SUPPLEMENTARY INFORMATION Supplementary online material for Bai et al., (2). EHS3D MT data collection Broadband magnetotelluric (MT) data were recorded on profiles P, P2 and P4 in the frequency band -.5 Hz using Phoenix Geophysics V5-2 instruments. On profile P3 (CZMT-WE) Metronix instruments were used and gave data in the frequency band 256-.2 Hz. On all profiles MT data were recorded with 3-4 instruments in simultaneous operation, allowing the application of remote reference time series processing. The MT time series data were processed to give estimates of the impedance tensor and vertical magnetic field transfer functions, using a conventional processing algorithm with robust estimation. 2. Determination of strike direction Tensor decomposition was then used to determine the geoelectric strike and results are summarized in Figure S and as follows: P4 (EHS3D-): Strike direction is N5ºW-N5ºE in the western part of the profile and changes to N5ºE to N25ºE east of the Xiaojiang fault (XJF). P2 (EHS3D-2): Strike direction is NºW-NºW. P3 (CZMT-WE) : West of the Longmenshan Fault (LMF) the strike direction is NºW to NºW while to the east in the Sichuan Basin it changes to N5ºE. P (EHS3D-3): Strike direction is NºW-N8ºW. nature geoscience www.nature.com/naturegeoscience
These directions are generally consistent with the regional geological strike of the major faults shown in Figure S. A geoelectric strike direction that varies along a profile is an indication of 3D induction effects and is not ideal for a twodimensional MT analysis. Thus the 2D inversion used a carefully chosen subset of the MT data. In the subsequent data inversions a strike angle corresponding to the dominant strike direction was used for each profile, as listed in Table. In a 2D situation the MT data can be separated into a transverse electric () mode and a transverse magnetic () mode. In the mode the electric currents flow parallel to the strike direction, while in the mode the electric currents flow at right angles to the strike direction. A fully 3D inversion could be applied to the EHS3D magnetotelluric data but the profiles are too widely spaced to sample structure between the profiles. 3. Examples of EHS3D MT data Typical apparent resistivity and phase curves are shown in Figures S2, S3 and S4. In each figure, the data are shown in a co-ordinate system that has been rotated to the regional strike direction (Table ). The black triangles in Figure S2 show the apparent resistivity measured with the independent time domain electromagnetic (TDEM) instrument (PROM EM67). The TDEM data were used to measure near surface resistivity structure with the goal of correcting static shifts present in the magnetotelluric data. Note the good agreement in Figure S2 between the slopes of the TDEM and MT apparent resistivity data. Typical shifts in apparent resistivity were around 2-% with the maximum value being around %. 2 nature geoscience www.nature.com/naturegeoscience
The data are shown in pseudosection format in Figures S5 and S6 for profiles P4 (EHS3D-) and P (EHS3D-3) respectively. This type of display is useful because it uses signal period on the vertical axis. Since increasing signal period corresponds to increasing penetration depth, the pseudosection gives a qualitative impression of how resistivity varies with depth in the Earth. In Figures S5 and S6, the magnetotelluric signals with the shortest periods (. s) sample near surface structure (upper crust) and the apparent resistivity values are relatively high, as expected for dry crustal rocks. At longer periods (- s) corresponding to deeper penetration, reduced apparent resistivities are observed, indicating a zone of lower resistivity (higher conductivity) at depth. The resistivity of this zone varies with distance along both profiles, as can be seen most clearly from the phases. On profile P4 (EHS3D-) note the high phases in both the and mode in the - s period band at distances of -5 km and -55 km. These features are responsible for the conductors in the model in Figure 2 at the same offsets. 4. Details of inversion of EHS3D MT data The MT data for each profile were converted into a resistivity model, taking topography into account, using a conventional 2D inversion algorithm (NLCG). The inversions used the following control parameters: error floor for apparent resistivity = 5%; error floor for phase = %; vertical to horizontal smoothing ratio α = ; trade-off parameter, τ =. The final root-mean-square (r.m.s.) misfits are listed in Table. Inversions started from halfspaces with resistivity 5, and Ωm gave essentially the same final model. To investigate the depth extent of the lower crustal conductors, constrained inversions with the model forced to be resistive below km depth were also implemented, as nature geoscience www.nature.com/naturegeoscience 3
discussed in the main paper. These inversions allow reliable estimates of the lower crustal conductance (depth integrated conductivity) to be made. Inversions with the constraint applied at 8 km gave similar models. The MT responses predicted by the models are shown by the continuous lines in Figure S2-S4 and can be compared with the measured MT data. The fit is also shown in pseudosection format in Figures S5 and S6. Note that there is generally good agreement between the measured MT data and the predicted response. 5. Interpretation The emphasis of this paper is the regional scale geoelectric structure, so the inversion models were forced to be as smooth as possible while still fitting the measured MT data. The interpretation in the main paper shows two linear zones with very high crustal conductance (>, S). Such conductance values are only found in a few locations worldwide and indicate the presence of crustal material with very unusual properties. The data on profile P3 (CZMT-WE) are important because they define the regional tectonic setting of the May 2 28 M = 7.9 Wenchuan earthquake. As outlined in the main paper, it is not clear if dynamic topography caused by crustal flow has contributed to motion on the faults on the west side of the Sichuan Basin. The inversion model in Figure 2 is shown in more detail in Figure S7. Note that the eastern margin of conductor B appears to extend to the surface, sub-parallel to inferred thrust faults in this area. 4 nature geoscience www.nature.com/naturegeoscience
Profile Strike direction Final r.m.s. misfit P4 (EHS3D-) N.43 P3 (CZMT-WE) N.96 P2 (EHS3D-2) N5ºW.8 P (EHS3D-3) N7ºW.98 Table : Summary of strike directions and final r.m.s. misfit for the 2D inversions of the EHS3D magnetotelluric data. nature geoscience www.nature.com/naturegeoscience 5
Figure S: Rose diagrams showing geoelectric strike direction from tensor decomposition. The rose diagrams are for a single MT station that was representative of that section of each profile. Geological labels are the same as in Figure. Locations of stations plotted in Figures S2-S4 are labelled. 6 nature geoscience www.nature.com/naturegeoscience
. EHS3D--85 M..... EHS3D--28 M.... Figure S2: Typical apparent resistivity and phase curves from the P4 (EHS3D- ) profile. The continuous lines represent the fit of the 2D inversion model in Figure 2 to the MT data. Station EHS3D--85 is located between the Red River Fault and the Xiaojiang Fault. Station EHS3D--28 on the east of the Xiaojiang Fault (Yangtze Block). The MT data has been rotated to the strike direction (Table ). nature geoscience www.nature.com/naturegeoscience 7
. EHS3D-3-23.... EHS3D-3-39... Figure S3: Typical apparent resistivity and phase curves from the P (EHS3D- 3) profile. The continuous lines represent the fit of the 2D inversion model in Figure 2 to the MT data. Locations of these MT stations are shown in Figure S6. Data has been rotated to the strike direction of N7ºW. 8 nature geoscience www.nature.com/naturegeoscience
. CZMT-WE-E2.... CZMT-WE-E8... Figure S4: Typical apparent resistivity and phase curves from the P3 (CZMT- WE) profile. The continuous lines represent the fit of the inversion model in Figure 2 to the MT data. CZMT-WE-E2 is located in the Sichuan Basin and the increasing apparent resistivity in the period range - s indicates the usual situation of a resistive crust and upper mantle. CZMT-WE-E8 is located above the centre of the conductor west of Longmenshan Fault and shows apparent resistivity that decreases in the period range - s, indicating that the crust has an unusually low resistivity (high conductivity). Data has been rotated to the strike direction (N). nature geoscience www.nature.com/naturegeoscience 9
Figure S5: The left column shows the pseudo-sections for MT data measured on profile P4 (EHS3D-). The right column shows the response of the 2D inversion model in Figure 2. Note the high values of the phase in both the and mode in the - s period band at distances of -5 km and - 55 km. These features are responsible for the conductors in the model in Figure 2 at the same offsets. The locations of stations from Figure S2 are shown (EHS3D--85 is station 79 from left and EHS3D--28 is station 2 from left) nature geoscience www.nature.com/naturegeoscience
Figure S6: The left column shows the pseudo-sections for MT data measured on profile P (EHS3D-3). The right column shows the response of the 2D inversion model in Figure 2. Note the high values of the phase in both the and mode in the - s period band. These features are responsible for the two conductors in the model in Figure 2. The locations of stations from Figure S3 are shown (EHS3D-3-23 is station 29 from left and EHS3D-3-39 is station 45 from left) nature geoscience www.nature.com/naturegeoscience