Question 3 of 9: Planning, Acquiring, Processing and Interpreting a Magnetotelluric Survey

Question 3: Planning, Acquiring, Processing and Interpreting a Magnetotelluric Survey (16.66% of paper)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

Planning. The magnetotelluric (MT) method uses naturally occurring time-varying electromagnetic fields (from ionospheric/magnetospheric currents at longer periods and worldwide lightning activity, "sferics," at shorter periods) as the source, so no transmitter is required — planning starts from the TARGET DEPTH, since the skin depth relation $\delta\approx503\sqrt{\rho/f}$ sets which period range must be recorded (deeper targets need lower frequencies/longer periods, hence longer recording times). Station spacing is chosen from the expected lateral scale of the target and whether a 1-D, 2-D or 3-D interpretation is intended (a 2-D profile needs stations along a line perpendicular to geoelectric strike; a 3-D survey needs an areal grid). Sites must be sited away from cultural EM noise (power lines, railways, pipelines, fences) and, where possible, a simultaneously recording remote reference station is planned tens of kilometres away to allow noise removal in processing. Audio-MT (AMT, higher frequency) may be added for shallow targets alongside conventional MT for deeper structure.

Acquisition. At each site, two orthogonal horizontal electric-field components ($E_x,E_y$, via grounded dipoles, typically 50–100 m) and the corresponding horizontal magnetic-field components ($H_x,H_y$, via induction coils or fluxgate magnetometers) are recorded simultaneously as continuous time series; the vertical magnetic field $H_z$ is often also recorded (it is zero for a perfectly 1-D earth, so a non-zero $H_z$ is itself diagnostic of lateral structure, expressed as a "tipper"). Recording must run long enough to capture enough low-frequency energy for the deepest target period, which can mean many hours to days for very deep crustal studies. GPS timing synchronizes all sites (essential for remote-reference and array processing).

Processing. The time series are Fourier-transformed (or processed with a robust/cascade-decimation spectral method) into the frequency domain, and the impedance tensor $Z$ relating $E$ to $H$ ($E=ZH$) is estimated at each frequency, typically by a robust regression that down-weights noise-contaminated frequency estimates; using a REMOTE REFERENCE station's magnetic field in place of (or alongside) the local $H$ field in this regression removes locally correlated noise that would otherwise bias the impedance estimate. From $Z$, apparent resistivity $\rho_a=\frac{1}{\omega\mu_0}|Z|^2$ and phase are computed as a function of period, along with the tensor's skew and dimensionality indicators to check whether the local geology is effectively 1-D, 2-D or genuinely 3-D. A static-shift correction (a frequency-independent, near-surface galvanic distortion of the apparent-resistivity curve, caused by small near-surface heterogeneities) is applied where independent constraints (TEM soundings, borehole resistivity) are available.

Interpretation. The processed apparent-resistivity and phase curves are inverted for a resistivity model: a simple 1-D layered inversion at an isolated sounding, or a 2-D inversion along a profile (using the TE and TM mode data separately or jointly, after first rotating the impedance tensor into the geoelectric strike direction), or a full 3-D inversion for an areal survey. The resulting resistivity-depth section is interpreted geologically using known property associations (clay/saline-fluid zones and graphitic/sulphide horizons are conductive; crystalline basement, resistive carbonates and permafrost are resistive), cross-checked against any independent constraint (wells, other geophysics, geological mapping) since MT inversion is inherently non-unique.