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

| Class | Purpose | |-------|---------| | pg.Mesh | Finite element meshes | | pg.DataContainer | Survey data and geometry | | pg.Inversion | Base inversion framework | | ert.ERTManager | ERT processing and inversion | | srt.SRTManager | Seismic refraction inversion |

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

| Code | Array | |------|-------| | wa | Wenner-alpha | | wb | Wenner-beta | | dd | Dipole-dipole | | pd | Pole-dipole | | pp | Pole-pole | | slm | Schlumberger | | gr | Gradient |

Data Formats

| Format | Extension | Description | |--------|-----------|-------------| | BERT/pyGIMLi | .ohm | Unified data format | | Syscal | .txt | IRIS export | | Res2DInv | .dat | 2D inversion format | | ABEM | .ohm | ABEM Terrameter | | SRT | .sgt | Seismic traveltimes |

When to Use vs Alternatives

| Scenario | Recommendation | |----------|---------------| | Standard ERT inversion with common arrays | pyGIMLi - simplest API, built-in array types | | Seismic refraction tomography (SRT) | pyGIMLi - integrated SRT manager | | Multi-method inversion (DC, magnetics, gravity, EM) | SimPEG - broader method coverage | | Commercial ERT processing with reporting | Res2DInv - industry standard, GUI-based | | Custom forward operators or research flexibility | SimPEG - more modular design | | FEM-based geophysical modelling | pyGIMLi - native FEM mesh support |

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/ertinversion.py](scripts/ertinversion.py) - Complete ERT inversion workflow

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: SteadfastAsArt
• Source: SteadfastAsArt/geoscience-skills
• License: MIT

Install and usage instructions live in the source repository linked above.