Magnetotelluric data sense the electrical resistivity of the Earth, a physical parameter particularly sensitive to the presence of low resistivity phases such as aqueous fluids, partial melts or metallic compounds. Fluid phases have electrical resistivities orders of magnitude lower than the rock matrix, and thus relatively small amount of fluids, when interconnected, can decrease bulk rock resistivity by several orders of magnitude 9 .Fluids additionally have a significant weakening effect on the rheology of rocks, even more so if fluids form an interconnected network 10 . Measurements of electrical resistivity can therefore be used to constrain the volume of subsurface fluids, their interconnectivity, and the rheology of the crust and mantle.We collected magnetotelluric data along seven profiles across the SAF near Parkfield and Cholame, covering the tremor concentration zone near Cholame and the transition from locked to creeping behaviour (Fig. 1a).The most prominent structure revealed by the MT data is a deep low-resistivity zone (1-5 ohm-m) centred 30-40 km southwest of the SAF below 15-20 km depth (Fig. 1b). This anomaly occupies a broad region in the lower crust and upper mantle between the surface traces of the Rinconada fault, a former strand of the SAF, and the modern-day SAF. Along profiles 4-7, which cross the SAF at the tremor concentration zone near the northern end of the locked segment 5,7 , crustal resistivities are in excess of 500 ohm-m for the profiles 5-7 across the tremor concentration zone near Cholame, the inversion reveals a deep low resistivity region which is bounded from above and laterally by resistive formations (Figs. 1b, 2b). Here, the tremor source region appears to coincide with the high resistivity rocks, adjacent to the less resistive and potential fluid source area. Inversions of the MT data confirm the resistive cap as a very robust feature of the models.Any low resistivity connection into the upper crust (as observed northeast of Parkfield) results in a significant increase of data misfit (see supplementary information).Tremor near Cholame separates spatially into a southwestern zone with periodic tremor episodes and a northeastern zone with aperiodic episodes 7 . Our resistivity models suggest that the fluid source (i.e. low resistivity) is located to the southwest of the periodic tremor zone. Ongoing fluid generation by mantle dehydration reactions 4 could result in high fluid pressures in the low resistivity zone, and consequently, in fluids continuously driven through fracture systems towards and into the more resistive tremor regions.Observations of high seismic reflectivity 21 (cf. Fig. 2b) at tremor source depths between the low resistivity fluid source and the more resistive tremor zone are consistent with deformed and fractured material.Lateral migration of fluids could be responsible for elevated fluid pressures in the tremor source region Tremor amplitudes along the central SAF vary by a factor of seven 6 and correlate with variations of resistivity by a factor ...

[1] High-resolutionseismictomography and magnetotelluric (MT) soundings of the shallow crust show strong changes in material properties across the Dead Sea Transform Fault (DST) in the Arava valley in Jordan. 2D inversion results of the MT data indicate that the DST is associated with a strong lateral conductivity contrast of a highly conductive layer at a depth of approximately 1.5 km cut-off at a position coinciding with the surface trace of the DST. At the same location, we observe a sharp increase of P wave velocities from <4 km/s west of the fault to >5 km/s to the east. The high velocities in the east probably reflect Precambrian rocks while the high electrical conductivity west of the DST is attributed to saline fluids within the sedimentary filling. In this sense, the DST appears to act as an impermeable barrier between two different rock formations. Such a localized fluid barrier is consistent with models of fault zone evolution but has so far not been imaged by geophysical methods. The situation at the DST is remarkably different from active segments of the San Andreas Fault which typically show a conductive fault core acting as a fluid conduit. INDEX TERMS: 1515

Robust estimates of magnetotelluric and geomagnetic response functions are determined using the coherency and expected uniformity of the magnetic source field as quality criteria. The method is applied on data sets of three simultaneously recording sites. For the data acquisition we used a new generation of geophysical equipment (S.P.A.M. MkIII), which comprises novel concepts of parallel computing and networked, digital data transmission. The data‐processing results show that the amount of noise on the horizontal components of the magnetic field varies considerably in time, between sites and over the frequency range. The removal of such contaminated data beforehand is essential for most data‐processing schemes, as the magnetic channels are usually assumed to be free of noise. The standard remote reference method is aimed at reducing bias in response function estimates. However, this does not necessarily improve their precision as our results clearly show. With our method, on the other hand, we can filter out source field irregularities, thereby providing suitable working conditions for the robust algorithm, and eventually obtain considerably improved results. Contrary to previous concepts, we suggest rejecting as much data as feasible in order to concentrate on the remaining parts of high‐quality observations.

S U M M A R YMagnetotelluric (MT) response function estimates can be severely disturbed by the effects of cultural noise. Methods to isolate and remove these disturbances are typically based on time-series editing, robust statistics, remote reference processing, or some combination of the above. Robust remote reference processing can improve the data quality at a local site, but only if synchronous recordings of at least one additional site are available and if electromagnetic noise between these sites is uncorrelated. If these prerequisites are not met, we suggest an alternative approach for noise removal, based on a combination of frequency domain editing with subsequent single site robust processing. The data pre-selection relies on a thorough visual inspection of a variety of statistical parameters such as spectral power densities, coherences, the distribution of response functions and their errors, etc. Extreme outliers and particularly noisy data segments are excluded from further data processing by setting threshold values for individual parameters. Examples from Namibia and Jordan illustrate that this scheme can improve data quality significantly. However, the examples also suggest that it is not possible to establish generally valid rules for selection as they depend strongly on the local noise conditions. High coherence, for example, can indicate a good signal-to-noise ratio or strongly correlated noise. However, we found that strong polarization of the magnetic field channels and the distribution of response function errors are two important parameters for noise detection.The magnetotelluric (MT) method is based on measuring time variations of orthogonal components of electric and magnetic fields at the surface of the Earth. The MT impedance tensor Z, which generally should be a time invariant quantity is the response of the Earth to electromagnetic induction and carries information about the conductivity distribution of the subsurface. In the frequency domain, the electromagnetic fields are assumed to be linearly related by the impedance tensor Z (e.g. Berdichevsky 1960Berdichevsky , 1964Tikhonov & Berdichevsky 1966):with E being the electric field in mV km −1 , B the magnetic field in nT and Z i j (i, j = x, y) the components of the impedance tensor Z in units of m s −1 . A similar relation can be postulated for the vertical magnetic field (e.g. Schmucker 1970):with T x and T y as the geomagnetic transfer functions. More generally, the relations above can be described by the following expression:Usually, the output channel X is associated with either E x , E y or B z and the input channels Y 1 and Y 2 with B x and B y , respectively (see eqs 1 and 2); Z 1 and Z 2 are response functions of a linear equation system. In general, the estimation procedure for the components of Z as well as T x and T y is based on least-squares (LSQ) methods; the parameter estimate is chosen in order to minimize the misfit between the predicted (right side of eq. 3) and observed output variable (left side of eq. 3) by mini...

S U M M A R YMagnetotelluric and seismic methods provide complementary information about the resistivity and velocity structure of the subsurface on similar scales and resolutions. No global relation, however, exists between these parameters, and correlations are often valid for only a limited target area. Independently derived inverse models from these methods can be combined using a classification approach to map geologic structure. The method employed is based solely on the statistical correlation of physical properties in a joint parameter space and is independent of theoretical or empirical relations linking electrical and seismic parameters. Regions of high correlation (classes) between resistivity and velocity can in turn be mapped back and reexamined in depth section. The spatial distribution of these classes, and the boundaries between them, provide structural information not evident in the individual models. This method is applied to a 10 km long profile crossing the Dead Sea Transform in Jordan. Several prominent classes are identified with specific lithologies in accordance with local geology. An abrupt change in lithology across the fault, together with vertical uplift of the basement suggest the fault is sub-vertical within the upper crust.The interpretation of geophysical models derived by inversion is a highly subjective part of any geologic study. Our incomplete knowledge of the subsurface, the spatially-varying resolution of the models, and the non-uniqueness of the geophysical inverse problem make it difficult to objectively interpret physical property models in terms of geologic structure. The problem is exacerbated by the many-tomany, or at best, many-to-one, relationship between geologic units and their physical properties. It is thus commonplace to use multiple methods to determine multiple physical properties over an area of interest in order to discriminate between the range of possible geologic/lithologic structures. The analysis of such complementary data, however, is rarely taken beyond a qualitative comparison. Attempts at a quantitative comparison are, for the most part, centered upon constitutive or empirical relations between physical properties, which tend to be limited in scale and applicability.Seismic and magnetotelluric (MT) methods are often favored for crustal studies as they provide images of acoustic velocity (V p , V s ) and electrical resistivity (ρ), respectively, on similar scales and with comparable spatial resolution (Jones 1987). By looking in tandem at velocity and resistivity we retain the strengths of each method, while lessening the susceptibility of our interpretation to their individual weaknesses. Seismic refraction, for example, has difficulty imaging vertical velocity contrasts. Along a similar vein, MT has difficulty resolving structure beneath strong conductors due to the large amount of energy dissipated within them. A properly formulated joint interpretation must take these variations in resolution into account, but unfortunately there exists no fundame...