Standalone toolkit for running Laterally Constrained Inversion (LCI) and Piecewise Inversion (PWI) on frequency-domain EMI (FDEM) data, extracted from the FDEM_LCI_PWI research repo. Two independent use cases:
Field/— run LCI/PWI directly on raw DUALEM-842s.xyzfield exports. Ready to run out of the box againstField/data/sample.xyz.2Layers/— the original synthetic 2-layer LCI/PWI scripts, kept as a reference implementation / CLI pattern. These expect their ownmodel_s*_c*_u*.npy/data_s*_c*_u*.npyfiles (not included here — see Notes below).
Both use cases share the forward-modelling code in src/FDEM1D.py.
conda env create -f environment.yml
conda activate fdem_lci_pwior, without conda:
pip install -r requirements.txt.
├── environment.yml / requirements.txt
├── src/
│ ├── FDEM1D.py forward model + pygimli Modelling/LCModelling classes
│ └── showStitched.py stitched 1D-model cross-section plotting
├── Field/
│ ├── README.md full usage guide (start here for field data)
│ ├── dualem_xyz.py DUALEM-842s .xyz reader
│ ├── run_lci_field.py LCI on raw .xyz field data
│ ├── run_pwi_field.py PWI on raw .xyz field data
│ └── data/sample.xyz small example file for smoke-testing
└── 2Layers/
├── run_lci.py LCI on synthetic 2-layer data
└── run_pwi.py PWI on synthetic 2-layer data
cd Field
python run_lci_field.py data/sample.xyz
python run_pwi_field.py data/sample.xyzSee Field/README.md for the full guide: input format,
all CLI options, output files, and how to use dualem_xyz.py directly.
The forward models in src/FDEM1D.py can use any subset of the instrument's
coils and offsets. Available coils and offsets (DUALEM-842s):
| Coil | Offsets [m] |
|---|---|
H |
2.0, 4.0, 8.0 |
V |
2.0, 4.0, 8.0 |
P |
2.1, 4.1, 8.1 |
Use InstrumentConfig.subset(coils=..., offsets=...). Offsets are matched on
their integer part, so 2 selects H/V at 2.0 m and P at 2.1 m.
from FDEM1D import FDEM1D, DUALEM842S
cfg = DUALEM842S.subset(coils=['H'], offsets=[2, 4]) # H at 2.0 and 4.0 m
cfg = DUALEM842S.subset(coils=['H', 'P'], offsets=[8]) # H 8.0 m, P 8.1 m
cfg = DUALEM842S.subset(coils=['V']) # V, all offsets
cfg = DUALEM842S.subset(offsets=[4, 8]) # all coils, 4 and 8 m
resp = FDEM1D(sgm, thk, config=cfg)For different offsets per coil, give them explicitly:
from FDEM1D import InstrumentConfig
cfg = InstrumentConfig(offsets={'H': [2.0, 4.0], 'V': [8.0], 'P': [4.1]})The config works with FDEM1D, FDEM1D_mdp, FDEM1D_field, FDEM1D_topo
and the FDEM1DModelling* classes (config=cfg).
Notes:
- The response vector keeps the order
[HOP, POP, VOP, HIP, PIP, VIP](FDEM1D_field:[HOP, VOP, POP, HIP, VIP]), containing only the selected channels. Select the observed data the same way, or the sizes won't match. FDEM1D_fieldnever includes 2 m in-phase (not recorded in field data), so H or V at 2 m only gives quadrature values.- Asking for a coil with no matching offset raises a
ValueError.
run_lci.py and run_pwi.py take a slope/contrast/unit triple and
load models/model_s{slope}_c{contrast}_u{unit}.npy and
data/data_s{slope}_c{contrast}_u{unit}[_nosort].npy relative to
2Layers/. Those synthetic dataset files aren't included in this toolkit —
supply your own in that shape (see the docstrings/--help in each script
for the exact CLI), or generate them with the model/forward-modelling
scripts in the original
FDEM_LCI_PWI repo.
cd 2Layers
python run_lci.py 1 1 c --help
python run_pwi.py 1 1 c --help