Smithery is now a part of Arcade.dev! Read more in our announcement here
Magnetotelluric data processing and modelling. Read EDI files, analyze MT responses, perform inversions, and visualize resistivity models...
Install the mtpy-v2 distribution, which imports as mtpy. The older mtpy
distribution has a different API; do not install both in one environment.
The operations below were executed with MTpy-v2 2.1.4 and mt-metadata 1.0.10.
from mtpy import MT
station = MT("station.edi")
station.read(get_elevation=False) # constructor alone does not load the file
frequency_hz = station.frequency
z = station.Z.z # complex tensor, shape (frequency, 2, 2)
sigma_z = station.Z.z_error # standard deviation, not EDI variance
rho = station.Z.resistivity # ohm m
phase = station.Z.phase # degrees, signed component phases
Confirm impedance units, time convention, axes, channel orientation, ZROT,
horizontal CRS and elevation reference from acquisition metadata. An EDI UNITS
field can describe distances rather than impedance. MT units are mV/km/nT;
the corresponding E/H impedance in ohms is multiplied by mu0 * 1000.
For MT units, apparent resistivity is 0.2 * abs(z)**2 / frequency_hz.
Do not call xy and yx TE/TM until a justified 2D strike coordinate system
has been established. A yx phase near -135 degrees can be valid under NED.
Read EDI conventions and missingness when handling empty values, variances or exports. mt-metadata can represent missing EDI values as zero, so inspect source masks before computing quality flags or rotating. Missing uncertainty is not an exact observation.
Use the QC helper for direct-impedance EDI after confirming mV/km/nT, positive time convention and north/east/down:
python scripts/mt_analysis.py station.edi --output-dir results \
--impedance-units mt --sign-convention + --plot
Resolve the script relative to this skill directory. It exports every frequency and all four components with missingness, uncertainty and diagnostic flags, plus a JSON provenance file. It reopens its CSV and refuses existing output files. The default 50% relative-error threshold is a recorded diagnostic choice; choose task-specific criteria before examining the desired interpretation.
An additional clockwise rotation is available as --rotation-deg 30; it requires
complete tensors and variances. It adds to the original orientation, not an
absolute strike estimate. The tested direct library equivalent is
station.rotate(30, inplace=True); its default in-place form returns None.
Marginal error propagation does not provide a full covariance model.
For complete data, write with station.write(fn="copy.edi"), then read back
coordinates, rotation and tensors. The native EDI writer maps numeric zeros to
its EMPTY sentinel; preserve the original and a separate mask-aware CSV when
physical zeros or missing components matter. Read
response plotting only when a figure is requested.
Compare tensor components, uncertainty, phase behavior and frequency coverage before dimensionality analysis. Phase-tensor skew is a diagnostic affected by noise and sampling, not a universal pass/fail test for 3D geology. Do not apply automatic strike or static-shift corrections without independent justification. Preparing ModEM/Occam inputs is separate from executing an installed solver.
The repository's near-surface workflow checks a field-derived upstream EDI sample, independent tensor algebra, synthetic responses and output readback. It does not establish raw time-series estimation, field inversion or geological uniqueness. Consult the installed-version MTpy-v2 API for other formats and survey collections; do not substitute older v1 method names.
