steadfastasart/geoscience-skills/pygimli

pyGIMLi - Geophysical Inversion

Quick Reference

import pygimli as pg
from pygimli.physics import ert, srt
# Load ERT data
data = ert.load("survey.ohm")
# Invert
mgr = ert.ERTManager(data)
model = mgr.invert(lam=20, verbose=True)
# View result
mgr.showResult()

Key Classes

Essential Operations

Load and View ERT Data

import pygimli as pg
from pygimli.physics import ert
data = ert.load("survey.ohm")
print(f"Measurements: {data.size()}")
ert.showData(data)  # Pseudosection

ERT Inversion

from pygimli.physics import ert
mgr = ert.ERTManager(data)
model = mgr.invert(
    lam=20,          # Regularization
    verbose=True
)
mgr.showResult()
resistivity = mgr.model

Seismic Refraction

from pygimli.physics import srt
data = srt.load("traveltimes.sgt")
mgr = srt.SRTManager(data)
model = mgr.invert(lam=30, zWeight=0.3)
mgr.showResult()

Create Custom Mesh

import pygimli as pg
from pygimli.physics import ert
data = ert.load("survey.ohm")
mesh = pg.meshtools.createParaMesh(
    data.sensors(),
    quality=34.0,
    paraMaxCellSize=5,
    boundary=2
)
pg.show(mesh)

Save and Export

# Save mesh and model
mgr.mesh.save("result_mesh.bms")
pg.save(model, "resistivity_model.vector")
# Export to VTK for ParaView
mgr.mesh.exportVTK("result", mgr.model)

Array Types

Data Formats

When to Use vs Alternatives

Choose pyGIMLi when: You need near-surface geophysical inversion (ERT, SRT, IP) with minimal code. Its manager classes (ERTManager, SRTManager) handle the full workflow from data loading to inversion to visualization in a few lines.

Avoid pyGIMLi when: You need methods beyond near-surface (use SimPEG), or you require a commercial-grade reporting pipeline.

Common Workflows

ERT data inversion and visualization

• [ ] Load ERT data file with ert.load("survey.ohm")
• [ ] Inspect data: check measurement count with data.size(), plot pseudosection
• [ ] Remove outliers or bad data points
• [ ] Create ERTManager with data
• [ ] Run inversion: mgr.invert(lam=20) (start with higher lambda)
• [ ] Check chi-squared value (target ~ 1)
• [ ] Visualize result with mgr.showResult()
• [ ] Export mesh and model to VTK for ParaView: mgr.mesh.exportVTK()
• [ ] Adjust lambda and zWeight if needed, re-invert

Inversion Tips

1. Start with higher lambda (50-100) and decrease
2. Check data quality - remove outliers before inversion
3. Use zWeight < 1 for layered structures
4. Check coverage - low coverage = poorly resolved
5. Chi-squared ~ 1 indicates good fit without overfitting

References

• [Geophysical Methods](references/methods.md) - Supported methods and workflows
• [Mesh Generation](references/mesh.md) - Mesh creation and quality control

Scripts

• [scripts/ert_inversion.py](scripts/ert_inversion.py) - Complete ERT inversion workflow