 3-D visualization of a typical marine electromagnetic (EM) survey. The magnetotelluric method utilizes the natural, farfield incident EM field, assumed to hit the Earth's surface as a plane wave ('northern light-like' green coloured area in the back). For controlled-source electromagnetic data, a deep-towed antenna is emitting a dipole-like EM field (white areas) above the seafloor receiver lines. Illustration modified from EMGS' image library, available on the company's website. The illustration is not-to-scale. Figure S2. Phase tensor ellipses pseudo-section colour-coded by skew angle values along transect EW4, for receivers on the Tiddlybanken Basin (Fig. 1). The two-fold, period-dependent strike angle characterizing the 'eastern' domain of the Barents Sea (Fig. 5a) is well expressed in the orientation change of the major axis of the phase tensor ellipse. Figure S3. Observed (left-hand panel) versus 2-D predicted (righthand panel) MT data plotted as pseudo-sections along EW3. 2-D resistivity model is shown in Fig. 8. For each pseudo-section, the horizontal axis represents receiver number along EW3. MT data are rotated to electromagnetic strike (Table 1) with subsequent TE and TM mode assignment. The RMS misfit is 1.89. SvB:,Sørvestsnaget Basin; VH, Veslemøy High; BFC, Bjørnøyrenna Fault Complex; LH, Loppa High.
Please note: Oxford University Press is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the paper.